Crystal structure prediction as a tool for chemical space exploration

Speaker: Prof. Graeme Day, University of Southampton
Series: Lennard-Jones Centre

The properties of molecular crystals depend on both molecular structure and crystal packing, making their discovery a challenge spanning chemical and structural space. Crystal structure prediction (CSP), coupled with property calculations, provides a route to anticipating materials’ behaviour before synthesis. I will introduce how energy-structure-function maps have guided the experimental discovery of functional molecular crystals [1,2]. These successes motivate a broader question: how can we efficiently identify promising molecules within the vast chemical space available for materials discovery?

I will present an evolutionary approach that embeds CSP within the search through chemical space [3], assessing candidate molecules using the predicted properties of their crystal structures. Applied to organic molecular semiconductors, the performance of this approach is demonstrated by optimising electron mobility across a space of substituted pentacenes. Incorporating crystal packing identifies candidates with substantially higher predicted mobilities than searches guided by molecular properties alone. Crucially, reduced-cost structure searches retain sufficient information to guide molecular evolution effectively, making the approach computationally practical. This work demonstrates how CSP can become an active component of molecular design, directing chemical-space exploration towards molecules whose packing preferences lead to the desired material function.

Functional materials discovery using energy-structure-function maps. Nature,543, 657–664 (2017).
Porous isoreticular non-metal organic frameworks, Nature, 630, 102–108 (2024).
Exploring organic chemical space for materials discovery using crystal structure prediction-informed evolutionary optimisation, Nature Communications, 16, 10540 (2025).