Surface & Coatings Technology 201 (2007) 8825 – 8829
www.elsevier.com/locate/surfcoat
Homoepitaxial silicon carbide deposition processes via chlorine routes
A. Fiorucci, D. Moscatelli, M. Masi ⁎
Dipartimento di Chimica Materiali e Ingegneria Chimica “Giulio Natta”, Politecnico di Milano, via Mancinelli 7, 20131 Milan Italy
Available online 3 May 2007
Abstract
The use of chlorinated precursors recently emerged as the most promising route in SiC CVD but the chemical mechanisms involved in are still
not completely defined. Thus, the homoepitaxial SiC film in horizontal hot wall CVD reactors was here analyzed by considering different
processes involving chlorinated species. A general but rather simple deposition mechanism involving the most important chemical species was
then developed. Reaction rate constants were estimated through classical kinetic theories while most significant reaction rate constants were
refined through quantum mechanical methods. The resulting mechanism was embedded into a simplified reactor model and the simulation results
were compared against experimental growth rate data.
© 2007 Elsevier B.V. All rights reserved.
Keywords: [A] Growth models; [C] Chemical vapour deposition; [D] Silicon carbide; [X] Epitaxial deposition
1. Introduction
Silicon carbide is commonly used in power electronics and
in optoelectronics, for applications like rectifiers (Schottky
diodes) and transistors (MOSFET, MESFET). However, SiC
market is still limited by difficulties both in growing highly
crystalline bulk material and in depositing good quality
epitaxial layers via CVD. Thus, to increase the SiC surface
quality up to electronic grade, the deposition of epitaxial layers
onto SiC substrates is still necessary. Unfortunately, this process
is very difficult to control both in terms of film composition and
of “flat area” portion of the reactor. Moreover, the traditional
recipe based on silane and light hydrocarbons shows many
practical problems related to the formation of particulate that
practically limits the growth rate values around the 20 μm/h. To
ensure both high quality films and high throughput production
processes, the use of chlorinated precursors recently emerged as
the most promising route in SiC technology where at least three
different pathways can be considered, like adding hydrochloric
acid to the standard precursors [1,2], using silicon chlorinated
silicon precursors [3] or even carbon chlorinated precursors [4].
However, because the carbon–chlorine bond is the weakest one
(i.e., C–Cl ≈ 327 kJ/mol vs. Si–Cl ≈ 381 kJ/mol), these last
⁎ Corresponding author. Tel.: +39 0223993131; fax: +39 0223993180.
E-mail address: maurizio.masi@polimi.it (M. Masi).
0257-8972/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2007.04.110
precursors starts to decompose at low temperatures and
therefore are of in interest for low temperature (e.g., around
1300 °C), low growth rates processes [4]. On the other hand, the
technology is shifting towards the direct use of SiCl4 or SiHCl3
because of their easier process handling with respect to the HCl
addition to the SiH4/hydrocarbons mixture.
Accordingly, starting from the previously developed
mechanisms involving both the HCl addition [5] and the direct
feed of SiHCl3 [6], a refined deposition mechanism was here
developed. The detailed mechanism originally developed
involving 60 gaseous, 19 adsorbed species connected through
233 reactions (75 surface); in this work the mechanism was
reduced because of the necessity to have a simpler figure of the
system and thus only the really main species were considered.
In [7] part of this work was done, but only the SiH4/
hydrocarbon/HCl/H2 system was examined to define the first
lumped mechanisms; here the addition of SiHCl3 and CH3Cl
was also addressed. This new lumped model can accept as entry
point the whole system of precursors usually adopted to grow
epitaxial SiC. This was possible because, considering that most
of the species considered in the detailed mechanism [5,7] are
present in concentrations approaching the delectability threshold, the species here included were only those present in
significant amount in the process conditions of interest for the
epitaxial deposition. In any case, the involved gas phase and
surface reactions were all microscopically reversible and their
8826
A. Fiorucci et al. / Surface & Coatings Technology 201 (2007) 8825–8829
Fig. 1. Complete thermodynamic analysis for the systems (A) SiHCl3/C2H4/H2 with H2 = 150 slm, Si/H2 = 0.040%, C/Si = 1.75 and (B7) SiH4/CH3Cl/H2 with
H2 = 150 slm, Si/H2 = 0.040%, C/Si = 1.75.
reaction rates were estimated through classical kinetic theories
(e.g., collisional and transition state theories) [8]. However, the
most significant reaction rate constants were refined also
through quantum mechanical methods. The resulting mechanism was finally embedded into a simplified reactor model and
the simulation results were compared against experimental
growth rate data.
chlorine side of the mechanism because the literature provides
detailed schemes for the typical carbon precursors (e.g., C3H8,
C2H4 and CH4) pyrolysis. For all the involved species,
thermodynamics data were taken from SANDIA Thermochemistry
database [11] while computations were performed through
2. Thermodynamic analysis
At typical process temperatures for the epitaxial SiC processes
(i.e., around 1600 °C), only a few reaction paths are really
important because the system conditions tends to approach the
thermodynamics equilibrium ones. The thermodynamic analysis
of SiC growth systems involving chlorinated precursors was
already performed also in [9,10], but the examined conditions
differ significantly from those examined here, by addressing PVT
and HTCVD processes. Thus, a preliminary thermodynamics
analysis, like the one illustrated in Fig. 1, can be a useful tool to
identify the key gaseous species to be included in the kinetics
analysis. In particular, this tool is important to analyze the silicon–
Fig. 2. Gas phase kinetic scheme.
8827
A. Fiorucci et al. / Surface & Coatings Technology 201 (2007) 8825–8829
Fig. 3. Sketch of the main reaction pathways involved in the deposition mechanism (silicon species adsorbed on surface carbon and vice versa).
Cantera software [12]. In particular, two different systems were
analyzed here (i.e., SiHCl3/C3H8/H2 and SiH4/CH3Cl/H2) to test
the necessity to include different species while changing the
chlorine carrying molecule.
Generally, by considering the results obtained for the SiHCl3/
C3H8/H2 system shown in Fig. 1A, the most important species
gaseous species on the solid SiC are H2, HCl, H, C2H2, CH4,
CH3, C2H4, Si, SiH2, SiCl and SiCl2. It is also very important to
notice that most abundant silicon–carbon species (i.e., Si2C and
SiC2) show molar fractions below 10− 7; however, because these
are the key species for SiC high temperature bulk growth they
were still included in the mechanism. Because these results are
coherent with other works [13], all of these species were thus
included into the lumped scheme developed below. Moreover,
the relative species abundance was checked also for the SiH4/
CH3Cl/H2 system, obtaining results comparable with those of
the previous system for analogous C/Si/Cl/H molar ratios, as
shown in Fig. 1B.
also the presence of adsorbed chlorine and chlorinated species,
reflecting more the chemistry of the silicon deposition from
chlorosilanes [14]. A particular feature of this surface
mechanism is to consider all the possible surface reactions
among the adsorbed species and not only a “sticking
coefficient” incorporation in the film of the gas phase ones.
There, all the gaseous species adsorb dissociatively, producing
the following adsorbed species H, C, CH, SiH2, SiH, Si, SiCl,
Cl where carbon containing species stick on a silicon atom and
vice versa for the silicon containing ones, while chlorine can be
bonded to both silicon and carbon. Then, these adsorbed species
react to desorb gaseous stable species or to form the three solid
species SiC, Si2C and SiC2, where, the last two are fictitious
ones representing the silicon rich and the carbon rich lattices.
