Periodic DFT calculation with B3LYP

[Abhishek Bagusetty]

Hi All,

I am trying to do a single point energy calculation of a graphene based system using B3LYP hybrid functional and have been constantly getting this warning. Although the WARNING is pretty descriptive but any corrections to the concerned section did not yield any fruitful results. 

A quick summary of the system and calculations - Graphene based system with PERIODIC in XYZ. Single Point Energy calculations are performed with B3LYP level of theory. I have looked into the test cases for hybrid functions but haven't made any progress with that. 

I have attached the input and output file and any pointers related to this issue is greatly appreciated. 

*** 23:45:43 WARNING in hfx_types:hfx_create_neighbor_cells :: Periodic   ***

*** Hartree Fock calculation requested without use of a truncated or      ***

*** shortrange potential. This may lead to unphysical total energies. Use ***

*** a truncated  potential to avoid possible problems. hfx_types.F line   ***

*** 1436                                                                  ***

Thanks !

Abhishek

[Rolf David]

Hi,

the commented part 

      # &INTERACTION_POTENTIAL

      #  POTENTIAL_TYPE SHORTRANGE

      #  OMEGA 0.33

      #  CUTOFF_RADIUS 0.11

      #  T_C_G_DATA /opt/pkg/cp2k/2.6/dist/cp2k/data/t_c_g.dat

      # &END INTERACTION_POTENTIAL

needs to be looked.

In your input, you don't have as said shortrange or truncated, you need to uncomment this part and put it into the HF

Or you can use a truncated one

    &INTERACTION_POTENTIAL

     POTENTIAL_TYPE TRUNCATED

     CUTOFF_RADIUS 6.0 (Need to be tuned and <=Lcell/2)

     T_C_G_DATA /PATH/TO/t_c_g.dat

    &END INTERACTION_POTENTIAL

[Abhishek Bagusetty]

Hi Rolf,

I have used the truncated interaction potential and that has addressed the WARNING but the issue currently I am facing is that the calculations seems to hang at the SCF wavefunction optimization. 

SCF WAVEFUNCTION OPTIMIZATION

 ----------------------------------- OT ---------------------------------------

 Allowing for rotations:  F

 Optimizing orbital energies:  F

  Minimizer      : DIIS                : direct inversion

                                        in the iterative subspace

                           using      : -   7 DIIS vectors

                                        - safer DIIS on

 Preconditioner : FULL_SINGLE_INVERSE : inversion of H + eS - 2*(Sc)(c^T*H*c+const)(Sc)^T

  Precond_solver : DEFAULT

  stepsize       :    0.08000000

 energy_gap     :    0.08000000

 eps_taylor     :   0.10000E-15

 max_taylor     :             4

  mixed_precision    : F

  ----------------------------------- OT ---------------------------------------

  Step     Update method      Time    Convergence         Total energy    Change

  ------------------------------------------------------------------------------

 Trace(PS):                                  192.000000000

 Electronic density on regular grids:       -191.9999999586        0.0000000414

 Core density on regular grids:              192.9999999888       -0.0000000112

 Total charge density on r-space grids:        1.0000000302

  Total charge density g-space grids:           1.0000000302

  HFX_MEM_INFO| Est. max. program size before HFX [MB's]:                    134

Any ideas.

Thanks !

[Rolf David]

Hi Abhishek

It comes from fact than the MOLOPT basis set is very diffuse, so for the HF exchange is very very long, if I remember correctly.

You can either use the GTH_BASIS or use the ADMM methods ((1) Guidon, M.; Hutter, J.; Vandevondele, J. J. Chem. Theory Comput. 2010, 6, 2348.), but with B3LYP, I have no idea how to use it, or you can go with the GAPW/All-Electron and "classical" basis sets (6-31+G*, for exemple)

Also you can read: 

(1) Guidon, M.; Schiffmann, F.; Hutter, J.; Vandevondele, J. J. Chem. Phys. 2008, 128.

(2) Guidon, M.; Hutter, J.; VandeVondele, J. J. Chem. Theory Comput. 2009, 5, 3010.

I think if you do a search on this googlegroup, you'll find more information

Y

On Friday, June 5, 2015 at 6:19:19 AM UTC+2, Abhishek Bagusetty wrote:

[Abhishek Bagusetty]

Thanks appreciate your help !