082108-3
Wistey et al.
Appl. Phys. Lett. 90, 082108 ͑2007͒
according to the reaction ͑GeH3͒2CH2͑gas͒→CH4͑gas͒
+2Ge͑solid͒+2H2͑gas͒. We note that the O content through
the film thickness was also at background levels. Minor C
and O contamination peaks, commonly observed in CVD
grown materials, can be also seen at the Ge/Si interface.
These C and O signatures vary from sample to sample indi-
cating that C ͑in particular͒ may not be related to any mea-
surable decomposition of ͑GeH3͒2CH2 in the vicinity of the
interface. However, some C detected at impurity levels could
be due to side reactions of the precursor with the reactive Si
surface. Nevertheless, the fact that oxygen levels track those
of carbon in all of our samples cannot be explained in terms
of the precursor decomposition.
FIG. 4. Low T, low P reaction of ͑GeH3͒2CH2 with H-covered Ge surface.
Sequence shows intact adsorption of GeH3 –CH2 –GeH3, surface decompo-
sition, and release of CH4. Bond enthalpies relative to gas phase
͑GeH3͒2CH2 ͑prior to reaction͒ are indicated in eV.
To confirm that ͑GeH3͒2CH2 does not deposit C we con-
ducted a control experiment in which an initiation layer of
ϳ250 nm was first grown on Si using the 1:15 reactant ratio.
The remainder of the film was then deposited in situ by
switching the gas source to pure Ge2H6 yielding a total layer
thickness of 800 nm. We note that the growth rate during the
latter stage using pure Ge2H6 is four to five times higher than
that using the 1:15 mixture. XTEM and AFM examinations
of the full sample thickness revealed a complete, continuous,
and monocrystalline layer with a flat surface ͑rms ϳ0.4 nm͒.
SIMS showed a constant Ge content throughout and the typi-
cal C and O impurity peaks located at the interface ͑Fig. 3͒.
In addition, small and distinct C and O signals can be seen
slightly above the noise at the switchover point, 250 nm
above the Si/Ge interface. These minor impurities likely
originate from exposure of the surface to the reactor ambient
during the ϳ1 h switchover. XTEM showed that the micro-
structure throughout the growth transition region is indistin-
guishable from the bulk material indicating that the layer-by-
layer growth is uninterrupted in the absence of ͑GeH3͒2CH2.
This demonstrates conclusively that the carbon from the pre-
cursor is continuously eliminated and does not accumulate
on the Ge surface. Furthermore this two step process yields
high growth rates, Ͼ20 nm/min, making the method viable
for large scale fabrication.
Recent growth studies of Ge on Si suggest that conven-
tional Sb or As surfactants alter the free energy of the surface
and promote layer-by-layer growth far beyond the critical
thickness.8 These surfactants remain on the surface where
they mediate chemisorption interactions reducing reactant
diffusion, and lower surface tension. In contrast, our two step
Ge growth described above demonstrates that any C origi-
nating from the precursor desorbs as CH4 and thus cannot
act as a conventional surfactant. Instead, the effective surfac-
tant mechanism likely involves the formation of surface
pseudosurfactant intermediate groups composed of
–GeH2–CH2–GeH2– ͑or –GeH2–CH3͒ submonolayers,
which are analogous in function to a conventional surfactant
͑Fig. 4͒. These groups are continuously replenished at the
growth front by the incident ͑GeH3͒2CH2 flux and serve as
site holders for the organized assembly of smooth films and
circumvent unintentional incorporation of residual As/Sb
surfactants at these low growth temperatures.
