Docstoc

Ge Stack Using 0Ge and 0O - JJAP

Document Sample
Ge Stack Using 0Ge and 0O - JJAP Powered By Docstoc
					                                                                                                                   SS10205 1 Total pages 6
Japanese Journal of Applied Physics 50 (2011) 04DA01                                                                            REGULAR PAPER
DOI: 10.1143/JJAP.50.04DA01


Isotope Tracing Study of GeO Desorption Mechanism
from GeO2 /Ge Stack Using 73 Ge and 18 O
Sheng Kai Wang1;2 Ã, Koji Kita1;2 , Tomonori Nishimura1;2 , Kosuke Nagashio1;2 , and Akira Toriumi1;2
1
    Department of Materials Engineering, The University of Tokyo, Bunkyo, Tokyo 113-8656, Japan
2
    JST-CREST, Bunkyo, Tokyo 113-8656, Japan
Received September 23, 2010; accepted December 1, 2010; published online April 20, 2011

        GeO desorption from a GeO2 /Ge stack is a critical concern in Ge metal oxide semiconductor field effect transistors (MOSFETs). In this
        contribution, we focus on a uniform-desorption region and unveil the GeO desorption mechanism from a GeO2 /Ge stack by 73 Ge and 18 O isotope
        tracing in thermal desorption spectroscopy (TDS) analysis, in which the Ge and O diffusion kinetics in GeO2 and the interfacial reaction kinetics
        have been investigated. Through 73 Ge isotope tracing, we have clarified that Ge in the desorbed GeO dominantly comes from the GeO2 surface.
        Moreover, the self-diffusivity of oxygen was evaluated to be much larger than Ge in GeO2 . Furthermore, owing to the difference among GeO
        desorptions from GeO2 /Ge stacks with various substrate orientations, the reaction at the GeO2 /Ge interface was attributed to the redox reaction
        kinetics. On the basis of our experimental findings, we have proposed an oxygen vacancy diffusion model of the GeO desorption mechanism.
        # 2011 The Japan Society of Applied Physics




