Thin Solid Films, 239 (1994) 225-228 225

Modelling of silicon epitaxy at low pressure

D. K. Pal and A. N. Daw

Centre for Microelectronics, Institute of Radio Physics and Electronics, University College of Science and Technology, 92 Acharya Prafulla
Chandra Road, Calcutta 700-009 (India)
(Received March 2, 1993; accepted September 16, 1993)

Abstract
A model for simulating the growth rate of epitaxial silicon in the vapour phase by pyrolysis of Sill 4 in a horizontal
reactor in the pressure range 10-500 Torr has been developed. The model takes into account the variation of
diffusion coefficient, gas density and surface reaction rate constant with pressure. Using the known dependences of
these parameters on pressure, an expression for the growth rate of an epilayer has been derived as a function of
temperature, pressure, flow rate, silane mole fraction, position of wafer, etc. The results of simulation under
appropriate conditions have been found to be in good agreement with experimental data as available in the
literature. \

1. Introduction atmospheric pressure. The primary factors responsible

For some time researchers have shown an interest in
low pressure silicon epitaxy in view of several advan-
tages which this technique possesses, such as reduced
autodoping, enhanced uniformity of films, decrease in
pattern shift, etc. A number of workers [1-4] have
studied in detail the dependence of the growth rate of
epitaxial silicon under low pressure as a function of
pressure, silane mole fraction, temperature etc. Not
much effort has, however, been made to develop simple
models suitable for simulating the growth of epilayers
under low pressure. Since availability of a satisfactory
model for any process is very useful for efficient process
design and better process control, especially for very-
large-scale integration manufacturing, an attempt was
made to develop such a model for simulation of growth
over the pressure range 10-500 Torr, for which experi-
mental data are available in the literature. In this paper
we report the results of such a study for a horizontal
reactor with silane as the source material. In Section 2,
the various relations for simulating the process quanti-
tatively are derived and in Section 3 the dependence o f
growth rate as predicted by these relations is described
and compared with available experimental data.
2. Theory
In order to derive an expression for growth rate in
the pressure range 10-500 Torr, we note that at such
pressures a number o f changes take place in the gas
flow dynamics within the reactor compared with that at
for such changes are the increase of diffusion coefficient
and decrease of gas density with lowering of pressure,
and a variation in the surface reaction rate constant
with pressure. Also, owing to reduction o f the thermal
conductivity of the gas layer, an appreciable difference
in temperature is observed between the substrate and
the susceptor.
It is well known that at atmospheric pressure the
kinetics of film growth is controlled by both the gas
phase mass transfer mechanism and chemical surface
reaction, their relative contributions depending on the
temperature at which the growth takes place. At ex-
tremely low pressure of the order of millitorr, the
growth of an epilayer is entirely controlled by chemical
reactions taking place at the surface. In the range
considered by us ( 1 0 - 5 0 0 T o r r ) , however, both these
mechanisms are in operation to varying degrees, de-
pending on the pressure and temperature inside the
reactor. We therefore propose to study the variation of
growth rate in the above pressure range using the
general expression for growth involving both the afore-
said mechanisms as suggested by us earlier [5], and to
incorporate therein such changes of the various physical
parameters as may be caused by the change of pressure.
In our earlier work referred to above, the reactor was
assumed to be a horizontal one. The gas space above
the susceptor was divided into two layers, a convective
layer and a boundary layer of varying thickness 6. The
velocity profile within the reactor above the boundary
layer was taken as a parabolic one a s suggested by
Shepherd [6]. Using the continuity equation within the
convective layer an expression for the flux along the Y

