Docstoc

Chemical Solution Deposition of Semi

Document Sample
Chemical Solution Deposition of Semi Powered By Docstoc
					       CHEMICAL SOLUTION
         DEPOSITION OF
      SEMICONDUCTOR FILMS



                             Gary Hodes
                        Weizmann Institute of Science
                              Rehovot, Israel




            Marcel Dekker, Inc.                            New York • Basel
       TM




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ISBN: 0-8247-0851-2

This book is printed on acid-free paper.

Headquarters
Marcel Dekker, Inc.
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540

Eastern Hemisphere Distribution
Marcel Dekker AG
Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-260-6300; fax: 41-61-260-6333

World Wide Web
http://www.dekker.com

The publisher offers discounts on this book when ordered in bulk quantities. For more information,
write to Special Sales/Professional Marketing at the headquarters address above.

Copyright © 2003 by Marcel Dekker, Inc. All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form or by any means, elec-
tronic or mechanical, including photocopying, microfilming, and recording, or by any information
storage and retrieval system, without permission in writing from the publisher.

Current printing (last digit):
10 9 8 7 6 5 4 3                 2   1

PRINTED IN THE UNITED STATES OF AMERICA




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
           To my parents, for their dedication to my education
                     and their faith in my abilities




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Preface




Chemical solution deposition (CSD); known also as chemical bath deposition
(CBD) and simply chemical deposition (CD, the form we will use in this book)
was first described in 1869, and it has been used since to deposit films of many
different semiconductors. It is probably the simplest method available for this pur-
pose—all that is needed is a vessel to contain the solution (an aqueous solution
made up of a few, usually common, chemicals) and the substrate on which depo-
sition is required. Various “complications,” such as some mechanism for stirring
and a thermostated bath to maintain a specific and constant temperature, are op-
tions that may be useful.
       In spite of this extreme experimental simplicity, understanding the mecha-
nisms involved in the deposition and the ability to widen the range of deposits ob-
tained—both in composition and the control of numerous other properties—is
usually not so simple. Also in spite of its simplicity, it has not been exploited as a
technique as much as might be expected. However, CD has experienced some-
what of a renaissance recently, due largely to its overwhelmingly successful use
in depositing buffer layers of CdS (and similar materials) in thin-film photovoltaic
cells. The deposition of the CdS, as with many other semiconductors that have
been deposited by CD, is often recipe oriented; there seem to be almost as many
different “recipes” as there are groups.
       Notwithstanding the wide interest and use of this technique, at the time
when the idea to write this book was conceived, there was no recent comprehen-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sive review or any general summing up of the field of CD. The first general re-
view on the subject is that by Chopra et al. in 1982 [1]. Nine years later, a review
by Lokhande [2] was published with an emphasis on describing the deposition of
the various semiconductors that had been deposited up to then. A comprehensive
and general review was published by Lincot et al. in 1998 [3] (about a year after
the writing of this book had commenced) and, in the same year, two more reviews,
one a more specific review by Nair et al. [4] on their extensive work in the field
connected with solar energy–related issues and the other, by Savadogo [5], de-
scribing CD (and electrodeposited) semiconductors used as solar energy materi-
als. In the last two years, two new reviews have appeared, by Mane and Lokhande
[6] and most recently by Niesen and DeGuire [7], the latter covering also other so-
lution deposition methods, such as SILAR (Successive Ion Layer Adsorption and
Reaction), electroless deposition and liquid phase deposition (which can be con-
sidered to be a subset of CD) and emphasizes oxides, although sulphides and se-
lenides are also covered.
       The driving force for this book is the perceived need for a detailed coverage
of a field that has expanded enormously in recent years. While the title of the tech-
nique suggests that the book is aimed mainly at chemists, this would be an incor-
rect impression. Many of those who should find this book useful will be “physi-
cists” or “engineers” who are dealing with thin-film photovoltaic cells. Some of
these readers may have only a superficial background in chemistry, and for these,
Chapter 1, Fundamentals, which deals with the science (largely chemistry) behind
the technique, will be very important background reading. In this chapter, mate-
rial will be found on such topics relevant to CD as principles of precipitation and
solubility product; nucleation; growth; colloids; aggregation and sticking. Even
those with a good chemistry background are advised to read this chapter, if only
to refresh their knowledge.
       In one respect, this book is organized somewhat differently than usual. It
contains a fairly comprehensive review of CD in the form of Chapter 2, General
Review. Most of the material in this review will be expanded on in the relevant
chapters, and one might ask why it is included at all. The reason is that most peo-
ple do not read a book of this type from cover to cover; they read those chapters
or parts of chapters they consider relevant to their purposes. In doing this, they are
likely to miss matter from other chapters that is also relevant. For this reason, it is
strongly recommended that all readers read through the first three chapters. Chap-
ter 1, as already noted, is to acquire or refresh the relevant scientific background.
Chapter 2 should give a good overview of what has been done without having to
go into too much detail. Apart from this, however, Chapter 2 contains detail not
found elsewhere; a short history of CD, details on the effects of substrate nature
and variation (if I were to rewrite this book, an additional chapter would be de-
voted solely to the substrate), some “recipes” for depositing certain films and,
very important, descriptions, and, where possible, explanations of the reasons for



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
a particular recipe and the expected effects of various changes to those recipes.
Also, other methods that are related to CD are treated briefly in this chapter. In or-
der to promote the ease of reading and “flow” in this chapter, references are, for
the most part, not given unless they are not clearly provided in the chapter and sec-
tion relevant to that particular topic elsewhere in the book. The importance of the
third “required” chapter, Chapter 3, Mechanisms of Chemical Deposition, should
be self-evident. Understanding the possible mechanisms of the deposition is the
best defense against what is commonly (and often with considerable justification)
thought of as a “recipe-oriented” field. Not that that a reading of this chapter will
automatically endow the reader with the ability to know what mechanism is actu-
ally operating in every case—unfortunately, often far from it. However, it is hoped
that it will help in the movement toward that goal.
       Having worked (it is hoped) through these “compulsory” chapters, most
readers will want more detail on specific aspects that are important to them. How
can films of X be made? How are the various experimental parameters expected
to affect the properties of this film? Why is CD so useful for photovoltaic cells?
How can nanocrystalline films of Y be made with a specific crystal size? Such
questions will be answered in subsequent chapters (or, if not answered, at least in-
formation will be given to allow the reader to plan experiments in order to find the
answers).
       The next five chapters deal with deposition of specific groups of semicon-
ductors. In Chapter 4, II–VI Semiconductors, all the sulphides, selenides, and
(what little there is on) tellurides of cadmium (most of the chapter), zinc (a sub-
stantial part), and mercury (a small part). (Oxides are left to a later chapter.) This
chapter is, understandably, a large one, due mainly to the large amount of work
carried out on CdS and to a lesser extent on CdSe. Chapter 5, PbS and PbSe, pro-
vides a separate forum for PbS and PbSe, which provided much of the focus for
CD in earlier years. The remaining sulphides and selenides are covered in Chap-
ter 6, Other Sulphides and Selenides. There are many of these compounds, thus,
this is a correspondingly large chapter. Chapter 7, Oxides and Other Semicon-
ductors, is devoted mainly to oxides and some hydroxides, as well as to miscella-
neous semiconductors that have only been scantily studied (elemental selenium
and silver halides). These previous chapters have been limited to binary semicon-
ductors, made up of two elements (with the exception of elemental Se). Chapter
8, Ternary Semiconductors, extends this list to semiconductors composed of three
elements, whether two different metals (most of the studies) or two different
chalcogens.
       The final two chapters deal with “applications” (in the scientific as well as
commercial sense) of CD films. As already mentioned, photovoltaic cells is the
one subject that has given CD a push in the last decade, while photoelectrochem-
ical cells was probably the main driving force for such studies in the decade be-
fore that. Chapter 9 deals with Photovoltaic and Photoelectrochemical Properties.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Finally, the tendency for CD films to be nanocrystalline and often to exhibit quan-
tum-size effects is treated in the final chapter, Chapter 10, Nanocrystallinity and
Size Quantization in CD Semiconductor Films.
       The layout of this book means that there will be some overlap between sec-
tions. However, this system should allow those readers interested in one or more
specific sections to skip the others, thereby making the book more efficient for the
individual reader. An example of this is the use of quantum-size effects to eluci-
date CD mechanisms. This is treated, with different emphasis, both in Chapter 3
(Mechanisms of Chemical Deposition) and in Chapter 10 (Nanocrystallinity and
Size Quantization in CD Semiconductor Films).
       This book covers a field that, from its title, may appear to be limited. In re-
ality, it is surprisingly multidisciplinary. Inorganic chemistry and film formation
are, of course, fields that are evident from the title. Certainly, those concerned
with the deposition of semiconductor films for any purpose should find this book
useful and informative. However, it will also be valuable to those working in other
fields. The considerable section on semiconductor quantum dots, for example,
will be of interest to those working with low-dimensional semiconductors. Scien-
tists and engineers working in thin-film solar cells will find a compendium of re-
search on CD buffer layers in these cells. There is much in the book relevant to
colloid scientists. Even biologists are not forgotten: the slow formation and depo-
sition of inorganic compounds characteristic of CD has a lot in common with
biomineralization, and hopefully this book will be useful to those working in the
field of biomimetics. And, of course, the original application of CD, for near- to
mid-infrared detectors, will attract those designing or using optoelectronic equip-
ment in this wavelength range.
       Finally, a word concerning the coverage of the literature. When starting out
on this enterprise, the intention was to try and cover the field more or less com-
pletely, with the exception of some of the earlier work, mostly on PbS. Of course,
some papers might occupy a considerable amount of book space, while, at the
other extreme, others might just be mentioned in a table. During the long course
of putting this book together, it became increasingly clear, from the “new” litera-
ture that appeared (not necessarily chronologically new but just new to the au-
thor), even toward the final stages of writing, that an appreciable number of rele-
vant papers would remain unknown to the author. For these, the author expresses
regret, not only to the authors of such works, but also to the readers of this book.
It is hoped that the expanding literature that is continually appearing on the sub-
ject does not render this book out of date too rapidly. From the objective scientific
viewpoint, it can be hoped that this process will not be too slow either.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
REFERENCES
1. KL Chopra, RC Kainthla, DK Pandya, AP Thakoor. In: Physics of Thin Films, Vol. 12.
   Academic Press, New York and London, 1982, pp 167.
2. CD Lokhande. Mater. Chem. Phys. 28:1, 1991.
3. D Lincot, M Froment, H Cachet. In: RC Alkire, DM Kolb, Eds. Adv. Electrochem. Sci.
   Eng., New York: Wiley-VCH, 1998, Vol. 6, p 165.
4. PK Nair, MTS Nair, VM Garcia, OL Arenas, Y Pena, A Castillo, IT Ayala, O Gomez-
   daza, A Sanchez, J Campos, H Hu, R Suarez, ME Rincon. Sol. Energy Mater. Sol. Cells
   52:313, 1998.
5. O Savadogo. Sol. Energy Mater. Sol. Cells 52:361, 1998.
6. RS Mane, CD Lokhande Mater. Chem. Phys. 65:1, 2000.
7. TP Niesen, MR De Guire J. Electroceram. 6:169, 2001.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Contents




Dedication
Preface

 1. Fundamentals


 2. General Review


 3. Mechanisms of Chemical Deposition


 4. II-VI Semiconductors


 5. PbS and PbSe


 6. Other Sulphides and Selenides


 7. Oxides and Other Semiconductors




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 8. Ternary Semiconductors


 9. Photovoltaic and Photoelectrochemical Properties


10. Nanocrystallinity and Size Quantization in
    Chemical Deposited Semiconductor Films




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
1
Fundamentals




The purpose of this chapter is to give sufficient background in the chemical prin-
ciples involved in CD. For those with a good background in chemistry, a quick read
through this chapter as a refresher course will probably be sufficient. However, the
chapter is written, to a large extent, keeping in mind that not all readers will have
a good chemistry background. The emphasis is on a qualitative or semiquantitative
understanding of the principles involved, sufficient to understand the concepts that
arise throughout the book. A deliberate policy has been made not to go too deeply
into the details of these fundamentals where it is considered unnecessary; refer-
ences to further reading will be given for those who wish such additional detail.


1.1 SOLUTION CHEMISTRY
1.1.1 Basic Terminology
The pH of a solution is the negative logarithm of the hydrogen ion concentration
in the solution:
      pH       log [H ]                                                        (1.1)
(Note for thermodynamic purists: Here and for the rest of this book, concentra-
tions are used in place of activities.) Concentrations are denoted by square brack-
ets; thus [H ] means the concentration of hydrogen ions.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
        The pH of pure water at 25°C is 7. Most (but not all) CD reactions take place
in basic solutions at typical pH values of 9–12. Since hydroxide intermediates are
often important in CD, it is worth noting that a pH of 10 is equivalent to a hy-
droxide ion concentration of 10 4 M at 25°C (since the ion product of water,
[H ][OH ], 10 14 at this temperature). As will be discussed shortly, this ion
product is very temperature dependent, and so the OH concentration at any par-
ticular pH varies considerably with temperature.
        The pH of pure (and also not so pure) water is very sensitive to small con-
centrations of acids and bases. One drop of concentrated sulphuric acid added to
a liter of water will change the pH by 4 pH units (from 7 to ca. 3). Solution pH can
be stabilized by a buffer (although there may be cases where a stable pH is not de-
sirable); addition of (not too large) quantities of acid or base to a buffered solution
will not affect the pH much. Buffers are usually mixtures of weak acids or bases
and their salts. A common example in CD is the use of an ammonium salt
(NH 4 X ) to control the pH of an ammonia solution. The equilibrium of ammo-
nia in water is given by
      NH3       H2O D NH 4           OH                                          (1.2)
      Since hydroxide ions are formed when ammonia dissolves in water, the pH
of an aqueous ammonia solution is alkaline. The value of pH can be calculated
from a knowledge of the equilibrium constant, K, of this equilibrium. The equi-
librium constant for an equilibrium of general type
      aA      bB            dD D eE        fF            hH                      (1.3)
is given by
              [E]e[F] ƒ . . . [H]h
      K       [A]a[B]b . . . [D]d                                                (1.4)

K for the ammonia dissolution, Eq. (1.2), is given by
              [NH 4 ][OH ]                     5
      K                          1.8      10       (at 25°C)                     (1.5)
                 [NH3]
For example, for a 1 M solution of ammonia ([NH3]         1), since the NH 4 and
OH concentrations are equal [from Eq. (1.2)], [OH ] can be calculated to be
4.2 10-3 M. Knowing that the ion product of water, [H ][OH ], 10 14 at this
temperature allows us to convert [OH ] to [H ] and thus to find the pH (which is
11.62).
       If an ammonium salt is added to the ammonia solution, the NH 4 concentra-
tion is now dictated by the concentration of ammonium salt added rather than by
that existing due to the weak dissociation of ammonia. Thus, assuming the same
1 M ammonia as before, but adding 0.1 M NH 4 (say, as NH4Cl), then, ignoring
the few percent correction due to the extra NH 4 arising from the NH3 dissocia-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tion, [OH ] is given by
                 1.8      10-5              4
      [OH ]                      1.8   10       M                               (1.6)
                       0.1
and the pH of the solution becomes 10.25.
       The buffering action of this solution can be understood by considering equi-
librium Eq. (1.2). If extra OH is added to the solution, the equilibrium is shifted
to the left; i.e., it tends to remove OH . It also removes NH 4 ; but if the concen-
tration of this ion is high to begin with, then this change will not affect the pH
greatly. (This is the reason that the combination of ammonia and ammonium ions
is a better buffer than ammonia by itself.) A similar argument can be made for ad-
dition of acid, through the equilibrium:
      NH3     H D NH 4                                                          (1.7)
In this case, hydrogen ions are consumed in converting ammonia to NH 4 .
       This buffering action requires nonionized base (or acid) to operate, hence
the requirement of a weak base or acid together with its salt
       As well as this buffering action, addition of ammonium ion also decreases
the pH of an ammonia solution as shown above. This is an important effect—more
important than the buffering action in many CD processes.

1.1.2 Hydrolysis of Metal Ions
Most cations are hydrated in aqueous solutions to a greater or lesser extent:
      Mx      nH2O D M(H2O)x
                           n                                                    (1.8)
The water is polarized and attracted by the positively charged cation. The greater
the positive charge on the cation and the nearer the water can approach the cation,
the greater will be this polarization and attraction. Thus small, highly charged
(high-valence) cations will in general be more strongly solvated than large, mono-
valent ones.
      Continuing the argument, the positive charge on the cation attracts electrons
from the oxygen of the water molecules. This, in turn, can result in the transfer of
electron density from the OMH bonds to the (now electron-deficient) oxygen, as
exemplified here:



                                                                                (1.9)



This will weaken the OMH bond and may even break it, resulting in formation of
a metal hydroxide and a hydrogen ion, the latter which will be hydrated by a



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
molecule of the surrounding water:
      M(H2O)x
            n        H2O D M(H2O)n 1OH(x         1)
                                                        H3O                    (1.10)
The H3O (a hydrated hydrogen ion) is acidic; therefore this equilibrium gener-
ates acidity in the solution. The more the cation attracts electron density from the
water, therefore, the more acidic is the cation. As with solvation, small, highly
charged cations should be more acidic than large cations with a small charge. The
charge in particular is a very important factor in determining the degree of acidity
of cations. Therefore monovalent cations are generally basic, while trivalent ones
are acidic. Tetravalent cations, such as Sn4 and particularly Ti4 are so highly
acidic that their simple cations either do not exist in water or do so only under very
acidic conditions.
       The hydrated metal hydroxy complex in Eq. (1.10) is a soluble species.
However, if the pH is sufficiently high, the metal hydroxide, which is relatively
insoluble for most metals (apart from the alkali group metals) will precipitate. The
pH value at which hydroxide precipitation occurs can be related to the acidity of
the cation and is approximately equal to the pKa of the cation, where the pKa is
minus the logarithm of the equilibrium constant of Eq. (1.10).

1.1.3 Solubility Product
A central concept necessary to understanding the mechanisms of CD is that of the
solubility product (Ksp). The solubility product gives the solubility of a sparingly
soluble ionic salt (this includes salts normally termed “insoluble”). Consider a
very sparingly soluble salt (say, CdS) in equilibrium with its saturated aqueous so-
lution:
      CdS(s) D Cd2        S2                                                   (1.11)
(where subscript s represents the solid phase). The CdS dissolves in water to give
a small concentration of Cd and S ions. This concentration is defined by the solu-
bility product, Ksp, the product of the concentrations of the dissolved ions:
      Ksp    [Cd2 ][S2 ]                                                       (1.12)
or more generally, for the dissolution:
      MaXb D aMn    bX m                                                       (1.13)
      Ksp [Mn ]a[Xm ]b                                                         (1.14)
The more soluble is the salt, the greater the ion product and the greater is Ksp.
However, Ksp also depends on the number of ions involved. Thus Bi2S3 has a
value of Ksp [Bi3 ]2[S2 ]3 10 100. The very low value is due, in part, to the
relatively large number of atoms in the Bi2S3 molecule and therefore of ions in-
volved in the equilibrium. A list of approximate values of Ksp for some of the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 1.1 Values of the Solubility Product (at 25 C) for Compounds Relevant to CD

Solid             K sp             Solid          K sp             Solid            K sp    Solid          K sp
                             50
Ag2S         3        10          CdSe       4     10 35      FeS                10 18           a
                                                                                           Pb(OH)2   10   15
                                                                                                            –10 20
                       54
Ag2Se            10               CdTe           10 42        HgS               6  10 53   PbS            10 28
                             10
AgCl         2        10          Co(OH)2    5     10 15      HgSe              4  10 59   PbSe           10 37
                             13
AgBr         8        10          CoS            10 21        In(OH)3           6  10 34   Sn(OH)2    5     10 28
                       16
AgI              10               CuOH       1     10 14      In2S3             6  10 76   Sn(OH)4    1     10 56
                             22
As2S3        2        10          Cu(OH)2    2     10 20      Mn(OH)2           5  10 13   SnS            10 26
                             31
Bi(OH)3      6        10          Cu2S           10 48        MnS                10 13     SnS2       6     10 57
                       100
Bi2S3            10               CuS        5     10 36      Ni(OH)2           3 10 16    SnSe       5     10 34
                       130
Bi2Se3           10               CuSe       2     10 40      NiS                10 21     Zn(OH)2        10 16
                             14
Cd(OH)2      2        10          Fe(OH)2    5     10 17      NiSe              2 10 26    ZnS        3     10 25
                       28
CdS              10               Fe(OH)3    3     10 39      PbCO3              10 13     ZnSe           10 27

There is often a large variation in values from source to source—in some cases, some orders of magnitude. For
this reason, only one significant figure (at most) is given before the exponent. A table of solubility products for
many sulphides based on a reevaluated value for the second dissociation constant of H2S is given in Ref. 1. The
values in that study are typically some orders of magnitude lower than the ones shown here.
a
  This is probably a hydrated lead oxide rather than a simple hydroxide.




     semiconductors and related salts encountered in CD is given in Table 1.1. The val-
     ues for oxides are not so readily available as for sulphides and selenides. However,
     it must also be kept in mind that deposition of oxides often occurs via a hydrox-
     ide or hydrated oxide, and therefore the relevant value is that of this hydroxide or
     hydrated oxide.
            Some explanation is required here concerning the S2 ion. In actual fact, in
     all but highly alkaline solutions (and the solutions used in CD, while mostly alka-
     line, are not that alkaline), most of the sulphur ion will be in the form of HS
     rather than S2 . This is due to the equilibrium between the two species:
             HS D S2                    H        Ka      10   17.3
                                                                                                     (1.15)
     or alternatively, in terms of hydroxide concentration which is related to the hy-
     drogen ion concentration through the ion product of water:
             HS             OH D S2              H2O          Ka         10   3.3
                                                                                                     (1.16)
     (at room temperature).
            Thus at a pH of 11 (a common value in CD), which gives a value for [OH ]
     at room temperature of 10 3 M, the S2 concentration will be
             [S2 ]           10   3.3
                                        [HS ][OH ]         10      6.3
                                                                         [HS ]                       (1.17)
     Therefore the main sulphur ion in solution will be HS .



     Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      Since Ksp is given in terms of [S2 ], we can write [S2 ] in terms of [H ]
using Eq. (1.15):
       [S2 ]      10   17.3
                              [HS ]/[H ]                                            (1.18)
and derive a solubility product of a sulphide, MxSy, in terms of the dominant HS
concentration:
       Ksp      [M2y/x ]x[S2 ]y            [M2y/x ]x(10   17.3
                                                                 [HS ]/[H ])y       (1.19)
A list of such solubility products for metal sulphides, as well as updated conven-
tional ones, has been given by Licht [1]. In this book, we will continue to use the
more conventional solubility products, partly because they are more common and
partly because the relevant equilibria are less unwieldly to describe.
       In applying the solubility product concept to CD, it is often useful to con-
sider it in terms of what concentration of ions is required in solution before pre-
cipitation occurs. Thus, for CdS, with a Ksp value of ca. 10 28 (from Table 1.1),
       [Cd2 ][S2 ]        10     28
                                      M.                                            (1.20)
While the concentration of each ion in this example will be equal when dissolu-
tion of the solid is considered, for formation of the solid from the ions, they may
be completely different; it is the product of the concentrations that is important.
Thus a solution 0.2 M in sulphide ion and 10 27 M in Cd2 will (in principle) pre-
cipitate CdS (the ion product will be greater than Ksp) while 0.1 M sulphide and
the same 10 27 M of Cd2 will (only just) not.
       Ksp can be derived theoretically from the free energies of formation of the
species involved in the dissolution equilibria. Thus, for the equilibrium
       MaXb(s) D aMc (aq)                  bXd (aq)                                 (1.21)
the free energy of the dissolution is given by
        G0       a G0(aMc (aq))              b G0(Xd (aq))            G0(MaXb(s))   (1.22)
And since
        G0         RT ln K                                                          (1.23)
then
                        G0
       ln Ksp                                                                       (1.24)
                       RT
Since Ksp is a thermodynamic quantity, the ion product that should result in pre-
cipitation may not necessarily do so for kinetic reasons (hence the term used ear-
lier to qualify precipitation: “in principle”). This would be a case of supersatura-
tion. In practice, however, the solubility product does give a fairly good idea of
when precipitation will occur in most cases.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       CD reactions sometimes proceed via a metal hydroxide intermediate; the
concentration of OH ions in the solution is particularly important in such cases.
Since almost all CD reactions are carried out in aqueous solutions, the pH of the
deposition solution will give this concentration. In translating pH into OH con-
centration, the very temperature-dependent ionization constant of water should be
kept in mind, as mentioned previously. The reason for this can be seen from Table
1.2, which gives the OH concentration in water at a pH of 10 (a typical pH value
for many CD reactions), calculated from the ionization constant of water, Kw from
the relation
        log Kw     log [H ][OH ]          log [H ]     log [OH ]
                                                                             (1.25)
                  pH log [OH ]
The OH concentration increases by nearly two orders of magnitude between 0
and 60°C.
      The OH concentration increases (decreases) by one order of magnitude for
every unit increase (decrease) in pH. This means that the formation of a metal
hydroxide (whether as a colloid or as a precipitate) in aqueous solution will be
strongly dependent on temperature when the product of the free metal ions and
OH ions is close to the hydroxide solubility product, although increase in Ksp
with temperature may partially offset this effect.

1.1.4 Complexation
Most CD reactions are carried out in alkaline solution. To prevent precipitation of
metal hydroxides, a complexing agent (often called a ligand, since complexing
agents to cations are electron donors) is added. The complexant also reduces the
concentration of free metal ions, which helps to prevent rapid bulk precipitation
of the desired product. This section gives the basics of the theory of complexation.


TABLE 1.2 Effect of Temperature on OH Concentration
in Water

Temp.                                  OH concentration
( C)              -log10 Kw              at pH 10
                                                       5
   0               14.944                 1.138   10
                                                       5
  10               14.535                 2.917   10
                                                       5
  20               14.167                 6.808   10
                                                       4
  30               13.833                 1.469   10
                                                       4
  40               13.535                 2.917   10
                                                       4
  50               13.262                 5.470   10
                                                       4
  60               13.017                 9.616   10




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       If a KOH solution is added to a solution of a Cd salt, Cd(OH)2 will precipi-
tate immediately. From the Ksp of Cd(OH)2 (2 10 14 at room temperature), and
assuming a pH of, say, 11 ([OH]          10 3 at the same temperature), from Eq.
(1.14), we can calculate that a Cd concentration above 2 10 8 M is enough to
initiate Cd(OH)2 formation.
       If ammonium hydroxide (ammonia in water)—a common complexant for
Cd in CD—is added to a suspension of Cd(OH)2, the Cd(OH)2 will redissolve, as-
suming enough ammonia has been added. How much is enough ammonia? This
can be calculated from the stability constant of the complex between ammonia and
Cd. The equilibrium of this reaction to form the cadmium tetraamine complex is
given by
      Cd2      4NH3 D Cd(NH3) 2
                              4                                                 (1.26)
and the stability constant of this equilibrium, Ks, by
              [Cd(NH3)2 ]
                      4
      Ks                         1.3   107                                      (1.27)
                 2
             [Cd ][NH3]4
As calculated previously, Cd(OH)2 will precipitate when the free Cd2 concen-
tration is larger than 2 10 8 M (at a pH of 11 and at room temperature). From
Eq. (1.27), for a total Cd concentration of 0.1 M, we can calculate that a free NH3
concentration of 0.79 M will result in a free Cd2 concentration of 2 10 8 M.
Add to this the ammonia tied up by complexation (0.1 4 M), the minimum NH3
concentration required to prevent precipitation of Cd(OH)2 is therefore 1.19 M. At
a Cd concentration of 0.01 M (more typical of many depositions), the corre-
sponding concentration of ammonia is ca. 0.5 M. At a deposition temperature of
60°C (CdS deposition is generally carried out at elevated temperatures, usually
  60°C), the ion product of water is 13, and therefore the OH concentration at a
pH 11 will be 10 2 M. Calculating the minimum concentration of ammonia re-
quired to prevent precipitation of Cd(OH)2 at 60°C and 0.01 M total Cd gives a
value of 1.44 M. The value of pH chosen is typical of these solutions. For higher
values of pH, and at higher temperatures at the same pH (both of which mean an
increased [OH ]), more ammonia will be required to prevent precipitation of
Cd(OH)2. This calculation ignored the decrease in the stability constant of the
complex with increasing temperature (see later) as well as the increase in Ksp that
normally occurs with an increase in temperature. These two effects act in opposite
directions; for most cases, their combined effect will be much smaller than that of
the temperature-dependent ion product of water. Another simplification is the as-
sumption of only one complex species; this simplification is reasonable for most
purposes.
       If a solution contains an excess of one of the ions of a sparingly soluble salt,
this will modify the solubility of the sparingly soluble salt according to the com-
mon ion effect. As an example of this effect, we might consider the precipitation



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of Zn(OH)2 by hydroxide according to the reaction
      Zn2      2OH D Zn(OH)2                                                    (1.28)
An excess of OH (the common ion) should shift the reaction to the right, i.e., to
more complete precipitation of the Zn(OH)2. This effect is a general one, but the
conclusions are not always valid; the example (deliberately) given here is one
where it is not valid. The reason is that OH can form a complex with Zn2
(Zn(OH)2 —the zincate ion), thus removing free Zn2 from solution and reduc-
          4
ing the degree of precipitation. For a sufficiently high concentration of OH ,
which can be calculated from the stability constant of the zinc–hydroxide (zincate)
complex, the Zn(OH)2 will completely redissolve.
       The stability constant of a complex is temperature dependent—increased
temperature generally leads to increased dissociation of the complex. Qualita-
tively, this can be explained by the Le Chatelier principle, which states that if there
is a change in a reaction parameter, the reaction will proceed in a direction that op-
poses that change. Thus an increase in temperature will cause the reaction to go in
the direction in which heat is absorbed, which is dissociation of the complex.
More quantitatively, the relation between equilibrium constant and temperature is
given approximately by
                 H0
      ln K              (   a constant)                                         (1.29)
                RT
where H0 is the standard enthalpy change in the process and R is the gas con-
stant. This is the integration of the van’t Hoff equation, hence the constant term.
(The derivation of this equation can be found in any elementary physical chem-
istry textbook and there is no need to repeat it here—the result is what is impor-
tant for us.) This equation is approximate for a number of reasons. One is that it
ignores changes in entropy that often will act in the opposite direction for complex
formation. However, the trend is generally correct.
       The stability constant of a complex does not, according to Eq. (1.27), depend
on the concentrations of the species comprising the complex. For very dilute solu-
tions, however, complexes become less stable than expected from their “literature”
stability complex. The reason for this lies in the fact that the equilibrium shown in
Eq. (1.26) is not strictly correct; a more accurate representation would be
      Cd(H2O)2
             6        4NH3 D Cd(NH3)2
                                    4           6H2O                            (1.30)
(hydrolysis and hydration of ammonia and ammonium ion is ignored, although for
an accurate representation, it should be considered—it will not affect the argu-
ment). Since ammonia is a much stronger ligand than water (water of hydration
can be considered as a ligand), it will exchange all the water as long as the am-
monia concentration is not too low. If it is very low, then not all the water will nec-
essarily be exchanged, and a different equilibrium (or mixture of equilibria) with



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
a different equilibrium constant will exist. A classic example of this effect is given
by the cobalt complex with thiocyanate, SCN :
      Co(H2O)2
             6        4SCN D Co(SCN)2
                                    4             6H2O                          (1.31)
          pink                          blue
                                    2
The pink color of the hydrated Co ion turns blue when a high concentration of
SCN is added; if diluted with water, this solution reverts to pink. This color tran-
sition is reversible. An aqueous solution of Co2 is pink, while anhydrous Co(II)
salts are typically blue (a fact well-known to chemists and even to schoolchildren
who have experimented with invisible inks). The color change to blue on addition
of SCN to a noncomplexed Co2 solution is caused by dehydration of the Co2
due to exchange of water with the SCN , a stronger ligand than water. However,
SCN is not a very much stronger ligand than water to Co2 , and therefore a rel-
atively high concentration is required to exchange all the water. At intermediate
concentrations of SCN , mixed aquo-thiocyanato complexes can be formed,
which are pink. The stronger the ligand relative to water, the less the concentra-
tion required to exchange the water.
       It is worth noting that a statistical effect (different combinations of the var-
ious complexants) may result in mixtures of complexants binding more strongly
to a cation than would be expected based on the individual stability constants of
the complexants [2].
       If a compound containing more than one cation is to be deposited, com-
plexation could be used to offset the difference in Ksp between the individual
metal compounds. As an example, consider the deposition of (Cd,Hg)S. From
Table 1.1, the value of Ksp for HgS is much lower than for CdS. This means that
under the conditions of CD, where the sulphide ion is slowly formed, we would
expect only HgS to form (until almost all the Hg was used up). Some Cd might be
incorporated into the deposit by adsorption, but the deposit should, according to
considerations of solubility product, be predominantly HgS.
       The concentration of Hg in the deposit can be decreased by choosing a com-
plex (or mixture of complexes) that complexes Hg more strongly than it does Cd. In
this case, since Hg forms very strong complexes with many ligands, there is a large
choice. Skyllas-Kazacos et al. deposited films of (Cd,Hg)S using a combination of
ammonia and cyanide (the latter is a strong complex for both cations, but more so
for Hg) [3]. In addition, the Hg concentration was much smaller than that of cyanide,
while the Cd concentration was larger. This means that there was enough cyanide
to complex the Hg but not enough for the Cd. A further factor that may have allowed
codeposition of Cd was the use of the chloride anion, which is a moderately strong
complex for Hg2 but only a weak one for Cd2 . This combination of factors al-
lowed codeposition of the Cd and Hg as sulphides, but the concentrations of Hg in
the films were larger (by a factor of ca. 4) than in the deposition solution. The addi-
tion of a large concentration of iodide to the deposition solution would probably



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
have been even better, since iodide is an extremely weak complexant for Cd and a
very strong one for Hg. It would therefore have removed most of the free Hg2 ions
while only changing the Cd2 concentration a relatively small amount.
       This judicious use of a complexant to allow codeposition of two cations
with widely differing values of Ksp is, unfortunately, not always useful. An exam-
ple is the deposition of (Cd,Zn)S—a material of interest particularly because of its
potential use in photovoltaic cells (see Chapter 9). The stability constants of Cd
and Zn ions are in most cases very similar for any particular complex (although,
of course, they vary greatly from one complex to another). This reflects the
similarity of the chemistry of these two ions (and the difference between them and
the Hg2 ion). Therefore complexation is very limited as a means to control the
concentration of one of these ions relative to the other.
       The use of complexation to allow codeposition of alloys is well known in
electroplating. The best-known example is that of brass (Cu/Zn) plating, where
cyanide, which is a stronger complex for Cu than it is for Zn, brings the deposi-
tion potentials of the two metals, originally far apart, to almost the same value.
There is a direct connection between this effect and the equivalent one for CD.
This arises from the fact that, for both CD and electrodeposition of alloys (we in-
clude mixed metal compounds in the term alloy), the effect of the complexant is
to lower the concentration of free cations. For CD this affects the deposition
through the solubility product, while for electrodeposition it affects the deposition
potential through the Nernst equation:
                   RT      [Ox]
       E E0            ln                                                      (1.32)
                   nF     [Red]
where the oxidized species, Ox, the cation in this case, is reduced in concentration,
resulting in a more negative deposition potential, E, compared to the standard po-
tential, E0. In the case of metal electrodeposition, the reduced species, Red, is the
metal that, since it is a solid, can be taken as unity concentration. n is the number
of electrons transferred per molecule of reaction (e.g., for Cd deposition from
Cd2 , n 2) and F is Faraday’s constant (ca. 96,500 coulombs/mole).
       The shift in potential due to complexation, E, ( E E 0) can be approxi-
mated by
        E (in mV)     60 log [cation]/n                                       (1.33)
From Eq. (1.27), we can write
      log Ks    log [complexed cation]      log [cation]   a log [ligand]     (1.34)
where a is the number of ligand molecules in the complex.
Combining Eqs. (1.33) and (1.34) we get
      log Ks      n E (in mV)/60
                                                                              (1.35)
                  log [complexed cation]      a log [ligand]



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
For all but very weak complexes, the concentration of complex ion (and often also
the free ligand) is normally very much larger than that of the free cation. With this
in mind, Eq. (1.35) can often be approximated by
      log Ks      n E (in mV)/60                                                (1.36)
This [or, more accurately, Eq (1.35)] allows us to calculate values of the stability
constant of a complexant from tables of electrochemical potentials. For example,
a shift of 300 mV in potential due to complexation gives an (approximate) value
for the stability constant of that complex of 105 (for n 1) or, for the more com-
mon case in CD, where divalent cations (n 2) predominate, 1010.


1.2 NUCLEATION AND GROWTH
CD can occur either by initial homogeneous nucleation in solution or by het-
eronucleation on a substrate, depending on the deposition mechanism (see Chap-
ter 3). For this reason, we consider both types of nucleation.

1.2.1 Homogeneous Nucleation
According to simple solubility considerations, a precipitate will be formed when
the product of the concentrations of anions and cations exceeds the solubility
product. From another viewpoint, phase transformation occurs when the free en-
ergy of the new phase is lower than that of the initial (metastable) phase. However,
there are many examples where the ion product exceeds Ksp, yet no precipitation
occurs—the phenomenon of supersaturation. The solubility product also does not
provide information on how the particles of the precipitate form—nucleation. Nu-
cleation involves various physical processes, and both thermodynamic and kinetic
aspects must be considered.
       Homogeneous nucleation can occur due to local fluctuations in the solution—
whether in concentration, temperature, or other variables. The first stage in growth is
collision between individual ions or molecules to form embryos (embryos are nuclei
that are intrinsically unstable against redissolution—see later). Embryos grow by col-
lecting individual species that collide with them. While these species may be ions,
atoms, or molecules in general, for CD, adsorption of ions on the embryo seems to
be the most probable growth mechanism. They may also grow by collisions between
embryos; however, unless the embryo concentration is large, this is less likely.
       These embryos may redissolve in the solution before they have a chance to
grow into stable particles (nuclei). Because of the high surface areas, and there-
fore high surface energies of such small nuclei, they are thermodynamically un-
stable against redissolution. They may, however, be kinetically stabilized by low
temperatures, which increase their lifetime, possibly enough for them to grow to
a size where they are thermodynamically stable. This is an important reason why



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
smaller particles can be formed at lower temperatures in a precipitation reaction;
the subcritical embryos last long enough to grow into stable particles, while at
higher temperatures they would redissolve, reducing the density of nuclei. This re-
sults in an increase of the particle size, since there is more reactant per nucleus.
       The critical radius, Rc, is the size where the embryo (nucleus) has a 50:50
chance of either redissolving or growing into a stable nucleus; it is determined by
the balance between the surface energy required to form the embryo,
      Es    4 R2        ( is the surface energy per unit area)                 (1.37)
and the energy released when a spherical particle is formed,
      Ev    4 R3 L/3
                                                                               (1.38)
            ( is the density of the solid and L is the heat of solution).
This balance is shown in Fig. 1.1. The typical size of Rc is about 100 molecules—
between 1 and 2 nm in diameter. Solvent molecules can adsorb on the embryos
and change their surface energy; the critical radius will therefore depend not only
on the material of the nucleating phase but also on the solution phase.

1.2.2 Heterogeneous Nucleation
In heterogeneous nucleation, subcritical embryos (or even individual ions) can
adsorb onto the substrate. The energy required to form an interface between the




FIG. 1.1 Energetics of nucleation. The critical radius, Rc, depends on the balance be-
tween surface and volume energies of the growing particle.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 1.2 Processes involved in heterogeneous nucleation on a surface.



embryo and the solid substrate will usually be less than that required for homoge-
neous nucleation, where no such interface exists. Therefore heterogeneous nucle-
ation is energetically preferred over homogeneous nucleation and can occur near
equilibrium saturation conditions, compared with the high degree of supersatura-
tion often required for homogeneous nucleation. These subcritical nuclei can
grow, either by surface diffusion or by material addition from solution. It should
also be noted that nuclei that are subcritical in solution may be supercritical when
adsorbed on a substrate. This is a consequence of reduced contact between nucleus
and solution as well as stabilization of the adsorbed nucleus. These processes are
shown schematically in Fig. 1.2.
       It was noted earlier that even individual ions may adsorb onto a surface.
More specifically, depending on the surface chemistry of the substrate, individual
ions or molecular species may actually be chemisorbed, creating a nucleus for re-
action and further growth.
       Pure homogeneous nucleation is probably less common that might appear
from the above discussion. Because of the greater ease of nucleation on a solid
phase than homogeneously, any solid matter in the solution will act as a nucleation
center. It is difficult to prepare solutions without some solid phase (usually dust
particles)—careful filtering is necessary to attain such particle-free solutions. That
this is so can be seen from the simple test of shining a laser beam (preferably a
green or blue laser, since scattering is greatly enhanced compared with a red one)
through an visibly “clear” solution; the resulting scattering by dust particles is al-
most always evident.

1.2.3 Crystal Growth
Once (stable) nuclei have formed, there are several ways in which they can in-
crease in size. One is a continuation of the process of embryo growth discussed
earlier: adsorption of ionic species from the solution onto the nucleus. Crystal



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
growth of this type can be considered a self-assembling process. Thus for CdS, ei-
ther Cd2 or S2 will adsorb (as discussed later, since a crystal, and in particular
a polar one, is made up of different faces, the adsorption properties of each may
be different, and therefore both types of ions may adsorb on any one crystal). The
next growth step will then be adsorption of the oppositely charged ion to give an
additional CdS molecule. This process can continue until either all the ions of any
one type are used up or growth is blocked, e.g., by aggregation or by blocking of
the crystal surface by a foreign adsorbed species. Also, growth may continue but
in a different geometric orientation, giving rise to twinning, polycrystallinity, etc.
       Another mechanism for crystal growth is known as Ostwald ripening. If a
small nucleus or embryo is close to a larger crystal, the ions formed by (partial)
dissolution of the smaller, less stable crystal can be incorporated into the larger
crystal. As the smaller crystal becomes even smaller, its dissolution will become
ever more favorable and eventually it will disappear. The result is that the larger
crystals grow at the expense of the smaller ones.
       If the concentration of particles is sufficiently high, then the probability
of collisions between these particles becomes high. This can result in either
aggregation or coalescence. When two particles approach each other, the van der
Waals force of attraction (see section 1.3.1) between them will often cause them
to stick together. This can continue until a large particle (large in relation to the
original particle size) comprising the individual particles has formed (Fig. 1.3A).
This is the process of aggregation, and the resulting large particle is called an
aggregate. (In colloidal chemistry, the alternative terms of flocculation and
floc are often encountered.) The properties of the aggregate may be similar
to those of the individual particles in some ways (such as X-ray diffraction




FIG. 1.3 Aggregation (A) and coalescence (B) of individual particles.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
peak broadening, quantum size effects) and very different in others (e.g., light
scattering, sedimentation).
      In an aggregate, there are grain boundaries between individual crystallites.
However, in some cases, particularly if the temperature is high enough to allow
appreciable diffusion of the crystal atoms, surface diffusion may occur where two
(or more) particles have aggregated, resulting in the formation of a neck. This is
termed coalescence. Coalescence may continue until one large particle is formed
from the original two or more particles (Fig. 1.3B).

1.2.4 Particle Size Distribution
If nucleation occurs in a very short time, whereas growth occurs separately, often
over a much longer time but without further nucleation, then the size distribution
is likely to be narrow, since all the original nuclei should be of similar size and
grow at the same rate. The opposite case, where nucleation and growth occur si-
multaneously, usually results in a wide size distribution.
       Homogeneous nucleation normally requires a supersaturated solution, while
growth can occur close to the saturation concentration. Therefore rapid nucleation
can occur if supersaturation is rapidly reached. This nucleation lowers the con-
centration of reactants below that needed to cause further nucleation. If one of the
reactants is supplied at a low concentration after nucleation has occurred (such as
by in situ homogeneous formation in the solution), then growth can occur without
further nucleation, resulting in a narrow size distribution.
       In CD, where the reaction is slow, it might be expected that nucleation and
growth will always occur together, resulting in a relatively wide size distribution.
This is indeed expected for heterogeneous nucleation on a substrate. However, for
mechanisms where homogeneous nucleation of an intermediate phase occurs
rapidly in the solution but conversion to the final compound is a slow process, nu-
cleation and growth can still be separated.


1.3 FORCES BETWEEN PARTICLES AND
    BETWEEN PARTICLES AND SURFACES
Once nanoparticles have been formed, whether in an early state of growth or in a
more or less final size, their fate depends on the forces between the individual par-
ticles and between particles and solid surfaces in the solution. While particles ini-
tially approach each other by transport in solution due to Brownian motion, con-
vection, or sedimentation, when close enough, interparticle forces will determine
their final state. If the dominant forces are repulsive, the particles will remain sep-
arate in colloidal form. If attractive, they will aggregate and eventually precipitate.
In addition, they may adsorb onto a solid surface (the substrate or the walls of the
vessel in which the reaction is carried out). For CD, both attractive particle–sur-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
face and particle–particle forces are required for film formation. In all the stages
of the CD process, except for a very few studies, that of adhesion of the film to the
substrate is probably the least understood; it is rarely even considered in mecha-
nistic studies. Why do some particles stick to a particular substrate and others not?
Also (and easier to understand, at least intuitively), why do the particles stick to
each other in building up a film?
       In order to understand these sticking phenomena so crucial for the CD pro-
cess, we consider the various forces involved—repulsive as well as attractive—
involving the particles. We discuss first the more obvious forces and then some
less obvious ones that nonetheless may be important in some cases. Since the
dominant force in CD is the van der Waals attraction, we will begin with this in-
teraction.

1.3.1 van der Waals Forces
The main interaction that determines whether and to what extent particles will ad-
here to each other and also (if there is no specific chemical interaction) to a sub-
strate is, in most cases, the van der Waals interaction. The van der Waals force of
attraction is a universal interaction that operates between all particles, whether
atoms, molecules, clusters, charged, or noncharged. The attraction is due to an in-
duced dipole–induced dipole interaction between particles. The dipoles arise from
fluctuations in electron density around the ion cores, resulting in transient changes
in the charge density distribution. This transient dipole in one particle induces an
equal and opposite dipole in the other one, resulting in an attraction. It may be
thought that all the transient dipoles in the randomly orientated particles would
cancel each other and average to zero. This is indeed the case for an ensemble of
particles. However, the correlation between the dipole in one particle and the in-
duced dipole in another at any time is not zero—the correlations, and therefore the
attractive interactions, do not average out to zero. Such charge fluctutations are a
universal property of matter and occur even in a completely nonpolar material. If
there is a permanent dipole in (some of) the interacting particles, these dipoles will
also contribute to the van der Waals interaction. For purely nonpolar particles, the
interaction is known as the London, or dispersion, energy.
       The van der Waals interaction between atoms or molecules, E, varies as the
inverse sixth power of the distance, d, between them:
            1
      E                                                                        (1.39)
           d6
(The minus sign signifies an attractive interaction.) For macroscopic (this includes
microscopic and nanoscopic) bodies, this interaction is much less short-range, and
the distance dependence varies both with the geometry of the interacting bodies
and with the distance of separation. For macroscopic bodies, it is usually assumed



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 1.3 Interaction Energies and Forces of Attraction Between Two
Bodies with Different Geometries

Geometry                               Energy                             Force

                                      16Ar 3r 3
                                           1 2
                                                                            3 3
                                                                       32Ar1 r 2
                                        9d 6                             3d 7




                                      16Ar6                            32Ar 6
                                       9d 6                             3d 7

                                         Ar1r2                             Ar 1r 2
                                      6(r1 r2)d                        6(r 1 r 2)d 2

                                       Ar                               Ar
                                      12d                              12d 2

                                      2Ar3                             2Ar3
                                      9d3                              3d 4

                                      Ar                                Ar
                                      6d                               6d 2
A is the Hamaker constant, r (r1, r 2) the radius of the spherical particles, and d the
distance between surfaces of the two bodies. Note that larger particles will interact
more strongly (more adherent films?).


that the interactions between all the different bodies are additive. Table 1.3 shows
values of this interaction for various geometries of two interacting bodies relevant
to CD. These can be divided into particle–particle attraction (formation and
growth of aggregates) and particle–plane surface (i.e., the substrate) interaction.
Two different distance scales are shown, depending on whether the separation is
considerably larger or smaller than the radius of the particle. Clearly there will
also be intermediate separations, where the separation and radius are comparable,
with intermediate dependence on the separation. Cases where two interacting
spheres are identical or of different size are also shown. Both are relevant for CD,
where initial aggregation will occur between elementary colloids of approxi-
mately the same dimensions but further interaction can occur between two parti-
cles of very different sizes.
       The Hamaker coefficient, A, is a measure of the interaction and is dependent
on the material of the particle as well as on the surrounding medium. Heavy atoms,
which are generally more polarizable (i.e., the electron distribution can be more



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
easily perturbed), generally have a larger value of A than lighter atoms and there-
fore a greater attractive interaction. A has units of energy and values that vary typ-
ically from several times the thermal energy, kT, to several tens of kT in air or a
vacuum and typically an order of magnitude less in liquid media.
       At large distances between particles, correlation between fluctuations in one
particle and the induced dipole in another breaks down. This occurs when the time
taken for the interaction (acting at the speed of light) is comparable to the charac-
teristic scale of the electron fluctuations, viz. the plasma frequency. The plasma
frequency ranges typically from 10 eV down to 2 eV (closer to the former for
many dielectrics and to the latter for metals), which translates into a length scale
of between 600 and 100 nm. At this distance scale, the (at this point, very weak)
van der Waals forces are termed retarded forces, because of the appreciable time
required for the transmitting dipole electromagnetic field to reach the receiving
species. At this distance, the attraction falls off approximately as the inverse sev-
enth power of the distance. It is probable that diffusional and convective motion
and electrostatic interactions will dominate at such distances and the van der
Waals interaction will be negligible in most cases.
       These various relationships between force and particle separation imply that
the attractive force between particles will become infinite when they touch. In re-
ality, other short-range forces will modify this relationship when r is very small,
in particular the repulsion from overlap of atomic orbitals. The van der Waals at-
traction will then be balanced by this overlap repulsion. At these short distances
(a few tenths of a nanometer), the van der Waals attraction will be strong enough
to hold the particles fairly strongly together. This balance between van der Waals
forces of attraction and overlap repulsion forces is shown schematically in Fig.
1.4, where the very steep repulsive interaction at atomic distances is due to the
overlap repulsion. Hydration forces (see section 1.3.3) may also result in repulsion
between surfaces at somewhat greater separations.
       Particle adhesion occurs when the distance between bodies is that of an
atomic spacing. From Table 1.3, the force between a sphere of radius r and a flat
surface at close approach is
        Ar
                                                                               (1.40)
       6d 2
At contact, d is the atomic spacing. For a solid where the van der Waals forces
dominate, the work needed to separate two unit areas from contact to infinity is
given by
        A
                                                                               (1.41)
      12 z2
where z is the atomic spacing [ d in Eq. (1.40) at contact]. This is the energy re-
quired to produce two new surfaces, i.e., 2 , where is the surface energy of the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 1.4 Resultant interaction energy between two particles with van der Waals attrac-
tive interactions and electron overlap repulsion interactions.



solid. Therefore,
             A
                                                                               (1.42)
           24 z2
Substituting Eq. (1.42) into Eq. (1.40) gives the force of adhesion at contact in
terms of the surface energy:
      F    4 r                                                                 (1.43)
This relation is clearly very simplified, being based on a number of approxima-
tions, such as the validity of the use of the Hamaker constant at such close dis-
tances and the particle and surface being of the same material. Also, the relation-
ship between surface force and van der Waals forces does not hold for many
solids, in particular for metals where metallic bonding is important. Nonetheless,
if taken as an indication of the forces holding particles to each other and to sur-
faces, it does give a feel for these forces.

1.3.2 Electrostatic Forces
For solid particles dispersed in a liquid medium, there exists, in most cases, a layer
of charge separation at the phase boundary—the electrical double layer. A num-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ber of processes can cause this double layer. One of the most common is adsorp-
tion of charged species at the solid–liquid interface. For example, in a colloidal sol
of CdS prepared by precipitation of a Cd salt with sulphide ions, Cd2 or S2
(HS ), depending on which is in excess, will adsorb at the CdS surface. Another
mechanism for formation of surface-charged species is surface dissociation. A
common example of this is the case of metal oxides in water; the water may dis-
sociate at the oxide surface as follows:


                        D MOH          OH
      MMO        H 2O   D                                                       (1.44)
                            MOMOH           H

leaving a positively or negatively charged surface. Such a reaction will, obviously,
be very pH dependent.
       Yet another possibility for formation (or change) in a double layer is by ac-
cumulation of one charge type. This may occur by doping the solid with an ion of
valence different from that of the solid (e.g., In3 in CdS) or by illumination with
super-bandgap illumination. In the first case, the (in this case) electron donor
(In3 ) is immobile, while the donated electron is mobile; if the electron is trans-
ferred to the liquid, then the solid will become positively charged. The same oc-
curs for an illuminated solid where electron/hole formation occurs, if one of the
charges is preferentially injected into the liquid. Even in the absence of charge in-
jection into the liquid, localized (e.g., trapped) charges will affect the double layer
if the countercharge is (relatively) delocalized over the particle. The double layer
close to a (near) surface-localized donor will be different than that for the rest of
the particle. Such an effect is probably not important in large particles, where such
fluctuations can be evened out. For very small particles, however, where only a
single “dopant” may exist, this effect may be appreciable. Even for a pure and per-
fectly stoichiometric particle, the double layer need not be homogeneous around
the particle. For example, a CdS particle will consist of different crystal faces.
Most notably, the opposite polar faces, consisting of only Cd or of S atoms, can
be expected to possess different double layers. Incidentally, these polar faces
might be expected to attract other polar faces of opposite polarity and repel those
of the same polarity. Such an effect would lead to some form of self-assembly.
However, any effect of this nature will be much smaller (if it exists at all) in solu-
tion compared to vacuum or air, due to adsorption of ions from solution onto the
polar faces, which will tend to neutralize this effect.
       The charge at the surface of a solid (including any adsorbed species) will be
balanced by a countercharge in the electrolyte; the double layer as a whole is elec-
trically neutral. The countercharges remain in the vicinity of the surface-adsorbed
charge but, due to thermal motion, do not accumulate at the surface but move in a
more or less diffuse cloud surrounding the particle. The extent to which this layer



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of countercharge extends into the solution—how diffuse it is—depends on the
concentration of charged species in the solution. This gives the solution screening
length. The screening length is analogous to the space charge layer width in a
semiconductor; in the same way, the screening length, , is a function of the
square root of the charge (ionic) concentration.
             2e2z 2n   1/2
               0                                                                  (1.45)
                 kT
where z is the ionic charge, n is the ionic concentration, 0 is the permittivity of
free space, and is the dielectric constant of the material. (The dielectric constant
is normally taken as constant. It should be pointed out, however, that for nanopar-
ticles of several nanometers or less, the value of decreases with particle size, as-
suming the particles are in a medium of smaller dielectric constant than the parti-
cles themselves, a reasonable assumption for our purposes. This effect is treated
in some detail by Lannoo et al. [4].) The thickness of this diffuse layer (also
known as the Gouy layer) is the inverse of ; i.e., the potential drop across the dif-
fuse layer, d, decays to d/e (where e in this case is the natural exponent, 2.718)
over a distance 1.
       Figure 1.5 shows a schematic representation of the double layer at a planar
solid–liquid interface. The potential drop across the Helmholz layer is shown as
linear (in the presence of specific adsorption, it will not be completely linear), fol-
lowed by a tailing-off of the potential into the diffuse layer. For concentrated so-
lutions ( 0.1 M) the diffuse layer is typically a nanometer or less, while for dilute
solutions it may be tens or even hundreds of nanometers.




FIG. 1.5 Schematic diagram of the electrical double layer at a solid–liquid interface.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       The crucial importance of the double layer when dealing with colloidal par-
ticles dispersed in a solution is due to the repulsion of one particle by another.
While overall the particles are neutral, because the diffuse layer can extend into
the solution, the unbalanced charge in the diffuse layer of one particle experiences
a repulsion by that of another particle. Normally, from the Coulomb law of elec-
trostatics, the force between two equal (in charge and in sign) particles is given by
               q1q2
       F 4                                                                     (1.46)
                  0 d

where F is the force acting between two charges, q1 and q2, separated by a distance
d. Because of the presence of the diffuse layer, however, the repulsion force be-
tween two particles is strongly dependent on the screening length, , and is ap-
proximately proportional to exp (- d); the force of repulsion between two colloids
will decrease exponentially with distance.
       Overlap of the diffuse layers of approaching particles prevents them from
getting close to each other, and the particles form a stable colloidal solution. More
relevant for our purposes, for a moderately concentrated electrolyte of the type
normally encountered in CD (on the order of 0.1 M or more total concentration),
this diffuse layer is around 1 nm or less. The diffuse layer screens the surface charge
and allows the like-charged particles to approach each other closely, to the point
where the van der Waals forces of attraction dominate, causing aggregation. This
is the basis of salting-out of a colloid; addition of a strong electrolyte to the colloid
reduces the thickness of the diffuse layer, allowing closer approach of the particles
to each other and eventual aggregation and precipitation. The competition between
the attractive van der Waals and repulsive electrostatic forces and the importance of
this competition in colloid stability is known as the DLVO theory, named for the
scientists who developed a theoretical analysis of the overall interaction (Derjaguin,
Landau, Verwey, and Overbeek). The resultant interaction is shown in Fig. 1.6. For
a dilute electrolyte, there is a relatively large barrier to aggregation where the
double-layer repulsion dominates the interaction to the greatest extent and the
interacting particles fall into what is known as the secondary minimum. At this
point they are kinetically stabilized against aggregation. As the electrolyte concen-
tration increases, the barrier becomes smaller and eventually disappears, resulting
in the particles becoming trapped in the primary minimum, i.e., aggregation.
       It is worth noting that the electrostatic force can be attractive as well as re-
pulsive, depending on the sign of the two charges. For the case where a single col-
loidal species is present in the solution, it will be repulsive, since all the particles
will have the same charge. Two different colloidal species of opposite charge
could conceivably be present in the deposition solution, either because the CD
process involves conversion of one species into another (e.g., metal hydroxide and
metal chalcogenide) or because two or more different cations, or even different
valence states of the same cation, are present.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 1.6 DLVO interactions showing the energetics of colloidal particles as a competi-
tion between electrostatic double-layer repulsion and van der Waals attractions. The pri-
mary minimum is due to strong short-range electron overlap repulsion (shown in Figure 1.4
but not shown here).


1.3.3 Entropic and Other Short-Range Forces
Apart from double-layer, van der Waals, and electron-overlap interactions, any
two bodies in a liquid medium (even if only one, or even neither, is charged) will
experience a (usually) repulsive component of force as they approach each other.
This is due essentially to entropic considerations. As two bodies approach each
other very closely, the species in solution have increasingly less room in which to
move; the entropy of these species therefore decreases, producing a repulsive
force between the two bodies. In its simplest form, this force is usually considered
to exist between two infinite flat plates. In the context of forces between colloidal
particles of the type common in CD, this entropic force should be less important,
since the solution species between two approaching particles can be relatively eas-
ily pushed out of the intervening space, both due to the small size and due to the
approximately spherical shape common in these systems. In fact, in this case there
may even be a weak attraction—the depletion interaction—due to the smaller den-
sity of solution species in the space where the two particles are closest to each
other (from which they have been pushed out) and the surrounding solution. The
entropic force of repulsion may be important, however, when considering the
sticking behavior of larger aggregates at plane surfaces.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       This entropic force is important where adsorption of polymers occurs on
colloidal particles. This is due to interaction between polymer chains on the inter-
acting particles: As the particles approach each other to the point where the poly-
mer chains of the two particles interact, there is an decrease in entropy due to con-
finement of the chains, in an analogous manner to the solution species discussed
earlier, with the same result—repulsion. This is the basis of polymeric stabiliza-
tion of colloids; it is generally undesirable in CD, since adhesion and aggregation
are preferred in this case. However, in view of the fact that the presence of such
polymers (and other stabilizing adsorbates) may prevent the aggregation needed
to build up a CD layer, it is important to be aware of the effect.
       There are other close-range forces related to entropy changes, including var-
ious interactions between solution species and a solid surface, such as solvation
(in water, hydration) forces. Hydration forces can occur when hydrated cations are
adsorbed at interacting surfaces. As these surfaces approach each other closely,
loss of water of hydration is necessary in order to allow closer approach. While
these forces can be repulsive, attractive or oscillating, they are most likely to be
repulsive under the conditions of CD. Such forces may be very important for CD,
which is almost always carried out in the presence of a high ionic concentration.
For example they could be a cause of poor adhesion of some CD films. Solvation
forces are treated in detail in Israelachvili’s book—see Further Reading at the end
of this chapter, “Forces” subsection.
       If this treatment of forces between particles fails to convince the reader that
it is natural for particles to stick together, one can resort to the more intuitive ap-
proach. It is well known that inorganic colloids require a stabilizing agent to pre-
vent their sticking together and eventually precipitating. In other words, precau-
tions usually have to be taken to prevent the natural tendency of these particles
eventually to stick to each other.

1.4 CHARACTERIZATION TECHNIQUES—
    SOME CAUTIONS
Many techniques have been used to characterize CD films. The purpose of this
section is not to review all these techniques, but only to draw attention to some of
them that are sometimes misinterpreted.
       Three common techniques used are transmission electron microscopy to-
gether with electron diffraction, powder X-ray diffraction, and optical absorption
(or transmission) spectroscopy.

1.4.1 Transmission Electron Microscopy/Electron
      Diffraction and X-Ray Powder Diffraction
Transmission electron microscopy (TEM) is used to image nanocrystal (lateral)
size, shape and size distribution. Electron diffraction (ED) provides information



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
on the composition of the deposit, crystal phase, and orientation. X-ray diffraction
(XRD) also provides similar information on composition (more accurately than
ED) and phase as well as crystallite orientation. In the last, ED is superior in many
ways, since a much smaller area can be selected (selected area diffraction—SAD)
and, in addition, azimuthal lattice alignment between deposit and substrate (epi-
taxy) can be determined from ED but not from the commonly used powder XRD
measurements; powder XRD reveals texturing (one particular crystal face parallel
to the substrate for all crystals for perfect texturing) but not orientation (crystal lat-
tices in any direction parallel to the substrate of all crystals aligned in the same
way). There are XRD measurements that can distinguish orientation, but while be-
coming somewhat more common, these are still rather infrequently used, at least
in CD studies, compared to the normal “ –2 ” measurement.
       CD films are often nanocrystalline. One very important use of XRD when
dealing with nanocrystals is to estimate crystal dimensions through the Scherrer
relationship:
                                    1.3
       crystal diameter                                                                       (1.47)
                                  (2 ) cos
where is the X-ray wavelength (0.1541 nm for Cu K radiation, a commonly
used source), (2 ) is the peak full width at half maximum (FWHM) in radians,
and is the peak position.* The shape of the crystal can also modify this relation-
ship, which is valid for a spherical crystal (close to the shape often encountered).
       As a rough and useful rule of thumb, a peak FWHM of 1° at an angle of
2      25° (a common approximate position) means a crystal size of ca. 12 nm (for
Cu K radiation), and the size is inversely proportional to the FWHM. Actually, to
be more precise, what is measured is not necessarily crystal size but coherence
length, the length over which the periodicity of the crystal is complete. An exam-
ple of a coherence length smaller than the crystal size is a twinned crystal; XRD
measures the size of each individual twin. Other causes for XRD peaks being
broader than expected based on crystal size is the presence of strain in the crystals
or other defects, such as dislocations, which destroy the long-range lattice order.
Separation between crystal size and strain can be made if several different peaks
are present, since the angular dependences of the two factors are different (see Ref.
5 for an example of this). Thus, the interpretation of XRD peak broadening should
be carried out with care and preferably using complementary TEM measurements.
The opposite case, where the peaks are narrower than expected based on crystal
size, does not occur; a narrow peak means a (relatively) large coherence length and


* Note that in the common /2 measurement, the in cos is half of the 2 value. For example, if the
peak being measured is at 2      25°, cos will be cos 12.5°. For small angles, the error in taking cos
2 instead of cos is not too large ( 10% at 2        25°, for example). However this error becomes
larger as the peak angle increases.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
therefore crystal size. However, even here, interpretation is not always straightfor-
ward. For example, the XRD pattern of a deposit of tall cylinders of small cross
section will give a peak width characteristic of the height but not of the cross sec-
tion (the latter will be seen in TEM images). Thus the TEM and XRD sizes will
necessarily be different in such a case. Another example is where there is a mixture
of large and small crystals. Even if the large crystals constitute a relatively small
fraction of the total material, they may in some cases dominate the XRD pattern,
since peak heights decrease with decreasing crystal size due to increase of peak
width and (ideally) constant peak area for the same quantity of material.
       If the crystal size becomes very small (a few nanometers), the XRD peaks
will be very wide and also relatively weak. There is no shortage of examples in the
literature where samples have been classified as “amorphous” or “poorly crys-
talline” either on the basis of the lack of any XRD peak or because the peak(s)
were very broad. When carrying out an XRD measurement on a CD film, in par-
ticular, a particularly thin one (some tens of nanometers or less), if no peaks are
seen in the measurement, it is advisable to repeat the measurement over a narrow
range (where a major peak is expected) and with a very slow scan (e.g., 10°/hr or
even slower). If a thin-film attachment is available, this will reduce the likelihood
of such misinterpretations. It is useful to remember that except for compounds that
are commonly amorphous, CD semiconductor films are rarely truly amorphous.

1.4.2 Optical Absorption
Optical absorption spectroscopy is often carried out on CD films to verify that the
films have a bandgap expected from the deposited semiconductor. Additionally,
since CD films are often nanocrystalline and the most apparent effect of very
small crystal size is the increasing bandgap due to size quantization (the effect is
visible to the eye if the bandgap is in the visible region of the spectrum), absorp-
tion (or transmission) optical spectroscopy is clearly a fast and simple pointer to
crystal size, since bandgap–size correlations have been made for a number of
semiconductor colloids and films.
      There are some potential problems that should be taken into account when
interpreting such spectra. A spectrophotometer measures transmission (and maybe
also reflection) but not absorption. What is measured as absorption is a transmis-
sion measurement that is mathematically manipulated to convert it to absorption.
Absorption is usually measured as absorbance, A, which by definition is given by
      A    log10 Io/I                                                           (1.48)
where I is the intensity of the transmitted light and I0 that of the incident light.
     The transmission, T, is
            I
      T                                                                         (1.49)
           I0



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
The spectrophotometer measures the transmission and, if an absorption measure-
ment is carried out, converts the transmission into absorbance using these equa-
tions. This conversion works fine for samples where there is no reflection, either
specular or diffuse, as is the case for nonturbid solutions. However, for films there
is invariably some reflection, which is often quite large, particularly for films of
high dielectric constant (or refractive index) materials, such as PbS and PbSe. Ad-
ditionally, if the films are not completely transparent, then scattering introduces
an extra element of reflection. Therefore, to measure the real absorption of a film,
a reflection measurement must also be carried out and correction for this reflec-
tion made. The correction will be approximate and depends on the nature of the
film itself. However, that most commonly used is
                          T
      Tcorr                                                                      (1.50)
                  1           R
where Tcorr is the corrected transmission, T is the measured transmission, and R is
the reflectance. This correction neglects reflection from the film/substrate inter-
face (assuming the front face of the film faces the illumination source), and it can
be calculated that this will give a value of Tcorr that is too small (therefore an ab-
sorption that is too high). Use of (1 R)2 in the denominator of Eq. (1.50) in-
cludes this reflection but tends to give a value of Tcorr that is too high. For very
thin films, where reflection cannot be assumed to originate from a surface (i.e.,
where the film thickness is not much greater than the depth from which reflection
can occur), the calculation is more complicated. Fortunately, reflection from such
thin films is also normally low, and therefore the correction is less important.
       To collect scattered transmission and correct for diffuse reflectance, a spec-
trophotometer with an integrating sphere should be used. This is important if films
are not very transparent.
       In many cases, the lack of correction for reflection will not affect the shape of
the optical spectrum very much, but it will just give an inflated value for absorption.
However, there are frequently cases where the shape of the absorption spectrum is
also appreciably changed after the reflection correction is carried out. Also, if the
primary absorption is a weak one, then correction for reflection, and in particular for
scattering, is crucial, since the absorption may be masked by these effects.
       The absorption coefficient, , of the semiconductor can be derived from the
absorption (or transmission) spectrum according to the Beer–Lambert equation
applied to solids:
                      t
      I    I0 e                                                                  (1.51)
where t is the film thickness.
      For bulk semiconductors, the relationship between derived from the ab-
sorption spectrum and the semiconductor bandgap, Eg, is given by:
      ( h )n          C(h         Eg)                                            (1.52)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
where n is 2 for a direct transition and 0.5 for an indirect one and C is a constant.
A plot of ( h )n vs. h should then give a straight line (over much of the absorp-
tion onset region), which extrapolates at a zero value of ( h )n to the value for Eg.
This calculation is based on the density of states in the valence and conduction
bands of the bulk semiconductor. For semiconductors in the quantum size regime,
however, the density of states may be quite different than in the bulk. Addition-
ally, a large size distribution, meaning distribution of bandgaps, will smear out the
onset. In practice, however, this extrapolation often appears to give a reasonable
value of the bandgap. The important thing is to be aware of the limitations of the
measurement. Just because a few points on a plot based on Eq. (1.52) give a
straight line does not automatically mean that the bandgap can be obtained from
the extrapolation of this line; not too infrequently, a weaker absorption onset at
longer wavelengths has been ignored, although consideration of the entire spec-
trum, together with the expected behavior of the material, would lead one to con-
clude that this weaker absorption determines the bandgap.

FURTHER READING
Solution Chemistry
Just about any textbook on inorganic chemistry. A particularly useful one, which
the author referred to on many occasions while writing this chapter, is:
G Wulfsberg. Principles of Descriptive Inorganic Chemistry. Mill Valley, CA: University
  Science Books, 1991.

For more extensive tables of solubility product than given here, Lange’s Hand-
book of Chemistry (J.A. Dean, 11th ed. New York: McGraw-Hill, 1999) gives an
extensive list. The standard CRC Handbook of Chemistry and Physics (CRC
Press) also gives a useful, if less extensive, list. Reference 1 also provides an ex-
tensive list of sulphides.

Nucleation and Growth
HC Freyhardt, ed. Crystals: Growth, Properties, and Applications. New York: Springer-
 Verlag, 1983, Vol. 9 (this is more mathematical than the other treatments).
HK Henisch. Crystal Growth in Gels. University Park: Pennsylvania State University
 Press, 1973.
AE Nielsen. Kinetics of Precipitation. New York: Pergamon Press, 1964.
BN Roy. Crystal Growth from Melts. New York: Wiley, 1992.

Forces
F Evans, H Wennerström. The Colloidal Domain: Where Physics, Chemistry, Biology, and
  Technology Meet. New York: VCH, 1994.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
PC Hiemenz. Principles of colloid and surface chemistry. New York: Marcel Dekker, 1986.
J Israelachvili. Intermolecular and Surface Forces. Orlando, FL: Academic Press, 1992.
H Ohshima, K Furusawa, eds. Electrical Phenomena at Interfaces: Fundamentals, Mea-
   surements, and Applications. New York: Marcel Dekker, 1998.


REFERENCES
1.   S Licht. J. Electrochem. Soc. 135:2971, 1988.
2.   P O’Brien, DJ Otway, D Smith-Boyle. Thin Solid Films 361:17, 2000.
3.   M Skyllas-Kazacos, JF McCann, R Arruzza. Appl. Surf. Sci. 22/23:1091, 1985.
4.   M Lannoo, C Delerue, G Allan. Phys. Rev. Lett. 74:3415, 1995.
5.   SB Qadri, JP Yang, EF Skelton, BR Ratna. Appl. Phys. Lett. 70:1020, 1997.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2
General Review




2.1 A (BRIEF) HISTORY OF
    CHEMICAL DEPOSITION
Chemical deposition (CD) of films is not a new technique. As early as 1835,
Liebig reported the first deposition of silver—the silver mirror deposition—using
a chemical solution technique [1]. The first reported CD of a compound semicon-
ductor film appears to be formation of “lüsterfarben” (lustrous colors) on various
metals from thiosulphate solutions of lead acetate, copper sulphate, and antimony
tartrate, giving films of PbS, Cu-S or Sb-S, which possessed “splendid” colors (in-
terference colors resulting from various thicknesses of the deposited films) [2].
More “recent” studies of this general process have invoked an electrochemical
mechanism for some thiosulphate depositions, based on the dependence of depo-
sition on either the nature (standard electrochemical potential) of the metal sub-
strate or on a contacting non-noble metal (which can be looked at as an internal
electrochemical deposition) [3–5]. However, while it is probable that an electro-
chemical or mixed electrochemical/chemical mechanism may be applicable on
some metal substrates, some of these depositions do appear to be true CD pro-
cesses. PbS is probably the clearest of these; others were Cu2S, Ag-S, Bi-S, Sb-S.
Fe, Ni, and Co all formed apparent sulphide films on Fe substrates, while ammo-
nium molybdate deposited a film from a thiosulphate solution that did not contain
S and was probably an oxide. Beutel and Kutzelnigg cover a wide range of depo-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sitions from thiosulphate solutions—both CD and electrochemical [5]. Only in a
few cases were these films characterized other than by color.
       In 1884, Emerson-Reynolds reported deposition of PbS films by reaction
between thiourea (thiocarbamide) and alkaline lead tartrate, where “the metallic
sulphide . . . became firmly attached as a specular layer to the sides of the vessel”
[6]. A wide range of substrates, apart from that just mentioned (a glass beaker),
was successfully used for this deposition; porcelain, ebonite, iron, steel, and brass
were specifically mentioned. Even more important, the deposits were very adher-
ent, as quantified by their ability to “withstand considerable friction with a wash-
leather, and under this treatment take a fine polish.”
       Infrared photoconductivity in CD PbS films was reported nearly a century
ago [7,8], and this application has been a central driving force for subsequent
investigations in CD lead chalcogenide films. The early literature invariably
mentions the pioneering work of Kutscher in Germany, during World War II, in
developing CD PbS and PbSe films for infrared detectors. However, the apparent
lack, in all these references to Kutscher’s studies, of any published papers might
suggest (to the overly suspicious reader) a possible military involvement in these
studies. These (and subsequent) studies succeeded to the extent that CD was,
and apparently still is, the main technique used in making commercial PbS and
PbSe infrared detectors (vacuum evaporation was the only competing technology)
[9,10].
       For a long time, CD was then essentially limited to PbS and PbSe. It was not
until 1961 that deposition of CdS, now the most widely studied material in
CD, was explicitly reported [11] (although CdS deposited from a thiosulphate so-
lution which “sticks obstinately to the glass” was already noted in 1912 [11a]).
The range of materials deposited by CD was gradually extended, particularly in
the 1980s, to include sulphides and selenides of many metals, some oxides, and
also many ternary compounds (Tables 2.1 and 2.3 in this chapter list films de-
posited by CD).
       Chemical deposition received a major impetus after CdS films, chemically
deposited onto CdTe (and, later, onto CuInSe2) films, were shown to give supe-
rior photovoltaic (PV) cells compared with the previously evaporated CdS. The
first reference to CD CdS used in thin-film PV cells appears to be from Uda et al.
[12], although no special importance was attached to the CD technique in that pa-
per. Birkmire et al. showed that CD CdS was as good as evaporated (Cd,Zn)S as
the heterojunction partner in CuInSe2-based thin-film cells, giving 10.6% effi-
ciency [13]. Two years later, the efficiency of CuInSe2 cells using CD CdS had in-
creased to 12.8% [14]. In 1991, Chu et al. used CD CdS to make high-efficiency
(13.4%) CdTe/CdS thin-film cells, explicitly stressing the beneficial role of the
CD CdS [15] and followed this a year later with a 14.5% cell using the CD CdS
[16]. Nowadays, Cd is almost universally used to form the CdS layer on both
CdTe and CuIn(Ga)Se2 thin-film PV cells.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Another cause of interest in this technique is due to the fact that the crystals
in most as-deposited CD films are very small. Considering the current interest in
nanoparticles, CD is an excellent technique to deposit nanocrystalline films. More
specifically, if the nanocrystals are small enough, they exhibit size quantization,
the most obvious manifestation of which is an increase in the optical bandgap with
decrease in crystal size, as was shown for CD CdSe [17] and later for CD PbSe
[18,19]. In fact, the changes in optical spectra that occurred in these films as a
function of nanocrystal size were exploited to provide information on the differ-
ent mechanisms of the deposition process [20].
       Chemical deposition has also been emphasized as a technique to form solar
control coatings. Solar control coatings are envisaged for use on windows in hot cli-
mates and possess the (ideal) characteristic of moderate to high visible transmission
to provide adequate lighting, together with high infrared (0.7–2.5 m) reflectance
to minimize heating by solar energy. CD is a potentially suitable method to prepare
these coatings on the large areas of glass that would be needed. Most of the work in
this field has been carried out by Nair and Nair in Mexico using various semicon-
ductor films, mainly PbS [21–23] and CuxS [23]. See also this group’s
recent review on this work [24]. These coatings are normally yellowish or neutral
by transmitted light and various shades of gold, blue, or purple by reflected light.


2.2 WHAT IS CHEMICAL DEPOSITION?
Chemical deposition refers to the deposition of films on a solid substrate from a
reaction occurring in a solution (almost always aqueous). Using the prototypical
CdS as an example, a Cd salt in solution can be converted to CdS by adding sul-
phide ions (e.g., as H2S or Na2S); CdS immediately precipitates (unless the solu-
tion is very dilute—a few millimolar or less, in which case CdS often forms as a
colloidal sol). Another pathway for CdS formation, one that does not require free
sulphide ions, is decomposition of a Cd-thiocomplex (a compound that binds to
Cd through a sulphur atom). In CD, the trick (or at least one of them) is to control
the rate of these reactions so that they occur slowly enough to allow the CdS ei-
ther to form gradually on the substrate or to diffuse there and adhere either to the
substrate itself (at the early stages of deposition) or to the growing film, rather than
aggregate into larger particles in solution and precipitate out.
       This rate control can be accomplished by generating the sulphide slowly in
the deposition solution. The rate of generation of sulphide, and therefore reaction
rate, can be controlled through a number of parameters, in particular the concen-
tration of sulphide-forming precursor, solution temperature, and pH.
       The CdS forms through a number of different possible pathways: simple
ionic reaction between Cd2 and sulphide ion; topotactic conversion of Cd(OH)2,
which may be present in the deposition solution, to CdS by sulphide; and decom-
position of a complex between Cd (whether as a free ion or as a Cd compound,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
e.g., Cd(OH)2) and the sulphide precursor (often thiourea, which, like other
chalcogenide precursors, also acts as a complexant for metal ions).
       Although CD can be carried out in both acidic and alkaline solutions, most
CD reactions have been carried out in alkaline solutions. This is necessary for se-
lenide deposition using selenosulphate (see later), which is unstable in acid solu-
tion (of course, the chalcogenide precursor must not be too stable under all condi-
tions, otherwise they will not work). Therefore to prevent (at least bulk)
precipitation of metal hydroxides in the deposition solution, the metal ion must be
complexed. There is a very wide range of possible complexing agents available;
the most used are intermediate in complexing strength—not too weak, in order to
prevent bulk precipitation of hydroxide, but not too strong, which may prevent de-
position of the desired film altogether.

2.3 WHAT MATERIALS CAN BE DEPOSITED
    BY CD?
In principle, CD can be used to deposit any compound that satisfies four basic re-
quirements.
      The compound can be made by simple precipitation. This generally, al-
         though not exclusively, refers to the formation of a stoichiometric com-
         pound formed by ionic reaction.
      The compound should be relatively (and preferably highly) insoluble in the
         solution used (except in a very few cases, this has been water).
      The compound should be chemically stable in the solution.
      If the reaction proceeds via the free anion, then this anion should be rela-
         tively slowly generated (to prevent sudden precipitation). If the reaction
         is of the complex-decomposition type, then decomposition of the metal
         complex should similarly occur relatively slowly (see Sec. 2.5 for a de-
         scription of reaction mechanisms).
Of course there are other specific factors that need to be taken into account, par-
ticularly whether the compound will form an adherent film on the substrate or not.
However, the preceding four factors are general requirements.

2.4 FORMATION OF THE (CHALCOGENIDE)
    ANIONS
One of the requirements for CD is either slow release of the anion—in most cases
a chalcogenide anion—or slow decomposition of a suitable complex containing a
chalcogenide atom. This section is confined to the former and discusses slow anion
release. The precursors used up to now, together with some of the reactions leading
to anion formation, will be briefly described (more details can be found in Chap. 3).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2.4.1 Oxide
Many oxides have been deposited by CD using many different techniques, some
of which are described below. Section 2.9.4.1 provides a wider overview of these
methods. Often, what is deposited is a hydroxide or hydrated oxide. In many
cases, hydroxide ion is generated slowly. There are a number of methods to do
this, the most common being hydrolysis of urea:
      (NH2)2CBO        2H2O → (NH4)2CO3                                        (2.1)
where the carbonate formed partially hydrolyses to give OH . Another technique
is reduction of nitrate to nitrite by an alkylamineborane:
      NO 3     H2O     2e (from alkylamineborane) → NO 2           2OH         (2.2)
Additionally, hydroxide may initially be present, but the reaction may be slowed
down e.g. by complexation of the metal ions or by carrying out the reaction at low
temperature. The hydroxide reacts with the metal ion to give the metal hydroxide,
a hydrated oxide or an oxide, depending on the chemistry of the particular metal-
(hydr)oxy system. The hydroxides or hydrated oxides can be heated in air or oxy-
gen to form the oxides.
      Another method used to deposit oxides, particularly those with a higher
oxidation state than the starting cation, uses persulphate, S2O 2 , a strong oxidiz-
                                                                8
ing agent. While the exact mechanism of oxide formation using persulphate
is unclear, it appears to involve internal electrochemical reactions; e.g., for
PbO2:
      Pb2      2H2O → PbO2        4H      2e                                  (2.3a)
      S2O28     2e → 2SO 2
                         4                                                    (2.3b)
However, it is possible that free radicals, such as · OH, are involved, since persul-
phate hydrolysis can proceed with formation of H2O2, which is itself sometimes
used to deposit oxides:
      S2O2
         8      2H2O → 2SO 2
                           4           H2O2    2H                              (2.4)

2.4.2 Sulphide
Thiourea (SC(NH2)2), the sulphur analogue of urea, is the most commonly used
sulphur precursor. There are a number of possible decomposition routes for
thiourea in aqueous solution (it is invariably used in alkaline solutions). Probably
the most important is
      SC(NH2)2       OH D HS           CN2H2     H2O                           (2.5)
which generates sulphide ion (the cyanamide, CN2H2, can hydrolyze further, but
this need not concern us at present). Actually, aqueous solutions of thiourea are



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
not very unstable; the presence of a cation that can precipitate an insoluble sul-
phide is necessary for the decomposition to proceed at a reasonable rate. This im-
plies that reaction (2.5) is actually an equilibrium; the metal ion removes sulphide
ion, which drives the equilibrium continually to the right.
       Thioacetamide (H3C.C(S)NH2) has also been commonly used in CD. It has
the advantage, compared with thiourea, that it works in both acid and alkaline so-
lution. A general decomposition reaction for sulphide formation is
      H3C.C(S)NH2       2H2O → CH3COOH            H2S     NH3                  (2.6)
In alkaline solution, the sulphide will be in the form of sulphide (HS and S2 )
ions.
                           2
       Thiosulphate (S2O3 ) was the original sulphur source in early CD pro-
cesses, and, while less commonly used nowadays, it still has a place in the mod-
ern CD literature. It is most commonly used in somewhat acidic solutions, al-
though it has also been employed in alkaline solution. Thiosulphate is unstable in
fairly acidic solutions and decomposes to give elemental sulphur, e.g.,
      S2O 2
          3     H →S         HSO 3                                             (2.7)
It is often suggested that the thiosulphate, a mild reducing agent, reduces this sul-
phur to sulphide.
       There are a number of other potential sulphide-forming reactions, depend-
ing on pH (see Sec. 3.2.1).
       It should be pointed out here that the SMS bond in thiosulphate is easily
broken. In view of the strong complexes it forms with some metal cations, the
probability of a mechanism whereby the SMS bond of the complex is broken,
leading to metal sulphide formation without formation of sulphide ion, should be
seriously considered (see Sec. 2.5).

2.4.3 Selenide
Selenourea, the selenium analogue of thiourea, which hydrolyzes in the same way
to give selenide ions, was once the most common source of Se. It is an unstable
compound that requires the presence of a reducing agent—usually Na2SO3.
Dimethylselenourea is more stable than selenourea but still difficult to work with.
The most common Se precursor used nowadays is sodium selenosulphate
(Na2SeSO3), which can be considered as the analogue of thiosulphate, with one S
atom substituted by Se. It is much more stable (and cheaper) than selenourea and
therefore simplified greatly the deposition of selenides. It can only be used in al-
kaline solutions (it decomposes at pH values lower than ca. 7 to precipitate ele-
mental red Se). Its alkaline hydrolysis is usually given as
      SeSO2
          3       OH → HSe           SO 2
                                        4                                      (2.8)




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
although it is very probable that SO 2 is the end product of a more complicated
                                     4
reaction. As for most other mechanisms discussed earlier, Eq. (2.8) is probably
also an equilibrium between selenide and some sulphur–oxygen intermediate,
with the metal cation removing the selenide as for the thiourea decomposition in
the presence of metal cations.

2.4.4 Telluride
There are only a very few reports of telluride deposition by CD. This is due to
a number of factors: the relative instability of the tellurium analogues of the S-
and Se-forming precursors; the strong reducing conditions necessary to form tel-
luride ions; and the rapid reaction of telluride ions with oxygen dissolved in the
solution. CdTe films have been formed by using hydrazine—a strong reducing
agent—in a deposition solution containing TeO2, and very thin films have been
observed to form when H2Te was added to a solution of a Cd salt. Recently, the
deposition of CdTe has been reported using the tellurium analogue of seleno-
sulphate (tellurosulphate). While Te is apparently only very slightly soluble in
sulphite, it is apparently enough to deposit tellurides. For both methods, stoi-
chiometric films are more difficult to obtain than for sulphides or selenides, with
elemental Te typically also formed. A very recent study has also described con-
version of CD Cd(OH)2 films to CdTe using a solution of Te in hydroxymethane
sulphinic acid.

2.4.5 Halides
Halides (confined at present to silver halides) can be deposited by hydrolyzing a
water-soluble halogeno-alcohol (halohydrin) to slowly form halide ions in the
presence of Ag ions:
      X(CH2)nOH       H2O D X        H      HO(CH2)nOH                        (2.9)
The solubility products of most halides are much higher in general than those of
chalcogenides. Those of the silver halides are fairly low, which allows these de-
positions to take place readily.

2.4.6 Other Anions
Although CD seems to have been limited to chalcogenides (including oxides and
hydroxides) and isolated cases of carbonates, silver halides, and elemental Se, it
should be possible to deposit salts of other anions. There are a number of other an-
ions that can be slowly and homogeneously generated. These are discussed in
Chapter 3.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2.5 MECHANISMS OF CHEMICAL DEPOSITION
2.5.1 Introduction
The mechanisms of CD processes can be divided into two different processes: for-
mation of the required compound by ionic reactions involving free anions, and de-
composition of metal complexes. These two categories can be further divided in
two: formation of isolated single molecules that cluster and eventually form a
crystal or particle, and mediation of a solid phase, usually the metal hydroxide. We
consider first the pathways involving free anions and defer to later those where a
metal complex decomposes.
       A starting point for discussing the mechanisms of CD is to consider a sim-
ple precipitation reaction. If H2S is added to an aqueous solution of a Cd salt, yel-
low CdS precipitates out immediately. H2S precipitates the sulphides of most
cations (the alkaline and alkaline earth sulphides are soluble in water); this is the
basis of the well-known (at least, in the author’s university days) inorganic ana-
lytical scheme. Such a precipitation will not, however, result in a film on a sub-
strate or on the walls of the reaction vessel (actually, it may do so to a very slight
extent but this film would be extremely thin). To form a visible film of CdS, con-
ditions must be chosen so that bulk precipitation is prevented or at least slowed
down drastically. This is the purpose of the chalcogenide precursors, discussed in
the previous section. They slowly generate the chalcogenide, allowing slow for-
mation of the metal chalcogenide (CdS in the present example).
       The formation of the film, based on the formation of chalcogenide ions, can
occur by two fundamentally different processes. We continue to use CdS as the
example.

2.5.2 Ion-by-Ion Mechanism
The simplest mechanism, often assumed to be the operative one in general, is
commonly called the ion-by-ion mechanism, since it occurs by sequential ionic re-
actions. The basis of this mechanism, illustrated for CdS, is given by
      Cd2      S2 → CdS                                                        (2.10)
If the ion product [Cd2 ][S2 ] exceeds the solubility product, Ksp, of CdS
(10 28; Table 1.1), then, neglecting kinetic problems of nucleation, CdS will
form as a solid phase (see Chap. 1). If the reaction is carried out in alkaline so-
lution (by far the most common case), then a complex is needed to keep the
metal ion in solution and to prevent the hydroxide from precipitating out (but
see later). Since the decomposition of the chalcogenide precursor can be con-
trolled over a very wide range (by temperature, pH, concentration), the rate of
CdS formation can likewise be well controlled. Of course, the CdS should form
a film on the substrate and (at least ideally) not precipitate in the solution. This



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
aspect of the ion-by-ion mechanisms (and all other mechanisms) is treated in
Section 2.6.

2.5.3 Hydroxide Cluster Mechanism
It was stated earlier that complexation of the Cd was necessary to prevent
Cd(OH)2 precipitation. However, very often (more often than realized), Cd(OH)2
(or metal hydroxides in general) are important reaction intermediates in the CD
process. If the complex concentration is not high enough to prevent completely the
formation of Cd(OH)2, then a relatively small amount of Cd(OH)2 may be formed,
not as a visible precipitate, but as a colloid. Since Cd(OH)2 is colorless and col-
loids typically do not scatter light, unless they aggregate to a large extent (in which
case a suspension is the result), this means that the Cd(OH)2 colloid may not be
visible to the eye.
       The CdS is then formed by reaction of slowly generated S2 ion with the
Cd(OH)2:
      Cd 2     2OH → Cd(OH)2                                                    (2.11)
followed by
      Cd(OH)2      S2 → CdS         2OH                                         (2.12)
Reaction (2.12) occurs because Ksp for CdS (10 28; Table 1.1) is much smaller
than that for Cd(OH)2 (2 10 14). Another way of looking at this is that the free
energy of formation of CdS is more negative than that of Cd(OH)2. It has also been
suggested that the hydroxide cluster can act as a catalyst for thiourea decomposi-
tion. In this case, sulphide formation will occur preferentially at the surface of the
hydroxide rather than nucleate separately in the solution. Such a course is logical
based on the previous discussion of the effect of metal cations on the equilibrium
of thiourea decomposition.
       It should be borne in mind that the mechanism may change in the course of
the deposition. As the metal is depleted from solution, the complex:metal ratio
will increase and may pass the point where no solid hydroxide phase is present in
the solution. In this case, the ion-by-ion process will occur (initially in parallel
with the hydroxide mechanism, later maybe exclusively) if the conditions are suit-
able.

2.5.4 Complex-Decomposition Mechanism
The chalcogenide precursors possess many talents. Apart from forming the
chalcogenide ions, they also form complexes with metal ions. As noted at the be-
ginning of this section, and ignoring the distinction between ion-by-ion and hy-
droxide cluster mechanisms treated previously, CD processes can be divided ac-
cording to two basic mechanisms: participation of free sulphide ions (the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
commonly accepted mechanism in most cases, although this does not necessarily
mean that it is always the correct one), and a pathway involving decomposition of
a metal/chalcogen-containing complex without formation of free sulphide. Un-
fortunately, it is usually difficult to distinguish between these two processes, and
the former is assumed more out of inertia than because of any clear proof. In the
specific case of the CdS deposition using thiourea, a complex-decomposition
mechanism has been proposed in a number of different investigations, based on
kinetic studies of the film formation process. Here we can revert to intuition and
suggest that, in the case of strong complexation between the chalcogen compound
and the metal ion (e.g., as occurs between thiosulphate and Hg, Ag, and Cu), it
may seem more logical for the fairly weak SMS bond to break than the very strong
metal–chalcogen bond. To be fair, these very strongly complexed cations are also
those whose chalcogenides have a very low solubility product, and therefore very
little free sulphide would be needed to form those metal chalcogenides.
        As for the free chalcogenide processes, the complex-decomposition mech-
anism can occur either by an ion-by-ion (or molecule-by-molecule, since free ions
need not be involved directly) pathway, e.g.,
 [SO3MSMHgMSMSO3]2                H2O → HgS       SO2
                                                    4       2H      S2O2
                                                                       3     (2.13)
(the molecule of HgS can interact with other HgS molecules to form clusters and
eventually crystals. Of course, they may also redissolve. These aspects are treated
in the following section), or by a solid-phase intermediate, e.g.,
     [Cd(OH)2]n SC(NH2)2 → [Cd(OH)2]nMSMC(NH2)2                              (2.14)
   [Cd(OH)2]nMSMC(NH2)2 → [Cd(OH)2]n 1CdS CN2H2                      2H2O    (2.15)
with eventual exchange of all the hydroxide in the Cd(OH)2 to CdS.


2.6 NUCLEATION, ADHESION, AND
    FILM GROWTH
Probably the least-known aspect of the CD process is what determines the nucle-
ation on the substrate and the subsequent film growth. In considering this aspect,
we will treat the ion-by-ion and hydroxide cluster mechanisms separately, al-
though there will be many features in common. The principles discussed should
be the same for both the free chalcogenide and the complex-decomposition mech-
anisms.

2.6.1 Ion-by-Ion Growth
For nucleation to occur homogeneously in a particle-free solution by the ion-by-
ion process, supersaturation, usually a high degree of supersaturation, is typically
required (Chap. 1, section 1.2.1). The presence of a surface (the substrate or the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
walls of the reaction vessel) introduces a degree of heterogeneity that facilitates
nucleation. For this reason, depositions that proceed via the ion-by-ion process
tend to occur mainly on the substrate or other surfaces, rather than involving a
large amount of precipitate typical of the hydroxide mechanism. The surface can
be considered a catalyst for the nucleation.
       As discussed in Chapter 1, the most important force involved in adhesion of
the deposit to a substrate in general is the van der Waals force of attraction. In the
initial stages of growth, there may be specific chemical interactions between the
deposit and substrate. For example, if gold is used as a substrate, S, Se, and many
of their compounds interact chemically with the gold to form S(Se)–Au bonds.
This would promote good adhesion of the deposit to the gold. There could also be
chemical and electrostatic interactions between surfaces of the individual crystals.
For example, the positive S(Se) face of polar crystals could bind to the negative
metal face of an adjacent crystal if the relative orientations are suitable (in prac-
tice, this will probably not occur, since the crystal faces will adsorb solution
species as they grow). However, the van der Waals interaction between the crys-
tals in the strongly ionic solution is enough in most cases to ensure adhesion of the
crystals to one another.
       The fact that reasonably adherent films can be grown on apparently unreac-
tive substrates, such as plastics, and even on such an inert and hydrophobic mate-
rial as Teflon suggests that while such specific interactions between the semicon-
ductor and substrate may improve adhesion to the substrate, they are not essential
for film formation.
       Once nucleation has begun on a substrate (this usually includes the inside
walls of the reaction vessel), it generally becomes easier for the film to grow, since
deposition usually occurs more readily on the nucleated surface than on the clean
surface. The crystals will continue to grow until blocked by some process, such as
steric hindrance by nearby crystals or adsorption of surface-active substances
from the solution. The former is probably the dominant reason for growth termi-
nation in most cases.

2.6.2 Hydroxide Mechanism
Nucleation of the chalcogenide is much simpler in this process, since a solid
phase—the metal hydroxide (or other solid phase)—is already present and the
process proceeds by a substitution reaction on that solid phase. In this case, the
initial step in the deposition is adhesion of the hydroxide to the substrate. This
hydroxide is then converted into, e.g., CdS, forming a primary deposit of CdS
clusters. More Cd(OH)2 and, as the reaction proceeds, CdS and partially con-
verted hydroxide diffuses/convects to the substrate, where it may stick, either to
uncovered substrate (in the early stages of deposition) or to already deposited
material. This is essentially the same process as aggregation, described in Chap-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ter 1 in the section on forces, and again is a consequence of van der Waals at-
traction.
       Since the initial nucleation of hydroxide occurs homogeneously in the solu-
tion, the CdS also is formed homogeneously and therefore usually precipitates out
in the solution to a large extent. This precipitation occurs if the isolated crystals
aggregate to a sufficient extent to form large flocs. Film formation occurs when
high-surface-energy particles (single nanocrystals or small aggregates) reach the
substrate (or any other surface) before they precipitate out in the form of large ag-
gregates. This aggregation and homogeneous precipitation can be minimized, in
some cases even prevented, by judicious choice of deposition parameters. Thus,
while extensive precipitation suggests a hydroxide mechanism (some precipita-
tion can occur in the ion-by-ion process), its absence does not always mean that
the ion-by-ion process occurs.
       An expected difference between ion-by-ion and hydroxide (or any other
cluster) mechanisms is that in the latter, since colloids from the solution stick to
the substrate surface, the crystal size is not expected to change greatly with film
thickness (it may increase to some extent, since the colloids themselves can grow
via an ion-by-ion process on the crystals). For ion-by-ion growth, it is likely that
crystal growth occurs on nucleii already present on the substrate, and therefore
crystal size can increase with increasing deposition.
       The foregoing description assumes adsorption of colloidal metal hydroxide
from solution onto the substrate as the primary nucleation step. However, hy-
droxides can also adsorb on solid surfaces at pH values below that of bulk hy-




FIG. 2.1 Schematic diagram showing the probable steps involved in the ion-by-ion
mechanism. A: Diffusion of Cd and S ions to the substrate. B: Nucleation of the Cd and S
ions facilitated by the substrate to form CdS nucleii. C: Growth of the CdS nucleii by ad-
sorption of Cd and S ions from solution and nucleation of new CdS crystals. D: Continued
growth of CdS crystals, which adhere to each other through van der Waals forces (possibly
also chemical interactions).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 2.2 Schematic diagram showing the probable steps involved in the hydroxide
mechanism. A: Diffusion of hydroxide colloidal particles to the substrate, where they ad-
here (B) and react with S ions (either generated homogeneously in solution or catalyzed by
the hydroxide surface). This reaction results in exchange of the hydroxide by sulphide,
probably starting at the surface of the colloid and proceeding inward (C). This reaction will
occur both at the surface-adsorbed colloids and at those dispersed in the solution. Reaction
will continue (as long as the supply of sulphide continues) until most of the hydroxide is
converted to sulphide (D); eventually the primary particles of CdS will adhere to each other
to form an aggregated film (E); usually the nonadsorbed particles will also aggregate and
precipitate out of the solution.




droxide precipitation, an effect which has been related to the presence of an elec-
tric field at the substrate/solution interface. This will certainly affect the nucle-
ation, since this can now occur only on the substrate and not in solution. It is not
so clear whether it will affect further crystal growth or not.
       The basic features of the ion-by-ion and hydroxide cluster film-forming
mechanisms are shown schematically in Figures 2.1 and 2.2, respectively. Film
formation involving complex decomposition will proceed in a similar manner
(Fig. 2.3 shows this for a molecule-by-molecule deposition).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 2.3 Schematic diagram illustrating possible steps in the complex-decomposition
mechanism. The complex (CdMSML, where L is a ligand or part of the S-forming species)
decomposes to CdS on the substrate (possibly catalyzed by the substrate) and, to a greater
or lesser extent, also homogeneously in the solution (A, B). The CdS nuclei formed grow
by adsorption and decomposition of more complex species (C) until a film of aggregated
crystals is formed (D) in the same manner as for the previous two mechanisms.


2.7 KINETICS OF THE DEPOSITION
Due to the different pathways that can occur in the CD process, the kinetics can
vary widely from one deposition to another, as reflected by the rather wide range
of activation energies (not commonly measured, but measured often enough to
draw some conclusions) found. Regarding the time taken for a deposition, some
depositions can be completed in a few minutes or less, while others can proceed
for days and still be far from termination. This section is only meant to give a gen-
eral picture; Chapter 3 should be consulted for more specific details and examples.
       Kinetic studies on the growth of CD films show, in most cases, an induction
period at the beginning of the process where no clearly observable growth occurs,
an approximately linear growth region, and a termination step where no further
growth occurs (see Fig. 2.4). Strangely enough, this type of growth kinetics often
occurs regardless of the deposition mechanism. For the ion-by-ion growth, it is
very simple to explain. Deposition begins only when the chalcogenide concentra-
tion is high enough to allow nucleation to occur—the induction time corresponds
to this buildup of chalcogenide concentration. Growth then occurs on these initial
nucleii, along with new nucleation—the approximately linear region of growth.
As the limiting reactant is used up, growth will start to slow down and eventually
stop due to depletion of the reactants.
       For the cluster mechanism, while growth and termination can be similarly
explained, the induction period is less obvious. The hydroxide cluster can start to
adsorb on the substrate immediately after immersion of the substrate in the depo-
sition solution, yet experiments have shown that film growth often does not occur
for some time. While the reason for this is not clear, it may be connected with the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
fact that the hydroxide particles often do not form a film, beyond some primary
adsorption on a surface; only when reaction to form the metal chalcogenide occurs
does film formation develop. In this case, we can again invoke the need for a min-
imum concentration of chalcogenide ion. Some studies of the deposition rate have
suggested that the rate-limiting step is a chemical rather than a diffusion process,
which supports the formation of the chalcogenide as this limiting step rather than
diffusion of cluster species to the substrate. Also, as described in the previous sec-
tion, metal hydroxide might deposit on the substrate under conditions where it will
not form in the solution, but this is likely to be confined to the surface layer. It
must be stressed, however, that due to the various possible processes involved in
CD, the results of one or even of several studies cannot automatically be extrapo-
lated to all other depositions of the same compound.
       If reaction is allowed to proceed until the termination stage is reached, the
terminal thickness of many CD films is typically several hundred nanometers, al-
though it may reach a micron or more in some cases. This terminal thickness de-
pends to a large extent on the deposition parameters. To take an extreme case, ad-
dition of sulphide to a solution of Cd ions will give an immediate precipitate of
CdS, but no (or at most an extremely thin) deposit on the walls of the deposition
vessel, which may thicken somewhat with time, but will not be visible (which
means a terminal thickness less than ca. 20 nm). For a normal CD reaction, if pre-
cipitation occurs homogeneously in solution, then that precipitate is lost for film




FIG. 2.4 Typical shape of the curve reprenting time dependence of film thickness dur-
ing growth.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
deposition, with resulting reduction in terminal thickness. Therefore, the ion-by-
ion process, with its lesser tendency for homogeneous precipitation, will usually
result in a larger terminal thickness than the cluster process, for comparable initial
reactant concentrations. Of course, film thickness can made as small as desired
simply by removing the substrate when the desired thickness has been reached (al-
though very thin films may not be homogeneous but, rather, clusters spread het-
erogeneously on the substrate). Alternatively, films thicker than the terminal
thickness may be obtained by repeated deposition (there will be a limit to the
thickness even here, since thick films tend to peel off the substrate).


2.8 SUBSTRATE
To a first approximation, films can be deposited by CD on any surface (this is one
of the advantages of CD). Of course, there will be certain obvious exceptions, such
as substrates that are unstable in the deposition solution (this is rarely a problem,
in practice) or “dirty” substrates. In several studies, CdS has been shown to form
quite adherent films on Teflon, and this attests to the ability of CD to form films on
a wide range of substrates. An important advantage of CD is that the shape of the
substrate is usually not important – very irregularly-shaped substrates can be used.
       To a second approximation, the nature of the substrate is usually important
in order to obtain an adherent film; some substrates result in more adherent films
than others. Rough substrates are better in this respect (in common with most de-
position procedures), probably due to the greater actual surface area of contact per
geometric surface area and the possibility of anchoring of the initial deposit in
pores of the substrate. Oxides [this includes glass, conducting oxides such as tin
oxide and indium tin oxide, and, to a lesser extent, silica (quartz)], in spite of ap-
pearing inert, are actually quite reactive in terms of their adsorption properties.
This is due to the presence of hydroxyl surface groups, which can form fairly
strong hydrogen bonds. Yet there can be very noticeable differences in the adhe-
sion to different glasses and between glass and silica (deposits tend to be less ad-
herent to silica than to glasses in general). An early study on CD PbS films using
different glass substrates found large differences in film formation on the differ-
ent substrates; no (or, at best, only patchy) films were formed on borosilicate glass
or on silica, whereas lead flint glasses, followed by zinc crown glass, resulted in
the best films [25]. The ability to form good PbS films on these latter glasses was
ascribed to the ability of the PbO or ZnO in these glasses to from insoluble sul-
phides. This would enhance binding to the depositing film. The possibility of ion
exchange between metal ions in the glass and those in solution may also play a
role in binding the initial CD film.
       Glass substrates can be sensitized, usually with a solution of SnCl2, which
hydrolyzes to give nuclei of tin hydroxide or oxide, on the surface. While in most
cases of CD, such sensitization is not used, and not required, there have been re-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ports of better layers (more adherent and/or homogeneous layers, faster deposi-
tion) using such sensitization. For example, it has been shown that, for PbSe de-
position from selenosulphate solutions, film formation begins immediately on
SnCl2-sensitized glass, instead of after an induction period as on plain glass. Ad-
ditionally, film formation occurs (at least initially) in the absence of bulk precipi-
tation in solution, in contrast to the parallel film deposition and bulk precipitation
that occurs when nonsensitized glass is used [26]. This suggests that a high su-
persaturation is required for nucleation to occur on untreated glass (as well in so-
lution) and that this is not the case when nuclei are already present on the glass
from the sensitization (or from previously deposited PbSe). It can be concluded
that such sensitization of glass (and probably many other surfaces) should be con-
sidered if satisfactory growth is not obtained without it. Although probably not
important for most requirements, it should also be kept in mind that such films will
contain a small amount of tin at the film/substrate interface.
       Metals make good substrates in general, either because chalcogenides tend
to adsorb strongly on the noble metals, in particular gold, or the non-noble metals
are covered with a (usually hydroxylated in the deposition solution) oxide layer.
In addition, if the metal in the deposition solution has a sufficiently negative po-
tential, an internal electrochemical reduction may occur (remember that elec-
trodeposition can often be carried out from CD solutions). This was suggested a
long time ago for deposition of various metal sulphides from thiosulphate solu-
tions on certain metal substrates [3,5].
       A large variety of CD films have been deposited on different polymer sur-
faces subjected to various activation treatments [27]. The most effective treatment
was immersing the substrate in KMnO4 solution for 24 hr, which formed a brown
Mn-O film and subsequent removal of this film with, e.g., conc. HCl. It was sug-
gested that the permanganate introduces carboxylic groups on the originally hy-
drophobic surface.
       Films have also been deposited on monolayers—both Langmuir and self-as-
sembled. In many cases, such depositions have studied nucleation of very thin lay-
ers of deposit (more accurately, scattered nanocrystals), but deposition of thicker
films, such as PbS [28] and Fe(O)OH [29] on self-assembled monolayers and var-
ious metal sulphides and selenides on Langmuir–Blodgett films [30] have been
studied, in some cases to understand how a well-defined—either chemically
and/or geometrically—substrate can control nucleation and growth geometry. De-
position does not occur on some monolayers, or at least considerably less readily
than on non-monolayer-covered substrate. An example of such a monolayer is oc-
tadecylphosphonic acid; CdS was found not to deposit on this monolayer but did
grow on the free areas of mica partially covered with the monolayer [31].
       The use of monolayers as substrates has been exploited to pattern deposits.
The principle behind this idea is that monolayers with either hydrophobic or
hydrophilic end groups can be patterned onto a substrate. Deposition will usually



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
occur only (or at least highly preferentially) on the hydrophilic endgroups.
Whether this is due to a simple physical interaction between solution (aqueous)
species and a hydrophilic surface, due to electrical charge on the solvated end-
group, or to some other specific interaction (or some combination of these effects)
is not clear. An example of this patterning is shown here for TiO2 deposition [32].
A long-chain thioacetate-terminated trichlorosilane is self-assembled on an
oxidized Si substrate (Fig. 2.5a and b). The thioacetate is somewhat hydrophobic.
Exposure of this monolayer through a mask (a fine mesh grid) to UV radiation




FIG. 2.5 Self-assembly of a long-chain thioacetate-terminated trichlorosilane on an ox-
idized Si substrate (a and b). Exposure of this monolayer through a fine mesh grid mask to
UV radiation (c) oxidizes the somewhat hydrophobic-thioacetate to hydrophilic sulphonate
endgroups (d). Deposition of TiO2 from an aqueous solution of TiCl4/HCl at 80°C on the
UV-exposed (hydrophilic) regions of the substrate (e). (From Ref. 32.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 2.6 SEM micrograph of patterned TiO2 deposit (left side) and Ti elemental map-
ping of this sample (right side). (From Ref. 32.)


oxidizes the thioacetate to hydrophilic sulphonate groups (c and d). Deposition of
TiO2 from an aqueous solution of TiCl4/HCl at 80°C occurred only on the UV-ex-
posed (hydrophilic) regions of the substrate (Fig. 2.6). See Section 4.1.5.3 and
Section 5.2.4.2) for other examples of patterned deposition (of CdS and PbS,
respectively) using different patterning procedures.
      For any particular substrate, the adhesion can also depend on the deposition
parameters, although little is reported on this aspect (for a good reason—it is not
really understood). The surface of glass (of whatever type) is hydroxylated in
aqueous solution and the concentration of the various species is clearly pH de-
pendent, as can be seen from Eq. (2.16):
      J                     J
      MSiMOMH
      J                D    MSiMO
                             J             H                                  (2.16)

[Since silica is very acidic, the dissociation to give a positively charged surface
(MSiMOH 2 ) will only occur in very acidic solution (pH 2) and is not com-
monly encountered in CD. Other less acidic oxide surfaces may, however, be pos-
itively charged in CD solutions.] This dissociation can affect interaction between
glass and various species in the solution as a function of the solution pH. Note that
water (also ammonia) may interact, through hydrogen bonding, with the hydrox-
ylated silica.
       In the end, adhesion as a function of deposition parameters (in contrast to
the nature of the substrate) can usually not be reliably predicted, and there is no
substitute for experimental experience.
       Apart from adhesion, the crystallographic properties of the CD film are
sometimes dependent on the nature of the substrate (although more often there
does not seem to be any dependence of this type). One example is epitaxial depo-
sition on a crystallographically ordered substrate [epitaxial here means a struc-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tural relationship between the crystal lattice of the deposit and of the substrate—
i.e., the crystal axes of the deposit are aligned (but not necessarily parallel to) the
crystal axes of the substrate]. Examples of this will be discussed later and in other
chapters. Also, even if the deposition is not epitaxial (and only a few cases have
been reported where epitaxy occurs), different texturing of the film may occur.
Texturing refers to the preferred orientation of the crystals of the deposit perpen-
dicular to the substrate; thus, textured films of hexagonal CdS often have the basal
face (the Cd or S face) pointing upward. If orientation occurs also in the plane of
the substrate (azimuthal orientation), then the deposit is called oriented. The term
orientation is often (erroneously) used where texturing is meant.
        A monolayer-covered substrate may dictate the crystallographic form of a
CD film, depending on the interaction between the monolayer endgroups and the
semiconductor. Thus, epitaxial growth of PbS has been accomplished at arachidic
acid monolayers where the PbS interatomic spacings along the (111) plane are
well matched with the monolayer packing [33]. Texturing has been observed in
some cases, even when the monolayers themselves were not well ordered.
        It will be obvious that the cluster mechanism of deposition is unlikely to
lead to an oriented film, since the clusters would have to align themselves with the
substrate lattice, either on adsorption or subsequently. Therefore an epitaxial film
is highly suggestive of an ion-by-ion growth, which is more likely to be directed
by the substrate.
        One example has been described of CD (PbS) on a poled ferroelectric sub-
strate. The PbS crystal size was larger (ca. 1 m) on the poled substrate than on
the unpoled (or a glass) substrate (ca. 0.3 m) [34]. Other changes in the electri-
cal properties of the films were noted. The differences were ascribed to the elec-
tric field and charge accumulation at the ferroelectric surface (more details can be
found in Sec. 5.2.4.3).


