Predicting Crystal System of Cathode Materials in Lithium-Ion Batteries Using Machine Learning Models

Student: Tyler Fu
Table: CHEM2
Experimentation location: Home
Regulated Research (Form 1c): No
Project continuation (Form 7): No

Display board image not available





[3] Raja, P. and Barron, A. Physical Methods and Nano Science. LibreTexts

[4] Manthiram, A. A reflection on lithium-ion battery cathode chemistry. Nature Communications 11, 1550 (2020).



Additional Project Information

Project website: -- No project website --
Presentation files:
Additional Resources: -- No resources provided --

Research Plan:

Background and Rationale:

Lithium-ion (Li-ion) batteries consist of three major parts: cathode, anode, and electrolyte. Each element engages in a redox reaction, in which certain reactants acquire electrons while others lose them. Functionally, the cathode acts as the oxidizing agent, seeking to grab electrons; conversely, the anode is the reducing agent, aiming to release electrons. The electrolyte, on the other hand, allows lithium ions to move between the cathode and anode. During the discharge process of a lithium-ion battery, positively charged lithium ions (Li+) travel from the anode to the cathode through the electrolyte [1]. The anode oxidizes lithium into lithium ions, which then bind to the cathode [2]. Simultaneously, electrons traverse from the cathode to the anode via a circuit, creating the flow of electric current. This research project focuses on the structural aspects of the cathode within Li-ion batteries.

            Li-ion batteries have many different types of cathodes that usually are crystals. A crystal is a repeating arrangement of atoms, and the smallest arrangement of atoms that can repeatedly produce a crystal structure is called a “unit cell”. There are 14 basic unit cells called Bravais lattices, falling within 7 main primitive crystal systems. The specific crystal system depends on factors like the distance between the corners of the unit cell and the angles between the edges of the unit cell [3]. Lithium ions can easily bind or unbind themselves from a crystal structure through the process of intercalation. Intercalation is one of the reasons lithium-ion batteries can discharge and recharge many times (Layered Structures and Intercalation Reactions). The aim of lithium-battery cathodes is to have the lowest reduction potential possible – a substance’s tendency to get reduced – while maximizing the reduction potential of the anode because the difference in redox energies determines the voltage of the battery [4]. 

The crystal system of a cathode significantly influences its electrochemical properties, directly impacting battery performance, such as capacity and voltage. The ability to accurately predict the crystal system is instrumental in estimating cathode performance for specific applications. Leveraging the capabilities of machine learning to handle complex data patterns and make accurate predictions, this research project sets out to achieve its primary objective: to develop a machine-learning model capable of accurately predicting the crystal system of a lithium-ion battery cathode based on the characteristics of the cathode material.

Research Questions: 
1) What is the best model that can predict the crystal system of a lithium-ion battery cathode based on the characteristics of the cathode material?
2) How to find/build the best model with highest prediction accuracy?
3) How to use the model to enrich understanding of the relationship between crystal systems and different cathode materials?

1) Research “The Materials Project [5]”, an open web-based platform with access.
2) Select an appropriate dataset from the Materials Project to use for this project.
3) Study and understand each data field collected in the dataset in Step 2.
4) Conduct exploratory statistical analysis to further understand the dataset including graphic display of data, identify any outliers, and correlation between each pair of variables.
5) Split dataset into training set and testing set.
6) Build preliminary models using machine learning.
7) Evaluate each model built in Step 6 and further fine tuning.
8) Identify the best model with highest prediction accuracy.
9) Write a report on my findings and identify limitations.

Risk and Safety:
Not applicable as these procedures do not involve any real experiment but rather use the data already exist from the Materials Project.

Questions and Answers

Iniitial project questions

1. What was the major objective of your project and what was your plan to achieve it? 

The major objective of my project was to predict the crystal system of a lithium-ion battery cathodes.  I planned to use machine learning and using public data to achieve that.

    a. Was that goal the result of any specific situation, experience, or problem you encountered?  

I learned at school that the system of crystal structure has major effect on the physical and chemical properties of lithium-ion battery cathodes.  Hence, the prediction of crystal system is important to estimate many other properties of cathodes. This motiviates me to research and build a model using machine learning to predict the crystal system reliably.

    b. Were you trying to solve a problem, answer a question, or test a hypothesis?

I was trying to solve a problem which was predicting the crystal system of a material using machine learning models.

2. What were the major tasks you had to perform in order to complete your project?

First, I had to find a suitable dataset, then I had to choose what kind of machine learning model to use, then I had to preproccess the data to train the model, and lastly I had to analyze the results and further fine tuning.  

    a. For teams, describe what each member worked on.

I did this by myself with no team members.

3. What is new or novel about your project?

    a. Is there some aspect of your project's objective, or how you achieved it that you haven't done before?

Yes.  This project was the first time I used XGboost classifier, a decision-tree based algorithm for machine learning so it was new to me.  I also developed coding skills throughout this project.

    b. Is your project's objective, or the way you implemented it, different from anything you have seen?

I read another paper that did something similar, but they used a random forest machine learning model, while I used XGboost, which is newer and better than random forest models. The model I made actually outperformed the one in the paper I read.

    c. If you believe your work to be unique in some way, what research have you done to confirm that it is?

I searched for other papers before starting the project to gain some initial knowledge about the topic. I found one other paper that used a different machine learning model. 

4. What was the most challenging part of completing your project?

   a. What problems did you encounter, and how did you overcome them?

The problems I encountered include finding an appropriate datasets with open-access to use and to develop computer programs to build the XGboost model which failed to run initially.

   b. What did you learn from overcoming these problems?

I learned to be persistent, do deep dive research, and repeat try-assess-revise cycles multiple times until I came up with a satisfactory final model with the best performance.

5. If you were going to do this project again, are there any things you would you do differently the next time?

I would reach out to more academic research centers to look for a larger, more comprehensive datasets so I can have enough data in both training and testing, as well as cross validation tests to prevent overfitting.  

6. Did working on this project give you any ideas for other projects? 

Working on this project gave me ideas for more projects related to energy storage such as methanol fuel cells and using microalgae to make biofuel for more sustainable energy generation.

7. How did COVID-19 affect the completion of your project?

COVID-19 did not affect the completion of my project because it was done in 2023 where most restrictions were already lifted.