[Pw_forum] ERROR dexx is negative

[Eduardo Menendez]

Hi
Thank you Paolo for your answer.
If the code is not ready for spin-orbit calculation, let me know, or
please, someone let me know.
For collinear calculation I had no problem. I also got the code working for
spin-orbit and using  nqx1=nqx2=nqx3=1.

I have found the error with the versions v.6.0 (svn rev. 13079), and
v.5.4.0, the ones that I have tried.

The error arises after the first step of refining the calculation for EXX,
i.e.,
1) The codes makes a non-hybrid calculation.
2) There is a message message
EXX: now go back to refine exchange calculation,
3) then there is a self-consistent calculation
4) the code prints energies for each k-point and the Fermi energy,
5) then comes the error message

the end is like the following :

     the Fermi energy is     4.0251 ev

!    total energy              =   -1494.22228872 Ry
     Harris-Foulkes estimate   =   -1528.64317856 Ry
     estimated scf accuracy    <       0.00000023 Ry

     convergence has been achieved in   6 iterations

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine electrons (1):
     dexx is negative!   Check that exxdiv_treatment is appropriate for the
system
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

     stopping ...
the same message many times, I guess one for every MPI process.

The error also happens using the default ecutrho=4*ecutwfc.




---------- Mensaje reenviado ----------
From: Paolo Giannozzi <p.giannozzi at gmail.com>
To: PWSCF Forum <pw_forum at pwscf.org>
Cc:
Date: Fri, 28 Oct 2016 07:50:53 +0200
Subject: Re: [Pw_forum] ERROR dexx is negative
Hi Eduardo

     Error in routine electrons (1):
>      dexx is negative!   Check that exxdiv_treatment is appropriate for
> the system
>

does this happen at the beginning of the calculation, or towards the end?
which code version are you using?

                     ecutwfc = 70.0 ,! same error if use 80.0
>                      ecutrho = 180.0 , ! also fails if commented out
>

not good: ecutrho should be 4*70=280. You may however play with the
following parameter to speed up the calculation:


>                      ecutfock =180.0,  ! also fails if commented out
>

I am not 100% sure that the code works with spin-orbit and hybrid
functionals, though:


>                      lspinorb=.true.,
>                      noncolin=.true.,
>



Eduardo Menendez Proupin
Departamento de Fisica, Facultad de Ciencias, Universidad de Chile
URL: http://www.gnm.cl/emenendez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20161029/20c634df/attachment.html>