2.9 DEPOSITION OF SPECIFIC
    SEMICONDUCTORS
This section will give an overview of the various semiconductors that have been
deposited by CD. The groups of semiconductors will be divided in much the same
way as in the rest of the book:
      II–VI
      IV–VI
      Other binary sulphides and selenides
      Oxides, other binary semiconductors, and elemental Se
      Ternary compounds
      A few specific examples of experimental details will be given in this sec-
tion, with explanation of the importance of the different variables.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2.9.1 II–VI Compounds
This comprises the most studied group and includes CdS (the single most studied
compound), CdSe, CdTe, ZnS, ZnSe, HgS, HgSe, and various mixtures of these
compounds.
2.9.1.1      CdS
       Film Deposition. We begin with a common “recipe” for CdS deposition
and follow the deposition process using this recipe. This is described in detail as
a way of understanding both the deposition process and which factors can affect
the final film.
       STEP 1: PREPARATION OF STOCK SOLUTIONS. Usually, stock solutions of the
solid reactants are prepared in advance, since many depositions are normally car-
ried out. The three reagents required are:
      CdSO4: Other water-soluble Cd salts can be used equally well, such as the
        chloride or acetate; there is no clear evidence in general that the nature of
        the anion is important, although there are a few studies that reported dif-
        ferences in the deposit depending on the anion used.
      NH4OH: Since this is required in high concentrations and is a liquid, a stock
        solution is not required, but it can be used directly from a bottle of con-
        centrated ammonia (concentrated ammonia is ca. 15 M).
      Thiourea (SC(NH2)2): This solution will slowly precipitate sulphur (seen
        as a fine white precipitate—white instead of the usual yellow of sulphur
        possibly due to size quantization of the finely divided precipitate?).
        However, it can usually be kept over a period of weeks or even months
        in a stoppered bottle without major adverse effects, although for opti-
        mum reproducibility a fresh solution may be preferred. If a prepared
        stock solution is used, filtration of this solution before use will mini-
        mize the presence of sulphur particles in the deposition solution; such
        particles can act as nucleation centers and accelerate precipitation in the
        solution.
       STEP 2: PREPARATION OF THE SUBSTRATE(S). While almost any substrate can
be used, we will use glass microscope slides in this example; this is a common
substrate and makes it easy to see the CdS film. The microscope slide can be cut
to whatever size and shape is convenient. The slide should be cleaned well, since
films usually do not adhere well to “dirty” surfaces. Suitable cleaning agents are
trichloroethylene or/and sulphochromic acid, and the slide should be well rinsed
with pure water. If the slide is clean, water dropped onto it will form a film (hy-
drophilic surface), while on a “dirty” (hydrophobic) slide the water will form
drops. Needless to say, the part of the slide where deposition is to occur should not
be touched with the hands after this treatment.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       STEP 3: PREPARATION OF THE DEPOSITION SOLUTION. Solution compositions,
even based on exactly the same constituents as stated earlier, vary widely from one
literature source to another. We give a typical “average” composition and discuss
the effect of variations from this composition.
       This “average” solution is made up of (concentrations in final solution):
      CdSO4 10 mM
      Aqueous ammonia 1M
      Thiourea 50 mM
      Solution pH 11
      Deposition temperature 70°C
Concentrated ammonium hydroxide is added to a stock solution of CdSO4 (or
other Cd salt). Initially, Cd(OH)2 precipitates, but this redissolves in excess am-
monia to give the cadmium ammine complex:
      Cd2      4NH3 D Cd(NH3) 2
                              4                                               (2.17)
This solution is heated to 70°C in a thermostated water bath. A stock solution of
thiourea is then added to bring the thiourea concentration up to 50 mM. The final
solution pH can be adjusted with KOH (more basic) or acetic acid (more acidic).
       STEP 4: DEPOSITION. The substrate(s) is immersed in the preceding solution
and placed in the water bath at 70°C. (Note that CdS usually also forms homoge-
neously in solution, and this can sediment onto the substrate, where it forms a
loose coating and may prevent growth of an adherent film. For this reason, the
substrate should be placed vertically in solution; or if placed at an angle or even
horizontally, only the underside of the substrate, where sedimentation does not oc-
cur, should be used.)
       The time of deposition is variable (although it should be reproducible from
one run to another under the same conditions), and it is difficult to give a particu-
lar time. Usually a thin film (tens of nanometers) will form in some minutes, and
this will slowly thicken over some tens of minutes to hours to a typical terminal
thickness of ca. 200 nm. The simplest way of determining the optimum time (and
“optimum” will depend on the application) is to simply look at the film. With a lit-
tle experience (or with the more quantitative help of a spectrophotometer), the ap-
proximate thickness can be estimated from how deep the yellow color is. Thick-
ness can also be estimated from the transmission spectrum by measuring the drop
in transmission over the near-bandgap region; this method is less dependent on re-
flection and scattering losses (as long as they are not too large) than simply mea-
suring absorption at a fixed wavelength. In any case, prior calibration of the spec-
tra with known thicknesses of the film is required.
       When the desired thickness is attained, the substrate is removed and rinsed
with water. This rinsing can also be carried out in an ultrasonic bath, which more
effectively removes loose particles.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       VARIATIONS IN THE PROCESS. There are many variables in this (experimen-
tally simple) process; concentration of the various reactants, pH, and temperature
are the main ones. Other, less important (in most cases) ones are stirring of solu-
tion and illumination of the solution during deposition. We will ignore these last
two here.
       The effect of reactant concentration can be divided into two separate influ-
ences. The simplest is obvious: Lower overall concentrations result in a slower rate.
This does not necessarily mean a thinner film, however—sometimes the opposite.
The reason for this is clear if we return to our introductory discussion on the CD
process—rapid precipitation. It is clear that if the reaction is too fast, it will termi-
nate with most of the product precipitating homogeneously in solution rather than
depositing on the substrate (which requires time to occur). This results in a very
thin film, if any film at all. Similarly, for the less extreme case of a CD reaction that
terminates, not within a second, but still in a short time, the final film thickness will
be small. At the other extreme, if the reaction is extremely slow, a thick film can be
built up, but it may take a very long time for this to occur (weeks, even months). It
is therefore evident that there is an optimum rate for the reaction, which can be con-
trolled by a combination of reactant concentrations, temperature, and pH.
       A separate effect of concentration is the ratio between the metal ion and the
complexant concentrations. This ratio determines, often more than the overall re-
actant concentrations, the reaction rate, since it controls the concentration of free
metal ions in solution. It can also determine the reaction pathway. Further discus-
sion of this factor will be left for the next example (CdSe), since it has been treated
in more detail for that case.
       The solution pH influences a number of factors, and it is not always simple
to predict its effect. Thus, thiourea decomposition (in alkaline solution) is gener-
ally faster at higher pH. The probability of the presence of a solid phase of
Cd(OH)2 and its concentration in the solution are both increased at higher pH
(higher OH concentration). The pH is determined, in the example given earlier,
by the concentration of ammonia. However, it can be adjusted independent of the
ammonia concentration. Addition of an ammonium salt, which with ammonia,
acts as a buffer, will lower the pH through the following equilibrium:

      NH3      H2O D NH 4        OH                                               (2.18)

Ammonia is alkaline in water because of this equilibrium, which produces hy-
droxide ions. The addition of an ammonium salt (NH 4 ions) will push the equi-
librium back toward the left, i.e., lower hydroxide concentration, therefore lower
pH. Increase of pH can be effected by addition of sodium or potassium hydroxide.
This explains why some CdS depositions based on the Cd/NH3/thiourea formula
include either an ammonium salt or an alkali metal hydroxide.
      Finally, the effect of temperature on increasing the reaction rate (possibly
also the mechanism—see next example) is again obvious, since the thiourea de-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
composition will be faster at higher temperature. Additionally, since the stability
constant of the complex is usually smaller at higher temperatures (see Chap. 1),
there will be a higher free metal ion concentration, which again translates into
faster rate (although, since sulphide generation is usually rate determining, this ef-
fect may not be large). Temperature programming during the deposition has also
been employed. This can be useful, for example, if the deposition solution changes
the substrate in some beneficial way (as occurs in some photovoltaic cell sub-
strates—see Chap. 9). The initial low temperature delays coverage of the substrate
and allows more time for this surface treatment to occur. If the temperature is then
increased, deposition occurs at a reasonable rate.
       There are other bath compositions based on different sulphide-generating
precursors and/or complexing agents. Thioacetamide and thiosulphate are two of
the former, while ethylenediamine is a common example of a complexant that has
been used instead of ammonia. The volatility of ammonia, and its gradual loss in
an open deposition bath, is circumvented by using a less volatile complexant, such
as ethylenediamine.

      Some Properties of CdS Films. It should be noted that all properties dis-
cussed in this chapter refer to as-deposited films (not annealed) unless specifically
stated otherwise. In general, annealing increases crystal size and reduces dark con-
ductivity. The latter obviously depends to a large degree on the annealing atmo-
sphere.

       STRUCTURE. CdS can exist in three different crystal structures: hexagonal
(wurtzite), cubic (zincblende)—both tetrahedrally coordinated and cubic (rock-
salt), which is sixfold coordinated. Except in a few cases, the rocksalt modifica-
tion of CdS has been observed only at very high pressures: CD films of this phase
have never been reported. The other two phases have been reported to occur in CD
films under various conditions. The wurtzite phase is thermodynamically slightly
more stable, and invariably forms if the zincblende phase is heated above
300–400°C. The low-temperature CD method therefore can allow the formation
of the zincblende phase, and this phase is commonly obtained in CD CdS films.
Very often, a mixture of wurtzite and zincblende phases has been reported in the
literature. There are many variables that affect the crystal structure, including the
nature of the complex, the substrate, and sometimes even stirring.

        ORIENTATION. In common with with other deposition techniques, a prefer-
ential texturing of the film in the (111) (for zincblende) or (0002) (for wurtzite) di-
rection is often reported. However, nontextured (or weakly textured) films are
probably more commonly obtained. As for crystal structure, the degree of textur-
ing depends on several factors, an important one being the nature of the substrate.
It is likely that highly textured deposits form by the ion-by-ion mechanism; a pure
solid-phase cluster mechanism is less likely to result in strong texturing.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In certain cases, epitaxial growth has been observed on crystallographically
well-defined substrates, such as single crystal InP and CuInSe2. This type of
growth, even more than textured growth, demands ion-by-ion growth as well as a
crystallographically well-defined substrate. Lattice parameters (at least at some
defined ratio and angular match) fairly close to that of CdS are also required for a
relatively defect-free interface, although there is some flexibility here. For exam-
ple, CdS with a high density of stacking faults has been epitaxially deposited on
GaP (7% mismatch).
       OPTICAL PROPERTIES. The most commonly reported optical properties are
optical transmission, with some studies also on photoluminescence. The impor-
tance of the optical transmission for CdS in particular lies in its use in photovoltaic
cells, where it acts as a window layer. The CdS should be as transparent as possi-
ble to the incoming radiation. The transmission is a function of thickness,
bandgap, and film structure (is the film transparent or scattering?). The bandgap
in most studies is constant (ca. 2.45 eV at room temperature), although somewhat
larger values have been obtained due to size quantization in very small crystallites.
       Photoluminescence of the films varies greatly, both in intensity and in spec-
tral shape, from one report to another. This is not surprising, since this property is
very dependent on the state of the surface of the individual crystals. A red (ca. 1.8
eV) defect emission is usually seen, but green, yellow, and infrared peaks have
also been reported. The various wavelengths are related to different defects in/on
the crystals; even the green emission is probably due to a shallow defect emission.
       ELECTRICAL PROPERTIES AND PHOTOCONDUCTIVITY. Electrical resistivity of
CD CdS films is commonly studied. Values for this (dark) resistivity vary over
many orders of magnitude from one film to another, usually for reasons that are not
understood. Values as high as 109 -cm and as low as 15 -cm have been reported
for undoped films (doped films have been reported with still lower resistivities).
       Since the films are often highly resistive, it is not surprising that they exhibit
strong photoconduction. Photoconduction occurs due to the formation of free car-
riers by illumination, and if the free carrier concentration is low to begin with (low
conductivity), then the photogenerated carriers will usually dominate the conduc-
tivity. This is in contrast to a relatively conducting semiconductor (high doping
level), when the extra photogenerated (majority) carriers are only a small pertur-
bation to those present in the dark. Light:dark photoconductivity ratios (sensitiv-
ity) as high as 109 have been reported. In many cases, the photocurrent decay time
is measured in hours, and this is explained (in a general way) by slow states; the
nature of these states is not usually known.
2.9.1.2      CdSe
     Film Deposition. Chemical deposition films of CdSe have been relatively
widely investigated, largely for photoanodes in photoelectrochemical cells (see



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Sec. 2.12.1.2). Here we will give an example of CdSe deposition, emphasizing the
differences between the procedures for CdSe and CdS. The most important single
difference in the deposition is, of course, the chalcogenide-forming reagent. A
procedure used commonly by the author will be described, where the complexant
is nitrilotriacetic acid (NTA), which is related to the more common EDTA
(ethylenediaminetetracetic acid) and has the chemical formula N(CH2COOH)3. It
is a strong complexant for Cd2 (and many other cations), although less so than
EDTA, but stronger than the often-used ammonia. It also has an advantage (in
most cases) over ammonia in that it is nonvolatile and is not lost during deposi-
tion.
      STEP 1: PREPARATION OF STOCK SOLUTIONS. The aqueous Cd solution is pre-
pared as for CdS. As for CdS, the nature of the anion does not appear to be very
important—the sulphate is commonly used.
      NTA. NTA itself is not very soluble in water. Either it is used as the Na or
        K salt, or the acid can be dissolved together with three equivalents of
        KOH or NaOH to form a solution of the salt. This salt will be referred to
        hereafter simply as NTA.
      Sodium selenosulphate (Na2SeSO3). This compound is made by dissolving
        elemental Se in an aqueous solution of sodium sulphite:
      Na2SO3      Se D Na2SeSO3                                                (2.19)
The prepared solution will slowly deposit Se as a black precipitate over a period
of weeks. Also, the solution is considerably less stable than thiourea. If kept out
of excessive contact with air, it will be usable for about a month if high repro-
ducibility of the deposition kinetics is not important. However, it is important to
be aware of the slow decomposition and loss of reactivity of this reactant. If a
freshly made Na2SeSO3 solution is used to deposit CdSe, the reaction will proceed
much faster than if an aged solution is used; this fact should be taken into account
in preparing the overall deposition solution.
       The solution is made up from an aqueous solution 0.4 M in Na2SO3 and 0.2
M in Se. This solution typically is stirred at 60°C for a couple of hours (the Se dis-
solves slowly, the reason a fresh solution is usually not made up every time). If
this solution is prepared over a longer time (maybe 6 hours and/or at a higher tem-
perature, even boiling), it will undergo accelerated aging. This results in more re-
producible deposition conditions on one hand, but at the cost of reduced reactiv-
ity and a shorter lifetime of the solution.
      STEP   2: PREPARATION OF THE SUBSTRATE(S).      The same as described earlier
for CdS.
      STEP 3: PREPARATION OF THE DEPOSITION SOLUTION. As for CdS, the condi-
tions can be very variable. The following is a typical deposition solution, but re-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
member that allowances should be made for the varying activity of the selenosul-
phate. If deposition occurs too rapidly (slowly), parameters should be changed to
slow down (speed up) the reaction [e.g., lower (higher) selenosulphate concentra-
tion, higher (lower) NTA:Cd ratio, lower (higher) temperature].
       The following procedure is followed to obtain a final concentration of 60
mM each Cd2 and Na2SeSO3 and ca. 100 mM NTA (the reason for the “ca.” will
be explained shortly).
       Take the calculated amount of stock Cd solution and add water. Then add the
stock NTA solution. If the concentrated Cd and NTA solutions are mixed together
without adding water between them, they may form a gel and the solution will then
be useless. The pH should then be adjusted to ca. 8 by adding KOH or NaOH. It is
important that the pH at this stage be greater than 7, since selenosulphate immedi-
ately decomposes to red selenium at a pH lower than 7. This solution is brought to
the temperature at which the deposition is to be carried out. The selenosulphate
(also at the same temperature) is then added. This will increase the pH (since the
selenosulphate is alkaline). KOH is added as necessary to bring the pH to ca. 10
(9.5–10 is a suitable range; it should be neither much lower nor much higher than
this). If the pH increases to much above 10 on the addition of selenosulphate, di-
lute acetic acid can be carefully added to lower the pH to the required range.
      STEP 4: DEPOSITION. This is the same as for CdS, except that the deposition
usually is slower and takes from hours to days, depending on conditions. Or-
ange/red coloration in the solution, corresponding to the start of CdSe formation,
should occur some minutes after adding the selenosulphate. If it occurs immedi-
ately or almost immediately, the reaction may be too fast and only a thin terminal
thickness may be obtained. On the other hand, if coloration has still not started af-
ter about 30 min, the deposition will probably be very slow.
       VARIATIONS IN THE PROCESS. The reader may have noticed that some of the
preparation details just given are rather vague. This reflects the varying activity of
the selenosulphate solution, both with time and from one batch to another. It is
more important to understand the effects of the various parameters and to be able
to vary them logically than to follow an exact recipe. These effects will be de-
scribed in detail in Chapter 3, and it is recommended that anyone wanting to de-
posit these films read that chapter before carrying out the deposition. Here, a brief
explanation of these various factors is given.
       The deposition temperature obviously will increase the reaction rate (and
the deposition rate, although, again, if the reaction is too rapid, only a thin film
will be obtained). Another effect of increased temperature is increased crystal size
in the film. The crystal size varies (assuming the NTA:Cd ratio is not too high),
typically from 4 nm at close to 0°C to ca. 8 nm at 80°C. This change can be seen
by a change in color from yellow-orange (low temperature) to red (high tempera-
ture) due to size quantization (see Sec. 2.12.2 and Chapter 10).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
        If crystal size is not an important issue, then higher temperatures are usually
more convenient, since the deposition is faster. Adhesion of the film to the sub-
strate also appears to be better in general at higher temperatures (larger crystal
size? See table 1.3).
        Solution composition is important not only because the reaction rate in-
creases with the concentrations of selenosulphate and/or Cd, but even more so
through the ratio between the NTA and Cd concentrations. The higher this ratio,
the slower the reaction, since the free Cd2 concentration is less. The optimum
value for this ratio will depend, among other factors, on the deposition tempera-
ture. At higher temperatures, a higher NTA:Cd ratio is required to prevent too
rapid reaction. Thus the 100 mM concentration of NTA (for 60 mM Cd) given ear-
lier is typical for ca. room temperature deposition, but would probably need to be
increased to 110 mM or even more at substantially higher temperatures.
        Even more important, if the NTA:Cd ratio increases above a certain value
(which depends on the other solution parameters but which is typically between
1.7 and 2.1), the mechanism of deposition changes from a hydroxide cluster mech-
anism to ion-by-ion deposition. The latter is considerably slower, does not deposit
homogeneously in solution (or much less than the cluster mechanism), and results
in larger crystal size (typically 8–20 nm, depending on deposition temperature).
Also, in contrast to the cluster mechanism, where deposition is (should be) homo-
geneous and the films transparent throughout the deposition, the ion-by-ion mech-
anism usually results in films that appear nonhomogeneous and highly scattering
in the early stages of deposition, but become homogeneous and transparent with
time (typically several days at lower temperatures, less at higher ones).
        If all the foregoing gives the reader the impression that CdSe deposition is
more of an art than a science, this would not be a gross misunderstanding. Expe-
rience is certainly useful here, more so than for CdS deposition, which is more re-
producible, probably due to the more stable thiourea (and possibly also because of
the shorter deposition time typical for the CdS deposition, based on one of Mur-
phy’s laws—the more time you allow for something to occur, the greater the op-
portunity that something will go wrong).
      Some Properties of CdSe Films.
       STRUCTURE. CdSe forms the same three crystal structures as described ear-
lier for CdS. The main difference between the CD films of the two materials is
that, while CdS can be commonly found in both the wurtzite and sphalerite forms,
CdSe is more commonly deposited in the cubic zincblende form. Mixtures of the
two forms have been reported in some cases, particularly when a visible Cd(OH)2
precipitate is present in the initial deposition solution.
       As for CdS, CdSe has been epitaxially deposited on single crystal InP. As
expected, epitaxy occurred only for the ion-by-ion mechanism, where individual
species could either adsorb on or migrate to the ideal lattice position.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      The CdSe crystal size in the hydroxide mechanism is probably determined
by the size of the CdOH)2 particles in the solution and on the substrate, while that
formed through the ion-by-ion mechanism will depend on the heterogeneous nu-
cleation on the substrate, and is invariably larger.
      OPTICAL PROPERTIES. CD CdSe films often exhibit size quantization, with
a blue shift in their absorption spectrum of as much as 0.6 eV.
      Photoluminescence of the films (in the absence of water vapor) usually is
dominated, as for CdS, by a broad defect emission varying from ca. 1.4 to 1.6 eV.
A (close to) band-to-band emission is often also observed, usually (but not al-
ways) at a lower intensity than the broad defect emission. In the presence of wa-
ter vapor, however, the band-to-band emission often dominates.
      ELECTRICAL PROPERTIES. Resistivity studies on CdSe are much less
widespread than on CdS films. The dark conductivity of undoped films is high
(108 -cm is typical), and the photocurrent sensitivity is less than for CdS films
(even under illumination, the films are normally very resistive).
2.9.1.3     CdTe
Only two different studies on true CdTe deposition by CD appear to exist. In
the first, CdTe was deposited from solutions containing TeO2 and hydrazine,
the latter presumably slowly reducing the TeO2 to telluride ion. In this study, the
main interest was not on the deposition itself but on the further use of the CdTe
films, and not much information was given on the films themselves. More
recently, CdTe was deposited using what was apparently the Te analogue of
selenosulphate, Na2TeSO3. In all these depositions, the films were not very
stoichiometric and included considerable amounts of Te. Some structural, optical,
and electrical properties of these latter films were given (described in Chap. 4).
In a very recent study, CD Cd(OH)2 films were converted to CdTe using a
solution of Te in hydroxymethane sulphinic acid, which acted as a telluride
source. The conversion of the hydroxide to CdTe was incomplete, but there
did not seem to be any free Te in the films. Properties of these films were also
described.
2.9.1.4     ZnS and ZnSe
The most important difference between CD of CdS(Se) and ZnS(Se) is related to
the difference between the solubility products of the hydroxide and chalcogenides
of the two metals. Considering the sulphides, the various values of Ksp are:
Cd(OH)2—2 10 14; CdS—10 28; Zn(OH)2—1.10 16; ZnS—3.10 25. The dif-
ferences between the pairs of Ksp are 2 1014 (for Cd) and 3 108 (for Zn). Since
hydroxide ions are present at much higher concentrations than sulphide, hydrox-
ide formation, and stability against sulphurization, is much more likely for Zn than
for Cd.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       ZnS is commonly deposited by CD. However, in many cases, the ZnS is not
stoichiometric and contains hydroxide in one form or another. ZnSe, on the other
hand, is more difficult to deposit from selenosulphate baths, in spite of the some-
what lower solubility product of ZnSe (4.10 26) compared with ZnS. In fact, a
strong reducing agent such as hydrazine is normally required. The hydrazine pre-
sumably reduces the selenosulphate to give a high enough selenide concentration
to allow ZnSe to form, although it is possible that other factors are also important.
(The sulphite, present in excess in selenosulphate, is itself a reducing agent, al-
though much weaker than hydrazine.) Activation energies for ZnS(Se) deposi-
tions are generally considerably lower than for CdS and PbS deposition, and this
suggests a different deposition mechanism (although, as we have already seen, the
deposition mechanism is largely a function of the deposition conditions and not
simply the material itself). Also, the crystal size of ZnS(Se) is usually smaller than
for CdS or PbS. Both these factors suggest that ZnS and ZnSe form by a pure clus-
ter mechanism. In fact, nowhere in the literature is there evidence for ion-by-ion
growth of ZnS or ZnSe.
2.9.1.5      HgS and HgSe
There are only several reports on the deposition of the mercury chalcogenides
(some ternaries containing Hg have also been described).
      HgS films have been deposited from a simple chemical precipitation reaction
between mercuric chloride and sodium sulphide. Under suitable conditions, a film
is formed along with the precipitate. The thickness of the films were ca. 0.7 m,
which is thicker than normal for CD films, even more so considering the rapid na-
ture of the reaction, which normally only leads to a very thin film, if any at all. The
films apparently deposit from the colloidal HgS formed on mixing the reactants.
      HgS films have also been deposited from “more conventional” deposition
solutions using thiourea and the tetraiodide complex of mercury—a strong com-
plex—in alkaline solution. Both these and the previous films showed an optical
absorption with a gradual absorption onset at 700 nm and a sharp one at 400 nm.
      The thiosulphate mercury(II) complex in ammoniacal solution has also been
used to deposit films of HgS, both in ammoniacal and nonammoniacal solutions.
The deposit from the former was predominantly -HgS (cinnabar). As with the pre-
vious films, a sharp optical absorption onset at ca. 400 nm was observed, together
with a more gradual one extending, in this case, to beyond 800 nm and dependent
on film thickness. The nonammoniacal solution gave crystal sizes (and optical
bandgaps) that varied with deposition temperature from 3 nm (2.4 eV) at 0°C to 8
nm (1.9 eV) at 85°C and corresponding resistivities between 104 and 103 -cm.
      HgSe was first deposited from an iodide-complexed solution using se-
lenosemicarbazide. Two other depositions were described, both using selenosul-
phate. One used an alkaline Hg–formamide complex. The as-deposited films did
not show an XRD pattern, suggesting that the material was either amorphous or



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
very small nanocrystalline. The latter in particular is supported by the large
bandgap measured from the absorption spectrum (1.42 eV compared to semimetal
bulk HgSe with a negative bandgap). Electrical conductivity measurements indi-
cated a midgap Fermi-level characteristic of intrinsic semiconducting material. In
the other selenosulphate method, HgSe films were deposited from an ammoniacal
bath onto polyester substrates. The films were strongly (111) textured. The crys-
tal size was 7.7 nm, resulting in a strong blue shift in the optical absorption spec-
trum, and a measured bandgap of 2.5 eV (compared to a negative—semimetal—
bandgap of bulk HgSe). The sheet resistance of these films (13 k -cm 2) was
relatively low, considering the small crystal size.

2.9.2 IV–VI Compounds
2.9.2.1      Deposition
As noted at the beginning of this chapter, most of the early studies in CD focused
on PbS, followed by PbSe, driven by their photoconducting properties. For opti-
mum use as photoconductors, the deposited films were annealed in an oxygen-
containing atmosphere. Most of this section will focus on nonannealed films, and
annealed films will be treated only very briefly. More details on the annealed as
well as as-deposited films will be given in Chapter 5.
       Many different bath compositions (different refers to bath constituents
rather than simply to different concentrations) have been used to deposit PbS and
PbSe, more so than the II–VI materials. Both alkaline and acid baths have been
described, although the former are much more common. The chalcogen sources
are similar to those used for the II–VI compounds, including thiourea, most com-
monly used for PbS and selenourea (originally the source of choice) and, more
commonly nowadays, selenosulphate for PbSe. The variety of complexing agents
used is large; various carboxylic acids (most commonly citrate), triethanolamine,
nitrilotriacetate, hydroxide, and even selenosulphate itself have all been used. Ad-
dition of thiosulphate to a citrate/selenosulphate bath for PbSe resulted in a de-
posit that did not show an XRD pattern and was assumed to be amorphous; blue
shifts in the optical spectra could be explained either by structural changes or, pos-
sibly, by size quantization in crystals too small to be seen in the XRD spectrum.
       PbS has been deposited from an acidic bath using thiosulphate. Also, very
thin films of PbS (maximum absorbance 0.015) have been grown on quartz im-
mersed overnight in a solution of Pb ions together with polyvinyl alcohol (ostensi-
bly to protect against aggregation of colloids) through which H2S has been passed.
2.9.2.2      Film Structure and Morphology
PbS and PbSe are almost always found in the rocksalt (RS) crystal form. All struc-
tural investigations on CD films have shown this form, with one exception; PbSe
deposited from hydroxide complex at high hydroxide concentrations and at rela-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tively high temperatures have produced a novel rhombohedral modification with
an external hexagonal morphology and crystals typically 1–3 m in size. For
PbSe, crystal size varies considerably, depending on deposition parameters, more
so than for the II–VI compounds. Crystals as small as 4 nm and larger than a mi-
cron can be grown. The main factors, as for the II–VI materials, are temperature
and complex:Pb ratio. The use of a hydroxide complex gives the widest range of
crystal sizes. PbSe often exhibits a bimodal distribution of crystal sizes. PbS, in
contrast, tends to form only relatively large crystals.
       Chemical deposition films of PbS and PbSe are generally not strongly tex-
tured. One report has described (200) textured PbS films on glass if H2O2 is pre-
sent in the deposition solution.

2.9.2.3     Electrical Properties
The IV–VI films are usually p-type, both as deposited and after annealing in air.
One study, where PbS was deposited from a bath containing hydrazine, found the
deposit on glass to be n-type temporarily but converted to p-type on air exposure.
By depositing the PbS on a trivalent metal coating (such as Al), the n-type con-
ductivity could be stabilized for a longer time.
       Electrical resistivity of the films, both PbS and PbSe, has often been re-
ported to be of the order of 105 -cm as deposited, with a reduction of about an
order of magnitude after annealing in air. However, this can vary considerably
from one type of deposition to another. Resistivities greater than 109 -cm have
been reported in some cases, which invariably drop to the k -cm range after air
annealing. These high-resistivity films are probably those with a very small crys-
tal size (small meaning ca. 10 nm or less).
       There are many studies on photoconductivity in these films, many of them
early ones and focused on annealed films (since air annealing is necessary for
optimal photoconductivity). The use of a chemical oxidant (which never seems
to be specified) gives much higher photosensitivity for as-deposited films than
for films deposited without oxidant, although even here annealing is used to ob-
tain maximum performance. Some studies on photoconductivity in as-deposited
PbSe films have shown shifts in photoconductivity spectral response, with on-
sets shifted to 2.2 m instead of the ca. 4.5 m more typical of annealed films.
As with optical absorption studies, these shifts can be attributed to size quanti-
zation.

2.9.3 Other Sulphides and Selenides
A large range of other metal sulphides and selenides have been deposited by
CD. Since these will be individually described in Chapter 6, it will be sufficient
here to list all binary sulphides and selenides (along with oxides) in Table 2.1,
along with up to three references to each compound.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 2.1 Binary Semiconductors Deposited by CD

Sulphides                                         Selenides                           Oxides

Ag2S [35–37]                                 Ag2Se [38, 39]                    Ag2O/AgO [40, 41]
As2S3 [42, 43]
Bi2S3 [44–46]                                Bi2Se3 [47–49]
CdS [50–52]                                  CdSe [20, 53, 54]                 CdO [55–57]
CoS [58]                                     CoSe [59]                         CoO [60, 61]
CuxS [62–64]                                 CuxSe [65–67]                     Cu2O [68]
                                                                               FeO(OH) [29]
                                                                               Fe2O3 [69]
                                                                               Fe3O4 [70]
HgS [71–73]                                  HgSe [38, 74, 75]
In2S3 [76–78]                                                                  In2O3 [79, 80]
MnS [81–83]                                                                    MnO2 [84, 85]
Mo-S [86,87]                                 Mo-Se [86-88]
NiS [89]                                     NiSe [89]                         NiO [90, 91]
PbS [92–94]                                  PbSe [53, 95, 96]                 PbO2 [97, 98]
Sb2S3 [43, 99, 100]                          Sb2Se3 [101–103]                  Sb2O3 [102]
SnS [104–106] (also Sn2S3)                   SnSe [107]
SnS2 [42, 106, 108]                                                            SnO2 [109–111]
                                                                               TiO2 [32, 112, 113]
TlS [114]                                    TlSe [115,116]                    Tl2O3 [84, 98]
ZnS [117–119]                                ZnSe [120–122]                    ZnO [123-125]

Only three references to pure CD deposition of tellurides, all of them for CdTe, have been found
[126–128] and therefore no column for tellurides is given here. SiO2, Y2O3, and ZrO2, not given in the
table, have also been deposited, as have Ag halides.




2.9.4 Oxides, halides, and elemental Se
2.9.4.1        Oxides
While most studies on CD have been on sulphides and selenides, considerable
work has also been carried out on oxide films. The films are most often formed by
reaction of hydroxide ions with a metal salt. While it might be expected that the
product is a hydroxide rather than an oxide, in many reported cases oxides are di-
rectly formed. This is probably due to two factors: Many of the metal ions used
(e.g., Pb, Sn, Tl,) do not readily, if at all, form simple hydroxides; others (Ag, Cu,
Mn) are very readily oxidized even in aqueous solutions. Ni(II) hydroxide is fairly
readily dehydrated, particularly in the presence of the persulphate ion used to de-
posit the oxides in some cases. Zn(OH)2, Cd(OH)2 and In(OH)3 are reasonably
stable; Zn(OH)2 can be easily dehydrated while the other two require annealing to
form the oxide.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Persulphates (also called peroxydisulphates) (S2O 2 ) are very strong oxi-
                                                           8
dizing agents and have been used to deposit oxides, in particular those with
oxidation states higher than those of the original cation. PbO2, MnO2, and Tl2O3
deposited from solutions of Pb2 , Mn2 , and Tl are examples of this type of
deposition. Sometimes a small concentration of Ag ions is needed in the
deposition solution as a catalyst. Ag is a known catalyst for oxidations
using S2O 2 (S2O 2 oxidizes it to Ag(III), and this is then the active oxidizing
           8       8
agent).
       Fe-oxides, ZnO and In2O3 have been deposited using dimethylamineborane
(DMAB) or trimethylamineborane and Zn or In nitrate. The nitrate anion is im-
portant and is believed to be reduced by the DMAB to nitrite and hydroxide. It is
not clear why ZnO, and not the hydroxide, is the final product. In the case of In,
the deposited films were In(OH)3 and required annealing at ca. 200°C to form the
oxide.
       Homogeneous precipitation using urea (which hydrolyzes to give an alka-
line solution) has been used extensively, and in a few cases films of basic salts
(sulphates of Al and Sn(IV) and formate of Fe) have been obtained. These are not
considered semiconductors in the conventional sense, but do provide examples for
extension of the CD method beyond the conventional sulphide-selenide-oxide
compounds.
       Many (hydr)oxides have been deposited by slowing down natural hydroly-
sis, usually by complexation (e.g., AgO from an alkaline triethanolamine-com-
plexed silver bath). Highly-acidic cations will readily hydrolyse even under acidic
conditions. Fluoro-complexes of some of these (e.g., Ti, Si) can be controllably
hydrolyzed by addition of boric acid, which reacts with the F, thereby destroying
the complex and allowing hydrolysis.
       A (not-closed) heated solution of ammonia gradually loses ammonia. If a
cation is complexed with ammonia, the free-cation concentration will gradually
increase as ammonia is lost (a rare example of slow cation release rather than
anion release). It will also increase with an increase in temperature, due to the
decrease in stability of complexes with increasing temperature. As one example
using this principle, thin films of mixed ZnO/Zn(OH)2, which converted to ZnO
on heating over 200°C, were deposited from a heated aqueous Zn-ion/ammonia
bath.
       Apart from these methods, there are others that are relevant to this section.
Aqueous solutions of permanganate will slowly decompose, forming a brown film
[MnO2 or possibly MnO(OH)] on the walls of the vessel in which they are stored.
Increase of either acidity or alkalinity of the solution can accelerate this decom-
position reaction. As another example of film formation due to slow solution de-
composition, the author possesses a glass bottle with a green, highly tenacious
film of (presumably) Cr2O3 resulting from years of storage of some (unfortunately
unknown) Cr-containing solution.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 2.2 Electrical and Optical Properties of CD Oxidesa

                      Resistivity                              Special      Conductivity   Absorption
Material               ( -cm)                                 conditions       type        onset (nm)   Reference
                                       1
Cu2O                          10                                                 p          ca. 600        68
Fe3O4                     2         103                                                     Black         129
Fe2O3                              2                         Annealed                       ca. 570       130
                                                               at 350 C
In2O3                             109                        As deposited        n          ca. 400         80
                                  33                         Annealed
                                                               at 200 C
                                           2
In2O3                     2.10                                                                              79
MnO2                                                         pH   6.3                                       84
                                                             pH   8
NiO                               105                                            p          ca. 700        90
NiO                       3         102                                                       430         131
PbO2                              10 3                                                           1.7       98
SnO2                              10 1                       Annealed            n            350         109
                                                               at 250 C
                                               3
                      2            10                        Annealed            n
                                                               at 400 C
                                           3
SnO2                              10                         Annealed            n            310
                                                               at 250 C
                                           4
(5% Sb)                           10                         Annealed            n
                                                               at 400 C
TiO2                     109                                                                               32
                                               4
Tl2O3                 4 10                                                                    600          98
ZnO              2    104 to 4                     102       Depending           n            380         132
                                                               on boron
                                                               content
ZnO                               10 2                                                        450         133
                              4                          2
ZnO:Al       2       10           to 2             10                                         340

In contrast to the rest of this book, optical properties are given as an approximate absorption edge in the
absorption (transmission) spectrum, rather than as a value of bandgap. This gives those who are not
familiar with semiconductors a better feel for the appearance of the film.
a
  As deposited, unless stated otherwise in the “Special conditions” column.


         Since one common use of oxide films is for transparent, conducting coat-
  ings, the resistivities of these films were usually measured. Table 2.2 shows some
  basic electrical and optical properties of some of these films.
  2.9.4.2            Halides
  Silver halides, in particular AgI and AgBr, have been deposited by hydrolysis of
  halogenoalcohols (halohydrins) to free halide ions in a solution of Ag under



  Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
acidic conditions. For AgCl, a better method was found to be simple precipitation
by adding a solution of NaCl to one of AgNO3. The concentration of the solutions
is important for this latter deposition—ca. 20 mM (much larger concentrations re-
sult in rapid aggregation and little film formation, while much lower concentra-
tions give a very slow deposition rate). A visible AgCl film (visible by scattering)
is formed very rapidly—in seconds—which is unique in CD, which normally re-
quires much longer to form a visible film. The crystal size of these silver halide
films is large compared to most CD films, rarely less than 100 nm and sometimes
as large as a micron.

2.9.4.3      Elemental Se
Virtually all the semiconductors deposited by CD are compound semiconduc-
tors, the one exception being elemental Se. This has been deposited from solu-
tions of selenosulphate, which rapidly form Se if acidified. By control of the pH,
this reaction can be controlled to allow Se deposition to occur. Se films have
also been deposited from colloids of Se (prepared by reducing SeO2 solutions)
by photodeposition, whereby the light activates the formation of films.

2.9.5 Coprecipitation of Metal Chalcogenides—
      Ternary Compounds
Chemical deposition is not limited to binary compounds. Ternary (and higher)
compounds can be deposited by this technique. For the same reason as for the non
II–VI and IV–VI compounds in Section 2.9.3, this section will suffice with a table
of ternary compounds reported up to now, with two additions. The first is a brief
consideration of the principles involved in the deposition of materials containing
three or more elements. The second is to identify, in the table, which deposits have
been clearly demonstrated to be a true single-phase solid solution rather than a
mixture of two or more phases.
       If, e.g., thiourea is added to a mixture of Cd and Zn ions complexed with
ammonia, then, depending on the mechanism and experimental conditions of de-
position, the deposit could be CdS, ZnS, a mixture of the two, a single-phase
CdxZnxS compound, or some combination of these. Structural characterization,
most commonly XRD, together with elemental analysis, can usually reveal the na-
ture of the product.
       There a number of factors involved in determining just what the product
turns out to be. An obvious one is consideration of the solubility of the various
products (and intermediates): The lower the value of Ksp of one binary product rel-
ative to another, the more likely that product is (in principle) to deposit preferen-
tially. This simple consideration is complicated by a number of other factors. One
is the tendency of metal ions to coadsorb on the (usually high-surface-area) pri-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mary deposit, even if that metal chalcogenide (or whatever compound) is much
more soluble than the primary one. Such adsorbed species can become occluded
in the growing crystal as subsequent layers are built up. Another possibility, most
likely for codeposition of structurally similar materials, is formation of a true solid
solution. Even if the relevant binary materials are not structurally similar, a true
single-phase ternary deposit may be obtained, although in this case the composi-
tion range is likely to be much narrower than in the previous case. For codeposi-
tion of two different cations, compound formation may occur, particularly if the
two cations have quite different acidic and basic properties. Chapter 8 describes
these principles in more detail.
       Table 2.3 lists ternaries that have been deposited, together with indication
of when clear single compounds formation was verified. While solid solution
formation is usually the goal of these studies, it should be kept in mind that
separate phases, either as a composite or as separate layers, may be required
for some purposes. For example, bilayers of CdS/ZnO and CdS/ZnS have
been deposited from single solutions. These depositions depend on the prefer-
ential deposition of CdS over ZnS and, in the case of the former, the often-en-
countered greater ease of formation of the oxide (hydroxide) than the sulphide
of Zn.

TABLE 2.3 Ternary Materials Deposited by CD

Material                   References                     Material                    References

(Cd, Bi)S            134                             In(S, OH)                  135, 136*, 137, 138
(Cd, Hg)S            139, 140*                       (Pb, Hg)S                  144–144*
(Cd, Pb)S            140, 145, 146–150*              Pb(S, Se)                  151*
Cd(S, Se)            152–154*, 155, 156a             (Sb, Bi)2S3                157* (after anneal)
(Cd, Zn)S            158, 159*, 160, 161,            Sn(S, OH)                  165
                     162–164*
(Cd, Zn)Se           166*, 167                       (Zn, Cd)O                  168*
CuBiS2               169                             Zn(O, OH),                 170
                                                       Zn(S, O, OH)
Cu3BiS3              171* (after anneal)             Zn(S, OH)                  170, 172
CuInS2               173, 174*b                      Zn(Se, OH)                 175
CuInSe2              176–180*                        Zn(S, Se)                  181, 182*
Bi2 (S, Se)3         183*                            (Pb, Cu)S                  184 (two phase)
(Pb, Sn)Se           185*                            (Cd, Sn)O                  133*
(Pb, Bi)S            186*

Where a predominantly single phase (even over only part of the composition range) was at least
reasonably clearly demonstrated or could be inferred from the results, at least at some composition, the
relevant reference is followed by an asterix, although sometimes this refers to annealed films.
a
  Probably solid solution, based on Ref. 154.
b
  See relevant section in Chapter 8.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2.10 NONAQUEOUS DEPOSITIONS
2.10.1 Introduction
Although almost all CD processes have been carried out in aqueous solution, there
are a few examples in nonaqueous media (specifically, in carboxylic acids). It is
worth considering the differences between aqueous and nonaqueous deposition in
general first.
       Solubility is an important criterion. The reagents involved need to be solu-
ble in the solvent (this is, of course, obvious, but sometimes the obvious needs
pointing out). This limits the choice of reagents, compared to aqueous solutions.
For example, sodium selenosulphate, the most common selenide-producing
reagent, is insoluble in most organic solvents at any useful concentration. Sele-
nourea (or one of its derivatives) is more useful in this case. When considering the
choice of metal salt, halides and particularly iodides as well as perchlorates tend
to be more soluble in organic solvents than most other common anions.
       Next there is the question of whether the anion-producing reaction, which is
normally a hydrolysis, can occur in the absence of water. Alternate reactions may
be needed. This is not a problem for depositions which occur by the complex de-
composition mechanism.
       Finally, even if these criteria are satisfied, there remains the question of
whether the product will adhere to form a film or just precipitate homogeneously
in the solution. This is the most difficult criterion to answer a priori. The hydrox-
ide and/or oxy groups present on many substrates in aqueous solutions are likely
to be quite different in a nonaqueous solvent (depending on whether hydroxide
groups are present or not). Another factor that could conceivably explain the gen-
eral lack of film formation in many organic solvents is the lower Hamaker con-
stant of water compared with many other liquids; this means that the interaction
between a particle in the solvent and a solid surface will be somewhat more in wa-
ter than in most other liquids (see Chapter 1, van der Waals forces). From the au-
thor’s own experience, although slow precipitation can be readily accomplished
from nonaqueous solutions, film formation appears to be the exception rather than
the rule. The few examples described in the literature are confined to carboxylic
acid solvents (see later).
       What are the advantages of deposition from nonaqueous solutions? One is
the possibility to form films of compounds that are soluble, or not sufficiently in-
soluble, in water. A (potential) example of this is formation of halides using, e.g.,
chlorohydrins, which are in general soluble in organic solvents, to generate Cl .
       This could be expanded to materials that tend to form hydroxy-chalcogenides
rather than pure chalcogenides, such as ZnS and In2S3. If water (including water of
hydration from the salts) is rigorously excluded, the hydroxy impurity cannot form
from solvents that do not contain hydroxy groups.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
2.10.2 Examples of CD from Nonaqueous
       Solutions
Films of various stoichiometries of Sn-S have been deposited from carboxylic
acid (acetic, propionic, butyric) solutions of elemental S and SnCl2. Depending on
deposition conditions, in particular whether some water was present and how
much, as well as the presence of complexing agents, films of approximate com-
position SnS, Sn2S3, or SnS2 could be formed. Interestingly, various Sn-S films
were also formed on the walls of the deposition vessel above the liquid level (by
several centimeters); this was attributed to reaction between volatile SnCl4 and
H2S, both formed in the deposition bath.
      Bi2S3 was deposited from glacial acetic acid solutions of Bi(NO3)3 mixed
with formaldehyde solutions of Na2S2O3 or thioacetamide. Sb2S3 films were de-
posited in a similar manner from acetic acid solutions.
      These depositions are described in more detail in Chapter 6.

2.11 RELATED DEPOSITION TECHNIQUES
The CD technique is based on either slow formation of a reactive anion or slow
decomposition of a complex compound. However, there are other techniques, not
involving these slow steps, that nevertheless are sometimes called chemical de-
position or chemical solution deposition or are closely related. These techniques
are dealt with very briefly in this section.

2.11.1 Successive Ion-Layer Adsorption and
       Reaction (SILAR) Process
The SILAR process is, as its name suggests, somewhat analogous to molecular
beam epitaxy (MBE), although the films obtained are most often not epitaxial.
Like MBE, SILAR proceeds via a layer-by-layer buildup of the film, except in so-
lution instead of in a vacuum. In the SILAR process, the substrate is immersed
first in a solution containing the metal cation, rinsed, then immersed in a solution
containing the desired anion, and again rinsed. This gives (ideally) one monolayer
of the deposit. The process is then repeated for as many times as needed to obtain
the required thickness. (It is probably not important if the anion or the cation is ad-
sorbed first; after the first cycle, the process should be the same, although it is con-
ceivable that differences in adhesion to the substrate may result, depending on the
initial order.) The rinsing steps are important, since without them relatively large
reservoirs of one ion would remain on the substrate, and clusters of the semicon-
ductor, rather than a film, would result. In fact, by omitting the rinsing step, films
can be built up much more quickly. Thus, by successively immersing a glass slide
in fairly strong solutions of Na2S and a Cd salt (say, 0.1M), visible yellow films




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 2.7 Simplified schematic diagram of an automated SILAR process for CdS.


of CdS appear after only several cycles. However, such films are less likely to be
homogeneous, compared with a properly prepared SILAR film. Cu2O, ZnO, and
SnS have been deposited in this way (see Ref. 187 and references therein). The
true SILAR process is based on the expectation that, after each rinse step, only one
monolayer of the previously adsorbed ion will remain, which will favor a layer-
type growth.
       The technique is slow and tedious, but automation of the process can be car-
ried out, whereby the substrate is attached to stepping motors that alternately im-
merse and remove it from a series of beakers. This is shown schematically in Fig.
2.7 for the example of CdS. The substrate is attached to an arm that can be moved
both vertically (to immerse/remove it) and horizontally (or in a circle) to position
it above different reaction vessels.
       While the layers are usually polycrystalline, epitaxial layers of both
zincblende and wurtzite CdS have been grown on various single-crystal substrates
with lattice parameters close to those of CdS, as might be expected for an ion-by-
ion growth.
       Besides CdS, many other semiconductors have been deposited by the
SILAR technique as well as organic conducting polymers, such as polypyrrole
and polyaniline. For representative references and to locate some groups working
in this field, see Refs. 188–195.

2.11.2 Pyrolysis of Precursor Films
This method is based on pyrolysis of a metal chalcogenide–containing precur-
sor. Heating CD hydroxide films to form oxides is a simple example and is com-
mon in the deposition of some oxides. This will be treated in Chapter 7, which



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
deals with oxides. Somewhat further from CD as understood in this book, the
method is also widely used as the well-known sol-gel method for oxide films, in
particular SiO2 and TiO2. In the sol-gel method, a more-or-less viscous colloidal
gel is deposited on a substrate either by dipping the substrate in the gel and
slowly raising it from the solution or by spin-coating. The film, whose thickness
can be controlled by the viscosity of the sol and either the rate of removal from
solution or the rotational speed of the spin-coater, is then pyrolyzed to form the
oxide.
       This technique is not so easily extended to nonoxides. Sulphur, for example,
does not form the cross-linking bonds needed to form the sol-gel as readily as does
oxygen. However, a related method has been used, albeit to a very small extent, to
form CdS films. It is based on the thermal decomposition (at ca. 300°C) of a
Cd–thiourea complex, which is formed as a film by slowly withdrawing the sub-
strate from a methanolic solution of a Cd salt and thiourea [196].
       This can certainly be extended to other metal sulphides, using other com-
plexes of sulphur (and also selenium). However, the complex and anion of the
metal salt need to be chosen so that all the by-products of the pyrolysis reaction
are volatile, otherwise the film will be contaminated with the nonvolatile by-prod-
ucts. For example, using cadmium nitrate and thiourea, all the by-products are
volatile:
      Cd(NO3)2     (NH2)2CS → CdS        CO2     2H2O      2N2O             (2.20)


2.11.3 Exchange Reactions
Exchange reactions to convert one material into another by immersion in suitable
solutions are well known. Such a reaction is the basis for formation of the once-
popular CdS/Cu2S photovoltaic cell, where a CdS film was immersed in a CuCl
solution and part of the film converted to Cu2S. Since the solubility product of
Cu2S is lower than that of CdS (or, in electrochemical language, Cu is more noble
than Cd), exchange is thermodynamically favorable. Examples of such exchange
reactions using CD films are CdS and CdSe to the corresponding silver salts [197]
and SnS2 to Ag2S [198].

2.11.4 Intermixing by Annealing Multilayers
By depositing two (or more) different layers and annealing them, intermixing of
the layers can lead to ternary and multinary compounds, although clear compound
formation does not always occur. Thus, annealing (at 150°C, a relatively low tem-
perature) ZnS-CuxS and PbS-CuxS films resulted in extensive interdiffusion of the
metallic elements but no XRD confirmation of solid solution formation [199]. On
the other hand, Sb2S3-CuS layers converted fully to CuSbS2 at 400°C, which ex-




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
hibited p-type conductivity and a direct bandgap of 1.5 eV [200]. By evaporating
In onto CD Sb2S3, InSb was formed after annealing at 300°C.

2.11.5 Gas–solution Interface Reaction
Films of semiconductors up to a few hundred nanometers in thickness have been
formed by the simple reaction between H2S or H2Se and the surface of an aqueous
solution of the metal ion. This technique was described long ago for PbS and PbSe
[201]. A more comprehensive description of the method, extended to a number of
different metal sulphides, has been given [202]. It is stressed that the gas phase is
passed over the solution surface and not bubbled through the solution, which would
break up the film. In fact, if the gas flow continues for too long (typically more than
a few minutes), the film tends to break up and precipitate. Since the substrate for
these films is a liquid surface, the films can be (carefully) picked up and transferred
to another surface or possibly even be self-supporting in small areas.


2.12 SPECIFIC TOPICS RELEVANT FOR
     CD FILMS
2.12.1 Solar Cells
2.12.1.1 Solid-State Photovoltaic Cells
Probably the most important factor responsible for the renewal of interest in the
CD technique is the almost universal use of CD CdS films in thin-film photo-
voltaic cells based on either Cu(Ga)InSe2 (abbreviated here as CIS, which in-




FIG. 2.8 Schematic diagram of the CIS/CdS and CdTe/CdS photovoltaic cells. The back
contact to the CdTe cell—Cu-doped carbon paste – is a commonly used one, but there are
several modifications to this contact as well as completely different ones in use.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
cludes Cu(In,Ga)Se2, usually abbreviated in the literature as CIGS, CuInSe2, and
CuInS2) or CdTe absorber films. Schematic diagrams of these cells are shown in
Figure 2.8. The common component of both is a thin (50–100 nm) film of CD CdS
called the window layer [since incident light (ideally) passes through it] or the
buffer layer (which points out the lack of understanding of exactly what this layer
does). The CIS cell is a substrate (frontwall) cell (light passes through the semi-
conductors in the direction of the substrate), while the CdTe cell is normally in the
superstrate (backwall) configuration (light passes first through the substrate).
       With only a few exceptions, the CdS is deposited from a standard ammo-
nia/thiourea bath at ca. 70°C, with variations in the concentrations of reactants, the
use of temperature programming, and some variation in pH (using an ammonium
chloride buffer). It is notable that, in spite of many attempts to substitute CdS by
another CD material (driven by the desire for a environmentally friendlier mate-
rial), CdS remains the best material to date for this purpose, both for CIS and CdTe
cells. Other materials deposited by CD include various Zn(OH,S), Zn(OH,Se),
and In(OH,S) “compounds” and In(OH)3. The three first materials appear to be in-
completely sulphided or selenided hydroxides, and it is not clear whether they are
a mixed or a single phase. Also, it is usually unclear whether oxide or hydroxide
also occurs [although one XPS study of In(OH,S) has demonstrated the absence
of either In(OH)3 or In2S3 in the film]. While some of these buffer layers approach
CdS in terms of cell efficiency, they are invariably inferior.
       Studies have been undertaken in an attempt to understand why CdS appears
to be so unique in this role and why CD is the best technique to deposit it. The CD
solution clearly plays an important role, not only in depositing the CdS, but also
in its effect on the absorber surface. In the case of CuInSe2, the solution removes
native oxides (of In, Cu, and Se), removes excess CuxSe, and forms an interface
that is Cu-deficient and contains Cd. In fact, Cd has been shown to diffuse a small
but appreciable distance (ca. 10 nm) into Cu(Ga)InSe2 films but not into CuInSe2
single crystals. It is not clear whether this is due to differences in composition or
in crystallinity. However, the diffusion of Cd into CIS is believed to be related to
the presence of a Cu-depleted region at the CIS surface; the CD bath, as already
noted, is instrumental in forming such a region. Additionally, it is thought that re-
moval of surface oxygen substituting for Se vacancies by replacement with S may
increase band bending by modification of the surface charge.
       Since the envisionaged application of a CD process in thin-film solar cells
is a large-scale one, efforts have been made to optimize the deposition process
used, particularly in minimizing the waste Cd-containing solutions. Dilute Cd so-
lutions (ca. 1 mM), a flow system with filtration, and a heated substrate have been
employed to this end. The heated substrate means that deposition occurs prefer-
entially on the substrate rather than on the cooler walls of the deposition vessel.
Also, ethylenediamine has been used as a complexant rather than the much more
volatile ammonia.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      There are a number of studies on other photovoltaic properties of CD films
(see Chapter 9). As an example, p-n junctions have been fabricated by depositing
PbS on a glass substrate partially coated with a trivalent metal, such as Al. The
PbS deposited on the glass is (as usual) p-type, while that deposited on the metal
coating was n-type, at least for some time. Photovoltages up to 0.1 V (at room tem-
perature) and 0.28 V (at 90 K) were measured (the bandgap of PbS is ca. 0.4 eV –
even lower at low temperatures so only small photovoltages are expected).
2.12.1.2 Photoelectrochemical Cells
One of the attractive features of CD is its simplicity (in terms of carrying out the
deposition, that is, not always in understanding the deposition itself). The same
property of simplicity is often ascribed to photoelectrochemical cells (PECs).
Therefore it is not surprising that CD has often been used to fabricate the semi-
conductor electrodes for PECs.
       A PEC is the liquid junction analogue of a solid-state Schottky cell. In its
simplest configuration, a semiconductor with an ohmic contact is immersed in an
electrolyte and connected via a load to a second counterelectrode (often platinum
or graphite). Super-bandgap light incident on the semiconductor creates elec-
tron/hole pairs that are separated by the electric field (space charge layer) formed
by the contact between the semiconductor and the electrolyte. One charge type
(holes, for the more common case of n-type semiconductors) is injected into the
electrolyte. The holes oxidize some electrolyte species, while the electrons are ex-
tracted through the ohmic contact and flow through an external circuit to the coun-
terelectrode, where a reduction occurs. If the same species is both oxidized (at the
semiconductor electrode) and reduced (at the counterelectrode), no net change oc-
curs in the electrolyte, and electrical energy is generated in the external circuit.
This type of cell is called a regenerative PEC. If different species are electrolyzed,
chemical energy can be produced. Most PECs made using CD semiconductor
films are of the former type.
       In nanocrystalline semiconductor films (commonly obtained in CD), the
crystal size may be too small to support an appreciable space charge layer.
Charges in that case are separated by differing kinetics between electron and hole
injection into the electrolyte. The upcoming discussion on nonannealed films
treats this in somewhat more detail. (Chapter 9 discusses PECs and their princi-
ples of operation more fully.)
       The majority of PEC studies have been carried out on either CdS or CdSe
films, although many other CD semiconductors have been demonstrated to exhibit
PEC activity. In most, though not all, cases, these films were annealed for opti-
mum PEC response. Films annealed at temperatures above ca. 300°C usually ex-
hibit a large degree of crystal growth, and therefore such films will be discussed
separately from as-deposited films, which, in most cases, are composed of crys-
tals 20 nm in size.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Annealed Films. For many years, the CdSe photoelectrode—polysulphide
electrolyte PEC—was probably the most studied system in PEC research. The
bandgap of (bulk—see later) CdSe is ca. 1.7 eV, which is close to the theoretical
optimum of 1.5 eV for photovoltaic cells in general; and in relative terms, that sys-
tem was fairly stable in terms of self-oxidation of the semiconductor film in the
electrolyte by the photogenerated holes.
       Simulated solar conversion efficiencies up to 6.8% on Ti substrates have
been reported for annealed CD CdSe films in polysulphide electrolyte based on a
low-ammonia-concentration–selenosulphate bath. Several successive depositions
were required to build up an optimum final film thickness of 2.5 m (when most
of the light was absorbed). The initial deposit was annealed to improve adherence
and the final multideposited film was annealed at 550°C in air, followed by etch-
ing and zinc ion treatment.
       By incorporating silicotungstic acid (STA) in the deposition bath, larger
particle size (after annealing at 430°C) and much better conversion efficiencies
(compared to the STA-free bath) were obtained—11.7% based on tungsten-halo-
gen illumination (solar efficiency is lower). The effect of the improved perfor-
mance was not clear. It was suggested that the STA improved charge transfer ki-
netics at the CdSe/electrolyte interface. The larger particle size may also be an
important factor.
       Chemical deposition CD CdS has shown much lower efficiencies in a PEC.
This is due to its higher bandgap, which allows only a small fraction of solar ra-
diation to be absorbed in the film.
       Nonannealed films. Although the conversion efficiencies are much lower
than those of annealed films, the PEC properties of as-deposited films show other
interesting properties, connected with their nanocrystalline and somewhat porous
morphology. As already noted, there is usually no appreciable space charge layer
in these nanocrystals. Since the films are porous and electrolyte can reach (most
of) the surface of all the nanocrystals, charge generation can be considered to take
place at the surface of the individual crystals—there is no need for a field to pro-
vide the driving force for charge drift. Electrons and holes are then separated by
different kinetics for electron and hole transfer to the electrolyte, which in turn is
affected by the relative trapping efficiencies of the charges at the nanocrystal sur-
face. If one charge (say, holes) is removed rapidly by the electrolyte, the electrons
can get to the back contact without recombination. In practice, much of the pho-
togenerated charge is lost, probably by injection of both carriers into the elec-
trolyte (indirect recombination).
       A unique property of these films results from this mechanism. CdSe, as de-
posited, behaves as an n-type semiconductor (holes are transferred to the elec-
trolyte, while electrons are extracted at the back contact). In this case, the holes
are preferentially trapped at the surface, and are more readily injected into solu-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tion. Mild etching in dilute HCl changes the distribution of trapping states at the
surface in such a way that electrons are now more readily trapped and are prefer-
entially injected into the (identical polysulphide) electrolyte, the film behaving as
a p-type semiconductor. It is notable that this behavior, typical for a crystal size of
4–5 nm, is not observed when the crystal size is 20 nm. The very high surface area
and high density of trapping states appear to be determining here.
       As-deposited CdS has been studied as a photoelectrode with various
dopants (Al, As, Cu) incorporated in the deposition bath. The emphasis in these
studies was on PEC efficiencies, which were very low in all cases, although dop-
ing, particularly with As, did have a beneficial effect.
       An interesting variation of photoactivity has been observed during the de-
position of several semiconductor films (CdSe, CdS, PbS, Bi2S3). Illumination
during deposition increases the deposition rate and, in some cases, increases the
crystal size somewhat. While some effects can be ascribed to local heating, in-
creasing deposition rate, the main effect is probably electron-hole generation by
super-bandgap light and reduction of the chalcogen species to chalcogenide by the
photogenerated electrons at the growing crystal surface, resulting in the formation
of more metal chalcogenide.
       Some other semiconductors have been investigated as photoelectrodes, in
all cases giving low but appreciable photoactivity. These include Bi2S3, PbSe,
Sb2Se3, SnS2, HgS, and Ag2S. Chemical deposition has also been used to form
both the photoelectrode (CdSe) and the storage electrode (Ag2S) in a photoelec-
trochemical storage cell, where the Ag2S acts as an in situ storage electrode.

2.12.2 Quantum Size Effects in CD Films
Films of materials deposited at or near room temperature (and in this respect
100°C is considered to be near room temperature) tend to have a small crystal size.
This is not surprising since high temperatures are normally required to impart suf-
ficient mobility to a freshly deposited species in order for recrystallization to oc-
cur. This small crystal size, which at one time was almost universally considered
to be a disadvantage, is increasingly considered to be an advantage as interest in
nanocrystalline and nanoparticle materials grows. The term nanocrystalline usu-
ally refers to materials with a crystal size from a nanometer up to hundreds of
nanometers (at this upper limit, the term microcrystalline starts to take over).
      The size of the crystals formed in CD films is often small enough that quan-
tum-size effects become apparent. The terms quantum size effect and size quanti-
zation are normally used to describe a material whose energy structure differs
from that of the bulk material. As crystal (or, more generally, particle) size de-
creases, charges (electrons and holes) in the particles are constrained in an in-
creasingly small volume. When the particle size becomes smaller than the Bohr
diameter of the charges in the bulk material (between 2 and 20 nm for many ma-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
terials), the quantum-size effect is manifested as an increase in the bandgap, Eg,
of the material and separation of the energy bands into discrete levels. The in-
crease in Eg is most commonly seen as a blue shift (i.e., to higher energy) of the
optical absorption spectrum (see Chapter 10 for more details on this topic.)
       This effect is shown in Figure 2.9 for CdSe films deposited from baths con-
taining Cd complexed with NTA (nitrilotriacetate) and Na2SeSO3 as a Se source.
The nanocrystal size, measured by both XRD and TEM, varied from ca. 3 nm up
to 20 nm with increase in temperature and/or change in mechanism from a cluster
mechanism to an ion-by-ion deposition. The optical bandgap shifts from 1.8 eV
(for bulk, zincblende CdSe) to ca. 2.4 eV for the smallest nanocrystals (ca. 3 nm).
       The main difference between the two mechanisms as they relate to crystal
size (discussed in Sec. 2.6) is that the cluster mechanism is three dimensional
while the ion-by-ion one is mainly two dimensional. Crystal size in the former is
limited largely by the amount of reactant per nucleus: The more nuclei, the smaller
the final crystal size, since the same concentration of reactants is divided over
more nuclei. Temperature affects this by stabilizing (kinetically) smaller nuclei as
temperature is lowered, thus increasing the number of nuclei at lower temperature,




FIG. 2.9 Transmission spectra of CD CdSe films deposited at various temperatures from
CdSO4/NTA/Na2SeSO3 baths. All samples deposited by hydroxide cluster mechanism ex-
cept 80°C HC (high complex), which proceeded via the ion-by-ion mechanism. The effective
bandgap can be approximated by the wavelength (photon energy) a little shorter (higher) than
the absorption onset. A second absorption knee, ca. 0.4 eV to higher energy of the initial on-
set, seen clearly in the 41°C and 80°C samples, is due to a transition from the spin-orbit va-
lence level to the lowest conduction level and is commonly observed in these films.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
resulting in a smaller final crystal size. For the two-dimensional ion-by-ion
mechanism, the final crystal size is probably determined by the number of nuclei
on the substrate (initially bare, eventually covered with deposit), which grow
until they touch other crystals and cannot grow further in the horizontal (to the
substrate) direction. Increased temperature also may allow increased coalescence
of the individual crystals into larger ones.
       Other CD semiconductors have been shown to exhibit size quantization.
PbSe shows the effect very clearly, since quantum size effects can be clearly seen
in this material, even in crystals up to several tens of nanometers in size (due to
the small effective mass of the excited electron-hole pair). Shifts of greater than 1
eV have been demonstrated, from the bulk bandgap of 0.28 eV to 1.5 eV.
       CdS, deposited from the usual ammonia/thiourea bath, normally gives films
that are not quantized, with crystal size larger than 10 nm. Some exceptions do oc-
cur, however. Small increases in bandgap have been found when high thiourea
concentrations were used. At high values of pH ( 12), smaller crystal sizes (down
to ca. 5 nm) have been obtained with blue spectral shifts corresponding to in-
creases in bandgap of up to a few hundred meV for very thin films. The crystal
size increased with increase in film thickness. Using NTA as a complex and work-
ing under conditions where the cluster mechanism is operative, 5-nm nanocrystals
of CdS exhibiting quantum size effects have also been obtained. This crystal size
did not vary much with film thickness. Using a process of electrochemically in-
duced CD of CdS (see Chapter 4), nanocrystalline CdS films were grown using 2-
mercaptoethanol as a strongly adsorbing growth-termination (capping) agent. By
increasing the concentration of mercaptoethanol, crystal size was reduced down
to 4.1 nm (and bandgap increased up to 2.69 eV).
       Both ZnS and ZnSe films have been grown that show moderate increases in
bandgap (up to a few hundred meV). The Zn chalcogenides generally exhibit
smaller increases in bandgap than the corresponding Cd compounds of the same
size, due to their larger effective masses.
       HgS, deposited from thiosulphate solution, has been described with crystal
sizes that depend on deposition temperature, from 3 nm to 8 nm and correspond-
ing variation in apparent bandgaps from 2.4 to 1.9 eV.
       HgSe has been deposited by CD exhibiting different bandgaps and ones
strongly shifted from the bulk value (bulk HgSe is a semimetal with a negative
bandgap). Values as high as 2.5 eV for 7.7 nm crystal size have been reported. An-
other bath composition gave a bandgap of 1.42 eV, although this was not ex-
plained through size quantization but because of an amorphous structure.
       Bi2S3 has been deposited with film thickness–dependent crystal sizes, from
5.2 to 8 nm and corresponding bandgaps, measured from the absorption spectra,
from nearly 2.3 down to 1.8 eV.
       Cu-S films, of various stoichiometries, have also shown small quantum size
effects.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Apart from these reported quantum size effects, there are a number of ma-
terials that, while quantum size effects have not been explicitly invoked, have
been reported to show blue spectral shifts that may be explained in the same man-
ner. These include Cu1.8S, various CuS compositions (Eg 1.8—2 eV), Sb2S3, and
SnS. These systems are all described in Chapter 10.
       While shifts due to size quantization have most commonly been seen in ab-
sorption spectroscopy, other spectroscopies, such as photoelectrochemical pho-
tocurrent, photovoltage (using a vibrating Kelvin probe), photoluminescence, and
photoconductivity spectroscopies have all shown quantum shifts in various CD
films.

2.13 APPLICATIONS (ACTUAL AND POTENTIAL)
     OF CD FILMS
We conclude this chapter with applications of CD semiconductor films, both those
that have been realized and potential uses.
        The most important use of CD films for many years was to make PbS and
PbSe films for photoconductive detectors [10]. Such detectors, made by CD, are
still in use today, although they are facing competition from photovoltaic III–V
detectors. It should be noted that for good photosensitivity, air-annealing of the
CD films is carried out, and this annealing treatment is connected with partial
oxidation of the PbS and PbSe surfaces.
        Today, the the most important “application” for CD films is the use of CD
CdS as the window (or buffer) layer in thin-film photovoltaic cells [16]. Both
CdTe- and CuInSe2-type absorber films use this procedure. Such cells have
reached the pilot plant stage, and there appears to be no obvious competitor for the
CD CdS at present.
        The study of CD semiconductors, and in particular CdSe and CdS, for use
as photoelectrodes in photoelectrochemical cells is connected with this use, al-
though much farther from likely application. This study was driven, to a large ex-
tent, by the simplicity of deposition of the films, a particularly sought-after re-
quirement for this purpose, both from the point of view of applications and
because it allowed many groups (usually chemists) who maybe did not have ready
access to conventional vacuum deposition systems to prepare the films.
        The success of CD CdS in photovoltaic cells has driven related research
with potential applications in other semiconductor devices. Since the CD process
seems to play a role in the favorable properties of the CdS windows by decreas-
ing interface recombination, studies of its passivation properties on other inter-
faces and surfaces have been carried out, with considerable success. For example,
when a very thin film (ca. 6 nm) was deposited between InP and SiO2, the result-
ing reduction of the interface state density led to improved electrical properties of
metal-insulator-semiconductor capacitors and field effect transistors (FETs)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
based on this interface [203,204]. These improvements were correlated with re-
moval of native oxides and protection against oxide regrowth by the CdS, as well
as with removal of phosphorus vacancies at the interface. Improvements were also
obtained in InAlAs/InGaAs transistors and metal-semiconductor-metal photode-
tectors by the same treatment [205]. It is clear that these effects are connected not
only with the properties of the CD films but also with reaction between the sub-
strate semiconductor surface and the CD solution.
       Another example of CD films applied toward electronic devices is the fab-
rication of thin-film field effect transistors by depositing CdS (50 nm) onto oxi-
dized n -Si and annealing at 400°C [206].
       Another potential application is for solar control coatings. Thin films of cer-
tain sulphides—in particular those of copper and lead—are reasonably transmis-
sive to visible solar radiation while reflecting most of the infrared radiation. If
used as a window coating, the heat rejection from the solar radiation will result in
reduced heating of the interior by solar radiation, compared to a noncoated glass
window [23,24]. Such coatings are commonly in use, particularly in large build-
ings, and often take on an attractive golden appearance by reflected light. They are
applied by vacuum techniques. If the material usage is sufficiently high (as has
been demonstrated for CD CdS used for photovoltaic cells), CD is an attractive al-
ternative to deposit these coatings.
       The high infrared reflectivity of these films implies highly doped (therefore
relatively highly electrically conducting) films. However, another type of solar
control coating, a solar shield for passive cooling, requires films that have a high
transmission in the mid-infrared. Passive cooling occurs when a surface emits
more radiation than it absorbs [207]. The (cloudless) atmosphere is relatively
transparent to mid-infrared radiation (between ca. 8–13 m—the atmospheric
window); therefore, radiation of these wavelengths can be emitted from a surface
at ambient temperature through the atmosphere to space, which is a sink at a very
low temperature (theoretically 4 K, in practice higher, but still very much lower
than the surface of the Earth). This leads to a cooling effect at night and explains
why car surfaces and grass become covered with much more dew during a clear
night than a cloudy one. However, during the day, solar radiation would swamp
this cooling effect. To prevent this, a shield is required that is transparent to the
8–13- m region but that blocks the complete solar spectrum. Low-bandgap semi-
conductors, such as PbS and PbSe on polyethylene (one of the very few materials,
readily available in large areas, that is transparent to the atmospheric window re-
gion), are suitable for this purpose [208], but they must be close to intrinsic to pre-
vent free carrier absorption and reflection in the infrared region. Nanocrystalline
films (and CD films are often nanocrystalline) are more likely to be intrinsic than
large-grain ones: unlike a macroscopic crystal, it is possible to obtain small
nanocrystals with zero doping concentration, i.e., truly intrinsic.
       Other potential applications for CD films have been suggested and studied.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Related to the solar control coatings just described, solar absorbing coatings can
be used, e.g., for water heating. Multilayered stacks of CD PbS and CdS on Ni-
coated Cu can be configured to minimize reflectivity in the solar region and min-
imize emittance in the thermal infrared region [209]. The relatively high electri-
cal conductivity of CD CuxS films has been exploited to form an electrical contact
to ferroelectric films, as a partially transparent conducting film on plastic and also
as a cupric ion sensor [210].
       Our brief overview has given uses for CD films and rationales for their
study. It should be clear that more uses are likely to develop as the present resur-
gence in interest increases the pool of knowledge in the field and allows deposi-
tion of better-quality films, higher reproducibility, new materials and old ones in
different forms than usual.


REFERENCES
  1.    J Liebig. Ann. Pharmaz. 14:134, 1835.
  2.    C Puscher. Dingl. J. 190:421, 1869.
  3.    E Beutel, Z Angew. Chem. 26:700, 1913.
  4.    E Beutel, A Kutzelnigg. Z. Elektrochem. 36:523, 1930.
  5.    E Beutel, A Kutzelnigg. Monats. 58:295, 1931.
  6.    J Emerson-Reynolds. J. Chem. Soc. 45:162, 1884.
  7.    G Rosenheim, W Stadler, VJ Mayer. Z. Anorg. Chem. 49:1, 1906.
  8.    G Rosenheim, W Stadler, VJ Mayer. Z. Anorg. Chem. 49:13, 1906.
  9.    RH Bube. Photoconductivity of Solids. New York: Wiley, 1960.
 10.    DE Bode. In: Physics of Thin Films, Vol. 3. Academic Press, New York and Lon-
        don, 1966, p 275.
 11.    SG Mokrushin, YV Tkachev, Kolloidn Zh. 23:438, 1961.
 11a.   ET Allen, JL Crenshaw, HE Merwin. Am. J. Sci. 34:341, 1912
 12.    H Uda, H Taniguchi, M Yoshida, T Yamashita. Jpn. J. Appl. Phys. 17:585, 1978.
 13.    RW Birkmire, BE McCandless, WN Shafarman, RD Varrin Jr. In: 9th ECPV Solar
        Energy Conf., Freiberg Germany, 1989, p 134.
 14.    RH Mauch, M Ruckh, J Hedström, D Lincot, J Kessler, R Klinger, L Stolt, J Vedel,
        H-W Schock. In: 10th ECPV Solar Energy Conf. Lisbon Portugal, 1991, p 1415.
 15.    TL Chu, SS Chu, C Ferekides, CQ Wu, J Britt, C Wang J. Appl. Phys. 70:7608,
        1991.
 16.    TL Chu, SS Chu, N Schultz, C Wang, CQ Wu. J. Electrochem. Soc. 139:2443, 1992.
 17.    G Hodes, A Albu-Yaron, F Decker, P Motisuke. Phys. Rev. B 36:4215, 1987.
 18.    G Hodes, A Albu-Yaron. Proc. Electrochem. Soc., 88–14: 298, 1988.
 19.    S Gorer, A Albu-Yaron, G Hodes. J. Phys. Chem. 99:16442, 1995.
 20.    S Gorer, G Hodes. J. Phys. Chem. 98:5338, 1994.
 21.    PK Nair, MTS Nair. J. Phys. D: Appl. Phys. 23:150, 1990.
 22.    PK Nair, M Ocampo, A Fernandez, MTS Nair. Sol. Energy Mater. 20:235, 1990.
 23.    PK Nair, MTS Nair, A Fernandez, M Ocampo. J. Phys. D: Appl. Phys. 22:829, 1989.
 24.    PK Nair, MTS Nair, VM Garcia, OL Arenas, Y Pena, A Castillo, IT Ayala, O



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      Gomezdaza, A Sanchez, J Campos, H Hu, R Suarez, ME Rincon. Sol. Energy Mater.
      Sol. Cells 52:313, 1998.
25.   HL Smith. J. Sci. Instrum. 4:115, 1927.
26.   AB Lundin, GA Kitaev. Inorg. Mater. 1:1900, 1965.
27.   P Pramanik, S Bhattacharya. J. Mater. Sci. Lett. 6:1105, 1987.
28.   FC Meldrum, J Flath, W Knoll. J. Mater. Chem. 9:711, 1999.
29.   PC Rieke, BD Marsh, LL Wood, BJ Tarasevich, J Liu, L Song, GE Fryxell. Lang-
      muir 11:318, 1995.
30.   JH Fendler. Isr. J. Chem. 33:41, 1993.
31.   ML Breen, JT Woodward, DK Schwartz, AW Apblett. Chem. Mat. 10:710, 1998.
32.   RJ Collins, H Shin, MR DeGuire, AH Heuer, CN Sukenik. Appl. Phys. Lett. 69:860,
      1996.
33.   XK Zhao, J Yang, LD Mccormick, JH Fendler. J. Phys. Chem. 96:9933, 1992.
34.   I Pintilie, E Pentia, L Pintilie, D Petre, C Constantin, T Botila. J. Appl. Phys.
      78:1713, 1995.
35.   H Meherzi-Maghraoui, M Dachraoui, S Belgacem, KD Buhre, R Kunst, P Cowache,
      D Lincot. Thin Solid Films 288:217, 1996.
36.   M Ristova, P Toshev. Thin Solid Films 216:274, 1992.
37.   SS Dhumure, CD Lokhande. Thin Solid Films 240:1, 1994.
38.   AA Velykanov, EK Ostrovskaya, NP Garina, VA Turacova, AA Tchurkan. Ukr.
      Chim. Zh. 49:764, 1983.
39.   B Pejova, M Najdoski, I Grozdanov, SK Dey. Mater. Lett. 43:269, 2000.
40.   AJ Varkey, AF Fort. Sol. Energy Mater. Sol. Cell 29:253, 1993.
41.   T Kocareva, I Grozdanov, B Pejova. Mater. Lett. 47:319, 2001.
42.   CD Lokhande. Mater. Chem. Phys. 28:145, 1991.
43.   JD Desai, CD Lokhande. Thin Solid Films 249:135, 1994.
44.   S Biswas, A Mondal, D Mukherjee, P Pramanik. J. Electrochem. Soc. 133:48, 1986.
45.   MTS Nair, PK Nair. Semicond. Sci. Tech. 5:1225, 1990.
46.   CD Lokhande, AU Ubale, PS Patil. Thin Solid Films 302:1, 1997.
47.   P Pramanik, RN Bhattacharya, A Mondal. J. Electrochem. Soc. 127:1857, 1980.
48.   RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:332, 1982.
49.   VM García, MTS Nair, PK Nair, RA Zingaro. Semicond. Sci. Technol. 12:645,
      1997.
50.   PC Rieke, SB Bentjen. Chem. Mat. 5:43, 1993.
51.   R Ortega-Borges, D Lincot. J. Electrochem. Soc. 140:3464, 1993.
52.   JM Doña, J Herrero. J. Electrochem. Soc. 144:4081, 1997.
53.   RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:277, 1980.
54.   H Cachet, H Essaaidi, M Froment, G Maurin. J. Electroanal. Chem. 396:175, 1995.
55.   RL Call, NK Jaber, K Seshan, JR Whyte Jr. Sol. Energy Mater. 2:373, 1980.
56.   M Ocampo, AM Fernandez, PJ Sebastian. Semicond. Sci. Technol. 8:750, 1993.
57.   M Najdoski, I Grozdanov, B Minceva. J. Mater. Chem. 6:761, 1996.
58.   PK Basu, P Pramanik. J. Mater. Sci. Lett. 5:1216, 1986.
59.   P Pramanik, S Bhattacharya, PK Basu. Thin Solid Films 149:L81, 1987.
60.   AJ Varkey, AF Fort. Sol. Energy Mater. Sol. Cells 31:277, 1993.
61.   B Pejova, A Isahi, M Najdoski, I Grozdanov. Mater. Res. Bull. 36:161, 2001.
62.   I Grozdanov, M Najdoski. J. Solid State Chem. 114:469, 1995.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
63.   MTS Nair, L Guerrero, PK Nair. Semicond. Sci. Technol. 13:1164, 1998.
64.   P Pramanik, MA Akhter, PK Basu. J. Mater. Sci. Lett. 6:1277, 1987.
65.   A Mondal, P Pramanik. J. Solid State Chem. 47:81, 1983.
66.   I Grozdanov. Synth. Metals 63:213, 1994.
67.   C Lévy-Clément, M Neumann-Spallart, SK Haram, KSV Santhanam. Thin Solid
      Films 302:12, 1997.
68.   I Grozdanov. Mater. Lett. 19:281, 1994.
69.   QW Chen, YT Qian, H Qian, ZY Chen, WB Wu, YH Zhang. Mater. Res. Bull.
      30:443, 1995.
70.   QW Chen, YT Qian, ZY Chen, Y Xie, GE Zhou, YH Zhang. Mater. Lett. 24:85,
      1995.
71.   KM Gadave, PP Kumbhar, CD Lokhande. Indian J. Pure Appl. Phys. 30:299, 1992.
72.   M Najdoski, I Grozdanov, SK Dey, B Siracevska. J. Mater. Chem. 8:2213, 1998.
73.   SS Kale, CD Lokhande. Mater. Chem. Phys. 59:242, 1999.
74.   P Pramanik, S Bhattacharya. Mater. Res. Bull. 24:945, 1989.
75.   B Pejova, M Najdoski, I Grozdanov, SK Dey. J. Mater. Chem. 11:2889, 1999.
76.   GA Kitaev, VI Dvoinin, AV Ust’yantseva, MN Belyaeva, LG Skornyakov. Inorg.
      Mater. 12:1448, 1976.
77.   T Yoshida, K Yamaguchi, H Toyoda, K Akao, T Sugiura, H Minoura. Proc. Elec-
      trochem. Soc. 97–20:37, 1997.
78.   R Bayón, C Guillén, MA Martínez, MT Gutiérrez, J Herrero. J. Electrochem. Soc.
      145:2775, 1998.
79.   RP Goyal, D Raviendra, BRK Gupta. Phys. Status Solidi (a) 87:79, 1985.
80.   M Izaki. Electrochem. Solid State Lett. 1:215, 1998.
81.   P Pramanik, MA Akhter, PK Basu. Thin Solid Films 158:271, 1988.
82.   CD Lokhande, A Ennaoui, PS Patil, M Giersig, M Muller, K Diesner, H Tributsch.
      Thin Solid Films 330:70, 1998.
83.   CD Lokhande, KM Gadave. Turkish J. Phys. 18:83, 1994.
84.   W Mindt. J. Electrochem. Soc. 118:93, 1971.
85.   KL Chopra, RC Kainthla, DK Pandya, AP Thakoor. In: G Hoss, MH Francombe,
      JL Vossen, eds. Physics of Thin Films, Vol. 12. Academic Press, New York and
      London 1982, p. 167.
86.   P Pramanik, RN Bhattacharya. J. Mater. Sci. Lett. 8:781, 1989.
87.   P Pramanik, S Bhattacharya. Mat. Res. Bull. 25:15, 1990.
88.   KC Mandal, O Savadogo. J. Mater. Chem. 1:301, 1991.
89.   P Pramanik, S Biswas. J. Solid State Chem. 65:145, 1986.
90.   P Pramanik, S Bhattacharya. J. Electrochem. Soc. 137:3869, 1990.
91.   AJ Varkey, AF Fort. Thin Solid Films 235:47, 1993.
92.   F Kicinski. Chem. Ind. 54, 1948.
93.   M Isshiki, T Endo, K Masumoto, Y Usui. J. Electrochem. Soc. 137:2697, 1990.
94.   KM Gadave, SA Jodgudri, CD Lokhande. Thin Solid Films 245:7, 1994.
95.   S Gorer, A Albu-Yaron, G Hodes. Chem. Mater. 7:1243, 1995.
96.   I Grozdanov, M Najdoski, SK Dey. Mater. Lett. 38:28, 1999.
97.   W Mindt. J. Electrochem. Soc. 117:615, 1970.
98.   RN Bhattacharya, P Pramanik. Bull. Mater. Sci. 2:287, 1980.
99.   KC Mandal, A Mondal. J. Phys. Chem. Solids 51:1339, 1990.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
100.   MTS Nair, Y Pena, J Campos, VM García, PK Nair. J. Electrochem. Soc. 145:2113,
       1998.
101.   P Pramanik, RN Bhattacharya. J. Solid State Chem. 44:425, 1982.
102.   RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:1642, 1982.
103.   RN Bhattacharya, P Pramanik. Sol. Energy Mater. 6:317, 1982.
104.   P Pramanik, PK Basu, S Biswas. Thin Solid Films 150:269, 1987.
105.   MTS Nair, PK Nair. Semicond. Sci. Tech. 6:132, 1991.
106.   RD Engelken, HE McCloud, C Lee, M Slayton, H Ghoreishi. J. Electrochem. Soc.
       134:2696, 1987.
107.   P Pramanik, RN Bhattacharya. J. Mater. Sci. Lett. 7:1305, 1988.
108.   CD Lokhande. J. Phys. D: Appl. Phys. 23:1703, 1990.
109.   D Raviendra, JK Sharma. J. Phys. Chem. Solids 46:945, 1985.
110.   K Tsukuma, T Akiyama, H Imai. J. Non-Cryst. Sol. 210:48, 1997.
111.   QW Chen, YT Qian, ZY Chen, GE Zhou, YH Zhang. Thin Solid Films 264:25,
       1995.
112.   QW Chen, YT Qian, ZY Chen, WB Wu, ZW Chen, GE Zhou, YH Zhang. Appl.
       Phys. Lett. 66:1608, 1995.
113.   H Shin, RJ Collins, MR De Guire, AH Heuer, CN Sukenik. J. Mater. Res. 10:692,
       1995.
114.   A Mondal, P Pramanik. Thin Solid Films 110:65, 1983.
115.   MJ Mangalam, KN Rao, N Rangarajan, CV Suryanarayana. Jpn. J. Appl. Phys.
       8:1258, 1969.
116.   RN Bhattacharya, P Pramanik. Bull. Mater. Sci. 3:403, 1981.
117.   R Ortega Borges, D Lincot, J Vedel. In: 11th ECPV Solar Energy Conf. Montreux,
       Switzerland, 1992, p 862.
118.   JM Doña, J Herrero. J. Electrochem. Soc. 141:205, 1994.
119.   P O’Brien, J McAleese. J. Mater. Chem. 8:2309, 1998.
120.   CD Lokhande, PS Patil, H Tributsch, A Ennaoui. Sol. Energy Mater. Sol. Cells
       55:379, 1998.
121.   JM Doña, J Herrero. J. Electrochem. Soc. 142:764, 1995.
122.   P Pramanik, S Biswas. J. Electrochem. Soc. 133:350, 1986.
123.   P O’Brien, T Saeed, J Knowles. J. Mater. Chem. 6:1135, 1996.
124.   M Izaki, J Katayama. J. Electrochem. Soc. 147:210, 2000.
125.   A Ennaoui, M Weber, R Scheer, HJ Lewerenz. Sol. Energy Mater. Sol. Cells
       54:277, 1998.
126.   GK Padam, SK Gupta. Appl. Phys. Lett. 53:865, 1988.
127.   RW Buckley. In: 11th ECPV Solar Energy Conf. Montreux, Switzerland, 1992, p
       962.
128.   VB Patil, PD More, DS Sutrave, GS Shahane, RN Mulik, LP Deshmukh. Mater.
       Chem. Phys. 65:282, 2000.
129.   M Izaki, O Shinoura. Adv. Mater. 13:142, 2001.
130.   B Pejova, M Najdoski, I Grozdanov, A Isahi. J. Mater. Sci.-Mater. Electron. 11:405,
       2000.
131.   B Pejova, T Kocareva, M Najdoski, I Grozdanov. Appl. Surf. Sci. 165:271, 2000.
132.   M Izaki, T Omi. J. Electrochem. Soc. 144:L3, 1997.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
133. D Raviendra, JK Sharma. J. Appl. Phys. 58:838, 1985.
134. S Misra, HC Padhi. J. Appl. Phys. 75:4576, 1994.
135. D Braunger, D Hariskos, T Walter, HW Schock. Sol. Energy Mater. Sol. Cells
     40:97, 1996.
136. D Hariskos, M Ruckh, U Ruhle, T Walter, HW Schock, J Hedstrom, L Stolt. Sol.
     Energy Mater. Sol. Cells 41–2:345, 1996.
137. R Bayón, C Mafftiotte, J Herrero. Thin Solid Films 353:100, 1999.
138. J Herrero, MT Gutierrez, C Guillen, JM Dona, MA Martinez, AM Chaparro, R
     Bayon. Thin Solid Films 361:28, 2000.
139. LP Deshmukh, KM Garadkar, DS Sutrave. Mater. Chem. Phys. 55:30, 1998.
140. M Skyllas-Kazacos, JF McCann, R Arruzza. Appl. Surf. Sci. 22/23:1091, 1985.
141. NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 42:383, 1977.
142. NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Mater. Res. Bull. 11:1109, 1976.
143. NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 62:97, 1979.
144. NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 157, 1979.
145. GB Reddy, DK Pandya, KL Chopra. Sol. Energy Mater. 15:383, 1987.
146. GA Kitaev, VF Markov, LN Maskaeva, LE Vasyunina, IV Shilova. Inorg. Mater.
     26:202, 1990.
147. GA Kitaev, LN Maskaeva, VF Markov, AY Kurkin, LE Vasyunina. Inorg. Mater.
     25:1065, 1989.
148. LN Maskaeva, VF Markov, GA Kitaev. Russ. J. Appl. Chem. 73:751, 2000.
149. BB Nayak, HN Acharya, GB Mitra. Bull. Mater. Sci. 3:317, 1981.
150. BB Nayak, HN Acharya. J. Mater. Sci. Lett. 4:651, 1985.
151. YS Sarma, HN Acharya, NK Misra. Thin Solid Films 90:L43, 1982.
152. RS Mane, CD Lokhande. Thin Solid Films 304:56, 1997.
153. RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 129:99, 1982.
154. GS Shahane, BM More, CB Rotti, LP Deshmukh. Mater. Chem. Phys. 47:263, 1997.
155. GS Shahane, KM Garadkar, LP Deshmukh. Mater. Chem. Phys. 51:246, 1997.
156. GS Shahane, LP Deshmukh. Mater. Chem. Phys. 70:112, 2001.
157. I Grozdanov. Semicond. Sci. Tech. 9:1234, 1994.
158. JM Doña, J Herrero. Thin Solid Films 268:5, 1995.
159. SA Al Kuhaimi, Z Tulbah. J. Electrochem. Soc. 147:214, 2000.
160. NC Sharma, RC Kainthla, DK Pandya, KL Chopra. Thin Solid Films 60:55, 1979.
161. T Nakazawa, S Kuranouchi, A Ashida, N Yamamoto. In: 12th ECPV Solar Energy
     Conf., Amsterdam the Netherlands, 1994, p 601.
162. GK Padam, GL Malhotra, SUM Rao. J. Appl. Phys. 63:770, 1988.
163. T Yamaguchi, Y Yamamoto, T Tanaka, Y Demizu, A Yoshida. Thin Solid Films
     281–282:375, 1996.
164. T Yamaguchi, Y Yamamoto, T Tanaka, A Yoshida. Thin Solid Films 344:516,
     1999.
165. D Hariskos, R Heberholz, M Ruckh, U Ruhle, R Schäffler, HW Schock. In: 13th
     ECPV Solar Energy Conf., Nice, France, 1995, p 1995.
166. KC Sharma, JC Garg. J. Phys. D: Appl. Phys. 23:1411, 1990.
167. DS Sutrave, GS Shahane, VB Patil, LP Deshmukh. Mater. Chem. Phys. 65:298,
     2000.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
168.   G Contreras-Puente, O Vigil, M Ortega-Lopez, A Morales-Acevedo, J Vidal, ML
       Albor-Aguilera. Thin Solid Films 361:378, 2000.
169.   LP Deshmukh, DS Sutrave, BM More, CB Rotti, KM Garadkar. Semicond. Devices
       421, 1996.
170.   K Kushiya, T Nii, I Sugiyama, Y Sato, Y Inamori, H Takeshita. Jpn. J. Appl. Phys.
       35:4383, 1996.
171.   PK Nair, L Huang, MTS Nair, H Hu, EA Meyers, RA Zingaro. J. Mater. Res.
       12:651, 1997.
172.   B Mokili, Y Charreire, R Cortes, D Lincot. Thin Solid Films 288:21, 1996.
173.   GK Padam, SUM Rao. Sol. Energy Mater. 13:297, 1986.
174.   GK Padam, GL Malhotra, SUM Rao. Phys. Status Solidi (a) 109:K45, 1988.
175.   A Ennaoui, M Weber, M Saad, W Harneit, MC Lux-Steiner, F Karg. Thin Solid
       Films 361:450, 2000.
176.   RN Bhattacharya. J. Electrochem. Soc. 130:2040, 1983.
177.   GK Padam. Mater. Res. Bull. 22:789, 1987.
178.   KR Murali. Thin Solid Films 167:L19, 1988.
179.   JC Garg, RP Sharma, KC Sharma. Thin Solid Films 164:269, 1988.
180.   PK Vidyadharan Pillai, KP Vijayakumar, PS Mukherjee. J. Mater. Sci. Lett.
       13:1725, 1994.
181.   GN Chaudhari, S Manorama, VJ Rao. J. Phys. D: Appl. Phys. 25:862, 1992.
182.   GN Chaudhari, S Manorama, VJ Rao. Thin Solid Films 208:243, 1992.
183.   AR Patil, VN Patil, PN Bhosale, LP Deshmukh. Mater. Chem. Phys. 65:266, 2000.
184.   R Suarez, PK Nair. J. Solid State Chem. 123:296, 1996.
185.   VM Markov, LN Maskaeva, LD Loshkareva, SN Uimin, GA Kitaev. Inorg. Mater.
       33:555, 1997.
186.   N Parhi, BB Nayak, BS Acharya. Thin Solid Films 254:47, 1995.
187.   M Ristov, G Sinadinovski, I Grozdanov, M Mitreski. Thin Solid Films 173:53,
       1989.
188.   YF Nicolau. Appl. Surf. Sci. 22/23:1061, 1985.
189.   YF Nicolau, JC Menard. J. Cryst. Growth 92:128, 1988.
190.   YF Nicolau, M Dupuy, M Brunel. J. Electrochem. Soc. 137:2915, 1990.
191.   YF Nicolau, S Davied, F Genoud, M Nechtschein, JP Travers. Synth. Met. 42:1491,
       1991.
192.   VP Tolstoy. Thin Solid Films 307:10, 1997.
193.   M Sasagawa, Y Nosaka. Phys. Chem. Chem. Phys. 3:3371, 2001.
194.   CD Lokhande, BR Sankapal, HM Pathan, M Muller, M Giersig, H Tributsch. Appl.
       Surf. Sci. 181:277, 2001.
195.   S Lindroos, A Arnold, M Leskela. Appl. Surf. Sci. 158:75, 2000.
196.   MK Karanjai, D Dasgupta. Mater. Lett. 4:368, 1986.
197.   CD Lokhande, KM Gadave. Mater. Chem. Phys. 36:119, 1993.
198.   CD Lokhande, VV Bhad, SS Dhumure. J. Phys. D: Appl. Phys. 25:315, 1992.
199.   L Huang, PK Nair, MTS Nair, RA Zingaro, EA Meyers. J. Electrochem. Soc.
       141:2536, 1994.
200.   Y Rodriguez-Lazcano, MTS Nair, PK Nair. J. Cryst. Growth 223:399, 2001.
201.   H Wilman. Proc. Phys. Soc. 60:117, 1948.
202.   SH Pawar, PN Bhosale. Mater. Chem. Phys. 11:461, 1984.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
203.   K Vaccaro, HM Dauplaise, A Davis, SM Spaziani, JP Lorenzo. Appl. Phys. Lett.
       67:527, 1995.
204.   HM Dauplaise, K Vaccaro, A Davis, GO Ramseyer, JP Lorenzo. J. Appl. Phys.
       80:2873, 1996.
205.   K Vaccaro, A Davis, HM Dauplaise, SM Spaziani, EA Martin, JP Lorenzo. J. Elec-
       tron. Mater. 25:603, 1996.
206.   JH Schön, O Schenker, B Batlogg. Thin Solid Films 385:271, 2001.
207.   M Martin. In: J Cook ed. Passive Cooling. Cambridge, MA: MIT Press, 1989.
208.   Y Mastai, D Katzen, KD Dobson, G Hodes. Unpublished.
209.   GB Reddy, V Dutta, DK Pandya, KL Chopra. Sol. Energy Mater. 5:187, 1981.
210.   I Grozdanov, CK Barlingay, SK Dey. Mater. Lett. 23:181, 1995.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
3
Mechanisms of Chemical
Deposition




In this chapter, we critically discuss the various mechanisms for CD. Since the ki-
netics of deposition are intimately related to the mechanism, this aspect is also
treated here. This chapter is divided into several sections dealing with different as-
pects of the mechanisms. However, there is a large degree of overlap of the mate-
rial in these different sections; indeed, it was often not obvious in which section a
specific topic should be placed. While the main discussion of any specific point
will be limited to the most relevant section, the same point will often be treated,
more briefly, in other sections. However, for those readers who are looking for a
specific topic in this chapter, it would be prudent not to be confined to only what
appears to be the relevant section. For example, elements of the mechanism, prob-
ably with different emphases, are dealt with also in the sections on nucleation and
on kinetics, while many issues of nucleation are considered in the mechanistic
studies.
       There is one example of a CD process (for deposition of tin sulphides) in
which elemental sulphur dissolved in a nonaqueous solvent is used as a source for
S. Since this appears to be the only example in the literature for this type of film
deposition, it will be discussed in Chapter 6 together with the relevant study on tin
sulphides. However, there is no reason to believe that this process may not be ap-
plicable to other materials. Metal sulphides (and selenides) are known to form, as
precipitates, by reacting certain metal salts with dissolved elemental chalcogen, al-
though visible film formation seems to be limited, up to now, to this one example.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
3.1 THE FOUR BASIC MECHANISMS
In spite of the fact that CD has been in use for a long time and that the reactions
involved appear to be quite straightforward, the mechanism of the CD process is
often unclear. There is a good reason for this: There are several different mecha-
nisms of CD. These can be divided into four fundamentally different types. Using
the common thiourea deposition of CdS as an example, these are:
       The simple ion-by-ion mechanism
       Cd(NH3) 2 D Cd2
               4                     4NH3
                                                                           (3.1)
                              (dissociation of complex to free Cd2 ions)
       (NH2)2CS        2OH → S2     CN2H2 2H2O
                                                                           (3.2)
                                               (formation of sulphide ion)
       Cd2        S2   → CdS   (CdS formation by ionic reaction)           (3.3)
       The simple cluster (hydroxide) mechanism
       nCd2    2nOH D [Cd(OH)2]n
                                                                    (3.4)
                             (formation of a solid Cd(OH)2 cluster)
       [Cd(OH)2]n nS2 [from Reaction (3.2)] →
                                                                    (3.5)
                               nCdS 2nOH (exchange reaction)
       The complex decomposition ion-by-ion mechanism:
       (NH2)2CS Cd2 D [(NH2)2CSMCd]2                                                          (3.6)
       [(NH2)2CSMCd]2  2OH → CdS CN2H2                                 2H2O                  (3.7)*
       The complex decomposition cluster mechanism:
       [Cd(OH)2]n [from Reaction (3.4)] (NH2)2CS D
                                                                (3.8)
                                  [Cd(OH)2]n 1(OH)2CdMSMC(NH2)2
       [Cd(OH)2]n 1(OH)2CdMSMC(NH2)2 →
                                                                (3.9)
                                   [Cd(OH)2]n 1CdS CN2H2 2H2O
       which continues until conversion of all the Cd(OH)2 to CdS.
       The first two mechanisms involve free sulphide ions (or other anions), while
the last two are based on breaking of a carbon–chalcogen bond and do not involve
formation of free chalcogenide. Most mechanistic studies have assumed the for-


* The use of the simple thiourea–Cd ligand in Eq. (3.7) is for simplicity. There are a number of dif-
ferent complexes involving Cd and thiourea, in particular, some containing hydroxy groups. As long
a a solid phase of Cd(OH)2 is not present, then such hydroxy–thiourea Cd complexes involve an ion-
by-ion-type of mechanism, as exemplified in Eq. (3.7).




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mation of free anions (not always justifiably). It is also possible that more than one
mechanism occurs in parallel or that conditions change during deposition so that
the deposition mechanism also changes.
       Since most mechanistic studies were carried out for CdS deposition from an
ammoniacal solution and using thiourea as a sulphide source, and a smaller num-
ber for CdSe deposition using selenosulphate as selenide source, this chapter will
mostly use these materials as examples. However, it should be stressed that the
concepts discussed in this chapter are, for the most part, valid for the deposition
of all semiconductors.


3.2 SLOW FORMATION OF THE
    CHALCOGENIDE IONS
The rate-limiting step in CD for the first two mechanisms is almost always for-
mation of the chalcogenide ion. This reaction should be slow; otherwise fast, ho-
mogeneous precipitation of the metal chalcogenide will occur with little film for-
mation. (Even rapid precipitation can lead to a film; however, this film will be
extremely thin and in most cases not visible.) Almost all the literature on CD is
limited to sulfides (mostly), selenides, and oxides (including hydrated oxides and
hydroxides). Anion-forming reactions are described in this section.

3.2.1 Sulphides
3.2.1.1      Thiourea
Thiourea (the sulphur analogue of urea) is the most commonly used sulphur pre-
cursor. In alkaline solution (in which depositions involving thiourea are carried
out), the first step of the hydrolysis gives sulphide ions and cyanamide:
      SC(NH2)2      OH D HS           CN2H2      H2O                           (3.10)
The cyanamide can hydrolyze further, with the overall reaction, if carried to com-
pletion, going via urea to ammonium carbonate:
      CN2H2  CO(NH2)2 2O (NH4)2CO3
             H2O         2H
              →           →                                                  (3.11)
Cyanamide can also react with ammonia to give guanidine:
             NH
              
      CN2H2 3→ (NH2)2CBNH                                                    (3.12)
All of these compounds can be (and have been) found as impurities in CD films.
However, the important step is the formation of sulphide ion.
      In neutral and acidic solutions, thiourea can be decomposed to thiocyanate
ion [1], which can be useful if the intention is to deposit thiocyanates:
      SC(NH2)2 → NH 4         CNS                                              (3.13)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       It should be kept in mind that many of these decomposition reactions are
equilibria. The decomposition of thiourea in the absence of a metal ion will nor-
mally be much slower than in the presence of such an ion. The metal ion removes
sulphide as metal sulphide—the less soluble the sulphide, the more effective the
removal at very low sulphide concentrations. This continuous removal of sulphide
shifts the equilibrium to the direction of more sulphide production. The same prin-
ciple holds for many other anion precursors.

3.2.1.2      Dimethylthiourea
Dimethylthiourea is much less commonly used than the more available thiourea:
     (CH3)NHC(S)NH(CH3)           H2O → (CH3)NHC(O)NH(CH3)            H2S   (3.14)

3.2.1.3      Thioacetamide
Thioacetamide has been used for a long time as an analytical reagent to precipi-
tate metal sulphides (see Ref. 2 for many relevant references). Thioacetamide can
be hydrolyzed over a wide range of pH and is often used for CD in acidic baths.
In a strongly acidic bath (pH ca. 2), H2S is formed:
      H3C.C(S)NH2        2H2O → CH3COOH          H2S     NH3                (3.15)
It has been shown [3] that this reaction can proceed by two pathways, one in which
the carbon–sulphur bond is broken first:
      H3C.C(S)NH2        H2O → H3C.C(O)NH2         H2S                      (3.16)
forming acetamide as an intermediate, or a pathway in which the carbon–nitrogen
bond is first broken to give thioacetic acid:
      H3C.C(S)NH2        H2O → H3C.C(S)OH        NH3                        (3.17)
which then is hydrolyzed to H2S and acetic acid. The H2S dissolves in water as
hydrosulphide ion:
      H2S     H2O D HS         H3O                                          (3.18)
In an alkaline bath, the overall reaction is:
      H3C.C(S)NH2        2OH → CH3COO            NH3     HS                 (3.19)
Hydrolysis in alkaline solution is considerably faster than in acid solutions.
        At intermediate pH values, particularly in weakly acidic solutions (pH 2),
metal sulphide formation using thioacetamide may proceed through decomposi-
tion of a metal ion (or solid phase)–thioacetamide complex rather than through in-
termediate formation of sulphide [4] (see Sec. 3.3.3). Thioacetamide in pure wa-
ter is fairly stable and does not readily hydrolyze at room temperature.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
3.2.1.4       Thiosulphate
The earliest CD processes were carried out using thiosulphate. Although thiourea
(and to some extent thioacetamide) are now more commonly used to deposit sul-
phides, thiosulphate is still sometimes used. While the reaction pathways listed
below are intended to suggest possibilities for reactions involving thiosulphate, it
must be noted that the mechanism(s) for these reactions is (are) still not clear.
Mechanisms have been proposed in the CD literature, but no convincing proof for
any particular one has been forwarded.
      Thiosulphate depositions are most often carried out in a weakly acidic bath
(pH 3). Several reactions are possible in such solutions:
      S2O 2
          3     H2O → H2S       SO 2
                                   4                                         (3.20)
with the H2S dissolving to give HS , as in Reaction (3.18):
      S2O 2
          3     H    →S      HSO3                                            (3.21)
      S2O 2
          3     2H → S       SO2     H2O                                     (3.22)
and in alkaline solution:
      S2O 2
          3     OH → HS            SO 2
                                      4                                      (3.23)
It has been suggested that the thiosulphate, a reducing agent, may act as an elec-
tron donor and reduce the elemental sulphur formed in Reactions (3.21) and
(3.22), forming sulphide ions:
      S    2e → S2                                                           (3.24)
        Because of the strong complexes thiosulphate forms with some metal ions,
it is very possible that these metal–thiosulphate complexes undergo a complex-de-
composition mechanism (Section 3.3.3). However, one early study on the forma-
tion of PbS on boiling Pb2 and thiosulphate in water found that PbS formed more
readily when excess thiosulphate was present [5], which suggests that decompo-
sition of thiosulphate to sulphide might be the dominant pathway under the con-
ditions of that study.

