Multiple airgap membrane structures and buried mushroom stripe

Document Sample
Multiple airgap membrane structures and buried mushroom stripe Powered By Docstoc
					   Invited contribution to the review book 'Vacuum Science & Technology', Volume 'Three-dimensional structurisation technology
   for semiconductor based heterostructure devices and circuitry', 2002, Research Signpost, Kerala, India

                 Multiple airgap membrane structures
                 and buried mushroom stripe lasers –
            Fabrication by dry and subsequent wet etching
                                              J. Daleiden and H. Hillmer

                             Institute of Microstructure Technologies and Analytics,
                              Heinrich-Plett-Strasse 40, D-34132 Kassel, Germany,

ABSTRACT                                                            INTRODUCTION

Real three-dimensional (3D) material structures                     During the last years miniaturization of micro-
enable enormous perspectives in the functionality                   electronic circuits has increased exponentially as a
of advanced electronic and optoelectronic III/V                     function of time. G. Moore based this trend in
semiconductor devices. We review typical fabrica-                   1965 on just 5 data pairs. In a semi-logarithmic
tion philosophies and fundamental technological                     plot this trend is linear and well-known as Moore’s
processes of surface-micromachined III/V semi-                      law, which still holds in our days. This trend can
conductor devices. Main application fields of                       be formulated as a function of time for different
compound semiconductor based micromachining                         properties: such as the number of transistor func-
are demarked from those of Si systems. Consider-                    tions per chip, or the degree of miniaturization, or
ing fabrication technology, the general principles                  the calculation speed, or the storage capacity, or it
can be reduced to three fundamental process steps:                  can be simply formulated in a single covering ex-
deposition of a layered heterostructure on a sub-                   pression: the complexity. Concerning the rele-
strate, vertical structurization and horizontal un-                 vance of this trend as a measure for the develop-
dercut by selectively removing sacrificial layers.                  ment of our information age, it is interesting to
Very useful quality control elements for precise                    notice that the complexity of micro-optics and mi-
process control are presented. The basic principles                 crosystem technology has been observed to grow
are applied and illustrated in detail presenting two                faster, i.e. superlinear, in a semi-logarithmic dia-
selected optoelectronic examples: (i) fabrication of                gram. Compared to micro-electronics and micro-
bent waveguide buried mushroom stripe lasers and                    optics, microsystem technology is a relatively new
(ii) vertical cavity based tunable single or multi-                 field, describing a combination of at least two of
membrane devices including airgaps. Finally the                     the three disciplines micro-electronics, micro-
potential of micromachined III/V compound semi-                     optics and micro-mechanics.
conductor devices are underlined by some selected
experimental results.                                               Our chapter is devoted to the combination of all
                                                                    the   three    disciplines:   micro-opto-electro-
                                                                    mechanical systems (MOEMS). In microsystem
technology, Si is the by far dominating material        branes and tooth wheels allow precisely controlled
class. However, it cannot be efficiently used in        motions and deformations. Complicated combina-
monolithic systems including active light emitting      tions of these basic elements have been shown in
devices, since the radiative recombination rate for     Si and allow very complex motions (e.g. Si comb
pure Si is by four orders of magnitude smaller than     actuator driven external cavity devices [1,2], digi-
the rate for III/V compound semiconductor materi-       tal micromirror devices [3], high duty factor mi-
als such as GaAs, InP and related ternary, quater-      cromechanical pendants of oscillating LC circuits
nary and pentanary compounds.                           [4], various free beam optical components [5-13],
                                                        microgrippers [14-16], cantilever tips [17] and nxn
Favorizing monolithic or hybride systems for            matrix switches [18-20]. These structures play key
MOEMS? This interesting question is still under         roles for micromachined Si sensors and actuators.
intense discussion. Several concepts based on sur-      The success of Si is mainly due to a few reasons:
face micromachining, bulk micromachining and            crystalline Si has excellent mechanical properties
hybride systems have been successfully demon-           and is the cheapest semiconductor material, poly-
strated in the past. Our chapter is devoted to fun-     Si and natural Si oxide combine fantastic material
damental fabrication principles of surface micro-       properties and extremely low fabrication costs. In
machining for monolithic systems using III/V            comparison, the technological difficulties are by
compound semiconductors. We implement the               far extended for III/V or II/VI semiconductor sys-
optoelectronic components within a single semi-         tems and successful pendants of poly-Si and natu-
conductor material class and use batch processing       ral Si oxide are not available for these compounds.
to avoid final micro mounting of individually im-       Therefore, compound semiconductor micro-
plemented parts. We demonstrate fundamental             machining is more expensive and by far less com-
process principles using two examples and two           plex at the moment as Si based micromachining.
heterostructure systems: GaInAsP and AlGaInAs,          However, micromachining using compound semi-
both on InP substrates.                                 conductors has consequently and logically devel-
                                                        oped into another very successful direction making
In our chapter, the basic principles are applied and    full use of the direct band structure and, thus, the
illustrated in detail presenting two selected opto-     efficient light generation in III/V or II/VI semi-
electronic examples: first, constricted ridge mesa      conductor heterostructures. In our opinion, direct
structures and second, micromachined membrane           compound semiconductor systems will always be
based tunable filters. These photonic devices are       by far superior to Si concerning radiative recombi-
very attractive for applications in ultra-high capac-   nation efficiency. This will even hold, although the
ity 1.55 µm optical communication systems based         radiative recombination efficiency can be en-
on dense wavelength division multiplex (DWDM).          hanced in element semiconductors using: intrava-
                                                        lence band transitions, Brillouin zone foulding in
                                                        Si/Ge superlattices, quantum cascades, porous Si,
1. FOUNDATIONS OF SURFACE                               quantum size effects in Si nano crystals embedded
   MICROMACHINING IN III/V                              in SiO2 [21] or dislocation loops around Boron,
   SEMICONDUCTOR                                        ionimplanted in Si [22]. III/V micromachining has
   HETEROSTRUCTURES                                     been successfully applied, e.g., in undercut ridge
                                                        mesa structures for high-speed edge emitting lasers
                                                        [23] or bent waveguide lasers [24-29]. Based on
                                                        cantilevers and suspended membranes tunable mi-
1.1 Material and geometric aspects
                                                        crocavity devices such as filters [30-44], vertical
                                                        cavity surface emitting lasers (VCSEL’s) [36,45-
Mesa, trenches, grooves, nozzles, pyramides,
                                                        51], LEDs [52] and photodetectors [53-56] using
curved surfaces as well as undercut ridges and
                                                        electrostatic or thermal actuation have been im-
stripes are static basic elements in micromachin-
                                                        plemented. Other material systems and techniques
ing. Further structures such as cantilevers, mem-
                                                        involving extremely short-wavelength light

sources [57,58] are also very attractive, however,     ZnS, InAs and others may become important.
not subject of our chapter.                            Typical material systems for layers B and C are
Suspended membrane structures and cantilevers          - AlzGa1-zAs, Ga1-xInxAsyP1-y,
are identified to be very attractive in micro system   - AlzGa1-x-zInxAs, Ga1-xInxAs1-vNv
technology, however, this rises questions of me-       - AlzGa1-x-zInxAsyP1-y, or AlzGa1-x-zInxAsuSb1-u .
chanical stability. Studying scaling properties of
fundamental physical properties underlines the
need of miniaturization for these concepts. We
studied physical aspects of the efficiency of elec-
trostatic actuation and the tremendous reduction of
the harmful influence of inertia and gravity forces
to avoid the fatigue of material and demonstrated                                       C
                                                                                        B          z
this in some detail in Ref. [36]. Summarizing,                                          A
miniaturization plays a key role for mechanical                                                x

stability and lifetime.                                         vertical

1.2 Micromachined fabrication
The micromachined fabrication of these structures
and the general principles can be reduced to three
fundamental process steps. In the following this is             horizontal
demonstrated for a three layer heterostructure (Fig.
1 a) consisting of the substrate A, layer B being
partly or entirely removed and finally the top layer
                                                                      d             d

