Abstract:The urgent global push towards decarbonization underlines the need for accelerated development of innovative energy storage technologies. In this context, the integration of data-driven research methodologies, such as machine learning and experimental automation, is emerging as a transformative strategy for expediting the discovery and optimization of electrochemical energy storage materials. This research work presents a pioneering approach within experimental materials science, combining automated thin-film experimentation with machine learning algorithms to significantly accelerate the development of novel battery materials while minimizing research time, resource consumption and overall costs. This contribution shows an autonomous materials acceleration platform developed and tailored for the rapid discovery of novel thin-film anode materials. This closed-loop system integrates high-throughput electrochemistry with sequential learning algorithms to dynamically determine the course of subsequent experiments. This data-driven optimization has facilitated rapid identification of optimal material compositions, resulting in a nearly 20-fold increase in the discovery speed and a more than 30% improvement in gravimetric energy density compared to conventional graphite anodes. Our platform incorporates thin-film material libraries synthesized via magnetron co-sputtering with a chemical composition gradient. These libraries are thoroughly electrochemically characterized in an argon glovebox using a scanning droplet cell (SDC) setup. This setup consists of a millimeter-scale three-electrode electrochemical cell, which is controlled by a potentiostat and driven by high-precision motors, with the capability of electrolyte exchange using a syringe pump. Additionally, the setup integrates a machine learning unit and an on-the-fly analysis module, all driven by a hierarchical orchestration framework. This framework facilitates real-time analysis of impedance spectra using physically interpretable equivalent circuits, calculating the specific capacity, energy density, and other critical parameters of electrode performance. In addition, the platform integrates additional high-throughput techniques such as µ-XRF and XPS for bulk and surface chemical analysis, respectively, and Raman spectroscopy for structural evaluation, enhancing our understanding of material properties and performance metrics. The novel Si-Ge-Sn ternary alloy system selected for this study, combines the advantages of the highest theoretical capacity of silicon, as well as mitigation of volume expansion during cycling and enhanced electrical conductivity of germanium and tin, making them suitable as fast-charging anodes. The material libraries were sputtered on 4" wafers, with a focus on optimizing the deposition process to achieve a wide compositional spread, high deposition yield, and amorphous phase across the material library, expected to enhance performance as battery anodes. Employing Bayesian optimization with a Gaussian process regressor, the material with the highest expected capacity was predicted after only 4 measurements, and the best performing material was identified after 14 trials. High-throughput Raman results highlighted shifts in Ge-Ge and Si-Si band positions that correlated with chemical composition changes, while XPS identified surface oxides that affect battery performance, explaining the discrepancy between theoretical and experimental results. The best materials have been selected for the scalability and long-term cycling tests at 1C dis-/charge rates, and their performance has been compared with the performance of pure elements. The results have validated the practicality of the high-throughput electrochemical experimentation approach, confirming the viability of the materials acceleration platform. In conclusion, this research demonstrates a transformative approach in battery material development by integrating machine learning with high-throughput characterization techniques. The autonomous materials acceleration platform developed herein significantly expedites the discovery and optimization of novel anode materials. This work highlights the pivotal role of AI and data-driven methodologies in advancing electrochemical energy storage technologies, paving the way for rapid, sustainable, and efficient advances, crucial for accelerating the discovery of novel solutions for advanced energy storage.

Automating Materials Exploration with a Semantic Knowledge Graph for Li‐Ion Battery Cathodes

Chemometric and Machine Learning Analysis of Lithium Concentration and Solvation Behavior in Li-Ion Battery Electrolytes

Machine Learning-Accelerated Discovery and Design of Electrode Materials and Electrolytes for Lithium Ion Batteries

Navigating Materials Chemical Space to Discover New Battery Electrodes Using Machine Learning

A document-level information extraction pipeline for layered cathode materials for sodium-ion batteries

Construction and Application of Materials Knowledge Graph Based on Author Disambiguation: Revisiting the Evolution of LiFePO4

Recent advance into understanding electro-chemo-mechanical behavior of lithium ion batteries by electron microscopy

Construction and Application of Materials Knowledge Graph Based on Author Disambiguation: Revisiting the Evolution of LiFePO 4

Machine Learning: An Advanced Platform for Materials Development and State Prediction in Lithium‐Ion Batteries

Deep learning of experimental electrochemistry for battery cathodes across diverse compositions

Combining High-Throughput Electrochemistry and Data-Driven Experimentation for Digital Discovery of Si-Ge-Sn Thin-Film Battery Anode Materials

Text to Insight: Accelerating Organic Materials Knowledge Extraction via Deep Learning

Solutions for Lithium Battery Materials Data Issues in Machine Learning: Overview and Future Outlook

Ab Initio Design of Ni‐Rich Cathode Material with Assistance of Machine Learning for High Energy Lithium‐Ion Batteries

Demonstrating Linked Battery Data To Accelerate Knowledge Flow in Battery Science

Application of Artificial Intelligence to Lithium-Ion Battery Research and Development

A comprehensive data network for data-driven study of battery materials

Materials Synthesis Insights from Scientific Literature via Text Extraction and Machine Learning

Unsupervised word embeddings capture latent knowledge from materials science literature

Harnessing artificial intelligence to holistic design and identification for solid electrolytes

Materials descriptors of machine learning to boost development of lithium-ion batteries