PhD Studentship: The Self-Driving Microscope: Predicting Stochastic Failure in Solid-State Batteries using Physics-Informed AI

The development of safe, high-energy solid-state batteries is a key UK national priority, but it faces a critical bottleneck: the microscopic flaws that cause catastrophic failure (such as dendrite penetration, severe structural cracking, and short-circuiting) are unpredictable, fleeting, and hidden from standard tests. Capturing these elusive, stochastic events is a recognised 'grand challenge' for scientists.

This PhD studentship is the foundational computational element (WP1) proposed for a major 8-year project ("The Self-Driving Microscope”). The project’s goal is to build an autonomous, AI-piloted X-ray imaging platform that intelligently hunts for these hidden failure points in real-time.

As the first PhD student and a founding member of the new BASE (Beamlines for Autonomous Science and Engineering) Laboratory, your goal is to design and build the core predictive engine for this platform, teaching microscopes how to think and helping develop the next generation of beamlines.

This is a cutting-edge computational project at the intersection of AI, physics-based simulation, and materials science. Your key objectives will be to:

• Build a Multi-Scale Training Dataset - You will use 3D X-ray tomograms of next-generation solid-state cells (e.g., Li-metal/Li₆PS₅Cl) acquired at the I13-2 beamline (Diamond Light Source). The dataset will capture cells at pristine, aged, and post-failure stages.
• Develop the Core AI Predictor - You will explore and train advanced models, such as Graph Neural Networks (GNNs), to solve the key challenge: distinguishing benign aging from true failure precursors, features retrospectively confirmed to be the origin of a crack.
• Integrate Physics-Based Signatures - You will go beyond simple image recognition by using high-performance computing solvers (like the open-source OpenImpala developed by our group) to transform static 3D porosity maps into dynamic 3D maps of local tortuosity and ionic flux. This provides a direct, physically-grounded signature for the AI to learn from.
• Augment Data with Generative AI - You will use physics-constrained generative models (Diffusion models/GANs) to create a diverse library of synthetic microstructures, improving model generalisation against noisy, live experimental data. 

This studentship is fully and securely funded by the University's £9M Research England-funded M³4Impact expansion programme. This project is a cornerstone of the Computational Science and Engineering Group’s (CSEG) goals and will proceed as a key foundational project for the new BASE Laboratory. You will be fully embedded within the M³4Impact doctoral cohort and co-supervised by Prof. Andrew Kao, whose group will provide validated simulation models to benchmark the AI's prediction and Dr. Mikhail Poluektov.

As the founding PhD student of the new BASE Laboratory, the successful applicant will join a dynamic, growing research group. The lab is structured with a computational core at the University of Greenwich and an experimental hub at the Rutherford Appleton Laboratory (RAL). Your supervisor, Dr James Le Houx, is the Faraday Institution Emerging Leader Fellow at RAL and co-leads the UK's Battery Imaging BAG at Diamond Light Source. You will benefit from this direct link to national facilities and be part of a team dedicated to developing open-source computational tools for advanced materials discovery.

Advert information