TURBOMOLE Users Forum

TURBOMOLE Modules => Jobex: Structure Optimization and Molecular Dynamics => Topic started by: greisen on January 10, 2011, 06:34:21 PM

Title: converged problems for open-shell system using BP86
Post by: greisen on January 10, 2011, 06:34:21 PM

Hi all,

I am optimizing a cluster of ions where I have first optimized using the highest spin state. After convergence I took those coordinates and transfered them into a new coordinate file. I then used a lower spin state but now I get problems with the scf convergence of the my cluster. I have used default settings for the scf-options and I am using RI in order to speed up calculation. Any flags to set that might help convergence of the cluster?

Thanks in advance

Title: Re: converged problems for open-shell system using BP86
Post by: antti_karttunen on January 12, 2011, 10:05:51 AM

Hi,

Try changing the following keywords (see the manual for details):
  $scfdamp start=4.000 step=0.100 min=0.500
  $scforbitalshift closedshell=.4
  $scfiterlimit 150

If this does not help, just increase the damping (and perhaps the shift, too):
  $scfdamp start=8.500 step=0.200 min=0.500
  $scforbitalshift closedshell=.5
  $scfiterlimit 200

For closed-shell systems, use automatic $scforbitalshift instead of closedshell.

These keywords are usually the most straightforward way of reaching convergence. Fermi-smearing can also be very helpful, especially when used with the "nue" option (see manual for $fermi)

Regards,
Antti

Title: Re: converged problems for open-shell system using BP86
Post by: amavr78 on January 12, 2011, 02:36:04 PM

In some cases, it may help convergence for low-spin, if you use the preconverged high-spin orbitals.