The discovery of new materials is an immense challenge, with a vast design space and numerous success criteria. Microsoft has recently demonstrated an advanced approach to machine learning-assisted material discovery, particularly in the realm of lithium-ion battery electrolytes.
They began by exploring all possible structure types, decorating these structures with various atoms, leading to a pool of millions of candidate materials. The screening process went beyond simple stability checks to encompass a broad range of criteria, including predicted properties, electrode stability, and cost. This was achieved through various layers of filtering, leveraging data from diverse calculations, ranging from costly DFT and MD simulations to lower-fidelity calculations.
Microsoft Azure wisely positioned the expensive calculations towards the end of the pipeline, focusing resources only on the most promising candidates. Furthermore, they partnered with the Pacific Northwest National Laboratory Pacific Northwest National Laboratory to synthesize the compounds identified.
In our Materialism Podcast, we'll delve into this process, the challenges faced, and the future opportunities in this field, in conversation with Chi Chen and Nathan Baker
links below
itunes: https://lnkd.in/gM2zqmUC
spotify: https://lnkd.in/gyPC7m6e
youtube: https://lnkd.in/g-vT7AZQ
