Synthesis of silicon-based infrared semiconductors in the Ge-Sn system using molecular chemistry methods

Article abstract of DOI:10.1021/ja0115058

Growth reactions based on a newly developed deuterium-stabilized Sn hydride [(Ph)SnD₃] with Ge₂H₆ produce a new family of Ge-Sn semiconductors with tunable band gaps and potential applications in high-speed, high-efficienc

Full text of DOI:10.1021/ja0115058

10980
J. Am. Chem. Soc. 2001, 123, 10980-10987
Synthesis of Silicon-Based Infrared Semiconductors in the Ge-Sn
System Using Molecular Chemistry Methods
Jennifer Taraci,† S. Zollner,‡,§ M. R. McCartney,| Jose Menendez,‡ M. A. Santana-Aranda,‡
D. J. Smith,‡,| Arne Haaland, Andrey V. Tutukin, Grete Gundersen, G. Wolf,† and
J. Kouvetakis*,†
Contribution from the Department of Chemistry, Arizona State UniVersity, Tempe, Arizona 85287,
Center for Solid State Science, Arizona State UniVersity, Tempe, Arizona 85287, Department of Physics
and Astronomy, Arizona State UniVersity, Tempe, Arizona 85287, Motorola Inc., Semiconductor Products
Sector, 2200 West Broadway Road, Mesa, Arizona 85202, and Department of Chemistry,
UniVersity of Oslo, P.B. 1033 Blindern, N-0315 Oslo, Norway
ReceiVed June 20, 2001
Abstract: Growth reactions based on a newly developed deuterium-stabilized Sn hydride [(Ph)SnD3] with
Ge2H6 produce a new family of Ge-Sn semiconductors with tunable band gaps and potential applications in
high-speed, high-efficiency infrared optoelectronics. Metastable diamond-cubic films of Ge1-xSnx alloys are
created by chemical vapor deposition at 350 °C on Si(100). These exhibit unprecedented thermal stability and
superior crystallinity despite the 17% lattice mismatch between the constituent materials. The composition,
crystal structure, electronic structure, and optical properties of these materials are characterized by Rutherford
backscattering, high-resolution electron microscopy, and X-ray diffraction, as well as Raman, IR, and
spectroscopic ellipsometry. Electron diffraction reveals monocrystalline and perfectly epitaxial layers with
lattice constants intermediate between those of Ge and R-Sn. X-ray diffraction in the θ-2θ mode shows
well-defined peaks corresponding to random alloys, and in-plane rocking scans of the (004) reflection confirm
a tightly aligned spread of the crystal mosaics. RBS ion-channeling including angular scans confirm that Sn
occupies substitutional lattice sites and also provide evidence of local ordering of the elements with increasing
Sn concentration. The Raman spectra show bands corresponding to Ge-Ge and Sn-Ge vibrations with
frequencies consistent with random tetrahedral alloys. Resonance Raman and ellipsometry spectra indicate a
band-gap reduction relative to Ge. The IR transmission spectra suggest that the band gap decreases monotonically
with increasing Sn fraction. The synthesis, characterization, and gas-phase electron diffraction structure of
(Ph)SnD3 are also reported.
Introduction
reports that Ge-Sn heterostructures with Sn concentrations up
to 15 at. % might exhibit tunable direct band gaps.4,5 From a
synthesis viewpoint, the major problem with Sn incorporation
into the tetrahedral Ge lattice is the large (17%) lattice mismatch
between the elements and the instability of the diamond-cubic
structure of tin (R-Sn) above room temperature.6 Accordingly,
the thermodynamic solubility limit of Sn in Ge is less than 1%
while that of Ge in Sn is practically zero.7 Alloy heterostructures
of Ge and Sn are thus unstable, and their synthesis requires
conditions that are far from equilibrium.
By using nonequilibrium synthetic methods, progress is being
made in growing Ge1-xSnx alloys via solid-source molecular
beam epitaxy (MBE).6,8-11 A major challenge facing MBE
growth of Ge1-xSnx is the propensity of Sn to segregate toward
the film surface because of the low surface energy of Sn relative
to that of Ge.12,13 It has been reported that Ge1-xSnx films can
The spectacular advances in integrated microelectronics over
the past 35 years have primarily relied on the unique properties
of silicon and its native oxide. Industrial ingenuity combined
with key technological breakthroughs has made it possible to
double the transistor density in electronic devices about every
2 years. The synthesis of band-gap-engineered epitaxial het-
erostructures on Si will continue to be important for future
generations of high-speed devices. Research and development
in this area has focused on the Si-Ge equilibrium system in
which complete solubility of the elements allows formation of
random Si1-xGex alloys with variable composition, band struc-
ture, and lattice constant.1-3 Recently, synthesis and develop-
ment of related Ge1-xSnx random alloys with similar diamond-
cubic structures has attracted considerable attention because of
† Department of Chemistry, Arizona State University.
‡ Department of Physics and Astronomy, Arizona State University.
§ Motorola Inc.
(4) He, G.; Atwater, H. A. Phys. ReV. Lett. 1997, 79, 1937.
(5) Ragan, R.; Atwater, H. A. Appl. Phys. Lett. 2000, 77, 3418.
(6) Gurdal, O.; Desjardins, R.; Carlsson, J. R. A.; Taylor, N.; Radamson,
H. H.; Sundgren, J.-E.; Greene, J. E. J. Appl. Phys. 1998, 83, 162.
(7) Bull. Alloy Phase Diagrams 1984, 5, 266 and references therein.
