Molecular synthesis of high-performance near-ir photodetectors with independently tunable structural and optical properties based on Si-Ge-Sn

Article abstract of DOI:10.1021/ja309894c

This Article describes the development of an optimized chemistry-based synthesis method, supported by a purpose-built reactor technology, to produce the next generation of Ge_1-x-ySi_xSn_y materials on conventional Si(100) and Ge(100) platforms at gas-source molecular epitaxy conditions. Technologically relevant alloy compositions (1-5% Sn, 4-20% Si) are grown at ultralow temperatures (330-290 C) using highly reactive tetragermane (Ge₄H₁₀), tetrasilane (Si₄H₁₀), and stannane (SnD₄) hydride precursors, allowing the simultaneous increase of Si and Sn content (at a fixed Si/Sn ratio near 4) for the purpose of tuning the bandgap while maintaining lattice-matching to Ge. First principles thermochemistry studies were used to explain stability and reactivity differences between the Si/Ge hydride sources in terms of a complex interplay among the isomeric species, and provide guidance for optimizing process conditions. Collectively, this approach leads to unprecedented control over the substitutional incorporation of Sn into Si-Ge and yields materials with superior quality suitable for transitioning to the device arena. We demonstrate that both intrinsic and doped Ge_1-x-ySi_xSn_y layers can now be routinely produced with defect-free microstructure and viable thickness, allowing the fabrication of high-performance photodetectors on Ge(100). Highlights of these new devices include precisely adjustable absorption edges between 0.87 and 1.03 eV, low ideality factors close to unity, and state-of-the-art dark current densities for Ge-based materials. Our unequivocal realization of the molecules to device concept implies that GeSiSn alloys represent technologically viable semiconductors that now merit inclusion in the class of ubiquitous Si, Ge, and SiGe group IV systems.

Full text of DOI:10.1021/ja309894c

Article
pubs.acs.org/JACS
Molecular Synthesis of High-Performance Near-IR Photodetectors
with Independently Tunable Structural and Optical Properties Based
on Si−Ge−Sn
Chi Xu,‡ Richard T. Beeler,† Gordon J. Grzybowski,† Andrew V.G. Chizmeshya,† David J. Smith,‡
Jose Menendez,‡ and John Kouvetakis*,†
́
́
†Department of Chemistry and Biochemistry, Arizona State University, Tempe, Arizona 85287-1604, United States
‡Department of Physics, Arizona State University, Tempe, Arizona 85287-1504, United States
ABSTRACT: This Article describes the development of an
optimized chemistry-based synthesis method, supported by a
purpose-built reactor technology, to produce the next generation
of Ge1−x−ySixSny materials on conventional Si(100) and Ge(100)
platforms at gas-source molecular epitaxy conditions. Technologi-
cally relevant alloy compositions (1−5% Sn, 4−20% Si) are grown
at ultralow temperatures (330−290 °C) using highly reactive
tetragermane (Ge4H10), tetrasilane (Si4H10), and stannane (SnD4)
hydride precursors, allowing the simultaneous increase of Si and Sn
content (at a fixed Si/Sn ratio near 4) for the purpose of tuning the
bandgap while maintaining lattice-matching to Ge. First principles
thermochemistry studies were used to explain stability and reactivity
differences between the Si/Ge hydride sources in terms of a
complex interplay among the isomeric species, and provide guidance for optimizing process conditions. Collectively, this
approach leads to unprecedented control over the substitutional incorporation of Sn into Si−Ge and yields materials with
superior quality suitable for transitioning to the device arena. We demonstrate that both intrinsic and doped Ge1−x−ySixSny layers
can now be routinely produced with defect-free microstructure and viable thickness, allowing the fabrication of high-performance
photodetectors on Ge(100). Highlights of these new devices include precisely adjustable absorption edges between 0.87 and 1.03
eV, low ideality factors close to unity, and state-of-the-art dark current densities for Ge-based materials. Our unequivocal
realization of the “molecules to device” concept implies that GeSiSn alloys represent technologically viable semiconductors that
now merit inclusion in the class of ubiquitous Si, Ge, and SiGe group IV systems.
The original synthesis of Ge1−x−ySixSny was conducted via
reactions of SiH3GeH3 with SnD4 using ultrahigh vacuum CVD
to produce the first generation of samples that allowed
thorough initial characterization of the intrinsic properties of
the new system to be performed.10 This approach exploits the
high reactivity of SiH3GeH3, which is also more compatible
with the low temperatures required for growth of single-phase
monocrystalline structures with high Sn contents (note that the
degree of Sn incorporation and the growth temperature are
typically inversely related).10 The routine use of SiH3GeH3 in
this process was, however, found to be problematic from a
device development perspective because the fixed Si/Ge ratio
in the precursor limited stoichiometry control needed to
precisely tune alloy compositions with respect to strain. In
addition, this compound is not yet commercially available in
sufficient quantities and purity to ensure routine large-scale
fabrication of thick layers with suitable microstructure and their
deployment on large-area platforms as required for high-
I. INTRODUCTION
For the past 50 years, simple molecular compounds such as
silane (SiH4) and germane (GeH4) have been the chemical
vapor deposition (CVD) precursors of choice in the develop-
ment of silicon and silicon−germanium technologies, which
have dominated the field of electronics. More recently, the need
for lower deposition temperatures and more complex group IV
materials, such as those incorporating Sn, has stimulated
research into higher-order silanes and germanes, most notably
digermane (Ge2H6), and trisilane (Si3H8), as well as Sn
compounds such as SnD4 and SnCl4, which are now
commercially available.1−3 One of the most intriguing by-
products of recent work in this area is the synthesis of
crystalline Ge1−x−ySixSny alloys, which represent the first viable
ternary semiconductor system among group IV elements with
independently tunable lattice parameter and electronic
structure.4−6 This flexibility confers the material a great
potential for multijunction solar cells in which the multiple
band gaps must be independently tuned to optimize efficiency,
while the lattice parameters must be kept the same across the
structure to minimize formation of deleterious defects.7−9
Received: October 6, 2012
Published: December 14, 2012
© 2012 American Chemical Society
20756
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
Figure 1. Schematic illustration of the “molecules, to materials, to functional devices” paradigm emphasized in this Article. The reactive Si4H10 and
Ge4H10 compounds shown in the top left are used to synthesize individual component layers (bottom left), which are then combined into seamlessly
matched stacks (bottom right), culminating in an integrated device architecture with desired operational properties.
performance low-cost devices such as multijunction photo-
voltaics.
for introducing deleterious impurities into complex architec-
tures of the grown devices, thereby diminishing their
performance potential. A common example includes the
incorporation of reactant residues from prior runs into
subsequently grown components of the device (i.e., residual
memory affects). In view of the novelty of the Si4H10 and
Ge4H10 precursors for the growth of these crystalline
semiconductors, we carried out ab initio thermochemistry
studies to elucidate reactivity trends and identify optimal
process conditions. These studies reveal a complex interplay
between the equilibrium properties of the various reacting
isomers at typical reaction conditions. Perhaps most
importantly, we demonstrate that the crystalline Ge1−x−ySixSny
systems grown using Si4H10 and Ge4H10 automatically lattice-
match the underlying Ge wafers (the quintessential substrate
for high efficiency PV) and that they possess defect-free
microstructure, superior morphology, and viable thickness as
needed for applications in multilayer devices. Furthermore, our
approach offers the potential for additional improvements: (a)
process scale-up to industry-standard large-area wafers, (b)
improved efficiency and cost by using smaller amounts of
expensive chemicals, (c) increased throughput by drastically
reducing the down time due to chamber preparation for
subsequent runs, and (d) simplified and consistent wafer
cleaning.
