Molcas Forum

freshman1234

Member

Registered: 2017-02-23

Posts: 18

Problem about frequency calculation of SA-CASSCF in Molcas

Can anyone help me in solving the following problem?
I am trying to calculate vibrational frequency of CASSCF optimized geometry of excited state(with CiRoot=8 can keep the active space), but I got the following problem.
*********** My Input ***************

>>COPY  /home/w123/opt/rea-T4-opt.RasOrb INPORB

&Seward
Expert
 Basis set
 H    / inline
          1.   1
* S-type functions
    4    2
             13.0100000
              1.9620000
              0.4446000
              0.1220000
      0.0196850              0.0000000
      0.1379770              0.0000000
      0.4781480              0.0000000
      0.5012400              1.0000000
* P-type functions
    1    1
              0.7270000
      1.0000000
H             -3.949437       1.714578       0.213015 Angstrom

.......

End of basis

&SCF
&RASSCF &END
TITLE
 use scforb
 Spin
   3
 Nactel
 10 0 0
 Ras2
 11
 Inactive
 86
 LumOrb
CiRoot
8 8 1
Rlxroot
 4
&McKinley
Perturbation
Hessian
End of Input

********************************* Problem *********************************

--- Stop Module:  rasscf at Wed Jun  6 19:40:04 2018 /rc=0 ---
--- Module rasscf spent 42 minutes and 20 seconds
--- Start Module: slapaf at Wed Jun  6 19:40:06 2018

  Slapaf requests the computation of gradients first!

--- Stop Module:  slapaf at Wed Jun  6 19:40:06 2018 /rc= _INVOKED_OTHER_MODULE_ ---
--- Start Module: alaska at Wed Jun  6 19:40:07 2018

  Alaska requests MCLR to be run before it starts again!

--- Stop Module:  alaska at Wed Jun  6 19:40:08 2018 /rc= _INVOKED_OTHER_MODULE_ ---
--- Start Module: mclr at Wed Jun  6 19:40:08 2018


()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
                                  MOLCAS executing module MCLR with 50000 MB of memory
                                              at 19:40:08 Wed Jun  6 2018
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()


 Ordinary integral handling
 No .RLXPT2  or MCKINT  , I hope that is OK
 Seward mode is assumed, reading perturbation from ONEINT
 OrdInt status: non-squared

    ******************************************************************************************************
     *                                                                                                    *
     *                                              Project:                                              *
     *                                                                                                    *
     ******************************************************************************************************


     Header of the ONEINT file:
     --------------------------
                         cc-pvdz.
      Integrals generated by seward 4.2.0  , Wed Jun  6 18:09:48 2018

                                     .
                                     .
                                     .
                                     .

--- Stop Module:  mclr at Wed Jun  6 20:45:10 2018 /rc=0 ---
--- Module mclr spent 1 hour and 5 minutes
--- Start Module: alaska at Wed Jun  6 20:45:16 2018


()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
                                 MOLCAS executing module ALASKA with 50000 MB of memory
                                              at 20:45:16 Wed Jun  6 2018
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()


--- Stop Module:  alaska at Wed Jun  6 22:00:25 2018 /rc=0 ---
--- Module alaska spent 1 hour and 15 minutes
--- Module auto spent 2 hours and 20 minutes
--- Start Module: slapaf at Wed Jun  6 22:00:25 2018

************************************************************************************
The work always executes module SEWARD,RASSCF,SLAPAF,ALASKA and MCLR, and doesn't execute module MCKINLEY. This is why?

Last edited by freshman1234 (2018-06-11 03:00:23)

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Problem about frequency calculation of SA-CASSCF in Molcas

That's because there are no analytical second derivatives, so MCKINLEY (you should see it before the first SLAPAF) calls SLAPAF instead, with the "Numerical frequencies" keyword. After that, it's all SLAPAF's doing, computing the gradient for different displaced geometries, and it should print the frequencies when it's done.

freshman1234

Member

Registered: 2017-02-23

Posts: 18

Re: Problem about frequency calculation of SA-CASSCF in Molcas

But when I remove the keyword CiRoot=8 and Rlxroot=4,the job can be done very well, and doesn't do SLAPAF and alaska,do the module as I set(Seward,Rasscf,Mckinley  and Mclr, it is regreated that the active space cannot keep and the excited state is not I wanted).  This is why?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Problem about frequency calculation of SA-CASSCF in Molcas

With CIRoot it's a state-averaged CASSCF. There are analytical frequencies for state-specific CASSCF, but not for state-averaged CASSCF, so with SA-CASSCF you get the numerical algorithm.

freshman1234

Member

Registered: 2017-02-23

Posts: 18

Re: Problem about frequency calculation of SA-CASSCF in Molcas

Thank you very much for your answer. I have looked for "Numerical frequencies" keyword in the manual but failed. Could you please provide me the exact input file for this? I need your assistance. Please help me in this matter.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Problem about frequency calculation of SA-CASSCF in Molcas

You don't need anything special, just call &MCKINLEY and it will do analytical (run MCKINLEY and call MCLR) or numerical (perform many energy + gradient + SLAPAF calculations) frequencies depending on the availability.

freshman1234

Member

Registered: 2017-02-23

Posts: 18

Re: Problem about frequency calculation of SA-CASSCF in Molcas

So there is no error in my input file:

&Gateway &End
&Seward
&SCF
&RASSCF &END
TITLE
 use scforb
 Spin
   3
 Nactel
 10 0 0
 Ras2
 11
 Inactive
 86
 LumOrb
CiRoot
8 8 1
Rlxroot
4
&McKinley
Perturbation
Hessian
End of Input

It calls &MCKINLEY and it will do numerical (perform many energy + gradient + SLAPAF calculations) frequencies .Does this mean that my original optimized geometry is not minimum? And whether the calculated frequency is according to my original optimized geometry?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Problem about frequency calculation of SA-CASSCF in Molcas

It does small displacements to the geometry to do the numerical differentiation, but it's not "optimizing" the initial geometry (the SLAPAF call is there only to generate the list of displaced geometries). Whatever you provide as an input, that's what the frequencies correspond to. Whether your optimized geometry is a minimum or not, depends on the result of the frequencies. If all frequencies are real, it is a minimum (assuming it the optimization converged); if some are imaginary (excepting possible translations/rotations) it is a saddle point.

freshman1234

Member

Registered: 2017-02-23

Posts: 18

Re: Problem about frequency calculation of SA-CASSCF in Molcas

If I set the CiRoot=8 and Rlxroot=4,does the work calculate eight roots' frequencies or just the fourth root's frequency?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Problem about frequency calculation of SA-CASSCF in Molcas

Only one, I believe. But why not try it for a smaller system where you can have your results in a few minutes?