Understanding Default Finnis-Sinclair Embedding Function in Silicon Simulations with PACE

LAMMPS LAMMPS Development
1

Howdy everyone,

I’m running a simulation in LAMMPS with Silicon using an PACE potential(Performant implementation of the atomic cluster expansion), and I encountered a series of warnings regarding the embedding function defaulting to Finnis-Sinclair. The potential file Si_npj_CompMat2021.ace is from Performant implementation of the atomic cluster expansion (PACE): Application tocopper and silicon. **
I am aware that the FS is for metallic system and considered as lowest order of ACE, which may not be suitable for Silicon simulations. However, I don’t find additional files to indicate embedding function for Silicon.
** Here’s the warning message I received:
--------------------------------------------------------------------------

LAMMPS (7 Feb 2024 - Development - patch_7Feb2024_update1-70-ge51a656)
using 1 OpenMP thread(s) per MPI task
Reading data file …
orthogonal box = (0 0 30) to (162.93 162.93 135)
2 by 2 by 2 MPI processor grid
reading atoms …
26167 atoms
read_data CPU = 0.063 seconds
ACE version: 2023.11.25
Recursive evaluator is used
Loading Si_npj_CompMat2021.ace
Notice! No embedding-function is specified, but potential has linear embedding, default embedding-function: FinnisSinclair would be assumed
Total number of basis functions
Si: 21 (r=1) 6806 (r>1)
--------------------------------------------------------------------------

Here’s the partial LAMMPS in file
--------------------------------------------------------------------------
units metal
dimension 3
boundary p p f
atom_style atomic

read_data sphere.lmp group group1

pair_style pace recursive
pair_coeff * * Si_npj_CompMat2021.ace Si

Could someone explain why LAMMPS defaults to this embedding function? I would really appreciate any help.

2

This error message is from the (P)ACE library and not from LAMMPS itself.

This is not a choice made by LAMMPS or the LAMMPS developers. Please understand that the “pace” pair style in LAMMPS is just a thin layer around the (P)ACE library and all the main work is done by that library (which is not part of the LAMMPS distribution but downloaded separately). Thus to get more insight into these details, you are looking in the wrong place.
Try asking here: GitHub - ICAMS/lammps-user-pace

3

Thank you for your guidance.