MDI drivers

Official Resources

• Homepage: https://molssi-mdi.github.io/MDI_Library/
• Documentation: https://molssi-mdi.github.io/MDI_Library/
• Source Repository: https://github.com/MolSSI-MDI/MDI_Library
• License: BSD 3-Clause License

Overview

The MolSSI Driver Interface (MDI) is a standardized API that allows different computational chemistry codes to communicate and exchange data during runtime. It enables interoperability between codes (e.g., a quantum chemistry code and a molecular dynamics driver) without requiring them to be linked into a single executable. MDI drivers facilitate complex workflows like QM/MM, advanced sampling, and machine learning integration.

Scientific domain: Interoperability, code coupling, multiscale modeling, QM/MM
Target user community: Developers of computational chemistry software, researchers needing coupled codes

Theoretical Methods

• Client-Server Architecture
• Runtime Data Exchange
• Coupled Simulations (QM/MM, AIMD)
• Force Bridging
• External Driver Control

Capabilities (CRITICAL)

• Standardized communication between simulation codes
• Language-agnostic interface (C, C++, Fortran, Python)
• Support for MPI parallelization within and between codes
• Enables QM/MM without monolithic codebases
• Allows Python drivers to control compiled HPC codes
• Dynamic library loading or TCP/IP socket communication

Sources: MDI documentation, J. Chem. Theory Comput. (MolSSI publications)

Inputs & Outputs

• Input formats: MDI command-line options (-mdi), driver scripts
• Output data types: Exchange of coordinates, forces, energies, virials, charges, etc.

Interfaces & Ecosystem

• Supported Codes (MDI-compliant): LAMMPS, Quantum ESPRESSO, Psi4, Q-Chem, Molcas, OpenMM, Tinker, etc.
• Python: MDI Python package for writing drivers
• Standards: Defined by MolSSI (Molecular Sciences Software Institute)

Workflow and Usage

1. Launch code A (e.g., QM code) in MDI driver mode: psi4 -mdi "role=DRIVER ..."
2. Launch code B (e.g., MD code) in MDI engine mode: lmp_mpi -mdi "role=ENGINE ..."
3. Codes handshake and exchange data as defined by the MDI standard
4. Driver controls the simulation loop

Performance Characteristics

• Minimal overhead for communication
• Enables heterogeneous computing (e.g., GPU MD + CPU QM)
• Flexible coupling strategies (tight vs. loose)

Application Areas

• QM/MM simulations
• Ab initio molecular dynamics (AIMD)
• Machine learning potentials coupled to MD
• Advanced sampling (Metadynamics driven by external code)
• Multiscale modeling

Community and Support

• Developed by MolSSI (US NSF-funded institute)
• Active development and standardization
• Workshops and hackathons
• Growing adoption in major codes

Verification & Sources

Primary sources:

1. Homepage: https://molssi-mdi.github.io/MDI_Library/
2. GitHub: https://github.com/MolSSI-MDI/MDI_Library

Confidence: VERIFIED

Verification status: ✅ VERIFIED

• Website: ACTIVE
• Documentation: COMPREHENSIVE
• Source: OPEN (GitHub)
• Development: ACTIVE (MolSSI)
• Applications: Interoperability, code coupling, standard API