MIT physicists have developed a new method that could someday provide a way to customize multilayered nanoparticles with preferred properties, potentially for use in cloaking systems, displays, or biomedical devices. It may also help physicists handle a range of thorny research issues, in ways that could in certain cases be orders of magnitude faster than present approaches.
The innovation employs computational neural networks, a form of artificial intelligence, to “learn” how a nanoparticle’s structure influences its behavior, in this case, the way it scatters various colors of light, based on numerous training examples. Then, having learned the association, the program can fundamentally be run backward to design a particle with a preferred set of light-scattering properties — a process known as inverse design.
The results are being published in the journal Science Advances, in a paper by MIT senior John Peurifoy, research affiliate Yichen Shen, graduate student Li Jing, professor of physics Marin Soljačić, and five others.
While the method could eventually result in practical applications, Soljačić says, the research is mainly of scientific interest as a way of predicting the physical properties of a range of nano-engineered materials without necessitating the computationally intensive simulation processes that are usually used to handle such issues.
Soljačić says that the objective was to study at neural networks, a field that has witnessed a lot of progress and produced excitement in recent years, to see, “whether we can use some of those techniques in order to help us in our physics research. So basically, are computers ‘intelligent’ enough so that they can do some more intelligent tasks in helping us understand and work with some physical systems?”
To examine the idea, they used a comparatively simple physical system, Shen explains.
In order to understand which techniques are suitable and to understand the limits and how to best use them, we [used the neural network] on one particular system for nanophotonics, a system of spherically concentric nanoparticles.
Yichen Shen, Research Affiliate, MIT
The nanoparticles are layered like an onion, but each layer is formed of a different material and has a varying thickness.
The nanoparticles have sizes similar to the wavelengths of visible light or smaller, and the way light of various colors scatters off of these particles relies on the wavelength of the incoming beam and on the details of these layers. Calculating all these effects for nanoparticles with a number of layers can be a rigorous computational task for many-layered nanoparticles, and the complexity becomes worse as the number of layers grows.
The scientists were keen to see if the neural network would be able to predict the way a new particle would distribute colors of light — not only by interpolating between identified examples, but by essentially figuring out some underlying pattern that permits the neural network to extrapolate.
“The simulations are very exact, so when you compare these with experiments they all reproduce each other point by point,” says Peurifoy, who will be an MIT doctoral student next year. “But they are numerically quite intensive, so it takes quite some time. What we want to see here is, if we show a bunch of examples of these particles, many many different particles, to a neural network, whether the neural network can develop ‘intuition’ for it.”
In reality, the neural network was able to predict quite well the exact pattern of a graph of light scattering versus wavelength — not flawlessly, but very close, and in a lot less time. The neural network simulations, “now are much faster than the exact simulations,” Jing says. “So now you could use a neural network instead of a real simulation, and it would give you a fairly accurate prediction. But it came with a price, and the price was that we had to first train the neural network, and in order to do that we had to produce a large number of examples.”
Once the network is trained, though, any simulations in the future would gain the full advantage of the speedup, so it could be a beneficial tool for situations needing repeated simulations. But the project’s real goal was to learn about the methodology, not just this specific application.
One of the main reasons why we were interested in this particular system was for us to understand these techniques, rather than just to simulate nanoparticles.
Marin Soljačić, Professor of Physics, MIT
The subsequent step was to fundamentally operate the program in reverse, to use a set of preferred scattering properties as the starting point and observe if the neural network could then figure out the precise combination of nanoparticle layers required to attain that output.
“In engineering, many different techniques have been developed for inverse design, and it is a huge field of research,” Soljačić says. “But very often in order to set up a given inverse design problem, it takes quite some time, so in many cases, you have to be an expert in the field and then spend sometimes even months setting it up in order to solve it.”
However with the team’s trained neural network, “we didn't do any special preparation for this. We said, ‘ok, let’s try to run it backward.’ And amazingly enough, when we compare it with some other more standard inverse design methods, this is one of the best ones,” he says. “It will actually do it much quicker than a traditional inverse design.”
The initial motivation we had to do this was to set up a general toolbox that any generally well-educated person who isn’t an expert in photonics can use. … That was our original motivation, and it clearly works pretty well for this particular case.
The speedup in specific kinds of inverse design simulations can be pretty significant. Peurifoy says, “It's difficult to have apples-to-apples exact comparisons, but you can effectively say that you have gains on the order of hundreds of times. So the gain is very very substantial — in some cases, it goes from days down to minutes.”
The study received support from the National Science Foundation, the Semiconductor Research Corporation, and the U.S. Army Research Office through the Institute for Soldier Nanotechnologies. Other people involved in the study are: Yi Yang, Fidel Cano-Renteria, John D. Joannopoulos, and Max Tegmark, all from MIT; and Brendan G. Delacy from U.S. Army Edgewood Chemical Biological Center.