A theoretical study in extracting the essential features and dynamics of molecular motions: Intrinsic geometry methods for PF(5) pseudorotations and statistical methods for argon clusters.

摘要

We studied the problem of understanding and computing the essential features and dynamics of molecular motions through the development of two theories for two different systems. First, we studied the process of the Berry Pseudorotation of PF5 and the rotations it induces in the molecule through its natural and intrinsic geometric nature by setting it in the language of fiber bundles and graph theory. With these tools, we successfully extracted the essentials of the process' loops and induced rotations. The infinite number of pseudorotation loops were broken down into a small set of essential loops called "super loops", with their intrinsic properties and link to the physical movements of the molecule extensively studied. In addition, only the three "self-edge loops" generated any induced rotations, and then only a finite number of classes of them. Second, we studied applying the statistical methods of Principal Components Analysis (PCA) and Principal Coordinate Analysis (PCO) to capture only the most important changes in Argon clusters so as to reduce computational costs and graph the potential energy surface (PES) in three dimensions respectively. Both methods proved successful, but PCA was only partially successful since one will only see advantages for PES database systems much larger than those both currently being studied and those that can be computationally studied in the next few decades to come. In addition, PCA is only needed for the very rare case of a PES database that does not already include Hessian eigenvalues.