                                                                                  Ge16 O2 /Ge18 O2 /Ge stack, we found that the oxygen in the
1. Introduction                                                                   desorbed GeO comes from the GeO2 surface. In addition, our
Ge is considered to be a promising candidate for replacing                        previous study revealed that the desorption of GeO from
Si for devices beyond scaling because of its intrinsic high                       GeO2 /Ge is divided into a uniform-desorption region
mobility.1) Dielectrics and dielectric/Ge interfaces with                         (Region I) and a nonuniform one (Region II) with void
good quality are required for high-performance Ge metal                           formation.11) Although the GeO desorption behavior from
oxide semiconductor field effect transistors (MOSFETs).                             GeO2 /Ge stacks has been partially clarified, the diffusion
However, the thermodynamically unstable nature of Ge                              species and detailed desorption mechanism remain unknown.
oxides on Ge hampers the development of Ge MOSFETs.                                  In this paper, 73 Ge and 18 O isotope labeling in GeO2 /Ge
GeO2 quality severely deteriorates during the thermal                             stacks, a diffusion kinetic study of Ge and O in GeO2 , and
treatment processes such as substrate dopant activation and                       the study of the interfacial reaction mechanism with Ge
dielectric post deposition annealing, because the desorption                      substrates with different orientations are discussed to
of GeO from GeO2 /Ge occurs even below 500  C.2)                                 elucidate the desorption mechanism of GeO desorption from
   To overcome the above challenge, much effort has been                           a GeO2 /Ge stack. The mechanism in the nonuniform
devoted to Ge interface passivation processes such as ozone                       desorption region, however, is much complex and we only
treatment,3) high-pressure oxidation,4) oxide capping,5) sur-                     focus on the GeO desorption mechanism in the uniform
face nitridation,6) rare-earth introduction,7) and atomic                         desorption region.
oxygen radical treatment.8) Albeit Ge MOSFETs electric
characteristics have been significantly improved by applying                       2. Experimental Procedure
the above-mentioned techniques, without sufficient knowl-                           Four kinds of samples, i.e., Nat GeO2 /73 Ge/SiO2 /Si denoted
edge of the desorption mechanism of GeO from a GeO2 /Ge                           by [A], 73 GeO2 /Nat Ge (100) denoted by [B], Nat Ge16 O2 /
stack, the process optimization and technology evolution                          Nat
                                                                                      Ge18 O2 /Nat Ge denoted by [C], and Nat Ge16 O2 /73 Ge18 O2 /
remain limited.                                                                   SiO2 /Si denoted by [D] were used in our experiments.
   Previously, we studied the desorption mechanism of GeO                         For sample [A], a 65-nm-thick amorphous 73 Ge layer was
from a GeO2 /Ge stack, because GeO2 /Ge is the most                               deposited on a SiO2 (1 m)/Si substrate by thermal
fundamental gate stack in Ge MOSFETs, even though the                             evaporation in vacuum,12) followed by 17-nm-thick Nat GeO2
gate dielectric film is not GeO2 but another high-k oxide.                         deposition onto a 73 Ge layer by sputtering. For sample [B],
GeO desorption is not observed from the decomposition of                          first, a gallium-doped p-type Ge(100) substrate was cleaned
GeO2 itself below 700  C but from the reaction of GeO2 with                      for 10 min in methanol using an ultrasonic cleaner, for 1 min
the Ge substrate.9) In our study, for GeO2 /Ge, GeO desorbs at                    in 25% HCl solution, for 30 s in H2 O2 /ammonia/H2 O
approximately 550  C whereas no GeO desorption was                               (1 : 0:5 : 100) solution, and for 3 min in 5% HF solution.
observed from a GeO2 /SiO2 /Si structure. Moreover, through                       After each cleaning process, the wafer was rinsed in de-
atomic force microscopy (AFM) observation from a line-                            ionized water for five cycles and finally dried with nitrogen
patterned GeO2 /Ge stack, we found that, in the GeO                               blow. Then, a 12-nm-thick 73 Ge layer was deposited by
desorption process in ultrahigh vacuum, the Ge substrate                          thermal evaporation in vacuum, followed by annealing in
underneath GeO2 is consumed via the following chemical                            1 atm O2 at 550  C for 15 min. The GeO2 thickness was
reaction: GeO2 þ Ge ! 2GeO.10) From the results of our                            confirmed to be 17 nm by ellipsometry. Comparing the
TDS analysis of GeO2 /Ge with various initial GeO2                                physical density of GeO2 with that of amorphous Ge,13) we
thicknesses, we concluded that GeO desorption from                                concluded that the 73 Ge layer is slightly overoxidized,
GeO2 /Ge is a diffusion-limited process.10) Furthermore, on                        including the Nat Ge substrate. In addition, the self-diffusivity
the basis of the results of our 18 O tracing experiment from a                    of Ge in Ge is very small ($2 Â 10À22 m2 /s) at 550  C,14)
                                                                                  and the GeO2 layer in sample [B] is mainly 73 GeO2 . For
    Ã
    E-mail address: skwang@adam.t.u-tokyo.ac.jp                                   sample [C], a Nat Ge16 O2 (17 nm)/Nat Ge18 O2 (17 nm)/Nat Ge
                                                                          04DA01-1                          # 2011 The Japan Society of Applied Physics
Jpn. J. Appl. Phys. 50 (2011) 04DA01                                                                                                S. K. Wang et al.


bilayer structure was prepared by thermal oxidation in 18 O2
ambient, followed by the Nat Ge16 O2 sputtering process. This
structure is quite similar to that used in our previous study.11)
For sample [D], a 20-nm-thick amorphous 73 Ge layer was
deposited on a SiO2 (1 m)/Si substrate by thermal
evaporation in vacuum, followed by annealing at 525  C in
18
   O2 ambient (18 O2 >97%) at 20 atm. The concentration of
18
   O in the as-grown GeO2 was estimated to be about 60% by
TDS measurement. Here, we define the concentration of 18 O
as the amount of 18 O enriched in the GeO2 layer [18 O=
ð16 O þ 18 OÞ ¼ 60%]. The incorporation of 16 O might come
from the adsorbed 16 O2 or H2 16 O from the inside wall of the
furnace. After that, Nat Ge16 O2 was deposited by sputtering.
To confirm the thicknesses of the 73 Ge18 O2 and Nat Ge16 O2
layers, we deposited the layers onto Si substrates. Their                                                         (a)
thicknesses were confirmed to be about 28 and 26 nm by
ellipsometry, respectively.
   Isothermal TDS measurements were performed on the
samples [A] and [B] at 540 and 555  C, respectively. For
non-isothermal TDS measurements, temperature was in-
creased from 25 to 800  C at a heating rate of 60  C/min. A
TDS measurement system with a lamp heating system and a
quartz sample holder (ESCO EMD-WA1000S/W) was used
in a multi-ion detection (MID) mode with a sensitivity of up
to 10À15 A. For samples [A] and [B], the mass numbers of
m=z ¼ 18, 40, 44, 86, 88, 89, 90, and 92 were chosen, in
which m=z ¼ 86, 88, 89, 90, and 92 were considered as
the signals from 70 Ge16 O, 72 Ge16 O, 73 Ge16 O, 74 Ge16 O, and
76
   Ge16 O, respectively. In the Ge-73 isotope tracing experi-
ment, the following calculation was applied to the raw data
treatment:
                                                                                                                  (b)
       INat GeO ¼ I70 GeO þ I72 GeO þ I73 GeOðNatÞ þ I74 GeO þ I76 GeO
                  I70     þ I72 GeO þ I74 GeO þ I76 GeO
                ¼ GeO                                    ;           ð1Þ
                   A70 Ge þ A72 Ge þ A74 Ge þ A76 Ge
 I73 GeOð pureÞ ¼ I73 GeOðTotalÞ À I73 GeOðNatÞ
              ¼ I73 GeOðTotalÞ À INat GeO A73 Ge ;                   ð2Þ

