In conventional Stokes polarimetry, where the polarimetric information is obtained from a series of intensity measurements, stable and accurate measurements typically require the optical elements to be carefully designed. Here, we propose a paradigm shift where deep neural network assisted polarimeters based on disordered photonic structures perform high quality polarimetric measurements while completely removing the need for specially designed polarization analyzers, demonstrating how disordered photonic structures fabricated without the use of any nanolithography techniques can enable accurate analysis of optical signals. We implement this concept with disorder-engineered nanoscatterers that allow for analyzing varying degrees of disorder and cellophane film that demonstrate the cost-saving potential of the concept. We demonstrate polarimetric performances, calibrated using deep neural networks, that are comparable to commercial polarimeters, does not require prior knowledge of the input wavelength and show a high degree of mechanical stability.

中文翻译：

---

具有无序光子结构的极化

在常规的斯托克斯偏振法中，从一系列强度测量中获得偏振信息，稳定而准确的测量通常需要仔细设计光学元件。在这里，我们提出了范式转换，其中基于无序光子结构的深层神经网络辅助旋光仪执行高质量的偏振测量，同时完全消除了对专门设计的偏振分析仪的需要，证明了在不使用任何纳米光刻技术的情况下制造的无序光子结构如何能够实现准确的定位光信号分析。我们通过无序工程纳米散射体来实现此概念，该纳米散射体可以分析不同程度的无序和玻璃纸薄膜，从而证明了该概念的成本节约潜力。