Linear scaling algorithm for tight-binding molecular dynamics simulations

Area of Science:

• Computational physics
• Materials science
• Quantum chemistry

Background:

• Linear scaling (O(N)) methods enable theoretical treatment of large atomic systems.
• Efficient algorithms are crucial for large-scale molecular dynamics simulations.

Purpose of the Study:

• To develop a novel linear scaling algorithm for tight-binding molecular dynamics.
• To enable accurate simulation of large systems without full matrix diagonalization.

Main Methods:

• A divide-and-conquer approach was implemented, dividing the system into subsystems.
• The method avoids constructing the density matrix or electronic density.
• Physical properties are accessed through subsystem calculations.

Main Results:

• The algorithm was applied to a tungsten metallic system.
• Results for atomic structures, melting point, defect formation energy, and electronic properties matched exact diagonalization.
• The method demonstrated linear scaling complexity and high parallel efficiency.

Conclusions:

• The proposed method is effective for large-scale tight-binding molecular dynamics simulations.
• It offers advantages in linear scaling, memory consumption, and parallel efficiency.
• This approach facilitates the study of complex, large material systems.