3.2.2 Selenides
3.2.2.1       Selenourea
Selenourea (SeC(NH2)2) was apparently first used for CD by Kutzscher’s group
in Germany during World War II. It appears that this work was not published; ref-
erences to it come through other sources [6]. The first published use of selenourea
for CD appears to be in 1949 by Milner and Watts [7] to deposit PbSe for photo-
conductive cells. It was the main reactant used to form selenide films (mainly
PbSe) through the 1960s, after which selenosulphate (see later) became the dom-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
inant selenide precursor. Other examples of its use are for ZnSe [8] and HgSe [9].
Selenourea is an unstable compound that requires the presence of a reducing
agent—usually Na2SO3—to minimize oxidation to elemental Se. The selenide
formation presumably parallels thiourea hydrolysis in Eq. (3.10):
      SeC(NH2)2      OH → HSe           CN2H2       H2O                        (3.25)

3.2.2.2      Dimethylselenourea
Dimethylselenourea (CH3)2NC(NH2)Se (for preparation see Ref. 10) was used to
deposit films of PbSe [11]. It is more stable than selenourea, although still not very
stable in solution and, like selenourea, was used together with Na2SO3 to prevent
oxidation. The optimum pH for this deposition was 9.8; bulk precipitation oc-
curred at a pH of 10.1, while the deposition slowed greatly as the pH decreased
even a few tenths of a pH unit. The activity of this solution, as measured by the
rate of deposition, was reported to increase with time (hours). This was explained,
in view of the measured reduction in the rate of reaction by sulphite (which was
assumed to complex with the dimethylselenourea), by reduction of the concentra-
tion of sulphite with time (by oxidation to sulphate). This means that, for repro-
ducible results, the age of the dimethylselenourea/sulphite solution should be
taken into account or an aged solution (at least 10 hr old) should be used (the
change in activity of the solution slows down to a large degree after this time).
3.2.2.3      Selenosemicarbazide
Selenosemicarbazide (H2NN(H)C(Se)NH2) was used by Velykanov et al. to make
Cd, Zn, Ag, and Hg selenide films and precipitates from aqueous alkaline solu-
tions [12]. As with the other Se precursors, Na2SO3 was used to stabilize the se-
lenosemicarbazide against rapid decomposition. This reagent was reputed to be
more stable than the previous two.
3.2.2.4      Selenosulphate
Selenosulphate (Na2SeSO3), which is the analogue of thiosulphate with the active
S atom substituted by Se, was used by Kitaev for deposition of PbSe [13] and
CdSe films [14]. Since it is more stable, simpler to prepare, and cheaper than se-
lenourea, it simplified the deposition of selenides and has for a long time, with
only a few exceptions, been the precursor invariably used to deposit selenides. It
can be prepared by dissolving elemental Se in an aqueous sulphite solution:
      Se    SO 2 D SeSO 2
               3        3                                                      (3.26)
The Se dissolves slowly (typically an hour or two at 60–70°C). Although more
stable than selenourea, it does slowly hydrolyze and becomes less active with
age—fairly rapidly for the first few days after preparation and then slowly; it is
still usable for weeks after preparation without any special storage demands, al-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
though at increasingly reduced activity. Storage under cold, oxygen-free condi-
tions will slow the aging further. This gradual change in activity should be taken
into account in depositions using this material, since it is not always practical to
make it fresh for each deposition.
      The mechanism for hydrolysis of selenosulphate is often given as
      SeSO 2
           3       OH → HSe           SO 2
                                         4                                       (3.27)
However, although    SO 2
                        4  is likely to be a final product, the reaction is probably
not as simple as this. A suggested first step for the hydrolysis is [15]
      2SeSO 2
            3       H2O → HSe          SeS2O 2
                                             6        OH                         (3.28)
      Selenosulphate is unstable in acidic solutions. If the pH of a fairly concen-
trated selenosulphate solution is reduced below ca. 7, red Se will precipitate out.
This property has been used to prepare films of Se by slightly acidifying dilute so-
lutions of selenosulphate [16].

3.2.3 Tellurides
Very few examples of telluride deposition have been reported. There are a number
of reasons why tellurides are more difficult to deposit by CD than sulphides or se-
lenides. One is the very negative redox potential of the telluride/tellurium couple
(E0      1.14 V). This means that a strong reducing solution is necessary. Another,
directly related reason is the instability of telluride ion (and possible tellurium pre-
cursors). Even dissolved oxygen will rapidly oxidize telluride. The strong reduc-
ing solution will need to reduce dissolved oxygen, preventing this reaction. Thus a
combination of a suitable Te source and a strong reducing agent with exclusion (or
reduction) of oxygen can succeed in forming tellurides. In fact, such conditions
have also been found necessary to form films of ZnSe using selenosulfate and are
preferable for deposition of ZnS films (see Sec. 3.5 and Chap. 4).
       CdTe has been deposited by hydrazine reduction of TeO2 [17,18]. The po-
tential of hydrazine oxidation is sufficiently negative to allow formation of tel-
luride:
      N2H4(aq)    4OH → N2            4H2O       4e        E0    1.16 V          (3.29)
      Te    2e D Te2                                       E0    1.14 V          (3.30)
especially because only very small concentrations of telluride need to exist at any
time, causing a positive shift of the Te reduction through the Nernst equation
(Eq. 1.32).
      Te was considered to be essentially insoluble in Na2SO3 under normal con-
ditions, although preparation of tellurosulphate, the analogue of selenosulphate, un-
der hydrothermal conditions has been reported [19]. A recent study has described
CdTe deposition using this reagent [20]. Apparently the solubility, while low, is suf-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ficient to be useful. At this stage, it is assumed that the tellurosulphate reacts anal-
ogously to selenosulphate (Eqs. 3.26–3.28). As with the TeO2 reduction just de-
scribed, because of the great ease of oxidation of telluride to elemental Te, the films
invariably contain elemental Te. (See Sec. 4.3 for details of this deposition.)
      While not true CD, a novel telluriding agent, Te dissolved in an alkaline so-
lution of hydroxymethanesulphinic acid, has been used to convert Cd(OH)2 films
to CdTe [21]. While there is some doubt as to the nature of the active telluriding
agent, from the description of the preparation process of this reagent—a change in
color from deep purple (characteristic of polytelluride ions, Te 2 ) to a faint pink
                                                                     x
(pure Te2 is colorless but will be colored this way if traces of Te 2 are present)
                                                                        x
as the preparation proceeds—it does appear to contain free telluride ion. It should
be noted that elemental Te can slowly dissolve in concentrated air-free alkaline
solutions, with the formation of polytelluride and its characteristic purple color.

3.2.4 Oxides
Oxides have been commonly deposited by CD. In many cases, the deposit is a hy-
droxide or hydrated oxide formed by reaction of the metal ions with slowly gener-
ated hydroxide. A variety of precursors has been used to generate the hydroxide.

3.2.4.1      Urea
Urea is slowly hydrolyzed in water to form ammonium carbonate:
      (NH2)2CBO         2H2O → (NH4)2CO3                                         (3.31)
Carbonates are alkaline since they dissociate to some extent to form hydroxide
ions:
      CO 2
         3      H2O D OH          HCO 3        (K    1.8    10 4)                (3.32)
The hydroxide so formed will react with the metal ion to form the metal hydrox-
ide, hydrated oxide, or oxide, depending on the relative stability of the various ox-
ides and hydroxides. (The resulting hydroxides or oxy hydroxides can be heated
in air or oxygen to form the oxides.) In addition, insoluble carbonates may also
form. The competition between hydroxide and carbonate will depend on their sol-
ubility products. Carbonates tend to be more soluble than hydroxides of the same
metal ion. On the other hand, the value of K for equilibrium (3.32) (1.8 10 4)
means that the concentration of hydroxide will be ca. four orders of magnitude less
than that of carbonate (assuming no other pH-determining species is present).
       While the urea method has been commonly used in the past to form bulk
precipitates of basic salts, including oxides, for analytical purposes, it was noted
that transparent, adherent films were typically formed on the walls of the beaker
in which the precipitation was carried out [22]. Notably “basic stannic sulfate,”
which was probably mainly SnO2 since very little sulphate was found in the ac-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
companying precipitate, was found to adhere so tightly to the walls of the beaker
that it was difficult to remove [23].
3.2.4.2 Dimethylamineborane (DMAB) and
        Trimethylamineborane
Dimethylamineborane (DMAB) and trimethylamineborane have been used to de-
posit Zn [24], In [25] and Fe [25a] oxides using the metal nitrates. The nitrate an-
ion is believed to be reduced by the DMAB to nitrite and hydroxide:
      (CH3)2NHBH3 2H2O → BO 2 (CH3)2NH                     7H      6e          (3.33)
          NO 3 H2O 2e → NO 2 2OH                                               (3.34)
              Mx  xOH → M(OH)x                                                 (3.35)
In the case of Zn the oxide forms spontaneously, while for In the hydroxide film
is heated to form the oxide. This mechanism is actually a mixed electrochemical
process involving partial anodic and cathodic reactions (i.e., charge exchange oc-
curs at an interface) rather than a pure chemical deposition process.
3.2.4.3       Persulphate
                                                    2
Persulphate (also called peroxydisulphate) (S2O 8 ) is a very strong oxidizing
agent and has been used to deposit oxides, often in a higher oxidation state that that
of the original metal ion. Films grown using this reagent include -PbO2, NiO,
MnO2, and Tl2O3. The two latter required a small concentration of Ag ions in the
deposition solution as a catalyst. Ag is a known catalyst for oxidations using
S2O 2 (it is oxidized by persulphate to Ag(III), which is then the active oxidation
     8
agent). It is probable that Pb and Ni also act likewise, while Mn and Tl do not (or
much less so). A study of the use of persulphate for the deposition of PbO2 has pro-
vided strong evidence that the deposition actually is a mixed electrolytic process
(similar to that proposed for dimethylamineborane depositions) rather than a pure
chemical deposition [26]. The partial electrolytic reactions were given as
      Pb2      2H2O → PbO2        4H      2e                                  (3.36a)
      S2O 2
          8     2e → 2SO 2
                         4                                                    (3.36b)
Persulphates hydrolyze to form (finally) sulphate and hydrogen peroxide; the
probable overall reaction can be given by
      S2O 2
          8     2H2O → 2SO 2
                           4           H2O2    2H                              (3.37)
This reaction occurs via various radical species, and it is also possible that some
oxide depositions occur by these radicals (the formation of which is probably ini-
tiated by fission of the S2O 2 ion to two sulphate radicals) or directly by the hy-
                             8
drogen peroxide, which itself usually involves a free-radical reaction. Hydrogen
peroxide itself has been used directly to form oxide films in a few cases.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
3.2.4.4       Other oxide-forming reactions
Aqueous solutions of acidic metal salts are usually inherently unstable and hy-
drolyze readily to oxides (the hydroxides of these metals tend to be relatively un-
stable), in some cases forming films. Such hydrolysis can be more readily con-
trolled by adding boric acid to fluoro-complexes of the metal, e.g.:
      H2TiF6     2H2O D TiO2       6HF                                     (3.38a)
followed by removal of the HF by the boric acid:
      H3BO3      4HF D BF4       H3O      2H2O                             (3.38b)
Ref. 27, which describes such reactions as liquid phase deposition, gives more de-
tails on this method.
       Finally, many ammoniacal or amine solutions of metal salts will form films
of hydroxides or oxides on heating or even long standing at room temperature.
Ammonia is very volatile and will gradually be lost (in an open system), resulting
in reduced complexation of the metal ions. The pH will also drop, but it is likely
that the increase in concentration of free metal ions due to loss of ammonia dom-
inates. Chapter 7 treats the deposition of oxides in detail.

3.2.5 Halides
Halide ions can be formed by hydrolysis of alkyl halides, as shown for AgCl pre-
cipitation [28]:
      RMCl      OH → Cl          ROH                                        (3.39)
Haloalcohols (halohydrins) which are more water soluble than alkyl halides, have
been used to generate chloride ions [29], and are therefore more suitable for aque-
ous CD. The hydrolysis of haloalcohols has been used to deposit films of AgBr
and AgI, e.g.,
      ICH2CH2OH        H2O → I       HOCH2CH2OH          H                  (3.40)

3.2.6 Other Anions
Although they do not appear to have been used (at least deliberately) to form
films, there are other slow anion-generating reactions. Although most of the com-
pounds formed in these reactions are not considered semiconductors in the nor-
mally accepted sense, they merit at least a mention here.
3.2.6.1       Sulphate
Sulphate is formed by reaction between persulphate and thiosulphate:
      S2O 2
          8     2S2O 2 → 2SO 2
                     3       4         S4O 2
                                           6                                (3.41)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
This would be limited to cations that form insoluble sulphates but soluble persul-
phates and thiosulphates (Ba and Sr were demonstrated by Lamer and Dinegar
[30]). Thiosulphate, in particular, forms soluble complexes with many cations and
therefore should (often) not present a problem in this respect, as long as the metal
sulphide is not formed under the conditions of the deposition. In addition, solvents
other than water can be used in principle, and therefore it might be possible to de-
posit sulphates that are soluble in water but insoluble in another solvent.
3.2.6.2      Phosphates
Phosphates have been formed by slow hydrolysis of trialkyl phosphates, hydrogen
phosphate ions, or metaphosphoric acid, which liberate phosphate ions (see p. 46
in Ref. 31). This may be of interest as a precursor for the preparation of phosphide
semiconductors.
3.2.6.3      Arsenates
Arsenates have been described in one case exploiting the fact that the zirconyl
cation forms a water-soluble arsenite but insoluble arsenate. By adding nitric acid
to a solution of zirconyl chloride and sodium arsenite, the arsenite was oxidized to
arsenate by the nitric acid, precipitating the insoluble zirconyl arsenate [32]. As for
phosphates (and probably more readily), arsenates might be reduced to arsenides.
       It should be stressed that these reactions were used to form precipitates and
not films. There is no guarantee that films can be formed using these reactions.
However, it is reasonable to expect that, under the right conditions, it may be pos-
sible to produce films of these compounds. It is left as an exercise for the curious
reader to find these “right” conditions.
       Details of these and other slow precipitations are given in Ref. 31.
       There are a number of examples of homogeneous precipitation of hydrox-
ides based on slow cation release, such as destruction of the Fe–EDTA complex
with H2O2 (see Ref. 33). In CD, the only well-defined example of this is heating
an ammonia complex (e.g. of Cd2 ). The loss of ammonia by volatilization will
gradually increase the concentration of free Cd2 ions.

3.3 MECHANISMS OF CHEMICAL DEPOSITION
Having dealt with slow formation of the reacting anion, we now turn to the various
mechanisms by which the CD compounds are formed. For the most part, the de-
tails of nucleation and film growth are left to the following section. Here we con-
centrate on the reactions taking place that form the semiconductor material. There
are four main mechanisms for the compound formation, as outlined in Sec. 3.1;
which one is operative depends on the specific process and reaction parameters.
      1.   Simple ion-by-ion mechanism
      2.   Simple cluster (hydroxide) mechanism



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      3.    Complex-decomposition ion-by-ion mechanism
      4.    Complex-decomposition cluster mechanism.

3.3.1 (Simple) Ion-by-Ion Mechanism
The conceptually simplest mechanism, often assumed to be the operative one in
general, is commonly called the ion-by-ion mechanism, since it occurs by se-
quential ionic reactions. The basis of this mechanism, illustrated for CdS, is given
by
      Cd2      S2 D CdS       (adsorbed on the substrate)                     (3.42)
If the ion product, [Cd2 ][S2 ], exceeds the solubility product, Ksp, of CdS (ca.
10 28; Table 1.1), then CdS can form as a solid phase, although a larger ionic
product may be required if supersaturation occurs. If the ion product does not ex-
ceed Ksp, no solid phase will form, except possibly transiently due to local fluctu-
ations in the solution, and the small solid nucleii will redissolve before growing to
a stable size. For that reason, the precipitation process is shown as an equilibrium
rather than as a one-way reaction.
       This reasoning, that no solid phase phase will form if the ion product does
not exceed Ksp, may not be true under certain circumstances. CdS formation might
occur at the substrate surface under conditions where none can be formed homo-
geneously. This is an important point; however, since it will be treated in more de-
tail in the following section, on hydroxide formation, it therefore will not be dis-
cussed further in this section.
       Equation (3.42) gives the overall reaction for formation of CdS. However,
the process is more complicated than this and comprises a number of reactions and
equilibria. The mechanism involves the formation of S2 ions and control of Cd2
concentration.
       The S2 can be formed by a number of methods, as already described. Here
we consider the most common one, the decomposition of thiourea by aqueous al-
kaline solution:
      (NH2)2CS      2OH → S2          CN2H2      2H2O                         (3.43)
Since the S2 concentration can be made as low as desired simply by controlling
the rate of Reaction (3.43) (e.g., using low temperatures and/or relatively low pH),
in principle, even at relatively high free-Cd2 concentrations, the deposition rate
should be easily controlled. Since an alkaline pH is required to decompose the
thiourea to sulphide, a complexing agent is needed to keep Cd2 in solution and
to prevent Cd(OH)2 from precipitating. As explained in Chapter 1 [Eqs. (1.26) and
(1.27) and calculation following these equations], using ammonia as a complex-
ant for Cd (0.1 M total Cd concentration) and at a pH of 11, a concentration of am-
monia 1.19 M will be needed to prevent formation of Cd(OH)2 at room tem-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
perature—more at the higher temperatures more commonly encountered in CdS
deposition. (We shall see, in the following section, that although this statement is
perfectly accurate, it does not mean that CdS deposition will not occur at lower
concentrations of ammonia at the same pH.)
       When reading the literature, in many (probably most) cases it is not clear
whether the deposition proceeds by an ion-by-ion process. The reason is that, un-
less another mechanism is specifically discussed, it is often assumed that the de-
position proceeds via the ion-by-ion mechanism. If the exact deposition parame-
ters are known, which mechanism is operative can, in most cases, be calculated.
Two criteria have often been cited in the literature as proof of deposition via the
ion-by-ion mechanism. One is epitaxial deposition of the CD film. (Epitaxy refers
to growth of one material on another in such a way as to result in coherence be-
tween the lattice of the substrate and the deposit. Often—although not necessar-
ily—the lattice of the deposit is aligned in the same direction as that of the sub-
strate.) This is based on the expectation that a cluster mechanism will not result in
an epitaxial film; for this to occur, clusters of maybe thousands of atoms would
need to be able to rearrange themselves on the substrate. Some examples of epi-
taxial growth are given in Sections 3.4.2 and 4.1.5.2.
       The second criterion, sometimes seen in the older literature, is that films
formed via the cluster mechanism should be poorly adherent and optically scat-
tering, while those formed via the ion-by-ion mechanism will be adherent and
transparent. Unlike the epitaxy, this criterion is faulty; films formed by the clus-
ter mechanism can be highly transparent and strongly adherent, while there are
examples of films that proceed via the ion-by-ion mechanism that are not well ad-
herent and are optically scattering. The degree of adhesion of the film and its
transparency say little, if anything, about the mechanism of the deposition. Thus,
following this reasoning, it was claimed that ion-by-ion deposition occurred un-
der conditions where a visible Cd(OH)2 suspension was present [34]. While the
presence of Cd(OH)2 does not exclude the possibility that ion-by-ion deposition
occurs, it is unlikely that ion-by-ion deposition would be dominant when a high
concentration of Cd(OH)2 is present.
       More recently, there have been a small number of studies that provide strong
evidence for the ion-by-ion mechanism. It must be pointed out, however, that
while it is not very difficult to distinguish between an ion-by-ion and cluster
mechanism in most cases, it is much more difficult to distinguish between a sim-
ple ion-by-ion and a complex-decomposition ion-by-ion mechanism. Therefore
most investigations that conclude an ion-by-ion mechanism is operative, while
usually assuming the simple ion-by-ion process, do not distinguish between the
simple and complex-decomposition pathways.
       One investigation has shown a clear-cut boundary in the crystal size of films
(CdSe, CdS, and, to a lesser extent, PbSe), depending on whether the deposition
occurred via an ion-by-ion or a cluster mechanism [15]. The solution conditions



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
were constant except for a change in the ratio between complexant and Cd. A
small crystal size was obtained under conditions where Cd(OH)2 was proven to
exist (as a colloid, not visible to the eye) in the deposition solution, while a larger
crystal size was obtained when no Cd(OH)2 was present as a solid phase. Thus the
separation between the conditions where the ion-by-ion and cluster mechanisms
occurred was clearly shown. This study is dealt with in more detail in the follow-
ing section.
       A recent investigation of CdS deposited from a thiourea bath found that the
initial deposit (on mica) formed as islands ca. 0.5 nm high and 10–40 nm across;
further growth led to an increase in height without changing the lateral dimensions
[35]. Such a growth mode supports an ion-by-ion mechanism, since a cluster
mechanism, whereby (presumably fairly symmetric) hydroxide clusters adsorb on
a substrate, is expected to lead to an equally symmetric initial growth mode. The
island size in this study was measured by atomic force microscopy, where the tip
size can determine the measured lateral dimensions of particles smaller than the
tip. For this reason, scanning tunneling microscopy (or electron microscopy for
lateral dimensions) would be more reliable in such measurements. This investiga-
tion was carried out using triethanolamine as complexant and under conditions
where deposition was slower than usual. As always the case, the results and inter-
pretation cannot be extrapolated to other bath conditions.

3.3.2 Hydroxide (Cluster) Mechanism
We noted in the earlier example that Cd(OH)2 precipitation should be avoided and
calculated the concentration of ammonia required to prevent this precipitation. In
reality, CDs are quite often carried out under conditions where a metal hydroxide
(or hydrated oxide) is formed. This might seem to imply that a precipitate of, e.g.,
Cd(OH)2 is formed at the start of such depositions. In fact, this is (usually) not the
case; the Cd(OH)2 is formed, but either as a colloid rather than a precipitate, or as
an adsorbed species on the substrate but not in the bulk of the solution. Since
Cd(OH)2 is colorless and colloids usually do not scatter light (otherwise it is
termed a suspension), this means that the Cd(OH)2 colloid is not apparent to the
eye. The CdS is then formed by reaction of S2 ion with the Cd(OH)2:
      Cd2      2OH → Cd(OH)2                                                    (3.44)
followed by
      Cd(OH)2      S2 → CdS         2OH                                         (3.45)
The driving force for Eq. (3.45) is the much lower value of Ksp for CdS (ca. 10 28;
Table 1.1) than for Cd(OH)2 (2 10 14), which reflects the more negative free
energy of formation of the former. This means that sulphide will readily substitute
for hydroxide in the case of Cd. An idea of the amount of sulphide needed to con-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
vert Cd(OH)2 into CdS can be calculated from the ratio of the two solubility prod-
ucts:
        Ksp (CdS)           [Cd2 ][S2 ]
      Ksp(Cd(OH)2)         [Cd2 ][OH ]2
                            [S2 ]                                           (3.46)
                                      10 28                   15
                                                    5    10
                           [OH ]2    2 10 14
At a pH of 11 ([OH ] 10 3 M at room temperature), this gives a value for [S2 ]
of 5 10 21 M. Thus it requires a very low concentration of S2 indeed to begin
to convert Cd(OH)2 into CdS. Even taking into account that most of the sulphide
species will be HS rather than S2 , at pH 11 the [S2 ]:[HS ] ratio is 10 6.3 (or
5 10 7) [Eq. (1.17)], and the required HS concentration will be
                 21
      5     10        10   14
                                M                                          (3.46a)
                 7
      5     10
still a very low concentration. At the higher temperatures usually used for CdS de-
position, the OH concentration will be about an order of magnitude higher (be-
cause of the strong temperature dependence of the ionic product of water), and the
required S2 concentration will be about two orders of magnitude higher than re-
quired at room temperature.
        The participation of Cd(OH)2 in the deposition of CdS (and other metal
chalcogenides) has been demonstrated or suggested on many occasions. Kitaev et
al. presented a theoretical thermodynamic treatment of the Cd2 /ammonia/
thiourea system to show when Cd(OH)2 should be present as a solid phase in the
deposition solution [36]. A graphical representation of this analysis is shown in
Figure 3.1. This graph is based on two equilibria: the solubility product of
Cd(OH)2 and the stability constant of the ammonia (ammine) complex of Cd.
Consider first the former:
      Cd2        2OH D Cd(OH)2         Ksp    2    10   14
                                                                            (3.47)
                                               2
Based on this equilibrium, we can express [Cd ] in terms of [OH ] for the case
where Cd(OH)2 will just precipitate; at higher pH it will precipitate and at lower
pH it will not. Since [OH ] can be converted into [H ] (and therefore pH) through
the ion product of water (see Chap. 1), a graph can be made of pH vs. [Cd2 ] (or
p[Cd2 ], which, analogously to pH, is equal to minus the logarithm of [Cd2 ]).
This is the hydroxide line in Figure 3.1. Its physical meaning is that above this
line, Cd(OH)2 will be present in the solution, while below it there will be no
Cd(OH)2.
       A similar (somewhat more complicated) calculation can be made based on
the stability constant of the Cd–tetrammine complex [see Eq. (1.27)] and using the
hydrolysis of ammonia (see Refs. 34 and 36 for details of the calculation), which



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 3.1 Regions of stability for the Cd–ammonia system for 0.1 M total Cd concentra-
tion and at room temperature. The hydroxide line separates conditions where Cd(OH)2 will
(above the line) or will not (below the line) thermodynamically exist. The complex line
gives the concentration of free Cd2 at any pH value (where the pH is determined only by
the ammonia concentration and not, e.g., by added alkali). (Adapted from Ref. 34.)



gives the free-Cd2 concentration as a function of pH. This is shown as the com-
plex line in Figure 3.1. (The calculation is made for a total-Cd2 concentration of
0.1 M, but it is only slightly dependent on this concentration; e.g., for a Cd2
concentration of 1 mM, the line will shift ca. 0.2 pH units to lower pH values.
Also, the presence of other ammine and hydroxy complexes should strictly speak-
ing be taken into account; however, these considerations will not result in major
changes to the overall picture.) Since the free-Cd2 concentration is a function
of ammonia concentration, the p[Cd2 ] can be also identified with a p[NH3]
(top axis).
       Considering both the hydroxide and complex lines together, for the left side
of the figure, where the complex line is above the hydroxide line, the concentra-
tion of free Cd2 will always be high enough to form Cd(OH)2. Where the com-
plex line is below the hydroxide line, however (the right side of the figure),
Cd(OH)2 will not form at pH values above the complex line but below the hy-
droxide line, since the Cd2 concentration will not be high enough to form



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Cd(OH)2. Extra OH ions (e.g., added KOH) are required to increase the pH and
form Cd(OH)2 in this region.
       This type of analysis (which can be extended to other metal ions and other
complexants) provides a basis for choosing conditions for depositions involving
hydroxide clusters. An elegant example of this was given by Kitaev et al [36] for
CdS deposition in the absence of ammonia (or any other complex, except for
thiourea itself, which is a weak complexant for Cd). Extrapolation of the hydrox-
ide line to p[Cd2 ] 1 (0.1 M free Cd2 ) gives a pH of 7.7, or, alternatively, a
pOH of 6.3; above this value of pOH, no Cd(OH)2 is formed. However, at higher
temperature, the ion product of water increases considerably (Table 1.2). Thus,
while pOH ( pH for pure water) at room temperature is 7, at 80°C it becomes
6.3. Thus, while no Cd(OH)2 is formed in a 0.1 M solution of a Cd salt (ignoring
effects of the Cd salt and thiourea on the solution pH), at 80°C it can thermody-
namically form. Kitaev et al. demonstrated experimentally that a solution of 0.1
M cadmium acetate and thiourea produced a film of CdS only at 90°C and above.
This analysis, of course, is based on the hydroxide cluster model. Deposition by
another mechanism could conceivably occur at lower temperatures given enough
time.
       Rieke and Bentjen made a detailed study of the role of Cd(OH)2 in the de-
position of CdS films from the ammoniacal thiourea bath [37]. They first studied
Cd(OH)2 formation in the absence of thiourea. Cd(OH)2 formation, as measured
visually by laser scattering (which shows up small amounts of turbidity in the so-
lution), began at a pH of ca. 10.4 (the solid phase, once formed, could exist for
quite long periods of time down to a pH as low as 10). Surface analysis (XPS)
showed that Cd(OH)2 formed on silicon substrates at a pH value as low as 9, even
though no Cd(OH)2 formed in the solution under those conditions. The Cd(OH)2,
about 1.5 nm average thickness, was apparently stabilized against dissolution by
the substrate. In the presence of thiourea, a SEM study of the early stages of de-
position showed that large, platelike crystals of CdS formed at a pH where
Cd(OH)2 was present in the solution but that this deposit was very nonadherent
and could easily be wiped off. However, at lower values of pH, where Cd(OH)2
had previously been shown to occur only on the substrate, strongly adherent CdS
spherules were observed. The density (in terms of surface coverage) of this de-
posit decreased with decreasing pH, and the coverage became low at low values
of pH, where surface-adsorbed Cd(OH)2 did not occur. Therefore, only films de-
posited under conditions where appreciable amounts of surface-adsorbed, but not
bulk, Cd(OH)2 occurred were of good quality (adherent and specularly reflecting).
       In connection with this study, in particular the suggestion that the Cd(OH)2
was stabilized by the substrate against dissolution, it has been shown that
Co(OH)2 can form at a solid (SiO2) surface at a pH lower than that necessary to
cause bulk precipitation of Co(OH)2 [38]. This was explained by the effect of the
electric field at the solid/liquid interface on the dielectric constant of the interface



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
region and, through this, on the free energy (and therefore the solubility product)
of the relevant hydroxide dissolution. It was noted that this field effect should in-
crease with an increase in the charge of the cation. In terms of CD, this would
mean that the difference in the ease of deposition of the hydroxide at a solid sur-
face and homogeneously in solution will increase with the charge on the cation.
Furthermore, this argument is not limited to hydroxides but should be valid for all
depositions, including chalcogenides (and therefore also ion-by-ion deposition),
as was briefly mentioned in the previous section.
       Betenekov et al. [39] used an isotopic tracer technique to show that, for their
range of solution compositions, the initial deposition involved adsorption of
Cd(OH)2 on the glass substrate. At the beginning of the reaction, only Cd was ob-
served to form on the substrate and this was interpreted to be due to Cd(OH)2,
since any other insoluble Cd compounds that might be formed from the deposition
solution (containing CdCl2, NaOH, NH4OH, and thiourea dissolved in water)
were expected to contain either S or C.* However, they concluded that the depo-
sition proceeded, not by reaction between Cd(OH)2 and sulphide formed by de-
composition of thiourea, but rather by decomposition of a Cd(OH)2–thiourea
complex (see Sec. 3.3.3.1).
       O’Brien and Saeed studied the CdS deposition, only using ethylenediamine
instead of ammonia as complexant because of the better-defined coordination
properties of the former [40]. In common with the former studies, they also found
the presence of Cd(OH)2 necessary for the deposition of good-quality CdS films
by comparing the conditions needed to obtain such films with those calculated us-
ing the relevant thermodynamic parameters.
       A short digression at this point is required to define the term good-quality
film. The use of the adjective good-quality depends very much on what is required
from the film. In the context of CD and thin-film preparation in general, good
quality almost always refers to two parameters: good adhesion and specular re-
flection (smooth). Of course, requirements can be envisaged where these proper-
ties, in particular the latter, are not desired, such as if a rough “scattering” film is
required. However, keeping this caveat in mind, we will continue to use the term
good-quality as it is almost always used in the CD literature.
       While most mechanistic investigations have been carried out on CdS, other
semiconductors, in particular CdSe, have also been studied with regard to the de-
position mechanism. Kainthla et al., in their study of the formation of CdSe films
from ammoniacal solutions containing sodium selenosulfate, noted that when a
visible precipitate of Cd(OH)2 was present in their solutions (obtained by adding


* It needs to be repeatedly emphasized that the mechanism deduced in any one investigation is not nec-
essarily valid for different experimental conditions. At the same time, it does appear likely that for this
very well-studied system of CdS deposition from ammonia/thiourea baths, Cd(OH)2 is, in many cases,
involved in the deposition.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
KOH), a thin terminal thickness (ca. 80 nm) of CdSe was obtained, compared with
several hundred nanometers obtained under optimal conditions (no precipitate)
[41]. If the pH was too high and Cd(OH)2 precipitation was extensive, no film was
observed to form. In fact, an analysis of their results suggests that films are ob-
tained only if Cd(OH)2 is present (not necessarily as a visible precipitate). While
they do not specifically consider Cd(OH)2 as a chemical intermediate in their re-
action, they do conclude that Cd(OH)2—both adsorbed on the substrate and in the
solution—acts as a nucleation center for CdSe formation.
       Gorer and Hodes carried out a study of CdSe deposition from selenosul-
phate solutions of Cd complexed with nitrilotriacetate—NTA [15], which is a
stronger complex for Cd than is ammonia and with which is easy to obtain condi-
tions where no Cd(OH)2 is present, even at relatively high values of pH. A change
in reaction mechanism, from hydroxide cluster to ion-by-ion, was observed by
monitoring the optical transmission spectra of the films deposited on glass. The
basis for this investigation is the change in semiconductor bandgap, and thus in its
spectrum, when the crystal size becomes very small—the quantum size effect
(discussed in detail in Chap. 10). The CdSe crystal size (and therefore the
bandgap) is not strongly dependent on deposition parameters within a fairly wide
range of parameters. However, it was observed that if the ratio between the NTA
and Cd concentrations was increased above a certain value—called the critical
complex ratio—the crystal size suddenly increased (for fixed values of tempera-
ture and pH). This was seen first by a red shift in the optical spectrum (see Chap.
10 for examples of these changes) and was subsequently verified by direct mea-
surements of the CdSe crystal sizes in the films using TEM and XRD.
       A variety of analytical techniques were then used to verify that Cd(OH)2
was present in the solution when the complex:Cd ratio was below the critical value
(Rc) but absent above it. Cd(OH)2 absorbs in the UV range of the spectrum, and
spectral monitoring of Se-free solutions showed that it was present only below Rc.
Light scattering by a blue laser also confirmed the presence of a heterogeneous
phase below Rc but not above it. Similar XPS analyses to those employed by Rieke
and Bentjen for CdS showed that Cd adsorbed on the glass substrate, immersed in
Se-free solutions, only below Rc. This is seen in Table 3.1: Appreciable amounts
of Cd (as Cd(OH)2) were seen only when the pH was sufficiently high and the
complex:Cd ratio relatively low.


TABLE 3.1 Cd:Si Ratio Measured by XPS

pH                9.0               10.0                10.5              11.0               11.0
Cd:Si             0.020              0.024               0.17              0.45               0.005

For glass substrates immersed in Cd2 /nitrilotriacetate solutions (no selenosulphate) at different pH
values for 5 min at 40 C. NTA:Cd ratio 1.63 except for the rightmost column, where it is 2.25.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Theoretical thermodynamic calculations of the conditions under which
Cd(OH)2 should form were also carried out, based on the solubility product of
Cd(OH)2, the stability constants of the Cd-NTA system, and the ion product of
water at different temperatures. The values of Rc computed from these calcula-
tions agreed with those measured experimentally.
       In agreement with Kainthla et al. and contrary to at least some conditions for
thiourea-based CdS deposition, Gorer and Hodes found that adsorption of a Cd
hydroxide species on the substrate occurred only under conditions where the solid
hydroxide formed also in the solution. This need not necessarily be interpreted as
contradictory; it may be due to the different conditions involved.
       An important factor in the reaction for CdSe deposition via the hydroxide
mechanism was an observed gradual increase in pH, ca. 0.8 pH units over the
course of the deposition. It is not clear what the cause of this increase is. However,
it means that while Cd(OH)2 may not be formed at the start of the deposition, it
may form during the deposition. This can then explain the induction time where
no apparent reaction takes place initially. For the preceding experiments, the stan-
dard conditions were to set the solution pH to 10. Coloration of the solution (in-
dicating formation of CdSe) occurred when the pH reached ca. 10.3 (depending,
of course, on the temperature and complex:Cd ratio). If the pH was adjusted to
10.3 at the beginning of the reaction, coloration began almost immediately rather
than after a more typical time of several minutes. This increase in pH did not oc-
cur for the ion-by-ion mechanism, probably because the excess NTA necessary in
this case also acts as a buffer. In any case, an increase in pH is not required for this
mechanism.
       These results for CdSe were extended to the deposition of PbSe (using cit-
rate as complex) and CdS (using NTA and thiourea) and found to be also applica-
ble for these cases [15]. For PbSe, the transition, while clearly evident, was not so
sharp, since citrate is a weak complex and a solid phase is always present, but to
greatly differing degrees depending on the conditions. The colloidal phase in this
case is a hydrated lead oxide, which, in a selenosulphate-free solution, adsorbs
strongly onto the substrate (see Sec. 3.4.3 and Fig. 3.5). This hydrated oxide is
rapidly selenized to PbSe, and the process involves breaking down of the rela-
tively large oxide crystals during reaction with the selenosulphate and recrystal-
lization of the PbSe product. This can be seen by following the reaction of pre-
cipitated hydrated lead oxide with selenosulphate by XRD (Fig. 3.2), where the
sharp peaks of the oxide transform into very broad peaks indicative of an almost
amorphous structure, which themselves vanish as sharp PbSe peaks appear.
       A quartz crystal microbalance study of the kinetics of CdSe deposition from
the preceding solution (nitrilotriacetate) showed an importance difference in the
mode of growth of the CdSe below and above Rc [42]. Below Rc (hydroxide mech-
anism), a periodicity in the deposition rate was observed, with a period corre-
sponding to a film thickness of ca. 5 nm—i.e., the approximate size of a single



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 3.2 XRD spectra showing the process of PbSe formation from the reaction of pre-
cipitated hydrated lead oxide with Na2SeSO3 solution. (a) Starting material; (b–e) after 1.5,
3, 4.5, and 6 mn reaction, respectively. (Adapted from Ref. 46.)



crystal. Above Rc, the period corresponded to 0.3 nm, a single CdSe monolayer,
as expected from an ion-by-ion growth. This interpretation is based on growth by
coverage of the surface by one layer—whether a layer of crystallites or a single
Cd-Se layer—at a time.
       The critical ratio concept already described was derived for a fixed concen-
tration of metal ions. As pointed out by O’Brien and McAleese [43], while useful
at high metal concentrations, it requires modification when low concentrations are
employed, due to the changes in stability of complexes when dilute. They devel-
oped a system whereby, on a plots of total metal concentration against total com-
plex concentration, a curve defining a constant value of free-metal concentration
was drawn. Figure 3.3 shows several such plots for the Cd–ethylenediamine sys-
tem studied by them. From these plots, the need for much higher complex:metal
ratios for dilute solutions is evident. The complex:metal ratio required for a con-
stant concentration of free metal, shown in the top curve, is reasonably constant
for higher concentrations but increases strongly at low concentrations.
       We postpone detailed explanation of what determines the different crystal
sizes in the two mechanisms to Section 3.4. At this point, it is enough to say that
the CdSe crystal size in the hydroxide mechanism will be determined mainly by
the size of the Cd(OH)2 particles in the solution and on the substrate, while that



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
formed in the ion-by-ion mechanism will depend on the heterogeneous nucleation
on the substrate.
       It should be repeated at this stage that this mechanism is dependent on a
large difference between the solubility products of the hydroxide and chalco-
genide of the required metal. The situation for ZnS, for example, is considerably
less favorable. The values of Ksp for the hydroxide and sulphide of Zn are 8
10 17 and 3 10 25, respectively. The same calculation for Zn as carried out ear-
lier for CdS shows that a S2 concentration of 4 10 15 M is required at pH
11 to convert the hydroxide into sulphide; this is a million times more than that re-
quired to form CdS. This does not tell us at this stage whether ZnS will form or
not, but only that it is less likely to than CdS. We will return to this problem when
we discuss the specific deposition of II–VI compounds in Chapter 4.
       It is quite possible that the mechanism will change in the course of the depo-
sition. As the metal is depleted from solution, the complex:metal ratio will increase




FIG. 3.3 Equivalent solution contour plots for solutions of Cd2 and ethylenediamine
[en] at 50°C. The curves represent conditions for a constant concentration of free Cd2
10 9M. Bottom curve (solid line): [total Cd] against [ethylenediamine]. Top curve (dotted
line): the ratio of [ethylenediamine] to [total Cd] against [ethylenediamine]. (Adapted from
Ref. 43.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
and may pass the point where no solid hydroxide phase is present in the solution.
In this case, the ion-by-ion process will occur if the conditions are suitable. This is
most likely if the initial conditions were close to the border between ion-by-ion and
cluster mechanisms. The opposite may also occur: initial deposition by an ion-by-
ion pathway followed by clusters, which either build up gradually in the solution
or change their aggregation properties, adhering to the film. This has been shown
to occur for CdS deposited from a thiourea bath using a combination of quartz crys-
tal microbalance (which measures the mass of the deposit) and electrochemical im-
pedence spectroscopy (which provides indirect structural information) [44].
Change in aggregation properties of the colloids present during the deposition was
suggested as the cause of the change in deposition. It is also possible that both
mechanisms occur in parallel. Thus, deposition might begin with nucleation of
Cd(OH)2, but growth might occur on this Cd(OH)2 via an ion-by-ion mechanism:
      Cd(OH)2      S2      Cd2 → Cd(OH)2 CdS                                    (3.48)
either by itself or together with conversion of the Cd(OH)2 to CdS, as in Eq.
(3.45). Alternatively, each mechanism may occur independent of each other. An
example of where this latter case appears to occur is in the deposition of PbSe
[45,46], where, under certain conditions, two spatially separated domains of small
crystals (cluster mechanism) and larger crystals (typical of the ion-by-ion mecha-
nism) are obtained. The relative concentrations of these two domains can, of
course, be varied by changing the solution composition, but the fact that they oc-
cur as separated domains on the substrate suggests that differences in the substrate
from one region to the other also plays a role. Another factor that could lead to a
change in the details of the mechanism with time is the buildup of reaction prod-
ucts in the solution. For example, the cyanamide formed in the decomposition of
thiourea [Eq. (3.10)] can react with Cd ions adsorbed on the surface of the CdS to
give the sparingly soluble CdCN2:
      (CdS)n Cd 2
                ads     CN2H2 D CdS CdCN2(ads)          2H                      (3.49)
The CdCN2 would then react with any sulphide present to form the more insolu-
ble CdS. This would be parallel to the more straightforward sulphidization of
(CdS)n Cd 2 to CdS.
            ads
       Before finishing this section, we present a few words concerning the effect
of deposition temperature on the terminal film thickness. Although exceptions
have been reported, it is a general observation that a lower deposition temperature
results in ultimately thicker films (of course, the deposition rate is slower). There
may be a number of reasons for this. One is that higher temperatures favor the hy-
droxide cluster mechanism due mainly to the higher hydroxide concentration (the
strongly temperature-dependent ion product of water) and also to the lower sta-
bility of complexes at high temperature (which may be offset by higher solubility
of the metal hydroxide). This effect has been shown for CdSe deposition, where a



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
higher complex:Cd ratio is needed to prevent Cd(OH)2 formation at higher tem-
peratures [15]. The ion-by-ion mechanism, since it often does not result in homo-
geneous precipitation, or at least less than in the cluster mechanism, tends to yield
films of larger terminal thickness. This will be valid even if the mechanism is a
mixture of the two mechanisms.
       Another effect of low temperature is (in the absence of stirring) reduced
mass transport. This can cause two opposing effects: reduction of reactants reach-
ing the substrate and reduction of aggregation (collisions between colloidal parti-
cles). The concentration of free chalcogenide ions will also increase essentially
exponentially with increasing temperature. However, it is difficult to predict the
effect this will have on terminal thickness; rates of homogeneous precipitation and
film formation will both increase. It is possible that thinner films will occur for the
ion-by-ion mechanism, since there is an increased probability of the occurrence of
homogeneous precipitation with increasing chalcogenide ion concentration (the
product of chalcogenide concentration and metal ion will increase—also because
the free-metal ion concentration will be somewhat higher at higher temperature
due to reduced stability of the metal complex).

3.3.3 Complex-Decomposition Mechanism
The complex-decomposition mechanism can, as with the free-anion mechanisms,
be divided into ion-by-ion and cluster pathways. However, since experimental
data relating directly to the complex decomposition mechanism is rather sparse, it
will be dealt with in one main section rather than two. The cluster pathway has
been more emphasized in these studies and will be dealt with first. It is very im-
portant to stress at this point that almost all experiment data described below could
be explained in terms of a simple (anion-mediated) mechanism. Equally valid,
most of the data described in the previous section could be explained by a com-
plex-decomposition mechanism.

3.3.3.1      Cluster Mechanism
The basis of this mechanism is that a solid phase is formed but, instead of react-
ing directly with a free anion, it forms an intermediate complex with the “anion-
forming” reagent. Continuing with CdS deposited from a thiourea bath as our ex-
ample, this would be given as
      MCd(OH)2        (NH2)2CS D Cd(OH)2 SC(NH2)2                               (3.50)
where MCd(OH)2 is one molecule in the solid-phase cluster. This complex, or a
similar one containing also ammine ligands, then decomposes to CdS:
      MCd(OH)2 SC(NH2)2 → MCdS               CN2H2       2H2O                   (3.51)
i.e., the SMC bond of the thiourea breaks, leaving the S bound to Cd.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Such a mechanism was suggested by Betenekov et al. based on their isotopic
tracer technique discussed in the previous section [39]. They suggested that
Cd(OH)2 forms initially on the substrate and catalyzes the thioureau decomposi-
tion. Of course the catalytic effect of the solid surface could be to decompose
thiourea to sulphide ion and not necessarily to catalyze the complex-decomposi-
tion mechanism.
       A similar catalytic effect of PbS on the decomposition of thiourea had been
suggested previously by Norr [47]. Kinetic measurements by Rieke and Bentjen
suggested that CdS likewise catalyzed thiourea decomposition [37]. Ortega-
Borges and Lincot also deduced such a mechanism based on kinetic measurements
of the CdS deposition using a quartz crystal microbalance [48]. In this case, the
measurements were found to fit best with a complex-decomposition model. Both
they and Rieke and Bentjen found optimum deposition to occur under conditions
where Cd(OH)2 was formed as a surface species on the substrate but not in the bulk
of the solution. Kinetic measurements also led Doña and Herrero to a similar con-
clusion of a complex-decomposition mechanism, but with the main difference that
the initial adsorbed species is not Cd(OH)2 itself but an ammine–hydroxide [49]:
      Cd(NH3) 2
              4      2OH D [Cd(OH)2(NH3)2]ads           2NH3                (3.52)
They based this modification on the known adsorbance of OH on glass and on
the common occurrence of transition metal mixed water–ammonia complexes
with coordination number of 4. Parallel structural studies of the deposited CdS
showed textured growth, supporting a mechanism whereby alternate Cd and S
species were involved, in an ion-by-ion process. Such a growth suggests adsorp-
tion of a molecular hydroxy-ammine species rather than a cluster. In fact, the
mechanism of Ortega-Borges and Lincot also does not differentiate between a hy-
droxide cluster and molecule.
       Unfortunately, it is nontrivial to distinguish reliably between the complex-
decomposition and sulphide-formation mechanisms. For example, in the study
of PbS (as a precipitate) formation from thiourea [47], the two main results used
to support complex decomposition were: (a) very little sulphide was formed in
alkaline solutions of thiourea and (b) addition of PbS powder catalyzed the re-
action, seen by the disappearance of the induction time for precipitation and
more rapid PbS formation when PbS was added at the start of the reaction. How-
ever, these results would also be obtained in a free-anion mechanism, for the fol-
lowing reasons:
      (a) Thiourea decomposition is an equilibrium reaction [see Eq. (3.10)].
          Formation of sulphide will shift the equilibrium back to the left. If a
          metal ion, which forms a sulphide with a low-solubility product, is pre-
          sent in the solution, however, it will remove even a very low concen-
          tration of sulphide continually as formed.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      (b) The presence of a large solid surface in the solution will reduce the in-
          duction time, even if the mechanism proceeds through free-sulphide
          formation, since the initial nucleation step will be facilitated.
These factors do not argue against the complex-decomposition mechanism, but
they should not be too readily interpreted, in the absence of other evidence, as ev-
idence against the sulphide mechanism. Granted, this is an old study, but it does
point up the difficulty in distinguishing between the two mechanisms. Kinetic
studies and subsequent fitting of the data from these studies to various models
[48,49] appear to be the best way of approaching this problem at present.
3.3.3.2      Ion-by-Ion Mechanism
Consider the complexation of free Cd2 by thiourea to give a Cd–thiourea com-
plex ion:
      Cd2      (NH2)2CS D [(NH2)2CSMCd]2                                      (3.53)
This ion could, in principle, hydrolyze by breaking the SMC bond to form CdS:
      [(NH2)2CSMCd]2          2OH → CdS          CN2H2      2H2O              (3.54)
This would lead to CdS formation in solution.
       If the Cd2 is adsorbed on the substrate (either directly or indirectly through
a hydroxide linkage) or on previously deposited CdS, then the same reaction
would occur. If the CdS so formed remained bound to the substrate (it is assumed
that CdS generated on previously deposited CdS would remain bound), the result
would be film growth by an ion-by-ion, complex-decomposition mechanism. As
with the cluster mechanism, it is difficult to distinguish experimentally between
the complex-decomposition mechanism and the free-anion pathway.
       Some studies have involved deposition (or precipitation) from acidic solu-
tions. It is reasonable to assume that no hydroxide is present under these condi-
tions for most metal ions commonly used in CD and that deposition occurs via an
ion-by-ion mechanism.
       Thioacetamide decomposition at intermediate pH values, particularly in
weakly acidic solutions (pH 2), has been suggested to occur through a thioac-
etamide complex rather than through intermediate formation of sulphide [4]. Of
course, this process may also occur in parallel with either acid or alkaline hydrol-
ysis of thioacetamide to (ultimately) sulphide at certain pH ranges. It is also pos-
sible that this complex-decomposition reaction occurs at both high and low pH
values in certain cases.
       The Cd–thiourea example that has been mainly used up to now in this sec-
tion is a very weak complex. However, there are examples where the chalcogenide
precursor is a strong complexant to the metal and may also be used as the com-
plexant. Depositions based on thiosulphate as a S source are good examples of



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
this. Also, thiosulphate depositions are most often carried out under at least some-
what acidic conditions, which means that they are more likely to involve an ion-
by-ion mechanism. Two old examples are the hydrolysis of silver thiosulphate to
give (bulk) silver sulphide [50]:
      Ag2S2O3      H2O → Ag2S         H2SO4                                      (3.55)
and the deposition of a number of metal sulphides, including PbS and CuxS by
heating thiosulphate solutions of the metal salts [51]. While less common nowa-
days as a S source than, e.g., thiourea, thiosulphate reactions have been used to de-
posit films of many different metal sulphides. The mechanism often suggested in
these studies is reduction of elemental S, formed by acidic decomposition of thio-
sulphate, by the thiosulphate itself, forming sulphide ions (see Sec. 3.2.1.4). How-
ever, no mechanistic studies of these reactions appear to have been undertaken.
While there is no convincing proof in the literature to distinguish which mecha-
nism is operative in these cases (sulphide-ion formation or complex decomposi-
tion), chemical intuition leads us to expect the latter where the metal–chalcogen
bond is strong (metal ions as Ag and Cu form very strong complexes through
the labile S atom of thiosulphate), and it seems reasonable to expect this
metal–sulphur bond to remain intact during the reaction rather than formation of
sulphide ion to occur. At the same time, those same metals from sulphides with
very-low-solubility products (understandably, since both the solubility product
and strength of complexation are related in the same way to the strength of the
metal–sulphur bond). Therefore, very low concentrations of free sulphide are
needed to form the metal sulphide.
       While thiosulphate is not very commonly used to deposit sulphides nowa-
days, its Se counterpart, selenosulphate, is the most common reagent for selenide
deposition. By analogy with thiosulphate, it might be argued that the mechanism
involves formation of selenide either through hydrolysis or through reduction of
Se, which forms even more readily in selenosulphate than does S in thiosulphate,
or by complexation with metal ion or metal hydroxide and breaking of the SeMS
bond (complex-decomposition mechanism). It can only be said that the mecha-
nism of selenide formation using selenosulphate has not been unambiguously de-
termined. An important difference between depositions using selenosulphate and
thiosulphate is that the former are carried out in alkaline solution, in contrast to the
(mainly) acidic conditions used for thiosulphate depositions. This means that both
ion-by-ion and cluster mechanisms can occur using this reagent, as has been
shown to be the case for CdSe, CdS, and PbSe [15,46]. Once again, it requires em-
phasizing that, although selenosulphate depositions are invariably assumed to oc-
cur via free selenide ions, this has not been proven, and a complex-decomposition
pathway cannot be excluded in selenosulphate depositions in general.
       Thermodynamic analyses of metal sulphide formation from thiosulphate
[52] and thiourea [53] and metal selenide formation from selenosulphate [13,54]



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
have been made. These analyses are based on the assumption that free chalco-
genide ions are formed.


3.4 NUCLEATION, ADHESION, AND
    FILM GROWTH
Probably the least-known aspect of the CD process is what determines the nucle-
ation on the substrate. Why do adherent films grow under some conditions and
poorly adherent films or even no film at all under others, even when slow precip-
itation occurs in solution? In considering this aspect, the two basic mechanisms—
hydroxide and ion-by-ion—may behave very differently, although there are also
features in common. When considering nucleation, the anion-mediated mecha-
nisms and the complex-decomposition mechanisms will behave similarly in most
cases. Some basic features of nucleation will first be considered, followed by is-
sues specific to each.
       Film growth can involve continuous nucleation, sticking of colloids from
solution and growth of individual crystals in the film. The last in particular is con-
sidered in Chapter 10, where it is an important factor in the context of quantum
size effects.

3.4.1 General Features of Nucleation and Adhesion
The basic science behind nucleation and forces between materials have been
treated in Chapter 1. For those interested in this section, it is assumed that this ba-
sic science is (more or less, at least) understood. However, the basics treated in
Chapter 1, while important to an understanding of film (as opposed to isolated crys-
tal) formation, are not enough by themselves to provide a phenomenological ex-
planation of film formation. We would ideally like to be able to predict in advance,
from fundamental principles, whether a particular bath formulation will result in
adherent films or not. We cannot! However, if we cannot reliably predict adhesion,
we can at least choose conditions so that the probability of adhesion is good.
       Considering first adsorption of metal ions or neutral species directly on the
substrate, there are a number of possible mechanisms for this process. Most sim-
ply, there will an equilibrium between metal species in solution and a solid sur-
face leading to dynamic adsorption of the metal. Adsorption of metal ions onto
solid surfaces has been extensively studied, to a large extent because of the use of
oxide surfaces to adsorb heavy metal ions and remove them from solution (see
Ref. 55 for an example and list of other references on this subject). This adsorp-
tion may go even farther with ion exchange between the solution metal ions and
ions in the substrate (again, glass is a good example of where this may occur).
       Coulombic attraction between charged species in solution and a surface may
play a part in initial adsorption on the surface. Under the high pH values more com-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mon in CD, most oxide surfaces (including glass) tend to be negatively charged be-
cause of the acid–base equilibrium of the oxide [see Eq. (2.16)]. Positive ions (e.g.,
Cd2 , Cd(OH) ) will be attracted to this surface by coulombic forces (in practice,
this attraction can be reduced by the solvation shell of the metal ion).
       Specific chemical interactions between primary particles or even reactants
and the substrate is another parameter that can aid in prediction of adhesion, at
least qualitatively. Chalcogenide ions will chemisorb to many metals, in some
cases forming a surface compound with the metal. Au, Ag, and Cu are the best ex-
amples. For Au in a solution of sulphide ions, the surface can be considered a gold
sulphide entity; for Cu, bulk sulphidization occurs, leading to eventual disintegra-
tion of the entire Cu to CuxS. However, since chalcogenide ion requires time to be
formed in a CD process, it is more likely that specific (or even nonspecific, as de-
scribed earlier) adsorption of the metal ions or species is dominant in the initial
adsorption process. A study of Cd2 adsorption on SiO2 from ethanol/cyclohex-
ane mixtures followed by CdS formation by reaction with H2S has concluded that
the CdS bonds to the silica via SiMOH linkages [56]. While not normally con-
sidered, it is possible that the chalcogen presursor is bonded, through the chalco-
gen atom, to the surface of some substrates, mainly metals.
       Considering now particle adsorption, the section on forces in Chapter 1 con-
cluded by stressing that it was normal for particles to stick together (the van der
Waals attractive forces eventually dominate in CD processes unless a protective
surface layer is present—and sometimes even then). Yet this property of “sticking
together” and also of “sticking to the substrate” clearly can vary greatly; sufficient
particle adhesion to cause aggregation may not be sufficient to form an adherent
film (adherent meaning that the particles stick both to the substrate and to each
other—both are necessary). However, since colloids do normally aggregate, the
main task in CD is to ensure that these aggregated particles (and therefore the pri-
mary ones attached to the substrate) do adhere well to the substrate.
       Why was the phrase and therefore the primary ones attached to the sub-
strate just used? Why not just aggregated particles? It is probably not a very in-
accurate generalization to state that adherent films of a reasonable thickness will,
in most cases, not form from a solution in which aggregation has already visibly
started and in which no new product is being formed before the substrate is im-
mersed. As a clear example of this, if, during CdSe deposition from a solution in
which the hydroxide mechanism is operative, the substrate is placed in the depo-
sition solution after precipitation starts, although some deposition will occur, it is
usually poor quality (poorly adherent and patchy). Clearly the initial adsorption
process is important, not just for the initial deposit but for the entire film. This will
not be a surprise to the many who deal with covering a surface—whether by elec-
trodeposition, vacuum coatings such as evaporation or sputtering, or, to take even
more common examples, painting or gluing, where the state of the initial surface
is very important. However, while a clean substrate is as important for CD as in
any coating process, there must be other factors involved.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       The fact that primary particles will stick where aggregated ones will not can
have (at least) two causes. One is that the primary particle may be different than
the final aggregate (e.g., in the hydroxide mechanism, the first stage is adsorption
of hydroxide particles). The other is explained by surface energy: A single parti-
cle has a larger external surface-to-volume ratio than an aggregate of the same par-
ticles, meaning a larger surface energy and therefore greater potential to stick to a
surface in which it comes in contact. While it should always be remembered that
adsorption from solution will certainly reduce this surface energy, perhaps drasti-
cally, the difference between single and aggregated particles remains valid.
       Finally, and this is likely to be important for nucleation in many cases, the
effect of the electric field (Helmholz layer) at the substrate/solution interface in
promoting formation of a deposit under conditions where none can form in the so-
lution (described in Sec. 3.3.2) should be considered. Whether or not this occurs,
and to what extent, can be experimentally measured.

3.4.2 Ion-by-Ion Mechanism
Most nucleation studies of CD have treated either the hydroxide or hydroxide-
complex mechanisms (see later) or have not clearly defined which mechanism
was, indeed, operative under the conditions of the experiments. Due to the paucity
of dependable experimental data, therefore, we consider nucleation and growth by
the ion-by-ion mechanism, to a large extent, from a theoretical viewpoint.
       Figure 2.4 showed a general-form curve of film formation as a function of
time. This form is valid in many cases regardless of the mechanism. For the ion-
by-ion mechanism, an induction period is generally necessary for sufficient
chalcogenide ion to build up and form a solid metal chalcogenide phase. It is prob-
able that some metal ions adsorb on the substrate, e.g., by an ion exchange, an
electrostatic mechanism, or simple equilibrium (see Sec. 3.4.1). However, while
this stage may be important for film initiation, it is not normally considered as a
growth stage in the usual sense of the word. Growth can be considered to begin
when stable clusters of the deposit begin to form on the substrate.
       Ideally, deposition occurs only on the substrate and not in solution. This is
possible due to the effect of a surface (even one in which no chemical interaction
occurs with any constituents of the deposition solution—and such interaction
probably does occur to a greater or lesser extent; water, for example, interacts with
many different surfaces). It is easier to nucleate on a surface than from a homoge-
neous solution. The possible effect of the electric field at the substrate/solution in-
terface on promoting nucleation has already been described. Additionally, some
surfaces are easier to nucleate on than others. This is the basis of the sensitization
of some surfaces, usually glass by a SnCl2 treatment. The SnCl2 hydrolyzes to
give nuclei of hydrated tin oxide on the glass surface, and these nuclei then form
nucleation centers for growth of the CD film. Such sensitization may reduce (even
to zero) the initial induction (nucleation) time of deposition. An example of this is


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the deposition of CdS on a SnO2/glass (conducting glass) surface that has already
been covered with a layer of CdS. Essentially no induction time was found for film
growth, compared with a few minutes induction time for deposition, from the
same solution and under otherwise identical conditions, on bare conducting glass
[57]. This was explained by catalytic decomposition of the thiourea by the prede-
posited CdS, and hence immediate film growth, compared to the slower decom-
position of thiourea in the absence of this catalytic surface.
       Once nucleation of clusters has begun, growth can occur, since in most
cases the depositing material will deposit on itself more readily than on a substrate
(since the reactants, almost by definition, chemically bond to the product). Thus
CdS will chemisorb sulphide and/or cadmium ions, depending on the absorption
properties of the CdS, in particular the crystal face involved. The crystal size of
deposits formed via the ion-by-ion process tends to be larger than those of the
same material deposited via the hydroxide mechanism (see Sec. 3.4.3). This can
be explained by slower nucleation, resulting in fewer nucleii; growth by homoge-
neous formation of chalcogenide ion tends to favor slower nucleation for the ion-
by-ion process. These nuclei therefore have more lateral room to grow in the plane
of the substrate.* In principle, large crystals (of the order of microns) could be ex-
pected in this case. While true in some cases, crystal size was usually much
smaller than this. Growth termination by adsorption of various species that are not
involved in the growth process or by defect formation are possible reasons for this.
       One possible measure of ion-by-ion growth is that, in contrast to a cluster
mechanism, deposition of fresh material will generally be preferred on already-ex-
isting deposit, as already noted. This means that the number of nucleii may not in-
crease greatly (after the very early stages of film formation) and that the film will
grow by growth of these initial crystals. This has been reported, for example, in the
case of Ag2S deposition from a thiourea/ammonia bath [58]. While the mechanism
of this deposition is not certain, the combination of this growth pattern together
with the strong complexation of the Ag by thiourea (strong AgMS bond in the
thiourea complex) suggests an ion-by-ion complex-decomposition mechanism.
       Growth of various semiconductors onto certain single-crystal substrates has
resulted in epitaxial growth in a number of cases. This epitaxy has been well stud-
ied for CdS deposition by Lincot et al. [59–63]. Although the epitaxy requires a
certain degree of lattice matching between semiconductor and substrate, chemical
interactions between the constituents of the deposition solution and the substrate
are important as well (discussed in more detail in Chap. 4). It is a reasonable as-
sumption that epitaxial deposition occurs via an ion-by-ion process. Indeed, it has



* It is a characteristic of the ion-by-ion mechanism for CdSe deposition from NTA solutions that thin
films are highly scattering, and this scattering decreases as film thickness increases. It is likely that this
scattering is due to voids in the thin films, which are a result of the low density of initial nuclei.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
been observed that epitaxy ceases when CdS colloids begin to appear in the solu-
tion, allowing a cluster growth to occur [63].
       Termination of growth due to depletion of reactants will eventually occur.
The reactants do not have to be completely used up. The concentration of free
metal ions will decrease not only due to decrease in total metal concentration, but
also because the ratio of the metal ion to complexant (the concentration of the lat-
ter usually remains constant throughout the reaction unless ammonia is used in an
open vessel) also decreases. Additionally, as the chalcogen source is used up, the
chalcogenide-forming (or chalcogen complexation) reaction slows down. Thus, at
some point, the rate of deposition will slow down to an unacceptable degree, even
though there may be an appreciable fraction of the reactants remaining in the de-
position solution. For the ion-by-ion growth, as the deposition slows and there is
more time for rearrangement of newly formed material to its most stable configu-
ration, the likelihood of larger and better-formed crystals may increase, as shown
in an early study of PbSe deposition [13].

3.4.3 Cluster Mechanism
The general shape of the growth of CD films as a function of time is often similar
for the cluster mechanism as for the ion-by-ion mechanism (Fig. 2.4). Figure 3.4
shows an actual example of CdSe deposition from a solution (containing nitrilo-




FIG. 3.4 Time dependence of CdSe (nitrilotriacetate bath, hydroxide mechanism, room
temperature) film growth measured by quartz crystal microbalance.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
triacetate as complex and under conditions where Cd(OH)2 is present as a colloid)
measured by a quartz crystal balance. This deposition was not continued to the in-
evitable termination point, but the initial stages of the induction period followed
by essentially linear growth are clear.
       Since there is a colloidal phase present in the solution from the very begin-
ning, why is there an induction period at all in this growth? Should these colloids
not stick to the substrate and build up a film? The answer, at least in this specific
case (and probably in many others), is yes and no; yes, the colloids will stick to
the substrate, and no, they need not build up to form a film. There are a number of
studies that show that Cd(OH)2 forms on the substrate from the start of the pro-
cess. An early radiochemical study of CdS deposition showed the initial presence
of Cd, free of S, on glass immersed in the (ammoniacal) deposition solution both
with and without thiourea present [39]. The amount of Cd (considered to be
Cd(OH)2) was constant with time (after an initial short time) until S began to be
detected in the deposit, at which point it started to increase. Cd(OH)2 was found
on the substrate, using XPS analyses, for CdSe deposition on glass [15] and for
CdS deposition on Si using an ammonia bath [37], only using deposition solutions
that did not contain the chalcogen presursor. The Cd(OH)2 only formed on the
substrate at a high-enough pH (typically 9 and above, although this value will de-
pend greatly on other solution parameters, in particular complex:Cd ratio and tem-
perature). In both cases, the coverage of the substrates by the Cd(OH)2 was usu-
ally sparse, as evidenced by the predominant substrate (Si) signals, and never
increased to a level where the Si was not predominant. TEM studies of the
Cd(OH)2 formation in the CdSe deposition, where selenosulphate was present,
confirmed this sparse coverage by Cd(OH)2[15]; CdSe formed only after some
time. Similar results were obtained for PbSe deposition; immersion of Au-coated
glass in an alkaline solution of (selenosulphate-free) citrate-complexed PbAc2 re-
sulted in adhesion of hydrated lead oxide (there is no stable simple lead hydrox-
ide) to the Au, which was not washed off by water (Fig. 3.5), although in this case,
PbSe formed more rapidly than did CdSe when selenosulphate was present [46].
       All these results show that Cd(OH)2 colloids do adsorb on a substrate (ei-
ther under conditions where Cd(OH)2 is present in solution or, according to the
studies of Rieke and Bentjen and Ortega-Borges and Lincot [48], even when it is
not present in solution but under solution conditions close to solid hydroxide for-
mation). The induction period when “no” deposition is seen in the hydroxide-clus-
ter deposition therefore is understood to mean that a fast and nongrowing
Cd(OH)2 adsorption has occurred, which is too fast and/or too little to measure by
the experimental methods used to make the kinetic curves, and that only when the
hydroxide starts to convert into the chalcogenide, by reaction of the slowly formed
chalcogenide ion with the hydroxide, does real film formation proceed.
       An obvious question at this point is: Why does CdS (CdSe) grow, i.e., con-
tinue to deposit and form a thick film, but Cd(OH)2 does not? Rieke and Bentjen



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 3.5 TEM image of Au film-on-glass immersed for 3 min in a solution of PbAc2 (60
mM) and trisodium citrate (160 mM) at pH 10.8. Electron diffraction showed the black
crystals to be hydrated lead oxide.


discussed the adsorption of various Cd and Cd-OH species onto the Si substrate.
The Si will be covered with an oxide (therefore it can be treated similarly to quartz
and even approximated to glass for the present purposes), and this oxide will be
negatively charged at the pH values involved. Adsorption of positive ions from so-
lution (Cd2 , Cd(OH) ), would be favored, and would eventually neutralize this
charge, after which further adsorption of positively charged species would no
longer occur. This scenario, however, was considered to conflict with the obser-
vation that greater amounts of Cd(OH)2 were found on the (Si) substrate at higher
pH values, where the concentrations of the positive species would be greatly re-
duced. Adsorption of neutral Cd(OH)2 was considered more likely. This adsorp-
tion could occur in different ways. Direct adsorption of Cd(OH)2 colloids from so-
lution was one possibility. However, since Cd(OH)2 could form, at least to some
extent, on the substrate from solutions where no Cd(OH)2 was present (see ear-
lier), surface-catalyzed adsorption was considered, such as [48].
      Cd(NH3) 2
              4       2OH       surface site D [Cd(OH)2]ads      4NH3         (3.56)
where the OH may or may not be that originally bound at the surface, or, in the
hydroxide–chalcogenide complex-decomposition mechanism, by adsorption of
the ammine-hydroxide species in Eq. (3.52).
       The various observations that, at relatively low pH, the coverage of the sub-
strate by Cd(OH)2 was poor indicate the dynamic equilibrium between adsorbed
Cd(OH)2 and the solution; the Cd(OH)2 was in a continual state of dissolution and
deposition. Since the concentrations of both free Cd2 and OH was constant (in



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the absence of large-scale depletion of the reactants, the case when no chalco-
genide precursor was added), this dynamic equilibrium resulted in a steady (and
low) concentration of Cd(OH)2 on the substrate.
       When chalcogenide ion begins to form and converts the hydroxide to the
chalcogenide (the simple hydroxide mechanism), or when the chalcogenide pre-
cursor forms a complex with the surface-adsorbed species followed by complex
dissociation (the complex-decomposition mechanism), this disturbs the equilib-
rium, allowing more surface-adsorbed hydroxide to form, resulting in film
growth. [Note that films of Cd(OH)2 (and other hydroxides) can be grown by CD
under conditions where generation of OH occurs; i.e., the system is not in equi-
librium]. This is one likely scenario for the linear part of the film growth. There
are other possibilities, and it is likely that more than one mechanism of growth oc-
curs. These other possibilities include direct adsorption of colloids from the solu-
tion and parallel ion-by-ion growth on the primary deposit.
       Direct adsorption of colloids can certainly occur. However, this mechanism
is often associated with poorly adherent and optically scattering films. It occurs
with these properties if the substrate is placed so that sedimentation of colloids
from the solution occurs directly onto the substrate. For this reason, the substrate
should be placed either vertical in the solution or, if non vertical, with only the
lower side of the deposit (which forms on both sides of the substrate) retained. In
principle, direct adhesion of colloids could result in adherent films, but it is more
likely that this involves isolated colloids or small aggregates that have greater con-
tact area than a large aggregate. Greater contact area here means that 200 (for ex-
ample) colloids that have adsorbed onto the surface one by one forming a single
aggregate would have a larger contact area to that surface than the same 200 col-
loid aggregate that adsorbed, as an aggregate, in a single step. There is also the
question of whether a Cd(OH)2 colloid would adhere better to a CdS surface than
to itself (for the case where hydroxide is present in solution). Since the hydroxide
is in a dynamic equilibrium, nonaggregated particles (or particles containing only
a small number of crystals) will be more likely to be present for the hydroxide than
for the chalcogenide (where the equivalent equilibrium is likely only for very tiny
nuclei).
       Parallel ion-by-ion growth might occur on previously deposited (by the hy-
droxide mechanism) film. However, at least in the simple hydroxide mechanism,
where a solid hydroxide is present in solution, this is not likely to be a major fac-
tor in the growth, except under solution conditions close to the transition between
hydroxide and ion-by-ion growth. The reason for this is that the chalcogenide is
being formed homogeneously throughout the solution. In the case where hydrox-
ide is present in solution, most of this chalcogenide will react with colloidal metal
hydroxide in solution (as seen in most cases by precipitated metal chalcogenide),
and the concentration at the substrate will be very low—enough to convert hy-
droxide to chalcogenide perhaps, but less likely to form a new phase.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       For the case where no hydroxide is present in solution but is formed on the
substrate while the homogeneously formed chalcogenide will not react with hy-
droxide in solution, some or even much of it may still not reach the substrate. This
is due to the relative lack of stability of the chalcogenide ions, which could be ox-
idized homogeneously, e.g., by dissolved oxygen. This will be more important for
selenide than for the considerably more stable sulphide; but even sulphide is not
stable in very low concentrations unless oxygen is rigorously excluded. This po-
tential problem will not exist for the complex-decomposition mechanism, where
there is a high concentration of the chalcogenide complex.
       The crystal size of the deposit obtained by the simple hydroxide mechanism
will be essentially that of the Cd(OH)2 clusters, which are converted into, e.g., CdS.
Such colloids tend to be very small (typically in the region of 5 nm), and therefore
the CdS crystal size should be similar. It may be larger if some form of ion-by-ion
growth occurs on these primary crystals. Also, it may be different if the hydroxide
is present on the substrate but not in solution, since the size of the hydroxide deposit
on the substrate will, at least in part, be affected by different factors than that formed
homogeneously in the solution. In general, crystal size in the simple hydroxide
mechanism (hydroxide colloid present in solution) is smaller than that formed in an
ion-by-ion process. This has been shown for CdS, CdSe, and PbSe, where typical
crystal sizes for the hydroxide (ion-by-ion) mechanisms were found to be (in nm):
CdS—5 ( 70); CdSe—5 (15); PbSe—5 (10–1000) [15]. In contrast to the ion-by-
ion mechanism, the crystal size of films formed via a cluster mechanism is not ex-
pected to grow greatly (some modest growth can and usually does occur) with in-
creasing film thickness, since film growth occurs by sequential addition of new
clusters. Thus, in our observation of CdSe growth via the hydroxide mechanism,
the color of the film (a sensitive measure of crystal size due to size quantization; see
Chap. 10) changes only a little during the deposition (color refers to spectral posi-
tion and not to depth of color, which, of course, does increase with film thickness).
       SEM micrographs in the study of Rieke and Bentjen [37] showed that, al-
though the number density of CdS nucleii/unit area on the substrate was constant
with time after deposition started, the size of the nucleii increased linearly with
time. Additionally, the size distribution of the nucleii, both in the early and later
stages of growth, was quite narrow. These film growth kinetics were identified
with burst-type nucleation, well known in homogeneous solution precipitation,
where a homogeneous reaction in solution causes sudden nucleation whenever a
critical concentration of one of the reactants is reached. This nucleation reduces
the concentration of this reactant so that further growth occurs only on existing nu-
cleii (nucleation usually requires a large supersaturation, while growth on an ex-
isting nucleus does not). This type of nucleation usually results in a narrow size
distribution, as seen here.
       In a quartz crystal microbalance investigation of CdSe film growth rate from
a selenosulphate/ammonia/triethanolamine bath with different Cd:selenosulphate



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ratios, two peaks were observed in the growth rate vs. time plot of all solutions
[64]. This was explained by a two-stage growth. From electron microscopic ex-
amination of the growing films, the first stage was attributed to instantaneous nu-
cleation and 2-D lateral growth to cover the substrate, while the second stage was
due to 3-D nucleation and growth at random sites on the first layer. Aggregation
of colloidal particles was invoked as the mechanistic pathway.
       After all the speculation involved in the foregoing discussion of film
growth, the termination step is as simple to explain as for the ion-by-ion mecha-
nism. Growth can occur as long as the concentration of chalcogenide anion is high
enough to allow Reaction (3.5), the conversion of the hydroxide to the chalco-
genide, to occur [or, for the complex-decomposition mechanism, sufficient
chalcogenide precursor as, e.g., in Eq. (3.7)]. It is also possible that depletion of
the metal hydroxide occurs first (in which case the mechanism may change to ion-
by-ion, as described earlier). In either case, termination is simply due to depletion
of the reactants. Typically in the cluster mechanism, most of the reactants are lost
in homogeneous precipitation.

3.4.4 Complex-Decomposition Mechanism
The initial nucleation stage of the complex-decomposition mechanism is probably
similar to the simple free-anion mechanism. Either ionic or molecular metal
species (ion-by-ion) or Cd(OH)2 (cluster) adsorbs on the substrate. However, in-
stead of conversion of the hydroxide to sulphide by topotactic reaction with sul-
phide ions, the chalcogenide precursor (in almost all studies of this mechanism,
that is thiourea) adsorbs on the Cd(OH)2 surface to form a hydroxide–thiourea
complex, which then decomposes to CdS.
       A possible difference between the simple hydroxide mechanism and the
complex decomposition is in the manner of crystal growth. We noted earlier that
crystal growth in the simple hydroxide mechanism may occur via an ion-by-ion
process but probably not to a large degree, since the concentration of chalcogenide
ion would be very low indeed (it would react rapidly with Cd(OH)2 in the solu-
tion). However, for the hydroxide-decomposition mechanism, the chalcogenide
reactant is not the free chalcogenide ion but the precursor, which is present in
much higher concentrations. Therefore once a solid phase capable of catalyzing
the chalcogenide precursor has formed, the crystal growth is quite likely to
switch over to a predominantly ion-by-ion process, such as in Reaction (3.51).
This means that, as for the pure ion-by-ion process, the crystal size might be
expected to be larger than for the pure hydroxide mechanism, since the ion-by-ion
growth favors, in principle, crystal growth rather than renucleation. Typical
crystal sizes for CdS prepared from the ammonia/thiourea bath, which appears un-
der many experimental conditions to proceed via the hydroxide-complex mecha-
nism, are in the region of 10 nm to several tens of nanometers, larger than the ap-
proximately 5 nm obtained for CdS and CdSe from the simple hydroxide-cluster
mechanism.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
3.5 KINETICS OF DEPOSITION
Kinetic studies have been a popular topic in CD, particularly for CdS. This would
suggest that the present section will be a large one to reflect this activity. In fact,
the reverse will be the case. This will be a relatively short section that will not try
to cover even a moderately large part of the kinetic studies. The reason is that ki-
netic measurements have been used, to a large extent, to study the mechanisms of
deposition, and this has been dealt with already (not the details of the kinetic stud-
ies, but the conclusions). Additionally, since a CD process can vary widely in
rate—from a few minutes to days and weeks—often depending strongly on small
changes in one or another concentration of a particular reactant, the important in-
formation to be learned from kinetic studies (apart from mechanistic diagnosis) is
how this rate depends on experimental conditions, and this can be done with a few
selected examples from the literature.
      Ortega-Borges and Lincot [48] carried out a detailed kinetic study of CdS
deposition from the standard ammonia (ammonium)/thiourea bath using a quartz
crystal microbalance to measure film thickness. They measured a deposition rate
with fractional values of reaction order
                   [Cd]0.6[Tu]0.8
      rate    K                                                                 (3.57)
                  [NH3]3.3[H ]1.5
They therefore concluded that several different rate-determining steps were in-
volved in the deposition. Figure 3.6 shows the dependence of the deposition rate
on the concentration of the reactants (Cd, thiourea, ammonia, and pH—the last
varied through introduction of ammonium ion) (a) as well as an Arrhenius plot of
the deposition (b) for the CdS deposition. From the kinetic data, they deduced the
hydroxide-complex-decomposition mechanism, given earlier in Eqs. (3.50) and
(3.51) and, more specifically, as
      Cd(NH3) 2
              4       2OH       surface site D Cd(OH)2(ads)        4NH3         (3.58)
This first step represents a reversible adsorption of Cd(OH)2 on the substrate. The
reaction order of 1.5 for hydroxide [given by H in Eq. (3.57)] implied the par-
ticipation of two hydroxide ions in the process. The next step was formation of a
complex between the adsorbed Cd(OH)2 and thiourea:
      Cd(OH)2(ads)       (NH2)2CS → Cd(OH)2 SC(NH2)2(ads)                       (3.59)
This complex then decomposes to CdS:
      Cd(OH)2 SC(NH2)2(ads) → CdS            CN2H2      2H2O                    (3.60)
probably by nucleophilic attack of the thiourea S atom on the Cd(OH)2. The acti-
vation energy of the deposition (85 KJ/mole), measured from the Arrhenius plot
in Figure 3.6b, is very similar to that of thiourea decomposition, suggesting the
slow nature of Reaction (3.60). Earlier measurements of PbSe deposition from a


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 3.6 (a) Log–log plots of reactant concentration (X) vs. deposition rate for CdS de-
posited from an ammonia/thiourea bath. Standard conditions: [Cd]          14 mM; [Tu]
(thiourea) 28 mM; [NH3] 1.74 M; T 60°C. Reactants are Cd (total concentration);
Tu; NH3; pH (adjusted by adding ammonium ion) that gives hydroxide concentration.
(b) Temperature dependence of deposition rate. (Adapted from Ref. 48).


selenourea bath gave an activation energy of 60 KJ/mole [65], with the lower
value compared to CdS deposition presumably reflecting the greater instability of
selenourea compared to thiourea.
       Using the same basic system and similar experiments, Doña and Herrero
measured reaction orders for the various species comparable to those measured in
the study of Ortega-Borges and Lincot, except for that of ammonia, which was 1.8
instead of 3.3 [49]:
                 [Cd]0.9[Tu]1.1[OH ]1.7
      rate   K                                                                   (3.61)
                        [NH3]1.8
This, together with the known tendency of metal ions to form mixed hydroxy–am-
mine complexes, suggested to them that two ammonia molecules were involved
in the first step and that the adsorbed species in Reaction (3.58) was a hy-
droxy–ammine species, viz. Cd(OH)2(NH3)2. Decomposition of the hydroxide–
ammine–thiourea complex was then assumed to occur by nucleophilic attack of an
ammonia species on the SBC bond of the thiourea.
       Based on this mechanism, a detailed theoretical model has recently been
proposed for CdS deposition from the thiourea/ammonia bath [65a]. Prediction of
different aspects of the deposition kinetics using this model provided a very good
fit with the relevant experimental data.
       Rieke and Bentjen [37] studied the kinetics of CdS deposition from an am-
moniacal thiourea bath using SEM. As discussed earlier, they found that good-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
quality films formed in a pH range where Cd(OH)2 formed on the substrate (Si, in
their experiments) but not in the bulk of the solution. Their kinetic study was made
in this pH range (specifically, pH 9.55). No CdS deposited initially, but the rate
of formation of CdS increased with time, eventually becoming more or less con-
stant over the time of their experiment. This is characteristic of an autocatalytic re-
action, where the initial deposit accelerates the rate of further deposition.
       O’Brien and Saeed, using ethylenediamine as compexant, higher deposition
temperatures, and glass as a substrate, found that the thickness of the CdS film in-
creased linearly with time (after an initial induction period) and also that there was
no increase in the size of the nucleii (both in contrast to the previous study) [40].
In spite of the different experimental conditions, the mechanism of the depositions
in both studies appears to be essentially the same, i.e., hydroxide-mediated catal-
ysis of thiourea decomposition.
       It must be kept in mind that the kinetics of CD, as with the deposition mech-
anism, can be very different from one system to another. Two connected exam-
ples of this are given here.
       In one study of PbSe deposition from a citrate-complexed selenourea solu-
tion containing hydrazine, the rate was proportional to the pH and to the sele-
nourea concentrations but independent of the Pb and citrate concentrations [65].
This was explained by a rate-determining step involving decomposition of sele-
nourea at the (catalytic) PbSe surface by hydroxide. It is noteworthy that the Pb
concentration was typically an order of magnitude less than that of selenourea.
Therefore the independence of the rate on Pb (or citrate, which determines the
concentration of free Pb2 ) concentration, suggests that formation of selenide ion,
and not a complex-decomposition mechanism, occurs.
       The second example is seen in the study of PbSe deposition by Kainthla et
al. from selenosulphate solution [41]. In most examples of CD from alkaline so-
lution, the deposition rate increases with increase in pH. This is due to both the
greater rate of decomposition of the chalcogenide precursor at higher pH (this de-
composition usually involves hydroxide ions) and, in many cases, the greater
probability of solid hydroxide formation (as long as this is not excessive). How-
ever, for PbSe deposition using citrate as complex for the Pb and selenosulphate
as Se precursor, the opposite occurs: The deposition rate decreases with increase
in pH. This is due to the specific hydroxy–citrate complex formed:
      Pb(OH)C6H5O 2 D Pb2
                  7                   OH       C6H5O 3
                                                     7                          (3.62)
Increase in pH ( increase in [OH ]) shifts the equilibrium to the left, resulting
in a lower concentration of free Pb2 ions and thus a slower reaction to give PbSe.
This means that, in contrast to the deposition from a selenourea bath described ear-
lier, the rate is dependent on Pb concentration and possibly independent of hy-
droxide concentration at a constant free-Pb2 concentration. This would then sug-
gest that the opposite mechanism, i.e., a complex decomposition, is effective for
the selenosulphate bath. It is stressed that these conclusions on selenide formation


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
or complex-decomposition mechanisms are indications of which mechanism is
taking place, but are far from being firm proof of this.
       For the deposition of ZnS and ZnSe, hydrazine is normally used to form
films at a reasonable rate. The role of the hydrazine is not obvious. It is tempting
to assume that hydrazine, being a strong reductant, reduces the chalcogen precur-
sor to chalcogenide ion, as was assumed for CdTe deposition (see Sec. 3.2.3).
However, this appears to be oversimplified. In a study of the effect of various
amines (including hydrazine) on the deposition rate (and composition) of ZnS
films deposited from ammonia/thiourea baths, Mokili et al. found a strong depen-
dence of the rate on the type of amine added [66], as shown in Figure. 3.7. While
it is difficult to separate the effect of concentration from the different types of
amines in this experiment, it is clear that an increase in rate is general on addition
of amine (apart from the initial induction time using triethanolamine). Since the
amines also act as complexation agents, they would, on this basis, be expected to
reduce the deposition rate (by reducing the free Cd2 concentration). The fact that
the opposite occurs implies that they must increase the thiourea (or selenosulphate
for selenides) decomposition. In this respect, all the amines used have pronounced
reducing properties, with redox potentials of 0.46 (triethanolamine), 0.56
(ethanolamine), and 1.16 (hydrazine), which parallels the order of increase
in deposition rate seen in Figure 3.7. O’Brien et al. discuss various theories for




FIG. 3.7 Evolution of ZnS film thickness with time. ZnCl2/NH4Cl/thiourea/NH3 bath at
85°C. The effect of various amines on growth. Triethanolamine—0.2 M; ethanolamine—
0.7 M; hydrazine—3 M. (Adapted from Ref. 66).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the accelerating effect of hydrazine on the CD process [67]. While agreeing
that the hydrazine increases thiourea decomposition, the specific details of the
effect of hydrazine (and other amines) on the CD process are still not fully
understood.
       The activation energies of the deposition for both ZnS [68] and ZnSe [69],
measured from Arrhenius plots, are 21 and 26 KJ/mole, respectively, much
smaller than the values for CdS (85 KJ/mole) or PbSe (60 KJ/mole) described ear-
lier. Stirring does not affect the deposition rate for either ZnS or ZnSe, so the de-
position is not under diffusion control. In interpreting activation energies for CD
processes, it is important to remember that what is measured is the film growth,
and this is not necessarily the same as the rate of formation of the metal chalco-
genide, much of which is usually formed homogeneously in the solution. The Zn
compounds were both probably formed by a cluster mechanism, in contrast to the
ion-by-ion complex-decomposition mechanism probable for the CdS. The “acti-
vation energy of deposition” for the ZnS(Se) therefore depends to a large extent
on the rate of sticking of clusters, although other factors could also be involved,
such as a parallel ion-by-ion (whether by complex decomposition or free chalco-
genide ions). To interpret such results correctly will require a study of activation
energies of different compounds deposited under different and controlled mecha-
nistic pathways and preferably also measuring the total amount of product formed
(in solution as well as on the substrate and other surfaces). It is also relevant that
the crystal size for ZnS and ZnSe deposits is typically smaller than for CdS or
CdSe deposited from an ion-by-ion bath; this supports a cluster mechanism for
these depositions. In fact, in contrast to CdS and CdSe, there are no cases in the
literature where ZnS or ZnSe have been clearly shown to have been deposited via
an ion-by-ion mechanism.


3.6 DEPOSITION FROM ACIDIC BATHS
While the majority of CD reactions have been carried out in alkaline baths, there
have been a number of sulphide depositions reported in acidic baths. These are all
(with one exception, Sn-S from acidic nonaqueous S baths [70]; see Chap. 6)
based on two sulphide precursors: thiosulphate (more commonly used in the very
early studies, but still sometimes used) and thioacetamide. Beutel and Kutzelnigg
[71] described a large selection of colored films deposited on various metals us-
ing CD (and also electrodeposition) from metal salt–thiosulphate solutions. No
characterization of these films (other than interference colors) was made. How-
ever, it is clear that CD of metal sulphides did occur in some of these cases (see
Chap. 6). Lokhande described a variety of sulphides (CdS, ZnS, Bi2S3, Sb2S3,
As2S3, Cu2S, Ag2S, SnS2, and PdS2) deposited using thiosulphate at a pH of typ-
ically ca. 3 [72]. Other studies on deposition from acidic thiosulphate baths are:
PbS [73], Ag2S [74], Bi2S3 [75], CuxS [76,77], Sb2S3 [78], SnS2 [79].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      The mechanism of the thiosulphate reaction is not clear. Lokhande, in his
studies, has suggested an internal reduction involving the reaction
      2S2O 2 → S4O 2
           3       6        2e                                               (3.63)
These electrons reduce elemental S formed in Reaction (3.21) or (3.22) to give
sulphide (or hydrosulphide) ions, as in Reaction (3.24), which react with the metal
ions. The general mechanism of sulphide generation has been assumed in most
studies using thiosulphate, e.g., Nair et al [78] for Sb2S3 deposition and Groz-
danov et al. in their study of CuxS deposition from Cu-thiosulphate solutions [76],
although in this latter study it is noted that the mechanism may be more compli-
cated in this Cu-S system.
       As has already been pointed out, in spite of the fact that the free sulphide
mechanism is invariably assumed in thiosulphate depositions, there is no evidence
up to now against the complex-decomposition mechanism. No thorough mecha-
nistic or kinetic studies have been made on this system. Since the studies on CD
using thiosulphate have not attempted to differentiate between these different
mechanisms (and since such differentiation may be difficult, this is not surpris-
ing), we are left with the conclusion that there is no clear consensus on which
mechanism is operative. Also, the mechanism may vary depending on conditions
(as for the alkaline baths), and of course a combination of mechanisms may be op-
erative in some cases.
       Other points to note when considering thiosulphate as a reagent in CD is that
thiosulphate is a strong complex for a number of metals and, since it is a fair re-
ducing agent, also may reduce the metal ions (as is known to occur for Cu2
to Cu ).
       Many sulphides have been deposited using thioacetamide in acidic solutions
(Chapter 6 describes most of these). For depositions using thioacetamide, as with
thiosulphate, there are no detailed mechanistic studies. Both H2S formation and
complex decomposition are possible in acid solutions, as discussed in Section
3.2.1.3. Deposition of CdS was accomplished using thioacetamide in acidic solu-
tion by exploiting electrolytic proton reduction to increase the pH locally at the
cathode (substrate), and the mechanism was believed to be a surface-catalyzed de-
composition of a Cd–thioacetamide complex [80].
       Because of the acidic conditions, with the exception of very acidic cations,
nucleation resulting from a solid colloidal phase is unlikely. For a deposition that
occurs through free (hydro)sulphide ion, it is probable that nucleation occurs
when the concentration of this sulphide ion is high enough to cause precipitation
of the metal sulphide, possibly catalyzed by the substrate surface. Similarly for a
complex-decomposition pathway, a high-enough concentration of the final prod-
uct to permit solid-phase formation will be required. It should be remembered that
the concentration of free metal ion will, in most cases, be higher than for alkaline
baths, due to the (usually) weaker complexing strengths of the sulphide precur-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sors. This will allow formation of a solid phase at lower concentrations of sulphide
than in more heavily complexed alkaline solutions.
       There appear to be no cases of selenide deposition from acidic baths. Se-
lenosulphate is not stable under even mildly acidic conditions, and all selenourea-
based baths have been alkaline ones.
       Oxides or hydroxides have been deposited from acid baths, in particular
readily hydrolyzable acidic metal ions. These have been discussed in Section
3.2.4.


3.7 EFFECT OF STIRRING
There have been only a few studies on the effects of stirring the deposition solu-
tion on the deposited film. Overall, stirring affects CD films mainly by preventing
deposition of loosely adhering, large aggregates. These loose deposits are readily
removed by the stirring action. This is important, since they block the substrate,
preventing normal adherent film growth. Such nonadherent deposits can also be
prevented without stirring by placing the substrate in the bath at an angle; the de-
posit on the upper surface, which will usually be a mixture of adherent and loosely
adherent material, can be removed (by wiping with a reagent that dissolves the
film, often dilute HCl), leaving the film on the lower surface, which does not col-
lect precipitated deposit.
       Such loosely adhering CdS films in nonstirred solutions have been reported
by Kaur et al. [34] and by Doña and Herrero [49]. The latter and also Ortega-
Borges and Lincot [48] found that the rate of deposition is affected by stirring only
at low stirring rates, and the effect is not large. There is no apparent difference be-
tween low and fast stirring rates. This implies that even slow stirring is enough to
prevent sticking of large, loosely adhering particles. Apart from deposition rate,
the study by Kaur et al. found that stirring could, in some cases, strongly affect the
film quality and that the effect of stirring was dependent on the concentration of
ammonia (relative to the Cd). For low ammonia concentrations, where a visible
Cd(OH)2 phase existed, and for high ammonia concentrations, where it is proba-
ble that no solid hydroxide phase occurred, strongly stirred solutions resulted in
formation of CdS films with predominantly wurtzite structure, while films de-
posited from unstirred solutions contained a large amount of zincblende CdS. Pre-
cipitate formed homogeneously in the solution was found to be zincblende, and
the effect of stirring was to minimize sticking of colloidal particles from solution,
thereby reducing the zincblende component in the films (and also resulting in
more adherent and specularly reflecting films—adsorption of large aggregated
colloids from solution caused light scattering and reduction in specular re-
flectance). For the intermediate case where there was just enough ammonia to dis-
solve the visible Cd(OH)2 (but a colloidal Cd(OH)2 probably was present in solu-
tion), stirring did not have any major effect—thick, powdery, and zincblende



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
films were obtained in both cases. This is unexpected, since these are conditions
where good films are often obtained.
       There is one unusual case where stirring had a very strong effect on the de-
position rate: Ag2S deposited from a thiourea bath [81]. The deposition was very
slow in the unstirred solution and increased, more rapidly at first, then linearly, up
to the maximum stirring speed of ca. 1100 rpm. The energy of activation of this
deposition was 20.4 kJ/mole, much less than values typical of reaction-controlled
depositions under similar conditions involving, e.g., CdS or PbS, and similar to
those found for ZnS and ZnSe. However, the Zn-S(Se) depositions were indepen-
dent of stirring. It therefore appears that the mechanism of this Ag2S deposition is
diffusion controlled and is unlike other mechanisms discussed previously.
       As pointed out by Ortega-Borges and Lincot, the relative independence
of the majority of CD processes on hydrodynamic conditions explains the
excellent lateral homogeneity characteristic of this technique, since the deposition
depends not so much on mass transport in the solution as on chemical reaction
rates.


REFERENCES
 1.   WHR Shaw, DG Walker. J. Am. Chem. Soc. 78:5769, 1956.
 2.   DF Bowersox, DM Smith, EH Swift. Talanta 3:282, 1960.
 3.   OM Peeters, CJ de Ranter. J. Chem. Soc. Perkin II 1832, 1974.
 4.   R Williams, PN Yocom, FS Stofko. J. Colloid Interface Sci. 106:388, 1985.
 5.   WH Perkins, AT King. J. Chem. Soc. 103:301, 1913.
 6.   WN Arnquist. Proc. IRE 47:1420, 1959.
 7.   CJ Milner, BN Watts. Nature 163:322, 1949.
 8.   GA Kitaev, TP Sokolova. Russ. J. Inorg. Chem. 15:167, 1970.
 9.   GA Kitaev, VM Makova. USSR Patent 377,445:C.A. 79:84806j, 1973.
10.   RA Zingaro, FC Bennett Jr, GW Hammar. J. Org. Chem. 18:292, 1953.
11.   RA Zingaro, DO Skovlin. J. Electrochem. Soc. 111:42, 1964.
12.   AA Velykanov, EK Ostrovskaya, NP Garina, VA Turacova, AA Tchurkan. Ukr.
      Chim. Zh. 49:764, 1983.
13.   GM Fofanov, GA Kitaev. Russ. J. Inorg. Chem. 14:322, 1969.
14.   GA Kitaev, TS Terekhova. Russ. J. Inorg. Chem. 15:25, 1970.
15.   S Gorer, G Hodes. J. Phys. Chem. 98:5338, 1994.
16.   GA Kitaev, GM Fofanov. Zh. Prikl. Khim. 43:1694, 1970.
17.   GK Padam, SK Gupta. Appl. Phys. Lett. 53:865, 1988.
18.   RW Buckley. In: 11th ECPV Solar Energy Conf: Montreux, Switzerland, 1992,
      p 962.
19.   AV Kalyakina, RI Pelyukpashidi. Tr. Khim. Met. Inst. Akad. Nauk. Kaz. SSR
      17:114, 1973.
20.   VB Patil, PD More, DS Sutrave, GS Shahane, RN Mulik, LP Deshmukh. Mater.
      Chem. Phys. 65:282, 2000.
21.   M Sotelo-Lerma, RA Zingaro, SJ Castillo. J. Organomet. Chem. 623:81, 2001.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
22.    L Gordon. Anal. Chem. 24:459, 1952.
23.    HH Willard, L Gordon. Anal. Chem. 25:170, 1953.
24.    M Izaki, T Omi. J. Electrochem. Soc. 144:L3, 1997.
25.    M Izaki, O Shinoura. Electrochem. Solid State Lett. 1:215, 1998.
25a.   M Izaki. O Shinoura. Adv. Mater. 13:142, 2001.
26.    W Mindt. J. Electrochem. Soc. 117:615, 1970.
27.    TP Niesen, MR DeGuire. J. Electroceram. 6:169, 2001
28.    DH Klein, L Gordon, TH Walnut. Talanta 3:177, 1959.
29.    L Gordon, JI Peterson, BP Burtt. Anal. Chem. 27:1770, 1955.
30.    VK LaMer, RH Dinegar. J. Am. Chem. Soc. 73:380, 1951.
31.    L Gordon, ML Salutsky, HH Willard. Precipitation from Homogeneous Solutions.
       New York: Wiley, 1959.
32.    JR Gump, GR Sherwood. Anal. Chem. 22:496, 1950.
33.    PFS Cartwright. Analyst 92:319, 1967.
34.    I Kaur, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:943, 1980.
35.    ML Breen, JT Woodward, DK Schwartz, AW Apblett. Chem. Mat. 10:710, 1998.
36.    GA Kitaev, AA Uritskaya, SG Mokrushin. Russ. J. Phys. Chem. 39:1101, 1965.
37.    PC Rieke, SB Bentjen. Chem. Mat. 5:43, 1993.
38.    RO James, TW Healy. J. Colloid Interface Sci. 40:53, 1972.
39.    ND Betenekov, VP Medvedev, AS Zhukovskaya, GA Kitaev. Sov. Radiochem.
       20:524, 1979.
40.    P O’Brien, T Saeed. J. Crystal Growth 158:497, 1996.
41.    RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:277, 1980.
42.    H Cachet, M Froment, G Maurin. J. Electroanal. Chem. 406:239, 1996.
43.    P O’Brien, J McAleese. J. Mater. Chem. 8:2309, 1998.
44.    D Lincot, RO Borges. J. Electrochem. Soc. 139:1880, 1992.
45.    S Gorer, A Albu-Yaron, G Hodes. J. Phys. Chem. 99:16442, 1995.
46.    S Gorer, A Albu-Yaron, G Hodes. Chem. Mater. 7:1243, 1995.
47.    MK Norr. J. Phys. Chem. 65:1278, 1961.
48.    R Ortega-Borges, D Lincot. J. Electrochem. Soc. 140:3464, 1993.
49.    JM Doña, J Herrero. J. Electrochem. Soc. 144:4081, 1997.
50.    EB Andersen. Z. Phys. Chem. B. 32:237, 1936.
51.    E Beutel, A Kutzelnigg. Z. Elektrochem. 36:523, 1930.
52.    GA Kitaev, AA Uritskaya. Russ. J. Appl. Chem. 72:592, 1999.
53.    LE Yatlova, AA Uritskaya, GA Kitaev, TI Dzyuba. Russ. J. Appl. Chem. 66:1534,
       1993.
54.    GA Kitaev, AZ Khvorenkova. Russ. J. Appl. Chem. 71:1325, 1998.
55.    J Liu, SM Howard, KN Han Langmuir. 9:3635, 1993.
56.    I Dekany, L Turi, G Galbacs, JH Fendler. J. Colloid Interface Sci. 195:307, 1997.
57.    D Lincot, J Vedel. In: 10th ECPV Solar Energy Conf. Lisbon Portugal: 1991, p 931.
58.    TP Bol’shchikovsa, GA Kitaev, VI Dvoinin, MV Degtyarev, LM Dvoskina. Izv.
       Akad. SSSR, Neorg. Mater. 16:387, 1980.
59.    D Lincot, R Ortega-Borges, M Froment. Appl. Phys. Lett. 64:569, 1994.
60.    M Froment, MC Bernard, R Cortes, B Mokili, D Lincot. J. Electrochem. Soc.
       142:2642, 1995.
61.    R Cortes, M Froment, B Mokili, D Lincot. Philos. Mag. Letts. 73:209, 1996.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
62.    D Lincot, B Mokili, R Cortes, M Froment. Microsc. Microanal. Microstruct. 7:217,
       1996.
63.    MJ Furlong, M Froment, MC Bernard, R Cortés, AN Tiwari, M Krejci, H Zogg, D
       Lincot. J. Crystal Growth 193:114, 1998.
64.    H Cachet, H Essaaidi, M Froment, G Maurin. J. Electroanal. Chem. 396:175, 1995.
65.    AB Lundin, GA Kitaev. Inorg. Mater. 1:1905, 1965.
65a.   M. Kostoglou, N. Andritsos, AJ Karabelas. Ind. Eng. Chem. Res. 39:3272, 2000.
66.    B Mokili, M Froment, D Lincot. J. de Phys. IV 5:261, 1995.
67.    P O’Brien, DJ Otway, D Smith-Boyle. Thin Solid Films 361:17, 2000.
68.    JM Doña, J Herrero. J. Electrochem. Soc. 141:205, 1994.
69.    JM Doña, J Herrero. J. Electrochem. Soc. 142:764, 1995.
70.    RD Engelken, HE McCloud, C Lee, M Slayton, H Ghoreishi. J. Electrochem. Soc.
       134:2696, 1987.
71.    E Beutel, A Kutzelnigg. Monats. 58:295, 1931.
72.    CD Lokhande. Mater. Chem. Phys. 28:145, 1991.
73.    KM Gadave, SA Jodgudri, CD Lokhande. Thin Solid Films 245:7, 1994.
74.    SS Dhumure, CD Lokhande. Thin Solid Films 240:1, 1994.
75.    CD Lokhande, AU Ubale, PS Patil. Thin Solid Films 302:1, 1997.
76.    I Grozdanov, CK Barlingay, SK Dey. Thin Solid Films 250:67, 1994.
77.    I Grozdanov, M Najdoski. J. Solid State Chem. 114:469, 1995.
78.    MTS Nair, Y Pena, J Campos, VM García, PK Nair, J. Electrochem. Soc. 145:2113,
       1998.
79.    CD Lokhande. J. Phys. D: Appl. Phys. 23:1703, 1990.
80.    K Yamaguchi, T Yoshida, T Sugiura, H Minoura. J. Phys. Chem. B 102:9677, 1998.
81.    H Meherzi-Maghraoui, P Cowache, D Lincot, M Dachraoui. J. Chim. Phys. 96:259,
       1999.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
4
II–VI Semiconductors




There has been a clear emphasis, in the CD literature, on II–VI semiconductors,
mostly CdS, some CdSe, and recently on ZnS. This being the case, the reader may
reasonably expect this chapter to be a voluminous one. On the other hand, many
of these studies have focused on deposition mechanisms and kinetics (which are
dealt with in the previous chapter), with photovoltaic cells, and, to a lesser extent,
with quantum size effects, both of which will be dealt with in subsequent chapters.
Two detailed descriptions of the experimental procedure (for CdS and CdSe) are
given in Chapter 2. This leaves the obvious question: “What’s left?” The present
chapter will answer that question. This includes properties of the films not ex-
plicitly discussed in other sections, such as crystal structure, optical and electrical
properties, as well as variants of the deposition process. Also, more detail will be
given on non-Cd chalcogenides. In short, there is indeed much left.


4.1 CdS
A point concerning CdS deposition. Many studies have used what is referred here
to as the standard deposition bath. This bath is made up of a Cd salt, ammonia
(sometimes with an ammonium salt to lower the pH) to complex the Cd and ad-
just the pH, thiourea, and deposition temperatures usually in the range of
60–90°C. Of course, this bath still allows for large differences in reactant concen-
trations (e.g., the Cd concentration varies from a low of 1 mM to as much as 100



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mM). If the reader has read the previous chapter, it should be obvious that not only
is the concentration of various reactants important, but so is the ratio between the
Cd (or other metal ion) and the complexant. With this caveat, we will use the term
standard bath to cover all concentrations, unless there is a specific reason to do
otherwise. Since the majority of studies on CdS used this standard bath in one
form or another, the films discussed in this section can be assumed to have been
deposited from such a bath unless otherwise stated. Also for this reason, it is more
natural to begin with properties of the films and afterwards to discuss variations
in deposition.

