[CP2K-user] [CP2K:22375] CP2K for Geometry Optimization of Rare-Earth Systems

[Thomas Kühne]

Dear Rubee,

yes, CP2K can in principle be used for geometry optimization of a
rare-earth-containing molecule on a surface, including spin-polarized DFT+U.
A system containing 500 to 1500 atoms can be feasible with the GPW method,
a semilocal functional, suitable pseudopotentials, and sufficient parallel
computing resources.

However, the presence of a rare-earth element makes the electronic-structure
setup more important than the geometry optimizer itself.

For every element you need a mutually compatible basis set and
pseudopotential. In particular, verify which rare-earth electrons are included
in the valence of the selected pseudopotential and whether this is appropriate
for the oxidation state, bonding, and magnetic properties of interest. Do not
select a potential only because its element symbol is available.

DFT+U is activated in the KIND section. For a localized f shell the angular
momentum is L=3, for example:

  &KIND Ln
    ELEMENT Ln
    BASIS_SET ...
    POTENTIAL ...
    &DFT_PLUS_U
      L 3
      U_MINUS_J [eV] ...
    &END DFT_PLUS_U
  &END KIND

Here `Ln`, the basis, the potential, and especially U_MINUS_J must be replaced
by values appropriate to the actual element and method. The U value is not a
universal atomic constant. It depends on the pseudopotential, basis,
functional, oxidation state, and target property, and should be justified or
computed for the intended system.

The calculation should normally be spin polarized. You should test the
multiplicity, initial atomic magnetizations, and possibly different initial
occupations of the localized shell. Rare-earth DFT+U calculations can converge
to different metastable occupation patterns, so SCF convergence alone does
not demonstrate that the physically relevant electronic state has been
obtained.

If spin-orbit coupling is important for the selected element or observable,
DFT+U by itself does not replace it. The Hamiltonian used for the structural
and STM calculations should be chosen consistently with the required physical
accuracy.

For a system of this size, I recommend first validating the setup on a smaller
model:

1. converge the basis sets, cutoffs, and pseudopotential;
2. compare relevant spin and DFT+U occupation states;
3. determine an appropriate U value;
4. check the slab thickness, vacuum, and lateral cell;
5. test whether Gamma-point sampling is sufficient;
6. only then scale to the complete 500-1500 atom system.

The subsequent STM calculation is also possible in principle, but CP2K's
current STM print section has important limitations. It uses a
Tersoff-Hamann-type post-processing of the Kohn-Sham states, is currently
implemented only for Gamma-point calculations, and sums the alpha- and
beta-spin contributions into one cube. Positive-bias images additionally
require enough unoccupied states through NLUMO.

A large real-space surface supercell may justify Gamma-point sampling, but
this must be checked against a smaller k-point-converged model. If explicit
k-point sampling is required for the electronic structure, CP2K's current STM
printer cannot directly generate the corresponding STM cubes.

Therefore, CP2K is a suitable candidate, but the answer depends on the actual
rare-earth element, pseudopotential, oxidation and spin state, functional,
U value, surface, and available hardware. Providing these details would make
it possible to recommend a more specific setup.

Best regards,
Thomas

> Am 06.08.2026 um 16:34 schrieb Rubee Soni <rubee1702 at gmail.com>:
> 
> Dear All,
> 
> Can we use CP2K for the geometry optimization of a system containing a rare-earth atom? I would like to optimize the geometry of a molecule containing one rare-earth element adsorbed on a surface. The calculations would need to include DFT+U to properly describe the highly localized electrons of the rare-earth atom.
> 
> The systems I am interested in range from approximately 500 to 1500 atoms. In addition to geometry optimization, I am also planning to perform STM simulations of these systems. Would CP2K be suitable for carrying out both the geometry optimization and subsequent STM simulations for systems of this size?
> 
> Thank you.
> 
> Best,
> 
> Rubee
> 
> 
> -- 
> 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+unsubscribe at googlegroups.com <mailto:cp2k+unsubscribe at googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/cp2k/2971e381-caca-4a2f-8c06-1b42d6768972n%40googlegroups.com <https://groups.google.com/d/msgid/cp2k/2971e381-caca-4a2f-8c06-1b42d6768972n%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/924764C5-218F-42A7-9AF3-00409095B5D2%40gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260806/7714158e/attachment.htm>