Skip to main content
Publication

How Machine Learning Can Extend Electroanalytical Measurements Beyond Analytical Interpretation

Authors

Mistry, Aashutosh; Johnson, Ian; Cabana, Jordi; Ingram, Brian; Srinivasan, Venkat

Abstract

Electroanalytical measurements are routinely used to estimate material properties exhibiting current and voltage signatures. Analysis of such measurements relies on analytical expressions of material properties to describe the experiments. The need for analytical expressions limits the experiments that can be used to measure properties as well as the properties that can be estimated from a given experiment. Such analytical relations are essentially solutions of the physics-based differential equations (with properties as coefficients) describing the material behavior under certain specific conditions. In recent years, a new machine learning-based approach has been gaining popularity wherein the differential equations are numerically solved to interpret the electroanalytical experiments in terms of corresponding material properties. Since the physics-based differential equations are solved, one can additionally estimate underlying fields, e.g., concentration profile, using such an approach. To exemplify the characteristics of such a machine learning assisted interpretation of electroanalytical measurements, we use data from the Hebb-Wagner test on a magnesium spinel intercalation host. As compared to the traditional analytical expression-based interpretation, the emerging approach decreases experimental efforts to characterize relevant material properties as well as provides field information that was previously inaccessible.Machine learning can simultaneously infer multiple physics-consistent material properties from electroanalytical tests, as well as describe underlying field variations.