Measurements of High-Temperature Silane Pyrolysis Using Sih4 IR Emission and SiH2 Laser Absorption

Article abstract of DOI:10.1021/jp0302663

The thermal decomposition of silane highly diluted in argon was observed behind reflected shock waves over the temperature range from 1060 to 1730 K and pressures between 0.6 and 5.0 atm, making this the first SiH₄ decomposition study to thoroughly investigate pressures above 1 atm. Silane and silylene time histories were monitored using the infrared emission of SiH ₄ near 4.7 μm and laser absorption of a SiH₂ A-X transition near 579 nm. Reaction rate coefficients for the first- and second-order forms of the decomposition reaction SiH₄ + M = SiH₂ + H₂ + M were determined from the species measurements for M = Ar. The effects of competing reactions were considered using a detailed reaction mechanism. The bimolecular rate constant over the pressure range considered herein was determined to be k_1a = 7.2 × 10¹⁵ exp(-E/RT), with k_1a in cm³/mol s and E = 45.1 ± 1.2 kcal/mol. This second-order rate coefficient describes the silane decomposition reaction over the entire range of temperatures and pressures herein and shows good agreement with the results of previous studies at similar temperatures as well as those at lower pressures and at temperatures above and below the values herein. The implication of the second-order reaction rate is that the silane decomposition is still in the low-pressure limit at total mixture pressures as high as 5 atm. A pressure-dependent rate is therefore not needed for SiH₄ decomposition over a wide range of conditions of practical interest when the reaction is simply expressed in the bimolecular form.

Full text of DOI:10.1021/jp0302663

10988
J. Phys. Chem. A 2003, 107, 10988-10995
Measurements of High-Temperature Silane Pyrolysis Using SiH4 IR Emission and SiH2
Laser Absorption
,†
‡
Eric L. Petersen* and Mark W. Crofton
Space Materials Laboratory, The Aerospace Corporation, El Segundo, California 90245, and Mechanical,
Materials & Aerospace Engineering, UniVersity of Central Florida, P. O. Box 162450, Orlando, Florida 32816
ReceiVed: February 27, 2003; In Final Form: August 14, 2003
The thermal decomposition of silane highly diluted in argon was observed behind reflected shock waves over
the temperature range from 1060 to 1730 K and pressures between 0.6 and 5.0 atm, making this the first SiH
decomposition study to thoroughly investigate pressures above 1 atm. Silane and silylene time histories were
monitored using the infrared emission of SiH near 4.7 µm and laser absorption of a SiH A˜ - X˜ transition
near 579 nm. Reaction rate coefficients for the first- and second-order forms of the decomposition reaction
SiH + M ) SiH + H + M were determined from the species measurements for M ) Ar. The effects of
competing reactions were considered using a detailed reaction mechanism. The bimolecular rate constant
4
4
2
4
2
2
1
5
over the pressure range considered herein was determined to be k1a ) 7.2 × 10 exp(-E/RT), with k1a in
3
cm /mol s and E ) 45.1 ( 1.2 kcal/mol. This second-order rate coefficient describes the silane decomposition
reaction over the entire range of temperatures and pressures herein and shows good agreement with the results
of previous studies at similar temperatures as well as those at lower pressures and at temperatures above and
below the values herein. The implication of the second-order reaction rate is that the silane decomposition is
still in the low-pressure limit at total mixture pressures as high as 5 atm. A pressure-dependent rate is therefore
4
not needed for SiH decomposition over a wide range of conditions of practical interest when the reaction is
simply expressed in the bimolecular form.
Introduction
reaction rate for reaction 1b within a larger chemical kinetics
mechanism; that is, certain rates quoted in the literature for
reaction 1b are only valid for the pressure of that study. The
pressure dependence can be estimated if the low- and high-
pressure limits were available and utilized in the method of
Because of its importance as a source of silicon in the
materials processing industry and in the manufacture of glass
containing silicon oxides, the thermal decomposition of gaseous
silane has received much attention in recent years. Among those
investigations are the early reaction vessel experiments of
1
9
Troe, for example, or if some other method were used as in
2
0
Coltrin et al. Unfortunately, most of the analytical studies of
silane pyrolysis in the falloff region and high-pressure limit are
based on the early shock-tube experiments of Newman et al.,3
whose rate coefficient (according to other researchers) appears
to be an order of magnitude larger than recent shock-tube
1
2
Hogness et al. and Purnell and Walsh, several high-temperature
shock-tube studies,3
-11
RRKM calculations,
11-13
and various
,4
other experimental works.1
4-16
Within these and related studies,
it was conjectured and proven that the initial decomposition of
SiH4 over the temperature range of interest (<2000 K) forms a
silylene radical and a hydrogen molecule2
1
0
experiments, further clouding the issue.
,17,18
When the conditions are such that the reaction is sufficiently
close to the low-pressure limit, the reaction rate constant, k1a,
of the bimolecular reaction is not pressure-dependent and can
be expressed in the usual Arrhenius form
SiH + M ) SiH + H + M
(R1a)
4
2
2
where M is any third-body collision partner when written as a
second-order reaction. However, the rate of silane decomposition
for pressures less than 1 atm is often expressed in the literature
in terms of the reaction rate for the unimolecular decomposition,
reaction 1b
k ) A exp(-E /RT)
(1)
i
a
In eq 1, ki is the rate constant of reaction i, A is the
preexponential constant, Ea is the activation energy in kcal/mol,
R is the universal gas constant, and T is the temperature in K.
