[QE-users] MPI error using using simple.x

[Anibal Thiago Bezerra]

Dear Quantum Espresso Users and Developers,

I'm using simple.x to get the dielectric function for a gold-Aluminium
alloy. With pure systems (Au and Al), I had no problems at all. For the
alloy, the supercell has 12 atoms and the calculation ran with the
parameters similar to the ones in the example. However, I got an
anisotropic response. I already evaluate such a structure using epsilon,
and it returned isotropic dielectric function as expected for a cubic
lattice (I had to use as many k-point as I could do the calculation - for a
coarse grid, epsilon.x returned anisotropic response too).

Therefore, I increased the k-point (from 20 20 20 to 30 30 30, and to 40 40
40) in the simple.x and simple_ip.x calculations, and I failed. Simple.x
finished the calculations with no errors (at least no error messages). The
simple_IP, however, was abruptly finished by the MPI manager with the
following message:
 k-point block:         1093

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   PID 13719 RUNNING AT quantum01=   EXIT CODE: 9
=   CLEANING UP REMAINING PROCESSES
=   YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
===================================================================================
   Intel(R) MPI Library troubleshooting guide:
https://software.intel.com/node/561764
===================================================================================
After a kick search, I've found that such kind of error occurs when some
an exception like division by zero happens. I've tried to tweak the
parameters, the number of bands (I increased the number of bands, I used
the same number of bands as the KS-states, I tried to use the non-local
commutator, I increased the parameter s, and also decreased it). Hence, I
had no success.

Do you have any hint? Any workaround?

Thanks a lot for your help!!

Best!

Anibal Bezerra
The Federal University of Alfenas
Alfenas
Brazil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200915/5ecd9308/attachment.html>