[Pw_forum] efermi?

Alessandra Satta alessandra.satta at dsf.unica.it
Thu Nov 18 13:36:36 CET 2004


To whom has already got this trouble and found out the solution:

To compute the Fermi energy in a relaxed doped silicon
(i.e. odd number of elects.) I perform a "scf" calculation
whith both "gaussian smearing" and "tetrahedra" methods.

It happens that when the run accidentally stops before
the convergence is reached (because the simulation cell is big and
the computing time per run is limited) the next run in the
"restart" mode stops in the second case (tetrahedra) with
the message:

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     from efermit : error #         1
     unexpected error
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

I figured out that in the restart mode the eigenvalues "et (nbnd, nks)" 
calculated
in the previous run do NOT arrive to this routine, since they result
to be  0.000000000000000E+000 and the sumkt is, as a consequence, also 
zero.
The check-point-"if"-condition is then satisfied calling the error.
 
All this, is also verified in the first case, using the gaussian smearing.
It does not stop only because the first guess of Eup and Elw is calculated 
adding/subtracting 2*degauss. Degauss is 'known' and the calculation 
goes on.

In conclusion:  If the scf calculation is performed in one-shot the
fermi energy is correct. On the contrary, I'm not sure.
In a quick&dirty solution I can comment the "if-line" but this is not 
the right way....

Hope to have made myself clear.....

Did someone of you run into such a problem? 
thanks a lot....

     alessandra

P.S.: I refer to the ESPRESSO package.

-- 
Alessandra Satta
Dipartimento di Fisica dell'Università di Cagliari
Cittadella Universitaria - SP Monserrato-Sestu Km. 0,700
09042 - Monserrato (Ca)
Tel.: +39 070 675 4847
Fax : +39 070 510171
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20041118/2b00efbe/attachment.html>


More information about the users mailing list