akohlmey at cmm.chem.upenn.edu
Tue May 29 15:50:38 CEST 2007
On Tue, 29 May 2007, Jess Kondor wrote:
JK> Hi, everyone,
JK> I have a question. Sometimes (at least, I found several times in the papers
JK> for ferroelectrics), we need to relax a system with very tight criteria for
JK> forces - up to 1 meV/Ang ( ~4.0d-5 in `pw (a.u.)` units). However, I found
JK> that it is hard to achieve this criteria. If I use 'bfgs' with conv_thr =
JK> 1.d-6 (or better) and upscale = 10.d0 , calculations often crashes with
if you want to converge the geometry tightly, you have to converge
the wavefunction even more tightly. also, you may run into 'ripples',
particularly when using ultra-soft pseudopotentials.
in that case, you may also need to crank up the density cutoff.
JK> message ''bfgs history was reset on the previous step...' and number of bfgs
JK> steps = 15, number of scf steps = 2 . If I switch a relaxation method to
JK> 'damp' with dt=70.0, after one (or two) ionic steps forces are small, but
JK> then 'Total force' increases rapidly (from 0.001 to 0.4 or something) and
JK> remains almost constant. So, are there any other parameters which can help
JK> in geometry relaxation?
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.
More information about the users