SiH ) SiH + H
2
(R1b)
4
2
3
The units of eq 1 are cm /mol s for bimolecular reactions and
Although most researchers express the silane decomposition
as R1b with some pressure dependence, there is some discrep-
ancy as to whether the reaction is in the pressure-dependent
-1
s
for unimolecular reactions. An overview of previous
measurements of SiH4 decomposition, including temperature and
pressure ranges and the resulting activation energies, is provided
in Table 1.
falloff region at pressures on the order of a few atm or
below.7
,9,11-13
Care must therefore be taken when using a
2
For the activation energy, Purnell and Walsh measured a
value of 51.2 kcal/mol (for k1b) over a temperature range from
*
To whom correspondence should be addressed. E-mail: petersen@
6
48-703 K and pressures from 0.05-0.30 atm, and Meunier
mail.ucf.edu.
1
4
†
‡
et al. found a value of 46.0 kcal/mol from their laser-induced
University of Central Florida.
The Aerospace Corporation.
chemical vapor deposition experiment (723-848 K). In the first
1
0.1021/jp0302663 CCC: $25.00 © 2003 American Chemical Society
Published on Web 11/14/2003

Measurements of High-Temperature Silane Pyrolysis
TABLE 1: Summary of Measurements of SiH Decomposition
J. Phys. Chem. A, Vol. 107, No. 50, 2003 10989
4
workers
ref
E
a
(kcal/mol)
temp. (K)
press. (atm)
method
Hogness et al. (1936)
Purnell and Walsh (1966)
Newman et al. (1978)
Newman et al. (1979)
Tanaka et al. (1983)
Votintsev et al. (1987)
Tanaka et al. (1987)
Meunier et al. (1987)
Koshi et al. (1991)
1
2
3
4
5
6
7
14
9
51.7
51.2
56.1
52.7
39.4
64.5
48.1
46.0
38.1
48.0
39.9
45.1
653-763
648-703
<1.0
reaction vessel
reaction vessel
shock tube
shock tube
shock tube
shock tube
shock tube
LI CVD reactor
shock tube
shock tube
0.05-0.3
5.0
1200-1300
1035-1184
1700-3900
1400-1800
1250-1600
723-848
5.3
0.1-0.5
1.0
0.25-1.3
0.006-0.013
0.18-1.2
0.7
0.001-0.004
0.6-5.0
1188-1574
1250-1715
623-748
Mick et al. (1993)
Han et al. (1996)
Petersen and Crofton (2003)
10
15
this study
reaction vessel
shock tube
1060-1730
shock-tube studies of silane decomposition, Newman et al.3,4
obtained (for k1b) a value of 56.1 kcal/mol in their earlier paper
Purity of the driven-section volume was maintained by a
1000-L/s Leybold TMP1000C turbomolecular pump, and ulti-
-7
(
1200-1300 K, 5.0 atm) and 52.7 kcal/mol in their later paper
mate pressures below 5 × 10 Torr were obtained via overnight
(1035-1184 K, 5.3 atm). At much higher temperatures and over
pumping. Typical pre-experiment fill pressures were at least
5
-6
a much wider range (1700-3900 K), Tanaka et al. found the
activation energy of k1a to be 39.4 kcal/mol. In a later study,
5 × 10 Torr with a worst-case combined leak and outgassing
-
5
rate of 2 × 10 Torr/min. To minimize the possibility of
impurities leaking into the test gas, experiments were routinely
performed within 30 s of filling the tube.
7
Tanaka et al. measured a value of 48.1 kcal/mol for temper-
atures between 1250 and 1600 K. Similarly, Koshi et al. in
9
their shock-tube study (1190-1570 K) noted activation energies
between 37.1 and 39.2 kcal/mol for k1b, depending on the total
mixture pressure. More recently, Mick et al.10 used atomic
resonance absorption spectroscopy of H and Si atoms to
determine an activation energy of 48.0 kcal/mol for k1a for
reflected-shock temperatures between 1250-1715 K at a single
pressure of 0.7 atm. These varied results for Ea may depend to
some extent on the temperature range, the pressure range, and
competing reactions.
Because the products of silane decomposition are condensed
silicon hydrides in the form of a light powder film, care was
taken to clean the inner walls of the shock-tube driven section
every 1-2 experiments. However, exploratory tests were still
performed to determine what impact, if any, the residual powder
on the shock-tube walls would have on subsequent experiments
if a rigorous cleaning procedure were not followed. It was
deduced that, because the initial silane concentrations were very
small (see below), the buildup of residual powder from up to a
half dozen previous experiments had a negligible impact on the
results.
The purpose of the present study was to perform a series of
shock-tube experiments over a range of temperatures (=1100-
1700 K) and pressures (=0.6-5.0 atm) to determine the reaction
Temperatures behind the reflected shock wave were deter-
mined from direct measurement of the incident-shock velocity.
Five fast-response (<1 µs) piezoelectric pressure transducers
(PCB 113A) were used to trigger four time-interval counters
(Fluke PM6666), providing four axial measurements of the
shock speed. Using the velocity extrapolated to the endwall in
conjunction with the standard one-dimensional shock relations
rate constant of silane pyrolysis and its pressure dependence.
To conduct the study, the authors utilized a combination of
established shock-tube techniques, nonintrusive optical diag-
nostics for species concentration measurements, and modern
chemical kinetics modeling. To the authors’ knowledge, the
present work was the first study to use a detailed kinetics model
of the secondary reactions while determining the rate of silane
pyrolysis from the shock-tube data. Details of the experimental
apparatus and procedure are described first, followed by an
analysis of the silane decomposition kinetics and competing
reactions. The results of the experiment are given in terms of
species concentration profiles and curve fits of the reaction rate
constant for silane decomposition, and the results are compared
with previous studies. A discussion of the apparent pressure
dependence is also included.