Step 1: Epitaxy or deposition of the individual lay-
ers (B, C, . . .) on the substrate A (Fig. 1 a):       Fig. 1: Three fundamental process steps of surface micro-
In case of mixed material systems, the substrate
may be, e.g., silica glass or a semiconductor sub-
strate (Si, . . .). Polymers may be used for layer B   Step 2: Vertical etching (Fig. 1 b):
and dielectrics or a metal for layer C. There exists
a large variety of combinations for mixed systems.     Next, the layered structure (A, B, C) is photolitho-
Note that the three fundamental process steps are      graphically masked. The uncovered areas are sub-
also characteristic for these systems.                 sequently dry etched to define the ridge mesa
Typical deposition technologies are for instance:      structures. Depending on the process, slightly
evaporation, vapour phase deposition, plasma en-       sloped sidewalls, vertical sidewalls or slightly un-
hanced chemical vapour deposition (PECVD),             dercut sidewalls can be obtained. The vertical
plasma or magnetron sputtering. Typical epitaxial      etching is shown schematically in Fig. 1 b can be
processes are for example: metal-organic vapour        obtained, e.g., by gas chopping [59] or by carefully
phase epitaxy (MOVPE), hydride vapour phase            balancing physical and chemical etching compo-
epitaxy, solid source molecular beam epitaxy           nents in plasma etching [60-63]. Strongly angled
(MBE), metal-organic molecular beam epitaxy            undercut structures can be obtained by pretilting
(MOMBE), gas source molecular beam epitaxy             the wafers and subsequent wafer rotation during
GSMBE) and chemical beam epitaxy (CBE).                etching. Preferentially an etching process is used
In the following we will emphasize pure semicon-       having a very low etching rate of the mask and
ductor heterostructures. Typical substrates are        high etching rates for materials B and C (high se-
GaAs and InP. However, also SiC, sapphire, GaSb,       lectivity) and at the same time a low selectivity

concerning the etching rates of materials B and C.        Note that the crystal orientation is crucial for the
In the schematic diagram the etching process is           etching velocity in the xy-plane. Therefore, the
stopped just after reaching the interface between         ridges, i.e. the masks, have to be properly oriented
materials A and B. However, in some cases also            with respect to the crystalline directions.
the substrate is partly etched during the vertical
structurization step. This may occur intentionally        1.3 Crucial control of strain and of stress
or unintentionally. In the latter case no serious in-
fluence on the third step exists.                         We believe it is worth to mention that the expres-
                                                          sion “strain” is generally used for crystalline mate-
Step 3: Horizontal underetching (Fig. 1 c):               rials to describe deviations in lattice constants
                                                          among the individual layers. Unstrained, compres-
In the last step, a highly selective wet-chemical         sively strained, tensile strained or strain relaxed
etching is used to underetch layer C by a distance        layers are used [64-67]. For amorphous het-
d. The relation of the ridge width W to the un-           erostructure materials (amorphous dielectrica,
deretching depth d defines the final structure:           amorphous Si (a-Si), polymers and others) the ex-
(i) If 2d≥W the central ridge is completely un-           pression “stress” is used to describe mechanical
     deretched. In this case layer B acts as a sacrifi-   tensions. Note that in micromachining literature
     cial layer and the remaining central bar of mate-    “stress” is used, since that field is dominated by Si
     rial C defines a cantilever (see Fig. 1 c). Note     and in many cases cantilevers and membranes
     that more complicated structures (e.g. sus-          consist of a-Si. However for compound semicon-
     pended membranes see Figs. 11 and 13) can            ductor micromachining, the historic crystallo-
     also be implemented, depending on the lateral        graphic expression “strain” should be preferred.
     structures of the used mask.                         Therefore, we use “strain” for strained semicon-
(ii) If 2d<W the central ridge is only partly un-         ductor layers.
     deretched. The remaining central bar of material
     C and the remaining central part of material B       In the case of amorphous materials, special care
     define an undercut ridge mesa (e.g. Fig. 4 a).       has to be taken either to avoid stress or to control
     Such an undercut ridge mesa can also be ob-          stress among the individual layers precisely. This
     served in the left and right hand-side outer sec-    guarantees in the first case the desired flat cantile-
     tions (Fig. 1c).                                     vers or membranes. In the second case this defines
(iii) If 2d is considerably less than the lateral di-     the desired bendings.
     mensions of the larger mesa, support posts or
     ridge posts can be fabricated (see , e.g., Fig. 13   However, also for crystalline semiconductor het-
     a).                                                  erostructures, in some cases the finally released
                                                          layer (here layer C) should include no strain gradi-
Thus, layer B acts as a sacrificial layer in case (i).    ents. If unbent cantilevers or membranes are de-
In case (ii) layer B is only partly underetched. In       sired prior to actuation, a good choice is a precise
the case of constricted mesa structures, layer B          lattice matching to the substrate. If an additional
includes the waveguiding confinement layers, the          relaxed buffer layer is applied, lattice matching to
active quantum well (QW) layers and the barrier           the last grown monolayers is preferred. However
layers. In this case, the precise final width of mate-    also distinct bendings can be desired in other
rial B is crucial for the proper lateral single- or       cases. This also requires precise control of total
multi-mode operation of constricted mesa                  strain and respective strain gradients in the indi-
waveguide laser devices.                                  vidual layers.

                                                                      E 2d c
                                                         σ grad =            δ              (1)
In the following we have selected typical process                   1 −ν l 2
and quality control elements which are batch fab-
ricated parallel to the micromachined structures.        where σgrad is the internal stress gradient, dc the
Quality control elements which allow nondestruc-         thickness of the layer C, E the elastic modulus of
tive observation and quantitative results, e.g. by       the beam, ν Poisson’s ratio, and l the length of the
light microscopy or white light interferometry are       cantilever and δ the deflection of the cantilever tip.
preferentially used, since cleaving of the wafers
can be avoided, which decreases yield and in-                                                                         C
                                                                                 reference plate
creases production costs. Two main points of in-                                                                      B
terest are isolated in our chapter: (i) the control of                                                                A
strain and stress and (ii) the precise control of lat-     δ
eral underetching.
2.1 Strain and stress control indicator elements:
Precise control of the strain and the stress in sus-                                                      x
pended layers is essential for the successful fabri-
                                                         Fig. 2: Schema of cantilever test structures used to measure
cation of micromachined devices. Generally the           the internal stain/stress gradient of surface micromachined
overall stress in a layer can be described by a gra-     layers (C).
dient part σgrad and a homogeneous part σhom. The
internal stress gradient is the variation of the lat-
eral tension (xy-plane) in z-direction. For instance,    2.1.2 Sensitively rotatable pointer structures –
unintentional variations of composition parameters             homogeneous strain/stress
(in z-direction) could be the reason for such a gra-
dient. The homogeneous stress is a volume prop-          For the characterization of the homogeneous
erty. Homogeneous or gradient stress (Pa) cause          strain/stress, the structure shown in Fig. 3 has been
elongation or contraction of the layers, described       used. It consists of two essentially symmetrically
by the strain value (ppm).                               arranged parts, each indicating two actuator beams
                                                         and one indicator beam. One end of each actuator
2.1.1 Multiple-length cantilever structures –            beam is fixed to the substrate by a support post and
      internal strain/stress gradients                   the other is connected to the indicator by a narrow
Figure 2 shows quality control structures used to
measure the internal strain/stress gradients of the                                                       Y
films consisting of a set of suspended cantilevers                                                            z
fixed to the substrate at one end (reference plate).                                                              x
After etching the sacrificial layer (layer B accord-
ing to Figs. 1 and Fig 2) the cantilevers (layer C)                          actuator
are released. An internal strain/stress gradient will
deflect the cantilever beam tip and the deflection       Fig. 3: Rotatable pointer structures used to characterize
can be measured as the height of the tip above the       the homogeneous strain/stress of surface micromachined
reference plate top level. The stress gradient σgrad
can be calculated from the cantilever tip deflection
δ according to the following formula [68]:
                                                         When the actuator beams are released by removing
                                                         the sacrificial layer wet-chemically, a homogene-

