density fitting auxbasis in cp2k

[yj jiang]

Hello everyone,

I am implementing Density Fitting (DF) for Hybrid DFT using the MOLOPT basis set.

My reference, PySCF, uses the Even-Tempered Basis (ETB) to generate the auxiliary basis set (auxbasis). This ETB is extremely large, often over 10 times the size of the primary basis, making the calculation highly inefficient.

I am seeking a method to generate a more compact auxbasis. I know MOLOPT basis come from CP2K, so come here for some help.

Could anyone provide references or specific details on:

CP2K's strategy for generating these compact MOLOPT auxiliary basis sets (i.e., the recipe that avoids the oversized ETB)?

Any specific parameters (like geometric progression rules) that could replace the default PySCF ETB for MOLOPT while maintaining accuracy?

Thanks for any guidance!

[Frederick Stein]

Hi,

I am not quite sure which MOLOPT auxiliary basis sets you are referring to. CP2K does not have preoptimized RI basis sets for HFX but for RI-MP2, RI-RPA, GW and separate ones for the LRIGPW-approach only. The recommendation is to use automatically generated basis instead (see https://manual.cp2k.org/cp2k-2025_2-branch/methods/dft/hartree-fock/ri_gamma.html). In case of all-electron calculations, you can employ those corresponding to the primary basis set from https://www.basissetexchange.org/.

HTH,

Frederick