Geometry optimization can't converge and output trajectory

[Anqi Qiu]

Hi everyone,

I'm running a geometry optimization simulation of a AlFeSi cell. The simulation will not converge, and when I try to output the trajectory to see what is wrong, it will not print the trajectory. Can anyone give me some hints on what might have gone wrong?

Best,

Anqi Qiu

[Krack, Matthias]

Hi Anqi

 

Such problems are often caused by improper atomic coordinates or cell lattice vector definitions. Check the output for small interatomic distances or other warnings.

 

Best,

 

Matthias

 

--
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 visit https://groups.google.com/d/msgid/cp2k/896faf58-eabd-4b87-bcfa-ed7b766ed7b5n%40googlegroups.com.

[聪刘]

Hi, what if we calculate the cell at a very high pressure, so that the atomic distance is very small? I recently tried high-pressure calculation for diamond at 1000 GPa and found the SCF never converged.

[Konstantin Tokarev]

Hi,

What basis set are you using? Small interatomic distances increase likeliness of linear dependence, so you have better chances of convergence when using smaller and less diffuse basis set. Also, in this case it might be worthwhile to try using plane wave basis set instead of Gaussian.

[聪刘]

Hi Konstantin, 

here is my input below and scf never converges.

&GLOBAL

  PROJECT C

  RUN_TYPE ENERGY

  PRINT_LEVEL MEDIUM

&END GLOBAL

&FORCE_EVAL

  METHOD Quickstep

  &DFT

    BASIS_SET_FILE_NAME  BASIS_SET

    POTENTIAL_FILE_NAME  POTENTIAL

    &MGRID

      NGRIDS 4

      CUTOFF 400

      REL_CUTOFF 100

    &END MGRID

    &QS

      METHOD GAPW

    &END QS

    &SCF

      SCF_GUESS ATOMIC

      EPS_SCF 1.0E-6

      MAX_SCF 50

      ADDED_MOS 10

      CHOLESKY INVERSE

      IGNORE_CONVERGENCE_FAILURE TRUE

      &SMEAR ON

        METHOD FERMI_DIRAC

        ELECTRONIC_TEMPERATURE [K] 300

      &END SMEAR

      &DIAGONALIZATION

        ALGORITHM STANDARD

      &END DIAGONALIZATION

      &MIXING

        METHOD BROYDEN_MIXING

        ALPHA 0.4

        BETA 0.5

        NBROYDEN 8

      &END MIXING

    &END SCF

    &XC

      &XC_FUNCTIONAL PBE

      &END XC_FUNCTIONAL

    &END XC

  &END DFT

  &SUBSYS

&KIND C

  BASIS_SET ORB DZVP-GTH-BLYP-q4

  POTENTIAL GTH-BLYP-q4

&END KIND

&CELL

      A 0 2.4195442594936767 2.4195442594936767

      B 2.4195442594936767 0 2.4195442594936767

      C 2.4195442594936767 2.4195442594936767 0

      PERIODIC XYZ

    &END CELL

    &COORD

      C 0 0 0

      C 0.60488606487341917 0.60488606487341917 0.60488606487341917

    &END COORD

  &END SUBSYS

  &PRINT

    &TOTAL_NUMBERS  ON

    &END TOTAL_NUMBERS

  &END PRINT

&END FORCE_EVAL

But if I change the position to below, it can converge quickly.
&COORD
      C 0 0 0
      C 0.35488606487341917 0.35488606487341917 0.35488606487341917
    &END COORD

best,

Cong

[聪刘]

On Sat, Oct 18, 2025 at 3:49 AM Konstantin Tokarev <ann...@gmail.com> wrote:

Hi,

What basis set are you using? Small interatomic distances increase likeliness of linear dependence, so you have better chances of convergence when using smaller and less diffuse basis set. Also, in this case it might be worthwhile to try using plane wave basis set instead of Gaussian.

--

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 visit https://groups.google.com/d/msgid/cp2k/aa82397f-be69-42b9-a191-b931782217d3n%40googlegroups.com.