AI speeds up design of light-manipulating nanostructures by tenfold

Abstract Deep learning has emerged as a key tool for designing nanophotonic structures that manipulates light at
sub-wavelength scales. Although a conventional approach of measuring the optical properties of a given
nanostructure is conceptually straightforward, inverse design remains difficult because the existence and
uniqueness of an acceptable design cannot be guaranteed. Furthermore, the dimensionality of the design
space is often large, and simulation-based methods quickly becomes intractable. Deep learning methods are
well-suited to tackle this problem because they effectively handle high-dimensional input data. Here we train
a conditional generative adversarial network model and use it for inverse design of plasmonic nanostructures
based on their extinction cross section spectra. Our results show that adding label projection and a label
embedding network to the model, improves the performance in terms of error estimates and requires fewer
epochs of training. The mean absolute error is reduced by 50% in the best case, and the training algorithm
converges up to ten times faster. This is shown for two network architectures, a simpler one using a fully
connected neural network architecture, and a more complex one using convolutional layers. We pre-train
a convolutional neural network and use it as a surrogate model to evaluate the performance of our inverse
design model. The surrogate model evaluates the extinction cross sections of the design predictions, and we
show that our modifications lead to equally good or better predictions of the original design compared to a
baseline model. This provides an important step towards more efficient and precise inverse design methods
for optical elements.