Applied Surface Science 224 (2004) 193–196

Local atomic structure in Czochralski-grown

Ge1
xSix bulk alloys
I. Yonenaga*, M. Sakurai, M.H.F. Sluiter, Y. Kawazoe
Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan

Abstract

We investigated the local atomic structure in GeSi alloy experimentally and theoretically. Extended X-ray absorption fine
structure measurements in Czochralski-grown Ge1
xSix bulk alloys showed that the Ge–Ge and Ge–Si bond lengths maintain
distinctly different lengths and vary linearly with alloy composition across the whole composition range 0 < x < 1. This is in
good agreement with the expectation derived from ab initio electronic structure calculations. The result indicates that the bond
lengths and bond angles are distorted with alloy composition in GeSi.
# 2003 Elsevier B.V. All rights reserved.

PACS: 61.10.Ht; 61.43.Dq; 61.66.Dk

Keywords: Germanium–silicon; Silicon–germanium; Bond length; Bond angle; Extended X-ray absorption fine structure; Bulk alloy

1. Introduction spectroscopy has been performed to determine bond

Ge1
xSix or SixGe1
x (where x indicates the mole
fraction of silicon) is a complete solid solution system,
having the diamond-cubic structure. The 4.2% differ-
ence in the lattice constants of the constituent Ge and
Si atoms leads to various unique alloying effects on the
electronic, optical, thermodynamic, and mechanical
properties. GeSi alloys, thus, are important for micro-
electronics and opto-electronics in view of the possi-
bilities of band gap engineering. Knowing accurately,
the local atomic structure in the context of the local
strain relaxation is essential in order to clarify the
origins of such properties and to utilize the device
potential of GeSi. Up to now, much work on the
extended X-ray absorption fine structure (XAFS)
*
Corresponding author. Tel.: þ81-22-215-2042;
fax: þ81-22-215-2041.
E-mail address: yonenaga@imr.edu (I. Yonenaga).
lengths for different types of neighbor atoms and the
corresponding fractional occupancy of each type of
neighbor [1–6]. However, the local structure of GeSi
crystal is still in question. Previously, the GeSi alloys
used for these studies were grown as thin films on Si
substrates by various epitaxial techniques. In such
hetero-structures, the large inherent biaxial stress,
even when relaxed through the induction of disloca-
tions, hinders the quantitative study of the native
material properties. From such a viewpoint, we started
XAFS investigations and reported preliminary results
on the bond length of the nearest-neighbor Ge–Ge and
Ge–Si atom pair in bulk GeSi crystals grown by the
Czochralski method across the whole composition
range [7].
This paper reports the results, including those on
Si–Si atom pairs recently obtained, in comparison
with those derived using ab initio electronic structure
calculations [8].

