New Nature Publication Alert: “Learning heterogeneous reaction kinetics from X-ray videos pixel by pixel”
This innovative paper builds on over six years of D3BATT’s groundbreaking research into reaction kinetics and phase separation in battery materials. Using machine learning, PDE-constrained optimization, and uncertainty quantification, we directly learn models for heterogeneous reaction kinetics from X-ray images. By utilizing every pixel in the X-ray data set, we quantitatively relate properties of the interfacial carbon coating to the lithium intercalation reaction rate. For more details, see the MIT News article, SLAC News article, interview with Brian Storey of TRI, and the full published paper.
For each particle pair, the left particle is the experimental data (interpolated to produce a smooth movie) and the right particle is a simulation of the learned model. The clock below the particles indicates the rate of charge/discharge.
D3BATT: Data-Driven Design of Li-ion Batteries
Our objective is to develop a novel multiscale-modeling framework for rechargeable batteries to accelerate materials discovery and design, guided by machine learning on large databases of experimental/simulated images/videos at matching length and time scales, electrochemical characterization, and atomistic calculations of material properties. Research Overview>
To learn more about the TRI Accelerated Materials Design and Discovery (AMDD) program, visit their website!