Shenghong Ju*, Hong Wang

Shanghai Jiao Tong University, Shanghai, 200240, China 

EXTENDED ABSTRACT: Exploring materials with ultra-high or ultra-low thermal transport properties has wide applications in the fields of thermal design of electronic devices and thermoelectric energy conversion. The traditional trial­and-error style is inefficient and costly. In this work,we introduce three successful cases by combining the machine learning and thermal transport calculations: (1) The exploring diamondlike latticethermal conductivity crystals via feature-basedtransfer learning, (2) The high-throughput screening of interfacial materials with high/low interfacial thermal conductance, (3) The designing of highthermal conductivity polymer monomers based on the high-throughput molecular dynamics simulations and interpretable machine learning. Those successful cases have shown great advantage of designing materals with ultimate thermal property via machine learning

Keywords: Machine learning; High-throughput screening; Polymers thermal conductivity; Interfacial thermal transport.Crvstal thermal conductivity.

REFERENCES

[l] S. Ju, R. Yoshida, C. Liu, et al., Physical Review Materials, 5(5), (2021) 053801.

[2] N. Wang, J. Jieensi, Z. Zhen, et al., 23rd International Conference on Electronic Packaging Technology(ICEPT), (2022) doi: 10.1109/ICEPT56209.2022.9872718.
[3] X. Huang, S. Ma, H. Wang, et al., International Journal of Heat and Mass Transfer 197, (2022) 123332.

[4] S. Ju, S. Shimizu, J. Shiomi, J. Appl. Phys., 128(16), (2020) 161102.