Molcas Forum

panwang

Member

Registered: 2020-02-19

Posts: 6

unsmooth MS-CASPT2 curver near the CI

Hello, everyone.

I perform a linearly interpolated internal coordinate scan between SA2-CASSCF(10,10) optimzied S1 and MECI geometries with MS-CASPT2 and CPCM solvation model. All the calculations finished normally. However, the obtained curver is unsmooth near the CI point. Especially, the ground state curver is smooth for all the points, and energies of last two points in the S1 state had a sudden change. After carefully check the output, I found in the output file of the last two points, there is "RF Root has flipped from  2  to  1" in CASSCF calculations after some iterations where the RFRoot has been set to 2 in the input file. So, is there a way to eliminate "RF Root has flipped from  2  to  1" to makes the curver more smooth. Thanks.

stefabat

Member

Registered: 2019-12-02

Posts: 12

Re: unsmooth MS-CASPT2 curver near the CI

I do not have experience with CPCM thus I cannot exclude that the issue is only due to CPCM. However, it is known that MS-CASPT2 might give you non-smooth potential energy surfaces around a CI (see e.g. [1]).
If the sudden jump you are observing is due to MS-CASPT2 (try to run your calculation without CPCM and see if the same happens again), then you might need to switch to another method: either XMS-CASPT2 [2] or the new XDW-CASPT2 [3].

To use XMS-CASPT2 in OpenMolcas, just change "MULTistate" to "XMULtistate" in your CASPT2 input and that should do it. Note that at the moment the CI coefficients printed out after the (extended) multistate section are wrong and so is the wave function passed over to RASSI (but the calculations is correct and the energies are also correct).
In order to get the correct wave function or use the new XDW-CASPT2 method, please have a look at this other thread:
https://molcasforum.univie.ac.at/viewtopic.php?id=568

Cheers
Stefano

[1] Granovsky, A. A. Extended Multi-Configuration Quasi-Degenerate Perturbation Theory: The New Approach to Multi-State Multi-Reference Perturbation Theory. J. Chem. Phys. 2011, 134 (21), 214113. https://doi.org/10.1063/1.3596699.
[2] Shiozaki, T.; Győrffy, W.; Celani, P.; Werner, H.-J. Communication: Extended Multi-State Complete Active Space Second-Order Perturbation Theory: Energy and Nuclear Gradients. J. Chem. Phys. 2011, 135 (8), 081106. https://doi.org/10.1063/1.3633329.
[3] https://pubs.acs.org/doi/10.1021/acs.jctc.9b01129

panwang

Member

Registered: 2020-02-19

Posts: 6

Re: unsmooth MS-CASPT2 curver near the CI

Thanks. It seems that the CPCM is not implemented completely for XMS-CASPT2 because the obatained results by XMC-CASPT2//PCM and MS-CASPT2//PCM are the same.
I will try to recompile OpenMolcas to include XDW-CASPT2 and to perform XDW-CASPT2 calculations. By the way, dose the CPCM has been implemented for XDW-CASPT2?

stefabat

Member

Registered: 2019-12-02

Posts: 12

Re: unsmooth MS-CASPT2 curver near the CI

I am not sure what you mean by the results of XMS-CASPT2//PCM and MS-CASPT2/PCM are the same. You mean with or without PCM are the same, but different between MS and XMS? In principles, MS and XMS are always different.
Perhaps it's easier if you post here your input.

Regarding the implementation of CPCM in the CASPT2 module, I have to take a look at the code before I can answer that.

panwang

Member

Registered: 2020-02-19

Posts: 6

Re: unsmooth MS-CASPT2 curver near the CI

I mean that without PCM, MS and XMS are different; with PCM, the obtained results by MS and XMS are the same.

The input file for the MS-CASPT2 is following (for XMS-CASPT2, the XMultistate is used):

&GATEWAY
  RICD
  coord = DIII-S0.xyz
  basis = 6-31g*
  group = C1
  RF-Input
   PCM-Model
   Solvent=acetonitrile
   CONDuctor version
  End of RF-Input
&SEWARD
&SCF
&RASSCF
 LumOrb
 ALTEr=1; 1 80 81
 Inactive=70
 NACTEL=10
 RAS2=10
 CIRoot=2 2 1
 RFRoot=1
&CASPT2
 Multistate= 1 1
 RFPert
&RASSCF
 LumOrb
 Inactive=70
 NACTEL=10
 RAS2=10
 CIRoot=2 2 1
 RFRoot=2
 NONEquilibrium
&CASPT2
 Multistate= 1 2
 RFPert

stefabat

Member

Registered: 2019-12-02

Posts: 12

Re: unsmooth MS-CASPT2 curver near the CI

Thanks for the input, I will look into this and come back with a reply as soon as possible.