Dear ADF-users,

I know it is an FAQ, but I'm having serious SCF convergence problems in a
geometry optimization of a rather large (58 atoms) metallacrown complex.

Studying the logfile, there is something I don't understand.
Sometimes, the SCF is only 'acceptably converged', but the calculation
just continues with the next geometry cycle:

  97 ErrMat 0.00545056 MaxEl 0.00073598
  98 ErrMat 0.00566045 MaxEl -0.00077417
 SCF acceptably converged
 Procedure halted (very slow convergence)
  99 ErrMat 0.00509185 MaxEl -0.00080625
  .
  .
  .
 SCF NOT COMPLETELY CONVERGED
 ... accepted
  .
  .
  .
 new geom cycle

In other cases, the calculation just stops, although the SCF is
'acceptably converged' (and the geometry cycles have not reached the
maximum number):

  93 ErrMat 0.00274725 MaxEl 0.00045096
  94 ErrMat 0.00582794 MaxEl 0.00094729
 SCF acceptably converged
 Procedure halted (very slow convergence)
  95 ErrMat 0.00802555 MaxEl -0.00122441
  .
  .
  .
 SCF NOT COMPLETELY CONVERGED
 STOP GEOMETRY ITERATIONS
 NORMAL TERMINATION
 END

Does someone know why this happens?

And more important, does someone have some more ideas to help the SCF
converge? What helps in your case? I've been playing around with lshift,
keeporbitals, smearq, mixing, more iterations, changes in the input
coordinates, a different basis set, a different functional, smaller
geometry optimization steps,... without succes.

As the symmetry of the system is C5, I have to use NOSYM. The charge (+2)
is given in the input.

Thanks in advance for all answers,
Els
Received on 2004-11-09 15:43:14