This repository comprises scripts used in the research paper titled 'Data-Driven Electrolyte Design for Anode-Free Sodium Metal Batteries'.

This repository contains the scripts used to perform a closed-loop Bayesian optimization of ether electrolyte formulations for anode-free sodium-ion batteries, plot the results using t-SNE, and test the robustness of this algorithm with a ground-truth model. The optimization model utilizes the volume fraction of ethers of a formulation (no chemical information) as Gaussian process model inputs to provide broad generalizability and accessibility to our methods. Experimental measurements were done in batches of 5 and follow the workflow shown below.

The data collected in this study is available in the Experimental Data folder to provide all tested formulation results to the broader scientific community. This consists of 45 formulations of varying mixtures of the electrolytes shown below.

Publication

• In Progress

t-SNE Map of Bayesian Optimization Progression

Whisker-Box Plot of GP Model Predictions and Measured Data

For detailed usage, navigate to the Jupyter Notebooks directory.

• scikit-learn v1.6.1
• numpy v1.26.4
• pandas v1.4.4
• scipy v1.7.3
• matplotlib v3.4.3

• Scalable Systems Lab
• Liu Lab
• Van Lehn Group