Extreme Scalability of DFT-Based QM/MM MD Simulations Using MiMiC

Area of Science:

• Computational Chemistry
• Biomolecular Modeling
• High-Performance Computing

Background:

• Accurate molecular simulations require combining quantum mechanics (QM) and molecular mechanics (MM).
• Existing QM/MM methods often face scalability challenges, limiting simulation length and complexity.
• Developing efficient multiscale modeling frameworks is crucial for studying large biomolecular systems.

Purpose of the Study:

• To present a highly scalable Density Functional Theory (DFT)-based QM/MM implementation within the MiMiC framework.
• To demonstrate the efficiency and scalability of this new approach for biomolecular simulations.
• To enable long-timescale molecular dynamics (MD) simulations using hybrid functionals.

Main Methods:

• Development of a loose-coupling QM/MM strategy using the MiMiC framework.
• Parallelization of electrostatic QM/MM interactions using distributed- and shared-memory approaches.
• Integration of CPMD (QM engine) and GROMACS (MM engine) for biomolecular simulations.

Main Results:

• The MiMiC framework demonstrates high scalability, achieving over 70% efficiency on up to 13,000 CPU cores.
• The loose-coupling strategy shows no significant computational overhead compared to tight-coupling schemes.
• Successful large-scale benchmark simulations of realistic biomolecular systems were performed.

Conclusions:

• The presented DFT-based QM/MM implementation offers a significant advancement in computational efficiency.
• This approach makes nanosecond-scale QM/MM MD simulations with hybrid functionals accessible.
• The MiMiC framework provides a flexible and powerful tool for advanced biomolecular modeling.