arXiv Science⌕ Search

arXiv subjects

Mathew Chow

Publications and source records attributed to Mathew Chow.

2 recordsLinked to original sources

Can AI Agents Reliably Optimize Computational Chemistry Codes?

Artificial intelligence (AI) approaches have opened new possibilities in computational and theoretical chemistry. AI agents can now execute computational workflows autonomously, and their use in scientific software development represents another promising direction. Herein, we present a general strategy for the controlled optimization of scientific software using a dual-agent harness. By using separate worker and controller agents for optimization and evaluation, respectively, we find that this approach can reduce computational costs while preserving numerical accuracy for the computational chemistry methods examined. We apply this strategy to TIP4P water simulations in LAMMPS, as well as to Coulomb and exchange matrix construction and linear-response time-dependent density functional theory (TDDFT) calculations in PySCF. Among these applications, the largest speedups are obtained for TDDFT. We then extend this strategy from optimizing established packages to developing efficient parallel CPU- and GPU-accelerated implementations of an initially serial molecular dynamics code, resulting in speedups of two to three orders of magnitude. Finally, we apply the protocol to open-ended algorithm development by constructing an adaptive self-consistent field approach that coordinates the use of multiple solvers. We find that the resulting approach can converge some challenging systems and occasionally reach lower-energy solutions than those obtained with the standalone solvers tested. These applications demonstrate that a strategy developed for improving code efficiency may also support the systematic development of computational methods. Overall, our benchmark study provides an example of AI-assisted advances in computational chemistry.

physics.chem-ph↗

Extended Lagrangian molecular dynamics on vibronic surfaces in the nuclear-electronic orbital framework

Proton transfer is central to many processes of chemical interest. The simulation of proton transfer dynamics requires the inclusion of nuclear quantum effects, such as zero-point energy, nuclear delocalization, and tunneling. Herein, we introduce methods within the nuclear-electronic orbital (NEO) framework, where specified nuclei are treated quantum mechanically on the same level as the electrons, for the simulation of proton transfer dynamics. Specifically, NEO density functional theory is used to treat the transferring protons quantum mechanically, and the other nuclei are propagated classically on the adiabatic vibronic ground-state surface. We formulate a NEO extended Lagrangian molecular dynamics (NEO-ELMD) approach to incorporate the motion of the nuclear basis function centers during such simulations. Density matrix extrapolation and purification are introduced as a means to accelerate the NEO self-consistent field procedure at each time step by reducing the number of iterations required for convergence. We demonstrate the fidelity and efficiency of NEO-ELMD by comparison to related dynamics methods for intramolecular proton transfer in malonaldehyde. We also use these accelerated techniques to simulate the nonequilibrium single and double proton transfer dynamics of proton-coupled electron transfer in much larger benzimidazole-phenol systems. This work provides a foundation for future methodologies to efficiently simulate proton transfer dynamics within the NEO-DFT framework while incorporating nonadiabatic effects between adiabatic vibronic states.

physics.chem-ph↗