[Pw_forum] dE0s is positive which should never happen

[Alexandru Bogdan Georgescu]

Dear Forum,

I am currently trying to relax multiple structures using QE5 using PBE
pseudos with spin included in the calculation. However most of my jobs fail
with the following message:
     from bfgs : error #         1
     dE0s is positive which should never happen

Some unexplainably work perfectly, others that are nearly identical will
fail. What I've tried to do to bypass this is simply taking the last
coordinates and restarting from there, however I am in a situation where I
am constantly babysitting my jobs for this particular error.
The parameters I use are:
  ecutwfc = 30.0
  ecutrho = 280.0
and k-points are around 6 per unit cube of 8 atoms (I am mostly working on
perovskites, so I have a metal A, metal B and three oxygens, but I may
divide by 2 say if I have two unit cubes on top of each other, and so on).
I use the fairly low parameters for the ecuts because I am using the same
numbers for a surface calculation with over 60 atoms.

The commands I used to load the libraries I used to compile QE are (in case
the libraries might have to do with it as I saw in some previous messages)
module load Compilers/Intel/11/11.1.069
module load MPI/OpenMPI/1.4.4
module load Libs/MKL/10.2

I'm sorry if my question is silly, I just started using this code properly
and other members in my group didn't run into this problem as often. They
are using QE 4.2 if that helps.

Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20130511/c6fec443/attachment.html>