Molcas Forum

Adrian

Member

Registered: 2020-07-05

Posts: 4

Nonadiabatic couplings (NAC) with MCPDft wave function

Dear all,
I was trying to calculate NACs in a MCPDFT run on top of a state-average CASSCF (3 roots, no symmetry).
At first, it looks like Alaska is ok (the keyword "Grad" is in the &MCPDFT input block).
But then, after the CI components of the derivative couplings are printed, it crashes all the way, before the CSF components are evaluated.
Here is what happens:

()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()

                                              &ALASKA

                                   only a single process is used
                       available to each process: 2.0 GB of memory, 1 thread?
                                             pid: 15761
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()


 ...
 ...
 ...

 Skipping Nuclear Charge Contribution

 DFT contribution computed for a moving grid.

 Conventional ERI gradients!

 Wavefunction type: MCPDFT  
 Functional type:   TPBE            


 *****************************************************
 *                                                   *
 *              CI derivative coupling               *
 *                                                   *
 *****************************************************

  Irreducible representation: a  
 ------------------------------------------------------------------------------------------
                              X                       Y                       Z
 ------------------------------------------------------------------------------------------
  O1                6.39045554437242E+00   -1.27584389562457E+00   -2.54265834400381E-13
  O2               -6.39045551335866E+00   -1.27584383820024E+00   -5.30817844048835E-13
  LI3              -3.10137459469353E-08    2.55168773382483E+00    7.85009314885681E-13
 ------------------------------------------------------------------------------------------

 ###############################################################################
 ###############################################################################
 ###                                                                         ###
 ###                                                                         ###
 ###    Location: get_dArray                                                 ###
 ###                                                                         ###
 ###                                                                         ###
 ###    Data not defined: D1ao-                                              ###
 ###                                                                         ###
 ###                                                                         ###
 ###############################################################################
 ###############################################################################
[ process      0]: xquit (rc =    128): _INTERNAL_ERROR_

Program aborted. Backtrace:
#0  0x103da882e
#1  0x103da9048
#2  0x103ed1323
#3  0x10313a99a
#4  0x10313aa83
#5  0x103135106
#6  0x1030ed328
#7  0x102f4d672
#8  0x102f09cbb
#9  0x102f0d065
#10  0x1031c8da8
--- Stop Module: alaska at Mon Feb  8 16:41:33 2021 /rc=_RC_INTERNAL_ERROR_ ---
--- Module alaska spent 1 minute 30 seconds ---

.########################################################.
.# Some internal inconsistency of the code was detected #.
.########################################################.

The error "Data not defined: D1ao-" is quite puzzling to me. Searching in the source, I kind of understood that it has something to do with reading from an old density matrix...

First of all: is it possible, to your knowledge, to use MCPDFT for my intent (i.e. NACs calculation)?
Then: anyone knows why Alaska does prompt such an error, and/or how to handle it?

For the sake of completeness, I post the relevant part of my input

&GATEWAY
 Coord = mycoord.xyz
 Basis = ANO-L-VTZP
 Group = nosym
 NoCD

&SEWARD
 Grid input
  Grid=ultrafine
  end of grid input

&RASSCF
 JobIph
 Symmetry = 1
 Spin = 1
 Charge = -1
 Nactel = 10
 Ras2
  7
 CIRoot
  3 3 1

&MCPDFT
 KSDFT = tPBE
 Grad

&ALASKA
 NAC = 1 3

Thank you very much, in advance.
Adrian

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,204

Re: Nonadiabatic couplings (NAC) with MCPDft wave function

As far as I know, NACs are not implemented for MCPDFT.

Adrian

Member

Registered: 2020-07-05

Posts: 4

Re: Nonadiabatic couplings (NAC) with MCPDft wave function

Thank you Ignacio.

Just to understand...
I noticed that Alaska has a "NOCSF" option with skip the evaluation of the CSF contributions. If I include it, the calculation runs without error as I get only the CI contributions to the NACs:

 *****************************************************
 *                                                   *
 *              CI derivative coupling               *
 *                                                   *
 *****************************************************

  Irreducible representation: a  
 ------------------------------------------------------------------------------------------
                              X                       Y                       Z
 ------------------------------------------------------------------------------------------
  O1                6.38162890253168E+00   -1.23806628216010E+00    5.09786269110132E-12
  O2               -6.39487625662683E+00   -1.31497925876010E+00   -1.30559556328576E-12
  LI3               1.32473540951565E-02    2.55304554092020E+00   -3.79215600524557E-12
 ------------------------------------------------------------------------------------------

               Energy difference:  5.377167E-01

 *******************************************************
 *                                                     *
 *              Total derivative coupling              *
 *                                                     *
 *******************************************************

  Irreducible representation: a  
 ------------------------------------------------------------------------------------------
                              X                       Y                       Z
 ------------------------------------------------------------------------------------------
  O1                1.18680135479579E+01   -2.30245093131580E+00    9.48057375439327E-12
  O2               -1.18926498563796E+01   -2.44548717836871E+00   -2.42803617538495E-12
  LI3               2.46363084217481E-02    4.74793810968452E+00   -7.05233092265360E-12
 ------------------------------------------------------------------------------------------

               norm:      17.7795
Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL
--- Stop Module: alaska at Wed Feb 10 12:14:35 2021 /rc=_RC_ALL_IS_WELL_ ---

They are very different from the RASSCF-only NACs, so I would exclude that Alaska has simply used the RASSCF wave function disregarding the on-top MCPDFT correction...
So, to your opinion, these numbers may be just "trash"?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,204

Re: Nonadiabatic couplings (NAC) with MCPDft wave function

I would imagine it's trash, but you may find more information in https://comp.chem.umn.edu/openmolcas/