Computational spin dynamics and visualisation of large spin systems

<p>The thesis commences with a detailed review of the background theory of spin dynamics simulations. State space restriction is introduced via a "top-down" approach. Common terms that make up the spin Hamiltonian are reviewed, and it is noted that the mathematical forms of these ter...

Deskribapen osoa

Xehetasun bibliografikoak
Egile Nagusiak: Charnock, G, Gareth Charnock
Beste egile batzuk: Kuprov, I
Formatua: Thesis
Hizkuntza:English
Argitaratua: 2012
Gaiak:
Deskribapena
Gaia:<p>The thesis commences with a detailed review of the background theory of spin dynamics simulations. State space restriction is introduced via a "top-down" approach. Common terms that make up the spin Hamiltonian are reviewed, and it is noted that the mathematical forms of these terms can be categorised in one of three ways. The review of the background theory complete, accounts are given of the following four areas of research:</p><p><ul><li>1. Formal conditions are established for the validity of state space restriction via spin order pruning, based on tracking the density matrix norm through spin order subspaces. The primary predictor for success is seen to be the ratio of the largest eigenvalue to the relaxation rate. The lower this ratio, the fewer spin orders are required.</li><li>2. Software based around the Spin XML format, suitable for constructing and visualising large spin systems, is presented. Both a functional specification and a discussion of the internals are given.</li><li>3. A potential application of state space restriction, called "direct structure fitting", (DSF) is explored. In DSF, a candidate chemical structure is optimised directly by minimising the difference between its predicted spectrum and an experimental spectrum. The following examples of successful fits are provided: cyanomethyl, propargyl, and tyrosyl radicals, in the liquid state, and, tyrosyl embedded in two ribonuclease reductase proteins in the powder state.</li><li>4. A new model of the pseudocontact shift, which assumes a delocalised electron, is presented. Mathematical subtleties are resolved that would otherwise lead to the failure of numerical evaluation if left untreated. Techniques to improve efficiency are discussed, and the resulting program runs comfortably on workstation-grade hardware on protein sized datasets. E. Coli DNA Polymerase III is investigated as an example, and evidence is presented that suggests that the new model would predict significant differences in structures if used in conjunction with molecular dynamics based structural refinement.</li></ul></p>