3. Kinetic scheme
Reaction
As underlined before, the aim of this work is to propose a
simple kinetic scheme for SiC growth via chlorinated precursors
that allows the prediction of growth rate and main gas phase
species. The proposed mechanism includes 19 gaseous and 9
surface species linked by means of 25 gas phase and of 61
surface reactions, as sketched in Figs. 2 and 3, whose reactions
and reaction rate constants are summarized in Tables 1 and 2 for
the gaseous and the surface reactions, respectively. Carbon
precursor reactions start from propane or ethylene because they
are the most commonly used precursors. However, the pathways to produce C2H2 or CH4, that are the most abundant
carbon containing species, are similar for all the light hydrocarbons. Main silicon precursors examined were SiHCl3 and
SiH4, producing SiHX and SiClX (X = 0, 1, 2) as their main
fragments. These species, in particular dichlorosilylene (SiCl2)
and gaseous Si, are the main responsible for silicon adsorption
on the surface and consequently for SiC growth when they
interact with the C-containing molecules. The same species are
also responsible for the formation of gaseous Si2C and SiC2. On
the surface side, the original mechanism, already developed for
the silane–hydrocarbons system [5,6], was upgraded to include
Table 1
Gas phase kinetic for SiC crystal growth; K = A·Tb·e− E/RT, with parameters
coherent with rates in [mol/cm3/s], gas concentrations in [mol/cm3] and
activation energy in [cal/mol]
2H + M_H2 + M
C3H8_C2H5 + CH3
C2H5_C2H4 + H
C2H4 + H_C2H5
C2H4_C2H2 + H2
CH4_CH3 + H
CH3 + H2_CH4 + H
2CH3_C2H5 + H
CH4 + CH3_C2H5 + H2
SiHCl3_SICl2 + HCl
SiH4_SiH2 + H2
Si + H2_SiH2
SiH2 + H_SiH + H2
Si + H_SiH
SiH_Si + H
2Si + CH4_Si2C + 2H2
2SiH + CH4_Si2C + 3H2
Si + C2H2_SiC2 + H2
SiH2 + C2H4_SiC2 + 3H
Si + HCl_SiCl + H
SiCl + HCl_SiCl2 + H
SiH3Cl_HCl + SiH2
SiH2 + HCl_SiCl + 3H
CH3Cl + H2_CH3 + HCl
R
R
F
F
R
R
R
R
R
F
R
R
R
F
F
F
F
F
F
F
F
R
F
F
Log10A
β
Ea [cal/mol]
Reference
18.73
22.42
10.8
8.92
26.5
16.92
2.361
21.06
13.01
14.64
28.2
12.08
14.05
13.3
6.18
19.75
21.3
15.1
13.35
15.4
12.75
14.99
13.27
15.1
−1.3
−1.8
0.4
1.49
−3.5
0
3.12
0
0
0
−4.79
0.5
0
0
0
0.23
0
0
0
−0.2
0
0
0
0
0
88,693.72
39,200
991.513
88,759.29
103,810
8711
26,377.4
22,850
73,683.9
60,450.0
0
15,000
20,000
32,135
5000
0
30,000
0
38,510
17,190
76,000
16,090
75,000
[5]
[5]
Modified
Modified
[5]
[5]
[5]
Modified
[5]
Modified
[5]
Modified
[5]
[5]
[5]
[5]
[5]
[5]
[5]
[5]
[5]
[5]
Modified
Modified
Gas reactions are equilibrium (R) or forward (F). Reaction constants obtained by
previous models or modified.