where INat GeO , I70 GeO , I72 GeO , I74 GeO , and I76 GeO in eq. (1) are
the TDS desorption integrated area of GeO with natural
abundance excluding the contribution of the pure 73 GeO
layer, and the GeO desorption integrated areas of 70 GeO,
72
   GeO, 74 GeO, and 76 GeO, respectively. I73 GeOðPureÞ ,
I73 GeOðTotalÞ , and I73 GeOðNatÞ in eq. (2) are the TDS desorption
integrated area of 73 GeO from the pure 73 GeO2 layer, and
the total 73 GeO desorption integrated area and integrated
area of 73 GeO that comes from the natural abundance
                                                                                                                  (c)
enriched layer. A70 Ge , A72 Ge , A73 Ge , A74 Ge , and A76 Ge are the
natural abundances of 70 Ge, 72 Ge, 73 Ge, 74 Ge, and 76 Ge,
                                                                            Fig. 1. (Color online) Non-isothermal TDS spectra of GeO desorption
respectively. In the 18 O isotope tracing experiment,                       from (a) Nat GeO2 /73 Ge/SiO2 /Si (sample [A]), (b) 73 GeO2 /Nat Ge (100)
m=z ¼ 89 and 91, corresponding to 73 Ge16 O and 73 Ge18 O,                  (sample [B]). (c) Nat Ge16 O2 /Nat Ge18 O2 /Nat Ge (sample [C]). Nat GeO and
                                                                            73
respectively, were selected for comparison. During the                         GeO desorption signals are plotted by dark-blue and red dots, respectively.
measurements, the background vacuum level of the main
chamber was about 8 Â 10À8 –5 Â 10À7 Pa.
                                                                            73
    Sample [D] was annealed in 1 atm N2 ambient for 10 s at                   Ge has the least natural abundance, we expected to obtain
600 and 650  C. After that, their depth profiles of Ge and O                the Ge diffusion coefficient with a higher accuracy.
were measured by secondary ion mass spectrometry (SIMS).
SIMS measurements were performed using a 2.0 kV Csþ                         3. Results and Discussion
ion beam. Five masses, 28 Si, 16 O, 18 O, 73 Ge, and 74 Ge, were            Figures 1(a)–1(c) show the non-isothermal TDS spectra of
simultaneously detected. Since the natural abundance of                     GeO desorption for samples [A], [B], and [C], respectively.
74
   Ge (36.7%) is the largest among those of Ge isotopes and                 Using such stacks, we can easily determine where Ge in the
                                                                     04DA01-2                          # 2011 The Japan Society of Applied Physics

                                                                                                                                                SS10205
Jpn. J. Appl. Phys. 50 (2011) 04DA01                                                                                      S. K. Wang et al.


