How to set up an input file that uses the PBE0 functional: my version runs extremely slowly.

Nat

unread,

Feb 26, 2020, 12:36:59 PM2/26/20

to cp2k

Dear all:

I would like to use CP2K to run NPT water simulations that use different functionals. I have so far used CP2K tutorials to set up simulations based on the PBE and PBE0 functionals. The simulations that used the PBE functional ran quite efficiently and I'm happy with the results. The problem is that the simulations based on PBE0 are running so inefficiently that I must have made a mistake somewhere in my input file (i.e., it will take decades for the simulations to finish). If anyone out there is experienced in using PBE0, I would be very grateful for any advice.

I used this tutorial to set up the simulation.

https://www.cp2k.org/exercises:2017_uzh_cp2k-tutorial:hybrid

Here is my input file. I have colored the exchange–correlation functional part blue. If anything looks a little suspect, it probably is because I don't really know what I'm doing as far as optimizing the solver settings or choosing the best PBE0 specifications.

&GLOBAL
PROJECT pbe0
RUN_TYPE MD
&END GLOBAL

&FORCE_EVAL
STRESS_TENSOR ANALYTICAL
METHOD Quickstep
&DFT
    BASIS_SET_FILE_NAME BASIS_MOLOPT
    POTENTIAL_FILE_NAME POTENTIAL_SET
    CHARGE 0
    MULTIPLICITY 1
    &MGRID
       CUTOFF [Ry] 800
       REL_CUTOFF [Ry] 60
    &END
    &QS
       METHOD GPW
       EPS_DEFAULT 1.0E-10
       EXTRAPOLATION ASPC
       EXTRAPOLATION_ORDER 4
    &END QS
    &POISSON
       PERIODIC XYZ
    &END
    &PRINT
       &E_DENSITY_CUBE OFF
       &END E_DENSITY_CUBE
       &MO_CUBES
          NLUMO 4
          NHOMO 4
          WRITE_CUBE .FALSE.
          &EACH
            MD 10
          &END
       &END
    &END
    &SCF
      SCF_GUESS ATOMIC
      MAX_SCF 30
      EPS_SCF 1.0E-6
      &OT
        PRECONDITIONER FULL_SINGLE_INVERSE
        MINIMIZER DIIS
      &END OT
      &OUTER_SCF
        MAX_SCF 10
        EPS_SCF 1.0E-6
      &END
      &PRINT
        &RESTART OFF
        &END
      &END
    &END SCF

    ! Exchange-correlation approach
    &XC
      ! Use PBE0
      &XC_FUNCTIONAL
       &PBE
         ! 75% GGA exchange
         SCALE_X 0.75
         ! 100% GGA correlation
         SCALE_C 1.0
       &END PBE
      &END XC_FUNCTIONAL
      &XC_GRID
        XC_DERIV SPLINE2_SMOOTH
      &END XC_GRID
      &HF
        ! 25 % HFX exchange
        FRACTION 0.25
        &SCREENING
          EPS_SCHWARZ 1.0E-6
          SCREEN_ON_INITIAL_P TRUE
        &END
        &INTERACTION_POTENTIAL
          POTENTIAL_TYPE TRUNCATED
          CUTOFF_RADIUS 6.0
          T_C_G_DATA ../t_c_g.dat
        &END
        &MEMORY
          MAX_MEMORY 4000
          EPS_STORAGE_SCALING 0.1
        &END MEMORY
      &END HF
      &VDW_POTENTIAL
         POTENTIAL_TYPE PAIR_POTENTIAL
         &PAIR_POTENTIAL
            PARAMETER_FILE_NAME dftd3.dat
            TYPE DFTD3
            REFERENCE_FUNCTIONAL PBE0
            R_CUTOFF [angstrom] 16
         &END
      &END VDW_POTENTIAL
    &END XC
&END DFT

&SUBSYS
    &CELL
      ABC 13.66 13.66 13.66
    &END CELL
    &COORD
      O    -7.6415350291705231E+00    6.0494971538864633E+00    8.1919697993062783E+00 H2O
      H    -6.6418362232982036E+00    5.9231018862335736E+00    8.4056822626120979E+00 H2O
      H    -8.0518670684031459E+00    6.3158684817228137E+00    9.0061271153555484E+00 H2O
      O    2.6825296782034535E+01    2.0466107496669252E+01    1.7802574406580192E+01 H2O
      H    2.6447741120699803E+01    2.0407102905755600E+01    1.6912157191231316E+01 H2O
      H    2.6439964847387159E+01    2.1295072206802320E+01    1.8100927322697789E+01 H2O

! I have omitted most of the atoms for clarity. I have 300.

