CAREER: CAS-Climate: Data-driven Coupled-Cluster for Biomimetic CO2 Capture

点击翻译按钮获取中文摘要

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).WIth support from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry, Konstantinos Vogiatzis of the University of Tennessee-Knoxville is developing data-driven computational methodologies for the biomimetic capture of carbon dioxide. Carbon dioxide (CO2) overload in the atmosphere generates a significant greenhouse gas (GHG) layer, a major contributor to climate change in the United States and around the globe. Climate change presents a growing challenge to human health and safety, quality of life, and economic growth. Direct air capture (DAC) refers to technologies that capture CO2 directly from the air. One approach to DAC agent design relies upon chemical compositions that lead to favorable CO2 binding. Computational studies can examine different chemical environments and suggest new CO2-philic groups. Dr. Vogiatzis and his research group will develop new hybrid quantum chemical/machine learning models for the exploration of novel DAC approaches that are based on how enzymes can selectively capture CO2. Dr. Vogiatzis will also develop a new course offered at the upper undergraduate or early graduate level that aims to bridge data science with chemistry and provide important skills to undergraduate and graduate students. This course aims to reach students from underserved groups and provide a stimulating view of chemistry while training students in more expansive use of data science in chemistry.The primary objective of this project is to develop computational methodologies that capitalize on recent progress in data science for expanding the applicability of accurate quantum chemistry methods. Dr. Vogiatzis’ approach is based on the recycling of molecular wave functions obtained at low computational cost to help train machine-learning models which will provide fast and reliable energies and geometries of complex molecular systems without loss of accuracy. Coupled-cluster singles-and-doubles with perturbative triples (CCSD(T)) is a wave function method that balances accuracy with efficiency. Dr. Vogiatzis and his research group will develop transferable machine learning models that learn highly accurate CCSD(T) wave functions by utilizing data from low-cost methods such as Hartree-Fock (HF) and second-order perturbation theory. This data-driven coupled-cluster (DDCC) scheme is based on electron correlation, a property that has a local, short-range character across all molecular species, independent of their size. DDCC models can effectively encode the local nature of electron correlation and, after thorough testing and benchmarking, can be used for the examination of CO2-oligopeptide systems for biomimetic CO2 capture. Furthermore, the advances made here in combining quantum chemical methods with machine-learning are expected to be applicable to a significant variety of other chemical challenges.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.