Machine Learning-Assisted Design of Doping Strategies for High-Voltage LiCoO2: A Data-Driven Approach

Man Fang, Yutong Yao, Chao Pang, Xiehang Chen, Wei Yutao, Fan Zhou, Xiaokun Zhang, Yong XiangBatteries 2025 · DOI

TL;DR

This paper leverages machine learning to predict the performance of doped LiCoO2 cathode materials for high-voltage lithium-ion batteries. By analyzing data from 158 doped systems, they developed models that accurately identify key doping elements and their properties (like ionic electronegativity) that enhance cycle capacity, paving the way for faster material discovery.

What Problem Does This Paper Address?

Lithium cobalt oxide (LiCoO2 or LCO) is a workhorse cathode material for lithium-ion batteries, offering high energy density. However, at high operating voltages, LCO suffers from detrimental phase transitions, leading to capacity fade and safety concerns. Doping LCO with other elements is a promising strategy to improve its high-voltage performance and stability. The challenge lies in the vast number of potential dopants and combinations, making traditional trial-and-error experimental approaches time-consuming and costly. First-principles calculations, while powerful, can also be computationally expensive. This paper aims to bridge this gap by using machine learning to efficiently predict the impact of doping on LCO performance, guiding experimental efforts towards more promising candidates.

Methods

The researchers employed a data-driven approach, collecting experimental data on 112 distinct doped LCO systems from existing literature. They then applied a multi-step machine learning methodology:

  • Feature Selection: To combat the "curse of dimensionality" with a relatively small dataset, they first performed feature selection. This involved a combination of mechanistic analysis and Pearson correlation analysis to identify the most relevant features (e.g., dopant ionic radius, electronegativity, doping amount, electrochemical testing parameters) that strongly correlated with discharge capacity.
  • Machine Learning Models: Five non-linear machine learning algorithms were used for modeling: Random Forest (RF), Gradient Boosting Machine (GBM), XGBoost, CatBoost, and LightGBM. These models were trained to predict the 1st discharge capacity (1DC) and 25th discharge capacity (25DC).
  • Model Optimization: To ensure optimal performance, hyperparameter tuning and data augmentation techniques were utilized.
  • Feature Importance Analysis: SHAP (Shapley Additive exPlanations) values were employed to understand the contribution of each feature to the model's predictions, revealing the most influential doping characteristics.
  • Experimental Validation: The insights gained from the ML models, particularly regarding the importance of ionic electronegativity, were experimentally validated by synthesizing and testing La-doped and Bi-doped LCO.

Datasets

The study utilized a dataset compiled from 13 published papers, comprising 112 distinct doped LiCoO2 systems with 23 different dopant elements. This dataset included 16 covariate variables (features) and 2 response variables (1DC and 25DC). No new benchmark dataset was introduced; rather, existing literature data was curated and analyzed.

Key Insights and Ideas

  • Machine Learning as a Material Discovery Accelerator: The study demonstrates that ML can effectively navigate the complex landscape of material doping, significantly accelerating the discovery of high-performance cathode materials compared to traditional methods.
  • Ionic Electronegativity is Crucial: Beyond ionic radius, which is a known factor, the research highlights ionic electronegativity as a critical, often overlooked, feature influencing LCO's high-voltage performance. Lower electronegativity appears to be beneficial for capacity retention.
  • Interpretable ML for Material Science: The use of SHAP values allows for a deeper understanding of why certain dopants perform better, moving beyond simple prediction to provide actionable insights for material design.
  • Combined Model Superiority: While individual models showed promise, the combination of RF for 1DC and XGBoost for 25DC prediction yielded the best results, suggesting that different algorithms might excel at predicting different performance metrics.

Key Results

  • High Prediction Accuracy: The developed ML models achieved impressive prediction accuracy. The RF model achieved an R² of 0.8882 for 1DC prediction, and the XGBoost model achieved an R² of 0.8318 for 25DC prediction. These are reported as the best-reported values to date for this type of prediction.
  • Identification of Key Features: The analysis identified ionic radius and ionic electronegativity as the top two features influencing discharge capacity.
  • Experimental Validation Success: Synthesized La-doped and Bi-doped LCO samples confirmed the ML model's predictions. La-LCO (with lower electronegativity) exhibited superior cyclability and capacity retention compared to Bi-LCO (with higher electronegativity), validating the ML-driven insights.

Limitations and Future Work

  • Dataset Size: While 112 systems is substantial for this specific problem, the authors acknowledge that larger datasets would further improve model accuracy and robustness.
  • Complex Interfacial Reactions: The study notes that the precise mechanisms behind the observed performance differences, particularly complex interfacial reactions, remain challenging to fully elucidate and are an area for future research.
  • Broader Application: The authors aim to develop more accurate and efficient ML algorithms to explore microscopic mechanisms and establish new paradigms for high-performance cathode material design, potentially extending this approach to other battery chemistries and material classes. They also plan to investigate multi-element doping strategies.

Summary updated

Abstract

Doping lithium cobalt oxide (LiCoO2) cathode materials is an effective strategy for mitigating the detrimental phase transitions that occur at high voltages. A deep understanding of the relationships between cycle capacity and the design elements of doped LiCoO2 is critical for overcoming the existing research limitations. The key lies in constructing a robust and interpretable mapping model between data and performance. In this study, we analyze the correlations between the features and cycle capacity of 158 different element-doped LiCoO2 systems by using five advanced machine learning algorithms. First, we conducted a feature election to reduce model overfitting through a combined approach of mechanistic analysis and Pearson correlation analysis. Second, the experimental results revealed that RF and XGBoost are the two best-performing models for data fitting. Specifically, the RF and XGBoost models have the highest fitting performance for IC and EC prediction, with R2 values of 0.8882 and 0.8318, respectively. Experiments focusing on ion electronegativity design verified the effectiveness of the optimal combined model. We demonstrate the benefits of machine learning models in uncovering the core elements of complex doped LiCoO2 formulation design. Furthermore, these combined models can be employed to search for materials with superior electrochemical performance and processing conditions. In the future, we aim to develop more accurate and efficient machine learning algorithms to explore the microscopic mechanisms affecting doped layered oxide cathode material design, thereby establishing new paradigms for the research of high-performance cathode materials for lithium batteries.
BibTeX
@inproceedings{fang2025machine,
  author = {Man Fang and Yutong Yao and Chao Pang and Xiehang Chen and Wei Yutao and Fan Zhou and Xiaokun Zhang and Yong Xiang},
  title = {{Machine Learning-Assisted Design of Doping Strategies for High-Voltage LiCoO2: A Data-Driven Approach}},
  booktitle = {Batteries},
  year = {2025},
  doi = {10.3390/batteries11030100},
}
Figure