Molcas Forum

PIjush Karak

Member

Registered: 2022-04-18

Posts: 10

Nonadiabatic coupling calculation using CASSCF and CASPT2 method

Dear Developers,

I want to compute the nonadiabatic coupling using the CASSCF and CASPT2 method with the help of ALASKA module. But after the successful completion of the CASPT2 section, there is an error regarding the computation of the NAC part. Numerical nonadiabatic coupling not implemented error is shown in the ALASKA module part. I have also attached my input file for your better understanding.

&GATEWAY
  coord=test.xyz
  basis = STO-3G
  group = NOSYM

&SEWARD
&SCF

&RASSCF
  Nactel= 14 0 0
  Inactive= 22
  Ras2= 10
  SPIN=1
  CiRoot= 3 3 1
  Rlxroot=3

&GRID_IT
All

&CASPT2
Multistate= 3 1 2 3

&ALASKA
 NAC = 3 2

I will be very grateful for your kind reply.

with regrads,
Pijush

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Nonadiabatic coupling calculation using CASSCF and CASPT2 method

Nonadiabatic couplings are not available with CASPT2 (yet)

PIjush Karak

Member

Registered: 2022-04-18

Posts: 10

Re: Nonadiabatic coupling calculation using CASSCF and CASPT2 method

Thanks for your kind reply and clearing my doubt.

I have another query about how to increase the speed of the above stated input file for calculating the NAC, where caspt2 method is off according to your reply.  Before the ALASKA module, GATEWAY, SEWARD and RASSCF modules are required. I have been using 6-311g** basis using RICD involving the (10e,80) casscf method. I have also run the code parallelly(MPI) using 64 processors with the distribution among 4 nodes. Still, the computation takes huge time, specifically in the two-electron integral calculation, RASSCF iteration and in the MCLR part. So, whether there is any way out to increase the speed of this computation.

with regards,
Pijush

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Nonadiabatic coupling calculation using CASSCF and CASPT2 method

I'd guess 64 processes is overkill, especially if there are only 4 physical nodes. I suggest you start by benchmarking a single iteration with 1 process, and 4 process (one per node).

PIjush Karak

Member

Registered: 2022-04-18

Posts: 10

Re: Nonadiabatic coupling calculation using CASSCF and CASPT2 method

Many thanks Ignacio.

I have run using the 4 process(one process per node) as you stated, and the program is running. But , whether using only 4 process, there will be enough speed to complete this program, and if i want ot use more processors and the node, then what will be the best combination of the processors as well as the nodes, so that the program run efficiently without any exit error.