8828
A. Fiorucci et al. / Surface & Coatings Technology 201 (2007) 8825–8829
Table 2
Surface kinetic scheme for SiC crystal growth ( is Si surface site, $2 is C surface
site, X⁎ are adsorbed species); K = A·T b·e− E/RT, with parameters coherent with
rates in [mol/cm2/s], surface concentrations in [mol/cm2] and activation energy
in [cal/mol]
Reaction
Log10A
β
Ea [cal/mol]
Reference
CH4 + $1_C⁎ + 2H2
CH3 + $1_CH⁎ + H2
C2H5 + 2$1_C⁎ + CH⁎ + 2H2
C2H4 + 2$1_2C⁎ + 2H2
C2H2 + 2$1_C⁎ + H2
SiH2 + $2_SiH2⁎
SiH4 + $2_SiH2⁎ + H2
SiH + $2_SiH⁎
Si + $2_Si⁎
SiHCl3 + 2$1 + 2$2_
SiCl⁎ + H⁎ + 2Cl⁎S
SiCl2 + 2$2_SiCl⁎ + Cl⁎C
SiCl2 + $2 + $1_SiCl⁎ + Cl⁎S
HCl + $1 + $2_CH⁎ + Cl⁎S
H2 + 2$2_2H⁎
2CH⁎_C⁎ + H2
H + CH⁎_C⁎ + H2
H + C⁎_CH⁎
CH⁎_0.5H2 + C⁎
CH⁎ + CH⁎_C2H2 + 2$1
CH⁎ + H2_$1 + CH3
CH⁎ + H⁎_H2 + $2 + C⁎
CH⁎ + 0.5H2_C⁎ + H2
2SiH⁎_2Si⁎ + H2
SiH2⁎_Si⁎ + H2
Si⁎_$2 + Si
SiCl⁎ + 0.5H2_Si⁎ + HCl
SiCl⁎ + H_Si⁎ + HCl
SiCl⁎ + HCl_SiCl2 + H + $2
SiCl⁎ + Cl⁎C_SiCl2 + 2$2
SiCl⁎ + Cl⁎S_SiCl2 + $1 + $2
2SiCl⁎_SiCl2 + Si⁎ + 2$2
SiCl⁎ + CH⁎_SiC + $1 + $2
H⁎ + SiCl⁎_HCl + Si⁎ + $2
2Cl⁎C + H2_2HCl + 2$2
Cl⁎C + H_HCl + 2$2
2Cl⁎C + SiCl2_SiCl4 + 2$2
2Cl⁎S + H2_2HCl + 2$2
Cl⁎S + H_HCl + $1
Cl⁎C + Cl⁎S + H2_2HCl + $2 + $1
Cl⁎C + Cl⁎S + SiCl2_SiCl4 + $2$1
CH⁎ + Cl⁎S_HCl + C⁎ + $1
Si⁎ + Cl⁎S_SiCl⁎ + $1
H⁎ + Cl⁎S_HCl + $2 + $1
CH⁎ + Cl⁎C_HCl + C⁎ + $2
Si⁎ + Cl⁎C_SiCl⁎ + $2
H⁎ + Cl⁎C_HCl + 2$2
H⁎ + H⁎_H2 + 2$2
HCl + SiC_SiCl⁎ + CH⁎
Si⁎ + C⁎_SiC + $1 + $2
2Si⁎ + C⁎_Si2C(s) + 2$2 + $1
Si⁎ + 2C⁎_SiC2(s) + $2 + 2$1
SiCl + $2_SiCl⁎
SiC2 + $1 + $2_2C⁎ + Si⁎
2C⁎ + Si⁎_SiC2
2Si⁎ + C⁎_Si2C
Si⁎ + CH⁎_SiC + $1 + $2 + H
SiH⁎ + CH⁎_SiC + $1 + $2 + 2H
SiH⁎ + C⁎_SiC + $1 + $2 + H
Si2C + 2$2 + $1_2Si⁎ + C⁎
SiH3Cl + 2$2_SiCl⁎ + H⁎ + H2
9.38
11.93
20.76
17.97
19.08
11.78
10.5
11.79
11.8
16.42
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0
0
0
0
0
0
18,678
0
0
0
Modified
[5]
[5]
Modified
[5]
[5]
[5]
[5]
[5]
Modified
19.49
19.49
12.55
11.36
23
12.55
12.55
23
23
11.36
23
12.36
25
25
13
15.37
15.52
10.55
19
19
19
17
19
12.37
12.5
11.51
12.37
12.5
12.37
10.51
19
17
19
19
17
19
24
10.55
17
23
23
11.62
20.1
13
13
19
19
19
19.57
16.42
0.5
0.5
0.5
0.5
0
0.5
0.5
0
0
0.5
0
0.5
0
0
0
0.5
0.5
0.5
0
0
0
0
0
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0
0
0
0
0
0
0
0.5
0
0
0
0.5
0.5
0
0
0
0
0
0.5
0.5
0
0
0
0
61,000
0
0
57,100
87,954
87,954
61,000
0
61,000
61,000
40,500
60,000
60,000
0
20,095
89,806
89,806
0
70,006
78,580
0
25,000
89,806
0
84,190
25,000
89,806
0
89,806
84,190
0
84,190
61,000
0
0
0
0
0
0
82,262
55,834
0
0
0
0
0
Modified
Modified
[5]
Modified
[5]
[5]
[5]
[5]
[5]
[5]
[5]
[5]
Modified
Modified
[5]
[5]
Modified
[5]
[5]
[5]
[5]
Modified
