Author Topic: Excited state geometry optimization with RICC2 (MP2)  (Read 9240 times)

JakubV

  • Full Member
  • ***
  • Posts: 39
  • Karma: +0/-0
Excited state geometry optimization with RICC2 (MP2)
« on: July 25, 2016, 01:50:40 PM »
Hello,

I am trying ground and excited state geometry optimization with RICC2 program for several methods - MP2, CIS(D), CIS(Dinf), ...
I started with MP2.

Code: [Select]
$ricc2
  mp2
  cis(d)
  maxiter=  500
  geoopt model=mp2   state=(x)
$excitations
  irrep=a  multiplicity=  3  nexc=  9  npre=  9  nstart=  9

For a ground state didn't work, but after removing
Code: [Select]
state=(x) it was ok.
I used

Code: [Select]
ridft > ridft.out
rdgrad > rdgrad.out
ricc2 > ricc2.out
jobex -level cc2 -rijk -ri -relax -gcart 4 -c 300
NumForce -level cc2 -rijk -ri -central > force.out

For an excited state I didn't use the "-ex" directive in jobex/NumForce, as I suppose the excited state choice is based on "state =" option in the control file.
I perform the no symmetry calculation, so all irreps are labeled "a" and I don't use "$exopt n" (as for TDDFT/TDHF/CIS). So the only change for excited state is

Code: [Select]
$ricc2
  mp2
  cis(d)
  maxiter=  500
  geoopt model=mp2   state=(a n)
$excitations
  irrep=a  multiplicity=  3  nexc=  9  npre=  9  nstart=  9

where "n" = 1,2,...,9 for the number of the root.

But I got the output with "job.last" ending with

"
cc_parse_states> inconsistency in state input!
 ricc2 ended abnormally
"


and not more than one cycle of optimization was performed for the excited state (n= 1). Where is the mistake in my input, please?

Best regards,
Jakub

JakubV

  • Full Member
  • ***
  • Posts: 39
  • Karma: +0/-0
Re: Excited state geometry optimization with RICC2 (MP2)
« Reply #1 on: July 25, 2016, 01:59:22 PM »
Hello,
I am sorry, I should have included longer part of the error message:

 ========================
  internal module stack:
 ------------------------
    ricc2
    cc_exci
    cc_excipp
 ========================

 cc_parse_states> inconsistency in state input!
 ricc2 ended abnormally
error in gradient step (1)


Also, the MP2 excited states correspond to the CCS (and those are equal to CIS)?

JakubV

Arnim

  • Developers
  • Sr. Member
  • *
  • Posts: 253
  • Karma: +0/-0
Re: Excited state geometry optimization with RICC2 (MP2)
« Reply #2 on: July 27, 2016, 04:15:57 PM »
Hi,

when you want to calculate gradients of triplet state the syntax would have to be like this.
$ricc2
  cc2
  geoopt model=cc2   state=(a{3} n)

MP2 will not yield any excited state results. You can use CCS, CIS(DINF), ADC(2), and CC2.
CCS excitation energies are equal to CIS, the properties could be slightly different. But you should look the details up in the literature.

Also, why do you run rdgrad? This will calculate a RI-HF gradient.

Best regards,

Arnim

christof.haettig

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 291
  • Karma: +0/-0
    • Hattig's Group at the RUB
Re: Excited state geometry optimization with RICC2 (MP2)
« Reply #3 on: October 06, 2016, 06:01:46 PM »
Just to make it clear:

   1) there is no MP2 for excited states ...

   2) gradients for CIS(D) are not implemented in ricc2

If you want to optimize geometries for excited states you can choose between CIS(D_inf), ADC(2) and CC2 and the SOS and SCS variants of them.
ADC(2) is for optimizations somewhat cheaper than CC2 but usually as accurate. The cheapest correlated method for ex. state geo. opts. is SOS-ADC(2) with the
Laplace transformation switched on (note: no SMP parallelization, only MPI).

Christof