Simulation (mofa.simulation)#

Interfaces for functions which run physics simulation codes

DFT (m.simulation.dft)#

Interfaces for performing DFT calculations

mofa.simulation.dft.compute_partial_charges(cp2k_path: Path, threads: int | None = 2)#

Compute partial charges with DDEC

Parameters:

  • cp2k_path – Path to a CP2K computation which wrote a CUBE file

  • threads – Number of threads to use for chargemol

mofa.simulation.dft.load_atoms_with_charges(cp2k_path) → Atoms#

Load the structure labeled with partial charges from a DDEC run

Assumes chargemol has already been run

Parameters:

cp2k_path – Path to the CP2K run

Returns:

Atoms object complete with charges

MACE (m.simulation.mace)#

Run computations backed by MACE

class mofa.simulation.mace.MACERunner(lammps_cmd: list[str] | None = None, model_path: Path | None = None, run_dir: Path = PosixPath('mace-runs'), traj_name: str = 'mace_mp', md_supercell: int = 2, device: str = 'cpu', delete_finished: bool = False, lammps_pkg: str = 'ml-iap')#

Bases: MDInterface

Interface for running pre-defined MACE workflows

Employs ASE for geometry relaxations and LAMMPS for NPT MD.

delete_finished: bool = False#

Whether to clear completed files from disk

device: str = 'cpu'#

Which device to use for the calculations

lammps_cmd: list[str] | None = None#

Command used to invoke MACE

lammps_pkg: str = 'ml-iap'#

Which LAMMPS pair_style package to use for running MACE.

Note: You will need to save the model in the appropriate format with mace_create_lammps_model

md_supercell: int = 2#

How large of a supercell to use for the MD calculation

model_path: Path = None#

Path to the LAMMPS-compatible model file. Ignored if using ASE

run_dir: Path = PosixPath('mace-runs')#

Path in which to store MACE computation files

run_md_with_lammps(name: str, atoms: Atoms, min_steps: int, timesteps: int, write_freq: int) → list[tuple[int, Atoms]]#
run_molecular_dynamics(mof: MOFRecord, timesteps: int, report_frequency: int) → list[tuple[int, Atoms]]#

Run NPT molecular dynamics

Start from the last structure of the previous trajectory

Parameters:

  • mof – Record describing the MOF. Includes the structure in CIF format, which includes the bonding information used by UFF

  • timesteps – How many total timesteps to run

  • report_frequency – How often to report structures

Returns:

Structures produced at specified intervals

run_optimization(mof: MOFRecord, level: str = 'default', structure_source: tuple[str, int] | None = None, steps: int = 8, fmax: float = 0.01) → tuple[Atoms, Path]#

Perform a geometry optimization computation

Parameters:

  • mof – Structure to be run

  • level – Name of the level of computation to perform

  • structure_source – Name of the MD trajectory and frame ID from which to source the input structure. Default is to use the as-assembled structure

  • steps – Maximum number of optimization steps

  • fmax – Convergence threshold for optimization

Returns:

  • Relaxed structure

  • Path to the run directory

run_single_point(mof: MOFRecord, level: str = 'default', structure_source: tuple[str, int] | None = None) → tuple[Atoms, Path]#

Perform a single-point computation at a certain level

Parameters:

  • mof – Structure to be run

  • level – Name of the level of computation to perform

  • structure_source – Name of the MD trajectory and frame ID from which to source the input structure. Default is to use the as-assembled structure

Returns:

  • Structure with computed properties

  • Path to the run directory

traj_name: str = 'mace_mp'#

Name to use for MD performed with this class

mofa.simulation.mace.load_model(device: str, level: str = 'default')#

Load a MACE calculator and cache it in device memory

Parameters:

  • device – Which device on which to load MACE

  • level – What level of MACE to use