ous stress causes elongation or contraction. The        ridges. In Figure 4 b, the underetching just reaches
indicator beams are not directly touching each          d*, hence, the mesa of smaller widths W2 is now
other, but are slightly separated. In the case of un-   completely underetched and breaks down. The
strained layers the ends of the two indicator beams     wider mesa ridges are still stable. This can be eas-
are facing each other perfectly without any lateral     ily observed by optical microscopy from the top
deflection. With increasing strain the lateral de-      and is the signal either to stop the wet-chemical
flection grows and, thus, represents a quantitative     underetching by deionized water rinsing or to pre-
parameter to measure homogeneous strain. This is        cisely look via IR transmission microscopy for the
due to a torque deflecting the indicators. The mag-     final desired undercut 2d. In cases, IR microscopy
nitude of the deflection corresponds to the magni-      is not used d*=d may be chosen. In summary, 2
tude of the internal homogeneous strain/stress, and     simple analytical process control steps: mixing of
its direction corresponds to the type of                at least two different wide mesa ridges, differing in
strain/stress, i.e. tensile or compressive              width by 2d* combined with IR transmission mi-
strain/stress. A tensile homogeneous strain/stress      croscopy allows precise process and quality con-
causes a contraction of the actuator beams. Conse-      trol of the above mentioned third micromachining
quently the indicators rotate clockwise. Compres-       step.
sive strain/stress forces the indicator to rotate
counter-clockwise. The sensitivity is doubled by
placing pairs of elements opposite to each other.        (a)
The homogeneous stress can be determined ac-                                                                   z
cording to the formula [68]:

                                                               W1       W1              W1        W2
            E s
σ hom   =             δ     (2)
          1 − ν 2lali

where σhom is the internal homogeneous stress, E
the elastic modulus, ν Poisson's ratio, s the dis-
tance between the anchor of the actuator beam and                                  d*        d*
the centre of rotation, la the length of the actuator
beam, li the length of the indicator beam, and δ the    Fig. 4: Principles of quantitative undercut control by mix-
deflection.                                             ing mesa ridges of different widths Wi in micromachined
                                                        batch processing.
2.2 Etch-stop indicator elements for lateral
                                                        For telecommunication lasers, stable single-mode
                                                        oscillation in lateral direction (y-direction) is
2.2.1 Mixing of mesa ridges of different widths
                                                        needed which requires, e.g. a final width W-2d of
                                                        about 1.6 µm for 1.55 µm emission wavelength.
For many applications mesa ridges of width W1
                                                        Exact values of W-2d depend on the waveguide
have to be undercut by 2d to precisely maintain a
                                                        geometry and refractive indices. Thus, a key proc-
residual width W1-2d of layer B. Here we review a
                                                        ess is the precisely defined selective wet-chemical
very simple procedure to enable an interruption of
                                                        underetching of the mesa at the position of the
the selective wet-chemical underetching when
                                                        quaternary material layers (see part d of Fig. 7).
reaching an undercut of 2d* (with 2d* slightly
                                                        This is obtained using batch-processing of some
smaller than 2d). Among the mesa ridges of width
                                                        5 µm wide mesa ridges among the majority of
W1 a few mesa ridges of width W2 = W1-2d* are
                                                        7 µm wide ridges for coarse indicating the passing
positioned (Fig. 2 a). In Figure 4 a, the underetch-
                                                        of underetching 2d*. The fine indication to pre-
ing is well below d*, hence, no complete un-
                                                        cisely approach 2d is controlled, e.g., by IR mi-
deretching is reached for both types of mesa

For suspended membranes it is crucial to know the        (a)
exact moment when the membranes just reach en-                 l
tire underetching. This is very important for free
motion of the membranes. If unstrained mem-
                                                                       d         d
branes are applied and no bending and buckling of
the membranes can be observed by, e.g., white                      w
                                                                           w 3
                                                                                         w   1

light interferometry, after entire underetching and
full release of the membranes, the above described       (b)
indicator elements are very welcome. In that case,
the widths of the mesa ridge is adjusted to the re-
spective diameter of the membrane.

2.2.2 Multiple-width cantilever structures

Accurate control of the underetching as a function     Fig. 5: Schema of cantilever structures with different
of time can be achieved by using several cantile-      widths. The structures are used to control the underetching
                                                       depth d. Layer C shows a strong negative strain/stress gra-
vers in sequence of distinctly increasing widths.      dient in (a) and a lower value in (b).
Figure 5 shows schematically a cantilever test-
structure for 4 different width w1, w2, w3 and w4.
In the example the cantilevers are underetched by
a distance d whereby w1 < w2 < 2d < w3 < w4. The       3. FABRICATION OF CONSTRICTED
bending of the two cantilevers w1 and w2 might be
                                                          RIDGE MESA STRUCTURES
observed by simple microscope inspection and the
upper (w3) and lower (w2) limit of d is determined.
                                                       Longitudinally single-mode oscillation is obtained
For high (negative) strain/stress gradients the
                                                       by distributed feedback (DFB) gratings including a
completely underetched cantilevers stick to the
                                                       phase shift of about λ/4. This phase shift is either
substrate promptly (Fig. 5a) whereas for low
                                                       abrupt and located close to the centre of the reso-
strain/stress gradients the cantilevers have to be
                                                       nator, or it is divided into two or three longitudi-
long enough so that the free cantilevers clearly
                                                       nally separated abrupt parts, or it is continuously
deviate from the attached cantilevers (Fig. 5b).
                                                       distributed in axial direction (chirped DFB grat-
Anisotropy can be monitored by differently rotated
                                                       ing). Figure 6 bottom schematically depicts a con-
cantilevers. This method is strongly related to the
                                                       tinuously phase shifted chirped DFB laser obtained
“undereching star” used as a very successfull qual-
                                                       by distinctly bent waveguides (Fig. 6, main dia-
ity control element in Si micromachining.
                                                       gram) [27, 29].
In the following the above mentioned foundations
                                                       3.1 Epitaxy
are applied for two selected examples: (i) fabrica-
tion of buried constricted mesa lasers and (ii) ver-
                                                       The GaInAsP/InP semiconductor heterostructure
tical micro resonator based devices, capable of
                                                       layers were grown by MOVPE (i), the Al-
micromachined tuning. In both cases, the above
                                                       GaInAs/InP structures by MBE (ii).
mentioned process and quality control structures
are used.
                                                       (i) Using MOVPE, the following layers were
                                                       grown on p-InP substrates: a nominally undoped
                                                       150 nm InP buffer, the p-sided GaInAsP confine-
                                                       ment layers (50 nm up to 80 nm, 5x1017cm-3 p),
                                                       the strain-compensated GaInAsP multiple quantum
                                                       well (QW) active region (typically 10 QW’s) and

the n-sided GaInAsP confinement layers (50 nm            DFB grating, waveguiding calculations were per-
up to 120 nm, 5x1017cm-3 also p-doped!). Then the        formed to determine e.g. the DFB grating depth as
growth is stopped to define the DFB grating (grat-       well as the distance between the grating and the
ing depth between 25 nm and 90 nm have been              active layer (laser structure design). The coupling
implemented, enabling to vary the coupling coeffi-       coefficients obtained by theoretical waveguiding
cient K). At this position, the pn-junction is located   calculations agree very well with experimental
before starting the regrowth of the grating by           values obtained from experimental spectra, which
MOVPE. 2 µm of 2x10 18cm-3 n-InP and 0.5 µm              have been simulated by theoretical model calcula-
5x1018cm-3 n-InP (representing layers C) are epi-        tions [69-71].
taxially grown.

