| The ubiquitous nature and unusual properties of water have prompted great interest in trying to better understand the structure and behavior of water and ice. Simulations on water using molecular mechanics for the potential energy surface and classical mechanics for the nuclear motion have been carried out for nearly 30 years, but although molecular mechanics potentials sometimes give good agreement with experiment for the properties against which they are parameterized, they often give poor results when used to calculate properties outside this set. As a result, there has been a great deal of work on using density functional theory (DFT) to calculate the potential energy for simulations. Since DFT is a quantum mechanical method the cost of these simulations is much higher; however, the added cost is rationalized by the hope that the energies obtained are more accurate and that the functionals are more transferable. Nevertheless, recent work on both water and ice has shown that DFT has not performed as well as hoped, raising the question of whether DFT-based simulations are worth their cost, and if they can be improved without making them more expensive. In Part I of this thesis a new density functional, PBE1W, which is specifically parameterized for water-water interactions, is developed and tested alongside a variety of density functionals commonly used in quantum chemistry and materials science simulations for its accuracy for small water and water-ion clusters. One alternative to using density functional theory in simulations is the use of wave function methods such as perturbation theory or coupled-cluster theory; however, the rapid scaling of the cost of these methods as system size increases makes them prohibitively expensive for use on large systems or in simulation. As a result, much work has focused on developing wave-function-based methods with favorable scaling for use on large systems and in simulation. In Part II of this thesis a fragment-based ab initio method for calculating the structures and energetics of large systems, the electrostatically embedded many-body (EE-MB) method, is presented, and extensions of the method to geometry optimization, frequency analysis, and molecular dyanamics simulations are discussed. |
