Project Details
Projekt Print View

Efficient and scalable global structure optimization

Subject Area Theoretical Chemistry: Electronic Structure, Dynamics, Simulation
Term from 2015 to 2025
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 280675984
 
Final Report Year 2025

Final Report Abstract

Many research areas in chemistry (and in other natural sciences) require global optimization, i.e., finding the best of many local optima. Unfortunately, finding a global optimum with certainty requires computer time that increases exponentially with problem size, making this approach viable only for small toy problems. However, while non-deterministic global optimization does not come with a guarantee to find the global optimum, method developments in the past decades (including many contributions from the work group of the present PI) have made this approach useful and applicable to many problems of practical importance. In this project, our particular approach to non-deterministic global optimization (evolutionary algorithms) was applied to several different subfields of computational chemistry in which finding the best possible molecular structures is important, with respect to low energy or to other properties. As real-life test case for molecular self-organization on surfaces, we employed global optimization to compute the best possible arrangements of many ethyl pyruvate molecules on the (111) surface of platinum, in a collaboration project with the work group of Prof. Schauermann (physical chemistry). Additionally, we simulated scanning tunneling microscopy images and infrared spectra of this system in excellent agreement with experiment. In further subprojects, we extended this approach to also covering surface reactions in heterogeneous catalysis, investigating glycerol oxidation on mixed copper-cobalt hydroxycarbonates and isopropanol oxidation on cobalt-(II,III) oxide, in collaboration with Prof. Behrens (inorganic chemistry). In a collaboration with the organic chemistry group of Prof. Lindhorst we analyzed the kinetics and the reaction mechanism of a nitroaryl migration at several different sugars. When we repeated these calculations in globally optimized explicit solvent clusters, we discovered important features of this reaction that are completely misrepresented when solvent molecules are not included explicitly. It also turned out that present-day standard tools for studying chemical reactions computationally are not well equipped to include large clusters of solvent molecules. Improving this underdeveloped area will be the topic of future research. As a final subtask, we further developed our global optimization approach to catalysis by electric fields. Using global E-field optimization via abstract charge clouds, we could dramatically lower the activation energy barrier for a Diels-Alder reaction. We explored several ways to extend this approach and to translate abstract optimal E-fields back to real molecular catalysts, including first successful attempts to achieve this by decorating covalent organic frameworks with suitable substituents.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung