Battflow - Automated Workflow for Properties Prediction of Liquid Electrolytes

Battflow is an automated workflow designed to identify and compute missing molecular dynamics (MD) and density functional theory (DFT) properties in a MongoDB-based liquid electrolyte database.

Overview

The workflow is composed of the following steps:

1. Query MongoDB for entries with missing properties.
2. Preprocess structures from SMILES information and setup force-fields and electrolyte composition.
3. Run MD simulations using GROMACS to compute diffusion properties and solvation-cluster statistics.
4. Perform molecular calculations using ORCA to obtain binding energies, HOMO, LUMO, and HOMO–LUMO gaps for individual components and the most populated solvation clusters.
5. Postprocess results and update MongoDB.

Installation

1. Clone the repository

git clone https://github.com/neubifx/Battflow.git

2. Create the Conda environment

This workflow relies heavily on ACPYPE for GAFF2 topology generation. Therefore, an Anaconda environment and a Linux-based system are required for successful execution.

cd Battflow
conda env create -f environment.yml
conda activate battflow_env

2. External software requirements

The workflow requires the following external software:

• Python ≥ 3.9
• MongoDB
• ASE
• GROMACS
• ORCA

3. Configuration

Ensure that gmx, orca, and any MPI executables are available in your system PATH when the Battflow environment is active. All configuration parameters are defined in:

config/default.yaml

This file can be modified to adapt the workflow to local machines or HPC environments.

MongoDB configuration

mongodb:
    host: localhost                   # Replace with your MongoDB Atlas host if preferred
    port: 27017
    database: working_db
    collection: working_collection

MD run environment

md_run_env:
    mdrun: "gmx mdrun"     # Command to run GROMACS
    mpiexec: ""            # Path to the MPI executor (e.g. output of "which srun")
    ncores: 8

The directory resources/md_run_files/ contains default GROMACS input files for minimisation, equilibration, and production runs.

DFT simulation configuration

dft_simulations:
    orca_profile: /usr/local/orca_5.0.3/orca    # Full path to the orca executable
    orca_input_block: "B3LYP 6-311+G(d,p)"      # You can change the level of theory and add additional parameters for ORCA simulations
    ncores: 8                                   # Number of cores
    li_energy: -203.567718                      # Energy of Li atom calculated at the B3LYP 6-311+G(d,p) level of theory. This value should match the methodology adopted on your simulations

Running the workflow

Typical usage:

python run_workflow.py --config config/default.yaml

The workflow extracts SMILES and concentration information from documents stored in MongoDB. It expects documents with the following structure:

{
  "_id": { "$oid": "681ba5b9aabcaf427f287f62" },
  "components": {
    "molecules": ["ec", "emc"],
    "anions": ["pf6"],
    "cations": ["methylimidazolium"],
    "ions": ["li"]
  },
  "smiles": {
    "molecules": ["C1COC(=O)O1", "CCOC(=O)C"],
    "anions": ["F[P-](F)(F)(F)(F)F"],
    "cations": ["C[N+]1=CNC=C1"]
  },
  "concentrations": {
    "molecules": ["1M", "0.8M"],
    "anions": ["1.2M"],
    "cations": ["0.2"],
    "ions": ["1M"]
  },
  "simulation_data": {
    "diffusion_coefficients": null,
    "transference_number": null,
    "solvation_statistics": null,
    "coordination_number": null,
    "pairing_percentage": null,
    "dft_energies": null
  }
}

Output fields populated by Battflow

1. Molecular dynamics transport and solvation properties

The field simulation_data will be filled with:

• Diffusion coefficients,
• Transference number,
• The three most common solvation structures,
• Their Cartesian coordinates (XYZ format),
• Coordination number for the solvated ion,
• Pairing percentage for the solvated ion

2. DFT energies of individual components

The field dft_energies will store:

• DFT energy,
• HOMO,
• LUMO,
• HOMO–LUMO gap

for each molecular component in the electrolyte formulation.