Divide-And-Conquer Extended Tight-Binding Molecular Dynamics: A General-Purpose, Very Large-Scale Quantum Molecular...

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Accurate quantum molecular dynamics simulations are crucial for understanding chemical reactions and material properties.
• Traditional methods struggle with large systems due to high computational cost.
• Fragmentation-based methods offer a path to scaling, but maintaining accuracy is key.

Purpose of the Study:

• To develop a general-purpose, large-scale quantum molecular dynamics simulation platform.
• To combine the linear-scaling divide-and-conquer (DC) method with the semi-empirical extended tight-binding (xTB) method.
• To enable efficient and accurate simulations of large molecular systems.

Main Methods:

• Integration of xTB Hamiltonian construction modules into the DCDFTBK package.
• Development of the DCxTBMD platform leveraging a buffer region mechanism for accuracy.
• Implementation of energy gradients for molecular dynamics simulations.

Main Results:

• The DCxTBMD platform demonstrates suitability for efficient calculations of large systems, including those with multiple heteroelements.
• Investigations on the Fugaku supercomputer confirmed good parallel scalability and computational efficiency.
• The method achieves high accuracy despite using grid-based fragmentation.

Conclusions:

• DCxTBMD provides a powerful tool for large-scale quantum molecular dynamics simulations.
• The combination of DC and xTB methods overcomes limitations of traditional approaches.
• The platform is well-suited for studying complex chemical systems and materials.