(8) Fitzgerald, E. A.; Freeland, P. E.; Asom, M. T.; Lowe; W. P.;
Macharrie, R. A., Jr.; Weir, B. E.; Kortan, A. R.; Thiel, F. A.; Xie, T.-H.;
Sergent, A. M.; Cooper, S. L.; Thomas, G. A.; Kimerling, L. C. J. Elect.
Mater. 1991, 20, 489.
| Center for Solid State Science, Arizona State University.
Department of Chemistry, University of Oslo.
(1) Kasper, E. Phys. Scr. 1991, T35, 232.
(2) (a) Bergman, C.; Chastel, R.; Castanel, R. J. Phase Equilib. 1992,
13, 113. (b) Dismukes, J. P.; Ekstrom, L.; Paff, R. J. J. Phys. Chem. 1964,
68, 3021.
(3) Jain, S. C. Germanium-Silicon Strained Layers and Heterostructures;
Academic Press: Boston, MA, 1994.
(9) Asom, M. T.; Fitzgerald, E. A.; Kortan, A. R.; Spear, B.; Kimerling,
L. C. Appl. Phys. Lett. 1989, 55, 578.
10.1021/ja0115058 CCC: $20.00 © 2001 American Chemical Society
Published on Web 10/16/2001

Si-Based Infrared Semiconductors
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10981
best be achieved by bombarding the layers with low-energy ions
during deposition.14,15 This process initiates better mixing of
the surface and subsurface atoms thereby facilitating incorpora-
tion of surface Sn atoms into the growing layer. A major dis-
advantage is the low thermal stability of MBE-grown Ge-Sn
materials.6,16-18 In some cases, annealing at temperatures as low
as 120-300 °C causes Sn precipitation and phase segregation.6
A plausible explanation is provided by the presence of Sn-Sn
vibrations in the Raman spectra which indicate that Sn is not
entirely incorporated in the lattice as isolated atoms but also as
small aggregates, possibly dimers, trimers, etc. These presum-
ably become the critical nuclei that initiate and facilitate forma-
tion of undesirable precipitates even at moderate processing
temperatures. One possible source of the Sn clusters in the film
might be the molecular beam generated from the Sn Knudsen
source which is likely to carry small gas-phase clusters of the
element which in turn are incorporated into the film as one unit.
Chemical vapor deposition (CVD) utilizing molecular precur-
sors that incorporate single Sn atoms in the structure should in
theory be a suitable route to Ge-Sn materials which contain
highly dispersed Sn atoms in isolated tetrahedral sites throughout
the diamond-like Ge lattice.19 Such an arrangement is expected
to be thermally more robust than the MBE-grown structures.
Until now, there have been no reports to our knowledge of
Ge1-xSnx crystalline films prepared by chemical methods (CVD)
due to lack of suitable sources. The CVD sources that are
normally used in industry to deposit silicon-based semiconduc-
tors are the classic hydrides SiH4, GeH4, Si2H6, and Ge2H6 and
their chlorinated derivatives. The analogous Sn compounds such
as SnH4 and SnH3Cl are highly unstable at room temperature
because of the significantly lower Sn-H bond energy.
In this paper we report the development of a simple Sn
hydride derivative that has the necessary thermal stability and
volatility to be a viable low-temperature CVD source. The
molecular formula is (Ph)SnD3 (where Ph ) C6H5 and D )
deuterium). Deuterium, the most common, inexpensive, and
readily available isotope of hydrogen (other than H), has been
previously used to stabilize Sb and Ga hydrides in order to
develop practical MOCVD precursors for III-V semiconduc-
tors.20 The deuterium atoms are known to increase the kinetic
stability of these molecules. For perdeuterated stannanes such
as SnD4 the enhanced stability provided by D is not yet
sufficient, and these compounds decompose at room temperature
to form Sn and D2 gas. However, replacement of one D by Ph
in SnD4 yields (Ph)SnD3 which has the desired properties, i.e.,
good thermal stability and volatility (3 Torr at 22 °C), to be a
suitable CVD source for Sn. This molecule decomposes at 250
°C in a UHV reactor to form pure Sn via elimination of volatile
Figure 1. (I) Equilibrium structure of (Ph)SnD3 obtained by DFT
calculations at the B3LYP/LANL2DZ level. (II) Transition state for
internal rotation of the SnD3 group.
and thermally robust benzene-d1 (DC6H5) and D2 byproducts
as illustrated by the following equation:
(Ph)SnD3 f D2 + D(Ph) + Sn
This paper describes the synthesis and gas-phase structure
of (Ph)SnD3 as well as reactions of the compound with Ge
hydrides to grow highly concentrated Ge1-xSnx heterostructures
with metastable diamond-cubic structures on Si(100) substrates.
These materials display unprecedented thermal stability, superior
crystallinity, and unique optical properties.
Results and Discussion
Synthesis and Structure of (Ph)SnD3. (Ph)SnD3 is prepared
by reaction of the corresponding trichloro(phenyl)stannane with
LiAlD4. It is isolated as a colorless liquid that is stable in air
for several hours, but longer exposures to atmosphere result in
decomposition to produce an unidentified polymeric solid. The
title compound was characterized by 129Sn NMR, gas-phase
FTIR, and mass spectrometry. Its molecular structure was
determined by gas electron diffraction (GED). The mass
spectrum shows an isotopic envelope centered at 197 amu as
the highest mass peak corresponding to [(D3-x)SnC6H5]+ ions.
