Nanophotonic Particle Simulation and Inverse Design Using Artificial Neural Networks

PI: Marin Soljačić and John D. Joannopoulos

We developed a method that could provide a way to custom-design multilayered nanoparticles with desired properties, potentially for use in displays, cloaking systems, or biomedical devices. It may also help physicists tackle a variety of thorny research problems, in ways that could in some cases be orders of magnitude faster than existing methods.

The innovation uses computational neural networks, a form of artificial intelligence, to “learn” how a nanoparticle’s structure affects its behavior, in this case the way it scatters different colors of light, based on thousands of training examples. Then, having learned the relationship, the program can essentially be run backward to design a particle with a desired set of light-scattering properties — a process called inverse design.

Nanophotonic particle simulation using NN: The NN architecture has as its inputs the thickness of each shell of the nanoparticle, and as its output the scattering cross section at different wavelengths of the scattering spectrum.


The researchers from this IRG are the first to use an artificial intelligence method in photonics. In addition to employing this method for its use in displays, cloaking systems, or biomedical devices, and physics research, this method permeates new opportunities for researchers in both optics and artificial intelligence communities. Since educating nearly 1,000 researchers at the 2017 Frontiers in Optics conference in Washington, D.C., and nearly 5,000 researchers in the 2017 Neural Information Processing Systems conference in Long Beach, CA, an accelerating number of photonic researchers have begun to explore new applications based on this method, including researchers from other institutions, such as Cornell and Northeastern University, whose employment of this method has reduced several orders of magnitude of computational cost.

Comparison of forward runtime versus complexity of the nanoparticle: The simulation becomes infeasible to run many times for large particles, while the NN’s time increases much more slowly. Conceptually, this is logical as the NN is using pure matrix multiplication—and the matrices do not get much bigger—while the simulation must approximate higher and higher orders. The scale is log-log. The simulation was fit with a quadratic fit, while the NN was a linear fit.


Peurifoy, J., Shen, Y., Jing, L., Yang, Y., Cano-Renteria, F., Delacy, B. G., Joannopoulos, J. D., Tegmark, M., and Soljačić, M. “Nanophotonic Particle Simulation and Inverse Design Using Artificial Neural Networks.” Science Advances, 4(6), 2018. <doi:10.1126/sciadv.aar4206>