Nonadiabatic dynamics: molecular dynamics with electronic friction

Continuum of states of metal surfaces may lead to significant nonadiabatic effects, which are not accounted for in classical MD simulations. One method to include such methods, while employing Born-Oppenheimer, ground state PES is molecular dynamics with electronic friction (MDEF). In MDEF, frictional and random forces are included within the equation of motion, to account for nonadiabatic effects.

To run MDEF simulation, we first load PES model. Here, we load a MACE model, just as in MLIP-calculators.

pes_model = mace_model(ml_model_path, ase_atoms)
calculator_pes = mace_calc.MACECalculator(model_path=ml_model_path, device="cpu", default_dtype="float32")

Then, ODF model can be loaded. Here, we load an LDFA model, as described in EFT-LDFA.

friction_ids = [length(ase_atoms)-1,length(ase_atoms)]
eft_model_ml = ace_model(eft_model_path, atoms_julip)
density_model = AceLDFA(eft_model_ml; density_unit=u"Å^-3")
eft_model = LDFAFriction(density_model, atoms; friction_atoms=friction_ids)

Finally, we combine the MLIP, and EFT models using NQCModels.FrictionModels.CompositeFrictionModel.

model = CompositeFrictionModel(pes_model,eft_model)
sim = Simulation{MDEF}(atoms, model, cell=cell, temperature=parse(Float64,temp_surf)*u"K")

To run the MDEF simulation we can then execute the following line, using the constructed simulation (sim) object, using initial conditions included in distribution variable.

ensemble = Ensembles.run_dynamics(sim, (0.0, 3000*u"fs"), distribution; dt=0.1*u"fs",
                        ensemble_algorithm=EnsembleDistributed(), saveat=(0.0:austrip(0.1*u"fs"):austrip(3000*u"fs")))

More instructions on how to run the MDEF simulations and how to use the simulation output within the NQCDynamics.jl can be found here).