Influence of HCl on the chemical vapor deposition and etching of Ge islands on Si(001)

Article abstract of DOI:10.1063/1.122307

When HCl is added during the growth of Ge islands on Si(001) by chemical vapor deposition, the reduced Ge surface diffusion impedes island development. There is a shift in the relative populations of different island types even when other conditions such as temperature, coverage, and growth rate, are unchanged. The effect of HCl on the net rate of deposition is proportional to the square of the HCl partial pressure, suggesting a surface reaction with the Ge. When larger islands are etched with HCl at high enough temperature, they revert to a shape characteristic of smaller islands, confirming the reversibility of transformations from one island type to another. It has not proved possible to use etching to produce smaller and more uniform islands.

Full text of DOI:10.1063/1.122307

Influence of HCl on the chemical vapor deposition and etching of Ge islands on Si(001)
T. I. Kamins, G. A. D. Briggs, and R. Stanley Williams
Citation: Applied Physics Letters 73, 1862 (1998); doi: 10.1063/1.122307
View online: http://dx.doi.org/10.1063/1.122307
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/73/13?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
ABAB organization of Ge ∕ Si ( 001 ) islands in multiplanes grown with low pressure chemical vapor deposition
J. Appl. Phys. 96, 6175 (2004); 10.1063/1.1806255
Influence of phosphine on Ge/Si(001) island growth by chemical vapor deposition
J. Appl. Phys. 94, 4215 (2003); 10.1063/1.1604957
Structural and optical properties of Ge islands grown in an industrial chemical vapor deposition reactor
J. Appl. Phys. 90, 2565 (2001); 10.1063/1.1389335
Effect of phosphorus on Ge/Si(001) island formation
Appl. Phys. Lett. 78, 2220 (2001); 10.1063/1.1361096
Influence of molecular hydrogen on Ge island nucleation on Si(001)
J. Appl. Phys. 88, 5113 (2000); 10.1063/1.1317245
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
129.22.67.107 On: Mon, 24 Nov 2014 11:42:25