The 119Sn NMR spectrum shows a septet centered at -346.15
ppm due to isotopic splitting of the Sn signal by deuterium atoms
1
in the SnD3 moiety. The 13C and H NMR spectra confirmed
the presence of the C6H5 ring in the structure of (Ph)SnD3. The
IR absorption bands corresponding to the Sn-D modes display
the expected isotopic shifts with respect to the Sn-H bands of
the normal hydride analogue. The Sn-H and Sn-D stretching
(10) Piao, J.; Beresfor, R.; Licata, T.; Wang, W. I.; Homma, H. J. Vac.
Sci. Technol., B 1990, 8, 221.
(11) Wong, S. S.; He, G.; Nikzad, S.; Ahn, C. C.; Atwater, H. A. J.
Vac. Sci. Technol., A 1995, 13, 216.
(12) Pukite, P. R.; Harwit, A.; Iyer, S. S. Appl. Phys. Lett. 1989, 54,
2142.
(13) Wegscheider, W.; Olajos, J.; Menczigar, U.; Dondl, U.; Abstreiter,
G. J. Cryst. Growth 1992, 132, 75.
modes are observed at 1903, 1883 and 1363, 1355 cm-1
,
respectively. Peaks corresponding to the C-H stretching modes
of the C6H5 ring are observed between 3090 and 3015 cm-1 in
the spectra of both the deuterated and the isotopically pure
compounds.
(14) Taylor, M. E.; He, G.; Atwater, H. A.; Polman, A. J. Appl. Phys.
1996, 80, 4384.
To further characterize (Ph)SnD3, a determination of the gas-
phase structure by gas electron diffraction (GED) was under-
taken at the University of Oslo. Structure optimization of
(Ph)SnD3 by density functional theory (DFT) calculations carried
out without symmetry restrictions converged to the structure I
shown in Figure 1 in which one of the deuterium atoms is lying
close to the C6 ring plane, the dihedral angle τ[C(2)C(1)SnD′]
being 4.8°. Bond distances and valence angles are listed in Table
(15) He, G.; Atwater, H. A. Appl. Phys. Lett. 1996, 68, 664.
(16) Bennett, J. C.; Egerton, R. F. Vacuum 1996, 47, 1419.
(17) Pukite, P. R.; Harwit, A.; Iyer, S. S. Appl. Phys. Lett. 1989, 54,
2142.
(18) Zhang, J.; Deng, X.; Swenson, D.; Hackney, S. A.; Krishnamurthy,
M. Thin Solid Films 1999, 357, 85.
(19) Taraci, J.; Tolle J.; McCartney, M. R.; Menendez, J.; Santana, M.;
Smith, D. J.; Kouvetakis, J. Appl. Phys. Lett. 2001, 78, 3607.
(20) Todd, M. A.; Bandari, G.; Baum T. H. Chem. Mater. 1999, 11,
547.

10982 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Taraci et al.
Table 1. Interatomic Distances (r), Root-Mean-Square Vibrational
Amplitudes (l), and Valence and Torsional Angles in (Ph)SnD3
Obtained by Quantum Chemical DFT Calculations and Gas Electron
Diffractiona
Distances
DFT
GED
re
l
ra
l
bond dists
Sn-C
213.3
5.3
213.1(7)
6.1(5)d
8.0/8.0c
4.4(3)e
Sn-D′/D′′
171.5/171.6 8.0/8.0 168.6/168.7(14)
141.8 4.6 140.3(2)
C(1)-C(2)
C(3)-C(2/4) 141.0/140.8 4.5/4.5 139.5/139.5(2) 4.4/4.3 (3)e
C(2)-H
C(3/4)-H
108.9
7.6
109.9(9)
9.6(12)f
108.8/108.8 7.5/7.5 109.8/108.8(9) 9.5/9.5(12)f
nonbonded dists
Sn‚‚‚C(2/6)
Sn‚‚‚C(3/5)
Sn‚‚‚C(4)
Figure 3. Experimental (dots) and calculated (lines) radial distribution
curves of (Ph)SnD3. Below: difference curves. Artificial damping
constants k ) 30 pm2.
310.6/311.6 8.0
444.0/444.6 7.3
308.6
440.1
493.4
7.5(7)
8.0(9)
498.1
6.8
8.7(20)
C(2‚‚‚C(4/6) 244.1/243.0 5.4/5.5 242.2/242.6
6.1/6.2(5)d
6.2/6.2(5)d
6.8(5)d
C(3)‚‚‚C(1/5) 246.2/243.4 5.5/5.4 242.3/241.2
C(1)‚‚‚C(4)
C(2)‚‚‚C(5)
structure. Calculations of the force field and normal coordinate
analysis yielded one imaginary frequency, indicating that this
Cs model corresponds to a small hump on the potential energy
surface, at least at the present computational level. Finally,
optimization of another model of Cs symmetry, II (see Figure
1), obtained by rotating the SnD3 group 30° into an orientation
where one Sn-D′′ bond is perpendicular to the ring plane,
yielded an energy 60 J mol-1 above the C1 equilibrium structure
and with one imaginary frequency. Whatever the symmetry of
the real equilibrium conformation, the calculations indicate that
the barriers to internal rotation of the SnD3 group are very small.