In this Article, we describe the entire process development
from “molecules through materials to prototype device” (see
Figure 1) to produce intrinsic and doped functional
components, which in turn allow the assembly of working
photodiodes on Ge that exhibit the expected enhancements in
optoelectronic performance. This is demonstrated by fabricat-
ing and measuring an entire series of photodetectors exhibiting
record low dark current densities as compared to similar Ge-
based devices grown on Si(100), as well as precisely tunable
absorption edges over a broad energy range in the IR between
0.87 and 1.03 eV, thereby creating exciting opportunities in
optoelectronics, including photovoltaics.
Most recently, it was shown that alloys with Sn
concentrations below ∼2% could be fabricated by replacing
the exotic SiH3GeH3 precursor with commercially available
Ge2H6 and Si3H8 sources, which enabled the development of
micrometer thick films with reproducible control of atomic
content and dopant densities directly on Si and Ge
substrates.7,11−13 Materials in this limited composition range
(%Sn < 2) were then used to fabricate rudimentary p−n
junctions that exhibited promising collection efficiencies
approaching 80%.7 This suggests that it may be possible to
create working devices with tunable optical response beyond
that of Ge by expanding the Sn content in the alloy without
resorting to the use of Si3H8/Ge2H6 mixtures or the
problematic SiH3GeH3. These limitations, and in particular
the lack of flexible delivery agents for Si and Ge constituents,
have therefore prevented the exploration of the full potential of
the ternary Ge1−x−ySixSny system as a viable semiconductor for
widespread device applications.
In this Article, we report a comprehensive study of a new
molecular strategy that allows synthesis of device quality group
IV Si−Ge−Sn semiconductors with demonstrated optoelec-
tronic capabilities well beyond those of state-of-the-art Ge-
based devices.14 We have shown that the lower-reactivity
Si3H8/Ge2H6 mixtures, which have so far limited the Sn
incorporation, can be replaced by higher reactivity Si4H10 and
Ge4H10 molecules. The latter are also more chemically
compatible with Sn hydrides such as SnD4 and react at
unprecedented low temperatures of 285−320 °C to produce
the next generation of Ge1−x−ySixSny alloys possessing the range
of Si/Sn composition (up to 6% Sn and 22% Si) required for
structure/property tuning. To exploit the enhanced reactivity of
Si4H10 and Ge4H10 in the context of device development, we
designed a custom single-wafer reactor operating under gas-
source molecular beam epitaxy (GS-MBE) conditions, thereby
enabling reproducible control of atomic content and thickness
at the monolayer level, necessary for building quantum well-
based structures. This approach also allows control and
ultimately the elimination of secondary reactions responsible
20757
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
Figure 2. Schematic representations of the single wafer gas-source MBE reactor showing the internal structure of the CVD module, which includes
the heater, wafer holder, and injection nozzle. A three-dimensional CAD rendering of the apparatus is shown in the inset.
and uniform radiant heating of the bulk substrate. Finally, the
reactor is also fitted with in situ diagnostics, including a
RHEED system to monitor bonding at the surface and a mass
spectrometer (300 amu) to characterize the gaseous species
present during growth and within the residual atmosphere at
UHV conditions.
II. RESULTS AND DISCUSSION
A. Growth and Structural Characterization. The growth
apparatus used in this study (see Figure 2) is a custom-built
deposition system with gas-source epitaxy capabilities in which
a low-pressure CVD module (heater, wafer stage, gas injection
manifold) is housed entirely within a UHV environment. The
heater and sample stage are components of a single-wafer
process unit that is attached to the top of a spherical stainless
steel chamber. This chamber is fitted with a pumping stack
comprised of high capacity turbo pumps backed by dry pumps.
A cryo pump is used to maintain background vacuum levels of
∼10−10 Torr, which are necessary for “flashing” Si wafers to
desorb the native oxide. The system is equipped with a load
lock that allows transferring the substrate wafers into the
chamber under UHV conditions. The sample stage comprises a
rotating wafer holder designed to accommodate up to 4-in.
substrates. Heating is provided by a coiled graphite element
enclosed within a cylindrical quartz jar that is differentially
pumped down to 10−10 Torr. This is a unique arrangement
designed to: (1) isolate the heater from the reactor ambient for
the purpose of protecting the filaments from the corrosive CVD
environment, and (2) completely eliminate cross-contamina-
tion. The injection nozzle is mounted at the bottom of the
chamber through a leak valve, and it is directed upward toward
the wafer surface. The nozzle is terminated with a showerhead
configured to dispense the gaseous reactants substantially
normal to the downward facing wafer positioned upside down
in the holder. The gas manifold is movable over an 8 in. length
scale in the vertical direction, allowing the outlet of the nozzle
to be precisely positioned below the wafer at a predetermined
operating distance. The latter is optimized to produce films at
the highest possible growth rate exhibiting thickness uni-
formities of 90% or better, depending on wafer size and
temperature. The position of the quartz enclosure can also be
adjusted in the vertical direction by a precisely controlled
mechanism that is configured to independently set the distance
between the heater and the back side of the wafer. The latter
can be reduced to distance as little as 2 mm to ensure efficient
Using the above reactor, we initially conducted a series of
control experiments on 4 in. diameter Si(100) and Ge(100)
substrates. The Ge wafers employed here are those typically
used as platforms in commercial multijunction photovoltaic
devices, 150 μm-thick, off-cut by 6° toward (111), and doped
with Ga acceptor atoms (p = 6 × 1017 cm−3). Each “epi-ready”
substrate was loaded as received in the reactor and then heated
to 550 °C for 5 min under high vacuum to remove the surface
oxide. In the case of the Si wafers, an RCA process was used to
remove organic/metallic impurities. The substrates were then
dipped in HF/methanol solutions to passivate the surface with
hydrogen, loaded into the CVD chamber, and outgassed at 450
°C until the pressure was restored to the base level (10−10
Torr), and finally flashed at 850 °C to remove residual
contaminants.
The Si4H10 compound was obtained from Voltaix Corp. and
was used as received. The Ge4H10 analogue is produced by
thermolysis of Ge2H6 in a flow system at 250 °C as described
by the sequence of idealized equations below. The compound
was isolated by distillation as a stable colorless liquid product in
gram quantities sufficient for SiGeSn deposition studies.
2Ge2H6 → Ge3H8 + GeH4
(1)
Ge2H6 + Ge3H8 → Ge4H10 + GeH4
(2)
Using ab initio thermochemistry (see below), we find that the
driving force in these reactions is the creation of GeH4, which is
more easily produced by decomposition of Ge3H8 above 300
°C yielding Ge4H10 (see eq 2). Stock gas mixtures were
prepared by combining the Si4H10, Ge4H10, and SnD4
precursors in a 3 L container. The molecular flux was admitted
into the chamber at a steady flow rate, controlled by a precision
20758
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
leak valve. A turbo pump (backed by a dry pump) was used to
maintain a constant pressure of 1 × 10−4 Torr during growth,
while the temperature of the samples was estimated from
accurate readings (285−320 °C) of a heater thermocouple
positioned inside a quartz enclosure, 5 mm removed from the
backside of the wafer. Accordingly, the actual film temperatures
are expected to be slightly lower. It should be noted that under
the significantly reduced thermal budgets employed here the
emissivity of Si and Ge is beyond the detection threshold of a
single color pyrometer typically used to determine the substrate
surface temperature.