2
1
and the Sandia thermodynamic database, the temperature and
pressure behind the reflected shock wave were determined. The
uncertainty in the test temperature is on the order of 5 K for
the conditions of this study.
Small amounts of research-grade SiH4 (300-1030 ppm)
highly diluted in ultrahigh purity argon were used as the test-
gas mixtures. These mixtures were produced manometrically
using partial pressures in a mixing tank to a total pressure of
approximately 6 atm. The overall uncertainty in the volume
fraction of silane was less than 1% using the Baratron vacuum
pressure transducers described above.
Apparatus and Procedure
A stainless steel shock tube was utilized for the experiments.
The pressure-driven shock tube operated with helium as the
driver gas and has a driver length of 3.5 m with an internal
diameter of 7.62 cm. The 10.7-m test section has an internal
diameter of 16.2 cm. Lexan diaphragms ranging from 0.1-0.5
mm thick were utilized to produce reflected-shock pressures
from approximately 0.4 to 5 atm. Typical test-section fill
pressures were on the order of 5-100 Torr and were monitored
with one of three MKS Baratron model PDR-C-1B pressure
transducers with ranges from 0-10, 0-100, and 0-1000 Torr.
All measurements, described below, were performed exclusively
behind reflected shock waves at a location 1.6 cm from the
endwall.
Species Measurements
Two species were monitored during the experiments: SiH4
and SiH2. For the SiH4 concentration measurements, an infrared-
emission technique was used. Emission from the well-known
ν3 Si-H vibrational stretch mode was monitored by a fast-
response (<1 µs), LN2-cooled InSb detector (Judson J10D) with
a 1-mm element. A filter centered at 4.7 ( 0.1 µm was used to
distinguish the SiH4 emission from background radiation. The
emitted light from the shock tube gases was focused onto the
detector through a CaF2 window and a 2-mm slit to improve
the time resolution of the measurement and minimize scattered
radiation from the shock-tube walls. A similar IR-emission

10990 J. Phys. Chem. A, Vol. 107, No. 50, 2003
Petersen and Crofton
Figure 1. Infrared emission trace for SiH
4
concentration showing
Figure 2. Typical normalized SiH
4
concentration time history and
prominent features.
comparison with silane pyrolysis model (Table 2).
measured intensity (in volts) is proportional to the species mole
fraction, assuming a simple harmonic oscillator model as
follows:
technique has been employed in other shock-tube measurements
of silane decompostion.5,7-9 These earlier studies have shown
that the possibility of interference emission from other SixHy
molecules is negligible over similar SiH4 concentration ranges.
Figure 1 gives a sample IR-emission measurement showing
key features. Arrival of the incident and reflected shocks is
evident in the emission trace. For the reflected-shock temper-
ature shown (1521 K), the silane quickly decomposes after
arrival of the reflected shock at time t ) 0. At longer times,
broadband interference emission from condensed SixHy particles
is evident. However, interference emission from the soot
particles (which do not form until most of the silane has
decomposed) was proven during the course of the experiments
not to affect the kinetics measurements that utilize the emission
data at earlier times.
XSiH4(t)
1
I′(t) ∝
(2)
[
]
eθv/T - 1)
RT
(
In eq 2, I′(t) is the instantaneous IR light intensity at time t, θv
is the characteristic vibrational temperature of the ν3 transition,
and XsiH4(t) is the mole fraction of silane at time t. Because eq
2
applies to any time t, and the temperature is constant to within
a few Kelvins over the shock-tube experiment for the highly
diluted mixtures utilized, one can normalize the I′(t) to the value
at some time t*:
Assurance that the soot emission did not interfere with the
determination of the decomposition rate was accomplished by
comparing rates obtained from the data traces in two ways: (1)
from the raw emission traces and (2) from the emission traces
with a correction for the soot emission. The soot correction was
performed by extrapolating the long-time emission trend (Figure
XSiH4(t)
I′(t)
∝
) constant XSiH4(t)
(3)
I′(t*)
XSiH4(t*)
In this fashion, the measured SiH -emission signals can be
compared directly with the chemical kinetics model to infer the
4
1
) back to time zero and subtracting it from the overall measured
reaction rate coefficient, discussed in more detail below and
displayed in Figure 2. Note that the time response of the
emission setup is only limited near time zero, during the passage
of the shock wave. The actual time response of the detector is
<1 µs otherwise.
signal level. From such calculations, the decomposition rates
obtained at the early times (typically < 100 µs) using both
methods were nearly identical within the stated accuracy of the
final rate herein.
Premature decomposition of the silane behind the incident
shock wave, prior to arrival of the reflected shock wave, was
not a problem for the temperature range of the present
experiments. For example, at the maximum reflected-shock
temperature and pressure of this study (1730 K, 5 atm), a 1000-
ppm mixture of SiH4 in argon would be exposed to incident-
shock conditions on the order of 900 K and 1 atm. As shown
in Figure 1, the typical duration of the mixture at the incident-
shock conditions prior to reflected-shock arrival is about 50 µs.
Kinetics calculations at these conditions and for this time
duration indicate that the decomposition of SiH4 is imperceptible
The other species measured was the key intermediate radical,
SiH . A conventional shock-tube laser-absorption setup based
2
on differential absorption and common-mode rejection, as in
22
Dean and Hanson, was employed in experiments separate from
the SiH emission measurements. A key SiH transition near
4
2
579 nm has already been identified in a similar laser-absorption
2
3
experiment by Markus and Roth and was duplicated herein.
r
The transition of interest was the Q (4) line in the A˜ - X˜
0,4
-
1
(0,2,0)-(0,0,0) band at 17 260.82 cm .
