Ultra-Low-Kappa: Identification of Crystalline Materials with Ultra-Low Thermal Conductivity Based on Machine Learning Study

Introduction

This repository represents the work we did for accelerating the search of low thermal conducvitiy materials in materials science.

Dataset

The dataset contains 5486 crystalline materials with thermal conductivity property in total, downloaded from the AFLOW repository. The original data was transformed into four categories with 75 dimensions to describe the crystal materials by data preprocessing. They are saved as csv format for the sake of accessibility and reproducibility. In addition to the thermal conductivity, other targets of thermal properties like Debye temperature , heat capacity at constant volume, heat capacity at constant pressure, and thermal expansion coefficient are proviede in the lables.2020.1.29.csv file.

Usage

The training data of csv file can be loaded as

import pandas as pd
import os
data_path='./data'
labels = pd.read_csv(os.path.join(data_path, 'labels.2020.1.29.csv')).to_numpy()
raw_train_data = pd.read_csv(os.path.join(data_path, 'td.2020.1.29.csv')).to_numpy()

In our work, the results with best performance of ML models are saved in the result file. The following is an example to load ptc_ab.pkl model

import pickle
with open('./models/ptc_ab.pkl', 'r') as f:
  optimized_model=pickle.load(f)

Prerequisites

The ML models were developed using

• -Sklearn
• -Keras
• -Tensorflow
• -XGBoost

Other packages used in the work including

• -Pandas
• -Numpy

Installation

# python dependencies
pip install -r requirements.txt

# clone repo
git clone https://github.com/xinming365/Ultra-Low-Kappa.git
cd Ultra-Low-Kappa

Citing the related works

Please cite the related works in your publications if it helps your research:

@article{wang2020identi,
author = {Wang, Xinming and Zeng, Shuming and Wang, Zhuchi and Ni, Jun},
title = {Identification of Crystalline Materials with Ultra-Low Thermal Conductivity Based on Machine Learning Study},
journal = {The Journal of Physical Chemistry C},
volume = {124},
number = {16},
pages = {8488-8495},
year = {2020},
doi = {10.1021/acs.jpcc.9b11610},
URL = {https://doi.org/10.1021/acs.jpcc.9b11610}
}