High-Efficiency Metalens Antenna Design Through a ControlNet Diffusion Generation Model

In this letter, we propose a ControlNet stable diffusion-based method for the inverse design of a high efficiency metalens antenna. This marks the first application of a text and image fine-tuning generative deep-learning model in metalens design, offering high design freedom with more diversity and precise control. The design process involves dataset generation, electromagnetic simulation, feature encoding, training, unit cell generation, and metalens design. The generated unit cells demonstrate a phase range of 290<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{\circ }$</tex-math></inline-formula> and high transmission magnitudes over 0.88, with 66% exceeding 0.95. A three-layer metalens antenna with an f/D ratio of 0.5, formed using the generated unit cells and a feeding horn, achieves a gain of 28.1 dBi and an efficiency of 51.4%, nearing the theoretical limit of 63.7%, and maintains minimal side-lobe levels of −22.3 dB and −22.7 dB in the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\phi =0^\circ$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\phi =90^\circ$</tex-math></inline-formula> planes.