This page lists several common errors you will most likely encounter in both Quantum-Espresso and VASP, as well as how some users have had success overcoming them.
S Matrix not positive definite
This error is usually caused by two atoms in your unit cell overlapping. Check the positions of atoms remembering that periodic boundary conditions are usually enforced. Rattling atoms in the unit cell has fixed this error for some users.
Switching the optimizer may also help, particularly it may help using QE's internal optimizer.
Alternately, you can try increasing dw, the charge density cutoff.
forrtl: severe (174): SIGSEGV, segmentation fault occurred
This error can be caused by the node running out of memory. You can try the following:
- Reducing NPAR (it should still be a number that evenly divides the number of cores on the node)
- Reducing ENCUT
- Reducing vacuum between periodic slabs
- Parallelizing over multiple nodes (with KPAR = # nodes)
- Make sure you're requesting as much memory as possible from the node. On sherlock, use #SBATCH --mem-per-cpu=4000
- Use a node with more memory (e.g. bigmem nodes on Sherlock)