Full text of DOI:10.1063/1.114104

In situ real time measurement of the incubation time for silicon nucleation on silicon
dioxide in a rapid thermal process
Y. Z. Hu, D. J. Diehl, Q. Liu, C. Y. Zhao, and E. A. Irene
Citation: Applied Physics Letters 66, 700 (1995); doi: 10.1063/1.114104
View online: http://dx.doi.org/10.1063/1.114104
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/66/6?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Effects of hydrogen surface pretreatment of silicon dioxide on the nucleation and surface roughness of
polycrystalline silicon films prepared by rapid thermal chemical vapor deposition
Appl. Phys. Lett. 69, 485 (1996); 10.1063/1.118148
Real time investigation of nucleation and growth of silicon on silicon dioxide using silane and disilane in a rapid
thermal processing system
J. Vac. Sci. Technol. B 14, 744 (1996); 10.1116/1.588708
Realtime process and product diagnostics in rapid thermal chemical vapor deposition using in situ mass
spectrometric sampling
J. Vac. Sci. Technol. B 13, 1924 (1995); 10.1116/1.588110
Growth characteristics of silicon dioxide produced by rapid thermal oxidation processes
Appl. Phys. Lett. 57, 881 (1990); 10.1063/1.104265
Silicon temperature measurement by infrared transmission for rapid thermal processing applications
Appl. Phys. Lett. 56, 961 (1990); 10.1063/1.102592
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:
137.189.170.231 On: Tue, 23 Dec 2014 20:03:19

In situ real time measurement of the incubation time for silicon nucleation
on silicon dioxide in a rapid thermal process
Y. Z. Hu, D. J. Diehl, Q. Liu, C. Y. Zhao, and E. A. Irenea)
Department of Chemistry, University of North Carolina, Chapel Hill, North Carolina 27599-3290
͑Received 15 August 1994; accepted for publication 7 December 1994͒
Real time ellipsometry and atomic force microscopy ͑AFM͒ were used to measure critical
nucleation parameters for polycrystalline silicon deposition on an amorphous SiO2 layer by rapid
thermal chemical vapor deposition ͑RTCVD͒ using disilane ͑5% in helium͒. A particularly important
parameter for selective epitaxial deposition is the time for nuclei to form, the incubation time.
Quantitation of the nucleation parameters, such as the nuclei density, nuclei growth rate, nuclei
coalescence, and an operational incubation time were determined from the real time ellipsometric
measurements and confirmed by AFM. For a substrate temperature of 700 °C and at a chamber
pressure of 0.2 Torr, the nuclei densities of 1.4ϫ1010 nuclei/cm2, incubation time of 26 s and nuclei
layer growth rates of 20 nm/min were obtained. © 1995 American Institute of Physics.
Interest exists for the selective epitaxial growth ͑SEG͒ of
silicon, because this process offers number of
Si RTCVD. Thus, the incubation times we report are longer
than the time to form a critical nucleus.
