Académique Documents
Professionnel Documents
Culture Documents
RESUME
ABSTRACT
the hydrogen partial pressure. Nevertheless, for SiHdHCI a quadratic approach fits
the experimental results better.
The obtained selectivity is explained b y the reacting gases decomposition
model and is not affected by the nature of the carrier gas as the growth rate is. These
results show the existence of experimental conditions under which the growth rate is
nearly temperature independent.
A new contribution to the VLSI challenge is provided by this type o f CVD
at reduced pressure and temperature for which abrupt junctions, selective epitaxial
layers and low temperature processes are required.
INTRODUCTION
7 8 9 1 0 11
lo4 / TI°K]
Figure 1.- The growth rate of silane and DCS in an Arrhenius plot
and compares quite well with the values obtained from the literature 114,151.
On the other hand, the measured Ea value for the DCS decomposition is
about 59 kcallmol which is slightly higher than previously published values (between
40 and 50 kcallmol 116,171). If a certain reaction is rate-limiting, our result
implies that this reaction is different from the one generally assumed:
The growth rate for both systems exhibits a reaction order A defined by
the equation:
where p~ stands for the partial pressure of SiH4 or DCS, K is a constant at a given T,
and A is close to one at 900°C. At 800°C and below, this A value is well below one
which is a clue to a Langmuir-Hinshelwood type adsorption mechanism 1181. For the
SiH4 case, we have already mentioned that the SiH2 reaction at the surface must be
involved. For the DCS case it seems that SiCI2 should not be involved due to Ea.
Figure 2.- The growth rate of silane and DCS as a function of hydrogen partial
pressure for two different temperatures.
JOURNAL DE PHYSIQUE
for which lshitani et al. 1201 have given the dissociation energy of 61 kcallmol,
close to our value of 59 kcallmol. Also P. Ho et al. 1211 mentioned the existence of
SiHCl as a product at atmospheric and reduced total pressures. This scheme does not
involve H2, thus qualitatively fitting our observations. As HCI is also a product , it
explains a very interesting feature of our process, namely selective epitaxy.
Selective epitaxial growth is obtained throughout the whole studied
temperature range (1100°C - 650°C) when using DCS diluted in Hp and no epi-layer
thickness limitation has been observed. The selectivity can be easily explained by the
silicon nuclei reduction on the SiO2 mask through the HCI by product. Increasing the
HCI concentration will produce a rapid linear decrease in growth rate These
results eleiminate any explanations based on: SiCI2 + H2 -> Si + 2HCI
1000 900 800 700 ["C]
lo4
74 kcallmol
.- Selectiv.
E without
Y
HCI
Selectiv.
with
2 sccm HCI
10'
7 8 9 10 11
10 / T [OK]
Figure 3.- Growth rate as a function of the reciprocal temperature for the
SiH4/HCI/H2 system showing the region where selectivity is obtained only with SiH4.
When using silane, as reacting gas, selectivity is also obtained at temperatures
above 950°C, and the addition of HCI is not needed, as shown in Figure 3
On the other hand, below that temperature, just a few percent of the SiH4
concentration of HCI is needed to obtain full selectivity. In this case 2 sccm of HCI
in 20 sccrn of SiH4 diluted in 2 slm of H2. Accordingly the activation energy Ea is
affected: from 38 kcal/mol for the SiH4/H2 system it goes to 74 kcal/mol for the
SiH4/HCI/H2 system and selectivity is obtained up to 650°C.
In order to study the chemical reactions in the SiH4/HCI/H2/He system,
the growth rate variation was measured as a function of pH2 and G was plotted
versus the square of this pressure as shown in Figure 4.
0 1 2 3 4
(pH2 [torr])
Figure 4.- The growth rate for the two systems DCS and SiH4/HCI diluted in a gas
mixture of H2 and He are plotted as a function of the H2 partial pressure.
which means that the growth rate should depend on the second power of the hydrogen
partial pressure and the limiting reaction will come from:
where "*" means adsorbed species or a vacant site. Curve (a) from Fig. 4 shows this
quadratic behaviour for H2 pressures above 1 torr. Curve (b) is the same qualitative
C5-526 JOURNAL DE PHYSIQUE
result already shown in Fig. 2 where G is fairly constant above 1 torr of Hp. Curve (c)
corresponds to the same system as (a) at 900°C, which follows a similar beh-aviour to
DCS at 900°C shown in Fig. 2.
What is striking in the present picture is the important decrease in
growth rate at low hydrogen partial pressure, i.e., when there is more He than H2 as a
carrier gas (the total pressure being constant) and in the presence of HCI. This
behaviour is nearly temperature independent. Curve (a), at 750°C, is in the activated
regime and curve (c), at 900°C, in the diffusional one. In all cases the G values
decrease dramatically under 100% He as a carrier gas, whatever the system, i.e.,
DCSIHe or SiH4lHCIlHe. The growth rate'is more affected at 900°C than at 750°C and
full selectivity is preserved in the two systems. To our knowledge the growth rate,
under 100% He as a carrier gas, is lower than under 100% Hp because of a depletion of
reactant in the gas phase. The presence of He enhances the homogeneous
decomposition of reacting gases and the products are carried away from the reactor.
A clear advantage of this system, SiH4/HCI/H2/He, is in the region where
the volumetric ratio for HpIHe is about one: the G values are almost the same for 750
and for 900°C (around 600 Atmin which is fairly high). This implies that the addition
of He is able to remove the pathological sensitivity towards T of the growth rate in
the SiH4IHCIlHp system.
CONCLUSION
Full selectivity has been obtained with only DCS diluted in Hp. However, in
the silane system, a small percentage of HCI in silane is sufficient to obtain
selectivity up to 650°C. This is the main characteristic of our system working in the
torr pressure regime. The etching of silicon by the presence of HCI is enhanced by
this reduced pressure regime.
The rapid thermal process applied to Si epitaxial growth has proven its
capability of growing selectively good epitaxial layers with a high degree of control
and under a wide spectrum of experimental conditions.
The fact that growth rates can be controllably varied from 10 to 1000
Almin makes this procedure a very good candidate for V and ULSl applications.
ACKNOWLEDGEMENTS
The authors would like to thank Mrs. M. Timmins for reading of the
manuscript and R. Carre for technical assistance.
REFERENCES