Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Solid State Communications Vol. 12, pp. 253—257, 1973.

Pergamon Press Printed in Great Britain

RAMAN SCATTERING IN Ge-Si ALLOYS*

W.J. Brya

Sandia Laboratories, Albuquerque, New Mexico 87115


(Received 28 September 1972 by A.A. Maradudin)

Room temperature Raman spectra are reported for polycrystalline Ge—Si


alloys as a function of composition. The frequencies and linewidths of the
three main peaks show some disagreement with previously reported work;
the results are discussed within the theories of local and resonance modes
of impurities, where possible. Minor scattering peaks and second-order
spectra, as well as anomalous behavior in several inhomogeneous samples,
aie also reported and discussed.

THE VIBRATIONAL spectrum of the Ge—Si alloys chill-cast alloy material;5 several samples of zone-
system has recently received considerable interest leveled (ZL) alloys were also studied for comparison
because of its unusual multimode behavior. In particu- with the results obtained in the HP materials. The
lar, Raman studies14 show three main lines which alloy compositionx was determined primarily from
have been tentatively ascribed to the vibrations of 7-ray attenuation measurements and/or electron
Ge—Ge, Si—Si and Ge—Si pairs. In this work, compre- microprobe analysis. The microprobe analysis also
hensive Raman spectra for a large number of alloys showed the Si- and Ge-rich HP materials and all the
over the full range of composition are presented; the ZL samples to be quite homogeneous in composition;
frequencies and linewidths for the three main peaks the HP materials in the region x ~0.5 showed, how-
show several disagreements with previously reported ever, large inhomogeneities,6 as will be discussed later.
work.3 For the Ge-rich materials, the spectra are in All the samples were polished for several hours with
good agreement with theories of local-mode behavior 0.OSji alumina to remove surface damage.
for isolated substitutional Si impurities in Ge; the Si-
rich materials show only fair correspondence with The unpolarized Rarnan spectra were obtained
the predictions for resonant and near-resonant modes in a near-back-scattering configuration using an Ar-ion
of the heavy Ge impurity in Si. The behavior with laser, double spectrometer, and photon-counting de-
composition of the minor scattering peaks and the tection; all measurements were made at room tern-
second-order spectra is also given. In addition, the ob- perature. The positions of the stronger lines were de-
served splitting of only one of the main peaks in in- termined from an average of their Stokes and anti-
homogeneous samples appears to indicate preferred Stokes components, while the Stokes components of
local environments for the constituents. weaker lines were located relative to these stronger
lines.
For this study the Ge
1_~Si~samples were pre-
pared from undoped, polycrystalline materials. A Typical spectra obtained for several alloy corn-
majority of the samples were fabricated from vacuum positions are shown in Fig. 1. The usual three promi-
hot-pressed (HP) slugs formed from fine powders of nerit peaks are seen, along with additional minor
structure which we shall discuss shortly. Here, the
* This work was supported by the U.S. Atomic Energy lines at 300 and 500cm’~are labelled Ge and Si,
Commission, respectively, since they are derived originally from the
zone-center phonon modes of the pure Ge and Si

253
254 RAMAN SCATTERING IN Ge—Si ALLOYS Vol. 12, No. 4

materials. The strong central line is labelled Ge—Si 52~ 15


since it is believed to be related to the vibrations of a Si line
pair of nei~iboringGe and Si atoms, at least in the 500
Ge- and Si-rich materails. In Fig. 2 are shown the pos-

itions
to
of the and
alloy linewidths
Si-line
composition;
_ of
in the HP these for
except
samples three
thepeaks
with as a 0.6,
multiple
0.3 ~x function
peaks ~a1O 5

-,
which arise from the large inhomogeneities in compos- ~
ition, the HP and ZL materials show essentially ident-
45C 0
ical behavior. Our results are generally in good agree- 4~ 16
ment with the work of Renucci eta!.3 (RRC) although
E
12 ~

