[CP2K:9497] how to reduce the memory usage for HFX

[S Ling]

Hi

This maybe the largest hybrid DFT calculation for a 3D periodic system that
I have ever heard of.

You may also check whether you really need a CUTOFF_RADIUS of 10 A in your
PBE0-TC-LRC functional. The density matrix of insulators decays
exponentially with SQRT(E_gap)*r12 (see Eqn. 16 of this paper,
http://aip.scitation.org/doi/10.1063/1.2931945). For system with a sizeable
band gap, you can usually get converged electronic properties with a cutoff
radius of 6 A or even smaller. I would suggest you to check the convergence
of your target properties with respect to CUTOFF_RADIUS in smaller cells,
and then choose the CUTOFF_RADIUS that gives the best balance of accuracy
and cost for your calculation on the large supercell. A smaller cutoff
radius may help you to reduce the memory usage of the HFX calculation.

In addition, if you use OT algorithm for SCF, you can usually reduce
the MAX_SCF
parameter in &SCF subsection to 20 or 30, in order to take advanrage of
the preconditioning. This will not reduce the memory usage, but it may help
to accelerate the SCF convergence.

SL



On 27 September 2017 at 15:56, Xiaoming Wang <wxia... at gmail.com> wrote:

> Hi all,
>
> Is there any way to reduce the memory usage for HFX calculations? I am
> dealing with a large scale HFX calculation with the box size of 47*47*47 A
> and 22560 atoms. After testing, I find that 30Gb mem is required for each
> MPI task. I am using the MPI/Openmp version. Following is my input, any
> suggestion on the improvement of the parameters is appreciated.
>
> &GLOBAL
>   PROJECT_NAME mysys
>   RUN_TYPE GEO_OPT
>   PRINT_LEVEL MEDIUM
> &END GLOBAL
> &FORCE_EVAL
>   METHOD QS
>   &DFT
>     CHARGE -1
>     LSD
>     BASIS_SET_FILE_NAME BASIS_MOLOPT
>     BASIS_SET_FILE_NAME BASIS_ADMM_MOLOPT
>     POTENTIAL_FILE_NAME GTH_POTENTIALS
>     WFN_RESTART_FILE_NAME pbe.wfn
>     &QS
>      EPS_PGF_ORB 1.0e-32
>     &END
>     &MGRID
>       CUTOFF 250
>       REL_CUTOFF 50
>     &END MGRID
>     &XC
>       &XC_FUNCTIONAL
>        &PBE
>         SCALE_X 0.75
>         SCALE_C 1.0
>        &END PBE
>        &PBE_HOLE_T_C_LR
>         CUTOFF_RADIUS 10.0
>         SCALE_X 0.25
>        &END PBE_HOLE_T_C_LR
>       &END XC_FUNCTIONAL
>       &HF
>        FRACTION 0.25
>        &SCREENING
>         EPS_SCHWARZ 1.0e-6
>        &END SCREENING
>        &INTERACTION_POTENTIAL
>         POTENTIAL_TYPE TRUNCATED
>         CUTOFF_RADIUS 10.0
>         T_C_G_DATA ./t_c_g.dat
>        &END INTERACTION_POTENTIAL
>        &MEMORY
>         MAX_MEMORY 32000
>         EPS_STORAGE_SCALING 0.1
>        &END MEMORY
>       &END HF
>     &END XC
>     &SCF
>       MAX_SCF 100
>       EPS_SCF 1.0e-6
>       CHOLESKY INVERSE
>       SCF_GUESS RESTART
>       &OT
>        ROTATION
>        PRECONDITIONER FULL_KINETIC
>        ENERGY_GAP 0.001
>       &END OT
>       &OUTER_SCF
>        EPS_SCF 1.0e-6
>        MAX_SCF 10
>       &END OUTER_SCF
>     &END SCF
>     &PRINT
>      &MULLIKEN OFF
>      &END MULLIKEN
>      &MO_CUBES
>       WRITE_CUBE F
>       NHOMO 2
>      &END MO_CUBES
>      &END PRINT
>     &AUXILIARY_DENSITY_MATRIX_METHOD
>      METHOD BASIS_PROJECTION
>      ADMM_PURIFICATION_METHOD MO_DIAG
>     &END AUXILIARY_DENSITY_MATRIX_METHOD
>   &END DFT
>   &SUBSYS
>     &CELL
>       ABC [angstrom] 47.216 47.216 47.216
>       ALPHA_BETA_GAMMA [deg] 90 90 90
>       PERIODIC XYZ
>       SYMMETRY CUBIC
>     &END CELL
>     &TOPOLOGY
>      COORD_FILE_FORMAT XYZ
>      COORD_FILE_NAME el.xyz
>     &END TOPOLOGY
>     &KIND Cs
>       ELEMENT Cs
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       AUX_FIT_BASIS_SET cFIT7
>       POTENTIAL GTH-PBE-q9
>     &END KIND
>     &KIND Pb
>       ELEMENT Pb
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       AUX_FIT_BASIS_SET cFIT6
>       POTENTIAL GTH-PBE-q4
>     &END KIND
>     &KIND Br
>       ELEMENT Br
>       BASIS_SET DZVP-MOLOPT-SR-GTH
>       AUX_FIT_BASIS_SET cFIT6
>       POTENTIAL GTH-PBE-q7
>     &END KIND
>   &END SUBSYS
> &END FORCE_EVAL
> &MOTION
>  &GEO_OPT
>   MAX_ITER 500
>   OPTIMIZER LBFGS
>  &END GEO_OPT
> &END MOTION
>
> --
> 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... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170928/b5b37bc5/attachment.htm>