a
advantages,1,2 such as a via hole filling method for planariza-
tion and a diffusion microsource for attaining accurate dop-
ant profiles. The conventional approach in chemical vapor
deposition ͑CVD͒ processes to obtain selective deposition is
the use of dichlorosilane with H2 and HCl.3 However, the
employment of ultrahigh vacuum chemical vapor deposition
͑UHV\CVD͒ systems restricts the use of chlorine-containing
Si source compounds in favor of silane or disilane ͑Si2H6͒ as
the reactants with H2.4 The use of Si2H6 in Si selective epi-
taxy by CVD is motivated by4 low process temperatures and
high growth rate ͑ten times higher than that of the SiH4
process5͒. The effectiveness of SEG is determined largely by
the nucleation of Si on SiO2. The key parameters that char-
acterize the nucleation process such as the time for stable
nuclei formation, the incubation time, the coalescence point,
and the nuclei saturation density are difficult to measure, and
a technique that allows real time monitoring of these param-
eters is valuable for process development. The conventional
in situ surface probes cannot be used outside the ultrahigh
vacuum environment ͑i.e., at process pressures͒ or in the
highly reactive and high-temperature ambient of a CVD re-
actor. However, ellipsometry has been shown to be capable
of making measurements of nucleation, with monolayer sen-
sitivity and quantification of microstructure.6–13
In this work, we investigated the initial rapid thermal
CVD, RTCVD, nucleation, and growth regime for Si on SiO2
film covered Si, from disilane using real time ellipsometry
and atomic force microscopy ͑AFM͒. The results are pre-
sented in terms of an operational incubation time, the coales-
cence point, nuclei density, size, and growth rate. The
RTCVD system operational incubation time ͑hereafter called
the incubation time͒ is measured from the turn on of the
heating system. An extrapolation method described later is
used to obtain the onset of nucleation. This incubation time
includes the system heat-up time as was done by others14 for
The custom built rapid thermal processing ͑RTP͒ system
consisting of a vacuum system and an ellipsometer was de-
scribed in greater detail elsewhere.15 Essentially, the vacuum
system consists of a load lock and a turbopumped, water-
cooled stainless-steel RTCVD chamber with a base pressure
of 10Ϫ8 Torr. A Baratron and a quadrupole mass spectrometer
are used to monitor the total and partial pressures, respec-
tively. The ellipsometer is also a custom built rotating ana-
lyzer ellipsometer similar to the system described by
Aspnes.16 The ellipsometry measurements are made in an
inverted configuration ͑i.e., measured from below the
sample͒. This configuration is advantageous for RTP, since
with the wafer inverted and heated from above, the wafer
blocks much of the direct radiation from the lamps. The in
situ spectroscopic ellipsometry ͑SE͒ measurements were
made with photon energies between 2.5 and 5.0 eV and at
room temperature after cooling the sample. The real time
single wavelength ellipsometry ͑SWE͒ was carried out at
288 nm ͑4.3 eV͒, since this wavelength is near a Si interband
absorption and has been found to be sensitive17–19 to Si
nucleation. All data are expressed as trajectories in the ⌬, ⌿
ellipsometric measurement plane. In modeling Si nucleation,
the measured ellipsometric ⌬, ⌿ data are compared to the
calculated values as obtained from models to be discussed
below. The nuclei are assumed to grow in a hexagonal net-
work. Upon coalescence of the nuclei, voids may form at the
interface, due to incomplete coalescence and/or on top of the
film to represent surface roughness.
Nucleation experiments were performed on SiO2 ͑ther-
mally oxidized in dry O2 to 24–28 nm thick͒ covered single-
crystal Si ͑c-Si͒ wafers ͓p-type, ͑100͒ oriented͔. A mixture of
5% Si2H6 in He was used. A low substrate temperature of
700 °C was chosen, since low-temperature selectivity is a
current issue in RTCVD processes.
In order to study the initial stage of Si deposition on
SiO2 covered Si, we carried out a series of depositions at
different times at constant substrate temperature after ramp
up, gas pressure, and flow rate. In principle, the deposition of
a new layer on an optically distinct substrate induces a
a͒
Electronic mail: irene@uncvx1.olt.unc.edu
700
Appl. Phys. Lett. 66 (6), 6 February 1995
0003-6951/95/66(6)/700/3/$6.00
© 1995 American Institute of Physics
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:
137.189.170.231 On: Tue, 23 Dec 2014 20:03:19

dielectric function of the Si nuclei which is approximated
from spectroscopic ellipsometry measurements of a thick de-
posited poly-Si film. Using this model, the ⌬, ⌿ trajectory
for the progressive growth of the Si nuclei on the SiO2 cov-
ered Si substrate can be calculated. Some modeling results
are plotted in Fig. 1͑a͒ for mean distances between nuclei
͑D͒ of 0, 50, and 90 nm ͑dashed lines͒. The data correspond
closely to the Dϭ90 nm which yields a nuclei density of
1.4ϫ1010 nuclei/cm2, which as shown below is close to AFM
values. It should be mentioned that assuming purely hemi-
spherical nuclei, i.e., without the benefit of AFM measure-
ments, we obtain Nn of 9.6ϫ1010 nuclei/cm2 or a factor of 6
larger.
Once critical Si nuclei form, their size increases with
time whereas the density remains constant up to coalescence.
