[Pw_forum] Strange behavior of scf calculation under finite-field

[Axel Kohlmeyer]

On Thu, 5 Jul 2007, lan haiping wrote:

HPL> Dear All,
HPL> 
HPL>   I am studying electric-enthalpy method implemented in pw.x with
HPL> example31.  There are some problems i cannot solve by myself.
HPL> Firstly,  when i set "ecutrho" explictly, calculations under finite-field
HPL> are failed without any complain information, such as CRASH report. My server

please provide an input that can reproduce this.

HPL> is of 8GB memory, and no such crash for other scf calculations . I had tried
HPL> 'cg' and 'david' methods, and cannot solve this  problems.
HPL> Secondly,  when i set occupations to 'mv' smearing scheme with degauss=0.03
HPL> a.u. , the scf calculation of finite-field situation
HPL> is very difficult to reach convergence.  I have tried to increase or

calculations under external field _are_ more difficult to converge.
how does your field strength compare to the value in example31?

HPL> decrease mixing_ndim,  but failed.  Besides, i have tried to increase
HPL> nbnd , which seems no effect on scf convergence.  Would you please give me
HPL> some suggestion on parameters setting for scf calculation under finite-field
HPL> ?

it would be much easier to detect problems or make suggestions,
if you'd post the relevant part of your input file (or another
input that has the exact same problem).

cheers,
   axel.

HPL> 
HPL> Thanks for your time and attention . Any suggestion or comment is
HPL> appreciated .
HPL> 
HPL> Best Wishes,
HPL> H.P
HPL> 
HPL> 
HPL> 
HPL> 

-- 
=======================================================================
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.