Optimization of Thermal Conductance at Interfaces Using Machine Learning Algorithms.

Area of Science:

• Materials Science
• Computational Physics
• Nanotechnology

Background:

• Efficient thermal transport is crucial for micro-/nanoscale electronic, photonic, and phononic devices.
• Compositional intermixing at interfaces can enhance interfacial thermal conductance (ITC), but optimal configurations are challenging to identify due to computational costs.
• Scalability and transferability of machine learning models for materials design remain challenges.

Purpose of the Study:

• To develop a scalable Bayesian optimization framework for optimizing heat transfer at material interfaces.
• To investigate the impact of intermixing on interfacial thermal conductance (ITC) at the silicon and aluminum (Si/Al) interface.
• To address the computational challenges in designing materials for enhanced thermal transport.

Main Methods:

• Implementation of a scalable Bayesian optimization framework.
• Leveraging dynamic job spawning via Message Passing Interface (MPI) for parallel molecular dynamics simulations.
• Simulation and analysis of heat transfer at the silicon and aluminum (Si/Al) interface.

Main Results:

• A maximum of 50% increase in ITC was achieved by introducing a two-layer intermixed region with a higher percentage of silicon.
• The magnitude of ITC increase varied due to the random nature of intermixing.
• Both intermixing homogeneity/heterogeneity and the stochastic nature of molecular dynamics simulations contributed to the observed variance in ITC.

Conclusions:

• The developed Bayesian optimization framework effectively optimizes heat transfer at material interfaces.
• A two-layer Si-rich intermixed region significantly enhances ITC at the Si/Al interface.
• Understanding intermixing characteristics and simulation stochasticity is key for predictable materials design.