Figure 3 shows the basic laser absorption setup. A Coherent
699-21 ring-dye laser operating on Rhodamine 590 dye,
pumped by all lines of a Coherent Innova I-20 argon-ion laser,
was the tunable light source. A sample of the beam was sent to
a Burleigh WA-4500 wavemeter to monitor the wavelength with
(,0.1%). (In fact, it takes over 200 µs for 0.1% of the silane
to decompose at the extreme conditions of this example.)
To minimize effects near time zero immediately after passage
of the reflected shock wave, the emission signals were normal-
ized to the measured light intensity at a later time t*. This
normalization procedure, defined in Figure 2 for a typical silane
decomposition measurement, is based on the fact that the
-
1
a resolution of 0.005 cm , and the main beam was focused
through the CaF2 windows in the shock tube to a diameter on
the order of 1 mm. The incident laser intensity, I0, and the
transmitted intensity, I, were monitored by Si photodiodes with

Measurements of High-Temperature Silane Pyrolysis
J. Phys. Chem. A, Vol. 107, No. 50, 2003 10991
TABLE 2: Reaction Mechanism Employed for Silane
Pyrolysis Calculationsa
rate coefficient
no.
reaction
+ Ar ) SiH + H
+ Ar ) Si + H + Ar
A
E
a
ref
1
1
1
1
1
1
1
1
1
1
1
1
1
1
5
3
4
5
0
4
3
4
4
4
5
4
4
2
1
2
3
4
5
6
7
8
9
1
1
1
1
1
a
SiH
SiH
4
2
2
+ Ar
7.2 × 10
9.1 × 10
7.8 × 10
5.5 × 10
5.2 × 10
2.2 × 10
1.3 × 10
6.5 × 10
4.8 × 10
1.6 × 10
1.5 × 10
4.0 × 10
4.8 × 10
7.9 × 10
45.1 this study
30.2 26
50.9 26
52.0 26
33.5 26
95.9 26
0.0 26
0.0 26
23.6 26
0.0 26
2
2
Si
3
Si
3
Si
2
H
H
H
8
8
6
) SiH
) H SiSiH + SiH
) SiH + SiH
2
+ Si
2 6
H
3
4
4
2
H
2
+ Ar ) H + H + Ar
SiH
SiH
4
+ SiH
+ SiH
2
) H
) Si
3
SiSiH + H
+ H
+ H
+ H
2
2
2
2
H
2
2
SiH + H
2
) SiH
) Si
) SiH + H
Si + SiH ) Si
Si + Si ) Si
SiSiH ) H SiSiH
2
0
1
2
3
4
SiH + SiH
Si + H
4
2
H
4
2
31.8 26
0.0 26
0.0 26
H
2
+ H
2
2
Figure 3. Laser absorption setup for the measurement of SiH
2
4
2
H
6
3
H
+ 2H
2
concentration.
2
H
3
2
2
10.5 26
a
3
k
i
a a
) A exp(-E /RT); cm , mol, s Units; E is in kcal/mol.
I0 - I
)
1 - exp(-k PXSiH2L)
(4)
ν
I0
Describing the line shape of the transition by a Voigt profile,
and assuming the frequency is at line center, it can be shown
that the absorption cross section at line center as a function of
0
T and P, σ (T,P), is given by
0
σ (T,P)
)
0
σ (T ,P )
0
0
0
Q (T )
T 3/2
Φ (T,P)
hc 1
1
T0
vib
0
exp -F(J′′)
-
(5)
[
k
(
T
)
]
(
T
)
0
Q (T)
vib
0
Φ (T ,P )
0
0
where T0 and P0 are the reference conditions defined as 1757
Figure 4. Typical SiH
with the silane pyrolysis model (Table 2).
2
mole fraction time history and comparison
K and 0.3 bar, respectively, and the reference cross section is
0
-17
2
23
σ (T0,P0) ) 1.3 × 10
cm per Markus and Roth. In eq 5,
-
1
a 1-MHz bandwidth (New Focus model 2031). A differential
preamplifier (SRS SR560) was used to take the differential
absorption signal, I0 - I, and this difference signal along with
I0 were sent to separate channels on a computer-based oscil-
loscope consisting of two CS512 boards from Gage Applied
Sciences (4 channels total).
A typical SiH2 absorption signal is displayed in Figure 4. In
all experiments over the temperature and concentration ranges
investigated herein, the SiH2 formed immediately after passage
of the reflected shock wave at t ) 0. At time zero, passage of
the requisite density gradient of the reflected shock wave through
the laser beam caused temporary beam steering and, hence, some
extinction of the laser light; this phenomenon is identified as
the Schlieren spike in Figure 4 and lasted approximately 3-5
µs in most experiments. Signal-to-noise ratios between 5 and
F(J′′) is the term energy calculated to be 108.61 cm (ref 23),
h is Planck’s constant, c is the speed of light, k is Boltzmann’s
constant, and Qvib is the vibration partition function of a
harmonic oscillator with vibrational constants ωi taken from
2
4
-1
-1
Colvin et al. (ω1 ) 2032 cm , ω2 ) 1004 cm , ω3 ) 2022
-
1
cm ).