Least-squares refinement of a molecular model with one
deuterium atom in the ring plane and overall Cs symmetry to
the GED data yielded structure parameters in good agreement
with the calculated; see Table 1. All bond distances and valence
angles fall in the expected range. The Sn-C bond distance is
not significantly different from that in tetraphenyl stannane, rg
) 213.7(2) pm,21 and the Sn-D bond distance is indistinguish-
able from the Sn-H bond distance in SnH4, r0 ) 171.1(1) pm.22
It is well-known that introduction of an electropositive sub-
stituent to a benzene ring tends to reduce the endocyclic valence
angle at the ipso C atom,23 and the C(2)C(1)C(6) angle is indeed
calculated to be 117.9°. The experimental structure is unfortu-
nately too inaccurate to confirm the distortion.
284.8
281.1
6.1
6.1
280.3
279.2
6.8(5)d
Valence Angles
e
a
CSnD′
CSnD′′
C(2)C(1)C(6)
C(1)C(2)C(3)
C(2)C(3)C(4)
C(3)C(4)C(5)
108.6
110.7, 111.0
113(4)
115(4)
117.9
121.1
120.1
119.6
119.6(8)
119.9(7)
120.5(2)
119.6c
Dihedral Angles
τe
τa
0c
C(2)C(1)SnD′
C(2)C(1)SnD′′
R-factorsb
4.8
125.1/-115.2
(120.2c
0.047 (50 cm), 0.110 (25 cm),
0.064 (tot.)
a Distances and amplitudes in picometers and angles in degrees.
Estimated standard deviations in parentheses in units of the last digit.
bR ) [Σ(w(Iobs - Icalc)2/Σ(wIobs)2]. c Not refined. d-f Sets of amplitudes
refined with constant differences.
Growth of Ge1-xSnx Using PhSnD3. Depositions were
carried out in a custom-built ultrahigh-vacuum chemical-vapor
deposition (UHV-CVD) system. This is a load-locked hot-wall
reactor equipped with a differentially pumped mass spectrometer
capable of in situ gas sampling through several micrometer size
orifices. Precisely controlled heating of the quartz reactor is
provided by a three-zone resistance furnace. The reactor is
passivated with a thin Si layer formed by pyrolysis of disilane
at 800 °C prior to use. A high-capacity cryopump provides
ultrahigh vacuum (1 × 10-9 Torr) before and after each
deposition, and a corrosion-resistant turbopump is used during
deposition. The single-crystal (100) Si substrates were prepared
for deposition in a clean room using a modified RCA process.
This method results in formation of an oxide layer of controlled
thickness. The substrates were then dipped in 10% HF until
their surface became hydrophobic. In a typical experiment, the
substrates were loaded into a Si-coated quartz boat, pumped in
Figure 2. Experimental (dots) and calculated (lines) modified molec-
ular intensity curves of (Ph)SnD3. Below: difference curves.
1. Experimental and calculated intensities are compared in
Figure 2, with experimental and calculated radial distribution
curves in Figure 3. Calculation of the molecular force field
followed by normal coordinate analysis yielded only real
vibrational modes, confirming that this C1 structure corresponds
to a minimum on the potential energy hypersurface. When the
SnD3 group was rotated 4.8° to place D′ in the ring plane and
the optimization continued under Cs symmetry, the resulting
energy was just 7 J mol-1 above that of the optimized C1
(21) Csa´kva´ri E.; Shishkov, I. F.; Rozsondai, B.; Hargittai, I. J. Mol.
Struct. 1990, 329, 291.
(22) Kattenberg, H. W.; Oskam, A. J. Mol. Spectrosc. 1974, 51, 377.
(23) Dominicano, A. In Stereochemical Applications of Gas-Phase
Electron Diffraction, Part B Structural Information for Selected Classes of
Compounds; Hargittai, I., Hargittai, M., Eds.; VCH Publishers: Weinheim,
Germany, 1988.

Si-Based Infrared Semiconductors
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10983
Figure 4. RBS aligned and random spectra for Ge0.88Sn0.12. Dechan-
neling in Sn signal increased with increasing the Sn content in the films.
Figure 5. X-ray diffraction data (2θ scan) for Ge-Sn alloy (Sn )
12%) on Si(100). The inset shows a rocking curve of the Ge-Sn(004)
peak. Note the presence of the Ge-Sn(002) Bragg reflection which is
not typically observed in samples with Sn content lower than 10 at. %.
The dotted line shows data for bulk Ge for comparison.
the load lock to 10-6 Torr, and then inserted into the reactor.
At this point the pumping was switched to the process
turbopump and the temperature was set at 350 °C under a flow
of ultrahigh-purity H2. The appropriate quantities of the reactants
were immediately introduced into the reactor, perpendicular to
the substrate surface, through separate mass flow controllers
which were carefully calibrated for each of the gaseous reactants.
The Sn precursor was diluted with a large excess of research
grade H2, and the digermane precursor (GeH3)2 which was
obtained as a 35% mixture in H2 was used as-received from
Voltaix Chemicals.
The reaction of (GeH3)2 and (Ph)SnD3 at 350 °C on Si(100)
resulted in growth of Ge1-xSnx samples with concentrations up
to 20 at. % Sn. The films adhered extremely well to the
substrate, and they were similar in appearance with the Si
substrate. Rutherford backscattering (RBS) was used to deter-
mine the elemental composition and to detect and identify any
possible D, H, or C impurities as well as to estimate the film
thickness. The Sn to Ge elemental concentration was obtained
by utilizing 2 MeV He2+ ions (Figure 4). Forward scattering
and carbon resonance [12C(R,R)12C] experiments at 4.265 MeV
confirmed the absence of deuterium and carbon impurities
indicating complete elimination of the Ph and D ligands (the
detection limits of C and D using these methods were 0.5%
and 1% respectively).
