Nano-structuring in SiGe by oxidation induced anisotropic Ge self-organization
Ethan Long, , Augustinas Galeckas, , Andrej Yu Kuznetsov, , Antoine Ronda, , Luc Favre, , Isabelle Berbezier,
and , and Henry H. Radamson
Citation: Journal of Applied Physics 113, 104310 (2013); doi: 10.1063/1.4794991
View online: http://dx.doi.org/10.1063/1.4794991
View Table of Contents: http://aip.scitation.org/toc/jap/113/10
Published by the American Institute of Physics
Articles you may be interested in
Oxidation rate enhancement of SiGe epitaxial films oxidized in dry ambient
Applied Physics Letters 83, 3713 (2003); 10.1063/1.1622439
Effects of Ge concentration on SiGe oxidation behavior
Applied Physics Letters 59, 1200 (1998); 10.1063/1.105502
Oxidation studies of SiGe
Journal of Applied Physics 65, 1724 (1998); 10.1063/1.342945
Nano-structuring in SiGe by oxidation induced anisotropic
Ge self-organization
Ethan Long,
1,a)
Augustinas Galeckas,
1
Andrej Yu Kuznetsov,
1
Antoine Ronda,
2
Luc Favre,
2
Isabelle Berbezier,
2
and Henry H. Radamson
3
1
University of Oslo, Blindern, 0316 Oslo, Norway
2
Univ. Aix Marseille, Campus St. J
er
^
ome, 13397 Marseille Cedex 20, France
3
Royal Institute of Technology (KTH), Electrum 29, 16440 Kista, Sweden
(Received 31 January 2013; accepted 25 February 2013; published online 14 March 2013)
The present study examines the kinetics of dry thermal oxidation of (111), (110), and (100)
silicon-germanium (SiGe) thin epitaxial films and the redist ribution of Ge near the oxidation interface
with the aim of facilitating construction of single and multi-layered nano-structures. By employing a
series of multiple and single step oxidations, it is shown that the paramount parameter controlling the
Ge content at the oxidation interface is the oxidation temperature. The oxidation temperature may
be set such that the Ge content at the oxidation interface is increased, kept static, or decreased. The Ge
content at the oxidation interface is modeled by considering the balance between Si diffusion in SiGe
and the flux of Si into the oxide by formation of SiO
2
. The diffusivity of Si in SiGe under oxidation is
determined for the three principal crystal orientations by combining the proposed empirical model with
data from X-ray diffraction and variable angle spectroscopic ellipsometry. The orientation dependence
of the oxidation rate of SiGe was found to follow the order: ð111Þ > ð110Þ > ð100Þ. The role of
crystal orientation, Ge content, and other factors in the oxidation kinetics of SiGe versus Si are
analyzed and discussed in terms of relative oxidation rates.
V
C
2013 American Institute of Physics.
[http://dx.doi.org/10.1063/1.4794991]
I. INTRODUCTION
There is significant research and industrial interest in
silicon-germanium (SiGe) based nano-structures and devi-
ces.
1
Among numerous examples of how SiGe, in general,
and Ge condensation by thermal oxidation of SiGe, in partic-
ular, may be used for fabrication of nano-scale devices are:
monolithically integrated optical interconnects and wave-
guides,
2
nano-antennas,
3
bolometers for uncooled infrared
photodetectors,
4,5
nano-crystals for use in high density non-
volatile memories,
6,7
multiple gate field effect transistors
(including FinFETs),
8–11
and nano-wires.
12,13
Achieving a
direct bandgap in SiGe core-shell nanowires depends on,
among other things, the nanowire’s orientation and shell
thickness.
14–16
Local oxidation of SiGe has long been pro-
posed as a method to manipulate the Ge content in the chan-
nel or source/drain regions of transistors, which, in addition
to the performance benefits, may help reduce manufacturing
costs and cycle times by eliminating steps from SiGe CMOS
processes.
