The code has two options to improve that. The first trick is due to Michel Cote and goes as follows. After the selfconsistent calculation has been performed at normal cutoff, a second calculation with higher cutoff is done, but WITHOUT allowing the charge density to relax. This turns out to be still a very good approximation, but the second calculation is somewhat expensive. The cutoff energy (in Rydberg) for the second run is set with the keyword
polished_energy_cutoff.
Example:
polished_energy_cutoff 50
Default: switched off
The second trick is to modify the kinetic energy according to
Bernasconi et al. You can set the height , position , and
width (all in Rydberg) of the step function that gets added:
with the keyword modify_kinetic_energy .
Example:
modify_kinetic_energy 20.0 10.0 1.0.
If you just do
modify_kinetic_energy on
a reasonable setting is assumed: , , .
Default: modify_kinetic_energy off