complexes was then calculated and the most plausible ex-
ample consistent with the absence of gas phase reactions at
low temperatures ͑350–400 °C͒ and pressures ͑10−5 Torr͒ is
shown in Fig. 4. Here the intact adsorption of ͑GeH3͒2CH2
produces a strained methylene-bridged –GeH2–CH2–GeH2–
intermediate complex localized at the growth front. The bond
strain due to the C/Ge size difference, then leads to either
immediate release of methane ͑net gain of −1.04 eV͒ or con-
version to the GeH2–CH3 complex shown in Fig. 4, with a
net gain of −0.52 eV. The latter intermediate then finally
desorbs methane with an energy gain of −0.52 eV. In either
case two –GeH3 groups are deposited as building blocks for
layer-by-layer growth. Thus, the net reaction enthalpy for the
deposition of Ge and release of CH4 is found to be
−1.13 eV which represents a strong driving force as de-
scribed by the equation ͑GeH3͒2CH2͑gas͒+2͓−H͔͑surface͒
→2͑−GeH3͒͑surface͒+CH4͑gas͒. We note that the presence
of excess metastable surface complexes is also likely respon-
sible for the “surface poisoning” as discussed earlier for
deposition of high molar ratios ͑1:2–1:5͒. Finally, the ten-
dency of the C atoms to form strong C–H terminated leaving
groups ultimately prevents the incorporation of C into the
Ge, particularly at our low T and P conditions.
In conclusion, we find that digermylmethane with
built-in pseudosurfactant capabilities enables growth of
atomically flat ͑rms 0.21 nm͒ and stress-free Ge films di-
rectly on Si. High growth rates are achieved at 350–420 °C
using Ge2H6 and produce perfectly pure films with threading
dislocation densities below 105 cm−2 ͓orders of magnitude
less than that attainable from the best competing process
available for Ge/Si͑100͒ structures͔.
This work was supported by the AFOSR MURI,
FA9550-06-01-0442 and NSF ͑DMR-0526604͒.
1J. Oh, J. C. Campbell, S. G. Thomas, S. Bharatan, R. Thoma, C. Jasper,
R. E. Jones, and T. E. Zirkle, IEEE J. Quantum Electron. 38, 1238 ͑2002͒.
2J. F. Liu, J. Michel, W. Giziewicz, D. Pan, K. Wada, D. D. Cannon, S.
Jongthammanurak, D. T. Danielson, L. C. Kimerling, J. Chen, F. O. Ilday,
F. X. Kartner, and J. Yasaitis, Appl. Phys. Lett. 87, 103501 ͑2005͒.
3Y. H. Kuo, Y. K. Lee, Y. S. Ge, S. Ren, J. E. Roth, T. I. Kamins, D. A. B.
Miller, and J. S. Harris, Nature ͑London͒ 437, 1334 ͑2005͒.
4H. Kroemer, T. Y. Liu, and P. M. Petroff, J. Cryst. Growth 95, 96 ͑1989͒.
5B. Cunningham, J. O. Chu, and S. Akbar, Appl. Phys. Lett. 59, 3574
͑1991͒.
To explain the propensity for ͑GeH3͒2CH2 to form inter-
mediate surface complexes, we calculated the relevant bond
enthalpies using density functional theory at the B3LYP/aug-
cc-pVQZ, yielding H–H ͑4.64 eV͒, C–H ͑4.56 eV͒, Ge–H
͑3.17 eV͒, C–C ͑3.60 eV͒, Ge–C ͑2.56 eV͒, and Ge–Ge
6M. T. Currie, S. B. Samavedam, T. A. Langdo, C. W. Leitz, and E. A.
Fitzgerald, Appl. Phys. Lett. 72, 1718 ͑1998͒.
7H. C. Luan, D. R. Lim, K. K. Lee, K. M. Chen, J. G. Sandland, K. Wada,
and L. C. Kimerling, Appl. Phys. Lett. 75, 2909 ͑1999͒.
8J. N. Stirman, P. A. Crozier, D. J. Smith, F. Phillip, G. Brill, and S.
͑1.74 eV͒. The reaction enthalpy of various possible surface
Sivananthan, Appl. Phys. Lett. 84, 2530 ͑2004͒.
131.230.73.202 On: Fri, 19 Dec 2014 01:33:37