[molpro-user] scf problem during optg

[Werner Győrffy]

Dear Pawel,

In a geometry optimization, tight convergence criteria should be used 
for the energy calculation to avoid noise in the gradients, and to be 
able to converge the geometry optimization. In other words, a geometry 
optimization might not converge with unreliable gradients. So it is in 
principle the correct strategy not to proceed further with a geometry 
optimization if the energy is not converged. [For some methods, Molpro 
can however continue a geometry optimization even if the energy 
calculation is not converged (controlled by the user or automatically by 
the program)].

In correlated electronic structure theory methods, a calculation may be 
performed without any problem even if the the SCF calculation is not 
converged. Depending on the method, the correlated energy value can of 
course have a very large error in this case.

Note that the lack of convergence in SCF may indicate that the 
theoretical model for your wave function is wrong (spin, symmetry, 
charge, initial guess for orbitals, reasonable starting geometry, etc.).

Please also check "Handling difficult cases: When SCF does not converge" 
in the User's Manual.

Regards,

Werner Gyorffy.

On 08/04/2013 02:47 PM, Pawel Kadlubanski wrote:
> I have a problem with scf convergence during geometry optimization. When
> I compute the single point on MP2 level of theory I do not have any
> problem but during during optimization process the program stops without
> doing even one optimization point due to lack of scf convergence. Could
> anyone help me ? I put my input file and error info. Below.
>
>
> Pawel Kadlubanski
>
>
> ***,FTALOCYJANINA-IN-MP2;
>
> Memory,600 MW;
>
> Gprint,basis; ! Print basis information
>
> Gprint,orbital; ! Print orbitals in SCF and MCSCF
>
>
> geometry=ftalo_In.xyz
>
>
> basis=LANL2DZ
>
>
> {uhf;shift,-1.0,-0.5;maxit,300,save,,2100.2}
>
> ump2
>
> optg,MAXIT=100,ENERGY=1d-5
>
> {frequencies
>
> thermo,sym=auto
>
> print,thermo}
>
> put,molden,ftalo_In.molden;
>
> -----
>
>
> PROGRAM * OPT (Geometry optimization) Authors: F. Eckert and H.-J. Werner
>
>
>
> BFGS update of hessian
>
>
> *** Long output written to logfile
> /home/pawel001/molpro/testFtaloMP2/ftalo_In.log ***
>
>
>
> Geometry optimization using default procedure for command UMP2
>
>
> Geometry written to block 1 of record 700
>
>
> Generated 165 symmetrical displacements
>
>
> Number of displacements for numerical gradient: 330
>
>
> ? Error
>
> ? No convergence
>
> ? The problem occurs in uhfpro
>
>
> GLOBAL ERROR fehler on processor 0
>
>
>
>
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>