X~02
1 Os
AGAIN I I 400
390 a 8 I
Ge-Si line
38C 4
310 Ge line

z
ILl
_______________________________________________________________

,. x O~5 :: ~ a
8

412

Z 450
~
55C
J 280
O
Ge
0,2 0.4
x
0.6 0,8
l~
1.0
Si
/

~ —~——--—~ ~ I ~ FIG. 2. Dependence on composition of the frequencies


LI 09 0
G4~’~
and linewidths for the three main scattering peaks in
Ge
1_~Si~alloys at room temperature. Open triangles
Ii
and circles show frequencies for HP and ZL materials,
respectively; solid symbols show linewidths for same.
Multiple data for the Si-line of the HP materails are
due to sample inhomogeneities (see text). Open
~ ~ squares at x = 0 and I show line positions for single-
L
~00
~
250 400 550
ition
crystalofGe
Ge—Si
and Si
peak
(reference
in Ge-rich
7) material
and extrapolated
(referencepos-
1).
ENERGY SHIFT (cm~)
generally in good quantitative agreement with the
FIG. 1. Typical room temperature Raman spectra for RRC results; here also, our data indicates a linear
zone-leveled (ZL) and hot-pressed (HP) Ge i~xS~x variation in peak position, even for the small x regime.
alloys. Inset for x = 0.5 shows splitting of the Si-line In addition, our linewidths are typically less than those
at 500 cm’ observed in the inhomogeneous HP
materials, reported by RRC, with the Ge- and Si-linewidths show-
ing essentially linear behavior with composition and
the Ge—Si width showing a slight change at x 0.5.
there are several qualitative and quantitative differences.
In our case, the Ge-line typically occurs at a slightly Considerably more insight into the nature of these
higher energy than that reported by RRC; more tin- lines, including the minor peaks and second-order
portantly, our data show a linear variation of line pos- spectra previously mentioned, can be obtained by a
ition with composition as opposed to their observed more detailed analysis of specific regions in the alloy
nonlinear behavior. For the Si-line (neglecting tempo- composition. For the Ge-rich materials the
7 a behavior
single
rarily the multiple HP data) our line positions are is fairly well understood. Pure Ge shows
Vol. 12, No.4 RAMAN SCAflERING IN Ge—Si ALLOYS 255

strong Ge-peak at 300.5 cnc’ which arises from the modes. For pure Si, the single strong Si-peak is ob-
served7 at 5205 cm” As Ge is introduced, well-
-
q = 0 phonon mode. As Si impurities are introduced,
the Ge-peak slowly decreases in intensity and frequency defined lines appear at 402 and 437 cm’’ and a weak
and a new line appears 389 cm” which increases in in- line occurs at 487 cnf’ on the shoulder of the Si-
tensity and frequency with increasing Si. This line was peak. With increasing Ge, the Si-line decreases in fre-
first observed by Feldman eta!.1 in Raman scattering quency and intensity while the Ge—Si peak at 402 cm’
and was associated with the local mode of an isolated slowly shifts to higher frequencies and gains in inten-
Si atom in the Ge lattice; more recently, this same sity. The 437 and 487 cm” lines show little change
Ge—Si local mode has been observed in infrared ab- in frequency and only slight increases in intensity.
sorption studies by Cosand and Spitzer8 (CS). The ob- Ashkin er al. 2 report similar multiple peaks in the
served frequency of the local mode is in good agree- R.aman spectra of the Si-rich materials; likewise, CS
ment with the predicted value of 394 cm”1 determined also fmd evidence for i.r. absorption near these peaks.
from the Dawber and Elliott theory9 for an isolated Maradudin11 has calculated the frequencies of resonant
substitutional Si atom in the Ge lattice; in addition, and near-resonant modes for an isolated Ge atom in
Feldman er a!, show that the linear increase in the local Si; his results indicate resonant modes near 453 and
mode frequency and the decrease in the Ge-mode fre- 490 cm’ and near-resonant modes around 116, 334
quency with increasing Si content is in qualitative and 380 cm”~’.Xinh1°has calculated the Raman cross-
agreement with the theory of Xinh’°for Raman scat- sections for these modes and shows that the 116 cm’
tering in crystals of the diamond structure containing line should be weak; his results also indicate the Si-
substitutional random-mass defects. line frequency should decrease with increasing Ge, as
is observed. The observed lines at 437 and 487 cnf’
Forx ~ 0.1, a new pair of lines appears at —~445 show reasonable agreement with the calculated res-
and 460 cnf’. These peaks were seen by Feldman onant modes at 453 and 490 cm’ ;however, Si pos-
et a!. and were tentatively attributed to localized sesses zone-boundary TO phonons in the region of
modes of nearest-neighbor pairs of Si atoms in Ge; CS 450—490 cm” and they might appear in the first-
also see evidence for weak i.r. absorption bands in this order spectra as the 487 cm’ line because of the
region. Our results and those of RRC show that the lattice disorder produced with the inclusion of Ge
460 cnf’ line increases in frequency and intensity impurities. The 402 cm~line shows no obvious
with increasing Si content, eventually becoming the correspondence with any of the predicted modes al-
Si-line in the Si-rich materials; the 445 cm’ peak, though it might be assoicated with the near-resonant
however, shows little change in frequency or intensity modes at 350 cm’ It is well known, however, that
.