desorbed GeO comes from. In Fig. 1(a), in the initial stage
of the measurement (520–530  C), the desorption rate of
Nat
    GeO is much higher than that of 73 GeO, indicating that Ge
in the desorbed GeO dominantly originates from the top
Nat
    GeO2 layer. A similar result is observed in the reversed
structure [B], as depicted in Fig. 1(b), where 73 GeO shows a
higher desorption rate than Nat GeO at the initial stage (550–
565  C). To demonstrate a complete image of the isotope
tracing experiments, we show the 18 O isotope experimental
result in Fig. 1(c), where O desorption from surface
Nat
    Ge16 O occurs through the 18 O isotope tracing experiment.
It is shown that the oxygen in the desorbed GeO comes from
the GeO2 surface. Thus, we conclude that, in the initial stage
of GeO desorption, both Ge and oxygen in the desorbed GeO
come from the GeO2 surface. Moreover, in Fig. 1(c), with                                              (a)
the increase in sweeping temperature, the Nat Ge18 O trace is
also found to approach the Nat Ge16 O trace. This could be
explained by considering the intermixing of oxygen within
the stacked GeO2 films during annealing.11) In Figs. 1(a) and
1(b), as the sweeping temperature increases, the desorption
rate of Nat GeO becomes closer to that of 73 GeO, suggesting
that the contribution of Ge from the substrate increases,
because it is inferred that the intermixing of Ge between
GeO2 and Ge substrate occurs gradually during the
annealing treatment.
   To obtain the complete image of the GeO desorption
kinetic, isothermal TDS measurements were performed on
samples [A] and [B], and the results are shown in Figs. 2(a)
and 2(b), respectively. In Fig. 2(a), the desorption rate of
Nat
    GeO is higher than that of 73 GeO, while in Fig. 2(b), the
desorption rate of 73 GeO is higher than that of Nat GeO in                                            (b)
region I. These results are quite consistent with the non-
isothermal TDS results. It is necessary to point out that, in      Fig. 2. (Color online) Isothermal TDS spectra of GeO desorption from
Fig. 2(b), in the initial stage in region I, the signal from the   (a) Nat GeO2 /73 Ge/SiO2 /Si (sample [A]) and (b) 73 GeO2 /Nat Ge (100)
Nat
    Ge substrate appears to have a low intensity. Concerning       (sample [B]). Nat GeO and 73 GeO desorption signals are plotted by dark-blue
this result, we think that there are two possible causes: (1)      and red dots, respectively. The GeO desorption spectra are divided into
                                                                   regions I and II by dash lines.
the overoxidized GeO2 ; (2) part of the signal coming from
Nat
    GeO2 on the sample edges. Although the sample edges
were cut before the TDS measurement, the edges were still
inevitably oxidized.
   The sample structures and the results of the isotope
tracing experiments are schematically summarized in Fig. 3.
Figure 3 clearly shows that, in the uniform desorption
region, both Ge and O in the desorbed GeO come from the
GeO2 surface. Thus, we conclude that the diffusion species
should not be the GeO molecule that directly diffused from
the GeO2 /Ge interface. Instead, Ge atom (or ion) diffusion
from the GeO2 /Ge interface to the GeO2 surface or oxygen
atom (or ion or molecule) diffusion from the GeO2 surface to
the GeO2 /Ge interface may explain the GeO desorption.
                                                                   Fig. 3. (Color online) Schematic sample structures and isotope tracing
   Next, we discuss the diffusion coefficients of Ge and O
                                                                   experimental results.
in GeO2 . Figures 4(a) and 4(b) show the SIMS profile of
sample [D] annealed at 600 and 650  C in 1 atm N2 for 10 s,
respectively. In the estimation of the amount of diffused
isotopes, the profiles near the surface and GeO2 /SiO2              the following calculations. The 18 O and 73 Ge fractions
interface were not used, because the profile in the surface         at each depth point determined as 18 O=ð18 O þ 16 OÞ and
                                                                   73
region might suffer from the surface-induced nonequilibrium            Ge=ðNat Ge þ 73 GeÞ from the results in Figs. 4(a) and 4(b)
condition of the measurement, and that near the GeO2 /SiO2         are shown in Figs. 5(a) and 5(b), respectively. In Fig. 5,
interface region was rather distorted by the matrix effects.        x ¼ 0 represents the initial interface between Nat Ge16 O2 and
                                                                   73
We assume that the profile in the region 3 nm from both                Ge18 O2 . The original point (x ¼ 0) was defined as the point
the surface and the GeO2 /SiO2 interface is valid in               (depth) corresponding to a 50% decrease in the 18 O intensity
                                                            04DA01-3                         # 2011 The Japan Society of Applied Physics

                                                                                                                                     SS10205
Jpn. J. Appl. Phys. 50 (2011) 04DA01                                                                                                   S. K. Wang et al.




