Functional materials by computational design
The microscopic structure of materials plays a key role in determining macroscopic functionality. In this research theme, we apply our expertise in ML-driven simulations and amorphous structures to understand functional materials for a range of technological applications.
From atomic scale to real-world impact
ML-driven molecular-dynamics simulations have greatly improved our ability to link material structures to their properties. They have allowed us to overcome limitations in small system sizes and short simulation timescales, enabling the modelling of more realistic systems with disorder and structural heterogeneity. By mapping material characteristics – such as structure and composition – to their functional properties, we are unlocking new possibilities for innovation in material science.
Connecting structure and properties
Our group's ongoing research spans diverse functional materials:
- Phase-change materials (PCMs): These materials can be switched back and forth between a crystalline and an amorphous state, encoding “one” and “zero” bits, respectively. They have been used in re-writeable optical disks, solid-state memories, and emerging brain-inspired computing techniques. We have pioneered the "device-scale" modelling of PCMs (Nature Electronics, 2023) and continue to push forward atomistic simulations in this field (Nature Communications, 2025).
- Perovskite materials: We study optoelectronic materials used in photovoltaic devices or catalysis, which can present challenges for traditional computational modelling. In the perovskites class, our group has created realistic polycrystalline models of the emerging lead-free perovskite material BaZrS3 (JMC A, 2025), and modelled the order–disorder transition in LaMnO3 (arXiv, 2025).
- Graphene oxide (GO): GO is a material of immense interest. Yet, fully characterising and controlling its atomic-scale structure remains a challenge. With previous funding from EPSRC, we have studied the atomic-scale structure (Angewandte Chemie, 2024) and the mechanical properties of GO materials (Chemical Communications, 2025) using ML-driven methods. We continue to be interested in the role GO can play as a test case for atomistic machine learning methodology, as well (Digital Discovery, 2025)!
What is next?
Our vision is to generate fully atomistic, device-scale models of a variety of materials that can be connected directly with experiments. (A popular term here is creating “digital twins”.) This could facilitate the design, synthesis, and manufacturing of amorphous materials for applications in batteries, solar cells, catalysts, and beyond. We are always looking for new and important materials to study, so please get in touch if you are interested in collaboration!