with inclusion of Si. Possibly the 445 cm’’ line is the calculated frequencies of these in-band resonances
associated with isolated pairs of interacting Si atoms, are very sensitive to the details of the chosen phonon
wheras the Si-peak results from larger clusters of Si, density-of-states curve for the host material,

With increased Si doping, the three main lines For x ~ 0.8 another new line appears at 285 cm’
develop marked asymmetries with long broad tails on which increases in intensity and frequency with in-
the low-frequency sides and sharp cutoffs on the high- creasing Ge and eventually becomes the Ge-line in the
frequency ends; generally, this occurs because the Ge-rich materials. l’his line probably results from
introduction of Si impurities breaks down the trans- clustering of Ge and would not be predicted by the
lational symmetry of the lattice (introduces disorder) previous theories which are applicable to isolated urn-
and phonons over the full Brillouin zone can make purifies.
contributions to the first-order Raman spectrum. In
the pure Ge sample one also observes a weak, broad In these Si-rich materials, the main lines again
two-phonon band at 575 cm~which arises from develop the previously mentioned asymmetries for
zone-boundary TO phonons. This band shows little x ~ 07, with the Ge—Si peak showing the largest
change with Si doping. effect. In addition, one observes weak, broad two-
phonon lines at 260 and 965 cm~for pure Si
For the Si-rich materails the situation is slightly which result from zone-boundary TA and TO modes;
different. Substitution of the heavy Ge atom for the they show only slight changes with Ge content. The
lighter Si is expected to produce in-band resonant 285 cnf’ Ge-line occurs in the midst of the 260 cm’
256 RAMAN SCATTERING IN Ge—Si ALLOYS Vol. 12,No. 4

band, but it is not possible to determine if the Ge-line The HP materials in this range of compositions
actually has its origin in this band as suggested by show essentially the same behavior as the ZL materials
Ashkin et a!.2 with one notable exception for 0.3 x ~ 0.6, the Si-

peak appears as a doublet with the intensities and pos-


For alloys in the range x 0.5, where little is itions of the lines showing only minor variations be-
known about the nature of the vibrational excitations, tween samples of the same composition. As seen in
the spectra show several features characteristic of both Fig. 2, the HP alloy atx = 0.5 shows Si-peaks charac-
the Ge- and Si-rich materials. In zone-leveled material teristic of x 0.3 and 03; a microprobe analysis of
ofx = 0.5, the three prominent peaks are of corn- one sample indicates sizeable regions of the material
parable intensity with the Ge—Si line showing a pro- with just such compositions. The surprising result,
nounced asymmetry relative to the other peaks. however, is that this large inhomogeneity shows no
Further addition of Ge or Si causes the intensity and effect on the Ge and Ge—Si peaks; in several samples
frequency of the Ge—Si peak to decrease, whereas the of the x = 0.5 composition, the peak positions and
Si (Ge) line increases (decreases) in frequency and in- their linewidths are essentially identical for the ZL
tensity with Si doping. Both minor peaks are observed and HP materials (see Fig. 2). Even though the nature
at 430 and 445 cm’ which occured in the Si- and of the vibrational excitations in the x 0.5 materials
Ge-rich materials, respectively; with increasing Ge (Si) is not understood, our results do seem to indicate a
content the 430(445) cm’ line gradually disappears. tendency for the constituents to favor preferred local
environments12 in samples of large macroscopic in-
Weak, broad bands at 240, 565 and 995 cm’, homogeneity. Such behavior would appear to strength-
which are derived from the two-phonon bands of the en further the arguments3’4 for the formation of an
pure Si and Ge materials, are all present. However, one ordered Ge
0,5 Si0.5 alloy and for the possible inter-
observes a new band at 790 cm’ which is probably pretation of several of the observed lines in the x
a two-phonon line also. It is not possible to determine 0.5 materials as local and resonant modes of Ge and!
whether this band arises from a combination of those or Si impurities in the ordered Ge0.5 Si0~compound.
phonons responsible for the 565 and 955 cm’ bands X-ray diffraction studies in single-crystal material of
or whether it is associated with new excitation present x = 0.5 would be of particular interest in this regard.
in the Ge0.5 Si0,5 alloy. The band is present only when
both the 565 and 955 cm’ bands occur. Acknowledgements Thanks are extended to R.N.