0
The Voigt profiles at line center, Φ , are defined in the usual
manner as functions of the Doppler line width, ∆νD, the collision
line width, ∆νc, and the Voigt function, V. The Voigt function
at line center, V(0,a), was taken from the procedure defined by
Pierluissi and Vanderwood. Finally, the absorption coefficient
from eq 4 is related to the collision cross section via
25
0
0
σ
) 7.337 × 10 σ atm-1 cm-1
21
kν )
(6)
RT
T
1
0 were achieved in the absorption experiments, corresponding
Further details on the spectroscopic model can be found in
Markus and Roth.
to a resolution of approximately 0.5-1.0% absorption or about
2
3
1
4
-5 ppm, depending on the conditions (about (3 ppm in Figure
). The peak SiH2 concentration in Figure 4 corresponds to
Chemical Kinetics
approximately 5% absorption.
A spectroscopic model of the SiH2 absorption, based primarily
on the model given in Markus and Roth,23 was developed to
convert the absorption signals to mole fractions. Applying Beer’s
law to the shock-tube experiment, the differential absorption
measurement is related to the absorption coefficient at frequency
ν (cm ), the total pressure behind the reflected shock, P, the
SiH2 mole fraction, XsiH2, and the optical path length, L (16.2
cm):
When measuring the rate coefficient of the silane decomposi-
tion, reaction 1a, modern chemical kinetics modeling techniques
can be employed along with the latest rate coefficients to include
the effect of competing reactions. This approach was employed
in the present study using the reactions and rate coefficients
for silane pyrolysis summarized in Woiki et al. and listed in
Table 2. All chemical kinetics calculations were performed using
the constant-pressure option in the shock-wave subroutine of
-
1
26

10992 J. Phys. Chem. A, Vol. 107, No. 50, 2003
Petersen and Crofton
Figure 5. Sensitivity of SiH
reactions in the silane pyrolysis mechanism. Conditions correspond to
the SiH measurement in Figure 2.
4
concentration to the most significant
Figure 6. Sensitivity of SiH
2
concentration to the most significant
reactions in the silane pyrolysis mechanism. Conditions correspond to
the SiH measurement in Figure 4.
4
2
the CHEMKIN software.27 Likewise, all sensitivity analyses
were performed using the SENKIN software.27
Similarly, a sensitivity analysis of the effect of the reactions
in Table 2 on the measured SiH2 concentrations was performed.
Figure 6 presents the dominant sensitivities corresponding to
the conditions of the SiH2 data trace of Figure 4. Although k1a
is again the dominant reaction rate influencing the production
of SiH2 for the present conditions, reaction 8 is almost as
influential as a source of SiH2 removal. The sensitivity of SiH2
to reaction 7 is within about 20% of the sensitivity to R1a, and
the silylene decomposition (R2) reaction rate also has a
sensitivity within 10%. Of course, the results of the sensitivity
analysis depend on the reaction mechanism defined in Table 2.
Although the entire mechanism in Table 2 was used in the
data analysis presented herein for completeness, only a few of
the 14 reactions are actually important over the temperatures
and concentrations of interest. Of primary importance is the
sensitivity of the measured species to reaction rate coefficients
of each of the competing reactions. Figure 5 shows the
sensitivity of SiH4 to the rate coefficients of each reaction (where
only the dominant reactions are shown). The conditions of the
sensitivities shown in Figure 5 were calculated for the conditions
corresponding to the data trace provided in Figure 2. The
normalized sensitivities are defined as (1/XSiH ,max)(∂XSiH /∂ki),
Because the reaction rate coefficients of the competing
reactions R2, R7, and R8 (Figure 6) are not well-known,26 the
use of the SiH2 concentration measurements to discern the rate
of the main silane decomposition reaction was considered
secondary to the use of the SiH4 measurements. As shown
below, the SiH2 concentration measurements were nonetheless
useful in verifying the value of k1a obtained herein, although
the resolution of the signals was not adequate enough to
simultaneously determine the rate of R8 (given the rate of R1a
from the SiH4 measurements). This issue is an ongoing one in
the authors’ laboratory, wherein an improved SiH2 diagnostic
based on FM absorption spectroscopy has been set up.
4
4
where ki is the reaction rate coefficient of the ith reaction.
From Figure 5, the SiH4 concentration is most sensitive to
the rate of the reaction of interest, i.e.
SiH + Ar ) SiH + H + Ar
(R1a)
4
2
2
Although the silane concentration is about six times more
sensitive to R1a, the closest competing reaction is the SiH4-
removal reaction 7 involving reactions between SiH4 and SiH2;
this reaction competes for SiH2 radicals with the SiH2 recom-
bination reaction 8
SiH + SiH ) H SiSiH + H
(R7)
(R8)
4
2
3
2
Results and Discussion
SiH + SiH ) Si H + H
2
A total of 40 silane concentration profiles were obtained over
a range of temperatures from 1063 to 1734 K and pressures
2
2
2
2
The dominance of reaction R1a in the silane-concentration
sensitivity supports the use of SiH4 concentration measurements
to discern the reaction rate coefficient of R1a. However,
inclusion of the best-known rates for the closest competing
reactions still improves the accuracy of the measurement to some
extent. The additional reactions tend to accelerate the decom-
position of SiH4, so if they were ignored, the apparent rate of
R1a would be mistakenly higher.
from 0.6 to 5.0 atm. Eleven experiments using the SiH laser-
absorption diagnostic were performed for temperatures between
1240 and 1643 K and pressures from 1.2 to 2.0 atm. Using the
mechanism of Table 2 and CHEMKIN, the reaction rate of R1a
2
was altered to fit each SiH profile for each test; the resulting
4
best-fit k1a was then defined as the measured k1a for that test.
Figure 2 shows a typical comparison between the best-fit kinetics
prediction and the actual SiH concentration data. The agreement
4
The mechanism herein assumes that the SiH2 molecules are
consumed by the decomposition reaction
shown is visibly good and representative of the entire data set.