17
SiGe-on-insulator (SGOI) is a viable replace-
ment for bulk Si in deep sub-micron CMOS applications,
18
and the fabrication of SGOI wafers using Ge condensation
by thermal oxidation
19
as well as by thermally induced Ge
dilution
20
has been suggested. Use of thermal oxidation for
SGOI fabrication may also allow for endotaxial growth of
high Ge content layers and Ge nano-crystals at the interface
between a buried oxide and a SiGe layer.
21
A sound under-
standing of the oxidation of SiGe in multiple crystallo-
graphic orientations will be required to develop processes for
using SiGe in such applications.
The two phenomena commonly discussed in the litera-
ture about oxidation of SiGe are the potential for Ge to act as
a catalyst or inhibitor for oxidation, and the formation of a
Ge-rich layer between the oxide and the underlying SiGe,
referred to as Ge condensation, pile-up, or snow plowing.
22–30
A common explanation for the presumed catalytic effect of
Ge relies on the dissociation energy for a Si-Ge bond being
lower than that of a Si-Si bond,
31–35
while others explain
Ge’s role as a catalyst in terms of the generation of vacancies
and interstitials in the SiGe layers.
22–25,36–38
However, con-
clusions about the role of Ge in determining the oxidation
rate vary widely, and the Ge content at the oxidation interface
is rarely characterized in a systematic way.
39
Furthermore,
except for an early study using (111) oriented material,
40
oxi-
dation of SiGe has been studied with an exclusive focus on
(100) material. The orientation dependence of oxidation of
Si
41–43
may be an indication that SiGe will exhibit similar
behaviour, but it is not obvious that SiGe and Si are perfectly
synonymous in this respect. It has been established that oxida-
tion enhanced diffusion of dopants in Si is tied to both point
defects and crystallographic orientation.
44,45
If point defects
play a role in Si diffusion in SiGe,
22,38,46
then it is likely that
any oxidation enhanced diffusion of Si in SiGe due to point
defects is also orientation dependent. By virtue of the depend-
ence of the Ge condensation on the diffusivity of Si in
SiGe,
26,27
any orientation dependence in the latter will have
a direct consequence on the Ge content at the oxidation
interface.
The present study evaluates the kinetics of oxidation of
SiGe with (111), (110), and (100) oriented thin epitaxial films
of SiGe. The possibility to increase, keep stable, or decrease the
Ge content at the oxidation interface is demonstrated by using
a)
Electronic mail: ethanl@smn.uio.no.
0021-8979/2013/113(10)/104310/7/$30.00
V
C
2013 American Institute of Physics113, 104310-1
JOURNAL OF APPLIED PHYSICS 113, 104310 (2013)
X-ray diffraction (XRD) characterization of SiGe samples after
multiple oxidations. Characterization of samples with a range
of oxide thicknesses and oxidation temperatures shows that the
Ge content in the pile-up region is strongly dependent on oxida-
tion temperature and only weakly dependent on the Ge content
in the underlying SiGe. Lower oxidation temperatures are
showntobelinearlycorrelatedtohigherGecontents,though
the linear temperature dependence of Ge content varies with
crystallographic orientation. The Ge content at the oxidation
interface is modeled by an empirical relationship which consid-
ers the balance between Si diffusion in SiGe and Si flux into
the oxide by formation of SiO
2
. The diffusivity parameters of
Si in SiGe under oxidation are determined for the principal
crystal orientations. The oxidation rates of both Si and SiGe are
found to be dependent on the crystallographic orientation as
well as the presence of Ge at the oxidation interface. The degree
of growth rate enhancement or reduction is discussed in terms
of oxidation rate ratios.
