IEEE Open Journal of Antennas and Propagation (Jan 2023)

A Diffusion Model for Multi-Layered Metasurface Unit Cell Synthesis

  • Chen Niu,
  • Mario Phaneuf,
  • Puyan Mojabi

DOI
https://doi.org/10.1109/OJAP.2023.3292149
Journal volume & issue
Vol. 4
pp. 654 – 666

Abstract

Read online

A deep learning approach based on a diffusion model is proposed to yield metasurface unit cell designs. This method takes desired two-port scattering parameters along with the frequency of operation in an attempt to synthesize three-layered metasurface unit cells. The core of this approach lies in casting the three-layered unit cell synthesis process as conditional three-channel binary image synthesis. The conditions are governed by the desired scattering parameters at a given frequency whereas the binary nature implies the presence and absence of metallic traces. Once synthesized, these unit cells are placed beside each other, without any further tuning, to form the final metasurface. The performance of the whole metasurface, for three different design scenarios, is then tested against full-wave simulation and/or experimental data.

Keywords