Record Atomistic Simulation of Crystalline Silicon: Bridging Microscale Structures and Macroscale Properties

Area of Science:

• Computational Materials Science
• High-Performance Computing
• Nanotechnology

Background:

• Molecular dynamics simulations are crucial for understanding material properties at the atomic level.
• Simulating large systems of covalent materials like silicon has been computationally intensive.
• Advancements in supercomputing offer opportunities for unprecedented simulation scales.

Purpose of the Study:

• To implement and achieve the fastest molecular dynamics simulation for covalent crystalline silicon.
• To leverage the Sunway TaihuLight supercomputer for large-scale atomistic simulations.
• To enable virtual experiments for predicting macroscale properties from microscale structures.

Main Methods:

• Utilized the CovalentMD 2.0 software package for molecular dynamics simulations.
• Employed bond-order potentials for simulating covalent crystalline silicon.
• Executed simulations on the Sunway TaihuLight supercomputer using up to 10.3 million cores.

Main Results:

• Achieved a sustained performance of 16.0 Pflops in double precision for crystalline silicon simulation.
• Simulated over 137 billion silicon atoms, demonstrating high parallel efficiency (>80%).
• Simulated system dimensions exceeded experimentally measurable ranges and scale-dependent property limits.

Conclusions:

• The implemented simulation represents a record high for atomistic simulations of covalent materials.
• This capability enables direct prediction of macroscale properties from microscale structures for nanostructured materials.
• Opens new possibilities in nanotechnology, information technology, electronics, and renewable energy research.