Dear all,
I have compiled the latest stable version of LAMMPS (3 Mar 2020) on three different machines and I am running small tests on a liquid system.
On one of the three machines simulations die with segmentation fault, on the other two I do not see problems. (Similar behavior also with LAMMPS 12 Dec 2018)
The issue seems to be with the long range electrostatic interactions (tests done on a different system that does not include them worked fine). All details are in the attached folder.
Any insight would be really helpful.
Best,
Davide
segfault.tar.gz (216 KB)
Apart from the segmentation fault, using long-range electrostatics, i.e. a kspace style with ReaxFF is incorrect.
ReaxFF uses a Wolf sum based approximation to coulomb interactions, so using it with a kspace style would double count the reciprocal space contributions.
The segmentation fault is due to some part of USER-REAXC not being fully ported to be used with pair style hybrid, which is a bad idea in the first place. Especially in your input where you seem to be mixing a gas phase parameterization with bulk water.
Axel.
Dear Axel,
Thanks for the reply.
I thought that when using pair_style hybrid the kspace_style would compute long range electrostatic only for atoms interacting via a consistent pair_style, now it’s clear.
Best,
Davide