8
$\begingroup$

The structure was created using Stillinger Weber with LAMMPS.

The command I used was the following: mpiexec -np $PBS_NP --machinefile $PBS_NODEFILE /opt/spack/opt/spack/linux-centos7-x86_64/intel-18.0.2/cp2k-develop-fw3esv3ywfndzwchh6qphm4nfp253aco/bin/cp2k.popt -o cp2kMD.out silicon64.inp

Can the Hamiltonian be nonconvex?

Here is the input file:

@SET SYSTEM silicon8
@SET LIBDIR ~/cp2k


&GLOBAL
  PROJECT ${SYSTEM}
  RUN_TYPE ENERGY
  PRINT_LEVEL MEDIUM
&END GLOBAL
&FORCE_EVAL
  METHOD Quickstep
  STRESS_TENSOR ANALYTICAL
  &DFT
    BASIS_SET_FILE_NAME ~/abtstSi/cp2kmd/PBE/BASIS_SET
    POTENTIAL_FILE_NAME ~/abtstSi/cp2kmd/PBE/GTH_POTENTIALS
    &POISSON
      PERIODIC XYZ
    &END POISSON
    &SCF
      SCF_GUESS ATOMIC
      EPS_SCF 1.0E-8
      MAX_SCF 500
    &END SCF
    &XC
      &XC_FUNCTIONAL PBE
      &END XC_FUNCTIONAL
    &END XC
  &END DFT
  &SUBSYS
    &KIND Si
      ELEMENT   Si
      BASIS_SET DZVP-GTH-PBE
      POTENTIAL GTH-PBE-q4
    &END KIND
    &CELL
      ABC 7.68 10.86116 15.36
      ALPHA_BETA_GAMMA  90.0 90.0 90.0
      PERIODIC XYZ
    &END CELL
    &COORD
Si 7.05927 6.64266 5.37576
Si 1.76237 9.252659999999999 5.21654
Si 1.22129 10.0626 10.6935
Si 2.6313200000000005 1.9667599999999998 12.6966
Si 6.8577 8.98551 4.63333
Si 1.83598 3.60553 7.161880000000001
Si 7.35508 0.528038 5.59927
Si 3.9881700000000007 9.96088 0.6837730000000001
Si 1.8054 0.803562 4.11245
Si 7.63652 2.6251399999999996 0.546681
Si 1.08848 0.709195 1.7654
Si 0.7939940000000001 7.6549499999999995 11.3353
Si 0.123084 0.9485609999999999 9.42303
Si 6.44041 2.60145 4.83899
Si 1.42496 4.16918 15.2986
Si 4.4713 7.25779 12.83
Si 2.46468 7.359639999999999 14.24
Si 6.9968 1.29906 14.058600000000002
Si 6.08778 10.2214 15.0148
Si 3.5213200000000002 4.75244 9.93462
Si 1.8154100000000002 3.8253399999999997 11.5148
Si 5.65175 4.03603 6.687490000000001
Si 7.52268 7.27055 0.699091
Si 4.51691 2.88637 2.9541
Si 3.3048 7.84857 4.20158
Si 7.156880000000001 9.50139 1.82017
Si 7.43061 3.1365499999999997 10.5857
Si 1.0672 5.350579999999999 13.299500000000002
Si 5.28622 9.39642 12.561000000000002
Si 1.1846200000000002 0.18134899999999998 12.9538
Si 3.1005 5.90203 7.56032
Si 6.701040000000001 10.0717 7.5551
Si 5.76607 5.066889999999999 11.0669
Si 4.519260000000001 6.08691 5.54111
Si 3.41978 7.47243 1.41607
Si 4.1169 0.37773099999999993 11.3865
Si 5.632110000000001 2.36772 12.2505
Si 4.60556 7.57004 8.80234
Si 3.19236 4.68417 3.79859
Si 3.38263 1.45833 1.21872
Si 1.4199 1.16354 7.33426
Si 3.31172 2.83036 14.804700000000002
Si 6.33074 3.99995 2.03448
Si 4.865850000000001 9.68236 3.13388
Si 7.478370000000001 4.1583 8.27792
Si 4.263060000000001 0.7377769999999999 4.33558
Si 0.7445280000000001 6.71793 7.63183
Si 1.7130400000000001 5.84304 1.83262
Si 4.86432 8.66803 6.547920000000001
Si 3.16261 6.99151 10.881300000000001
Si 7.483300000000001 8.274009999999999 13.7653
Si 6.364810000000001 9.18195 10.1728
Si 0.0874567 4.91559 3.78355
Si 6.81633 6.657199999999999 9.667010000000001
Si 1.46051 3.19541 4.84721
Si 1.2459 9.04408 7.58636
Si 6.28422 5.63362 13.235
Si 5.39731 3.9856199999999995 14.5655
Si 5.41372 6.109999999999999 0.813683
Si 4.09035 2.86163 8.28733
Si 1.7804 9.85183 15.001300000000002
Si 3.10574 9.46592 9.22525
Si 4.91753 0.752929 9.04135
Si 3.5791200000000005 0.0752674 6.74137
    &END COORD
  &END SUBSYS
&END FORCE_EVAL

