I am interested in using the interatomic potential described in DOI: 10.1021/jp982211k, which utilized GULP. The Buckingham, Coulombic, and three-body terms are already implemented in LAMMPS, though the core-shell term described in the paper is massless. Some digging through the mailing list suggests that people have attempted to implement a massless shell using the CORESHELL package by assigning a low mass to the shell particle, though there were reported issues regarding inter-shell interactions. Does anyone have any advice regarding implementing a core-shell potential with a massless shell in LAMMPS?
I am interested in using the interatomic potential described in DOI:
10.1021/jp982211k, which utilized GULP. The Buckingham, Coulombic, and
three-body terms are already implemented in LAMMPS, though the
core-shell term described in the paper is massless. Some digging
through the mailing list suggests that people have attempted to
implement a massless shell using the CORESHELL package by assigning a
low mass to the shell particle, though there were reported issues
regarding inter-shell interactions. Does anyone have any advice
regarding implementing a core-shell potential with a massless shell in
LAMMPS?
have you tried following the instructions for the existing core-shell
implementation?
the massless core/shell model as found in GULP is not implemented in the
current CORESHELL package. Technically one would need to add a
self-consistent optimization of the shell position following each MD step and
treat the core/shell pairs as frozen (fixed core/shell distance and orientation)
while propagating the system in time.
In my opinion such an implementation is possible in the LAMMPS framework,
however requires some effort. As Steve already wrote, the adiabatic core/shell
model is an approximation to the massless one. Therefore, it should also be
possible to test the same parametrization with the adiabatic core/shell model
(as has been done in literature: [1] vs [2]). Hereby you would need to assign a
suitable core/shell mass relation [1].
Keep in mind that typically the adiabatic core/shell model additionally requires
a shorter time step, although this reduced timestep is usually still less
computationally costly than the shell position optimization procedure.
Thank you for your advice. Would you be able to clarify what you mean by adding in a self-consistent optimization for the shell positions? I take that to mean that there needs to be some sort of minimization between the cores and their respective shells after each step, though this would not treat the core/shell pairs as frozen.
what I mean is a geometry optimization of the shell positions while all core positions fixed.
This optimization is performed in between every MD-step. Based on the forces on the cores
after such a shell-optimization the core positions are then propagated for the next time step.
The initial core/shell relative positions for each shell-optimization are usually taken from the
last shell-optimization performed. I labeled this condition as ‘frozen’ - I can see why this
might have been confusing.