A wide range of single-layer reference samples, both intrinsic
and P-doped, was initially produced, on both Si(100) and
Ge(100) substrates, for the purpose of establishing reliable and
reproducible reaction conditions. The deposition temperature
and molar ratio of the coreactants were optimized to ensure the
growth of films with the desired compositions and dopant
levels, while exhibiting large thicknesses, flat surfaces, and high-
quality microstructure. Subsequently, all device-related develop-
ments were pursued exclusively on samples grown on Ge
substrates. Using the optimized reference samples conditions, a
range of device stacks of intrinsic and n-layers were then
codeposited on off-cut Ge wafers to assemble Ge1−x−ySixSny/
Ge(100) pin diodes that were processed into photodetector
devices. The intrinsic layers in these structures were synthesized
by reactions of Si4H10, Ge4H10, and SnD4 to yield nominal
stoichiometries at unprecedented low temperatures as shown in
Table 1. The n-type overlayer in each sample was grown under
components did not react or decompose even when stored
for an extended period of time. The relative amount of H2 in a
given formulation did not influence the overall film
stoichiometry, although higher diluents resulted in lower
growth rates, as expected. The latter also depended on
temperature and were estimated to be as high as 10 nm/min
for the 315 °C depositions. Finally, we note that in the above
list of samples in Table 1, the Si and Sn content is increased
while keeping the Si/Sn ratio close to 3.7 for the purpose of
tuning the band gap above that of Ge (0.8 eV), while
maintaining the lattice parameter near that of germanium
(5.658 Å). This ensures close lattice matching of the device
stack with the underlying wafer.
All samples grown on Si(100) or Ge(100) (e.g., both
reference layers and device stacks) were initially examined using
Nomarski microscopy and found to be optically featureless.
Rutherford backscattering spectroscopy (RBS), high-resolution
X-ray diffraction (HR-XRD), atomic force microscopy (AFM),
and cross-sectional electron microscopy (XTEM) character-
ization were then performed to establish that micrometer-thick
films (single and double layers) could be readily produced with
atomically flat surfaces.
Figure 3 shows the RBS random and aligned plots of a
Ge0.95Si0.04Sn0.010 reference sample deposited on Si(100) at 320
Table 1. Compositions of Representative Samples Grown on
a
Ge, and Processed into Photodetectors
T (oC)
%Ge
%Si
%Sn
320
320
315
315
300
290
285
285
92.5
91.3
90.0
88.6
84.0
83.0
78.0
75.0
6.3
7.2
1.2
1.5
2.0
2.0
3.0
4.0
4.0
5.0
8.0
9.4
13.0
13.0
18.0
20.0
a
The corresponding growth temperatures (first column) were
obtained from thermocouple readings.
the same conditions in situ by introducing an appropriate
amount of the single source P(GeH3)3 into the reaction
mixture, to yield virtually identical Si, Ge, and Sn compositions
and active dopant levels in the range of 2 × 1019/cm3. We
found that the atomic content in the alloys was controlled by
the relative amounts of the gaseous precursors in the reaction
mixtures. The latter were formulated in a manner in which the
amount of Ge4H10 was always kept constant at 1.5 Torr per unit
volume, which represents its maximum vapor pressure at 22 °C.
The pressure of Si4H10 and SnD4 was then simultaneously
adjusted to achieve an optimal ratio, which gave the target Si
and Sn fraction in the layers reproducibly. Perhaps surprisingly,
a significant excess of Si4H10 was typically found necessary to
systematically increase the alloy’s Si concentration in the 6−
20% Si regime of interest (ab initio thermochemistry is used to
explore the origin of this behavior, see below). In all cases, the
gas mixture was diluted with research grade H2 at a total final
pressure of 30−40 Torr, and checked prior to growth by gas
infrared absorption, which confirmed that the individual
Figure 3. XTEM micrograph of a Ge0.95Si0.04Sn0.01 layer grown directly
on Si(100) and annealed at 700 °C showing the presence of defects
resulting from the lattice mismatch of the epilayer with Si(100).
(Inset) Random and channeled RBS spectra indicating a high level of
epitaxial alignment and single phase structure of the epilayer.
°C with an Si/Sn ratio of ∼4. The data indicate single-phase
material in full epitaxial registry with the underlying Si as shown
by the high level of RBS channeling. The film is strain-free
relative to Si(100) as indicated by the relaxation line passing
through the center of the off axis (224) Bragg peaks (not
shown). The corresponding XTEM image of the entire layer
(Figure 3) shows occasional defects arising from the lattice
mismatch between the epilayer and the substrate. In general,
reference layers (or devices) grown on Ge under the same
reactions conditions exhibited stoichiometry identical to those
on Si, indicating that the ratio of coreactant on the growth front
20759
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
controls the final composition irrespective of the growth
platform employed.
Figure 4 shows typical RBS and XRD plots from a
Ge0.78Si0.18Sn0.04 device film as-grown on Ge at 310 °C with a
reflection of the n-Ge0.78Si0.18Sn0.04 epilayer appears as a
shoulder on the combined peak due to the Ge wafer and i-
Ge0.78Si0.18Sn0.04 epilayer. This indicates the expected reduction
in lattice parameter due to intentional enhancement in Si
content for the purpose of creating a higher bandgap capping
layer, to optimize overall device performance. Perhaps most
importantly, the width of the “device” peak is similar to that of
bulk Ge, indicating minimal mosaic spread and flawless
heteroepitaxial alignment, which is further corroborated by
the presence of thickness fringe patterns on both sides of these
2θ/ω features. Collectively the XRD plots indicate superior
crystallinity in the as-grown material and close similarity in the
lattice constants to those of Ge (a0 = 5.658 Å), as evidenced by
a close correspondence of the epilayer and substrate peak
maxima. These results are a testament to the precise nanoscale
control of both stoichiometry and microstructure afforded by
our ultralow temperature synthesis using designer hydrides with
compatible reactivities.
Secondary ion mass spectrometry (SIMS) depth profiles of
selected samples showed a homogeneous distribution of the Si,
Ge, and Sn atoms. The plots also revealed a sharp transition in
the P profile across the heterojunctions, reaching a constant
value of ∼2 × 1019/cm3 within the top layer. AFM scans
revealed a root-mean-square (RMS) roughness of ∼ less than
0.7 nm for 20 × 20 μm2 areas in samples across the entire
composition series, indicating a homogeneously smooth surface
in all cases. The latter is corroborated by XTEM observations,
which revealed the presence of monocrystalline layers with
atomically flat surfaces and cubic diamond-like structures akin
to that of the underlying Ge platform, as illustrated in Figure 5
by micrographs of two representative GeSiSn/Ge pin samples.
The top and bottom panels of the figure are diffraction contrast
images obtained from 3% and 5% Sn films, respectively, both
showing a flat surface and uniform layer morphology with
typical thickness variations appearing as alternating bright and
dark contrast bands. In both cases, the entire layer is found to
be completely devoid of threading dislocations on a lateral scale
of many micrometers. Virtually identical defect-free micro-
structures are observed by XTEM within full layer segments
throughout the cross-sectional samples. The middle panel of
Figure 4. (Top) Standard 2 MeV RBS spectra of a 750 nm thick
Ge0.78Si0.18Sn0.04 epilayer illustrating a high degree of crystallographic
alignment with the Ge wafer. (Bottom) XRD 2θ/ω plot of the same
sample showing a strong (004) peak due to overlapping contributions
from the Ge wafer and i-GeSiSn components of the device as well as a
weak shoulder attributed to the n-GeSiSn overlayer. (inset) The (224)
reciprocal space map due to i-GeSiSn/Ge(100) is perfectly aligned in
the vertical direction with that of the n-GeSiSn overlayer. Collectively
the XRD data reveal a very close degree of lattice matching and similar
crystalline quality between the device structure and the Ge platform.
