How can one guess a reasonable cutoff value?
This is for the impatient CP2K user: As a rule of thumb one can multiply the largest exponent of the Gaussian basis sets employed by 40 to obtain a reasonable guess for the cutoff value in Rydberg. The first column of the CP2K basis set format contains the Gaussian exponents, e.g. for the oxygen DZVP-MOLOPT-SR-GTH basis set you may find a value of about 10.4 and the corresponding hydrogen DZVP-MOLOPT-SR-GTH basis set employs a largest exponent of about 10.0. Thus a cutoff value of 10*40 = 400 Rydberg would be a reasonable guess for a cutoff value to start with. The following values
Total charge density on r-space grids: 0.0000000000 Total charge density g-space grids: 0.0000000000
in the CP2K output should be smaller than 10-8. Use the PRINT_LEVEL medium
&GLOBAL PRINT_LEVEL medium ... &END GLOBAL
to get this printout after each SCF iteration step. The default is to print this information when the SCF iteration is converged which possibly might never happen with a too small cutoff value.
For a more fine-tuned setting of the cutoff and the relative cutoff value you may check the this How-to.