4.1.1 Crystallography
In several cases where epitaxial growth occurs involving the ion-by-ion mecha-
nism, the crystal structure is dictated by the substrate structure. This is treated sep-
arately in Section 4.1.5.
       Many papers state that one or the other crystal modification is obtained,
without giving either diffraction data or/and where the data is ambiguous. The en-
ergy difference between the hexagonal (wurtzite) and cubic (zincblende) phases is
very small (the former is slightly more stable); hence both are often found to-
gether. This commonly leads to the presence of twins and stacking faults in the
crystals. The density of stacking faults in films deposited from a standard bath in-
creases with increased thiourea concentration or decreased Cd concentration, and
is typically 1011–1012 cm 3 [1]. If the cubic phase is annealed at ca. 400°C or
above, the hexagonal phase is normally obtained. In view of the lack of a con-
vincing explanation of why one or the other crystal structure is formed, a sampling
of reported crystal structures is given in list form, together with differences in
preparation from the standard bath (if any) that may give clues to the crystal mod-
ification obtained.
4.1.1.1 Hexagonal (Wurtzite)
      When insufficient NH3 was added and Cd(OH)2 was present as a clearly vis-
        ible suspension, hexagonal CdS formed on the substrate if the solution
        was well stirred; i.e., the precipitate in solution, which was cubic, was not
        allowed to accumulate on the substrate [2].
      A high resolution transmission electron microscopy (HTEM) study of the
        early stages of CdS deposition on a carbon-coated TEM grid showed only
        hexagonal CdS to be formed, while hexagonal with some cubic CdS was
        formed by precipitation in the solution [3].
      Using CdI2 in a standard bath, hexagonal CdS was obtained. (If CdCl2 was
        used, the deposit appeared to be zincblende, although it may also have
        been highly textured wurtzite) [4].)
      Using CdCl2 (Cd(Ac)2 gave no XRD) hexagonal CdS with moderate texture
        (0002) was deposited on glass [5].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      With ethylenediamine as complexant and with Cd(OH)2 present in solution,
        some hexagonal CdS was formed, although cubic CdS might also have
        been present [6].
      A citrate/ammonia bath gave predominantly hexagonal CdS [7].
      Some particularly clear examples of predominantly hexagonal formation
        were from an acid bath using thioacetamide [8], a triethanolamine bath
        under conditions where ion-by-ion deposition was believed to occur and
        the deposition rate was slow [9], and a nitrilotriacetate bath, where depo-
        sition was also slow but a hydroxide cluster deposition was shown to take
        place [10]. (Ion-by-ion growth, under conditions similar to the last exam-
        ple, but with a high enough complex concentration to prevent Cd(OH)2
        formation, also showed apparently highly textured hexagonal CdS, al-
        though in this case the predominant presence of this form was not unam-
        biguous [10].)
4.1.1.2 Cubic (zincblende)
      If just enough ammonia was added to dissolve Cd(OH)2, the cubic form was
obtained [2] (regardless of stirring—see Ref. 2).
      From a standard solution although the concentrations of reactants used were
         not given [11].
      Mainly cubic obtained from a standard bath on SnO2/glass over a range of
         conditions (including with and without ammonium buffer and using
         ethylenediamine instead of ammonia) [12].
      From a bath with low Cd concentration (1 mM) and high ammonia concen-
         trations (2 M) suggesting that the conditions were such as to favor ion-by-
         ion deposition [13]. Another study with low Cd concentration (2–5 mM),
         ammonia concentrations ca. 300 times higher than the Cd concentration,
         and added ammonium ions (which reduces the pH and therefore favors
         ion-by-ion deposition) likewise found only the cubic phase.
      Preferential (111) texture of cubic CdS on ITO/glass [14].
4.1.1.3 Mixed Hexagonal/Cubic
A mixture of phases was often reported. This is not surprising considering the
small energy difference between them. Some examples follow.
      Standard deposition giving thick films (close to 1 m thick) on glass re-
        sulted in films that were ca. 90% cubic and 10% hexagonal [15].
      Either not enough NH3 to dissolve the Cd(OH)2 and not stirred or a large ex-
        cess of NH3 and stirred [2].
      Standard deposition on tin oxide/glass [16].
      The precipitate formed in solution was predominantly cubic, with some
        hexagonal [17].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      Detailed analysis of different XRD techniques led to the conclusion that
        CdS used in CdS/CdTe PV cells was polytype, with essentially random
        stacking of cubic and hexagonal structures in individual crystals [18].
        This study goes a long way to explaining the wide variation in apparent
        crystal structure.
      A triethanolamine/ammonia bath gave a mixed deposit [19,20].
       Some comments on the role of the anion of the Cd salt are in order. Due to
the small energy difference between the two phases, small changes in adsorption
of solution species onto the growing crystals may be enough to dictate the final
crystal structure. Early studies have shown differences in the crystal structure of
CdS precipitates, depending on the anion of the Cd salt. Halides resulted in hexag-
onal CdS, sulphate gave cubic, while nitrate could give either, depending on tem-
perature and pH [21].
       Films deposited from a typical NH3/thiourea bath, using either the iodide or
chloride salt of Cd, were studied by both XRD and ED [4]. The films deposited us-
ing CdCl2 were highly textured (111) zincblende (with crystal size ca. 15 nm). Those
deposited from CdI2 showed sharp hexagonal reflections that were not highly tex-
tured. In addition, these sharp peaks rode on broad peaks, which, while not dis-
cussed, suggest that most of the film is made up of a much smaller crystal size, which
might be cubic. Using acetate or chloride anions, a well-defined peak [(probably
(111)] and some much smaller peaks were found for acetate and very weak (but nar-
row) peaks for chloride [22]. If ammonium salts were added (therefore lowering the
pH), well-defined and strongly textured (111) peaks were obtained for both anions.
       The presence of foreign ions can obviously affect the crystal structure. More
noticeable, however, was the effect of these ions on the crystallinity. Adding Cu (by
CuCl in the deposition solution) caused a decrease in the intensity of the (zincblende)
XRD peaks with increasing Cu concentration [23,24]. However, there was no ap-
parent change in peak width, implying that the crystal size did not change apprecia-
bly, since a reduction in crystal size—whether by reduction of the coherence length
of a fixed crystal size by defects or by actual change in crystal size—would result in
peak broadening. Similar results were obtained for films doped using CuCl2 and sub-
jected to annealing in air at 300°C [25]. Doping with a variety of cations (Cu, Ag, As,
In) was in all cases reported to result in loss of the XRD pattern [15].

4.1.2 Crystal Size
There are a number of factors that determine crystal size. Probably the two most
important are the deposition mechanism (ion-by-ion growth, in general, will result
in larger crystal size than the hydroxide mechanism, discussed in detail in Chap.
3) and specific adsorption of anions onto the growing crystal (this can affect both
crystal structure and size).
       In most cases, the CdS crystal size from the standard bath was typically
10–20 nm, although sometimes it could be several times larger than this, particu-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
larly from lower-pH (buffered) solutions. This is larger than the typical size of the
accompanying CdS precipitate, which tends to be between 5 and 10 nm. For ex-
ample, a crystal size of some tens of nanometers was deposited on a carbon-coated
TEM grid, but the precipitate in the same solution was 3–6 nm [3]. This was at-
tributed to an ion-by-ion mechanism for the film vs. a hydroxide cluster mecha-
nism in solution. Some large (typically 100 nm but reported up to 1 m) hexago-
nal-shaped thin crystals of Cd(OH)2 were also formed in the solution precipitate.
If the growth begins by a cluster mechanism but ion-by-ion growth (whether by
free sulphide or by complex decomposition) occurs in parallel, then an intermedi-
ate crystal size is a logical outcome, since the ion-by-ion growth can occur on the
small hydroxide-formed clusters, leading to crystal growth.
       The nature of the anion of the Cd salt was found to affect the crystal size in
some cases, although it appears that such effects are not universal but related to
other variables in the deposition process. In one report, the use of CdCl2 gave large
crystals (probably 100 nm), but with CdAc2 no XRD pattern was observed [5].
From the optical spectrum of the CdS deposited from an acetate bath, a crystal size
of ca. 5 nm can be inferred based on size quantization. Another study [22] found
fairly narrow XRD peaks (crystal size at least 20 nm) using CdAc2. CdS deposited
from a CdI2 solution gave an XRD pattern of sharp peaks (see the previous sec-
tion) on a broad background [4]. This, together with the blue-shifted optical spec-
trum, suggests that most of the film is made up of a much smaller crystal size. The
crystal size of films deposited using CdCl2 in the same study was ca. 15 nm (esti-
mated from the XRD pattern).
       Using a nitrilotriacetate (NTA) bath, the deposition mechanism could be
more easily controlled than from the ammonia bath; NTA is a much stronger com-
plexant than ammonia, allowing pure ion-by-ion deposition if the NTA:Cd ratio
is high enough. The crystal size from such an ion-by-ion deposition was 70 nm
(instrument broadening limited), while from a hydroxide-mediated NTA bath it
was 5 nm [10].
       A crystal size of ca. 10 nm was reported from an acidic thioacetamide bath
[8]. The only other acidic bath where crystal size could be extracted was the pho-
todeposition method using a thiosulphate solution, where, from the XRD, a size of
   10 nm could be estimated [26]
       If a comparison of crystal size and structure is made from this and the pre-
vious section, a general trend appears suggesting that crystals that grow in the
hexagonal modification are, in general (and there are exceptions), larger than
those that are cubic.

4.1.3 Optical Properties
4.1.3.1 Transmission/Absorption Spectroscopy
Many studies present optical absorption or transmission spectra of the resulting
films. (A reminder that a spectrophotometer measures transmission, not ab-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sorbance: The “absorbance” measured by a spectrophotometer is a mathematical
manipulation of the transmission and, if reflection is present, will not be accurate
without reflection correction. See Sec. 1.4 for details.) The purpose of these spec-
tra is usually to show that the deposits are of high quality (usually interpreted to
mean transparent in the subbandgap region; in most cases, scattering is undesir-
able, although there may be exceptions, depending on the intended application of
the films) and are indeed made of the material claimed (as seen from the bandgap
value, which can be estimated from these spectra). Scattering is usually caused by
optically large (comparable to the wavelength of the light) nonhomogeneous ag-
gregates; this often occurs by sedimentation of colloidal aggregates onto the sub-
strate. However, it can also occur even if no colloidal phase is present in the solu-
tion. There have been few studies on control of scattering in CD films. One study
reported more aggregates and lower transmission at lower deposition tempera-
tures from a citrate/ammonia bath [27]. However, another study, using a tri-
ethanolamine bath, reported more aggregates at higher deposition temperatures,
although not in a regular manner, resulting in generally lower transmission at
higher deposition temperatures [28]. Therefore, as is generally the case, such spe-
cific results should not automatically be applied to all CD CdS films. Application
of a magnetic field perpendicular to the substrate caused an increase in transmis-
sion of the film [29], although it is not clear whether this is due to increased spec-
ular reflectance of the field-free films or decreased scattering of the films de-
posited with the field. It may be indirectly inferred from the transmission spectra
that there is no less (maybe even more) scattering in the films deposited with a
field; but Atomic Force Microscope (AFM) morphology studies give the impres-
sion that these films are smoother, which would be one (but not a unique) indica-
tion of less scattering and greater specular reflectance.
       Most CD films reported are fairly transparent to very transparent, typically
between 60 and 90% transparent in the subbandgap region, although lower values
are not infrequently seen. Since the optical spectra are most often not corrected for
specular reflection, this reflection will reduce the transmission, but the “quality”
of such films can be very high. In fact, a high specular reflectance is indicative of
“good-quality” films, since films with a considerable degree of scattering exhibit
low specular reflectance.
       The other type of information that can be extracted from optical spectra via
the bandgap is an estimation of crystal size if the semiconductor is in the size-
quantized domain. This is due to the blue spectral shift caused by size quantiza-
tion: The smaller the crystal size, the larger the blue shift and the larger the
bandgap. This is discussed in more detail in Chapter 10. Here we note briefly some
studies where such shifts have been seen.
       In most cases, the standard bath gives a crystal size larger than the largest size
that will show an appreciable blue shift (for CdS, this value is ca. 6 nm). There are
some exceptions, however, with minor modifications of the deposition conditions.
The use of iodide [4] and acetate [5] as anions of the Cd salt resulted in blue shifts


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of 0.1 eV or more. Blue shifts of ca. 0.2 eV were measured in CdS films deposited
from citrate/ammonia baths [7,27].
       Transmission spectra can be modified by doping. Thus, if Cu (as CuCl) is
added to a triethanolamine/ammonia bath for CdS deposition, the effective
bandgap measured from the spectra shifted from 2.35 eV (no Cu) to just over 2.0
eV [23,24]. The less steep onset and more pronounced absorption tail of the
Cu:CdS, together with the very low value of bandgap, suggests that this large shift
is due either to a subbandgap transition arising from Cu impurity states in the CdS
or even to absorption in a separate phase of Cu-S, which would probably itself be
quantized. In a somewhat similar study, but with films annealed at 300°C in air, a
drop in bandgap due to Cu doping from 2.48 eV to 2.38 eV was measured [25]. In
contrast, another study reported an increase in bandgap from 2.4 to 2.48 eV upon
doping with either Cu, Ag, As, or In (as well as a less steep onset). In principle,
doping can affect the measured absorption spectrum in different ways. The most
obvious is introduction of levels in the gap (which would result in an apparent
lowering of the bandgap). Amorphization (as seen by the loss in the XRD pattern)
is commonly found to occur upon doping; the resulting disorder could cause tail-
ing of the states near the band edges. This tailing would normally be seen as a de-
crease in the bandgap, although an increase has been explained by splitting of the
tailed levels from the bands [15]. Increase in bandgap can also occur by filling the
lower-lying conduction (valence) band levels with electrons (holes), thereby re-
quiring a larger photon energy to promote an electron from the valence to the con-
duction band (the Burstein–Moss shift).
       Antimony doping has been shown to have strong effects on the optical spec-
tra [30]. The bandgap decreased from 2.47 eV (pure CdS) to 1.7 eV (nominally
0.075% Sb) and then increased to 2.86 eV for 0.1% Sb. The strong drop in
bandgap (the absorption was strong) for moderate Sb levels suggests an impurity
band, while the increase for higher doping levels could, by itself, be explained by
the Burstein–Moss shift. However, these explanations would require effective re-
moval of the impurity band at some intermediate Sb level, which would not nor-
mally be expected. The crystallinity of the films does not appear (from the XRD
spectra) to change appreciably with doping, thus removing amorphicity as a pos-
sible explanation for the effects.

4.1.3.2 Photoluminescence

Photoluminescence (PL) is light emitted when photogenerated electrons and holes
recombine. In that sense, it is the opposite of absorption. However, while an opti-
cal absorption spectrum in the great majority of cases shows the valence–conduc-
tion band transition (i.e., the bandgap), photoluminescence spectra are much more
complex as a rule. The bandgap emission (sometimes called band-to-band emis-
sion) may or may not occur (it often does not), and subbandgap transitions of
longer wavelength are commonly seen. These transitions are from various surface


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
or bulk states in the bandgap, and they therefore can give information about these
states, which are not seen, or at best are seen very weakly, in absorption spectra.
In the majority of cases, most of the energy of photogenerated electron–hole pairs
is not emitted in radiative transitions, but converted to heat in nonradiative transi-
tions (i.e., the luminescence efficiency is commonly low).
       It is outside the scope of this book to go into detail on the explanations for
the various PL spectra measured in CD CdS films; several examples are given
with possible origins for the various spectral peaks.
       The pH of deposition (adjusted by adding NH4OH, therefore pH increased
but free Cd2 decreased) affected the PL spectra of the CdS films deposited from
a standard solution [31]. A broad, red luminescence (ca. 1.2–2.0 eV with peak at
1.68 eV) was characteristic of all the spectra, regardless of deposition pH. At
pH 11.5, a narrow (0.18 eV half-width) green peak (2.255 eV) appeared, but it
did not occur above or below this pH value). This peak, ca. 0.2 eV less than the
bandgap, could be either a shallow-donor-to-shallow-acceptor transition or a
band-to-fairly shallow interband state transition.
       Different crystal sizes (some in the quantum size regime) were obtained by
varying the film thickness. Three interconnected PL peaks at ca. 1.83, 1.35, and 1.06
eV were obtained (no green emission) [32]. A model of transitions from a deep donor
(Cd–O complex) level to various other levels was suggested to explain these peaks.
       Another deposition, probably from a standard solution (although the details
of the deposition were not complete), gave a dominant peak in the green region
(2.38 eV) and a broad low-energy shoulder extending to ca. 1.5 eV [33]. Decon-
voluting the spectrum revealed, besides the green peak, a small yellow peak (2.25
eV), attributed to a Cd interstitial–Cd vacancy (iCd–VCd) complex and red band
(1.80 eV) associated with sulphur vacancies.
       PL spectra of CdS deposited from two different acidic baths have been re-
ported. From an acid thioacetamide bath, a broad band centered around ca. 1.5 eV
was obtained [8]. The most likely cause for this luminescence was suggested to be
valence band hole–S vacancy recombination. Films deposited under illumination
from a thiosulphate solution exhibited a broad band from ca. 1.46–2.0 eV (peak at
ca. 1.66 eV) [26].
       The wide range of different PL spectra obtained shows just how much the
various films vary from each other and the sensitivity of the (mainly surface) elec-
tronic structure of the CdS to the deposition parameters.

4.1.4 Resistivity and Photoconduction of
      As-Deposited CdS
4.1.4.1 Dark Resistivity
Note that only as-deposited films are considered here. As will be evident to any-
one familiar with semiconductor processing, annealing of these films can be car-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ried out to drastically change their electrical and photoconductive properties, de-
pending on the annealing conditions. While a considerable amount of work has
been expended on studying the various effects of annealing on CD CdS films, this
is mostly outside the scope of this book. (See, however, the end of Sec. 4.1.7,
which gives some information on the effects of rapid thermal annealing and sub-
sequent removal of oxygen on the electrical properties. Also, since it is relevant
for PbS(Se) photoconductors and for photovoltaic cell use, both of which usually
require some annealing, it will be treated somewhat in Chap. 5 and 9.) Two points
are worth noting here. One is that, in general, annealing in hydrogen or vacuum
invariably reduces the dark resistivity of CD CdS, while reannealing in air or oxy-
gen increases it again. Oxygen can chemisorb on the CdS surface, extracting elec-
trons from the conduction band, and therefore decrease the free-electron concen-
tration (hence increase in resistivity). The second point is that, for photovoltaic
cell use, a lower resistivity does not necessarily mean a better cell; in fact the op-
posite may even be true.
       The dark resistivity of CD CdS is often, although by no means always, very
high. This may be reasonably attributed to the high degree of stoichiometry usu-
ally obtained with CD films. This stoichiometry is certainly expected for ion-by-
ion growth, and is probable also for hydroxide-mediated growth (both simple and
complex) as long as all the hydroxide has been converted. It is likely that the cases
where low resistivity has been reported can be explained by nonstoichiometry. In
one study [34], the activation energy of the dark conductivity (measured at and
above room temperature) was found to be 1 V, from which it can be inferred that
the CdS Fermi level is very close to the bandgap center, meaning that the CdS is
highly intrinsic and free of common bulk defects, in particular S vacancies. Weak
n-type behavior with very low donor concentration (1012 cm 3) and considerably
higher deep trap densities (1015–1016 cm 3) were found on standard films de-
posited on quartz using space charge–limited current measurements [35].
       This high resistance may be responsible for the fact that the commonly used
Ag contacts to the CdS behave in an ohmic manner [36], although Ag is not nor-
mally considered to be a good ohmic contact to CdS in general. Two reasons can
be given for this. One is that the high resistivity of the CdS means that even an ap-
preciable contact resistance may be negligible. Another factor is that, since the
high-resistance CdS is often close to intrinsic (i.e., the Fermi level is close to the
bandgap center), which by definition means a higher value of work function, even
a high-work-function metal is less likely to form a Schottky (blocking) contact to
the CdS.
       This is a good point to bring up briefly a property of very small crystals (of-
ten obtained in CD), which is dealt with in more detail in Chapter 9. The crystal
size is, in most cases, much smaller than the size of any space charge layer that
would be formed. This means that in an isolated nanocrystal, unless the doping
level is very high (usually it is not, as attested to by the high resistivities more of-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ten obtained), there will not be much built-in electric field. The situation of a film
of aggregated nanocrystals is not so obvious, but it is likely that a space charge layer
of the normal type will not be obtained. This has important implications in consid-
ering contacts both between nanocrystal and metal contact and between nanocrys-
tals. As far as (photo)conducting properties are concerned, terms such as Schottky
contact or grain boundary barriers need to be considered with this point in mind.
       Values for the resistivities of various CD CdS films are given in Table 4.1.
The first thing that can be observed is the lack of any obvious correlation between
the resistivity and deposition conditions in most cases. Only some conditions are
given in the table; the details of the concentrations are not always available; even
if they are, it would be oversimplistic to try to compare them based only on con-
centration. For example, the ratio of Cd salt to complexant is no less (probably



TABLE 4.1 Dark Conductivity and Photoconductivity of CD CdS Films

                                                        Resistivity ( -cm)

Bath conditions                                      Dark                Light             Reference
                                                        9
CdCl2 90 C                                           10                   700                38
CdSO4 70 C                                           108                 1000                39
CdAc2 or CdCl2 80 C                                106–107            103–104                5
CdAc2                                              104–106             10 –1000              16
Cd(NO3)2 room temperature                            1012            Small effect            40
CdSO4 90 C                                           104                  —                  11
(80–95 C) pH 10–11                                 340–600                —                  22
(80–95 C) pH 9–10                                    15–150                                  22
CdCl2                                              106–108            102–104                4
CdI2                                                108–1010          103–105                4
CdAc2 NH 4 85 C                                    103–104               —                   41
Triethanolamine 26 C                                 109                1000                 28, 36
Triethanolamine 75 C                                 109                   1                 28, 36
Citrate/ammonia 60 C                                  5.107             2000                 27
Citrate/ammonia 75 C                                 108                   5                 27
Citrate/ammonia 50 C                               2 108                 200                 7
Citrate/ammonia 70 C                               2 108                   4                 7
Thioacetamide (pH 8) 40 C                            104                 —                   37
  Temp. of minimum resistance 63 C                   10                  —
  Temp. of minimum resistance 78 C                 400                   —
CdCl2 85 C No magnetic field                       2 105                 —                   42
CdCl2 85 C Magnetic field                            60                  —

The first column gives the Cd salt and deposition temperature. If not specifically noted, the bath is a
standard ammonia/thiourea bath.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
even more) important than the absolute concentrations themselves. Also, so many
parameters are interrelated.
       In an attempt to say something intelligent about these resistivities, there ap-
pears to be some correlation between the pH and resistivity, with low resistivity
obtained when the pH is relatively low (only a few experiments have been carried
out at relatively low values of pH; also note Ref. 22, which describes an anoma-
lously low resistivity even at “normal” values of pH). The bath described by Ito
and Shiraishi [37] is very different from the previous ones, for three reasons: the
relatively low pH ( 8), the use of thioacetamide instead of thiourea, and the flow
system used in this deposition. Very low values of dark resistivity were obtained
with this bath and with an unusual temperature dependence (a minimum of 10 -
cm was found at 63°C, which increased on either side of this temperature value).
It was suggested that Cl, from the NH4Cl buffer, acted as a dopant; however, other
chloride baths gave much higher resistivities.
       Some weak correlation between film morphology and resistivity was noted
for films, deposited from a pH 9.5 bath: Films deposited from a closed system (no
evaporation or loss of ammonia) were more uniform and had somewhat lower
(two to three times) resistivity (ca. 1.5 104 -cm) than films deposited from an
open bath [43].
       The triethanolamine bath gives consistently high resistivities, independent
of the deposition conditions, and the citrate/ammonia bath behaves similarly, al-
though with somewhat lower resistivities. It should be noted that chemisorption of
oxygen on the CdS is known to increase the resistivity, and some (many?) differ-
ences may well be due to different surface chemistries of the CdS crystals. Thus,
by definition the complexant can bind to surface Cd and cover the surface in some
cases. This can lead to (at least) two opposing effects: “insulation” of one crystal
from another and prevention of oxygen chemisorption (this latter need not, how-
ever, necessarily lead to lower resistivity but the opposite if the adsorbed complex
acts in a similar way to the oxygen). This surface adsorption may explain, e.g.,
both the high resistivity of the triethanolamine-bath films and their relative inde-
pendence from other deposition conditions. It is worth noting that most (although
not all) CD CdS films have a rather similar crystal size; therefore this factor does
not seem to be as important as might have been expected.
       Films deposited from a standard bath but under application of a magnetic
field perpendicular to the substrate exhibited resistivities that decreased strongly
as the strength of the magnetic field increased [42]. The resistivity, which for a
film deposited without the field was 2 105 -cm, was as low as 60 -cm on ap-
plication of a 77-mT magnetic field. From the dark resistivity/temperature values,
a level with an activation energy of 0.68 eV was found for the field-free films and
a shallow level of 0.053 eV and others at 0.17 eV for those deposited in the pres-
ence of the field. The shallow level at 0.053 eV was suggested to be due to excess
Cd. It was suggested that the magnetic field might affect the rate of arrival of cad-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mium and sulphide ions at the substrate and that more sulphur vacancies are
formed in the presence of the field, hence the lower resistivity (see Sec. 4.1.6.8 for
more details on this deposition).
4.1.4.2      Photoconductivity
CD CdS films are usually strongly photoconducting as deposited. This is in con-
trast to most other CD films, which tend to be only weakly or moderately photo-
conductive in the as-deposited state [44]. Values of the resistivity under illumina-
tion are given in Table 4.1 where available. Most experiments have used
illumination intensities comparable to solar irradiation.
       As is the case for the dark resistivity, the dependence of the sensitivity of the
photoconductivity (defined here as the ratio between light and dark conductivity)
on the deposition parameters is far from clear-cut. Some observations can be
made, however. The first (obvious) one is that for a high sensitivity, the dark re-
sistivity must be high. Apart from this, there does seem to be a general trend
(clear-cut in the triethanolamine and citrate baths and seen also by the lack of ap-
preciable photoconductivity in the one low- (room-) temperature-deposited film
reported [40]) of an increase in photosensitivity (due to decrease in light resistiv-
ity) with increasing deposition temperature.
       For the standard baths, the sensitivity varies in most cases between 103 and
   6
10 . A CdI2 bath resulted in somewhat greater sensitivity (as well as dark resis-
tivity) than a CdCl2 bath [4]. The deposition temperatures of these two baths were
different, but it was reported that the film properties were independent of the de-
position temperature.
       Another study found no appreciable difference in either dark or light resis-
tivity between acetate and chloride baths [5]. Interestingly, there was apparently a
large difference in crystal size between the two baths (see Sec. 4.1.2 on crystal
size), which implies that the crystal size is not an important factor in determining
the resistivity or photoconductivity, at least for this bath.
       The triethanolamine bath showed a distinct trend of the photosensitivity
with deposition temperature [28,34,36,45,46]. The photosensitivity was higher
(which, since the dark resistivity was temperature independent, means the light re-
sistivity was lower) for higher temperatures, with the major change occurring be-
tween 30 and 45°C.
       The citrate/ammonia bath has much in common with the triethanolamine/
ammonia bath with high light sensitivities, particularly at higher deposition tem-
perature, and little temperature dependence of dark resistivity. This is in spite of
the very different bath compositions and concentrations. In particular, the citrate
bath contained much lower concentrations of Cd and, as a result, was more highly
complexed.
       One study showed a very strong dependence of dark resistivity on measur-
ing temperature and a much weaker dependence of the resistivity under illumina-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tion, with the result that the sensitivity was very temperature dependent [15] (see
Sec. 4.1.4.3).
      The decay time of the photoconductivity, , is another important parameter
for which, as with the dark resistivity and photoconductivity, there is often no ob-
vious correllation with the deposition parameters. In many cases, the decay is very
slow (hours), particularly for the triethanolamine bath, where it can be greater than
10 hr. Decay times of hours have been reported for the standard bath [4,16], al-
though if CdI2 was used instead of CdCl2 in the latter, dropped to seconds (the
deposition temperature was different for the chloride and iodide bath, but report-
edly this did not affect the properties). For films deposited from the citrate bath,
temperature-dependent decay times from 1 min (deposition at 75°C) to tens of
minutes (60°C) were reported in one study [27] and tens of minutes with only a
small deposition temperature dependence in the other [7]. Differences in the de-
position conditions of these two studies were described earlier. From the tri-
ethanolamne bath, much longer decay times were observed at higher deposition
temperatures, as shown in Figure 4.1 [34]. Another study, using ammonium-
buffered standard films (possibly ion-by-ion deposition), found decay times of
seconds for single films, which increased to several minutes for multiply de-
posited films [47]. The photosensitivity of these latter films was less than for most
others (ca. 30 for the single films and 10 for the multiple ones).




FIG. 4.1 Time dependence of photoconductivity of CdS films deposited from tri-
ethanolamine/ammonia/thiourea bath at two different temperatures (26°C and 75°C). The
two plots at each temperature differ by the ratio between the Cd and thiourea concentra-
tions: [Thiourea]:[Cd] 0.25 for the upper plots at each temperature and 0.5 for the lower
plots. (Adapted from Ref. 34 with permission from Elsevier Science).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       It is important to note that oxygen adsorption has a strong effect on the pho-
toconductivity decay of the triethanolamine films [36]; it is probable that this is
not limited only to these films. Oxygen greatly decreases the decay lifetime, as
seen by the increase in for aged films compared with freshly deposited ones. In
this same study, the photoconductivity sensitivity and decay time both decreased
greatly with increase in measuring temperature (the former due mainly to the ac-
tivated decrease in dark resistivity with increase in temperature, but also due
partly to decrease in resistance under illumination). The possible effects of oxy-
gen on photoconducting parameters are discussed in this reference (general back-
ground on photoconductivity is given in Ref. 34). Here it is enough to note that ad-
sorbed oxygen is believed to extract electrons from the CdS conduction band and
also to introduce deep trapping centers (interband surface states) that increase the
carrier lifetimes, thereby increasing the photoconductivity decay time. Oxygen
can also affect electron mobility between grains by modification of grain bound-
ary barriers between crystals.
       A comparison was made between films deposited from standard baths using
either CdCl2 or CdAc2 [48]. While for most measurements these films were an-
nealed at 300°C, and therefore are not compared with the as-deposited films here,
thermally stimulated current (TSC) measurements were carried out on as-
deposited films. Such TSC provides an indication of the density and energy of
trapping centers: The magnitude of the current, obtained by heating the sample
and exciting charges out of traps into a band, is an indication of the trap density
and the temperature at which the current is generated is a measure of the trap en-
ergy. The trap density was much higher and the traps considerably deeper for the
acetate-prepared films than for the chloride ones (after annealing, the trap density
was higher for the chloride films, seen as a large increase in TSC for the chloride
film after heat treatment).
       We can conclude this section with the insight, gained from this overview of
the electrical and photoconductivity properties of these films, that, in spite of the
many studies already carreid out, a comprehensive and systematic study of these
properties and their correlation with a wide range of deposition parameters is still
needed in order to understand what determines these properties. These studies
should also include postdeposition treatments—not so much annealing, which has
been carried out, but surface treatments (e.g., immersion in triethanolamine),
which could show the importance (or lack of it) of the crystal surface condition.
4.1.4.3      Electrical Properties of Doped CdS Films
Doping can be divided into two parts: native doping (e.g., S vacancies) and ex-
trinsic doping by foreign elements. This section deals with the latter, not because
it is more important but because there is little in the literature to link native dop-
ing with the electrical properties of CD films. It will be enough to note that the few
measurements of ND (donor density) carried out tend to give values typically



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
                                                                    ¯¯¯
around 1016 cm 3 (e.g., Ref. 49 for epitaxial, hexagonal CdS on (1 1 1) InP from
C–V measurements) or even larger. This is somewhat surprising, since, if these
films are so highly stoichiometric as expected, lower values might be expected.
Two comments here. One is that the doping may originate from the surface, since
the surface-to-volume ratio of the CdS crystallites is high. The other, also related
to the high surface area, is that errors in interpreting C–V measurements com-
monly used to derive ND may arise because of the lack of knowledge of the true
surface area.
      We will now consider individual dopants used in CD CdS.
       Boron. Boron, substituted for Cd, is a donor in CdS. B-doped films were
deposited by adding boric acid to a standard deposition bath, with the B:Cd ratio
varying from 10 5 to 10 2 [16]. The boron was assumed to occur in the form of
borate ions (BO 2 ). The dark resistivity dropped nearly three orders of magnitude
                  3
with optimum B content (B:Cd ratio in solution of 0.001), from 2.104 to 30 -cm.
At higher B concentrations, the dark resistivity again increased until, at a B:Cd ra-
tio of 0.01, the original resistivity of the undoped CdS was regained, and it did not
change with increased B content. However, the resistivity under illumination de-
creased to ca. 3 -cm, almost independent of the B content. Thus high B content
increased the photosensitivity of the CdS, although only by a factor of 2–3. The
rate of decay of the photocurrent was greatly reduced by B doping, from ca. one
hour for undoped films to as much as several tens of hours for doped ones. This
suggests a deep trap resulting from the B, separate from the shallow donor that is
responsible for the drop in resistivity.
      Nitrogen. Nitrogen ions (N with energy of 130 KeV) were implanted
into CdS deposited from a triethanolamine/ammonia bath [20]. The resistivity of
the as-deposited films was ca. 108 -cm and dropped, depending on ion dose, up
to seven orders of magnitude for an ion dose of 1017 ion/cm2. Even more notable,
the conductivity type changed from n-type (the normal type for CdS) to p-type, as
measured by hot probe. An acceptor level, 0.6 eV above the valence band, was in-
troduced by the ion implantation.
       Copper. Copper was doped into triethanolamine/ammonia films by
adding CuCl in the deposition solution [23,24]. Resistivity dropped from 109 -
cm (undoped) to a minimum of 0.5 -cm for optimum doping (the Cu content of
the CdS was not measured) and the conductivity was p-type. Both higher and
lower Cu concentrations in solution gave higher resistivities; it was surmised that
too high concentrations of Cu in the deposition solution resulted in rapid precipi-
tation of the Cu as CuxS, depleting the solution of Cu. While the high-conductiv-
ity p-type CdS was, not surprisingly, not photoconductive, films with smaller
amounts of Cu were photoconductive (although the photoconductivity gain was
less than for nondoped CdS, and the response time of the photoconduction, both



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
rise and fall, was much shorter than for the undoped film). The Cu-induced states
resulted in recombination centers rather than long-lived trap states characteristic
of CD CdS films. Another study of Cu-doped CdS films reported different results,
although there were differences in the preparation; in particular the films were an-
nealed in air at 300°C, and CuCl2 (rather than CuCl) was used [25]. In this work,
dark resistivity did not vary greatly for low Cu concentrations (ca. 107 -cm), but
increased by nearly two orders of magnitude for high Cu concentration. Also, the
photoconductivity response increased with Cu content. The annealing carried out
in this study calls for caution in any comparison with the previous one. A third
study (no variation in Cu content) found a decrease in dark resistance (from ca.
108 to ca. 104 -cm upon Cu (also Ag, As, and In) doping [15]. The dark resis-
tivity was very highly temperature dependent, especially for the undoped CdS (ca.
106 -cm at 35°C and 104 -cm at 50°C). Since the resistance under illumination
( 105 -cm for undoped and 104 -cm for doped) was much less temperature
sensitive, the photoconductivity response was very temperature dependent, being
more pronounced at lower temperatures.
       Lithium. Lithium can act as an acceptor in CdS. Shikalgar and Pawar stud-
ied electrical [50] and photoconducting [51] properties of Li-doped CdS [standard
deposition with 0.1% (by weight of CdS) Li salt added to the deposition solution].
The addition of Li increased the dark resistivity by a factor of 3–4 (resistivities
were given as ohms and the exact geometry of the measuring system was not de-
scribed: however, specific resistivities could be estimated to be of the order of 107
   -cm). The room-temperature energy of activation for both doped and undoped
films was ca. 22 meV, i.e., shallow donor conduction. Above ca. 60°C, the resis-
tivity dropped much more rapidly as a function of temperature, with an activation
energy of ca. 1.2 eV (intrinsic conductivity). In addition, the Li-doped film ex-
hibited an intermediate level at 0.16 eV in the temperature range of ca. 40–65°C,
ascribed to a Li acceptor level. The Li-doped films, like the undoped ones, were
n-type; CdS is difficult to dope p-type due to self-compensation, and since the re-
sistivity of CD CdS films are normally very high, it is not surprising that acceptor
doping does not increase this resistivity very greatly.
       The light:dark conductivity ratio of these Li-doped films was not explicitly
given, although an order of magnitude value of 106 could be inferred from the re-
sults. The photocurrent–time behavior for the CdS:Li (the equivalent data for the
undoped films were not given) was history dependent. Initially, the photocurrent
increased linearly with time (over a maximum measured time of 10 min), but in-
creased more rapidly and exponentially with larger photocurrents after light–dark
cycling. In all cases, the decay was multiexponential and slow, typically tens of
minutes. These measurements were carried out in vacuum; if air was introduced,
the steady-state photocurrent decreased, attributed to oxygen adsorption on the
surface of the CdS crystals, resulting in extraction of electrons from the conduc-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tion band. Interestingly, the peak of the photocurrent of the CdS:Li was blue-
shifted (480 nm) compared to the undoped CdS (520 nm); no explanation could
be given for this effect.
       Aluminum. Aluminum, as a trivalent ion, should be an n-type dopant for
CdS. A small decrease in resistivity (by a factor of 2) to ca. 103 -cm was found
when Al2(SO4)3 was added to a standard bath [52]. In another study, Al was added
as Al2(SO4)3 to a thiourea bath (85°C) of relatively low pH (9.5) [53]. The resis-
tivity of the undoped CdS was ca. 3 105 -cm and decreased at least an order
of magnitude on doping with Al. Codoping with chlorine (as NH4Cl in the solu-
tion) decreased the resistivity almost another order of magnitude. In both studies,
an excess of Al resulted in an increase of resistivity. This was explained by excess
Al3 occupying interstitial positions: however, interstitial Al3 might be expected
to increase the n-type doping and therefore decrease the resistivity, and it is not
clear why the resistivity should increase. It may be that an insulating Al(OH)3
phase occurs if too much Al is added.

4.1.5 Substrate-Dependent Growth and Epitaxy
4.1.5.1      Introduction
There are a number of studies that report the effects of the substrate on the CdS
films. With the exception of epitaxial deposition, which will constitute the main
part of this section, it is usually difficult to explain any specific substrate effect.
Also, it should be borne in mind that each specific study is confined to one depo-
sition bath and that a substrate effect obtained for one bath need not necessarily be
obtained for a different one.
       Some examples of substrate effects on the film can be given. Strong (0002)
or (111) texturing was obtained on glass substrates but much weaker texturing on
SnO2 /glass [16]. Much poorer crystallinity (this may also mean smaller crystal
size) of the CdS was obtained on Si than on glass or ITO/glass [54]. Using XRD
peak shifts and optical absorption spectroscopy, the presence of strain in as-grown
CdS on both glass and ITO/glass was inferred [55]. The strain was greater for the
films deposited on the ITO, and this was attributed to mismatch strain between the
CdS and ITO.
4.1.5.2      Epitaxy
Various investigations into the epitaxial deposition of CdS onto different single-
crystal substrates have been carried out by Lincot et al. On InP, which is closely
lattice matched to CdS ( 0.1% difference), epitaxial deposition (c-axis of hexag-
                                                          ¯¯¯
onal CdS perpendicular to the substrate) occurs on the (111) P polar face of the InP
but polycrystalline deposition on the (111) In face [49,56]. This difference was
clearly due to differing chemical or electrostatic interaction between the InP faces



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
and the constituents of the deposition solution, since the lattice spacing is the same
for both polar faces. The degree of epitaxy on the P face was also dependent on
the deposition conditions: in particular, higher temperatures resulted in better epi-
taxy, as might be expected due to the higher mobility of the depositing species on
the surface. The epitaxy was maintained up to at least 100 nm film thickness (the
maximum thickness studied).
       On (100) etched (Br-MeOH) InP, cubic, fairly well-oriented CdS was
formed, although with many small-angle grain boundaries. In the absence of the
etch treatment (using only H2SO4 to remove native oxide, as was also employed
following the Br-MeOH treatment), only polycrystalline CdS was deposited,
which showed that not only the crystal face, but also the manner in which that face
was pretreated, is important [57]. The formation of the cubic phase in contrast to
                                 ¯¯¯
the hexagonal phase formed on(111) InP was attributed to the lack of lattice match
between the (cubic) (100) face of InP and hexagonal CdS.
       GaP has a much larger mismatch with CdS ( 7%) compared with InP. Yet
                                                               ¯¯¯
a fair degree of epitaxy was obtained for CD CdS on the (111) GaP surface [58].
In this case, a mixture of cubic and hexagonal CdS with a large density of stack-
ing faults, presumed due to strain relaxation arising from the large mismatch, was
obtained.
       Because of the importance of the junction between CD CdS and CuInSe2
(CIS) for thin-film photovoltaic cells (see Chap. 9), as well as the relatively small
mismatch between CdS and CIS ( 0.7%), deposition onto oriented CIS films has
also been studied [59,60]. Two different CIS faces were studied—(100) and (112).
As with the (100) P face of InP, because there was no lattice match between this
face and any hexagonal CdS face, cubic, epitaxial CdS was deposited. On (112)
CIS, which matches either (111) cubic or (0001) hexagonal CdS, a mixture of both
phases was deposited, with a moderate degree of epitaxy, which improved if the
CIS was first subjected to a cyanide treatment (cyanide removes excess CuxSe and
various Se species and generally cleans up the surface). The epitaxy also improved
with increase in temperature. The transition temperature was quite abrupt: Below
60°C, the films were polycrystalline, while above this temperature they were epi-
taxial, with increasing perfection as the temperature increased. Since the deposi-
tion at lower temperatures was much slower than at higher ones (for a temperature
difference of ca. 40°C, the deposition rate increased 30 times), this implied that
the increased mobility of the depositing species on the surface was not necessar-
ily the main factor in the temperature dependence of epitaxy. It was suggested that
faster and more complete decomposition of reaction intermediates was an impor-
tant factor in determining the epitaxy. If so, this is presumably true for epitaxial
deposition on other substrates.
       The composition of the deposition solution was important in order to obtain
epitaxy on CIS, which, in contrast to the epitaxy on InP, was not obtained using
“standard” solutions. Instead, low Cd concentrations (maybe more important, low



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Cd:NH3 ratios) and lower than usual pH (adjusted by adding ammonium ions)
were necessary. These are the factors that favor an ion-by-ion mechanism over
one involving Cd(OH)2, which may explain the need for this modified solution
(most commonly used solutions operate in or near the region where Cd(OH)2 can
exist, at least on the substrate). This explanation still leaves open the question of
why good epitaxial growth can be obtained on InP under some conditions but not
CIS from a “standard” solution. While there is no present answer to this question,
considering the sensitivity of the epitaxy to the chemical properties of the surface
(such as etch or differences between P and In polar faces of InP), this should not
be too surprising.
       The epitaxy was maintained for CdS thicknesses up to 100 nm, after which
the deposit became polycrystalline. This transition coincided with the visual for-
mation of CdS in the solution, which resulted in a switch of the mechanism from
an ion-by-ion growth, necessary to obtain epitaxy, to one involving colloidal
species. Since, in principle, conditions can be chosen so that only an ion-by-ion
growth occurs, it can be expected that much thicker epitaxial films are obtainable
from CD on suitable substrates.
4.1.5.3 Deposition on Monolayers: Selective
Growth and Patterning
If a substrate is not clean, films either do not grow or grow with poor adherence
on the “dirty” parts of the substrate. This has been exploited by partially covering
the substrate with a monolayer. When a mica substrate was incompletely covered
by a monolayer of octadecylphosphonic acid, CdS growth was found to occur
preferentially on the mica [9]. This was shown also for CdS deposited on an oc-
tadecyltrichlorosilane- (OTS)-coated Si substrate and was used to pattern the CdS
deposit by applying the OTS onto the Si using a patterned stamp [61]. Either the
OTS could be removed by sonication or, even without removal of the OTS, depo-
sition occurred only on the bare Si if the CdS was not too thick (ca. 50 nm). Edges
with a variation of ca. 100 nm could be deposited by this method.

4.1.6 Variations in Preparation
4.1.6.1      Variation in pH
While deposition rate normally increases with increase in pH for the standard
bath, using an ammonium salt to lower the solution pH resulted in the opposite be-
havior; i.e., increased pH led to slower deposition [22]. The pH in these experi-
ments was increased by adding NH4OH. Increased pH (and NH4OH) results in
two opposing effects: Thiourea decomposition increases but free [Cd2 ] de-
creases. Since the deposition rate for the solution with no added NH 4 increases
with increase in pH, the former apparently outweighs the latter. When extra am-
monium ion is added, much more ammonium hydroxide is needed to increase the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
pH to the original value, and the decrease in free [Cd2 ] dominates the reaction
kinetics. More generally, addition of an ammonium salt (apart from the hydrox-
ide, which increases the pH) increases the complexation of Cd without increasing
the pH (it actually decreases the pH of an NH4OH solution due to buffer action).
This results in a decrease in deposition rate due to a lower free-Cd2 concentra-
tion and, if the pH decreases, reduction in the thiourea decomposition rate. A
slower deposition allows the formation of thicker films, since less CdS will be
formed homogeneously in the solution. This has been shown in many studies (e.g.,
Refs. 22, 40, and 43). In addition to slower thiourea decomposition, lower pH will
also decrease the likelihood of Cd(OH)2 formation and will therefore favor (rela-
tively) an ion-by-ion rather than cluster mechanism.
      Only a few acidic baths have been described (see also sections 4.1.6.6 and
4.1.6.7 below). In one, thiosulphate was used at a pH of between 2 and 4 and a tem-
perature of 85°C [131]. The bandgap was 2.35 eV and the resistivity 104–105 -cm.
Thioacetamine has also been used at a pH of 5 [8]. The films from this bath were
clearly hexagonal. The rationale for using an acid bath is to prevent the formation
of hydroxy species; this is a major problem for ZnS but much less so for CdS.
4.1.6.2      Variation in Complexant
Cyanide, a stronger complexant than others used, has been employed as a com-
plex for CdS deposition [62]. Except for the fact that thicker films could be ob-
tained (ca. 1 m compared to a few hundred nanometers with the standard
method), the properties of the films made with cyanide [crystal structure, crystal
size, bandgap (measured to be 2.2 eV, an anomalously low value but the same as
that of films deposited from a standard bath in the same study)] were the same as
those of the standard bath. Solution composition details were not given.
       Films up to 3 m thick were obtained with the triethanolamine/ammonia
bath [19]. It is probable that this larger-than-normal thickness is due to deposition
occurring via an ion-by-ion mechanism, due to the additional complexing by the
triethanolamine, the somewhat lower pH than usual (10), and, for the 3- m film,
the low deposition temperature employed (30°C), factors that reduce free-Cd2
and/or OH concentrations, thereby favoring the ion-by-ion mechanism. This
would result in thicker films since no (or at least less) CdS is wasted as a homo-
geneous precipitate in the solution.
       Ethylenediamine has been used as a complexant [6]. It is a stronger com-
plexant than ammonia and therefore only needs to be used in low concentration
(between two and four times that of Cd).
4.1.6.3      Variation in Thiourea Concentration
Increase in the concentration of thiourea clearly leads to an increase in deposition
rate. Additionally, it has been seen that the defect density (measured from TEM
micrographs as structural defects such as stacking faults) decreased greatly with



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
an increase in thiourea concentration from 1011–1012cm 2 for a thiourea concen-
tration of 28 mM to 1010cm 2 for a concentration of 100 mM [12].
4.1.6.4      Anion Effects
Some effects of anions have been noted previously. These are often not consistent
and, in general, it is difficult to attribute the effect of the anion of the Cd salt to any
specific effect with any degree of confidence. Two studies on the effect of the anion
on the rate of deposition did find small but significant differences (a factor of 2),
which more or less were in agreement [63,64]. The latter found the rate to increase
in the order: CdI2, CdSO4, Cd(NO3)2, Cd(CH3COO)2, and CdCl2; this series corre-
sponds approximately with the decrease in (negative) electrode potential and corre-
spondingly to the decreased strength of complexation between Cd2 and the re-
spective anion. Additionally, the rate decreased with increase in concentration of the
anion (added as an alkali metal salt). These observations suggest that the effect may
be due to mild additional complexation compared to that of ammonia alone.
4.1.6.5      Surfactants
The addition of surfactants to the standard CdS bath resulted in a reduction in the
rate of deposition and an increase in the terminal thickness [65]. Surfactants adsorb
onto surfaces (both the substrate and colloidal particles in the solution), and there-
fore it is not surprising that the growth rate is reduced. The adsorption of the sur-
factant onto CdS colloidal particles also can prevent flocculation and precipitation,
thereby increasing the CdS available for deposition (hence, presumably, the in-
creased terminal thickness). At the same time, it is possible that the presence of sur-
factants (or any strongly adsorbed species) might prevent sticking of the colloidal
particles to the substrate and to each other, in the same way as they prevent floc-
culation, which is exactly the sticking together of the colloids. However, there is
no evidence that this actually occurs in this study. Another effect of the surfactant,
inferred from the slightly higher than usual bandgap (2.52 eV), is the small size of
the CdS crystals; although not measured, it can be assumed to be ca. 5 nm from the
bandgap shift, presumably due to size quantization (see Chap. 10). This is not sur-
prising since crystal growth is in competition with adsorption of the surfactant.
4.1.6.6      Electrochemical/Chemical Deposition
Yamaguchi et al. described an interesting extension of the CD process for CdS us-
ing a parallel electrochemical step [66]. They termed this process electrochemi-
cally induced chemical deposition. It is based on electroreduction of protons in so-
lution, which results in an increase in pH locally at the electrode. They used
thioacetamide as a sulphur source. In the acid solutions in which the deposition is
carried out (pH between 1.6 and 4.6), no film deposition of CdS occurs (although
it does precipitate in the solution) in the absence of the electrochemical proton re-
duction. In the presence of proton reduction, CdS films were formed. These films



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
were characterized by relatively large, hexagonally faceted wurtzite CdS crystals
(crystal sizes from a few tens of nanometers up to 300 nm, the larger sizes being
formed at lower pH). The CdS precipitated in solution was quite different; it was
not faceted and had smaller crystal size (ca. 15 nm). It was suggested that the film
growth proceeded by a surface-catalyzed decomposition of a Cd–thioacetamide
complex and that the electrochemical proton reduction affected the surface prop-
erties (presumably the surface of both substrate and growing CdS) in such a way
as to reduce the activation energy needed for the deposition reaction. The growth
was an atom-by-atom (or ion-by-ion) process, leading to larger crystal size than
normally obtained by the hydroxide-mediated particle growth.
       By adding a strongly adsorbing species (2-mercaptoethanol) to such a de-
position bath, they were able to reduce the crystal size by varying amounts due to
surface capping of the growing crystals, preventing further crystal growth but al-
lowing nucleation to proceed. Thus, the film thickness was not strongly affected
provided the mercaptoethanol concentration was low ( 10 mM); above this con-
centration, film growth was prevented, as would be expected, since adhesion be-
tween crystals and substrate or between different crystals would probably be poor
in the presence of an adsorbed coating [67]. The resulting nanocrystalline films
exhibited quantum size effects (see Chap. 10 for more details).
4.1.6.7      Illumination-Induced Growth
There have been a few experiments related to the effect of illumination of the
growth of CdS films. Simple heating of the deposition bath by absorption of the
radiation is one obvious factor that can affect the deposition [68]. However, even
in this case, other effects occur, since the color of the bath was reported to darken
if UV (sunlight) illumination was employed. Based on previous studies of illumi-
nated CdS colloids when elemental Cd was formed, both as a film and in solution
[69], as well as the known tendency of ZnS to undergo reduction to metallic Zn
under UV illumination, this darkening may be assumed to be caused by elemen-
tal Cd. There are several possible mechanisms that may explain such an effect; re-
duction of the CdS by photogenerated electrons is one possibility.
       A variant of CD was based on illumination of a solution containing thiosul-
phate and cadmium ions by UV light [26,70,71]. CdS was deposited only on the
illuminated portion of the substrate. Since only light absorbed by thiosulphate
(wavelength shorter than 300 nm) was effective, the effect was attributed to pho-
todecomposition of thiosulphate to elemental S and solvated electrons and subse-
quent reaction with Cd2 .
               S2O 2
                   3     h   D S SO 23                                         (4.1)
              2S2O 2
                   3     h   → S4O 6
                                   2
                                       2e                                      (4.2)
         2
      SO 3     S2O 2
                   3     h   → S3O 6
                                   2
                                       2e                                      (4.3)
           Cd2
                  S     2e   → CdS                                             (4.4)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
The optimal pH was ca. 3.5 (at lower values, the film was contaminated with ele-
mental S, which forms spontaneously in the dark), while the deposition rate
slowed down at higher values). The CdS was sphalerite, and, from an examination
of the XRD, a crystal size of 10 nm could be estimated. The bandgap was 2.42
eV, the literature value for CdS.
4.1.6.8 Deposition Under the Influence of an External
Magnetic Field
Deposition has been carried out from a standard bath with a magnetic field ap-
plied, both parallel and perpendicular to the substrate [29]. Differences were
found in the film properties only for a field perpendicular to the substrate. The
transmission of the films in the nonabsorbing region was ca. 10% higher (see Sec.
4.1.3.1). The films deposited with the field were up to three times thicker than
those deposited using the same conditions but in the absence of a field. The dark
resistivity of the field-applied films was much less than that of the field-free ones
(see Sec. 4.1.4.1). The cause of these effects is not clear.
       Similar measurements were carried out using an external electric field [72].
Some differences in morphology and optical properties were measured, depend-
ing on the direction of the field with respect to the substrate. It is not clear, from
the experimental setup, why the field should influence the deposition, since the
field is external and should drop across the air and the glass walls of the reaction
vessel.
4.1.6.9      Deposition on/in Porous Silicon
Porous silicon is under extensive study, largely due to its luminescence properties.
For electroluminescence, however, some form of contact has to be made with the
Si, and this necessitates deposition of another phase inside the pores of the Si in
order to contact as much as possible of the internal area of the high-surface-area
Si. With this in mind, CdS has been deposited inside the pores of porous silicon
via a two-stage method [73]. Cd(OH)2 was deposited from an ammoniacal bath at
pH 8, followed by conversion of the Cd(OH)2 to CdS by treatment with thioac-
etamide at pH 8. This was repeated several times until the pores were essentially
filled with CdS. The reason that this two-stage process was needed is that either
the Si was unstable at the temperatures and pH values needed to deposit CdS from
a thiourea solution, or CdS was formed in solution rather than on the Si surface us-
ing thioacetamide.
4.1.6.10       Bath Geometry
One of the disadvantages of the CD process as usually carried out is the large
waste of materials (for example, in CdS deposition, most of the Cd—often over
90%—is unused in the film deposition because it deposits homogeneously in so-
lution and/or on the walls of the reaction vessel). Probably more important than



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the material loss is the environmental concern of disposing of this Cd (or other
heavy metal) if it is not recycled. The same goes for the ammonia often used as a
complexant—much is lost to the atmosphere and, in common with other com-
plexants, much is wasted to tie up the heavy metal ions.
       Use of low concentrations of metal ion (ca. 1 mM) presents a partial solu-
tion to this problem. However, for any industrial process, a continuous-flow sys-
tem seems the best option. Ito and Shiraishi flowed a solution of thioacetamide
and a Cd salt into a 0.5-mm-thick flow space [37]. A detailed flow system has
been described for CdS by Boyle et al. [74]. There are several features in this sys-
tem, shown schematically in Figure 4.2. Probably the most important is the locally
heated substrate. Since CD reactions are usually very temperature dependent, by
heating only the substrate (in this case, by resistive heating), deposition is limited,
to a large extent, to the substrate. This system also uses ethylenediamine instead
of ammonia, which greatly decreases loss by evaporation as occurs with ammonia
in an open system. Filters are employed in the flow system to remove any colloidal
matter formed. Fresh reagents can be added, as required, to the recirculating
closed-loop system. In connection with the foregoing studies, a batch process,
whereby the solution was filtered after deposition and complete reaction and
reused (with the addition of more reactants as required), was also shown to be fea-
sible [75]. While the deposition rate slowed down for successive depositions, this
could be compensated for by increasing the concentrations of various reactants
from run to run. The photovoltaic parameters of Cu(In,Ga)Se2/CdS solar cells fab-
ricated using this approach was not found to vary from deposition to deposition.




FIG. 4.2 Flow system for continuous deposition using a locally heated substrate. (After
Ref. 74 with permission from Elsevier Science.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       By minimizing the spacing between substrates, reagent utilization can be
maximized by increasing the ratio of substrate surface to solution volume [27].
The maximum film thickness was reduced if the substrate spacing was too close—
depending on the deposition parameters, and this was related to a “critical thick-
ness” of reagent layer at the substrate surface connected with the presence of col-
loidal particles in the solution. Typically some 1- to 10-mm spacing was necessary
to obtain the maximum thickness (200–400 nm in this case). This study was car-
ried out under conditions where a hydroxide cluster mechanism was operative (the
solution was already turbid in the early stages of deposition), and the results can-
not be extrapolated to other mechanisms.

4.1.7 Impurities in Chemical Deposition CdS
There have been a number of studies involving impurities in the CdS films, with
various results. It must be emphasized that if the films are not very well rinsed af-
ter preparation (and possibly even if they are), some of the ions involved in the
preparation may be present as adsorbed species. The most comprehensive study,
involving a range of analytical techniques, sums up the probable impurity situa-
tion for films deposited from standard baths [76].
       The main impurity, not unexpectedly, is oxygen (ca. 11 atomic %). Evi-
dence was presented to show that this O was probably mainly in the forms of car-
bonate and adsorbed water. The carbonate could come from two sources: dissolu-
tion of atmospheric CO2 and (see Eq. (3.11)) from decomposition of thiourea.
       Nitrogen (ca. 5 at. %) occurs as carbon–nitrogen bonds, probably mainly
cyanamide (NCN2 ), although other C–N bonded compounds were also believed
to be present. If cyanamide is present as the Cd salt, this would tie up 5% of the
Cd. The Cd:S ratio was found to be only slightly higher than unity (ca. 1.02), and
some of the Cd may be bound to carbonate. Therefore other C–N species are likely
also to be present, e.g., cyanide, several of which could adsorb to one Cd or even
to a CdS moiety. By reducing the concentration of thiourea in the bath, C–N im-
purities in the CdS film could be reduced almost to zero [77]. Whatever the nature
of the C–N impurity, much of it could be removed simply by dissolution in water
at 60°C [78]
       It should be pointed out that this deposition was carried out for films ca. 50
nm thick; the study was carried out with CdS window layers for solar cells in
mind, which are usually thin. It is possible that much longer depositions result in
different impurities. Thus the sparingly soluble cadmium carbonate and
cyanamide will be converted to CdS if enough sulphide ion is formed with time
(or, for the complex-decomposition mechanism, if enough adsorbed thiourea de-
composes on the surface of the solid phases). Of course, longer time also means
more thiourea decomposition products.
       Another study found much smaller concentrations of oxygen in the films
( 4 at.%) [79], and most of this oxygen was attributed to bound water. Although



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the deposition rates were comparable for this study and that in Ref. 76, the depo-
sition solutions and conditions were quite different.
       A decrease in the O content, measured by XPS, on Ar ion sputtering to-
gether with a Cd:S ratio close to unity led Danaher et al. to propose that the O (not
quantified) was present as surface sulphate [38]. In this same investigation, SIMS
analysis (much more sensitive than XPS, which is limited to ca. 0.1% concentra-
tion) found a variety of impurities, including decomposition products of thiourea,
CdO, and Cd(OH)2, but these were not seen in XPS, showing that they were pre-
sent in very low concentration.
       It is worth noting that analysis of the deposition solution after deposition
was complete, and after filtration of the solid precipitate showed the presence of
urea and guanidine, but not cyanamide, and that the amounts of these compounds
were less than those stoichiometrically expected, suggesting further decomposi-
tion of urea to ammonia and carbonate [75] (see Sec. 3.2.1.1).
       Films deposited from chloride and iodide (otherwise standard) baths were
compared [80]. I (ca. 3 at.%) was found in the iodide-deposited films but 1% Cl
in the chloride-deposited ones. About 5 at.% O was also found in both films. The
excess Cd was believed to occur as Cd–O, Cd–OH, and, additionally in the iodide
films, Cd–I or Cd–(I–O) species.
       In most cases, the Cd:S ratio in these films was slightly greater than unity
(usually between 1.02 and 1.1). A ratio of less than unity (0.92) was found for
multiple layers (i.e., two or more layers deposited one on the other [47]; for a
single layer using the same deposition solution, the ratio was unity). The oxy-
gen concentration varied from 8% (from an iodide bath) to 10–12% (from a
chloride or sulphate bath). As in the previous study, the oxygen was believed to
be present mainly as Cd(OH)2 or CdO. Even larger concentrations of oxygen
were found at low concentrations of ammonia (up to 18%) or at lower deposi-
tion temperature. Another (XPS) study, however, found the ratio to be typically
1.3 [81],
       Some general conclusions can be drawn concerning oxygen in the films, in
spite of the large spread reported in different studies, both in amounts and in in-
terpretation of its source. The first thing to note is that, since the crystal size of the
CdS is often in the region of 10 nm, around 10% of all the atoms will be located
at a crystal surface. Thus adsorption of either oxygen or water will already show
a relatively large amount of oxygen. However, unless this oxygen substitutes for
sulphur, such adsorption will not change the Cd:S ratio. Other sources of oxygen,
such as hydroxide, carbonate, and oxide (the last is less likely) will increase this
ratio. As discussed earlier, carbonate can form from either dissolution of atmo-
spheric CO2 or from the decomposition products of thiourea. Cd(OH)2 is more
likely to be formed when the pH is high (unbuffered solutions) or the ammonia
concentration is low (less complexation, which probably outweights the slightly
lower pH). The effect of temperature on Cd(OH)2 formation is complicated. A



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
higher temperature means a higher hydroxide concentration at a constant pH and
less effective complexation on one hand, but also faster decomposition of thiourea
resulting in more efficient conversion of the hydroxide to sulphide on the other.
From the results of Nakada et al. [81] discussed earlier, where more oxygen is
found at lower temperatures, it appears that the latter effect is dominant.
       Rapid thermal annealing (RTA) in vacuum of CD CdS films has been
shown to remove most of the oxygen that occurs in these films [82]. Typical an-
nealing conditions were: heating rate—100°C/min; maximum temperature—
600°C for 1 min; cooling rate—50°C/min. X-ray photoelectron spectroscopy
(XPS) showed that, apart from the immediate surface, oxygen was effectively re-
moved from CdS films (deposited from a thiourea bath). Dark resistivity was dras-
tically reduced after this treatment, from ca. 107 -cm for the as-deposited film to
ca. 1–10 -cm after the RTA treatment. As can be expected, the photoconduction
sensitivity also decreased drastically, from a light:dark resistivity ratio of ca. 104
to ca. 1.6 after annealing. This decrease in resistivity was attributed to removal of
electron traps that originated from the adsorbed oxygen. The authors also sug-
gested that, unlike conventional annealing, which results in a loss of stoichiome-
try, e.g., by formation of Cd vacancies if annealed in vacuum, the RTA process
does not change stoichiometry.
       Pronounced thickness effects on resistivity were also noted in these RTA
CdS films. For example, a 95-nm-thick film showed a resistivity of 15 -cm,
which decreased to 0.2 -cm for a thickness of 150 nm (with no pronounced
change for even thicker films). Also, storage (in a dessicator, presumably in air)
increased the resistance of the thin films, about an order of magnitude for the 95
nm film after 50 days, with continuing increase but, for a 250 nm-thick film, only
a small, initial increase (ca. 50%). RTA of the stored films decreased the resistiv-
ity to their original value before storage. The effect of storage was attributed to
oxygen adsorption.


4.2 CdSe
4.2.1 A Mechanistic Introduction
Before going into details of the various aspects of specific CdSe depositions, and
although it is not intended to deal with mechanistic aspects here (they have been
considered already in Chap. 3), it bears mentioning that, although in contrast to
CdS, the complex-decomposition mechanism has not been discussed with respect
to CdSe deposition, it is still possible that this mechanism does occur in some, or
even many, cases. If there is no evidence specifically in favor of this mechanism
in general, there is also none against it. This point is stressed here since, in the lit-
erature on CdSe (and selenides in general), it is automatically assumed that the re-
action proceeds via free selenide ions.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
4.2.2 Selenosulphate as Se Source

CdSe was deposited using a selenosulphate source as far back as 1970 [83]. Mir-
ror films of CdSe were reported to form from an ammoniacal solution of Cd2
only under conditions where Cd(OH)2 was present in the solution and at a pH
11.75. Under the conditions of those experiments, it was reported that this was the
minimum pH required to convert Cd(OH)2 to CdSe. At higher values of pH, the
rate of conversion to CdSe increased, but so did homogeneous precipitation, with
the result that the films were thinner.
       A detailed study of CdSe deposition was carried out using an ammonia-
complexed solution with selenosulphate [84]. Most of this study was concerned
with kinetic measurements, already discussed in Chapter 3. Two different types of
solution were considered: a clear solution where there was no visible Cd(OH)2 and
one with added KOH to give a visible Cd(OH)2 suspension. The former required
heating to at least 45°C for deposition to occur (although it is likely that deposi-
tion would occur even at room temperature after enough time). The CdSe was of
the zincblende structure. With a visible Cd(OH)2 suspension present, deposition
occured at room temperature, but the terminal thickness was only ca. 80 nm. The
higher the pH, the lower the terminal thickness, since more of the Cd precipitated
in the solution. The CdSe from this bath was a mixture of wurtzite and zincblende
structures. The deposition rate and terminal thickness of the films were somewhat
dependent on the nature of the substrate, both somewhat larger for Ge and Si than
for glass.
       A modification of this method used lower concentrations of ammonia (0.2
M for a Cd concentration of ca. 50 mM) in a sealed vessel, thus preventing irre-
producibilities due to escape of ammonia vapor [85]. Treatment of the Ti and
stainless steel substrates by soaking in a suspension of Cd(OH)2 improved the ho-
mogeneity of the films. At an ammnonia concentration of 0.3 M, no deposition
occurred (at least within the time frame of these depositions—about one hour).
       A triethanolamine/ammonia bath has been used for CdSe [19]. While this
system resulted in thick films for CdS (up to a few microns), CdSe films deposited
under the same conditions, only using selenosulphate instead of thiourea, were
thinner [although films of 500 nm were obtained at 30°C that did not show signs
of satuation (of thickness) after 25 hr—the longest time measured]. Ethylenedi-
amine has also been used as a complex for Cd, with both precipitates and films
formed [86]. In this case, the emphasis was on the precipitates, and no character-
ization was carried out on the films.
       Nitrilotriacetic acid (NTA) [N(CH2COOH)3] is a complexant for many
metal ions (see Sec. 2.9.1.2 for information on this compound). The sodium or
potassium salts of NTA have been used to complex Cd for CD of CdSe from se-
lenosulphate solutions [10,87]. The rate of film growth depends on many factors,
as discussed in Chapter 3; experimental details for CdSe deposition from this so-
lution are given in Chapter 2. However, growth times are generally longer than for


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ammonia-based baths—typically from a few hours to a few days for film thick-
nesses in the range of 100—300 nm. The most notable property of these films is
their change in color with deposition conditions—from yellow if deposited at tem-
peratures below 0°C and under conditions where the hydroxide mechanism is op-
erative, to very deep red (thick films appear black by reflected light) for high-tem-
perature, ion-by-ion depositions. This variation in color is a consequence of size
quantization, discussed in detail for these films in Chapter 10, and the CdSe (al-
ways zincblende) crystal size varies from 3 to 20 nm. This color variation can be
translated into a variation of CdSe bandgap from ca. 2.3 eV for the yellow films
to the bulk value (for zincblende CdSe) of ca. 1.8 ev. Annealing the films causes
crystal growth and therefore loss of the size quantization effects. The major crys-
tal growth, which corresponds to the phase change from zincblende to wurtzite,
occurs between 300 and 400°C [88,89].
       Since the optical spectra of these films are so sensitive to the crystal size (a
change of 10% in crystal size can result in an easily measured spectral shift), mea-
surements of the spectra provide a sensitive technique to investigate the effect of
different deposition parameters on crystal size. Thus, while the crystal size is not
strongly dependent on the various concentrations of reactants (apart from the
NTA:Cd ratio in the region where the mechanism changes), small increases (of the
order of 10–20%) in crystal size are observed if the Cd and/or selenosulphate con-
centration is decreased considerably in the hydroxide cluster mechanism regime
[90]. This can be rationalized, in a general way, by the greater likelihood of small,
thermodynamically unstable CdSe nuclei growing to a stable configuration if the
reactant concentrations are greater (see Section 1.2 for a discussion of nucleation
and growth), since the growth rate will be faster. The greater the concentration of
nuclei, the smaller the final crystal size for a fixed reactant concentration. This
reasoning can also explain the observation that the average crystal size increases
somewhat as deposition proceeds and reactant concentration decreases [89,91], al-
though this growth might also occur by deposition of new CdSe on previously de-
posited crystals; probably both mechanisms are operative to a greater or lesser ex-
tent. From broadening of photoluminescence peaks with increasing deposition
time, it was inferred that the crystal size distribution also increased with deposi-
tion time [91].
       Illumination of the solution during deposition by supra-bandgap light af-
fects crystal growth, probably via photoelectrodeposition of CdSe on the growing
crystals [90,92]. The solution used for CD can also be used for electrodeposition
of CdSe. Light absorbed by the individual CdSe crystals forms electron/hole pairs,
and the electrons can reduce the solution at the CdSe surface in the same way as
those supplied by an external power supply. Figure 4.3 shows the transmission
spectra of CD films deposited in the dark and under illumination for two different
deposition temperatures. The red shift in the spectra of the illuminated films indi-
cates a larger crystal size of these size-quantized samples (by ca. 1 nm) compared
to the nonilluminated ones. Additionally, the shape of the spectrum changes for


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 4.3 Transmission spectra of CdSe films deposited from a selenosulphate/NTA bath
in the dark and under illumination (tungsten halogen lamp) at two different temperatures
(6°C and 55°C).



the low-temperature illuminated film; the onset becomes less steep. This spectral
shape is typical for films electrodeposited from this solution [87]. Since the rate
of electrodeposition is essentially temperature independent while that of CD is
strongly temperature dependent, the effect of the illumination (through the rela-
tive amount of photoelectrodeposited CdSe) will be greater for low-temperature
films, seen particularly clearly by the pronounced change in shape of the low-tem-
perature film. The growth does not occur for very weak illumination, suggesting
that the photoelectrochemical deposition is not very efficient, and other processes
(electron/hole recombination or parasitic electrochemical reactions) dominate.
Also, the crystal size saturates for light intensities above a certain level. This was
interpreted to mean that one electron/hole pair was sufficient to influence the
growth process [92]. This would not be expected for a photoelectrochemical
growth process, as described earlier. A more logical explanation for the saturation,
particularly in view of the probable low quantum efficiency of the photoelectro-
chemical CdSe deposition discussed earlier, is that one of the charges is removed
more rapidly into the solution than the other. If, e.g., this is an electron, then a hole
will be left (probably trapped) on the crystal. Absorption of another photon will
form another electron/hole pair, which will then recombine rapidly by Auger re-
combination (three-body interaction) rather than form more CdSe.
       While the pH of the deposition solution (based on the cluster mechanism)
has been found to increase by as much as 0.8 during the deposition (see Sec.
3.3.2), this increase was found to be considerably greater, up to 2.2 pH units, un-
der illumination [92]. This could provide some clues about the mechanism of the
CdSe formation under illumination. A possible pathway that could account for



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
this increase in pH (i.e., overall generation of OH ) is given as
      Cd(OH)2      SeSO 2
                        3      2e → CdSe        SO 2
                                                   3      2OH                  (4.5)
This reduction has, of course, to be balanced by the local oxidation reaction. The
most likely reaction, oxidation of sulphite, leads to an increase in acidity that
would cancel out the rise in pH of Eq. (4.5). Probably other oxidation reactions
that do not generate acidity occur that result in a net increase in pH (an example
would be photocorrosion of the CdSe by photogenerated holes).
       Another possibility that could explain the effect of illumination is a change
in the electric double layer surrounding the CdSe particles, either adsorbed on the
substrate or in the solution, which could lower a potential barrier to adsorption and
coalescence, as suggested previously for film formation from Se colloids under il-
lumination [93]. Partial coalescence would reduce the blue spectral shift due to
size quantization. However, the spectral shape is not expected to undergo a fun-
damental change in this case. The photoelectrochemical explanation therefore ap-
pears more reasonable.
       Addition of silicotungstic acid (STA) to a selenosulphate/ammonia/tri-
ethanolamine bath resulted in a reduced rate of deposition but a larger final thick-
ness (greater than 1 m could be obtained). This was attributed to adsorption of
the STA on the individual CdSe crystals (typically 4–5 nm in size), which impedes
aggregation of the invidual crystals [94]. Reduced aggregation will slow down
both film growth (which relies on aggregation if the mechanism is one of colloidal
growth as appears to be the case here) and precipitation of CdSe in solution, which
will result in loss of CdSe for film formation. From the XRD pattern of such CdSe
films reported in an earlier study [95], a crystal size of ca. 4 nm could be esti-
mated. The STA may also play a role in limiting the crystal size by capping, al-
though even without the STA, small crystal sizes ca. 5 nm are usually obtained
from similar deposition baths.

4.2.3 Selenourea Source
Films have been deposited using selenourea and an ammonia-complexed solution at
65°C [96]. Zincblende CdSe was obtained with an optical spectrum corresponding
to a bandgap of 1.84 eV (the bulk room-temperature bandgap of zincblende CdSe is
ca. 1.8 eV). Analysis of electrical conductivity measurements indicated charge trans-
fer occurred via a variable hopping mechanism through fairly deep states (a level
0.29 eV below the conduction band was found from these measurements).

4.2.4 N,N-Dimethylselenourea Source
N,N-Dimethylselenourea was used with ammonia baths and additional citrate or
tartrate complexation at pH of 11.3 (citrate bath) or 10.4 (tartrate) to deposit CdSe
on glass at room temperature [97]. No XRD of the films was detected. From the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
optical spectra, the bandgaps were high (ca. 2.1 eV, even larger for thinner films)
suggesting crystal sizes of ca. 5 nm or less (see Chap. 10). The XRD spectra of
such a small crystal size can be missed in normal powder XRD if special care is
not taken, mainly slow scanning. Electrical and photoconductive properties of
these films are described in Section 4.2.7.2.
      Another study of these films concentrated on the particle size of the films
and is discussed in Chapter 10 [98].

4.2.5 Selenosemicarbazide Source
One example is given in the literature where selenosemicarbazide was used to de-
posit CdSe from solutions containing different complexants [99]. This reagent
was apparently more stable than selenourea or N,N-dimethylselenourea, and it is
surprising that it does not appear to have been subsequently used. The CdSe films
were specular and had a resistivity of 108–109 -cm, which dropped about an or-
der of magnitude on illumination.

4.2.6 Epitaxial Deposition
Using a nitrilotriacetate solution and a complex:Cd ratio high enough to prevent
Cd(OH)2 formation (ion-by-ion mechanism), epitaxial growth of zincblende CdSe
                                                                   ¯¯¯
was obtained from a selenosulphate solution on both (111) and (111) polar faces
of single-crystal InP [100]. (For lower complex:Cd ratios, in the regime of the hy-
droxide cluster mechanism, the deposits were always polycrystalline, as expected
for this mechanism.) The degree of epitaxy improved with increasing temperature
and was high at 90°C. Additionally, there was a high density of twins in the de-
posits obtained at low temperatures, but less in those obtained at 90°C. The addi-
tion of silicotungstic acid to the deposition solution destroyed the epitaxial
growth, presumably due to blocking of the InP surface (and also the growing
CdSe) by the strongly adsorbed silicotungstic acid.
       This study also reported that films deposited on carbon membranes at tem-
peratures 80°C were of hexagonal (wurtzite) structure, with a high density of
planar defects, in contrast to the zincblende obtained from both hydroxide and ion-
by-ion mechanisms at lower temperatures and to the epitaxial films on InP at all
temperatures.

4.2.7 Some Specific Optoelectronic Properties
4.2.7.1     Photoluminescence (see also Sec. 10.2.3)
Photoluminescence of CdSe films deposited from the selenosulphate/nitrilotriac-
etate bath varied both in intensity and in spectral shape. As an example of the for-
mer, a sample on glass, broken from a glass slide when the film thickness was ca.
30 nm, gave a much stronger signal than the original sample left in the deposition



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
solution until the thickness reached ca. 100 nm. Such an effect, however, was not
reproducible (unpublished results). Regarding spectral shape, both bandgap emis-
sion with weak, if any, subbandgap signal [87,91] and deep subbandgap signal
with relatively weak bandgap emission [89,101] have been observed with no ap-
parent difference between the samples. (Note that the use of the term bandgap
emission refers to emission close to the bandgap energy. The emission may be
from very shallow traps.) Although the measurement temperatures for these dif-
ferent experiments were not all the same, temperature does not usually have a very
pronounced effect on the ratio between the two peaks, except below 50 K, and
cannot explain the observed differences. In humid atmosphere, where water vapor
is adsorbed on the CdSe, the predominant emission is close to bandgap [101] (see
Sec. 4.2.7.3). Most studies do not state whether measurements were carried out in
the open atmosphere (low-temperature measurements, of course, are not). An in-
vestigation on films deposited from an N,N-dimethylselenourea/citrate/ammonia
bath showed both bandgap and lower-energy (ca. 1.75 eV) peaks [98]. The latter
was attributed to larger crystal size (see Chap. 10), but it is likely that the low-en-
ergy peak is a subbandgap response arising from surface states.
      The temperature dependence of the emission spectra provides useful infor-
mation on the source of the emission. Figure 4.4 shows a series of emission spec-




FIG. 4.4 Photoluminescence spectra taken at different temperatures of a CdSe film
deposited from a selenosulphate/NTA bath. Crystal size of CdSe ca. 4 nm. (From ref. 101
with permission from Elsevier Science).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tra of a (4-nm-crystal-size) CdSe film taken at different temperatures. Each spec-
trum is made up of a bandgap peak (at high photon energy) and a subbandgap peak
at lower energy. The bandgap peak becomes weaker as the measurement temper-
ature increased. This is normal for a band-to-band transition. The subbandgap
peak, however, initially increases up to ca. 50 K, and then decreases with further
increase in temperature. This behavior (and also incident light intensity depen-
dence of the emission) is typical of donor–acceptor recombination. In the present
case, the nanocrystals are essentially intrinsic and are not expected to contain bulk
dopants. However, the surface states (see following section), after trapping
charge, can behave in much the same way, and the recombination can be ex-
plained by recombination of surface-trapped electrons and holes with the emission
red-shifted ca. 0.5 eV from the bandgap [101]. In the study by Trojanek et al. [89],
this emission is shifted nearly 0.7 eV from the bandgap and the parallel increase
in emission energy as a function of crystal size with the theoretical (effective mass
approximation) conduction band shift interpreted to mean that the emission oc-
curred from shallow-trapped electrons to deep-trapped holes.
       Time-resolved measurements of photogenerated (very intense illumination,
up to 0.56 GW/cm2) electron/hole recombination on CD (selenosulphate/NTA
bath) CdSe of different crystal sizes has shown that the trapping of electrons, prob-
ably in surface states, occurs in ca. 0.5 ps, and a combination of (intensity-depen-
dent) Auger recombination and shallow-trapped recombination occurs in a time
frame of ca. 50 ps. A much slower (not measured) decay due to deeply trapped
charges also occurred [102]. A different time-resolved photoluminescence study
on similar films attributed emission to recombination from localized states [103].
In particular, the large difference in luminescence efficiency and lifetime between
samples annealed in air and in vacuum evidenced the surface nature of these states.
       A photoluminescence study of CdSe deposited from a selenourea/ammonia
solution onto glass at 80°C and relatively low pH (7–8) was made [31]. An emis-
sion peak centered at 1.445 eV (860 nm) was observed with a tail to the low en-
ergy side. Such an emission must be due to deep traps, since the shift from the
bandgap emission is ca. 0.4 eV, a value close to, or somewhat less than, that for
films deposited from selenosulphate solution (see earlier). Annealing in air shifted
the emission to higher energies.
4.2.7.2      Photoconductivity
There appears to be a fundamental difference between films deposited from a se-
lenosulphate source and those deposited from a dimethylselenourea source (most
of the detailed photoconductivity studies are on the latter). As-deposited films us-
ing a dimethylselenourea source had a resistivity of ca. 2        1012 /sq ( 108
   -cm for a film thickness of 0.5 m). The resistivity dropped only a little under
illumination, with a maximum decrease of less than an order of magnitude. For




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
films deposited at a higher temperature (50°C), the resistivity was up to an order
of magnitude lower (ca. 107 -cm), which dropped ca. six times under illumina-
tion. Air-annealing increased the photosensitivity greatly (up to ca. 107 at 450°C),
in contrast to the case for CdS, where good photosensitivity was obtained for as-
deposited films (but always deposited at high temperatures) and was reduced on
air-annealing. The dark resistivity of the CdSe films increased by nearly an order
of magnitude after annealing, again in contrast to the decrease usually obtained for
CdS films. This behavior is also in contrast to CdSe films deposited from seleno-
sulphate/triethanolamine/ammonia solutions, where the relatively low resistivity
of the as-deposited film (5 103 -cm) dropped to a few ohm-centimeters on
annealing (430°C in air) [95], or to another study of selenosulphate/ammonia
films with dark resistivity of 107–108 -cm, which dropped to 1–10 -cm after
annealing at 280°C in vacuum [104]. On the other hand, in another study on se-
lenosulphate/ammonia films, which reported a dark resistivity of 105–106 -cm
as deposited, the dark resistivity initially increased on heating in air (ca. four times
at 180°C) then decreased only an order of magnitude on heating at 340°C [105].
The photosensitivity of these latter films was low but appreciable in the as-de-
posited state (varying from 5 to 50) and increased up to ca. 103 after annealing at
180°C before decreasing again at higher annealing temperatures. While it is clear
that CdSe films deposited from dimethylselenourea possess higher photosensitiv-
ity after annealing compared to those deposited from a selenosulphate bath, the
reasons for the difference are not understood. One possibility that might explain
this difference was suggested, based on a comparison of CdSe films deposited
from a dimethylselenourea bath with CdS films. It was hypothesized that the dif-
ference in the effect of annealing on dark resistivity was due to the formation of
conducting CdO due to the oxidation of CdS, which did not occur in CdSe [106].
The ease of oxidation of the Cd chalcogenides is normally CdTe CdSe CdS;
however, this usually refers to oxidation of the chalcogen, e.g., CdTe to CdTeOx,
rather than to CdO. More details on the effect of annealing on the photoconduc-
tion properties are given in Refs. 106 and 107. Another characteristic of these
films, different from most CD CdS films is the fast photocurrent decay (no more
than a few seconds at most) of the films, both as deposited and annealed. Clearly,
the trapping centers in these films are also efficient recombination centers, which
may, at least in part, explain the low photosensitivity of the as-deposited films.
CdSe films prepared by the selenosulphate/citrate process were less photosensi-
tive than those deposited by the dimethylselenourea method, although details were
not given [106].
       Some studies of these films were made after immersion in solutions of Hg
or Cu ions, when ion exchange reactions occurred, converting the surface of the
crystallites to (partial) Hg–Se and Cu–Se compounds. As expected, such treat-
ments could affect the photoconductivity of the films greatly.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
4.2.7.3      Surface States on CdSe Films
An important aspect of semiconductor films in general with regard to electronic
properties is the effect of intrabandgap states, and particularly surface states, on
these properties. Surface states are electronic states in the “forbidden” gap that ex-
ist because the perfect periodicity of the semiconductor crystal, on which band
theory is based, is broken at the surface. Change of chemistry due to bonding of
various adsorbates at the surface is often an important factor in this respect. For
CD semiconductor films, which are usually nanocrystalline, the surface-to-
volume ratio may be very high (several tens of percent of all the atoms may be sit-
uated at the surface for 5 nm crystals), and the effects of such surface states are ex-
pected to be particularly high. Some aspects of surface states probed by photolu-
minescence studies are discussed in the previous section.
       Surface treatments of CD CdSe films deposited from selenosulphate/NTA
solutions have a pronounced effect on various optical, electrical, and optoelec-
tronic properties of the films, due to interaction with or modification of such sur-
face states. Mild etching (dilute HCl) of the films reverses the direction of current
flow both in CdSe/polysulphide photoelectrochemical cells [108] and in Kelvin
probe surface photovoltage (SPV) measurements in air [109]. These studies are
discussed in more detail in Chapter 9, in Section 9.2 on photoelectrochemical
cells. At this point, it is sufficient to state that the effect is believed to be due to
preferential trapping of either electrons or holes at surface states that are modified
by the etching process.
       The adsorption of water vapor on these CdSe films acts to passivate, at least
to a large extent, some of these surface states. In particular, strong subbandgap sig-
nals have been observed in SPV measurements of these CdSe films (as deposited)
only when measured in a dry ambient; in normal atmospheric humidity, no such
signal occurs and only suprabandgap light gives rise to an SPV signal [109]. Par-
allel results have been observed in photoluminescence measurements, which are
particularly sensitive to surface states: The predominant subbandgap emission that
occurs in a dry atmosphere changes to a predominantly (near) bandgap emission
in a humid atmosphere [101]. The asymmetrical nature of these states, seen in op-
tically detected magnetic resonance (ODMR) spectroscopy, is further evidence
for their surface nature; bulk states are expected to be symmetric [110]. It is im-
portant to note that these effects are seen only in small-crystal-size nanocrystalline
films (the foregoing experiments were carried out on 4- to 5-nm-crystal-size
films). No such effects were observed if the crystal size was ca. 20 nm; the sur-
face-to-volume ratio is already much smaller for this size. Current–voltage spec-
troscopy of individual CdSe quantum dots deposited mostly by electrodeposition,
but also by CD, using a conducting AFM (atomic force microscope) tip also
showed directly the presence of surface states in a dry atmosphere but not in a hu-
mid ambient [111].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
4.3 CdTe
There appear to be only two independent reports in the literature on CD CdTe
(which is the only telluride reported in the CD literature). In the first one, the CdTe
was deposited with the main purpose of carrying out electron spin resonance and
morphological studies of the effect of annealing, while another paper, based ap-
parently on the first, described the deposition, with the main purpose of further use
in photovoltaic cells. Therefore only limited detail on the actual deposition or
properties of the as-deposited films were given. This is surprising in view of the
fact that the first report is the first case of a CD telluride.
       In the original study by Padam and Gupta [112], the deposition solution con-
tained triethanolamine/ammonia-complexed CdAc2 and the Te source was TeO2
with hydrazine hydrate as a reducing agent. The nature of the TeO2 solution was
not clear, since TeO2 is only slightly soluble in water; it may have been dissolved
in a hydroxide solution, in which it is much more soluble. The deposition was car-
ried out at 90°C. Electron diffraction of the as-deposited films showed both
zincblende and wurtzite phases of CdTe (the zincblende phase is the more stable
and commonly encountered one).
       Buckley used the same technique to deposit films for photovoltaic cells
[113], only with CdCl2 as Cd source and apparently a lower Cd concentration.
Electron diffraction of these films showed a predominantly zincblende structure
with some wurtzite phase. The films were p-type with a resistivity of 20              5
  -cm. These values, and the subsequent photovoltaic cells, apparently refer to as-
deposited films; no reference to annealing was made in this study.
       The second, more recent investigation described Te dissolved in sodium
sulphite as a source of telluride (tellurosulphate) [114]. It was previously believed
that Te was insoluble in sulphite solution under normal conditions, although there
is one previous reference in the literature to this reagent prepared under pressure
at high temperatures [115]. The CdTe deposition described in Ref. 114 indicates
that the solubility is sufficient to allow deposition of tellurides. The deposition
was carried out in a solution of CdSO4 containing triethanolamine, ammonia, and
NaOH. Both deposition rate and film thickness were maximal at 75°C deposition
temperature. As with the previous TeO2 deposition, both zincblende and wurtzite
(dominant) phases of CdTe were obtained. Elemental analysis showed a small Cd
excess. This appears to be in contradiction to the XRD analyses, which showed
considerable amounts of Te (also TeO2, particularly after heating at 100°C). Al-
though the films were apparently highly scattering, making bandgap measurement
more difficult, the bandgap (direct), measured from the optical spectrum, was ca.
1.4 eV, close to the literature value. The (room-temperature) resistivity was ca.
106 -cm and the conductivity n-type. The carrier density was ca. 1019 cm 3.
       In experiments carried out by us to repeat these two methods, we have been
able to deposit CdTe, although stoichiometry control was difficult, with Te oc-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
curring in the hydrazine method and mainly TeO2 impurity in the tellurosulphate
one. Probably careful attention to removal of oxygen from the solution and maybe
addition of hydrazine to the tellurosulphate-based solution would improve the sto-
ichiometry. It is clear that other telluride films, in particular with metals that form
very insoluble chalcogenides, such as Pb, Cu, and Hg, should be accessible using
these methods.
      Although not strictly CD, CdTe films were very recently prepared by treat-
ing CD Cd(OH)2 films with a telluriding solution [116]. The Cd(OH)2 films were
deposited from an alkaline H2O2 bath containing citrate-complexed Cd. These
films were treated with a solution of elemental tellurium dissolved in hydrox-
ymethanesulphinic acid. It appears that this solution contains telluride ions, al-
though it has not yet been well characterized. The Cd(OH)2 was converted
(incompletely) into CdTe films. The bandgap was ca. 1.5 eV (approximate due to
the highly scattering nature of the films). The dark resistivity was ca. 5          108
                                         7
   -cm, which decreased to ca. 7 10 -cm upon illumination.


4.4 ZnS
4.4.1 Introduction
Chemical deposition of ZnS has been the subject of considerable activity, the main
reason for which is its hoped-for substitution for CdS in thin-film photovoltaic
cells. Since the chemistries of Zn and Cd are similar in many ways, it might be ex-
pected that deposition of their chalcogenides is also similar. However, there is a
dominant difference in their properties that results in the fact that ZnS is consider-
ably more difficult to deposit by CD than CdS. This difference is manifested by the
difference in solubility products between the respective hydroxides and chalco-
genides. Considering, for example, the sulphides, the relevant values of Ksp are:
                          14
      Cd(OH)2 2      10          CdS 10 28
                          17                   25
      Zn(OH)2 8      10          ZnS 3 10
The deposition for mechanisms proceeding through hydroxide clusters is depen-
dent on a large difference between the solubility products of the hydroxide and
sulphide, since the sulphide exchanges the hydroxide. The situation for ZnS is
therefore much less favorable than that for CdS: About a million times higher con-
centration of sulphide is required to form ZnS than CdS [see Eq. (3.46)], as shown
in graphical form in Figure 4.5. More sulphide is also required at higher tempera-
tures because of the strongly increasing ion product of water with increasing tem-
perature, resulting in higher hydroxide concentrations for any particular value of
pH. (It is again stressed that in dealing with values of solubility products, there are
large variations, sometimes of orders of magnitude, between one source and an-
other. Therefore calculations based on these values are correspondingly impre-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 4.5 Steady state concentration of sulphide ion needed to convert the hydroxides of
Cd and Zn into the corresponding sulphides at 25°C and 60°C.



cise. However, to take the present case as an example, whether the difference in
solubility product ratios between the Zn and Cd hydroxides and sulphides is 106
or an order or so larger or smaller in magnitude does not qualitatively alter the
conclusion.)
       The obvious solution to this problem is to deposit ZnS at a lower pH when
the OH concentration will be lower. However, with thiourea, for example, low-
ering the pH results in slower hydrolysis of the thiourea and therefore a lower sul-
phide concentration (presumably also a reduced decomposition rate if the
thiourea–hydroxide complex-decomposition mechanism is effective). This has
been circumvented, as described later, by working with alternative sulphiding
agents at low values of pH.
       Deposition by a pure ion-by-ion mechanism should also solve this problem,
since no hydroxide is involved. However, in this case we encounter the problem
of the high Ksp of ZnS compared to CdS, which again means that more sulphide is
needed. For thiourea, this means a higher pH, which again means that strong com-
plexation is needed to prevent Zn(OH)2 formation, by reducing the free [Cd2 ].
However, this will also reduce the rate of ZnS deposition. While there are many
examples in the literature of cluster deposition of ZnS, there does not even seem
to be one unambiguous case of ion-by-ion deposition of this semiconductor.
       As already implied, most depositions of ZnS have been carried out under
conditions where Zn(OH)2 can be formed. From the forgoing general discussion,
this means that, even if ZnS is identified (by XRD, for example), there is in most
studies no evidence for the absence of some hydroxide species. More recent stud-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ies have addressed this problem, and there are some papers with “Zn(OH,S)” in
the title. This probably usually refers to a mixture of ZnS and unreacted Zn(OH)2
and not to a true ternary (more accurately, quaternary) compound. For that reason,
such films are discussed in this chapter rather than in Chapter 8 (Ternary Semi-
conductors).
       A recent condensed review on ZnS deposition, with emphasis on the differ-
ences in CD of CdS and ZnS, is given in Ref. 117.
       The majority of studies involved ammonia-complexed baths with thiourea as
a source of S. Hydrazine was also used in most cases, although there were several
studies where ZnS was obtained without hydrazine. Amines (triethanolamine or
ethanolamine) were also used, again both with and without hydrazine (which is it-
self a type of amine). One effect of the hydrazine is to speed up the deposition. Hy-
drazine, a strong reducing agent is expected to increase the rate of sulphide forma-
tion. However, it seems that this is only part of the picture. A study of the effect of
various amines on the rate of ZnS deposition showed that although hydrazine gave
the fastest rate, other amines (ethanolamine and triethanolamine) also increased the
deposition rate [118] (see Fig. 3.7). The amines all act as complexants; therefore
they would reasonably be expected to reduce the deposition rate by reducing the
free-Cd2 concentration. Amines have reducing properties (redox potentials of hy-
drazine, ethanolamine, and triethanolamine are 1.16, 0.56, and 0.46, respec-
tively). It appears that the amines accelerate thiourea decomposition. However, the
mechanism of this effect is not yet clearly understood [119].
       Kinetic studies have found a value for the activation energy of ZnS film for-
mation from thiourea/ammonia-based baths of 5 kCal/mole (21 kJ/mole), too low
to be chemical reaction controlled (unlike CdS, which is thiourea-decomposition
controlled) but not diffusion controlled, since stirring the deposition solution has
no effect [120,121]. This was explained by a rate-determining step of Zn-ligand
dissociation from a hydrazine complex. However, addition of an additional com-
plexant into the solution will only act to reduce the free-Zn2 concentration, not
only due to the extra complexing power of the hydrazine, but also because of the
statistical factor that complexes comprising more than one ligand will be further
stabilized by the fact that different combinations of the ligands can be found in the
complexes [119]. There is clearly a fundamental difference in deposition mecha-
nism of these ZnS depositions compared with that normally encountered with
CdS. More details on the difference in activation energies of deposition for CdS
and ZnS (ZnSe) are discussed in Section 3.5.

4.4.2 Specific Studies of ZnS Deposition

Some specific properties of ZnS films grown from thiourea/ammonia/hydrazine
baths, sometimes with added compounds, are given next.
       Doña and Herrero noted that hydrazine was not essential for growth but that
it speeded up the growth rate (by a factor of ca. 3) and improved the homogeneity


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
and the specular reflectance (a consequence of better microhomogeneity) [120].
No XRD pattern was observed for these films, indicating either highly disordered
and/or very small crystals. However, TEM/ED did show that the sphalerite phase
was formed with a crystal size of 10 nm. The stoichiometry measured by EDS
was Zn:S       1:1. This implies that either pure ZnS or a hydroxy sulphide
(Zn,OH,S) was formed.
      Optical transmission of the films was a spectral average of ca. 85% beyond
the bandgap. The bandgap measured from the spectrum was 3.76 eV (literature for
zincblende ZnS 3.6 eV), suggesting size quantization, although the measured
crystal (ca. 10 nm) is at least twice the size required for such an effect.
      Resistivity was nearly 109 -cm and independent of bath composition.
Temperature dependence of resisitivity gave an activation energy of 0.95 eV,
which was ascribed, based on previous studies, to an acceptor level above the va-
lence band due to Cu impurity.
      Using various amines added to the ammonia bath (in most cases with added
hydrazine), sphalerite ZnS films were obtained with a crystal size of ca. 3 nm [118].
Rutherford Backscattering Spectroscopy (RBS) analyses showed that there was
about twice as much Zn in the films as S. (More basic solution and more hydrazine
gave more stoichiometric films). Extended X-ray Absorption Fine Structure (EX-
AFS) and Fourier Transform Infra-red (FTIR) spectroscopy showed that the films
did not have Zn-O groups but rather Zn-OH ones [122] and that there is probably
a mixture of ZnS and unreacted Zn(OH)2, quite likely as a ZnS shell around a
Zn(OH)2 core. Optical spectra gave a bandgap of ca. 3.85 eV, considerably blue-
shifted from the bulk value of 3.6 eV, as expected from such small crystals.
      Nucleation studies of ZnO and ZnS on glass and SnO2-glass from ammo-
nia/thiourea baths (sometimes also with hydrazine) were carried out [123]. The de-
position conditions, mainly pH, were varied. On glass, both ZnS and ZnO could be
deposited, depending on conditions. On SnO2-glass, however, only ZnO was
formed (a few percent S could be obtained at high pH). This suggested that a sur-
face-activated mechanism was important for nucleation of ZnS and less so for ZnO.
      Most depositions were carried out at high temperatures. One exception was
a room-temperature deposition from an ammonia/thiourea bath (no hydrazine)
[124]. No structural characterization was made, but the optical spectra were con-
sistent with ZnS. The spectra show high transmission and low absorption in the
suprabandgap region, along with low specular reflectance. Some samples, which
had been deposited for a relatively long time, exhibited strong absorption (and
somewhat increased reflectance) in the IR region, which could be ascribed to free
carriers. The bandgap, at between 3.7 and 3.8 eV, was slightly higher than the bulk
value for sphalerite ZnS. Some room-temperature depositions were also carried out
from a thiourea/ammonia/ammonium sulphate/hydrazine bath. Again, high trans-
mission in the suprabandgap region and a bandgap of 3.75 eV were measured.
      A bath without ammonia, using triethanolamine/hydrazine/thiourea and
added NaOH to a pH ca. 12.3, was described [125]. Contrary to most other depo-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sitions, the ZnS had the wurtzite structure, measured by XRD. Optical spec-
troscopy gave a bandgap of 3.68 eV Transmission above the bandgap was ca.
30%, implying a large amount of scattering (specular reflection would not be as
high as 70%). The resistivity was ca. 106 -cm, and the films were n-type.
       Thioacetamide has also been used to deposit ZnS. In this case, depositions
could be carried out under acid or neutral conditions as well as alkaline ones.
Three different studies were carried out using thioacetamide. In one [126], two
different baths were investigated—one with and the other without ammonia (the
latter was probably slightly acidic and certainly had a considerably lower pH than
the ammonia bath). Zincblende ZnS was obtained from both baths. The crystal
size (TEM) was 6–8 nm (ammonia bath at 90°C); from the ammonia-free bath, the
crystal size was not given, but from the sharpness of the ED pattern, it is probably
larger. The bandgap (temperature independent) for films deposited from the am-
monia-free bath was 3.6 eV ( literature value). For films deposited from the
ammonia bath, values of bandgap up to 4.05 eV (5°C deposition), which depended
on deposition temperature (3.8 eV at 90°C; 3.95 eV at 30°C), were measured.
Urea was used in the ammonia-free bath to improve adherence, although it is not
known why this affected adherence.
       ZnS was grown from a thioacetamide bath using triethanolamine to com-
plex the Zn2 and buffered by NH3/NH 4 to a pH of 10 [127]. Glass coated with
these ZnS films was found to be a good substrate for other CD semiconductors that
might otherwise exhibit poor adhesion, and this was the purpose of this study. It
was noted from optical transmission that for films no thicker than ca. 0.15 m,
scattering was negligible, but it became increasingly marked for thicker films.
XPS depth-sputtered analysis of these films indicated that the ZnS–glass interface
was not sharp, and it was suggested that Zn diffuses into the glass to some extent,
explaining the good adhesion of the films [128]. A more complete characteriza-
tion of these films was subsequently carried out [129]. No XRD pattern was found
for the as-deposited films, implying amorphous or very small crystal size. Even
after a short anneal at 500°C, a crystal dimension as low as 13 nm (depending on
crystal orientation) was measured by XRD, implying a much smaller size in the
as-deposited film. In line with the crystal size, optical absorption showed
bandgaps between 3.85 and 3.95 eV, higher than the bulk value, presumably due
to size quantization.
       These films were not photoconductive as deposited but became so on air-an-
nealing with an optimum annealing temperature of 388°C, when the photosensi-
tivity increased to 104. The film resistivity decreased from ca. 5 107 -cm (as
deposited) to ca. 104 -cm after air-annealing at 400–500°C.
       The third study used a solution of ZnCl2 and thioacetamide at a pH of 2.45
[130]. Films were deposited on ITO/glass at 70°C. In spite of the acid conditions,
which precluded formation of Zn(OH)2, a precipitate formed in the solution in




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
parallel to film formation. The crystal size for both film and precipitate was ca. 3
nm. The energy of activation was measured for both the precipitate and the film
formation (usually, it is only measured for film formation) and found to be the
same for both, with a value of 9.5 kCal/mole (40 kJ/mole); this value is interme-
diate between that usually found for CdS deposition, where the rate-determining
step is a chemical reaction, and for ZnS deposited from an alkaline thiourea bath
[120] (21 kJ/mole; see Sec. 3.5). Whether this difference is due to the expected
different mechanisms (ion-by-ion probable for the acidic bath and cluster mecha-
nism for the alkaline one) or to differences in thiourea and thioacetamide decom-
position is unknown at present.
       Thiosulphate has been used in an acid bath (pH 2–4) at 85°C [131]. The
films were poor (not uniform, powdery, and nonadherent). The only other charac-
terizations given were the bandgap (3.4 eV) and resistivity (106–107 -cm).
       Thiosulphate has also been used to deposit ZnS using a photochemical re-
action, in the same manner as used previously for CdS (see Sec. 4.1.6.7 for a de-
scription of the proposed mechanism). In brief, UV light of wavelength shorter
than 300 nm (from a Hg lamp) forms hydrated electrons and elemental S from
thiosulphate solution, and this reacts with Zn2 to give ZnS [132]. The ZnS was
highly nonstoichiometric with excess Zn. This is not surprising, since ZnS itself
is sensitive to UV light, with the formation of elemental Zn due to the strong re-
ducing action of photogenerated electrons.
       While referring to precipitation in solution and not CD of films, it is of in-
terest to mention that N-allylthiourea was used as a sulphur source in a Zn2 /am-
monia bath [133]. Pure ZnS (as detected by XRD) was obtained at a pH of ca. 11.0
at 90°C. At lower pH values (and at 80°C), only ZnO was obtained; at higher val-
ues, a mixture of ZnO and ZnS was formed. The ZnS was the wurtzite form in all
cases. The ZnS crystal size was ca. 5 nm at pH 11.0 and slightly smaller (ca. 4
nm) at higher pH values, which gave a mixed phase of ZnS and ZnO. The ZnO
crystal size was much larger (ca. 200 nm). The ZnS fraction increased as the am-
monia concentration (at constant pH) increased (lower [Zn2 ]) or as the pH de-
creased (at constant ammonia concentration) from 13 to 11 (lower [OH ]). Such
experiments should help in choosing optimum conditions for CD.
       A variant of CD has been described where ZnS was precipitated as a gel by
adding concentrated S2 to a concentrated solution of a Cd salt. The pH was then
reduced by HNO3 to a value between 5 and 7, when a semitransparent sol formed.
Heating this sol between 100 and 200°C (in an autoclave) resulted in the forma-
tion of zincblende ZnS films [134]. If the S2 was not in excess (twice the Cd con-
centration), some ZnO, together with some wurtzite ZnS, also formed. Addition
of a CuCl2 solution to the pH-adjusted sol and heating at 140°C resulted in Cu-
doped ZnS (particle size 60       10 nm) that showed several photoluminescence
bands [135].




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
4.5 ZnSe
4.5.1 Introduction
In contrast to CdSe, which in most cases was prepared using selenosulphate as a
Se source, ZnSe was deposited in the majority of studies using selenourea. One re-
port was given where ZnSe (among other selenides) was deposited from an am-
moniacal solution at 20°C [99] using selenosemicarbazide as a Se source. Other
than that the ZnSe films were specular in nature, no details on their properties
were given. There are two reported studies where selenosulphate was used. In one
[136], it was noted that hydrazine was essential to obtain film formation; for ZnS,
while it was usually preferable to include hydrazine in the deposition solution, it
was apparently not essential and films could be made without it. This difference
probably reflects the higher solubility of ZnSe compared to CdSe, the lower solu-
bility of the hydroxide of Zn compared to that of Cd and either the faster forma-
tion of selenide ion from selenourea compared to selenosulphate or a different de-
position mechanism for the different Se sources. The other study reportedly
obtained ZnSe without using hydrazine (triethanolamine and ammonia were used
together as complexants) [137]. Unusually high-concentration selenosulphate so-
lution (0.8 M) was used. No structural or analytical characterization was carried
out on the films, although the bandgap measured from the optical absorption was
in the correct region for ZnSe.
       Hydrazine was also used in all the studies employing selenourea (but not
where N,N-dimethylselenourea was used—see later). It is not clear to what degree
hydrazine was essential in these studies. In an early theoretical study of ZnSe de-
position from Zn–ammine/selenourea baths, the use of hydrazine is not mentioned
[138]. On the other hand, in a patent by the same authors describing ZnSe deposi-
tion, the use of hydrazine hydrate was the main issue in the claims [138a]. In most
of these studies, ammonia was used as a complexant. The conditions varied some-
what from study to study, but it appears, either explicitly or by resorting to edu-
cated guesses, that in all cases the deposition occurred by a hydroxide cluster
mechanism. The pH was usually ca. 11.5, and deposition temperatures varied
from 50°C to 80°C. The exception was the deposition described in the patent by
Kitaev and Sokolova, where only ZnCl2, hydrazine hydrate, and selenourea (
acetic acid to acidify the very alkaline solution to a pH of ca. 9) were used [138a].
It was claimed that deposition could be obtained over a temperature range of
10–70°C. No characterization or properties of these particular films were given.

4.5.2 Depositions Using Selenourea
4.5.2.1      Structure
XRD of the films gave very broad, ill-defined peaks [139,140]. In one study, the
precipitated powder from the deposition solution was found to be wurtzite, while



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
grazing-angle XRD (more sensitive for thin films) gave a poorly defined, very
broad spectrum [141]. In another study [140], while no normal XRD pattern was
obtained, grazing-angle XRD showed a mixture of wurtzite and zincblende ZnSe
together with a little ZnO. TEM/ED showed zincblende ZnSe in one study [139]
and predominantly wurtzite in another [140]. Crystal size was measured in two
cases by TEM: 2–2.5 nm [139] and ca. 10 nm [140].
4.5.2.2      Composition
Elemental analyses found such films to be Zn rich (with respect to Se) [139,142].
The films were oxygen rich, and it is probable that, as was often found with ZnS,
the films are a mixture of ZnSe and Zn(OH)2, together with a little ZnO in some
cases [140]. An XPS depth profile study found the surface of the films to be more
stoichiometric (although still Zn rich) than near the substrate [143].
4.5.2.3      Optical Spectroscopy (Bandgap)
Optical spectroscopy was used to measure bandgap (bulk value ca. 2.6 eV). Such
values could be correlated with the crystal size, with values of 2.7 eV [140] and 2.9
eV [139,142], the latter due to size quantization in very small crystals (ca. 2.5 nm).
One study noted that the film was a transparent white (colorless) turning to orange,
presumably due to Se formation on exposure of excess selenourea to air [141]. The
white film, when annealed in air at 300°C, showed a broad ZnSe diffraction peak
(by 400°C, the film was converted to ZnO). This, together with the very broad
peaks (of both the 300°C annealed sample and of the powder precipitated in the so-
lution) suggested that the white color was due to size quantization (white implies a
bandgap of 3 eV—bulk ZnSe is pale yellow). The presence of oxide/hydroxide
in these films would result in a weakening of the yellow color of pure ZnSe but
would not be expected to change the spectral position.
4.5.2.4      Electrical Resistivity
Electrical resistivity was measured in only one case, with a value of 2           108
  -cm [140].

4.5.3 Deposition Using N,N-Dimethylselenourea
A modification of the general procedure used a substituted selenourea (N,N-
dimethylselenourea) instead of selenourea and no hydrazine. Citrate was used, to-
gether with ammonia, presumably as a co-complexant, although it may also have
functioned as a mild reducing agent, and the deposition was carried out at a rela-
tively low temperature of 50°C [144].
      No XRD pattern of the films was found, but the powder precipitated in the
solution was zincblende phase with very broad peaks ( 3 nm coherence length).
In contrast to other measurements, where the Zn:Se ratio was greater than 1, the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Zn:Se ratio of these films was less than 1 (0.85). As with the other films, there
were considerable amounts of oxygen (this was measured by XPS and includes
normal surface-adsorbed oxygen). A value for the bandgap of 2.63 eV was mea-
sured from the optical spectrum. This (very slightly) higher value than the bulk
value was explained by the small crystal size. (Size quantization for a crystal size
of 3 nm would be expected to give a larger blue shift than this). The films were n-
type with a resistivity of ca. 3 107 -cm, within an order of magnitude of the
value measured from a selenourea bath [140].

4.5.4 Deposition Using Selenosulphate
There is one reported brief study of ZnSe deposition using selenosulphate [136].
Considering the understandable preference for using selenosulphate rather than
selenourea for CdSe depositions in most cases (selenosulphate is more stable and
simpler to make and to handle), it is surprising that this is not also the case for
ZnSe. It is possible that this is simply a case of “inertia”; i.e., most researchers
follow essentially the same recipe (although the selenosulphate technique de-
scribed here predated the other studies).
       Hydrazine was used (in its absence, no ZnSe was formed). A mixture of both
triethanolamine and ammonia, together with NaOH, was used as complexant/pH
adjuster; no explanation of the solution composition was given. The deposition was
carried out at 100°C (the highest deposition temperature of all these processes).
       The films were light yellow (characteristic of ZnSe), which changed to a
light reddish color. The reddish color might have been due to formation of Se un-
der the conditions of the deposition, although Se would be expected to dissolve in
the selenosulphate. X-ray diffraction showed the films (presumably the reddish
ones) to be wurtzite ZnSe. A bandgap of 2.62 eV was measured from the optical
spectra. The films were n-type with a relatively low resistivity of ca. 104 -cm.