A linear regression of the data on an Arrhenius plot is
displayed in Figure 7. Each symbol in Figure 7 corresponds to
a different initial SiH4 concentration in argon. The resulting rate
coefficient in the form of eq 1 is
SiH + Ar ) Si + H + Ar
(R2)
2
2
In recent experiments yet to be published, the authors have found
that R2 is particularly important as well, although still secondary
in its impact on SiH4 time histories compared to R1a.
1
5
3
k ) 7.2 × 10 exp(-45.1/RT) cm /mol s
(7)
1a

Measurements of High-Temperature Silane Pyrolysis
J. Phys. Chem. A, Vol. 107, No. 50, 2003 10993
Figure 7. Arrhenius plot of the present data for the bimolecular
Figure 8. Effects of uncertainties in k1a and k
measurements of SiH concentration.
8
on the present
decomposition reaction rate coefficient k1a
.
4
2
The r of this equation is 0.98, and the uncertainty in the
activation energy is (1.2 kcal/mol. Referring to Figure 1, the
present value of Ea1 is near the middle of the span seen from
the literature.
Equation 8 provides a rate that agrees well with those of
4,5.7,9
earlier shock-tube experiments,
within about 30% or better.
Conversely, the current measurement of k1a is higher than but
1
0
within a factor of 2 of the measurement by Mick et al., which
has an activation energy of 48.0 kcal/mol. Their experiments
were performed in highly diluted mixtures (≈1 ppm SiH4) and
employed measurements of Si and H atoms using absorption
spectroscopy. They used a reaction set including only R1 and
R2 to fit their data and infer rates for each.
Possible errors and their impact on the measured k1a were
investigated in an attempt to justify the current Ea1 in light of
the wide range of values seen in previous studies (Table 1).
For example, interference from soot emission at longer times
would lead to an artificially larger SiH4 signal at later times,
resulting in a lower decomposition rate than in eq 7. Impurities
in the shock-tube test section such as oxygen and/or residual
particles tend to speed up the decomposition of silane. Hence,
the effects of impurities would be higher for the lower-
concentration SiH4 mixtures, hence making these more-dilute
mixtures show an artificially faster decomposition rate. How-
ever, this effect is also not seen in the data of Figure 7, where
the 300-ppm SiH4 mixtures produced decomposition rates
similar to those of the 1000-ppm mixtures within the accuracy
of the experiment.
Figure 9. Effects of uncertainties in k1a and k8 on the present
measurements of SiH concentration.
2
agreement between the present SiH2 results and the Table 2
model may be because some of the rates in Table 2 were
obtained using the particle-tainted technique. The results of these
later experiments are forthcoming in a future publication by
the authors.
Although shown above to be of secondary importance, the
impact of competing reactions and their lesser-known ki can
likewise be estimated. Figure 8 shows in comparison with the
measured SiH4 time history the model prediction with the rate
coefficient of reaction 8 being higher and lower than the value
in Table 2 by a factor of 2. The impact of such an uncertainty
in k8 is not significant. Similarly, the impact of k8 on the
prediction of a typical SiH2 time history is shown in Figure 9.
As expected, because the SiH2 concentration is particularly
sensitive to the rate of reaction 8 (Figure 6), uncertainty in its
reaction rate can have a larger predicted impact on the SiH2
profiles than on the SiH4 profiles (although still within the
uncertainty of the SiH2 measurement).
The present silane-decomposition kinetics model agrees well
with the measured SiH2 profiles. A typical comparison is shown
in Figure 4, where the agreement between the current silane
decomposition rate (and Table 2 model) is well within the
resolution of the SiH2 measurement. This result is typical of
most of the measured SiH2 profiles, so further refinement of
either the kinetics model or the SiH2 spectroscopic model was
not performed. The impact of differences in the measured k1a
by as much as a factor of 2 is shown in a representative SiH4
profile (Figure 8) and a representative SiH2 profile (Figure 9).
Unfortunately, the laser absorption measurements herein and
23
in Markus and Roth are not immune to the presence of particle
formation, which tends to increase the level of laser extinction,
making the level of SiH2 appear artificially higher. Recent
experiments by the authors using a SiH2 diagnostic that is
insensitive to particle formation show results that are somewhat
different from those herein for the SiH2 time histories. The fair
However, as mentioned above, the authors have noticed that
the value inferred for k1a from the present data depends on the
reaction mechanism employed. Nonetheless, even omissions of
and drastic changes to the rates in Table 2 (by more than ×10)
led to k1a correlations within about 50% of the final value

10994 J. Phys. Chem. A, Vol. 107, No. 50, 2003
Petersen and Crofton
extended in this paper. This robustness is testimony to the
dominance of R1a seen in Figures 5 and 8.
Of particular interest is the fact that the bimolecular form of
the silane decomposition reaction, R1a, can represent all of the
data. This result is significant in that, for the range of total
pressures up to at least 5 atm, the silane decomposition rate
over the temperature range of this study (≈1000-1800 K) is
in the low-pressure limit and defined by the 2nd-order reaction
R1a. Therefore, over the range of most practical applications
of silane combustion, the rate coefficient of the primary
decomposition step is not pressure dependent. One check on
this result is to correlate the data for only the lower-pressure
points. When the authors excluded all data above 2 atm, the
same rate coefficient expression as eq 7, within 0.2 kcal/mol
and with the same preexponential factor, was obtained.
Quite often in the literature, however, silane decomposition
is expressed in its unimolecular form
Figure 10. Arrhenius plot of the measured unimolecular decomposition
rate coefficient k1b at various pressures.
