LAMDA: Aiding Visual Exploration of Atomic Displacements in Molecular Dynamics Simulations

TL;DR

LAMDA tackles the challenge of analyzing state-to-state transitions in large molecular dynamics datasets by integrating a data-reduction, hierarchical clustering workflow with multi-view visual analytics and annotation provenance. It aligns transitions, encodes displacement information with tensor-based superquadrics, and provides an interactive pipeline (Reduction → Selection → Cluster Windows) to identify, compare, and interpret transition ensembles. The approach is validated through a case study on nanoparticle transitions, showing rapid clustering, detailed inspection of clusters, and actionable insights via the Scratchpad/export features. The work contributes a cohesive tool that enables analysts to move from coarse transition ensembles to detailed, interpretable unit processes, with practical implications for understanding material behavior under dynamic conditions.

Abstract

Contemporary materials science research is heavily conducted in silico, involving massive simulations of the atomic-scale evolution of materials. Cataloging basic patterns in the atomic displacements is key to understanding and predicting the evolution of physical properties. However, the combinatorial complexity of the space of possible transitions coupled with the overwhelming amount of data being produced by high-throughput simulations make such an analysis extremely challenging and time-consuming for domain experts. The development of visual analytics systems that facilitate the exploration of simulation data is an active field of research. While these systems excel in identifying temporal regions of interest, they treat each timestep of a simulation as an independent event without considering the behavior of the atomic displacements between timesteps. We address this gap by introducing LAMDA, a visual analytics system that allows domain experts to quickly and systematically explore state-to-state transitions. In LAMDA, transitions are hierarchically categorized, providing a basis for cataloging displacement behavior, as well as enabling the analysis of simulations at different resolutions, ranging from very broad qualitative classes of transitions to very narrow definitions of unit processes. LAMDA supports navigating the hierarchy of transitions, enabling scientists to visualize the commonalities between different transitions in each class in terms of invariant features characterizing local atomic environments, and LAMDA simplifies the analysis by capturing user inputs through annotations. We evaluate our system through a case study and report on findings from our domain experts.

Figures (7)

• Figure 1: Our alignment scheme: the transition $T'$ (blue) is being aligned to $T$ (orange). Both transitions contain an initial and final state, where atoms are colored by their changes in bond length. These changes indicate the areas that should be aligned. Pseudo positions ($q_i$) are calculated based on a transition's $\Delta f$ matrix. A point set alignment algorithm is then used to calculate a rotation matrix $R$ that results in an appropriate correspondence indicated by the dotted lines. The matrix $R$ is then applied to $T'$.
• Figure 2: An example of the superquadric visualization; it displays the local displacement around each atom without the need for animations. The top row of figures illustrates a typical atomic visualization of a transition, while the bottom row is the same transition visualized as a superquadric, colored with different strain invariants (\ref{['invariants']}). $K_1$ is the default value used for coloring superquadrics; the others are included for illustrative purposes as they contain higher-order information about the deformation.
• Figure 3: An overview of LAMDA's workflow. Initially, an ensemble of transitions and scalars of interest are provided as input. a.) Analysts interactively remove duplicate transitions in the ensemble and cluster them using the Reduction Window. b.) Analysts examine the results using the dendrogram and the heatmap to identify broad groups of transitions (T1). The Selection Window supports in-depth exploration of both clusters and transitions through clicking on the dendrogram and heatmap, respectively. (c. & d.) Clicking the dendrogram displays a Cluster Window that provides views and interactions that characterize the cluster (T2) and help evaluate its quality (T3). e.) Transitions and clusters from elsewhere in LAMDA can be stored in the Scratchpad, a centralized location for examining and organizing the analyst's selections to generate insights that can be later exported.
• Figure 4: A visual example of how three distinct transitions are aggregated into the Group Displacement visualization. This visualization is intended to provide overviews for multiple transitions, facilitating intra-cluster comparisons (R2) and providing a visual marker for cluster quality (R3). Saturated colors indicate high displacement correlations between members, while gray spheres indicate the correlation for that point is below an adjustable threshold.
• Figure 5: The Selection Window, where experts can explore a transition ensemble organized by cluster. The dendrogram (1) is used to explore the dataset. Experts can click its branches to open a Cluster Window to view individual transitions within their hierarchical context. The gray horizontal line is used to update the heatmap (2), which represents the distances between all transitions in the ensemble. The Scratchpad (3) allows experts to organize and export their insights through a simple WYSIWYG editor (R4), adding annotations and grouping objects of interest together from selections made across the interface. (4) provides interactions that change the visualizations in the Scratchpad; (5) shows LAMDA's export options (R5).