Machine Learning Assisted Electrode Development for Next Generation Batteries

In response to the growing demand for sustainable battery technologies, we developed a novel methodology to optimize and fabricate electrodes for next-generation metal-ion batteries. Our approach combines high-throughput experimentation with machine learning (ML) algorithms to accelerate the optimization of specific capacity and cycle life of batteries. To develop a proof of concept, we focused on nanoporous polymer-carbon (NPC) materials, known for their high surface area and tunable properties. We prepared 150 different NPC cathode combinations using automated tools, allowing us to generate a diverse set of samples quickly and efficiently for property evaluation. Each of these samples was then subjected to experimental testing to assess their electrochemical, spectral, and morphological properties using cyclic voltammetry, Raman spectroscopy, and electron microscopy. These techniques provided a wealth of data on the behavior and characteristics of each NPC combination under various conditions.<br/><br/>The experimental data collected from the testing phase was used to train a newly developed ML framework. This model was designed to analyze the data, identify patterns, and predict the most promising NPC combinations for aluminum-ion storage. Key features of the ML framework included training on both literature and experimental data to enhance its predictive accuracy, refining synthesis conditions to optimize the electrochemical properties of the NPC materials and identifying new combinations of precursors and synthesis methods to identify high-performance cathodes. The ML-guided approach led to the discovery of several new high-performance electrode materials, including fluorine-based precursors, multi-metallic structures, and polymer composites. Using the synthesis conditions and material combinations suggested by the ML framework, we experimentally fabricated NPC-based hybrid cathodes, which were then tested to validate their performance.<br/><br/>The ML-optimized NPC cathodes demonstrated high performance in aluminum-ion storage systems, achieving a specific capacity ranging from 197 to 250 mAh g^-1 at a current density of 500 mA g^-1, a high coulombic efficiency of 97% over 10,000 cycles, and long-term stability with performance maintained above 150 mAh g^-1 for over 5,000 cycles. Our novel methodology, combining high-throughput experimentation with machine learning, has proven effective in developing high-performance NPC cathodes for next-generation Al-ion batteries. This approach not only accelerates the discovery and optimization of new materials but also provides a framework for future advancements in battery technology. Continuous refinement and application of this machine learning-assisted methodology can help develop efficient, durable, and high-performing electrodes for next-generation electrochemical energy storage devices commercially.