4.5.5 Deposition Using Selenosemicarbazide
Selenosemicarbazide was used in one study to make, among other materials, ZnSe
films and precipitates from aqueous ammoniacal solutions at 20°C [99]. The films
were specular, but no further information was provided.

4.5.6 Miscellaneous Methods
A novel CD technique used metallic Al to reduce Se. Elemental Se and metallic
Al foil, together with ZnCl2, were dissolved in NaOH solution and heated to 80°C
in an autoclave with substrates of Teflon or alumina [145]. (Metal chalcogenide
films have been chemically deposited onto Teflon in a number of reports.) It is no-
table that the films deposited onto alumina were reported to be 1.8 m thick—
much thicker than ZnSe CD from other solutions. The films were sphalerite ZnSe



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
with a crystal size greater than 10 nm, estimated from XRD peak widths. The
Zn:Se ratio was essentially unity, in contrast to other reports, where the films were
invariably nonstoichiometric. (At deposition temperatures of 125°C and above,
ZnO was also formed.) The mechanism of this process was suggested to be:
                          3Se  6OH → 2Se2    SeO 2
                                                 3               3H2O          (4.6)
      2Al    SeO 2
                 3      3H2O 2OH → 2Al(OH) 4 Se2                               (4.7)
                     Se2
                            Zn(OH) 2 → ZnSe 4OH
                                   4                                           (4.8)
Since the presence of the Al was essential to obtain films, Reaction (47) or other
reactions, possibly involving reduction of Se species with nascent hydrogen
(formed by the dissolution of Al in the NaOH solution), were probably the im-
portant steps in the deposition.
      Another method of ZnSe deposition is not true CD but is related and worthy
of mention. H2Se, prepared by decomposition of CdSe with HCl, was passed over
an aqueous solution of zinc acetate [146,147]. The thin ZnSe film (ca. 50–100 nm
thick) formed at the gas–solution interface could be lifted up and placed on any
desired substrate. The films deposited at 80°C were found to be zincblende ZnSe
by both XRD and ED (no XRD pattern was observed for films deposited at 2°C).
The optical bandgap was 2.62 eV, and resistivity was 107 -cm.


4.6 HgS
HgS possesses a very low value of Ksp (6 10 53) and therefore is expected to be
relatively simple to deposit. In fact, apart from some relatively early literature on
ternary mercury sulphides with lead [148,149] and cadmium [150], which will be
discussed in Chapter 8, only three separate studies on CD HgS have been found.
       Perakh and Ginsburg [151] deposited HgS films using two different tech-
niques. One was a standard CD method using thiourea and HgCl2 complexed with
iodide (iodide is a strong complexant for Hg2 ) in an alkaline solution. The other
technique was simply reaction of a low concentration (ca. 2 mM) of HgCl2 with
(at least three times the Hg concentration) Na2S solution, which precipitates HgS
as a colloid. HgS deposited slowly over many hours. It is interesting that while
some film deposition is expected by this second method, thicknesses of at least 0.7
  m were obtained—much thicker than would intuitively be expected. The tem-
perature dependence of the growth (at least for the thiourea method) was compli-
cated and depended on other parameters. Optimum temperatures were around
room temperature; temperatures higher than 25°C resulted in rough films. The ab-
sorption spectra of both types of films were rather strange—a gradual absorption
onset at somewhat less than 700 nm (ca. 1.8 eV) and a sharp onset at 400 nm (3.1
eV). HgS occurs in (at least) two forms: red (distorted rocksalt, cinnabar—the sta-
ble form at normal temperatures) with a bandgap of ca. 2.0 eV and a black form



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
(zincblende) with a bandgap of ca. 0.5 eV (although large variations of this value
have been reported—see Hg ternaries in Chap. 8). On precipitation of a solution
of Hg2 ions with sulphide solution, a black precipitate is formed (it is not clear
if this is the same as the zincblende form), which eventually should convert to the
more stable red form, although this change might take a long time. From the ab-
sorption spectra of the films given in this study, with no absorption at wavelengths
greater than 700 nm, the films were clearly not of the black form. The sharp onset
at 3.1 eV could conceivably be due to absorption by very small crystals However,
it is more likely that it is a higher transition of HgS, since HgS that is not quan-
tized also shows this feature (see later). The gradual onset from ca. 1.8 nm could
be due to absorption in a distribution of larger sizes also containing a low con-
centration of “black” heavily quantized HgS (low concentration since the absorp-
tion below 2 eV is very weak). It must be stated that no structural characterization
of these films was made; however, it is unlikely that anything other than HgS was
formed under the deposition conditions.
       The other studies used thiosulphate as a sulphur source. One was carried out
in a simple mixture of HgCl2 and Na2S2O3, presumably under acidic conditions
[152]. Crystal size measured by XRD was reported to range from 3 nm to 8 nm
over a temperature range of 0–85°C (film thickness also varied over this temper-
ature range from 50 to 180 nm), and this size regime was confirmed by TEM. The
bandgap estimated from optical spectra varied monotonically as a function of de-
position temperature from 2.3–2.4 eV at 0°C to ca. 1.9 eV at 85°C, due to size
quantization. The resistivity of the films varied from 104 to 103 -cm over the
same temperature range and generally decreased strongly with increasing mea-
surement temperature to low values (very approximately 100 -cm) at tempera-
tures over 150°C.
       A similar deposition, only carried out under alkaline conditions (with added
ammonia) at pH 11 and at 65°C, was described [153]. According to this study,
simple mixing of a mercuric salt solution with thiosulphate results in immediate
formation of a black precipitate of HgS. By first treating the Hg2 solution with
aqueous ammonia, a white precipitate formed by the following reaction:
      Hg(NO3)2      2NH3 → (NH2Hg)NO3          NH4NO3                          (4.9)
This precipitate dissolves in thiosulphate to form a thiosulphate complex, which,
in common with other metal thiosulphate complexes, decomposes when heated to
the metal sulphide (see Sec. 3.3.3). Besides direct decomposition of the thiosul-
phate complex, another possibility suggested in this study is formation of sulphide
ion by alkaline hydrolysis of thiosulphate [Eqs. (3.20) to (3.24)] and reaction with
Hg2 to form HgS. The substrates were glass precoated with a very thin film of
CD PbS (presumably this improved adherence and/or homogeneity).
      The films deposited by this process were golden yellow if thin (ca. 100 nm)
and became red (the normal cinnabar color) if thick (ca. 500 nm). The terminal



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
thickness of the films was 80–90 nm; thicker films were obtained by repeated
depositions.
       X-ray diffraction showed that the deposit (both as a film and the precipitated
material) was cinnabar HgS. The diffraction peaks were sharp (crystal size at least
20 nm, possibly much greater).
       Optical transmission spectra showed a sharp absorption at ca. 400 nm (3.1
eV) and a weaker one at 600 nm (2.0–2.1 eV), the latter probably correspond-
ing to the HgS bandgap, since little or no size quantization should be observed in
these films due to the relatively large crystal size. However, there was a gradual
loss in transmission to beyond 800 nm, which was more pronounced for thicker
films. It is not clear to what extent this is due to scattering (the thicker films were
reported to be less reflecting than the thinner ones, implying increased scattering)
or to absorption.
       The resistivity of the films was greater than 104 -cm (calculated from a re-
ported sheet resistivity of 109 and an assumed thickness of 100 nm).


4.7 HgSe
Three studies in total have been made on CD HgSe. In the earliest, selenosemi-
carbazide was used as a Se source and the Hg was complexed with iodide [99].
The deposition was carried out at 20°C. No details of the films were given other
than that they were specular.
       In the other two studies, selenosulphate was used. In one, a formamide com-
plex of Hg, made by dissolving HgO in formamide, was used [154]. The solution
was made ca. 0.5 M in NaOH, and a trace of polyvinyl pyrollidone was added. The
deposition was carried out at room temperature. The polyvinyl pyrollidone slowed
the deposition somewhat and apparently improved film uniformity and adherence
as well as slightly increased terminal thickness (500 nm). It was noted that films
were not obtained with the usual complexants, such as ammonia, triethanolamine,
and cyanide. It is not mentioned in which way these complexants were unsuitable;
ammonia and triethanolamine might be too weak, resulting in immediate precipi-
tation in solution. Also, addition of ammonia to some mercuric salts tends to lead
to precipitation of insoluble products. Cyanide, however, is a very strong com-
plexant and would be expected to control such bulk precipitation better than for-
mamide. Iodide, a strong complex for Hg2 (and successfully used to deposit HgS,
as described earlier), resulted in film deposition but with poor reproducibility.
       No XRD pattern was found for the films, and on this basis they were be-
lieved to consist of amorphous HgSe. Based on more recent XRD studies of
nanocrystalline materials, the lack of an XRD pattern was likely due to very small
crystal size (supported by the increased bandgap; see later). Annealing at 200°C
“crystallized” the HgSe to an extent that it was clearly identified by XRD. Opti-
cal spectroscopy gave a bandgap value of 1.42 eV. Bulk HgSe is a semimetal with



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
a bandgap of -0.15 eV (i.e., overlapping valence and conduction bands). This sug-
gests very strong size quantization (see HgSe in Chap. 10). The room-temperature
resistivity was ca. 300 -cm.
       In the other study [155], ammonia-complexed Hg(NO3)2 was mixed with
the selenosulphate solution. As for the corresponding HgS deposition, a white pre-
cipitate formed on addition of ammonia to the Hg(NO3)2 [Eq. (4.9)]. This precip-
itate dissolved partly in the excess ammonia used, due to formation of various am-
mine complexes, and completely when the selenosulphate solution was added,
due to additional formation of selenosulphate (and maybe sulphite from the excess
sulphite in the selenosulphite solution) complexes. It is likely that mixed ammine-
selenosulphate/sulphite complexes were formed. The deposition was carried out
on polyester substrates (the transparencies used in overhead projectors) at 10°C.
Deposition occurred over several hours to a terminal thickness of ca. 250 nm. Bulk
precipitation occurred in parallel with the deposition, suggesting that the cluster
mechanism was dominant.
       X-ray diffraction of the powder precipitated in solution confirmed it to be
HgSe (sphalerite). The spectrum of the film showed a strong (111) peak and vir-
tually no other reflection, suggesting a high degree of texture of these films. From
the peak broadening, a crystal size of 7.7 nm was calculated.
       From optical spectra, a bandgap of 2.5 eV was estimated (based on an indi-
rect gap), and this increase from the negative bandgap of bulk HgSe (see earlier)
was attributed to size quantization.
       The sheet resistivity was measured to be 13 k -cm 2. Although the film
thickness was not given, based on a thickness of 250 nm, this translates into a spe-
cific resistivity of 1 -cm. Annealing the films at low temperatures (ca. 100°C)
results in a decrease in resistivity up to as much as an order of magnitude (the crys-
tal size, as measured by XRD, increases only slightly via this treatment).



REFERENCES
  1.   D Lincot, M Froment, H Cachet. In: RC Alkire, DM Kolb, eds. Adv. Electrochem.
       Sci. Eng., Vol. 6. New York: Wiley-VCH, 1998, p 165.
  2.   I Kaur, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:943, 1980.
  3.   D Lincot, R Ortega-Borges, M Froment. Philos. Mag. B 68:185, 1993.
  4.   T Nakanishi, K Ito. Sol. Energy Mater. Sol. Cells 35:171, 1994.
  5.   M Rami, E Benamar, M Fahoume, F Chraibi, A Ennaoui. Solid State Sci. 1:179, 1999.
  6.   P O’Brien, T Saeed. J. Crystal Growth 158:497, 1996.
  7.   MTS Nair, PK Nair, RA Zingaro, EA Meyers. J. Appl. Phys. 75:1557, 1994.
  8.   DS Boyle, P O’Brien, DJ Otway, O Robbe. J. Mater. Chem. 9:725, 1999.
  9.   ML Breen, JT Woodward, DK Schwartz, AW Apblett. Chem. Mat. 10:710, 1998.
 10.   S Gorer, G Hodes. J. Phys. Chem. 98:5338, 1994.
 11.   S Kohle, SK Kulkarni, AS Nigavekar, SK Sharma. Sol. Energy Mater. 10:47, 1984.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
12.   ME Özsan, DR Johnson, M Sadeghi, D Sivapathasundaram, D Lincot, B Mokili,
      M Froment, J Vedel, LM Peter, G Goodlet, RC Walker. In: 13th ECPV Solar En-
      ergy Conf. Nice, France, 1995, p 2115.
13.   MA Martínez, C Guillén, J Herrero. Appl. Surf. Sci. 140:182, 1999.
14.   R Castro-Rodríguez, AI Oliva, V Sosa, F Caballero-Briones, JL Peña. Appl. Surf.
      Sci. 161:340, 2000.
15.   NR Pavaskar, CA Menezes, APB Sinha J. Electrochem. Soc. 124:743, 1977.
16.   TL Chu, SS Chu, N Schultz, C Wang, CQ Wu. J. Electrochem. Soc. 139:2443, 1992.
17.   PC Rieke, SB Bentjen. Chem. Mat. 5:43, 1993.
18.   PN Gibson, ME Özsan, D Lincot, P Cowache, D. Summa. Thin Solid Films 361:34,
      2000.
19.   A Mondal, TK Chaudhuri, P Pramanik. Sol. Energy Mater. 7:431, 1983.
20.   KL Narayanan, R Rajaraman, MC Valsakumar, KGM Nair, KP Vijayakumar. Mat.
      Res. Bull. 34:1729, 1999.
21.   WO Milligan. J. Phys. Chem. 38:797, 1934.
22.   NG Dhere, DL Waterhouse, KB Sundaram, O Melendez, NR Parikh, B Patnaik.
      J. Mater. Sci. 6:52, 1995.
23.   PJ Sebastian. Appl. Phys. Lett. 62:2956, 1993.
24.   PJ Sebastian, M Ocampo. J. Appl. Phys. 77:4548, 1995.
25.   D Petre, I Pintilie, E Pentia, T Botila. Mater. Sci. Eng. B. 58:238, 1999.
26.   M Ichimura, F Goto, E Arai. J. Appl. Phys. 85:7411, 1999.
27.   A Arias-Carbajal Reádigos, VM García, O Gomezdaza, J Campos, MTS Nair,
      PK Nair. Semicond. Sci. Technol. 15:1022, 2000.
28.   MTS Nair, PK Nair, J Campos. Thin Solid Films 161:21, 1988.
29.   O Vigil, Y Rodriguez, O Zelaya-Angel, C Vazquez-Lopez, A Morales-Acevedo,
      JG Vazquez-Luna. Thin Solid Films 322:329, 1998.
30.   LP Deshmukh, SG Holikatti, BM More. Mater. Chem. Phys. 39:278, 1995.
31.   JM Gracia-Jiménez, G Martínez-Montes, R Silva-González. J. Electrochem. Soc.
      139:2048, 1992.
32.   KK Nanda, SN Sarangi, S Mohanty, SN Sahu. Thin Solid Films 322:21, 1998.
33.   R Lozada-Morales, O Zelaya-Angel. Thin Solid Films 281–282:386, 1996.
34.   PK Nair, MTS Nair, J Campos, LE Sansores. Solar Cells 22:211, 1987.
35.   JH Schön, O Schenker, B Batlogg. Thin Solid Films 385:271, 2001.
36.   PK Nair, J Campos, MTS Nair. Semicond. Sci. Technol. 3:134, 1988.
37.   K Ito, K Shiraishi. Sol. Energy Mater. Sol. Cells 35:179, 1994.
38.   WJ Danaher, LE Lyons, GC Morris. Sol. Energy Mater. 12:137, 1985.
39.   C Guillén, MA Martínez, J Herrero. Thin Solid Films 335:37, 1998.
40.   M Nagao, S Watanabe. Jpn. J. Appl. Phys. 7:684, 1968.
41.   IO Oladeji, L Chow. J. Electrochem. Soc. 144:2342, 1997.
42.   O Vigil, O Zelaya-Angel, Y Rodriguez, A Morales-Acevedo. Phys. Status Solidi (a)
      167:143, 1998.
43.   JA Akintunde. Phys. Status Solidi (a) 179:363, 2000.
44.   PK Nair, MTS Nair, VM Garcia, OL Arenas, Y Pena, A Castillo, IT Ayala, O
      Gomezdaza, A Sanchez, J Campos, H Hu, R Suarez, ME Rincon. Sol. Energy Mater.
      Sol. Cells 52:313, 1998.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
45.   PK Nair, MTS Nair, J Campos. In: CM Lampert, ed. Proc. SPIE, San Diego, CA,
      1987, Vol. 823, p 256.
46.   PK Nair, MTS Nair, J Campos. Sol. Energy Mater. 15:441, 1987.
47.   IO Oladeji, L Chow, JR Liu, WK Chu, ANP Bustamante, C Fredricksen,
      AF Schulte. Thin Solid Films 359:154, 2000.
48.   L Pintilie, E Pentia, I Pintilie, D Petre. Mater. Sci. Eng. B 44:403, 1997.
49.   M Froment, MC Bernard, R Cortes, B Mokili, D Lincot. J. Electrochem. Soc.
      142:2642, 1995.
50.   AG Shikalgar, SH Pawar. Sol. State Commun. 32:361, 1979.
51.   AG Shikalgar, SH Pawar. Thin Solid Films 61:313, 1979.
52.   CD Lokhande, SH Pawar. Sol. State Commun. 44:1137, 1982.
53.   JA Akintunde. J. Mater. Sci.-Mater. Electron. 11:503, 2000.
54.   AI Oliva, R Castro-Rodriguez, O Ceh, P Bartolo-Perez, F Caballero-Briones,
      V Sosa. Appl. Surf. Sci. 148:42, 1999.
55.   D Bhattacharyya, MJ Carter. Thin Solid Films 288:176, 1996.
56.   D Lincot, R Ortega-Borges, M Froment. Appl. Phys. Lett. 64:569, 1994.
57.   R Cortes, M Froment, B Mokili, D Lincot. Philos. Mag. Letts. 73:209, 1996.
58.   D Lincot, B Mokili, R Cortes, M Froment. Microsc. Microanal. Microstruct. 7:217,
      1996.
59.   MJ Furlong, D Lincot, M Froment, R Cortés, AN Tiwari, M Krejci, H Zogg. In: 14th
      ECPV Solar Energy Conf. Barcelona, Spain, 1997, p 1291.
60.   MJ Furlong, M Froment, MC Bernard, R Cortés, AN Tiwari, M Krejci, H Zogg,
      D Lincot. J. Crystal Growth 193:114, 1998.
61.   YK Hwang, SY Woo, JH Lee, DY Jung, YU Kwon. Chem. Mat. 12:2059, 2000.
62.   RL Call, NK Jaber, K Seshan, JR Whyte Jr. Sol. Energy Mater. 2:373, 1980.
63.   GA Kitaev, SG Mokrushin, AA Uritskaya. Kolloid Zh. 27:51, 1965.
64.   R Ortega-Borges, D Lincot. J. Electrochem. Soc. 140:3464, 1993.
65.   V Polpescu, EM Pica, I Pop, R Grecu. Thin Solid Films 349:67, 1999.
66.   K Yamaguchi, T Yoshida, T Sugiura, H Minoura. J. Phys. Chem. B 102:9677, 1998.
67.   K Yamaguchi, T Yoshida, N Yasufuku, T Sugiura, H Minoura. Electrochem.
      67:1168, 1999.
68.   PK Nair, MTS Nair. Sol. Energy Mater. 15:441, 1987.
69.   V Weiss, AA Friesem, A Peled. Thin Solid Films 218:193, 1992.
70.   F Goto, M Ichimura, E Arai. Jpn. J. Appl. Phys. 2 36:L1146, 1997.
71.   M Ichimura, F Goto, E Arai. J. Electrochem. Soc. 146:1028, 1999.
72.   JGea Vázquez-Luna. J. Crystal Growth 187:380, 1998.
73.   M Gros-Jean, R Herino, D Lincot. J. Electrochem. Soc. 145:2448, 1998.
74.   DS Boyle, A Bayer, MR Heinrich, O Robbe, P O’Brien. Thin Solid Films 361:150,
      2000.
75.   D Hariskos, M Powalla, N Chevaldonnet, D Lincot, A Schindler, B Dimmler. Thin
      Solid Films 387:179, 2001.
76.   A Kylner, J Lindgren, L Stolt. J. Electrochem. Soc. 143:2662, 1996.
77.   A Kylner. J. Electrochem. Soc. 146:1816, 1999.
78.   A Kylner, M Wirde. Jpn. J. Appl. Phys. 1 36:2167, 1997.
79.   DW Niles, G Herdt, M AlJassim. J. Appl. Phys. 81:1978, 1997.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 80. Y Hashimoto, T Nakanishi, T Andoh, K Ito. Jpn. J. Appl. Phys. 2 34:L382, 1995.
 81. T Nakada, H Fukuda, A Kunioka, S Niki. In: 13th ECPV Solar Energy Conf. Nice,
     France, 1995, p 1597.
 82. R Jayakrishnan, JP Nair, BA Kuruvilla, SK Kulkarni, RK Pandey. J. Mater. Sci.
     7:193, 1996.
 83. GA Kitaev, TS Terekhova. Zh. Neorg. Khim. 15:48, 1970.
 84. RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:277, 1980.
 85. RA Boudreau, RD Rauh. J. Electrochem. Soc. 130:513, 1983.
 86. GA Kitaev, AA Uritskaya, LE Yatlova, TS Terekhova, TI Dzyuba. Zhurn. Neorg.
     Khim. 35:3072, 1990.
 87. G Hodes, A Albu-Yaron, F Decker, P Motisuke. Phys. Rev. B 36:4215, 1987.
 88. S Gorer, G Hodes, Y Sorek, R Reisfeld. Mater. Lett. 31:209, 1997.
 89. F Trojanek, R Cingolani, D Cannoletta, D Mikes, P Nemec, E Uhlirova, J Rohovec,
     P Maly. J. Cryst. Growth 209:695, 2000.
 90. G Hodes. Isr. J. Chem. 33:95, 1993.
 91. R Garuthara, G Levine. J. Appl. Phys. 80:401, 1996.
 92. P Nemec, D Mikes, J Rohovec, E Uhlirova, F Trojanek, P Maly. Mater. Sci. Eng.
     B 69:500, 2000.
 93. M Perakh, A Peled. Thin Solid Films 50:293, 1978.
 94. H Cachet, H Essaaidi, M Froment, G Maurin. J. Electroanal. Chem. 396:175, 1995.
 95. KC Mandal, O Savadogo. J. Mater. Sci. Lett. 10:1446, 1991.
 96. R Lozada-Morales, M Rubin-Falfan, O Portillo-Moreno, et al. J. Electrochem. Soc.
     146:2546, 1999.
 97. MTS Nair, PK Nair, H Pathirana, RA Zingaro, EA Meyers. J. Electrochem. Soc.
     140:2987, 1993.
 98. BK Rai, HD Bist, RS Katiyar, MTS Nair, PK Nair, A Mannivannan. J. Appl. Phys.
     82:1310, 1997.
 99. AA Velykanov, EK Ostrovskaya, NP Garina, VA Turacova, AA Tchurkan. Ukr.
     Chim. Zh. 49:764, 1983.
100. H Cachet, R Cortés, M Froment, G Maurin, N Shramchenko. J. Electrochem. Soc.
     144:3583, 1997.
101. E Lifshitz, I Dag, I Litvin, G Hodes, S Gorer, R Reisfeld, M Zelner, H Minti. Chem.
     Phys. Lett. 188, 1998.
102. XC Ai, R Jin, CB Ge, JJ Wang, YH Zou, XW Zhou, XR Xiao. J. Chem. Phys.
     106:3387, 1997.
103. P Maly, J Kudrna, F Trojanek, D Mikes, P Nemec, AC Maciel, JF Ryan. Appl. Phys.
     Lett. 77:2352, 2000.
104. K Rajeshwar, L Thompson, P Singh, RC Kainthla, KL Chopra. J. Electrochem. Soc.
     128:1744, 1981.
105. RC Kainthla, DK Pandya, KL Chopra. Sol. State Electron. 25:73, 1982.
106. VM Garcia, MTS Nair, PK Nair, RA Zingaro. Semicond. Sci. Technol. 11:427, 1996.
107. MTS Nair, PK Nair, RA Zingaro, EA Meyers. J. Appl. Phys. 74:1879, 1993.
108. G Hodes, IDJ Howell, LM Peter. In: Photochemical and Photoelectrochemical Con-
     version and Storage of Solar Energy. ZW Tian, Y Cao, eds. Int. Acad. Publishers,
     Beijing, China. 1993, p 331.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
109.    L Kronik, L Burstein, M Leibovitch, Y Shapira, D Gal, E Moons, J Beier, G Hodes,
        D Cahen, D Hariskos, R Klenk, HW Schock. Appl. Phys. Lett. 67:1405, 1995.
110.    E Lifshitz, I Dag, I Litvin, G Hodes. J. Phys. Chem. B 102:9245, 1998.
111.    B Alperson, I Rubinstein, G Hodes. Phys. Rev. B 6308:1303, 2001.
112.    GK Padam, SK Gupta. Appl. Phys. Lett. 53:865, 1988.
113.    RW Buckley. In: 11th ECPV Solar Energy Conf. Montreux, Switzerland, 1992, p
        962.
114.    VB Patil, PD More, DS Sutrave, GS Shahane, RN Mulik, LP Deshmukh. Mater.
        Chem. Phys. 65:282, 2000.
115.    AV Kalyakina, RI Pelyukpashidi. Tr. Khim. Met. Inst. Akad. Nauk. Kaz. SSR
        17:114, 1973.
116.    M Sotelo-Lerma, RA Zingaro, SJ Castillo. J. Organomet. Chem. 623:81, 2001.
117.    P O’Brien, J McAleese. J. Mater. Chem. 8:2309, 1998.
118.    B Mokili, M Froment, D Lincot. J. de Phys. IV 5:261, 1995.
119.    P O’Brien, DJ Otway, D Smith-Boyle. Thin Solid Films 361:17, 2000.
120.    JM Doña, J Herrero. J. Electrochem. Soc. 141:205, 1994.
121.    IO Oladeji, L Chow. Thin Solid Films 339:148, 1999.
122.    B Mokili, Y Charreire, R Cortes, D Lincot. Thin Solid Films 288:21, 1996.
123.    J McAleese, P O’Brien. In Mater. Res. Soc. Symp. Proc., 1998; Vol. 485; p 255.
124.    IC Ndukwe. Sol. Energy Mater. Sol. Cells 40:123, 1996.
125.    S Biswas, P Pramanik, PK Basu. Mater. Lett. 4:81, 1986.
126.    R Ortega Borges, D Lincot, J Vedel. In: 11th ECPV Solar Energy Conf., Montreux,
        Switzerland, 1992, p 862.
127.    PK Nair, MTS Nair. Semicond. Sci. Technol. 7:239, 1992.
128.    PK Nair, MTS Nair, O Gomezdaza, RA Zingaro. J. Electrochem. Soc. 140:1085, 1993.
129.    OL Arenas, MTS Nair, PK Nair. Semicond. Sci. Technol. 12:1323, 1997.
130.    K Yamaguchi, T Yoshida, T Sugiura, H Minoura. J. Mater. Res. 13:917, 1998.
131.    CD Lokhande. Mater. Chem. Phys. 28:145, 1991.
132.    M Ichimura, F Goto, Y Ono, E Arai. J. Cryst. Growth 199:308, 1999.
133.    GA Kitaev, AA Uritskaya, LE Yatlova, VR Mirolyubov. Russ. J. Appl. Chem.
        67:1415, 1994.
134.    QW Chen, YT Qian, ZY Chen, L Shi, XG Li, GE Zhou, YH Zhang. Thin Solid Films
        272:1, 1996.
135.    QW Chen, XG Li, Y Qian, JS Zhu, G Zhou, WP Zhang, YH Zhang. Appl. Phys.
        Lett. 68:3582, 1996.
136.    P Pramanik, S Biswas. J. Electrochem. Soc. 133:350, 1986.
137.    RK Nkum, AA Adimado, H Totoe. Mater. Sci. Eng. B 55:102, 1998.
138.    GA Kitaev, TP Sokolova. Russ. J. Inorg. Chem. 15:167, 1970.
138a.   GA Kitaev, TP Sokolova. USSR Patent 356, 319 1972.
139.    CD Lokhande, PS Patil, A Ennaoui, H Tributsch. Appl. Surf. Sci. 123:294, 1998.
140.    JM Doña, J Herrero. J. Electrochem. Soc. 142:764, 1995.
141.    AM Chaparro, MA Martinez, C Guillen, R Bayon, MT Gutierrez, J Herrero. Thin
        Solid Films 361:177, 2000.
142.    A Ennaoui, M Weber, M Saad, W Harneit, MC Lux-Steiner, F Karg. Thin Solid
        Films 361:450, 2000.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
143.   J Herrero, MT Gutierrez, C Guillen, JM Dona, MA Martinez, AM Chaparro,
       R Bayon. Thin Solid Films 361:28, 2000.
144.   CA Estrada, PK Nair, MTS Nair, RA Zingaro, EA Meyers. J. Electrochem. Soc.
       141:802, 1994.
145.   C Wang, XF Qian, WX Zhang, XM Zhang, Y Xie, YT Qian. Mat. Res. Bull.
       34:1637, 1999.
146.   GN Chaudhari, SD Sathaye, P Singh, VJ Rao, V Manorama, RS Bhide, SV Bho-
       raskar. J. Mater. Sci. Lett. 11:1097, 1992.
147.   GN Chaudhari, S Manorama, VJ Rao. J. Phys. D:Appl. Phys. 25:862, 1992.
148.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Mat. Res. Bull. 11:1109, 1976.
149.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 42:383, 1977.
150.   M Skyllas-Kazacos, JF McCann, R Arruzza. Appl. Surf. Sci. 22/23:1091, 1985.
151.   M Perakh, H Ginsburg. Thin Solid Films 52:195, 1978.
152.   SS Kale, CD Lokhande. Mater. Chem. Phys. 59:242, 1999.
153.   M Najdoski, I Grozdanov, SK Dey, B Siracevska. J. Mater. Chem. 8:2213, 1998.
154.   P Pramanik, S Bhattacharya. Mat. Res. Bull. 24:945, 1989.
155.   B Pejova, M Najdoski, I Grozdanov, SK Dey. J. Mater. Chem. 11:2889, 1999.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
5
PbS and PbSe




5.1 HISTORICAL BACKGROUND
PbS holds the honor of being the first reported compound to be deposited by CD.
In 1869, Puscher described a “new and cheap process, without using dyes, to coat
various metals with splendid lustrous colors” [1]. This involved deposition from
a thiosulphate solution of lead acetate (and also, in the same paper, from Cu and
Sb salts to give presumably corresponding sulphides). These shiny, colored coat-
ings prompted further studies in this process, both to expand the process to other
metal sulphides and to understand the process. These studies are discussed in Sec-
tion 5.2.1.
       The common thiourea process for CD was also first used for PbS [2]. The
thiourea method became the preferred one for the emerging development of PbS
photoconductive cells as infrared (IR) detectors during the Second World War.
Obviously much of this early work, carried out by German groups for military pur-
poses, was secretive and was not published at the time. Later, photoconductive
cells using CD PbS (which gave better cells than did the more conventional, evap-
orated PbS films [3]) became commercial as IR detectors, and, together with CD
PbSe, they remain so until this day. An early description of such cells is given by
Kicinski [4]. This application then became the driving force for CD studies, which
were limited almost entirely to PbS and PbSe up to the start of the 1960s. Section
5.2.6 gives a brief overview of the operation of photoconductive cells in general
and lead salt cells in particular.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Since the expansion of CD to II–VI compounds in the 1960s and subse-
quently to other compounds, work on the IV–VI materials has not kept pace with
this expansion, and there has been only a relatively limited amount of work since
then, mainly on the use of PbS for solar control coatings. This chapter does not as-
pire to comprehensively cover the older literature, although it is intended that the
important aspects of these studies be included. Previous reviews that deal with
these materials (mostly reviews on infrared detectors) include Refs. 3 and 5–8.

5.2 PbS
5.2.1 Deposition Using Thiosulphate
The original CD process involves decomposition of a thiosulphate solution of
PbAc2 [1]. From the outset, it must be noted that the mechanism of this deposition
has not been unambiguously elucidated up to the present time. There are two main
possibilities for the reaction.
      1. Decomposition of a thiosulphate complex of Pb, a possible reaction being
      Pb(S2O3) 2
               2       H2O → PbS        S2O 2
                                            3      H2SO4                           (5.1)
      2. Decomposition of       S2O 2
                                    3to give free sulphide ion, which then reacts
with free Pb2 . It has been suggested [9,10] that this can occur by reduction of el-
emental S, which forms in acidic solutions of S2O 2 3

      S2O 2
          3      H →S         HSO 3                                                (5.2)
by electrons formed in the half cell reaction:
      2S2O 2 → S4O 2
           3       6          2e                                                   (5.3)
(An internal electrochemical mechanism was proposed long ago for deposition on
certain metal substrates, since the rate of deposition sometimes depended on the
nature of the substrate [11].) The standard potential of Reaction (5.3) is 0.08 V,
considerably more positive than the reduction potential of S to S2 ( 0.45 V).
Free sulphide, if formed, would be in a very low concentration, since it will be re-
moved continually by precipitation of PbS; this will move the S reduction poten-
tial strongly positive according to the Nernst equation [Eq. (1.32)]. This positive
shift will be even greater than normal because of the non-Nernstian behavior of
the S2 /S couple when [S] [S2 ] (at least in alkaline solution) [12]. In opposi-
tion to this, the solubility of S in the (slightly acidic) aqueous solutions is very low,
which will move the potential in the opposite direction. Add to this the very small
concentration of S2 in acid solution [Eq. (1.15)], and it becomes clear that it is
not trivial to estimate the feasibility of the formation of PbS by free sulphide. The
non-Nernstian behavior of the sulphur-rich S2 /S couple and the lack of knowl-
edge of the solubility of free S in the deposition solution are the two factors that
complicate what would have been a tractable thermodynamic calculation.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      The decomposition of metal thiosulphates to the corresponding sulphide is
well known and is generally assumed to occur by breaking of the S–S bond in thio-
sulphate. Thus, it has been shown, by a radioactive tracer method, that if thiosul-
phate is prepared by dissolving radioactive S in sulphite solution:
      S*    SO 2 D S*MSO 2
               3         3                                                     (5.4)
and this solution is reacted with Ag to form a precipitate of Ag2S2O3, then the
AgS formed on boiling this precipitate in water contains only labeled S, showing
that the metal bonds only to the S which is not attached to oxygen [13]. Admit-
tedly, even if the AgS was formed by free sulphide, the same result would be prob-
able due to the lability of the SMS bond; however, such reactions have been stud-
ied for many metal cations, and it is invariably assumed (and this does not
necessarily make the assumption correct) that the reaction occurs by simple fis-
sion of the SMS bond in the thiosulphate. Additionally, since this reaction occurs
most readily for metals with a high affinity for S (Ag, Cu(I), Hg and Pb), there is
no reason to expect that the MMS bond present in the thiosulphate (or complex
thiosulphate) would be broken, as would be the case were the formation of metal
sulphide to occur by free sulphide, although, as has already been pointed out, these
same metal sulphides have a very low-solubility product and therefore very little
sulphide is required to precipitate them.
       One potentially useful piece of information that can be explained more read-
ily based on a free-sulphide generation comes from an early study on the forma-
tion of PbS by boiling Pb2 and thiosulphate in water, when it was found that PbS
formed much more readily when excess thiosulphate was present [14].
       To sum up, the mechanism of formation of PbS using thiosulphate is still not
defined, either in general or even for any specific case. Some dedicated research
to solve this question is clearly required.
       While a number of early studies described PbS (and other sulphides) for-
mation from thiosulphate solutions [1,11,15,16], these early studies provided lit-
tle characterization of the films other than the interference colors obtained (due to
different thicknesses of the films). There appear to be only two modern studies
that provide some general characterization of these PbS films. In one [10], films
were deposited on glass at 80°C from a solution of PbAc2 and Na2S2O3 at a pH
between 5 and 6. X-ray diffraction (XRD) showed sharp peaks of polycrystalline
PbS. Although the thickness of the films was not given, they were clearly thick by
CD standards, since the optical transmission over the range from 400 to 2000 nm
was less than 1%. Electrical resistivity was measured to be ca. 105–106 -cm. In
another study [17] (also by the same group on Cu2S-coated glass [18]), the condi-
tions were similar (the pH range was between 4 and 6, and acetic acid was added,
if necessary, to reach this pH range; the bath temperature was slightly lower—ca.
70°C—and the reactant concentrations were lower). The PbS films were thinner
(0.1 m), as could be seen by the greater optical transmission ( 60% at 2.5 m,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
dropping gradually to ca. 7% at 500 nm and more rapidly at shorter wavelengths).
Such semitransparent films are considered useful for solar control coatings (see
Sec. 2.13). The films were p-type with a very low resistance (4 k /sq, equivalent
to 4 10 2 -cm). The large difference between the resistivities in the two stud-
ies seems worthy of further investigation.


5.2.2 Deposition Using Thiourea
With the exception of the few cases just described, all the modern (and most of the
earlier) studies on PbS deposition were carried out using thiourea as a source of
sulphide, as first described by Emerson-Reynolds in 1884 [2]. In the original
study, the lead was present as a strongly alkaline solution of lead tartrate (proba-
bly a mixed tartrate–hydroxide complex of lead). When this solution was heated
with a solution of thiourea, “at about 50° a specular layer forms at the bottom and
sides of the vessel. When the beaker is thoroughly clean in the first instance, the
adhesion is uniform and strong.” Most of the PbS formed as a precipitate. The
competing effects of temperature were noted: At room temperature the deposition
was very slow (days) and the film tended to be less even (more heterogeneous to
the eye), while at higher temperatures thin (due to excessive precipitation) brown-
ish films were obtained. Deposition was reported on a number of different mate-
rials (including porcelain and ebonite as well as iron and steel, although films on
the metals had poorer adhesion).
       The same basic technique was used to study the deposition of PbS on dif-
ferent types of glass substrates. The quality of the film varied greatly from very
poor, partially formed films that were not adherent (quartz and borosilicate glass),
through occasionally good but irreproducible films (on sodalime glass), to homo-
geneous, adherent films on flint (lead containing) and Zn-containing (crown)
glasses [19]. The good adherence on the Pb and Zn glasses was explained by for-
mation of insoluble sulphides by these metals; such sulphides would form a good
binding site for further deposition of PbS (presumably also for other compounds).
       Lead acetate was later usually employed as the Pb salt (e.g., Refs. 4 and 20).
The thiourea acts not only as a source of S but also as a complexing agent, as
shown by Kicinski. In the absence of thiourea, hydroxide can also complex Pb2 ,
but a considerable excess is usually required in practice. Small amounts of lead
oxide and hydroxide were detected in these films by XRD [4].
       There does not seem to be a clear consensus as to the mechanism of PbS
formation in these and similar studies. It is often stated that the formation of PbS
occurs via decomposition of a Pb–thiourea complex species [4,21,22]. This was
often based on the absence of any measurable concentration of sulphide on alka-
line hydrolysis of thiourea. However, as discussed in Section 3.3.3.1, this is not a
valid criterion for the absence of a sulphide-mediated reaction. Even today, it can-
not be stated categorically which mechanism is operative or even dominant.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       A microscopic study of the PbS formed in these films deposited on glass at
50°C showed the formation of large (0.2–1 m) cubic crystals from relatively
concentrated solutions (the films were specularly reflecting) and somewhat larger,
but less well-defined crystals from more dilute solutions (resulting in more highly
scattering films) [23]. Addition of very small amounts of CuSO4 to the dilute de-
position solution resulted in specularly reflecting films with smaller, more evenly
sized crystals. From XRD line broadening, the crystal size of PbS deposited by the
method described by Kiciniski was estimated at ca. 70 nm [24]. Another report us-
ing films prepared by Kicinski on pyrex described a preferred orientation with the
(001) faces parallel to the glass surface and a crystal size larger than 15 nm and in
some cases ca. 50 nm [25].

5.2.3 Variations in Deposition
This section deals only with thiourea-based baths. There is little variation in the
thiosulphate baths that have been reported.

5.2.3.1 NH4OH (in place of NaOH or KOH)
Ammonium hydroxide has also been used in place of NaOH or KOH [20,24,26].
In reported contrast to films deposited from alkali metal hydroxide, these films,
prepared at or slightly above room temperature, were photoconductive (photosen-
sitivity ca. 10) as deposited without need for air-annealing [26]. The crystal size
of films deposited at different temperatures was measured (XRD line broadening)
to be 10–15 nm (30°C), 17 nm (40°C), and 39 nm (50°C) [24,26]. The presence
of strain in the crystals was inferred from the same XRD measurements [24].
5.2.3.2 Addition of Halides
A comprehensive study has been made on the effect of added ammonium halides
to deposition from solutions of citrate-complexed Pb2 containing NH4OH [27].
The effects of the ammonium halides were both strong and varied. The deposition
rate decreased with increased halide concentration. This is expected, if only due
to the lower pH of the buffered solution. Also, the retardation effect of the am-
monium halide increased from Cl to I. This could be due to partial removal of Pb
by the sparingly soluble halides (the iodide is the least soluble and therefore will
most effectively remove the Pb). It is also possible that the more strongly adsorbed
iodide retards growth due to adsorption and surface capping. This is supported by
the gradual decrease in crystal size (measured by electron microscopy) from ca.
0.8 m (no halide) through 0.5 m (Cl), 0.3 m (Br), and 0.2 m (I), which par-
allels the increase in adsorption power of the halide ions from Cl to I. On the other
hand, substitution of NH4I by KI or NaI does not have the same effect (the crys-
tal size remains ca. 0.8 m). The films were preferentially (111) textured in the
absence of ammonium halide, and this texture decreased as the ammonium halide



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
concentration, as well as its atomic weight, increased, again as expected based on
adsorption of halide; but again this texture-reducing effect did not occur (at least
not to the same extent) with KI or NaI. Other effects of the different halides, given
only for the ammonium halides (i.e., it was not known whether NaI or KI had a
similar effect or not), were change in shape of the crystals from cubic (Cl) to
spherical (I) (which could also be explained by strong adsorption of iodide on all
crystal faces) and an energy of activation of the deposition varying from 38 (Cl)
to 67 (I) kJ/mole; Both are characteristic of a chemical rate-determining step, but
it is clear that there is an important difference in the two mechanisms. It seems
that, while anion adsorption may play a role in these effects, the main role, at least
in those effects where ammonium and alkali metal halides were compared, has an-
other explanation. A possible one, in particular for the effect on deposition rate, is
the lower pH of the ammonium-buffered solutions.
       Another property of the PbS films deposited from the ammonium-buffered
solutions was the relatively high photosensitivity obtained for the as-deposited
films, particularly from the Cl bath. Increases in photosensitivity of close to two
orders of magnitude were obtained, compared to films deposited from solutions
containing no ammonium halide. It was suggested that the halide compensated
the PbS (low-temperature measurements showed quasi-intrinsic conduction in
these films), resulting in high photosensitivity. Additionally, there is a general
inverse relation between the photosensitivity and the grain size/degree of textur-
ing. This relationship between grain size and photosensitivity is often seen and
implies a major role of the grain surfaces in the photoconduction mechanism.
Note that this is in contrast to photoconductivity in the II–VI semiconductors,
where there was no obvious correlation between sensitivity and grain size (see
Chap. 4).
       A related study, using KBr as an additive [28], found a modest increase in
crystal size with increasing KBr concentration together with an increase in crys-
tal height (up to 0.4 mM Br concentration followed by a subsequent decrease with
further increase in KBr concentration) and a slight (200) preferred orientation. It
was suggested that this crystal growth was due to retarded growth of small nuclei
due to the complexing power of the bromide. Note that the pH of this bath was ca.
12, probably higher than the buffered ammonium bath described earlier.
       As with the ammonium halide baths, the photosensitivity of the resulting
PbS films was found to increase with added KBr in the bath (no Br was found
in the layers themselves, although very small “doping” concentrations might not
be detected in the analyses), up to a maximum of 0.4 mM KBr, and then de-
creased, which correlates with the crystal height. The increase in photosensitiv-
ity was attributed to disorder at the grain boundaries. The study using ammo-
nium halides also attributed a major role of the grain surface to the
photoconductivity, but the relation between grain size and sensitivity is very dif-
ferent in the two studies.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
5.2.3.3 Oxidizing Agents in the Bath
PbS films deposited from most basic baths are only weakly photoconducting. For
maximizing photoconductivity response, air-annealing is usually carried out. By
employing an oxidizing agent in the bath, films can be deposited that have rela-
tively high response as deposited (explained by introduction of sensitizing centers
in the PbS; see Sec. 5.2.6). One study [29] has described the affect of an oxidizing
component in the bath on the sensitivity of the resulting films (ca. 1 m thick). Un-
fortunately, neither the bath conditions (probably a Pb2 /hydroxide/thiourea bath)
nor the identity of the oxidizing agent was revealed, other than that they were “vari-
ations of a commercially available material.” The films were all p-type with car-
rier density between 1016 and 1017/cc. The detectivity of a film deposited without
the oxidant was lower (by orders of magnitude at room temperature, less at low
temperatures) than for a film prepared with an excess of oxidant or a standard (pre-
sumed commercial) film. The absorption spectra of the films deposited with and
without oxidant were also quite different (see Sec. 5.2.5). It is notable that this dif-
ference in absorption was not reflected in the photoconductivity response shape
near the response onset; the samples behaved similarly with a photoconductivity
onset of ca. 0.4 eV, although at higher photon energies the response of the oxidant-
free film decreased strongly with increasing photon energy, in contrast to a mod-
est decrease for the oxidant samples. It was also noted that the film with a high ox-
idant concentration appeared more porous (in SEM micrographs) than the
oxidant-free film and that the particle size was ca. 70 nm (somewhat less in the ox-
idant-containing films than in the oxidant-free ones). A similar study on the effect
of (the as-usual unspecified) oxidant on PbS films deposited from a thiourea/hy-
droxide bath found similar effects to the foregoing study, namely, better sensitiv-
ity, smaller particle size (from almost a micron in the absence of oxidant dropping
to 0.2 m with a high oxidant concentration) and more porous with increasing
oxidant concentration [30]. In this case, the sensitivity peaked at a certain oxidant
concentration, followed by a decrease. Also, while the oxidant films were p-type,
those deposited without the oxidant were n-type. The oxygen content was constant
and independent of the oxidant concentration (as found also in the previous study
[29]), and it was concluded that the origin of the active sensitizing centers was re-
lated not to the oxidizing agent itself but to the structure of the film.
       Apart from considerations of photoconductivity, one oxidant, H2O2, was
shown to exert considerable influence on the crystallographic texture of PbS de-
posited from a PbAc2/NaOH solution with thiourea at close to room temperature
[31]. While the (111) reflection was fairly dominant in the absence of H2O2, the
(200) reflection became very dominant in its presence. The H2O2 was added, not
with the other reactants, but some time after film formation became visible, and
the degree of texturing reached a maximum at a certain time, after which it again
decreased. No explanation for these effects was given.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
5.2.3.4 Miscellaneous Variations in Deposition
      Addition of Inhibitors. Addition of salts of Sb, Sn, or As to a basic
PbAc2 /thiourea/OH bath resulted in a slower deposition, typically by 2–4 times
[21]. These metal ions all form complex sulphides in aqueous solution, and it was
believed that they caused “peptization” of growing nuclei; peptization converts
large particles into colloidal ones. Probably what is meant is some sort of capping
action due to adsorption of the retarding ions on the growing PbS crystals, pre-
venting (more correctly, retarding) further growth.
       Illumination-Induced Growth. As for CdS, illumination has been found to
increase the growth rate of CD PbS films [32]. The rate increase was attributed to
a combination of heating by light absorption and the formation of charge carriers
in the PbS film, the latter resulting in activation of the deposition. As for CD of
CdSe under illumination, described in Chapter 4, such activation may result from
either a photochemical reaction by the photogenerated electrons and/or holes on
the PbS surface, as suggested earlier (the most likely scenario), and/or by a change
in the electrical double layer at the surface of the PbS particles (adsorbed on the
surface or as a colloid in the solution near the substrate), which might lower a po-
tential barrier to adsorption and coalescence of the PbS colloids.
      Alkali-Metal-Free Solutions. Films of CD PbS are usually p-type as de-
posited. One early suggestion to explain this was that the alkali metal ions used
in the deposition solution (as NaOH or KOH) act as a p-type dopant [33]. Based
on this supposition, Bloem deposited PbS from a solution of PbAc2, hydrazine
hydrate, and thiourea (free of Na or K). The as-deposited films were initially n-
type but changed to p-type on exposure to air. Attempts to dope the films perma-
nently n-type by adding trivalent ions to the deposition solution were unsuccess-
ful. However, by depositing the films on a substrate coated with trivalent ions
(such as Al, In, Ga), n-type behavior could be maintained for a considerable time.
PbS p-n junctions were fabricated using this approach (see Chap. 9).

5.2.4 Substrate Effects (See also Sec. 5.2.2)
5.2.4.1 Epitaxial Deposition
There have been a few reports on epitaxial deposition of PbS on various single-
crystal substrates. PbS (n-type) was epitaxially deposited on (111) Ge (5.4% mis-
match) from a Pb(NO3)2 /KOH/thiourea solution at room temperature with (111)
orientation [34] (although another study using apparently the same conditions
found the deposit to be p-type and polycrystalline with some (100) preferred ori-
entation [35]). From a similar solution (with addition of some ethanol), PbS was
deposited on single-crystal CdS (ca. 6.6% mismatch) with varying degrees of epi-
taxy [36]. On the (0001) faces of CdS, the growth was (111) [(111) cubic corre-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sponds to (0001) hexagonal] and on the (1120) face, it was (220) oriented. The de-
gree of epitaxy was moderate on the Cd (0001) face, better on the S (0001) face,
and very high on the (1120) face.
       Using the same reactants (but at higher concentrations), also at room tem-
perature, epitaxial PbS growth was observed on (100) InP, (100) Ge, and also
(although with rougher morphology) on (111) Ge [37]. In these experiments, the
films, ca. 300 nm thick, were all p-type. The energy of activation for film forma-
tion in this study was 65 kJ/mole—similar to CdS deposition from thiourea solu-
tions but different from the value of activation energy measured for ZnS deposi-
tion from a thiourea bath (see Sec. 4.4.1). Homogeneous precipitation also
occurred gradually, but the films were removed before it interfered by adhesion of
clusters (presumably before it became excessive).
       PbS films grown on single-crystal GaAs (lattice mismatch ca. 5%) were
polycrystalline or only somewhat oriented [38], and those on Si (9% mismatch)
were polycrystalline [39]. Some of these heterojunctions with CD PbS are dis-
cussed in Chapter 9.
5.2.4.2     Deposition on Monolayers
PbS has been deposited from hydroxide/thiourea solutions onto Au and thiol
monolayer–coated Au substrates [40]. The quality and texturing of the films var-
ied according to the specific substrate used and also on the hydroxide concentra-
tion. For dilute OH solutions (0.1 M OH ; 0.01 M Pb2 ), good deposition was
obtained only on some surfaces, while for more concentrated OH (0.25 M OH ;
0.01 M Pb2 ), the deposition was not very surface dependent. Various degrees of
(111) or (100) texture were obtained, depending on the substrate. On bare (111)
Au, a (100) texture was obtained. It was noted that a high degree of texture was
obtained on some of the monolayers, which themselves were poorly ordered. The
crystal size varied according to both film thickness and other deposition condi-
tions. For very thin films, very small crystal sizes were obtained (from a few
nanometers up). For thick films, relatively large, more or less well-defined cubic
crystals of between 50 and several hundred nanometers were obtained. In general,
the depositions from the more concentrated OH solutions gave larger, better-
defined crystals. The differences were rationalized in terms of hydroxide and ion-
by-ion mechanisms predominating in the low- and high-concentration OH solu-
tions, respectively.
       The ability of certain monolayer-coated surfaces to enhance or retard film
growth was exploited to pattern CD PbS films (see also Secs. 2.8 and 4.1.5.3).
This patterning was based on the UV-induced oxidation of a thiol linkage to Au
(a strong bond) to a weakly bound sulphonate group that could be rinsed away
[41]. Figure 5.1 shows the processes involved in this patterning. A long-chain
mercapto-carboxylic acid (16-mercaptohexadecanoic acid) was self-assembled on
a Au substrate and exposed to short-wavelength UV radiation through a patterned



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 5.1 Scheme for patterned deposition of PbS. (a) Self-assembled monolayer (SAM)
of long-chain mercapto-carboxylic acid. On exposure to UV radiation through a mask, the
exposed thiol group is oxidized to a sulphonate group (b) that is weakly bound to the
Au substrate and can be easily rinsed away (c). Subsequent formation of a long-chain alkyl
thiol SAM occurs only on the exposed Au (d). CD of PbS occurs only on hydrophilic car-
boxylate endgroups and not on hydrophobic methyl groups (e). (See Ref. 41).


mask, in this case a TEM grid (a). The parts of the monolayer exposed to the UV
were oxidized to sulphonates (b) that were then rinsed away, leaving the substrate
patterned with mercapto-carboxylates and exposed Au (c). A long-chain alkyl
thiol (16-mercaptohexadecane) was then self-assembled onto the (exposed Au
portions of the) substrate (d). Chemical deposition of PbS (Pb2 , NaOH, thiourea)
onto this patterned substrate resulted in PbS deposition on the hydrophilic car-
boxylate endgroups and almost not at all on the hydrophobic alkyl endgroups (e).
Figures 5.2a and b show the resulting deposit: A dense PbS deposit formed on the
parts that were unexposed to the UV radiation (dark grid areas in 5.2a and lower
part of 5.2b), while only scattered particles were found on the exposed areas.
5.2.4.3      Ferroelectric Substrate
Deposition of PbS onto a poled ferroelectric substrate (a complicated oxide of
mainly Pb, Zr, and Ti) from a Pb(NO3)2/NaOH/thiourea bath (containing also hy-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 5.2 SEM image of (a) PbS (dark areas) growing on patterned carboxylate-termi-
nated regions of substrate and (b) boundary between PbS-covered carboxylate region
(lower part of micrograph) and almost bare thiol-terminated region. (From Ref. 41 with
permission from Elsevier Science).

droxylamine hydrochloride as a reductant and a trace of Bi3 ; the reason for these
additives was not given) resulted in films with larger particle size ( 1 m) com-
pared to that of films deposited onto an unpoled ferroelectric or glass (particle size
ca. 0.3 m) [42]. The larger particle size was explained by the electric field pre-
sent at the substrate surface, which attracts ions from the solution and increases
the growth rate. It was not stated whether the film thickness was greater for the
poled substrates (as should be for a faster growth rate). In any case, a faster growth
rate does not necessarily mean a larger crystal size; the opposite is often true.
       The electrical properties of the films also depended on the poled state of the
substrate. The resistivity of the PbS on the poled surfaces (10 -cm on the posi-
tively poled face and 20 -cm on the negatively poled face) was overall lower than
that on the unpoled ceramic (varied between 10 and 100 -cm) or on glass (50 -
cm). The photoconductivity response varied over almost an order of magnitude, de-
pending on substrate and temperature. The room-temperature response was high-
est on glass and lowest on the poled substrate, but this order was reversed at liquid
N2 temperatures. The films were all p-type except for that deposited on the nega-
tively poled face, which was n-type. All these results were explained by the effects
of the electric field and the surface charge on the depositing film. The temperature
dependence of the photoconductivity effects was attributed to trapping dominating
the photoconductivity at lower temperatures. It should also be noted that the parti-
cle size may also affect the electrical properties; e.g., fewer grain boundaries would
result in a lower resistivity (assuming identical bulk properties).
5.2.4.4      Deposition on Liquid
Although not CD in the usual sense of the technique, it is worth mentioning a re-
port of formation of PbS films by passing H2S over PbAc2 aqueous solution
slightly acidified with acetic acid [25]. The films were picked up on a gauze and



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
were analyzed by electron diffraction. The (001) faces were found to be more or
less parallel to the solution surface, and the crystal size was ca. 25 nm. It is clear
that this technique should be applicable to almost all metal sulphides (also se-
lenides—see Sec. 5.3—and tellurides).
       Deposition from colloidal sol, a related deposition technique (although not
depending on a particular substrate surface) also resulted in PbS films using
gaseous H2S. A very thin PbS film was formed when a quartz plate was immersed
overnight in a solution of Pb(ClO4)2 and poly(vinyl alcohol) through which H2S
had been bubbled [43]. The absorption spectrum of this film was similar to that of
the PbS sol and consisted of several absorption peaks with an absorption onset of
ca. 630 nm (strongly blue shifted from the PbS bulk bandgap). The XRD crystal
size of the precipitate was ca. 3 nm (see Chap. 10 for more details).

5.2.5 Optical Absorption/Transmission/Reflection
      of PbS Films
Some general remarks on the optical absorption of lead chalcogenides are in order
here, since transmission spectra of thin films of these compounds are open to mis-
interpretation. If a transmission spectrum of, say, a thin ( 100 nm) PbS film is
taken, a sharp drop in transmission will be seen in the red–near IR region (usually
between 600 and 800 nm). It is easy to translate this into the bandgap of PbS, and
this has been done, even in careful studies. Thus, one thorough study of films of the
Pb chalcogenides deposited both by CD and by evaporation (for PbTe, only by
evaporation) has reevaluated their bandgaps upwards (e.g., 1.3 eV for PbS instead
of 0.4 eV) [44]. It was later clarified that the bandgaps were indeed much lower
than originally believed (e.g., Ref. 45). These materials have an absorption coeffi-
cient in the region of their bandgap, and for considerably higher photon energies,
that is only moderate (ca. 104 cm 1), and only at much higher photon energies does
this absorption coefficient increase substantially (by about an order of magnitude),
giving the apparent bandgap onset measured in thin films. In thick films (microns
and up), the absorption close to the bandgap is high. However, for much thinner
films, this absorption is weak. Furthermore, it is often masked by reflection.
       It is particularly important when dealing with films that absorb weakly and
that have high reflectivity (such as PbS) to stress that transmission measurements
that are uncorrected for reflection should not be directly converted to absorption,
since the differences in transmission with change in wavelength may well be due
to differences in reflection rather than in absorption. Since the lead chalcogenides
have high dielectric constants (therefore high refractive indices and high reflec-
tion), masking of weak or even moderate absorption by reflection is probable.
       Although many studies of the optical properties of CD PbS films have been
made, most of them do not extend to wavelengths corresponding to the bulk PbS
bandgap (ca. 3 m). The study noted by Gibson [44], which was corrected for re-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
flection, showed a strong absorption beginning at ca. 800 nm and a relatively weak
absorption extending out to at least 6 m.
       In another study that does show the extended IR region for 1- m-thick
films, the absorption edge is very dependent on the presence or absence of an ox-
idizing agent in the bath (the absorption coefficients at energies well above the on-
set were similar) [29]. The films prepared with oxidant had a very gradual onset
starting well below the normal bandgap of PbS (ca. 0.4 eV), while that deposited
without oxidant exhibited a sharp onset at ca. 0.52 eV, considerably above the nor-
mal bandgap. Various possibilities might explain these differences. The lower on-
set of the “oxidant” films might be caused by bandgap tailing due to a high con-
centration of states in the gap near the band edges or to electric field effects on the
bandgap due to these or other states. The larger bandgap of the nonoxidant films
could be due to size quantization (the crystal size—as opposed to particle size—
was not measured), band filling (Burstein–Moss shift), or, as suggested in the pa-
per, a high concentration of oxygen in the PbS (although, as pointed out in Sec.
5.2.3.3, the use of oxidant in the deposition solution did not affect the oxygen con-
centration in the film).
       Most other studies show the optical spectra of thin (often 100-nm) films,
with emphasis on their solar control properties, and limited to a long-wavelength
limit of 2.5 m [46–48] (Ref. 49 shows reflection spectra to longer wavelengths).
These films usually have an apparent absorption onset in the region of 600–800
nm, the longer wavelengths characteristic of thicker films. The spectra at longer
wavelengths are typically dominated by reflection rather than by absorption. This
would suggest bandgaps of between 2 and 1.5 eV. Figure 5.3 shows typical ex-
amples of the transmittance and specular reflectance of PbS films deposited from




FIG. 5.3 Optical transmittance (a) and near-normal specular reflectance (b) of CD PbS
films of different thicknesses. The thickness increases from A to F over an estimated range
of ca. 50 nm to 200 nm. (Adapted from Ref. 46 with permission from IOP Publishing
Ltd.).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
a Pb(Ac)2/triethanolamine/NaOH/thiourea bath; the absence of absorption in the
near-IR region, where PbS would normally absorb, is evident from a comparison
of the transmission and reflection data. In fact, observation of various CD PbS
(from thiourea baths) spectra shows that there is absorption starting around 1.5–2
  m, but it is usually weak. Partly this is due to the fact that much of the literature
deals with thin films (often 100 nm), but the absorption does seem to be smaller
than expected. In the study by Pop et al. [49], moderately strong absorption still
occurs in multilayer films (some hundreds of nanometers thick) at wavelengths
longer than 2 m.
       The small amount of information on films deposited from thiosulphate baths
suggest that the absorption of these films in the near-IR range may be higher than
those deposited from thiourea baths. Gadave et al. measured a transmission under
1% over the entire measured range from 2 m to 400 nm (in fact the transmission
increased slightly towards shorter wavelengths) [10]. No reflectivity or thickness
data were given: however, the films must have been very thick (by CD standards)
to give such a low transmission at 2 m. Reference 17 shows a gradually de-
creasing transmission from the maximum wavelength measured (2.5 m) for a
film of unknown thickness, with a transmission of ca. 20% at 800 nm, while a 60-
nm film deposited by a similar process on ultrathin Cu2S (which, according to the
data of Ref. 17, did not much affect the spectrum) showed a transmission of 11%
at 800 nm (lower than most layers of comparable thickness deposited from a
thiourea bath) but with a poorly defined long-wavelength onset.
       To sum up, while there is too little information available to draw any firm
conclusions, it appears that films deposited from most thiourea baths are weakly
absorbing in the near-IR region and that films deposited from thiosulphate solu-
tions (which are mildly acidic) may possess different optical properties in general
than those deposited from (alkaline) thiourea baths. In this respect, and if this dif-
ference is real, it would be interesting to deposit PbS from thioacetamide baths,
which can be both acidic and alkaline.

5.2.6 Photoconductivity of CD PbS (and PbSe):
      General Considerations
Specific photoconductive properties of PbS films have already been treated. This
section deals with more general aspects of CD PbS (for the most part, also rele-
vant for PbSe) films.
       The bandgap of PbS at room temperature is 0.4 eV, corresponding to a
wavelength of 3.1 m, which more or less gives the long-wavelength detection
limit. The bandgap increases with increasing temperature, in contrast to the nor-
mal semiconductor bandgap dependence on temperature. Therefore the long-
wavelength detection limit of detectors made using these films shifts to shorter
wavelengths with increasing temperature. A low temperature of operation there-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
fore not only increases the sensitivity of the detectors (reduction of noise) but also
widens the spectral response.
       The air-annealing (typically between 300 and 400°C) of PbS and PbSe used,
in most cases, to maximize the photoconductive response results in some surface
oxidation, probably to PbSO3 and PbSO4. Such compounds have been detected in
PbS precipitate, prepared from a typical CD process, after air-annealing at 300°C
or higher [50]. Oxidizing agents are often added to the CD bath to give photocon-
ductive films in the as-deposited state [29,30]. In fact, commercial photoconduc-
tive films used a proprietary oxidant as a matter of course. These commercial
films, prepared from alkaline solutions of a lead salt and thiourea, were made up
of several layers of CD PbS (total thickness ca. 1 m) to maximize the photore-
sponse, in particular the long-wavelength response.
       The cause of photoconductivity in these films has been thoroughly studied.
Here only a very condensed account of the theory is given. References 51 and 52
give more detail.
       Electrical conductivity, , is given by the product of free-carrier concentra-
tion, n, carrier mobility, , and carrier charge, e:
           ne                                                                    (5.5)
Therefore an increase in conductivity upon illumination (photoconductivity) can
be due to either an increase in carrier concentration and/or an increase in mobil-
ity. In general, it is believed that an increase in carrier (hole) concentration is the
dominant cause for room-temperature photoconductivity for the lead chalco-
genides and that an increase in mobility becomes increasingly important at low
temperatures. The dark conductivity of films deposited with or without added ox-
idant were similar; the difference in photoconductivity between them was as-
cribed to the formation of sensitizing centers (interband states) due to the oxidant.
       Finally, it is worth mentioning a comment made in a paper describing junc-
tions between Ge and CD PbS [34]. It was noted that evaporated epitaxial PbS
films were poorly, if at all, photoconducting, while CD films, with mobility lower
by two orders of magnitude and much “poorer” structure, were much superior in
this respect. In a way, this should not be surprising since, for good photoconduc-
tivity, low dark conductivity (and therefore either low mobility and/or low carrier
concentration) is necessary.

5.3 PbSe
PbSe has a considerably more recent history than does PbS; having been deposited
by CD only about 60 years ago, instead of more than 130 years for PbS. The ear-
liest work appears to have been carried out by the Germans during World War II
and, as for PbS, was shrouded in secrecy. While PbSe has been deposited by the
Se analogues of thiosulphate and thiourea, as for PbS, in contrast to the history of



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
PbS deposition, the early work used mostly selenourea (or selenourea deriva-
tives), while the more recent studies used the selenosulphate baths. As for PbS,
these two baths will be treated separately. Optical and photoconducting properties
will not be treated separately as for PbS, for which there is a larger body of results,
but are included with the description of the various deposition baths.

5.3.1 Selenourea-Based Baths
The first apparent report in the open literature of CD PbSe for photoconductive de-
tectors was in 1949 [53]. The PbSe was deposited from a solution of PbAc2 and
selenourea onto a predeposited (from PbAc2 and thiourea) layer of PbS. The PbS
layer acted as a seed layer, presumably to obtain faster deposition (it was noted
that the PbSe deposition was much slower than that of PbS). The photoconduc-
tivity of this film exhibited a broad maximum between 3 and 4 m, giving a rea-
sonable response out to beyond 4.5 m (PbS drops off at 3 m).
       A detailed description of PbSe deposition using N,N-dimethylselenourea
(DMS) was presented in 1964 [54]. DMS was used instead of selenourea because
of its greater stability. Even so, Na2SO3 was added to inhibit decomposition to Se.
The Pb (as nitrate) was complexed with citrate to keep it in solution in the alka-
line conditions used (pH 9.8 using ammonia). This pH was fairly critical: At 9.5
(and at 25°C) the deposition was much slower, and at 10.1 rapid bulk precipita-
tion occurred. Counter intuitively (and contrary to the case using selenosulphate),
the reaction rate was faster for an aged DMS solution than for a fresh one (after
aging for 15 hr, little additional increase in rate occurred on further aging). This
was explained by inhibition of the DMS decomposition to selenide by the sulphite
(Sec. 3.2.2.2). It was noted that PbSe formation occurred more rapidly in the ab-
sence of sulphite and that adding fresh sulphite to an aged DMS solution reduced
the reaction rate. The best films were obtained if a fresh solution was used to de-
posit PbSe on glass slides that had been precoated with a thin PbSe layer. In this
case, film growth started almost immediately rather than after an induction period.
       The resistivity of the films varied from 106 to 107 . Since the thick-
nesses of the films were not given, it is difficult to convert these values to a spe-
cific resistivity. However, it does appear that the variations in resistivity were due
mainly to film thickness and an upper value of 100 -cm seems reasonable.
       A detailed study of both the mechanism [55] and the kinetics [56] of PbSe
deposition from selenourea baths has been carried out. The Pb was complexed
with citrate, and hydrazine was used to control alkalinity (remember that hy-
drazine can function also as reducing agent and as complexant). As with the fore-
going study, immediate deposition, with no induction time, occurred on glass on
which PbSe had been precoated (or on glass that was sensitized with SnCl2 solu-
tion, resulting in formation of tin hydroxide/oxide nuclei). On untreated glass, an
induction time for film deposition, which paralleled that for homogeneous pre-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
cipitation in solution, was observed. The pH was somewhat lower than that of the
preceding study (typically ca. 8.8) and evidently less critical, although the depo-
sition rate increased with increasing OH concentration. The rate was also de-
pendent on the selenourea concentration but independent of the Pb or citrate con-
centration. Based on these observations, it was suggested that the rate-determining
step was decomposition of selenourea by OH to selenide ion:
      (NH2)2CSe      OH → HSe            CH2N2      H2O                         (5.6)
The selenide could then react with whatever Pb species were present, either in so-
lution or as a solid phase, if present. The activation energy of this process, 60
kJ/mole, was consistent with a chemical rate-determining step.
       Films deposited from a selenourea bath (details not provided) were annealed
at 350°C in air and changes in their resistivity and photoconductivity measured
[57]. The resistivity of the films (between 1 and 4 m thick) dropped from a few
hundred k (assumed lateral conduction across the film) to about 1 k after heat-
ing for two minutes, with a peak in the resistivity after heating for about a minute.
The photoconductivity maximum shifted from 1.5–1.8 m for the as-deposited
film to 3.6 m for the annealed one. These phenomena were attributed to crystal-
lization of a matrix of amorphous or nanocrystalline PbSe surrounding larger crys-
tals, which crystallize or grow with annealing time. The crystal size of as-de-
posited films in this and other studies by the same group was typically some tens
of nanometers.

5.3.2 Selenosulphate-Based Baths
Sodium selenosulphate was used to precipitate PbSe from Pb2 solutions by mix-
ing PbAc2 and Na2SeSO3 solutions [58]. The precipitation was rapid, and film
formation did not occur (at least to any visible extent). Complexation was required
to slow the reaction. Films of PbSe were first deposited using selenosulphate from
a hydroxide-complexed (plumbite) bath [59]. By complexing the Pb [with citrate,
Rochelle salt (tartrate), or oxalate] and adding alkali (NaOH, KOH, or NH4OH) to
an optimum pH of 11.05, mirror films of PbSe were obtained [60]. The mecha-
nism proposed, for both precipitate [58] and films using carboxylic acids [60], was
formation of PbSeSO3:
      PbAc2     Na2SeSO3 → PbSeSO3           2NaAc                              (5.7)
followed by decomposition of the PbSeSO3 to PbSe by water:
      PbSeSO3      H2O → PbSe        H2SO4                                      (5.8)
No evidence was presented for this mechanism, although it is a reasonable route,
nor for the assumption of an identical mechanism for precipitate and film forma-
tion (although that too is likely. In fact, in another study, based on a kinetic study



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of PbSe formation from a similar bath employing glass powder (particle size ca.
80 m) as a “substrate,” it was concluded that the same mechanism was operative
both for PbSe formation as a precipitate in solution and for that deposited on the
glass surface [61]). On the other hand, Fofanov and Kitaev assumed a mechanism
whereby selenide ion, formed by hydrolysis of selenosulphate, reacted with Pb2
via the ion-by-ion mechanism [59].
       The thickness of the films using the carboxylate complex was not dependent
on the Pb concentration (at least from 10 mM up). It was, however, very depen-
dent on the pH value, with a sharp maximum, the position of which was depen-
dent on the anion of the lead salt and, more so, on the hydroxide used (Na, K,
NH4). The thickness varied from 50 nm to 300 nm. The film thickness was also
moderately dependent on deposition temperature, increasing with increase in tem-
perature (in contrast to the more usual decrease in terminal thickness with in-
creasing temperature). It is not certain, however, whether these thicknesses (and
others measured in this study) were actually terminal thicknesses or only the
thickness measured after a certain time.
       The bandgap of these films was measured to be 0.28 eV—the same as the
literature value.
       Films were also deposited from similar solutions, only using selenourea in-
stead of selenosulphate and at somewhat lower pH ( 9). The quality (of the mir-
ror surface) and thickness of the films deposited from selenosulphate solutions
were similar to those deposited from the (at that time) conventional selenourea
ones.
       Kainthla et al. carried out an investigation on the parameters that affected
deposition rate [62]. The rate increased, as expected, with increase in temperature
and selenosulphate concentration. However, it decreased with increase in pH. This
was explained on the basis of the expected dominant hydroxy-citrate complex of
Pb, [Pb(OH)C6H5O7]2 . The equilibrium involving this complex is
      [Pb(OH)C6H5O7]2 D Pb2            OH       C6H5O 3
                                                      7                       (5.9)
A greater hydroxide concentration will shift the equilibrium to the left, decreasing
the free-Pb2 concentration. This explanation means that the rate is dependent on
the lead concentration, in contrast to the previous studies discussed earlier, which
found the rate to be essentially independent of this parameter.
       It was also noted that the optimal pH was temperature dependent: At low
deposition temperature it was 9, while at high temperatures it was 10. This fol-
lows from the inverse dependence of rate on pH. The optimal pH is a balance be-
tween slow-enough formation of PbSe (to prevent precipitation in solution) but
not too slow to prevent formation of a film in a reasonable time. At higher tem-
peratures the rate is faster, and therefore the optimal pH should be lower.
       Films (between 0.4 and 2.5 m thick) deposited from selenosulphate so-
lution were characterized by optical spectroscopy and (photo)conductivity [63].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
There appeared to be a very large variation in optical properties from film to
film; in one case, little absorption (corrected for specular reflection) was mea-
sured from 2 m toward increasing wavelength, while another sample showed a
sharp absorption onset close to 5 m. The films had a resistivity of between
3.103 and 3.104 -cm and a photoconductivity maximum varying between 1.2
and 1.9 m.
       An investigation of PbSe deposited from selenosulphate baths was carried
out using various complexants and conditions and with emphasis on the mecha-
nisms and, in particular, on the crystal size and morphology of the deposits [64].
Citrate, nitrilotriacetate (NTA), and hydroxide (in order of increasing complexing
strength with Pb) were used under conditions where a solid hydrated oxide phase
was either present in solution (cluster) or absent (ion-by-ion). The presence of this
phase, and a semiquantitative estimation of its relative concentration, was mea-
sured by UV absorption and light-scattering measurements. As for CdSe, the mor-
phologies of the two types of films were very different, with the ion-by-ion films
having a larger crystal size. However, the difference depended on the complexant
used as well as on other deposition conditions. A wide variety of crystal sizes was
obtained. Figure 5.4 shows some TEM micrographs of various PbSe films. The
conditions of deposition for each film (the important factors are given in the fig-
ure legend) are not important here; the purpose is to demonstrate the wide range
of crystal sizes and morphologies in PbSe films obtained using different com-
plexants, solution compositions, and deposition temperatures.
       Under certain conditions, relatively large and small crystals coexisted in the
same films. For example, in “cluster” deposition from citrate complex, both small
(ca. 4 nm) and medium-sized (6–12 nm) crystals were formed, although only the
small crystals formed at the beginning of the deposition, and the larger ones ap-
peared in thicker films, again, an indication of both mechanisms operating (ion-
by-ion growth normally is slower than cluster growth, therefore it takes longer for
the larger crystals to appear). Additionally, the small and larger crystals were not
deposited together but in separate regions; i.e., regions only of larger crystals were
formed surrounded by (most of the deposit) only small crystals. It was suggested
that the larger crystals formed in regions when no previous small-crystal deposi-
tion had occurred (i.e., in pinholes in the originally thin film).
       For low selenosulphate concentrations, only the small crystals were formed,
even in thicker films, and this was rationalized by the lower steady-state selenide
concentration, which would favor cluster growth over ion-by-ion formation (the
product of free lead and selenide ions needs to be larger than the solubility product
of PbSe for ion-by-ion deposition to occur). An important difference between the
citrate depositions and the NTA or hydroxide ones is that, even in the “ion-
by-ion” citrate deposition, some low concentration of colloidal hydrated oxide was
present, due to the relatively low complexing strength of citrate. The pH of the hy-
droxide baths ( 13) was much higher than that of the citrate or NTA baths (10.8).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 5.4 TEM micrographs of PbSe films deposited under different conditions. A: from
citrate/0°C/hydroxide mechanism. B: nitrilotriacetate/60°C/ion-by-ion mechanism. C: hy-
droxide/60°C/ion-by-ion mechanism. D: hydroxide/80°C/ion-by-ion mechanism. All films
are as deposited.



       In common with CdSe deposition from selenosulphate baths, cluster depo-
sition of PbSe normally resulted in specular films, while ion-by-ion films were ini-
tially highly scattering as thin films but eventually (usually) became specularly re-
flecting with increase in thickness. As for CdSe, the development of specularity
with thickness of ion-by-ion films could be explained by filling in of voids be-
tween the large crystals.
       The wide range of very small crystal sizes in these films gives rise to strong
blue shifts in their optical spectra due to size quantization [65]. This aspect of
these films is dealt with in detail in Chapter 10.
       Other complexants have been used for PbSe deposition. Triethanolamine
was used in one study [66]. While deposition occurred over a wide range of tem-
peratures, optimum results (in terms of rate of deposition and film thickness) were
obtained at a deposition temperature of 75°C. In another study, lead nitrate was
dissolved in an excess of hydroxide and excess selenosulphate was also used as an
additional complexant [67]. The pH was 10 (adjusted with acetic acid), and depo-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sition was carried out both at room temperature and at 70°C. The substrate (glass
or polyester) was pretreated with SnCl2 solution—the standard tin sensitization;
presumably this improved the adherence and/or homogeneity of the films. The re-
sistivity of these films was ca. 104 -cm as deposited. Annealing in air at 100°C
for 24 hr decreased this resistance to 3 10 3 -cm, a drastic change for such a
mild treatment and one that suggests that the electrical properties of these as-de-
posited films may change greatly with time, even under ambient conditions.

5.3.3 Variations in PbSe Deposition
X-ray amorphous PbSe has been deposited by addition of Na2S2O3 (at least half
the selenosulphate concentration) to a room-temperature selenosulphate/citrate
bath [68]. The thiosulphate increased the induction time for PbSe formation in the
solution, although eventually film growth was even faster than in the absence of
thiosuphate. It was noted that the initial, very thin deposit was yellow. A clear
XRD pattern was obtained (for thicker films) after annealing at 350°C for one
minute, giving a crystal size of 13 nm (which grew with continuing annealing). It
is interesting to note that, in contrast to films deposited in the absence of thiosul-
phate, those made using thiosulphate did not peel off, even after remaining a long
time in the depositing solution. Chemically deposited films will often peel off if
left too long in solution, probably due to stress in the thickening film, and this is
expected to be less or absent in an amorphous film. The amorphous structure was
attributed mainly to the increased deposition rate caused by the thiosulphate.
Since the growth rate was apparently in the range of hours, this by itself would not
be expected to result in an amorphous film, although it might contribute to the ef-
fect. A thorough investigation of this deposition would be desirable, including
TEM imaging (to see whether the film is, indeed, amorphous, since XRD by itself
is not a good enough verification of this) and, particularly important, composi-
tional analysis (to see if there is an appreciable amount of S in the film, thiosul-
phate itself forming PbS).
       From the optical spectrum, an approximate bandgap of 1.5 eV could be esti-
mated, although the interpretation of the spectrum is open to ambiguity. A photo-
conductivity maximum at 1.1 m (1.1 eV) was measured. These anomalously high
energy values were attributed to size quantization (Chap. 10). Transmission in the
mid-IR was also measured for these films. Although there were some absorption
bands, probably due to adsorbed species from the deposition bath, the films were
essentially transparent at wavelengths longer than ca. 5 m, the transmission de-
creasing gradually at shorter wavelengths. While annealing at 350°C did change
the IR spectral shape, particularly the intensity of some of the absorption bands,
overall the transparency beyond 8 m was maintained (in some regions, even in-
creased). This implies that the films, both as deposited and annealed, were very in-
trinsic, since free carriers would result in absorption in this spectral region.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       The (presumed lateral) resistance of the films was ca. 108 , which de-
creased 4–5 orders of magnitude after annealing at 350°C.
       In analogy with PbS, a film of PbSe has been formed by passing H2Se over
PbAc2 solution acidified with acetic acid, giving specularly reflecting films that
could be picked up from the surface of the solution [25]. Electron diffraction of
these films showed a tendency to grow with (001) planes parallel to the solution
surface. The average crystal size of these films was 25 nm.
       A variation of the CD process for PbSe involved deposition of a basic lead
carbonate followed by selenization of this film with selenosulphate [64]. White
films of what was identified by XRD as 6PbCO3 3Pb(OH)2 PbO (denoted here as
PbMOHMC) were slowly formed over a few days from selenosulphate-free so-
lutions that contained a colloidal phase and that were open to air (they did not form
in closed, degassed solutions). CO2 was necessary for film formation—other than
sparse deposits, no film formation occurred of hydrated lead oxide under any con-
ditions attempted in this study. Treatment of these films with selenosulphate so-
lution resulted in complete conversion to PbSe at room temperature after 6 min.
The selenization process of this film was followed by XRD, and it was seen to pro-
ceed by a breakdown of the large PbMOHMC crystals to an essentially amor-
phous phase of PbSe with crystallization of this phase to give finally large (ca. 200
nm) PbSe crystals covered with smaller (15–20 nm) ones as well as some amor-
phous material.
       Hydrazine added to a selenourea//Pb2 bath was found to strongly affect the
electrical and photoconductive properties of the PbSe films [69]. As the hydrazine
concentration increased, the dark resistivity increased in a very non-monotonic
way, from ca. 0.1 -cm (low hydrazine concentration) to 104 -cm. Photocon-
ductivity was observed, as might be expected, only for the high-resistivity films.
The effect of the hydrazine was attributed to an increase in the deposition rate by
an increase in pH; however, the pH values of the solutions were not reported, and
hydrazine would be expected to affect the reaction as a reducing agent apart from
pH considerations. As an interesting aside, it was noted that only p-type films
could be deposited by this technique (and this is the case for most CD PbS), while
evaporated PbSe films were normally n-type (due to Se vacancies). The absorp-
tion edge of the films was ca. 4.2 m, close to to the literature bandgap of
0.28 eV.

5.3.4 Comparison of Films Deposited from
      Selenourea and Selenosulphate Baths
A series of studies comparing films deposited from selenourea and selenosulphate
baths were carried out, with emphasis on photoconductivity behavior and the ef-
fects of annealing. A broad photoconductivity maximum occurred at ca. 1.6 m
for the selenosulphate film, decreasing strongly beyond 2 m. A similar spectrum,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
but shifted 0.2–0.4 m to longer wavelengths, was observed for the selenourea
film. Annealing at 350°C in air broadened both spectra to ca. 5 m (approximately
the bulk bandgap of PbSe), although the the photoconductivity response of the se-
lenosulphate films was initially degraded (mainly due to increased noise), and a
long (ca. 50 hr) annealing time was required for good response [70].
      Photoluminescence spectra of the films were measured (77 K) and com-
pared with an epitaxial PbSe layer in the same study. Blue shifts in the spectra
(greater for the selenourea films) were attributed to quantum size effects (see
Chap. 10). The crystal size was reported to be 40–60 nm (the selenourea ones be-
ing somewhat smaller than the selenosulphate ones), growing to 100–150 nm af-
ter annealing.
      The changes in resistivity with annealing of films deposited from selenourea
and selenosulphate baths, as well as evaporated films, were compared [71,72]. Al-
though there were small differences between the various films, no major differ-
ence was found. Additionally, the resistivity of as-deposited films, deposited from
both selenourea and selenosulphate baths, does not change with time over a period
of months in air. However, after annealing in air at 350°C when the resistivity in-
creases, there is a gradual decrease in room-temperature resistivity (and also in
photoconductivity response) with time [73]. These variations were related to for-
mation of PbSeO3 and adsorbed oxygen on the surface of the annealed crystals.


REFERENCES
 1.   C Puscher. Dingl. J. 190:421, 1869.
 2.   J Emerson-Reynolds. J. Chem. Soc. 45:162, 1884.
 3.   DE Bode. In: Physics of Thin Films. Vol. 3, Academic Press, New York, NY, 1996,
      p 275.
 4.   F Kicinski. Chem. Ind. 54, 1948.
 5.   TS Moss. Proc. IRE 43:1869, 1955.
 6.   WN Arnquist. Proc. IRE 47:1420, 1959.
 7.   DJ Lovell. Am. J. Phys. 37:467, 1969.
 8.   TH Johnson. Proc. SPIE 60, 1983.
 9.   CD Lokhande. Mater. Chem. Phys. 28:1, 1991.
10.   KM Gadave, SA Jodgudri, CD Lokhande. Thin Solid Films 245:7, 1994.
11.   E Beutel, A Kutzelnigg. Z. Elektrochem. 36:523, 1930.
12.   PL Allen, A Hickling. Chem. Ind. 51:1558, 1954.
13.   EB Andersen. Z. Phys. Chem. B. 32:237, 1936.
14.   WH Perkins, AT King. J. Chem. Soc. 103:301, 1913.
15.   E Beutel. Z. Angew. Chem. 26:700, 1913.
16.   E Beutel, A Kutzelnigg. Monats. 58:295, 1931.
17.   I Grozdanov. Semicond. Sci. Tech. 9:1234, 1994.
18.   M Najdoski, B Mincevasukarova, A Drake, I Grozdanov, CJ Chunnilall. J. Mol.
      Struct. 349:85, 1995.
19.   HL Smith. J. Sci. Instrum. 4:115, 1927.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
20. O Hauser, E Biesalski. Chem. Ztg. 1078, 1910.
21. G Bruckman. Kolloid Z. 65:1, 1933.
22. MK Norr. J. Phys. Chem. 65:1278, 1961.
23. H Pick. Zeit. Phys. 126:12, 1949.
24. HN Acharya, NK Misra. J. Phys. D: Appl. Phys. 4:1968, 1971.
25. H Wilman. Proc. Phys. Soc. 60:117, 1948.
26. HN Acharya, HN Bose. Phys. Status Solidi (a) 6:K43, 1971.
27. VF Markov, LN Maskaeva, GA Kitaev. Inorg. Mater. 36:657, 2000.
28. EM Larramendi, O Calzadilla, A Gonzalez-Arias, E Hernandez, J Ruiz-Garcia. Thin
    Solid Films 389:301, 2001.
29. GH Blount, PJ Schreiber, DK Smith, RT Yamada. J. Appl. Phys. 44:978, 1973.
30. GP Kothiyal, B Ghosh, RY Deshpande. J. Phys. D: Appl. Phys. 13:869, 1980.
31. IC Torriani, M Tomyiama, S Bilac, GB Rego, JI Cisneros, ZP Argüello. Thin Solid
    Films 77:347, 1981.
32. PK Nair, VM Garcia, AB Hernandez, MTS Nair. J. Phys. D: Appl Phys. 24:1466,
    1991.
33. J Bloem. Appl. Sci. Res. Section B 6:92, 1956.
34. JL Davis, MK Norr. J. Appl. Phys. 37:1670, 1966.
35. G Guizzetti, F Filippini, E Reguzzoni, G Samoggia. Phys. Status Solidi (a) 6:605,
    1971.
36. S Watanabe, Y Mita. J. Electrochem. Soc. 116:989, 1969.
37. M Isshiki, T Endo, K Masumoto, Y Usui. J. Electrochem. Soc. 137:2697, 1990.
38. BL Sharma, SN Mukerjee. Phys. Status Solidi (a) 2:K21, 1970.
39. H Sigmund, K Berchtold. Phys. Status Solidi 20:255, 1967.
40. FC Meldrum, J Flath, W Knoll. J. Mater. Chem. 9:711, 1999.
41. FC Meldrum, J Flath, W Knoll. Thin Solid Films 348:188, 1999.
42. I Pintilie, E Pentia, L Pintilie, D Petre, C Constantin, T Botila. J. Appl. Phys. 78:1713,
    1995.
43. S Gallardo, M Gutierrez, A Henglein, E Janata. Ber. Bungenges. Phys. Chem.
    93:1080, 1989.
44. AF Gibson. Proc. Phys. Soc. B 63:756, 1950.
45. JGN Braithwaite. J. Sci. Instr. 32:10, 1955.
46. PK Nair, MTS Nair, A Fernandez, M Ocampo. J. Phys. D: Appl. Phys. 22:829, 1989.
47. PK Nair, MTS Nair. J. Phys. D: Appl. Phys. 23:150, 1990.
48. PK Nair, M Ocampo, A Fernandez, MTS Nair. Sol. Energy Mater. 20:235, 1990.
49. I Pop, C Nascu, V Ionescu, E Indrea, I Bratu. Thin Solid Films 240: 1997.
50. GA Kitaev, LG Protasova, VG Kosenko, VF Markov, LV Dyadyuk, GV Stepanova.
    Inorg. Mater. 29:1182, 1993.
51. S Espevik, C-H Wu, RH Bube. J. Appl. Phys. 42:3513, 1971.
52. E-H Lee, RH Bube. J. Appl. Phys. 43:4259, 1972.
53. CJ Milner, BN Watts. Nature 163:322, 1949.
54. RA Zingaro, DO Skovlin. J. Electrochem. Soc. 111:42, 1964.
55. AB Lundin, GA Kitaev. Inorg. Mater 1:1900, 1965.
56. AB Lundin, GA Kitaev. Inorg. Mater. 1:1905, 1965.
57. LP Biró, A Darabont, P Fitori. Europhys. Lett. 4:691, 1987.
58. NI Glistenko, AA Eremina. Zh. Neorg. Khim. 5:1003, 1960.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
59.   GM Fofanov, GA Kitaev. Russ. J. Inorg. Chem. 14:322, 1969.
60.   NI Glistenko, IN Shramchenko, AN Kosilova. Zhurn. Prikl. Khim. 45:1356, 1972.
61.   GA Kitaev, AZ Khvorenkova. Russ. J. Appl. Chem. 72:1520, 1999.
62.   RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 127:277, 1980.
63.   RM Candea, N Dadarlat, R Turcu, E Indrea. Phys. Stat. Sol. (a) 90:K91, 1985.
64.   S Gorer, A Albu-Yaron, G Hodes. Chem. Mater. 7:1243, 1995.
65.   S Gorer, A Albu-Yaron, G Hodes. J. Phys. Chem. 99:16442, 1995.
66.   P Pramanik, S Biswas, PK Basu, A Mondal. J. Mater. Sci. Lett. 9:1120, 1990.
67.   I Grozdanov, M Najdoski, SK Dey. Mater. Lett. 38:28, 1999.
68.   LP Biro, RM Candea, G Borodi, A Darabont, P Fitori, I Bratu, D Dadarlat. Thin Solid
      Films 165:303, 1988.
69.   DH Roberts, JE Baines. J. Phys. Chem. Solids 6:184, 1958.
70.   D Dadarlat, RM Candea, R Turcu, LP Biro, Zasavitskii, II, MV Valeiko, AP Shotov.
      Phys. Status Solidi (a). 108:637, 1988.
71.   RM Candea, R Turcu, P Margineanu, N Dadarlat. Phys. Status Solidi (a) 96:337,
      1986.
72.   RM Candea, R Turcu, G Borodi, I Bratu. Phys. Status Solidi (a) 100:149, 1987.
73.   RM Candea, LP Biro, N Dadarlat, G Borodi, A Darabont, P Fitori, R Turcu. Phys.
      Status Solidi (a). 108:233, 1988.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
6
Other Sulphides and Selenides




6.1 INTRODUCTION
This chapter covers sulphides and selenides not included in Chapters 4 and 5, i.e.,
all metals except for Zn, Cd, Hg, and Pb. Some of these materials, e.g., the sul-
phides of Bi, Cu, and Ag and Cu-Se, have been the subject of many investigations.
There are others, however, on which as little as one paper has been published al-
together.
       In order to minimize, as much as possible, making this chapter into a list of
recipes and properties, its layout will be somewhat different than that used up to
now. The chapter is divided into sections alphabetically (by English name rather
than by chemical symbol), designated by the name of the metal. Basic data for those
compounds for which there are at least several papers will be listed in tabular form
together with references. The data in the first column of the table (the deposition
bath composition) will be given as sulphide or selenide source/complex/tempera-
ture and pH. The specific metal ion is not given in most cases, because this will be
clear from the location of the table and the metal salt used will normally be found
in the specific description of the study in question. It should be kept in mind that
the chalcogen source itself is a complexant, often fairly weak but sometimes strong.
Resistivity is given as specific resistivity; if not defined, then the resistance re-
ported is given in parentheses. Specific details outside of these basic data will be
treated separately in the order in which the particular study appears in the table and,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
if connected with a specific datum in the relevant table, will be signified by an as-
terix after this datum. Each discussion of this type will include the reference num-
ber to identify it. Abbreviations used in the “Solution” column are: DMSeU—N,N-
dimethyl selenourea; EDTA—ethylenediaminetetra-acetate; S2O3—thiosulphate;
SeSO3—selenosulphate; TA—thioacetamide; TEA—triethanolamine; TU—
thiourea; RT—room temperature. The abbreviations in the “Bandgap” column are:
dir.—direct; indir.—indirect.
       Before discussing the specific studies, a general overview will be given for
each metal. This overview relates only to information of direct relevance to the
subject and will be relatively lengthy in some cases and very brief or even nonex-
istent in others, according to need.


6.2 ANTIMONY
6.2.1 Sb2S3 (See Table 6.1)
All reported cases of CD antimony sulphide involve the trisulphide, Sb2S3. Sb2S3
is soluble in hydroxide to give antimonates and in excess sulphide to give thio-
complexes. The latter is not a problem in CD since free sulphide, if it exists, does
so in a very low concentration and is rapidly taken up to precipitate Sb2S3. How-
ever, the solubility in alkaline solutions limits the pH of the deposition solutions.
Sb2S3 exists in two forms: so-called amorphous Sb2S3, which varies in color from


TABLE 6.1 Antimony Sulphide

                                   Bandgap         Resistivity     Conduct.
Solution                            (eV)            ( -cm)           type       Ref.

TA/NH3/TEA/RT                     1.85 ind.         5   104           n-       3
Annealed, 300 C                   1.74 ind.          250
TA/NH3/TEA/RT                     1.86 ind.         4 108             n-       4
Annealed, 300 C                   1.74 ind.    5   106 (5 108 )                5
TA/tartrate/RT/pH 9.5             1.62 ind.            107                     6
                                                     5
TA/SbCl3 in CH3COOH               1.75 dir.    7   10 (130 C)         n-       7
  (nonaqueous)/RT*/pH 1.5
Annealed 200 C                    1.62 dir.    2   105 (130 C)        n-
TU/RT/pH 1–1.2                                                                 8
Annealed 200 C                                         108            n-
S2O3/SbCl3/EDTA/RT/pH 2–3         1.82              104–105                    9
S2O3SbCl3 in CH3 COOH             2.48* dir.           107            n-       10, 11
Annealed 170–200 C                1.76 dir.            105
S2O3/SbCl3 in acetone/10 C/pH 5   2.21 dir.            107                     12
Annealed 250 C                    1.79 dir.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
yellow to red, and black crystalline Sb2S3, with a bandgap between 1.7 and 1.8 eV.
Thioacetamide is the most commonly used S source for this compound.
       Antimony sulphide deposition onto metallic substrates, together with PbS
and Cu-S, was first reported in the original paper of Puscher [1] using thiosulphate
and antimony tartrate. No characterization of this film was carried out nor prop-
erties given. The same method was also described “recently,” in 1931, using a
number of different metals as substrates [2]. It was noted that SbCl3, when mixed
with thiosulphate, reacted too rapidly, hence the use of the tartrate (tartaric acid is
a complexant). Again, no characterization of the films was made.
6.2.1.1      Thioacetamide
In Refs. 3–5, potassium antimonyl tartrate was used as a source of Sb. While the
deposition conditions appeared to be almost identical in all cases (other than the
silicotungstic acid added in some cases in the latter two studies), there were some
unexplained differences in structural and electrical properties (the latter shown in
the table). The lower resistivities obtained in the first study were mirrored by
higher carrier concentration (2 1015 cm 3 vs. 1012 cm 3) and slightly higher
mobility (14 vs. 10 cm2 V 1 sec 1).
       The as-deposited film gave either no peaks or very broad ones in the XRD
pattern, from which it is difficult to extract a crystal size, since there are so many
close peaks in the spectrum. After annealing, well-defined peaks were observed.
The films in the first study (deposited on glass) had a grain size (measured by
SEM, not necessarily crystal size) of 0.12 m as deposited, growing to nearly
4 m after annealing. From Ref. 5, the grain size of the annealed films on SnO2-
coated glass or steel (both with well-defined crystal surfaces) was close to 1 m,
but for films deposited on glass it was 0.08 m. The films in the first study were
reported to be close to stoichiometric, while those in the other studies were more
or less S rich.
       Addition of silicotungstic acid to the bath [4,5] reportedly improved the sto-
ichiometry and increased the grain size somewhat (as well as introducing a sepa-
rate WO3 phase) of the annealed films. There was no major effect of the STA on
the electrical resistance. The resistivity of the annealed (not the as-deposited)
films was also very substrate dependent [5]: ca. 5 108 -cm for films on glass
or steel and two orders of magnitude less on SnO2-coated glass. Therefore there is
no direct relationship between the resistivity and the grain size. Photoconductiv-
ity in these layers was studied [5]. The ratio between dark and light resistivities
was as high as 104 (for annealed films deposited with silicotungstic acid), but this
ratio decreased, and the decay time increased, with exposure to air. The ratio for
as-deposited films was ca. 10. Photoconductivity decay times varied from 1 sec
to several seconds.
       In a similar study [6], SbCl3 was complexed with tartaric acid. Also,
NaOH was used instead of ammonia, and the solutions were more concentrated



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
than the previous one. The films were close to stoichiometric. A lower value of
indirect bandgap was measured for this study (1.62 vs. 1.85 eV). It is notewor-
thy that while the films exhibited no XRD pattern, the precipitate in solution
gave sharp XRD peaks. Whether this is due to a different mechanism for the two
routes or an effect of the glass substrate, two suggestions made by the authors,
is an open question.
       A nonaqueous (acetic acid) bath is described in Ref. 7. While the deposition
was carried out mostly at room temperature, the bath was initially heated to 80°C
for 10 min to speed up the deposition. Unlike other films of Sb2S3, the as-
deposited films showed an XRD pattern of well-defined peaks (although these be-
came more numerous and with mostly different reflections on annealing). It is in-
teresting to note, in possible connection with this observation, that the as-de-
posited films were described as pink, which is not typical of any form of Sb-S and
suggests a mixture of a white material with a reddish one.
6.2.1.2      Thiourea
Reference 8 is the only study in which thiourea has been used to deposit Sb2S3. A
methanolic solution of SbOCl at a very low pH (ca. 1) was used. This seems to be
a unique case of deposition using thiourea at low pH and suggests that the reaction
may proceed through some complex-decomposition reaction, since free sulphide
is not expected to form under such conditions. The as-deposited films were highly
scattering, white, and nonadherent, but converted to adherent, still scattering films
with the typical orange color of as-deposited Sb2S3 after heating at 120°C. X-ray
diffraction of this film showed no pattern; annealing at over 200°C converted the
film to gray Sb2S3 with a well-defined XRD spectrum. While optical spectra were
given, it is difficult to interpret them, due to the large degree of scattering.
6.2.1.3      Thiosulphate
Reference 9 involved the first thiosulphate deposition after the original reports
of Refs. 1 and 2 and gives some properties of the films but no structural charac-
terization.
       In Refs. 10 and 11, aqueous Na2S2O3 was added to SbCl3 in glacial acetic
acid (SbCl3 hydrolyzes in water unless complexed or the solution is moderately
acidic or strongly alkaline). A pH of ca. 3 was optimum; below 2.5, adhesion
was poor; above 4, basic antimony salts precipitated. The solution was kept be-
low room temperature to prevent rapid bulk precipitation. No XRD pattern was
found for the as-deposited film, which was presumed to be amorphous. Anneal-
ing at 170°C crystallized the film, at least partly. The bandgap of the as-de-
posited film was reported to be 2.48 eV and that of the annealed film 1.76 eV.
Photoconductivity was exhibited by the annealed film but not by the as-de-
posited one.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       A similar thiosulphate bath, only using acetone instead of acetic acid to dis-
solve the SbCl3 was described in Ref. 12. It was reported that the films made us-
ing acetic acid tended to flake off, and use of acetone prevented this. Photocon-
ductivity studies showed that the photosensitivity was poor for as-deposited films
(ca. a factor of 2 increase in conductivity) but increased to between 102 and 103
after optimum annealing at 300°C in N2 (annealing at 250°C or less had little ef-
fect on the photosensitivity).

6.2.2 Sb2Se3
Only one group has reported CD of Sb2Se3. The solution used was potassium an-
timonyl tartrate, complexed with triethanolamine and ammonia. Selenosulphate
was used as the Se source. No XRD pattern was found, as for the sulphide de-
posited under equivalent conditions. The bandgap was 1.88 eV, and resistivity
  107 -cm [13,14]. Continued study of this deposition showed the effect of var-
ious parameters on deposition rate and film thickness (the latter typically reaching
1 m) [15]. This study also described some photoelectrochemical behavior of
these films (Chap. 9).


6.3 ARSENIC
There are at least three sulphides of As: As4S4, As4S6, and As4S10. The dimeric ar-
senous (III) sulphide, As4S6 (often given as As2S3) and arsenic (V) pentasulphide
(As4S10) can be precipitated by H2S as yellow solids from acidic solutions of the
respective As salts. These sulphides are soluble in alkaline solution and in
(poly)sulphide solutions and must therefore be deposited from acidic or at most
neutral solution. The pentasulphide is not a very stable compound; it is hydrolyzed
by boiling water to arsenious acid and also is unstable as a solid in air above
ca. 100°C.
      Only two papers on CD of an arsenic chalcogenide (arsenic sulphide) were
found. Films were obtained using thiosulphate in an EDTA solution of As2O3 at
room temperature and a pH of 2–3 [9]. No compositional information was given.
A bandgap of 2.0 eV and a resistivity of 104–105 -cm were measured.
      As2S3 was deposited at room temperature (27°C) from an acidic (pH 2)
thioacetamide bath containing As2O3 dissolved in concentrated HCl (and in some
cases complexed with EDTA) [16]. The terminal thickness (which reached a max-
imum and then decreased with time) was studied as a function of various deposi-
tion parameters. Well-defined XRD peaks were obtained showing the monoclinic
structure (notable since this compound has a tendency to be amorphous or nearly
so as deposited). A direct bandgap of 2.42 eV (similar to the standard value for
As2S3) was estimated from the optical spectrum. The resistivity was ca. 105 -cm.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
6.4 BISMUTH
6.4.1 Bi2S3 (See Table 6.2)
Bismuth sulphide, Bi2S3, has been rather extensively investigated. Bi3 is readily
hydrolyzed in aqueous solution and is either used in acid solution or strongly com-
plexed. Note the very low-solubility product of this compound—10 98 (Table
1.1). The very low value is due largely to the large number of ions involved (five).
However, even apart from this, the solubility (given as the concentration of free
Bi3 and S2 ions) is low, and the solid is very readily precipitated.
       A glance at the various values of bandgap shown in the table is enough to
see how wide this range is. These differences are often attributed either to an
amorphous structure or to size quantization. However, as discussed in more detail
in Chapter 10, many of these values are not dependable, and the values given in
this chapter are, in many cases, more a measure of the shape of the transmission
(absorption) curve than of any specific bandgap value.
       Deposition of Bi2S3 was first reported in 1931 from a thiosulphate bath onto
metal substrates, although no details of the deposit properties were provided [2].
6.4.1.1     Thiosulphate
In Ref. 18, the initial deposition temperature was 60°C, which was then lowered
to 27°C. The properties of the films were found to vary considerably with film
thickness. The crystal sizes and bandgaps are discussed in Chapter 10. Films 220




TABLE 6.2 Bismuth Sulphide

                                 Bandgap         Resistivity    Conduct.
Solution                          (eV)            ( -cm)          type       Ref.

S2O3 /EDTA/60–27 C*/pH 2      2.22–1.62* dir.    5.103–106                 9, 17, 18
S2O3 /HCHO/CH3COOH/             1.9 dir.            106            n-      19
  7 C/pH 1.4
S2O3 /HCHO/CH3COOH/            1.27 dir.            105                    20
  25 C/pH 1.5
TA/TEA/NH3                     1.7 indir.            103           p-      21
TA/TEA/25 or 50 C/pH 8.5       1.9              ca. 106*                   22, 23
                               1. 6 (illum.)
TA/( EDTA)/pH1–2               1.84 dir.            105                    24
TU/TEA/NH3 /                   ca. 1.5*           107 -cm          n-      25, 26
  100 C- RT*/pH 8–10
TU/TEA/NH3 /95 C/pH 9.5        1.6                  105–106        n-      27
TU/NH3 /up to 90 C             1.76                                        28




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
nm thick had a resistivity of 5 103 -cm, compared to 53-nm-thick films with
a value of ca. 3 104 -cm.
       Reference 19 described a nonaqueous deposition using a mixture of acetic
acid (in which the Bi(NO3)3 was dissolved) and formaldehyde (in which the
sodium thiosulphate, which was insoluble in acetic acid, was dissolved). Deposi-
tion was carried out at 7°C because the resulting films were more homogeneous
than at higher temperatures. No clear XRD pattern was found for the films, al-
though one was for the precipitated powder.
       A more recent investigation of the previous deposition [24] showed an XRD
pattern of moderately sharp peaks similar (in peak widths) to the precipitated pow-
der but weaker (presumably due to less material) and with only some of the re-
flections (suggesting texturing of the film).
6.4.1.2     Thioacetamide
The majority of the various studies on Bi2S3 used thioacetamide as the source of
sulphur. In the first of these studies [20], very broad peaks were observed in the
XRD spectrum and the film was classified as amorphous. (A radial distribution
function analysis of these films allowed a structure to be proposed [29]). Much
sharper XRD peaks were obtained after a mild annealing at 150°C (for 6 hr). The
broad peaks in the as-deposited film seem, for the most part, to have different
values of 2 from those of the annealed film.
       In Ref. 21, it was noted that use of ammonia (or NaOH) resulted in films
containing particulate deposits and incorporation of Bi(OH)3 into the films.
Therefore the deposition solution was ammonia free, and, although alkaline, the
pH was relatively low. It was also noted that use of thiourea instead of thioac-
etamide gave very thin films. No difference in either optical or electrical proper-
ties was found for films deposited at 25 or 50°C. The dark resistivity dropped from
ca. 106 -cm to as low as 0.3 -cm (this value was apparently very variable from
sample to sample) after air-annealing at 200°C, although it increased steeply again
at higher annealing temperatures. The photoconductivity of these films was stud-
ied. The sensitivity was ca. 20 for the as-deposited films, increasing to ca. 100 or
more after (optimum) annealing at 150°C. This temperature corresponded to the
temperature at which the dark resistivity began to decrease strongly; above this
temperature, the dark resistance dropped and therefore so did the photoconduc-
tivity response. The effect of various annealing treatments on the films was stud-
ied in more detail in a separate work [30]. Here, XRD showed no peaks whatso-
ever for the as-deposited films but strong and sharp peaks after annealing at
200°C. Specific studies of the effect of annealing on the electrical and photocon-
ducting properties of these films annealed in argon or hydrogen [31] and in oxy-
gen [32] have been described but will not be discussed further here.
       The rate of deposition of these films increased on exposure to light. This
phenomenon is often observed in CD films and is believed to be due to photogen-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
erated charge formation in the growing particles, leading to an electrochemical de-
position process in parallel with the CD one. This phenomenon has been exploited
to generate photographic images by shining light through a mask (in this case, a
photographic negative) onto the depositing film [22]. The films in this case were
deposited on a predeposit of ZnS to improve adhesion. The contrast (mainly in
transmission but also seen in reflection) between the thinner and thicker regions
of the film reproduced the image on the negative. Changes in bandgap due to size
quantization may also have contributed to the contrast (see Chap. 10).
       It was mentioned that better adhesion was obtained on glass if ZnS was pre-
deposited first [22,33]. Improved adhesion was also obtained from films deposited
from this bath if the glass substrates were first treated with an organosilane [22].
The silane binds to the glass surface, and the growing film anchors to terminal
amino or thiol groups, both of which (the thiol in particular) bind strongly to metal
chalcogenides.
       A fairly strongly acidic thioacetamide bath (pH between 1 and 2) was de-
scribed in Ref. 23, both with and without EDTA (sodium salt) as a complexant
While some structural differences as well as variations in film thickness were
noted between the films deposited from baths with and without EDTA, they were
not highly significant. As with most other acidic baths, the films were clearly crys-
talline, showing defined XRD peaks that allowed a crystal size of the order of 10
nm to be estimated.
6.4.1.3      Thiourea
Triethanolamine was used in Ref. 25 as complexant together with ammonia, the
latter to slow the reaction, and it also apparently improved adherence. The solu-
tion was heated to boiling for 40 min and then left at room temperature for 4 hr.
As for other depositions using this strategy, no rationale was given for this regime,
although some hints as to its reason might be gleaned from the follow-up paper
([26]; see next paragraph): It can be presumed that it provided better films than
those obtained by simply depositing at one temperature. No XRD pattern was ob-
tained as for other alkaline bath depositions.
       In a follow-up investigation [26], the initial pH was 10.17, and this dropped
somewhat during the deposition. The deposition was dependent on the age of the
Bi(NO3)2/triethanolamine solution. A solution aged for ca. 5 hr before making up
the deposition solution gave films about twice as thick (0.1–0.2 m) as a freshly
made one. However, if aged for 24 hr, the color of the heated solution, instead of
turning brown, became white, and no film formed. This was explained by in-
creased hydrolysis of the Bi/triethanolamine solution on aging; it seems that some
hydrolysis is good for the process, but too much prevents adherence and therefore
film formation.
       A solution similar to the previous ones was used in Ref. 27. One major dif-
ference between the films obtained in this study and the previous ones is the film



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
thickness—ca. 1 m thick after 30 min at 95°C, compared to an order of magni-
tude less after 40 min in a boiling solution, then 4 hr at room temperature. The
most likely difference, in the absence of a knowledge of the triethanolamine con-
centration used in the present study, is that this triethanolamine concentration was
lower in the present study and therefore thicker films could be obtained in a
shorter time, since the the Bi would be less complexed. This is supported by the
fact that the Bi(NO3)3 was triturated with the triethanolamine, which suggests that
only enough triethanolamine was added to dissolve the Bi(NO3)3.