                                      (a)                                                                               (a)




                                      (b)
                                                                                                                        (b)
                                            73
Fig. 4. (Color online) SIMS profiles of      Ge and 18 O for sample [D]
annealed at (a) 600 and (b) 650  C in 1 atm N2 for 10 s, respectively.        Fig. 5. (Color online) The concentration-distance relationship of (a) 73 Ge
                                                                               and (b) 18 O for sample [D] annealed at 600 and 650  C in 1 atm N2 for 10 s,
                                                                               respectively. x ¼ 0 represents the initial interface between Nat Ge16 O2 and
                                                                               73
                                                                                  Ge18 O2 . The original point (x ¼ 0) was defined as the point (depth)
in the SIMS profile of the as-prepared Nat Ge16 O2 /73 Ge18 O2 /                corresponding to a 50% decrease in the 18 O intensity in the SIMS profile of
                                                                               the as-prepared Nat Ge16 O2 /73 Ge18 O2 /SiO2 /Si structure.
SiO2 /Si structure. It is clear that 73 Ge shows a much sharper
slope at x ¼ 0 than 18 O, strongly indicating that the diffusion
coefficient of Ge in GeO2 is much lower than that of O.
   Figure 5(a) also shows the depth profile of 18 O measured                              Table I. Diffusion coefficients of Ge and O in GeO2 .
before annealing. The 18 O concentration profile of the starting
structure (red line) in Fig. 5(a) is a convolution of the real                                      Ge diffusion coefficient          O diffusion coefficient
                                                                                Temperature
                                                                                                             DGe                            DO
interface with an unknown instrumental broadening func-                            ( C)
                                                                                                         (10À18 m2 /s)                 (10À18 m2 /s)
tion.15) If the Nat Ge16 O2 /73 Ge18 O2 interface is assumed to be
                                                                                     600                   0:1 Æ 0:02                     1 Æ 0:3
abrupt (a step function) and broadening function is assumed to
                                                                                     650                   0:4 Æ 0:2                      8Æ2
be Gaussian,16) the resulting convolution will be an error
function. Under these assumptions, the 2 width, defined by
the positions of 83–16% of the nominal 60% 18 O concentra-
tion, is estimated to be about 3 nm. If the original Nat Ge16 O2 /             motion of O should be dominant in GeO2 . Furthermore, by
73
   Ge18 O2 interface does not move, and is defined as x ¼ 0, the                taking the surface initiated desorption into consideration, the
real 18 O and 73 Ge concentrations are described by17)                         diffusion of O must proceed by exchanging with O at its
                                                                             neighbor sites. In this case, the O diffusion process within
                              1             x
                   Cðx; tÞ ¼ C0 erfc pffiffiffiffiffiffi ;                 ð3Þ             the GeO2 network can be regarded as the oxygen vacancy
                              2           2 Dt
                                                                               diffusion. This result also eliminates the possibility of a
where C0 is the isotope concentration in the starting 73 Ge18 O2               paired Ge–O vacancy diffusion, because of the large
layer.                                                                         difference between their diffusion coefficients. Thus, we
   By using eq. (3), we can roughly estimate the diffusion                      conclude that, in the uniform-desorption region, the diffu-
coefficients of Ge and O in GeO2 . The calculated results are                    sion species for GeO desorption from the GeO2 /Ge stack are
listed in Table I. The diffusion coefficient of O is 10–20                        mainly oxygen vacancies from the GeO2 /Ge interface, albeit
times larger than that of Ge at 600 and 650  C. Therefore, the                Ge diffusion may also provide a small contribution. The
                                                                          04DA01-4                        # 2011 The Japan Society of Applied Physics

                                                                                                                                                    SS10205
Jpn. J. Appl. Phys. 50 (2011) 04DA01                                                                                      S. K. Wang et al.




