Stable and Efficient Linear Scaling First-Principles Molecular Dynamics for 10000+ Atoms

Area of Science:

• Computational Physics
• Materials Science
• Quantum Chemistry

Background:

• Linear-scaling (O(N)) methods have advanced density functional theory (DFT).
• Expectations exist for applying O(N) techniques to first-principles molecular dynamics (FPMD) for complex systems.
• However, O(N) FPMD simulations are scarce, with limited data on their accuracy and reliability.

Purpose of the Study:

• To demonstrate the feasibility of efficient and robust O(N) FPMD simulations.
• To establish reliable calculation conditions for accurate O(N) FPMD.
• To enable practical, self-consistent FPMD simulations for very large systems.

Main Methods:

• Combined the extended Lagrangian Born-Oppenheimer molecular dynamics method with the density matrix method.
• Utilized the linear-scaling DFT code Conquest.
• Investigated reliable calculation conditions for O(N) FPMD.

Main Results:

• Demonstrated that efficient and robust O(N) FPMD simulations are achievable.
• Established conditions for accurate O(N) FPMD calculations.
• Successfully performed practical, reliable self-consistent FPMD simulations on a system with 32,768 atoms.

Conclusions:

• Efficient and robust O(N) FPMD simulations are now possible.
• This enables accurate, large-scale simulations for complex systems.
• The methodology allows for practical, reliable self-consistent FPMD of significantly larger systems.