Metasurfaces (MS) are widely accepted in the devices, such as absorbers, toimprove their performances. MS provide new design freedoms, however, theyalso result in complexity in designs because of the high-dimensional topologicalspace of meta-atoms, and high computational costs in the optimizationprocedure. To alleviate this challenge, a generative meta-atom model thatgenerates the pattern and corresponding electromagnetic (EM) responses ofmeta-atoms, specialized for absorbing applications is developed. The modelis established by the convolutional variational autoencoder (CVAE), and adeep neural network (DNN). The model is verified by designing differenttypes of absorbers with the evolutionary algorithm, and an ultrabroadbandlower profile absorber at low microwave frequencies is achieved. The realizedmetasurface-based absorber (MSA) covers the frequency range from 1.4to 18 GHz at the criteria of return loss (RL) ?10 dB with a thickness of 8 mm,which is validated by experiments. This work provides an effective and highlyefficient way to design high-performance MSA, which can be easily extendedto other metasurface-based functional devices.
展开▼