Navigating the Space of Chemical Reactions from First Principles
Abstract
Traditional theoretical chemistry is hypothesis-driven. For example, a particular reaction may be studied to determine a mechanism and then one might start from this mechanism to investigate chemical functional group substitutions that could accelerate or decelerate the reaction by altering the barrier height. This strategy has sometimes proven quite successful, but it has drawbacks. Alternative competing reactions might exist. Optimal reactions may not be simply related to known reactions. An alternative discovery-based strategy uses simulation methods to discover new species and to “harvest” large numbers of potential reactions. Covering a large swath of chemical space potentially leads to unimagined reactions and intermediates. This approach could provide leads for new synthetic routes, such as novel “click” reactions, and elucidate complex chemical reaction networks, such as combustion. The chemical space spanned by combustion reactions includes many species and interconversion pathways. A discovery based approach could identify many of these, whose signatures could then be searched for experimentally. Technical Approaches: We have developed a new approach to chemical simulation that leverages GPU-acceleration of first principles dynamics methods. This new approach is termed the ab initio nanoreactor and consists of placing a number of molecules (?200 atoms) in a nanoscopic sphere (with spherical reflecting boundary conditions). The system is then heated and pressurized and reactions are allowed to occur freely, according to the laws of classical mechanics as dictated by the interatomic forces obtained by solving the electronic Schrodinger equation at each time step. Our previous work introduced the concept and here we propose to advance it by introducing feedback loops between discovery and microkinetic models, new artificial forces to enhance discovery, new machine learning methods to improve accuracy and efficiency of discovery, and automated database storage of discovered reactions. Furthermore, we will apply the reaction discovery workflow in the context of photoredox catalysis and molecular decomposition. Anticipated Outcome and ONR Impact: This new nanoreactor methodology is a completely different approach for discovering new synthetic pathways and/or modeling complex reaction chemistry. The development of a kinetic model allows one to reach long time scales, which is very difficult from a fully first principles dynamics approach. The key problem with the usual kinetic modeling approach is that it depends on the identification of all important reactions and intermediate species. Our nanoreactor discovers these and can thus make kinetic modeling approaches robust and less sensitive to unknown reactions and/or intermediates. We anticipate this approach will have a transformative impact on attempts to synthesize new materials and modeling of complex reaction processes like combustion and corrosion.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 06, 2021
- Source ID
- N000142112151
Entities
People
- Todd Martinez
Organizations
- Office of Naval Research
- Stanford University
- United States Navy