To determine the quality of the epitaxial growth, and to
confirm substitutionality of Sn in the Ge diamond-cubic lattice,
the random and aligned (channeled) RBS spectra were recorded
and compared. Figure 4 shows the two spectra for a sample
containing 12% Sn. The most striking feature is that both Sn
and Ge atoms channel remarkably well despite the large
difference in lattice dimensions between the epilayer and the
Si substrate. The channeling of Sn is unique in our CVD grown
films and provides unequivocal proof that this element must
occupy substitutional sites in the diamond-cubic Ge structure.
The extent of substitutionality can be qualitatively assessed by
comparing the measured values of ømin (where ømin is the ratio
of the aligned vs random peak height) between Ge and Sn in
the same sample. This is nearly the same, for both Ge and Sn
in a 2 at. % Sn sample and in a 6 at. % Sn sample suggesting
that the entire Sn content is substitutional. However, the value
of ømin is higher for Sn than Ge in a 12 at. % Sn sample (see
Figure 4) meaning that a small fraction of the Sn does not
occupy random sites. Since high-resolution transmission electron
microscopy and a range of analytical methods including X-ray
diffraction (see below) rule out the presence of elemental Sn,
we believe that the discrepancy in the channeled spectra is due
to local ordering of Ge and Sn in random domains within the
crystal. Preliminary data from RBS angular scans also point to
a possible ordering. Another explanation for the observed
dechanneling is a slight distortion of the Sn atoms from their
equilibrium tetrahedral positions caused by an increase in
structural strain as the Sn content in the alloy increases. The
overall ømin for all samples is relatively high compared to the
practical limit of about 5% for a perfect Si crystal. This is
possibly due to some mosaic spread in the crystal and the
presence of misfit dislocations at the interface arising from the
lattice mismatch between the Ge-Sn layer and the Si substrate.
Nevertheless, the channeling data indicate that the sample is
monocrystalline with substantial epitaxial character and rela-
tively few structural defects.
X-ray measurements in the θ-2 θ mode using a rotating
anode generator (50 kV, 200 mA) show a single, strong, and
sharp alloy peak corresponding to the (004) reflection of the
diamond lattice. In-plane rocking scans of the (004) reflection
have a full-width-at-half-maximum of slightly less than 0.5°
indicating a tightly aligned spread of crystal mosaics. Similar
widths of the rocking curves are found for other relaxed
heteroepitaxial films with a large lattice mismatch relative to
the substrate. Since there is only one (004) peak from the
epilayer in the X-ray data, we conclude that the Sn and Ge are
completely mixed and that no phase separation between the
GeSn alloy and excess Sn has taken place (no intermetallic
compounds with intermediate compositions are reported in the
phase diagram).
Finally, it is interesting to note that another X-ray peak
emerges for samples with Sn contents higher than about 12 at.
%. The peak corresponds to the (002) reflection of the same
Ge-Sn crystal lattice, and the intensity increases substantially
with increasing Sn content. This reflection is presumably due
to the breaking of the inversion symmetry in the zinc blende
structure, and it could be related to charge transfer from Sn to
Ge. Also, this peak would be expected to appear if the symmetry
is broken due to local distortions. One would expect that a Sn
atom displaces the Ge nearest neighbors, which would activate
the GeSn (002) reflection. This supports local ordering with
the Sn atoms perhaps being arranged as second nearest neighbors
in the crystal. Figure 5 shows representative X-ray diffraction
patterns for a Ge-Sn alloy with 12 at. % Sn.

10984 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Taraci et al.
Figure 6. Cross-sectional electron micrographs of Ge0.94Sn0.06 showing
nearly perfect epitaxy and relatively defect-free coherent growth.
Complementary, cross-sectional transmission electron mi-
croscopy revealed that all samples were monocrystalline and
heteroepitaxial. Periodic arrays of misfit dislocations and
occasional {111} stacking faults originating at the interface were
observed but the remainder of the layer was remarkably defect-
free. A typical cross-sectional electron micrograph of the
interface region demonstrating heteroepitaxial growth of high
quality Ge0.94Sn0.06 is shown in Figure 6. Selected-area electron
diffraction patterns (SAED) confirm that the material has the
diamond-cubic structure. There is a marked splitting of the
diffraction spots (see Figure 6), even those corresponding to
low index planes, confirming the large difference in lattice
dimensions between the substrate and the alloy. The lattice
parameters obtained from measurements of the SAED patterns
from this sample was 5.70 Å which is intermediate to those of
pure Ge (5.646 Å) and of the metastable phase of diamond-
like R-Sn (6.489 Å). A similar value of 5.696 Å was measured
by X-ray diffraction. Optical diffractogram analysis (by Fourier
transform of the lattice-fringe images) gave similar values and
also showed that the lattice parameter did not vary locally
throughout this sample.
Energy-dispersive X-ray microanalysis (EDX) with a probe
size smaller than 1 nm was used to determine whether the
elemental composition was homogeneous on the nanometer scale
and to further verify phase purity. A detailed survey of the
compositional profiles in cross-section and plan-view geometries
showed that the constituent elements were evenly distributed
throughout the sample. Typical elemental profiles, the corre-
sponding line scans, and dark-field images showing the analyzed
region of the sample are shown in Figure 7.
