Abstract

Aqueous solutions provide the essential medium for countless physical, chemical, and biological processes. Yet, despite decades of study, the microscopic fluctuations that underpin their thermodynamic and dynamical behavior remain elusive. In this talk, I will present recent efforts from our group to move beyond purely intuition-based approaches toward data-driven frameworks for understanding water’s complexity. Specifically, I will introduce an unsupervised learning protocol designed to quantify and interpret high-dimensional fluctuations in liquid water. In chemistry, we often rely on dimensionality reduction to construct simplified conceptual models that shape how we interpret experimental observations. Here, I will discuss the validity and limitations of such representations in two longstanding contexts: the proposed coexistence of high- and low-density liquid water, and the structural motifs of the excess proton (Eigen versus Zundel). I will conclude with broader reflections on how integrating data science with chemical physics can help reconstruct and refine our foundational notions of liquids and their collective behavior.