[CPMD-list] CPMD+PAW?

[Axel Kohlmeyer]

On Wed, 12 Mar 2008, Charles Chen wrote:

CC> Dear CPMD users,

dear charles,

CC> I am just a beginner here in CPMD.

nothing bad about that. ;)

CC> I am working some TiO2 surface system and I noticed there are quite 
CC> a lot of papers on this topic. In the attached two, the authors 
CC> mentioned the simulation is done with MD within CP approaches. The 
CC> electron-ion interaction is described by ultrasoft pseudopotentials. 
CC> Plane wave basis set cutoff for the smooth part is 25Ry while for 
CC> the augmented density is 200 Ry. When I check the CPMD manual, there 
CC> is one cutoff value need be specified and it seems to be the former 
CC> on (25Ry) for USPP. So, where to specify the other one? Or the "MD 

the keyword you are looking for is "DENSITY CUTOFF"
in the &SYSTEM section (check out the file sysin.F).
alternately you can specify the ratio between wavefunction
and density cutoff through the DUAL keyword. for norm-conserving
pseudopotentials there is little reason to change this from the
default value of 4.0, but for USPPs the range of 5.0 to 10.0 is
more adequate.

CC> within CP approaches" does not refer to CPMD?

since car-parrinello molecular dynamics is abbreviated
CPMD it can be confusing, whether somebody is referring
to the CPMD code or the CP-MD method. there are several 
implementations of the CP-MD method, some even have
the same origin as the CPMD code (so fundamental internal 
variable names tend to be similar to some degree).

you have to check the references of the paper whether 
the authors used CPMD or one of the other implementations.

please note, that the ultrasoft-support in CPMD is limited
to basic functionality (most of the time you'll get a proper
warning and the code will abort) and that we occasionally 
run into "undocumented features". so i'd like to encourage
you to pay special attention to your outputs and double
check everything. when running in parallel, you can frequently
run into the situation where using norm-conserving potentials
is actually faster (provided you have enough parallel nodes
available) since there is much less overhead associated with
them and you can parallelize better across the larger basis set.

cheers,
   axel.
 
CC> BTW: I am still with CPMD 3.9.1
CC> 
CC> Thanks!
CC> 
CC> Charles
CC> 
CC> 

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