ISSN Online: 2377-424X
ISBN Print: 978-1-56700-474-8
ISBN Online: 978-1-56700-473-1
International Heat Transfer Conference 16
DESIGNING NANOSTRUCTURES FOR HEAT TRANSPORT VIA MATERIALS INFORMATICS
Résumé
As the length scales of materials are reduced to nanoscale, it becomes possible to tune heat carrying phonons transport by manipulating the nanostructures. However, it is rather difficult to identify the detail optimal structure for heat transport due to the various parameters and coupled interference/resonance effects. The key next-generation technology can be materials informatics, which is a new interdisciplinary research to provide efficient tools to accelerate the materials discovery and design. In this work, we present two successful materials informatics approaches in designing nanostructures for heat transport: Bayesian optimization and Monte Carlo tree search. Bayesian optimization is good at dealing cases with limited number of candidates and finding optimal candidate as quickly as possible, while Monte Carlo tree search can handle cases with huge or unlimited number of candidates. We apply both algorithms to design the Si/Ge interfacial alloy structures that minimize/maximize the interfacial thermal conductance across Si-Si and Si-Ge interfaces. The result indicates
that using Bayesian optimization the optimal structures can be obtained by calculating only a few percent of the
total candidates, considerably saving the computational resources. In comparison to Bayesian optimization, the
Monte Carlo tree search algorithm has shown competitive search efficiency and superior scalability for targeting
candidates that are approaching the global optimal ones. The present work has shown great advantage of materials informatics in designing nanostructures to control heat transport, which can be extended to other nanostructures and properties.