This assumption has been justified by Claassen et al.21 at
higher growth temperatures using scanning electron micros-
copy to analyze the nucleation step. We believe that we do
not observe the critical nuclei, because within a few seconds
of the appearance of nuclei, the saturation density is reached.
We can then consider a schematic model of the nucleation
process where the first step is the formation of critical nuclei
separated by a distance D, and in order to render the calcu-
lation simple, the nuclei are assumed to be distributed on a
hexagonal network. The mean nuclei density which is readily
FIG. 1. ͑a͒ ⌬, ⌿ trajectories from real time single wavelength ellipsometry
͑␭ϭ288 nm͒ measurement ͑symbols͒ at 700 °C and 0.2 Torr of Si2H6 ͑5% in
He͒ along with simulations ͑dashed lines͒ assuming a spherical segment
nuclei shape model ͑growth rate ratio: Kϭ0.35͒. ͑b͒ Initial evolution of ⌬ as
a function of time.
calculated from the nuclei distance
D
is 1.4ϫ1010
nuclei/cm2. The incubation time is one of the important pa-
rameters for Si selective epitaxy and it is difficult and tedious
to measure using ex situ observations. However, real time
ellipsometry monitoring of ⌬ versus time as is shown in Fig.
1͑b͒ provides an efficient method to measure this time. From
Fig. 1͑b͒ the minimum change of ⌬ which can be observed is
about 0.5°. From simulations of an optical model for the
caplike shape ͑Kϭ0.35͒ at a wavelength of 288 nm this
value corresponds to about 1, 3, or 5 nm nuclei layer thick-
nesses for the distances between nuclei of 10, 50, or 90 nm,
respectively. This means that by ellipsometry we could not
detect a change of up to 5 nm nuclei layer thickness for a 90
nm nuclei separation. The AFM sensitivity is ͑conserva-
tively͒ less than 1 nm vertically and 2–3 nm horizontally.
The nearly linear evolution of the nuclei layer thickness with
deposition time, obtained from the experimental ellipsomet-
ric trajectory and the best-fit optical model is shown in Fig.
2͑a͒. Assuming a linear relationship from the beginning of
the deposition, we can extend the line to the time axis and
obtain t0 corresponding to the zero thickness, in order to
obtain the incubation time which for our present purposes is
from the start of the lamp power which includes warm up
movement of the ⌬, ⌿ point as the layer thickens. It is
known6 that the growth interface between deposited poly-
crystalline silicon ͑poly-Si͒ and SiO2 becomes rough with
the deposition temperatures higher than about 615 °C. When
the roughness is much smaller than the wavelength of light,
an effective medium approximation ͑EMA͒ has been found
to give an adequate description of the rough layer optical
properties.16 The Bruggeman EMA with isotropic
screening17 was used to calculate the effective dielectric
function of the layer ͑at each thickness͒ from the volume
fractions of the two components, c-Si and voids.
Figure 1͑a͒ shows the ⌬, ⌿ trajectories measured by real
time ellipsometry for a Si film deposited onto a SiO2 covered
Si substrate from Si2H6 ͑5% in He͒ at 700 °C. The data de-
viate significantly from the layer-by-layer film growth model
͑Dϭ0 nm, dashed line͒. The optical model used for simula-
tions of Si nucleation on amorphous oxide consists of a hex-
agonal network of spherically shaped nuclei having the bulk
Si—Si bond density with an average distance D between
nuclei, the radius of the nuclei r(t) increases with time10,20
up to the point where the nuclei come into contact, i.e., coa-
lescence. AFM shows the nuclei geometry to be spherically
shaped but smaller than a hemisphere, and based on these
observations the model used for this study is also shown in
Fig. 1͑a͒. In this model, the growth ratio K, which is defined
as the ratio of the growth rate in the vertical ͑h͒ to the lateral
growth ͑r͒ direction, is 0.35 as obtained as an average from
the AFM observations ͑Kϭ0 and 1 correspond to layer-by-
layer and hemispherical models, respectively͒. From this
model, the Si volume fraction is given as:
fSiϭ␲h2(3ϩK3͒/3•31/2K2D2, where h is the thickness of the
nucleation layer. Modeling requires a prior knowledge of the
and obtain t
0ϭ26 s. This extrapolation somewhat alleviates
the problem of the sensitivity of ellipsometry to small sparse
nuclei.
