[CPMD-list] input and output files for calculating chemical potential

[Axel Kohlmeyer]

On Wed, 20 Sep 2006, fafa chiker wrote:

FC> Hello
FC>   Thank your for your reply.
FC>   So the version of the cpmd code which I use is 3.11.0,

are you certain about that? 
according to your outputs, you are using version 3.9.1!

FC>   For each studied bulk structures (GaAs, As, and Ga) I made three
FC> steps of running programs separately with the same restart (the
FC> input and output files were attached bellow):

thanks for doing it in this structured way. this makes
is easy to understand what you are doing.

FC>   1-       Optimize Wave function
FC>   2-       Optimize geometry
FC>   3-       Kohn-sham energies

FC>   So I remarked several changes when I increased the cutoff plane
FC> waves and the K-points (I have tested for 13, 40, and 100Ryd also
FC> for MP-K-Points for 2x2x2, 4x4x4 and 6x6x6 also for 8x8x8), so the
FC> problem why I find different values of chemical potential in each
FC> case, then I have tested for bulk As and bulk Ga with one atom in
FC> unit cell and we have compared the occupied energies with the
FC> following tables

FC>   http://physics.nist.gov/PhysRefData/DFTdata/Tables/31Ga-T.html
FC>   http://physics.nist.gov/PhysRefData/DFTdata/Tables/33As-T.html

those tables are reference data for _single_ atoms by solving the 
the schroedinger equation on a radial grid. similar to what is done
in atomic codes as part of the generation of pseudopotential.

FC>   so I find some agreement, but the questions is, its right to keep
FC> one atom in the unit cell of the bulk Ga or As???

how many atoms you need as a minimum in the unit cell is of course
determined by the space group of your lattice. 

note that if you do the calculations like this,
you have fixed occupations and no smearing, so 
your k-point convergence will be slow. as long 
as you have a large band gap and a large system
- which is probably the most typical application
of the CPMD code - this is ok. for semiconductors,
and even more so for metals, you need to use the 
FREE ENERGY FUNCTIONAL module which in turn 
requires to use lanczos diagonalization. 

also, since you use goedecker-hartwigsen pseudopotentials,
you should keep in mind, that those usually require
a somewhat large plane wave energy cutoff than
the more frequently used troullier martins type.


FC>   best regard 
FC>   fafa


FC>                 *******  *******   **********  *******  
FC>                ***       **   ***  ** **** **  **   *** 
FC>                **        **   ***  **  **  **  **    ** 
FC>                **        *******   **      **  **    ** 
FC>                ***       ******    **      **  **   *** 
FC>                 *******  **        **      **  *******  
FC>                  ******  **        **      **  ******   
FC>    
FC>                             VERSION 3.9.1

see. it says 3.9.1 here.


[...]


cheers,
    axel.

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