Ceilidh ... Re^3: scf seems to converge, but I obtain wildly fluctuating mp2 gradients

PC GAMESS/Firefly-related discussion club

Re^3: scf seems to converge, but I obtain wildly fluctuating mp2 gradients

Dear Prof. Monev,

You can specify tighter convergence criteria for SCF procedure.
Alternatively, you can loosen value of the tolsym keyword
of $moorth group (the default is 1d-10). This parameters controls
various symmetry-related checks of MOs quality, etc...

Other useful symmetry-related keywords of this group are:

  SYMS  (logical).  If set, additional extra symmetrization 
                    of the overlap matrix will be performed.

  SYMDEN (logical). If set, additional symmetrization 
                    of the density matrix will be 
                    performed during SCF procedure.

  SYMVEC (logical)  Requests tighter symmetrization of MOs after SCF.

 E.g., you can try to add 
 
 $moorth syms=.t. symden=.t. symvec=.t. tolsym=3d-8 $end 

 to your input.

Regards,
Alex Granovsky

>I tried using C2 symmetry. After several optimization steps, I obtained different versions of the following error, depending on the starting geometry (see attached file):
>MP2 TRANSFORMATION CANNOT ASSIGN A DEFINITE ORBITAL SYMMETRY TO MO 185
> JOB ABORTS. YOU MUST EITHER
> A) PROVIDE A SYMMETRY ADAPTED MO SET BY GUESS=MOREAD
> B) ENTER NOSYM=1 TO USE THESE SYMMETRY BROKEN MO-S,

>Trying c2v symmetry (i.e. z values zero), equilibrium geometry was rapidly reached – only to prove, when tested with a hessian calculation, to have imaginary frequencies. I did obtain a non-symmetric optimized geometry but finding a way to utilize symmetry for the calculations, as well as obtaining a symmetric optimized geometry, preferably c2, would have been very satisfying… Any hints?

Sat Aug 8 '09 8:01pm

This message read 1548 times