This crashes, and I don't understand why

[Jonathan Torstensen]

Kind regards, 

Jonathan 

[Thomas Kühne]

Dear Jonathan,

the calculation does not fail because of the geometry optimization itself.

The input requests both

  ADDED_MOS 20

and orbital transformation through the &OT section. CP2K does not implement

OT with ADDED_MOS /= 0 and therefore aborts with

  OT with ADDED_MOS/=0 not implemented

For a ground-state geometry optimization, remove ADDED_MOS. OT optimizes the

occupied subspace and does not require additional virtual orbitals:

&SCF

  MAX_SCF 200

  EPS_SCF 1.0E-8

  SCF_GUESS ATOMIC

  &OT

    PRECONDITIONER FULL_SINGLE_INVERSE

    MINIMIZER DIIS

  &END OT

&END SCF

If you actually need 20 unoccupied orbitals, remove the &OT section and use

conventional diagonalization instead, keeping ADDED_MOS 20.

There is also a separate periodicity mismatch in the input. Because &POISSON

uses PERIODIC NONE, add PERIODIC NONE to &CELL as well:

&CELL

  A 30.0 0.0 0.0

  B 0.0 30.0 0.0

  C 0.0 0.0 30.0

  PERIODIC NONE

&END CELL

For this isolated, elongated molecule I would also center the coordinates in

the box using &TOPOLOGY / &CENTER_COORDINATES.

I reproduced the abort and verified that removing ADDED_MOS allows CP2K to

pass the failing OT initialization. No source-code change is required.

If the abort message did not appear in NL.out, capture both stdout and stderr,

for example:

  mpiexec -n 2 cp2k.psmp -i NL-geo-opt.inp > NL.out 2>&1

Best regards,

Thomas

Am 20.07.2026 um 16:12 schrieb Jonathan Torstensen <jona...@gmail.com>:

Kind regards, 

Jonathan 

--
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/CABEx5tT9GD7ZSrDbFua9c5-zaiEghRUNHU6FsBZ-vezkOCOdCg%40mail.gmail.com.
<NL14.xyz><BASIS_MOLOPT><NL.out><NL-geo-opt.inp>