A bug for CELL_OPT with DFT-D3?

[贾建峰]

Dear all,

I use the cp2k (5.1) to optimize a cell of TiO2. Below is the main input,
&GLOBAL
  RUN_TYPE  CELL_OPT
&END GLOBAL
&MOTION
  &CELL_OPT
    OPTIMIZER  BFGS
   &END CELL_OPT
&END MOTION
  &FORCE_EVAL
  METHOD  QS
  STRESS_TENSOR  ANALYTICAL
  &DFT
          BASIS_SET_FILE_NAME ${DATAPATH}/BASIS_MOLOPT
          POTENTIAL_FILE_NAME ${DATAPATH}/GTH_POTENTIALS
          &SCF
                   &OT  T
                 ..
                  &END OT
          &END SCF
          &QS
                  METHOD  GPW
          &END QS
          &MGRID
                  CUTOFF   400
          &END MGRID
          &XC
                  &XC_FUNCTIONAL  NO_SHORTCUT
                          &PBE  T
                          &END PBE
                  &END XC_FUNCTIONAL
                   &vdW_POTENTIAL
                    DISPERSION_FUNCTIONAL PAIR_POTENTIAL
                           &PAIR_POTENTIAL
                           TYPE DFTD3
                               CALCULATE_C9_TERM .TRUE.
                                 REFERENCE_C9_TERM .TRUE.
                                 LONG_RANGE_CORRECTION .TRUE.
                                  PARAMETER_FILE_NAME ${DATAPATH}/dftd3.dat
                                  VERBOSE_OUTPUT .FALSE.
                                  REFERENCE_FUNCTIONAL PBE
                                  R_CUTOFF [angstrom] 9.0
                                  EPS_CN 1.0E-6
                          &END PAIR_POTENTIAL
                               &END vdW_POTENTIAL
      &END XC
     &END DFT
  &SUBSYS
          &CELL
                  ABC 4.59400 4.59400 2.95900
                  ALPHA_BETA_GAMMA 90.0 90.0 90.0
                  PERIODIC  XYZ
                  SYMMETRY TETRAGONAL
                  MULTIPLE_UNIT_CELL  3 3 4
          &END CELL
          &COORD
                  Ti     0.0000    0.0000    0.0000
                  Ti     2.2970    2.2970    1.4795
                  O      1.4003    1.4003    0.0000
                  O      3.1937    3.1937    0.0000
                  O      0.8967    3.6973    1.4795
                  O      3.6973    0.8967    1.4795
          &END COORD
          &KIND Ti
                  BASIS_SET DZVP-MOLOPT-SR-GTH
                  POTENTIAL GTH-PBE-q12
          &END KIND
          &KIND O
                  BASIS_SET TZVP-MOLOPT-GTH
                  POTENTIAL GTH-PBE-q6
          &END KIND
          &TOPOLOGY
                  MULTIPLE_UNIT_CELL  3 3 4
          &END TOPOLOGY
  &END SUBSYS
&END FORCE_EVAL

It is convergent after 9 step. The energy changes as below in these steps:
  -6514.202116775491959
  -6514.107348731997263
  -6514.100302314682267
  -6514.121769318782754
  -6514.123121857887782
  -6514.122521243046322
  -6514.121260371050994
  -6514.120835064291896
  -6514.120835064288258
It is clear that the first step has the lowest energy. When I change the vDW correction to below:
 &vdW_POTENTIAL
    DISPERSION_FUNCTIONAL NON_LOCAL
    &NON_LOCAL
      TYPE DRSLL
      VERBOSE_OUTPUT
      KERNEL_FILE_NAME /lustre/home/jiajf/software/cp2k30/cp2k/cp2k/data/vdW_kernel_table.dat
      CUTOFF  80
    &END NON_LOCAL
 &END vdW_POTENTIAL
The energy changes as below,
  -6501.667309011268117
  -6501.552098176282016
  -6501.683944202200109
  -6501.684511798672247
  -6501.691681789144241
  -6501.691989256092711
  -6501.692075269643283
  -6501.692160170975512
  -6501.692235978556710
  -6501.692261498596054
  -6501.692262177740304
  -6501.692262178249621

So, I think that there is a bug for CELL_OPT with DFT-D3 correction.

Jianfeng Jia



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180323/0267316f/attachment.htm>