Application of Machine Learning on Designing Thermal Radiation Devices

Shenghong Ju*,Dezhao Zhu, Hong Wang

Shanghai Jiao Tong University, Shanghai, 200240, China

EXTENDED ABSTRACT: Thermal radiation which transmits energy through electromagnetic waves has wide applications in thermophotovoltaics, thermal sensing, thermal camouflage, and thermal radiative cooling. However, due to the multi-degrees of materials selection and structure parameters, it is extremely difficult to quickly design nanostructures with ideal thermal radiation properties using trial-and-error method. In this work, we introduce three successful thermal radiation devices design cases by combining the machine learning and electromagnetic wave transport calculations: (1) Designing aperiodic multilayered metamaterials with ultranarrow-band wavelength-selective thermal emission via Bayesian optimization. (2) Designing thermal radiative cooling metamaterials with quantum annealing algorithm. (3) Designing narrow-band thermal radiation structures via conditional generation countermeasure network model. The above cases have shown that the design of thermal radiation devices can be significantly improved by using machine learning, which has important guiding significance for the future thermal radiation  applications.

Figure 1.  Thermal radiation devices design via cCGAN

REFERENCES

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

[2] K. Kitai, J. Guo, S. Ju, S. Tanaka, et al., Phys. Rev. Res., 2(1), (2020) 013319. 

[3] J. Guo, S. Ju, S. Ju, S. Shiomi, Opt. Lett., 45(2), (2020) 343-346.

[4] A. Sakurai, K. Yada, T. Simomura, S. Ju, et al., ACS Central Science 5(2), (2019) 319-326.