Ab Initio molecular Dynamics with Quantum Nuclear Effects: Potential Surfaces and Gradients from on-the-fly Graph-Theory-Based Molecular Fragmentation Methods

Srinivasan S. Iyengar of Indiana University is supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry. Many problems at the forefront of energy, environmental and biological research demand the quantum mechanical treatment of electrons and nuclei, but the detailed quantum-mechanical description of such problems is much too complex even in today’s high performance computing environments.

This is because the computational complexity in these problems grows exponentially with system size, which makes them intractable. Iyengar and his research group are developing new computational methods to address these issues. These methods are based on a mathematical idea called graph-theory that allows Iyengar and co-workers to partition a molecular system into regions that communicate through an idea called electron correlation.

This is very similar to Google maps, where cities are connected through highways, and in the same way, in Iyengar’s formalism molecular domains are connected through similar roads and bridges that provide pathways for electrons to communicate through a concept called electron correlation. Unfortunately, while electron correlation allows electrons to communicate and has a critical role in all chemical processes, this concept is also responsible for the catastrophic computational complexity of obtaining accurate molecular properties.

By creating such graph-theoretic methods, Iyengar will help to reduce the computational complexity of these problems, to allow state of art calculations. These methods are poised to have major impact on the study of a wide class of problems in fields ranging from enzymology to atmospheric chemistry to materials science, including the study of hydrogen transfer in polymer electrolyte fuel cells.

In addition, the methods are also poised to allow innovative implementations on a mixed set of hybrid quantum and classical computing systems. The methods being developed by Iyengar are at the intersection of modern computational quantum chemistry and chemical physics. Hence students in the group have the opportunity to learn and develop new theoretical methods and apply these methods to important problems.

The results, involving computer codes as well as novel scientific ideas are to be disseminated to the scientific community. Specifically, the computer programs developed by Iyengar will appear as part of the NSF-funded SEAGrid science gateway. Furthermore, as a member of the quantum science center at Indiana University, and as the director of the university-wide scientific computing program, Iyengar will be involved in the organization of summer workshops for middle- and high-school teachers from the local Bloomington, Indiana area to provide cross-disciplinary training in chemistry, physics and computer science.

These workshops will focus on the quantum nature of matter, providing a unified treatment of problems in physics, chemistry and biochemistry; furthermore, modeling these problems is then to be done through connections to computational algorithms. Through involvement in the Holland Hudson Scholars Program (HHSP) and the Indiana Louis Stokes Alliance for Minority Participation (LSAMP) program, the PI will work to recruit students from under-represented groups.

Ab initio molecular dynamics (AIMD) is appealing, since it does not need a priori fitted potentials. This allows application of AIMD as a self-contained black box. But this advantage is deeply affected by the cost of evaluating the electronic potential and forces.

Hence, most applications of AIMD are limited to density functional theoretic (DFT) treatment. While there has been substantial progress in developing accurate DFT functionals, fundamental challenges remain. This proposal deals with the development and application of on-the-fly graph-theoretic techniques to compute accurate, low-scaling AIMD trajectories that are in agreement with post-Hartree-Fock electronic structure, but at the cost of DFT.

These developments are applicable for both cluster studies as well as periodic condensed phase problems, such as reactions on surfaces. In addition, during a single AIMD step, the approach can integrate multiple electronic structure packages. Current capabilities include the ability to use Gaussian, ORCA, Psi4, Quantum Espresso and OpenMX within a single AIMD umbrella.

There are three specific aims in this proposal: (1) to implement the team's graph theory-based approach in an asynchronous fashion on novel hybrid, interleaved, quantum/classical computing hardware. This will allow the steep scaling aspects of our method to be treated on quantum hardware, the lower scaling aspects and graph-theoretic decomposition of molecular structure on classical hardware and provide a new thrust for studying reactive chemical problems; (2) to study hydrogen transfer reactions on the surface of water.

The systems studied are in the condensed phase, and of critical importance in atmospheric chemistry. The reactions considered deal with isoprene-based hydroxy-peroxy radicals, thought to be pivotal on hydroxyl radical concentrations in the atmosphere. (3) The graph theory-based approach will be used to construct multi-dimensional potential surfaces for hydrogen-transfer reactions to gauge quantum nuclear effects.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.