the figure is a high-resolution image of the Ge0.75Si0.20Sn0.05
/
Ge(100) interface region showing a sloped heterojunction
profile (indicated by arrows) relative to the (111) planes, as
expected from the off-cut wafer geometry. The overlayer is fully
conformal to the underlying terraced surface of the wafer, while
the two materials exhibit a flawless epitaxial registry due to
close lattice matching.
thickness of ∼750 nm. The random RBS plot (red trace) is
used to determine the composition and estimate the thickness,
while the channeled analogue (blue trace) indicates single-
phase monocrystalline material grown with epitaxial alignment
on the underlying wafer. The RBS analyses reveal a uniform Si/
Sn content throughout the films with a margin of error of 0.1%
Sn and 1% Si. The larger uncertainly in the Si fraction is due to
the overlap of the Si signal with the dominant Ge wafer
background, precluding a precise fit of the Si step height in the
spectrum using the program RUMP.15 In this case, XRD
measurements were employed to confirm the Si content using
the measured lattice constant of the epilayer in conjunction
with Vegard’s Law. The 2θ/ω plots and (224) reciprocal space
maps typically contain a single Bragg reflection in each case
corresponding to overlapping contributions from the intrinsic
and n-type overlayers. However, we find that even a minor
deviation in Si between the two layers of the device (not
detectable by RBS) is manifested by a slight separation of their
XRD peak maxima, as expected due to the highly resolved
spectral features, as shown in Figure 4. In this sample, the (004)
B. Reactivity Studies of the Ge4H10 and Si4H10
Precursors. The SiGeSn films with 1−5% Sn compositions
described above are specifically targeted to lattice match Ge by
maintaining a ratio of Si/Sn ∼ 3.7, while the corresponding Si/
Ge ratios vary systematically from about 4−27%. In these
depositions, the latter ratio is controlled by adjusting the partial
pressures of the Si and Ge molecular precursor sources. We
find that a considerable excess of Si4H10 as compared to Ge4H10
is needed to incorporate target levels of silicon into the growing
films in the range of 4−20%. For instance, maintaining a partial
pressure ratio of 2:1 between the Si4H10 and Ge4H10 at 300 °C
in the gas-phase reactant mixture yields a sample containing
12% Si and 85% Ge. This suggests that the Si4H10 and Ge4H10
possess dramatically different reactivities. In the case of Ge4H10,
some insight into its reactive behavior was provided by our
prior work in which this compound was used as the source in
20760
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
pentasilane, Si(SiH3)4, on silicon to form −Si(100)−Si−
(SiH3)3 via elimination of SiH4. In view of this analogy, and
even conceding the existence of a small reactivity difference of
iso-Si4H10 and iso-Ge4H10 on a Ge(100) surface (due to the
slightly stronger Si−H bonds), the requirement of a large
excess of the Si4H10 source in our film growth is difficult to
explain. At the low deposition pressures mentioned above (10−4
Torr), gas-phase reactions between the Si4H10 and Ge4H10
sources are likely to be negligible. Thus, the required Si4H10
excess must be explained by the equilibrium thermodynamics of
the reactant isomers present at the growth front. For example, a
large concentration of the n-Si4H10 and g-Si4H10 isomers with
reduced reactivities over the i-Si4H10 might explain the need for
large excess of Si4H10 to produce the desired alloy
compositions.
To examine our conjecture on a more robust footing, we
pursed a targeted, and comparative, first principles thermo-
chemistry study of Si4H10 and its isomers to determine their
proportions at growth conditions. For consistency, and to
facilitate straightforward comparisons with our prior work on
the related Ge4H10 molecules,16 we here adopt the same
quantum simulation methods and theoretical thermochemistry
procedures. Specifically, all of our calculations are based on
density functional theory (DFT) as implemented in the
Gaussian 03 code,18 using the B3LYP hybrid exchange-
correlation functional in conjunction with a standard 6-311G
++(3df,3pd) basis set. All static molecular ground-state
structures (see Table 2) were converged using an RMS force
criterion of 10−6 (designated using the “VeryTight” keyword in
Gaussian 03), “ultrafine” integration grids (75 radial shells and
203 angular points per atom), and no symmetry constraints
imposed during optimization. In all cases, the harmonic
normal-mode frequencies calculated for the ground state
molecular structures are found to be positive definite, indicating
that the molecules are dynamically stable. A symmetry analysis
of our final converged structures yielded C3v, C2h, and C2 point
groups for the i-Si4H10, n-Si4H10, and g-Si4H10 molecules,
respectively, which are identical to those obtained for the
corresponding Ge4H10 species.16
As in the case of the Ge4H10, the Si4H10 bond lengths
calculated from our simulations are in general excellent
agreement with those obtained by other authors, and in our
prior work and follow expected trends.19,20 The shortest Si−Si
bond lengths (2.351 Å) typically occur between −SiH3 and
−SiH2− moieties, while slightly longer values (2.355 Å) are
found between −SiH3 and −SiH−, or adjoining −SiH2−
groups. In analogy with the Ge analogue, the Si−H bond
lengths are also found to follow the expected increasing trend
1.483, 1.486, and 1.490 Å for the −SiH3, −SiH2−, and −SiH−
moieties, respectively. The bond angle trends in these
molecules also exhibit systematic patterns with typical ∠Si−
Si−Si angles of ∼113° among the isomers having a branched
character, but a more tetrahedral value of 110.9° in the case of
the i-Si4H10 species possessing C3v symmetry. The torsion angle
among the Si atoms adopts trivial values for the high symmetry
species and a value of 65° for the g-Si4H10 isomer, close to the
typical 63° value found in the gauche isomer of butane. We
note here that a smaller value of 58° was found in our
simulation of the corresponding g-Ge4H10 isomer.16 The dipole
moments of the isomers (listed in Table 2) indicate a maximum
value of 0.105 D for the i-Si4H10 molecule, 0.06 D for g-Si4H10,
and a vanishing dipole moment for the n-Si4H10, as expected.
Figure 5. XTEM micrographs of p−i−n device structures with
nominal compositions of Ge0.84Si0.13Sn0.03/Ge (top panel) and
Ge0.75Si0.20Sn0.05/Ge (bottom panel) as grown at 300 and 290 °C,
respectively. The layers in both cases possess defect-free micro-
structures, flat surfaces, and viable thicknesses up to 1 μm as needed
for applications in multilayer devices. High-resolution image (middle
panel) of the sloped heterointerface marked by arrows indicating
highly conformal and fully commensurate growth of the
Ge0.75Si0.20Sn0.05 device on the off-cut Ge wafer.
ultralow temperature growth of Ge films.16 There we showed,
using state-of-the-art first principles thermochemistry simu-
lations, that iso-Ge4H10 is the dominant isomer, as compared to
the “normal” (n) and “gauche” (g) analogues, at typical
synthesis conditions similar to those used in the present work.
The isomeric speciation predicted theoretically was corrobo-
rated by measured gas-phase FTIR spectra of the compound,
whose main features are very well accounted for using the
weighted sum of the theoretically calculated isomeric spectra.
