[Pw_forum] Pseudo-potential cutoff recommendations

[Axel Kohlmeyer]

On Wed, 29 Oct 2008, JR Schmidt wrote:

JRS> Is there an simple mechanism that can be used to estimate a reasonable 
JRS> cutoff energy for a given pseudo-potential?  I realize that these values 
JRS> need to be checked for convergence, but I would like to at least get a 
JRS> reasonable starting point.  It appears that there are fields stored in 
JRS> the pseudo-potential files that are supposed to give "reasonable" 
JRS> defaults for both the wavefunction and density KE cutoffs, but they seem 
JRS> to be always zero in every case I have checked.

yep. these cannot be easily detected automatically. one can derive
some estimates during the generation procedure though and UPF files
generated in version 2 file format contain this value.

however, different properties converge differently with the size of 
the basis set (wfc and density cutoffs). and obviously nobody cared 
to edit the old files.

in general there are some rough guidelines:

ultra-soft pseudopotentials require usually about 
30-50 ry wavefunction cutoff and 6-10 times the
density cutoff. 

norm-conserving pseudopotentials generally require
higher wavefunction cutoffs, say, 60-120ry sometimes
even more, sometimes (e.g. Si) less and there is no 
gain from changing the density cutoff.

forces generally converge faster than total energies
or the stress tensor. energy differences also tend
to converge fast.

JRS> I'm guessing a possible answer is to run a calculation on the isolated 
JRS> atom for each atom type.  However this seems like something that would 

usually a good estimate is done with a small bulk system 
with a proper k-point sampling. checking the E/V curve 
for different cutoffs is a frequent test for pseudos.

JRS> have been done previously for each pseudo-potential and tabulated somewhere.

feel free to start such a table. 

JRS> Please forgive me if this question comes up frequently, but I couldn't 
JRS> find an easy mechanism to search the archives.

you can (ab)use google. adding "site:democritos.it" to your query
will search only the server where the mailing list is currently
archived. there also is a FAQ on the wiki at http://www.quantum-espresso.org

cheers,
   axel.

JRS> 
JRS> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.