Fortune for his technical assistance and to G.L.


McVay and E.H. Farnum for supplying the samples.

REFERENCES
1. FELDMAN D.W., ASHKIN M. and PARKER J.H., Jr.,Phys. Rev. Lert. 17, 1209 (1966).
2. ASHKIN M., PARKER J.H., Jr., FELDMAN D.W. and RUBENSTEIN M., Bull. Am. Phys. Soc. 14, 343 (1969).
3. RENUCCI M.A., RENUCCI J.B. and CARDONA M.,Light Scattering in Solids, p.326. Flammarion, Paris
(1971).
4. RENUCCI J.B., RENUCCI M.A. and CARDONA M., Solid State Commun. 9, 1651 (1971).
5. LEFEVER RA., BAUGHMAN R.J. and MCVAY G.L., (to be published).
6. This behavior results from a segregation of the initially homogeneous alloy during the chill-cast process. For
Si- and Ge-rich materials, the segregation is small enough that the subsequent vacuum hot-press technique can
restore the homogeneity; for x 0.5, the greater segregation is difficult to remove.
~-

7. PARKER J.H., Jr., FELDMAN D.W. and ASHXIN M.,Phys. Rev. 155, 712 (1967).
8. COSANDA.E. and SPITZERW.G.,J. appi. Phys. 42, 5241 (1971).
9. DAWBERP.G. and ELLIOTT R.J.,Proc. R. Soc. (London) 273A, 222 (1963);I~oc.R. Soc. (London) 81,
453 (1963).
Vol. 12, No.4 RAMAN SCAUERING IN Ge—Si ALLOYS 257

10. XINH N.X., Localized Excitations in Solids, p.167. Plenum, New York (1968).
11. MARADUDIN A.A., Solid State Physics, Vol. 18, p. 273, Academic Press, New York (1966).
12. Measurements of zone-boundary phonon energies in the Ge—Si system suggest that the minority atoms in an
alloy do not occupy neighboring lattice sites. See LOGAN R.A., ROWELL J.M. and TRUMBORE F.A, Phys.
Rev. 136, A175l (1964).

CoobIJJa~JTcs~PaslaHa Cfl~KT~binp~KoMHaTHoi~ITeMneparype w1~


flOJ1MK~HCT~J1Jll4M~CKHX Cn.naBOB Ge—Si B 3~ffl1CMMOCTM OT cocrana.
~aCTOTbI H IIIMPMHbI .JII4HI1ii ~pex OCHOBHbIX flMKOB He B flOJllIOM
cor~auieaisn c npe~uiiywei~paöoro~i. Pe3y.nbTamI, r~e BO3MO)KHO,
oOcyxaaioTc~ Ha OCHOB~ TeopMi~ J1OK~JIbHb!X H ~~3OH~HCHblXP~~MMOB
npHMeceii. Tax~eH3JI~F8JOTC~~ flMK~paccesi~~siH
BTO~H4HbI~Cfl~KT~blTaX ~ K~K H ~HOM~J1bHO~ none~e~~e
H~KOTO~bIX H~O~1HO~O~HbIX o6pa3L~oB.

You might also like