Molcas Forum

Jacopo

Member

From: Siena

Registered: 2024-01-24

Posts: 2

Website

MCLR function of OpenMolcas 24.02

Hi everyone,

being interested in optimizations and QM/MM trajectories, I was studying the output of &MCLR module called by &ALASKA (version 24.02). I see there is a “properties” section which is quite the same of the “properties” in the “&RASSCF” and also in the “&CASPT2” modules. At first, I will narrow the discussion to a comparison between &RASSCF and &MCLR since it is sufficient to correctly state the question. Suppose you have to do a 2 roots optimization with rlxroot = 2.

Whereas in &RASSCF output one will find:

      Expectation values of various properties for root number:  1
      -----------------------------------------------------------
          
       
++    Molecular properties:     
      ---------------------
…properties and their values…

In &MCLR called by &ALASKA one will simply find one state, without any root specified:

++    Molecular properties:     
      ---------------------
…properties and their values…

But the values of the properties of the latter do not match neither the 1st, nor the 2nd root. However, they are very similar to that of the 2nd (this I understand why, being ALASKA referring to the 2nd thanks to the rlxroot = 2).

If I now compare the &CASPT2 output provided with the GRDT keyword, but neither the PROPerties nor the DENSity ones, I see that it prints updated properties at the CASPT2 level only for that very state indicated by rlxroot, but again the following &MCLR following it presents values which are different from both RASSCF and CASPT2. In this latter case, both &CASPT2 and &MCLR outputs are again in the shape I posted via the last chunk of code: no root specified, on the contrary of &RASSCF.

Any clue on what wavefunction is &MCLR is referring to? I wonder if I can get some useful information in the future by exploiting it. Thank a lot to everyone interested, have a great day.

Sincerely,
Jacopo

---

Computational researcher on light-responsive biological systems, exploiting OpenMolcas and Gromacs as QM and MD tools.

Posted description and CV in my LinkedIn profile: https://www.linkedin.com/in/jacopo-d-ascenzi-213349124/

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: MCLR function of OpenMolcas 24.02

As far as I know, the properties printed in RASSCF are expectation value of the corresponding operators, while those printed in MCLR are energy derivatives (possibly numerical) with respect to an external field. with CASPT2, since there is no direct access to the 1st order perturbed wave function, the properties are only approximate (and there are reports and indications that the DENSITY keyword is buggy). MCLR after CASTP2 is implemented/tested for nuclear gradients, I wouldn't be surprised if the properties there end up being just a fancy random number generator...