Development of Density-Functional Tight-Binding Methods for Chemical Energy Science

Density-functional tight-binding (DFTB) method is an approximation to the popular first-principles density functional theory (DFT) method. Recently, DFTB has gained considerable visibility due to its inexpensive computational requirements that confer it the capability of sustaining long-timescale reactive molecular dynamics (MD) simulations while providing an explicit description of electronic structure at all time steps. This capability allows the description of bond formation and breaking processes, as well as charge polarization and charge transfer phenomena, with accuracy and transferability beyond comparable classical reactive force fields. It has thus been employed successfully in the simulation of many complex chemical processes. However, its applications for chemical energy science, particularly for the development of chemical energy storage systems and industry-scale catalysts are limited due to the lack of carefully validated parameters, specially for transition metal elements. In this dissertation, my previously developed DFTBparaopt semi-automatic parameterization was therefore employed for the development of two sets of urgently needed DFTB parameters involving prominent transition metals, (1) for phosphine-stabilized nanoscale gold clusters, and (2) for platinum nanoparticles in the vacuum and on TiO2 support. Performances of the newly developed DFTB parameters were validated extensively against DFT geometries, binding energies, electronic structure, and chemisorption for gold and platinum clusters. With the new parameters, the DFTB method can be used to investigate Au-cluster or Pt-clusters catalysts, and can be expected to pave the way for the future development of improved catalytic systems. In addition to these parameterizations, I focused on the development of an accurate DFTB-based linear-scaling method capable of describing zwitterionic systems for the simulation of electrolytes. To realize this method, the combination of the fragment molecular orbital (FMO) approach with the recently developed long-range corrected DFTB (LC-DFTB) method had been selected and implemented. Initial benchmarks and demonstration applications are presented. A satisfactory accuracy and computational performance of FMO-LC-DFTB has been demonstrated. The new method offers advanced capabilities for the first principles-based modeling of large-size ionic liquids, ions interaction with electrodes, and other chemical energy storage systems. The lessons learned in this dissertation point to areas for future improvements of the DFTB method that will enhance its accuracy and the transferability of its parameters even further.