    &END COORD
    &TOPOLOGY
      CONNECTIVITY GENERATE
      &GENERATE
        BONDLENGTH_MAX 7
      &END
    &END
    &KIND H
      BASIS_SET DZVP-MOLOPT-SR-GTH
      POTENTIAL GTH-NLCC-PBE-q1
    &END KIND
    &KIND O
      BASIS_SET DZVP-MOLOPT-SR-GTH
      POTENTIAL GTH-NLCC-PBE-q6
    &END KIND
&END SUBSYS
&END FORCE_EVAL

&MOTION
&MD
   ENSEMBLE NPT_I
   TEMPERATURE [K] 400.
   TIMESTEP [fs] 0.5
   STEPS 8000000
   &THERMOSTAT
     &NOSE
       LENGTH 3
       YOSHIDA 3
       TIMECON 1000
       MTS 2
     &END NOSE
   &END THERMOSTAT
   &BAROSTAT
     PRESSURE 1000.
   &END BAROSTAT
&END MD
&PRINT
    &CELL
    &END
    &RESTART_HISTORY OFF
    &END
&END PRINT
&END MOTION

Thank you.

Krack Matthias (PSI)

unread,

Feb 26, 2020, 12:57:33 PM2/26/20

to cp...@googlegroups.com

Hi

PBE0 is indeed much more expensive than PBE, especially with MOLOPT basis sets if you do not take advantage of the ADMM implementation (4^th task in the tutorial).

Matthias

--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/72a8b76f-1c3b-4ae3-b09a-1f72da1a036d%40googlegroups.com.

Nat

unread,

Feb 26, 2020, 1:09:06 PM2/26/20

to cp...@googlegroups.com

Dear Matthias,

Thank you very much for your reply. I saw that section in the tutorial and didn't understand the significance of it. If you have time, do you know exactly where in the input file to put these lines (below) from the ADMM section? Should anywhere in the &XC section be OK?

I knew a PBE0 simulation would take longer than PBE, but I'm very surprised to find it's so much longer. I think I would have to live to 100 for my current simulation to finish! Thank you.

    ! use ADMM
    &AUXILIARY_DENSITY_MATRIX_METHOD
      ! recommended, i.e. use a smaller basis for HFX
      ! each kind will need an AUX_FIT_BASIS_SET.
      METHOD BASIS_PROJECTION
      ! recommended, this method is stable and allows for MD. 
      ! can be expensive for large systems
      ADMM_PURIFICATION_METHOD MO_DIAG
    &EN

Krack Matthias (PSI)

unread,

Feb 26, 2020, 4:13:04 PM2/26/20

to cp...@googlegroups.com

Hi

Check CP2K_INPUT/FORCE_EVAL/DFT/AUXILIARY_DENSITY_MATRIX_METHODand have a look at

https://manual.cp2k.org/cp2k-7_1-branch/CP2K_INPUT/FORCE_EVAL/SUBSYS/KIND.html#BASIS_SET and

https://manual.cp2k.org/cp2k-7_1-branch/CP2K_INPUT/FORCE_EVAL/DFT.html#list_BASIS_SET_FILE_NAME or

as well to define the auxiliary basis set using BASIS_SET AUX <FIT…>

Matthias

Von: cp...@googlegroups.com <cp...@googlegroups.com> Im Auftrag von Nat
Gesendet: Mittwoch, 26. Februar 2020 19:09
An: cp2k <cp...@googlegroups.com>
Betreff: Re: [CP2K:12930] How to set up an input file that uses the PBE0 functional: my version runs extremely slowly.

Dear Matthias,

Thank you very much for your reply. I saw that section in the tutorial and didn't understand the significance of it. If you have time, do you exactly where in the input file to put these lines (below) from the ADMM section? Should anywhere in the &XC section be OK?

I knew a PBE0 simulation would take longer than PBE, but I'm very surprised to find it's so much longer. I think I would have to live to 100 for my current simulation to finish! Thank you.

    ! use ADMM

    &AUXILIARY_DENSITY_MATRIX_METHOD

      ! recommended, i.e. use a smaller basis for HFX

      ! each kind will need an AUX_FIT_BASIS_SET.

      METHOD BASIS_PROJECTION

      ! recommended, this method is stable and allows for MD.

      ! can be expensive for large systems

      ADMM_PURIFICATION_METHOD MO_DIAG

&EN

--

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.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/84c8c781-102b-45d7-888b-377e0930690c%40googlegroups.com.

Nat

unread,

Feb 26, 2020, 4:15:39 PM2/26/20

to cp2k

Thank you, Matthias.

Reply all

Reply to author

Forward