Material Design.
Supercharged with AI.
Next-generation material design, supercharged with AI
Prototype less
Gain deeper insights
Make better materials
Reduce costs
Case studies
Capturing accurate 3D images of materials is challenging. 2D techniques are quicker, cheaper, and can often capture more. Using Polaron, 2D to 3D reconstruction is possible with a single image. This was validated against a 3D volume collected with XCT and FIB-SEM.
With Polaron, an NMC Li-ion battery electrode was optimised. A process-structure model was trained that learned the relationship between mixing and calendering, and microstructure using just 9 training images. The performance was optimised for a range of conditions.
Our Mission
We envision a future where AI not only enhances the discovery and design process but also helps create materials that are crucial for building greener infrastructure, reducing waste, and supporting the transition to renewable energy.
Our team
We’re hiring!
FAQs
Segmentation and 2D to 3D reconstruction models can be trained using a single micrograph, as long as it is large enough to capture the features you are interested in. Optimisation studies require between 3 and 50 images, depending on the number of process parameters to be explored. Micrographs must be cross sectional images of your material, but can be collected using a range of different imaging techniques e.g SEM, optical microscopes, EBSD. There are no specific magnification or resolution requirements - any features captured in your data will be analysed and included in the models.
Our algorithms have been validated on a broad range of materials, including alloys, battery materials, ceramics and more. The model learns solely based on the features in your data, and is thus largely material agnostic. See microlib.io for examples of reconstructions.
Segmentation allows users to extract statistical information about phase morphology, such as volume fractions, surface area, and correlation functions. Reconstruction allows users to perform 3D electrochemical and mechanical modelling to better understand how and why materials behave differently. New microstructures with a specific microstructural property (e.g. volume fraction) can also be predicted. Finally, optimisation allows users to directly determine the best manufacturing parameters for their material processing routes.
No - we understand that micrographs and processing data often constitute core IP for manufacturers. Any data you upload to Polaron is only used to train your private models, and is isolated from any users outside your organisation. Unlike many AI companies, Polaron does not develop centralised foundational models, instead training bespoke models from scratch on small datasets. This approach was designed to match the unique requirements of material manufacturing.
You can book a demo with one of our material science experts by following the link. We can then give you access to a demo and help you to understand how our tool can be integrated into your R&D workflows.
Request your personalised demo today!
Don’t just imagine the possibilities—experience them.