Inverse Design of Nanoarchitected Materials
What is a nanostructured material's geometry with a given thermal conductivity tensor? To answer this question, we recently introduced a shape optimization methodology [arXiv(2022)], based on the sensitivity analysis of the phonon Boltzmann transport equation and automatic differentiation.A key aspect of our method is the "Transmission Interpolation Model" (TIM), a technique that hacks the transmission coefficients into a proxy for material density. For educational purposes, we also developed a Web App that solves the standard Fourier law in periodic materials. The project is funded by MIT-IBM Watson AI Lab"". Covered in MIT News.
Differentiable Solar Cell Simulator
The computational design of solar cells entails repeatedly solving the drift-diffusion model, a system of nonlinear partial differential equations. We introduced ∂PV [J. Comput. Phys. (2022)] to make this task amenable. This open-source software directly provides the conversion efficiency's sensitivity to crucial material properties, such as energy gap and electronic affinity.The solver, implemented in Python/JAX, is hosted on GitHub and can be readily run in the cloud with Google Colab. Thanks to a permissive license, any design identified with ∂PV can be freely commercialized, thus accelerating decarbonization. Project funded by MITEI. Select outlets: MIT News, IEEE Spectrum, Photonics, pv-magazine, Sean Mann's MIT News.
Modeling multiscale heat conduction
Nanoscale heat transport may be exploited for heat management and thermal energy harvesting. We developed a computational framework that solves for the ab-initio phonon Boltzmann transport equation [ J. Heat Transfer (2015) , arXiv (2022), arXiv (2021), arXiv (2020), J. Heat Transfer (2018) ], implemented in the open-source code OpenBTE [ github ].
The code has been employed to guide measurements [ J. Appl. Phys. (2021), Nanoscale (2018), Sci. Rep. (2017) ], to predict the effective thermal conductivity of several porous materials [ Int. J. Heat Mass Transf (2022), Phys. Rev. B (2017), Sci. Rep. (2017), Appl. Phys. Lett. (2017), Phys. Rev. B (2016), Appl. Phys. Lett. (2014), J. Comput. Electron. (2014) ], and to benchmark reduce-order and machine-learning models [ Phys. Rev. Res. (2022), Int. J. Heat Mass Transf. (2022) Phys. Rev. B (2019) ]. Partially funded by DOE and NASA-JPL.