Understanding long timescale phenomena in atomic systems from accelerated molecular simulation methods

Metadynamics, Ascent Dynamics, Advanced Sampling, Molecular Dynamics, Energy Landscape

Walter, Nathan Peller

关键词: Metadynamics, Ascent Dynamics, Advanced Sampling, Molecular Dynamics, Energy Landscape;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/106166/WALTER-DISSERTATION-2019.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Conventional molecular dynamics simulations have been proven instrumental to the understanding of material behaviors. However, the temporal constraints of molecular dynamics simulations have limited attempts to capture long timescale dynamics of material systems, such as phase transitions or dynamics of supercooled liquids. This class of phenomena often requires the crossing by thermal activation of large energy barriers, which represent transition states separating two stable or metastable states of a system. Therefore, we propose two methods to simulate the escape of a system from a metastable state to a transition state, which, when repeated, leads to atomic trajectories of extremely slow or rare phenomena of non-equilibrium matter. The first method, all-atom Metadynamics, is a version of the popular advanced sampling method Metadynamics. By biasing over all atoms rather than collective variables, all-atom Metadynamics samples the potential energy landscape in all-atom dependent variable space. All-atom Metadynamics utility is displayed by studying the nucleation and crystal growth of a model Lennard-Jones Argon system, while also displaying the computational downside of Metadynamics, the scaling over simulation time. Thus, the second method, Varied Steepness Ascent Dynamics (Ascent Dynamics), is introduced as an alternative method with constant computational scaling over simulation time. Ascent Dynamics, based on former surface walking methods, is first verified and validated on several mathematical problems and on vacancy diffusion in a Lennard-Jones system, respectively, followed by application of the method to a two-dimensional polydisperse model liquid. Ascent Dynamics is shown to accurately sample the potential energy landscape without the large computational overhead.