(ii) Using MBE, the following layers were grown
on p-InP substrates: a 150 nm lattice-matched Al-
GaAs buffer, the p-sided AlGaInAs confinement
layers (25 nm up to 80 nm, 5x1017cm-3 p), the ac-
tive strain-compensated AlGaInAs active multiple
quantum well (QW) active region (typically 10
QW’s) and the n-sided AlGaInAs confinement
layers (50 nm up to 120 nm, 5x1017cm-3 also p-
doped!). Then the growth is stopped to define the
DFB grating (grating depths between 20 nm and
60 nm have been implemented, enabling to vary
the coupling coefficient K). At this position, the
pn- junction is located before starting the regrowth
of the grating using MOVPE. 2 µm of 2x10 18cm-3
n-InP and 0.5 µm 5x10 18cm-3 n-InP (being com-
prised in layer C) are epitaxially grown. In some
cases the top layer is replaced by a 200 nm highly       Fig. 6: Centre: Schematic view of a bent waveguide to ob-
Be-doped MBE grown n-GaInAs layer (1019cm-3).            tain chirped DFB gratings. Top inset: waveguide parts ori-
                                                         ented perpendicular to the grating lines (ϑ=0 ) provide the
The number of QW’s and the thickness of the con-         smallest grating periods Λ=Λx, and the largest angle occur-
finement layers have been considerably varied in         ing in the bending ϑmax provides the largest effective grating
                                                         periods Λ=Λx/cosϑmax. Lower inset: cross sectional view of
different structures and new asymmetric structures       the structure shown in the centre in axial z-direction. Since
have been applied to enhance the modulation              the waveguide includes largest local angles ϑ(z) in the cen-
bandwidths [23]. 26 GHz bandwidth have been              tre of the waveguide, the effective grating periods conse-
obtained in our fastest structures, including            quently show largest values there. The lower inset addition-
10 QW’s, short p-sided confinement layers                ally includes the epitaxially regrown grating (layers C).
(25 nm) and longer n-sided confinement layers
(112 nm).                                                To fabricate buried constricted mesa laser devices
                                                         or buried mushroom lasers, constricted ridge mesa
3.2 DFB grating definition and regrowth                  structures (Fig. 7 a ) have to be implemented. Fig-
                                                         ure 6 centre depicts the ridge of layers B (accord-
To define a first-order DFB grating in the top con-      ing to Fig. 1). The mushroom cap layer (layer C) is
finement layer (corrugation pitch length close to        not displayed, in order to show the DFB grating
Λ=232.5 nm) electron beam lithography and Ar/O2          and the bent waveguide. Note, that this structure is
ion milling are used. MOVPE is applied to regrow         created by the undercut as described now and ex-
the grating by a 2 µm n-InP layer (Si, 5⋅1017cm-3)       ists during no fabrication step without the cap
and to terminate the structure by a 0.5 µm n-InP         layer C!
layer (Si, 2⋅1018cm-3). Prior to the definition of the

3.3 Vertical and horizontal etching                         The mesa is realized by N2+ dry etching using
                                                            H2O bypass [76]. The Si3N4 and the n+-InP layer
To provide lateral optical and carrier confinement,         are etched down to a total depth of 2.7 to 3 µm,
the mushroom-stripe laser structure (constricted            hence the DFB grating, the confinements layers
mesa) [72-75] shown in Fig. 7 is used. Part a in            and the QW active layers are passed during the
Fig. 7 shows the preparation of the metal mask              etching process (part c of Fig. 7). Now, the Ti
used for the mesa fabrication by dry etching. A Ti          mask is removed carefully by HF in order to pre-
stripe (lateral width 7 µm, thickness 100 nm)               serve the Si3N4 layer.
shown in part b of the figure is defined on top of a
sputtered Si3N4 layer (about 80 nm thick).                  (i) First the micromachined undercut of the
                                                            GaInAsP/InP ridge mesa structures is described:
                                                            Layer B comprises the QW active layers, the bar-
                                                            rier layers and the confinement layers. To obtain
                                                            stable single-mode oscillation (lateral direction, y-
                                                            direction) a final width W1-2d between 1.4 and
                                                            1.7 µm was found to be optimum by detailed theo-
                                                            retical and experimental studies. Thus, a key proc-
                                                            ess is the precisely defined selective wet-chemical
                                                            underetching of the mesa at the position of the
                                                            quaternary material layers (see part d of Fig. 7).
                                                            This is obtained by two analytical process control
                                                            steps: batch-processing of some W2=5 µm wide
                                                            mesa ridges among the majority of W1=7 µm wide
                                                            ridges combined with IR microscopy. Wet-
                                                            chemical etching by 3H2SO4:1H2O2:1H2O at a
                                                            temperature of 38°C is used for the undercut of the
                                                            ridge mesa. Just at the moment of breaking down
                                                            the 5 µm wide ridges, the remaining material layer
                                                            B has a width of slightly less than 2 µm. The re-
                                                            sidual active layer width is precisely controlled by
                                                            IR light transmission microscopy during the etch-
                                                            ing. As indicated in the figure, the light is scattered
                                                            and reflected at the inclined interfaces, which thus
                                                            yields a lateral intensity contrast. Semi-insulating
                                                            InP (Si-InP) is used in the next step (part e in Fig.
                                                            7) to fill the undercut volumes and to planarize the
                                                            mesa structure. Hydride vapour phase epitaxy is
                                                            used providing a Fe concentration above 1018cm-3.
                                                            The Si3N4 layer is used to preserve the n-InP con-
                                                            tact layer on top of the mesa and is removed wet-
                                                            chemically by HF after this process step.

                                                            A cross sectional view of our mushroom DFB la-
Fig. 7: Various steps of the technological fabrication of   ser structure is depicted in Fig. 8 by a scanning
buried mushroom stripe lasers (constricted ridge mesa
                                                            electron micrograph where the undercut volume is
                                                            partly refilled by semi-insulating InP (SI-InP). The
                                                            device shown in the figure originates from a pre-
                                                            run where the undercut volume is only partially
                                                            cladded on both sides by SI-InP. The small circle

in the centre indicates the location of the GaInAsP              and low-cost wavelength definition of the individ-
waveguiding layers (horizontal bright stripes)                   ual channels of laser arrays for DWDM [77].
which embed the active GaInAsP QW’s (dark
lines). The five dark horizontal stripes indicate the            In the case of bent and differently tilted mesa
position of the 5 QW’s. The right hand side circu-               ridges [24-29, 77] the etching rate has to be known
lar inset shows the result of an image processing of             or measured as a function of the angle in the xy-
the left inset.                                                  plane (anisotropy of the etching rate) prior to the
                                                                 device design and the definition of the maximum
                                                                 tilt and bending angles. If all the local tilt angles
                                                                 are kept below ±6°, the variation of the wet-
                                                                 chemical selective underetching rate was found to
                                                                 depend less than 5% on the variation of the angles.
                                                                 In those cases the weak anisotropy can be ne-
                                                                 glected. If larger local tilt angles are desired the
                                                                 anisotropy can be easily compensated by a local
                                                                 variation of the ridge widths.

