Molcas Forum

chipmunk

Member

Registered: 2015-11-12

Posts: 23

[SOLVED] Vibrational frequencies at CASPT2 level of optimization?

Hi all,

I found something from Molcas BillBoard about CASPT2 frequencies. I wonder if we can use MCKINLEY to obtain vibrational frequencies in CASPT2 optimization calculations or use the bellow tip?

Thanks a lot.

Hi there, jhkim!
The expert here is Roland, but he was mistaken about the test input.
Now he is travelling, so I thoght I would be able to give at least an
amateurish answer: I concocted an input, which I submit below.

This input works reasonably well on our development version, but also
on 6.4. Howevere, it is far from ideal.

I have first used a similar input to optimize geometries. Now, I use the
optimized geometries in a new run, and tell Slapaf to evaluate numerical
force constants, and to do 0 iterations in the geometry optimization. The
report from the numerical force constant calculation includes also frequencies.

Why is this far from ideal??
Because this facility in the Slapaf was intended only to give a good starting
Hessian for a subsequent geometry optimization. This way of running works, but
the output becomes a bit confusing. Ignore that, and just search the output for
the word "frequencies"! Also ignore *small* differences in the result you get from
those quoted in the input.
Also, this way of computing them can be slow -- depends entirely on your problem,
but grows quadratically with number of atoms.

Things are a bit better on the development version, and will be made more
convenient in Molcas 7. Anyway, here is the input I ran on 6.4:

>>>>>>>>>>>>>>>>>>> DO while <<<<<<<<<<<<<<<<<<<<
&SEWARD &END
Title
Water, OH=0.972815 A, HOH=104.32 deg, ANO-L 3s2p1d/2s1p
Symmetry
X
Basis set
H.ANO-L...2s1p.
H1 0.000 1.4517697852 -1.1277891106
H2 0.000 -1.4517697852 -1.1277891106
End of basis
Basis set
O.ANO-L...3s2p1d.
O 0.000 0.000 0.000
End of basis
End of input
*-------------------------------------------------------------------------------
>>>>>>>>>>> if ( iter = 1 ) <<<<<<<<<<<<<<
&SCF &END
Title
Water
OCCUPIED
4 1
End of Input
>>>>>>>>> endif <<<<<<<<<<<<<<<<<<<<<<<<
*-------------------------------------------------------------------------------
!ln -s $Project.ScfOrb INPORB
&RASSCF &END
Title
Water Cs Act orb 6,2 Act el 8
Energy: -76.1759831786 (CASSCF)
NACTEL
8 0 0
Inactive
1 0
Ras2
6 2
ITER
100 100
LUMORB
End of Input
*-------------------------------------------------------------------------------
&CASPT2 &End
* Water
* Energy: -76.2639685877 (CASPT2)
MaxIterations
20
End of Input
*-------------------------------------------------------------------------------
&Alaska &End
End of Input
*-------------------------------------------------------------------------------
&Slapaf &End
* Water
* Harm freq (cm-1) : 1623.06 3785.49 3917.13
* IR inten (km/mol): 70.6 23.5 184.0
Numerical Hessian
Iterations
0
End of Input
>>>>>>>>>>>>> ENDDO <<<<<<<<<<<<<<<<<<<<<<<<<<<<<
*-------------------------------------------------------------------------------

Last edited by chipmunk (2016-08-02 11:12:59)

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: [SOLVED] Vibrational frequencies at CASPT2 level of optimization?

Yes, you can simply add &MCKINLEY in your input to compute frequencies (fully numerically at CASPT2 level). It will call &SLAPAF under the hood as in the above sample. To avoid trouble, do not include this inside a loop:

&GATEWAY
  ...
&SEWARD
  ...
&RASSCF
  ...
&CASPT2
  ...
&MCKINLEY
  ...

chipmunk

Member

Registered: 2015-11-12

Posts: 23

Re: [SOLVED] Vibrational frequencies at CASPT2 level of optimization?

Yes, you can simply add &MCKINLEY in your input to compute frequencies (fully numerically at CASPT2 level). It will call &SLAPAF under the hood as in the above sample. To avoid trouble, do not include this inside a loop:

&GATEWAY
  ...
&SEWARD
  ...
&RASSCF
  ...
&CASPT2
  ...
&MCKINLEY
  ...

Thanks lots for your quick reply.

Insofar as I understand, we need to perform a second calculation based on the optimized geometry and orbitals. That's correct?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: [SOLVED] Vibrational frequencies at CASPT2 level of optimization?

That's not strictly needed, you could also add &MCKINLEY directly after the optimization (after the ">> EndDo"). But since the numerical differentiation will involve many more calculations, having an extra one at the beginning for the reference geometry does probably not represent a significant overhead, and will help having things cleaner (in my opinion).

chipmunk

Member

Registered: 2015-11-12

Posts: 23

Re: [SOLVED] Vibrational frequencies at CASPT2 level of optimization?

Thanks a lot Ignacio.

Have nice time.