Clarification on k-point sampling, slab passivation, and convergence in CP2K

[usman ul muazzam]

Dear CP2K Users and Developers,

I am trying to model Ga and O adatom diffusion barriers on substrates like diamond and sapphire using CP2K (periodic DFT + NEB), and I have a few questions regarding best practices:

1. k-point Sampling:
Some CP2K publications seem to use only Gamma-point, even for periodic systems. In contrast, codes like Quantum ESPRESSO require k-point convergence for accuracy.

• Is k-point convergence generally not needed in CP2K.

• When should one use explicit k-point grids in CP2K.

2. Slab Model Passivation:
In plane-wave codes, bottom surface passivation (e.g., with hydrogen) is common in slab models to avoid spurious states. But I rarely see this in CP2K-based surface studies.

• Is bottom-layer passivation not necessary in CP2K.

3. Convergence Parameters:
Is optimizing CUTOFF and REL_CUTOFF typically sufficient for accurate periodic DFT and NEB calculations in CP2K.

• Are other parameters like basis sets (as here we use Gaussian instead of plane wave) or grid levels also important to converge.

• Any recommended references for CP2K convergence strategies.

Any guidance or pointers to best practices would be greatly appreciated.

Best Regards,

Usman

[FS]

Hi
If possible, I would add on these questions as I had overlapping/similar ones.

A. SCF Convergence:

a-What are the typical variables in cp2k that make sense to play around with, with trailing convergence? Are there some rule of thumbs that work well for values/ratios for ALPHA and NBROYDEN in the SCF/MIXING block?

b-I notice that i run into convergence issues with the SCCS solvation (and to an lesser extend the gCP correction, guess noise). Any tips for parameters for the SCCS block here?

&SCCS on
&ANDREUSSI
RHO_MAX 0.001
RHO_MIN 0.0001   

&END ANDREUSSI
      DIELECTRIC_CONSTANT 40
      EPS_SCF 1.0e-3
      EPS_SCCS 3.0E-6
      MAX_ITER 300
    &END SCCS

c-As I come from non-periodic calcs: fermi smearing for convergence, what are acceptable temperatures to use?

B. Adding to the Fermi Smearing:

Similar to molecular systems with strong static electron correlation (SEC) I looked into in the past, it might make sense for me to use FT-DFT as described in Grimme's FOD papers at smear temps of 20000K*%HFexchange + 5000K in geometry optimizations, freqs and single points to account for strong SEC effects. Is there anything in cp2k that would make FT_DFT behave differently/unexpectedly for these calcs as in programs like ORCA?

C. Is there like a good practice reference I could look into for periodic calcs, similar to https://onlinelibrary.wiley.com/doi/10.1002/anie.202205735 for mostly non-periodic calcs?

All the best