                                                                 4. Fabrication of micromachined
                                                                    membrane-based tunable filters

                                                                 4.1 Fundamental principles

                                                                 Optical microcavity devices such as tunable Fabry-
                                                                 Perot filters comprise two high reflective distrib-
Fig. 8: Cross-sectional view of the mushroom DFB laser           uted Bragg reflectors (DBR’s) separated by an air
structure. The waveguiding and active layers have been
                                                                 cavity of the thickness L = m· λ/2 (m = 1,2,…). In
wet-chemically undercut and partially refilled. Large left
circle: magnification of the central small circle. Large right   our case the microcavity is vertically oriented. The
circle: image processed graph.                                   lower DBR and the substrate represent layer A
                                                                 (according to Fig.1). The sacrificial layer B finally
(ii) In the case of the AlGaInAs/InP laser struc-                defines the air cavity. The upper DBR corresponds
tures, the quaternary layers plus AlInAs buffer are              to layer C.
comprised in layer B. All the steps are similar to               Fig. 9 schematically shows the top view as well as
those used for the GaInAsP/InP lasers, except the                the cross section of the tunable air gap filters. The
selective wet-chemical undercut. For the quater-                 top DBR is a circular membrane suspended in the
nary Al-containing layers, citric acid was used. In              lateral xy-plane by 2, 3 or 4 suspensions (support
some cases CO ion beam etching was used for the                  beams) of small widths. The suspensions are con-
ridge mesa definition.                                           nected to support posts on the opposite side. The
                                                                 cross sectional view intersects from the left to the
Note that the Ti mask has to be removed in order                 right: the left hand side support post, the left hand
to allow IR light passing the heterostructure for                side suspensions, the membrane, the right hand
transmission microscopy. This fabrication tech-                  side suspension, and the right hand side support
nique has been successfully used also for bent and               post (section XX’).
tilted waveguide DFB lasers having shown record                  Continuous tuning of the filter transmission wave-
values in wavelength tunability (3-section DFB                   length is achieved by electrostatic or thermal ac-
laser) [25, 26], increase of functionality, low-cost             tuation of the upper DBR (layer C) and thus
chirped DFB grating fabrication [28] and precise                 changing the cavity length L. Depending on L, the
                                                                 filter can be adjusted to be transparent for only one

of the DWDM channels (different vertical arrows)                        (i) InP/airgap filters and
while blocking the others.                                              (ii) Si3N4/SiO2/airgap filters.
In the case of electrostatic tuning, the top DBR is
p-doped, the sacrificial layer is intrinsic and the                     (i) InP/airgap filters
lower DBR is n-doped, forming a p-i-n structure.                        In the case of InP/airgap filters we made use of an
By reverse biasing this p-i-n diode, the suspended                      advanced multi-airgap structure achieved by a se-
top DBR moves towards the substrate. The advan-                         quence of several membrane (C) layers and sacri-
tage of this principle is the high continuous tuning                    ficial (B) layers on top of a substrate (A):
range obtained by a single control parameter and                        ABCBCBC…
the low actuation power due to the low leakage                          This sequence is an alternation of Ga0.53In0.47As
current [36, 37, 39].                                                   (B) and InP (C) lattice matched to the InP sub-
                                                                        strate (A). In this design the function of the airgaps
(a) top view:                                                           is twofold:
                                      suspension                        - The mDBR· λ/4 (mDBR=1,3,5..) air layers inside the
                                                   support post
                                                                        DBR implicate an ultra-high refractive index con-
                    X                         X'                        trast with respect to the InP layers resulting in a
                                                                        very high reflectivity.
                                                                z       - The mcavity· λ/2 (mcavity=1,2,3..) air layer defines
                                      filter membrane
(b) section X-X':
                                                                    x   the air cavity and includes a tuning capability.

                        untuned                         tuned
                                  C                                 _
                L                 B                 L-∆L


Fig. 9: (a) Top view and (b) cross section of the tunable air
cavity filter. (b) shows the electromechanical tuning princi-
ple: untuned resonator (left), actuated cavity (right).

In the case of thermal actuation thin-film heaters
[42, 78, 79] are additionally defined. The tuning of
the top DBR is achieved by a heating current and                        Fig. 10: Scanning electron micrograph of InP cantilevers.
                                                                        The structures are bend upwards due to a strain gradient
by thermal expansion of the suspensions clamping                        caused by As incorporation inside the InP.
the mirror membranes. These air cavity structures
also allow a wide continuous tuning with a single
control parameter.                                                      MOVPE is used to grow the multilayer structure
                                                                        lattice-matched to (100) InP substrates: alternating
4.2 Device fabrication                                                  λair/4 GaInAs sacrificial layers (defining finally
                                                                        the position of the air-gaps) and 3λInP/4 InP mem-
The fabrication steps of those MOEMS devices are
                                                                        brane layers. Special care has to be taken to reach
based on the foundations described in section 2. In
                                                                        lattice matching very well since compressive strain
the following detailed process descriptions are
                                                                        and vertical strain gradients are very critical, lead-
emphasized. We will focus on two material sys-
                                                                        ing to serious and undesired bendings of the un-
                                                                        deretched membranes. It has been observed that
                                                                        InP layers grown by MOVPE contain arsenic as an
                                                                        impurity whenever they are grown after an As con-

taining layer [80]. This results in InP cantilevers            sequent reactive ion etching (20sccm CH4, 70sccm
(Fig. 2) that bend upwards due to strain caused by             H2, 35mTorr, 200W, 13.56 MHz) is used to etch a
a gradient in the As distribution within the InP               depth down to ~5.5 µm. Using a Si 3N4 mask in-
layer (Fig. 10). The homogeneous strain in these               stead of a resist mask guarantees higher selectivity
structures is compressive causing InP bridge struc-            during the dry-etching and prevents polymer depo-
tures to buckle. This problem can be solved either             sition which is highly undesired at the mesa side-
by optimizing the MOVPE growth procedure [44]                  walls. This non-selective process provides vertical
or by replacing InP with GaxIn1-xAsyP1-y with a                side-walls with high aspect ratio and sharp,
minimum strain/stress gradient and a slight tensile            straight edges (Fig. 12). In this case we achieved a
homogeneous strain [80].                                       selectivity between the III/V material and the
                                                               Si3N4 mask of more than 100.
On the masks different membrane geometries are
included as well as strain/stress indicating struc-
tures (Figs. 2 and 3) and indicator elements for
lateral undercut (Fig. 5). For the four-suspension
filter structure (Fig. 11), the membranes have di-
ameters of 40 µm, the suspensions have 10 µm
widths and lengths between 10 µm and 80 µm.
The suspensions are oriented in equivalent <100>
direction. The borders of the square support posts
(150 µm x 150 µm) are parallel to <110> direc-

                                                               Fig. 12: Scanning electron micrograph of mesa structures
                                                               in InP/GaInAs etched by CH4/H2-RIE.

