Shadow Molecular Dynamics and Atomic Cluster Expansions for Flexible Charge Models.

Area of Science:

• Computational Chemistry
• Materials Science
• Theoretical Physics

Background:

• Accurate molecular dynamics simulations require efficient and precise interatomic potentials.
• Flexible charge models are crucial for systems with charge transfer, but computationally demanding.
• Existing methods often involve significant computational overhead for electronic structure calculations.

Purpose of the Study:

• To present a novel shadow molecular dynamics scheme for flexible charge models.
• To integrate the linear atomic cluster expansion (ACE) for efficient potential modeling.
• To demonstrate the scheme's stability and accuracy using uranium oxide and liquid water systems.

Main Methods:

• Developed a shadow Born-Oppenheimer molecular dynamics (BOMD) scheme based on extended Lagrangian (XL) BOMD.
• Modeled interatomic potentials, including electronegativities, using the linear atomic cluster expansion (ACE).
• Emulated self-consistent charge density functional tight-binding (SCC-DFTB) dynamics with a second-order charge equilibration (QEq) model trained via ACE.

Main Results:

• The ACE+XL-QEq molecular dynamics simulations showed stability across a wide temperature range for both UO2 and water.
• The method precisely samples Born-Oppenheimer potential energy surfaces.
• Simulations of UO2 yielded accurate ground Coulomb energies, within 1 meV of SCC-DFTB results.

Conclusions:

• The proposed shadow molecular dynamics scheme with ACE provides a computationally efficient and accurate approach for flexible charge models.
• This method offers a viable alternative to computationally expensive machine learning potentials.
• The scheme successfully emulates complex electronic structure dynamics for diverse material systems.