II. EXPERIMENTAL
Epitaxial layers of Si
1X
Ge
X
were grown on (111),
(110), and (100) oriented Si substrates by molecular beam
epitaxy (MBE). The as-grown SiGe layers were composed of
20% Ge, while a supplementary set of (100) oriented sam-
ples contained 15% Ge. Additionally, a set of (100) oriented
samples with Si
0:8
Ge
0:2
layers were grown by chemical
vapor deposition (CVD). The CVD grown samples were
used exclusively for experiments involving repeated oxida-
tions. Those samples which were subjected to repeated oxi-
dations had their oxides removed by a timed buffered
hydrofluoric acid etch between each oxidation. All as-grown
Si
1X
Ge
X
layers had thicknesses of 80 nm. Bare Si sub-
strates were used as reference sa mples for all oxidation runs.
The thermal oxidations were carried out at ambient pres-
sure (1 atm) in a tube furnace flushed with dry O
2
.Forany
given oxidation time and temperature, all samples were proc-
essed simultaneously in order to ensure identical oxidation con-
ditions between samples with various characteristics (i.e., SiGe,
Si, crystal orientation). Oxidations for (111), (110), and (100)
oriented samples were carried out at 900, 950, and 1000
C
with oxidation times chosen to target 20, 40, 60, 80, and
100 nm thick oxides. Supplementary (100) oriented Si
0:85
Ge
0:15
and Si
0:80
Ge
0:20
samples were oxidized at 780, 820, 870, 920,
or 960
C to grow oxides between 0 and 60 nm thick.
XRD measurements were made with a diffractometer in
double axis configuration. The incident beam was composed
of Cu-K
a1
radiation, while Cu-K
a2
and Cu-K
b
radiation was
removed with a G
€
obel mirror and Ge monochromator. The
peaks for the 2h-x scans were chosen according to sample ori-
entation, i.e., the (004) peak for (100), the (333) peak for
(111), and both (022) and (044) peaks for (110) oriented mate-
rial. The profiles from the 2h-x scans were fit using a 3-layer
model, lattice constants from Dismukes et al.
47
and the LEPTOS
simulation software. Reciprocal space maps of a limited num-
ber of samples confirmed that the SiGe layers were pseudo-
morphically strained before and after oxidation.
Oxide thicknesses were measured by variable angle
spectroscopic ellipsometry. Measurements were recorded at
65
; 70
,and75
with photon energies varied between 1.39
and 3.25 eV in increments of 0.01 eV. Oxide thicknesses were
determined using a multi-layer model, optical constants for
SiO
2
, Si, and SiGe from literature,
48,49
and the COMPLETEEASE
software.
III. RESULTS AND DISCUS SION
A. Ge content in the pile-up
A series of multi-step oxidations was performed to high-
light the relative influence of temperature and initial Ge con-
tent on the pile-up of Ge at the oxidation interface. Figure 1
shows XRD scans for Si
0:8
Ge
0:2
samples subjected to one,
two, and three separate oxidations at progressively lower
temperatures. The XRD scans are aligned to the Si substrate
peak at 69:13
. The peak at 68:10
arises from the as-grown
SiGe layer and reflects the 20% Ge content of the layer. The
left most peaks correspond to the Ge pile-up layers that form
as a result of the oxidations. After oxidation, the intensity of
the XRD peak for the as-grown layer will be reduced as a
result of the thinning of the layer. For the oxidized samples
in Fig. 1, the oxide and pile-up layers were thick enough so
that any extant signal from the as-grown layer is obscured.
The shift in the 2h position of the pile-up peaks from high to
low angles indicates an increase in the Ge content of the
pile-up layer, X
pu
. The first sample was s ubjected to a single-
step oxidation at 1000
C, which resulted in X
pu
¼ 0:310.