SiH ) SiH + H
(R1b)
4
2
2
When written in this fashion, k1b must be pressure dependent
TABLE 3: Rate Coefficient Constants for k1b
)
19
for all total pressures lower than the high-pressure limit. When
expressed in the form of R1b, the authors have found that the
bimolecular and unimolecular rate coefficients of SiH4 decom-
position are related by the total concentration over the range of
the present study, i.e, k1b ) k1a[M] (with M ) Ar). This result
implies that the pressure dependence of k1b is linear for the
present study, which agrees with the simple relationship between
k1a and k1b derived by ignoring competing reactions
A exp(-E /RT) (s-1) at Various Pressure Groupings
a
a
P
A
E
a
r2
1
1
0
0
0
.6
.0
.0
.5
1.1 × 10
3.0 × 10
2.2 × 10
5.0 × 10
40.7
42.8
40.3
41.0
0.983
0.991
0.987
0.943
1
2
4
10
10
a
a
E is in kcal/mol, and P is in atm.
d[SiH4]
)
-k [SiH ][M] ) -k [SiH ]
(8)
1a
4
1b
4
dt
Other investigators have also concluded that k1b is linearly
15
related to the pressure, namely Han et al. and Meyerson et
2
8
15
al. As discussed in Han et al., the rate coefficient of R1b
2
0
was postulated by Coltrin et al. to follow the form
â
k ) (R + bP)A T exp(-E /RT)
(9)
1
b
1b
1b
However, Coltrin et al.20 found a value for R close to zero (i.e.,
0
.0504) and a value for b close to one (i.e., 0.9496) using an
3
,4
RRKM analysis of the experimental values of Newman et al.
for pressures from 0.006 to 1.0 atm. As a result, eq 9 predicts
k1b to be essentially linearly related to the pressure, as observed
in the present experiments over a range of pressures from 0.6
7
to 5.0 atm. Tanaka et al. were also able to correlate their shock-
Figure 11. Unimolecular falloff plot for the k1b data (Table 3) in
comparison with a second-order rate coefficient, k1a[M].
tube data by a single rate expression for k1a (for pressures
1
/2
between 0.25 and 1.3 atm), although they observed a P
pressure dependence for k1b.
for each temperature as lines. The good agreement between the
k1b results and the k1a[M] approximation (i.e., the bimolecular
result) is further confirmation that the present results are all
within the low-pressure limit of silane decomposition and not
within the falloff region, even at total pressures as high as 5
atm.
An Arrhenius plot of k1b obtained from the data is given in
Figure 10. The data fall within four, average total-pressure
groupings: 0.6, 1.0, 2.0, and 4.5 atm. The rate coefficients k1b
in the form of eq 1 for each pressure group are summarized in
Table 3 and plotted as solid lines in Figure 10. When compared
with the result for k1a in eq 7, the activation energies for k1b in
Table 3 are about 4 kcal/mol lower and closer to 40 kcal/mol.
Another useful plot is that of k1b versus pressure to discern
if the reaction rate is within the falloff region between the low-
and high-pressure limits from unimolecular decomposition
theory. Using the curves from Table 3, the results of the present
experiments are plotted versus pressure in Figure 11 for 4
different temperatures: 1100, 1300, 1500, and 1700 K. The
Table 3 results (representing the measured data) are displayed
as points, whereas the estimate that k1b ) k1a[M] is displayed
Finally, the present results for k1a are compared with the
results of previous investigators in Figure 12. When the literature
results were given in terms of k1b, they were plotted in Figure
12 in terms of k1a by dividing by [M] for that set of experiments.
When plotted over the same temperature range, eq 7 is in good
7
agreement with the shock-tube results of Tanaka et al., Koshi
9
10
et al., and Mick et al. Previous investigators have pointed
4
out that the shock-tube data of Newman et al. are an order of
magnitude higher than the other, more-recent shock-tube data.
However, this misconception is because their measurements

Measurements of High-Temperature Silane Pyrolysis
J. Phys. Chem. A, Vol. 107, No. 50, 2003 10995
described adequately by the bimolecular form, implying that
the reaction is still within the low-pressure limit at total pressures
as high as 5 atm.
Acknowledgment. This work was supported by The Aero-
space Corporation and The Air Force Space and Missile Systems
Center under Contract No. F04701-00-C-0009. The assistance
of Brian Brady in obtaining references and the invaluable help
of Carrol Gardner (Aerospace) in the laboratory are recognized.
References and Notes
(
1) Hogness, T. R.; Wilson, T. L.; Johnson, W. C. J. Am. Chem. Soc.
936, 58, 108-112.
2) Purnell, J. H.; Walsh, R., Proc. R. Soc. London, Ser. A 1966, 293,
1
5
1
(
43-561.
(3) Newman, C. G.; Ring, M. A.; O’Neal, H. E. J. Am. Chem. Soc.
978, 100, 5945-5946.
(4) Newman, C. G.; O’Neal, H. E.; Ring, M. A.; Leska, F.; Shipley,
N. Int. J. Chem. Kinet. 1979, 11, 1167-1182.
(
5) Tanaka, H.; Ishigaki, A.; Asaba, T. Shock Tubes and WaVes; Sydney
Figure 12. Comparison between the present results for k1a and
measurements from the literature.
Shock Tube Symposium Publishers: Sydney, 1983; pp 672-683.
(6) Votintsev, V. N.; Zaslonko, I. S.; Mikheev, V. S.; Smirnov, V. N.
Kinet. Catal. 1987, 27, 843-846.
(7) Tanaka, H.; Koshi, M.; Matsui, H. Bull. Chem. Soc. Jpn. 1987, 60,
were at a relatively high pressure (5.3 atm, Table 1) and were
given in terms of k1b. When the Newman et al. value is put in
terms of k1a in Figure 12 by dividing by the [M] of their study,
there is good agreement with the rest of the shock-tube results.
When extrapolated to temperatures between 2000 and 4000
K, the present rate coefficient expression agrees well with the
3
519-3523.
(8) Koshi, M.; Yamauchi, M.; Koseki, K.; Matsui, H. Shock Tubes
and WaVes; VCH: Weinheim, Germany, 1988; pp 443-449.
9) Koshi, M.; Kato, S.; Matsui, H. J. Phys. Chem. 1991, 95, 1223-
227.
10) Mick, H. J.; Smirnov, V. N.; Roth, P. Ber. Bunsen-Ges. Phys. Chem.
(
1
(
1993, 97, 793-798.
(11) Mick, H.-J.; Roth, P.; Smirnov, V. N.; Zaslonko, I. S. Kinet. Catal.
994, 35, 439-451.
12) Roenigk, K. F.; Jensen, K. F.; Carr, R. W. J. Phys. Chem. 1987,
91, 5732-5739.
13) Moffat, H. K.; Jensen, K. F.; Carr, R. W. J. Phys. Chem. 1991, 95,
5
shock-tube results of Tanaka et al. At lower temperatures, the
1
present correlation agrees more with the reaction-vessel data
(
2
15
of Purnell and Walsh than with the results of Han et al. and
the laser-induced chemical vapor deposition experiments of
Meunier et al.14 The poor agreement of the latter studies is
probably due to the indirect natures in which the rate of silane
decomposition was inferred.
(
145-154.
(
14) Meunier, M.; Flint, J. H.; Haggerty, J. S.; Adler, D. J. Appl. Phys.
1
987, 62, 2812-2821.
(15) Han, J. H.; Rhee, S.-W.; Moon, S. H. J. Electrochem. Soc. 1996,
43, 1996-2002.
1
(
16) Onischuk, A. A.; Strunin, V. P.; Ushakova, M. A.; Panfilov, V. N.
Summary
Int. J. Chem. Kinet. 1998, 30, 99-110.
(
17) Neudorff, P.; Jodhan, A.; Strausz, O. P. J. Phys. Chem. 1980, 84,
Silane decomposition at elevated temperatures is important
in many applications such as the deposition of silicon-containing
compounds, the manufacture of various glasses, and aerospace
propulsion. To determine the temperature and pressure depen-
dence of the primary silane decomposition reaction, experiments
were performed behind reflected shock waves over a range of
initial silane concentrations in argon between 300 and 1000 ppm,
temperatures between 1060 and 1730 K, and pressures from
3
38-339.
(
18) Purnell, J. H.; Walsh, R. Chem. Phys. Lett. 1984, 110, 330-334.
(19) Troe, J. J. Phys. Chem. 1979, 83, 114-126.
(
20) Coltrin, M. E.; Kee, R. J.; Miller, J. A. J. Electrochem. Soc. 1986,
1
33, 1206-1213.
21) Kee, R. J.; Rupley, F. M.; Miller, J. A. SAND87-8215B, Sandia
National Lab.: Albuquerque, NM, 1990.
22) Dean, A. J.; Hanson, R. K. J. Quantum Spectrosc. Radiat. Transfer
989, 42, 375-384.
23) Markus, M. W.; Roth, P. J. Quantum Spectrosc. Radiat. Transfer
(
(
1
1
(
0.6 to 5.0 atm. Reaction progress was monitored by measuring
994, 52, 783-789.
(24) Colvin, M. E.; Grev, R. S.; Schaefer, H. F.; Bicerano, J. Chem.
the concentration of SiH4 as a function of time via the emission
from the ν3 vibrational stretch mode near 4.7 µm and the
concentration of SiH2 using laser absorption of a rotational line
in the A˜ - X˜ transition at a wavelength near 579 nm. By
considering the kinetics of simultaneous reactions, the rate
coefficient of SiH4 + M ) SiH2 + H2 + M (M ) Ar) was
Phys. Lett. 1983, 99, 399-405.
(
25) Pierluissi, J. H.; Vanderwood, P. C. J. Quantum Spectrosc. Radiat.
Transfer 1977, 18, 555-558.
(26) Woiki, D.; Catoire, L.; Roth, P. Ceram. Process. 1997, 43, 2670-
2
678.
(
27) Kee, R. J.; Rupley, F. M.; Miller, J. A.; Coltrin, M. E.; Grcar, J.
F.; Meeks, E.; Moffat, H. K.; Lutz, A. E.; Dixon-Lewis, G.; Smooke, M.
D.; Warnatz, J.; Evans, G. H.; Larson, R. S.; Mitchell, R. E.; Petzold, L.
R.; Reynolds, W. C.; Caracotsios, M.; Stewart, W. E.; Glarborg, P.; Wang,
C.; Adigun, O. Chemkin Collection, release 3.6; Reaction Design, Inc.: San
Diego, CA, 2000.
28) Meyerson, B. S.; Ganio, E.; Smith, D. A.; Nguyen, T. N. J.
Electrochem. Soc. 1986, 133, 1232.
15
3
determined to be k1a ) 7.2 × 10 exp(-45.1/RT) cm /mol s.
This result compares favorably with established measurements
in the literature, even at temperatures above and below those
of the present experiments. Over the range of pressures studied,
it was found that the rate of silane decomposition can be
(