                                                               Next, the etch mask is removed wet-chemically
                                                               (HF/H2O) providing the pure semiconductor mesa
                                                               (Fig. 11b). For the underetching of the InP mem-
                                                               branes the GaInAs sacrificial layers are selectively
                                                               removed using FeCl3/H2O providing excellent se-
                                                               lectivity and good semiconductor/air interfaces
                                                               (Fig. 11c). For a temperature of T=21°C, an opti-
                                                               mized etching time of 35min was found for the
                                                               above described structures. Note that the etching is
Fig. 11: Technological implementation of InP/airgap filters.   anisotropic, showing an about 5 times higher etch-
(a) mask definition, (b) mesa definition by dry-etching, (c)   ing rate in <100> directions compared to <110>
wet-chemical underetching (sacrificial layer removal), (d)     directions. This has been taken into account during
side view of underetched filter membranes.
                                                               the definition of the filter orientation. The mem-
                                                               branes and suspensions have to be underetched
First, Si3N4 is deposited by PECVD on the epi-                 completely to enable free vertical motion. How-
taxial structure, then a positive resist mask is used          ever, the 4 support posts are also underetched. This
in combination with reactive ion etching (5.1sccm              can be avoided by additional masking techniques.
Ar, 3.5sccm CHF3, 50mTorr, 100W, 13.56 MHz)
to define the mask. Figure 11a shows a cut with a
single filter element (400 filters are fabricated on a
10mm x 10mm size) including the black Si3N4
mask. Outside the black area the semiconductor
surface is uncovered, ready for dry-etching. Sub-

After etching the sacrificial layer the sample has to        white light interferometry (Zygo LOT). Note the
be dried in a way that the capillary forces due to           homogeneous behavior of the four support posts
the liquid inside the “airgaps” are minimized. Oth-          and the four stripe-shaped suspensions among
erwise the released parts tend to collapse and stick         equivalent elements and lateral positions, respec-
together. CO2 critical point drying avoids the for-          tively. The four support posts are covered by a
mation of a liquid meniscus between microstruc-              protection layer avoiding underetching. A structure
ture surfaces [81] and gives excellent results [44].         including 3 spiral suspensions is shown in Fig.
Membranes with extreme aspect ratios have been               13c.
successfully fabricated. InP membranes with di-
mensions of 0.37 µm, 20 µm, 100 µm (thickness,
width, length) with air gaps of 0.38 µm thickness
have been suspended without sticking problems.

Figure 11d depicts a 3D side view of the multi-
membrane filter structure. The 3 membranes of the
top DBR are clearly separated from the 3 mem-
branes of the bottom DBR. The larger air gap in
the centre is the air cavity.

                                                             Fig. 14: White light interferometric image (Zygo LOT),
                                                             80 µm long suspensions.

                                                             Although the all-semiconductor devices require
                                                             rather expensive materials and epitaxial processes,
                                                             the devices have a potential of high optical
                                                             performance. However, the filter is only a low-cost
                                                             product if a very large number of devices is batch-
                                                             fabricated simultaneously. If in addition cheaper
                                                             materials, deposition processes and sacrificial lay-
                                                             ers are involved, the total price can be considera-
                                                             bly reduced.

                                                             (ii) Si3N4 / SiO2 /airgap filters
Fig. 13: Scanning electron micrographs: (a) 4 suspensions
per filter, (b) detail of b, (c) 3 suspensions per filter.
                                                             The use of Si3N4 / SiO2 DBR’s in combination
Fig. 13a displays a scanning electron microscope             with a polymer sacrificial layer enables low-cost
image of such a four-suspension filter structure             filters which also have been fabricated and studied.
(resonator details in Fig. 13b). The InP membranes           The dielectric DBR’s were deposited by plasma
have a thickness of 3/4λ. The optical quality of the         enhanced chemical vapor deposition (PECVD) at
two surfaces of each membrane is defined by the              low temperatures. The compositions and layer
quality of the epitaxial heterointerfaces. The sur-          thicknesses of the structure have been optimized
face micromachining fabrication process requires             and controlled using ellipsometry in combination
no micro-mounting since all is fabricated using a            with transfer matrix model calculations. The mate-
batch process. Furthermore we have a monolithic              rial stress has been controlled by frequency inter-
implementation in the GaInAsP/InP material sys-              lacing during PECVD [82]. The polymer sacrifi-
tem. According to Fig. 14 the bending of the sur-            cial layer has been deposited on top of the lower
face membranes can be very well observed using               dielectric DBR via spin coating. Subseqently the

top DBR was deposited by low temperature                   4.3 Device results
PECVD (60°C).
                                                           The InP/airgap filters show extremely high per-
For the lateral structurization the above mentioned        formance due to the very high reflective DBR’s.
mask set has been used as well. Additionally me-           The narrowest full width at half maximum
ander-shape like thin film heaters are defined on          (FWHM) with only 2.5 period InP/airgap DBR’s
the suspensions to allow for thermal actuation of          and a λ/2 air cavity was 0.4 nm [39]. A maximum
the membranes. In this case the geometric design           tuning range of ∆λ=62 nm with an actuation volt-
of the microstructures is of crucial importance.           age of 14V and insertion loss of –3dB has been
The mesa structure was realized using a photore-           achieved. In this case the suspension length was
sist (AZ1518) mask followed by reactive ion etch-          40 µm.
ing (5.1sccm Ar, 3.5sccm CHF3, 35mTorr, 200W,
13.56 MHz) of the Si3N4 / SiO2 top DBR. The se-            The first demonstrator Si3N4 / SiO2 filters (5.5 pe-
lectivity of Si3N4 / SiO2 to the PR was approxi-           riods top DBR and 5 periods lower DBR) show a
mately 5. The sacrificial polymer layer acts as an         FWHM of 15 nm at the moment. Thermal tuning
etch stop for this process step. Plasma etching            of this filter shows ∆λ = 12 nm for 1.7 V (for I =
(Technics Plasma 2.45 GHz barrel reactor,O2 ,              1 mA, 40 µm suspension length) [36].
0.7 mbar, 250 W) is also used to etch both the PR-
mask and the polymer sacrificial layer. Thereby a          4.4 Outlook
special drying technique as for the InP/airgap fil-
ters is not necessary. Figure 15 shows a white light       Both concepts, the InP/airgap filter and the
interferometric measurement of a dielectric filter.        Si3N4/SiO2 filter are capable to be monolithically
The upper Si3N4 / SiO2 DBR shows bending-up of             integrated with GaInAsP-based active layers, re-
the suspensions, whereas the circular filter mem-          quired, e.g., for VCSEL’s or photodetectors for
brane remains nearly flat. Controlling the stress we       1.55 µm applications.
are able to form half-symmetric cavities matching
to a single spatial mode. Depending on lateral de-
sign and PECVD stress calibration we are able to           7. CONCLUSIONS
fabricate membrane curvatures between 0.5 mm
and 20 mm, suitable for single spatial mode de-            In this chapter we presented typical fundamental
vices at λ = 1550 nm (spot size 10 µm with cavity          technological processes required for surface-
lengths between λ and 10λ).                                micromachined III/V semiconductor devices. The
                                                           main application field of compound semiconductor
                                                           based micromachining is put in contrast to the
                                                           widely applied and highly developed Si based sys-
                                                           tems. Considering fabrication technology, the gen-
                                                           eral principles were reduced to three fundamental
                                                           process steps: deposition of a layered heterostruc-
                                                           ture on a substrate, vertical structurization and
                                                           horizontal undercut by selectively removing sacri-
                                                           ficial layers. The basic principles were applied and
                                                           demonstrated by two specially selected optoelec-
                                                           tronic examples: (i) bent waveguide buried mush-
                                                           room stripe lasers and (ii) vertical cavity based
                                                           tunable single or multi-membrane devices includ-
Fig. 15: White light interferometric image (Zygo LOT) of
Si3N4/SiO2 membrane suspended by four arms (80 µm long).
                                                           ing airgaps. The potential of these techniques has
                                                           been underlined by remarkable experimental de-
                                                           vice results. Advances in functionality and capa-
                                                           bility of telecommunication devices have been