Figure 7. Plan-view electron micrographs of Ge0.95Sn0.05 (top) and EDX
profile (bottom) showing that the elements are homogeneously dis-
tributed throughout the sample.
in situ EDX and electron diffraction. Nanoprobe analysis
revealed homogeneous mixing of Sn and Ge, but no Sn
precipitation or any other type of phase separation was detected
up to about 600 °C. In situ electron diffraction showed that the
annealed material remained highly crystalline. This result is in
direct contrast to earlier studies of MBE-grown Ge1-xSnx films
which showed that annealing at temperatures ranging between
100 and 300 °C caused precipitation and Sn segregation
particularly toward the film surface.6
The higher thermal stability of the CVD grown samples is
attributed to the absence of Sn clusters in the films as
demonstrated by the complete lack of Sn-Sn lattice modes in
Raman spectra (see below). The lack of Sn-Sn bonding and
the higher deposition temperatures compared to MBE imply a
different growth mechanism for our CVD process. The fact that
we deposit at 350 °C compared with -25 to 150 °C for MBE
and yet we are able to incorporate Sn into the lattice indicates
that the Sn mobility on the surface must be significantly
suppressed. This is likely to be due to the presence of large
amounts of adsorbed H and D atoms which function as
surfactants to slow the surface diffusivity of Sn and thus
facilitate incorporation of the atoms into the structure. This
process also occurs in CVD of SiGe layers on Si where adsorbed
H is well-known to suppress diffusion of Ge and thus promote
layer-by-layer growth.24 In our case, the slightly higher kinetic
stability of adsorbed D must make this species a more effective
surfactant than H. The net result is complete incorporation of
Sn into the lattice rather than surface segregation and absence
To determine the thermal stability of this material, in situ
annealing experiments between 350 and 650 °C were made in
a high-resolution analytical electron microscope. The sample
was heated in the specimen holder for 1 h at increasing 50 °C
intervals and then cooled to room temperature for analysis by
(24) Kahng, S.-J.; Ha, Y. H.; Monn, D. W.; Kuk, Y. Appl. Phys. Lett.
2000, 77, 981.

Si-Based Infrared Semiconductors
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10985
of Sn clusters and thus an overall increase in thermal stability
of the deposited material.
High-resolution electron microscopy of the sample annealed
between 600 and 650 °C revealed a distinct variation in
morphology and microstructure. The film remained crystalline
and heteroepitaxial near the interface. However, the film near
the surface was dominated by an array of crystalline and highly
coherent islands which seemed to have grown epitaxially from
the underlying layer (Figure 8). The lattice fringes of the islands
were commensurate with the (111) lattice planes of the layer.
Nanoprobe analysis and line scans revealed that the main layer
consisted of a uniform distribution of Ge with small concentra-
tions of dissolved Sn. The elemental profile was reversed inside
the island, with Sn being in higher concentration (Figure 8).
Diffractogram analysis of the lattice fringes revealed a lattice
constant for the Sn-rich particles close to that of R-Sn. These
results indicate that annealing at 650 °C has caused phase
separation with particles near the surface becoming Sn-rich
Sn1-xGex alloys with diamond-cubic structure while the bulk
layer remained a Ge-rich alloy of the same structure.
High-resolution Raman spectroscopy was used to identify and
characterize the local bonding environment of the Sn atoms.
The Raman spectrum of the Sn-Ge films was measured using
the 488 nm line of an argon laser. Each sample showed a strong
peak assigned to Ge-Ge lattice vibrations which is downshifted
relative to the Raman peak of pure Ge (Figure 9). A weak peak
around 240 cm-1 was also observed. Since the Sn-Ge band in
Sn-rich Sn-Ge alloys has been observed near 220 cm-1 and a
higher frequency is expected in Ge-rich alloys, we assign the
240 cm-1 peaks to the Sn-Ge vibrations in our sample.25 The
observation of Sn-Ge bands provides strong spectroscopic
evidence for the presence of tetrahedral Ge-Sn coordination
in our films and supports the formation of an extended random
alloy with the diamond-cubic structure. Any additional peaks
corresponding to Sn-Sn vibrations were not detected in the
Raman spectra, which is consistent with the incorporation of
isolated Sn atoms during CVD growth using (Ph)SnD3.
In addition to the structural information obtained from the
Raman frequencies, Raman spectroscopy can be used to study
the electronic structure of the alloys by measuring the intensity
of the Raman peaks as a function of the laser wavelength. In
pure Ge, it has been shown that the Raman cross section
undergoes a resonant enhancement for laser photon energies
near the E1 and E1 + ∆1 gaps. A broad peak about 2.22 eV,
halfway between the E1 and E1 + ∆1 transitions, has been
measured by Cerdeira et al.26 Figure 9 shows similar measure-
ments for our Ge0.95Sn0.05 alloy and a Ge reference. To extract
the Raman cross-section from the measured intensity, a cor-
rection must be applied which depends on the dielectric function
of the material.27 Of course, the dielectric function of SnGe
alloys is not known. Hence we adopted the following proce-
dure: we rigidly shifted the known dielectric function of pure
Ge (see ref 28) to lower energy by an amount ∆E and used the
shifted values to compute the correction for the Raman data.28
The value of ∆E was chosen so that it self-consistently agreed
with the shift in the peak of the resonant profile. As seen in
Figure 9, we find ∆E ) 100 meV. The error in the shift
Figure 8. (a) Cross-sectional electron micrograph of Ge0.95Sn0.05
annealed at 650 °C showing segregation of an Sn-rich island near the
film surface. (a) High-esolution electron micrograph showing lattice
fringes of diamond-cubic Ge1-xSnx layer and diamond-cubic Sn1-xGex
island. (c) EDX elemental profile across the Ge-rich and Sn-rich sections
of the layer.
