Data Science Approaches for Tracking Electrochemical Reactions and Phase Transformations

Abstract

Battery characterization techniques give critical information about what is happening inside the battery. However, fully utilizing the data obtained via battery characterization, particularly battery cycling can be difficult, because so many data points are produced during battery cycling. A more recent innovation in battery characterization is to leverage data science to more fully utilize information obtained during various battery cycling protocols. Herein, we introduce two examples of using data science approaches combined with the battery characterization to maximize the value of the characterization data: total differential capacity (dQ/dV) plots and galvanostatic intermittent titration technique (GITT). Total differential capacity plots are a powerful tool for understanding battery degradation when combined with various characterization techniques. However, due to the large amount of data, it is difficult to obtain quantitative conclusions via visual analysis, the typical approach for total differential capacity analysis. As a result, a data science approach was proposed to support quantitative conclusions for total differential capacity analysis. Herein, a software package has been developed to quantitatively analyze the battery cycling data from the dQ/dV plots. In addition, a user interface was built up and some documentations and demonstrations were set up to help the users have a better understanding of the codes and make it more accessible to use. Besides the application in tracking the electrochemical reactions within the battery, data science can also be combined with GITT to extract thermodynamic and kinetic data from a battery. This work focuses on generalizing a previously developed novel GITT model that accurately determined the lithium ion diffusions coefficient for a two-phase lithium iron phosphate electrode, so that the model can readily be applied to any phase transformation battery electrode system. We validate our generalized tool both by replicating the data from the previous work and using the generalized tool to determine the sodium ions diffusion coefficients in the two-phase region during the sodium antimonide battery charge-discharge cycling. In the standard GITT diffusivity model, the diffusion coefficients can appear to be many of magnitude lower in the two-phase region. This apparent drop occurs because phase transformation is not considered in the GITT diffusivity model. Therefore, the novel phase transformation GITT diffusivity model that we develop into a tool in this work takes both the diffusion and the interface migration into account. By solving the 1-D diffusion partial differential equations (PDEs) with the corresponding moving boundary conditions, the concentration profile, and the diffusion coefficients in both α and β phase can be determined. The diffusion coefficients obtained by the new model in the two-phase region are about 10^(−15)~ 10^(−14) cm2/s for α and β phase, which is validated by the similarity of the diffusion coefficients calculated by the standard model in the single-phase region.