obtained by the combination of III/V compound          [12] Tien, N.C., Solgaard, O., Kiang, M-H.,
semiconductor photonic devices and 3D MEMS.            Daneman, M., Lau, K.Y., and Muller, R.S., 1996,
                                                       Sensors and Actuators A52, 76.
                                                       [13] Daneman, M.J., Solgaard, O., Tien, N.C.,
ACKNOWLEDGMENTS                                        Lau, K.Y., and Muller, R.S., 1996, IEEE
                                                       Phot.Technol. Lett. 8, 396.
Support by the German DFG, German Telekom              [14] Volland, E., Heerlein, H., and Rangelow,
TOPBUS and BMBF funding is gratefully ac-              I.W., 2002, Microelectronic Engineering
knowledged. Parts of the work were performed           [15] Volland, E., Heerlein, H., Kostic, I.,and
under the umbrella of the European TUNVIC and          Rangelow, I.W., 2000, Microelectronic Engineer-
MOEMS research project. The authors thank C.           ing 57-58, 641.
Prott, A. Tarraf, S. Irmer, F. Römer, V. Rangelov,     [16] Rangelow, I.W., 2000, Vacuum 62, 279.
S. Schüler, M. Strassner, N. Chitica, P. Meissner,     [17] Oesterschulze, E., 1998, Appl. Phys. A 66, 3.
H. Halbritter, F. Riemenschneider, E. Ataro, D.        [18] Lin, L.Y., Goldstein, E.L., and Tkach, R.W.,
Gutermuth, H. Schröter-Hohmann, I. Wensch, H.          1998, IEEE Photon. Technol. Lett. 10, 525.
Burkhard, S. Hansmann, B. Kempf, R. Lösch, W.          [19] Neilson, D.T., 2000, Proc. Techn. Digest OFC
Schlapp, E. Kuphal, R. Göbel, B. Hübner and B.         2000, PD 12-1.
Klepser, for technical support, simulation and         [20] Marxer, C., de Rooij, N. F., 1999, IEEE J.
measurement results and stimulating discussions.       Lightwave Technol. 17, 2.
                                                       [21] Pavesi, L., 2000, Nature, 408, 440.
                                                       [22] Homewood, K., 2001, Nature, 410, 192.
REFERENCES                                             [23] Hillmer, H., Greiner, A., Steinhagen, F.,
                                                       Lösch, R., Schlapp, W., Binder, E., Kuhn, T., and
[1] Uenishi, Y., Tsugai, M., and Meregany, M.,         Burkhard, H., 1996, SPIE Proc. series, 2693, 352.
1995 J. Micromech. Microeng. 5, 305.                   [24] Hillmer, H., Magari, K., and Suzuki, Y., 1993,
[2] Uenishi, Y., Tsugai, M., and Meregany, M.,         Photonic Technol. Lett. 5, 10.
1995 Electron. Lett. 31, 965.                          [25] Hillmer, H., Zhu, H.-L., Grabmaier, A.,
[3] Hornbeck, L.J., 1996, IEEE Digest 1996 Sum-        Hansmann, S., Burkhard, H., and Magari, K., 1994,
mer Topical Meetings, MOEMS, 96TH8164, 7.              Appl. Phys. Lett. 65, 2130.
[4] Tilmans, H.A.C., 1996, J. Micromech. Micro-        [26] Hillmer, H., Grabmaier, A., Hansmann, S.,
eng. 6, 157.                                           Zhu, H.-L., and Burkhard, H., 1995, IEEE J. of
[5] Lee, S.S., Lin, L. Y., and Wu, M.C., 1995,         Selected Topics in Quantum Electronics, 1, 356.
Appl.Phys. Lett. 67, 2135.                             [27] Hillmer, H., Grabmaier, A., Zhu, H.-L.,
[6] Lee, S.S., Lin, L. Y., Pister, K.S.J., Wu, M.C.,   Hansmann, S., and Burkhard, H1995, IEEE J.
Lee, H.C., and Grodzinski, P., 1995, IEEE Phot.        Lightwave Technol. 13, 1905.
Technol. Lett. 7, 1031.                                [28] Hillmer, H., Burkhard, H., Seebald, E., and
[7] Lin, L. Y., Shen, J. L., Lee, S.S., and Wu,        Kiesel, K., 1995, J. Vac. Sci. Technol. B 13, 2853.
M.C., 1996, Optics Letters 21, 155.                    [29] Hillmer, H., Grabmaier, A., and Burkhard, H.,
[8] Lin, L. Y., Shen, J. L., Lee, S.S., Wu, M.C.,      1997, IEE Optoelectronics, 144, 256.
and Sergent, A.M., 1996, IEEE Photon. Technol.         [30] Spisser A., Ledantec R., Seassal C., Leclercq
Lett. 8, 101.                                          J.L., Benyattou T., Rondi D., Blondeau R., Guillot
[9] Lin, L. Y., Shen, J. L., Lee, S.S., and Wu,        G., and Viktorovitch P., 1998, IEEE Photonics
M.C., 1997, IEEE Photon. Technol. Lett. 9, 345.        Technol. Lett. 10, 1259.
[10] Toshiyoshi, H., Piyawattnanmetha, W., Chan,       [31] Tayebati P., Wang P.D., Vakhshoori D., and
C.T., and Wu, C. M., 2001, IEEE J. Microelectro-       Sacks R.N., IEEE Photonics Technol. Lett., vol.
mech. systems 10, 205.                                 10, no.3, 394.
[11] Solgaard, O., Daneman, M., Tien, N.C.,            [32] Vail, E.C., Wu, M.S., Li, G.S.,Eng, L. and
Friedberger, A., Muller, R.S., and Lau, K.Y.,          Chang-Hasnain, C.J., 1995, Electronics Letters 31,
1995, IEEE Phot.Technol. Lett. 7, 41.                  228.

