|Institution:||The Ohio State University|
|Keywords:||Chemistry; Physical Chemistry; SAPT; perturbation theory; many-body expansion; chemistry; quantum chemistry; theoretical chemistry; ab initio; electronic structure theory; density functional theory|
|Full text PDF:||http://rave.ohiolink.edu/etdc/view?acc_num=osu1457973344|
We discuss the development and application of a number of fragmentation methods focused on understanding of intermolecular interactions in different systems. The advantage of fragmentation methods is to avoid the exponential growth of required computational power for the most advanced and accurate quantum chemistry theories which preclude the application in systems with large number of atoms and molecules. In those fragmentation methods, the full chemical system is partitioned into different subsystems, circumventing the exponential scaling computational cost. How this partitioning is performed and applied appropriately is the principal emphasis of this work. One of the fragmentation methods developed by our group, called extended XSAPT, combines an efficient, iterative, monomer-based approach to computing many-body polarization interactions with a two-body version of symmetry-adapted perturbation theory (SAPT). The result is an efficient method for computing accurate intermolecular interaction energies in large non-covalent assemblies such as molecular and ionic clusters, supramolecular complexes, clathrates, or DNA – drug complexes. As in traditional SAPT, the XSAPT energy is decomposable into physically-meaningful components. Dispersion interactions are problematic in traditional low-order SAPT, and the empirical atom-atom dispersion potentials are introduced here in an attempt to improve this situation. Comparison to high-level ab initio benchmarks for biologically-related dimers, water clusters, halide – water clusters, supremolecular complexes, methane clathrate hydrates, and a DNA intercalation complex illustrate both the accuracy of XSAPT-based methods as well as their limitations. The computational cost of XSAPT scales as third to fifth order with respect to monomer size, depending upon the particular version that is employed, but the accuracy is typically superior to alternative ab initio methods with similar scaling. Moreover, the monomer-based nature of XSAPT calculations makes them trivially parallelizable, such that wall times scale linearly with respect to the number of monomer units. XSAPT-based methods thus open the door to both qualitative and quantitative studies of non-covalent interactions in clusters, biomolecules, and condensed-phase systems. Advisors/Committee Members: Herbert, John (Advisor).