TDDFPT using PBE0 in version 9.1 versus in 4.1

[Maria Fumanal]

Dear CP2K community,

I wanted to reproduce the TDDFPT results of a small system obtained with PBE0 with CP2K 4.1 and 9.1,
but I have seen that the TDDFPT input section has changed a lot and I do not manage to make it work to obtain the same results.

with CP2K 4.1 the simplest TDDFPT input I used to use is (see pbe0_41.inp/out):

  &TDDFPT
     NSTATES     3             
     MAX_ITER    2000             
     CONVERGENCE [eV] 1.0e-4    
     &MGRID
        CUTOFF 200  
     &END
  &END TDDFPT

This above does not work with the 9.1 version anymore.
The only way I have manage to make it work without failing is specifying the &XC subsection this way:

  &TDDFPT
     &XC
       &XC_FUNCTIONAL
          &PBE
            SCALE_X 0.75
            SCALE_C 1.0
          &END PBE
       &END XC_FUNCTIONAL
     &END XC
     NSTATES     3             
     MAX_ITER    2000             
     CONVERGENCE [eV] 1.0e-4    
     &MGRID
        CUTOFF 200  # separate cutoff for TDDFPT calc
     &END
  &END TDDFPT

It works but the TDDFPT results I obtained are very far from the ones I get with the older version. (see pbe0_91_XC.inp/out)

If I do not specify this &XC section it fails (see pbe0_91.inp/out, error "derivatives bigger than 1 do not work correctly ")

If I specify the same &XC settings than for the Ground State including the &HF section, &vdW_POTENTIAL, etc it fails (see pbe0_91_XC_HF.inp/out, error "ADMM is not implemented for a TDDFT kernel XC-functional which is different from the one used for the ground-state calculation. A ground-state 'admm_env' cannot be reused.)

Does anyone know why this is happening and how to solve it?

Thanks a lot in advance,

Maria

[Jürg Hutter]

Hi

this input should work with the current Trunk version of CP2K. You can find many
examples in tests/QS/regtests-tddfpt-*
Please note that we recently fixed a bug in the HFX-tddft code.

best, Juerg

&FORCE_EVAL
METHOD Quickstep
&PROPERTIES
&TDDFPT
KERNEL FULL
NSTATES 5
MAX_ITER 50
CONVERGENCE [eV] 1.0e-7
&END TDDFPT
&END PROPERTIES
&DFT
&QS
METHOD GPW
&END QS
&AUXILIARY_DENSITY_MATRIX_METHOD
METHOD basis_projection
ADMM_PURIFICATION_METHOD none
&END
&XC
&XC_FUNCTIONAL PBE0
&END
&END XC
........

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Maria Fumanal <mfum...@gmail.com>
Sent: Tuesday, May 24, 2022 5:22 PM
To: cp2k
Subject: [CP2K:17029] TDDFPT using PBE0 in version 9.1 versus in 4.1

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/a07282d7-2ccf-4847-9213-2ff78a3a09bdn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/a07282d7-2ccf-4847-9213-2ff78a3a09bdn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Maria Fumanal]

Thanks a lot for your answer, this input works well  with 9.1 and gives (almost) same results than with the 4.1 version

I see however that now HF truncation is no longer specified, is it still considered though?

Another question I have is if it would be possible to use Longe Range correction and how to specify it?

Many thanks,

[Jürg Hutter]

Hi

you can specify a truncated potential:

&INTERACTION_POTENTIAL
POTENTIAL_TYPE TRUNCATED
CUTOFF_RADIUS 5.0
T_C_G_DATA t_c_g.dat
&END

The long range correction only makes a significant contribution for very small
cutoff radius. I would not use it.

regards

Juerg Hutter

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Maria Fumanal <mfum...@gmail.com>

Sent: Thursday, May 26, 2022 10:59 AM
To: cp2k
Subject: Re: [CP2K:17041] TDDFPT using PBE0 in version 9.1 versus in 4.1

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/376870b1-e57d-4984-b755-73d05732634dn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/376870b1-e57d-4984-b755-73d05732634dn%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Maria Fumanal]

Hello,

If I add the truncation section to the previous input that works well (pbe0_jh.inp/out) as:

&FORCE_EVAL
METHOD Quickstep
&PROPERTIES
&TDDFPT
KERNEL FULL
NSTATES 5
MAX_ITER 50
CONVERGENCE [eV] 1.0e-7
&END TDDFPT
&END PROPERTIES
&DFT
&QS
METHOD GPW
&END QS
&AUXILIARY_DENSITY_MATRIX_METHOD
METHOD basis_projection
ADMM_PURIFICATION_METHOD none
&END
&XC
&XC_FUNCTIONAL PBE0
&END

&HF

&INTERACTION_POTENTIAL
POTENTIAL_TYPE TRUNCATED
CUTOFF_RADIUS 5.0
T_C_G_DATA t_c_g.dat

&END INTERACTION_POTENTIAL

&END HF

&END XC

it complains with the "derivatives bigger than 1 do not work correctly" error (see attached pbe0_jh_hf.inp/out)

I understood that for periodic systems it is necessary to include this truncation of the HF with cutoff radius half of the smallest cell vector...

Is this different in this version?

Many thanks,

Maria

[Anna Hehn]

Dear Maria

the choice of GGA exchange functionals for the first-order correction term of ADMM is restricted. Rerunning your input choosing EXCH_CORRECTION_FUNC PBEX in the ADMM section however works fine for me. 

Recommendations regarding the cutoff radius for exchange are the same as before; standardly half of the smallest cell vector is a good start. 

Best regards

Anna 

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/0cb361af-307a-4216-ab94-038d6b0de624n%40googlegroups.com.

[Maria Fumanal]

Dear Anna,

Thanks a lot, this solved my problem! Now works perfectly.

Best regards,

Maria