Modified
[5]
[5]
[5]
[5]
[5]
[5]
Modified
[5]
[5]
Modified
Modified
Modified
Modified
Modified
[5]
Modified
Modified
Modified
Modified
Modified
[5]
[5]
[5]
[5]
[5]
[5]
[5]
Table 2 (continued)
Reaction
Log10A
β
Ea [cal/mol]
SiH⁎ + $2_Si⁎ + H⁎
17
0
0
Reference
[5]
All surface are forward reactions. Reaction constants obtained by previous
models or modified.
The final solid composition was then obtained by considering
their relative abundance.
4. Reactor model
The developed kinetic scheme was embedded into a
simplified CVD reactor model [15,16]. This model solves the
mass balances for any involved species and the energy balance,
in a heterogeneous ideal plug flow reactor scheme, to evaluate
the thermal and the composition profile along the reactor axis.
Its use is justified by considering the resulting very high Peclet
value (about 1000) in the experimental reactor where the
validation data are taken. Here, the mass transport coefficient
was estimated by the combination of that calculated for flow in
rectangular ducts and that induced by the susceptor rotation.
Because the slow rotation speed usually adopted, the former
phenomenon dominates and thus the susceptor rotation has its
main effect in the azimuthal averaging of the growth rate.
5. Results and discussion
The experimental data taken for the model validation were
obtained in LPE ACISOne reactor. This reactor holds 6 × 2″ or
3 × 3″ wafers, with an inlet carrier flow rate of 150 slm [3].
The experimental runs were performed at different temperatures
(between 1550 and 1650 °C) and with different values of C/Si,
Cl/Si and Si/H2 inlet ratios as summarized in Tables 3 and 4
where all the reported runs exhibit mirror like surfaces and
lower defect density, compared with traditional processes
[3,17]. Runs examined both SiH4/C2H4/HCl/H2 and SiHCl3/
C2H4/H2 systems. As illustrated in Fig. 4, the calculated growth
rate values matches the experimental ones within the 5% error,
thus well inside the experimental error (e.g., about 10%). It is
important to remark that with respect to the starting lumped
mechanism [5,7] no corrections were introduced in the reaction
rate values besides the addition of the SiHCl3 decomposition to
SiCl2, that is a well known reaction with published rate constant
[18], and the CH3Cl decomposition, here estimated. The rate
Table 3
Main experimental conditions tested in the ACISOne reactor
1
2
3
4
5
0.00012
0.00022
0.00040
0.00045
0.00063
Si/H2
C/Si
2.7
2.5
1.75
1.5
1.5
150
150
150
150
H2 flow (slm) 150
Pressure
100
100
100
100
100
(mbar)
Temperature 1550
1600
1600
1650
1650
(°C)
Process operated with SiHCl3 + C2H4 + H2.
