High-performance gas sensors through predictive material design

Emerging applications in gas sensing, such as robotic olfaction for autonomous systems, remote spatial mapping and wearable point-of-care diagnostics, require fast and selective chemical recognition. Gas sensors are often based on nanostructured materials, but the diverse gas interaction pathways that occur on their surfaces obscure how structure governs sensing behaviour. Advances in isolated single-atom variants enable distinct analyte–surface interactions to link sensor performance directly to surface structure. Furthermore, machine learning-assisted modelling is now capable of simulating realistic sensing conditions. Here we propose a twin-loop strategy that integrates these advances into a predictive design framework. Interconnected inner loops link sensor performance, mechanism discovery and data-driven screening by tying the sensor metrics (such as response, selectivity, stability and speed) to experimental insights and atomistic models. An outer loop feeds the application requirements back into synthesis, enabling rigorous sensing material and surface design. Integrating atomic-scale control, modelling and iterative refinement facilitates the predictive discovery of next-generation sensing materials and the set-up of material libraries to bring an end to prevalent trial-and-error approaches.