Molcas Forum

nhlist

Member

Registered: 2019-11-30

Posts: 1

XMS-CASPT2 and RASSI

Dear Molcas community,

I have a question regarding XMS-CASPT2 and subsequent RASSI calculations.

I have tried to compute XMS-CASPT2 states and subsequent RASSI calculations to get oscillator strengths across a series of interpolated structures (connecting a minimum and a CI, both endpoints computed at the same level of theory).

While the energies appear smooth, the oscillator strengths make unexpected jumps. If I switch to MS-CASPT2, I do not observe any jumps.

In addition to a possible input error, I was wondering if XMS-CASPT2 and RASSI are compatible, e.g., is the XMS zeroth-order basis correctly accounted for by RASSI?

Thank you very much!

Nanna H. List

>> export MOLCAS_MOLDEN=ON
&GATEWAY
 coord = geom.xyz
 Group = NoSym
 Basis = CC-PVDZ
 RICD
&SEWARD
>>> COPY valence_rasorb.RasOrb INPORB
&RASSCF
 LUMORB
 nactel = 14 0 0
 inactive = 12
 ras2 = 12
 ciroot = 3 3 1
 cholesky
 cimx = 200
 iter = 200 50
&CASPT2
 maxiter = 300
 xmul = 3 1 2 3
 shift = 0.3
 frozen = 0
>>COPY $Project.JobMix JOB001
&RASSI
NROFJOBIPHS
1 3
1 2 3
EJob
DIPR
1.0e-8

stefabat

Member

Registered: 2019-12-02

Posts: 12

Re: XMS-CASPT2 and RASSI

Dear Nanna,

In principle there is no problem in using RASSI after a XMS-CASPT2 calculation.
The odd behavior observed is due to a bug concerning the XMS-CASPT2 code: the wave function printout after the result section is actually wrong, as is the wave function passed over to RASSI (i.e. the one in the jobiph/jobmix or h5 file).

I have fixed the bug, however it is still not available in the master branch of the official repository.
For the time being, you can access the working code on my public fork of OpenMolcas:

https://gitlab.com/stefabat/OpenMolcas.git

by checking out the xdw-caspt2 branch.

We will try to merge this branch into the main repository as soon as possible.

Best regards,
Stefano

stefabat

Member

Registered: 2019-12-02

Posts: 12

Re: XMS-CASPT2 and RASSI

Hi,

I just wanted to say that finally the bugfix for XMS-CASPT2 has been merged into the master branch of OpenMolcas, along with the new XDW-CASPT2 method.

Best
Stefano