APPLIED PHYSICS LETTERS
VOLUME 73, NUMBER 13
28 SEPTEMBER 1998
Influence of HCl on the chemical vapor deposition and etching
of Ge islands on Si„001…
T. I. Kamins,a) G. A. D. Briggs,b) and R. Stanley Williams
Hewlett-Packard Laboratories, 3500 Deer Creek Road, Palo Alto, California 94304-1392
͑Received 4 June 1998; accepted for publication 24 July 1998͒
When HCl is added during the growth of Ge islands on Si͑001͒ by chemical vapor deposition, the
reduced Ge surface diffusion impedes island development. There is a shift in the relative
populations of different island types even when other conditions such as temperature, coverage, and
growth rate, are unchanged. The effect of HCl on the net rate of deposition is proportional to the
square of the HCl partial pressure, suggesting a surface reaction with the Ge. When larger islands
are etched with HCl at high enough temperature, they revert to a shape characteristic of smaller
islands, confirming the reversibility of transformations from one island type to another. It has not
proved possible to use etching to produce smaller and more uniform islands. © 1998 American
Institute of Physics. ͓S0003-6951͑98͒03039-3͔
The primary factors affecting the morphology of germa-
early with GeH4 partial pressure GeH4→Geϩ2H2. The etch-
ing of Ge by HCl produces mainly gaseous GeCl2:
Geϩ2HCl→GeCl2ϩH2. If the rate-limiting step involves the
simultaneous reaction of Ge with two HCl molecules or two
Cl atoms ͑either in the gas phase or on the surface͒ to pro-
duce GeCl2, then the rate should be proportional to the
square of the HCl partial pressure. On the other hand, if Cl
atoms react with the Ge sequentially, or if two Cl atoms were
already paired on the surface, then the rate should depend
linearly on the HCl partial pressure. The net deposition rate
is
nium islands grown on Si͑001͒ substrates are the coverage
and the temperature.1,2 Stranski–Krastanov growth occurs,
with coherent islands forming after a wetting layer a few
monolayers thick has been deposited.3 If these islands could
be made small enough, and of sufficiently uniform size, then
they might be candidates for optoelectronic applications and
even future quantum electronic devices. Such devices might
use a combination of patterning and self-assembly.4 Chemi-
cal vapor deposition is attractive because it builds on the
extensive foundation of silicon technology. Since Ge quan-
tum dots are metastable with respect to Ge/Si alloy
formation,5 patterned self-assembly is likely to be achieved
by selective deposition on a partially processed silicon wafer
in order to minimize the time during which the islands are at
high temperature. The growth occurs on exposed silicon but
not on surrounding oxide. The problem with this is that the
consequent buildup of either gaseous or adsorbed Ge where
growth is inhibited leads to additional Ge deposition nearby,
so that the amount of deposited material depends on the sur-
roundings. The addition of HCl may inhibit this process for
Ge/Si quantum dots just as it does in the growth of Si layers,6
and so we need to understand how their growth is affected by
the presence of HCl. The scope for controlling the size dis-
tribution of Ge islands by varying the temperature is limited:
if the temperature is too low, then reaction rates are too slow;
and if the temperature is too high, then alloying with the
silicon substrate becomes significant. We need to know
whether HCl can provide an additional degree of control
over the island size and type distributions, either by growing
in the presence of HCl gas, or by etching immediately after
growth.
RϭA GeH ͒ϪB HCl͒n,
͑
͑
4
where nϭ1 or 2. By examining the dependence of the depo-
sition rate on the concentration of HCl, we can determine n
and gain some insight into the reaction mechanism.
In our experiments, Ge islands were formed by chemical
vapor deposition at 130 Pa in a lamp-heated reactor with the
wafer sitting on a support plate of moderate thermal mass.7
The amount of Ge deposited was measured by Rutherford
backscattering ͑RBS͒ and is expressed as ‘‘equivalent mono-
layers’’ (1 eq-MLϭ6.27ϫ1018 mϪ2), corresponding to the
equivalent coverage if the Ge in the islands were uniformly
distributed on the surface.
Germanium growth was carried out over a range of ger-
mane partial pressure ͑0.036–0.126 Pa͒, HCl partial pressure
͑0, 1.0, and 1.5 Pa͒ and substrate temperature ͑600, 625 and
650 °C) at a total pressure of 1.33 kPa. The deposition rate
without HCl increased only slowly with increasing tempera-
ture ͑and the incubation time between introducing the gas
and the start of growth decreased͒.7 The effect of HCl on the
net deposition rate decreased with increasing temperature,
and this may be because the increasing chemical reaction of
Cl with Ge is more than offset by a decrease in the amount of
Cl on the surface available for the reaction. At 600 °C add-
ing a partial pressure of 1.5 Pa HCl to the 0.07 Pa GeH4
partial pressure decreased the deposited thickness by a factor
of 2. At 650 °C the same addition of HCl decreased the
deposited thickness by only 10%. However the key point is
that the reduction in deposited thickness depended on the
square of the HCl flow. This is illustrated graphically in the
The deposition of Ge from a mixture of GeH4 and H2 in
the presence of HCl can be considered as simultaneous depo-
sition of Ge from the pyrolysis of GeH4 and etching of Ge by
HCl. If the surface adsorption sites are not saturated with
adsorbed GeH4 molecules, the deposition rate increases lin-
a͒
Electronic mail: kamins@hpl.hp.com
On sabbatical leave from the Department of Materials, Oxford University,
Oxford OX1 3PH, England.
b͒
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
0003-6951/98/73(13)/1862/3/$15.00 1862 © 1998 American Institute of Physics
129.22.67.107 On: Mon, 24 Nov 2014 11:42:25