0169-4332/$ – see front matter # 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.apsusc.2003.08.035
194 I. Yonenaga et al. / Applied Surface Science 224 (2004) 193–196
2. Experimental
High quality bulk crystals of Ge1
xSix alloys across
the whole composition range 0 < x < 1 were grown
by the Czochralski technique at very low pulling rates
ranging from 1 to 8 mm/h in a flowing Ar gas atmo-
sphere [9–13]. The composition of the alloy was
determined by means of energy dispersive X-ray
(EDX) analysis.
X-ray absorption measurements were conducted by
utilizing an XAFS station BL10B in the photon fac-
tory at the High Energy Acceleration Research Orga-
nization (KEK, Tsukuba) with a double Si(1 1 1)
monochromator at 20 K. Ge K-edge XAFS of the
GeSi alloys were measured in transmission mode at
20 K together with a comparison of that for pure Ge
single crystal. The details are described elsewhere [7].
3. Results and discussion
Fig. 1 shows the Fourier transform of the Ge K-edge
XAFS data, k3w(k), at 20 K as a function of Si content
x for the Ge1
xSix alloys. The intensity of the main
peak around r ¼ 2 Å, which is due to nearest-neighbor
atoms around Ge, decreases and becomes asymmetric
with decreasing Ge content. The peak position is also
Fig. 1. Fourier transform |F(r)| for the Ge K-edge XAFS data,
k3w(k), of GeSi for various Si content at 20 K.
5
Coordination number
4 N-Ge(Ge)
3
N-Si(Ge)
2
1
0
0 0.5 1
Si content
Fig. 2. Ge and Si coordination numbers around Ge atoms derived
from XAFS data as a function of Si content. The symbol (^) is for
pure Si crystal. The dashed lines show the coordination numbers
predicted from a random mixture of Si and Ge atoms.
shortened. The first-shell Ge and Si coordination
numbers around Ge atoms derived from the XAFS
data are shown as a function of alloy composition in
Fig. 2. The dashed lines show the coordination num-
bers predicted for a random mixture of Si and Ge
atoms. The XAFS results are in good agreement with
the random-site occupancy of Si and Ge atoms across
the whole composition range. Diffraction studies of
GeSi thin films grown by various epitaxial methods
indicate the existence of ordering [14]. Aubry et al. [5]
pointed out the tendency of clustering of Ge–Ge
dimers within Si-rich GeSi thin films. Conversely,
there is no report detecting an ordered structure
within bulk GeSi crystals [15]. In the present study,
the ordering parameter ð0:25NGe
ð1
xÞÞ=x (Ref.
[16]) estimated from the Ge coordination number NGe
around the Ge atoms in bulk Ge1
xSix crystals
(0:44 < x < 0:82), is around 0.06–0.11, but not close
to 1. This means that no preferential ordering of the
Ge–Ge dimer seems to occur in bulk GeSi crystals
across the whole composition range. These XAFS
results are in good agreement with the prediction by
ab initio calculation [8].
The Ge–Ge bond length and the Ge–Si bond length
in GeSi alloys are plotted against the alloy composi-
tion in Fig. 3. Although the uncertainty on the derived
bond length is less than 0.015 Å, it seems that the
Ge–Ge and Ge–Si bond lengths in the GeSi alloys are
distinctly different and vary in a linear fashion as a
function of Si content. At the composition x ¼ 0:5 the
Ge–Si bond length is 2.40–2.41 Å, corresponding to
 I. Yonenaga et al. / Applied Surface Science 224 (2004) 193–196
Distance of 1st nearest atom (nm)
0.25
Ge => Ge
0.24
Ge => Si
Si => Si
0.23
0 0.5
Si content
Fig. 3. Ge–Ge bond length (Ge ! Ge), Ge–Si bond length (Ge !
Si), and Si–Si bond length (Si ! Si) in GeSi alloys as a function of
Si content. The dashed lines are the predicted compositional
dependence of the bond lengths from ab initio electronic structure
calculation.
the sum of the Si and Ge covalent radii. The Si–Si
bond length increases with decreasing Si content,
though the TEM–electron energy loss spectroscopy
data is limited to Si-rich GeSi [17]. Possibly the Si–Si
bonds have the same dependence on composition as
the Ge–Ge and Ge–Si bonds.
The observed feature is similar to other semicon-
ductor alloys as GaInAs [18], ZnCdTe [19], etc. Our
results confirm that the bonding in crystalline GeSi
alloys grown by the Czochralski method is close to,
but not completely corresponding to, the so-called
Pauling limit [20] rather than the Vegard limit [21].
The dependence of the bond length on the bonding
types indicates that bond bending occurs. These
results agree rather well with those theoretically esti-
mated from ab initio electronic structure calculations.
The dotted lines in Fig. 3 are the slope of the calcu-
lated results of [8].
However, the bonds of Ge–Ge, Ge–Si, and possibly
Si–Si show a dependence of the bonding length on the
alloy composition, i.e. they do not follow a strict
Pauling model, which is consistent with other XAFS
studies on GeSi alloys [4,5]. The fact that their
dependence has the same slope can be explained with
regard to the topological rigidity model, similar to the
results by Woicik et al. [4] but contrary to those by
Aubry et al. [5].
From the slope of the bond length dependences on
the composition shown in Fig. 3, we can estimate the
topological rigidity parameter a how a central atom
195
bonded to four neighbor atoms pushes the nearest-
neighbor atoms away. A roughly estimated topological
rigidity parameter a according to the model by Cai
and Thorpe [22] is 0.60 in agreement with the values
of 0.63–0.70 reported experimentally [3,5] and found
theoretically [23,24] for the Ge–Ge bond in GeSi
alloys. Since the parameter a varies from 0 in the
Vegard limit to 1 in the Pauling limit, our results mean
that both the bond lengths and bond angles are dis-
torted by changes in the composition of GeSi alloys,
1 similar to the reported results for most ternary alloy
semiconductors. Typically for GaInAs a  0:78 is
estimated from the change of the InAs and GaAs bond
lengths across the alloy composition range [25]. Thus,
the strain in GeSi alloys is accommodated mostly by
changes of both the bond angle and bond length.
4. Conclusion
We have investigated the local atomic structure of
GeSi alloys by XAFS measurements in comparison
with ab initio electronic structure calculation. From Ge
K-edge spectra measured at 20 K Ge–Ge and Ge–Si
bond lengths of Ge1
xSix grown by the Czochralski
method in the whole composition range 0 < x < 1 have
been determined. The bond lengths are found to main-
tain distinctly different lengths and vary linearly with
alloy composition and are close to, but not completely
at, the Pauling limit in good accordance with the
theoretical expectation. The estimated topological
rigidity parameter 0.6 suggests that both the bond
lengths and bond angles distort with alloy composition.
Acknowledgements
This work was partially supported by JSPS Grants-
in Aid (no. 13450001).
References
[1] M. Matsuura, J.M. Tonnerre, G.S. Cargill III, Phys. Rev. B 44
(1991) 3842.
[2] H. Kajiyama, S. Muramatsu, T. Shimada, Y. Nishino, Phys.
Rev. B 45 (1992) 14005.
[3] D.B. Aldrich, R.J. Nemanich, D.E. Sayers, Phys. Rev. B 50
(1994) 15026.
196 I. Yonenaga et al. / Applied Surface Science 224 (2004) 193–196