Fig. 6. (Color online) Non-isothermal TDS spectra of GeO desorption
from 30 nm GeO2 /Ge stacks with (100), (110), and (111) substrate
orientations. The sweeping rate is 60  C/min.




diffusion coefficients of Ge and O in GeO2 were estimated
on the assumptions that the original Nat Ge16 O2 /73 Ge18 O2          Fig. 7. (Color online) Schematic of the redox reaction at the GeO2 /Ge
interface (x ¼ 0 before anneal) follows an ideal step function        interface.
and the interface does not move during the annealing
process. This may result in a slight overestimation of the
diffusion coefficient especially in the case of DGe at 600  C.
However, this overestimation does not affect the fact that Ge
diffuses much more slowly than O in GeO2 .
   To further probe the reaction at the GeO2 /Ge interface,
a 30 nm GeO2 layer was deposited on Ge(100), Ge(111),
and Ge(110) substrates by sputtering. Non-isothermal TDS
measurements were performed for these three samples at a
sweeping rate of 60  C/min. Figure 6 shows the GeO
desorption results from three kinds of 30 nm GeO2 /Ge
stacks with different substrate orientations. This clearly
shows the substrate orientation dependence. For GeO2 /
Ge(111), the GeO desorption occurs at the lowest rate and
the highest peak temperature, while for GeO2 /Ge(110), it
showed the highest rate and resulted in a shift in the                Fig. 8. (Color online) Schematic of the GeO desorption mechanism from
desorption peak temperature towards lower values. By                  GeO2 /Ge stacks in the uniform desorption region.
comparing the desorption kinetics with the oxidation
behavior for three substrate orientations,18) it is recognized
that the GeO desorption from GeO2 /Ge systems is in a good            and GeO2 . Since the equilibrium concentration of oxygen
agreement with their oxidation behavior. This suggests that           vacancies is much higher for the interfacial region than for
the interfacial reaction between GeO2 and the Ge substrate            the GeO2 bulk, the oxygen vacancies diffuse into the GeO2
can be regarded as an oxidation process. When a Ge                    surface region. Consequently, GeO desorbs from the surface
substrate is oxidized such as in back-bond oxygen insertion,          region. The schematic of this GeO desorption model is
the oxygen must come from the GeO2 network. Thus, the                 shown in Fig. 8.
interfacial region at the GeO2 will always become oxygen-                Finally, we will briefly mention the nonuniform-deso-
deficient, while the partially oxidized Ge can also be treated         rption region. Concerning Figs. 2(a) and 2(b), in region I,
as an oxygen-deficient network. From the viewpoint of                  the GeO desorption is considered to occur uniformly.11) For
thermodynamic equilibrium, the oxygen-deficient network                the desorption peaks in region II of the isothermal TDS
must include a certain concentration of oxygen vacancies.             spectra, however, it is attributed to the nonuniform
Thus, it is inferred that the interfacial redox reaction is the       desorption with void formation from the AFM observa-
source of oxygen vacancy generation. The schematic of the             tions.11) To further discuss this result, we define  as
interfacial reaction is depicted in Fig. 7.
                                                                                                           IN
   On the basis of the experimental results obtained so far,                                         ¼                                   ð4Þ
we propose an oxygen vacancy diffusion model to explain                                                     I0
the mechanism of GeO desorption from the GeO2 /Ge stack               to evaluate the nonuniform degree of GeO desorption,
in a uniform desorption region. First of all, oxygen vacancies        where, IN and I0 are the integrated areas of region II and the
are generated from the interfacial redox reaction between Ge          total area, respectively. Figure 9(a) shows the isothermal
                                                                 04DA01-5                      # 2011 The Japan Society of Applied Physics

                                                                                                                                    SS10205
Jpn. J. Appl. Phys. 50 (2011) 04DA01                                                                                                     S. K. Wang et al.


                                                                               uniform-desorption region. On the basis of the experimental
                                                                               results that O shows a much higher diffusion coefficient than
                                                                               Ge in GeO2 , the oxygen vacancy diffusion model has been
                                                                               proposed to explain the desorption mechanism of GeO from
                                                                               a GeO2 /Ge stack in the uniform-desorption region. By
                                                                               using this model, we can successfully explain most of the
                                                                               experimental observations obtained so far. In addition, two
                                                                               kinds of GeO desorption (uniform and nonuniform) has been
                                                                               demonstrated, and the uniform one is likely to occur at
                                                                               relatively lower temperatures.
                                                                               Acknowledgements
                                                                               The authors would like to thank T. Okabe for his helpful
                                                                               comments on diffusion kinetics. This work was partly
                                          (a)                                  supported by a Grant-in-Aid from the Ministry of Education,
                                                                               Culture, Sports, Science and Technology of Japan (MEXT),
                                                                               and partly performed in collaboration with Semiconductor
                                                                               Technology Academic Research Center (STARC). S. K.
                                                                               Wang is grateful to the China Scholarship Council (CSC)
                                                                               and the Global Centers of Excellence (GCOE) program by
                                                                               MEXT for their financial support.



