[Pw_forum] PAW in 4.0CVS3

[Axel Kohlmeyer]

On Wed, 16 Apr 2008, Chao Cao wrote:

CC> -----BEGIN PGP SIGNED MESSAGE-----
CC> Hash: SHA1
CC> 
CC> Dear Lorenzo,
CC>   Is there any major change or bug fix for PAW method between CVS2 and
CC> CVS3? I used exactly the same set of PP-generation input and PW input,
CC> but somehow got very different result in total energy and force (differs
CC> by several eV).

lorenzo is probably (hopefully) asleep, so i'll give it a shot...

the fundamental constants and unit conversions were updated to the 
latest CODATA 2006 recommended values. if you enter data that needs
to be converted to internal units, e.g. coodinates or cell dimensions
in angstrom, that may have the effect that you see.

i've seen even larger discrepancies in some cases where the code
(not QE) was even further off. makes you think about the real
accuracy of what we are doing here, don't it?

cheers,
  axel.

 
CC>   Also, if I do structural relaxation using PAW in 4.0cvs3, and use the
CC> relaxed structure to do another SCF, the total energy/stress/force from
CC> SCF calculation can be substantially different from the final step in
CC> structural relaxation calculation. For example, I used the PAW dataset
CC> provided in pseudo/ directory to calculate Cu2O bulk, and the last step
CC> of relaxation gives me a total energy of -592.22777692 Ry and P=-0.03
CC> kbar; but if I start from the final structure and do one SCF, the result
CC> is -592.22732012 Ry and P=-0.42 kbar. Since I set conv_thr to be 1.0d-7,
CC> a 0.0004 Ry difference in total energy looks really big to me...
CC> 
CC> - --
CC> Chao Cao
CC> 
CC> Quantum Theory Project and Department of Physics
CC> University of Florida, Gainesville, FL 32611
CC> U.S.A.
CC> -----BEGIN PGP SIGNATURE-----
CC> Version: GnuPG v2.0.7 (GNU/Linux)
CC> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
CC> 
CC> iD8DBQFIBqBzFfX+kymy7P4RArAhAJ9Lp9H8kQykEbHi5P4FodimdVHSFgCfVFyO
CC> Quu1E1XTCoWIKWhoPS9iBCA=
CC> =JTOk
CC> -----END PGP SIGNATURE-----
CC> _______________________________________________
CC> Pw_forum mailing list
CC> Pw_forum at pwscf.org
CC> http://www.democritos.it/mailman/listinfo/pw_forum
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.