Subject: Parallelizing fix sgcmc with MTP in LAMMPS for Hybrid MD/MC simulation

LAMMPS LAMMPS Development
1

Hi everyone,

I am currently working with LAMMPS (version 22DEC2022) to conduct hybrid MD/MC simulations on a 5-element high entropy alloy system using Moment Tensor Potentials (MTP). I’ve built LAMMPS with the MC package and the USER-MLIP package necessary for MTP.

However, I’ve encountered a error when attempting to run simulations across multiple processors. The error message I receive is: ‘ERROR: Cannot run fix vcsgc with naive total energy calculation and more than one MPI process. (…/fix_sgcmc.cpp:251)’. Running the calculation on a single processor is possible but extremely time-consuming, taking over two days to complete just one timestep.

Could anyone advise on how I might parallelize the fix sgcmc process while using MTP forcefield?

Thank you

2

This is significantly outdated. If you want to do code development, you should be using the latest feature release 7 Feb 2024 or use the “develop” branch of the git repository.

Please have a look at the two functions of the fix: FixSemiGrandCanonicalMC::computeEnergyChangeEAM() and FixSemiGrandCanonicalMC::computeEnergyChangeGeneric()

To run in parallel with multiple MPI ranks, you would have to add a function that works similar to the former and produces the same results as the latter.