[molpro-user] convergence problems with RHF

Lee E.P.F. E.P.Lee at soton.ac.uk
Fri Jan 29 10:27:17 GMT 2010

Grant is correct. Looking at the occupied mos, a lot of them have large +ve mo energies, suggesting that it is a highly excited state at the last SCF iteration. I have no time to look at the mos in details. Nevertheless, my guess is that your basis set for non-U atoms are too poor. Consequently, e's get into mos of U instead of non-U mos. 

Grant''s suggestion of improving the initial guess of the wavefunction is in the right direction. However, your system has C1 symmetry. Hence all mos can mix. Consequently, you can not use occ to fix the occupancies. My feeling is that if you do not improve yur basis set of the non-U atoms, you will get this kind of troubles. In general, the basis set used should be balanced. That is the U basis set and non-U basis sets should be of the same quality. Using imbalance basis set will get strange results.

Regards, Ed.

From: molpro-user-bounces at molpro.net [molpro-user-bounces at molpro.net] On Behalf Of Grant Hill [jghill at wsu.edu]
Sent: 28 January 2010 18:16
To: Dimitri Cavalcanti
Cc: molpro
Subject: Re: [molpro-user] convergence problems with RHF


It is no use playing around with the convergence of the geometry optimizer when your initial SCF energy is bouncing around like crazy. Assuming your basis set etc. is ok, perhaps you can try to improve your initial guess. One way would be to run multi first, but I've no experience of doing this on such a large system.

Additionally, why turn symmetry off and then use occ,15,24?



On 28 Jan 2010, at 06:48, Dimitri Cavalcanti wrote:

> Hello molpro users,
>    Its me again, with a help of one molpro user i solve my memory problem, but, the convergence problems with RHF, still stands. I already try numerical gradients, reduce the thrshold, reduce the convergence criteria for energy and gradient. Almost every thing that is in the manual. and i still have the same problem:
> ? Error
>  ? No convergence
>  ? The problem occurs in rhfpro
> I need to perform this calculation with RHF, so, no change in the method. Below goes my input and output:

Molpro-user mailing list
Molpro-user at molpro.net

More information about the Molpro-user mailing list