Hi,
Has anyone experienced that SCF convergence may be dependent on the
position of the molecule if cartesian coordinates are used. And also
that the converged geometry is dependent on the position of the
molecule, and on the choice of coordinates, cartesian or internal?

I am doing geometry optimizations on some large cluster models of Cr on
silica with small hydrocarbons bonded to Cr, e.g. CrSi6O19H11-C2H6. A
large part of the cluster is kept frozen during the optimization. I
first noticed that SCF convergence was easier achieved when the cluster
was located close to origo. Also, internal coordinates give lower
geometry optima than cartesian coordinates. The energy difference is
usually in the range of 2-10 kJ/mol. Bond lengths are similar, but
angles may differ up to 10-20 degrees. Varying the convergence criteria
of the cartesian coordinates optimization does not change the results.

I have done checks on smaller molecules like H2O, here there is no
problem.

I would be grateful for any input!
Sindre Lillehaug
Received on 2002-02-19 11:06:39