Graph-based quantum response theory and shadow Born-Oppenheimer molecular dynamics.

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Extended Lagrangian Born-Oppenheimer molecular dynamics (MD) simulations face challenges with sensitive chemical systems and unsteady charge solutions.
• Accurate electronic structure calculations are crucial for reliable MD simulations, especially for complex systems.

Purpose of the Study:

• To adapt graph-based linear scaling electronic structure theory for advanced molecular dynamics simulations.
• To enable stable simulations of complex chemical systems with fractional molecular-orbital occupation numbers.
• To develop efficient quantum response calculations for electronic states with fractional occupation.

Main Methods:

• Adaptation of graph-based linear scaling electronic structure theory to recent shadow potential formulations.
• Integration of preconditioned Krylov subspace approximation for extended electronic degrees of freedom.
• Introduction of graph-based canonical quantum perturbation theory for response calculations.

Main Results:

• Stable simulations of sensitive complex chemical systems with unsteady charge solutions are achieved.
• Linear scaling complexity and natural parallelism are maintained for both ground state and response calculations.
• Demonstration of accelerated self-consistent field calculations and stable MD simulations using self-consistent charge density-functional tight-binding theory.

Conclusions:

• Graph-based techniques combined with semi-empirical theory provide stable simulations for large chemical systems (tens of thousands of atoms).
• The developed methods are particularly suitable for semi-empirical electronic structure theory.
• This advancement significantly improves the capability for simulating complex chemical dynamics.