Our society is confronted with major challenges which call for innovative technological solutions. Some of the most pressing questions are : How can we produce chemicals from sustainable feedstocks instead of fossil resources? How can we maintain healthy indoor air quality? How can we provide potable water also in water-scarce regions? Nanoporous materials, including zeolites, metal-organic frameworks, covalent organic frameworks, and related porous materials, offer unique opportunities to address these challenges as they have the ability to selectively adsorb, separate, store, and transform small molecules. Their tunable pore architectures and chemical functionalities raise an intriguing question: can we design such materials predictively for a desired application? In this talk I will specifically discuss the role that molecular modelling can play in facing this challenge.
Over the past decades, molecular modelling has evolved from a primarily explanatory tool towards a more powerful predictive discipline that increasingly guides the interpretation of experiments and the discovery of novel materials. Significant progress has been achieved and it is now become possible to quantitatively predict properties such as adsorption isotherms, phase-transition temperatures, and other experimental observables.
Despite these advances, truly predictive design remains a formidable challenge and put enormous demands on various components of the molecular modelling exercise. Real materials are never perfect, they possess defects and disorder. Furthermore the behaviour of the material is strongly dependent on the operating conditions, for example active sites only appear at given conditions or structural changes are induced by the working conditions. This requests for a molecular modelling approach that captures the time behaviour across various scales and is able to follow the whole trajectory when a feed of molecules is sent over a nanoporous material. Capturing such complexity requires simulations that span multiple length and time scales while maintaining quantum-mechanical accuracy in the underlying energies. Some properties like selectivities in catalysis or adsorption capacities are extremely sensitive to small energetic differences, placing stringent demands on the electronic structure methods. Recent developments in machine learning potentials have opened new possibilities to bridge accuracy and scale. Yet major challenges remain, particularly for reactive systems, complex chemical transformations and systems with difficult electronic structures.
In this talk, I will discuss recent progress towards the predictive design of nanostructured materials. Examples will be drawn from technologically important applications, including carbon capture, photocatalytic chemical conversions and the conversion of emerging feedstocks such as CO₂ into value-added chemicals. These examples illustrate how advances at the interface of quantum mechanics, statistical physics, and machine learning are bringing molecular modelling closer to experimental observables and moving the field towards predictive design. However, it will become clear that further progress requires close interactions between multiple disciplines, including advanced characterization techniques, fundamental developments in electronic-structure theory, precise experimental synthesis, and multiscale modeling approaches. I will argue that molecular modelling starting from the atomic scale, can take us a long way towards predictive design, but that realizing this vision ultimately requires a synergistic effort across a broad range of scientific disciplines.