Appl. Phys. Lett., Vol. 73, No. 13, 28 September 1998
Kamins, Briggs, and Williams
1863
FIG. 1. 1 ␮mϫ1 ␮m tapping-mode atomic force micrographs of 11 eq-ML Ge/Si͑001͒ films deposited for 180 s. The inset graph shows the model for the
thickness, th, with square law (nϭ2) dependence on HCl flow and empirical temperature dependence for the coefficients A and B.
inset to Fig. 1, where by using an exponent of nϭ2 ͑together
with an empirical temperature dependence͒ it is possible to
fit all deposition thickness results to a single master relation-
ship. In order to elucidate the effect of the HCl on island type
and size, further samples were grown with the GeH4 flow
adjusted ͑using this relationship͒ to keep the growth rate con-
stant as the HCl flow varied so that the time available for
surface diffusion at a given coverage was the same.
than in Fig. 2͑a͒: if this is correct then it would suggest that
the Cl inhibits transfer of Ge from the wetting layer to the
islands.
A more comprehensive map of the effects of HCl and of
temperature is shown in Fig. 1. The Ge coverage is 11.5
Ϯ0.6 eq-ML in each case. At this coverage the proportion of
pyramids is less than at 8 eq-ML, but at 600 °C a few are
present, and the proportion increases systematically with
The effect on island formation of HCl during growth can
be seen by comparing the two atomic force micrographs in
Fig. 2. The coverage is 8 eq-ML. Figure 2͑a͒ shows islands
grown at 600 °C without HCl. Two kinds of island are vis-
ible. The smaller islands are pyramid shaped; other studies2
indicate that these have ͕105͖ facets and are the first to form
at low coverage. The larger islands are dome shaped; these
have additional facets and form after pyramids appear, even-
tually superseding them. Figure 2͑b͒ shows islands grown at
the same temperature in the presence of HCl. The Ge depo-
sition rate and total coverage are the same, but the appear-
ance is quite different. Those domes that are present appear
slightly smaller than in Fig. 2͑a͒, but the dominant effect is
that the relative distributions of pyramids and domes is quite
different, with a much higher proportion of pyramids. A
simple estimate from the AFM data indicated that the vol-
FIG. 2. 1 ␮mϫ1 ␮m atomic force micrographs of 8 eq-ML Ge/Si͑001͒
films deposited for 131 s at 600 °C ͑a͒ without HCl ͑b͒ with HCl.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
ume of material in the islands was somewhat less in Fig. 2͑b͒
129.22.67.107 On: Mon, 24 Nov 2014 11:42:25

