Silicon for the direct process to methylchlorosilanes
Doctoral thesis
View/ Open
Date
1992Metadata
Show full item recordCollections
- Institutt for kjemi [1352]
Abstract
The available literature in the field of the direct process to methylchlorosilanes has
been reviewed and commented.
An apparatus and operation procedure have been developed to study the effect of
structure on the direct process to methylchlorosilanes. The reproducibility was found
to be better than ±5% on both reactivity and selectivity. However, a very careful
control is required to be able to obtain such a reproducibility. Reproducible stirring
and vibration (movement in the bed) conditions were found to be the most critical
factors.
There is a significant difference between the Low Aluminium (LA, 0.12 wt% Al) and
the High Aluminium (HA, 0.6 wt% Al) samples. HA is approximately 30% higher in
reactivity and 15% poorer in selectivity (Tri/Di) compared to LA. The effect of higher
Al on reactivity is reduced by serious agglomeration problems. These agglomeration
problems occur not only at start-up, but during the whole run. Al is expected to be
necessary in the start up period, but as the silicon is activated, lower levels can be
accepted. Heavy deposition of AICI3 in the tubing was observed for the HA samples
and is expected partly to be the reason for the poor selectivity.
The two samples with low Al had very different gram structures. Surprisingly, the slow
cooled sample had the finest gram structure, which was believed to be caused by a
change in the crystallisation path from columnar growth to equiaxed growth due to
the very slow cooling conditions. Compared to the quickly cooled sample (LAR) the
slow cooled sample (LAS) had 8% improved reactivity and 1 1 % improved selectivity.
Metallographic examination of the LAS sample showed the gram structure similar to a
rapid cooled silicon, but the impurity distribution typical for an annealed silicon.
Between the two high Al samples only minor differences were observed. No
significant difference was observed in reactivity between the rapid cooled (HAR) and
the slow cooled (HAS) samples. But it should be noted that HAS showed improved
selectivity at higher silicon conversion compared to HAR. Clogging and
agglomeration was a problem for both samples, but especially for sample HAS,
probably due to larger intermetallic phases.
This examination shows that a finer grained material has improved performance, but
the effect of structure is more important for qualities with low Al content. The effect of
structure is much less important than the effect of chemical composition, and it does
not make sense to start optimisation of structure if the chemical composition and
metallic impuritiy content are out of control.
The gram boundaries were found to react much faster than the rest of the silicon,
being responsible for the improved reactivity in a fine grained material. Areas with
large amounts of defects are expected to behave similarly since they, as gram
boundaries, have a high dislocation density.
It is shown that it is possible to produce a small grained silicon quality with slow
cooling. The mechanism for this process is not fully understood. Very slow cooling,
that probably is required, is difficult to realise in industrial production.
Annealing of silicon was found to change the shape of the intermetallic phases and
the defect structure. The long thin bands of impurities normally observed in silicon
were changed to more spherical inclusions. The defect density was increased by
annealing, probably due to the stress stored in the solidified ingot, but the amount of
cracks are lower in the annealed sample. Thick castings showed a structure similar to
the annealed samples. Annealing is expected to be beneficial for the performance of
the silicon, botn due to the increased defect density increasing reactivity, and the
round intermetallic phases and change in intermetallic phase composition improving
selectivity.
CaSi2 was found to be detrimental to performance in the direct process, mainly
because it surrounds the other intermetallic phases and thereby reduces the
availability of Al. The start-up period of such a quality is especially critical because it is
mainly then the Al is required. High Ca is also associated with a higher amount of slag
in the silicon due to the lower viscosity and higher density of the slag phase.
Oxygen compounds in the methyl chloride were found to be very detrimental to the
reactivity and selectivity. If these compounds are present during activation of the
silicon a poor distribution of the catalyst on the silicon surface is observed. The
catalyst seems to agglomerate rather than deposit on the silicon particles. As in other
studies, higher temperature was found to increase the reactivity. But in addition it
was found that the increase in reactivity depends on whether the methyl chloride was
purified or not. With purified methyl chloride the reactivity increased by 40% when
the temperature was increased from 300 to 320°C. With unpurified methyl chloride
an increase of 130% was observed for the same temperature rise. This result
indicates that the operating temperature should be adjusted depending on the
amount of oxygen present in the methyl chloride. A pure methyl chloride should
allow a lower reaction temperature than a less pure methyl chloride. It was also found
that lower temperature improves the selectivity, indicating that very pure methyl chloride and low reaction temperature should give the best results in the direct
process.
Oxygen in silicon was found to be less important. The main influence was the lower
amount of active Al and Ca due to the presence of these impurities in the slag phase
instead of in intermetallic phases. However, the results obtained here are from a
batch reactor and it is likely that other results will be observed in a continuous reactor
due to build up of inert particles in the reactor.
A ternary chloride melt of CuCI, ZnCtø and SnCtø was found likely to be present and
the synergy effect of Sn and Zn is probably due to the reduction of the melting point.
Other chloride combinations that lower the melting point and do not form stable
bonds with silicon are likely to be potential promoter systems. Elements giving
complexation effects with CuCI are also potential promoters. The surface oxygen was
not examined in this study.
The intermetailic phases that were found to be without beneficial influence on
performance when mixed into the contact mass, were found to behave different if
they are present in the silicon. The intermetallic phases will distort the silicon lattice
and act as nucleation points for the catalyst deposition. The small cracks created by
the intermetallic phase are very active areas. The round intermetallic phases formed
upon annealing seem to generate more cracks than the thin bands. Again the long
term influence of the impurities has not been examined, but a study is in progress to
determine the influence of the main impurities.
A chain mechanism involving a surface temperature higher than the bulk gas
temperature was considered to explain the sudden start up of the direct process.
Probably due to the heat generated in the chlorination of the impurities like Al.