The second sample was subjected to a two-step oxidation:
the same oxidation at 1000
C and a subsequent second oxi-
dation at 900
C, resulting in X
pu
¼ 0:466. The third sample
underwent a three-step oxidation at 1000, 900, and then
800
C, resulting in X
pu
¼ 0:572. Despite the Ge content at
the oxidation interface increasing with multiple oxidations at
progressively lower temperatures, T, these results are con-
sistent with what is predicted by empirical relations for
X
pu
ðTÞ that are based on single oxidations of Si
0:80
Ge
0:20
and
FIG. 1. XRD scans of the (004) peaks of (100) oriented SiGe samples after
multi-step oxidations with decreasing temperatures. The 2h position for the
as-grown sample is marked for reference.
104310-2 Long et al. J. Appl. Phys. 113, 104310 (2013)
Si
0:85
Ge
0:15
alloys.
27
That is, the value of X
pu
depends crit-
ically on the oxidation temperature, and is largely independ-
ent of the Ge content in the underlying SiGe. In the case of
multiple oxidations at progressively lower temperatures, the
Ge content at the oxidation interface, X
pu
, is primarily deter-
mined by the temperature of the last oxidation performed,
despite the progressively increasing X
pu
.
Figure 2 shows XRD scans of Si
0:8
Ge
0:2
samples sub-
jected to a similar scheme of multi-step oxidations. A set of
four samples was first oxidized at 1000
C in order to create
a thick pile-up layer with X
pu
¼ 0:310. Three samples were
subsequently subjected to an additional oxidation step at
1120, 1000, or 900
C. These temperatures were chosen to
induce a decrease, no change, and an increase in X
pu
by
following the previously published analysis for single
oxidations of SiGe(100).
27
Indeed, the XRD scans in Fig. 2
reveal that the secondary oxidations at 1120, 1000, and
900
C have caused X
pu
to shift from 0.310 to 0.217, 0.331,
and 0.466, respectively. As stated above, X
pu
is determined
primarily by the temperature of the last oxidation conducted.
However, the Ge content at the oxidation interface is
increased from 0.20 to 0.31 after the first oxidation at
1000
C. The higher Ge content at the oxidation interface at
the start of the second oxidation had the consequence of
increasing X
pu
by 2% after the second oxidation at
1000
C. This effect is evident in the empirical relation for
X
pu
ðT; N
SiGe
Þ
27
(also in Eq. (1)), where N
SiGe
is the Si density
in the pr imary SiGe layer.
An additional series of oxidations was conducted on
SiGe and Si samples to investigate the influence of crystallo-
graphic orientation on the formation of the pile-up region
and on the oxidation kinetics of SiGe. These oxidation runs
involved a single oxidation of as-grown Si
0:8
Ge
0:2
and Si
samples, though a variety of oxidation temperatures and
times were used for different oxidation runs. Figure 3 shows
the oxide thickness versus oxidation time for 900, 950, and
1000
C. The oxidation rates are ordered as ð111 Þ > ð110Þ >
ð100Þ for both Si and SiGe. Most of the oxidation runs per-
formed at 900 and 1000
C result in SiGe oxidizing faster
than Si, but the longer oxidations at 950
C and the 360 min
oxidation at 900
C show Si oxidizing faster than SiGe.
Figure 4 shows typical results of XRD measurements
performed to quantify X
pu
for the samples described in Fig.
3. There are three distinct peak positions: the substrate peak
at 95
, the peaks at 93:8
from the primary SiGe layers,
and the leftmost peaks corresponding to the pile-up layers.
The pile-up layer peaks are distinguished by their separation
according to oxidation temperature, while oxide thickness
does not have a profound influence on X
pu
.
The dependence of X
pu
on crystallographic orientation
and temperature is illustrated in Fig. 5. Even though X
pu
ðTÞ
is orientation dependent, linear fits to the measured values
reveal nearly identical slopes for all three orientations.
FIG. 2. XRD scans of the (004) peaks of (100) oriented SiGe samples after
various multi-step oxidation schemes. The 2h positions of the peaks indicate
an increase, no change, and a decrease in X
pu
. The scan for the as-grown
sample is omitted for clarity, but its 2h position is marked for reference.
