Visualizing Energy Landscapes through Manifold Learning

Energy landscapes provide a conceptual framework for structure prediction, and detailed understanding of their topological features is necessary to develop efficient methods exploration. The ability visualise these surfaces essential, but the high dimensionality corresponding configuration spaces makes this difficult. Here we present Stochastic Hyperspace Embedding Projection (SHEAP), method energy landscape visualisation inspired by state-of-the-art algorithms reduction through manifold learning, such as t-SNE UMAP. performance SHEAP demonstrated its application Lennard-Jones clusters, solid-state carbon, quaternary system C+H+N+O. It produces meaningful interpretable low-dimensional representations landscapes, reproducing well known funnels, providing fresh insight into layouts. In particular, an intrinsic low in distribution local minima across space revealed.