Design of terahertz metasurface structures for biosensing applications based on deep learning methods

With the advancement of terahertz band applications, extensive research has been conducted on terahertz devices. Terahertz band biosensors find widespread use in biomedical microdetection; however, predicting terahertz spectrum and designing structures pose complex and time-consuming challenges. This article proposes an efficient deep learning method for optimizing the design of terahertz metasurface biosensors. The method employs three neural networks, utilizing spectral response as an intermediary to effectively map customized performance indicators onto geometric structural parameters. Test results demonstrate that the proposed design scheme can generate suitable structural parameters based on the required frequency and bandwidth of the analyte. These output parameters were subsequently simulated and validated using electromagnetic simulation software, yielding results consistent with predictions. This method of using neural networks instead of electromagnetic simulation can be applied to the study of spectrum prediction and inverse design of terahertz devices, providing more possibilities for the future application of terahertz devices.