The following is the CP2K output

I can't include all the output but the following is the SCF part:

   497 P_Mix/Diag. 0.40E+00    1.4   576.42755204        63.8866362031  4.07E+00
   498 P_Mix/Diag. 0.40E+00    1.4   628.09173001        53.9808222782 -9.91E+00
   499 P_Mix/Diag. 0.40E+00    1.4   976.73936963        69.7309890316  1.58E+01
   500 P_Mix/Diag. 0.40E+00    1.4  1278.54363778        52.4883144534 -1.72E+01

  Leaving inner SCF loop after reaching   500 steps.


  Electronic density on regular grids:       -255.9999999043        0.0000000957
  Core density on regular grids:              255.9999999562       -0.0000000438
  Total charge density on r-space grids:        0.0000000520
  Total charge density g-space grids:           0.0000000520

  Overlap energy of the core charge distribution:               0.00000000021129
  Self energy of the core charge distribution:               -656.51151540102558
  Core Hamiltonian energy:                                    290.84105129210911
  Hartree energy:                                             501.21656005088914
  Exchange-correlation energy:                                -83.05778148873662

  Total energy:                                                52.48831445344740

 *** WARNING in qs_scf.F:542 :: SCF run NOT converged ***
$\endgroup$
3
$\begingroup$

After modifying your file, I was able to get a converged energy of -250.37 hartrees, using CP2K v7.0 `(git:d15a9c53f), however I think this can be run on any CP2K version beyond v2.5.

    14 OT DIIS     0.15E+00    0.4     0.00000001      -250.3728024712 -3.24E-11
    15 OT DIIS     0.15E+00    0.4     0.00000001      -250.3728024712 -2.69E-11
    16 OT DIIS     0.15E+00    0.4     9.7858E-09      -250.3728024712 -3.97E-11

  *** SCF run converged in    16 steps ***
  Electronic density on regular grids:       -255.9999999988        0.0000000012
  Core density on regular grids:              255.9999999562       -0.0000000438
  Total charge density on r-space grids:       -0.0000000426
  Total charge density g-space grids:          -0.0000000426

  Overlap energy of the core charge distribution:               0.00000000021129
  Self energy of the core charge distribution:               -656.51151540102558
  Core Hamiltonian energy:                                    135.62949829361440
  Hartree energy:                                             345.88853381770627
  Exchange-correlation energy:                                -75.37931918172779

  Total energy:                                              -250.37280247122129

  outer SCF iter =   13 RMS gradient =   0.98E-08 energy =       -250.3728024712
  outer SCF loop converged in  13 iterations or  256 steps

The modifications are:

      &DFT
    BASIS_SET_FILE_NAME ./GTH_BASIS_SETS
    POTENTIAL_FILE_NAME ./GTH_POTENTIALS
    &POISSON
      PERIODIC XYZ
    &END POISSON
    &SCF
      MAX_SCF  20
      EPS_SCF     1E-08
      SCF_GUESS  RESTART
      &OT  T
        MINIMIZER  DIIS
        PRECONDITIONER  FULL_KINETIC
      &END OT
      &OUTER_SCF  T
        EPS_SCF     1E-08
        MAX_SCF  20
      &END OUTER_SCF
      &PRINT
        &RESTART  SILENT
          BACKUP_COPIES  1
        &END RESTART
      &END PRINT
    &END SCF
    &XC
      &XC_FUNCTIONAL PBE
      &END XC_FUNCTIONAL
    &END XC
  &END DFT
  &SUBSYS
    &KIND Si
      ELEMENT   Si
      BASIS_SET DZVP-GTH-q4
      POTENTIAL GTH-PBE-q4
    &END KIND

One of my assumption here is that you are trying to use DZVP-GTH Basis for Si atom (works for generic GGA functionals such as PBE). The significant change here is the utilization of the Orbital Transform (OT) optimizer. I would recommend a quick overview of literature on OT methods for CP2K, as these are really crucial for a faster SCF convergence with CP2K.

$\endgroup$
1
  • $\begingroup$ Impressive! +1 good work getting the user's calculation to converge! I'll make some small edits later to correct a couple typos but thanks for getting this working for the user! $\endgroup$ Dec 8 '20 at 22:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.