| Molecular dynamics is a powerful computational modelling tool developed to study the behavior of complex chemical and biological processes, such as protein folding and crystalline polymorphism. It is a numerical method that samples the configurational space of these complex systems with correct statistic weight. Aided by statistical mechanics, important thermodynamic quantities can be calculated from the simulation trajectories.;One of the most challenging problems in computer simulation is how to efficiently explore the rough energy landscapes. Brute-force molecular dynamics converges slowly. Several enhanced sampling methods have been proposed.;In the first part of this thesis, a new method that extends the Crystal-adiabatic free-energy dynamics is presented. It is based on the use of order parameters as collective variables for enhanced sampling. The crystalline polymorphism of xenon under high temperature and pressure is studied. Besides the fcc and bcc structures, new structures like hcp is predicted by this method.;The second part of this thesis describes the adaptation of the recently introduced stochastic resonance-free multiple time algorithm for collective variable-based enhanced sampling methods, such as adiabatic free-energy dynamics/ temperature-accelerated molecular dynamics, and unified free-energy dynamics. An one-dimensional harmonic oscillator, and fully solvated alanine di- and tripeptide systems are studied to demonstrate the significantly improved sampling efficiency obtained with this new techniques.;In the last part of this thesis, a novel method that efficiently calculates the free-energy differences associated with isotopic substitution using path integral molecular dynamics is discussed. A mass switching function and a set of free energy derivative estimators are introduced to reduce the computational cost of thermodynamic integration. This method is applied to study several systems including microhydrated lithium cation. The results show the well performance of this approach. |
