Computer technologists and research scientists from the Massachusetts Institute of Technology have devised a computational neural network which is capable of learning how the structure of a nanoparticle influences the way the particle scatters light. This inference is drawn from thousands of examples being processed by the machine intelligence. The new method is expected to aid physicists in tackling research problems and doing so in ways that are considerably faster than any existing method is capable of achieving.
The goal of the research is to find ways to develop custom-designed, multilayered nanoparticles with properties of use for a range of different applications. Potential uses include displays, cloaking systems, and biomedical devices. There is also an on-going research dimension, in that the technology could assist physicists in addressing long-standing and complex research problems.
These new insights can arise as the artificial intelligence studies the structure and behaviors of nanoparticles, especially the ways that light of different colors are scattered light. As insights are gained about the relationships between light, color and structure, the platform can be run backwards to propose novel design for nanoparticles (this is a process called inverse design). An important advantage with the approach is that the network needs to be trained on only a small sampling of the data to approximate the simulation to high precision.
According to lead researcher Professor Marin Soljacic, a key aim is 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?”
The study is published in the journal Science Advances, under the titled “Nanophotonic particle simulation and inverse design using artificial neural networks”.
