Problems in geometry optimization

[Chn]

Dear experts in CP2K,

I am a learner of CP2K package and I am really thankful if my naive question can be explained.

I tried to do geometry optimization toward Pt bulk or Pt(111) by set the keyword "RUN_TYPE" as GEO_OPT and the *.inp file was referred to the exercises "Adsorption of acetylene on an intermetallic surface" about PdGa at CP2K website. But when the calculation done after several hours, I found that the atoms in cell distorted seriously. I think it should be obvious incorrect but I have no idea about how to fix it. Many thanks in advance.

The imported .xyz was prepared by using materials studio and VESTA. I have tried to use larger supercell (3*3*3) or to set the keyword "MULTIPLE_UNIT_CELL" as 3 3 3 and tried to optimize other structures like TiO2 but no use..is it that the cell I used still too small?

The attachments were the input files I used.

Regards,

chn

[Daniele Passerone]

Dear Chn

Apart from the size of the cell that could give you some trouble, the problem is the keyword

CHARGE 1

that you added in the &DFT section, meaning that you are simulating a charged system. Removing that would work.

Indeed in the output we see:

                          Hirshfeld Charges

  #Atom  Element  Kind  Ref Charge     Population                    Net charge

      1       Pt     1      18.000         17.750                         0.250

      2       Pt     1      18.000         17.750                         0.250

      3       Pt     1      18.000         17.750                         0.250

      4       Pt     1      18.000         17.750                         0.250

  Total Charge                                                            1.000

Also not necessary is the constraint on the position of one atom that you have in the optimization section.

In general you should also vary the cell volume and find what is the equilibrium volume for the PBE functional.

Hope this helps

Daniele

[Chn]

Dear Daniele,

Thanks for your answer! I annotated the keyword "CHARGE" and the constraint part as you mentioned, but it still doesn't work. Could you please tell me some reasons that why does this situation happen in general? I am wandering that whether it is caused by the input file.

Does the variation of cell volume you mentioned above mean that I need to do the CELL_OPT before GEO_OPT?

thanks again,

chn

在 2019年8月28日星期三 UTC+8下午5:26:58，Daniele Passerone写道：

[Patrick Gono]

Dear Chen,

A CELL_OPT calculation automatically finds the optimal atomic configuration as well. In general, the DFT lattice parameter will not be identical to the experimental one. Therefore, it makes sense to run a CELL_OPT calculation to determine the correct unit cell parameters and the atomic structure. In practice, such calculations do not take much more time than a GEO_OPT simulation. Note that in the case of bulk Pt, you know that the cell is cubic, and thus can constrain the angles during CELL_OPT by switching the option KEEP_ANGLES in the &CELL_OPT subsection of the &MOTION section to true.

But regarding your original issue: CP2K, by default, only uses a single k-point, the Gamma point. Whereas in a plane-wave code you would converge any results with respect to the k-point mesh density, in CP2K you must converge the results with respect to the supercell size. I would suggest running a CELL_OPT calculation with at least a 2x2x2, preferably a 3x3x3 supercell, and compare the lattice parameters with literature (still, the number will be off due to the functional used, so compare with PBE results).

Note that you must specify the MULTIPLE_UNIT_CELL keyword in both the &TOPOLOGY and the &CELL subsections!

Of course, get rid of that extra charge, you most probably want to study a neutral system. And if you are dealing with bulk systems, keeping fixed the position of an atom is not necessary, as the system will not float around. This might only be relevant in the case of e.g. molecules or 2D sheets in vacuum.

Yours sincerely,

Patrick Gono

--
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/7092cb9a-be30-4f63-8246-b988dea87c67%40googlegroups.com.

[Chn]

Dear Patrick,

Thanks a lot for your patient response! All these advices is really of great help to me, and the problem seems to be solved by setting &KPOINTS in &DFT section.

regards,

chn

在 2019年8月28日星期三 UTC+8下午8:30:20，Patrick Gono写道：

To unsubscribe from this group and stop receiving emails from it, send an email to cp...@googlegroups.com.

[Chn]

Dear Travis,

Thanks for your guidance about the input files! To set KPOINTS is really helpful about the distortion problem though the crystal optimization results have about 0.0x Ångstrom error compared with the result found in Materials Project. Hope this could be a reasonable result.

Regards,

chn

在 2019年8月28日星期三 UTC+8下午11:09:03，Travis写道：

Hi,

As others have mentioned, change your CHARGE to 0. Additionally, you'll want to use KPOINTS,

&FORCE_EVAL
 METHOD QUICKSTEP
 STRESS_TENSOR ANALYTICAL
 &DFT
  UKS T
  CHARGE 0
  MULTIPLICITY 1
  BASIS_SET_FILE_NAME data/BASIS_MOLOPT
  POTENTIAL_FILE_NAME data/POTENTIAL
  &MGRID
   CUTOFF 800
   NGRIDS 5
   RELATIVE_CUTOFF 50
  &END MGRID
  &QS
   EPS_DEFAULT 1.0E-12
   METHOD GPW
   EXTRAPOLATION USE_GUESS
  &END QS
  &SCF
   EPS_SCF 1e-06
   MAX_SCF 200
   SCF_GUESS RESTART
   ADDED_MOS  400
   &SMEAR  T
    METHOD  FERMI_DIRAC
    ELECTRONIC_TEMPERATURE     3.0000000000000000E+02
   &END SMEAR
   &MIXING  T
    METHOD  BROYDEN_MIXING
    ALPHA     4.0000000000000002E-01
    BETA     1.5000000000000000E+00
    PULAY_ALPHA     5.0000000000000003E-02
    PULAY_BETA     1.0000000000000000E+00
    NMIXING  5
    NBUFFER  8
   &END MIXING
  &END SCF
  &XC
   FUNCTIONAL_ROUTINE NEW
   DENSITY_CUTOFF 1.0e-12
   GRADIENT_CUTOFF 1.0e-12
   TAU_CUTOFF 1.0e-12
   &XC_FUNCTIONAL
    &PBE
     PARAMETRIZATION Orig
    &END PBE
   &END XC_FUNCTIONAL
   &XC_GRID
    USE_FINER_GRID  T
   &END XC_GRID
  &END XC
  &POISSON
   POISSON_SOLVER  PERIODIC
   PERIODIC  XYZ
  &END POISSON
 &KPOINTS
   SCHEME  MONKHORST-PACK  8  8  8
   FULL_GRID  .TRUE.
  &END KPOINTS
 &END DFT
 &SUBSYS
  &CELL
   ABC     3.976771     3.976771     3.976771
   ALPHA_BETA_GAMMA   90.000000  90.000000  90.000000
   PERIODIC XYZ
   MULTIPLE_UNIT_CELL  1  1  1
  &END CELL
 ...
 &END SUBSYS
&END FORCE_EVAL

Structure was taken off Materials Project which is pre-optimized with similar settings in VASP. Usually takes <10 CELL_OPT iterations to optimize.

-T