Observation of the nuclei layer morphology by atomic
force microscopy at progressive time steps in the nucleation
process gives parameters, such as nuclei density Nn average
nuclei height, hn , and mean nuclei area An ͑where Anϭ␲r2n͒.
Figure 3͑a͒ is the AFM image corresponding to the point A in
Fig. 1͑b͒ at tϭ25 s ͑less than the incubation time t0͒, and it
shows no nuclei. Figure 3͑b͒, corresponding to the point B in
Fig. 1͑b͒ at tϭ31 s ͑5 s after t0͒ which shows no change in ⌬
Appl. Phys. Lett., Vol. 66, No. 6, 6 February 1995
Hu et al.
701
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:
137.189.170.231 On: Tue, 23 Dec 2014 20:03:19

have deduced the average nuclei geometry to be a spherical
segment or cap with Kϭ0.35. Figure 3͑e͒, corresponds to the
point E͑82 s͒ in Fig. 1͑a͒ which was about 4 s after nuclei
coalescence as was calculated from the optical model shown
in Fig. 1͑a͒. The mean nuclei area of 5.9ϫ103 nm2 ͑rϭ43
nm͒ is very close to the ellipsometry result of 45 nm. The
good comparison of nuclei density and average nuclei height
as a function of time from AFM and ellipsometry optical
modeling are shown in Figs. 2͑a͒ and 2͑b͒.
In conclusion, we have shown that in situ real time el-
lipsometry can be used to characterize the nucleation of Si
on SiO2 in a RTCVD system using Si2H6 ͑5% in He͒. With
the use of AFM we have been able to measure several nucle-
ation parameters which are then used in the Bruggeman ef-
fective medium approximation for optical data analysis. The
main nucleation parameters such as: prenucleation incuba-
tion time, saturation nuclei density, mean nuclei size, and
nuclei layer growth rate are obtained. The best-fit model for
the poly-Si nucleation on SiO2 at 700 °C and 0.2 Torr is a
cap-shaped optical model with ratio of Kϭ0.35 and nuclei
distance of about 90 nm. From this model a incubation time
of 26 s, nuclei layer growth rate of 20 nm/min, coalescence
nuclei size of 45 nm, and nuclei coalescence were obtained.
¨
FIG. 2. ͑a͒ Initial nuclei layer thickness as a function of time from real time
ellipsometry and comparisons with AFM. ͑b͒ Comparison of nuclei density
from real time ellipsometry and AFM.
¨
The authors would like to thank Dr. M. C. Ozturk and K.
E. Violette for helpful discussions and suggestions. This
work was supported in part by the National Science Founda-
tion ͑NSF͒, Engineering Research Center in North Carolina
State University, and Northern Telecom Electronics ͑NTE͒
Ltd.
from the ellipsometry measurement, gives Nnϭ5.2ϫ109
nuclei/cm2, hnϭ2.3 nm and Anϭ7.2ϫ102 nm2. The smallest
nucleus observed in Fig. 3͑b͒ is about 2 nm in height and 15
nm base diameter. This size is much larger than that
expected22 for a critical nuclei in CVD which often consists
of only a few atoms. This may imply that there is rapid
lateral growth of nuclei from the critical nuclei size by lateral
attachment to the nuclei from surface diffusion. Figures 3͑c͒
and 3͑d͒ are the AFM images corresponding to the points of
C͑50 s͒ and D͑53 s͒ in Fig. 1͑b͒ which yield nuclei densities
of 9.2 and 13.0ϫ109 nuclei/cm2, respectively. The average
heights are about 10 and 12 nm and mean nuclei areas are
2.7ϫ103 nm2 and 2.9ϫ103 nm2 ͑rϭ30 nm͒, from which we
1 K. Baert, P. Deschepper, J. Poortmans, J. Nijs, and R. Mertens, Appl.
Phys. Lett. 60, 442 ͑1992͒.
2 M. L. Cheng, A. Kohlase, T. Sato, S. Kobayashi, J. Murota, and N. Mi-
koshiba, Jpn. J. Appl. Phys. 28, L2054 ͑1989͒.