[33] Jayaraman V., Goodnough T. J., Beam T.l .,         [47] Li, M.Y., Yuen, W., Li, G.S., and Chang-
Ahedo F.M., and Maurice R.A., 2000, IEEE Pho-           Hasnain, C.J., 1998, IEEE Photonics Technology
ton. Technol. Lett., vol.12, no.12, 1595.               Letters 10, 18.
[34] Peerlings, J., Dehe, A., Vogt, A., Tilsch, M.,     [48] Vakhshoori, D., Tayebati, P., Chih-Cheng Lu,
Hebeler, C., Langenhan, F., Meissner, P., and           Azimi, A., Wang, P., Jiang-Huai Zhou, and Ca-
Hartnagel, H.L., 1997, IEEE Photonics Technolo-         noglu, E., 1999, Electronics Letters 35, 1.
gy Letters 9, 1235.                                     [49] Chitica, N., and Strassner, M., 2001, Appl.
[35] Azis, M., Pfeiffer, J., Wohlfahrth, M., Luber,     Phys. Lett. 78, 3935.
C., Wu, S., and Meissner P., 2000, IEEE Phot.           [50] Chang-Hasnain C.,2000, IEEE Journal on
Technol. Lett. 12, 1522.                                Selected Topics in Quantum Eletctronics, vol.6,
[36] Hillmer, H., Daleiden, J., Prott, C., Römer, F.,   no.6, 978.
Tarraf, A., Irmer, S., Rangelov, V., Schüler, S.,       [51] Sugihwo F., Larson M.C., and Harris J.S., Jr.,
Strassner, M., 2002, SPIE Proc. series 4646.            1998, Appl. Phys. Lett. 72, 10.
[37] Strassner, M., Daleiden, J., Chitica, N., Ke-      [52] Larson, M.C., and Harris Jr., J.S., 1995, IEEE
iper, D., Stålnacke, B., Greek, D., and Hjort, K.,      Photonics Technology Letters 7, 1267.
2000, Sensors and Actuators 85, 249.                    [53] Riemenschneider R., Peerlings J., Pfeiffer J.,
[38] Chitica N., Daleiden J., Bentell J., Andre J.,     Dehe A., Vogt A., Meissner P., Hartnagel H.L.,
Greek S., Pasquariello D., Karlsson M., Gupta R.,       Chitica N., Daleiden J., Streubel K., Künzel H.,
Hjort K., 1999, Physica Scripta, Vol. T79, 131-         Görtz W., 1998, SPIE Photonic West 98, Vol.
134.                                                    3276.
[39] Daleiden, J., Chitica, N., Strassner, M.,          [54] Peerlings J., Riemenschneider R., Naveen
Rondi, D., Goutain, E., Peerlings, J., Pfeiffer, J.,    Kumar V., Strassner M., Pfeiffer J., Scheuer V.,
Riemenschneider, R., Hjort, K., Dantec, R., Ben-        Daleiden J., Mutamba K., Herbst S., Hartnagel
yattou, T., Spisser, A., Leclercq, J.L., and Vik-       H.L., Meissner P., 1999, IEEE Photonics Technol-
torovitch, P., 1999, Proc. of the Conf. on InP and      ogy Letters, Vol. 11, No.2.
related Materials, ISBN 0-7803-5562-8, p 285.           [55] Strite S., Ünlü M.S., 1995, Electron. Lett.,
[40] Daleiden, J., Chitica, N., Strassner, M., Prott,   vol. 31, no.8, 672.
C., Roemer, F., Tarraf, A.,and Hillmer, H., Summer      [56] Christenson G.L., Tran A.T.T.D., Zhu Z.H.,
School and European Optical Society Topical             Lo Y.H., Hong, Mannaerts J.P., and Bhat R., 1997,
Meeting on Semiconductor Microcavity Photonics,         IEEE Photonics Technol. Lett. 9, 724.
October 21-25, 2000, Technical Digest, Ascona,          [57] Kipp, L., Skibowski, M., Johnson, R.L.,
Switzerland.                                            Berndt, R., Adelung, R., Harm, S., and Seemann,
[41] Larson, M.C., Pezeshki, B., and Harris Jr.,        R., 2001 Nature.
J.S., 1995, IEEE Photonics Technology Letters 7,        [58] Müller, A., Göttert, J., Mohr, J., Rogner, A.,
382                                                     1996, Microsystem Technologies 2, 40.
[42] Hillmer, H., Prott, C., Strassner, M., Chitica,    [59] Volland, B., Shi, F., Hudek, P., Heerlein, H.,
N.,and Daleiden, J., Proceedings „Optic in comput-      and Rangelow, I.W., 1999, J. Vac. Sci. Technol. B
ing“, Hagen Sept.2000, ISSN 1437-8507, 75.              17, 2768.
[43] Chitica N., Daleiden J., Strassner M., and         [60] Daleiden J., Kiefer R., Klußmann S., Kunzer
Streubel K., 1999, IEEE Photonics Technol. Lett.        M., Manz C., Walther M., Braunstein J., and
11, 584.                                                Weimann G., 1999, Microelectronic Engineering
[44] Viktorovitch P., Leclercq J.L., Goutain E.,        45/1, pp 9-14.
Rondi D., 2000, SPIE Microfabrication Sympo-            [61] Daleiden J., Czotscher K., Hoffmann C., Kie-
sium 2000, Santa Clara CA.                              fer R., Klußmann S., Müller S., Nutsch A., Plet-
[45] Wu, M.S., Vail, E.C., Li, G.S.,Yuen, W. and        schen W., Weisser S., Tränkle G., Braunstein J.,
Chang-Hasnain, C.J., 1995, Electronics Letters 31,      and Weimann G., 1998, J.Vac.Sci.Technol. B
1671.                                                   16(4), 1864-1866.
[46] Larson, M.C., Massengale, A.R., and Harris,
J.S., 1996, Electronics Letters 32, 330.

[62] Daleiden J., Eisele K., Keller R., Vollrath G.,   [81] Maboudian R., Howe R. T., 1997, J. Vac. Sci.
Fiedler F., Ralston J.D., 1996 , Optical and Quan-     Technol. B 15(1).
tum Electronics 28, 527-532.                           [82] Tarraf, A., to be published.
[63] Daleiden J., Eisele K., Sah R.E., Schmidt
K.H.and Ralston J.D., 1995, J.Vac.Sci.Technol. B
13 (5), 2022-2024.
[64] Hillmer, H.,. Lösch, R., and Schlapp, W.,
1997, Journal of Crystal Growth 174, 1120.
[65] Hillmer, H., Lösch, R., Schlapp, W., and
Burkhard, H., 1995, Phys. Rev. Rapid Commun.,
52, R17025.
[66] Hillmer, H., Lösch, R., Steinhagen, F.,
Schlapp, W., Pöcker, A., and Burkhard, H., 1995,
Electron. Lett. 31, 1346.
[67] Hillmer, H., review paper, 1997, Research
Trends 3, 159.
[68] Ericson F., Greek S., Söderquist J., Schweitz
J.-A.,1997, J. Micromech. Microeng. 7, 30-36.
[69] Hillmer, H., Hansmann, S., and Burkhard, H.,
1995, Optical Engineering 34, 2985.
[70] Hillmer, H., Hansmann, S., Burkhard, H.,
Walter, H., Krost, A., and Bimberg, D1994, IEEE
J. Quantum Electron., 10, 2251.
[71] Hansmann, S., Walter, H., Hillmer, H., and
Burkhard, H1994, IEEE J. Quant. Electron. 30,
[72] Burkhard, H., and Kuphal, E1985, IEEE J.
Quantum Electron., QW-21, 650.
[73] Bowers, J.E., Hemenway, B.R., Gnauck,
A.H., Bridges, T.J., Burkhardt, E. G., Milt, D. P.,
and Waynard, S., 1985, Appl. Phys. Lett. 47, 78.
[74] Liau, Z.L., and Walpole, J.N., 1982, Appl.
Phys. Lett. 40, 568.
[75] Broberg, B., Koentjoro, S., Furuya, K., and
Suematsu, Y., 1985, Appl. Phys. Lett. 47, 4.
[76] Kempf, B., Göbel, R., Dinges, H.W., and
Burkhard, H., 1990, Proceedings of 22nd Confer-
ence on Solid State Devices and Materials, Sendai,
Japan, Aug. 22 1990.
[77] Hillmer, H., "Optoelectronic Multiwavelength
component", US Patent Nr. 5,600,743, Feb. 4,
[78] Klepser, B., and Hillmer, H., 1998, J. Light-
wave Technol., 16, 1888.
[79] Hillmer, H., and Klepser, B., 1998, Proceed-
ings of SPIE 3688, 308.
[80] Chitica, N., Strassner, M., and Daleiden, J.,
2000, Appl. Phys. Lett. 77, 202.


Shared By: