Time-resolved cavity ring-down spectroscopic study of the gas phase and surface loss rates of Si and SiH3 plasma radicals

Article abstract of DOI:10.1016/S0009-2614(02)00802-3

Time-resolved cavity ring-down spectroscopy (CRDS) has been applied to determine gas phase and surface loss rates of Si and SiH₃ radicals during plasma deposition of hydrogenated amorphous silicon. This has been done by monitoring the temporal decay of the radicals densities as initiated by a minor periodic modulation applied to a remote SiH₄ plasma. From pressure dependence, it is shown that Si is reactive with SiH₄ [(1.4 ± 0.6) × 10^-16 m^-3 s^-1 reaction rate constant], while SiH₃ is unreactive in the gas phase. A surface reaction probability β of 0.9 < β ≤ 1 and β = 0.30 ± 0.05 has been obtained for Si and SiH₃, respectively.

Full text of DOI:10.1016/S0009-2614(02)00802-3

3
July 2002
Chemical Physics Letters 360 (2002) 189–193
www.elsevier.com/locate/cplett
Time-resolved cavity ring-down spectroscopic study of the
gas phase and surface loss rates of Si and SiH plasma radicals
3
J.P.M. Hoefnagels, A.A.E. Stevens, M.G.H. Boogaarts, W.M.M. Kessels,
M.C.M. van de Sanden *
Department of Applied Physics, Center for Plasma Physics and Radiation Technology, Eindhoven University of Technology,
P.O. Box 513, 5600 MB Eindhoven, The Netherlands
Received 22 March 2002; in final form 13 May 2002
Abstract
Time-resolved cavity ring-down spectroscopy (CRDS) has been applied to determine gas phase and surface loss rates
of Si and SiH
toring the temporal decay of the radicals densities as initiated by a minor periodic modulation applied to a remote SiH
3
radicals during plasma deposition of hydrogenated amorphous silicon. This has been done by moni-
4
À16
À3 À1
plasma. From pressure dependence, it is shown that Si is reactive with SiH
constant], while SiH
has been obtained for Si and SiH
4
½ð1:4 Æ0:6ÞÂ10
m
s reaction rate
3
is unreactive in the gas phase. A surface reaction probability b of 0:9 < b 6 1 and b ¼0:30 Æ0:05
3
, respectively. Ó 2002 Elsevier Science B.V. All rights reserved.
For basic understanding and modeling of plas-
ma deposition processes, information on the den-
sity as well as the surface reactivity of the plasma
species is essential. The surface reaction probabil-
ity b of the species has often been obtained indi-
rectly, e.g., from measurements of a macroscopic
compound value of b [1], or under process condi-
tions different from the actual plasma deposition
process, e.g., by measurements in an afterglow
plasma [2–4] or from a molecular beam scattering
experiment [5]. In this Letter, we will present a
technique to obtain the gas phase and surface loss
rates of radicals during plasma deposition, here
particularly in the case of high rate deposition of
hydrogenated amorphous silicon [6]. In this tech-
nique, the highly sensitive cavity ring-down spec-
troscopy (CRDS) method [7] is used to map the
temporal decay of an increased radical density by
means of time-resolved absorption measurements.
This so-called ‘additional radical density’ is cre-
ated in a continuously operated remote silane
plasma by applying a minor periodic modulation
to the applied plasma power. Although time-re-
solved CRDS has been employed previously to
obtain gas phase loss rates of radicals [8], in this
work we apply the technique also to determine
surface loss rates of the radicals. This yields si-
multaneous information on the surface reaction
probability b and the density of the radicals under
the specific plasma conditions. Using this method,
it is shown that Si is mainly lost in the gas phase to
*
Corresponding author. Fax: +31-40-2456442.
E-mail address: m.c.m.v.d.sanden@tue.nl (M.C.M. van de
Sanden).
0
009-2614/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved.
PII: S0009-2614(02)00802-3

190
J.P.M. Hoefnagels et al. / Chemical Physics Letters 360 (2002) 189–193
SiH
reactions at the surface.
4
, whereas SiH
3
is only lost via diffusion to and
the radicals generated by the rf pulse is obtained
from the difference in absorption at some point Dt
in the rf afterglow and at a point long after the
influence of the rf pulse has extinguished (see Fig.
1b). Every CRDS trace is handled separately by
means of a new ‘state-of-the-art’ 100 MHz, 12 bit
In the expanding thermal plasma technique
(
is created in which SiH precursor gas is injected
ETP) (Fig. 1a) a remote expanding Ar–H
2
plasma
4
downstream. To detect (low-density) radicals such
as SiH and Si in this highly-diluted SiH plasma
the CRDS technique has previously been em-
data acquisition system [13] and an averaged A is
3
4
rf
obtained as a function of the time Dt in the after-
glow of the rf pulse.
2
0
~
ployed; SiH
2
X A
3
has been identified at the A A
broadband transition ranging from ꢀ200 to
1
~
1
A typical s-CRDS measurement for Si and SiH
is shown in Fig. 2. A duty cycle of 2.5% has been
3
ꢀ
260 nm [9], whereas Si radicals have been probed
3
23
at the 4s P0;1;2
3p P0;1;2 transition around 251
nm [10]. This has yielded typical densities of SiH
used in order to obtain a good signal-to-noise ratio
3
3
in the additional Si and SiH absorption, while
18
À3
15
À3
and Si of ꢀ5 Â10
m
[11] and ꢀ10
m [12],
possible powder formation due to the ‘anion con-
fining’ rf plasma sheath is suppressed [14]. Fig. 2
shows that both Si and SiH decrease single ex-
respectively. In the time-resolved CRDS (s-CRDS)
measurements, a minor periodic modulation of the
radical densities is produced by applying 5 Hz,
3
ponentially, which is expected from the radicals’
mass balance when the loss term is linear in the
radical density [15]. The corresponding loss rate
2
.5% duty cycle rf pulses to the substrate in addi-
tion to the continuously operating ETP. The
power in the rf pulse is 63 W and the bias voltage is
between )40 and )120 V with a fall off time of less
then 20 ls. The additional absorption Arf due to
À1
s
hand and the loss due to diffusion to and reactions
depends linearly on the gas phase loss on one
at the surface on the other hand [15]:
D
À1
s
¼k n þ
:
ð1Þ
r
x
2
K
In this equation, k is the gas phase reaction rate
r
x
with species x with density n , D is the diffusion
Fig. 2. A typical measurement during the rf afterglow of the
additional absorption Arf of Si and SiH generated by the rf
3
Fig. 1. (a) The ETP setup equipped with the cavity ring-down
spectroscopy system, which is aligned at 5 mm in front of the
substrate. A pulsed rf bias can be applied to the substrate
holder. (b) A timing scheme illustrating the modulation of the
radical density and the synchronization of the CRDS laser
pulses.
pulse during the continuously operating ETP plasma. Every
data point is an average of 128 CRDS traces and the statistical
error in the decay time obtained from least-squares analysis of
the Si and SiH
at Dt < 0Þ, the densities of Si and SiH
value of 2 Â10 and 3 Â10 , respectively.
3
decay curves is less then 5%. In the rf pulse (i.e.,
3
have reached a saturated
15
18
À3
m

J.P.M. Hoefnagels et al. / Chemical Physics Letters 360 (2002) 189–193
191
3
reveals a reaction rate constant of SiðP) with SiH
4
À16
À3 À1
of k m s . In the error
r
¼ð1:4 Æ0:6ÞÂ10
analysis of this reaction rate the small statistical
error in the loss rates as well as the following
systematic uncertainties have been considered: The
variation in the diffusion term in Eq. (1) can be
neglected for the small SiH dilutions used in Fig.
4
3
, while the local SiH density correction affects k
4
r
r
only slightly. The uncertainty in k is mainly de-
termined by the gas temperature, which has been
estimated to be (700 Æ200) K from plasma model
calculations [11]. The value of k
r
is somewhat
À16
Fig. 3. The Si and SiH
3
loss rate as a function of the SiH
4
lower than the value of k
r
¼ð3:5 Æ1:0ÞÂ10
density at constant total pressure of 0.27 mbar. The error bars
represent the statistical error in the decay time as shown in
Fig. 2.
À3 À1
À16 À3 À1
m s
and k
r
¼ð4:4 Æ1:0ÞÂ10
m
s
as
obtained, respectively, by Tanaka et al. [3] at ꢀ350
K in an rf plasma and Takahara et al. [17] at 295 K
in a photolysis experiment. The value of Tanaka et
al. is in relatively good agreement with our value
considering the assumptions made (in both cases)
on, e.g., the gas temperature and thermalization of
Si, while the unaccounted presence of disilane in
continuous rf plasmas could also easily account
for the discrepancy. The comparison with the va-
lue of Takahara et al. which has been determined
in a somewhat indirect way using a pre-assumed
chemistry model, might not be so straightforward
since it was obtained under a ꢀ50 times higher He
background pressure.
coefficient for the specific radical in the
2 4
Ar–H –SiH mixture [16], and K is the effective
diffusion length of the radical. The latter depends
on diffusion geometry and on the radical’s surface
reaction probability b [15].
As revealed by Fig. 2, Si is lost typically a factor
3 2 4
of 10 faster than SiH in the Ar–H –SiH plasma.
As will be shown below, this is mainly due to the
difference in the gas phase loss processes that need
to be considered first before surface loss rates of
the radicals can be deduced. For Si and SiH in the
3
ETP plasma the only candidate for a significant
gas phase loss is SiH [16]. To investigate this re-
In order to deduce the surface reaction proba-
4
3
bility b of Si and SiH from the diffusion term in
action channel, the loss rate of Si and SiH
been obtained as a function of the SiH density
Fig. 3) keeping the total pressure and thus the
diffusion term in Eq. (1) nearly constant. The SiH
density has been calculated from the SiH partial
3
has
Eq. (1), a semi-empirical expression for the cylin-
der symmetrical geometry has been assumed to
determine the effective diffusion length K as has
been proposed by Chantry [15]. However, the
diffusion geometry in the ETP reactor, although
cylinder symmetric, is not as well defined as for
example in an rf parallel plate reactor. In our ap-
proach we have assumed that the diffusion of the
radicals in radial direction can be neglected in re-
spect to the diffusion in axial direction. Although
this assumption introduces the main uncertainty in
b, larger than the uncertainty due to the radical’s
temperature, the assumption is validated by the
fact that within the 5 ms duration of the rf pulse
the additional radicals do not have sufficient time
to reach the outer reactor wall, while the addi-
tional density saturates due to axial diffusion to the
substrate (see Fig. 2). Moreover, on the basis of
4
(
4
4
pressure using a kinetic gas temperature of 700 K
and including a correction for the local SiH4
1
consumption . The loss rate of SiH in Fig. 3 is
independent of SiH
gas phase loss of SiH
3
4
density, which indicates no
, while the loss rate of Si
3
4
increases linear with the SiH density. The slope
1
From the mass balance it follows that the production term
of a radical (proportional to the local SiH4 density and the
electron density) in the rf pulse under steady state conditions is
equal to the radical density over its loss rate (i.e., n=sÞ. The
electron density remains fairly constant for the conditions used
yielding therefore information on the local SiH4 density.

192
J.P.M. Hoefnagels et al. / Chemical Physics Letters 360 (2002) 189–193
arguments such as plasma model calculations [11]
and light emission measurements, it is assumed
that the additional radicals are produced and
confined close to the substrate and that the
transport to the substrate is diffusion dominated.
Consequently, from the pressure dependence of
the decay time of SiH (Fig. 4) and the SiH dif-
generally assumed for the Si radical on the basis of
its hydrogen deficiency [4].
In conclusion, it has been shown that the time-
resolved cavity ring-down experiment applied on a
remote plasma is capable of obtaining radical loss
rates during the actual plasma deposition process.
From partial and total pressure series, it has been
shown that Si is highly reactive in the gas phase
3
3
fusion coefficient [16] the parameters describing
the axial diffusion geometry can be obtained: Fig.
with SiH
b is 0:9 < b 6 1. SiH
active with SiH and has a surface reaction prob-
4
whereas its surface reaction probability
4
time of SiH on the total pressure and therefore
reveals a good linear dependence of the decay
3
on the other hand is unre-
3
4
Eq. (1) has been fitted to the data points setting the
gas phase loss term to zero. The surface reaction
probability follows from the offset in Fig. 4 yield-
ability of b ¼0:30 Æ0:05 which is in good
agreement with previously reported values. In the
future, the geometrical model will be refined such
that the b values of these and other silane radicals
can be determined at very high accuracy and as a
function of plasma and surface conditions.
ing a value of b ¼0:30 Æ0:05 for SiH in the ETP
3
plasma. This value is in good agreement with the
3
estimated b for SiH from indirect measurements
as obtained previously with the ‘aperture-well as-
sembly’ method applied under similar conditions
in the same setup [1]. The value is also in good
Acknowledgements
agreement with b of SiH
3
of b ¼0:28 Æ0:03 as
The authors gratefully acknowledge the con-
tribution of D.C. Schram as well as the Technical
Laboratory Automation group of the Eindhoven
University of Technology, the technical assistance
by M.J.F. van de Sande, J.F.C. Jansen, H.M.M.
de Jong, and A.B.M. H u€ sken and the experimental
work performed by Y. Barrell. This research was
supported by The Netherlands Foundation for
Fundamental Research on Matter (FOM:
determined by Perrin et al. [2] by time-resolved
threshold ionization mass spectrometry in an rf
plasma with a 100% on–off modulation. Further-
more, by applying the same axial diffusion geom-
etry to the measurements of Si (while taking into
4
account the gas phase loss to SiH ), a lower limit
of 0.9 has been obtained for b of Si, yielding
therefore 0:9 < b 6 1. Although this is the first
direct experimental evidence for a nearly unity
surface reaction probability of Si, a b of ꢀ1 is
98PR1781 and 99PSI03) and the Royal Nether-
lands Academy of Sciences and Arts (KNAW) by
the fellowship of one of the authors (W.K.).
References
[
1] See e.g., W.M.M. Kessels, M.C.M. van de Sanden, R.J.
Severens, D.C. Schram, J. Appl. Phys. 87 (2000) 3313, and
references therein.
[
[
[
[
[
2] J. Perrin, M. Shiratani, P. Kae-Nune, H. Videlot, J. Jolly,
J. Guillon, J. Vac. Sci. Technol. A 16 (1998) 278.
3] T. Tanaka, M. Hiramatsu, M. Nawata, A. Kono, T. Goto,
J. Phys. D 27 (1994) 1660.
4] P. Kae-Nune, J. Perrin, J. Guillon, J. Jolly, Plasma Sources
Sci. Technol. 4 (1995) 250.
5] P.R. McCurdy, K.H.A. Bogart, N.F. Dalleska, E.R.
Fisher, Rev. Sci. Instrum. 68 (1997) 1684.
6] W.M.M. Kessels, R.J. Severens, A.H.M. Smets, B.A.
Korevaar, G.J. Adriaenssens, D.C. Schram, M.C.M. van
de Sanden, J. Appl. Phys. 89 (2001) 2404.
Fig. 4. The SiH
3
decay time as a function of the total pressure
19
À3
for a constant SiH
4
density of 3:7 Â10
m .

J.P.M. Hoefnagels et al. / Chemical Physics Letters 360 (2002) 189–193
193
[
7] K.W. Busch, M.A. Busch (Eds.), Cavity-Ringdown Spec-
troscopy: An Ultratrace-Absorption Measurement Tech-
nique, ACS, Washington DC, 1999.
[12] W.M.M. Kessels, J.P.M. Hoefnagels, M.G.H. Boogaarts,
D.C. Schram, M.C.M. van de Sanden, J. Appl. Phys. 89
(2001) 2065.
[
[
8] D.B. Atkinson, J.W. Hudgens, J. Phys. Chem. A 101
[13] Technical Laboratory Automation Group, Eindhoven
University of Technology, Den Dolech 2, 5600 MB
Eindhoven, The Netherlands.
(1997) 3901, and references therein.
9] P.D. Lightfoot, R. Becerra, A.A. Jemi-Alade, R. Lesclaux,
Chem. Phys. Lett. 180 (1991) 441.
[14] C. Courteille, J.L. Dorier, C. Hollenstein, L. Sansonnens,
A.A. Howling, Plasma Sources Sci. Technol. 5 (1996) 210.
[15] P.J. Chantry, J. Appl. Phys. 62 (1987) 1141.
[16] J. Perrin, O. Leroy, M.C. Bordage, Contrib. Plasma Phys.
36 (1996) 1.
[
[
10] A.A. Radzig, B.M. Smirnov, References Data on At-
oms, Molecules and Ions, Springer, Berlin-Heidelberg,
1985.
11] W.M.M. Kessels, A. Leroux, M.G.H. Boogaarts, J.P.M.
Hoefnagels, M.C.M. van de Sanden, D.C. Schram, J. Vac.
Sci. Technol. A. 19 (2001) 467.
[17] A. Takahara, A. Tezaki, H. Matsui, J. Phys. Chem. 103
(1999) 11315.