Researchers from Nottingham Trent University, in collaboration with the University of Nottingham and Sony Europe B.V., have developed a novel method for real-time programmable nonlinear wavefront shaping using a silicon metasurface driven by a genetic algorithm (GA). This innovative approach, published in Engineering, aims to enhance optical technologies by enabling dynamic control of wavefronts without the need for multiple fabrications of metasurfaces.

Nonlinear metasurfaces, composed of arrays of nano-atoms, have emerged as a versatile platform for nonlinear generation processes such as second-harmonic generation (SHG), third-harmonic generation (THG), and high-order harmonic generation (HHG). These metasurfaces can strongly confine light and enhance nonlinear generation through multipolar resonances and plasmonics. However, once fabricated, the nonlinear wavefront generated by a metasurface is fixed, limiting its flexibility and necessitating the fabrication of different metasurfaces for different wavefronts. This process is both time-consuming and inefficient.

To address this challenge, the research team combined evolutionary algorithms with spatial light modulators (SLMs) to dynamically control wavefronts using a single metasurface. The key innovation lies in the use of a GA to optimize the phase distribution of the SLM, allowing the nonlinear wavefront to be shaped according to input reference images. The GA evolves based on the captured nonlinear emission from a charge-coupled device (CCD) camera, establishing a real-time response.

The experimental setup involves a femtosecond laser beam with a wavelength of 1510 nm, which is reflected from a mirror to the SLM. The SLM manipulates the phase distribution of the laser beam, converting it into a linear signal. This signal is then converted and denoised by the silicon metasurface through a nonlinear optical process, generating a nonlinear signal. The cubic relationship between the THG emission and the infrared input reduces noise in the diffractive patterns produced by the SLM, allowing for precise experimental engineering of the nonlinear emission patterns with fewer alignment constraints.

The silicon metasurface used in the experiment consists of a square array of silicon nanodisks on a silicon dioxide substrate. The metasurface supports multipolar Mie resonances that strongly enhance light-matter interactions, thereby significantly boosting THG emission at resonant positions. The researchers demonstrated the effectiveness of their approach by generating various nonlinear emission patterns, including single-point, double-point, and triple-point focusing patterns, with the GA optimizing the SLM phase distribution in real-time.

This method eliminates the need to position the SLM in the Fourier plane, increasing tolerance to optical alignment errors. The cubic response of the THG in the silicon metasurface acts as a low-signal filter, suppressing diffraction signals from the SLM and serving as an effective denoising operator. The researchers believe that integrating neural networks with their optical system could further reduce processing time and enhance the fidelity of generated patterns.

The study’s findings pave the way for self-optimized nonlinear wavefront shaping, advancing optical computation and information processing. The dynamic control of wavefronts using a single metasurface has the potential to revolutionize applications in optical imaging, quantum light sources, and ultrasensitive sensing.

The paper “Real-Time Programmable Nonlinear Wavefront Shaping with Si Metasurface Driven by Genetic Algorithm” is authored by Ze Zheng, Gabriel Sanderson, Soheil Sotoodeh, Chris Clifton, Cuifeng Ying, Mohsen Rahmani, Lei Xu. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.04.023. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.