Researchers from CEMES-CNRS and LAAS-CNRS conceived and realized complex photonic nanostructures fulfilling two arbitrary properties, thanks to a numerical evolutionary approach mimicking the process of natural selection of species. The simultaneous optimization of multiple properties is illustrated by the example of double-resonant color pixels, whose different colors can be addressed by the polarization of light. Production by electron beam lithography of polarization-encoded micro-images demonstrates that the method is ready for practical use. An article on the study was published in October 24th 2016 in the scientific journal “Nature Nanotechnology”.
Designing photonic nanostructures to obtain particular optical properties is usually based on the systematic variation of very simple models. Unfortunately, such anticipating approach rapidly becomes impracticable due to large numbers of free parameters in the description of more complex geometries. Finally, when multiple properties are to be fulfilled concurrently, it usually fails altogether. In this context, biology-inspired evolutionary techniques, mimicking the natural selection process, can be a promising approach.
In the present study, an evolutionary algorithm for multi-objective optimization was coupled to a numerical framework for the simulation of optical properties: An initial, random “population” of nanoantenna geometries was submitted to a process of evaluation, selection and reproduction, in which the weakest individuals are successively eliminated, keeping only the fittest candidates (the nanoantennas that exhibit the best optical performances regarding the target criteria). After a defined number of evolutionary cycles, the strongest individual was chosen from the final population.
For the demonstration of this procedure, dual-color silicon nanoantennas were automatically designed by the computational scheme. The two resonant wavelengths of these color pixels were designed such, that they could be addressed individually by means of the incident light polarization (see figure). The geometries found by evolutionary optimization were then converted into a lithographic mask, which was used to fabricate silicon nanostructures by electron beam lithography. The optical characterization of the fabricated samples yielded an excellent agreement with the predictions of the optimization algorithm.
This work emphasizes the tremendous potential of evolutionary optimization in photonics and nano-optics. It can be easily adapted to target any other optical property like the directivity of light scattering, for the design of broadband light harvesting antennas and color filters, or for nonlinear optics.
Evolutionary Multi-Objective Optimisation of Colour Pixels based on Dielectric Nano-Antennas
P. R. Wiecha, A. Arbouet, C. Girard, A. Lecestre, G. Larrieu, and V. Paillard
Nature Nanotechnology (2016)
Contacts Researchers :
Arnaud Arbouet - Researcher CNRS - CEMES-CNRS
Guilhem Larrieu - Researcher CNRS - LAAS-NCRS
Vincent Paillard, Professor Université Toulouse Paul Sabatier - CEMES-CNRS
Peter Wiecha, PhD student Université Toulouse Paul Sabatier - CEMES-CNRS
Contact communication INSIS :
insis.communication at cnrs.fr