6.4.2 Bi2Se3 (See Table 6.3)
There are only a few papers dealing with Bi2Se3, and therefore they will all be
treated in one section. The first report was based on the early Bi2S3 depositions,
using a triethanolamine/ammonia bath and selenosulphate [34]. Films were de-
posited both with and without hydrazine. The deposition was faster with hy-
drazine, as may be expected, although in this case the films lost adherence if left
in the solution more than ca. 30 min. Also, while the films deposited from the
hydrazine bath were single phase, traces of elemental Se were found in the hy-
drazine-free bath. The large difference in resistivity between films deposited from
the two baths is interesting, although no reason for this difference was suggested.
In contrast to the corresponding deposition of Bi2S3, where the age of the Bi/tri-
ethanolamine solution was important (see earlier), the age of the Bi solution is not
critical to obtaining a good film; in fact, thicker films were obtained using even
strongly aged Bi solutions [26]. This was ascribed to the low temperature used in
the formation of the selenide compared to the sulphide.
       Films were also made using N,N-dimethylselenourea as Se source [35].
Na2SO3 was added, as usual for selenoureas, to minimize oxidation of the sele-
nourea. X-ray diffraction showed only a very broad and ill-defined spectrum of
the as-deposited film. As for Bi2S3, annealing at a relatively low temperature
(200°C) was sufficient to crystallize the film and show well-defined peaks. The


TABLE 6.3 Bismuth Selenide

                                Bandgap         Resistivity
Solution                         (eV)            ( -cm)           Type         Ref.

Na2SeSO3 /TEA-NH3 /              1.15*         ca. 104–105       n-           26, 34
  30 C/pH 9.9 0.1
  hydrazine                      1.03*           5    102
Na2SeSO3 /NH3 /RT                1.42                                         28
DMSeU*/TEA/RT—40 C              1.7–1.4*             107                      35
Annealed 200 C                 1.57–1.07              0.1        n- dir.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
crystal size in the annealed film, estimated from the peak widths, was ca. 12 nm.
The properties, in particular the bandgap calculated from the optical spectra, were
dependent on the film thickness. Thus the range of bandgaps given in the table
vary from a film 0.09 m thick (higher bandgap) to 0.15 m (lower bandgap).
This is a very large range for such a small change in thickness. Since the varia-
tions are probably due to size quantization, a measure of the XRD peak widths for
the different thicknesses (of minimally annealed films) would be of interest here.
The as-deposited films were very photosensitive (light-to-dark ratio up to 70), but
the annealed films were much less so.


6.5 COBALT
       In the early work by Beutel and Kutzelnigg on film formation on various
metals from hot thiosulphate solutions of various metal compounds, one of those
that resulted in apparent CD was Co [2]. Since no characterization of the films was
given and it is possible that the coloration was due to reaction between the thio-
sulphate and the substrate metals, this author made a single experiment, mixing
Co2 with excess thiosulphate and heating in a glass vial. A black film formed on
the glass (this was also carried out for Ni and Fe; see later). Although no charac-
terization was made of this film, it indicates that Co-S was indeed formed in the
experiment of Ref. 2.
       CoS was deposited at room temperature from a triethanolamine/ammonia-
complexed solution of CoCl2 using thioacetamide as sulphur source [36]. Both
compositional analysis (CoS1.035) and XRD analysis showed the formation of
CoS. From the optical spectrum, a direct bandgap of 0.62 eV was found. The films
were p-type with a resistivity of ca. 106 -cm.
       Metallic grey-brown films of Co3S4 were deposited from a CoCl2/NH3 so-
lution using thiourea at temperatures between room temperature and 50°C [36a].
The film resistivity was approximately 105 -cm.
       CoSe was deposited using a similar composition to that for CoS, except that
selenosulphate was used in place of thioacetamide, NaOH and hydrazine were
also added to the solution, and the deposition was carried out close to 100°C [37].
In contrast to CoS, no XRD pattern was observed for the as-deposited films; an-
nealing at 280°C crystallized the films to give a defined pattern corresponding to
CoSe. A direct bandgap of 0.45 eV was estimated from the optical spectrum. The
films were p-type and the resistivity 104 -cm.


6.6 COPPER
Copper chalcogenides can be readily deposited by CD. There is a strong affinity
between Cu and S or Se; metallic Cu exposed to elemental S dissolved in a sol-
vent (e.g., dimethyl sulphoxide) will quickly turn black due to formation of cop-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
per sulphide. This strong affinity between the elements is manifested as a low-sol-
ubility product of the various Cu chalcogenides. A complicating factor in this de-
position, however, is the large number of different phases and stoichiometries that
can exist. Four well-defined room-temperature phases of the CuxS system are
known: chalcocite (x 2); djurleite (x 1.96); digenite (x 1.8), and covellite
(x 1). Mixed phases can also occur, of course. Since only CuS gives an identi-
fiable XRD pattern in the CD Cu-S films, a knowledge of which phase(s), other
than CuS, exists is unknown, although the average composition has been mea-
sured in many cases. The various phases can, in some cases, be changed from one
to another after deposition. This has been shown for Cu-Se using electrochemical
polarization [38] and by aging and then reversed by heating [39]. Treatment of
CuxS film with gaseous H2S apparently changes the composition (possibly also
the phase) toward a more S-rich film [40].
       Cu-S shares with PbS and Sb-S the distinction of being the first published
CD compound [1]. This and (for many decades) subsequent reports involving CD
Cu-S described decomposition of thiosulphate solutions of Cu salts to give Cu-S
films. These (and other, mainly PbS) films were known as lüsterfarben (lustrous
colors) due to the varied interference colors obtained on metal substrates by de-
position of PbS or Cu-S (see Sec. 2.1 and Sec. 5.2 for more details of the history
of these lüsterfarben). As for PbS, very little characterization of these deposits was
reported in those early papers apart from their various colors.
       If this early work on CD of Cu-S was driven by the attractive colors they im-
parted to metallic substrates, more recent studies were driven initially by their po-
tential use in Cu2S/CdS photovoltaic cells (these cells are no longer studied to any
extent due to their perceived instability, although, with what has been learned
about Cu-containing chalcopyrite, such as CuInSe2, thin-film cells over the past
couple of decades, it would not be surprising if such studies were again pursued)
and later in solar control coatings (see Sec. 2.13).
       Thiosulphate and sulphite are sufficiently reducing to reduce Cu2 to Cu .
Therefore the Cu in solutions of Cu2 containing sufficient thiosulphate, seleno-
sulphate, or sulphite should be predominantly in the monovalent form. This would
lead to the expectation that the main product will be something close to Cu2S(e).
While this is often the case, CuS(e) is deposited in some cases. However, it is ar-
guable whether this reduction of Cu2 is, in fact, important in practice. The rea-
son is based on an XPS study that showed that Cu in its compounds with S, Se,
and Te is normally in the monovalent state; it is the chalcogenide ion (or polyion)
that is believed to change oxidation states in these compounds [41].
       An interesting characteristic of CD Cu-S films deposited from thiosulphate
solution is the range of compositions that can be obtained by varying the deposi-
tion conditions [40]. Elemental analyses of the precipitated CuxS powders ob-
tained by heating a solution 0.1 M each in CuSO4 and Na2S2O3 showed that the
composition varied from x 1.7 to x 1.0, with longer reaction times and higher



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
temperature giving lower values of x. X-ray diffraction of these end-members was
consistent with a phase where 1.86 x 1.96 (therefore the Cu1.7S was a mix-
ture of predominantly this phase, presumably djurleite, and other phases) and x
1.0, respectively. XPS of films deposited from the same solutions showed the sul-
phur of the CuS to have a higher binding energy (more oxidized) than that of the
Cu1.7S, in agreement with a variable valence state of the chalcogen in these com-
pounds. Variation of the Cu:thiosulphate ratio also affected the composition.
Compositions corresponding to Cu2S (Cu:thiosulphate 1:2); Cu1.8S (1:1); Cu1.4S
(1:2.5), and CuS (1: 3) were obtained, although in film form, only CuS gave an
XRD pattern [42,43]. The variation in composition is not a monotonic function of
the Cu:thiosulphate ratio. For equimolar Cu and thiosulphate, there is not enough
thiosulphate to both reduce all the Cu2 and form the sulphide, hence a partially
reduced Cu2 is formed; while for a ratio of 1:2, there is enough thiosulphate to
both reduce all the Cu2 to Cu and to form the sulphide. More difficult to un-
derstand at first sight is why an increasing concentration of thiosulphate appar-
ently oxidizes the Cu-S to an increasingly greater amount of CuS, until at a ratio
of 1:3, only CuS is formed. However, it must be kept in mind that CuS is not sim-
ply composed of divalent Cu, but is either 2Cu S 2 or a mixed-valence com-
                                                      2
pound with both S2 and S 2 groups. Thus it may be more useful to consider the
                             2
effect of the thiosulphate on the sulphur species. A possible hypothesis is that ex-
cess thiosulphate results in the formation of elemental S, which can react with S2
to give the polysulphide ion, S 2 , which exists in CuS. More generally, and re-
                                  2
gardless of the specific mechanism, a higher concentration of S relative to Cu can
be expected to favor more S in the final product.
       The various Cu-S and Cu-Se films generally exhibit similar optical spectra
for comparable thicknesses. Figure 6.1 shows some such spectra. The peak at ca.
0.6 m is characteristic of these films, and the drop in transmission at longer




FIG. 6.1 Optical transmission spectra of various CuxS films. (a) Cu2S; (b) Cu1.8S; (c)
Cu1.4S; (d) CuS. (Adapted from Ref. 43, with permission from Elsevier Science (USA)).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 wavelengths is presumably due to absorption and reflection by free holes
 (CuxS(e), as for the other Cu-S(e) compounds, is p-type and in most cases highly
 degenerate, therefore relatively highly conductive). The CuS (covellite) phase
 tends to be somewhat more conductive than the other sulphide phases.
        CD Cu-S(e) films have been proposed for a number of different potential
 applications. Solar control coatings, where the visible and IR transmission and re-
 flectivity can be varied, is probably the most studied, e.g., Refs. 44 and 45. The
 relatively high conductivity and the partial transmittance in the visible spectrum
 are useful for transparent conductors [46]. Other possible applications are for
 Cu2 sensor electrodes and electrical contacts for ceramic devices [46].
        In the tables for both Cu-S and Cu-Se (Tables 6.4 and 6.5), the column de-
 noting conductivity type has been deleted (these semiconductors are always p-
 type), and, in its place, the phase (and/or composition) has been given. In some
 cases, particularly for the sulphides, where no XRD pattern was seen (except for
 CuS), no phase (or composition) was proven and therefore no entry is given in the
 table. This was not a problem for Cu-Se, since XRD spectra were always clear and
 definitive.

 6.6.1 Cu-S (See Table 6.4)
 In Ref. 9, no structural or compositional characterization was given; Cu2S was as-
 sumed. The films were not uniform and did not adhere to the substrates. The high
 values of resistivity are unusual for this material: it may be that 10 3–10 4 was
 intended.

TABLE 6.4 Copper Sulphide

                            Bandgap         Resistivity           Phase or
Solution                     (eV)            ( -cm)             composition     Ref.

S2O3 /RT/pH ca. 2         1.2                104–105*                          9
S2O3 /60 C/pH 0.5         2.4 dir.             10 4                            47
S2O3 /60 or 70 C/pH 2–5                     10 2–10 4*       Varied            40
S2O3 /50 C/pH 5 (AcH)     1.7–2.0          10 3 (CuS)        Varied            42, 43
                                           3 10 3 (Cu2S)
S2O3-dimethylthiourea     1.55 indir.          8 10 3        CuS               48
Anneal 300 C              1.55 indir.          8 10 4        Cu1.8S
Anneal 400 C              1.4 indir.           10 3          Cu1.96S
TU/NH3 /30°C              ca. 2.26 ind.          0.25                          49
TU/TEA-acetate-NH3 /      2.58 dir. and        3 10 3        Cu1.86S           50
  40 C/pH 9.4                ca. 1.7*
CuCl/TU/EDTA/NH2OH/       1.45              10 3–10   4
                                                             CuxS              51
  RT—80 C/pH 8.5–11.5                                        1.83 x     1.85
TU/TEA-NH3 /RT            ca. 1.5 indir.    0.004     1                        52




 Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In Ref. 47, as before, no structural or compositional characterization was
given, and the films were classified as CuxS. Average grain size, measured by
scanning electron microscopy (SEM), was ca. 70 nm.
       In Ref. 40, films were deposited on various polymer substrates that in
most cases required pretreatment with either various organosilanes or poly-
(ethyleneimine) in order for good film formation to occur. This technique appears
to constitute a case of adsorption of colloids from the solution rather than forma-
tion of Cu-S directly on the substrate. For example, MNH groups on the imine-
coated substrate become protonated in the acidic solution and attract the appar-
ently negatively-charged Cu-S particles. The composition varied with
temperature and time of deposition (see the previous general discussion). The re-
sistivity of the films varied with the deposition conditions and substrate; CuS was
more conducting than compositions closer to Cu2S.
       In Refs. 42 and 43, composition depended on the Cu:thiosulphate ratio (see
the earlier general discussion). Only CuS gave a measurable XRD pattern. CuS
was less transmitting in the IR region than CuxS (x         1.4–2.0), and for CuS,
greater thiosulphate concentration (constant [Cu]) resulted in less transparency in
the IR, although film thickness was fairly constant. This is expected from the data
of Ref. 40, where greater thiosulphate concentration resulted in a deposit closer to
CuS in composition and with higher electrical conductivity (therefore less trans-
parent in the IR).
       A combination of thiosulphate and dimethylthiourea was used in Ref. 48.
Using only thiosulphate resulted in slower deposition and, more importantly,
poorly adherent films. The resistivity of the as-deposited CuS dropped to
3 10 4 -cm on annealing at 200°C (in N2) without a change in phase or com-
position. Annealing at 300 and 400°C resulted in loss of S and phase changes. The
crystal size was 11 nm (200°C anneal), 19 nm (300°C), and 20 nm (400°C). Pre-
sumably no clear XRD pattern was obtained for the as-deposited film of CuS.
       In Ref. 49, the composition of the films was given as Cu1.8S. However, no
XRD pattern was found and no compositional analysis given, and therefore it is
unclear just what the actual composition and phase were.
       A potentiometric technique was used in Ref. 50 to measure composition,
found to be Cu1.86S. In this deposition, stirring the deposition solution resulted in
nonuniform and poor-quality films, while good films were obtained in unstirred
films. The bandgap was measured to be direct, with a value of 2.58 eV. This is a
particularly high value. Examination of the transmittance spectrum showed a
sharp drop in transmission at ca. 1.7 eV, which is more likely to be the true
bandgap.
       In Ref. 51, the composition was CuxS, with 1.83 x 1.85. The films were
thicker than most others—several microns. The terminal thickness was very pH
dependent; at pH 8.5, it was ca. 0.5 m, and at pH 11, an order of magnitude




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
greater. A pH of 10 was optimum in terms of maximal thickness without exces-
sive bulk precipitation.
      No structural or compositional data was given in Ref. 52. Although a
value of bandgap was not given, from the optical spectra, an indirect bandgap of
between 1.5 and 1.6 eV could be estimated. The specific resistivity dropped with
increasing film thickness by more than two orders of magnitude, 1 -cm for
a thickness of 0.15 m to 4         10 3 -cm for 0.35 m. It was suggested that
this was due to increasing nonstoichiometry as deposition proceeded. If a sec-
ond deposition was carried out on a previous one, the resistivity (measured
across the film) was characteristic of a single layer, suggesting that a relatively
insulating layer was deposited on the first layer in the early stages of the second
deposition.
      This group published a number of other papers on CuxS, using the same ba-
sic deposition solution, with emphasis on variations of their spectral properties for
possible use in solar control coatings. Some examples are given in Refs. 44, 45,
and 53 (the last deals with films deposited on Kapton foil). This last showed some
weak CuS (covellite) peaks in the XRD spectrum; no inference of whether this
was the major phase or not could be made.


6.6.2 Cu-Se (See Table 6.5)

The three main phases encountered in CD Cu-Se films are Berzelianite (Cu2 xSe,
where x is typically ca. 0.2); Umangite (often written as Cu3Se2 but may be con-
siderably lower in Cu) and Klockmannite (CuSe).
       In Ref. 54, XRD showed the deposit to be hexagonal CuSe. Analysis of the
absorption spectrum gave a direct bandgap of 2.02 eV. As commonly seen for
these compounds, there was still strong absorption at lower energies (e.g., at 1.9
eV, the absorption coefficient was 7 104 cm 2), possibly due to an indirect
transition but likely due, at least in part, to free-carrier absorption. From Hall mea-
surements, the doping (acceptor) density was ca. 1022 cm 2 (heavily degenerate)
and the mobility ca. 1 cm2V 1sec 1. The dependence of film thickness and depo-
sition rate on the deposition parameters has been studied in a separate paper [62].
       Nitrilotriacetate was used as complexant in the deposition in Ref. 55. Cu-Se
could be both electrodeposited and chemically deposited from this solution. The
electrodeposited film was Cu1.8Se with the berzelianite structure, while the CD
one was Cu1.2Se with the umangite structure. The XRD pattern of the CD films
showed sharp peaks (instrument broadening) with no preferential texture. Elec-
tron microscopy of these films (Fig. 6.2) shows large (micron scale) particles that,
from their faceted shape and together with the sharp XRD peaks, appear to be sin-
gle crystals. This is a particularly large crystal size for a CD film; from this and




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 6.5 Copper Selenide

                            Bandgap           Resistivity
Solution                     (eV)              ( -cm)                        Phase           Ref.
                                                            3
SeSO3/TEA NH3/RT          ca. 2.0 dir.*            10                CuSe                    54
SeSO3/NTA*/RT/pH 9                                                   Cu1.2Se umangite        55
SeSO3/TEA NH3/            1.20 dir.       10 3–2            10   2
                                                                     Cu1.86Se berzelianite   56
   95 C/pH ca. 10
SeSO3/NH3/ 45 C/                          ca. 10        -sq.         Cu2Se                   57
  pH 10
SeSO3/pH 10                               3    10 3–10 4*            Cu2Se or CuSe*          58
SeSO3/TEA NH3 /75 C                                                  Cu2 xSe berzelianite*   38
                                                        4
SeSO3/NH3/RT                2.36 dir.     2    10                    Cu2 xSe berzelianite    59
                            1.9 indir.*
SeSO3/citrate/5–27 C/     See text        ca. 2 10 3                 Cu2 xSe berzelianite    39
  pH 7                                      (for all films)          Cu3Se2 umangite
SeSO3/citrate/60 C
  pH 9                    2.37                                       Cu3Se2 umangite         60
  pH 12                   2.0                                        CuSe klockmannite
DMSeU/tartrate/50 C                       10 3–10 4*                 CuSe                    61
DMSeU/tartrate/50 C       ca. 2.13          2 10 4                   CuSe klockmannite       59



 the other studies on CD Cu-Se, it can be seen that Cu-Se has a tendency to form
 relatively large crystals. This may be due, at least in part, to the high mobility of
 Cu (although the tendency is much less for Cu-S) and the relatively low melting
 points of the Cu-Se compounds in general.




 FIG. 6.2 Scanning electron micrograph of a Cu1.2Se film deposited from a selenosul-
 phate solution of Cu2 complexed with nitrilotriacetate.



  Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       The main difference between Ref. 56 and Ref. 54 is the higher deposition tem-
perature of the former. Elemental analysis gave a composition of 65% Cu and 35%
Se (at.%) corresponding to Cu1.86Se. X-ray diffraction confirmed the cubic
berzelianite phase.
       The deposition in Ref. 57 appears similar to those of Refs. 54 and 56, except
that only ammonia was used to complex the CuSO4 and the deposition tempera-
ture was ca. 45°C. The films were deposited on polyester (overhead transparency)
films. From XRD, a composition of Cu2Se (no specific phase given) was assigned.
The transmission spectrum was similar to those of Cu-S in general, with a strong
free-carrier absorption beginning in the near IR and a strong absorption (presum-
ably bandgap) onset at ca. 700 nm.
       Reference 58 involved a similar deposition as the previous case, but with-
out ammonia. The composition of the Cu-Se was Cu2Se (using a Cu:selenosul-
phate ratio of 1:1) and CuSe (1:5). (This is the same trend as found for sulphides
deposited from thiosulphate solutions.) The transmission spectrum of the Cu2Se
was similar to that in the previous study, while that of the CuSe showed an ab-
sorption onset at ca. 600 nm. Both films showed strong apparent free-carrier ab-
sorption starting in the near IR (for the CuSe, even at somewhat shorter wave-
lengths). However this absorption appeared to be stronger for the Cu2Se than for
the CuSe, although the resistivity of the former (3 10 3 -cm) was higher than
that of the CuSe (10 4 -cm).
       Copper acetate was used in Ref. 38; it was noted that if chloride was used
instead of acetate, no deposition occurred, and this was attributed to adsorption of
chloride on the substrate (Pt). The berzelianite phase with a small amount of
umangite impurity was obtained. The composition and phase of the film could be
altered by electrochemical cathodic polarization (in an aqueous K2SO4 solution).
Initially, there occurred an increase in lattice parameters and decrease in x
(Cu2 xSe). With continued polarization, a phase change occurred until eventually
only orthorhombic Cu2 xSe was present in the film. The umangite phase also dis-
appeared, and it was believed that this impurity phase catalyzed the phase trans-
formation. The change in composition during cathodic polarization was attributed
to reduction of zerovalent Se to Se2 , which was dissolved in the solution. Based
on the study of Folmer and Jellinek [41] discussed earlier, this explanation can be
interpreted as reduction of Se 2 (“monovalent” Se) to Se2 (divalent Se).
                               2
       Reference 59 provides a comprehensive explanation of the optical spectra
and extracted bandgaps. The direct bandgap of ca. 2.36 eV is compared to the lit-
erature value of ca. 2.2 eV and explained by size quantization in the fairly small
(20 nm) crystals. An indirect bandgap of 1.9 eV was measured (literature value
1.4 eV), but it was stressed that this provided an upper limit only, since the ab-
sorption in this region was dominated by free-carrier absorption, which masked
the indirect absorption. Annealing decreased the conductivity and the free-carrier
absorption and changed the indirect gap to 1.3 eV.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In Ref. 39, citrate was used a complexant and the pH was lower than in other
similar studies. The depositions were carried out at around room temperature or,
if the deposition was too fast, at lower temperatures (no difference in the nature of
the films was found with different temperatures). The composition and structure
of the deposit was found to be pH dependent: At a pH ca. 7 or lower, berzelianite
Cu2 xSe was deposited, while at a slightly higher pH (7.8), the product was uman-
gite (ca. Cu3Se2). The exact compositions varied with change in ratio between Cu
and selenosulphate concentrations. As with the other Cu-Se and Cu-S films,
bandgap determination was complicated by the strong free-carrier absorption. Di-
rect bandgaps of 2.2 eV (Cu2 xSe) and 2.8 eV (Cu3Se2) were measured from the
transmission spectra. However, from examination of these spectra, it can be in-
ferred that a strong absorption, not arising solely from free carriers, occurred at
lower energies. An approximate reanalysis of the transmission spectra, taking into
account free-carrier absorption, allowed estimation of indirect bandgaps of
1.5–1.6 eV (Cu2 xSe) and 2.0–2.1 eV (Cu3Se2).
       The berzelianite phase was subsequently found to slowly transform to the
umangite one under ambient conditions [63]. By heating at 140°C in air, this phase
transformation could be reversed. These phase changes could be repeated in a
cyclical manner.
       In Ref. 60, the differences in the two solutions giving Cu3Se2 and CuSe were
the lower pH (9) and higher Cu and citrate concentrations (6 mM) for Cu3Se2,
compared to pH 12 and Cu (and citrate) concentrations of 4 mM (constant se-
lenosulphate concentration of 30 mM in both cases). The films were deposited on
flexible polyester substrates. It was noted that the deposition was paralleled for the
most part by bulk precipitation. The average crystal size of both Cu3Se2 and CuSe
was 42 nm.
       In Ref. 61, N,N-dimethylselenourea was used (together with CuCl2 instead
of CuSO4 commonly used in the selenosulphate depositions). Film (specific) re-
sistivity dropped as thickness increased.
       Reference 59 is similar to the previous study. Values for film resistivity
were not given, but it was noted that the films were less conductive than Cu2 xSe
films (2 10 4 -cm) made in the same study using selenosulphate instead of
the selenourea.


6.7 INDIUM
Indium is very readily hydrolyzable, with a pKa of 4.0, and forms the hydroxide
even in moderately acidic solutions (see Sec. 1.1.2). This means that unless depo-
sition is carried out in strongly acidic solution, some hydroxide is likely to be pre-
sent in any chalcogenide formed by CD. This is indeed the case in most studies of
In2S3 deposition reported up to now.
       Films of In2S3 on glass deposited from a solution of InCl3 and thioac-
etamide were described as early as 1976 by the Kitaev group [64]. Broad XRD


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
peaks corresponding to -In2S3 with a crystal size of ca. 5 nm were obtained. A
bandgap of 2.45 eV could be estimated from the optical spectra. The resistivity
was very high—ca. 1012 -cm at room temperature—with an activation energy
of 1.2 eV (implying a midgap Fermi level and highly intrinsic material). Films
annealed at 250°C were photosensitive, with a photoconductivity maximum at
ca. 500 nm (2.5 eV). It was noted that this was blue-shifted by 0.13 eV, com-
pared to single-crystal In2S3. The small size of the crystals suggests that size
quantization occurs here.
       In2S3, or, as shown later more probably In(OH,S), was deposited by CD for
use as a buffer layer in photovoltaic cells [65,66]. The deposition bath was again
InCl3 and thioacetamide operating at a temperature of 70–80°C. Analysis of these
films by XPS showed that oxygen was present in the films, presumably as hy-
droxide [67]. Importantly, the results were inconsistent with a mixture of sulphide
and hydroxide (which might be expected from this bath) and suggested rather
some compound formation. (Details of photovoltaic cells using these films are
given in Chap. 9.)
       In the first of a series of studies, the same basic bath as previously, but us-
ing acetic acid to adjust the pH (to ca. 3, probably somewhat lower—see follow-
ing reference) was used [68]. The main purpose of the acetic acid is probably to
lower the pH and therefore to reduce the In3 hydrolysis, although its (weak) com-
plexing ability with In3 might also play some role in minimizing this hydrolysis.
The color of the film (and homogeneous precipitate) varied from whitish yellow
to yellow (the bandgap of In2S3 is ca. 2.4 eV), and XPS analysis showed that the
S:In ratio was always less than that expected for stoichiometric In2S3 (1.5), al-
though it was higher when the thioacetamide and acetic acid concentrations were
increased. Higher thioacetamide concentration increased the concentration of sul-
phide formed, while more acetic acid decreased the hydrolysis to hydroxide. The
resulting films were therefore believed to be composed of both sulphide and hy-
droxide, designated as Inx(OH)ySz. The bandgaps (indirect) varied between 2.0 eV
(highest S content) to 2.5 eV. For the film with the highest S content, the value of
2.0 eV is probably an underestimation, particularly taking into account the yellow
color typical of a film with a bandgap of at least 2.3 eV. The films were highly re-
sistive, between 107 and 108 -cm.
       An XPS investigation of these films was carried out [69]. The pH was more
accurately measured to be between 2.2 and 2.5. Also it was noted that, although
higher concentrations of acetic acid minimized the codeposition of hydroxide,
above 0.1 M acetic acid, the films were not homogeneous and poorly adherent.
From the XPS spectra, it was concluded the films were of the composition
In(OH)S, with small variations in the S:OH ratio. Sulphate, probably as surface
oxidized In-S, was also present.
       Structural (XRD and microscopic) studies of the films allowed more defi-
nite assignments to be made as to the identity of the films [70]. The structure was
dependent on the thioacetamide and acetic acid concentrations. At low concentra-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tions a composition identified as In5S4 was obtained, while at higher concentra-
tions a mixture of cubic - and -phases resulted. The former possessed a string-
like morphology, while the latter was typically composed of more or less spheri-
cal granules. The XRD peaks of the In5S4 deposit were broad, equivalent to a
crystal size of ca. 5 nm, verified also by TEM. The In3S3 deposits were apparently
of somewhat larger crystal size ( 10 nm).
      A study of the species present in these solutions and the mechanism of the
deposition has been presented [71]. Under the conditions of the depositions, the
main solution indium species (in the absence of thioacetamide) are In-Cl (mainly
[InCl2] ) complex species. Only ca. 1% of the total In content is present as free
In3 . No In(OH)3 or hydroxy-complexes were calculated to be present if acetic
acid was present (in the absence of acetic acid, the hydroxide could form). From
a kinetic analysis of the deposition reaction, it was concluded that the deposition
occurred by direct reaction between the thioacetamide and the chloro-indium
complexes. It was noted that thioacetic acid was the main by-product and that no
acetamide was detected (see Sec. 3.2.1.3 for a description of the possible mecha-
nisms and by-products of thioacetamide hydrolysis). Acetonitrile (CH3CN), a less
common by-product, was also detected at the higher pH values (these depositions
took place between a pH of 2 and 3) but not at the lower ones.
      A different study, using essentially the same deposition solution (InCl3
thioacetamide) at a pH of 3.1 or lower, has been described [72]. The films, de-
posited on ITO/glass adhered to the ITO side but not well to the glass. In this
case, compositional analyses showed the films, which gave electron diffraction
patterns and XRD spectra characteristics of In2S3 (mainly the -phase, possibly
together with the -phase) to be slightly S rich (the precipitate formed in solu-
tion was more or less stoichiometric). Thus, although there was no evident dif-
ference in the deposition parameters, these films appear different than the mixed
sulphide-hydroxide ones described previously. Microscopic investigation
showed the films to consist of a mixture of round particles and needles, forming
a porous, spongelike morphology. The films exhibited an increase in bandgap,
together with decrease in crystal size with decreasing deposition temperature,
due to size quantization (see Chap. 10 for more details). Decrease in solution pH
also resulted in a decrease in the bandgap. The bandgap varied between ca. 2.3
and 2.7 eV.

6.8 IRON
No clear-cut example of Fe-S has been described. The closest is in the report of
Ref. 2, where, among other metal salts, a boiling aqueous solution of iron thiosul-
phate imparted coloration to an iron substrate [2]. No confirmation was given as
to the composition or nature of this coloration, and a few attempts by the author
to deposit Fe-S on glass by heating thiosulphate solutions of iron salts were un-
successful (unlike the corresponding Co and Ni cases).


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
6.9 MANGANESE
The stoichiometry of Mn-S precipitated from solution is normally MnS. The sta-
ble form of MnS is green -MnS, which has the rocksalt structure. However, the
pink form, which is the form that usually precipitates from solution, is a mixture
of -MnS (zincblende) and -MnS (wurtzite), both of which are metastable.
MnSe behaves analogously.
       MnS was deposited from a room-temperature solution of Mn(II) acetate
complexed with triethanolamine and buffered with NH4Cl [73]. Thioacetamide
was used as a sulphur source, and hydrazine was also used (it was not specified
whether the reaction proceeded in its absence). No XRD pattern was seen in the
as-deposited (grey-pink) film; annealing at 500°C in an inert atmosphere gave a
pattern corresponding to MnS. A bandgap (indirect) of 3.25 was measured from
the optical spectrum. The film was p-type with a resistivity of ca. 105 -cm.
       Optimization of the film growth from the foregoing bath was carried out
[74]. In contrast to many other CD reactions, the growth rate decreased slightly
with increasing temperature (also, the terminal thickness was greater at lower tem-
perature—a common occurrence due to reduced bulk precipitation). It may be that
bulk precipitation was so rapid at higher temperatures that the thickness of the film
deposited at the higher temperatures was less by the time of the first thickness
measurement (10 min). While, as with the previous study, no XRD pattern was
seen in the as-deposited film on glass, a clear pattern was observed for films de-
posited on SnO2-conducting glass, showing a mixture of the cubic - and hexag-
onal -phases. However, TEM/ED showed the presence of 3- to 4-nm-sized MnS
crystals on glass. The optical bandgap (for the film on glass) was estimated to be
ca. 3.0 eV and direct. There does not seem to be a clear-cut value in the literature
for the bandgap of MnS, but it is has been given as 3.0 0.2 eV. The films showed
only weak photoconductivity.
       MnS has also been deposited from an alkaline (pH 9.7–9.8) thiosulphate
bath, using MnCl2 [75]. The deposition was carried out at room temperature after
initial heating at 70°C (this initial heating step was noted to be essential, although
no explanation for this was given). The XRD spectrum was barely indistinguish-
able from the noise; there was a possible correlation with the spectrum of -MnS.
Optical studies showed a direct bandgap of 3.1 eV, and the resistivity was mea-
sured to be between 107 and 108 -cm.


6.10 MOLYBDENUM
There are two main sulphides of Mo. The stable form is the black, layered MoS2,
commonly used as a lubricant. Precipitation from (acidic) solution normally gives
the amorphous MoS3, which converts to MoS2 on heating. An important issue
when using molybdates as a source of Mo is that solutions of molybdates do not
precipitate the sulphide (selenide) when sulphide (selenide)—either as H2S(Se) or


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
as an alkaline sulphide (selenide)—is reacted with the molybdate solution, but
rather form soluble thio(seleno)molybdate ions, such as MoS 2 .4
       Chemical deposition of both MoS2 and MoSe2 has been reported from am-
monium molybdate solution [76,77]. For the sulphur and selenium sources,
thioacetamide and selenosulphate were used, respectively. Ammonium hydroxide
was added to the sulphide solution, while an acetic acid/ammonium acetate buffer
was used with the selenide solution (pH values were not given). Reducing agents
(either hydrazine [76] or sodium dithionite [77]) were added to the baths. Deposi-
tion was started at 90–100°C, followed by lowering to room temperature.
       No XRD pattern was observed for the Mo-S deposit, but after heating (ap-
parently in the deposition solution) in an autoclave at 300°C, the XRD pattern of
MoS2 was obtained. The XRD pattern of MoSe2 was obtained for the as-deposited
film. It is possible that the as-deposited Mo-S was MoS3, which is often obtained
in an amorphous form from solution reactions at relatively low temperatures and
converts to crystalline MoS2 on annealing.
       The estimated bandgaps for the two materials were 1.17 eV indirect (MoS2)
and 1.14 eV direct (MoSe2). The latter is unusual, since this is the approximate
value of the indirect gap of MoSe2; the direct gap is substantially higher.
       While the role of the reducing agents (hydrazine and dithionite) was not ex-
plicitly discussed, it must be assumed that they play an essential role in forming
the Mo chalcogenides rather than the soluble thio(seleno)molybdate ion.
       MoSe2 was deposited from a Mo(VI) (the source used was not specified) so-
lution complexed with ammonia to give a hexammine complex and mixed with hy-
drazine and selenosulphate at 40°C [78]. The as-deposited films were XRD amor-
phous but converted to crystalline MoSe2 after annealing in N2 at 380°C. Elemental
analysis showed the as-deposited films to be nearly stoichiometric MoSe2. A direct
bandgap of 1.48 eV (1.36 after annealing) was measured. The films were n-type
with a resistivity of ca. 4 103 -cm (ca. 1 -cm after annealing).

6.11 NICKEL
Ni-S behaves rather similarly to Co-S (see Sec. 6.5, Cobalt). Note that the freshly
precipitated monosulphides of both metals transform in solution to a more insol-
uble form–possibly M(OH)S.
       As for cobalt (see earlier), an early study of the coloration of metals by im-
mersion in boiling metal salt–thiosulphate solutions resulted in coloration of the
metals [2]. Also as for cobalt, a single experiment by the author repeating (simi-
lar, not necessarily identical conditions) this experiment, only on glass, rather than
on metals which might be colored by the thiosulphate alone, resulted in a black
film on the glass.
       Other than the foregoing, only one other paper was found dealing with NiS
(and also NiSe) [79]. The baths were based on NiSO4, triethanolamine, and am-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 monia. For the sulphide, thioacetamide was used, while for the selenide, seleno-
 sulphate was the Se source, and NaOH and hydrazine were added to the bath. De-
 positions were carried out at room temperature. XRD confirmed the formation of
 NiS and NiSe films. The bandgaps (direct) were 0.35 eV (NiS) and ) 0.23 eV
 (NiSe). The films were both p-type, with resistivities of 10 -cm (NiS) and 0.1
   -cm (NiSe).


 6.12 SILVER
 6.12.1 Ag2S (see Table 6.6)
 The Ag ion forms strong complexes with thiourea (log K              12.7 for the
 Ag(thiourea) 3 complex and with thiosulphate (log K 13). The strong binding
 of Ag to thiosulphate is exploited in the use of Na2S2O3 solution to remove ex-
 cess Ag during the developing of photographic films. For this reason, Ag can



TABLE 6.6 Silver Sulphide

                            Bandgap                Resistivity    Conduct.
Solution                     (eV)                   ( -cm)          type      Ref.

S2O3 /RT/pH 2–3               1.2                  106 –107                  9
S2O3 /RT/pH 2.6 (2.2)*      0.95 dir.               103–104                  80
  EDTA                      0.73 dir.               104–105
S2O3 /NH3 /50 C/pH 9–11       2.2                     10                     11
        2
TU/S2O 3 /RT/pH 10.1*       2.3 dir.                  10 2           p-      81
TU/8–55 C/pH 8–10 (NH3)         —                  103–105*          n-      82, 83
  EDTA                      ca. 0.8 dir.            2 104
TU/NH3, kinetic study                                                        84
TU/NH3, thermodynamic                                                        85
  analysis
TU/NH3 /NH 4 /pH ca. 11,                                                     86
  structural study
TU/40–80 C/pH 9,                                                             87
  study of mechanism
TU/40–80 C/pH 9             ca. 1.0*                                         88
TU (NH4OH) Hg2 doped        0.8 (from            ca. 103 (dark)      n-      89, 90
                               photocond.)
TA/RT/8–55 C                0.95 dir.            ca. 105             n-      91
  EDTA                                           ca. 104
Dip technique Ag/S2O3 /RT   0.83 indir.             (100   )         n-      92
  TU/80 C/pH 8–11*




 Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
be kept in solution under alkaline conditions without the need for another com-
plexant. The strong binding of Ag to sulphur in thiourea and thiosulphate sug-
gests that the mechanism of Ag2S formation may be of the complex-decompo-
sition type rather than through formation of free sulphide. Thus any bond
breaking involving the Ag-complex is intuitively expected to occur at the SMC
or SMS bonds of the thiourea or thiosulphate, respectively (see Sec. 3.3.3.1,
Eqn. 3.5.5 and discussion following for more details on this topic).
       Two forms of Ag2S exist— -Ag2S (acanthite), a monoclinic form, and -
Ag2S (argentite), which is cubic. -Ag2S is the form that is stable at room tem-
perature and is invariably the one that occurs in CD films.
       Thiosulphate was used in an acid bath in Ref. 9. This study covered many
sulphides deposited using thiosulphate, and little detail was given on the deposi-
tion or on the films themselves.
       Equimolar quantities of Ag and thiosulphate were used in Ref. 80 (as for
the previous study), so the complexation of the Ag by thiosulphate was not as
strong as it would have been in an excess of thiosulphate. A suggested mechanism
for the deposition was reduction of elemental S to sulphide, formed in the acidic
thiosulphate solution, by the moderately reducing thiosulphate. It was stressed
that the thiosulphate was slowly added to the AgNO3 solution with heavy stirring,
with the implication that otherwise the thiosulphate would be oxidized. The films
were reported to be rough rather than the smooth specular films often obtained by
CD. Differences in properties were obtained if EDTA was added as an additional
complexant. The films with EDTA were somewhat thinner (0.14 instead of 0.19
  m), and XRD of the EDTA-free films gave sharp peaks, while those deposited
from an EDTA-containing bath apparently showed no pattern, hence were proba-
bly either very small crystalline or amorphous. As seen from Table 6.6, both ap-
parent bandgap and resistivity changed on addition of EDTA to the bath. The
higher resistivity with EDTA was explained by the smaller crystal size. The lower
bandgap is less obvious; very small crystal size would increase the bandgap, and
an amorphous semiconductor has often (although not always) a higher bandgap
than the crystalline form.
       The deposition in Ref. 11, from an alkaline thiosulphate bath, was reported
in the context of a general description of deposition of various materials by CD,
and only a little characterization was reported. X-ray diffraction showed some
Ag2S peaks. Optical spectroscopy showed a gradual decrease in transmission over
a wide spectral range, and it would be difficult to extract a reliable value for the
bandgap from the spectrum.
       Note that the deposition in Ref. 81 used thiosulphate as a complexing agent
and not ostensibly as a source of S. The thiourea concentration is critical. Thiourea
is added to the Ag /S2O3 just until some solid Ag2O is formed. Too little thiourea
results in thin, scattering films. Too much results in films that in their as-deposited
state are good, but after annealing (300°C), voids form. The pH is also critical:



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
Under 10 the deposition is slow, while above 10.2 it is too fast and bulk precipi-
tation dominates.
       In Refs. 82 and 83, this deposition was employed both with and without
EDTA, a strong complexant. The deposition was studied at various deposition
temperatures. Better adhesion was obtained at low temperature (8°C). The resis-
tivity of the films was dependent on deposition temperature: 2.103 -cm (8°C)
and 1.5 105 -cm at 25°C. For a film deposited from an EDTA-containing bath
at 8°C, it was ca. 2 104 -cm.
       The reaction kinetics were analyzed in an early study [84]. Of particular
note is the unusually high activation energy (160 kJ/mole)—about twice the nor-
mal value for reaction-controlled CD processes. This contrasts with the much
smaller value (20.4 kJ/mole) measured in Ref. 87, although there were several dif-
ferences in the deposition: The present solution contained ammonia, the pH was
higher—probably between 11 and 12 (compared to the borate buffered solution
with pH 9 of Ref. 87), and the activation energies were measured at lower tem-
peratures. It is interesting that the pH could not be measured directly in this study
using a pH meter since the pH electrode was apparently rapidly coated with Ag2S.
Based on the kinetic study, the overall reaction proposed was
      2Ag(N2H4CS) 3       2OH → Ag2S          5N2H4CS      NH2CN       2H2O (6.1)
While both ammonia and thiourea were present, it is probable that with sufficient
thiourea present, the main complexant was thiourea (see the next paragraph). This
would also explain the observed dependence of deposition rate on thiourea con-
centration: Initially the deposition increased (since thiourea is the source of S),
reached a maximum, and then decreased (due to increasing complexation).
       Reference 85 presents the thermodynamic side of the previous paper. It is
pointed out that although both ammonia and thiourea are present in the solution,
because of the much higher stability constant of the Ag-thiourea complexes com-
pared to the Ag-ammines, essentially all the Ag will be present as a thiourea com-
plex. In this case, it can be assumed that the role of ammonia is only to control pH.
       An interesting observation in Ref. 86 was that the density of nuclei formed
in the early stages of film deposition did not change with time. The film developed
by growth of the initially relatively small (ca. 20 nm) nuclei. This suggests an ion-
by-ion type of growth rather than a cluster one.
       Reference 87 is a mechanistic study of Ag2S deposition from a thiourea bath
(buffered to pH 9 with a borate buffer). There are some unusual properties of this
deposition. One, the unusually strong effect of stirring on the deposition rate, has
already been dealt with (Sec. 3.7) and, together with the measured activation en-
ergy of 20.4 kJ/mole, suggests a rate-determining diffusion step in the deposition.
Another observation is that the films reach a maximum thickness (between 0.5 and
1.6 m) after 30 min of deposition (the thickness increases with temperature but
the 30 min is, surprisingly, temperature independent) and then become thinner



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
with further time in the bath, presumably because of loss of adhesion of the film
due to increasing stresses in the film. The pH is fairly critical: 8.8 slows the
growth down greatly, while 9.4 results in a fast homogeneous precipitation in
solution (such precipitation occurs, even at the optimum pH of 9, in parallel with
film formation).
       The properties of the foregoing films (only without the borate buffer) were
described in Ref. 88. The XRD peaks of the as-deposited films were narrow and
sharp, evidencing relatively large crystal size. The bandgap of the as-deposited
film, measured from transmission spectra, was ca. 1.0 eV but varied somewhat with
deposition temperature: At 40°C it was 0.91 eV and reached a maximum of 1.02 at
60°C and then decreased slightly at still higher temperatures. The absorption onset
was sharper at higher temperatures, which was interpreted as being due to denser
films. The refractive index was also slightly higher for higher deposition tempera-
ture, again explainable by the same rationale. The films, annealed in N2 at 250°C,
were photoconducting, and the photoconductivity spectrum was similar to the ab-
sorption spectrum. Time-resolved microwave conductivity measurements were
carried out on the films. Fast decay times and moderately good mobilities were
found from these measurements. In particular, the mobility was very temperature
dependent, and the highest value (5.3 cm2V 1sec 1) was obtained for annealed
films that had been deposited at 60°C; both higher and lower deposition tempera-
tures gave much lower mobilities. While not understood, this dependence empha-
sizes the need to optimize these films specifically for any particular application.
       A study of the photoconductivity of Ag2S doped with Hg2 or Au3 was de-
scribed in Ref. 89. Illumination decreased the resistivity typically 2–3 times. Pho-
toconductivity spectra showed best results for Hg doping; Au doping gave a
higher peak sensitivity but a narrower spectrum, with lower sensitivity at shorter
wavelengths compared even to undoped films [90].
       Reference 91 involved an acid bath (although pH was not given) using
thioacetamide as the source of S. The terminal film thickness was greater for
lower-temperature deposition; films thicker than 3 m were obtained at 8°C. In
contrast to the acid thiosulphate bath, the use of EDTA decreased the resistivity,
as did deposition at lower temperatures. Photoelectrochemical activity was found
for these films (see Chap. 9).
       Reference 92 describes not a normal CD process, but one closer to the
SILAR technique described in Sec. 2.11.1. However, while the SILAR method in-
volves dipping the substrate in a solution of one ion (e.g., sulphide), rinsing to re-
move all but (ideally) a monolayer of adsorbed ions and then dipping in a solution
of the other ion (e.g., Ag ), the present technique omits the intermediate rinsing
step. This means that a relatively large amount of solution can remain on the sub-
strate between dips, and layer formation proceeds much more rapidly than for
SILAR, albeit with less control. A typical rate was 4 nm/dip cycle. In this case, a
visible layer of Ag2S formed after several dips. Since interference colors were ob-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tained, the films were smooth, because such colors are not seen on rough, highly
scattering films. The film thickness/number of dips increased with increasing pH;
however, the best films were obtained at pH 9.

6.12.2 Ag2Se
Ag2Se films were first deposited from a bath using selenosemicarbazide as a Se
source and thiourea to complex the Ag at 20°C [93]. The films were specular and
had a resistivity of between 2 and 20 -cm.
       The only other true CD of Ag2Se describes films deposited on polyester sub-
strates from an ammoniacal AgNO3 solution with selenosulphate at 0°C and a pH
of 10–11 [94]. They were strongly (111) textured, with a crystal size of 9 nm. The
optical bandgap (direct) was estimated to be 1.8 eV, compared to the normal value
of ca. 1.3 eV. This was attributed to size quantization. The absorption spectra
showed considerable absorption (scattering?) at longer wavelengths, which could
be due to a lower, indirect bandgap if not to scattering. The resistivity of the films
was ca. 2 10 3 -cm (200 -sq).
       A dip technique in which metallic Ag films were converted into Ag2Se was
described [95]. The Ag film was made by successive dipping of glass substrates in
a AgNO3 solution, followed by dipping in a solution of formaldehyde, and was con-
verted to the sulphide by treatment with a solution of SeO2. The films were rough
and apparently poorly adherent. The resistivity of the films was ca. 103 -cm.


6.13 THALLIUM
The few cases reported for CD sulphides and selenides of Tl all reported the
monosulphide (selenide)—TlS or TlSe. Tl can be monovalent or trivalent, and
these apparently divalent compounds are believed to be mixed-valence com-
pounds, with both Tl(I) and Tl(III) present.

6.13.1 TlS
TlS was deposited from a solution of TlNO3, ammonia, and thiourea at room tem-
perature (26°C) [96]. X-ray diffraction showed the formation of TlS. Optical spec-
troscopy (both transmission and diffuse reflection) allowed an approximate
bandgap of 1.0 eV to be estimated. The films were p-type, with resistivity of ca.
2 103 -cm. Photoconductivity was measured (although not quantified) with a
peak at ca. 1.2 eV (ca. 1 m) and extending from 1.0 eV to beyond 1.4 eV.

6.13.2 TlSe
TlSe was first deposited from a solution of thallium(I) acetate and selenosulphate
with added NaOH and hydrazine at room temperature [97]. The initial films were



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mirror-like but became thick (ca. 3–5 m) and matte black with increasing depo-
sition. X-ray diffraction confirmed the film to be tetragonal TlSe. The main pur-
pose for making these films was to study their photoconductivity. The films were
p-type, with a resistivity of 500 -cm, which decreased by a factor of 2 after an-
nealing (150°C in air). Indium doping (by adding In3 to the deposition bath) in-
creased the resistivity (of the annealed film) by an order of magnitude, probably
by introduction of compensating donors. The films, particularly the annealed,
doped ones, were highly photoconducting, with a maximum photosensitivity
(change in conductivity on illumination/dark conductivity) of 107 at a wave-
length of 1.1 m. The response extended to ca. 1.5 m (0.83 eV) (low-energy
side) and ca. 0.7 m (high-energy side).
       TlSe was also deposited from a solution of Tl2SO4 complexed with tri-
ethanolamine and ammonia and selenosulphate at 30°C [49]. Tetragonal TlSe was
identified by XRD. The bandgap was estimated at 1.12 eV; however, the absorp-
tion spectrum appears to show two transitions—one (possibly indirect) at 0.9 eV
and another at 1.3 eV. The films were p-type, with a resistivity of 105 -cm.
Considering the high carrier concentration measured (almost 1020 cm 3), this re-
sistivity value appears unusually high.


6.14 TIN
There are three sulphides of tin: SnS (grey, metallic; usually nonstoichiometric),
Sn2S3 (black), and the layered, yellow SnS2. SnS and SnS2 are formed when hy-
drogen sulphide is passed into solutions of Sn(II) and Sn(IV), respectively. The
analogous selenides also exist, although the existence of Sn2Se3 is apparently in
some doubt.
      Tin forms soluble thio(seleno)anions. The sulphides tend to be soluble in
very alkaline solutions.

6.14.1 Sn-S (See Table 6.7)
Deposition of Sn-S (from a thiosulphate bath) was claimed as far back as 1870 [98].
This was based on deposition from a boiling solution of “zinnsalz” [probably tin
chloride, but not clear whether Sn(II) or Sn(IV)] complexed with tartrate and us-
ing thiosulphate as the source of sulphur. Unfortunately, the substrate was brass;
since brass will slowly convert to a dark-colored sulphide upon immersion in boil-
ing thiosulphate solution and no characterization of the film was made other than
its color (which would vary initially according to thickness, due to interference ef-
fects), there is no evidence that the films were really a sulphide of Sn. In fact, an at-
tempt to reproduce these results concluded that there was no Sn in the layer [2]. On
the other hand, SnS2 films were later obtained using a thiosulphate bath, although
the solution composition was different (see later discussion of Ref. 99.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 6.7 Tin Sulphide

                                  Bandgap         Resistivity
Solution                           (eV)            ( -cm)        Composition    Ref.

S in RCOOH/SnCl2/ 90 *           See text                        SnS (p-)        100
                                                                            *
                                                                 Sn2S3; SnS 2
S2O3/SnCl4/RT/pH 1–2             2.35          103–104 (n-)      SnS2             99
TA/SnCl2/AcH/TEA NH3/RT          1.51 indir.   2.5 1010 (n-)     SnS (anneal)    101
TA/SnCl2·2H2O in acetone         1.3*          2 104–107 (p-)    SnS             102
  TEA NH3/RT—75 C
TA/SnCl4/EDTA NH3                2.3 dir.      1.2 (n-)          SnS*
                                                                    2            103
  hydrazine/RT/pH ca. 10




       Reference 100 described a different technique than usual, in that it used el-
emental S dissolved in a carboxylic acid. Propionic acid was most often used, al-
though other carboxylic acids could also be employed. It was noted that best re-
sults were obtained when the SnCl2 was added as a powder to a freshly prepared
S solution; aged solutions reacted much more slowly, if at all. It was surmised that
this was due to changes in the nature of the dissolved Sn2 with time, such as tin
oxide, or propionate formation and loss of HCl.
       The stoichiometry of the deposit was dependent on the water content of the
solution and on the presence of a complexant. In anhydrous solution and without
complexant, a uniform, brown film with the approximate composition Sn2S3 was
formed. Addition of ca. 1% of water resulted in a uniform, slate-gray film of
Sn1 xS. This was explained by the increased ionization of Sn species in water and
therefore increase in the concentration of Sn2 . A substantial increase in the wa-
ter concentration resulted in patchy films of an irreproducible nature, with both
brown and gray regions forming. Complexation of the Sn reduced the reaction rate
and allowed more time for the growing films to react with dissolved S, resulting
in films of approximate composition SnS2 (yellow) together with some Sn2S3. In
this case, some water was needed to form good films. An interesting and unusual
characteristic of this deposition method is that if the deposition is allowed to pro-
ceed for a long time, the amount of precipitate formed in the reaction is decreased
and the films grow thicker. This was explained through dissolution and reprecip-
itation of Sn-S solid phases via soluble thiostannate species. Dissolution was as-
sumed to occur preferentially in the bulk precipitate due to its greater accessible
surface area compared to that of the film.
       Another unique characteristic of this process is that a band of Sn-S above
the level of the deposition solution was frequently observed. This above-solution
film was ascribed to reaction of volatile SnCl4 and H2S. The presence of the for-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mer could be explained by reactions of the type
      2SnCl2     S → SnCl4      SnS                                             (6.2)
while the H2S (which could be detected separately) might be formed by SnCl2-re-
duction of S.
       X-ray diffraction showed the gray (low-water-concentration) films to have
the pattern of SnS, while compositional analysis showed them to be Sn deficient
by typically 10%. Crystals were large (micron-sized or larger). The other films
tended to be mixtures of approximate compositions Sn2S3 and SnS2, with a nee-
dle-like morphology (typically 1 m long by 0.1 m wide).
       The SnS had an indirect bandgap of 1.0–1.3 eV and was p-type. It was more
difficult to estimate the bandgaps of the other films due to their mixed nature.
However, approximate bandgaps of 1.8 eV (Sn2S3) and 2.4 eV (SnS2) could be es-
timated from the optical spectra.
       The possible mechanisms of this unique deposition are not considered in
Chapter 3 and therefore will be done so here. The reaction of metal salts with el-
emental S in nonaqueous solvents in which S dissolves is known, even if the
mechanism is not clear. In the present case, two plausible mechanisms can be
given. The Sn(II)/Sn(IV) redox potential is relatively negative ( 0.15 V vs. SHE
in aqueous solution; much more negative in alkaline solution, although this is not
relevant in the acidic conditions used here). Since only Sn(II) was added to the so-
lution, any Sn(IV) present will be formed in the solution and is likely to occur in
low concentration under most conditions. This means that, from the Nernst equa-
tion (see Chap. 1), the potential of the solution will be more negative than the stan-
dard potential, possibly by a large amount. Another, more negative potential that
may be relevant is the Sn2 /Sn0 ( 0.14 V; see later). The S/S2 standard poten-
tial is 0.45 V. Since only the oxidized form of this couple was added to the so-
lution, the redox potential (which for our purposes means the potential where
some appreciable concentration of S2 will be formed) will be considerably pos-
itive of this value. Add to this the known non-Nernstian behavior of the S/S2 cou-
ple when [S] [S2 ], when the potential shifts strongly positive to a greater ex-
tent than expected from the Nernst equation [104], and it is feasible that the Sn(II)
may reduce S to S2 in sufficient concentration to form Sn-S.
       By esentially the same reasoning, it could be argued that the Sn(II) might re-
duce itself (disproportionate) into elemental Sn (at low concentration). As just
noted, the standard potential of Sn(II)/Sn(0) is 0.14 V. It might be argued that
thermodynamically, this is more likely than the reduction of S (although the non-
Nernstian behavior of S/S2 will at least reduce this difference). Metals immersed
in nonaqueous solutions of S can react to form a layer of the metal sulphide (the
rate depending on the metal and on the temperature—e.g., Cu will readily sulphide
at room temperature, while sulphidization of Zn will proceed slowly even at high
temperature). The small Sn0 nucleii that may be formed in the disproportionation



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of Sn(II) would be chemically very active and more likely to react with the S in
solution.
      In connection with this mechanism, it has been reported that the reaction be-
tween elemental S and Sn in alkylammonium compounds or amines under hy-
drothermal conditions gives various organic tin-suphide species and even, under
some conditions (low pH), SnS2 [105]. A possible mechanism for this process was
proposed based on nucleophilic attack of the basic amine or hydroxide on the S8
chain:
      S8    RNH2 → SMS          SMNH2R                                         (6.3)
resulting in formation of polysulphide ions. This mechanism is unlikely in the
acidic solutions used for the CD process. However, according to the principles just
discussed based on the Nernst equation, elemental S in solution may be expected
to contain a (very low) equilibrium concentration of (poly)sulphide. If the sul-
phide is removed by reaction (with Sn2 , in this case), then even this very low
concentration may be enough to sustain the formation of a Sn-S solid phase.
Clearly, the lower the solubility product of the metal sulphide, the more likely this
process is to occur.
       Considering that homogeneous precipitation of metal chalcogenides
(mainly sulphides) by reaction between metal ions and dissolved chalcogen is well
established, the main difference between this deposition and similar reactions
seems to be that the products adhere to a substrate to give a visible film (in this
case) rather than only precipitate. Whether this is connected with the redissolu-
tion/redeposition process that occurs with the Sn-S system or has some other
explanation is important. If the former, it may be limited to only those systems that
behave similarly. Otherwise it is not unreasonable to expect that other metal sul-
phides and selenides (possibly also tellurides, although tellurium tends to be much
less soluble, if at all, in such solvents) may be deposited as films in this manner.
       In Ref. 99, yellow-gold films were obtained that gave no XRD pattern, but
the chemical composition (as well as color) was consistent with SnS2. Increase in
pH (more than 2) reduced the deposition rate, while increase in temperature led to
precipitation in solution and therefore thinner films.
       In Ref. 101, the films in the deposition were deep brown. No XRD pattern
was observed, but after annealing in an inert atmosphere at 410°C the pattern of
SnS was obtained and the stoichiometry confirmed by elemental analysis. The
bandgap (1.51 eV, indirect transition) was higher than the literature value (1.3
eV), and this was rationalized as resulting from the apparent amorphous structure.
The room-temperature conductivity (4.10 9 S-cm 1 a resistivity of 2.5 108
  -cm) is low for a relatively low-bandgap material, suggesting either a very stoi-
chiometric and intrinsic material or high grain-boundary resistance. The films
were photoconductive (the sensitivity was not given), with a spectral range from
550 to 1050 nm and a peak at ca. 850 nm.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In Ref. 102, the solution should be turbid; if too much triethanolamine was
added and the solution was clear, no deposition occurred. Too much NH3 (pH not
given) led to incorporation of hydroxide. The bandgap was given as ca. 1.3 eV;
however, from the optical spectra, there is a variation in the bandgaps, depending
on the deposition conditions and film thickness. In particular, the lower the tem-
perature of deposition, the larger the bandgap (which analyses of the spectra show
to be indirect and to vary from 1.45 to 1.2 eV). This behavior is typical of size
quantization. The resistivity varied strongly with film thickness; the values shown
in Table 6.7 are for 0.35 m (107 -cm) and 1.2 m (2 104 -cm). The films
were mildly photoconductive (maximum sensitivity 10).
       These SnS films (in one study, propylene glycol was used instead of acetone
to dissolve the SnCl2), coated with CD CuxS, were shown to possess spectral char-
acteristics favorable for various solar control purposes [106,107]. Depending on
the thicknesses of the two layers, films with varying absorption and reflection could
be obtained that might be suitable for solar collectors or for window glazing.
       In Ref. 103, the pH of the bath was rather critical; films deposited at pH
10.5 were powdery and poorly adhering, while at pH 9.5, no deposition was ob-
served. No structural or compositional characterization was given, but from the
transmission spectrum it could be assumed that the film was SnS2 (and also highly
scattering).
       Films designated as Sn(O,S) were deposited from a solution containing
tin(IV) acetate, HCl, and thioacetamide (the concentration of the latter two com-
ponents determining the S:O ratio) [108]. These films were prepared as buffer lay-
ers for photovoltaic cells (see Chap. 9), and little characterization of the films
themselves, other than some XPS, was reported. The XPS results suggested that
the films were a mixture of SnO2 and some Sn-S species.
       While not strictly CD, SnS has been deposited by an immersion technique
whereby a glass substrate was immersed in a cold sulphide solution, followed,
without rinsing, by immersion in a hot SnCl2 solution, and this cycle was repeated
to increase the film thickness [109]. The film properties, in particular the electri-
cal resistivity, were very dependent on the pH of the SnCl2 solution.

6.14.2 Sn-Se
Only one example of Sn-Se has been reported [110]. Films were deposited from a
room-temperature selenosulphate solution of SnCl2 complexed with tri-
ethanolamine and added NaOH. Polyvinylpyrollidone (PVP) was also added and
in general slowed down the deposition. At an optimum concentration of PVP, a
maximum terminal thickness was obtained (although no comparison with films
deposited from PVP-free solutions was given). No XRD pattern was observed for
the as-deposited films; heating in an inert atmophere at ca. 330°C gave the pattern
of SnSe. The bandgap was 0.95 eV (indirect). The films were n-type, with a re-
sistivity of ca. 10 -cm



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
REFERENCES
  1.   C Puscher. Dingl. J. 190:421, 1869.
  2.   E Beutel, A Kutzelnigg. Monats. 58:295, 1931.
  3.   KC Mandal, A Mondal. J. Phys. Chem. Solids 51:1339, 1990.
  4.   O Savadogo, KC Mandal. J. Electrochem. Soc. 139:L16, 1992.
  5.   O Savadogo, KC Mandal. Sol. Energy Mater. Sol. Cells 26:117, 1992.
  6.   JD Desai, CD Lokhande. Thin Solid Films 237:29, 1994.
  7.   RS Mane, BR Sankapal, CD Lokhande. Thin Solid Films 353:29, 1999.
  8.   LP Deshmukh, SG Holikatti, BP Rane, BM More, PP Hankare. J. Electrochem. Soc.
       141:1779, 1994.
 9.    CD Lokhande. Mater. Chem. Phys. 28:145, 1991.
10.    I Grozdanov, M Ristov, G Sinadinovski, M Mitreski. J. Noncryst. Solids 175:77,
       1994.
11.    I Grozdanov. Semicond. Sci. Tech. 9:1234, 1994.
12.    MTS Nair, Y Pena, J Campos, VM García, PK Nair. J. Electrochem. Soc. 145:2113,
       1998.
13.    P Pramanik, RN Bhattacharya. J. Solid State Chem. 44:425, 1982.
14.    RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:1642, 1982.
15.    RN Bhattacharya, P Pramanik. Sol. Energy Mater. 6:317, 1982.
16.    JD Desai, CD Lokhande. Thin Solid Films 249:135, 1994.
17.    JD Desai, CD Lokhande. Ind. J. Pure Appl. Phys. 31:152, 1993.
18.    CD Lokhande, AU Ubale, PS Patil. Thin Solid Films 302:1, 1997.
19.    JD Desai, CD Lokhande. Mater. Chem. Phys. 34:313, 1993.
20.    RS Mane, BR Sankapal, CD Lokhande. Thin Solid Films 359:136, 2000.
21.    S Biswas, A Mondal, D Mukherjee, P Pramanik. J. Electrochem. Soc. 133:48, 1986.
22.    MTS Nair, PK Nair. Semicond. Sci. Tech. 5:1225, 1990.
23.    PK Nair, MTS Nair, O Gomezdaza, RA Zingaro. J. Electrochem. Soc. 140:1085,
       1993.
24.    JD Desai, CD Lokhande. Mater. Chem. Phys. 41:98, 1995.
25.    P Pramanik, RN Bhattacharya. J. Electrochem. Soc. 127:2087, 1980.
26.    RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:332, 1982.
27.    LP Deshmukh, KV Zipre, AB Palwe, BP Rane, PP Hankare, AH Manikshete. Sol.
       Energy Mater. Sol. Cells 28:249, 1992.
28.    RK Nkum, AA Adimado, H Totoe. Mater. Sci. Eng. B. 55:102, 1998.
29.    D Mukherjee, AK Dutta. J. Mater. Sci. Lett. 8:511, 1989.
30.    PK Nair, J Campos, A Sanchez, L Banos, MTS Nair. Semicond. Sci. Technol. 6:393,
       1991.
31.    ME Rincon, PK Nair. J. Phys. Chem. Solids 57:1937, 1996.
32.    ME Rincon, R Suárez, PK Nair. J. Phys. Chem. Solids 57:1947, 1996.
33.    PK Nair, MTS Nair. Semicond. Sci. Technol. 7:239, 1992.
34.    P Pramanik, RN Bhattacharya, A Mondal. J. Electrochem. Soc. 127:1857, 1980.
35.    VM García, MTS Nair, PK Nair, RA Zingaro. Semicond. Sci. Technol. 12:645,
       1997.
36.    PK Basu, P Pramanik. J. Mater. Sci. Lett. 5:1216, 1986.
36a.   FC Eze, CE Okeke. Mater. Chem. Phys. 47:31, 1997.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
37. P Pramanik, S Bhattacharya, PK Basu. Thin Solid Films 149:L81, 1987.
38. C Lévy-Clément, M NeumannSpallart, SK Haram, KSV Santhanam. Thin Solid
    Films 302:12, 1997.
39. M Lakshmi, K Bindu, S Bini, KP Vijayakumar, CS Kartha, T Abe, Y Kashiwaba.
    Thin Solid Films 370:89, 2000.
40. T Yamamoto, K Tanaka, E Kubota, K Osakada. Chem. Mat. 5:1352, 1993.
41. JCW Folmer, F Jellinek. J. Less Common Metals 76:153, 1980.
42. I Grozdanov, CK Barlingay, SK Dev. Thin Solid Films 250:67, 1994.
43. I Grozdanov, M Najdoski. J. Solid State Chem. 114:469, 1995.
44. MTS Nair, PK Nair. Semicond. Sci. Technol. 4:599, 1989.
45. PK Nair, VM Garcia, AM Fernandez, HS Ruiz, MTS Nair. J. Phys. D:Appl. Phys.
    24:441, 1991.
46. I Grozdanov, CK Barlingay, SK Dev. Mater. Lett. 23:181, 1995.
47. KM Gadave, CD Lokhande. Thin Solid Films 229:1, 1993.
48. MTS Nair, L Guerrero, PK Nair. Semicond. Sci. Technol. 13:1164, 1998.
49. RN Bhattacharya, P Pramanik. Bull. Mater. Sci. 3:403, 1981.
50. E Fatas, T Garcia, C Montemayor, A Medina, E Garcia Camarero, F Arjona. Mater.
    Chem. Phys. 12:121, 1985.
51. AJ Varkey. Sol. Energy Mater. 19:415, 1989.
52. MTS Nair, PK Nair. Semicond. Sci. Technol. 4:191, 1989.
53. J Cardoso, O GomezDaza, L Ixtlilco, MTS Nair, PK Nair. Semicond. Sci. Technol.
    16:123, 2001.
54. A Mondal, P Pramanik. J. Solid State Chem. 47:81, 1983.
55. G Hodes, D Cahen. Ternary adamantine materials for low-cost solar cells (Project
    No. IL-2-04132-1), Quarterly Status Report to Solar Energy Research Institute,
    Golden, CO, 1985.
56. GK Padam. Thin Solid Films 150:L89, 1987.
57. I Grozdanov. Chem. Lett. 3:551, 1994.
58. I Grozdanov, CK Barlingay, SK Dey. Integrated Ferroelectrics 6:205, 1995.
59. VM Garcia, PK Nair, MTS Nair. J. Cryst. Growth 203:113, 1999.
60. B Pejova, I Grozdanov. J. Solid State Chem. 158:49, 2001.
61. CA Estrada, PK Nair, MTS Nair, RA Zingaro, EA Meyers. J. Electrochem. Soc.
    141:802, 1994.
62. A Mondal, P Pramanik. J. Solid State Chem. 55:116, 1984.
63. M Lakshmi, K Bindu, S Bini, KP Vijayakumar, CS Kartha, T Abe, Y Kashiwaba.
    Thin Solid Films 386:127, 2001.
64. GA Kitaev, VI Dvoinin, AV Ust’yantseva, MN Belyaeva, LG Skornyakov. Inorg.
    Mater. 12:1448, 1976.
65. KO Velthaus, J Kessler, M Ruckh, D Hariskos, D Schmid, HW Schock. In: 11th
    ECPV Solar Energy Conf., Montreux, Switzerland, 1992, p 842.
66. D Braunger, D Hariskos, T Walter, HW Schock. Sol. Energy Mater. Sol. Cells
    40:97, 1996.
67. D Hariskos, M Ruckh, U Ruhle, T Walter, HW Schock, J Hedstrom, L Stolt. Sol.
    Energy Mater. Sol. Cells 41–2:345, 1996.
68. R Bayón, C Guillén, MA Martínez, MT Gutiérrez, J Herrero. J. Electrochem. Soc.
    145:2775, 1998.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 69.   R Bayón, C Mafftiotte, J Herrero. Thin Solid Films 353:100, 1999.
 70.   R Bayón, J Herrero. Appl. Surf. Sci. 158:49, 2000.
 71.   R Bayón, J Herrero. Thin Solid Films 387:111, 2001.
 72.   T Yoshida, K Yamaguchi, H Toyoda, K Akao, T Sugiura, H Minoura. Proc. Elec-
       trochem. Soc. 97–20:37, 1997.
 73.   P Pramanik, MA Akhter, PK Basu. Thin Solid Films 158:271, 1988.
 74.   CD Lokhande, A Ennaoui, PS Patil, M Giersig, M Muller, K Diesner, H Tributsch.
       Thin Solid Films 330:70, 1998.
 75.   CD Lokhande, KM Gadave. Turkish J. Phys. 18:83, 1994.
 76.   P Pramanik, RN Bhattacharya. J. Mater. Sci. Lett. 8:781, 1989.
 77.   P Pramanik, S Bhattacharya. Mater. Res. Bull. 25:15, 1990.
 78.   KC Mandal, O Savadogo. J. Mater. Chem. 1:301, 1991.
 79.   P Pramanik, S Biswas. J. Solid State Chem. 65:145, 1986.
 80.   SS Dhumure, CD Lokhande. Thin Solid Films 240:1, 1994.
 81.   AJ Varkey. Sol. Energy Mater. 21:291, 1991.
 82.   SS Dhumure, CD Lokhande. Mater. Chem. Phys. 27:321, 1991.
 83.   SS Dhumure, CD Lokhande. Sol. Energy Mater. Sol. Cells 28:159, 1992.
 84.   GA Kitaev, TP Bol’shchikovsa. Izv. Akad. SSSR, Neorg. Mater. 2:65, 1966.
 85.   GA Kitaev, TP Bol’shchikovsa, TA Ust’yantseva. Izv. Akad. SSR, Neorg. Mater.
       3:1080, 1967.
 86.   TP Bol’shchikovsa, GA Kitaev, VI Dvoinin, MV Degtyarev, LM Dvoskina. Izv.
       Akad. SSSR, Neorg. Mater. 16:387, 1980.
 87.   H Meherzi-Maghraoui, P Cowache, D Lincot, M Dachraoui. J. Chim. Phys. 96:259,
       1999.
 88.   H Meherzi-Maghraoui, M Dachraoui, S Belgacem, KD Buhre, R Kunst, P Cowache,
       D Lincot. Thin Solid Films 288:217, 1996.
 89.   MJ Mangalam, KN. Rao, N Rangarajan, CV Suryanarayana. Brit. J. Appl. Phys.
       (J. Phys. D), Ser. 22:1643, 1969.
 90.   MJ Mangalam, KN. Rao, N Rangarajan, CV Suryanarayana. Ind. J. Pure and Appl.
       Phys. 7:628, 1969.
 91.   SS Dhumure, CD Lokhande. Mater. Chem. Phys. 28:141, 1991.
 92.   M Ristova, P Toshev. Thin Solid Films 216:274, 1992.
 93.   AA Velykanov, EK Ostrovskaya, NP Garina, VA Turacova, AA Tchurkan. Ukr.
       Chim. Zh. 49:764, 1983.
 94.   B Pejova, M Najdoski, I Grozdanov, SK Dey. Mater. Lett. 43:269, 2000.
 95.   AB Kulkarni, MD Uplane, CD Lokhande. Thin Solid Films 260:14, 1995.
 96.   A Mondal, P Pramanik. Thin Solid Films 110:65, 1983.
 97.   MJ Mangalam, KN Rao, N Rangarajan, CV Suryanarayana. Jpn. J. Appl. Phys.
       8:1258, 1969.
 98.   C Puscher. Dingl. J. 195:375, 1870.
 99.   CD Lokhande. J. Phys. D: Appl. Phys. 23:1703, 1990.
100.   RD Engelken, HE McCloud, C Lee, M Slayton, H Ghoreishi. J. Electrochem. Soc.
       134:2696, 1987.
101.   P Pramanik, PK Basu, S Biswas. Thin Solid Films 150:269, 1987.
102.   MTS Nair, PK Nair. Semicond. Sci. Tech. 6:132, 1991.
103.   AJ Varkey. Int. J. Mater. Prod. Technol. 12:490, 1997.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
104. PL Allen, A Hickling. Chem. Ind. 51:1558, 1954.
105. T Jiang, GA Ozin, RL Bedard. Adv. Mater. 6:860, 1994.
106. PK Nair, MTS Nair. J. Phys. D:Appl. Phys. 24:83, 1991.
107. MTS Nair, PK Nair. J. Phys. D:Appl. Phys. 24:450, 1991.
108. D Hariskos, R Heberholz, M Ruckh, U Ruhle, R Schäffler, HW Schock. In: 13th
     ECPV Solar Energy Conf., Nice, France, 1995, p 1995.
109. M Ristov, G Sinadinovski, I Grozdanov, M Mitreski. Thin Solid Films 173:53,
     1989.
110. P Pramanik, RN Bhattacharya. J. Mater. Sci. Lett. 7:1305, 1988.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
7
Oxides and Other
Semiconductors




Most of the compounds deposited by CD have been sulphides and selenides. Apart
from a very few examples of tellurides (and some related telluride experiments)
and with a very few exceptions, discussed at the end of this chapter, what is left is
confined to oxides (including hydrated oxides and hydroxides and two examples
of basic carbonates.) This chapter deals mainly with these oxides. In addition, as
noted in Chapter 3, there are a number of slow precipitations that result in precip-
itates, rather than films, of various other compounds, not necessarily semicon-
ductors in the conventional sense. These potential CD reactions, briefly discussed
in Chapter 3, will be somewhat expanded on in this chapter.
       Oxide films are often deposited because of their electrical (resistance) and op-
tical properties. A selection of such properties of CD oxides is given in Table 2.2.
       The reader is strongly urged to read Section 3.2.4 (precursors for oxide de-
position) before reading this chapter or at least to refer back to it when necessary.

7.1 GENERAL CONSIDERATIONS FOR
    OXIDE/HYDRATED OXIDE/HYDROXIDE
    DEPOSITION
The old analytical chemistry literature is rich with methods involving homoge-
neous precipitation from solution, the purpose being to obtain dense (therefore
easily filterable), contamination-free precipitates for purposes of analyses. The



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
urea method, in particular, has been extensively used in the past to form precipi-
tates of oxides and basic salts for analytical purposes. Urea hydrolyzes to ammo-
nia and (bi)carbonate, and the ammonia hydrolyzes further to give OH ions, with
a subsequent increase in pH. This leads to the formation of hydroxides, hydrated
oxides, carbonates, and basic salts.
       The formation of films has often been noted in precipitations using urea.
Thus, Gordon [1] noted: “The precipitation of basic salts with urea is character-
ized by the formation of thin transparent films of precipitate which strongly ad-
here to glass surfaces.” Also, in Ref. 2 Gordon et al. wrote (p. 39): “Basic thorium
formate adheres tenaciously to glass surfaces in the manner characteristic of the
basic salts precipitated by the urea method.” Basic sulphates of Al [3] and Ga [4],
which under suitable conditions contain very little sulphate (and are probably ox-
ides or hydroxides), have been observed to form on glass using urea precipitation.
       This film formation was an undesirable side effect—for accurate analysis,
the film needed to be removed and added to the precipitate. However, as pointed
out by Gordon [1], “To those who have worked with the urea method, the exis-
tence of these films will always be a reminder of Willard’s* fond hope that he will
someday find a way of making all the precipitate adhere to the beaker so that it
will only be necessary to dry and weigh the beaker after discarding the solution.”
       Film formation is, in retrospect, not surprising, since the slow reaction char-
acteristic of many homogeneous precipitation reactions is normally required
(among other factors) for appreciable film formation to occur. Furthermore, the
films tended to be very adherent. In their book Precipitation from Homogeneous
Solutions [2], Gordon et al. write on the films formed by precipitation of basic tin
sulphate: “The films cannot be removed by scraping with a policeman. However,
by adding a few milliliters of hydrochloric acid . . . the films are easily dissolved.”
(For those readers who, like the author, found this first sentence evoking amusing
mental images, a policeman is (or was) a glass rod with a piece of rubber attached
to the end, used to scrape deposits out of reaction vessels.) A notable exception to
this “easy” dissolution was deposition of “basic stannic sulfate” using urea. This
material formed such an adherent film that “it poses a difficult removal problem”
[2,4a]. The films could be peeled off in relatively large, transparent sheets by
warming for 30 min in a solution of (NH4)2SO4 and NaOH at pH of 9 0.5, a fact
that suggests using CD for the preparation of thin, self-supporting films. (This also
meant that the film particles adhered strongly not only to the substrate, but also to
each other.) Actually, though not analytically well defined, this “basic sulfate” ac-
tually contained very little sulphate and was probably mostly SnO2 (see the be-
ginning Sec. 7.2.14 on SnO2). It should be remembered that most of these prod-
ucts form white precipitates; therefore, if nonscattering, the films may not even be


* H. H. Willard was one of the pioneers of homogeneous precipitation for chemical analysis.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
visible. Thus film formation may have occurred in cases where it was not reported
or even recognized.
       As a general point, it might be expected that the product of reaction between
metal ions and hydroxide is a hydroxide (or basic salt) rather than an oxide. In
many reported cases, oxides are formed directly. This is probably due to two fac-
tors. Many of the metal ions used (e.g., Pb, Sn, Tl, Ti, Zr, Si) do not readily, if
at all, form simple hydroxides; most of these cations have a greater tendency to
form what may be called oxide polymers, involving condensation to chains of
OMMMOMMMO species at the reaction pH. Some hydroxides (e.g., Ag, Cu,
Mn) are not very stable and are quite readily converted to the oxide, even in aque-
ous solutions. In some cases, simple hydroxides do form and need to be heated to
dehydrate to the oxide.
       Very acidic (high valent) cations will readily hydrolyse in aqueous solution,
often even at low pH. These cations tend to form the polymeric metal oxide chains
mentioned previously. This hydrolysis can be controlled by addition of boric acid
(see Sec. 3.2.4.4) and forms the basis of a technique referred to as liquid phase de-
position. This method can be reasonably included in the more general term of
chemical solution deposition, and is treated, although not comprehensively, in
this book. Ref. 5 deals more thoroughly with this technique and describes many
cases of SiO2 as well as some examples of several other oxides not covered in this
chapter.

7.2 SPECIFIC OXIDES AND HYDROXIDES
7.2.1 Antimony Oxide (Sb2O3)
Films of Sb2O3 (more strictly, Sb4O6) have been deposited from a room-tempera-
ture solution of potassium antimonyl tartrate and sodium selenosulphate [6]. The
films showed a clear XRD pattern, and compositional analysis confirmed the
composition. It is interesting that the selenide was not formed from this solution
(it was formed if the Sb solution was mixed with triethanolamine and ammonia
before adding the selenosulphate; see antimony selenide in Chap. 6). The most
likely explanation for this is that the more alkaline solution, containing tri-
ethanolamine and ammonia, keeps the oxide, which might tend to form, in solu-
tion, both because Sb4O6 is soluble in sufficiently strong alkali and because of
complexation by the triethanolamine and ammonia. Hydroxide would be present
in much larger concentrations than selenide, even under mildly alkaline condi-
tions. The resistivity of the films was of the order of 109 -cm.

7.2.2 Cadmium (Hydr)oxide (Cd(OH)2, CdO)
Films of what was presumably Cd(OH)2 were deposited by heating to 80°C an al-
kaline cyanide solution of Cd2 containing H2O2 [7]. After heating the white as-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
deposited films at 250–300°C, they turned brown to give CdO with resistivities of
typically a few k /sq. (CdO is normally a degenerate semiconductor with a low
resistivity.)
       H2O2 was used in two other studies to deposit cadmium hydroxides or hy-
drated oxides. A mixed CdMOMOH film was deposited from a Cd-ammine so-
lution at pH       10 and at various temperatures onto glass and quartz [8]. The
deposited films exhibited clear XRD peaks that were identified with
Cd(O2)0.88(OH)0.24 and were (111) textured. However, if KBr was added to the de-
position solution, the preferred texture became (200) and the rate of film growth,
as well as the terminal thickness, decreased. Annealing the films at ca. 200°C con-
verted them into CdO with the same texture present as in the as-deposited films.
From the optical spectra, a bandgap of ca. 2.6 eV was obtained (slightly higher for
the Br-free bath and slightly lower for the Br-containing bath). The resistivity of
the films varied from 2 to 20 k /sq ( 2 10 2 -cm). In this case, thicker films
had a higher resistivity, and this was ascribed to cracking of the thicker films, ob-
served in optical microscopy studies. The second study used a similar solution and
obtained films, identified by XRD as CdO2, with a high resistivity (106–107 -
cm) that converted to low-resistivity (10 3 -cm) CdO on annealing in air at
450°C with a direct bandgap of 2.3 eV [9].
       Deposition without H2O2 has also been described, using ammonia-com-
plexed Cd. In one study [10,11], deposition was carried out at either room temper-
ature or 50°C. The as-deposited Cd(OH)2 films on glass were annealed at 400°C in
either air or an inert atmosphere to convert them to CdO. The as-deposited film was
X-ray amorphous, while the annealed film was polycrystalline CdO. A bandgap of
2.2 eV (a little lower than the standard value of 2.4 eV) was obtained from the op-
tical spectra of the CdO films, and the transmission in the nonabsorbing region was
high (up to 90%). The resistivity of the as-deposited Cd(OH)2, as expected, was
high (107 -cm), while that of the CdO was 10 3 -cm. In the other study, a higher
ammonia concentration was used and deposition was carried out at room temper-
ature [12]. It was noted that the uniformity of the films was better than when de-
posited at higher temperature. The Cd(OH)2 was heated in air at 150°C to convert
it to CdO (the previous films required heating to ca. 275°C to convert the Cd(OH)2
to CdO [11]). (This is a not insignificant difference for what is assumed to be iden-
tical material.) The bandgap was 2.3 eV, and, as in the previous study, the films
were very transparent to photons with less energy than the bandgap. The resistiv-
ity of the films was between 2 10 2 and 5 10 2 -cm.

7.2.3 Cobalt (Hydr)oxide and Hydroxy-Oxide
      (Co(O)OH, Co(OH)2, CoO, Co3O4)
Co(O)OH was deposited from an ammonia-complexed solution of CoCl2 [13].
The Co(II) ammonia complex was allowed to oxidize for two days in air to the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
more stable trivalent cobaltic (III)–ammine complex. Heating the cobaltic–am-
mine complex to 65–70°C resulted in deposition of ca. 0.1 m of an adherent,
brown CoO(OH) deposit on a glass substrate after 4 hr. Spectral measurements al-
lowed an estimation for the bandgap of this film of ca. 2.4 eV. The Co(O)OH was
oxidized to an adherent Co3O4 film by heating over 300°C in air.
       Films of CoO were deposited from a somewhat similar bath (but not left to
oxidize) after annealing in O2 at 350°C; Co3O4 started to appear only at 500°C
[13a].
       Hydrolysis of urea to increase the pH by formation of ammonia was used to
deposit Co(OH)2 by heating a solution of Co2 with urea at 100°C [14]. The pink
film was shown by XRD to be a mixture of phases of Co(OH)2. Heating the films
at 350–400°C converted the hydroxide to Co3O4.

7.2.4 Copper Oxide (Cu2O)
Cu2O films were deposited by treating a thiosulphate-complexed solution of
Cu(NO3)2 with NaOH [15]. This was based on an early study where a glass sub-
strate was alternately and repeatedly dipped in NaOH and then Cu-thiosulphate
solutions [16]. The thiosulphate (S2O2 ) both reduced the Cu(II) to Cu(I) and
                                           3
acted as a complexing agent. The films were deposited at 60–70°C, resulting in a
thickness of ca. 0.3 m in one hour. The substrates—glass slides or polyester
film—were precoated with very thin CuxS films by immersion in copper thiosul-
phate solution at 40°C. While the role of this prelayer was not clear, it was implied
that in its absence, the Cu2O films were not uniform.
       While the mechanism for the deposition was not discussed, the instability of
the copper hydroxides (the hydroxide of Cu(I) probably does not even exist) to-
ward dehydration, together with the reducing action of the thiosulphate, leads to
the expectation that Cu2O will be the product of the hydrolysis of Cu(I) in alka-
line solution. It should be noted, however, that the Cu-thiosulphate solution itself
is not very stable and apparently forms predominantly CuxS in the absence of
NaOH.
       X-ray diffraction showed the film to be Cu2O, with no detectable amount of
CuO and with a crystal size, estimated from the peak widths, of 20 nm. Optical
transmission measurements of the films gave a value of (indirect) bandgap of 2.28
eV (literature room-temperature bandgap 2.1 eV but is rather variable).
       The electrical resistivity of the films (between 0.1 and 0.25 m thick), mea-
sured through Au contacts, was ca. 2.5 k /sq (ca. 5 10 2 -cm). This value in-
creased with air-annealing (250°C, 20 min) up to 60 k /sq. The relatively low re-
sistivity was attributed to incorporation of S, either from the CuxS prelayer or by
hydrolytic decomposition of S2O2 to S2 . Treatment of the films with Na2S so-
                                     3
lution decreased the resistivity by nearly two orders of magnitude, and S was
found in the films. It is likely that the surface of the Cu2O crystals was partially



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
converted to CuxS, and surface conduction via this surface layer was responsible
for the enhanced conductivity.

7.2.5 Indium Oxide (In2O3)
In2O3 was deposited from a solution of InCl3 that was slowly hydrolyzed to form
the hydroxide [17]. The rate of hydrolysis was slowed down sufficiently to pre-
vent rapid bulk precipitation by a combination of a freezing agent (sodium citrate)
and a relatively low pH (7.5). Ag (as AgNO3) was added, supposedly as a cata-
lyst (although it is not clear what needed to be catalyzed) and to improve adher-
ence. The citrate will also act as a complexant for the In3 , which may be an im-
portant factor. Addition of SnCl4 to the deposition solution allowed doping with
Sn [tin-doped indium oxide or indium tin oxide (ITO)]. The film grew to a termi-
nal thickness of ca. 360 nm in 30 min. At higher solution pH, the terminal thick-
ness decreased. Heating at 200°C for 2 hr in a vacuum resulted in conversion of
the hydroxide(s) to crystalline oxide, with an average grain size of ca. 25 nm
(In2O3) and ca. 54 nm (ITO).
       Figure 7.1 shows the optical transmission and reflectance spectra of the two
films. The main difference is an increase in the mid-IR reflectance of the doped
film compared with the undoped one, due to the high free electron concentration




FIG. 7.1 Transmittance and reflectance spectra of In2O3 films. Broken lines: undoped
In2O3; solid lines: ITO (10% Sn). (Adapted from Ref. 17.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
in the conduction band. The bandgaps measured from the spectra were ca. 3.5 eV
(undoped) and ca. 3.6 eV (ITO).
       Electrical measurements of the In2O3 (ITO) films gave temperature-inde-
pendent resistivities ( ) of 2 10 2 (10 3) -cm, carrier concentrations (Nd) of
2 1020 (1021) cm 3, and mobilities ( ) of 3 (17) cm2V 1sec 1. The tempera-
ture independence of the resistivity indicated that the films, even the nominally
undoped ones, were degenerate semiconductors.
       In2O3 has been deposited on Sn/Pd-activated glass by first depositing a film
of In(OH)3 and then heating in air at a temperature of 200°C or more [18]. The
In(OH)3 was deposited using a solution of dimethylamineborane and indium ni-
trate maintained at 60°C. The deposition rate was dependent on the borane con-
centration up to a limiting concentration of 0.03 M, and the film thickness was
proportional to deposition time, with final thicknesses of ca. 1 m. X-ray diffrac-
tion showed mainly one sharp peak corresponding to the (210) plane of In(OH)3,
which, after annealing, converted to polycrystalline In2O3.
       Optical transmission spectroscopy of the In2O3 film showed a high trans-
mission at 800 nm, gradually decreasing with decreasing wavelength, character-
istic of a somewhat scattering film. The bandgap was estimated from the spectrum
to be 3.6 eV.
       Electrical conductivity measurements of the as-deposited In(OH)3 showed
an expectedly high resistivity of ca. 109 -cm. That of the annealed oxide film de-
creased to 33 -cm (carrier concentration 1.85 1016 cm 3; mobility 10
cm2V 1sec 1). The resistivity is high compared to many other In2O3 films (which
are often used as transparent conductors), mainly due to the low carrier concen-
tration, implying a high degree of stoichiometry.
       Finally, although no attempt was made to convert the film to oxide,
In(OH)3, for use as a buffer layer on PV cells (see Chap. 9), was deposited from a
thiourea-based solution of InCl3 at a pH of 3.3 [19]. Apparently no sulphide was
formed, possibly due to the relatively high (for In) pH, which favored hydroxide
formation.

7.2.6 Iron Oxides and Hydroxy-Oxide
The instability of Fe(III) compounds toward hydrolysis has been exploited to form
Fe(O)OH films [20]. The substrate in this study was a sulphonated-vinyl termi-
nated self-assembled monolayer (SAM). Deposition was accomplished by heating
Fe(NO3)3 solutions. The pH of the solution was rather critical; a pH of 2.0 or
slightly higher was necessary. At lower values of pH, hydrolysis did not occur; at
appreciably higher values, rapid hydrolysis occurred, resulting in precipitation
rather than film deposition [e.g., at a pH of ca. 3, only very thin films (ca. 5 nm
thick) of colloidal Fe(O)OH particles formed]. The films were columnar, with a
column diameter of ca. 20 nm, and the columns were composed of lamellae ca. 2



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
nm in size. Nucleation of the (noncolloidal) film occurred by binding of Fe species
to the sulphonate endgroups (one Fe to two sulphonate groups, measured by XPS)
[21,22]
          -Fe2O3 was deposited on Si (111) or Si (100) by using hydrolysis of urea
at high temperatures [23]. An aqueous solution of Fe(NO3)3 and urea (pH between
5 and 6), together with the Si substrate was heated to between 100 and 200°C (pre-
sumably in an autoclave) for 4–24 hr. X-ray diffraction showed the formation of
  -Fe2O3 with some (101) texture. From SEM measurements, the films were ca.
100 nm in thickness, with a morphology depending on the crystal face of the Si.
For (111) Si, the grains were spherical and 10 nm in size; for (100) Si, columnar
grains, 30 5 nm, were obtained.
        If reduced Fe powder was added to the preceding solution at an optimum pH
of 6–7, magnetite (Fe3O4) was deposited onto Si (100) or -Al2O3 at ca. 140°C
over several hours [24]. No other phase was found in the XRD spectrum. It was
suggested that the Fe3O4 formed by reaction between Fe(OH)3 (presumably
formed by hydrolysis of the ferric nitrate) and Fe(OH)2 formed by hydrothermal
oxidation of the Fe powder. Particle sizes of 150 nm (on Al2O3) and 50 nm (on Si)
were measured by SEM.
        By adding Co(Cl)2 to the deposition solution and heating the resulting films
in air (2 hr at 400°C) conversion of the magnetite to -Fe2O3, doped with Co, oc-
curred [25]. The grains were needle-shaped (50 10 nm). The films exhibited
good magnetic properties.
        A variation of the foregoing urea method was used to deposit -Fe2O3 on
SnCl2-sensitized glass substrates [26]. A solution of FeCl2 and urea at pH 3 was
heated at 90°C for 2 hr. The as-deposited film was probably FeO(OH) (the hy-
droxide group was seen in FTIR studies). On annealing at 350–400°C (presum-
ably in air), -Fe2O3 was formed, with a crystal size of 22 nm. Optical spec-
troscopy of the as-deposited film showed a direct bandgap of 3.2 eV and a weak
(possibly indirect) absorption starting at ca. 2.2 eV. The bandgap of the oxidized
films was 2.0 eV. Resistivity was ca. 2 -cm, which dropped by a factor of up to
four when exposed to high humidity, suggesting possible use as a sensor for wa-
ter vapor.
        Magnetite was also deposited on glass by the dimethylamine borane tech-
nique described for In2O3, using a solution of Fe(NO3)3 and dimethylamine bo-
rane with a pH 3.5 at 20°C [27]. At higher deposition temperatures, Fe(O)OH
and Fe2O3 were apparently also formed, and the magnetite content decreased, un-
til, at 60°C, no magnetite was observed and the films were yellow (the magnetite
films were black). The formation of pure Fe(III) oxide (hydroxide) at higher tem-
peratures, compared to the mixed Fe(II)/Fe(III) magnetite at lower temperatures,
is likely due to more facile oxidation of Fe(II) to Fe(III) at higher temperature. The
borane is a reducing agent, and this is presumably the reason that the mixed va-




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
lency magnetite, which can be envisaged as Fe(III)2O3 Fe(II)O, is formed at lower
temperature. The resistivity of the magnetite was ca. 2 k -cm—much higher than
usual for magnetite. This was explained by a lower concentration of Fe(III) than
expected from stoichiometry, which was obtained in these films. Magnetic prop-
erties of the films were described.

7.2.7 Lead Oxide (PbO2)
A solution of Pb2 ions can be oxidized to PbO2 by persulphate [28]:
      Pb2      S2O2
                  8       2H2O → PbO2        2SO 2
                                                 4      4H                       (7.1)
This reaction normally resulted in a precipitate of PbO2. However, this process
was subsequently modified to give films of PbO2 without precipitation in solu-
tion, apparently based on the observation of the authors that films of PbO2 were
sometimes observed to form on the walls of the glass beaker using this reaction
[29].
       The film deposition was carried out at room temperature from an aqueous
solution of plumbous acetate, ammonium acetate, and ammonium persulphate, us-
ing NH4OH to bring the pH to 6. A trace of AgNO3 was added as a catalyst for re-
action 7.4 [29]. A film of PbO2 ca. 50 nm thick was formed in an hour. Once this
initial film was deposited, thicker films could be built up, usually at a somewhat
higher pH, in the absence of the AgNO3. The initial film formation appears to be
a pure CD reaction. However, electrochemical studies of further film buildup
showed that an electroless deposition mechanism, involving two partial electro-
chemical reactions, was responsible for film formation.
       A few words on the difference between CD and electroless deposition are in
order here. Electroless deposition is related to electrodeposition, except that in-
stead of charge being supplied by an external power supply, it is supplied inter-
nally by oxidation (reduction) of a strong reducing (oxidizing) agent; the two par-
tial electrochemical reactions occur at different sites on the substrate (the substrate
is initially a sensitized solid and subsequently the deposited material itself). This
implies a reasonable electrical conductivity of the material to be deposited. For
this reason, electroless deposition is used mainly for metals, but can be used for
electrically conducting compounds, of which PbO2 is an example. Since charge
transfer is involved, a change in valence state of the metal cation normally occurs
between solution and film.
       Mindt [29] described some properties of these films (thicker electroless
films, not the initial purely CD ones). Electron diffraction showed that the film
was -PbO2. The crystal (more correctly the particle) size was found, by electron
microscopy, to be ca. 200 nm. The carrier density, measured by the Hall effect,
was ca. 1021 cm 3. The resistivity was somewhat dependent on the pH of deposi-




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
tion, varying from 2 10 3 (pH 7) to 3 10 2 -cm (pH 10), although the car-
rier density was not found to vary appreciably with pH (implying that the mobil-
ity did vary with pH).
       In a similar study, the deposition conditions were modified [pH of 8 (by am-
monia), no Ag catalyst, and a deposition temperature of 80°C] [30]. A primary
thin film was deposited, followed by a second deposition, resulting in films sev-
eral microns thick. Optical absorption spectroscopy gave a bandgap of 1.7 eV. The
film resistivity was 1.3 10 3 -cm (carrier density 8 1019 cm 3; mobility
   50 cm2V 1sec 1).
       White films of 6PbCO3 3Pb(OH)2 PbO (from XRD analysis) were slowly
formed over a few days from alkaline-complexed Pb2 solutions that contained a
colloidal hydrated oxide phase and that were exposed to air [31]. This was due to
reaction with CO2 in the air (see Sec. 5.3.3).

7.2.8 Manganese Oxide (Mn2O3, MnO2)
Aqueous solutions of permanganate will slowly oxidize, forming a brown film
[not clear if this is Mn2O3, MnO2, or Mn(O)OH] on the walls of the vessel in
which it is stored. Increase of either acidity or alkalinity of the solution can accel-
erate this decomposition reaction. This reaction has been used to treat polymer
substrates prior to chemical deposition, for improved adhesion [32]. In that work,
the MnMO film was dissolved before the CD process, and the improved adhesion
was probably due to some increase in the hydrophilic character of the mostly hy-
drophobic polymers.
       Mn2O3 films have been deposited on glass from an ammoniacal solution of
Mn2 [33]. Ammonium chloride was added to decrease the pH and slow down the
rate of hydrolysis. The initial product was believed to be Mn(OH)2, which oxi-
dized in air to Mn2O3.
       The persulphate technique used for PbO2 described earlier was extended to
MnO2 deposited on glass, using manganous acetate in place of lead acetate, to-
gether with ammonium persulphate and AgNO3 as a catalyst [34]. Adherent films
up to 0.5 m could be obtained. No XRD pattern was found for the films, imply-
ing that the deposit was amorphous or made up of very tiny nanocrystals (nonde-
fected crystal size of 2 nm or more can usually be detected by careful XRD). The
resistivity of the films was very dependent on the solution pH, with values of 4
102 -cm (pH 8) and 2 104 -cm (pH 6.3).

7.2.9 Molybdenum Oxide
An early attempt to deposit Mo—S on various metal substrates using ammonium
molybdate and thiosulphate resulted in films that were found to be sulphur free
and believed to be an oxide, although this was not investigated further [35].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
7.2.10 Nickel Oxide (NiO)
A brief communication on the deposition of NiO using persulphate was described
[36]. An ammoniacal solution containing NiSO4 and potassium persulphate was
used to deposit black NiO on glass at room temperature. It was suggested, based
on the importance of the NH4OH (NaOH or KOH did not give NiO; other alkaline
reducing species, such as amines, did) that a mixture of NiO and Ni2O3 formed
and that the higher-valent Ni2O3 was reduced by the ammonia
       The NiO was confirmed by XRD. Optical absorption spectroscopy was used
to estimate a direct bandgap of 1.75 eV (see later). The films were p-type (ther-
moelectric probe), with a resistivity of 105 -cm.
       In another study, an ammoniacal solution of Ni2 was heated at 60–80°C to
deposit a green-gray film of what was reported to be NiO [37]. This was assumed
to form via the hydroxide, although no structural or compositional characteriza-
tion of this deposit (or of the annealed film) was given. The deposition was car-
ried out in a beaker, and no deposition occurred at room temperature; this suggests
that deposition occurred by loss of ammonia. Heating this film in air at 280°C
formed Ni(O)OH. From the optical transmission and reflection spectra of the NiO
and Ni(O)OH, it appears that their absorption spectra were very similar, with a
weak absorption (possibly also scattering) in the visible and strong absorption in
the near-UV region. The resistivity of the Ni(O)OH was 800 -cm, while that of
the as-deposited NiO was apparently too high to measure.
       Ni(OH)2 was deposited from a solution of urea and Ni2 ions at an initial
pH of 6 and a temperature of 100°C. The hydroxide was then annealed in air at
350–400°C to convert it to NiO [38]. Films almost 1 m thick were obtained af-
ter ca. 2 hr, with an average crystal size (from XRD) of 13 nm. Optical absorption
spectroscopy of the annealed films gave a direct bandgap of 3.6 eV, somewhat
lower than the rather variable literature values of 3.7–4.0 eV. It should be men-
tioned that it is not as simple to correlate the band structure of NiO with its opti-
cal and electrical properties as it is for most of the other semiconductors dealt with
here. This feature, common to many transition metal compounds, is a conse-
quence of electrons in (often-narrow) d-bands that are relatively localized by elec-
tron–electron repulsions. Thus, although the 3d band in NiO (which in the pure
state is green, like the hydrated Ni2 ion) is only partially filled, pure NiO is in-
sulating due to the localized 3d electrons. The conductivity (and black color) of
NiO as it is normally obtained is due to nonstoichiometry leading to doping. The
room-temperature resistivity of the annealed NiO in this case was several M /sq.
(several hundred -cm).

7.2.11 Silicon Oxide (SiO2)
SiO2 films were deposited on soda lime glass from a silica gel–saturated solution
of hydrofluorosilicic acid (H2SiF6) and boric acid [39]. The boric acid reacts with



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the H2SiF6 to form SiO2 by removal of HF formed in the H2SiF6 hydrolysis equi-
librium:
      H2SiF6 2H2O D 6HF           SiO2                                        (7.2)
       H3BO3 4HF D BF4            H3O       2H2O                              (7.3)
The SiO2 formed gradually deposited on the substrate. The deposition rate was
typically 10–20 nm/ hr, depending on the boric acid concentration and solution
temperature. Film thickness was ca. 100 nm.
      Infrared measurements showed that the films had a higher concentration of
SiMOMSi bonds than some other silica films made by different techniques. This
was interpreted to mean that the silica network of the films was more orderly, a
property that was evidenced by greater stability of the films against chemical etch-
ning and good blocking properties to sodium diffusion from the soda glass, com-
pared to many other silica films.
      This is the first of many SiO2 depositions using essentially the same tech-
nique. The others are given in Ref. 5.

7.2.12 Silver Oxide (AgO and Ag2O)
Oxides of Ag have been deposited by deposition from a triethanolamine-com-
plexed Ag solution at a pH        11.5 [40]. At room temperature, a black deposit
formed over some hours that converted into a brown film when air-annealed at
150°C. Based presumably on the color of these films, the as-deposited film was
assumed to be AgO, which turned into Ag2O on annealing (no structural charac-
terization of the films was reported). From the optical spectra, a bandgap of 2.25
eV was estimated for the annealed films ([literature (direct) bandgap of Ag2O
1.2 eV; a pure semiconductor with a direct bandgap of 2.25 eV should be orange].
Both films were insulating (no values given) but became much more conducting
when high voltages ( 1 kV) were applied to two laterally spaced Ag electrodes
on the (ca. 500-nm-thick) films.
       Using a similar solution, films of either Ag or AgO were deposited on both
glass and polyester film [41]. Addition of triethanolamine to a Ag solution
caused initial precipitation (silver oxide or hydroxide), which redissolved in ex-
cess triethanolamine. Deposition from a solution where some precipitate remained
resulted in AgO (possibly with some Ag2O), while a solution where this precipi-
tate was completely redissolved gave metallic Ag. The reducing action of the free
triethanolamine present in the latter case may be the cause of the formation of
metallic Ag.

7.2.13 Thallium Oxide (Tl2O3)
The same persulphate technique described earlier for PbO2 and MnO2 was also
used to deposit Tl2O3, with thallous acetate in place of the other metal acetates



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
[34]. Adherent films could be deposited up to a thickness of 10 m—very thick
films for the CD method. X-ray diffraction confirmed that the films were cubic
Tl2O3. The film resistivity was 5 10 4 -cm.
      The modifications employed by Bhattacharya and Pramanik for the PbO2
deposition (higher pH and temperature, no Ag ) were also used for Tl2O3 [30].
Film resistivity was similar to that of the previous study (3.7   10 4 -cm),
                                            20     3
with a carrier concentration of 4        10 cm       and a mobility of ca. 50
cm2V 1sec 1.