One of the outcomes of equilibrium thermodynamic and
reactivity studies of the Ge4H10 isomers is that the iso-Ge4H10
compound plays a key role in the film deposition. This notion
is supported by the isotropic character of its molecular structure
and the presence of the highly reactive −Ge−H moiety
(tertiary site), which represents a facile mechanism for binding
the molecule onto the Ge (or Si) substrate surface. The
resulting −Ge(100)−Ge−(GeH3)3 intermediate is then com-
pletely analogous to that proposed17 in the reaction of neo-
20761
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
a
Table 2. Summary of Structural and Energetic Results for the Three Si4H10 Isomers
iso-tetrasilane (i-Si4H10)
n-tetrasilane (n-Si4H10)
g-tetrasilane (g-Si4H10)
point group
C3v
C2h
C2
bond lengths (Å)
Si−Si
(Si−H)T
(Si−H)S
2.353
2.351(×2), 2.354
1.483(×3)
1.486
2.351(×2), 2.355
1.483(×3)
1.486
1.483(×9)
1.490*
bond angles (deg)
∠Si−Si−Si
(H−Si−H)T
(H−Si−H)S
110.9
112.6
112.9
108.5(×2), 108.6
108.5, 108.6(×2)
107.5
108.5, 108.6(×2)
107.6
dipole moment (D)
thermochemistry
0.1046
0.0000
0.0591
E0 (au)
−1164.10846
−1164.01457
−1164.01363
−1164.05994
97.5
−1164.10750
−1164.01383
−1164.01288
−1164.05935
97.8
−1164.10690
−1164.01329
−1164.01235
−1164.05907
98.3
(T = 298 K, P = 1 atm)
E0 + Eth (au)
E0 + Hth (au)
E0 + Gth (au)
S (cal/(mol K))
a
The subscripts “T” and “S” refer to terminal and sagittal H−Si−H bond angle species. The asterisk on the (Si−H)S bond length in i-Si4H10 indicates
that the −Si−H moiety is distinct from the other sagittal −SiH2− moieties. Note: Thermochemistry functions in this table are not corrected for
internal rotor contributions.
The standard thermochemistry output from Gaussian 03 at T
= 298 K and P = 1 atm is also summarized in Table 2 and
includes the static electronic energy of the Si4H10 isomers (E0),
as well as its thermally corrected counterparts for internal
energy (Eth), enthalpy (Hth), and free energy (Gth). In analogy
with the results for the Ge4H10 isomers16 the i-Si4H10 species
possessing C3v symmetry is found to have the lowest energies
(E0, E0 + Eth, E0 + Hth, and E0 + Gth), followed by the
corresponding higher energies of n-Si4H10 and g-Si4H10. The
distinctions and similarities in the free-energy trends of the
Ge4H10 and Si4H10 isomers are summarized in Figure 6, where
species, in contrast to the Ge4H10 case, in which the largest
entropy was obtained for the n-Ge4H10 isomer, due to its larger
moment of inertia. An analysis of the rotational and vibrational
contributions indicates that the total entropies are controlled by
rotations in Si4H10 and vibrations in Ge4H10. Ultimately, as
discussed below, the playoff in the temperature dependence of
the vibrational and rotational contributions to the free energy
leads to manifestly different thermodynamic behavior in Si4H10
and Ge4H10 isomers.
The isomeric mixture in Si4H10 is comprised of linear and
gauche conformational isomers (n-Si4H10 and g-Si4H10) and the
positional isomer i-Si4H10, which is predicted to possess the
smallest free energy. In our earlier studies of butane-like Si−Ge
hydride isomeric mixtures,21 we introduced the concept of
fitting a linear combination of calculated isomeric vibrational
spectra to the experimental spectrum of a mixture, to ascertain
the relative proportions in the experimental gas-phase mixture.
Here, instead we apply an ab initio thermodynamic approach
based on free-energy minimization for the three isomerization
reactions:
i‐Si4H10 ⇒ n‐Si4H10 ΔG0R,1(298 K) = +1.56 kJ/mol
(3)
n‐Si4H10 ⇒ g‐Si4H10 ΔG0R,2(298 K) = +0.74 kJ/mol
(4)
g‐Si4H10 ⇒ i‐Si4H10 ΔG0R,3(298 K) = −2.30 kJ/mol
Figure 6. (Left panel) Relative Gibbs free energies of Ge4H10 and
Si4H10 isomers, indicating a common ordering on the energy scale but
much smaller energy differences among the Si-based isomers. (Right
panel) Molecular structures of the Si4H10 isomers. The calculated Si−
H bond lengths shown indicate a systematic dilation from the −SiH3
to −SiH2, to −SiH, implying that the latter is likely the most thermally
labile.
(5)
where the reaction free energies are obtained directly from the
E0 + Gth data listed in Table 2. All thermodynamic functions
were obtained assuming ideal gas behavior and mixing, and the
neglect of possible internal rotor contributions. The automati-
cally generated analysis of the latter corrections provided by
Gaussian 03 for these molecules (and Ge4H10 analogues)
identified the related normal modes as involving torsions in the
molecular backbone, and rotations of the terminal −SiH3
groups. Typical corrections to the thermal energy are estimated
to be on the order of ∼0.1−0.2 kJ/mol at ∼300 K, and we note
that in the present system similar rotor contributions appear on
both sides of the reactions listed above (eqs 3−5), suggesting a
significant cancellation of error. Accordingly, our analysis below
neglects these rotational corrections.
the i-Si4H10 values are used as a reference (Ge4H10 values taken
from ref 16). The most notable distinction between the relative
free energies of the Ge4H10 isomers, and their Si counterparts,
is the dramatically reduced energy separation in the latter,
which is comparable to kBT at 300 K (∼2.5 kJ/mol). This
implies that the iso-, n-, and g-Si4H10 mixture should be roughly
equimolar at room temperature. Finally, Table 2 lists the
molecular entropies of the Si4H10 isomers obtained at T = 298
K, which are found to be virtually independent of the isomer
20762
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
Figure 7. (a) Calculated temperature dependence of the Si4H10 (dashed lines) and Ge4H10 (solid lines) isomer reaction free energies up to 600 K,
and (b) that of the resulting equilibrium mole fractions predicted for both Si4H10 and Ge4H10 isomers. Note the predominance of iso Ge4H10 at 300
°C as compared to a nearly equimolar distribution in Si4H10. (c) Composite room temperature IR spectrum of Si4H10 obtained by linearly combining
the calculated IR spectra of its isomers according to the predicted equilibrium mixture. Scale factors of 0.995 and 0.975 were applied to the calculated
spectra in the low- and high-frequency regimes, respectively. Theoretical and experimental spectra are oriented upward and downward, respectively,
as indicated within the plots in part (c).
The temperature-dependent equilibrium mixture is described
by a simple system of (cyclic) coupled equations, described in
detail in ref 16. At 298 K, the equilibrium constants obtained
shown in Figure 7c, where the theoretical and experimental
spectra are oriented upward and downward, respectively, the
plots indicate that the equilibrium molar fractions account well
for the principal character (frequencies and intensities) of the
corresponding experimental spectrum. Most importantly, and
in contrast to Ge4H10,16 we find that the observed spectrum
cannot be accounted for theoretically using any individual
isomer spectrum alone. Changes in the proportions of as little
as ∼5% among the isomeric contributions produces large
discrepancies in the intensity ratios between the calculated and
experimental spectra. In the case of Si4H10, several of the low-
frequency features (<800 cm−1) in the IR spectrum can be
clearly assigned to a specific isomer, as indicated in the plots
using the designations “i”, “n”, and “g”. On the other hand, the
most intense vibrations (∼870 cm−1) exhibit almost the same
(common) frequency in all three isomer gas-phase species.