[4] J.C. Woicik, K.E. Miyano, C.A. King, R.W. Johnson, J.G. [15]
Pellegrino, T.-L. Lee, Z.H. Lu, Phys. Rev. B 57 (1998) 14592. [16]
[5] J.C. Aubry, T. Tyliszczak, A.P. Hitchcock, J.-M. Baribeau, [17]
T.E. Jackman, Phys. Rev. B 59 (1999) 12872. [18]
[6] S. Wei, H. Oyanagi, K. Sakamoto, Y. Takeda, T.P. Pearsall,
Phys. Rev. B 62 (2000) 1883. [19]
[7] I. Yonenaga, M. Sakurai, Phys. Rev. B 64 (2001) 113206.
[8] M.H.F. Sluiter, Y. Kawazoe, Mater. Trans. 42 (2001) 2201.
[9] I. Yonenaga, A. Matsui, S. Tozawa, K. Sumino, T. Fukuda, J. [20]
Cryst. Growth 154 (1995) 275.
[10] I. Yonenaga, M. Nonaka, J. Cryst. Growth 191 (1998) 393. [21]
[11] I. Yonenaga, J. Cryst. Growth 198/199 (1999) 404. [22]
[12] I. Yonenaga, J. Mater. Sci.: Mater. Electr. 10 (1999) 329. [23]
[13] I. Yonenaga, J. Cryst. Growth 226 (2001) 47. [24]
[14] J.Z. Tischler, J.D. Budai, D.E. Jesson, G. Eres, P. Zschack, J.M.
Baribeau, D.C. Houghton, Phys. Rev. B 68 (1995) 10947. [25]
D. Stekamp, W. Jäger, Philos. Mag. A 65 (1992) 1369.
G.S. Cargill, F. Spaepen, J. Non-Cryst. Solids 43 (1981) 91.
S. Muto, I. Yonenaga, unpublished work.
J.C. Mikkelsen Jr., J.B. Boyce, Phys. Rev. Lett. 49 (1982)
1412.
A. Balzarotti, in: S.K. Deb, A. Zunger (Eds.), Ternary and
Multinary Compounds, Materials Research Society, Pittsburg,
1987, p. 333.
L. Pauling, The Nature of the Chemical Bond, Cornell
University Press, Ithaca, New York, 1967.
L. Vegard, Z. Phys. 5 (1921) 17.
Y. Cai, M.F. Thorpe, Phys. Rev. B 46 (1992) 15872.
N. Mousseau, M.F. Thorpe, Phys. Rev. B 48 (1993) 5172.
P. Venezuela, G.M. Dalpian, A.J.R. da Silva, A. Fazzio, Phys.
Rev. B 64 (2001) 193202.
J.L. Martins, A. Zunger, Phys. Rev. B 30 (1984) 6217.