Presentation Information
[10p-E207-2]Machine learning potential driven prediction of thermal transport across substrate-Graphene/TMD-metal interfaces
〇(D)Wang Weitao1, Yunhui Wu1, Xin Wu1, Jiaqi Yang2, Daniel Capolat Palomar2, Marianna Sledzinska2, Masahiro Nomura1 (1.The Univ. of Tokyo, 2.ICN2)
Keywords:
Machine Learning Potential,Thermal Boundary Resistance,2D materials
The heat generated within three-dimensional integrated circuit architectures needs to be efficiently transferred to the package surface through multilayer interfaces in the out-of-plane direction. Also, reverse heat flow needs to be impeded to prevent the failure of thermally sensitive components is, however, challenging [1]. Here, we report the interfacial thermal transport in two kinds of van der Waals heterostructures including metal-2D materials-substrate. Unlike the traditional approach to predict the thermal transport at 2D materials interface with molecular dynamic (MD) simulation by empirical potential [2], we combine machine learning potential driven MD simulation for prediction and thermo-reflectance measurement for verification. For the 2D interlayer in these stacking structures, we including both monolayer graphene/WS2/MoS2 and bilayer twist graphene/WS2/MoS2/MoS2-WS2. The frequency-domain thermo-reflectance (FDTR) are performed on Au-graphene-SiO2 system, while time-domain (TDTR) are conducted on Al-TMD-Al2O3 system. From the phonon transmission level, the spectral thermal conductace and phonon wave packet simulations are further revealed the interfacial phonon behaviors. This work provides valuable insights into predict the interfacial thermal transport crossing substrate-Graphene/TMD-metal interfaces and useful guidance for thermal management in integrated circuit architectures.