0040-6090[94/$7.00 © 1994-- Elsevier Sequoia. All rights reserved

SSD1 0040-6090(93)02924-3
226 D. K. Pal, A. N. Daw / Modelling of Si epitaxy at low pressure
direction in the convective layer directed towards the
boundary layer was obtained. This flux was equated to
the flux of reactants diffusing along the Y direction in
the boundary layer. Noting that only a portion of the
Sill4 molecules at the surface are reduced to silicon
depending on the values of the surface reaction rate
constant, the expression for the growth rate G of the
epitaxial layer was obtained as
K H exp( - 2 ) N M
G- x 60 × 104 lam min -1 (1)
K + H exp( -2)N1
where K is the surface reaction rate constant (cm s-l),
N the number of molecules per cubic centimetre of the
gas, M the mole fraction of Sill4 in the SiH4-Hz
mixture, N1 the atomic density of the silicon substrate
and H the gas phase mass transfer coefficient (cm s -l)
given [7] by
D D
H - 6 - (tlX/pV) ~/2 (2)
where 6 is the thickness of the boundary layer (cm), t/
the viscosity and p density of the gas, V the gas velocity
and X the coordinate of the point of study along the X
direction, i.e. along the length L of the reactor under
consideration, which has been assumed to be placed in
the horizontal direction, and D is the diffusivity of
silane through the carrier gas Hz. The term 2 in eqn. (1)
is given [5] by the relation
12D X A To
- (3)
b6 FoFT
where b is the height of the reactor tube above the
susceptor, F the input flow rate, T the temperature in
kelvin of the substrate, A the cross-sectional area of the
reactor tube, TO the standard temperature in kelvin
(273 K) and F0 a constnat for a particular reactor. The
reaction rate constant K, as is well known, can be
written as
K = K0 e x p ( - E / R T ) (4)
where R is the universal gas constant and E the activa-
tion energy.
Under low pressure, eqn. (1) cannot be applied
straightaway for calculation of growth rate since, as
already stated, a number of changes that take place
with lowering of the pressure have to be taken into
consideration.
In order to assess the effect of the aforementioned
changes, let us first consider the pressure dependence of
the gas phase mass transfer coefficient H. The depen-
dence :of this parameter on pressure would obviously be
governed by the variation of the diffusion coefficient D
and the b.oundary layer thickness 6 with pressure. It is
known that D varies approximately as lIP [8] and is
given by D = Do Po/P, where Do is the diffusion co-
efficient at atmospheric pressure P0. Regarding g, we
note that at low pressure, the variation of q with
pressure being negligible [9], the pressure dependence of
6 is solely determined by that of the parameter p. Since
p is directly proportional to pressure P [8], it follows
readily that
= 6o (5)
where 60 is the thickness of the boundary layer at
atmospheric pressure P0. In other words, 6 is inversely
proportional to p1/2. Thus, considering the variation of
both D and 6, we come to the conclusion that H would
be inversely proportional to p1/2. In other words, H can
be written as
Let us now consider the pressure dependence of the
surface reaction rate constant K. Of the two parameters
that determine K, namely K0 and E, earlier workers
have measured the activation energy E at different
pressures and have found E to be dependent on pres-
sure [1]. From their data it is observed that, as the
pressure rises from low values, the activation energy
decreases monotonically up to a certain value and
thereafter its variation with pressure becomes notice-
ably slow up to atmospheric pressure. The observed
variation of the activation energy can be expressed by
the following curve-fitting relation:
E = 32.15 + 15.50 exp[ - (P -- 10) x 0.014] (7)
for pressures from 10 to 100 Torr, and
E = 17 + 19.55 exp[ - (P - 100) x 0.005] (8)
for pressures from 100 to 760 Torr.
The parameter K0 in the expression for K has been
assumed to be independent of both pressure and tem-
perature b y earlier workers. This presupposes that a
single reaction is occurring in the process, which may
not always be valid. According to absolute rate theory
of reactions, even for an elementary reaction K0 may be
a function of temperature [10]. Temperature depen-
dences of the form K0 = aT" are suggested by absolute
rate theory [10]. In gas phase reactions, the value of n
is generally small but in condensed phases n may be
quite large. We also feel that the reaction in this case
comprises a series of steps and there may also be
parallel pathways. For such a situation we expect Ko to
show both temperature and pressure dependence. Tak-
ing both these dependences into consideration, we sug-
gest the parameters K0 to be of the form
{To\A'/" po\A2
 D. K. Pal, A. N. Daw I Modelling of Si epitaxy at low pressure
where K*, A 1 and A 2 a r e constants. Substituting eqns.
(4), (6) and (9) in eqn. (1) the modified expression for
growth rate as a function of pressure can be written as
r
) exPL-' °t,T )
G- \T) \P/
[--E(P)] Do/]D \1/2
/ T o \ A , / P o \a2
) expL-Xot ) _IN,
where
12D o X A To
~o broFoFT
In order to calculate the value of G from eqn. (10), it
would obviously be necessary to know the values of the
constants K*, A1 and A 2. Since these values are not
available in the literature, they can be determined for a
given system, provided experimental values of the growth
rate at low pressure for two temperatures are available.
At low pressure, the growth rate would be solely con-
trolled by the parameter K. The values of the constants
K* and A1 can then be found if the values o f G at a given
low pressure are known for two different temperatures.
The value of A 2 can then be determined from the value
of G at any one of these two temperatures.
It should be noted that in the expression for growth
derived above, apart from the reaction rate constant,
which is a function of temperature, the parameter H is
also a function of temperature through the terms D, ~/,
p and V as was mentioned in our earlier work [11]. In
