Session: Research Posters

Paper Number: 113737

113737 - Atomistic Simulation of Gan/diamond Interface Through Direct Bonding With Amorphous Interlayer and Novel Nanostructures 

With GaN being widely used for high power electronic applications, the output power of GaN-based transistors is limited by the high channel temperature caused by local heating that will affect its performance and reliability. Since diamond has the highest thermal conductivity among the natural materials, it is considered to be a good material to be integrated with GaN to enhance heat dissipation of the device. Previous experimental work has shown a very low thermal resistance of GaN/Diamond interface which is bonded with an amorphous Si interlayer. Therefore, it is of interest to understand the mechanism that enhance phonon transmission across such a bonded interface though phonon modeling and nanoscale heat transfer measurement.

Molecular Dynamics (MD) simulation is a commonly used method to study thermal transport at the nanoscale, yet its results depend heavily on the chosen interatomic potential in the simulations. On the other hand, First principle calculation is accurate, but it is also computationally expensive and is hard to be applied to a large interface system. With the recent developments in machine learning method, an interatomic potential can be trained based on DFT data and later be used in MD simulation. Such a potential is able to predict results in a quantum-accurate level, while has computation speed close to an empirical potential.

In this study, Molecular Dynamics simulations has been used to study the interfacial thermal transport across GaN/Diamond interface that is bonded by an amorphous Si interlayer. A machine-learning interatomic potential is trained and used in the simulation to provide quantum-accurate results. Non-Equilibrium Molecular Dynamics (NEMD) simulations is performed and interface thermal boundary conductance (TBC) is calculated. The effective TBC of the system, which includes two interfaces and an amorphous interlayer, from NEMD simulations is in good agreement with the previous experimental results, proving the accuracy of our machine-learning potential used in the MD simulation. It is also found out that the amorphous interlayer actually contributes most to the total thermal resistance of the interface. Mode analysis is performed to understand the phonon transport across the interface. Besides the amorphous interlayer studied, novel nanostructures and isotope effects are also explored to enhance the TBC of an interface system. Based on the results, isotope near the interface will have negligible effect on the TBC, while the length and thickness of a nanostructure at the interface will have impacts on the TBC. An optimal combination of these two parameters can eventually lead to an improvement in the interfacial thermal transport.

Presenting Author: Yang Li Massachusetts Institute of Technology

Presenting Author Biography: Yang Li is a PhD student advised by Prof. Asegun Henry in the Mechanical Engineering Department at Massachusetts Institute of Technology.

Authors:

Yang Li Massachusetts Institute of Technology
Luke Yates Sandia National Laboratories
Asegun Henry Massachusetts Institute of Technology