Interpolating moving least-squares methods for fitting potential energy surfaces: using classical trajectories to...

Area of Science:

• Computational Chemistry
• Chemical Physics
• Theoretical Chemistry

Background:

• Accurate potential energy surfaces (PES) are crucial for simulating chemical reactions.
• Traditional methods for PES construction can be computationally expensive, requiring numerous electronic structure calculations.
• Developing efficient PES generation techniques is essential for advancing molecular dynamics simulations.

Purpose of the Study:

• To develop and illustrate two new approaches for growing a fitted PES using the interpolating moving least-squares (IMLS) technique.
• To demonstrate the efficiency of these methods by calculating cis-->trans isomerization trajectories of nitrous acid (HONO).
• To assess the computational cost and accuracy of the proposed PES fitting strategies.

Main Methods:

• Utilizing the interpolating moving least-squares (IMLS) technique for PES fitting.
• Implementing two distinct approaches: IMLS-accelerated direct dynamics and dynamics-driven IMLS fitting.
• Employing classical trajectories and ab initio forces (HF/cc-pVDZ) for HONO isomerization calculations.

Main Results:

• Achieving convergence of the isomerization rate constant to approximately 10% with as few as 300 ab initio calculations.
• Demonstrating that neither method requires preliminary electronic structure calculations or initial PES approximations.
• Showing that Hessians are not required for the PES fitting process.

Conclusions:

• The developed IMLS-based approaches offer efficient and accurate methods for constructing potential energy surfaces.
• These techniques significantly reduce the computational burden of ab initio molecular dynamics.
• The methods are suitable for various applications, including classical trajectory simulations and other computational chemistry tasks.