LAMMPS Calculator and MD

class flare.md.lammps.LAMMPS_MD(atoms, timestep, trajectory=None, **kwargs)

Run MD with LAMMPS based on the ase.md.md.MolecularDynamics. It includes using LAMMPS_MOD to run multiple steps, and supports Bayesian active learning with flare.

Parameters:parameters (dict) – LAMMPS input commands.

backup(curr_trj)

Back up the current trajectory into .xyz file. The atomic positions, velocities, forces and uncertainties are read from lammps trajectory. The step, potential energy and stress are read from thermo.txt

Parameters:curr_trj (list[ase.Atoms]) – lammps trajectory of current run read by ASE.

step(std_tolerance, N_steps)

Run lammps until the uncertainty interrupts. Notice this method neither runs only a single MD step, nor finishes all the N_steps. The MD exits only when 1) the maximal atomic uncertainty goes beyond std_tolerance, or 2) all the N_steps are finished without uncertainty beyond std_tolerance.

Parameters:

• std_tolerance (float) – the threshold for atomic uncertainty, above which the MD will be interrupted and DFT will be called.
• N_steps (int) – number of MD steps left to run.

class flare.md.lammps.LAMMPS_MOD(**kwargs)

A modified ASE LAMMPS calculator based on ase.lammpsrun.LAMMPS, to allow for more flexible input parameters, including compute, fix/nvt, fix/npt etc.

Supported customized commands for LAMMPS input: - mass (set by arg masses) - package - atom_style, bond_style, angle_style, dihedral_style, improper_style, kspace_style - units (default: metal) - boundary - neighbor - newton - kim_interactions - pair_style (default: lj/cut 2.5) - pair_coeff (default: * * 1 1) - *compute - *region - *dump - group - fix - timestep - minimize - run

Note: - additional commands needed at the beginning can be specified in the arg model_init - additional commands needed after “pair_coeff” can be specified in the arg model_post

Non-customized input commands: - atom_modify sort 0 0.0 - read_data - fix fix_nve all nve - dump dump_all all custom trj_file (dump_period) id type x y z vx vy vz fx fy fz - thermo_style custom (thermo_args) - thermo_modify flush yes format float %23.16g - thermo 1

Customized parameters: - dump_period - thermo_args - specorder

calculate(atoms=None, properties=None, system_changes=None, set_atoms=False)

Modify parameters

flare.md.lammps.check_sgp_match(atoms, sgp_calc, logger, specorder, command)

Check if the lammps trajectory or calculator matches the SGP predictions

flare.md.lammps.get_kinetic_stress(atoms)

LAMMPS stress tensor = virial + kinetic, kinetic = sum(m_k * v_ki * v_kj) / V. We subtract the kinetic term and keep only the virial term. In the calculator, results[“stress”] += kinetic_atoms gives virial.