inductively heated horizontal reactors, since there is a
temperature gradient normal to the surface of the sub-
strate, the variations of the aforesaid parameters with
temperature have to be taken into consideration in
making an accurate prediction of growth rate as a
function of temperature. Also, the flow velocity inside
the reactor is not uniform all over the reactor and this
factor has also to be considered while developing an
accurate model of growth by vapour phase epitaxy as
has been done earlier [5].
3. Results and discussion
In order to test the validity o f eqn. (10), it is obvi-
ously necessary to calculate the values of G as a func-
tion of pressure as predicted by eqn. (10) and compare
these with available experimental results. For this pur-
pose, the experimental data reported by Duchemin et
al. [2] were used by us. Their reactor, which was a
horizontal one, was heated by induction at a frequency
of 50 kHz and its walls were air cooled. The control gas
panel was similar to those used for systems working at
atmospheric pressure. Pressure control was carried out
by means o f a mechanical gauge and temperatures were
227
measured by an optical pyrometer, taking the emissivity
of silicon into consideration. The graphite susceptor
used contained three substrates, each 40 mm in diame-
. . . .
x 60 x 104 pm min -1 (10)
r / p ',,1/2-I
ter, and the H 2 flOW was kept constant at the rate of
18 1 min -1, measured under standard conditions of
pressure and temperature.
As mentioned earlier, the first step for calculating the
value of G would be to determine the values of the
constants K*, A 1 and A2. These values were calculated
using the growth rate values reported by Duchemin et
al. [2] at 10 Torr for T equal to 1203 K and 1323 K
and were obtained as 9.183 x 1016cms -1, 14.59 and
0.82 respectively. The values of the various geometrical
and physical parameters used in calculating G were:
D 0 = 0 . 2 c m 2s-1, P 0 = 7 6 0 T o r r , 60=1.49cm and
~0 = 5.87 × 10 -3 for X = 30 cm and T = 1203 K. With
these values, G was calculated as a function of pressure
using eqn. (10) for temperatures of 1203 K and 1323 K
respectively. These values are plotted in Fig. 1 for a
silane mole fraction of 2.1 x 10 -3. The experimental
values of Duchemin et al. are shown as dots for 1203 K
and as crosses for 1323 K in this figure. Satisfactory
agreement between the two is clearly observed.
It is easy to observe that in the high pressure region,
where the parameter H is expected to control the
growth rate [2], the growth rate at constant mole
fraction M would decrease with increase of pressure,
following a p-1/2 law as may be seen from eqn. (6). The
experimental dependence of G on P is in agreement
with this expectation.
0-5
0-4
•0-3
.c
~o.
10 io s'o 16o 260 560 --
Pressure in t o r t
Fig. 1. Growth rate as a function of pressure with temperature as a
parameter for a silane mole fraction of 2.1 x 10-3 and H2 flowrate of
18 1min-~.
228 D. K. Pal, A. N. Daw / Modelling of Si epitaxy at low pressure
Instead of keeping the mole fraction constant, if the
pressure dependence of growth rate is studied keeping
the partial pressure of the reactant gas constant, the
growth rate according to eqn. (1) would become pro-
portional to Hp/P at high pressure and proportional to
Kp/P at low pressure (p being the partial pressure).
Since H varies inversely as p~/2, it follows readily that
the growth rate would be proportional to p - 3 / 2 in the
high pressure region.
In the low pressure region, although the theoretical
relation does not suggest any simple relationship be-
tween growth rate and pressure, experimental values of
growth rate appear to obey an approximate pl/2 depen-
dence when the mole fraction is kept constant. Under
conditions of constant partial pressure, the growth rate
would obviously be expected to be proportional to
p-~/2.
From Fig. 1 it can be seen that the growth rate
increases with increasing temperature in the range 10-
200 Torr while it becomes independent of temperature
for pressures greater than 200 Torr. Since at low pres-
sure the growth rate is solely controlled by the surface
reaction rate constant K, which in turn is strongly
dependent on temperature, the observed dependence of
the growth rate on temperature is expected. At higher
pressure, since the growth rate is controlled by the mass
transfer coefficient H, which does not depend apprecia-
bly on temperature, the growth rate appears to depend
marginally on temperature.
Duchemin et al. [2] have given an experimental plot
of growth rate vs. silane mole fraction with pressure as
a parameter. All these plots are linear in nature. This is
in agreement with eqn. (1) which shows clearly that
growth rate G is directly proportional to mole fraction
M.
Acknowledgments
One of the authors (D. K. Pal) is thankful to the
Ministry of Human Resource and Development, Gov-
ernment of India for the award of a Senior Research
Fellowship. Thanks are also due to Professor P. Roy of
the Department of Chemical Engineering, Calcutta
University, for some helpful discussions.
References
1 M. Ogirima, H. Saida, M. Suzuki and M. Maki, J. Electrochem.
Soc., 124 (1977) 903.
2 M. J. Duchemin, M. M. Bonnet and M. F. Koelsch, J. Elec-
trochem. Soc., 125 (1978) 637.
3 E. Krullmann and W. L. Englo, IEEE Trans. Electron Devices, 29
491 (1982).
4 K. F. Jensen and D. B. Graves, J. Electrochem. Soc., 130 (1983)
1950.
5 D. K. Pal, M. K. Kowar and A. N. Daw, Int. J. Electron., 72
(1992) 103.
6 W. H. Shepherd, J. Electrochem. Soc., 112 (1985) 988.
7 A. S. Groves, Physics and Technology of Semiconductors Devices,
Wiley, New York, 1967.
8 J. Bloem and L. J. Gilling, Mechanisms of the chemical vapour
deposition of Si, Curt. Top. Mater. Sci., I (1972) 219.
9 J. F. O'Hanlon, A User's Guide to Vacuum Technology, Wiley,
New York, 1989.
10 K. J. Laidler, Chemical Kinetics, McGraw-Hill, New York, 1965.
11 A. N. Daw, D. K. Pal and M. K. Kowar, Microelectron. J., 21
(1990) 29.