Re: ESPF and XField

Posted by Nicolas Ferré on March 22, 2010 at 18:16:14:

In Reply to: ESPF and XField posted by Ignacio on March 22, 2010 at 12:09:49:

Hi Ignacio,

I can understand for what concerns ESPF.

: If I understand it correctly, in Molcas 7 there are at least to ways to include an external set of point charges (point charges is all I'm interested in at the moment) that polarize the electron density. One is adding XField in GATEWAY, the other is the ESPF module. Is there any difference between them that I should be aware of? Is the charges' effect properly included in all subsequent calculations (RASSCF, CASPT2, ALASKA) in both cases?

Yes, the 1-e hamiltonian is modified to include directly the interaction between the electrons and the external electrostatic potential. As you noticed, the ESPF method is approximate in the sense that it is based on a truncated multipolar expansion of some electrostatic interaction operator. However, from the QM/MM numerical point of view, there are evidences that such errors are not larger than the ones you get in any QM/MM calculation.

: As for the final interaction energy between the nuclei+electrons and the external point charges, this value was not provided with XField in Molcas 6, and it seems to be still the case (one has to calculate the potential with EPot in SEWARD). With ESPF an interaction energy is given after RASSCF, but I understand is an "approximated" energy, calculated with the fitted mono- and dipoles of the QM molecule. Is there any easy way te get the "real" electrostatic interaction energy between the QM molecule and the external charges?

There is no real electrostatic interaction energy, it depends only on the form of the operator you retain. In that respect, the ESPF electrostatic interaction energy is the real, with respect to the model. In other words, this electrostatic interaction energy is uniquely defined by the model.

: Finally, the "ESPF analysis" is reported after a RASSCF calculation, but not after CASPT2. Is there some keyword to add for CASPT2?

I've never put it into the code. Please fill a bug report about that.

Best regards,

Follow Ups:

Re: ESPF and XField Ignacio 11:44:09 3/24/10 (0)

Post a Followup

Name:
E-Mail:

Subject:

if B is 1s22s22p1, what is Li?

Passfield:

Comments:
: Hi Ignacio, : I can understand for what concerns ESPF. : : If I understand it correctly, in Molcas 7 there are at least to ways to include an external set of point charges (point charges is all I'm interested in at the moment) that polarize the electron density. One is adding XField in GATEWAY, the other is the ESPF module. Is there any difference between them that I should be aware of? Is the charges' effect properly included in all subsequent calculations (RASSCF, CASPT2, ALASKA) in both cases? : Yes, the 1-e hamiltonian is modified to include directly the interaction between the electrons and the external electrostatic potential. As you noticed, the ESPF method is approximate in the sense that it is based on a truncated multipolar expansion of some electrostatic interaction operator. However, from the QM/MM numerical point of view, there are evidences that such errors are not larger than the ones you get in any QM/MM calculation. : : As for the final interaction energy between the nuclei+electrons and the external point charges, this value was not provided with XField in Molcas 6, and it seems to be still the case (one has to calculate the potential with EPot in SEWARD). With ESPF an interaction energy is given after RASSCF, but I understand is an "approximated" energy, calculated with the fitted mono- and dipoles of the QM molecule. Is there any easy way te get the "real" electrostatic interaction energy between the QM molecule and the external charges? : There is no real electrostatic interaction energy, it depends only on the form of the operator you retain. In that respect, the ESPF electrostatic interaction energy is the real, with respect to the model. In other words, this electrostatic interaction energy is uniquely defined by the model. : : Finally, the "ESPF analysis" is reported after a RASSCF calculation, but not after CASPT2. Is there some keyword to add for CASPT2? : I've never put it into the code. Please fill a bug report about that. : Best regards, : NF

[ Follow Ups ] [ Post Followup ] [ Molcas user's WWWBoard ]