Using AI to create thinner, more efficient anti-reflective coatings


What the research is:

An AI-based approach for thinner anti-reflective (AR) coatings, which are used to maximize the light available to photovoltaic solar panels. Developed by researchers from Facebook and the University of Auvergne in France, this technique optimizes the pattern and structure of an AR coating based on its thickness by applying a biologically inspired AI algorithm to the design process. The overall goal is to refract, rather than reflect, incoming light at a wide range of frequencies and angles.

How it works:

Previous research and manufacturing techniques used mathematical functions such as Gaussian and exponential functions to determine the structure of AR coatings. Evolutionary algorithms have the potential to achieve even greater levels of optimization. Inspired by the evolutionary mechanics of organisms, these algorithms repeatedly generate solutions and test their fitness. The class of algorithm that we used, called differential evolution (DE), has proved particularly successful for optimizing complex systems, such as the intricate photonic structure of a nanoscale-size coating. Thinner coatings are potentially better at letting light through.

By giving the DE algorithm a specific thickness to optimize for, we were able to outperform traditional techniques. In one case, it found a structure with AR qualities equivalent to a Gaussian-based approach but at half the thickness. To demonstrate the practical manufacturing applications, we used a vapor deposition technique to produce a multilayer coating measuring 200 nanometers.

A 200-nanometer-thick coating on a silicon wafer

To demonstrate the real-world utility of AI-designed anti-reflective coatings, researchers from Facebook and the University of Auvergne deposited a 200-nanometer-thick coating on a silicon wafer.

Why it matters:

This research is an example of employing AI concepts and theory to advance materials science. Our results demonstrate the feasibility of nanostructured AR coatings. Though this work could have implications for lenses and other components that rely on controlling reflectivity to increase optical efficiency, the most important benefits could be in applying AR coatings to the silicon in photovoltaic (PV) solar panels. Such coatings already provide a significant boost in efficiency for PV cells, and by replicating that improvement with fewer layers and less overall thickness, we hope to contribute to making solar generation even more cost-effective.

Read the full paper:

Optimal anti-reflective coatings using differential evolution