AI unlocks the manufacturing secrets of LMFP
A highly successful Faraday Institution Industry Sprint project has demonstrated the power of AI and machine learning to inform and accelerate battery materials development. A collaboration between WMG, University of Warwick and Imperial spin-out Polaron has optimised the manufacturing of a promising cathode material – LMFP- for increased cell performance. For Polaron the project has provided access to real world data and validation of the company’s technology, which will be vital to the company’s future growth.
LMFP in battery cathodes
LMFP has been under the microscope for the last decade as a potential alternative to cathodes commonly used today in commercial electric vehicle batteries.
Dr Gerard Bree, Assistant Professor at WMG, comments,
LMFP can give you more energy density than LFP, which means a greater range for your electric vehicle. Unlike NMC because LMFP does not contain any cobalt, it is a lower cost material and has a more secure supply chain. It will also provide a superior lifetime.”
| By the numbers | |
|---|---|
| £1M | Manchester Prize won by Polaron |
| 9 | Polaron employees |
| 27 | Electrodes manufactured by WMG on its production line |
| 3 | Cathode manufacturing parameters optimised |
Glossary:
LMFP – lithium manganese iron phosphate
LFP – lithium iron phosphate
NMC – lithium nickel manganese cobalt oxide
But use of LMFP is not without its problems. The material has challenging ionic and electronic conductivity, meaning the cathode must be manufactured using very small particles of LMFP active material to decrease diffusion distances. This leads to challenges with battery production.
Dr Bree explains,
When you’re mixing a cathode slurry, you want the particles to be well dispersed, so that the conductive and binding additives combine well with the LMFP across the cathode. But the small LMFP particles like to clump together in big agglomerates, which is a problem for electrode performance.”
These challenges around LMFP manufacturing were uncovered as part of the Faraday Institution Degradation project, which is co-led by WMG’s Professor Louis Piper.
Dr Bree says,
The manufacturing of LMFP is much less mature than LFP and NMC, and there was a desire to optimise this process, preferably in a way that doesn’t involve making thousands of different formulations with thousands of different mixes and cells.”
Simplified graph demonstrating the performance advantage of the electrode optimised via machine learning.
The manufacture of battery electrodes involves many complex, interdependent processes. Characterising the effect of process parameters on electrode properties and performance is crucial for the development of next-generation materials. Overcoming processing challenges using traditional design-of-experiment approaches is often inefficient, requiring extensive physical prototyping, incurring significant cost and slowing innovation.
Unleash the AI
WMG turned to Polaron for help. Founded in 2023 by Dr Isaac Squires, Dr Steve Kench, and Dr Sam Cooper, the company spun out of Imperial College London to commercialise technology developed as part of the Faraday Institution Multi-scale Modelling Project.
Polaron is helping companies design higher performing materials for battery applications and beyond (for example, in advanced alloys). The spin-out’s AI technology promises to enable advanced material manufacturers to accelerate their design process by modelling the complex relationships between processing parameters, the microstructures of the materials produced, and the performance of the resulting products.
The company’s AI material design platform uses microscopy images to “help research teams understand how to optimise the way that they make materials,” says Dr Kench, who is now Polaron’s CTO and is a former Faraday Institution PhD researcher.
The platform is able to characterise images of a component, looking at the different materials within it, and performing reconstructions to try and show the 3D structures of those materials.”
The resulting 3D images can deliver useful information around tortuosity, transport properties, particle sizes, and other factors that inform the way that a material behaves, potentially enabling greater optimisation.
Dr Steve Kench won the Faraday Institution Community Award for Innovation in 2024.
Polaron has backing from a major venture capital firm and several notable angel investors, and was the winner of the inaugural £1 million Manchester Prize, a government scheme rewarding companies developing AI for public good. This backing has helped it grow its team (which now stands at nine) and further develop its technology.
Sprinting ahead
The collaboration between WMG and Polaron took the form of a Faraday Institution Industry Sprint, which asked the question “what combination of LMFP manufacturing parameters gives us an optimum cathode performance?”
Dr Bree says,
We focused on two mixing parameters – the ratio of solid to solvent used in mixing, and the mixing time. These are two of the levers we can adjust during the mixing process to effectively disperse the LMFP. The third parameter we looked to optimise was electrode press density.”
A sample scanning electron microscope image (top left), processed by Polaron through machine-learning enhanced image processing (segmentation) to generate a 3D reconstruction. Active LMFP particles, carbon-binder domain (CBD) and pores can be independently visualised.
The WMG team (Rebecca Sellers and Daniela Proprentner) manufactured 27 electrodes on its pilot production line, using various combinations of these three parameters. The researchers then took SEM images of the electrodes and sent these to Polaron for 3D reconstruction. Using the images, Polaron was able to recommend the optimal parameter set to manufacture LMFP electrodes to maximise cell performance. The WMG team then assessed the recommendations via electrochemical testing in coin cells, validating the accuracy of Polaron’s predictions.
Dr Bree continues,
It saved us a lot of time and money. In the past you might have tested ten different densities and selected the best performing one. But in this study, Polaron recommended the co-optimisation of three manufacturing parameters (taking into account interdependencies), which we were able to validate through testing.”
Dr Bree says the project will be vital in informing the ongoing research and scale up of the material as part of the Faraday Institution Degradation project and elsewhere. The power of the technique for use in studying the manufacturability of other battery chemistries has also clearly been established.
When studying cell behaviour and cell degradation mechanisms, the data you can get are so much more meaningful if you’re working with the optimised industry-standard cells. The Industry Sprint has given us a much deeper understanding of the role of mixing parameters and densification on microstructure, and its effect on performance. This is something that can feed into all our research into LMFP, and give us clues about other materials.”
For Polaron, the access to real world data and validation of the company’s technology that the Industry Sprint has provided will be vital to its growth.
Steve Kench of Polaron concludes,
The benefits for us are humongous. It’s something we can show potential customers as validation of our technology. To be able to say we’ve helped make real cells perform better is a huge stake in the ground for our team.”
Case study published in September 2025.