AI-assisted nanophotonic inverse design of optical sensors

Nanomaterials have optical properties that strongly depend on their geometry and are therefore are often used as tunable signal transducers in optical sensors. For example, palladium nanoparticles are capable of detecting H2 gas by absorbing hydrogen within its metallic lattice, leading to a modified electronic structure and optical appearance.

Recently, we have shown that periodic arrays of palladium nanoparticles can exhibit enhanced H2 sensitivities and lower limits of detection compared to conventional random arrays [1]. This advantage is due to the existence of a collective optical resonance originating from the interference of the light scattered in the sample plane by adjacent nanoparticles.

The sensitivity of our metasurfaces depends critically on the array geometry and finding the best configuration can be like finding a needle in a haystack. We solve this problem using a particle swarm optimization (PSO) algorithm that evaluates the performance of a range of random configurations and lets them evolve using a predefined figure of merit. So far we focused on sensing H2, but our method is applicable to any process that leads to a shift in the optical signal of our nanostructures.

The aim of the current project is to design periodic metasurfaces capable of detecting different gasses, such as CO and NOx, whose presence even in trace amounts can be dangerous. When these gasses absorb onto the surface of metal nanoparticles they can lead to broadening of their optical resonances, a process that we will exploit to design ultra-sensitive metasurfaces.

Student tasks:

– perform PSO-assisted finite-difference time-domain (FDTD) electrodynamic simulations of periodic metasurfaces 

– provide structural parameters for the nanofabrication of sensing metasurfaces for CO and NOx

– nanofabricate the sensing metasurface and test it in our home-built flow cell setup

[1] Nugroho et al., Nature Communications 2022, 13, 5737

Leave a Comment