[CPMD-list] Re: [CPMD-list]

srb at zurich.ibm.com srb at zurich.ibm.com
Tue May 28 16:11:41 CEST 2002


Hi,

with your settings, CPMD should write a RESTART file when convergence or the 
maximum time/number of steps are reached, whichever comes first.

By using steepest descent for ions and electrons at the same time, you are 
selecting another optimization scheme (the ionic positions are updated every 
wavefunction optimization cycle rather than only after convergence of the 
wavefunction). Did your job finish normally?

You have three possibilities:

1. Use the very efficient DIIS scheme for electron optimization instead of 
steepest descent. This will most likely save you many cycles but can fail in 
very rare cases.

2. Keep combined steepest descent scheme and specify
   STORE
     20
(or any other frequency) to write the RESTART file on the fly (in case your job 
does not finish normally)

3. Use steepest descent for all degrees of freedom but return to the 
conventional scheme which writes the RESTART file each time the wavefunction is 
converged. In rgmopt.F, you have to change the line
      IF(TSDP .AND. TSDE) THEN
into
      IF(.FALSE.) THEN
and recompile the code.

I'd try option 1 first.

Best wishes,
Salomon



More information about the CPMD-list mailing list