The CASINO forum

Dear all,

I performed VMC calculations with the CI trial wave function
,which is made without all pseudo potential and
of which basis set is gaussian basis basis set.
But I was not able to get correct vmc energy:
First, I performed the calculation with 'cusp correction', and
vmc energy was about 10000 Hartree higher than SCF energy.
Second, I performed the calculation without 'cusp correction', and
vmc energy was about 7 Hartree higher than SCF energy.

Could you tell me how to get correct vmc energy?

Ichibha

Dear Tomohiro Ichibha.

It would be very useful to describe the steps to reproduce your results, including the version of software used, input files, and possibly additional instructions.
Otherwise it is impossible to understand what happened in your case.

Vladimir.

One obvious point: a difference of 10,000 Hartree (=0.27 MeV) indicates that something has gone completely wrong in the calculations!

Best wishes,

Neil.

Neil Drummond wrote:One obvious point: a difference of 10,000 Hartree (=0.27 MeV) indicates that something has gone completely wrong in the calculations!

Best wishes,

Neil.

I managed several times to get such a big difference in energy by setting jastrow U TERM cutoff to >100 au in varmin_linjas optimisation, but sample variance was even greater in this case.
I do not know the details of the Ichibha calculation, so do not want anything to suppose.

Vladimir.

Dear all,

Sorry for my late response.
By substituting gpcc for normal cusp_correction, The calculation runs correctly.

Thank you very much, Vladimir and Neil.

Ichibha

It would be interesting to see why that is. Can you send me the input files?

Cheers
Mike

Hi,

Can you also state what molecule you are using? I have seen this problem before as well, where the gaussian_cusp_correction gives clearly erroneous values for the VMC calculation but the general purpose scheme gives reasonable values. I don't have specific data on hand, but I seem to recall it happening on molecules that contain several hydrogen atoms, such as water clusters or a benzene or anthracene molecule.

Thanks,
Mike

Hi guys,

I'm now been given access to these files - I'll try to find time to investigate the alleged problem this week.

Cheers,
Mike