C. Device Fabrication and I−V Characteristics. Photo-
diodes were fabricated with compositions listed in Table 1
using previously developed procedures for related GeSn IR
devices.22,23 The above were processed as-grown with no
further thermal treatments as typically employed by previous
workers to improve defect microstructure in related materials.22
Devices are produced in a circular mesa geometry by etching
the surrounding material down to the p-type Ge wafer (∼3 ×
1017/cm3). The top and bottom metal contacts consist of
annular Cr/Au layers with thickness of 20/200 nm. A 100 nm-
thick SiO2 coating is first deposited. The mesas are then
patterned via photolithography and etched by reactive ion
plasma using BCl3. They are subsequently passivated with a 100
nm-thick SiO2 antireflective coating deposited by plasma-
enhanced CVD (PECVD). Metal contact pads are then defined
using the AZ5214 resist, and the samples are baked at 115 °C
for 90 s to facilitate metal lift-off. The SiO2 layer in the contact
areas is removed by a buffered oxide etchant, and the metals are
deposited via e-beam evaporation. The formation of electrical
pads is finalized by lift-off in acetone, and the resultant devices
are finally cleaned in an oxygen plasma.
from the free energies listed in Table 2 are K0P,1 = 0.5329, KP0,2
=
0.7425, and K0P,3 = 2.5274 yielding isomer mole fractions ni =
52%, nn = 28%, and ng = 20%. This roughly equimolar
distribution of the Si4H10 isomers is dramatically distinct from
16
the results obtained for Ge4H10 in which the same
computational methods and procedures yielded ni = 85%, nn
= 9%, and ng = 6%. This dichotomy can be traced to the small
calculated free-energy differences among the Si4H10 isomers, in
comparison to their Ge4H10 counterparts. Additional insight is
afforded by examining the temperature dependence of the
reaction free energies, which is obtained here by re-evaluating
the thermochemistry for reactions 3−5 over the appropriate
range of 200−600 K, because this includes our deposition
conditions of 285−320 °C. The resulting thermodynamic
behavior is summarized Figure 7, which shows plots of the
isomerization reaction free energies for Si4H10 as well the
corresponding results for Ge4H10 (taken from ref 16). We note
in Figure 7b that while ordering in the reaction free energies for
both Si4H10 and Ge4H10 is qualitatively the same (ΔGr[i→n] >
ΔGr[n→g] > ΔGr[n→g]) over the indicated temperature range,
their rates of change with temperature are typically opposite in
sign (where the Ge species increase the Si decreases and vice
versa). In particular, the model predicts that at the growth
conditions, an equimolar mixture of Si4H10 and Ge4H10
contains twice as much of the reactive i-Ge4H10 isomer in
proportion to its i-Si4H10 counterpart. This reduction in the
availability of the most reactive Si source may partially account
for the observed excess in Si4H10 needed to achieve the desired
GeSiSn film compositions.
To validate the simple thermodynamic model above, we next
attempt to reproduce the observed room temperature IR
spectrum of Si4H10 by combining the calculated spectra of i-
Si4H10, n-Si4H10, and g-Si4H10 in the 52%, 28%, and 20%
proportions predicted by the equilibrium calculation above. As
20763
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
Current−voltage (I−V) measurements were performed on
mesas with diameters 100, 200, 300, 500, and 1000 μm, and
their respective dark currents were determined at a nominal
bias of −1 V. Figure 8 (top) shows representative I−V plots for
Figure 9, which shows that for the above bias range these
energies span from 0.578 to 0.478 eV, respectively. The low-
Figure 9. Arrhenius plots of the dark current densities at selected
reverse bias values for the Ge0.925Si0.063Sn0.012/Ge diode. The activation
energies are obtained from the slopes of the lines.
voltage value is significantly higher than Eg/2, where Eg is the
fundamental band gap for this alloy, estimated to be close to
∼0.70 eV following the procedure described in ref 7. The
corresponding activation energies at −0.5 V for
Ge0.913Si0.072Sn0.015, Ge0.886Si0.094Sn0.02, Ge0.84Si0.13Sn0.030
,
Ge0.78Si0.18Sn0.04, and Ge0.75Si0.20Sn0.05 are found to be 0.524,
0.557, 0.524, 0.493, and 0.486 eV, respectively, which are also
higher that the expected values of Eg/2 for these. These results
suggest that the dark current is not completely dominated by
Schottky−Read−Hall recombination but has a significant
diffusion component as well, underscoring the high crystal
quality of the ternary alloy layers in all samples studied to date,
regardless of composition and intrinsic layer thickness.
D. Optical Response and Direct Gap Determination.
The spectral response of the fabricated devices was measured
using monochromatized light originating from a tungsten
halogen lamp. The light was modulated with a mechanical
chopper and then transmitted onto the SiO2 diode window
using an optical fiber equipped with a focusing lens whose spot
size (30 μm) is significantly smaller than all measured devices.
The resultant photocurrent induces a voltage on a 100 Ω
resistor that was measured by a lock-in amplifier. The voltage
and resistance were then used to calculate the photocurrent,
which was divided by the power of the incident light to obtain
the responsivity R.
Figure 10 shows the spectral dependence of the external
quantum efficiency (EQE) of a Ge0.78Si0.18Sn0.04 sample
measured from a device 300 μm in diameter at 0.0 V bias
[EQE = 1240 nm × R/λ (nm)]. The two absorption edges at
1590 nm (0.78 eV) and 1230 nm (1.02 eV) nm are assigned to
direct-gap absorption in the Ge substrate and alloy epilayer,
respectively (the direct gap E0 in bulk, intrinsic Ge is at 0.805
eV at room temperature; the small downshift in our sample is
due to band gap renormalization in the highly doped Ge wafer).
The significant photoresponse observed in this case is
attributed to the improved crystallinity of the material grown
directly on bulk Ge under pseudo homoepitaxy conditions. The
signal of the plot is significantly reduced beyond 1600 nm
because only indirect absorption in Ge can contribute to the
photocurrent. The EQE becomes essentially zero at 1800 nm,
near the fundamental Ge band gap.
Figure 8. Current density curves for GeSiSn photodiodes. Top panel
shows the plots of a Ge0.925Si0.063Sn0.012, devices with diameters ranging
from 100 to 1000 μm. The bottom panel compares the curves
obtained from selected samples measured from 300 μm mesas,
indicating lower dark currents relative to Ge on Si devices with similar
geometries.22
a Ge0.925Si0.063Sn0.012 sample indicating remarkably low dark
current densities of ∼1 × 10−3 J(A/cm2), which are essentially
independent of the device size. Figure 8 (bottom) compares a
representative set of current densities as a function of
concentration. The magnitudes of the dark currents in this
case very slightly increase from 1 × 10−3 to 2 × 10−3 J(A/cm2)
with Si/Sn content. Alloy devices with 1−2% Sn content
exhibit lower ideality factors at n = 1.1−1.2 than 3−5%
analogues, which show values in the n = 1.3−1.4 range.
We note that the above device dark currents are much lower
than those of Ge and GeSn on Si(100) counterparts by at least
an order of magnitude.22 They are also reduced from those
observed in our Sn-doped Ge materials and diluted Ge1−ySny
alloys [5 × 10−2 J(A/cm2)].23 The lower dark current in
GeSiSn can be attributed to its higher fundamental band gap,
the structural perfection of the film/substrate interface, and the
superior crystal quality. In the case of Ge or GeSn grown on
Si(100), the lattice mismatch produces misfit dislocations of
which some cores propagate into the bulk layer as deleterious
threading defects. This effect seems to be removed in the
ternary system, in which lattice matching between the GeSiSn
and the underlying Ge can be readily achieved for Si/Sn ratios
in the vicinity of 3.7.
Arrhenius plots of the dark current densities as a function of
temperature were used to estimate their activation energies for
a series of voltage measurements conducted between 0.5 and
2.0 V. The data for a Ge0.925Si0.063Sn0.012 device are illustrated in
Figure 11 shows EQE versus wavelength plots comparing the
optical response of photodiodes containing ∼1−4% Sn. The
plots were derived from photocurrent measurements made on
20764
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
gap, as in the case for the Ge0.78Si0.18Sn0.04 and Ge0.83Si0.13Sn0.04
sample pair. In this regard, the latter effect is also observed for
the Ge0.900Si0.08Sn0.02 and Ge0.886Si0.094Sn0.02 samples (left panel)
in which a slight increase in Si (from 8% to ∼10%) while
keeping Sn constant (2%) produces the expected trend of
decreasing wavelength. These findings indicate that small
changes in the Sn content have a profound and systematic
effect on the optoelectronic properties of the material and
suggest that the tunability of the responsivity is highly precise,
representing a powerful new design tool in the development of
devices with targeted optoelectronic properties.