3 J. A. Friedrich, M. Kastelic, G. W. Neudeck, and C. G. Takoudis, J. Appl.
Phys. 65, 1713 ͑1989͒.
4 K. Aketagawa, T. Tatsumi, and J. Sakai, J. Cryst. Growth 111, 860 ͑1991͒.
5 C. H. Hong, C. Y. Park, and H. J. Kim, J. Appl. Phys. 71, 5427 ͑1992͒.
6 F. Hottier and R. Cadoret, J. Cryst. Growth 56, 304 ͑1982͒.
7 R. W. Collins and J. M. Cavese, J. Non-Cryst. Solids 97&98, 269 ͑1987͒.
8 R. W. Collins and B. Y. Yang, J. Vac. Sci. Technol. B 7, 1155 ͑1989͒.
9 R. W. Collins, Y. Cong, H. V. Nguyen, L. An, K. Vedam, T. Badzian, and
R. Messier, J. Appl. Phys. 71, 5287 ͑1992͒.
10 M. Li, Y. Z. Hu, J. Wall, K. Conrad, and E. A. Irene, J. Vac. Sci. Technol.
A 11, 1886 ͑1993͒.
11 M. Li, Y. Z. Hu, J. F. Wall, D. Diehl, and E. A. Irene, presented at 40th
National Symposium of AVS, Orlando, FL, Nov. 15–19, 1993.
12 M. Li, Y. Z. Hu, E. A. Irene, L. Liu, K. N. Christensen, and D. M. Maher,
J. Vac. Sci. Technol. B 13 ͑in press͒.
13 B. Drevillon, C. Godet, and S. Kumar, Appl. Phys. Lett. 50, 1651 ͑1987͒.
14
¨
¨
K. E. Violette, M. K. Sanganeria, M. C. Ozturk, G. Harris, and D. M.
Maher, J. Electrochem. Soc. 141, 3269 ͑1994͒.
15 K. A. Conrad, R. K. Sampson, H. Z. Massoud, and E. A. Irene, J. Vac. Sci.
Technol. B 11, 2096 ͑1993͒.
16 D. E. Aspnes and A. A. Studna, Appl. Opt. 14, 220 ͑1975͒.
17 D. E. Aspnes and A. A. Studna, Phys. Rev. B 27, 985 ͑1983͒.
18 D. E. Aspnes, J. B. Theeten, and F. Hottier, Phys. Rev. B 20, 3292 ͑1979͒.
19 D. E. Aspnes, Thin Solid Films 89, 249 ͑1982͒.
FIG. 3. AFM image of initial nucleation stage of rapid thermal CVD Si
nuclei on SiO2 surface ͑a͒, ͑b͒, ͑c͒, ͑d͒, and ͑e͒ corresponding to A, B, C, D
in Fig. 1͑b͒ and E in Fig. 1͑a͒. ͑a͒ tϭ25 s, none; ͑b͒ tϭ31 s, Nnϭ5.2ϫ109
nuclei/cm2, hnϭ2.3 nm, Anϭ7.2ϫ102 nm2; ͑c͒ tϭ50 s, Nnϭ9.2ϫ109
nuclei/cm2, hnϭ10 nm, Anϭ2.7ϫ103 nm2; ͑d͒ tϭ53 s, Nnϭ1.3ϫ1010
nuclei/cm2, hnϭ12 nm, Anϭ2.9ϫ103 nm2; ͑e͒ tϭ82 s, Nnϭ1.0ϫ1010
nuclei/cm2, hnϭ14 nm, Anϭ5.9ϫ103 nm2.
20 A. M. Antoine B. Drevillon, and P. R. Cabarrocas, J. Appl. Phys. 61, 2501
͑1987͒.
21 W. A. P. Claassen and J. Bloem, J. Electrochem. Soc. 127, 194 ͑1980͒.
22 J. Bloem, J. Cryst. Growth 50, 581 ͑1980͒.
702
Appl. Phys. Lett., Vol. 66, No. 6, 6 February 1995
Hu et al.
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:
137.189.170.231 On: Tue, 23 Dec 2014 20:03:19