(25) Mene´ndez, J. Characterization of Bulk Semiconductors using Raman
Spectroscopy. In Raman Scattering in Materials Science; Weber, W. H.,
Merlin, R., Eds.; Springer: Berlin, 2000; Vol. 42, p 55.
(26) Cerdeira, F.; Dreyrodt, W.; Cardona, M. Solid State Commun. 1972,
10, 591.
(27) Richter, W. Resonant Raman Scattering in Semiconductors. In
Springer Tracts in Modern Physics; Høhler, G., Ed.; Springer-Verlag:
Berlin, Heidelberg, New York, 1976; Vol. 78, p 121.
determined from this procedure is probably quite large, but it
is apparent that a downshift of the E1 and E1 + ∆1 transitions
takes place. Similarly we observe a red shift of the E2 critical
point of about 50 meV from a derivative analysis of spectro-
(28) Aspnes, D. E.; Studna, A. A. Phys. ReV. 1983, B27, 985.

10986 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Taraci et al.
Figure 10. IR transmission spectra of the Si substrate and Ge-Sn
samples. At intermediate energies (6000-9000 cm-1) the transmission
through the films decreases monotonically indicating that the band gap
decreases with increasing Sn concentration.
result in increased absorption at a fixed photon energy above
the band gap. The data also imply that incorporation of Sn in
Ge makes the material a more sensitive infrared photodetector.
Further experiments and data analysis aimed toward determi-
nation of the absolute band gaps and absorption coefficients
are in progress.
Concluding Remarks
We have developed for the first time a molecular method to
deposit highly concentrated semiconductors in the Ge-Sn alloy
system with tunable band gaps in the infrared. Transmission
FTIR measurements of layers grown on Si show that transmis-
sion decreases (absorption increases) monotonically with in-
creasing Sn content indicating that the band gap of the alloys
decreases with increasing Sn fraction. Our synthetic approach
is based on reactions of a newly prepared deuterium-stabilized
Sn hydride, (Ph)SnD3. This compound is designed to have the
necessary volatility and reactivity to interact at low temperatures
with common reagents such as (GeH3)2 to yield pure products
via elimination of extremely stable gaseous byproducts, includ-
ing benzene, hydrogen, and deuterium. This method is currently
applied to explore formation and systematic structural charac-
terization of a wide range of concentrations across the entire
Sn-Ge compositional range. This includes growth of epitaxial
alloys on suitable substrates and buffer layers that are predicted
to have direct band gaps.
Figure 9. (Top) Raman spectrum showing Ge-Ge and Ge-Sn
vibrations of Ge0.95Sn0.05. The dotted line shows data for pure Ge for
comparison. (Bottom) Excitation profiles for the Raman peak in pure
Ge (squares) and the Ge-Ge Raman peak in the Ge0.95Sn0.05 alloy
(circles). The Raman cross section was extracted from the measured
intensities using a correction discussed in the text. The downshift of
the resonant maximum for the alloy is attributed to a decrease of the
E1 and E1 + ∆1 transition energies as a function of the Sn concentration.
This indicates a reduction of valence band-conduction band energy
separation in the alloy relative to pure Ge.
scopic ellipsometry data.29 These data suggest that the band gap
of the alloys is also reduced relative to Ge, although it is not
possible from this information to determine whether the expected
indirect-direct transition has taken place.
This narrowing in band gap prompted us to conduct transmis-
sion IR measurements to examine the effect of increasing Sn
concentration on absorption. This should provide information
of possible band gap variation with composition in the alloy.
The transmission spectra of four Ge-Sn films with Sn content
ranging from 9 to 18% are shown in Figure 10. The spectrum
of the Si substrate is also provided as a reference. Note that the
absorption results are very similar for low energies, below the
GeSn band gap, and at high energies, above the Si band gap,
as expected. However, the transmission drops monotonically
with increasing Sn content at intermediate energies. The simplest
explanation for these results is a decrease of the band gap of
Ge1-xSnx as a function of x, since such a reduction is likely to
Experimental Section
Synthesis (Ph)SnD3. A solution of (Ph)SnCl3 (10 g, 33 mmol) in
20 mL of ethyl ether was added, dropwise, to a solution of LiAlD4
(3.5 g, 83 mmol) in 150 mL of ethyl ether which was cooled to -70
°C. The resulting solution was stirred for 30 min at -70 °C and at
-20 °C for an additional 60 min during which time the solution turned
to a dark brown mixture. The mixture was filtered under N2, and a
clear filtrate was isolated and separated by distillation. The volatiles
from the filtrate were passed through traps held at -20 and -196 °C
which retained (Ph)SnD3 (2 g, 10 mmol, 30% yield) and ether,
respectively. (Ph)SnD3 has a vapor pressure of 2-3 Torr at 22 °C.