E. Direct Gap Dependence on Composition. In this
work, we determined the direct band gap E0 of the
Ge1−x−ySixSny alloy using a simple procedure, which allows
for a straightforward determination of the direct absorption
threshold. We compute numerically the energy derivative of the
EQE, and we fit the resulting line shape with a Gaussian. An
example of the derivative spectrum is shown as a blue trace in
Figure 10. We find that this method gives values of E0 in
excellent agreement with earlier ellipsometric determinations of
the band gap.24 The energies extracted from our derivative fits
are corrected for minor residual strain effects using the
measured strain from our X-ray data and deformation potentials
from ref 6. The corrected values are shown (as ●) in Figure 12
as a function of the non-Ge fraction X = x + y. The strain shifts
are very small, ranging from 2 meV to a maximum of 10 meV.
In applying this correction, it is assumed that the absorption is
dominated by transitions from the heavy-hole band. The
systematic error associated with this assumption is estimated to
be on the order of 1 meV. The error bars on the data in Figure
12 are mostly smaller than the size of the circles in the figure.
In earlier work, we wrote the compositional dependence of
the E0 gap in Ge1−x−ySixSny alloys as6
Figure 10. External quantum efficiency (EQE) and its derivative (red
and blue curves, respectively) as a function of energy for a
heterostructure diode with composition Ge0.78Si0.18Sn0.04 measured at
zero bias. The data show the absorption edges for the Ge substrate at
1580 nm and the epilayer at 1230 nm. The vertical lines indicate the
position of the direct gap, determined as discussed in the text.
devices with mesa size 300 μm in diameter at 0.0 V bias. The
peak maxima were normalized to facilitate comparisons of the
responsivity among devices with varying active layer thickness
from 0.5 to 1 μm depending on composition. The Ge wafer
photoresponse is not shown here to emphasize minor shifts of
the absorption peaks as a function of alloy composition. In
general, the plots show that the simultaneous increase of Sn
(1−4%) and Si (6−18%) while keeping the Si/Sn ratio at 4
1
produces a systematic and significant shift of the absorption
wavelength by ∼200 nm within this range. For example, in the
case of the Ge0.925Si0.063Sn0.012 and Ge0.913Si0.072Sn0.015 samples,
their absorption edges (at left panel) are clearly resolved given
the very slight difference in their respective stoichiometries,
providing unequivocal evidence that the simultaneous incor-
poration of Si and Sn produces a measurable opening of the
band gap. In contrast, the plots for samples Ge0.84Si0.13Sn0.03
and Ge0.83Si0.13Sn0.04 (at right panel) show that increasing the
Sn at fixed Si content reduces the gap while keeping Sn
constant and increasing Si leads to significant opening of the
E0(x, y) = E0Gez + E0Six + E0Sny − bGeSixz − bGeSnyz
− bSiSnxy
(6)
where z = 1 − x − y, E0Ge, E0Si, and E0Sn are the E0 gaps in Ge, Si,
and α-Sn, respectively, and the b’s are the bowing parameters in
the corresponding binary alloy systems. This expression
Figure 11. (left) Normalized responsivity versus wavelength plots of heterostructure diodes with Sn contents of 1−4% measured at zero bias. The
absorption edge is found to shift to lower IR wavelengths by simultaneously increasing Sn and Si incorporation over a very narrow composition
range. The % values on each curve refer to the Si and Sn contents, respectively. (right) EQE plots comparing the absorption edges of several selected
GeSiSn samples in which increasing the Sn at fixed Si content reduces the gap while keeping Sn constant and increasing Si results in a significant
opening of the gap.
20765
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
Because of the practical importance of GeSiSn alloys with the
range of compositions used in this Article, in ref 14 we have
also fit the measured E0 values with a purely empirical
expression that is bilinear in the Si and Sn compositions.
Following the same procedure with our extended set of
samples, we now obtained (in eV):
E0(x, y) = 0.803 + (1.86 0.34)x − (2.40 1.4)y
(7)
We notice that the linear terms in eq 6 are ES0i − E0Ge − bGeSi
=
3.08 eV and E0Sn − EG0 e − bGeSn = −3.42 eV. The much lower
experimental x-coefficient (1.86 eV vs 3.08 eV) is thus the main
reason why eq 6 systematically overestimates the E0 energies.
The predictions from eq 7 for our samples are also plotted as
empty circles (red ○) in Figure 12, and we see that their
agreement with experiment is excellent. However, given the
entirely empirical nature of the bilinear expression (eq 7), it is
not possible to assess its validity outside the Sn and Si
compositional ranges explored in this work. For completeness,
we also show in Figure 12, as a solid line, the prediction from
eq 7 for GeSiSn alloys exactly lattice-matched to Ge.
Figure 12. Direct gap E0 energies in Ge1−x−ySixSny alloys as a function
of the non-Ge fraction X = x + y. Solid black circles correspond to
experimental data. The blue squares represent the prediction from eq
6. The red circles are the results from a bilinear (in x and y) fit of the
experimental energies shown in eq 7. The red line is the prediction
from eq 7 for the case of exact lattice match between GeSiSn and Ge.
III. CONCLUSION
In summary, a series of Ge1−x−ySixSny photodiodes was grown
on conventional Ge wafers using a specially developed CVD
approach based on ultralow temperature depositions of highly
reactive Ge4H10, Si4H10, and SnD4 hydrides. Ab initio
thermochemistry calculations using Gaussian 03 indicate that
an equimolar mixture of Si4H10 and Ge4H10 contains about
twice as much of the reactive i-Ge4H10 isomer in proportion to
its i-Si4H10 counterpart, in qualitative agreement with the
experimental observation that an excess in Si4H10 is needed to
incorporate target amounts of silicon. Devices with composi-
tions ranging from 5% to 20% Si and 1−5% Sn were fabricated
and found to exhibit a clear and systematic shift in responsivity
to higher IR energies relative to Ge, from 0.88 to 1.03 eV, while
preserving the benefit of nearly strain-free lattice matching to
bulk Ge. The latter has afforded superior microstructure in the
as-grown devices, leading to the generation of low dark currents
with a significant diffusion component, and enhanced
responsivities relative to previous Sn-based group IV devices
grown on Si. The results are particularly encouraging for
photovoltaic applications of the ternary Ge1−x−ySixSny alloy,
especially in the case of high efficiency multijunction designs
requiring ∼1 eV gap materials lattice-matched to the Ge
platform. Finally, the development of analogous Ge1−x−ySixSny
devices integrated on low cost Si wafers is envisioned by using
intermediate Ge buffers to mitigate the impact of mismatch
defects on diode performance.
neglects the contribution of ternary-specific terms in the
polynomial expansion of the band gap energy, such as those
proportional to xyz. Its validity can be tested experimentally by
performing measurements in the three binary alloy systems. In
our case, however, the Si1−ySny alloy is very poorly known, and
the compositional dependence of its E0 gap has not been
measured. Thus, eq 6 was fit to Ge1−x−ySixSny data in ref 24
using bSiSn as an adjustable parameter, and it was shown that a
reasonable good fit of the data was obtained using bSiSn = 13.2
eV. The remaining parameters in eq 6 are known from work on
binary alloys and elemental semiconductors,25 and in Figure 12
we show as blue □ the resulting prediction using the above
value for bSiSn and E0Ge = 0.803 eV, E0Si = 4.093 eV, E0Sn = −0.42
eV, bGeSn = 2.2 eV, and bGeSi = 0.21 eV. We notice that the
predicted energies follow the experimental trend quite well, but
there is a systematic upward shift of about 75 meV with respect
to the observed values (solid symbols). The discrepancy is due
to the fact that in ref 24, eq 6 was fit to the data over a very
broad range of compositions with X < 0.6, whereas in the
present work the data points cover a much narrower range 0.06
< X < 0.2. As indicated above, our new data (solid symbols in
Figure 12) overlap very well with the ellipsometric data in ref
24 (not shown here) over the same compositional range,
leading to the conclusion that the observed ∼75 meV deviation
is not a systematic error that can be attributed to the different
experimental techniques used to measure E0, but is rather a
limitation of eq 6 and/or the parameters used for that
expression in Figure 12. We already noted in ref 24 that very
large bowing parameters, such as bSiSn, tend to be compositional
dependent, and this could be the reason for the discrepancy.
