Molecular Dynamics Workflow

Overview

This directory contains all updated scripts relevant in basic thermophysical and structural analysis of ionic liquids. The directory organization is kept in summary.txt and created by running `python summary.py'.

To-do

• create dataset workflow to save analysis into ML/AI-readable (e.g., SQL?)
• high throughput MD workflow (Galaxy, Materials Project)
• snakemake slurm workflow
• charge fitting automation
• implement polymerized ILs
• clean up TI code for OPLS generality and remove excess/redundant code
• benchmark IL TI calculations for # of H2O molecules and eq. time

Doing

• compare OPLS with SAPT FF

Done

PACE Consulting

• High-throughput Molecular Dynamics (MD) Overview

• typical MD calculation of 1 molecular system is atleast a GPU job and ~100 MB - 10 GB
• memory footprint of 1 system depends on simulation time (10-300 nanoseconds) and molecule size (roughly, 1G/10ns)
• 1 system may have ~100 parameters to study (temperature, concentration, etc.)
• studying ~1000s of systems quickly becomes ~10-100 TB
• manually submitting these jobs is prohibitive

• Solutions / Alternatives to NERSC Software

• FireWorks: Package many small jobs into a single large job (e.g., automatically run 100 serial workflows in parallel over 100 cores)
• MongoDB

Atomate2 / OpenMM

1. OPLS HTMD (See results):

• extend download_opls_xml() to use BOSS v5.1 (via Docker / docker-py)

  • See run_ligpargen for docker / apptainer implementation

• support for merging xml file outputs from ligpargen (see merge_xml_files)

  • Implement OPLS_LJ combination rules (if OPLS=True: OPLS_LJ(system) in openmm.jobs.core.py, see implementation)

• user support for binary/tertiary mixtures going from concentration --> count, i.e., openff.utils.create_list_of_mol_spec_lists()

• water_models

2. Modifying openmm.core/openmm.jobs

• equilibration_checker() (add arguments to OpenMMFlowMaker? Or same as EquilibrationDocs? Emmet?)

  • See count_direction_changes(x, y) and smoothness_index

  • Create analogous d_density_dt() and dU_dt() functions

  • Suppose I am running a production simulation of a new ionic liquid with an unknown equilibration timescale, under NVTMaker.run_openmm, one might opt to set the n_steps based on a physical parameter (i.e., see Openmm Issue#4756:

       def run_openmm_conditionally(self, sim: Simulation, dir_name: Path, dU_dt=False, density=False,...) -> None:
           """Evolve the simulation with OpenMM for self.n_steps.
   
           Parameters
           ----------
           sim : Simulation
               The OpenMM simulation object.
           """
           
   	should_continue = True
   	# Run the simulation
           while should_continue(sim):
   		sim.step(self.n_steps)
  
       def should_continue(self, sim: Simulation):
   	
   	# computes dU_dt
   	# computes d_density_dt
   	if property no within tol:
   		return True
   	else:
   		return False			
   	

• Interface Simulations (eventual surface tension calculations)

  • Can we add ase.buld.add_vacuum(atoms, vacuum) into openff.core creating topology via pack_box()

  • See pressure_reporter

3. Eventual support for SAPT-FF

• automate partial atomic charges via gdma from Psi4