Abstract Optical metasurfaces have transformed nanoscale light control, but conventional platforms remain predominantly rigid and fabrication intensive. Polymer-based metasurfaces offer a complementary design space combining scalable manufacturing, mechanical compliance, chemical tunability, three-dimensional structuring, and functional molecular and nanoparticle incorporation. This Review examines polymer materials, fabrication strategies, and emerging applications in sensing, imaging, reconfigurable optics, light emission, and inverse-designed photonics, highlighting opportunities and challenges for scalable, adaptive, and multifunctional nanophotonic systems.
On-chip silicon light sources are highly desirable for integrated photonics but remain challenging due to silicon’s indirect bandgap and low radiative efficiency. Amorphous silicon supports photoluminescence via band-tail states, yet lacks scalable cavity platforms for efficient light extraction and control. Here, we demonstrate solution-synthesized amorphous silicon microspheres that simultaneously serve as the luminescent medium and spherical optical cavity within a compact ~1 μm³ volume. The spheres exhibit whispering-gallery-mode-coupled photoluminescence, with resonances spanning 800–1100 nm and a maximum measured quality factor of 522, producing narrow and spectrally resolved emission peaks at room temperature. The emission exhibits active tunability through temperature, surrounding refractive index, and excitation power, reflecting the sensitivity of the WGM resonances to thermal and dielectric perturbations. Moreover, the microspheres are readily integrated with three-dimensionally printed waveguides, enabling on-chip routing of emitted light. These features position amorphous silicon microspheres as a versatile, room-temperature platform, offering cavity-enhanced emission and broad tunability for next-generation integrated photonic systems.
Diffractive deep neural networks (D2NNs), composed of multiple layers of trainable diffractive neurons, perform deep network operations entirely through the diffraction and interference of light. This approach has been widely studied in classification, 3D perception, and edge learning due to its zero-power consumption, ultrafast processing, and intrinsic parallelism of light. This review presents fundamental principles, design methodologies, and recent advances in D2NNs for next generation optical computing.