                                                                                1) S. M. Sze: Physics of Semiconductor Devices (Wiley, New York, 1981)
                                                                                     2nd ed., p. 789.
                                                                                2) K. Prabhakaran and T. Ogino: Surf. Sci. 325 (1995) 263.
                                                                                3) D. Chi, C. O. Chui, K. C. Saraswat, B. B. Triplett, and P. C. McIntyre:
                                                                                     J. Appl. Phys. 96 (2004) 813.
                                                                                4) C. H. Lee, T. Tabata, T. Nishimura, K. Nagashio, K. Kita, and A. Toriumi:
                                                                                     Appl. Phys. Express 2 (2009) 071404.
                                                                                5) S. Balakumar, K. M. Hoe, G. Q. Lo, R. Kumar, N. Balasubramanian, Y. L.
                                          (b)
                                                                                     Foo, S. Tripathy, and D. L. Kwong: J. Electron. Mater. 37 (2008) 944.
                                                                                6) T. Maeda, T. Yasuda, M. Nishizawa, N. Miyata, Y. Morita, and S. Takagi:
Fig. 9. (Color online) (a) Isothermal TDS spectra of GeO desorption                  J. Appl. Phys. 100 (2006) 014101.
from Nat GeO2 /73 Ge/SiO2 /Si stacks at 540, 524, and 507  C. The regions I    7) T. Tabata, C. H. Lee, K. Kita, and A. Toriumi: ECS Trans. 16 [5] (2008)
and II are divided by broken lines. (b)  as a function of annealing                 479.
temperature.                                                                    8) M. Kobayashi, T. Irisawa, B. M. Kope, Y. Sun, K. Saraswat, H.-S. P.
                                                                                     Wong, P. Pianetta, and Y. Nishi: Tech. Dig. VLSI Symp., 2009, p. 76.
                                                                                9) K. Kita, S. Suzuki, H. Nomura, T. Takahashi, T. Nishimura, and A.
                                                                                     Toriumi: Jpn. J. Appl. Phys. 47 (2008) 2349.
TDS spectra of GeO desorption from sample [A], in which                        10) S. Wang, K. Kita, T. Nishimura, K. Nagashio, and A. Toriumi: Ext. Abstr.
regions I and II are divided by broken lines.  as a function                        Solid State Devices and Materials, 2009, p. 1002.
of annealing temperature is shown in Fig. 9(b). This                           11) S. K. Wang, K. Kita, C. H. Lee, T. Tabata, T. Nishimura, K. Nagashio, and
                                                                                     A. Toriumi: J. Appl. Phys. 108 (2010) 054104.
indicates that the lower temperature is more likely to follow                  12) Natural abundances of the five Ge isotopes: 70 Ge (20.38%), 72 Ge (27.31%),
the uniform-desorption mechanism (a smaller ). The                                  73
                                                                                       Ge (7.77%), 74 Ge (36.71%), 76 Ge (7.84%). In our experiment, the purity
temperature-dependent tendency of  suggests that the                                of the 73 Ge isotope was more than 92%.
                                                                               13)   P. Viscor: J. Non-Cryst. Solids 101 (1988) 170.
uniform desorption and nonuniform-desorption originate
                                                                               14)   E. Huger, U. Tietze, D. Lott, H. Bracht, D. Bougeard, E. E. Haller, and H.
from different mechanisms.                                                            Schmidt: Appl. Phys. Lett. 93 (2008) 162104.
                                                                               15)   J. C. Mikkelsen, Jr.: Appl. Phys. Lett. 45 (1984) 1187.
4. Conclusions                                                                 16)   C. W. Magee and R. E. Honig: Surf. Interface Anal. 4 (1982) 35.
                                                                               17)   J. Crank: The Mathematics of Diffusion (Claredon Press, Oxford, U.K.,
The GeO desorption mechanism has been further clarified                               1975) 2nd ed., p. 14.
using a 73 Ge and 18 O labeling technique. We conclude that                    18)   T. Sasada, Y. Nagakita, M. Takenaka, and S. Takagi: J. Appl. Phys. 106
the GeO desorption initiates from the GeO2 surface in the                            (2009) 073716.




                                                                        04DA01-6                           # 2011 The Japan Society of Applied Physics

                                                                                                                                                    SS10205

				
hkksew3563rd hkksew3563rd http://
About