Massachusetts Institute of Technology

Research Scientist

Computational Materials Science for Energy Applications

|Home | Contacts | Publications | Codes | Web Apps | Bio | Mentoring | Research Interests | Twitter | Media Gallery |

- Multiscale Modeling. The development of models capturing the interplay of physics at different scales is paramount to the simulations of modern devices.
We recently developed an approach that models multiscale heat transport by combining knowledge
from the atomistic scale and the phonon Boltzmann transport equation.
We plan to extend this approach to other domains, such as electronic transport, and to develop more synergetic integration between the atomistic and continuum description of matter.

- Inverse Design. What's the geometry of material that gives tailored physical properties? Key components
to this task, commonly referred to as inverse design, are differentiable solvers.
Within this space, we recently developed a method, termed the Interpolation Transmission Method (TIM), that enables the inverse design of structures for nanoscale thermal transport applications. As TIM is species-agnostic,
we also plan to apply it to diffusive electronic transport. Other efforts in this area include the inverse design of resilient multiscale materials.

- Machine-Learning-Assisted Experiments.In some cases, inverse design predicts an optimal
configuration in terms of quantities that are not directly accessible experimentally, e.g., the effective electron density in the drift-diffusion model
of a solar cell. To mitigate this issue, we plan to deploy artificial neural networks that fill this knowledge gap together with differentiable physics solvers.
In practice, this amount to end-to-end supervised learning, where, e.g., the samples are the

- Open-source software.
The implementation of free and widely accessible software enables a far-reached adoption of the
developed models and opens up an opportunity for contribution from the community. We are strongly committed to this
open-science approach and sharing data backing up our research. Efforts in this space include OpenBTE, a solver for nanoscale
heat transport, and ∂PV, a solver for the inverse design of solar cells.
To further democratize physics simulations, we also develop decentralized Web Apps (e.g., the run in the user's browser)
such as Heat Opt.