7.2.14 Tin Oxide (SnO2)
Very strongly adhering films of “basic stannic sulfate,” which was probably SnO2
(the Sn:SO 2 ratio was 44:1) were reported using an aqueous solution of SnCl4,
            4
urea, H2SO4, (NH4)2SO4, and HCl at pH 0.5 [4a]. The films adhered to the walls
of the glass deposition vessel so strongly that HF was one of the few reagents
available to remove them. Films which are mainly (hydr)oxides of Al, Ga and Th
have also been reported in the early literature using the urea method [2]. Tin salts,
in particular those of Sn(IV), are readily hydrolyzed, and the stable product is
SnO2. This has been exploited in a number of studies, most with minor differences
between them, to deposit SnO2 films.
      SnO2 films were deposited using SnCl4 and NH4F (the latter apparently as
a complexing agent to slow down hydrolysis of the Sn4 by the alkaline solution)
[33]. The resistivity of the as-deposited film was 200 -cm (10 2 -cm for In
doped).
      A similar method was described a few years later in which the NH4F was
used as a freezing agent to slow down the rate of hydrolysis [42]. AgNO3 was
added as a catalyst(?) and to improve the film adherence, although it is not clear
why a catalyst was needed or even desirable, since the objective was to slow down
the reaction. NaOH was used instead of ammonia to adjust solution pH to between
7.5 and 8.5.
      The growth rate was linear and decreased with decreasing pH, with a limit-
ing thickness that increased with decreasing pH. The growth rates and terminal
thicknesses were similar to those for ZnO deposited by the same technique (see
Fig. 7.2 in Sec. 7.2.17), only the rates were two to four times slower. These rela-
tionships were explained in the same way as for ZnO.
      The films were found to be SnO2 with the rutile structure (by XRD), with a
grain size of 20–30 nm (by TEM). Optical transmission and reflectance spec-
troscopy showed that the films were close to 80% transmitting up to ca. 1 m and
highly reflecting in the mid-IR. These spectra were similar to those of ZnO shown
in Figure 7.3, (Sec. 7.2.17), except for the absorption in the UV of the undoped
ZnO due to the lower bandgap. The bandgap (direct) was 3.56 eV. Electrical mea-
surements gave a resistivity ( ) of 0.1 -cm, carrier concentration (Nd) of 1019



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
cm 3, and mobility ( ) of ca. 6 cm2V 1sec 1. Annealing in vacuum decreased
to a minimum of 2 10 3 -cm at 375°C.
       Antimony-doped SnO2 films were deposited by adding SbCl3 to the depo-
sition solution. Sb is a well-known n-type dopant used to increase the conductiv-
ity of SnO2 films. The Sb concentration in the films increased linearly with that in
the deposition solution and was somewhat less than the solution concentration
(e.g., 6% Sb in solution gave ca. 4% in the film). The Sb doping increased both
the visible/near-IR transmission and mid-IR reflectance of the films, compared to
the undoped films. These spectra are similar to those for doped ZnO (Fig. 7.3), and
the effect of doping can be explained in the same way. The bandgap increased to
4.1 eV, compared to 3.56 eV for the undoped film, explained through band filling
by free electrons.
       The conductivity, Nd, and all increased with increasing Sb concentration
up to 5% and then decreased again with increasing Sb (Nd leveled off). This was
due to the obvious increase in Nd with increased doping, the measured increase in
grain size with doping (30–65 nm), resulting in increased followed by segrega-
tion of dopant at the grain boundaries at greater Sb concentrations, which again
decreased . The resistivity of the 5% Sb films decreased from 10 3 -cm as de-
posited to almost as low as 10 4 -cm on annealing at 375°C.
       In another preparation [43], ammonia gas was passed through a solution
of SnCl4, the precipitate rinsed well (to remove Cl, which caused the final films
to be porous—a useful observation since porous films are sometimes preferred
over compact ones), and redispersed in concentrated HNO3 to give a semitrans-
parent sol at a pH of 5–7. Films were deposited from this solution onto Si (100)
by heating at 60–100°C for 4 hr, followed by 100–200°C for 6–12 hr. The ini-
tial lower-temperature step was necessary to obtain nucleation; if omitted, no
film was deposited. Films ca. 200 nm thick were obtained with a crystal size of
3.5–4 nm.
       As with the original deposition of “basic stannic sulphate,” urea was used in
another deposition to slowly increase the pH of the initially strongly acidic (by
HCl) solution, thereby hydrolyzing the SnCl4 [44]. Besides slowing down the hy-
drolysis of the SnCl2 due to increased acidity, the HCl also complexed the tin as
hexachlorostannate, (SnCl6)2 , further slowing down the hydrolysis. From the op-
tical spectrum, a bandgap of 4.0 0.1 eV was estimated. These films were used
as buffer layers on CuInSe2 solar cells (see Chap. 9).
       SnO2 was deposited on hydrolyzed Si and on Si coated with sulphonate-
terminated self-assembled monolayers from a solution of SnCl4 in dilute HCl at
80° [45]. The films, up to 65 nm thick and consisting of a dense-packed aggregate
of SnO2 nanocrystals (5–10 nm) together with some amorphous basic tin oxide,
contained ca. 3 at.% Cl. They were adherent on all substrates, although the adher-
ence and homogeneity on Si was less reproducible than on the monolayer-coated
Si. Films were also deposited using a continuous flow system. The films were sim-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ilar to those using a static solution, but the deposition rate was considerably faster
( 20 nm hr 1 compared to an average of several nm hr 1 for the static solution),
and thicker films (160 nm) could be obtained.
       Films of SnO2 up to 0.5 m thick were grown by hydrolysis of aqueous
SnF2 (a change from the usual SnCl4) solutions, optimally at ca. 60°C on various
glasses and on Si [46]. The as-deposited films contained 6–16 mole% F, an n-type
dopant in SnO2. Films annealed at 300°C in air contained almost no F and were
crystalline SnO2 (cassiterite). It was suggested that the as-deposited material pos-
sessed a tin–oxygen polymer structure, with tin–fluorine bonds substituting for
some of the tin–oxygen bonds. Conductivities of the order of 10 2 -cm were
measured for films annealed at 500°C.
       Deposition of SnO2 onto sulphonated polystyrene at 40°C has been briefly
described in a review by Bunker et al., although without experimental details [47].
The point was made that the deposition of SnO2 involves hydrolysis and conden-
sation reactions involving poorly characterized species as opposed to precipitation
of an ionic salt. Also, the deposition is very dependent on pH—a change in pH of
one unit can change the solubility of the products by four orders of magnitude. In
spite of these factors, which imply difficulty in depositing films in the absence of
bulk precipitation in solution, they apparently succeeded in achieving preferential
nucleation of SnO2 on the sulphonated polystyrene. The films were dense and
composed of cassiterite, with a grain size of 4 nm.

7.2.15 Titanium Oxide (TiO2)
Sulphonate-terminated self-assembling monolayers on Si were used as substrates
for TiO2 deposition [48]. The deposition solution consisted of TiCl4 in 6M HCl at
80°C (considerably more dilute HCl solutions resulted in immediate bulk precip-
itation, while much stronger solutions were stable against hydrolysis and therefore
no deposition occurred.)
       The films, ca. 50 nm thick, comprised small (2–4 nm) nanocrystals of
anatase TiO2, possibly in an amorphous matrix, and were uniform, adherent, and
pore free. In contrast, only a small amount of irregular deposit was formed on bare
Si. The role of the sulphonate endgroups was believed to promote nucleation of
the nanocrystals and/or facilitate attachment of TiO2 clusters in solution to the
substrate. Hydrolysis of TiCl4 proceeds through various titanium hydroxy and
chloro-hydroxy complex cations. The anionic sulphonate groups could thus pro-
mote attachment and nucleation of these cationic complexes.
       Annealing in air increased the crystal size (up to 25 nm at 600°C) without
damage to the film while retaining the anatase structure (the rutile structure was
barely noticeable after 2 hr at 600°C).
       Using this deposition technique, TiO2 was deposited onto patterned self-
assembled monolayers [49]. Thioacetate-terminated trichlorosilane monolayers



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
were self-assembled onto oxidized Si substrates. Illumination through a grid-pat-
terned mask with grid openings of ca. 10 m resulted in photolysis of the ex-
posed, somewhat hydrophobic thioacetate groups to hydrophilic sulphonate
groups. TiO2 was then deposited from the TiCl4 solution onto the irradiated hy-
drophilic regions of the substrate (see Figs. 2.5 and 2.6). The resistivity of this
TiO 2 was ca. 109 -cm.
      TiO2 was deposited on (100) Si from a solution of TiO 2 prepared by dis-
                                                              4
solving Ti metal in an aqueous solution of ammonia and H2O2 [50]. An initial low-
temperature–final high-temperature regime similar to that described earlier for
SnO2 by the same group was used at a pH between 6 and 7. As before, the low-
temperature stage was necessary for film formation. A film formed in the low-
temperature stage but was X-ray amorphous. After the high-temperature stage,
pure anatase-phase TiO2 was obtained in the form of square platelets ca. 10 10
nm in size. The film was highly (112) textured.

7.2.16 Vanadium Oxide
Vanadium oxide films were deposited by dissolving V2O5 in aqueous HF and im-
mersing an Al plate in the solution (the Al acts as a scavenger for F in place of
the more commonly-used boric acid) [50a]. The brown film was X-ray amorphous
but crystallized on heating in air through a mixed V(IV)- V(V) oxide to V2O5 (in
an inert atmosphere, VO2 was formed on annealing). The as-deposited material
was believed to contain mainly V(IV).

7.2.17 Yttrium Oxide (Y2O3)
Basic yttrium carbonate [Y(OH)CO3] was deposited by CD and subsequently an-
nealed in air at 600°C to Y2O3 [51]. Si wafers and self-assembled monolayers with
sulphonate endgroups were used as substrates. An aqueous solution of YNO3 and
urea was heated at 80°C in sealed vials. The increase in pH, together with gener-
ation of carbonate from hydrolysis of urea (Sec. 3.2.4.1), resulted in formation of
the basic carbonate.
      The film thickness was 35 nm. Various analytical techniques were used to
confirm that the deposit was amorphous Y(OH)CO3. Annealing at 600°C was nec-
essary to convert the film to crystalline Y2O3 (amorphous oxide was formed be-
tween 300 and 400°C), with a film thickness of 25 nm and crystal size of ca. 20
nm.

7.2.18 Zinc Oxide (ZnO)
ZnO is the most studied of all the oxides deposited by CD. This is largely due to
its use as a transparent, electrically conducting layer.
       The first description of CD ZnO arose from the observation that deposition
of mixed (Cd,Zn)S resulted in large amounts of ZnO, and this led to development


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
of a technique for depositing pure ZnO [7]. The method is similar to that used for
CdO—heating an alkaline cyanide solution of a zinc salt to 80–90°C (although, in
contrast to the Cd case, no H2O2 was added). The films, a few hundred nanome-
ters in thickness, were shown by XRD to be ZnO. Sheet resistivities were ca. 108
   /sq, which dropped three orders of magnitude after heating at 350°C for 10 min
in forming gas.
        ZnO was deposited by CD using the freezing technique described previ-
ously for In2O3 [52]. The substrates (glass or quartz) were immersed in an aque-
ous solution containing ZnCl2, NH4F (as freezing agent—probably also acting as
a complexing agent for the Zn2 ), and Ag catalyst (as for In-O and SnO2, it is
not clear for what purpose), and NaOH was then added to a pH of between 7.5 and
8.5. The resulting Zn(OH)2 films were then annealed for a few hours in air or vac-
uum at 180–200°C. The rate of deposition and terminal thickness of the ZnO films
are shown in Figure 7.2. The films grow faster at higher pH but with a lower ter-
minal thickness than at lower pH. At a pH substantially lower than 7.5, no growth
occurs; the OH concentration is too low to precipitate Zn(OH)2. As the pH is in-
creased, the rate of formation of Zn(OH)2 increases. This results in increasing ho-
mogeneous precipitation in the solution, leading to loss of reactant and thinner
films, until, at a pH substantially greater than 8.5, essentially all the Zn is precip-
itated and no (or, more likely, only an ultrathin) film formation occurs.
        Al-doped ZnO films were also deposited by adding AlCl3 to the deposition
solution. The amount of Al in the films (given as at.% with respect to the Zn con-
centration) was somewhat smaller than that in the deposition solution but was pro-
portional to the concentration in solution (up to the maximum measured concen-
tration in the films of 5.5%).
        The films (both ZnO and ZnO:Al) were wurtzite structure with a preferen-
tial texturing (c-axis ⊥ substrate). No Al2O3 was found in the XRD spectra, sug-
gesting either dispersal of the Al in the ZnO matrix or its presence as very tiny
crystals of (hydr)oxide on the ZnO surface. TEM measurements showed an aver-
age grain size of 25 nm (ZnO) and 45 nm (ZnO:Al).
        Due to the possible application of these films for transparent electrically
conducting or infrared-reflecting purposes, the optical and electrical properties of
the films were the subject of careful study. Figure 7.3 from 268 shows the optical
transmittance and reflection spectra of both undoped and Al-doped films. The
doped films have a higher visible/near-IR transmittance. The transmittance cutoff
is blue-shifted for the doped film due to conduction band filling by electrons from
the Al dopant; this results in an increase in the effective optical bandgap (the
Burstein–Moss shift) and therefore an absorption blue shift. (The bandgaps mea-
sured from the spectra were 3.40 and 3.98 eV for the undoped and doped films, re-
spectively. The value of 3.4 eV for the undoped sample is considerably higher
than the usual value of 3.2 eV; these films were already quite highly conducting—
see the electrical measurements later and Table 7.1). The mid-IR reflectance of the
doped film also increased due to free-electron reflection.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 7.2 ZnO film thickness vs. deposition time for different values of solution pH.
(Adapted from Ref. 52.)




FIG. 7.3 Transmittance and reflectance spectra of ZnO films. Broken lines: ZnO; solid
lines: ZnO:A1 (4 at.%). (Adapted from Ref. 52.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 7.1 Variation of Electrical Parameters of ZnO Films with Al Doping and
Annealing Conditions

                       200°C                        225°C oxygen                 350°C vacuum

at.% Al        N                               N                             N

  0           0.67        7.7     12          2.5      10.4       2.5        2.4      10.2      2.5
  4           9.0        11.9      0.60      19        16.7       0.21      18        16.7      0.20
  5.5         9.4         1.5      4

200 C is the basic anneal that converts the hydroxide into the oxide. The films were then reannealed.
The two conditions given here are those that give the most highly conducting films. Higher tempera-
tures in oxygen decrease the conductivity, while higher temperatures in vacuum have no further effect.
N: free electron concentration ( 1020 cm 3); : mobility (cm 2V 1s 1); : resistivity 10 3 -cm.
(From Ref. 52.)



       The resistivity of the films decreased with increasing A1 content up to 4%
and then increased again to the maximum measured A1 content of 5.5%. This is
due to a combination of increase of carrier concentration at 4% A1, which then lev-
els off, and increase in mobility at 4% A1 followed by a sharp decrease at 5.5% A1.
The increase in carrier concentration is clearly due to the doping. Increase in mo-
bility was linked to the increased grain size of the doped films, while the decrease
in mobility (and conductivity) at doping levels 4% A1 was attributed to grain
boundary segregation, resulting in higher intergrain barriers. The relevant electri-
cal parameters are listed in Table 7.1. Further decrease in resistivity was obtained
upon annealing (see Table 7.1). In all cases where maximum conductivity was ob-
tained (350°C in vacuum or 225°C in oxygen), the decreased resistivity was due
mainly to increased carrier concentration, although a moderate increase in mobil-
ity was also measured. For vacuum-annealing, these two effects were explained by
an increase in oxygen vacancy concentration and desorption of oxygen from grain
boundaries (therefore decrease in grain boundary barrier), respectively. The reason
for the increase in carrier concentration and mobility after annealing in oxygen at
225°C is not clear, although the decrease in these parameters on annealing at higher
temperatures in oxygen follows naturally from the foregoing explanation.
       Hexagonal ZnO films were deposited on glass or SnO2 /glass using a solu-
tion of zinc acetate complexed with ethylenediamine and with the pH adjusted to
between 10.5 and 11.0 with NaOH [53,54]. The formation of the films occurred
only under conditions where Zn(OH)2 was calculated to be present in the deposi-
tion solution. Good-quality, adherent films were obtained only on glass and within
narrow pH and composition ranges. Films could be obtained outside these condi-
tions, but they were then poorly adherent. In general, the adherent films formed
under conditions where the deposition was slow (relatively low pH, relatively
high complex:Zn ratio).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       It was proposed that deposition of adherent films proceeded on Zn clusters
bound via OH groups to the glass surface. The poorer adherence on a SnO2 sur-
face was explained by a lower concentration of hydroxy groups on this surface.
       In a related study, McAleese and O’Brien showed how ZnO nucleated on
glass and SnO 2/glass from solutions of zinc acetate, ammonia, and thiourea (and
sometimes also hydrazine) [55]. Such a deposition solution is normally used to de-
posit ZnS, and this study showed that ZnO can form together with, or even instead
of, ZnS.
       From the optical spectra of the films deposited from the ethylenediamine-
complexed solutions, a bandgap of 3.15 eV was calculated (literature value 3.2 eV).
       ZnO was deposited on Si (100) by heating an ammoniacal solution (at pH
    7) of ZnAc2 [56]. The heating regime was important: First the solution was
heated to between 60 and 100°C for 6 hr and then the temperature was raised to
between 100 and 200°C for 6–12 hr (in an autoclave). The initial lower tempera-
ture was necessary to obtain deposition, and it was suggested that nucleation of
ZnO on the substrate occurred during this step. No XRD pattern was obtained af-
ter this lower-temperature stage; wurtzite ZnO was clearly seen by XRD after the
high-temperature stage. The films, ca. 65 nm thick, were smooth, dense, and ho-
mogeneous, with some (10.0) texture.
       ZnO films for use as buffer layers in photovoltaic cells (see Chap. 9) have
been chemically deposited from aqueous solutions of ZnSO4 and ammonia [57].
The solution was heated to 65°C, and adherent, compact Zn(OH)2 ZnO films
were formed after one hour. Low-temperature annealing converted the hydroxide
to oxide. The solution composition will be important in this deposition. On one
hand, increased ammonia concentration will increase the pH and therefore the ho-
mogeneous Zn(OH)2 precipitation in solution. However, further increase in am-
monia concentration will redissolve the hydroxide as the ammine complex. There
will clearly be an optimum ammonia (and zinc) concentration where Zn(OH)2
does form, but slowly enough to prevent massive homogeneous precipitation. The
use of ammonia in (hydr)oxide deposition derives, in part at least, from its grad-
ual loss by evaporation if the system is not closed [58]. Any open solution of an
ammonia-complexed metal ion (which forms an insoluble hydroxide or hydrated
oxide) should eventually precipitate the (hydr)oxide for this reason alone.
       The borane technique, described earlier for In2O3 preparation, was also used
to deposit ZnO on Sn/Pd-activated glass using a solution of Zn(NO3)2 and
dimethylamine borane at 50°C [59]. The films were randomly oriented polycrys-
talline ZnO, and the crystals were hexagonal shaped, with a typical size of 0.2 m,
at moderately high dimethylamine borane concentrations; at low concentrations,
the grains were smaller and more irregular.
       Optical transmission spectra gave an estimated bandgap of 3.3 eV. From the
spectra, the films showed some scattering, with the most transparent films having
an approximate integrated transmission over the visible region of 70%, obtained
from a solution containing 0.05 M/l dimethylamine borane. This correlated with



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the most regular morphology measured by SEM. A more recent study by the same
group, using higher Zn concentrations and slightly higher deposition temperature
(60°C), showed increased transmission up to nearly 80% over the visible range
with a borane concentration of 0.1 M [60].
      Electrical measurements of films as a function of boron doping were also re-
ported in this study. The amount of boron in the films varied over only a small
range, from 0.01% at very low borane concentrations in solution (a minimum con-
centration was needed for the deposition to proceed) to 0.02% at a borane con-
centration in solution of 0.1 M. The resistivity of the films was high in all cases,
varying from ca. 20 k -cm to 0.4 k -cm with increase in boron content. Mobil-
ity and carrier concentration measurements showed low values of both (maximum
values:       1 cm2V 1 sec 1, N 1.8 1016 cm 3). The low carrier concentra-
tion implies that these films are highly stoichiometric. While unfavorable for
transparent conducting purposes (although annealing would probably improve the
conductivity), the ability to make relatively insulating ZnO may be advantageous
for other purposes.


7.2.19 Zirconium Oxide (ZrO2)
A process similar to that described earlier for TiO2 was used to deposit ZrO2 films
[61]. Self-assembled monolayers with terminal sulphonate or methyl groups on Si
were used as substrates; no film growth occurred on bare Si. The deposition solu-
tion was Zr(SO4)2 dissolved in an aqueous HCl solution at 70°C.
       The film growth was slow—15 nm after 4 hr and with a limiting thickness
of 40 nm after ca. 20 hr, although it was faster at the beginning of the deposition.
The films were composed of a mixture of crystalline tetragonal ZrO2 and amor-
phous material—probably a basic zirconium sulphate. The (thicker) films varied
in their thickness, from up to 10-nm-sized crystals near the substrate to 2- to 3-nm
crystals, together with a greater proportion of amorphous basic sulphate, toward
the film surface. It was suggested that both electrostatic forces (between the neg-
atively charged sulphonate surface groups and positively charged zirconium ox-
ide and basic sulphate colloids) and van der Waals attractive forces cause the ob-
served good adhesion between the films and sulphonate monolayers while the
inferior adhesion to the uncharged methyl-terminated monolayers was due solely
to van der Waals forces.
       Annealing the films for 2 hr at 500°C, though causing pyrolysis of the
monolayer, did not damage the film (or its adhesion to the substrate), which re-
mained tetragonal ZrO2 with a crystal size ca. 10 nm and with some sulphate, the
latter disappearing after prolonged annealing. Annealing at 600°C or higher re-
sulted in a change to the monoclinic phase of ZrO2
       Yttrium-doped ZrO2 was deposited by adding Y2(SO4)3 and urea and de-
positing at 80°C (see deposition of Y2O3 described earlier). A higher pH (2.5–3.0)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
was needed to cause coprecipitation of Y2O3, and the Y concentration in the film
was ca. 25% of that in the deposition solution. Basic carbonates and sulphates
were also present in the films. Annealing in air at 500°C for 2 hr resulted in com-
plete crystallization of the films to yttrium-stabilized zirconia.

7.2.20 Ternary Oxides
Deposition of two ternary oxides—(Cd,Sn)O and (Zn,Cd)O—will be mentioned
very briefly next. These will be treated more fully in Chapter 8. Also, a brief de-
scription of a related technique (SILAR; see Sec. 2.11.1) that has been used for
complex oxides and related compounds will be given.

7.2.20.1 Cadmium Stannate (Cd2SnO4)
The ammonium fluoride technique used by Raviendra and Sharma for ZnO (de-
scribed earlier) has also been used by them to deposit cadmium stannate using a
mixture of CdCl2 and SnCl4 [52]. After annealing at over 200°C, Cd2SnO4 was ob-
tained. Optical and electrical properties of these films are described in Chapter 8.

7.2.20.2 Zinc Cadmium Oxide (Zn,Cd)O
Films of ZnxCd1 xO, with varying values of x, were deposited from ammoniacal
solutions of Cd and Zn chlorides containing hydrogen peroxide at 45°C followed
by annealing in air at 500°C, presumably to convert the hydroxides to oxide [62].
The optical and electrical properties of these films are described in Chapter 8.

7.2.20.3 SILAR Deposition of Metal Oxides,
         Hydroxides, and Peroxides
In SILAR (successive ionic-layer and reaction) deposition, discussed in Section
2.11.1, successive compound layers are built up from reaction between adsorbed
(ideally mono-) layers and a reactive solution. This technique has been applied to
sulphides, selenides, and oxides, including hydroxides, peroxides, and ternary ox-
ides. As an example, films of LaxNbOy were deposited [63]. This technique was
used to deposit many other oxides, hydrated oxides, and peroxides (see references
in Ref. 63). In view of the uncertain purpose of H2O2, often used in CD of oxides,
the role of this chemical as explained in this work is of interest. Except at a pH be-
low ca. 2, the surface of oxidized Si (Si was used as a substrate in these experi-
ments) in aqueous solution, and also glass, is composed of SiMO groups [see Eq.
(2.16)]. These groups can attract metal cations to give an uncharged (for monova-
lent cation) or positively charged (for a higher-valent cation) surface. Thus, for a
divalent cation:
      SiMO        M2 D SiMOMM                                                   (7.4)


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
This surface reacts with H2O2 to give
      SiMOMM           H2O2 D SiMOMMMOOH                 H                     (7.5)
which, in neutral or alkaline solution, dissociates to give a negatively charged sur-
face again:
      SiMOMMMOOH D SiMOMMMOO                         H                         (7.6)
Additionally, the ability of H2O2 to oxidize a metal to a higher-valent state, re-
sulting in a more insoluble hydroxide (higher-valent metal hydroxides are more
insoluble at a particular pH than the hydroxide of the metal in a lower-valent state)
has been pointed out in this study.


7.3 OTHER SEMICONDUCTORS: Se AND SILVER
    HALIDES
The remaining semiconductors (apart from ternaries, which are treated in Chap. 8)
that have been deposited by CD are elemental Se and silver halides. The little done
on these materials will be discussed here.

7.3.1 Se
Elemental Se exists in several forms, including amorphous (red), monoclinic
(red), and gray (hexagonal). Gray Se is the most stable form. There are several re-
ports on CD Se. The earliest is based on the fact that selenosulphate is unstable in
acidic solution and, if made acid, will immediately precipitate red Se (note that its
analogue, thiosulphate, behaves similarly but that a much lower pH is required to
precipitate elemental S). Se was deposited by slightly acidifying dilute (10–50
mM) selenosulphate solution [64]. Films of amorphous, red Se were obtained at
10–15°C, while gray Se was obtained at 30°C or higher. Using weak acids (citric
or ascorbic), amorphous Se films were likewise deposited at 0°C [65]. These con-
ditions slow the formation of Se enough to allow films (ca. 50 nm thick) to form.
A direct bandgap of 2.0 eV was measured for these films. Heating at 85°C trans-
formed the films to gray, hexagonal Se.
       Another technique, which is not strictly true CD but close enough to war-
rant inclusion, is photodeposition from an amorphous Se colloid [66–68]. The Se
colloid was prepared by reduction of a solution of SeO2. Illumination of a sub-
strate in this solution with light that was absorbed by the Se (bandgap 2.05 eV) re-
sulted in film formation on a substrate. Film formation occurred in the absence of
illumination but was extremely slow, particularly at lower temperatures (at tem-
peratures above 25°C, gray Se began to be formed). No XRD structure was found
for the films (deposited below 25°C). Raman spectroscopy revealed the presence
of Se chains and rings in the film. The mechanism of the deposition was not com-
pletely understood, but it was clearly connected with the photogeneration of


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
charges in the Se, since only superbandgap illumination was effective. The
bandgap of the films was measured, from the optical spectra, to be 2.05 eV [69].

7.3.2 Silver Halides (AgI, AgBr, AgCl)
Section 3.2.6 begins with the sentence “Although they do not appear to have been
used (at least deliberately) to form films, there are other slow anion-generating re-
actions.” And near the end of this section is written “It should be stressed that
these reactions were used to form precipitates and not films. There is no guaran-
tee that films can be formed using these reactions. However, it is reasonable to ex-
pect that, under the right conditions, it may be possible to produce films of these
compounds. It is left as an exercise for the curious reader to find these ‘right’ con-
ditions.”
       For most of the period while this book was being written, the halides were
included in that section of Chapter 3. However, the very act of writing down that
it should be possible to produce films of compounds containing these anions led
me (after a couple of years, when the opportunity arose to spend an extended pe-
riod of time in the laboratory during a visit to the university of Bern) actually to
try to do this. For a number of reasons, halides were the anion of choice, specifi-
cally silver halides. The slow generation of halides is discussed in Section 3.2.5
and is based on the slow hydrolysis of haloalcohols.
       AgI was deposited by hydrolysis of 2-iodoethanol (ca. 50 mM):
      ICH2CH2OH        H2O D I        H      HOCH2CH2OH                         (7.7)
in an aqueous solution containing AgNO3 (ca. 10 mM) and a small amount of tri-
ethanolamine (ca. 0.5 mM). The triethanolamine, originally added as a complex-
ant, reduced the Ag to Ag if used in high concentration. However, small amounts
were found to give a more homogeneous film than if no triethanolamine was used.
The deposition works best at room temperature—heating results in excessive pre-
cipitation in solution. The iodopropanol usually contains some free iodide, the
amount of which increases with age, and this can deleteriously affect the deposi-
tion if present in too high quantities. The yellow AgI films exhibited sharp XRD
peaks (no line broadening), showing them to be a mixture of wurtzite and spha-
lerite AgI. The films (other than very thin ones) scatter light moderately strongly,
and transmission spectra were taken using an integrating sphere. Figure 7.4 shows
the spectrum of such a film. The strong absorption onset at 440–450 nm is due to
the direct bandgap of AgI (ca. 2.8 eV). AgI also has an indirect absorption at
longer wavelengths, and the decrease in transmission over this region is due partly
to this absorption and probably partly to scattered radiation not collected in the in-
tegrating sphere.
       The deposition occurs in parallel with homogeneous precipitation, suggest-
ing that film formation is due to adhesion of crystals from the solution. This is sup-
ported by SEM pictures that show scattered crystal formation, with gradual den-

Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 7.4 Total transmission spectrum (measured with an integrating sphere) of a CD
AgI film on glass.


sification of the crystal structure as the deposition proceeds. Figure 7.5 A shows
such a micrograph of a AgI film on glass. The average crystal size is a few hun-
dred nanometers, and, as might be expected from such a morphology, the film ad-
hesion is poor—the films can be wiped off with a tissue, although they will usu-
ally stand up to cleaning in an ultrasonic bath. Adhesion is much better on




FIG. 7.5 TEM micrographs of (A) a CD AgI film on glass and (B) a similar film on
SnO2-coated glass.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
SnO2-coated glass; the films are not removed by rubbing with a tissue. The film
morphology is different than for the films on glass; the crystals are somewhat
smaller and tightly aggregated (Fig. 7.5B).
       AgBr films could be made in the same manner using various bromo-alco-
hols. AgBr is colorless but is usually slightly brownish due to photolytic forma-
tion of small particles of metallic silver (this occurs much more strongly if depo-
sition is carried out in room light, but formation of some brown coloration is still
noticeable even if deposition is carried out in the dark). This coloration masked
the optical absorption, which occurs mainly in the UV for AgBr. For bromides, tri-
ethanolamine was not needed.
       AgCl was not so readily deposited, and while occasionally some film for-
mation did occur, it was nonhomogeneous, thin, and irreproducible. AgCl could
be deposited by immersing a substrate in an aqueous solution of AgNO3 to which
NaCl solution was added, resulting in a cloudy precipitate/sol of AgCl. The con-
centrations of both reactants were important: ca. 20 mM each, preferably with a
slight excess of Ag. Much lower concentrations resulted in little deposition (at
least within a reasonable time), while even twice that concentration resulted in im-
mediate coagulation of the AgCl and no film formation. What was unique about
this deposition was that on two occasions (among five or six experiments in total),
visible film formation occurred virtually immediately on mixing the solutions.
This is the only example known to the author where visible film formation occurs
in a rapid precipitation and contradicts the “conventional wisdom,” which other-
wise seems to be valid, that the reaction leading to formation of the compound
must be slow in order for appreciable film formation to occur.
       More details on this work are given in Ref. 70.


7.4 EXTENSION OF CD TO OTHER MATERIALS?
To conclude this chapter, we look back at the earlier literature in hopes of widen-
ing both the potential deposition methods and the materials that can be deposited.
As well as oxides and related compounds, other anions are considered. The re-
sulting compounds do not necessarily fall under the common heading of semi-
conductors, but they are relevant in the hope of expanding the scope of chemical
deposition.
        Table 7.2 summarizes a range of homogeneous precipitation reactions. De-
tails of all these reactions can be found in Ref. 2 (this book, in spite of its age, is
required reading for anyone wishing to pursue this line; more recent books may
exist, but will probably not reduce its value). It should first be stressed that the ma-
terial in this final section relates to precipitates rather than to films. However, with
some effort (in some cases only a little or none, as seen from the common film for-
mation occurring in urea precipitations), it is reasonable to expect extension to
form films of the same materials in at least some cases.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 7.2 Homogeneous Precipitation Reactions

Precipitant                         Reagent                                Elements precipitated

Hydroxide           Urea                                              Al, Ga, Th, Fe(III), Sn, Zr
                    Acetamide                                         Ti
                    Hexamethylenetetramine                            Th
                    Metal chelate H2O2                                Fe(III)
Phosphate           Triethyl phosphate                                Zr, Hf
                    Trimethyl phosphate                               Zr
                    Metaphosphoric acid                               Zr
Oxalate             Dimethyl oxalate                                  Th, Ca, Am, Ac, rare earths
                    Diethyl oxalate                                   Mg, Zn, Ca
                    Urea and an oxalate                               Ca
Sulphate            Dimethyl sulphate                                 Ba, Ca, Sr, Pb
                    Sulphamic acid                                    Ba, Pb, Ra
                    Potassium methyl sulphate                         Ba
                    Ammonium persulphate                              Ba
                    Metal chelate persulphate                         Ba
Sulphide            Thioacetamide                                     Pb, Sb, Bi, Mo, Cu, As, Cd,
                                                                      Sn, Hg, Mn
Iodate              Iodine chlorate                                   Th, Zr
                    Periodate ethylene diacetate                      Th, Fe(III)
                      (or -hydroxyethyl acetate)
Carbonate           Trichloroacetate                                  Rare earths, Ba, Ra
Chromate            Urea dichromate                                   Ba, Ra
                    Potassium cyanate dichromate                      Ba, Ra
                    Cr(III) bromate                                   Pb
Chloride            Ag–ammonia complex chloride                       Ag
                      hydroxyethylacetate
Arsenate            Arsenite nitric acid                              Zr
Fluoride            Fluoboric acid                                    La

Source: Modified from: I. M. Kolthoff and P. J. Elving, eds. Treatise on Analytical Chemistry, Part 1,
Vol. 1. New York: Interscience Encyclopedia, 1959, p 741.



       The list in Table 7.2 may appear incomplete to the modern chemist utilizing
or studying chemical deposition; e.g., only thioacetamide is noted as a sulphide
source and selenides are not included. However, when we reflect that the vast bulk
of the work carried out on CD concerned just sulphides, selenides and oxides, this
“old” table might point the way to a major expansion of the CD technique, both
for semiconductors and for other compounds. Further processing may be expected
to extend the types of material even further. For example, arsenates and phos-
phates may be reducible in some cases to the better-known (to the semiconductor
community) arsenides and phosphides.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
REFERENCES
 1.    L Gordon. Anal. Chem. 24:459, 1952.
 2.    L Gordon, ML Salutsky, HH Willard. Precipitation from Homogeneous Solutions.
       New York: Wiley, 1959.
 3.    HH Willard, NK Tang. Ind. Eng. Chem. Anal. Ed. 9:357, 1937.
 4.    HH Willard, HC Fogg. J. Am. Chem. Soc. 59:40, 1937.
 4a.   HH Willard, L Gordon. Anal. Chem. 25:170, 1953.
 5.    TP Niesen, MR De Guire. J. Electroceram. 6:169, 2001.
 6.    RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:1642, 1982.
 7.    RL Call, NK Jaber, K Seshan, JR Whyte Jr. Sol. Energy Mater. 2:373, 1980.
 8.    M Najdoski, I Grozdanov, B Minceva. J. Mater. Chem. 6:761, 1996.
 9.    M Ortega, G Santana, A Morales-Acevedo. Solid State Electron. 44:1765, 2000.
10.    M Ocampo, AM Fernandez, PJ Sebastian. Semicond. Sci. Technol. 8:750, 1993.
11.    M Ocampo, PJ Sebastian, J Campos. Phys. Status Solidi (a) 143:K29, 1994.
12.    AJ Varkey, AF Fort. Thin Solid Films 239:211, 1994.
13a.   FC Eze. J. Phys. D: Appl. Phys. 32:533, 1999.
13.    AJ Varkey, AF Fort. Sol. Energy Mater. Sol. Cells 31:277, 1993.
14.    B Pejova, A Isahi, M Najdoski, I Grozdanov. Mater. Res. Bull. 36:161, 2001.
15.    I Grozdanov. Mater. Lett. 19:281, 1994.
16.    M Ristov, G Sinadinovski, I Grozdanov. Thin Solid Films 123:63, 1985.
17.    RP Goyal, D Raviendra, BRK Gupta. Phys. Status Solidi (a) 87:79, 1985.
18.    M Izaki. Electrochem. Solid State Lett. 1:215, 1998.
19.    KO Velthaus, J Kessler, M Ruckh, D Hariskos, D Schmid, HW Schock. In: 11th
       ECPV Solar Energy Conf., Montreux, Switzerland, 1992, p 842.
20.    PC Rieke, BD Marsh, LL Wood, BJ Tarasevich, J Liu, L Song, GE Fryxell. Langmuir
       11:318, 1995.
21.    BJ Tarasevich, PC Rieke, J Liu. Chem. Mat. 8:292, 1996.
22.    PC Rieke, R Wiecek, BD Marsh, LL Wood, J Liu, L Song, GE Fryxell, BJ Tarase-
       vich. Langmuir 12:4266, 1996.
23.    QW Chen, YT Qian, H Qian, ZY Chen, WB Wu, YH Zhang. Mater. Res. Bull.
       30:443, 1995.
24.    QW Chen, YT Qian, ZY Chen, Y Xie, GE Zhou, YH Zhang. Mater. Lett. 24:85, 1995.
25.    QW Chen, XG Li, YT Qian, YH Zhang. Mater. Lett. 31:247, 1997.
26.    B Pejova, M Najdoski, I Grozdanov, A Isahi. J. Mater. Sci.-Mater. Electron. 11:405,
       2000.
27.    M Izaki, O Shinoura. Adv. Mater. 13:142, 2001.
28.    P Ruëtschi, B Cahan. J. Electrochem. Soc. 104:406, 1957.
29.    W Mindt. J. Electrochem. Soc. 117:615, 1970.
30.    RN Bhattacharya, P Pramanik. Bull. Mater. Sci. 2:287, 1980.
31.    S Gorer, A Albu-Yaron, G Hodes. Chem. Mater. 7:1243, 1995.
32.    P Pramanik, S Bhattacharya. J. Mater. Sci. Lett. 6:1105, 1987.
33.    KL Chopra, RC Kainthla, DK Pandya, AP Thakoor. In: Physics of Thin Films, Vol.
       12, Academic Press, New York and London, 1982, p 167.
34.    W Mindt J. Electrochem. Soc. 118:93, 1971.
35.    E Beutel, A Kutzelnigg. Monats. 58:295, 1931.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
36.    P Pramanik, S Bhattacharya. J. Electrochem. Soc. 137:3869, 1990.
37.    AJ Varkey, AF Fort. Thin Solid Films 235:47, 1993.
38.    B Pejova, T Kocareva, M Najdoski, I Grozdanov. Appl. Surf. Sci. 165:271, 2000.
39.    H Nagayama, H Honda, H Kawahara. J. Electrochem. Soc. 138:2013, 1988.
40.    AJ Varkey, AF Fort. Sol. Energy Mater Sol. Cell 29:253, 1993.
41.    T Kocareva, I Grozdanov, B Pejova. Mater. Lett. 47:319, 2001.
42.    D Raviendra, JK Sharma. J. Phys. Chem. Solids 46:945, 1985.
43.    QW Chen, YT Qian, ZY Chen, GE Zhou, YH Zhang. Thin Solid Films 264:25, 1995.
44.    D Hariskos, R Heberholz, M Ruckh, U Ruhle, R Shäffler, HW Schock. In: 13th
       ECPV Solar Energy Conf., Nice, France, 1995, p 1995.
45.    S Supothina, MR De Guire. Thin Solid Films 371:1, 2000.
46.    K Tsukuma, T Akiyama, H Imai. J. Non-Cryst. Sol. 210:48, 1997.
47.    BC Bunker, PC Rieke, BJ Tarasevich, AA Campbell, GE Fryxell, GL Graff, L Song,
       J Liu, JW Virden, GL McVay. Science 264:48, 1994.
48.    H Shin, RJ Collins, MR Deguire, AH Heuer, CN Sukenik. J. Mater. Res. 10:692,
       1995.
49.    RJ Collins, H Shin, MR DeGuire, AH Heuer, CN Sukenik. Appl. Phys. Lett. 69:860,
       1996.
50.    QW Chen, YT Qian, ZY Chen, WB Wu, ZW Chen, GE Zhou, YH Zhang. Appl. Phys.
       Lett. 66:1608, 1995.
50a.   S Deki, Y Aoi, Y Miyake, A Gotoh and A Kajinami. Mater. Res. Bull. 31:1399, 1996.
51.    M Agarwal, MR DeGuire, AH Heuer. Appl. Phys. Lett. 71:891, 1997.
52.    D Raviendra, JK Sharma. J. Appl. Phys. 58:838, 1985.
53.    T Saeed, P O’Brien. Thin Solid Films 271:35, 1995.
54.    P O’Brien, T Saeed, J Knowles. J. Mater. Chem. 6:1135, 1996.
55.    J McAleese, P O’Brien. Mater. Res. Soc. Symp. Proc. Vol. 485. 1998, p 255.
56.    QW Chen, YT Qian, ZY Chen, YH Zhang. Mater. Lett. 22:93, 1995.
57.    A Ennaoui, M Weber, R Scheer, HJ Lewerenz. Sol. Energy Mater. Sol. Cells 54:277,
       1998.
58.    AJ Varkey. Int. J. Mater. Prod. Technol. 10:94, 1995.
59.    M Izaki, T Omi. J. Electrochem. Soc. 144:L3, 1997.
60.    M Izaki, J Katayama. J. Electrochem. Soc. 147:210, 2000.
61.    M Agarwal, MR De Guire, AH Heuer. J. Am. Ceram. Soc. 80:2967, 1997.
62.    G Contreras-Puente, O Vigil, M Ortega-Lopez, A Morales-Acevedo, J Vidal, ML Al-
       bor-Aguilera. Thin Solid Films 361:378, 2000.
63.    VP Tolstoy. Thin Solid Films 307:10, 1997.
64.    GA Kitaev, GM Fofanoy. Zh. Prikl. Khim. 43:1694, 1970.
65.    B Pejova, I Grozdanov. Appl. Surf. Sci. 177:152, 2001.
66.    M Perakh, A Peled, Z Feit. Thin Solid Films 50:273, 1978.
67.    M Perakh, A Peled. Thin Solid Films 50:283, 1978.
68.    M Perakh, A Peled. Thin Solid Films 50:293, 1978.
69.    A Peled. Philosoph. Mag. B 53:171, 1986.
70.    G Hodes, G Calzaferri. Adv. Funct. Mater. in press.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
8
Ternary Semiconductors




8.1 GENERAL CONSIDERATIONS FOR
    PRECIPITATION AND CHEMICAL
    DEPOSITION OF TERNARY COMPOUNDS
Mixed compositions are of interest mainly because they allow tuning of the semi-
conductor properties (most commonly bandgap and, therefore, spectral sensitiv-
ity). This is useful for various device applications. Photoconductive detectors,
where a certain spectral sensitivity range is desired, is probably the main applica-
tion that drove many studies on CD of ternary semiconductors.
       Mixed metal chalcogenides have been deposited by CD. According to sim-
ple fundamental considerations, the deposition should proceed according to the
solubility products of the two separate metal chalcogenides; the one with a smaller
Ksp should precipitate first, and only after, when the concentration of free (first)
metal ion was low enough, would the other chalcogenide precipitate, assuming a
sufficient supply of chalcogenide ions. If the difference in Ksp was large, then the
solution would be almost entirely depleted of the low-Ksp chalcogenide before
precipitation of the second would start. Note that this discussion relates to CD
where deposition is slow; for rapid precipitation, kinetic factors might be more im-
portant, and differences in concentrations of the two cations are likely to play a
more dominant role.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       This picture is in many (probably most) cases oversimplified. There are
other factors that may become important in such mixed products. The literature is
full of examples of precipitation of mixtures/compounds, and we will consider
some of these, keeping in mind that they refer to precipitation and not CD. One is
coadsorption of one cation on the precipitated compound of the other. A well-
known example of this phenomenon is adsorption of Cr 3 on ferric hydroxide pre-
cipitated from solutions containing Fe3 and small concentrations of Cr 3 . The
chromium is precipitated together with the Fe(OH)3 at a pH where chromium hy-
droxide itself would not readily precipitate [1]. This is due to strong adsorption of
the “impurity” cation (Cr 3 ) on the precipitating Fe(OH)3. The ability of some
precipitates to either cause coprecipitation of other metal salts that would, by
themselves, not precipitate under the same conditions—an effect that was known
as induced precipitation—or often even incorporate the “soluble” metal salt if the
freshly precipitated insoluble salt was exposed to a solution containing the more
soluble one—has been known for a long time; many examples from the old liter-
ature are given in a review by Kolthoff and Moltzau [2].
       Another, and on the face of it, rather different example, is the coprecipita-
tion of solid solution compounds, such as CuInS2 and CuInSe2—semiconductors
of particular interest due mainly to their applicability for photovoltaic cells. It was
shown, by X-ray diffraction, that the precipitate resulting from reaction between
H2S and an aqueous solution containing both Cu and In3 ions was, at least in
part (depending on the concentrations of the cations), single-phase CuInS2 [3].
Two factors were found to be necessary for this compound formation: (1) the pres-
ence of sulphide on the surface of the initially precipitated colloidal solid metal
sulphide and (2) one of the cations being acidic and the other basic. The monova-
lent Cu cation is relatively basic, while the trivalent In3 cation is relatively
acidic. It is not clear what the physical reason is for this latter requirement. A dif-
ference in practice between acidic and basic cations is that, in an aqueous solution
of both cations, the acidic cation is more likely to be in the form of some hydroxy
species (not to be confused with hydrated cations), while the basic cation is more
likely to exist as the free cation.
       In a subsequent study of AgMGaMS precipitation by the same group, a
mechanism for solid solution formation was proposed [4]. Colloidal Ag2S, with
its lower solubility product, formed initially. The surface of the colloid adsorbed
both sulphide and Ga (or Ga-hydroxy; see earlier) ions to form a gallium sulphide
layer. This then would be similar to induced precipitation. It was suggested that
this GaMS then diffused into the Ag2S, where solid solution formation occurred.
In view of the high mobility of Ag ions, it seems more likely that Ag diffused
outward rather than the Ga-species diffusing inward. Of course, precipitation is a
highly nonequilibrium process, while CD, depending on which mechanism is op-
erative, is closer to an equilibrium process (an ion-by-ion deposition occurs close
to equilibrium, while the initial hydroxide formation in a hydroxide mechanism is



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
more like a precipitation reaction; the subsequent chalcogenide exchange is again
closer to equilibrium). Thus, extension of precipitation reactions to CD processes
may be useful, but the comparison should be made cautiously.
       Various possibilities for precipitation of two cations, M1 and M 2, by sul-
phide are shown in Figure 8.1. The case where both metal sulphides have similar
solubility products (Cd,PbS would be an example of this) is shown in the left-hand
process, where a particle of M1M2S is formed (to simplify things, no information
on stoichiometry is implied here). This particle may be either a mixed phase or a
solid solution, depending on the miscibility of the two sulphides and the kinetics
of the precipitation. The middle process shows the case for two sulphides with dif-
ferent solubility products. The metal sulphide with the lower-solubility product
(assumed to be M1) precipitates as a separate phase, which adsorbs both sulphide
ions and M2 ions (the latter either onto adsorbed S or onto lattice S). Eventually a
shell of M 2S (probably containing M1S) will form. Depending on the driving
forces involved, this core-shell structure may remain in that state, or diffusion into
a mixed two-phase particle or single-phase solid solution may occur. A third pos-
sibility, shown on the right-hand side, is that only M1S is formed. If the M1 is suf-
ficiently depleted by precipitation of M1S, and if sulphide formation continues,
then M2S will eventually precipitate.
       These processes have been shown for free metal ions. However, if a cluster
mechanism based on metal hydroxide colloids is involved, they are equally appli-
cable to the formation of the solid hydroxide species. The degree of conversion of




FIG. 8.1 Scheme of various possibilities for coprecipitation of two metal (M1 and M2)
sulphides.


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the hydroxides to sulphide will then depend on the differences between the solu-
bility products of the hydroxides and sulphides as well as the hydroxide concen-
tration in solution (pH and temperature). If a complex-decomposition mechanism
is involved, then the most important factor is probably the differences between the
strengths of the metal–sulphur (keeping sulphide as our example) bonds, and this
is similar to the differences in solubility products of the sulphides.
       These processes have been described for rapid precipitation reactions. How-
ever, they should also be valid in general for slow precipitation—i.e., for CD—
with possible differences due to the very different kinetics involved. Thus, if free
sulphide is involved, since it is always present in very low concentration, the
lower-solubility product metal sulphide is more likely to deposit first, compared
to rapid precipitation. Solid-state diffusion processes have much more time to oc-
cur in CD (although they may occur in rapid precipitation after the precipitation
itself), increasing the probability of solid solution formation.
       Probably the most basic question to be asked when depositing ternaries is: Is
the as-deposited material a true single-phase solid solution? A look at Table 2.3
shows that this is often not the case. In some studies, clear structural characteriza-
tion (usually XRD) shows that the film is (at least predominantly) a single-phase
solid solution. Others state that the product is not a solid solution or contains a large
component of other phases. However, some studies claim that the films are solid
solution, without presenting clear evidence. In most cases, these are older studies,
based on XRD spectra which, by today’s standards, are not clear. The XRD spec-
trometers of today are a lot better than those of not so long ago. Additionally, nowa-
days it is recognized that nanocrystalline films (as usually deposited by CD) often
require more care in sampling, to avoid the common danger of incorrectly pro-
nouncing them “amorphous” or “poorly crystalline.” Films that are not a true solid
solution as deposited may often be converted to a solid solution by annealing. Sim-
ilarly, solid solution formation may sometimes occur if two separate films are de-
posited, one on top of the other, and then annealed to effect interdiffusion. While
the original intention of this chapter was to confine the contents to genuine solid
solutions (or at least those that might be solid solutions), in some cases ternary
films that clearly are not solid solutions are included, and this is then made clear.
       Another consideration is whether the deposited film is homogeneous
throughout its thickness. If the composition is a function of the relative solubility
products of the individual binary compounds, then the metal ion that has the
lower-solubility product with the anion will deposit preferentially at first, but, due
to depletion, it may become lower in concentration in the film as deposition pro-
ceeds. Thus for complete characterization of these films, compositional analysis
should ideally be made as a function of spatial position in all three dimensions;
this is rarely carried out in practice.
       The semiconductors described in this chapter are divided into two types:
those composed of two different metal cations (most of the studies) and those with
two different anions. In their 1982 review [5], Chopra et al. give a list of eight dif-


Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ferent ternary compounds that had been deposited by CD at that time. The work
on some of those compounds has apparently not been published, other than in the-
ses, and details of their preparation and characterization are not given; therefore
those studies are not discussed here. These include the sulphides of Cd,Zn and
Cd,Hg and the selenides of Cd,Hg, Cd,Pb, and Pb,Hg. Additionally, two papers on
bilayer formation are treated separately in Section 8.4.

8.2 MIXED METAL COMPOUNDS
8.2.1 (Cd,Zn)O
(Cd,Zn)O films were deposited on glass at 45°C from solutions of Cd and Zn chlo-
rides to which ammonia and then H2O2 were added (the purpose of the H2O2 was
not given); they were then annealed at a final temperature of 500°C [6]. (No de-
scription of the as-deposited films was given; they were presumably mixed hy-
droxides.)
       Optical spectroscopy showed that the optical bandgap shifted strongly to the
red with increasing Cd concentration (at fixed Zn concentration) for small Cd:Zn
ratios in the deposition bath, but it was affected only to a small extent by varia-
tions in this ratio when the Cd:Zn ratio was greater than ca. unity. From this it can
be inferred that Cd(OH)2 was preferentially deposited, even though the solubility
product of Zn(OH)2 is lower (by ca. 50 times). This could be explained by the
greater strength of the Zn-ammine complex compared to the corresponding Cd
complex (two orders of magnitude higher), resulting in a hundred times lower
free-Zn2 concentration compared to Cd (for the same total concentration of
each). This more than offsets the lower-solubility product of the Zn(OH)2, al-
though the difference is not large, resulting in preferential deposition of Cd(OH)2.
This provides a good example of the need to consider all the relevant parameters
when trying to understand the specifics of the depositions.
       The resistivity of pure CdO was 3 10 3 -cm (CdO is normally a de-
generate n-type semiconductor), which increased approximately linearly (on a
semilog scale) with increasing solution Zn content up to 107 -cm (at 60% Zn)
and then tailed off to a value of ca. 108 -cm for very Zn-rich films.

8.2.2 (Cd,Zn)S
By far the greatest interest and effort in the CD of ternary semiconductors has been
focused on cadmium zinc sulphide (Cd,Zn)S. This interest has been driven by the
expected improvement in performance of thin-film photovoltaic cells (CdTe- and
CIS (CuInSe2-based cells) using (Cd,Zn)S rather than the presently used CdS.
This expectation arises mainly from the increased bandgap of the Zn-containing
solid solution, resulting in increased transparency to shorter wavelengths of light
(see Sec. 9.1.4.5 for more details). Another consideration for heterojunction for-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
mation is the decrease in electron affinity of the semiconductor with increase in
Zn content (ZnS has a smaller electron affinity than CdS). The electron affinity of
a semiconductor is a measure of the position of the conduction band with respect
to the vacuum energy level; a lower value means a higher conduction band. Thus
the alignment of the conduction band of the (Cd,Zn)S with that of the second
semiconductor can be controlled to a large extent by varying the film composition.
       In spite of the overall chemical similarity of Cd and Zn, however, it has not
proven simple to deposit true solid solutions of the sulphides. There are a number
of reasons for this, some of which have been treated in detail in Chapter 4 in the
discussion on ZnS deposition. We sum them up here.
       While CdS is less soluble than ZnS, Cd(OH)2 is more soluble than Zn(OH)2.
For this reason, ZnS is more difficult to deposit than CdS, since Zn(OH)2 tends to
form instead of, or together with, ZnS. (Although the solubility product of ZnS is
lower than that of Zn(OH)2, the concentration of hydroxide in any typical aque-
ous solution will be much higher than that of sulphide). In an alkaline solution (the
most common medium for CD), CdS deposition will be preferred over ZnS.
       The concentration of free Cd 2 should be much lower than that of Zn2 in
order for ZnS to deposit according to simple solubility product considerations.
However, the strength of complexation of most ligands is comparable for both Cd
and Zn (ammonia and hydroxide give stronger complexes with Zn). We know of
no ligand that will complex Cd enough to bring its free-ion concentration in solu-
tion the orders of magnitude lower than that of Zn in the same solution that is re-
quired (cyanide is maybe the closest to this ideal, but the difference is still not
enough, and cyanide is such a strong complexant that deposition might be rela-
tively difficult from solutions containing it in large amounts). Because of the
lower solubility of Zn(OH)2, it should be possible to adjust the complexant con-
centration so that Zn(OH)2 is present in solution but Cd(OH)2 is not, ignoring the
possibility of induced coprecipitation. CdS would therefore be formed by the
(usually slower) ion-by-ion mechanism, while ZnS might be formed by the (usu-
ally faster) hydroxide cluster mechanism. This is probably not as ideal as it may
sound, both because Zn(OH)2 does not readily methasize to ZnS (due to the much
higher hydroxide than sulphide concentration in the solution) and because, if it did
occur, it is more likely that separate phases would be formed.
       One point in favor of a single-phase solid solution deposition is that CdS and
ZnS do readily form solid solutions in general. Thus, if the two sulphides can be
simultaneously deposited, there is a good chance that they will form a solid solu-
tion if the temperature is high enough. Put another way, annealing of a well-mixed
two-phase mixture of CdS and ZnS will form a solid solution if the temperature is
high enough. For finely divided precipitates (as normally occurs in CD), this tem-
perature is expected to be relatively low.
       There are a number of reports on CdMZnMS deposition where the CD film
was either clearly shown to be mixture of phases [7] or there was insufficient ev-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
idence to support solid solution formation [8,9]. The first well-characterized de-
position of a true (Cd,Zn)S alloy film was described by Padam et al. [10]. Cd and
Zn acetates were used in various ratios complexed with ammonia and tri-
ethanolamine and thiourea at 90–95°C to deposit (Cd,Zn)S over the complete
composition range onto glass substrates. Interestingly, the Zn will be more heav-
ily complexed than the Cd in this solution, which shows that the mechanism of de-
position is not one based solely on solubility products of the sulphides. In fact,
from the crystal size measurements of a similar deposition described in Ref. 11
(see later), it is possible that the deposition mechanism is different for the two
cations (the small crystals of pure ZnS and the larger ones of pure CdS suggest a
cluster mechanism and an ion-by-ion mechanism, respectively).
       The films were characterized by a variety of techniques. Elemental analysis
(EDS) showed that the Zn:Cd ratio in the film was almost equal (slightly less) than
that in the deposition solution. X-ray diffraction and ED were used for phase and
compositional analyses. All the compositions up to 80% Zn were wurtzite struc-
ture, while pure ZnS was sphalerite. Interestingly, while most of the films gave
ring ED patterns showing nonoriented growth, some showed a degree of orienta-
tion, in spite of the glass substrate.
       The bandgap, calculated from optical absorption spectroscopy, varied al-
most linearly with composition between that of CdS (2.4 eV) and ZnS (3.6 eV),
providing further evidence for solid solution formation.
       The films were all n-type (hot probe) with resisitivity that varied linearly
(on a log scale) from 109 -cm (CdS) to 1012 -cm (ZnS). Doping by In (as
InCl3 in the deposition solution) reduced ; e.g., for a Cd0.8Zn0.2S film, dropped
linearly (on a log scale) with In content from ca. 1010 -cm (undoped) to ca. 105
   -cm (1.5% In—the In ratio in the film was similar to that in the solution). At high
In ratios, increased, explained by a decrease in mobility due to scattering by In.
Annealing in H2 at 200°C also decreased . For example, a Cd0.8Zn0.2S:1.5% In
film showed a minimum value for of ca. 10 -cm, presumably due to loss of S.
       This same method was more recently repeated with very similar results
[11]. It was additionally found that the films were strongly textured (only one
XRD peak—either (0001) wurtzite or (111) sphalerite), although this texture
was lost if a subsequent layer was deposited to produce thicker films. The crys-
tal size (measured from XRD peak width) varied from 20 nm (CdS) to 9 nm
(ZnS). The bandgap varied between the same limits as found in the previous
study, but changed more rapidly for high Zn content. The resistivity of the films
varied (linearly on a log scale) from 109 -cm (CdS) to 1014 -cm (ZnS)—
the latter higher than the value measured by Padam et al. Boron doping (addi-
tion of boric acid to the solution) decreased the resistivity of CdS by three or-
ders of magnitude.
       Using ammonia-complexed metal iodides and thiourea at pH 10, films were
formed whose properties depended on the temperature–time regime of the depo-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
sition solution [12]. If the reagents were mixed at room temperature and deposi-
tion occurred while heating the solution to 80°C, only CdS was deposited from
various mixtures of Cd and Zn iodides, even where the Zn was in large excess. If,
however, the reagents were mixed at a higher temperature (60°C), followed by
heating the bath to 80°C during deposition, Zn was incorporated into the films, al-
though the Zn content in the films was much lower than in the solutions; the Zn
content increased slowly up to ca. 80% Zn in solution and then rapidly at higher
Zn concentrations. This was seen by both optical spectra and XRD, the latter in
particular supporting solid solution formation. A reason for this dependence on
temperature programming was not suggested. A possible clue may be obtained
from a consideration of the temperatures at which solid Cd(OH)2 and Zn(OH)2 are
formed (from thermodynamic calculations based on the equilibrium constants of
the Cd- and Zn-ammine complexes, the solubility products of the hydroxides, and
dependence of hydroxide concentration in water on temperature). At room tem-
perature, no solid hydroxide phase is calculated to be present in the various Cd/Zn
solutions. At approximately 40°C, hydroxide will form (slightly lower tempera-
ture for Cd, slightly higher for Zn, although the difference is not large, and, in
view of the approximations used in these calculations, as well as kinetic factors, it
is not certain that Cd(OH)2 will, indeed, form first). Also, it has been shown that
Cd(OH)2 can form on the substrate in some cases before it forms in solution (dis-
cussed in Chap. 3), but the equivalent experiment has not been done for Zn. How-
ever, if we assume that Cd(OH)2 will form before Zn(OH)2 as the temperature is
raised, then this might explain, at least in part, the formation of only CdS. For the
case where deposition was started at 60°C, then both hydroxides are present to be-
gin with and it is more likely that Zn will be incorporated into the final films. This
reasoning is based on the expectation that cluster deposition will be much faster
than ion-by-ion deposition.
       The resistivity of the films decreased with increase in Zn content (from 1010
   -cm for CdS (a very high value for CdS) to ca. 106 -cm for 90% (solution con-
centration) Zn and then increased to ca. 109 -cm for pure ZnS. No explanation
for this effect was given. The films were photoconductive, with the resistivity de-
creasing in a somewhat sporadic manner as a function of composition, up to a
maximum dark:light ratio of 5 103 for the 90% Zn films.
       Similar conditions, but at a lower pH of 8.4, were also used by the same
group [13]. The ammonia concentration was reported to be important in forming
the solid solutions, although this concentration was not given. Only Zn-rich solu-
tions were described in this study (between 80 and 99% Zn in solution). As for the
previous study at higher pH, the Zn concentrations in the films were quite differ-
ent from those in the solution (except for very Zn-rich solutions); the films were
richer in Cd up to at least 92% Zn in solution, and then the Zn concentration in the
films increased rapidly with further increase in solution Zn concentration up to
99% Zn, from which deposited films with very little Cd.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       X-ray diffraction showed well-defined peaks that shifted in position with
change in composition, while the optical absorption spectra gave values of Eg that
also varied gradually with composition. The well-defined shift in XRD peaks and,
to a somewhat lesser extent, the gradual change in estimated bandgap with com-
position provide good evidence for true solid solution formation.
       Unlike other studies as well as the higher-pH studies by the same group, the
electrical resistivity did not vary much with composition, being ca. 5 107 -cm.
An overview of the variation of resistivity of some of these (Cd,Zn)S films with
composition is given in Figure 8.2. There is quite a large variation, both in resis-
tivity values and in their compositional dependence.
       The question must be asked: Why are solid solutions formed in some cases
and not in others? A common denominator in the successful films and their dif-
ference from the unsuccessful ones (success being defined as formation of a solid
solution) is the higher temperature used in the former (80–95°C). Higher temper-
ature will facilitate intermixing of the codeposited CdS and ZnS.

8.2.3 (Cd,Zn)Se
Two selenosulphate baths have been described for (Cd,Zn)Se. In the first [14],
solid solution formation was claimed, at least for annealed (300°C) samples, al-




FIG. 8.2 Resistivity data for some (Cd,Zn)S films. The data from the two Yamaguchi
papers were modified to show resistivity values as a function of approximate film compo-
sition, rather than solution composition as given in the original papers.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
though XRD results of the (annealed) samples were difficult to understand (the
“a” lattice parameter increased as Zn concentration increased). The effects of
composition and annealing on the bandgaps and electrical conductivity were de-
scribed. In the other [15], solid solution formation over part of the composition
range was claimed. However, the XRD spectra of the films (relatively sharp peaks
characteristic of ZnO), together with the visual description of the pure ZnSe films
(white with a slight greyish tinge), suggest mixed-phase formation containing
ZnO.

8.2.4 (Cd,Hg)S
Solid mixtures of CdS and HgS have been shown to form solid solutions after
treatment with certain solutions, such as concentrated ammonium sulphide [16].
This may be due partly to the very similar ionic radii of the two cations and
(maybe more important) the ability of Hg to diffuse readily in solids. Therefore it
is probable that solid solutions can readily form in this system.
       (Cd,Hg)S was deposited from a solution containing CdCl2 and HgCl2 com-
plexed with a low concentration of cyanide (15 mM CN to 50 mM CdCl2—the
Cd concentration was fixed and the Hg was varied), thiourea, and ammonia to-
gether with KOH at 80–85°C onto Ti substrates [17].
       The mole fraction of Hg in the films was ca. four times its concentration in
the solution. This was expected based on the lower-solubility product of HgS
compared to CdS. The maximum mole fraction of Hg in the films (composition
measured by atomic absorption spectroscopy) was 0.18; attempts to increase this
value by adding more HgCl2 resulted in rapid precipitation in the solution and lit-
tle film formation. Increased Hg concentration in the films could probably be ob-
tained by optimization of the conditions, e.g., by reducing the Cd concentration
and/or by using a specific complex for Hg, such as iodide. The bandgap of the
films (annealed at 320°C for 3 hr in air), measured by photoelectrochemical pho-
tocurrent spectroscopy, decreased with increasing Hg content down to 1.8 eV for
0.18 mole fraction Hg. The shape of these spectra suggested that the (annealed)
films were solid solutions, although no structural characterization was made. The
main purpose for making these films seems to have been to study their photoelec-
trochemical properties, which are described in Chapter 9.
       Triethanolamine was also used as a complexant to deposit these films from
thiourea baths [18]. As with the previous study, there was a maximum Hg content
in the bath (0.05 mole fraction—absolute concentrations were not given), which
led to a 0.18 Hg mole fraction in the films, above which, although films were
formed, the Hg content decreased, also explained by rapid precipitation of HgS in
the solution. X-ray diffraction showed the formation of a single phase, up to a Hg
content (in the bath) of 0.15, and two-phase formation at higher concentrations.
The optical bandgap dropped from 2.4 eV (pure CdS) to 1.76 eV (0.05 Hg in bath,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
0.18 in film) and then slowly increased again with increasing Hg in the bath to ca.
1.9 eV.

8.2.5 (Cd,Pb)S
Solid solutions of Cd and Pb sulphides have been popular in ternary CD. The crys-
tal structures of the individual sulphides are different: PbS crystallizes in the rock-
salt structure, while CdS forms tetrahedrally bonded sphalerite or wurtzite struc-
ture (a rocksalt form of CdS exists but normally only at high pressures). This
suggest that the solubility range of the alloys will be limited.
        Acetates of Cd and Pb were mixed with ammonia and thiourea and films de-
posited on glass at room temperature [19]. The concentration of Cd in the films
was a little higher than in the solution. The lattice constant increased with increase
in Cd but was greater than that of PbS (which has a slightly larger lattice constant
than CdS) at all levels measured. This is not typical of solid solution formation,
although it does imply a single phase. It was suggested that lattice expansion oc-
curred due to PbS entering as an interstitial into the CdS. It seems more probable
that, if interstitial expansion is the correct explanation, Cd2 , which is the lower-
concentration component, will go into interstitial sites in the PbS. The morphol-
ogy of these layers was investigated as a function of composition [20].
        In a similar study, (Cd,Pb)S was deposited on Ti at ca. 75°C [17]. The ratio
of Cd to Pb in the films was found (by atomic absorption spectroscopy) to be very
similar to that in the solution, a consequence of the similar values for the solubil-
ity product of the two sulphides. Since the main purpose for investigating these
films was to study their photoelectrochemical properties (see Chap. 9 for details),
little no characterization, other than compositional and photoelectrochemical, was
made. Photoelectrochemical spectroscopy (of films annealed at 460°C in air for 4
hr) showed a decrease in bandgap with increasing Pb content down to 1.6–1.7 eV
for Cd0.82Pb0.18S, although very nonlinearly—a strong drop in bandgap, of
0.5–0.6 eV, occurred between 0.1 and 0.18 mole fraction Pb.
        Films deposited from mixed Cd/Pb solutions complexed with ammonia (for
Cd) and hydroxide (for Pb), both in a minimum amount to effect dissolution, at pH
values between 10 and 13 and deposition temperatures between 60 and 80°C were
concluded, from consideration of the XRD, TEM (which showed two different
crystal sizes), and optical spectra, to be mixtures of the two sulphides rather than
a solid solution [21]. A study of these films for solar-selective surfaces was car-
ried out.
        A study of variation of the composition of (Cd,Pb)S films as a function of
the free-[Cd2 ]:[Pb2 ] ratio (i.e., the uncomplexed metal ions, which could be cal-
culated from the concentrations of total metal ions, concentration of complex, and
the respective stability constants of the metal complexes) showed a linear increase
in Cd content of the films up to a ratio value of 10, followed by a sudden decrease



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
in Cd content above this value (the details of the solution composition were not
given) [22]. The maximum amount of Cd that could be incorporated as a solid so-
lution with PbS was 15%. It was suggested that for [Cd2 ] greater than this, the
rate of CdS formation was so great that it became more favorable for the CdS to
form a separate phase than a solid solution.
       The crystallographic texture of the films was dependent on the Cd content.
Up to 3 at.%, the films were (111) textured, while for higher Cd concentrations
they became (200) textured. The crystal size (measured from electron mi-
croscopy) was of the order of some hundreds of nanometers (somewhat smaller
for larger Cd content) but increased again to ca. 1 m for maximum Cd content
just before phase separation.
       The resistivity of the films increased from 106 -cm for very low Cd con-
tent to a maximum of ca. 108 -cm at 6% Cd and then slowly dropped again with
increasing Cd. This maximum correlates with the minimum crystal size, suggest-
ing a dominant role of grain boundaries in the conduction mechanism. The spec-
tral response of the photoconductivity blue-shifted with increase Cd content up to
a peak response at 1.35 m for 8.4% Cd.
       The presence of cyanamides of Cd and Pb in films of (Cd,Pb)S was con-
firmed by thermal desorption mass spectrometry [23]. Cyanamide (H2CN2) is a
product of the decomposition of thiourea and forms sparingly soluble metal salts.
The metal cyanamide content of the film varied from ca. 5% up to ca. 20% (by
weight). The presence of the cyanamides decreased the intensity of the XRD re-
flections, presumably due to poorer crystallization of the sulphides. Interestingly,
the photosensitivity of the films increased with higher metal cyanamide content,
although whether this was due specifically to the presence of the cyanamide or to
its effect on the crystal growth was not known.
       The same group also deposited (Cd,Pb)S using a flow system [24]. In this
case, metal cyanamides were not detected by XRD, presumably because the
flow system removed the cyanamide. The rate of flow affected the crystal size:
Larger flow rate resulted in finer-grained deposits. Elemental analysis and XRD
showed the incorporation of Cd in the films, again up to ca. 10%, as a solid
solution.


8.2.6 (Hg,Pb)S
Films of (Hg,Pb)S were grown on glass at 30°C from a solution of PbAc2, HgCl2,
thiourea, and NaOH at pH 10 [25,26]. It was noted that the pH had to be criti-
cally controlled to obtain good-quality films. Additionally, the order of mixing of
the solutions was unusual (it was not stated if this was critical or not) in that the
PbAc2 and thiourea were first mixed, the NaOH was then added until a light brown
color appeared in the solution, and only then was the HgCl2 solution added, fol-
lowed by adjustment of pH to 10.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       For films deposited at temperatures up to 45°C, ED showed the presence of
a single phase, with increasing lattice spacing with increased Hg content up to 4%
Hg. Above this concentration, aggregates of -HgS (metacinnabar) were found in
the deposit, which originated from the colloidal solution. While not clearly de-
fined, the transmission spectra shifted to the blue (increasing bandgap) with in-
crease in Hg concentration. This increase in bandgap, together with the ED data,
suggests that the films are alloys of PbS with -HgS (cinnabar, bandgap ca. 2 eV).
For deposition temperatures greater than 45°C, the lattice parameters decreased
with increasing Hg, and the -HgS phase was formed (the lattice parameters of -
HgS are ca. 1.5% smaller than those of PbS).
       If FeCl3 was added to the HgCl2 solution, the film properties, even for those
deposited at low temperatures, were similar to those grown without FeCl3 at
higher temperatures. Single-phase films were obtained up to 50 at.% Hg with the
  -HgS phase, and the lattice spacing decreased with increasing Hg content. The
transmission spectra, while again mostly not well defined, shifted to the red with
addition of Hg. The bandgap of -HgS was taken to be ca. 0.1 eV in these stud-
ies; other values of ca. 0.5 eV have also been measured for this phase. The FeCl3
was believed to stabilize the -phase.
       A specific study of the optical and electrical properties of these Pb1 xHgxS
films was carried out with an emphasis on the difference between the - and
  -phase alloys deposited at 30°C as described earlier [27]. A linear increase in
bandgap (up to 0.9 eV for x 0.33) for the -phase alloy and a linear decrease of
the -phase down to ca. 0.18 eV for x 0.33 was measured. The resistivities of
the alloys (x 0.14) were higher than for the pure PbS (ca. 10 -cm) by a factor
of ca. 5 ( -phase) and of 10 ( -phase). Both the photoconductivity response and
the thermoelectric power of the alloys were greater than for the pure PbS. The
electrical properties were believed to be controlled mainly by intergrain barriers.
       Using the FeCl3-containing solution, epitaxial films of -Pb1 xHgxS were
grown on (111) Si or Ge single crystals, where x varied between 0 and 0.33 [28].
The conditions to obtain epitaxy were low temperature ( 20°C), relatively dilute
solution (concentrations not given, but the typical concentrations were high—
metal concentration probably several hundred mM), and relatively thin films
( 80 nm; above this thickness, -phase deposition occurred). The films on Ge
                                               
were (111) oriented, while those on Si were (112). The requirement for low tem-
perature and relatively low concentration of reactants, both of which slow the de-
position process, suggest that the epitaxy occurs if enough time is allowed for
crystal growth to occur. Decreased temperature will decrease both crystal growth
and the rate of attainment of epitaxy (probably a surface diffusion process); for
epitaxy to be preferred at lower temperatures implies that the effect of tempera-
ture on the former is greater than on the latter. The attainment of epitaxial growth
is strong evidence for an ion-by-ion mechanism, even though parallel homoge-
neous precipitation occurred in the solution.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Terminal thickness usually decreases with increase in temperature, due to
faster homogeneous precipitation in the solution in parallel with film deposition.
In this study, the terminal thickness increased with increasing temperature. This
trend also suggests that ion-by-ion growth dominates, since homogeneous precip-
itation is less likely in an ion-by-ion process than in a cluster one. Although such
homogeneous precipitation was observed, it seems likely that it occurred to a
lesser extent than it would in a cluster process.

8.2.7 (Cd,Bi)S
Films of BiMCdMS were deposited from triethanolamine/ammonia-complexed
nitrates of Cd and Bi using thiourea [29]. The films were deposited at a pH of ca.
10 on glass or Si (111) at 95°C (for 90 min), followed by cooling to room tem-
perature and continuing deposition for 24 hr. Rutherford Backscattering (RBS)
analysis showed that both Cd and Bi were incorporated into the films very roughly
in proportions similar to those present in the solution. No structural data was given
to support solid solution formation. The gradual variation of bandgap with com-
position of annealed films (450°C in argon for 2hr), from 1.65 eV (for pure Bi2S3)
to 2.43 eV (for pure CdS), suggested solid solution formation, at least for the an-
nealed films.

8.2.8 (Cu,Pb)S
Two-phase films of PbSMCuS were deposited on glass from a triethanolamine/
thiourea bath at room temperature [30]. As deposition proceeded, the films be-
came Pb-rich as Cu was depleted by more rapid formation of sulphide. The resis-
tivity of the films was 10 -cm (1 m /sq. for a film thickness of ca. 0.1 m).

8.2.9 (Cu,Bi)S (and (Cu,Sb)S)
Films were deposited from a triethanolamine/thiourea bath containing CuSO4 and
Bi(NO3)3 [31]. No compositional or structural characterization was given; there-
fore there is no evidence that this was a solid solution or even a mixture of Cu and
Bi sulphides.
       An example of solid solution formation by separate deposition of binary lay-
ers followed by annealing to interdiffuse the two layers is given for Cu3BiS3 de-
position [32]. Bi2S3 (film thickness ca. 90 nm) was deposited at room temperature
from a Bi(NO3)3/triethanolamine/thioacetamide bath onto glass slides. CuS
(300–600 nm thick) was then deposited on this film from a CuCl2/tri-
ethanolamine/ammonia/NaOH/thiourea bath at room temperature. The films were
annealed at 250°C for 1 hr. Formation of the Cu3BiS3 phase could be seen from
the XRD pattern. Measurement of precipitated powders (prepared by putting the
Bi2S3 precipitated in the first deposition in the CuS deposition solution) annealed
at 300°C showed more clearly the formation of the solid solution.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Transmission spectra of the annealed films showed an approximate bandgap
of 1.9 eV. The films were p-type semiconductors with fairly low resistance (de-
pending on annealing conditions; even the as-deposited films had a sheet resis-
tance of 7 k /sq., which dropped to ca. 100 /sq. on mild annealing at 150°C).
       In a similar manner, Cu3SbS4 was formed by depositing a layer of CuS (as
earlier) onto a previously deposited film of Sb2S3 (from a thiosulpahte bath) and
annealing in N2 at 250°C [33]. These films (typically between 0.1 and 0.3 m
thick) were highly conducting (some tens of /sq). The films were evaluated for
solar-control purposes and exhibited good IR reflectivity/low IR transmittance
with sufficient visible transmittance.

8.2.10 (Pb,Sn)Se
Films were deposited from solutions of lead and tin salts (the salts used were not
specified) with ammonium acetate, ethylenediamine, and selenourea at a pH 9
(probably at least 11) [34]. To obtain thicker films, deposition was repeated a
number of times and the films were annealed; therefore it is not known if solid so-
lution formation occurred in as-deposited films. In annealed films, Pb1 xSnxSe
solid solutions with x up to 0.11 were verified by XRD. The spectral response of
the photoconductivity of the (annealed—as-deposited films were not photosensi-
tive) films shifted from a peak at ca. 4 m (pure PbSe) to ca. 7.5 m (11% Sn),
supporting solid solution formation of the annealed films. The room-temperature,
dark resistance of the (probably annealed, but not certain) films varied from 1 to
300 k , depending on deposition conditions.

8.2.11 (Pb,Bi)S
From solutions of Bi and Pb nitrates, complexed with triethanolamine and ammo-
nia, mixed sulphides were deposited with thiourea on glass at pH values between
9.5 and 11 and at 100°C (initially) followed by slow cooling in the solution [35].
Elemental analyses showed the presence of both metals in the films. It is not clear
whether solid solution formation occurred in the as-deposited films, although the
lattice parameters did vary non-monotonically, depending on composition.

8.2.12 CuInS2 and CuInSe2
CuInS2 (and, even more, CuInSe2) are strong candidates for thin-film photovoltaic
cells. For this purpose, the chalcopyrite structure (which is an ordered lattice) is
preferred over the disordered, zincblende form. Due to the large absorption coef-
ficients of these materials, a 1- m-thick film is more than enough to absorb al-
most all the suprabandgap radiation. Somewhat thicker films are generally used,
due to problems of pinholes, which commonly occur in thinner films. A number
of methods have been used to deposit these films. Surprisingly, very few (pub-
lished) attempts have been made to deposit them by CD.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       A note of caution is necessary when dealing with these materials. It is not
trivial to distinguish between CuInS(Se)2 and some phases of CuMS(Se). Diffrac-
tion and optical properties may be similar. Elemental analysis is particularly im-
portant to verify inclusion of indium in the films and in the correct ratio. A fin-
gerprint of the chalcopyrite XRD is the presence of a weak peak at 2         17–18°,
corresponding to the (101) chalcopyrite reflection. This is often not seen, although
this could be either because the deposit is not chalcopyrite or because weak peaks
are usually not seen in nanocrystalline materials with particularly small crystal
size.

8.2.12.1 (Cu,In)S
Deposition of single-phase, chalcopyrite CuInS2 was claimed using a solution of
CuCl2, InCl3, complexed by triethanolamine and ammonia, with thiourea [36].
(Note that this claim was contested, based on the diffraction and compositional
data [37,38].) The best films were somewhat Cu rich (typically Cu1.08In1S1.5) and,
more importantly, very S deficient. The excess Cu is not surprising, considering
(1) the greater concentration of Cu in the solution compared to In and (2) the lower
solubility of CuMS compounds compared to InMS ones (see, however, the dis-
cussion on CuMInMS compound precipitation in Sec. 8.1). The fact that films
with excess In can be obtained with In:Cu ratios in solution of less than 1 suggests
that the coprecipitation is more complicated than expected based solely on solu-
bility products or even taking into consideration adsorption of In on CuMS as in-
ferred from early studies on this and similar systems described in Section 8.1 [3,4].
       Overall, it appears likely that the films contained chalcopyrite CuInS2
mixed with other phases with similar diffraction patterns. Separate microstruc-
tural characterization (EDS) of films with varying composition (ca. 10% excess
Cu or In) showed the formation of separate phases of Cu2S and In2S3, respec-
tively, along with the CuInS2 [39]. The best films were obtained at high deposi-
tion temperatures (80°C) and with stirring. Lower deposition temperature resulted
in poorer stoichiometry (less S), and stirring improved film uniformity. Grain size,
measured by TEM (which does not necessarily show crystal size) was 100–400
nm.
       From the optical transmission spectra, a bandgap of 1.50 eV was found for
the most stoichiometric (in terms of Cu:In 1; the S content was always found to
be low) films. This value dropped slightly for nonstoichiometric films [39]. Re-
sistivities varied with composition, from ca. 50 -cm for In-rich films down to ca.
0.1 -cm for very Cu-rich ones [36].

8.2.12.2 (Cu,In)Se
Three studies used essentially the same baths—ammonia and triethanolamine-
complexed Cu and In salts—and selenosulphate as Se source [40–42]. In all cases,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
a Cu:In ratio close to unity in the solution resulted in an optimum (and ca. unity)
ratio in solution. In fact, the ratio in the film was not strongly dependent on the ra-
tio in solution (at least, over the narrow range measured) [42]. This is not obvious
based on solubility product considerations and suggests some form of compound
formation, as described in the related precipitation experiments of Rudnev et al.
[3,4].
       In the earliest study, the deposition was carried out at room temperature; no
elemental analysis was made, and the diffraction data do not show the presence of
chalcopyrite CuInSe2, although the sphalerite phase might be present. The other
studies used higher temperatures, a parameter that appears to be important. More
stoichiometric films were found at 90°C than at 50°C [41], and in one case, good,
adherent films were obtained only at 85°C; at lower temperatures, powdery, non-
adherent films were obtained [42]. Both these studies reported chalcopyrite
CuInSe2 as deposited. A bandgap of 1.08 eV was calculated from the optical spec-
trum for the low-temperature deposition of Ref. 40 and one of 0.9 eV for the high-
temperature (90°C) deposition of Ref. 41. All films that were characterized for
conductivity type were found to be p-type, with resistivities that varied between
0.08 and 500 -cm, depending mainly on the excess Cu content, which resulted
in low-resistivity films [41,42]. These are low values and suggest appreciable Cu
excess, although elemental analysis showed some of the films to be close to stoi-
chiometric.
       A similar deposition, using ammonia but with citrate complexant instead of
triethanolamine and at 40°C, was also reported [43]. From XRD measurements
(not shown in the study), predominantly chalcopyrite CuInSe2 was reported if the
Cu:In:Se ratio in solution was ca. 1:1:2. EDS analysis confirmed this approximate
ratio in the films. From the absorption spectra, a bandgap of 1.4 eV was measured,
which decreased to 1.15 eV on annealing to 520°C (the literature bandgap of
CuInSe2 is ca. 1.0 eV).
       Resistivity and Hall measurements of these films as a function of composi-
tion are interesting. The resistivity increased to a sharp maximum of ca. 108 -cm
at a Cu:In ratio of 1.5; lower values of Cu:In resulted in lower-resistance p-type
films, while higher values (more Cu) gave low-resistance n-type films (Fig. 8.3).
This is unexpected in that (a) In-rich films of CuInSe2 are normally n-type, while
Cu-rich films are p-type and (b) the highest resistivity would be expected for
Cu:In 1 if the material is CuInSe2. These results, together with the higher-than-
usual bandgap of the (mainly annealed) films, suggest that the films are not sim-
ply single-phase CuInSe2, but either a mixture of phases or a different composi-
tion.
       The most recent investigation is closest to the previous one, but using only
citrate as a complexant (no ammonia), a lower pH than the other studies (8 instead
of ca. 10), and deposited at room temperature [44]. The films from this deposition
were not adherent as deposited and required annealing (300°C) to become adher-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 8.3 Resistivity vs. Cu:In ratio for CuMInMSe films (annealed at 520°C). (Adapted
from Ref. 43 with permission from Elsevier Science.)



ent. The films again appeared, from XRD, to be chalcopyrite. The XRD peaks of
the as-deposited films were rather broad (crystal size of ca. 14 nm). As with the
previous study, the bandgap of the as-deposited films was anomalously high (1.3
eV), which, considering the small crystal size, may be a quantum size effect. Af-
ter annealing, the crystal size almost doubled and the bandgap dropped to the nor-
mal CIS value of 1.02 eV. The films were p-type and highly conducting: 10 3 -
cm as deposited and ca. 2 -cm after annealing. Again, these low resistivities
suggest that free CuMSe species were present.

8.2.13 (Cd,Sn)O
Cadmium stannate (Cd2SnO4) was deposited from a solution of CdCl2 and SnCl4
using NH4F, ostensibly as a freezing agent, although it is probable that it also
functioned as a complexant (see this technique for SnO2 deposition in Chap. 7)
along with a small amount of AgNO3 as catalyst (not clear for what) and NaOH
to adjust pH to between 7.5 and 8.5 [45]. The film grew to a maximum thickness
of 0.8 m in 40 min at a pH of 7.5 (faster deposition but lower terminal thickness
at higher pH). X-ray diffraction of films annealed at 200°C or above showed them
to be Cd2SnO4 with a grain size of 25 nm (20 nm before annealing).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 8.4 Transmittance and reflectance spectra of CdMSnMO (Cd2SnO4) films: as de-
posited, annealed in H2 and in vacuum, both at ca. 200°C. (Adapted from Ref. 45.)



       The optical transmittance and reflectance spectra of the as-deposited film
and films annealed in H2 and in vacuum at ca. 200°C are shown in Figure 8.4. All
the films show high transmission in the visible/near-IR region, in particular the an-
nealed ones. A high transmittance is important if the films are to be used for their
transparent, conducting properties. The blue shift of the absorption after anneal-
ing was explained by the increase in bandgap due to a high free-carrier concen-
tration (the Moss–Burstein shift caused by filling of the lower part of the conduc-
tion band by free carriers). The increased carrier concentration after annealing is
also the cause of the shift in the reflectance spectrum to shorter wavelengths upon
annealing; annealing in H2 results in a larger free-carrier concentration due to re-
moval of more oxygen (heavier n-type doping). The bandgaps, calculated from the
transmission spectra, were all indirect and were 2.7 eV (as deposited), 3.1 eV
(vacuum-annealed), and 3.2 eV (H2-annealed).
       The resistivities of the films were 4 10 1 (as deposited), 10 2 (vacuum-
annealed), and 4 10 3 -cm (H2-annealed). The decrease in resistivity was due
mainly to increase in free-electron concentration (2 1018, 5 1019, 1020 cm 3)
for the three films; the mobility increased by a factor of two between the as-de-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
posited and H2-annealed films. The increase in electron concentration was due to
oxygen removal, as mentioned earlier. The increased mobility was possibly due to
desorption of surface-adsorbed oxygen.


8.3 MIXED CHALCOGEN COMPOUNDS
A number of mixed sulphide/hydroxides have been deposited, mainly in the
search for improved window layers for photovoltaic cells (Chap. 9). These are
mostly probably mixed-phase films, although in one case, In(OH)S, experimental
evidence suggests true compound formation [46]. Most of these films have been
dealt with in previous chapters (see Chap. 4 under ZnS and Chap. 6 under In and
Sn sulphides). One study (described from the viewpoint of its properties in pho-
tovoltaic cells in Chap. 9) has not been described previously and will be men-
tioned briefly here. This deals with Zn(O,OH) and Zn(O,OH,S) deposited from
Zn-ammine solutions, the latter film from solutions also containing thiourea [47].
It is of interest to note that the Zn(O,OH) films did not deposit on glass but did on
both ZnO- and CuInSe2-type substrates. Even after annealing at 300°C, hydrox-
ide groups were still present in those films.

8.3.1 Cd(S,Se)
In two of the studies made of Cd(S,Se) deposition [48,49], the solutions and con-
ditions of deposition were similar. In both cases, Cd was complexed with ammo-
nia, a mixture of thiourea and selenosulphate was used, and the deposition was
carried out on glass at 75°C.
       From the study of Kainthla et al. [48], XRD of the films showed clearly that
solid solution formation occurred; the (predominantly sphalerite) diffraction
peaks shifted with change in composition. For compositions with S concentration
   60%, only zincblende structure formed; the amount of wurtzite increased with
increasing S content but was always low. The concentration of S in the films was
somewhat greater than that in the deposition solution; i.e., S deposited preferen-
tially. This is not surprising since CdS deposition is normally faster than that of
CdSe. The concentration of ammonia was increased as the thiourea:selenosul-
phate ratio increased, ostensibly to slow down the rate of formation of CdS
through decreased Cd2 concentration (although the rate of CdSe formation is
also dependent on this same factor).
       Optical spectra showed a gradual shift of the onset with composition, as ex-
pected for a true solid solution. Additionally, the refractive index (and therefore
the dielectric constant) increased gradually as the Se content increased, mirroring
the larger dielectric constant of CdSe compared with CdS.
       The study of Ref. 49 gave similar results, although XRD peaks of the films
were very weak and difficult to interpret. In addition, the electrical resistivity of



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the films was measured. The resistance of the S-rich and slightly Se-rich films was
typically 1–3 106 -cm, while for high Se concentrations the resistivity fell to
a value of 7 104 -cm for pure CdSe.
       Another study of Cd(S,Se) deposition was similar to the preceding ones,
with the differences that triethanolamine was used instead of ammonia (solution
pH 10.4) and the deposition was carried out at 55°C [50]. The main difference
in the films was that a true single-phase solid solution formed over only part of the
composition range (for values of x in CdSexS1 x between 0 and 0.4 and between
0.85 and 1). This was paralleled by sharp absorption onsets for the solid solutions
and more gradual ones for the mixed-phase systems. Crystal sizes varied in a non-
monotonic manner from 27 nm (CdS) to 9 nm (CdSe). The resistivities were of the
same order (for CdS, slightly less) as those of the previous study, although the
variation of resistivity with composition was somewhat different. However, since
the variation in resistivity over the entire composition range was only a little more
than an order of magnitude, which is not particularly large, such differences need
not be very meaningful.
       The same group also studied In-doped (by addition of InCl3) CdS1 xSex
[51]. As the In concentration increased, the degree of crystallinity (measured by
the height of the XRD peaks) and crystal size increased, reached a peak, and
then decreased. These structural changes correlated with other properties: The
bandgap and resistivity were minimum when the crystallinity and crystal size
were maximum. The concentration of In in the films was much higher than that
in solution ( 2% of the Cd concentration, compared to 0.1% in the solution)
[52].

8.3.2 Zn(S,Se)
Zn(S,Se) has been deposited on both glass and on single-crystal GaAs (110) from
a hydroxide-complexed solution of Zn2 using, as for Cd(S,Se), a mixture of
thiourea and selenosulphate [53,54]. Apparently conditions were chosen to give
the composition ZnS0.056Se0.944 because of its perfect lattice match with the GaAs
substrate. The composition did not appear to be dependent on the deposition tem-
perature.
       Room-temperature deposition resulted in films with very broad peaks,
which sharpened considerably with increasing deposition temperature to give a
crystal size of ca. 20 nm at a deposition temperature of 90°C. The high-tempera-
ture films on GaAs exhibited a fairly high degree of epitaxy, as seen by the spots
in the electron diffraction pattern.

8.3.3 Pb(S,Se)
Pb(S,Se) was deposited from a hydroxide-complexed solution of Pb(NO3)2 us-
ing, as before, a mixture of thiourea and selenosulphate [55]. The films were



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
confirmed to be single phase by XRD (Debye–Scherrer photographs). In con-
trast to the case for Cd and Zn, Se was preferentially deposited. (Small seleno-
sulphate concentrations in solution resulted in much larger Se concentrations in
the film. For example, from a solution containing 0.5 mM selenosulphate and
500 mM thiourea, the film composition was ca. PbS0.75Se0.25.) This was ex-
plained by the large difference in solubility products of PbS and PbSe (nearly 10
orders of magnitude; see Table 1.1). For Cd and Zn, this difference is some or-
ders of magnitude less. Thus, while for Cd and Zn it seems that the faster de-
composition of thiourea compared to selenosulphate more than compensates for
the lower-solubility products of the selenides, for Pb the difference in solubility
products between sulphide and selenide becomes the main composition-deter-
mining factor.
       From XRD line widths, it was noted that the crystal size increased as the S
concentration increased, although values were not given. A simple but useful
characterization technique to quantify film thickness homogeneity was used here.
The transmittance of a focused light spot scanned across the sample showed ex-
cellent homogeneity.

8.3.4 Bi2(S,Se)3
In this case, thioacetamide was used as the sulphur source, instead of thiourea as
for the previous mixed sulphides-selenides (selenosulphate, as before, was used as
the Se source) [56]. Bi(NO3)3 was complexed with triethanolamine and the pH ad-
justed with ammonia to 8.2. The deposition was carried out at 55°C. The compo-
sition was varied by varying the thioacetamide/selenosulphate ratio. Although it
is not clear what the elemental compositions of the various films were, from the
limited XRD data given, it seems that solid solution did occur. The crystal sizes
increased from 6 nm (pure sulphide) to 13 nm (pure selenide), and bandgap val-
ues decreased over the same range from ca. 1.9 to 1.0 eV.


8.4 BILAYERS OF DIFFERENT SEMICONDUCTORS
This chapter has dealt with true ternary compounds, with the underlying implica-
tion that deposition of separate phases is undesirable. However, it needs to be
stressed that what is undesirable for one purpose may be preferable for another
(examples being small crystal size and a large amount of scattering). So, too, a
composite of different phases may be the goal of a particular deposition. This is-
sue does not appear to have been dealt with in CD.
      Taking the principle of separation of phases one step further, separate lay-
ers may also be deposited, one on top of another. This has been done in a number
of cases and should present no problem (taking into consideration that there may
be some cases where deposition of the second layer will destroy or change in some



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
way the first one). There are, however, two examples in the literature of deposi-
tion of a bilayer from a single deposition solution.
       In the first [57], a bilayer of CdS/ZnS was formed using the electrochemi-
cally assisted technique described for CdS (Sec. 4.1.6.6). In this technique, a mix-
ture of CdCl2 and ZnCl2 was mixed with thioacetamide at a pH of 2.45 (added
HCl) and deposited on ITO/glass at 70°C. In the absence of electrochemical po-
larization, only ZnS, with a low percentage of CdS, deposited after an induction
period. CdS does not deposit by itself under these acidic conditions. When sub-
jected to cathodic polarization ( 0.65 V vs. S.C.E.), CdS preferentially deposited
due to local increase in pH at the cathode (by water electrolysis), while eventually,
as the Cd concentration dropped, ZnS formation became more favorable. Depth
analysis of the films showed that Zn was formed preferentially at the surface while
CdS formed preferentially near the substrate. This order is expected, considering
the lower-solubility product of CdS.
       The second example, which was a pure CD process, produced ZnO on top
of CdS [58]. The principle is based on the facts that CdS deposits much more read-
ily than does ZnS (see Sec. 4.4.1) and that ZnO (or Zn(OH)2, which readily con-
verts to ZnO) tends to deposit readily, more so than ZnS unless under conditions
of high active sulphur concentration (whether sulphide ion or sulphur-containing
complex) and low pH. The solution contained Cd and Zn ions (the latter in excess)
complexed with ethanolamine and ammonia (therefore at least fairly high pH) and
thiourea. Cross-sectional microprobe analysis showed that the film contained the
more readily deposited CdS at the substrate (ca. 0.3 m thick), covered with a 2-
  m ZnO layer, which formed as the solution conditions (probably mainly the drop
in Cd concentration) favored the ZnO deposition. The films were pale yellow, and
optical spectroscopy showed two transitions—one at the CdS bandgap (ca. 2.6 eV,
greater than the bulk bandgap, suggesting that the CdS crystal size was small—ca.
4–5 nm) and the other corresponding to ZnO at 3.2 eV. The films were photosen-
sitive, about an order of magnitude more sensitive than ZnO deposited by itself.
       A different form of bilayer can be formed using topotactic exchange reac-
tions. This type of exchange is well known, e.g., for the conversion of CdS into
Cu2S by immersing in a hot CuCl solution, used in the past for fabricating
CdS/Cu2S photovoltaic cells (see Sec. 9.1.2). It has been used more recently to
convert CD films of one semiconductor into another, e.g., CdS and CdSe into
Ag2S and Ag2Se [59] and SnS2 into Ag2S [60]. While these studies describe con-
version of one semiconductor into another, it is clear that, if carried out in a con-
trolled fashion, partial exchange can occur, leading to the expected formation of a
shell of the exchanged semiconductor around a core of the original semiconduc-
tor for each individual crystal in the film (assuming the film to be at least some-
what porous, as it invariably will be). This process therefore can lead to films of
core/shell nanocrystals.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
8.5 CONCLUSION
It is fair to state that the understanding of deposition of ternary compounds lags
behind that of binaries. A better understanding of the factors that control codepo-
sition, as well as solid solution formation, is needed. However, it is also clear that
there is scope for deposition of a wide range of compounds, not only ternaries, but
quaternaries and even higher-multinary materials. Additionally, the scope for de-
position of mixed-phase films, either as consecutive layers (as shown earlier) or
as composites, is great, and this aspect of CD will undoubtedly be pursued.


REFERENCES
 1.   CS Brooks. Metal Finishing 88:21, 1990.
 2.   IM Kolthoff, DR Moltzau. Chem. Rev. 17:293, 1935.
 3.   NA Rudnev, GI Malofeyeva. Talanta 11:531, 1964.
 4.   NA Rudnev, IV Melikhov, AM Tuzova. Zh. Anal. Khim. 28:635, 1973.
 5.   KL Chopra, RC Kainthla, DK Pandya, AP Thakoor. In: Physics of Thin Films. Vol.
      12, Academic Press, New York and London, 1982, p 167.
 6.   G Contreras-Puente, O Vigil, M Ortega-Lopez, A Morales-Acevedo, J Vidal, ML Al-
      bor-Aguilera. Thin Solid Films 361:378, 2000.
 7.   JM Doña, J Herrero. Thin Solid Films 268:5, 1995.
 8.   NC Sharma, RC Kainthla, DK Pandya, KL Chopra. Thin Solid Films 60:55, 1979.
 9.   T Nakazawa, S Kuranouchi, A Ashida, N Yamamoto. In: 12th EC Photovolt. Solar
      Energy Conf. Amsterdam, the Netherlands, 1994, p 601.
10.   GK Padam, GL Malhotra, SUM Rao. J. Appl. Phys. 63:770, 1988.
11.   SA Al Kuhaimi, Z Tulbah. J. Electrochem. Soc. 147:214, 2000.
12.   T Yamaguchi, Y Yamamoto, T Tanaka, Y Demizu, A Yoshida. Thin Solid Films
      281–282:375, 1996.
13.   T Yamaguchi, Y Yamamoto, T Tanaka, A Yoshida. Thin Solid Films 344:516, 1999.
14.   KC Sharma, JC Garg. J. Phys. D: Appl. Phys. 23:1411, 1990.
15.   DS Sutrave, GS Shahane, VB Patil, LP Deshmukh. Mater. Chem. Phys. 65:298, 2000.
16.   ES Rittner, JH Schulman. J. Phys. Chem. 47:537, 1943.
17.   M Skyllas-Kazacos, JF McCann, R Arruzza. Appl. Surf. Sci. 22/23:1091, 1985.
18.   LP Deshmukh, KM Garadkar, DS Sutrave. Mater. Chem. Phys. 55:30, 1998.
19.   BB Nayak, HN Acharya, GB Mitra. Bull. Mater. Sci. 3:317, 1981.
20.   BB Nayak, HN Acharya. J. Mater. Sci. Lett. 4:651, 1985.
21.   GB Reddy, DK Pandya, KL Chopra. Sol. Energy Mater. 15:383, 1987.
22.   GA Kitaev, VF Markov, LN Maskaeva, LE Vasyunina, IV Shilova. Inorg. Mater.
      26:202, 1990.
23.   GA Kitaev, LN Maskaeva, VF Markov, AY Kurkin, LE Vasyunina. Inorg. Mater.
      25:1065, 1989.
24.   LN Maskaeva, VF Markov, GA Kitaev. Russ. J. Appl. Chem. 73:751, 2000.
25.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Mat. Res. Bull. 11:1109, 1976.
26.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 42:383, 1977.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
27.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 62:97, 1979.
28.   NC Sharma, DK Pandya, HK Sehgal, KL Chopra. Thin Solid Films 59:157, 1979.
29.   S Misra, HC Padhi. J. Appl. Phys. 75:4576, 1994.
30.   R Suarez, PK Nair. J. Solid State Chem. 123:296, 1996.
31.   LP Deshmukh, DS Sutrave, BM More, CB Rotti, KM Garadkar. Semicond. Devices
      421, 1996.
32.   PK Nair, L Huang, MTS Nair, H Hu, EA Meyers, RA Zingaro. J. Mater. Res. 12:651,
      1997.
33.   MTS Nair, Y Pena, J Campos, VM García, PK Nair. J. Electrochem. Soc. 145:2113,
      1998.
34.   VM Markov, LN Maskaeva, LD Loshkareva, SN Uimin, GA Kitaev. Inorg. Mater.
      33:555, 1997.
35.   N Parhi, BB Nayak, BS Acharya. Thin Solid Films 254:47, 1995.
36.   GK Padam, SUM Rao. Sol. Energy Mater. 13:297, 1986.
37.   D Cahen. Sol. Energy Mater. 15:225, 1987.
38.   GK Padam, SUM Rao. Sol. Energy Mater. 15:227, 1987.
39.   GK Padam, GL Malhotra, SUM Rao. Phys. Status Solidi (a): 109:K45, 1988.
40.   RN Bhattacharya. J. Electrochem. Soc. 130:2040, 1983.
41.   GK Padam. Mat. Res. Bull. 22:789, 1987.
42.   KR Murali. Thin Solid Films 167:L19, 1988.
43.   JC Garg, RP Sharma, KC Sharma. Thin Solid Films 164:269, 1988.
44.   PK Vidyadharan Pillai, KP Vijayakumar, PS Mukherjee. J. Mater. Sci. Lett. 13:1725,
      1994.
45.   D Raviendra, JK Sharma. J. Appl. Phys. 58:838, 1985.
46.   D Hariskos, M Ruckh, U Ruhle, T Walter, HW Schock, J Hedstrom, L Stolt. Sol. En-
      ergy Mater. Sol. Cells 41–2:345, 1996.
47.   K Kushiya, T Nii, I Sugiyama, Y Sato, Y Inamori, H Takeshita. Jpn. J. Appl. Phys.
      35:4383, 1996.
48.   RC Kainthla, DK Pandya, KL Chopra. J. Electrochem. Soc. 129:99, 1982.
49.   RS Mane, CD Lokhande. Thin Solid Films 304:56, 1997.
50.   GS Shahane, BM More, CB Rotti, LP Deshmukh. Mater. Chem. Phys. 47:263, 1997.
51.   GS Shahane, KM Garadkar, LP Deshmukh. Mater. Chem. Phys. 51:246, 1997.
52.   GS Shahane, LP Deshmukh. Mater. Chem. Phys. 70:112, 2001.
53.   GN Chaudhari, S Manorama, VJ Rao. Thin Solid Films 208:243, 1992.
54.   GN Chaudhari, SN Sardesai, SD Sathaye, VJ Rao. J. Mater Sci. 27:4647, 1992.
55.   YS Sarma, HN Acharya, NK Misra. Thin Solid Films 90:L43, 1982.
56.   AR Patil, VN Patil, PN Bhosale, LP Deshmukh. Mater. Chem. Phys. 65:266, 2000.
57.   K Yamaguchi, T Yoshida, T Sugiura, H Minoura. J. Mater. Res. 13:917, 1998.
58.   SJ Castillo, M Sotelo-Lerma, RA Zingaro, R Ramirez-Bon, FJ Espinoza-Beltran, R
      Guillemette, MA Dominguez. J. Phys. Chem. Solids 62:1069, 2001.
59.   CD Lokhande, KM Gadave. Mater. Chem. Phys. 36:119, 1993.
60.   CD Lokhande, VV Bhad, SS Dhumure. J. Phys. D: Appl. Phys. 25:315, 1992.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
9
Photovoltaic and
Photoelectrochemical
Properties




A large number of studies on CD have been driven by two related potential uses:
photoelectrochemical (PEC) cells, mostly the earlier studies, and, more recently,
photovoltaic (PV) cells. This chapter is devoted to these two topics where CD
films have been used.

9.1 PHOTOVOLTAIC CELLS
9.1.1 Introduction
Since almost all thin-film (CdTe- and CuInSe2-type) cells today use CD films
(mostly CdS), there is no attempt here to be comprehensive regarding the litera-
ture. Rather, studies that emphasize the CD film itself are discussed. Before dis-
cussing the role of the CD layer specifically, a brief overview of the relevant cells
will be given.
       There are three main thin-film PV cells under development at present: amor-
phous Si, CdTe/CdS, and CI(G)S/CdS [CI(G)S refers to copper indium (gallium)
selenide]. Of these, the last two are polycrystalline (as opposed to amorphous),
and both normally employ CD CdS. Crystalline Si cells are not thin films, being
at least tens and usually hundreds of microns in thickness, compared to the few
microns of active thickness of the thin-film cells.
       Schematic diagrams of the CdTe and CI(G)S cells are shown in Figure 9.1.
The main difference in their construction is that the CdTe cell is a superstrate



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
(backwall) cell (illuminated through the conducting glass substrate), while the
CI(G)S cell is a substrate (frontwall) cell (illuminated through the front surface).
The CI(G)S cell is a development of the original CuInSe2 /CdS (CIS/CdS) cell,
with Ga added to increase the bandgap. Pure CuGaSe2 /CdS has also been investi-
gated, although considerably less than CI(G)S, due to its (at present) considerably
lower conversion efficiency. In the following, CIS will be used where CuInSe2 is
intended, while CI(G)S refers to CuInxGa1 xSe2, with 1 x typically 0.2 0.1.
Some studies have also been made on CuInS2, which has a higher bandgap than
CIS and in principle should give a better cell (in practice it is inferior, although
somewhat better than CGS).
       In both cells, the absorber layer (CdTe or CI(G)S) is a few microns thick,
while the CD CdS (or other CD layer) is typically 50–100 nm thick. The CD layer
is often called the “buffer” layer, a term that serves to show the lack of under-
standing of its role. Nominally, the CdS is the n-type part of the p-n junction. The
basic mode of action of a p-n PV cell is shown in Figure 9.2. The short-circuit cur-
rent (denoted in this chapter as ISC) is the current flowing in the illuminated cell
when the two sides (terminals) of the cell are shorted (full band bending, left fig-
ure) while the open-circuit voltage (denoted as VOC) is the voltage generated be-
tween the terminals at open circuit when no external current flows (right figure).
The absorber CdTe and CI(G)S are always p-type in these cells; n-type absorbers
have been little investigated, mainly because suitable high-bandgap p-type mate-
rials are not readily available. A high bandgap of the buffer layer is necessary be-
cause light passes through this layer on the way to the absorber, and some of the
light absorbed in the buffer layer is lost for current generation. The superiority of
CD CdS over evaporated CdS in both types of cell suggests that something other
than loss of photons absorbed in the buffer layer is involved. The factors believed
to contribute to this superiority, in particular the effects of the deposition solution
on the absorber, are discussed later.




FIG. 9.1 Schematic cross section of CI(G)S/CdS and CdTe/CdS PV cells.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 9.2 Band diagrams of p-n cell in the dark (or under illumination at short circuit)
and under illumination at open circuit.



       The foregoing explanation of the operation of these cells, while very basic,
will be almost sufficient for our purposes (there are many sources explaining the
mode of action of PV cells in more detail). One other process, which plays an im-
portant role in PV cells in general, should be described: electron–hole recombina-
tion. This is central to PV cell operation. Photogenerated electrons and holes are
ideally separated at the p-n junction and flow in opposite directions to give an ex-
ternal current. However, there are many pitfalls awaiting these charges on their
way to the terminals where at least one of them can be extracted. These pitfalls,
which cause the electrons and holes to recombine before external current flows,
are various forms of recombination centers. They can occur at the interface be-
tween the p- and n-type semiconductors, at grain boundaries, or in the bulk of the
semiconductor crystals. A major part of PV cell research is devoted to minimiz-
ing such recombination centers.

9.1.2 CuxS/CdS Cells
Before describing studies on the CIS and CdTe cells, there are two CD-related pa-
pers on the Cu2S/CdS cell, which was intensively investigated around 20 years
ago and was eventually abandoned because of perceived insoluble stability issues,
a perception that, it should be noted, while widely held, is not undisputed. Should
this cell make a comeback, CD is likely to be a method that will be considered for
either of the two semiconductors or even for both.
       One study utilized CD CdS, built up from several layers (probably a total
thickness of the order of a micron) and annealed [1]. The CuxS layer was formed
by the usual (for this type of cell) topotactic reaction between a CuCl solution and
the CdS substrate. While the cell performance varied over a wide range, depend-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ing on the CD process, the maximum efficiency obtained was 0.13%; all parame-
ters of the cell were very poor.
       In the other study, CuxS was chemically deposited on (presumably evapo-
rated) CdS films from a triethanolamine/ammonia/thiourea bath (see Chap. 6,
copper sulphides) [2]. Very low currents and poor fill factor were obtained, al-
though the VOC was reasonable (ca. 0.5 V), with an efficiency of ca. 0.5%. The sto-
ichiometry of the Cu-S was not ideal for PV cell use, although this could be var-
ied to an extent by electrochemical treatment.
       In view of the sensitivity of the Cu2S/CdS cell to the nature and phase of the
Cu-S, it is likely that much better performance can be obtained if an effort is made
to do so.


