期刊文献+

Deep-learning-driven design of multiplexed meta-array couplers for broadband HEMT terahertz detection

原文传递
导出
摘要 On-chip couplers are essential for coupling free-space electromagnetic waves into sub-wavelength semiconductor devices and enhancing light-matter interactions.However,the couplers used in existing single field-effect transistor(FET)detectors exhibit poor response over wide frequency ranges,making the detection of ultra-wideband weak signals highly challenging.In this work,we introduce a meta-array coupler for terahertz detectors based on field-effect transistors,which demonstrates exceptional coupling asymmetry beneath the gate.To optimize the meta-array structure for high gain across an ultra-wide frequency range,we propose a hybrid long short-term memory–multi-layer perceptron(LSTM-MLP)neural network for the detector’s on-chip coupler.This method enables accurate and rapid prediction of the enhanced spectra of meta-atoms,facilitating efficient conversion from structural parameters to resonance frequencies.Compared to traditional dipole couplers,the proposed metacoupler generates strong localized electric field enhancement through dipole resonance,achieving a mixing factor asymmetry up to 246 times higher near 660 GHz and an antenna factor enhancement of approximately 50 times.The meta-array coupled FET terahertz detector achieves ultra-wideband resonance enhancement from 350 to1000 GHz.This hardware architecture and hybrid LSTM-MLP neural network are expected to overcome the limitations of current FET-based detectors in coupling efficiency,bridging the performance gap toward ultra-sensitive detection applications.
出处 《Photonics Research》 2025年第5期1290-1301,共12页 光子学研究(英文版)
基金 National Key Research and Development Program of China(2023YFB3207800) National Natural Science Foundation of China(61927804)。
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部