Space-Time Multiresolution Approach to Atomistic Visualization: Application to Silicate Liquids

Ever-larger amounts of data related to the structural, electronic and mechanical properties of Earth materials (in their solid and liquid phases) are routinely produced by massively parallel computer simulations. The simulated data, which are time-dependent and three-dimensional in the nature, are not illuminating by themselves; gaining insight into them is, however, a non-trivial task. In order to take advantage of maximal information contained in various data related to modeling of materials including those of direct geophysical relevance, we have been developing a scalable and adaptive visualization framework, which aims to fulfill domain-specific needs thereby justifying the effectiveness of visualization for fundamental interdisciplinary science. In this presentation, we will talk about our recently developed interactive atomistic visualization scheme. We integrate the complex analysis and rendering of a given atomic position-time series by using an efficient space-time multiresolution approach. On the fly extracted data, which are rendered in an interactive manner, represent a given system at diverse length- (e.g., nearest or second-nearest neighbors or beyond) and time- (instantaneous, finite intervals, finite-time-span or overall averages) scales. In particular, the coordination environments and cluster structures are visualized to gain insight into the structural behavior whereas a variety of displacement data and covariance matrices are visualized to understand the dynamical behavior. It is shown that the interactive frame rates are achievable for systems consisting of up to a thousand atoms in a normal desktop environment. Such system sizes are typical in the case of today's common first-principles molecular dynamics simulations of complex materials systems including geophysically relevant silicate and oxide phases for which we justify the effectiveness and usefulness of our proposed scheme.