schrodinger.application.matsci.mlip.integrator module¶

Integrators for the MLIP nanoreactor MD engine.

Implements Velocity Verlet and BAOAB Langevin integrators for use with the MDEngine.

Tensor shapes use M for the number of replicas and N for the number of atoms in each replica.

class schrodinger.application.matsci.mlip.integrator.Integrator(engine, dt)¶

Bases: object

Base integrator class providing the run loop for MDEngine.

Subclasses must implement the step() method.

__init__(engine, dt)¶

Parameters:

engine (MDEngine) – The MD engine to integrate
dt (float) – Timestep in femtoseconds

step()¶: Perform a single integration step. Must be overridden by subclasses.

run()¶: Run the integration loop until the engine signals stop, then close any async writers.

class schrodinger.application.matsci.mlip.integrator.VelocityVerlet(engine, dt=0.5)¶

Bases: Integrator

Velocity Verlet integrator for NVE dynamics.

__init__(engine, dt=0.5)¶

Parameters:

engine (MDEngine) – The MD engine to integrate
dt (float) – Timestep in femtoseconds

step()¶: Perform a single Velocity Verlet step, updating positions and velocities on mdv.

class schrodinger.application.matsci.mlip.integrator.BAOABLangevin(engine, dt=0.5, temperature=300.0, gamma=1.0, random_seed_generator=None)¶

Bases: Integrator

BAOAB Langevin integrator for NVT dynamics.

Implements the BAOAB splitting scheme (Leimkuhler & Matthews, 2013) with configurable friction and temperature.

__init__(engine, dt=0.5, temperature=300.0, gamma=1.0, random_seed_generator=None)¶

Parameters:

engine (MDEngine) – The MD engine to integrate
dt (float) – Timestep in femtoseconds
temperature (float) – Target temperature in Kelvin
gamma (float) – Friction coefficient in 1/fs
random_seed_generator (torch.Generator or None) – PyTorch random number generator for reproducibility

step()¶: Perform a single BAOAB Langevin step with Ornstein-Uhlenbeck velocity randomization, updating positions and velocities on mdv.