NPAR and NSIM affecting the stress tensor

[tomic]

I am working on the CaFe2As2 and I've been doing some benchmarking yesterday in order to determine the best NPAR/NSIM settings. Unexpectedly, I have discovered that resulting stress tensor varies when NPAR and NSIM are changed. I find this very strange and unsettling because some computations I am supposed to run rely heavily on reliable stress tensor evaluation. Does anyone know why this is happening and how to avoid it?

Thanks.

[admin]

for basis sets that are sufficiently high this should never happen. Please check if all ionic steps are fully converged electronically (increase EDIFF to higher accuracy if necessary) before the forces and stress tensors are calculated for each ionic step. If the default number of 60 el. steps per ionic step is not sufficient to reach convergence, please increase NELM