Molcas Forum

vkj

Member

Registered: 2019-04-10

Posts: 29

Compute the effect of Xfield on previous JobMix

Dear molcas users,

Is it possible by some trick to compute the effect of a Xfield on a previously computed MS-CASPT2/XMS-CASPT2 or even SS-CASPT2 computation?

To elaborate lets say I have some computation at SS-CASPT2/(M,X,R)S-CASPT2 level. Now I want a new computation with the same input with the only change being the addition of an Xfield section. However I do not want to repeat the RASSCF+PT2 computation with the xfield. A common use case might be when xfield is a solvent or flexible part config and I want to average over hudreds/thousands of configs.

I could just put all of the configs in a single file + scale charges, but this results in a large xfield file and the seward section takes extremely long to finish.\

andrewshyichuk

Member

Registered: 2020-02-13

Posts: 86

Re: Compute the effect of Xfield on previous JobMix

I think you'd need a new run of Seward anyway (with the XField).
Then, likely, repeat the PT2 part with the RasOrbs (or RASSCF JobIph) from the run without the XField.

As for the long Seward run - maybe Cholesky might help?

vkj

Member

Registered: 2019-04-10

Posts: 29

Re: Compute the effect of Xfield on previous JobMix

I am already using RICD, and that is not the rate determining step but its the PT2 part which adds up if i want it for hundreds of Xfields config.  I would say my more specific question is if I want to keep the  wavefunction (CI cofficienets/and orbitalas fixed which i have prev. computed and is an input now)  fixed and just want the effect of changed Xfield on this "frozen wavefuction" , is it possible to do it very quickly without having to repeat RASSCF/PT2 over and over.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Compute the effect of Xfield on previous JobMix

What exactly is the effect you expect if you don't allow the wavefunction to change? Do you simply want the electrostatic interaction energy between your "frozen electron density" and an arbitrary external charge distribution (i.e. without allowing the electrons to relax in the presence of the external charges)? In that case, I guess what you want is EPOT instead of XFIELD. But you may need to add PROP to your CASPT2, and maybe you can just save the natural orbitals and compute the EPOT with SEWARD only.

vkj

Member

Registered: 2019-04-10

Posts: 29

Re: Compute the effect of Xfield on previous JobMix

Just to avoid any xy problem issues i will state my intention more clearly. Well my ultimate goal is to have accurate energies of the QM part in a statistical distributions of MM(described by point charges) part. For accurate statistics I should ideally do 100~500 snapshots which will be too many to single-points CASPT2. I am assuming  the different  charge distributions should not change the wave function too much amongst each other. Rather the changes between them could be described by just a "perturbation" effect to a reference wavefunction.

I will compute the reference in a certain charge configuration and then for all other charge configurations I hoped, I can reasonably get good enough energies, if I just compute the effect of new charge configuration to the frozen reference wave function. Indeed i think it should be just the electrostatic energy. I will try the recipes  mentioned by Ignacio.