Joining the Dots between AI, Machine Learning and Materials Advances in Green Technologies

In collaboration with Johnson Matthey Technology Centre you will research/design workflows and digital tools to establish/evaluate links between ionomer distributions in catalyst layers (CL) and performance in fuel cells. Your aim: extract numerical parameters autonomously from Secondary Electron Hyperspectral Imaging data as inputs for predictive modelling, preparing the ground for a digital twin of the CL system. Your chance to enable novel R&D on materials for generation/use of “Green” hydrogen. 

Project Description:

This PhD project is co-designed with our industrial partner, Johnson Matthey (JM) – a global leader in sustainable clean technologies with over 200 years of scientific excellence at the forefront of sustainable technology innovation. JM’s expertise in emission control, green hydrogen, fuel cells, and recycling vital raw materials provides real-world application context for this research. The project is expected to develop novel autonomous imaging capabilities and data analysis tools aimed at closing knowledge gaps that hamper the optimisation of key nanostructures vital for fuel cell performance. The motivation for this project is to speed up R&D on materials for generation/use of “Green” hydrogen by enabling digital optimisation processes. Why is this project vital? Green hydrogen can be generated from electrolysers when powered by renewable electricity and is used in fuel cells to generate electricity to provide electrical power for zero emission transport applications such as trucks, cars, and trains. To maximise the efficiency of these energy conversions, the correct catalyst layer (CL) structures are vital. A key component of the CL is the ionomer; the proton conducting polymer that acts as a binder and the electrolyte that supports the electrochemical reactions. Optimisation of the ionomer amount and distribution within the CL allows tuning of the device to minimise the use of expensive platinum group metals (PGMs) and adjustment of the behaviour to suit different types of operating conditions. Visualizing and quantifying the ionomer distribution in a way that can directly link it to CL performance by digital means is the challenge this project tackles as a first step towards digital enabled optimization.

Why Materials 4.0?

Distinguishing the ionomer from its nanoscale carbon support requires chemical imaging with nm spatial resolution without damaging/altering the ionomer distribution during the imaging process. The supervisory team has established that cryo-Secondary Electron Hyperspectral Imaging (SEHI) offers such capability. To describe the complex ionomer distributions on all relevant scales a large number of SEHI data volumes needs to be acquired and inspected, rendering manual analysis impractical. Supervised Machine Learning (ML) is widely used for automated analysis of such data but non-viable as it necessitates even larger data volumes for training, can introduce errors due to user bias in training in the absence of any ground truth. Thus, this project will focus on unsupervised ML. In summary, the project aims to establish new workflows for cryo-SEHI data acquisition with increased throughput with autonomous ML. ML and data acquisition changes need to be researched, designed and tested in tandem, as faster acquisition tends to result in more noise which could affect chemical identification. A suite of compatible ML tools will be developed to autonomously extract: (1) localised chemical information for different length scales down to single voxel level as dictated by any digital models for which our data can be used as inputs, (chemical identification down to single voxel), spatial relationship extraction between specific chemical components, for instance ionomer thickness variations across CL thickness, ionomer connectivity parameters etc.

This project sits firmly in the area of Imaging and Characterization, thus underlies other areas of research in the Henry Royce Institute. We expect some of the workflows and autonomous digital analysis tools to be transferable to the areas such as Biomedical Materials, Electrochemical Systems and potentially input to the Modelling and Simulations. We will encourage the student on this project to use their Materials 4.0 CDT intra- and inter-cohort relationships as a platform for cross fertilisation between different areas that could benefit from methods and tools created in our imaging and characterisation project targeted at fuel cells. 

Supervisors:

1. Cornelia Rodenburg ([email protected])
2. Lyudmila Mihaylova ([email protected])

If you are interested in this PhD, we encourage you to contact the project supervisor(s) directly. 

Application Deadline:

Applications open until successful candidate is recruited (no later than Summer 2025)

Funding Notes:

This is a fully funded project, part of cohort 2 of the EPSRC CDT in Materials 4.0. CDT. The studentship covers fees (home & international), a tax-free stipend of at least £19,237 plus London allowance if applicable, and a research training support grant.

Candidates of all nationalities are welcome to apply; up to 30% of studentships across the CDT can be awarded to outstanding international applicants. Early applications from interested overseas candidates are encouraged.

The Materials 4.0 CDT is committed to Equality, Diversity and Inclusion. Five countries are represented in cohort 1. We would like to see a more gender-balanced cohort 2, so we strongly encourage applications from female candidates.

Enquiries:

For application-related queries, please contact Sharon Brown ([email protected]). Please note that each partner of the CDT in Materials 4.0 will have its own application process.

Application Webpage:

https://www.sheffield.ac.uk/postgradapplication/login.do

After the personal details, you need to ‘add research course’, and select ‘Doctoral Training Course’, and then ‘Developing National Capability for Materials 4.0’.