Abstract: Phase transitions, metastability, and hysteresis in porous materials remain some of the most elusive phenomena in adsorption science — difficult to predict, and even harder to control. Yet controlling them is essential: hysteresis limits the efficiency of carbon capture, water harvesting, and adsorption-based cooling cycles. In this talk, I present a molecular-level framework for understanding and designing adsorption processes. Using Transition Matrix Monte Carlo (TMMC), we directly access free energy landscapes and characterize adsorption mechanisms beyond conventional isotherms, providing a broader picture that reveals the molecular origins of hysteresis and quantifies the associated free energy barriers. Building on this understanding, 

I show how we can move from analysis to design. Using a GPU-accelerated implementation combined with optimization techniques, we systematically tune host–guest interactions to control adsorption behavior. We demonstrate how modifying the distribution of interaction sites alters nucleation pathways and enables direct control over hysteresis: its width, shape, and ultimately its mitigation. Finally, I show how these concepts translate to realistic materials, such as metal–organic frameworks (MOFs), with applications in water harvesting, carbon capture under humid conditions, and adsorption cooling. Together, this work demonstrates how molecular simulations can not only explain adsorption phenomena but guide the design of next-generation functional materials.