1864
Appl. Phys. Lett., Vol. 73, No. 13, 28 September 1998
Kamins, Briggs, and Williams
HCl flow. As the temperature is increased, the domes be-
come larger and fewer in number in the absence of HCl; the
presence of HCl still has the effect of increasing the propor-
tion of pyramids, but to a markedly reduced degree at higher
temperatures.
lower temperature, whereas in etching they appear at higher
temperature, and it is hard to see how such a contradiction
could be accounted for by equilibrium effects alone. If the
HCl were to inhibit shape change at a given coverage, for
example by obstructing diffusion on island facets, then we
might expect to see transformation from domes to pyramids
upon etching only at higher temperature, as indeed we do,
but we might also expect to see larger pyramids when grow-
ing with HCl, which we do not. However, if the dominant
effect of the HCl is to inhibit diffusion of Ge between is-
lands, then the coarsening process whereby the population
shifts from pyramids to domes during growth as the coverage
increases should be delayed just as is observed at 600 °C.
Such a barrier to diffusion is expected to diminish in effect
as the temperature is raised ͑both because of thermal activa-
tion over any barrier, and because the Cl will remain on the
surface for less time͒, which is again consistent with what is
seen. Given that Cl is active on the surface, we observe that
the exponent nϭ2 should be interpreted at least in part in
terms of a surface reaction.
The relationship between kinetic and energetic con-
straints is a key issue in the formation of nanostructures on
surfaces. We believe that the reason that HCl changes the
relative distribution of pyramids and domes during growth is
because it impedes the surface diffusion of adsorbed Ge at-
oms. It does not appear to affect the size at which the tran-
sition from pyramids to domes occurs during growth, nor to
delay the shape transition once an island is large enough. On
the other hand, the reversion of domes to pyramids upon
etching at 600 °C is very strong evidence that pyramids are a
stable island type, and that the transition to them is revers-
ible. Can etching with HCl at lower temperatures be used to
obtain domes that are small enough and have a sufficiently
uniform size distribution to be useful for quantum devices?
Unfortunately not. The domes after etching at 550 °C are
rather flat in shape. The etching process reduces their height
but not their diameter, and the uniformity of their size distri-
bution is no better than that of pyramids at a similar coverage
grown without HCl.
How are we to account for these observations? A gas
phase reaction between Ge and Cl containing species could
have the effect of slowing down the actual rate of deposition
of Ge on the surface. Added Cl can certainly form stable
chlorogermanes,8 which has the effect of reducing growth
rates in chemical vapor deposition ͑CVD͒ of alloys, and the
effect on nonuniformity of growth close to an area of the
surface covered in oxide is consistent with mechanisms
based on gas phase transport. However the effect is less for
Ge than it is for Si, and most of the studies on alloy films
have been performed at higher temperatures than those used
here.9 A slower growth rate could give more time for coars-
ening of the island population from pyramids to domes10 ͑the
opposite of the observed effect of HCl͒. Slower growth also
allows more time for alloying with silicon from the substrate,
thus increasing the equilibrium size of each island type and
causing a reversion to a higher proportion of pyramids at a
given coverage5 ͑though this would be more significant at
higher temperatures͒. But since the average growth rate was
constant for all the samples in Fig. 1, these considerations do
not apply.
Because gas phase reactions can be eliminated, the
changing island distributions must be caused by the presence
of Cl on the growing surface. This could be due either to
equilibrium effects, if the HCl acts as a morphactant chang-
ing the relative energies of different crystallographic facets
under strain, or due to kinetic effects, if the HCl inhibits
transformation of shape from a pyramid to a dome or trans-
port of Ge from one island to another. Wet chemical treat-
ment with HCl can make a Ge͑111͒ surface stable in air,11
and it is possible that a Ge͑001͒ surface can be similarly
stabilized by Cl, presumably by forming a monochloride ter-
mination by analogy with the monohydride ͑the Ge–Cl bond,
2.23 eV, is stronger than Ge–H, 1.67 eV͒. Therefore if the
surface has a high exposure to Cl, it is quite possible that the
Ge dangling bonds are saturated with Cl, thus either chang-
ing relative surface energies and/or inhibiting Ge diffusion.
To provide further evidence to help resolve these issues,
we performed a series of etching experiments, in which HCl
was introduced into the H2 carrier gas flow shortly after the
deposition was completed; since the etch rate of Ge is orders
of magnitude greater than Si under these conditions, little if
any of the Si is etched. Two samples were grown at 600 °C
with 11 eq-ML Ge. One sample was etched at 600 °C, and
the other was etched at 550 °C, with very different results.
At both temperatures material is removed from the islands,
so that their average size is substantially reduced. At 600 °C
the islands change shape from domes to pyramids as they get
smaller. At 550 °C the domes do not transform to pyramids;
their height decreases faster than their width but they remain
domes.
The authors thank D. Ohlberg for assistance with the
AFM measurements, Dr K. Nauka for RBS measurements,
and the Hewlett–Packard ULSI Laboratory for use of growth
facilities.
1 M. Tomitori, K. Watanabe, M. Kobayashi, and O. Nishikawa, Appl. Surf.
Sci. 76/77, 322 ͑1994͒.
2 G. Medeiros-Ribeiro, A. M. Bratkovski, T. I. Kamins, D. A. A. Ohlberg,
and R. S. Williams, Science 279, 355 ͑1998͒.
3 I. Goldfarb, J. H. G. Owen, P. T. Hayden, K. Miki, and G. A. D. Briggs,
Surf. Sci. 394, 105 ͑1997͒.
4 T. I. Kamins and R. S. Williams, Appl. Phys. Lett. 71, 1201 ͑1997͒.
5 T. I. Kamins, G. Medeiros-Ribeiro, D. A. A. Ohlberg, and R. S. Williams
͑unpublished, 1998͒.
6 A. Ishitani, N. Endo, and H. Tsuya, Jpn. J. Appl. Phys., Part 2 23, L391
͑1984͒.
7 T. I. Kamins, E. C. Carr, R. S. Williams, and S. J. Rosner, J. Appl. Phys.
81, 211 ͑1997͒.
8 T. I. Kamins, J. Appl. Phys. 74, 5799 ͑1993͒.
9 T. I. Kamins, D. W. Vook, P. K. Yu, and J. E. Turner, Appl. Phys. Lett.
61, 669 ͑1992͒.
The appearance of pyramids upon etching would be con-
sistent with a morphactant effect, although we should then
also expect larger pyramids upon growth in the presence of
10 F. M. Ross, J. Tersoff, and R. M. Tromp, Phys. Rev. Lett. 80, 984 ͑1998͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
HCl. But during growth with HCl there are more pyramids at
11 Z. H. Lu, Appl. Phys. Lett. 68, 520 ͑1996͒.
129.22.67.107 On: Mon, 24 Nov 2014 11:42:25