9.1.3 CdTe/CdS Cells
While CD CdS is commonly used for CdTe-absorber cells, there is relatively lit-
tle work that emphasizes the CdS film.
       A 13.4% efficient cell fabricated by close space sublimation of CdTe on CD
CdS was reported in 1991 [3], followed by a 14.5% cell a year later by the same
group [4]. The CdS thickness was between 50 and 150 nm. The cells were illumi-
nated through the tin oxide/glass, which was used as the substrate for the CdS de-
position, and this geometry has been used ever since for these cells.
       The most comprehensive study of the effect of the CdS deposition parame-
ters on the resulting CdTe/CdS appears to have been made for electrodeposited
CdTe [5]. The most simple variable is CdS film thickness. Clearly a minimum
thickness is required for junction formation and to prevent shunting from the
CdTe through the CdS to the substrate (usually conducting glass). On the other
hand, an increase in this thickness leads to a decrease in ISC due to light absorp-
tion in the CdS, which is clearly seen as a decrease in the short-wavelength re-
sponse of the cell. The optimum CdS thickness was found to be ca. 70 nm, al-
though good cells were also made with more than twice this thickness.
       Other deposition parameters affected mainly VOC and fill factor rather than
ISC. These included an increase in thiourea concentration and the use of buffered
(ammonium ion, lower pH) solutions; both these factors resulted in higher S:Cd
ratios, therefore more stoichiometric films (CD CdS films are often Cd-rich; this
does not necessarily mean n-type doped but is more likely due to the presence of
other Cd species, e.g., Cd(OH)2). The use of chloride as an anion in solution rather
than sulphate also gave better cells. It was believed that all these factors influenced
the nature of both the CdS and the CdTe films after annealing the cell. Specifi-
cally, it was thought that small grain size and high defect density (the CdS was be-
lieved to be polytype with a large density of stacking faults [6]) in the cubic CdS
film was beneficial for the resulting recrystallization process and for intermixing
between the CdS and CdTe during the recrystallization and phase change (to



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
hexagonal phase) of the CdS during the annealing step. In this respect, it was noted
that if the CdS was recrystallized prior to CdTe deposition (and the cell then re-
annealed as usual), the resulting cells were very inferior in all output parameters
to normal cells.
       The VOC of electrodeposited CdTe on CD CdS cells was studied as a func-
tion of the CdS deposition parameters [7]. While there were a number of different
variables involved, it was clear that conditions leading to thicker films (ca. 180
nm), such as lower pH, high thiourea:Cd ratio (or possibly higher thiourea con-
centration), and repeated deposition of the films, resulted in the highest voltage
(ca. 0.7 V). The short-wavelength response was poorer for thicker films, not only
due to absorption in the CdS, but also at wavelengths longer than the CdS ab-
sorption onset, suggesting recombination at interface states.
       A clear effect of CD CdS on heterojunction formation has been shown for
                                                                       ¯¯¯
CdTe that was electrodeposited onto CdS films on single crystal (111) InP [8].
                                           ¯¯¯
CdTe electrodeposited directly onto the (111) InP shows some degree of epitaxy
but also considerable polycrystallinity (the latter not surprising, considering the
9.5% lattice mismatch between the two materials). If, however, a thin (20–30 nm)
CdS film was chemically deposited onto the same InP surface, the epitaxy of CdTe
electrodeposited onto the CdS/InP was found to be very good. The CdS was
largely epitaxial with the InP (the lattice parameters of CdS and InP are very
close), with ca. 15% polycrystallinity. Interestingly, the CdTe deposited on the
CdS exhibited an even higher degree of epitaxy than that of the CdS itself, show-
ing that the small but appreciable amount of polycrystalline CdS did not substan-
tially degrade the epitaxy of the CdTe. It was suggested that the improvement in
epitaxy due to the CdS was caused by a graded interface. Similarly, an XRD com-
parison of CdTe electrodeposited onto SnO2/glass (activated by a cathodic treat-
ment) or onto CdS chemically deposited on SnO2/glass showed better crystallinity
(narrower XRD peak) for the CdTe deposited on the CdS and also better texture
(only the (111) reflection was seen, while a small additional (220) peak was evi-
dent for the CdTe deposited directly on the SnO2) [9].

9.1.4 CdS/CI(G)S Cells
9.1.4.1      General Considerations
A considerably greater body of work with more emphasis on the CD buffer layer
exists for this cell. Much of this involves the specific effects of the CD process at
the interface, and this will be discussed in a later section.
       An experimental measurement of the band lineup between the CdS and so-
lar-grade polycrystalline CIGS has been made using contact potential difference
(Kelvin probe) measurements in air [10]. This lineup is shown in Figure 9.3. In
particular, it shows that no spike was found in the conduction band. The presence
of such a spike (believed to occur from previous studies either on single crystals



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 9.3 Band offset between CIGS and CdS. (After Ref. 10.)



or on polycrystalline films in vacuum) would constitute a barrier to electron flow
over the interface.
      From the practical point of view, reuse of the deposition solution after fil-
tering out the precipitated CdS and addition of fresh reagent was shown to have
no effect on the device properties of CdS/CIGS cells (see Sec. 4.1.6.10 for more
details of the deposition) [11].
9.1.4.2 Chemical Reaction and Diffusion at the
CdS/CI(G)S Interface
There are a number of studies on the effects of the CD process on the surface of
the CIS or CIGS. In many of these studies, the absorber surface was treated with
partial CD solutions, in particular, ammonia or ammonia Cd2 [12–19]. There
are several reactions that occur during these treatments, and these will be dis-
cussed in general before specific results from the different studies are treated.
       Aqueous ammonia removes surface oxides from the CI(G)S, in particular,
indium (possibly due to the tendency of the CIS surface to be enriched in In-O).
In this respect, it can be considered as an etchant, although the etching is limited
to the near surface region and does not continue (although in one case, it was
found that, if oxygen is present, etching can continue, presumably due to contin-
ual oxidation of the freshly exposed surface [13]).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Cd2 dissolved in ammonia has a particularly strong effect on the absorber
surface. The effect of ammonia Cd with respect to the In and Cu concentrations
(decrease in Cu) is opposite to that of ammonia by itself (decrease of In), and it
was noted that the surface reaction with all components present in the CD bath
would depend on the relative kinetics of the various partial reactions. Probably
more important, however, is that Cd rapidly substitutes for Cu for between several
nanometers and 20 nm into the absorber. This has been found to occur on single
crystals of CIS [15,16] as well as on polycrystalline films, and therefore is not sim-
ply grain boundary diffusion, as might at first be suspected. In contrast to these re-
sults, one study found no Cd substitution for Cu in CIS single crystals and related
this to the supposed absence of a Cu-deficient layer in the single crystals, com-
pared to films [19]. It is not clear whether Cd indiffusion promotes Cu outdiffu-
sion or vice versa. It should be noted that Cu is readily complexed by the ammo-
nia at the surface and therefore is easily removed. Also, the ionic radii of Cd 2 and
Cu are almost identical, facilitating exchange. It is believed that the indiffused
Cd type converts CIS to n-type and that the junction is a buried n-p one rather than
located at the CdS/CIS interface. There is also the question of how abrupt the junc-
tion is; the Cu-poor surface region has been considered a separate phase—
CuIn3Se5—known as an ordered vacancy compound (OVC). Exchange of Cu and
Cd was shown to be easier for single-crystal CuIn3Se5 than for CIS [16].
       The effect of this Cd/NH3 treatment on the PV properties are very marked.
While cells fabricated without a buffer layer [ZnO sputtered directly on the
CI(G)S] are very poor, with all parameters very low, the same cells, but subjected
to the Cd/NH3 treatment before ZnO deposition, are very much better, and in fact
the efficiencies are only a little lower than CD CdS cells, due to lower VOC (ISC is
actually often higher due to the better blue response in the absence of CdS). This
is a particularly important result since it shows that the main role of the buffer
layer is not related to the specific properties of the CdS itself, but rather to near-
surface modification of the CI(G)S. Substitution of Zn for Cd in the Cd/NH3 treat-
ment gave comparable results [15]. This is in contrast to the use of CD ZnS, which
was inferior to that of CdS, although not necessarily by much (see Section
9.1.4.5).
       However, the presence of the CD CdS is still required in order to obtain op-
timal efficiency, and therefore the CdS itself does play some role, possibly to pre-
vent sputter damage to the absorber (although Cd/NH3-treated absorbers, which
then have evaporated CdS deposited, are still not as good as CD CdS cells). The
conformal coverage of the irregular absorber surface by the CD CdS is another
factor often invoked. Incomplete coverage (e.g., pinholes) could lead to some
shunting between the ZnO and CI(G)S, either due to the lack of CdS at the pin-
holes and/or because of sputter damage at the pinholes. Also in connection with
the nature of the CdS itself, studies on the effects of CdS thickness and impurity
content on CdS/CIGS cell parameters have been made [20]. The impurity content



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
was controlled by the concentration of thiourea in the bath (see Sect. 4.1.7 for de-
tails) while maintaining the CdS thickness constant. After a very thin CdS film
had been deposited, the main effect of increasing both thickness and impurity con-
centration was to increase the VOC. This suggests that these effects are related to
the “bulk” properties of the CdS rather than to effects of the different solutions on
the absorber surface. It should be kept in mind, however, that the predeposited ul-
trathin layer either may not have totally covered the CIGS or may have been
porous enough to allow contact between the second deposition solution and the
CIGS. In connection with impurity considerations, a general property of CD CdS
is its relatively large oxygen content. While it is not clear whether this is of any
importance for its use in PV cells, it is often studied in this respect. The subject of
oxygen impurities in CdS is treated in some detail in Section 4.1.7.
        Photoluminescence (PL) measurements to monitor the changes due to the
Cd/NH3 treatment of CIGS have been carried out by two groups. Both show ma-
jor, though very different, effects of the treatment on the PL spectra. In one case
[14], two shallow subbandgap peaks, attributed to donor–acceptor pair recombi-
nation, were found in the nontreated CIGS. After treatment, the higher-energy peak
was quenched and a strong new, lower-energy peak appeared. This could be inter-
preted as the removal of states that caused interface recombination, although the
nature of the strong lower-energy luminescence was not understood. Notably, sim-
ilar behavior was obtained when CdS was deposited, providing further evidence
for the dominating role of the Cd/NH3 treatment in the CdS buffer layer formation.
        In the second study, the bandgap luminescence was found to increase 15
times after the Cd/NH3 treatment [16]. This could be interpreted as a passivation
of recombination centers. The luminescence from CD CdS/CIGS was stronger (by
ca. two times) than from CdS evaporated on Cd/NH3-treated CIGS and much
stronger (nearly 20 times) than from CdS evaporated on nontreated CIGS. From
these results, it was believed that both the Cd/NH3 treatment and the CdS deposi-
tion were important, although, again, it appears that the major effect is from the
Cd/NH3 treatment.
        On an In-rich CIS surface (one from which good-quality cells could be
made), the change in surface composition was followed upon deposition of CdS
from a complete bath [12]. An important observation was that initially, while Cd
was found at the surface, the S concentration was low, and stoichiometric CdS
only formed later in the deposition. It was stressed, however, that since the depo-
sition was begun at room temperature and then gradually heated to 60°C, it might
be that the Cd exchange process occurred preferentially at low temperature, where
CdS deposition was still very slow, while the thiourea decomposition, which was
strongly temperature dependent, occurred more readily as the deposition temper-
ature increased. The initial Cd species were believed to act as nucleation centers
for CdS formation. This heating regime also slowed down precipitation in solu-
tion and resulted in a lower terminal thickness (ca. 75 nm compared to 100 nm



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
for a preheated solution [13]), which would make the deposition time to obtain the
required thickness (ca. 50 nm) less critical.
       An investigation of CdS deposition on CIGS (In Ga rich—good PV cell
quality) substrates was carried out using two different baths: a “standard” bath and
a low-thiourea (3 mM) one [17]. Cd was preferentially incorporated into (onto) the
substrate compared to S in the initial stages of the deposition, in agreement with
the previous study. In contrast to the previous study was the conclusion that no
major preferential removal of any of the substrate constituents occurred (In, Ga,
and Se native oxides were removed). In addition, no clear evidence of compound
formation between the Cd and substrate was found, although it is clear that Cd was
incorporated into the substrate. Although the low-thiourea bath resulted in CdS
with lower impurity (mainly N compounds) levels and more Cd incorporation into
the CIGS, no improvement in cell performance resulted from such baths. While
the nature of the Cd incorporation in the early stages of the deposition could not
be unambiguously defined, it was suggested that this Cd was in the form of
Cd(OH)2, which converted to CdS, either as the thiourea slowly hydrolyzed to sul-
phide or as decomposition of a Cd(OH)2–thiourea complex occurred, both of
which are very temperature dependent.
       An explanation for the beneficial effect of the CD process on CI(G)S cells
was suggested based on the known effects of oxygen treatment on these cells [21].
Annealing in oxygen removes Se vacancies, which in turn decreases recombina-
tion at grain boundaries, surfaces and at the CdS/CI(G)S interface. In opposition
to this beneficial effect, the oxygen treatment also has been shown to reduce pos-
itive charge, and therefore band bending at the interface, and to increase Cu dif-
fusion into the bulk of the CI(G)S, thus reducing the acceptor doping; both these
effects are detrimental to cell performance. The CD deposition was believed to re-
store the positive charge to the interface (with which the deposition solution was
in contact) but not to the grain boundaries (where it did not reach) by creation of
Cd on Cu vacancies (CdCu) and possibly also removal of oxygen on Se vacancies
(OSe). This removes the detrimental effect of the oxygen treatment at the interface
but not the beneficial effect at the grain boundaries.

9.1.4.3     Epitaxy of CdS on CIS
The surface cleaning of the CIS also affected the mode of deposition of the CdS.
The CdS was found to grow to a greater or lesser extent of epitaxy on single-crys-
tal (heteroepitaxial layer) CIS [22]. Very good epitaxy of cubic CdS was found for
cyanide-treated CIS; somewhat lower epitaxy was found for ammonia-treated sur-
faces and poorer epitaxy obtained for untreated surfaces that contained consider-
able oxides. Additionally, the epitaxy was only obtained at higher deposition tem-
peratures ( 70°C); at lower temperatures, the growth was polycrystalline.
       Epitaxial growth of cubic CdS {111} on the CIGS {112} was also found to
occur using a room-temperature bath, with gradual heating to 80°C 18,19]. Again,



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the epitaxy was explained by ammonia cleaning of the CIGS followed by ion-by-
ion growth.
9.1.4.4      CuGaSe2-based cells
CuGaSe2 (CGS) has also been studied as a PV material, although efficiencies of
cells based on this semiconductor are, at present, much lower than those using
CIS or CIGS. There is one study on the specific interaction between CD CdS
and CGS [23]. Two different CD baths were used to deposit CdS in this inves-
tigation: one at 60°C and the other at 80°C (the latter with a higher ammonia
concentration to slow down the deposition). Several pronounced differences
were found between the two baths, in spite of the relatively small difference be-
tween them. For Cu-rich CGS, Cu-S inclusions in the CdS were formed in the
high temperature bath, due to interaction between Cu in Cu-rich CGS and CdS,
but not in the low-temperature bath. Such inclusions could lead to shunting. For
Ga,In-rich films (which gave better cells), the 80°C deposition resulted in Se be-
ing found in the CdS layer. Additionally, the higher-temperature CdS was less
defected and formed a less defected interface with the CGS. The 80°C deposi-
tion gave better PV properties for Ga-rich films (up to 9.3% efficiency). The
60°C deposition, however, was better for the (poorer cell quality) Cu-rich films,
which could be explained by shunting through the Cu-S inclusions in the 80°C
Cu-rich CGS devices. This investigation clearly shows the necessity for opti-
mization of the CD process, not only for every specific absorber material, but
even for different types of the same absorber.
9.1.4.5      Cd-Free Buffer Layers
While most of the reported studies on CD buffer layers deal with CdS, there have
been a number of attempts to chemically deposit other materials. There are sev-
eral reasons for this. One is the desire, particularly prominent in Europe, to find a
Cd-free cell, for obvious environmental reasons (this clearly is relevant for the
CIS and not for the CdTe cell). Another reason is that part of the light absorbed in
the CdS is lost for current generation (a point that does not seem to have been rig-
orously investigated but that probably is due to a high recombination rate for holes
photogenerated in the CdS). A higher-bandgap material will therefore allow more
short-wavelength photons to reach the CIS absorber and generate photocurrent.
Since the CD process is believed to be largely responsible for the beneficial prop-
erties of the CdS films, it has been anticipated that other CD layers would behave
likewise. Other possible factors are the band lineup between CIS and CD layer and
(photo)conductivity (or maybe better, resistivity, of the buffer layer, since it ap-
pears that a high resistivity is beneficial).
      ZnS. ZnS, with a high bandgap of 3.7 eV, is an obvious choice for alter-
nate buffer layers. ZnS films, prepared using thioacetamide from an ammonia-free



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
bath (see Sec. 4.4.2 for a description of the films) were deposited onto CIS films
[24]. The higher short-circuit current expected for the higher-bandgap ZnS was
obtained, although the overall efficiencies were lower than for CdS deposited on
similar substrates, due to lower photovoltages. Efficiencies as high as 9% were
obtained.
       Most ZnS baths probably give films containing some hydroxide (at higher
temperature, maybe oxide), as discussed in Section 4.4.1. The previous bath may
be an exception, since it was presumably carried out at relatively low pH. Using
Zn(O,OH,S), efficiencies as high as 12.8% were obtained on CIGS substrates
[25]. This efficiency was obtained after ca, 60-mn illumination and reversibly de-
creased when kept in the dark. Spectral response measurements showed the ex-
pected increase in short-wavelength response. Such an efficiency can be com-
pared with the best reported value over 16% obtained at that time; however, since
the highest reported values are considerably higher than those more routinely ob-
tained, then the value of 12.8%, while lower than that obtainable using CdS, is not
too far below it. Ideally, such experiments should be compared with state-of-the-
art CdS deposited on (as closely as possible) identical substrates.
       A somewhat more recent study reported 15.1% for ZnS/CIGS (compare
with 17.0% state-of-the art CdS on the same CIGS substrates) [26]. This study
showed the differences in optimization of the cell, depending on whether CdS or
ZnS was used. Thus, while a 40 nm-thick ZnS layer led to a large improvement in
all cell parameters, as was also the case for CdS, in contrast to CdS, increasing
thickness of the ZnS to 90 nm reduced the cell performance, giving in particular a
very low fill factor. This was explained by the higher resistivity of the ZnS (by two
orders of magnitude), compared to CdS. In addition, while a high-resistivity sput-
tered ZnO layer was deposited on the CdS prior to the conducting ZnO:A1 to ob-
tain maximum VOC, such a layer degraded the ZnS-buffer cell. Since this ZnO was
sputtered in an oxygen-containing atmosphere, and no such degradation occurred
for the conducting ZnO sputtered in an oxygen-free environment, it was believed
that the degradation was caused by negative oxygen ions or energetic neutral par-
ticles. Finally, while the short-wavelength response of the ZnS cell was better than
that of the CdS cell, this improvement did not compensate for the lower quantum
efficiency at longer wavelengths for the ZnS cell.
      ZnSe. ZnSe, with a bandgap of ca. 2.7 eV, is another obvious substitute for
CdSe. As with ZnS, there is a tendency for CD ZnSe to contain hydroxide.
Zn(Se,OH) deposited from a selenourea bath was deposited on CIGS [27] (see
Sec. 4.5.2 for details of the Zn-Se films). Efficiencies up to 13.7% (highest litera-
ture value 17% using CdS) were found. Spectral response measurements
showed the expected improvement in short-wavelength response. The optimal
thickness of the CD layer was 7–8 nm; a layer ca. twice that thickness resulted in
a drop in efficiency (to 10.4%), mainly due to a drop in fill factor, probably be-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 9.4 Band diagram of the CIGS/Zn(Se, OH)/ZnO cell. (After Ref. 27.)




cause of the higher resistivity of the thicker film. From XPS measurements, the
band alignment could be estimated; this is shown in Figure 9.4. Tunneling of elec-
trons through the large barrier at the junction between the CIGS and Zn(Se,OH)
conduction bands was presumed to occur readily for the thin films but much less
so for the thicker ones.
      ZnSe, deposited by the same method, was also used as a buffer layer for
CuInS2 cells [28]. Higher currents and voltages but lower fill factor were obtained,
compared to CdS, with a slightly lower overall efficiency. The band lineup for the
ZnSe/CuInS2 junction was also measured by XPS for this system.
      ZnO. ZnO, which is normally used as the conducting window layer on the
CIS-type cell, has also been used as a buffer layer for CuInS2 cells by annealing
CD Zn(OH)2 (deposited from Zn2 /ammonia solutions) [29] (see Sec. 7.2.18,
ZnO). The efficiency was much lower (3.8%) compared with comparable CdS
buffer layers (8.6%). The difference was due to much lower open-circuit voltage
and resulting lower fill factor; the photocurrents were similar using both buffer
layers. The increase in photocurrent at short ( 500 nm) wavelengths due to the
higher-bandgap ZnO was offset by lower photocurrents over the rest of the active



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
spectrum. While the thickness of the ZnO was not given, it was noted that the In
XPS signal was still visible after the ZnO deposition. Comparing this with the pre-
vious study by the same group using Zn(Se,OH), where the In peak disappeared
when the CD layer was 7 nm and was already weak at ca. 4 nm, suggests either
that the ZnO layer was thinner than optimal or that the coverage was not homo-
geneous. In view of the previous experiments on CI(G)S, which showed that a
Zn2 /ammonia treatment was as effective as a Cd2/NH3 treatment and both not
much less effective than optimally deposited CdS, the Zn2 /NH3 bath used here
was considerably inferior to CdS buffer layers, even assuming an overly thin film
deposited. Whether this difference is due to the different substrate (CuInS2 instead
of CI(G)S), to the annealing treatment, or to some other reason remains to be in-
vestigated.
       In(S,OH). Various compounds of In have been used, with some success,
for buffer layers. In(OH)3 was grown on CIS (In-rich) films from a solution of
InCl3 with thiourea (which possibly acted to gradually increase pH rather than as
a source of S) [30]. In spite of the higher blue response compared to a control CdS
cell [In(OH)3 is colorless as a film], the red response was poorer, leading to a
somewhat reduced overall photocurrent. The fill factor was also less. A best effi-
ciency of 9.5% was obtained, compared to 11.9% for the control using the same
batch of substrates. In the same paper, the deposition of In2S3 [based on later stud-
ies, this may have been In(S,OH)] from a thioacetamide bath at a pH between 1
and 2 was described, but with considerably poorer results (both photocurrent and
fill factor were much lower).
       Using essentially the same method for depositing In(S,OH) on CuInS2
films, the same group found slightly better performance (11.4%) with this buffer
layer than for CdS (10.8%), due to increase in photocurrent and also in open-cir-
cuit voltage [31]. Since CuInS2 has a higher bandgap (1.5 eV) than CIS or CIGS,
the fractional increase in photocurrent due to the improved blue response is larger.
       These same In(S,OH) films were also investigated on CIGS substrates [32].
XPS measurements (on CuGaSe2 used to prevent interference by In from the CIGS)
showed that the deposited film was not a mixture of In2S3 and In(OH)3, as might
be reasonably expected, but was a compound containing In, O, and S. Higher
thioacetamide concentrations resulted in better device performance, while the In
concentration was not found to influence the performance in any reproducible way.
The completed cells (with ZnO window layer) required an anneal (2 min at 200°C)
and light soak; the performance without this anneal was much poorer. The best cell
gave an efficiency of 15.7%. In comparison with CdS buffers, the VOC increased,
the ISC decreased, and there was no change in the fill factor. Overall, the efficien-
cies were only slightly less than for cells using CdS. For CIGS substrates deposited
on Corning “Pyrex-type” glass (good cells are usually deposited on soda-lime glass
and are beneficially affected by Na diffusion through the film), the In(S,OH) cells



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
were actually better than the CdS ones. In fact, in contrast to CdS cells, the effi-
ciency of the In(S,OH) ones were independent of the type of glass used for the sub-
strate. Capacitance–voltage measurements indicated a 10 times increase in accep-
tor concentration of the absorber layer and narrowing of the space charge layer,
compared to CdS. This explained the decrease observed in the red response (the
blue response was better) and the increase in the VOC, and also the independence
on the glass, since acceptor doping occurred by the CD process rather than by Na
diffusion. Since the CD process did not change the bulk of the CIGS, it was as-
sumed that the doping changes were due to surface (and grain boundary) effects.
       A list of cells made on CIS, CuInS2, and different CIGS substrates using
In(OH,S) buffer layers is given by Hariskos et al. [32]. They note that the fill fac-
tor of these cells drops with time when the cells are kept in the dark but that illu-
mination (by light absorbed by the In(OH,S) film) reverses the degradation. This
behavior suggests that the degradation is due to adsorption of some species (oxy-
gen, water?) that reacts with photogenerated electrons and/or holes.
      Sn(O,S). In the same paper, films of SnO2 and Sn(O,S)2—the latter ap-
parently a mixture of oxide and sulphide—were used as buffer layers on CIS and
CIGS but with all cell parameters lower than when using CdS and efficiencies
lower than those obtained using In(OH,S)

9.1.5 Other Cells and Related Studies
9.1.5.1      Other Heterojunctions and Devices
There are several studies on heterojunctions, other than those already described,
formed with at least one CD semiconductor. These will be described briefly here.
      The earliest of these studies was on PbS. PbS can have either p- or n-type
conductivity, although CD PbS is usually p-type. Based on the belief that the p-
type conductivity may be due to alkali metal cations from the deposition solution,
an alkali metal—free deposition, using lead acetate, thiourea, and hydrazine hy-
drate was used [33]. While initially n-type, the film converted to p-type in air. At-
tempts to stabilize the p-type material by adding trivalent cations to the deposition
solution were unsuccessful. However, deposition of the PbS on a trivalent metal,
such as Al, did stabilize the n-PbS, at least for a time. In this way, p-n junctions
were made (the PbS close to the trivalent metal was n-type, while the rest of the
film was p-type). Photovoltages up to 100 mV were obtained from these junctions
at room temperature and almost 300 mV at low temperatures (90 K).
      PbS was also deposited on single-crystal n-and p-type Ge [34]. The PbS was
epitaxial (111) with the (111) Ge (Ge has a 5% smaller lattice spacing than PbS).
A photovoltage was measured from the junctions. The photoresponse extended to
1.75 m for the junctions on p-type or intrinsic Ge and to 3.35 m on n-type Ge.
The difference could not be explained, although it can be noted that these values



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
correspond approximately to the bandgaps of Ge and PbS, respectively. Hot probe
measurements indicated that the PbS layers were n-type; however, it was men-
tioned that this measurement may be affected by the junctions and that PbS chem-
ically deposited onto glass by the same method normally gives p-type layers.
       Investigations of junctions formed by CD of PbS on n-Si [35] and n-Sb2S3
on p-Si [36] and on p-Ge [37] have been made. In particular, the Sb2S3 junctions
were found to be much more PV active if a small amount of silicotungstic acid
(STA) was added to the deposition solution. Conversion efficiencies of 5.2% on Si
and ca. 4% on Ge could be calculated from the photocurrent–voltage characteris-
tics. The STA resulted in formation of WO3 in the Sb2S3 film, and it was believed
that this may, at least in part, be responsible for the improvement. Of particular note
was the relatively large open-circuit voltage (nearly 0.7 V) obtained from the junc-
tion with Ge; this value is almost as large as the Ge bandgap. It is tempting to won-
der if this junction was not closer to a Schottky (metal–semiconductor) junction,
where the Ge behaves as a metal, in which case, the maximum photovoltage is lim-
ited by the Sb2S3 bandgap (ca. 1.7 eV). The p-Ge was highly doped (4 1018
cm 3), although less than might be expected if it were to behave as a metal.
       CuInSe2, deposited by CD (see Chap. 8), has been screened for photovoltaic
activity. In one study, CdS was evaporated on CIS that was chemically deposited
onto conducting glass [38], while in the other the CIS was chemically deposited
onto single-crystal Si [39]. The cells gave low activity, although the CdS/CIS cell
gave a short-current photocurrent of 4 mAcm 2 (AM1 illumination), which is
quite appreciable, if still low, and suggests further studies in this direction might
be fruitful, in particular also using CD CdS.
       CdO, a degenerate n-type semiconductor, was chemically deposited on sin-
gle-crystal p-type Si [40]. The junction showed clear diode behavior, and, al-
though no photovoltaic effect was observed, photocurrent was generated under re-
verse bias. From the spectral response of the photocurrent, almost all of the current
generation occurred in the Si.
       CdS/SnxS PV cells have been fabricated where the CdS was deposited by
CD and the SnS deposited by a variant of CD where the substrate is dipped first
in a solution of one of the ions and then in the other without rinsing in between, as
would be the procedure for SILAR (see Sec. 2.11.1) [41]. While the cells showed
very low conversion efficiencies, the main emphasis was on Ag-doping of the CdS
in order to increase the conductivity and the effect of this doping on the PV cells.
An increase in efficiency from 0.03% to 0.08%, mainly as a result of an increase
in short-circuit current, was obtained by doping the CdS with Ag. The doping was
carried out by an ion exchange process whereby the undoped CdS film was im-
mersed in a solution containing Ag complexed by thiosulphate.
       A heterojunction between two different CD semiconductors—n-Ag2S de-
posited on top of p-PbS—was fabricated [42]. Some photoconductivity was found
at wavelengths longer than that corresponding to the Ag2S bandgap (ca. 0.9 eV).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
However, at shorter wavelengths, where light could be absorbed in the Ag2S, neg-
ative photoconductivity (increase in resistivity with light) was found. This was ex-
plained by a combined Dember effect/electrostatic attraction of electrons by Ag
ions.
       Metal–CD semiconductor Schottky junctions have been examined as solar
cells. As for the p-n junctions described earlier, the addition of STA to both Sb2S3
and CdSe improved the cell parameters greatly [43]. AM1 efficiencies (for an-
nealed films) of 7.2% and 5.5% were obtained using Pt contacts on annealed CdSe
and Sb2S3, respectively. Again, the improvement due to the STA was attributed to
the presence of WO3. In this study, it was suggested that the WO3 might introduce
favorable interface states in the devices. Whatever the reason for the improvement,
the strong effect of the STA treatment and its applicability to two different semi-
conductors (as well as to both solid-state and liquid junction cells) warrants further
investigation. A similar study, using CD CdSe (with STA) deposited on poly(3-
methylthiophene), the latter electropolymerized (or the polythiophene electrode-
posited onto CdSe), was carried out [44]. Poly(3-methylthiophene) can be prepared
either p-semiconducting or doped to a metallic conductivity. While the undoped p-
n junctions gave poor photoresponse, the Schottky-type doped thiophene-CdSe
junction gave conversion efficiencies of 2.7%, which were stable for at least 72 hr
of illumination. CdSe deposited without STA resulted in lower efficiencies (1.3%).
       Thin-film transistors have been fabricated by depositing 50 nm of CdS onto
SiO2-covered n Si and evaporating two A1 electrodes (source and drain) onto the
CdS [45]. Similar devices were also made using CdS deposited on polyimide sub-
strates with three (source, drain, and gate) evaporated metal electrodes and vari-
ous sputtered insulator layers for the gate electrode.
9.1.5.2      Passivation Studies
Various sulphiding treatments have been known to passivate the surfaces of III–V
semiconductors like GaAs and InP. Similar effects have been found with CD of
very thin CdS films on InP surfaces. By deliberately oxidizing InP surfaces to pro-
duce In and P oxides, it was shown that a standard CdS solution (only with a higher-
than-usual concentration of ammonia) removed these oxides and prevented oxide
regrowth [46]. The same treatment also removed P vacancies deliberately intro-
duced by annealing in H2, as seen by photoluminescence studies, presumed due to
filling of P vacancies by S. Deposition of CdS (5–7 nm) onto InP both improved
the C–V behavior and lowered the interface state density of SiO2/InP junctions by
an order of magnitude. Further studies found that a pretreatment of the InP surface
by a solution of NH4OH/thiourea improved the passivation but that CdS deposition
(3–4 nm) was still important [47]. The NH4OH/thiourea treatment was believed to
produce a stable In-S terminated InP surface [48]. This latter study also reported
that the SiO2/InP junction quality, measured mainly by the interface state density,
was maximum at a very low CdS thickness (ca. 1 nm) and that this quality degraded



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
gradually as the CdS thickness increased, presumably due to interface states intro-
duced by the thicker CdS. It appears that the NH4OH/thiourea treatment passivates
the surface, while the very thin CdS protects it from reoxidation. This treatment
was extended to various device structures based on ternary semiconductors
(InA1As and InGaAs) with similar improvements due to removal of interface states
[49]. These studies are clearly related to similar ones on CI(G)S.
      CD ZnSe has also been demonstrated to passivate surface states, 0.92 eV be-
low the conduction band edge (measured by thermally stimulated exoelectron
emission) on single crystal GaAs. This passivation resulted in bandgap lumines-
cence from the originally non-luminescent GaAs [49a].


9.2 PHOTOELECTROCHEMICAL CELLS (PECs)
9.2.1 Introduction and Background
As noted earlier, there are numerous studies on the photoelectrochemical (PEC)
properties of CD films. Many, if not most, of these studies describe the prepara-
tion of the films and some PEC properties. In such cases, rather than describe each
study separately, it is more useful and efficient to tabulate the results, providing
important cell parameters together with the reference. Additional relevant infor-
mation will be given separately for each individual reference. However, no at-
tempt is made to cover all the individual studies in any detail, but rather to give
enough information to allow the reader to decide whether it may be worthwhile to
refer to the original reference. Some studies that treat the PEC properties of these
films in a more fundamental way will be discussed separately. An important issue
is whether the films have been annealed or not and under what conditions; an-
nealed films usually give better performance (normally much better) than nonan-
nealed ones.




FIG. 9.5 Schematic diagram of a PEC.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       We start with a brief introduction to PECs. Figure 9.5 presents a schematic
diagram of a PEC showing a semiconductor film on a substrate—the photoelec-
trode—connected through an external meter and/or load to a second electrode (the
counterelectrode). The two electrodes are immersed in an electrolyte, and the semi-
conductor film is exposed to illumination. If the substrate is transparent, conduct-
ing glass, the light can pass first through the substrate and then to the semiconduc-
tor (and the glass can also function as the window of the cell); this configuration is
know as a backwall cell, in contrast to the normal frontwall cell (shown in Fig. 9.5),
where light is incident directly on the nonsubstrate side of the semiconductor film.
       The following discussion assumes that the semiconductor crystal size is
large enough so that charge transport is dominated by a space charge layer in the
semiconductor. This is typically the case when chalcogenide films have been an-
nealed at temperatures of ca. 400°C or more, where the crystal size is typically of
the order of hundreds of nanometers. This assumption is usually not valid for as-
deposited CD films or for those annealed at low temperatures (e.g., 250°C or less).
The mode of operation of such films is treated in Section 9.2.2.2.
       Figure 9.6a shows the band diagram of a semiconductor–electrolyte junc-
tion (with an n-type semiconductor; most studies described here have used films
that give n-type response). The band bending represents the situation in the dark
at equilibrium (flat Fermi level) or under illumination at short circuit (i.e., the pho-
toelectrode (or photoanode) is short-circuited to the counterelectrode). (We ignore
complications of quasi-Fermi levels due to any nonequilibrium situation in the
light; it does not change the simple picture for our purposes). Photogenerated elec-
trons (in the conduction band) and holes (in the valence band) are spatially sepa-
rated by the space charge layer and flow in opposite directions. The holes flow to
the semiconductor/electrolyte junction and oxidize some species in the elec-
trolyte. The electrons flow through the back (ohmic) contact to the external cir-
cuit. Electrons thus flow to the counterelectrode and reduce some electrolyte
species. (For a p-type semiconductor, a photocathode, the direction of charge flow
is opposite: Electrons flow to the semiconductor/electrolyte junction and holes to
the back contact.) The resulting current flowing is the short-circuit current (ISC).
For a regenerative PEC, the electrolyte species that are oxidized and reduced form
a single redox couple. A common example is the polysulphide redox solution. The
reactions occurring at the photoanode are:
  S2        2h → S;      S     S2 → S2
                                     2          etc. up to approximately S2
                                                                          5       (9.1)
while at the counterelectrode, the reverse occurs:
       S2
        2     2e → 2S2           or      S    2e → S2                             (9.2)
The result is no net change in the PEC (hence the term regenerative) and electric-
ity is produced.
       The regenerative PEC has been the type predominantly studied using CD



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
A




B

    FIG. 9.6 Band bending in a PEC (a) in the dark or under illumination at short circuit and
    (b) under illumination at open circuit.




    films. However, there is another important type, where the anodic and cathodic re-
    actions occur with different redox species. The holy grail for this type of cell has
    been to photoelectrolyze water to hydrogen and oxygen. While this goal has at-
    tained only limited success, the search has led to very decided success in other, re-
    lated directions, usually connected with photo-oxidizing adsorbed layers of pollu-
    tants or bacteria on TiO2 [50].
           If the photoanode is not connected to the counterelectrode, then current
    cannot flow, and instead the potential of the photoanode (the Fermi level) rises
    until balanced by recombination of the photogenerated charges (Fig. 9.6b). The
    difference between this potential under illumination and the original potential
    (more correctly, the potential of the counterelectrode, which ideally is equal to
    that of the nonilluminated photoelectrode) is the open-circuit voltage (VOC) of the
    cell.



    Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In practice, in order to generate electrical power, the cell must operate un-
der conditions where both current and voltage are generated, as with the photo-
voltaic cells described earlier. This situation is shown in Figure 9.7, which gives
the photocurrent/photovoltage characteristics of the cell. The maximum power
(Pmax) is generated when the load is such that the product of the current (IP) and
voltage (VP) is a maximum. The shape of the photocurrent/photovoltage charac-
teristic, which determines Pmax, is quantified as the fill factor (FF), which is de-
fined as the ratio between Pmax and the product of ISC and VOC, i.e.,
                Pmax
      FF                                                                           (9.3)
             ISC VOC
and is given either as a fraction or, commonly, as a percentage.
       An important feature of photoanodes is that the photogenerated holes, which
are normally very strongly oxidizing, may oxidize the semiconductor instead of, or
as well as, the electrolyte species. This phenomenon is known as photocorrosion.
For the purposes of the limited explanation of PECs given here, it is enough to note
that continuous photocorrosion will destroy the photoelectrode. (This need not nec-
essarily occur if corrosion is confined to the semiconductor surface.)
       The substrate on which the semiconducting photoelectrode is deposited is
important, not just as an ohmic contact to extract charge (usually electrons, since
most photoelectrodes are photoanodes), but also because the substrate should be
electrocatalytically as inactive as possible toward the electrolyte species. The rea-
son for this is explained in Figure 9.8. The heavy solid line shows a typical pho-




FIG. 9.7 Current–voltage characteristic of a photoanode showing maximum power
point (Pmax). The fill factor is given by the product VP IP divided by the product of VOC
and ISC.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 9.8 The effect of exposed substrate on a PEC. The thick solid line gives the light-
induced current–voltage characteristic. The thin solid line gives the net current–voltage
characteristic when the substrate has the electrocatalytic activity given by the broken line
(fair electrocatalyst). If a substrate with poor electrocatalytic activity is used, there is little
effect of the substrate on the photocharacteristics in the fourth quadrant.



toanode response; it is assumed that dark currents are negligible. We now consider
the dark (or under illumination—there will be no difference) current–voltage
characteristic of any exposed substrate. This is shown for a poor electrocatalyst
and a better (still not good) electrocatalyst. The higher the electrocatalytic activ-
ity of an electrode, the greater the current at a given bias. For a photoanode, the
current in the fourth quadrant is the most important (for a photocathode, it is that
in the second quadrant). If the positive current of the (no dark current) photoan-
ode and the negative current of the exposed substrate are now added, the effect of
the poor electrocatalyst is negligible while that of the better electrocatalyst is to
strongly reduce both the VOC (from VOC to VOC(1)) and the fill factor of the PEC.
This is the reason that Ti, a poor electrocatalyst for most redox systems, is used so
commonly as a substrate for photoanodes. Another substrate sometimes used is Cr
(also a poor electrocatalyst in general), e.g., for Bi2S3 [51] and Sb2Se3 [52]. This
is in addition to the fact that these metals tend to form satisfactory ohmic contacts
to n-type semiconductors. Note that for nanocrystalline films where a space
charge layer may not exist, the concept of ohmic or nonohmic contact is not nec-
essarily the same as for bulk semiconductors (see Sec. 9.2.2.2).




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
9.2.2 CdSe
9.2.2.1 Annealed Films
CdSe has been the most extensively studied semiconductor for PEC purposes.
This is due to its fairly favorable bandgap for solar cell use (1.73 eV), compared
to the higher-bandgap CdS, although there are many, usually less detailed, PEC
studies on CdS as well. Cell details are given in Table 9.1, specific comments fol-
low.
      In Ref. 53, the main purpose of the study was to investigate PEC properties
of CD CdSe films rather than to optimize the actual solar cell (e.g., CdSe anneal-
ing was carried out at the relatively low temperature of 280°C and the film thick-
ness was only ca. 0.9 m), hence the cell parameters are not as high as they could
be. The CdSe/polysulphide junction was characterized by a number of techniques.
The effects of an surface layer of CdS, due to exchange of Se by S from the poly-
sulphide, were considered.
      Preliminary PEC results in Ref. 54 were previously described [60]. This
study was directed to optimization of the PEC parameters. The deposition was
based on the ammonia/selenosulphate bath. The Ti substrates were treated with a
suspension of Cd(OH)2 and the deposition carried out in a sealed tube to prevent
loss of ammonia and thereby to improve reproducibility. Several layers (about
five) were deposited to give a total thickness of ca. 2.5 m, which was found to
be optimum. The first layer was annealed at 500°C in air (to improve film adher-
ence) and the final film at 550°C, also in air. The annealed films were etched (9M
HCl) and treated with a ZnCl2 solution. The maximum efficiency was obtained at
50 mWcm 2 illumination (6.8%). At higher illumination intensities, the effi-


TABLE 9.1 PEC Parameters of CD CdSe

                        Isc          Voc                                   Illumination
Efficiency (%)       (mAcm 2)        (V)       FF         Electrolyte       (mWcm 2)      Refs.

Annealed
0.15                     1.92       0.23        0.34   Polysulphide        100            53
6.3                     15.3        0.66    ca. 0.42   Polysulphide         67            54
4.9a                    16.0        0.66        0.46   Polysulphide        100 (AM1)      55
4.4                      8.4        0.85        0.62   Polysulphide        100            56
11.7                    11.8        0.59        0.67   Selenosulphate       40            57, 58
Not annealed
0.1                      0.06       0.1         0.28   Polysulphide        Not defined    59
a
    Higher efficiency (5.5%) was found for lower solar illumination (42.5 mWcm 2).




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
ciency dropped, due to a sublinear increase in ISC with illumination at higher in-
tensities.
       The effects of annealing and film thickness on the PEC properties were in-
vestigated in Ref. 55. The optimum annealing conditions were 470°C for one hour
in air. ISC in particular was found to drop strongly both at lower and higher an-
nealing temperatures. ISC and VOC, and therefore efficiency, increased strongly
with film thickness up to ca. 1.5 m and then more gradually up to ca. 2 m, af-
ter which no further change was observed. Such thicknesses result from a number
of successive depositions. Larger-area photoelectrodes of 18 cm2 were also made;
the efficiency of these cells dropped to 3.2% compared with small cells, mainly
due to a drop in fill factor and ISC.
       A triethanolamine/ammonia/selenosulphate bath was used in the experi-
ments of Ref. 56. Three depositions were employed to give a final film thickness
of 4 m. The films were annealed at 500°C in N2 and etched in dilute aqua regia.
The efficiency at this stage in a polyselenide electrolyte was ca. 0.5%. This study
concentrated on the effects of a treatment consisting of a 50% HCl etch, which re-
sulted in a black matte surface (indicating a high surface roughness), followed by
a dip in an acidified ZnCl2 solution (a Zn2 dip has been previously been used
beneficially for CdSe photoelectrodes [61]). This surface treatment improved all
cell parameters. Using Kelvin probe studies, it was shown that the surface poten-
tial (measured against a Pt vibrating probe in nitrogen ambient) changed by 0.31
V after the treatment, interpreted as an increase in band bending due to change in
the surface charge. This change was the same as the increase in VOC of the cell.
Analysis of the dark current–voltage characteristics of the cell showed a decrease
in both the reverse saturation current and the ideality factor after the surface treat-
ment. That Zn was present in the samples was verified by separate XPS measure-
ments [62]. A spectral response study of these films was carried out, with an em-
phasis on subbandgap response, which might be related to surface states [63]. The
surface treatment decreased the subbandgap response (although it should be re-
marked that the subbandgap response of the untreated film was unusually high,
extending out to 1 m and with relatively high quantum yields for a subbandgap
response). It also preferentially increased the short-wavelength response, inter-
preted as a reduction in surface recombination. The beneficial effect of the surface
treatment was found to occur not just in polyselenide electrolyte, but also in a fer-
rocyanide electrolyte. This suggests that the effect of the Zn is not related to for-
mation of ZnSe at the surface, as might be thought. On the other hand, a large part
of the improvement in polyselenide electrolyte was due to the HCl matte etch, and
it is not clear how much of the improvement in the ferrocyanide electrolyte was
caused by this part of the treatment.
       In Ref. 57, a small concentration ( 10 5 M) of silicotungstic acid
(H4SiW12O40) was added to a triethanolamine/ammonia/selenosulphate bath,




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
deposition carried out at 40°C and annealing in air at 430°C (the CdSe remained
cubic, in contrast to the hexagonal form usually obtained after annealing at such
temperatures). A large increase in the PEC performance was obtained, compared
to photoelectrodes deposited without the silicotungstic acid. All cell parameters
were increased, but the major effect was on ISC. The stability of the PEC also im-
proved: It was stable (at 40 mW-cm 2 illumination) for greater than 2800 C/cm2,
although the selenosulphate electrolyte is unlikely to be stable in a PEC over the
long term. A follow-up study of these films [58] showed the presence of WO3 in
the films. From the dark current–voltage characteristics, the reverse saturation
current, ideality factor, and series resistance all decreased as a result of the modi-
fication. Two possibilities were put forward to explain the effect of the silico-
tungstic modification. One was that formation of a CdSe/WO3 heterojunction oc-
curred, improving the charge transfer at the semiconductor/electrolyte interface.
Another possibility, if a true heterojunction exists, is reduction in majority carrier
(electron) transfer to the electrolyte, as in a metal/insulator/semiconductor device.
In another study, it was suggested that the improved PEC response was due to
charge transfer catalysis by the W-containing groups adsorbed at the CdSe crys-
tal surfaces [64]. It should also be noted that the silicotungstic acid modification
also improved the performance of solid-state cells (see Sec. 9.1.5.1) where no
charge transfer to an electrolyte is involved, suggesting (although not proving)
that electrocatalysis is not the reason for the improved PEC behavior.
9.2.2.2 Nonannealed Films
In Ref. 59, films 0.8 m thick were deposited from an ammonia/selenosulphate
bath. Various configurations of PECs were studied.
       A study of the mode of operation of nanocrystalline CdSe photoelectrodes
was carried out [65]. This was based on the expectation that nanocrystals, being
(usually) much smaller than typical space charge layer widths, would not sup-
port such a space charge layer and therefore that some other mechanism for
charge separation should be considered. It was originally reported that nanocrys-
talline CD CdS [66] and CdSe [67] photoelectrodes, which normally gave pho-
tocurrent–voltage behavior characteristic of an n-type semiconductor, gave p-
type behavior after etching in dilute HCl, an example of which is shown in
Figure 9.9. The study by Kronik et al. [65] used surface photovoltage (SPV)
spectroscopy and X-ray photoelectron spectroscopy (XPS) to investigate this ef-
fect. It was concluded that the CdSe itself was close to intrinsic, as might be ex-
pected for very small nanocrystals [the crystal size in this study was 4–5 nm; no
such apparent reversal of semiconductor type was found for considerably larger
(ca. 20 nm) crystal size]. The direction of photocurrent flow, rather than being
determined by an electric field (the space charge layer) in the semiconductor,
was determined by trapping of charge carriers at the individual nanocrystal sur-
faces. Both electrons and holes could be trapped at the surface; however, the



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 9.9 Current–voltage characteristics under chopped illumination ( AM1) of a
nanocrystalline (4–5 nm) CdSe film, deposited by CD, in a polysulphide electrolyte. The
two characteristics are for as-deposited CdSe (top) and after etching with dilute HCl (bot-
tom). (After Ref. 67.)




charge that was preferentially trapped (longer lifetime of trapped charge, which
usually correlated with deeper trap sites) would preferentially be transferred to
the electrolyte. Such a scenario is reasonable due both to the longer residence
time of the charge at the surface as well as better overlap of a deeply trapped
carrier with the redox species, which, for the polysulphide system used in these
experiments, is located deep in the bandgap. The presence of such charge traps
and the effect of water on them was further substantiated in later studies using
photoluminescence [68,69] and scanning probe spectroscopies [70]. See Section
4.2.7.1 for a description of surface trapping in CD CdSe measured by various
photoluminescence studies.
        The photovoltage of such a cell comes not from neutralization of a built-in
space charge layer, but from change in the Fermi level in the almost intrinsic
nanocrystal upon illumination. This change in Fermi level is also determined by
the relative trapping of electrons and holes. If no trapping occurred and no charge
extraction took place, then illumination would not change the potential of the
semiconductor appreciably, since both electron and hole Fermi levels would
change by (close to) the same amount and in opposite directions. However, to take
a simple example, if the holes are deeply trapped and the electrons are not trapped
at all, then the shift in the overall Fermi level is determined almost completely by



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
the electron concentration; i.e., the Fermi level rises on illumination, similar to
what happens in an n-type semiconductor. The opposite will occur if electrons are
trapped and holes are free. In practice, both charges are usually trapped to a greater
or lesser extent, and the shift in Fermi level, and therefore the photovoltage direc-
tion and magnitude, will be dependent mainly on the relative trapping depths of
the two carriers.
       While it has not been studied, it is probable that there exists a range of crys-
tal sizes, possibly quite wide, where both surface trapping and space charge layer
effects contribute to the PEC functioning of the photoelectrode.
       While the cell efficiencies of these films were not specifically investigated,
best parameters of 2 mAcm 2 (ca. AM1 illumination; quantum efficiencies in-
creased with decreasing illumination intensity due to diffusion limitations in the
nanoporous film); 0.5 V and ca. 50% fill factor were obtained. However, great
variation in these parameters were obtained; one reason for this can be seen from
a consideration of Figure 9.9. If a CdSe film is etched, but less than optimally
(shorter time, more dilute HCl), it is clear that after a certain, unique etch treat-
ment, zero net photocurrent will be obtained. The actual photocurrent (and other
output parameters) of the film is a balance between photoanodic and photoca-
thodic currents.
       A word on “ohmic” contacts to space charge layer–free nanocrystalline
films. The ability of a contact to function as an ohmic contact to such films is de-
termined by the offset between the metal contact Fermi level and the semicon-
ductor energy bands, rather than by a potential barrier in the form of an electric
field in the semiconductor. It seems that, in practice, most conductors act as good
sinks for photogenerated charges in nanocrystalline semiconductors. One reason
for this may be that accumulation of one type of charge (the other is usually re-
moved rapidly by the electrolyte at the very high contact area between nanocrys-
tals and electrolyte) will charge the particle and raise the energy levels until the
accumulated charge can flow to the sink.

9.2.3 CdS
In contrast to CdSe, most studies on CdS involved either nonannealed films or
films annealed at relatively low temperatures. Relative efficiencies (i.e., taking
into account that the efficiency of the higher-bandgap CdS will be lower than that
of CdSe due to its lower light absorption) are therefore low. Additionally, most
PEC studies on CD CdS involve doped films, where ions of the dopant were added
to the deposition solution.
       In Ref. 71, Al doping (0.1 wt% Al added to the deposition bath; not neces-
sarily the same percentage in the films) improved the PEC properties. Annealing
(in H2) at 200°C, which removes S and should make the films more n-type, im-
proved the cell parameters of both doped and undoped films [72,76]. Stability data



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 9.2 PEC Parameters of CD CdS

                        Isc         Voc                              Illumination
Efficiency (%)       (mAcm 2)       (V)       FF      Electrolyte     (mWcm 2)      Ref.

0.003                     0.06    ca. 0.11    0.33   Polysulphide        100        71
0.0075 (Al-doped)         0.125   ca. 0.11    0.30   Polysulphide        100
0.04 (Al-annealed)        0.5         0.24    0.35   Polysulphide        100        72
0.01                      0.17        0.17    0.44   Polysulphide        100        73
0.05 (In-doped)           0.4         0.23    0.53   Polysulphide        100        73
0.022                     0.3         0.175   0.42   Polysulphide        100        74
0.05 (As-doped)           0.5         0.22    0.44   Polysulphide        100        74
—                     ca. 0.45    0.58–0.6           Polysulphide         60        75
                                                       or sulphite



in polysulphide electrolytes were presented. In doping (optimum 0.01% In) also
improved the PEC characteristics [73].
       The study in Ref. 74 followed essentially the same deposition procedure as
before, but on Cr-plated steel (compared with steel in the previous studies; see
Sec. 9.2.1 for an explanation of the effect of Cr plating based on the poor electro-
catalytic activity of Cr) and with a CuCl:KCN etch at 90°C (the reason for this
specific etch was not explained). These modifications led to improved PEC re-
sponse, which was further improved using As-doped CdS, deposited by adding
AsCl3 to the deposition bath.
       In general, there is an optimum doping concentration, which varied from
dopant to dopant. The dopants studied here were donors (increased n-type CdS).
Increase in doping density can increase all parameters; but if too much dopant is
present, the parameters can degrade due to a narrower space charge layer (poorer
ISC due to decreased collection efficiency in the red) and increased recombination
due to impurity centers. For nanocrystalline nonannealed (or low-temperature-an-
nealed) films, where there may be no space charge layer as discussed previously,
the effect of recombination centers will still be valid—maybe even more impor-
tant due to the lack of a space charge layer to separate charges.
       Cu doping of CdS has been investigated [77]. Since the light intensity was
not specified, the cell parameters are not given in Table 9.2. The doping caused a
small increase in ISC and an equally small decrease in VOC, with no appreciable
change in efficiency, although it is arguable if these changes are significant. These
electrodes were also used with a Ag2S storage electrode in a photoelectrochemi-
cal storage cell.
       Photoelectrochemical characterization was also carried out on CdS films us-
ing different sizes of CdS nanocrystals [75]. VOC increased with decreasing crys-
tal size from 0.58V (75 nm) to 0.68 V (5 nm). Surprisingly, ISC was not dependent



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
on size, as would be expected due to the increasing bandgap and therefore low-
ered light absorption with decreasing crystal size.
       Besides evaluating photoelectrodes for use in PECs, photoelectrochemical
characterization can be used for other purposes. For example, photocurrent spec-
tra of CD CdS has been used to measure the semiconductor bandgap (as I2 vs. h ),
and agreement between the bandgap values measured by this method and by ab-
sorption spectroscopy for as-deposited and annealed films was found [78].

9.2.4 Other Photoelectrodes
A number of other CD semiconducting materials as photoelectrodes have been re-
ported, the basic PEC characteristics are given in Table 9.3. Further details can be
found in the original references.
       Most semiconductors described in this chapter gave n-type response (as al-
ready shown, for nanocrystalline semiconductors this does not necessarily mean
that the semiconductors are actually n-type but rather that the net photogenerated
hole current is to the electrolyte while that of the electrons is to the substrate). One
example of a “p-type” photoelectrode is nanocrystalline CD PbSe [83]. Figure 9.10
shows the response of such a film. The crystal size of the PbSe in this film is ca. 4
nm, with regions of crystals two to three times larger (the crystal size distribution
is bimodal—see Ref. 88). Interestingly, a single crystal of PbSe was not found to
give any photoresponse at all under the same conditions, and the current–voltage
behavior of the single crystal was essentially ohmic rather than the clearly asym-




FIG. 9.10 Chopped illumination (100 mW-cm2) current–voltage characteristics of a
nanocrystalline PbSe film deposited from a citrate/selenosulphate bath at 60°C. The elec-
trolyte is the original solution from which the film was deposited. (After Ref. 83.)



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 9.3 PEC Parameters of Miscellaneous Semiconductor Films

                                Isc           Voc                                          Illumination
Efficiency (%)               (mAcm 2)         (V)         FF         Electrolyte            (mWcm 2)       Ref.

Sb2S3
0.8 (as-dep.)                   2.0          0.44        0.37      Polyiodide                  40          79
2.0 (ann.)a                     4.0          0.48        0.44
3.9 ( STA)b                     5.6          0.54        0.52
0.008c                          0.14         0.155       0.41      Ferrocened                  100         80
Sb2Se3 (nonannealed)
0.06                            0.45         0.37       ca. 0.3    Polyiodide           80 (IR filter)     52
Bi2S3 (nonannealed)
0.12                            0.34         0.39        0.55      Polysulphide        60 (water filter)   51
Ag2S (nonannealed)
––                              0.1          0.1          ––       Polysulphide                100         81
0.088                           0.25         0.09      ca. 0.25    Polysulphide                200         82
PbSe (nonannealed)
  p-type response
                                                                                   e
0.029                           1.8          0.08        0.4                2
                                                                   Pb2 /SeSO3                  100         83
SnS2 (nonannealed)
0.006                           0.019        0.45        0.43      KCl                         100         84
(Cd, Pb) S
0.14 ann.f                     1.4            0.2        0.39      Polysulphide                75          85
0.83g no ann.                  ca. 2.8       ca. 0.5     0.3       Polysulphide                50          86
(Cd, Zn) S
  (nonannealed)h
0.04                            0.065        0.31        0.43      Polysulphide                22          87
(Cd, Hg)S (annealed
  at 320 C in air)i
0.36                            1.6          0.51        0.33      Polysulphide                75          85
a
  Annealed at 300 C in N2.
b
  With silicotungstic acid and annealed at 300 C in N2.
c
  As deposited.
d
  In dimethylsulphoxide solution.
e
  The solution from which the PbSe was deposited.
f
  Annealed at 320 C in air; 0.11 mole fraction Pb––efficiency decreased as Pb increased.
g
  For Cd0.925Pb0.075S.
h
  For Cd0.7Zn0.3S.
i
  For Cd0.82Hg0.18S.




     Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
metric diode behavior of the nanocrystalline film. Note also the p-type response of
etched CdSe (also CdS) nanocrystalline photoelectrodes discussed in Section
9.2.2.2. Photoresponse spectra of PbSe films of different crystal sizes, reflecting
the varying bandgap due to size quantization, are shown in Figure 10.6.
       Some ternary compounds have also been used as photoelectrodes, with op-
timum efficiencies reported in Table 9.3. For Cd1 xPbxS, the efficiency increased
from 0.3% (pure CdS) to 0.83% (x 0.075) and then decreased sharply to give an
even lower efficiency than for pure CdS (at x           0.1), continuing to drop more
slowly at higher Pb fractions [86], although another study on this material found
the efficiency of all alloys to be less than that of the pure CdS [85]. A similar trend
of initial increase in efficiency was found for Cd1 xZnxS, with an increase in ef-
ficiency from 0.01% (CdS) to a maximum of 0.04 (Cd0.7Zn0.3S), dropping to the
CdS value at x 0.4 and continuing to drop slowly at higher Zn values [87]. The
initial increase in efficiency for the (Cd,Pb)S film with increasing Pb content was
probably due to the lower bandgap and therefore increased absorption. The reason
for the subsequent decrease in efficiency with further increase in Pb cannot be ex-
plained so simply, although it may be noted that CD PbS is not very photoactive,
possibly due to its relatively high conductivity. For the (Cd,Zn)S electrodes, since
the bandgap of ZnS is much higher than that of CdS, the initial increase in effi-
ciency (ISC also increases) is unexpected, although a possible reason is better dop-
ing characteristics, resulting in an optimum resistivity; The series resistance of
these photoelectrodes was a minimum at x 0.3. For (Cd,Hg)S films, while the
bandgap decreased with addition of Hg, the efficiency of films with a little (0.04
mole fraction) Hg was much lower than those of pure CdS but increased with fur-
ther addition of Hg until efficiencies close to those of pure CdS were reached at
0.18 mole fraction Hg [85].

9.2.5 Coupled Photoelectrodes
Some studies involving coupled photoelectrodes of two or even three semicon-
ductors, with at least one deposited by CD, have been reported. A relatively pop-
ular subject is sensitization of nanocrystalline wide-bandgap semiconductors with
lower-bandgap semiconductors. This is due to the wide interest in dye-sensitized
TiO2 solar cells [89]. A range of semiconductors has been deposited on wide-
bandgap semiconductors using a variety of techniques, most commonly by dip-
ping the wide-bandgap semiconductor film (usually TiO2) into solutions of a
metal salt and then a chalcogenide solution, but also using electrodeposition and
chemical vapor deposition (Refs. 90 and 91 give some examples). Charge separa-
tion efficiency often is improved compared to the absorbing semiconductor itself,
usually due to electron injection into the conduction band of the large-bandgap
semiconductor, which reduces electron–hole recombination. The stability of the
semiconductor is also often improved.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       In such a coupled system, CdSe has been chemically deposited onto TiO2
(the latter prepared by both spray-painting and by screen printing) from a tri-
ethanolamine/ammonia/selenosulphate bath [92]. The films were annealed at
400°C in air. The quantum efficiencies of such films using spray-painted TiO2 and
polysulphide electrolyte, were found to be ca. 20 times higher than films of CdSe
by itself (maximum quantum efficiency was reported to be over 0.6, and light-to-
electricity conversion was 1%). Much lower values were found for the CdSe-sen-
sitized (lower-surface-area) screen-printed TiO2 films, for nonannealed films, and
in ferro/ferricyanide electrolyte. The spectral response of the CdSe/screen-printed
TiO2 electrode exhibited a pronounced response beyond the bulk bandgap of
CdSe. This subbandgap response was absent in the CdSe electrode. It also was not
apparent in the absorption spectrum of the coupled electrode, although this ab-
sorption spectrum was not clearly defined (it was obtained from diffuse re-
flectance measurements).
       CdSe was deposited on CdS (both deposited by CD) and subjected to dif-
ferent annealing temperatures [93,94]. The purpose was to see if the CdS/CdSe
heterojunction affected the PEC properties. The main effect of the coupled system
compared to only CdSe was to improve the stability of the photoelectrode in
ferro/ferricyanide electrolyte (a partially stabilizing electrolyte for CdSe). The
spectral response of the coupled system (measured, as usual, at low light intensi-
ties) was closer to pure CdS than to CdSe, although the values of ISC were similar
for both at solar intensities, indicating an illumination-dependent behavior. Depo-
sition of ZnO (by dipping in ZnAc2/methanol and annealing at 400°C) improved
the stability of the system greatly, although with a decrease in ISC, so it is difficult
to know how much of the increase in stability was due to the lower ISC and how
much to other factors (stability normally will increase at lower currents). Changes
in surface morphology, related to formation of Cd2Fe(CN)6, were measured on all
samples. Various hypotheses were put forward to explain the effects of the cou-
pled system and of the ZnO on the stability.

REFERENCES
 1.   RL Call, NK Jaber, K Seshan, JR Whyte Jr. Sol. Energy Mater. 2:373, 1980.
 2.   E Fatas, T Garcia, C Montemayor, A Medina, E Garcia Camarero, F Arjona. Mater.
      Chem. Phys. 12:121, 1985.
 3.   TL Chu, SS Chu, C Ferekides, CQ Wu, J Britt, C Wang. J. Appl. Phys. 70:7608, 1991.
 4.   TL Chu, SS Chu, N Schultz, C Wang, CQ Wu. J. Electrochem. Soc. 139:2443,
      1992.
 5.   ME Özsan, DR Johnson, M Sadeghi, D Sivapathasundaram, D Lincot, B Mokili, M
      Froment, J Vedel, LM Peter, G Goodlet, RC Walker. In: 13th ECPV Solar Energy
      Conf., Nice, France, 1995, p 2115.
 6.   PN Gibson, ME Özsan, JM Sherborne, D Lincot, P Cowache. In: 2nd World Conf. &
      Exhib. on PV Sol. Energy Conversion. Vienna, Austria, 1998, p 1089.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
 7.   J Tousková, D Kindl, L Dobiásová, J Tousek. Sol. Energy Mater. Sol. Cells 53:177,
      1998.
 8.   D Lincot, A Kampmann, B Mokili, J Vedel, R Cortes, M Froment. Appl. Phys. Lett.
      67:2355, 1995.
 9.   A Kampmann, P Cowache, B Mokili, R Ortega-Borges, D Lincot, J Vedel. In: 12th
      ECPV Solar Energy Conf., Amsterdam, the Netherlands 1994, p 664.
10.   L Kronik, L Burstein, M Leibovitch, Y Shapira, D Gal, E Moons, J Beier, G Hodes,
      D Cahen, D Hariskos, R Klenk, HW Schock. Appl. Phys. Lett. 67:1405, 1995.
11.   D Hariskos, M Powalla, N Chevaldonnet, D Lincot, A Schindler, B Dimmler. Thin
      Solid Films 387:179, 2001.
12.   D Lincot, R Ortega-Borges, J Vedel, M Ruckh, J Kessler, KO Velthaus, D Hariskos,
      H-W Schock. In: 11th European PVSEC. Montreux, Switzerland, 1992, p. 870.
13.   J Kessler, KO Velthaus, M Ruckh, R Laichinger, HW Schock, D Lincot, R Ortega, J
      Vedel. In: 6th Int. PV Sol. Energy Conf., New Delhi, India, 1992, p 1005.
14.   K Ramanathan, RN Bhattcharya, J Granata, J Webb, D Niles, MA Contreras, H Wies-
      ner, FS Hasoon, R Noufi. Proc. 26th IEEE PVSC. Anaheim, CA, 1997, p 319.
15.   K Ramanathan, H Wiesner, S Asher, D Niles, RN Bhattcharya, J Keane, MA Contr-
      eras, R Noufi. In: 2nd World Conf. & Exhib. on PV Sol. Energy Conversion: Vienna,
      Austria, 1998, p 477.
16.   T Wada, S Hayashi, Y Hashimoto, S Nishiwaki, T Sato, Z Negami, M Nishitani. In:
      2nd World Conf & Exhib. on PV Sol. Energy Conversion. Vienna, Austria, 1998, p
      403.
17.   A Kylner. J. Electrochem. Soc. 146:1816, 1999.
18.   T Nakada, A Kunioka. Appl. Phys. Lett. 74:2444, 1999.
19.   T Nakada. Thin Solid Films 361:346, 2000.
20.   A Kylner. J. Appl. Phys. 85:6858, 1999.
21.   L Kronik, U Rau, JF Guillemoles, D Braunger, HW Schock, D Cahen. Thin Solid
      Films 361:353, 2000.
22.   MJ Furlong, D Lincot, M Froment, R Cortés, AN Tiwari, M Krejci, H Zogg. In: 14th
      ECPV Sol. Energy Conf., Barcelona, Spain, 1997, p 1291.
23.   V Nadenau, D Hariskos, HW Schock, M Krejci, FJ Haug, AN Tiwari, H Zogg, G
      Kostorz. J. Appl. Phys. 85:534, 1999.
24.   R Ortega Borges, D Lincot, J Vedel. In: 11th ECPV Sol. Energy Conf., Montreux,
      Switzerland, 1992, p 862.
25.   K Kushiya, T Nii, I Sugiyama, Y Sato, Y Inamori, H Takeshita. Jpn. J. Appl. Phys.
      35:4383, 1996.
26.   T Nakada, K Furumi, K Sonoda, A Kunioka. In: 2nd World Conf. & Exhib. on PV
      Sol. Energy Conversion, Vienna, Austria, 1998, p 1173.
27.   A Ennaoui, M Weber, M Saad, W Harneit, MC Lux-Steiner, F Karg. Thin Solid Films
      361:450, 2000.
28.   A Ennaoui, M Weber, CD Lokhande, R Scheer, HJ Lewerenz. In: 2nd World Conf.
      & Exhib. on PV Sol. Energy Conversion, Vienna, Austria, 1998, p 628.
29.   A Ennaoui, M Weber, R Scheer, HJ Lewerenz. Sol. Energy Mater. Sol. Cells 54:277,
      1998.
30.   KO Velthaus, J Kessler, M Ruckh, D Hariskos, D Schmid, HW Schock. In: 11th
      ECPV Sol. Energy Conf., Montreux, Switzerland, 1992, p 842.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
31.   D Braunger, D Hariskos, T Walter, HW Schock. Sol. Energy Mater. Sol. Cells 40:97,
      1996.
32.   D Hariskos, M Ruckh, U Ruhle, T Walter, HW Schock, J Hedstrom, L Stolt. Sol. En-
      ergy Mater. Sol. Cells 41-2:345, 1996.
33.   J Bloem. Appl. Sci. Res. Section B 6:92, 1956.
34.   JL Davis, MK Norr. J. Appl. Phys. 37:1670, 1966.
35.   H Sigmund, K Berchtold. Phys. Stat. Sol. 20:255, 1967.
36.   O Savadogo, KC Mandal. Appl. Phys. Lett. 63:228, 1993.
37.   O Savadogo, KC Mandal. J. Phys. D Appl. Phys. 27:1070, 1994.
38.   KR Murali. Thin Solid Films 167:L19, 1988.
39.   JC Garg, RP Sharma, KC Sharma. Thin Solid Films 164:269, 1988.
40.   M Ortega, G Santana, A Morales-Acevedo. Solid State Electron. 44:1765, 2000.
41.   M Ristova, M Ristov. Sol. Energy Mater. Sol. Cells, 53:95, 1998.
42.   MJ Chockalingam, KN Rao, N Rangarajan, CV Suryanarayana. Phys. Status Solidi
      (a) 2:K17, 1970.
43.   O Savadogo, KC Mandal. J. Electrochem. Soc. 141:2871, 1994.
44.   C Sene, HN Cong, M Dieng, P Chartier. Mater. Res. Bull. 35:1541, 2000.
45.   JH Schön, O Schenker, B Batlogg. Thin Solid Films 385:271, 2001.
46.   K Vaccaro, HM Dauplaise, A Davis, SM Spaziani, JP Lorenzo. Appl. Phys. Lett.
      67:527, 1995.
47.   HM Dauplaise, K Vaccaro, A Davis, GO Ramseyer, JP Lorenzo. J. Appl. Phys.
      80:2873, 1996.
48.   A Davis, K Vaccaro, HM Dauplaise, WD Waters, JP Lorenzo. J. Electrochem. Soc.
      146:1046, 1999.
49.   K Vaccaro, A Davis, HM Dauplaise, SM Spaziani, EA Martin, JP Lorenzo. J. Elec-
      tron Mater. 25:603, 1996.
50.   A Fujishima, K Hashimoto, T Watanabe. TiO2 Photocatalysis: Fundamentals and Ap-
      plications; BKC Inc., Tokyo: 1999.
51.   RN Bhattacharya, P Pramanik. J. Electrochem. Soc. 129:332, 1982.
52.   RN Bhattacharya, P Pramanik. Sol. Energy Mater. 6:317, 1982.
53.   K Rajeshwar, L Thompson, P Singh, RC Kainthla, KL Chopra. J. Electrochem. Soc.
      128:1744, 1981.
54.   RA Boudreau, RD Rauh. J. Electrochem. Soc. 130:513, 1983.
55.   RC Kainthla, JF McCann, D Haneman. Sol. Energy Mater. 7:491, 1983.
56.   KC Mandal. J. Mater. Sci. Lett. 9:1203, 1990.
57.   KC Mandal, O Savadogo. J. Mater. Sci. Lett. 10:1446, 1991.
58.   O Savadogo, KC Mandal. Mater. Chem. Phys. 31:301, 1992.
59.   SS Dhumure, CD Lokhande. Sol. Energy Mater. Sol. Cells. 29:183, 1993.
60.   DR Pratt, ME Langmuir, RA Boudreau, RD Rauh. J. Electrochem. Soc. 128:1627,
      1981.
61.   G Hodes, D Cahen, J Manassen, M David. J. Electrochem. Soc. 127:2252, 1980.
62.   KC Mandal, O Savadogo. J. Mater. Sci. 27:4355, 1992.
63.   KC Mandal, KSV Santhanam. J. Mater. Sci. 26:3905, 1991.
64.   M Froment, H Cachet, H Essaaidi, G Maurin, R Cortes. Pure Appl. Chem. 69:77, 1997.
65.   L Kronik, N Bachrach-Ashkenasy, M Leibovitch, E Fefer, Y Shapira, S Gorer, G
      Hodes. J. Electrochem. Soc. 145:1748, 1998.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
66.   G Hodes, A Albu-Yaron. Proc. Electrochem. Soc. 88-14:298, 1988.
67.   G Hodes, IDJ Howell, LM Peter. In: Photochemical and Photoelectrochemical Con-
      version and Storage of Solar Energy. Tian, Z. W., Cao, Y., eds. Int. Acad. Publishers,
      Beijing, China, 1993, p 331.
68.   E Lifshitz, I Dag, I Litvin, G Hodes, S Gorer, R Reisfeld, M Zelner, H Minti. Chem.
      Phys. Lett. 188, 1998.
69.   E Lifshitz, I Dag, I Litvin, G Hodes. J. Phys. Chem. B 102:9245, 1998.
70.   B Alperson, I Rubinstein, G Hodes. Phys. Rev. B 6308:1303, 2001.
71.   CD Lokhande, SH Pawar. Mat. Res. Bull. 18:1295, 1983.
72.   CD Lokhande, SH Pawar. Sol. State Commun. 49:765, 1984.
73.   SH Pawar, LP Deshmukh. Mater. Chem. Phys. 10:83, 1984.
74.   LP Deshmukh, AB Palwe, VS Sawant. Solar Cells 28:1, 1990.
75.   KK Nanda, SN Sarangi, S Mohanty, SN Sahu. Thin Solid Films 322:21, 1998.
76.   CD Lokhande, MD Uplane, SH Pawar. Ind. J. Pure & Appl. Phys. 21:78, 1983.
77.   CD Lokhande, MD Uplane, SH Pawar. Sol. State Commun. 43:623, 1982.
78.   ME Özsan, DR Johnson, M Sadeghi, D Sivapathasundaram, G Goodlet, MJ Furlong,
      LM Peter, A Shingleton. J. Mater. Sci. Mater. Electron. 7:119, 1996.
79.   O Savadogo, KC Mandal. J. Electrochem. Soc. 139:L16, 1992.
80.   LP Deshmukh, SG Holikatti, BP Rane, BM More, PP Hankare. J. Electrochem. Soc.
      141:1779, 1994.
81.   SS Dhumure, CD Lokhande. Mater. Chem. Phys. 28:141, 1991.
82.   SS Dhumure, CD Lokhande. Sol. Energy Mater. Sol. Cells 28:159, 1992.
83.   S Gorer. PhD dissertation, Weizmann Institute of Science, Rehovot, Israel, 1996.
84.   CD Lokhande. J. Phys. D: Appl. Phys. 23:1703, 1990.
85.   M Skyllas-Kazacos, JF McCann, R Arruzza. Appl. Surf. Sci. 22/23:1091, 1985.
86.   LP Deshmukh, BM More, CB Rotti, GS Shahane. Mater. Chem. Phys. 45:145, 1996.
87.   LP Deshmukh, CB Rotti, KM Garadkar. Mater. Chem. Phys. 50:45, 1997.
88.   S Gorer, A Albu-Yaron, G Hodes. J. Phys. Chem. 99:16442, 1995.
89.   B O’Regan, M Gratzel. Nature 353:737, 1991.
90.   D Liu, PV Kamat. J. Phys. Chem. 97:10769, 1993.
91.   R Vogel, P Hoyer, H Weller. J. Phys. Chem. 98:3183, 1994.
92.   ME Rincon, O Gomez-Daza, C Corripio, A Orihuela. Thin Solid Films 389:91, 2001.
93.   ME Rincon, M Sanchez, A Olea, I Ayala, PK Nair. Sol. Energy Mater. Sol. Cells
      52:399, 1998.
94.   ME Rincon, M Sanchez, J Ruiz-Garcia. J. Electrochem. Soc. 145:3535, 1998.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
10
Nanocrystallinity and Size
Quantization in Chemical
Deposited Semiconductor Films




Chemical deposition is a low-temperature technique compared with most other
semiconductor film deposition methods. This has both advantages and disadvan-
tages. An obvious advantage is the simple processing often inherent in a low-tem-
perature technique. What may be a more important advantage, however, is that
low-temperature deposition techniques usually (although not always) result in
small crystal size. As recently as a decade ago, this would have been considered a
decided disadvantage—large crystal size was almost always desired then, e.g., for
photovoltaic cells in order to minimize grain boundary recombination. However,
with the increasing emphasis on nanostructured materials over the past decade,
this characteristic of CD films is now often considered an advantage.
       This chapter deals mainly with quantum size effects in CD nanocrystalline
films. However, another, quite separate property of such films is related to the
large percentage of atoms located on the surface of the nanocrystals of these films,
e.g. 50% for a crystal size of a few nm; this is the effect of adsorption of molec-
ular and ionic species on the nanocrystal surfaces. This aspect has been dealt with
much less than has size quantization; therefore, it constitutes only a very small part
of this chapter, mainly Section 10.2.3, which discusses the effect of adsorbed wa-
ter on CD CdSe films. Section 9.2.2.2 deals in somewhat more detail with this par-
ticular issue.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
10.1 THE QUANTUM SIZE EFFECT:
     BACKGROUND
Quantum size effects in semiconductor nanocrystals became an important field of
research in the 1980s, when a number of groups, notably those of Brus at Bell
Labs and Henglein at the Hahn Meitner Institute, published seminal papers on the
effects of the size of semiconductor colloids on their optical properties and corre-
lated crystal size with changes in electronic band structure.
       Quantum size effects in semiconductor nanocrystals have been seen before,
although the effect presumably was not realized. Early references to precipitates
formed when alkali, cyanide-containing selenide solution was added to ammoni-
acal, cyanide-containing Cd2 described them as orange-yellow when precipi-
tated in the cold and changing to red-brown when heated, and also noted that
finely divided Cd-selenide varies in color from yellow to red-brown [1–3]. As is
described later, these color changes from normal dark-brown or black CdSe are
the most obvious and visual manifestations of the quantum size effect (or size
quantization).
       The terms nanocrystals and quantum dots are often used interchangeably.
Quantum dots, as used here, are invariably nanocrystals (amorphous materials
could, in principle, also exhibit quantum size effects as long as some electronic
separation between different particles occurs) that show quantum effects, while
nanocrystals may or may not be small enough to exhibit such effects.
       Three-dimensional size quantization is due to localization of electrons and
holes in a confined volume—e.g., a semiconductor nanocrystal—resulting in a
change of the energy band structure. As the crystal size decreases below a certain
limiting size, associated with its exciton Bohr diameter, the spacing between lev-
els in the bands becomes larger [the energy structure changes from a quasi-con-
tinuum (band) to discrete, quantized levels] and the bandgap increases. This latter
change is manifested as a blue shift in the optical spectrum of semiconductor
quantum dots (the “quantum size effect”). A simple way to visualize this effect is
to consider a silicon atom (silicon being the best-known and most common semi-
conductor), with two electrons in the 3s level and two in the 3p levels (Fig. 10.1).
sp3 hybridization of the single s and three p levels leads to formation of four de-
generate (i.e., equal-energy) sp3 levels, each containing one electron. Interaction
of these levels with neighboring atoms results in splitting into bonding ( ) and an-
tibonding ( *) orbitals, with all electrons in the bonding orbitals and none in the
antibonding ones. Up to here, the situation in Figure 10.1 is shown for a single
atom. However, as additional atoms are added, each atom contributes its own or-
bitals and the localized orbitals of the single atom are gradually broadened into a
range of molecular orbitals and, eventually, when the number of atoms becomes
very large, into the familiar filled valence band and empty conduction band char-
acteristic of a semiconductor. If we consider the process from the other direction



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 10.1 Scheme showing (from left to right) how the relevant energy levels of silicon
hybridize, interact with other atoms, split in a cluster, and eventually broaden into bands.
(Adapted from Fig. 4 in LE Brus. Nouv. J. de Chem. 11:23, 1987.)


(right to left), then the bandgap increases and the levels within the bands, which
in the bulk semiconductor are extremely (almost infinitesimally) close to neigh-
boring levels, open up into discrete levels, as described previously.
       This picture is reasonably valid for covalent silicon but rather simplistic for
many of the semiconductors common in CD, which are usually mixed covalent
and ionic. However, it serves to give a feeling for size quantization. For those
readers who would prefer a more realistic interpretation for semiconductors with
considerable ionic character, it is suggested that they construct a similar scheme
for purely ionic materials and then “imagine” the required combination of ionic
and covalent character.
       There are many theoretical models to correlate the increase in semiconduc-
tor bandgap with crystal size. However, for our purposes we will show only the
original model, known as the effective mass model, since this is the easiest to un-
derstand, in spite of its limited accuracy.
       The effective mass model is based on the energy of the lowest-energy par-
ticle-in-a-box configuration, taking into account that the relevant mass term is
given by a reduced effective mass, , where is given by
       1     1       1
                                                                                     (10.1)
             me      mh
The effective masses of electrons (me) and holes (mh) represent the masses that
these charges appear to have when moving in the solid rather than in free space,
and these vary from material to material. (In the size quantized regime, they can
also vary with crystal size, particularly for small quantum dots, hence the limita-
tions of the effective mass model).



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
      The increase in bandgap, E, of a semiconductor due to size quantization is
then given by
               2   2
                        1.79e2
        E          2                                                             (10.2)
             2 R           R
where the first term on the right-hand side is the localization energy [the particle-
in-a-box energy of the charges in a box (more correctly, in this case, a sphere) of
radius R, modified by the reduced effective mass term) and the second term rep-
resents a reduction in the energy increase due to coulombic interaction between
the electron and the hole and is a function of the dielectric constant of the semi-
conductor, . The increase in bandgap is inversely dependent on both the reduced
effective mass and on the square of the radius. The bandgap should therefore in-
crease as a parabolic function of the decrease in size. In practice, the rate of in-
crease is less than this, and an exponent considerably lower than 2 gives a better
fit of the bandgap increase with decrease in crystal size.
        An obvious importance of this size quantization is that a single semicon-
ductor can possess a range (sometimes a wide range) of bandgaps, which can be
controlled if the semiconductor crystal size is controlled. This “bandgap tailoring”
allows (ideally) control of all properties that depend on the bandgap. The most ob-
vious is the optical transmission (absorption) spectrum, and this is the property
most often measured in quantifying size quantization, since the bandgap can be es-
timated from this spectrum.
        In this chapter, size quantization effects in CD films are described. Since the
majority of reports on size quantization in CD films mention the effect but do not
go into detail on this aspect, as with many other chapters in this book, it will be
more efficient to tabulate the relevant literature and to deal with individual stud-
ies that provide additional results of interest outside of what is included in the table
or require further discussion. CdSe and PbSe will be dealt with in a more inte-
grated manner, since films of these materials, in particular CdSe, have been the
most intensively studied from the viewpoint of their nanocrystallinity and quan-
tum size effects.
        Some words of caution in interpreting optical transmission (absorption)
spectra (see also Sec. 1.4.2). Since the energy structure of nanocrystals in the quan-
tum size domain is more like an atomic structure, with separate levels, than the band
structure of bulk semiconductors, the derivation of bandgap from the absorption
spectra using the ( h )n vs. h plot, which is based on the density of states (band
structure) of the bulk semiconductor, is not really valid. Nonetheless, it is often
used to give an approximation. Since there is always a distribution of crystal sizes
(and therefore of bandgaps) in these size-quantized films, there is in any case no
single “correct” bandgap, and any value measured will be an approximation.
        Another area where nanocrystallinity is important is that of surface effects
due to the high real surface areas of such films (often tens of percent of all atoms



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
are located at a surface). Since these films are invariably porous to a greater or
lesser extent, much of this surface is accessible to modification and treatment by
liquids and gases. This aspect has been less dealt with for CD films than quantum
size effects, but some examples do exist and are briefly discussed in this chapter,
with references to other sections of this book where they are of specific relevance.
      The factors that influence crystal size in CD films are discussed in Section
10.2 on CdSe, since they have been most studied for that material. However, the
principles involved are general for all semiconductors.


10.2 CdSe
10.2.1 Historical Background
This chapter is the one closest to the interests of the author. These interests, both
in chemical deposition and in size quantization, were a direct result of a single
serendipitous observation. As a historical and personal interlude, this “experi-
ment,” involving CD CdSe, will be briefly described and its rationale explained
(in reverse order).
        Accepting an invitation to spend some time in Campinas, Brazil, I planned
to utilize my experience in electrodeposition of CdSe films. Shortly before this
trip, a paper appeared describing a method of electrodeposition of CdSe, based on
the use of selenosulphate instead of the commonly used SeO2 (Ref. 4, which itself
was based on an earlier study [5]). The trip to Campinas and the resulting ability
to spend a lot of time in the lab seemed a good opportunity to try this method. The
method proved to be simple to reproduce. More relevantly, however, in keeping
with my overall philosophy on life, instead of clearing away the “finished” ex-
periment, the beaker containing the electrodeposition solution was left sitting over
the ensuing weekend on the laboratory table to be taken care of at a more suitable
(i.e., later) time. On returning to the lab after the weekend, I found the beaker in
question to contain a bright red precipitate and also a similarly colored, transpar-
ent film on the inside of the glass walls. My first thought was that this was ele-
mental Se (which is usually bright red in the freshly precipitated state), based on
my understanding that selenosulphate is not very stable and can readily form ele-
mental Se under certain conditions. A simple chemical test (treatment with Na2S
solution, in which Se dissolves) showed that the precipitate was not Se. To make
a fairly long story short, the next two ideas (occurring more or less at the same
time if my memory does not fail me) were either that I had discovered a new amor-
phous form of CdSe or that the CdSe was size quantized, and some further exper-
iments were all that were needed to verify the latter hypothesis.
        Had I been more familiar with chemical deposition at that time, I might well
have ignored the “red” CdSe as something obvious; an earlier study on CD CdSe
had noted that the as-deposited films were red [6].



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Just to complete this history, it should be mentioned that the weekend in
question occurred shortly after I had spent several days in Rio de Janeiro during
Carnival. I do not make any (overt) claims that this experience affected my abil-
ity to interpret the experiment in any way.
       The results of this experiment and the subsequent investigation were pub-
lished in Ref. 7. At the time, it was not obvious that quantum size effects would
be seen in strongly aggregated nanocrystals. Subsequently, this was found to oc-
cur quite commonly (as seen in this chapter). While some degree of electronic iso-
lation between crystals is needed for size quantization to be exhibited, this isola-
tion need by no means be absolute. CdSe is a particularly attractive material to
show the phenomenon of size quantization, since its color can vary from very deep
red (even black in powder form) to yellow (almost white for very tiny crystals).
To give a feeling for the size dependence of the color, crystals of 6 nm will be red,
4 nm orange, and 3 nm yellow, a wide range in color for a small range in crys-
tal size. This variation in color can be seen from the transmission spectra shown
in Figure 2.9. The spectral changes show how the absorption onset (equal to the
start of the transmission decrease) moves to the red as the deposition temperature
increases and also as the mechanism changes from a cluster mechanism to an ion-
by-ion one (see later). The bandgap range, from ca. 2.3 eV to 1.8 eV, parallels a
change in crystal size from ca. 3 nm to over 12 nm (the rightmost spectrum is for
a film with crystal size ca. 20 nm, but this spectrum is reached by the time the crys-
tal size is close to 12 nm, where it can be considered to be bulk from the point of
view of size quantization).

10.2.2 What Determines the Size of Nanocrystals
       in CD Films?
The factors that determine crystal size, a topic of particular relevance to this chap-
ter, have been discussed to some extent in Section 3.4. There are two main factors
that generally affect crystal size for any particular material: the deposition mech-
anism and the deposition temperature. The hydroxide cluster mechanism is ex-
pected to give a crystal size similar to that of the original hydroxide cluster (with
some growth possible as deposition proceeds). That size depends mainly on tem-
perature, both because higher temperatures allow more grain growth and, possi-
bly more important, lower temperatures kinetically stabilize very small nuclei in
solution that are thermodynamically unstable. For example, in the hydroxide clus-
ter mechanism, where crystal size is believed to be controlled mainly by the size
of the Cd(OH)2 colloids, the relevant equilibria are

      Cd2      2OH D Cd(OH)2                                                   (10.3)
A number n of these molecules can form a cluster, [Cd(OH)2]n. This cluster can




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
continue to grow by collecting a variety of solution species, one possibility
being
             [Cd(OH)2]n      2OH D [Cd(OH)2]n.2OH                                  (10.4)

      [Cd(OH)2]n.2OH           Cd2 D [Cd(OH)2]n        1                           (10.5)
This may continue until eventually the cluster is large enough to be thermody-
namically stable (i.e., will not redissolve). However, if the cluster is smaller than
the critical nucleus size, then there is the possibility that the nucleus will redis-
solve. The lifetime of the nucleus will then depend on its size and also on the tem-
perature; lower temperatures will slow the redissolution step. Thus lower temper-
ature increases the chance that a subcritical nucleus will eventually grow to a
stable size rather than redissolve. This kinetic stabilization of small nuclei results
in a greater total density of nuclei and therefore smaller crystal size, since the to-
tal quantities of reactants are fixed.
       For the ion-by-ion reaction, nucleation is generally slower and the density
of nuclei smaller. Additionally, growth occurs (ideally) only at a solid surface;
therefore nucleation is confined to two dimensions, in contrast to three dimensions
for the cluster mechanism. The crystal growth may terminate when adjacent crys-
tals touch each other or by some other termination mechanism, e.g., adsorption of
a surface-active species. These factors should be valid regardless of whether the
mechanism proceeds via free chalcogenide ions or by a complex-decomposition
mechanism.
       The effect of temperature and mechanism on the optical spectra, through the
crystal size, is clearly seen in Figure 2.9. In particular, the difference in crystal size
between the two rightmost spectra, both deposited at 80°C but one through the
cluster mechanism and the other (HC) through the ion-by-ion mechanism, is rela-
tively large: 8.5 nm for the film deposited by the cluster mechanism and 20 nm for
that deposited by the ion-by-ion mechanism. While the effect of temperature is
gradual, that of mechanism is sudden. It is determined by the conditions that sep-
arate the formation of metal hydroxide colloids from a solution with no metal hy-
droxide phase. Decrease in the complex:metal concentration ratio and increase in
temperature and pH will all favor hydroxide formation. This sudden transition on
varying the complex:metal ratio is shown, for two different solution temperatures,
in Figure 10.2. The spectra are independent of the NTA:Cd ratio (NTA, nitrilotri-
acetate, the complex used) until the transition between hydroxide-containing and
hydroxide-free solutions is reached, whereupon they suddenly undergo a red shift
(increase in crystal size, decrease in bandgap) and then no further change as the
NTA:Cd ratio increases further. The crystal sizes in these films were also shown,
by XRD, to change only at the NTA:Cd ratio where the spectrum changes and
the experimental results agreed with thermodynamic calculations on the region of




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 10.2 Optical transmission spectra of CdSe films deposited at 10°C (top) and 80°C
(bottom) with varying NTA:Cd molar ratios (shown in the figures). The bandgaps esti-
mated from the spectra are indicated by the thin vertical lines. (Adapted from Ref. 8.)


existence of Cd(OH)2 as a function of temperature, pH, and solution composition
[8].
       Illumination, by light that is absorbed by the growing semiconductor crys-
tals, has been shown to increase the crystal size somewhat, seen as a red shift in
the optical spectrum and decrease in bandgap by as much as 0.2 eV [9–11] (see
Fig 4.3). This is probably due to photoelectrochemical reactions taking place at
the crystal surface. The chemical deposition solution can also be used to elec-
trodeposit CdSe. Electrons (either from an external source, as in the case of elec-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
trodeposition, or generated by illumination during CD) can reduce selenosulphate
to selenide, which reacts with Cd ions, causing (photo)electrodeposited CdSe to
form on the crystals and thereby increasing the crystal size. Only light that is ab-
sorbed by the semiconductor can cause this effect, as expected based on the pho-
toelectrochemical mechanism. In addition, a certain minimum intensity is needed
to cause a measurable red shift, but the effect saturates at high light intensities. At
low intensities, recombination was assumed to remove the photogenerated elec-
tron before it had time to reduce selenosulphate, hence the threshold. As intensity
increases, there is an increasing likelihood of photoelectrochemical reduction. The
saturation was explained by assuming that only one electron/hole pair was effec-
tive; further increase in intensity has no further effect [10,11]
       Since size-quantized CdSe can undergo visible color changes with change
in crystal size, this effect of illumination can clearly be used to form patterns on
the film by illuminating the film through a mask. We have even observed inter-
ference fringes at the edges of these patterns, alternating between orange (nonil-
luminated) and brown-red (illuminated), corresponding to destructive and con-
structive interference, respectively, at the pattern edges.
       It is a general observation when quantum size effects are observed in CD
films that the blue shift is reduced somewhat as the film thickness is increased, and
this has been shown clearly for CdSe [9,11–13]. From absorption spectra, a dif-
ference of ca. 0.08 eV (a crystal size difference of 0.26 nm for a crystal size of ca.
4 nm) was shown to occur between thin and thick films (growth time between 10
and 190 hr) [11]. In another study, using the same basic deposition solution
(NTA/selenosulphate), red shifts in the photoluminescence spectra could be cor-
related with a change in crystal size from ca. 4.5 nm (2.1 eV) to ca. 8 nm (1.85 eV)
[12].* The widths of the photoluminescence peaks increased as the deposition
time (and therefore crystal size) increased. This was explained as an increase in
size distribution of the crystals as deposition proceeded. Since the change in
bandgap with crystal size increases as crystal size decreases, wider peaks in the
photoluminescence spectra, if due to increase in size distribution, means that the
size distribution increases greatly with increasing deposition time. If no change in
the size distribution were to occur, the peak widths should actually decrease with
increase in average crystal size. While photoluminescence peak widths of size-
quantized films may certainly be influenced by size distribution, other factors, in
particular recombination from surface sites with various spatial separation, will
also affect the width. A microscopic study of such films would be the most de-
pendable way to measure size distribution.



* The values of crystal size reported in this study varied between 6.6 and 10.6 nm. These values were
based on the effective mass approximation, which overestimates the bandgap increase with decreasing
crystal size, particularly for small sizes.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       Increase in size distribution with increasing film thickness (or deposition
time) is expected for a number of reasons. One is the obvious one that, since film
growth can involve both crystal growth and new nucleation, the chances of any
particular crystal growing increases with time (film thickness), due to deposition
of new material on the crystal. Also, as deposition proceeds, the solution compo-
sition changes, and this can lead to changes in crystal size. There are two main
causes of this. One is that, while the Cd concentration decreases, that of the com-
plex does not change (this may not be the case where a volatile complexant, like
ammonia, is used in an open bath). Therefore the complex:Cd ratio increases dur-
ing deposition, and at some point the mechanism may change from hydroxide
cluster to ion-by-ion; the latter normally gives larger crystals and may also occur
on already existing crystals. The second cause is that, even assuming no change in
deposition mechanism, crystal size grows slightly with decrease in reactant con-
centrations; reducing the concentration of all reactants to one-half the original
concentration resulted in a bandgap increase of between 0.05 and 0.10 eV, corre-
sponding to a size increase of ca. 10% [9]. This can be explained based on the con-
cept of kinetic stabilization of small nucleii. As described earlier, small nucleii are
stabilized at lower temperatures, thus providing a greater chance of growth to a
size where the crystal will be thermodynamically stable and resulting in a smaller
crystal size. However, the lower the concentration of reactants (selenosulphate
and free Cd2 , the latter of which will decrease due to both decrease in total Cd
concentration and increase in complex:Cd ratio), the slower will be this growth
and therefore, as with increased temperature, the greater likelihood of redissolu-
tion of the small nuclei, resulting in a larger final crystal size. Of course, this con-
centration effect is important not only in the context of varying film thickness (de-
position time) but also as a means for further control of crystal size.
       Thicker films with crystal size more characteristic of thin films could be
formed by depositing on an existing thin film from a new solution and repeating
this to the desired thickness. This suggests that the main reason for increase in
crystal size with continuing deposition is not simply because of deposition on al-
ready-deposited crystals, but because of changes in the composition of the depo-
sition solution. This is also borne out by a comparison of modulated electrotrans-
mission (ET) and electroreflection (ER) spectroscopies [where modulation of the
potential of the film in an electrolyte results in corresponding changes in the ab-
sorption and reflection) of a CD CdSe film (Fig. 10.3)]. The bandgap of the film
is ca. 2.1 eV, but the ET spectrum is broader and shifted mainly to the high-energy
side compared to the ER spectrum. Since the ER spectrum samples the near-sur-
face region of the film and the ET spectrum the total film, this difference suggests
that the film is composed of smaller crystals close to the substrate and larger ones
toward the surface.
       The effect of film thickness was very evident for films deposited from baths
based on N,N-dimethylselenourea and complexed with both ammonia and either



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 10.3 Modulated electrotransmission (ET) and electroreflectance (ER) spectra of a
CD CdSe film deposited from an NTA/selenosulphate bath at 30°C. The experiments were
carried out in an electrolyte containing sulphide and sulphite (the latter to prevent forma-
tion of colored polysulphide) at a pH of 10 (buffered with NaH2PO4).




citrate or tartrate [14]. Bandgaps estimated from the transmission spectra varied
by at least 0.2 eV over a thickness range from under 100 nm to a few hundred
nanometers. (Similar thickness effects were seen using photoluminescence in sim-
ilarly prepared films [13].) Films from the tartrate bath were deposited more
rapidly than those from the citrate bath [15], and the higher bandgap found for the
citrate-based films in this study was probably due to the fact that those films were
thinner than the tartrate ones. For comparable film thickness, the bandgaps were
similar (derived from a comparison of deposition rate [15] and spectra as a func-
tion of deposition time [14]). No XRD patterns were found for these films; there-
fore crystal size was not directly measured. It was noted that the deposition solu-
tion color changed during deposition from colorless through turbid yellow,
orange, to orange-red. This color change is typical for low-temperature, hydrox-
ide-cluster-mechanism CD of CdSe in general when carried out at relatively low
temperatures.
       Annealed films deposited from a N,N-dimethylselenourea/citrate/ammonia
bath were shown to exhibit a (0001) XRD reflection at 2         13°, a reflection nor-
mally forbidden in hexagonal CdSe [the (0002) reflection is the one normally
seen] [13]. This was explained by a breaking of the selection rules due to the small
crystal size. Interestingly, this peak was very weak in thin films and prominent in



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
thicker ones. A broad Raman band at 250 cm 1, which is not observed normally
in CdSe and disappeared after annealing, was also observed in these films and at-
tributed to a surface optical mode.

10.2.3 Photoluminescence (See also Sec. 4.2.7.1)
Photoluminescence was mentioned earlier in connection with studies in change of
crystal size with film thickness. On a more general note, a number of photolumi-
nescence studies have been reported on size-quantized CD CdSe films. Spectra
showing both bandgap luminescence (this is probably not true bandgap but re-
combination from shallow surface states) [7,12,16] and dominant deep-trap lumi-
nescence [10,11,17,18] have been reported. One study, using dimethylselenourea
instead of the more common selenosulphate, found both “bandgap” luminescence
(which red-shifted on increasing film thickness, explained by increasing crystal
size) and a lower-energy peak at ca. 1.75 eV, attributed to larger, weakly quantum-
confined crystals. It is very possible that this low-energy peak arises from surface
states, since there was no evidence of a bimodal size distribution that would lead
to two separate peaks
        The role of water adsorbed on the surface of these nanocrystals in passivat-
ing surface states was discussed in Section 9.2.2.2. This is seen in luminescence by
change in the spectrum from deep-trap-dominated in a dry ambient to near-
bandgap-dominated in a humid ambient [17]. A study of the deep-trap lumines-
cence showed that the luminescence originated from recombination of trapped
charges [17] (see also surface photovoltage spectroscopy measurements in Ref.
18). From a consideration of the optically detected magnetic resonance (ODMR)
signals, it was shown that the recombination occurred from sites of low symmetry,
i.e., at the surface of the nanocrystals [19]. However, based on time-resolved pho-
toluminescence and transient absorption measurements, even the near-bandgap lu-
minescence was believed to result from shallow-trapped carriers. The dynamics of
the photogenerated charge trapping for these films showed, using transient ab-
sorption measurements, fast (subpicosecond) electron trapping to shallow surface
states [10,20] and slower, but still relatively fast (ca. 50 ps) emptying of these shal-
low traps, either to the ground state or, more likely, to deeper traps [20]. Even tak-
ing into account this strong effect of water adsorption, in the author’s experience
the luminescence from these CdSe films can be highly variable, not only in whether
it is dominated by near-bandgap or deep-trap recombination but, maybe even more
so, in the intensity of the luminescence, varying from relatively strong lumines-
cence (visible to the eye) to no measurable signal at all. While the reason for this
large variation is unknown, it is reasonable to assume that it is related to the nature
of the surface of the individual nanocrystals.
        A collection of studies on size quantization in CD CdSe films, together with
some relevant data, is presented in Table 10.1.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
TABLE 10.1 Possible Quantum Size Effects in CD CdSe

Max. Eg            Min. crystal
(eV)                size (nm)                     Miscellaneous                          Refs.

   2.3             ca. 3.5                 NTA/selenosulphate                       7, 8, 9, 17, 19
   2.3                                     Citrate or tartrate/ammonia/             14
                                             DMSea
   2.1                                     Triethanolamine/ammonia/                 21
                                             selenosulphate
    2.28                                   Citrate/ammonia/DMSe                     15
    2.06                                   Tartrate/ammonia/DMSe                    15
    2.06                                   Citrate/selenosulphate                   15
ca. 2.1            ca. 4.5 (6.5)b          NTA/selenosulphate                       12
ca. 2.05c              6.4d                NTA/selenosulphate                       20
 ca. 2.3                                   Citrate/ammonia/DMSe                     13
    2.3                4.25                NTA/selenosulphate                       11
    2.28               4.34                NTA/selenosulphate                       10
Bulk bandgap 1.73 eV (wurtzite), ca. 1.8 eV (zincblende). All bandgap values given in this chapter
are room-temperature values.
a
  N,N-Dimethylselenourea.
b
  This reported size is probably overestimated; simple effective mass approximation used in its esti-
mation.
c
  The optical absorption spectrum consists of a reasonably sharp onset corresponding to a bandgap of
ca. 2.05 eV and a broad tail to longer wavelengths (onset of between 1.8 and 1.9 eV).
d
  The films in this study were thick (ca 1 m), and therefore the average crystal size may be expected
to be larger than most films, which are usually in the range of 100–200 nm thick.



10.2.4 Annealed Films
It appears that most reported CD CdSe films are size-quantized, with crystal sizes
of 10 nm. There are some exceptions. Films deposited via an ion-by-ion mech-
anism at high temperature possess larger crystal size and show no size effects [8].
Films deposited from an ammonia/selenosulphate bath were reported with a
bandgap, measured from the absorption spectrum, typical of bulk CdSe (ca. 1.74
eV) [22]. The CdSe in this study was grown at 80°C from a solution containing 48
mM Cd and 2.1 M NH3. Taking into account the additional complexing power of
the selenosulphate, such a solution may be close to the transition between a hy-
droxide mechanism and an ion-by-ion one.
      One technique commonly used to illustrate size quantization in these films
is annealing them to increase crystal size. This results in a gradual red shift in the
spectra until eventually “bulk” CdSe (ca. 11 nm in size) is reached, after which no
further shift, at least not one due to size, is seen (see Refs. 7, 13, 14, 15, and 23).
The increase in crystal size with annealing depends on temperature and time of an-



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
FIG. 10.4 TEM micrographs of a nanocrystalline CdSe film as deposited (upper left)
and after air-annealing for 20 mn at increasing temperatures up to 500°C (lower right). The
50-nm scale is the same for all micrographs except for the 450°C and 500°C ones, which
are marked by the 100-nm scale. (From S. Gorer and G. Hodes, unpublished results.)



nealing as well as on the material annealed and on the annealing atmosphere. For
CdSe and CdS, which are usually annealed in air or sometimes in an inert atmo-
sphere, as a rule of thumb, increase in crystal size is slow up to a temperature of
ca. 300°C and increases greatly at a temperature somewhere between 300 and
400°C (with further growth at higher temperatures), together with a phase trans-
formation (if the original film is sphalerite) to wurtzite structure. This can be seen
in Figure 10.4, which shows a CdSe CD film as deposited and after sequential an-
nealing treatments. (See also Fig. 1a in Ref. 24, which shows XRD spectra of es-
sentially the same process). Although the size increase at low annealing tempera-
tures is small, the red spectral shift occurs mainly in this region, since only a small
increase in crystal size is necessary to obtain an appreciable red shift.

10.3 CdS
Films described in Table 10.2 are assumed to be deposited using a standard bath
(Cd2 /ammonia/thiourea/relatively high temperature) unless described otherwise.



Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
       The first report of size quantization in a CD semiconductor film was in 1981
for very thin films of CdS deposited from CdSO4 and thioacetamide [25]. Quan-
tum shifts were measured by photoluminescence (absorption spectra in the
bandgap region could not be measured, probably because the films were ex-
tremely thin). Compared to an exciton luminescence peak at ca. 507 nm (2.45 eV)
for bulk material, a peak at 468 nm (2.65 eV) was obtained for ultrathin films (par-
ticles of ca. 10 nm separated from each other by ca. 20 nm) and between 470 and
495 nm (2.64–2.51 eV) for thicker films (thickness not defined). The peak of the
ultrathin films did not change if the film was heated at 800°C in H2S, a treatment
that would almost certainly result in crystal growth well beyond the size quanti-
zation limit for thicker films. The 10-nm particle size is at least twice the size
needed to see the observed blue shift in the spectra, suggesting that either the par-
ticles were composed of a few aggregated crystals or the vertical dimensions of
the particles were much smaller than the measured lateral ones.
       Blue shifts in very thin films ( 3 nm average thickness) were also more re-
cently measured for CD CdS films deposited in an ultrasonic bath [34].
       In general, very thin films would be expected to exhibit size quantization if
the film thickness is less than the Bohr diameter, since quantization will occur in
at least one dimension. However, in most cases, such very thin CD films can be



TABLE 10.2 Possible Quantum Size Effects in CD CdS

Max. Eg                    Min. crystal
(eV)                        size (nm)             Miscellaneous                Refs.

   2.66                                    From thioacetamide bath,              25
                                             ultrathin layer
   2.63                                    CdI2                                  26
   2.5                                     Eg larger with increase in            27
                                             thiourea concentration
   2.64                    See text        Citrate/ammonia bath                  28
                           ca. 5           NTA bath; cluster mechanism            8
ca. 2.7; see text          4.8             Standard bath                         29
    2.60                     20 nm         65–85°C                               30
    2.5                                    Cd acetate                            31
    2.69                   4.1             Electrochemical-induced and           32
                                             thiol capping agent
   2.58                                    Citrate/ammonia bath                  33
   3.2                                     Ultrathin layer (2–3 nm),             34
                                             deposited in ultrasonic bath
                           5.0                           —                       35

Bulk bandgap    2.43 eV.




Copyright © 2002 by Marcel Dekker, Inc. All Rights Reserved.
expected to be composed of islands, as occurs in the films described earlier, and
size quantization is not then expected if the individual crystal size is greater than
the Bohr diameter.
       In Ref. 26, optical transmission spectra of films deposited using CdI2 in a
NH3/NH4 /thiourea bath were substantially blue-shifted (to 2.63 eV) from the nor-
mal CdS absorption. While no crystal size was given, from the XRD spectrum
given in the paper it appears that the crystal size was considerably larger than 10
nm, and therefore the cause of the blue shift is not clear. No such shift was seen
for films deposited under similar conditions but using CdCl2 instead of the iodide.
       In Ref. 27, small increases in bandgap were found when high thiourea con-
centrations were used. This was explained, in general terms, as a decrease in grain
size with increase in deposition rate.
       In Ref. 28, using a citrate/ammonia bath, clear increases in bandgap (up to
2.64 eV) were observed. A crystal size of ca. 4 nm was calculated from the XRD
data, and this size would be consistent with the observed optical spectra. Most of
the XRD peaks, while riding on a strong and noisy background, do appear to be
much sharper than would be expected for a 4-nm size. On the other hand, one
higher-angle peak [(11.0)] is considerably wider, although it remains wide even
after annealing at 450°C for one hour, a treatment that normally will cause crystal
growth into the size range of hundreds of nanometers. The main differences be-
tween this bath and many standard baths are the use of citrate together with am-
monia (citrate is a weaker complexant for Cd than is ammonia, although it might
act as a surface blocking agent through adsorbed carboxylate groups) and the high
thiourea:Cd ratio (ca. 17).
       In Ref. 8, crystals ca. 5 nm in size were deposited from a nitrilotriacetate
(NTA)-complexed bath (no ammonia) at 40°C (a lower temperature than most
CdS depositions). The composition of the bath was such that Cd(OH)2 was pre-
sent as a colloidal phase (cluster mechanism–see Chap. 3). Under conditions
where no hydroxide phase was present and the reaction proceeded via an ion-by-
ion mechanism, much larger crystals ( 70 nm) and a red-shifted spectrum were
found. See Section 10.2.2 for more detail on the dependence of crystal size on the
deposition mechanism.
       In Ref. 29, bandgaps up to 3 eV were reported. From the optical spectra, it
is more probable that the maximum bandgap value is ca. 2.7 eV. The bandgap de-
creased strongly with increase in film thickness up to 250 nm, at which stage the
bandgap was close to the bulk value. While most films were deposited at a pH of
11.7, it was noted that smaller particle size was obtained at pH 12 and larger ones
at pH 9.
       In Ref. 30, small increases in bandgap (up to a bandgap of 2.6 eV) were ob-
tained for CdS deposited in the presence of a