Concurrently, the neglect of third- or fourth-order terms in the
polynomial expansion may not be entirely justified. The
elucidation of this problem will require additional measure-
ments over a broad compositional range as well as a detailed
study of the binary alloy Si1−ySny. Very thin (<30 nm) films of
Si1−ySny have been demonstrated on GeSn-buffered Si using
Si3H8 and SnD4 precursors.26 The introduction of higher-order
silanes may represent an opportunity to grow thicker films
suitable for detailed studies of their optical properties.
AUTHOR INFORMATION
Corresponding Author
jkouvetakis@asu.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was supported by the U.S. Air Force under contract
AFOSR FA9550-12-1-0208 (Dr. Gernot Pomrenke), and by
the U.S. Department of Energy under contract DE-FG36-
08GO18003. We thank Emcore Corp. for providing the Ge
wafers.
20766
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767

Journal of the American Chemical Society
Article
(25) D’Costa, V. R.; Cook, C. S.; Birdwell, A. G.; Littler, C. L.;
Canonico, M.; Zollner, S.; Kouvetakis, J.; Menendez, J. Phys. Rev. B
2006, 73, 125207.
REFERENCES
■
(1) Bauer, M.; Taraci, J.; Tolle, J.; Chizmeshya, A. V. G.; Zollner, S.;
Smith, D. J.; Menendez, J.; Hu, C.; Kouvetakis, J. Appl. Phys. Lett.
2002, 81, 2992.
(2) Vincent, B.; Gencarelli, G.; Bender, H.; Merckling, C.; Douhard,
B.; Petersen, D. H.; Hansen, O.; Henrichsen, H. H.; Meersschaut, J.;
Vandervorst, W.; Heyns, M.; Loo, R.; Caymax, M. Appl. Phys. Lett.
2011, 99, 152103.
(26) Tolle, J.; Chizmeshya, A. V. G.; Fang, Y. Y.; Kouvetakis, J.;
D’Costa, V. R.; Hu, C. W.; Menen
Lett. 2006, 89, 231924.
́
dez, J.; Tsong, I. S. T. Appl. Phys.
(3) Gencarelli, F.; Vincent, B.; Souriau, L.; Richard, O.; Vandervorst,
W.; Loo, R.; Caymax, M.; Heyns, M. Thin Solid Films 2012, 520, 3211.
(4) Sun, G.; Cheng, H. H.; Menendez, J.; Khurgin, J. B.; Soref, R. A.
Appl. Phys. Lett. 2007, 90, 251105.
(5) Menendez, J.; Kouvetakis, J. Appl. Phys. Lett. 2004, 85, 1175.
(6) D’Costa, V. R.; Fang, Y. Y.; Tolle, J.; Kouvetakis, J.; Menen
́
dez, J.
Thin Solid Films 2010, 518, 2531.
(7) Beeler, R. T.; Smith, D. J.; Kouvetakis, J.; Menendez, J. IEEE J.
Photovolt. 2012, 3, 434−440.
(8) Chroneos, A.; Jiang, C.; Grimes, R. W.; Schwingenschlogl, U.;
̈
Bracht, H. Appl. Phys. Lett. 2009, 95, 112101.
(9) Fang, Y.-Y.; Xie, J.; Tolle, J.; Roucka, R.; D’Costa, V. R.;
Chizmeshya, A. V. G.; Menendez, J.; Kouvetakis, J. J. Am. Chem. Soc.
2008, 130, 16095.
(10) Bauer, M.; Ritter, C.; Crozier, P. A.; Ren, J.; Menen
G.; Kouvetakis, J. Appl. Phys. Lett. 2003, 83, 2163.
(11) Xie, J.; Tolle, J.; D’Costa, V. R.; Chizmeshya, A. V. G.;
Menendez, J.; Kouvetakis, J. Appl. Phys. Lett. 2009, 95, 181909.
́
dez, J.; Wolf,
́
(12) Xie, J.; Chizmeshya, A. V. G.; Tolle, J.; D’Costa, V. R.;
Menendez, J.; Kouvetakis, J. Chem. Mater. 2010, 22, 3779.
(13) D’Costa, V. R.; Fang, Y.; Mathews, J.; Roucka, R.; Tolle, J.;
Menendez, J.; Kouvetakis, J. Semicond. Sci. Technol. 2009, 24, 115006.
(14) Beeler, R. T.; Xu, C.; Smith, D. J.; Grzybowski, G.; Menen
J.; Kouvetakis, J. Appl. Phys. Lett. 2012, 101, 221111.
(15) Doolittle, L. R. Nucl. Instrum. Methods Phys. Res., Sect. B 1985, 9,
344.
́
dez,
(16) Grzybowski, G.; Beeler, R.; Chizmeshya, A. V. G.; Watkins, T.;
Jiang, L.; Kouvetakis, J.; Menendez, J. Chem. Mater. 2012, 24, 1619−
1628.
(17) Chung, K. H.; Yao, N.; Benziger, J.; Sturm, J. C.; Singh, K. K.;
Carlson, D.; Kuppurao, S. Appl. Phys. Lett. 2008, 92, 113506.
(18) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;
Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.;
Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.;
Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.;
Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.;
Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao,
O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J.
B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R.
E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.;
Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J.
J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.;
Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman,
J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.;
Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.;
Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.;
Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen,
W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03, revision
C.02; Gaussian, Inc.: Wallingford, CT, 2004.
(19) Weng, C.; Kouvetakis, J.; Chizmeshya, A. V. G. J. Comput. Chem.
2011, 32, 835−853.
(20) Ortiz, J. V.; Mintmire, J. W. J. Am. Chem. Soc. 1988, 110, 4522.
(21) Ritter, C. J.; Hu, C.-W.; Chizmeshya, A. V. G.; Tolle, J.; Nieman,
R.; Tsong, I. S. T.; Kouvetakis, J. J. Am. Chem. Soc. 2006, 128, 6919.
(22) Roucka, R.; Mathews, J.; Weng, C.; Beeler, R.; Tolle, J.;
Menendez, J.; Kouvetakis, J. IEEE J. Quantum Electron. 2011, 47, 213.
(23) Roucka, R.; Beeler, R.; Mathews, J.; Ryu, M.-Y.; Yeo, Y. K.;
́
Menendez, J.; Kouvetakis, J. J. Appl. Phys. 2011, 109, 103115.
(24) D’Costa, V. R.; Fang, Y. Y.; Tolle, J.; Kouvetakis, J.; Menendez,
J. Phys. Rev. Lett. 2009, 102, 107403.
20767
dx.doi.org/10.1021/ja309894c | J. Am. Chem. Soc. 2012, 134, 20756−20767