IR (gas phase): 3079 (w), 3067 (w), 3025 (w), 1363 (m), 1355 (st),
726 (w), 691 (w), 489 (vst), 458 (w) cm-1 119Sn NMR [toluene-d8,
.
referenced to Sn(CH3)4]: -346.15 ppm (sept, J ) 294 Hz, SnD3). The
seven peaks observed due to isotopic splitting from deuterium are
-350.87, -348.25, -347.72, -346.15, -344.57, -342.99, and
(29) Junge, K. E.; Lange, R.; Dolan, J. M.; Zollner, S.; Dashiell, M.;
Orner, B. A.; Kolodzey, J. Appl. Phys. Lett. 1996, 69, 4084. Vina, L.;
Hochst, H.; Cardona, M. Phys. ReV. B 1985, 31, 958.

Si-Based Infrared Semiconductors
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10987
-341.41 (ppm). EIMS (m/e): isotopic envelopes centered at 120
(DSn+), 197 (DSnC6H5+), and 78 (C6H6+). For comparison (Ph)SnH3
was prepared using the same procedure described above for the (Ph)-
SnD3 analogue. IR (gas phase): 3087 (w), 3067 (w), 3013 (w), 1903
(m), 1883 (m), 726 (w), 687 (s), 691 (w), 505 (w), 469 (w).30 The
peaks corresponding to Sn-H stretching and bending vibrations display
the calculated isotopic shifts with respect to those for (Ph)SnD3.
Gas-Phase Structure. The structure of PhSnD3 was optimized using
the GAUSSIAN program package, the B3LYP functional, and a
standard LANL2DZ basis.31 Molecular force fields were scaled as
described in ref 32 and transferred to the program ASYM40 for
calculation of root-mean-square amplitudes of vibration and vibrational
correction terms, D ) re - ra, at the temperature of the electron
diffraction experiment.33 The gas electron diffraction pattern was
recorded on the Baltzers KDG2 unit at the University of Oslo34 with a
metal inlet system at 22 °C. Exposures were made at nozzle-to-plate
distances of about 25 cm (3 plates) and 50 cm (3 plates) using the
BAS 1800 II imaging plates of Fuji. The plates were scanned on a
BAS 1800 II imaging plate reader and the intensity data processed using
a program written by S. Samdal, D. Shorokhov, and T. G. Strand.
Atomic scattering factors were taken from ref 35. Backgrounds were
drawn as least-squares adjusted polynomials to the difference between
the total experimental intensity and the calculated molecular intensity,
and the molecular structure was refined by least-squares calculations
on the intensity data using the program KCED 26 written by G.
Gundersen, S. Samdal, H. M. Seip, and T. G. Strand.
Inspection of the equilibrium structure obtained by DFT optimization
of PhSnD3 without imposition of symmetry showed that the (Ph)Sn
fragment adheres closely to C2V symmetry, the deviation in bond
distances being less than 0.1 pm and that in angles less than 0.1°.
Similarly the SnD3 fragment was found to approach Cs symmetry. Least-
squares structure refinement to the GED data were therefore based on
a model of Cs symmetry in which one Sn-D bond is lying in the ring
plane. Six independent structure parameters were refined. The first four
were the following: (i) the Sn-C bond distance; (ii) a common scale
factor for the three symmetry-inequivalent C-C bond distances (each
C-C bond distance assumed to be equal to the calculated bond distance
multiplied by this scale factor); (iii) a common scale factor for the
three symmetry-inequivalent C-H bond distances; (iv) a common scale
factor for the two symmetry-inequivalent Sn-D bond distances. After
the C-C bond distances have been fixed, the structure of the C6 ring
is determined by two valence angles, e.g. the C(2)C(1)C(6) and C(3)C-
(4)C(5) valence angles. The former was refined as an independent
parameter, the latter was fixed at the calculated value. The last structure
parameter to be refined was the CSnD angle.
In addition to the six structure parameters we refined six vibrational
amplitudes as indicated in Table 1. Nonrefined amplitudes were fixed
at the calculated values. Introduction of vibrational correction terms
(D) led to considerably poorer fit between calculated and experimental
intensities. The refinements were therefore carried out on a geo-
metrically consistent ra structure. The refinements proceeded without
difficulties to yield the best values listed in Table 1. The estimated
standard deviations have been multiplied by a factor of 2.0 to include
added uncertainty due to data correlation and further expanded to
include an estimated scale uncertainty of 0.1%.
(30) Kuivila, H. G.; Beumel, O. F. J. Am. Chem. Soc. 1961, 83, 1246.
(31) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria G. E.; Robb,
M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.;
Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A.
D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi,
M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.;
Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick,
D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.;
Ortiz, J. V.; Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz,
P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-
Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe,
M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.;
Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian
98, revision A.7; Gaussian, Inc.: Pittsburgh, PA, 1998.
(32) Scott, A. P.; Radom, L. J. Phys. Chem. 1996, 100, 16502.
(33) Hedberg, L.; Mills, I. M. J. Mol. Spectrosc. 1993, 160, 117.
(34) Zeil, W.; Haase, J.; Wegmann, L. Z. Instrumentenkd. 1966, 74, 84.
(35) Ross, A. W.; Fink, M.; Hilderbrandt, R. L. In International Tables
for Crystallography; Kluwer Academic Publishers: Dordrecht, The Neth-
erlands, 1992; Vol. C, p 245.
Acknowledgment. The work at ASU was supported by the
National Science Foundation (Grant DMR-9902417) and the
Army Research Office (ARO). We are grateful to the Norwegian
Research Council (Program for Supercomputing) for a grant of
computer time. M.A.S.-A. acknowledges partial support from
CONAC y T-Mexico.
JA0115058