Multiwfn forum

gloria

Member

Registered: 2023-06-21

Posts: 4

Acceleration calculation

Hi all,

I want to use Multiwfn to analysis electron-hole transfer of excitation. I calculated 500 excited states. The analysis process is really slow. Could you tell me any methods to improve analysis speed?( I use no GUI version in supercomputer, I have tried to change nthreads to 24 in setting.ini. It doesn't seem to work.

Thanks in advance for your help.

All the regards,

Gloria

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,419

Website

Re: Acceleration calculation

Dear Gloria,

Please make sure that Multiwfn indeed works in parallel mode with your specified number of threads. It can be seen when Multiwfn just boots up:

 ( Number of parallel threads:  24  Current date: 2023-06-22  Time: 00:09:26 )

Another way is reducing quality of grid, or use IOp(9/40=3) instead of IOp(9/40=4) to reduce the number of configuration coefficients taken into account. I cannot give you more detailed advise without having more information about your system and computational level.

In addition, please always use latest version, parallel efficiency of grid data calculation has been notably improved in the recent update.

Best regards,

Tian

gloria

Member

Registered: 2023-06-21

Posts: 4

Re: Acceleration calculation

Thanks a lot, sobereva! I'm sure I use 24 threads just showing like yours. My system is a heterostructure combined two layers, which has more than 700 atoms. Now I used DZVP and PBE in CP2K to do some tests. High accuracy is needed for my further calculation(eg: HSE06, PBE0) At beginning, I used 0.5 of grid and it was so slow. Now I use 1 for grid and it become much faster. However, it still needs 20-30 mins. What can I do to further improve the speed?

Thanks in advance!

Best regards,
Gloria

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,419

Website

Re: Acceleration calculation

Dear Gloria,

DZVP-GTH or DZVP-MOLOPT-SR-GTH? They are quite different in calculation speed in Multiwfn.

It is best to paste your CP2K input file here (without coordinate part)

Best,

Tian

gloria

Member

Registered: 2023-06-21

Posts: 4

Re: Acceleration calculation

This is my input file. Thanks a lot  for your advice and help!
___________________________________________________________

&FORCE_EVAL
  METHOD Quickstep
  &DFT
    UKS
    BASIS_SET_FILE_NAME ${data}/BASIS_SET
    BASIS_SET_FILE_NAME ${data}/BASIS_MOLOPT
    POTENTIAL_FILE_NAME ${data}/GTH_POTENTIALS
       &MGRID
      CUTOFF 600
      REL_CUTOFF 60
      NGRIDS 4
    &END MGRID
    &QS
      EPS_DEFAULT 1.0E-15
      #      EPS_PGF_ORB 1.0E-06
    &END QS
    &POISSON
     PERIODIC XYZ
    &END POISSON
    &SCF
     MAX_SCF 500
     SCF_GUESS RESTART
     EPS_SCF   5.0E-6
     CHOLESKY INVERSE
     &OT
        PRECONDITIONER FULL_SINGLE_INVERSE
      #  ENERGY_GAP 0.001
        MINIMIZER DIIS
     &END OT
    &PRINT
       &RESTART
         &EACH
           MD 20
         &END EACH
       &END RESTART
     &END PRINT
    &END SCF
    &XC
      &XC_GRID
         XC_DERIV  SPLINE2
         XC_SMOOTH_RHO NN50
    &END XC_GRID
    &XC_FUNCTIONAL PBE
    &END XC_FUNCTIONAL
    &VDW_POTENTIAL
         POTENTIAL_TYPE   NON_LOCAL
         &NON_LOCAL
           CUTOFF 300.
           KERNEL_FILE_NAME ${data}/rVV10_kernel_table.dat
           TYPE RVV10
         &END NON_LOCAL
    &END VDW_POTENTIAL
    &END XC
   &PRINT
     &MO_MOLDEN #Exporting .molden file containing wavefunction information
       NDIGITS 9 #Output orbital coefficients if absolute value is larger than 1E-9
     &END MO_MOLDEN

   &END PRINT
  &END DFT
  &PROPERTIES
    &TDDFPT
      NSTATES   500
      MIN_AMPLITUDE 0.01
      MAX_KV 100
     &END TDDFPT
  &END PROPERTIES

  &SUBSYS
    &KIND C
      ELEMENT   C
      BASIS_SET DZVP-MOLOPT-SR-GTH-q4
      POTENTIAL GTH-PBE
    &END KIND

    &KIND H
      ELEMENT   H
      BASIS_SET DZVP-MOLOPT-SR-GTH-q1
      POTENTIAL GTH-PBE
    &END KIND

    &KIND S
      ELEMENT   S
      BASIS_SET DZVP-MOLOPT-SR-GTH-q6
      POTENTIAL GTH-PBE
    &END KIND
    &CELL
      A 34.76132 0.0 0.0
      B -17.380659999999992 30.10418618908008 0.0
      C 0.0 0.0 33.5796
      MULTIPLE_UNIT_CELL  1 1 1
       PERIODIC XYZ
    &END CELL
    &TOPOLOGY
        COORD_FILE_NAME X.xyz
        COORD_FILE_FORMAT xyz
        MULTIPLE_UNIT_CELL  1 1 1
    &END TOPOLOGY
  &END SUBSYS
&END FORCE_EVAL
&MOTION
  &GEO_OPT
      TYPE MINIMIZATION
      MAX_DR    1.0E-03
      MAX_FORCE 1.0E-03
      RMS_DR    1.0E-03
      RMS_FORCE 1.0E-03
      MAX_ITER 500
      OPTIMIZER BFGS
      &BFGS
#              RESTART_FILE_NAME Cu2O_SM26-BFGS.Hessian
#              RESTART_HESSIAN
      &END BFGS
      #      &CG
      #         MAX_STEEP_STEPS  0
      #         RESTART_LIMIT 9.0E-01
      #      &END CG
   &END GEO_OPT
&END MOTION

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,419

Website

Re: Acceleration calculation

MOLOPT is a fully generally contracted basis set, which is quite inefficient for wavefunction analysis purpose in Multiwfn.

I suggest using pob-DZVP-rev2 or 6-31G* instead, which are segment contracted basis sets, the cost in Multiwfn is significantly lower. If you have to use a GTH pseudopotential basis set, DZVP-GTH or TZVP-GTH can be considered, which are also cheaper in Multiwfn than DZVP-MOLOPT-SR-GTH.

gloria

Member

Registered: 2023-06-21

Posts: 4

Re: Acceleration calculation

Thanks a lot, Tian!
Have a nice day!

Gloria