FIG. 3. Oxide thickness versus oxidation time at (a) 900, (b) 950, and (c) 1000
C. The data are for (111), (110), and (100) oriented Si
0:8
Ge
0:2
and Si.
104310-3 Long et al. J. Appl. Phys. 113, 104310 (2013)
B. Diffusivity of Si in SiGe and the oxidation rate
As detailed in earlier publications,
26,27
the magnitude of
X
pu
results from the diffusion induced flux of Si towards the
oxidation front, J
pu
, and the flux of Si into the oxide due to
formation of SiO
2
, J
ox
, being balanced such that J
ox
=J
pu
¼ 1.
Thus, changes to the oxidation rate must be matched by
changes to the diffusion of Si in SiGe, which appears as a
change in X
pu
. It is well established in the literature that the
oxidation rate of Si depends on its crystallographic
orientation,
41,50–53
and the data in Fig. 3 confirm that this is
also true for SiGe. Furthermore, the orientation dependent
diffusivity of dopants observed in Si under oxidation
44,45
may indicate that the diffusivity of Si in SiGe is also orienta-
tion dependent. Consequently, the orientation dependence of
both the oxidation rate of SiGe and the diffusivity of Si in
SiGe will alter the flux balance, J
ox
=J
pu
¼ 1, and thus, mod-
ify X
pu
ðTÞ.
The diffusivities of Si in (111), (110), and (100) oriented
SiGe are determined by comparing values for X
pu
as meas-
ured by XRD to values calculated with the empirical
relation
27
X
pu
¼
k
B
Tln
4N
2
SiGe
D
0
t
pN
2
ox
z
2
ox
E
Si
E
m
; (1)
where z
ox
is the oxide thickness from ellipsometry, T is
the oxidation temperature, t is the oxidation time, N
ox
¼ 2:21 10
22
cm
3
is the atomic density of Si in SiO
2
,
N
SiGe
is the Si density in the primary SiGe layer, and k
B
is
the Boltzmann constant. The diffusivity of Si in SiGe is
described by an Arrhenius relation, D ¼ D
0
exp½ðE
m
X
pu
þ E
Si
Þ=ðk
B
TÞ, where the same activation energy for Si self-
diffusion, E
Si
¼ 4:76 eV,
54
is used for all three crystallo-
graphic orientations. The diffusion parameters D
0
and E
m
were determined independently for the (111), (110), and
(100) orientations by fitting the calculated and measured val-
ues of X
pu
using the method of least squares; the results are
summarized in Table I. The correlation between measured
and calculated results for X
pu
is shown in Fig. 6.
The apparent linearity of X
pu
ðTÞ in Fig. 5 can be under-
stood if one models both the diffusivity of Si in SiGe and the
oxidation rate by Arrhenius relations. Although more refined
oxidation models exist, for the range of oxide thicknesses
considered here, a simple Arrhenius relation is consistent
with the literature
41,51–53
and appears as an obvious choice
when evaluating the balance of Si fluxes, J
ox
=J
pu
¼ 1. Thus,
FIG. 4. XRD 2h-x scans of the (333) peaks of (111) oriented Si
0:8
Ge
0:2
oxi-
dized at various temperatures and times. Five samples with oxide thick-
nesses between 20 and 100 nm are shown for each temperature.
FIG. 5. XRD measurements of the Ge content in the pile-up layer, X
pu
, ver-
sus oxidation temperature, T, along with linear fits to the data.
TABLE I. Parameters for diffusivity of Si in SiGe for different orientations.
Orientation E
m
(eV) D
0
ðcm
2
=sÞ
111 1.81 199
110 1.89 219
100 1.70 239
FIG. 6. Correlation between X
pu
values measured by XRD and those calcu-
lated by Eq. (1). The diagonal line indicates where the measured and calcu-
lated values are exactly equal and is drawn for visual guidance only.
104310-4 Long et al. J. Appl. Phys. 113, 104310 (2013)