8829
A. Fiorucci et al. / Surface & Coatings Technology 201 (2007) 8825–8829
Table 4
Main experimental conditions tested in the ACISOne reactor
Si/H2
Cl/Si
C/Si
H2 flow (slm)
Pressure (mbar)
Temperature (°C)
1
2
3
4
5
6
7
0.0010
2
1.5
150
100
1650
0.0020
1
1
150
100
1650
0.0030
1
1
150
100
1650
0.0035
1
1
150
100
1650
0.0040
1
1
150
100
1650
0.0045
1
0.8
150
100
1650
0.0060
1
0.8
150
100
1650
Process operated with SiH4 + HCl + C2H2 + H2.
constant basis set was obtained by means of consolidated
reaction rate estimation methods while for the most critical ones
(e.g., for reactions forming gaseous SiCl2) a quantum chemical
based refinement was performed [19]. It is important to notice
that the HCl addition produces processes maintaining high
quality surfaces even at higher growth rates than the pure
chlorosilanes feed. This point can be related to the reduction in
the process degrees of freedom about the Cl/Si ratio. Thus, it
is advisable that the industrial process will shift towards the
SiHCl3/HCl recipe.
The obtained results start to produce a moderate optimism
about the understanding of the main features for the deposition
mechanism involving chlorinated precursors. Both the addition
of HCl and the direct introduction of SiHCl3 were already
tested. Because the growth recipes reported in the literature are
not significantly different from those here examined it is
reasonable to forecast that the reported mechanism will match
also experimental data of different sources. Unfortunately, up to
now, these data were not available in the open literature and we
hope a further test of this mechanism by also other authors. In
particular the extension to the HTCVD conditions would be
very interesting.
Acknowledgments
The financial support of LPE Epitaxial Technology is kindly
acknowledged.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
Fig. 4. Experimental growth rate values (♦) and calculated ones (———) for
the points of (A) Table 3. SiHCl3 + C2H4 + H2 system. (B) Table 4. SiH4 + HCl +
C2H2 + H2 system; experimental growth rate data were collected along the whole
susceptor and then averaged; however, uniformity for showed points was always
below 10%; the same averaging procedure was adopted also for simulated
points.
O. Kordina, et al., Phys. Status Solidi, B Basic Res. 202 (1) (1997) 321.
D. Crippa, et al., Silicon Carbide and Related Materials 2004, 2005, p. 67.
S. Leone, et al., Mater. Sci. Forum 527529 (2006) 179.
Y. Koshka, et al., J. Cryst. Growth 294 (2) (2006) 260.
A. Veneroni, M. Masi, Chem. Vap. Deposition 12 (8–9) (2006) 562.
A. Veneroni, F. Omarini, M. Masi, Cryst. Res. Technol. 40 (10–11) (2005)
967.
M. Masi, A. Veneroni, ECS Transac. 2 (7) (2007) 11.
S.W. Benson, Thermochemical Kinetics, Wiley Interscience, New York,
1976.
D. Chaussende, et al., Chem. Vap. Deposition 12 (8–9) (2006) 541.
S. Nigam, et al., J. Cryst. Growth 284 (1–2) (2005) 112.
SANDIA Thermochemistry database, www.sandia.gov.
www.cantera.org.
D.D. Avrov, et al., J. Cryst. Growth 199 (1999) 1011.
G. Valente, et al., J. Cryst. Growth 230 (1–2) (2001) 247.
S. Carra, M. Masi, Prog. Cryst. Growth Charact. Mater. 37 (1) (1998) 1.
M. Masi, S. Kommu, Silicon Epitaxy, 2001, p. 185.
F. La Via, et al., Chem. Vap. Deposition 12 (8–9) (2006) 509.
S.P. Walch, C.E. Dateo, J. Phys. Chem., A 105 (10) (2001) 2015.
K. Raghavachari, J.B. Anderson, J. Phys. Chem. 100 (31) (1996) 12960.