Low-power Consumption Reconfigurable-encoded 16×16 Polymer Waveguide Switch Array

This work aims to design， fabricate， and characterize a low-power， reconfigurable 16×16 thermo-optic switch array on a polymer-based photonic platform， addressing the critical challenges of high consumption and thermal crosstalk associated with scaling large-scale photonic integrated circuits. The primary goal is to develop a reliable hardware platform that demonstrates the feasibility of utilizing the unique material properties of polymers， specifically a high thermo-optic coefficient and low thermal conductivity for efficient and dense on-chip dynamic optical path control. The overarching purpose is to provide a foundational device solution to enable future low-power， flexible， and intelligent photonic systems for applications such as data center optical interconnects， photonic neural networks， and secure optical communications. The central research questions involve validating the performance metrics of a scalable polymer waveguide switch architecture and establishing its advantages over traditional inorganic platforms.The study employed a comprehensive methodology integrating theoretical design， numerical simulation， process fabrication， and systematic experimental characterization. The core architectural innovation was the design of a cascadable unit cell termed the Directional Coupler-Honeycomb （DC-HC） structure， which combines a DC for power switching and a honeycomb HC interferometer for phase modulation. Multi-segment S-bend waveguides interconnect these units to achieve the required 127 μm port pitch for fiber array compatibility. A low-loss waveguide structure was meticulously designed and simulated. The core layer utilized a fluorinated photosensitive polymer， FSU-8， chosen for its low optical absorption at the 1 550 nm communication wavelength. The cladding material was an Organic-Inorganic Grafting Polymethyl Methacrylate （OIG-PMMA）， selected for its enhanced thermal stability， self-planarization capability， and precise refractive index tunability. The optimized cross-section featured a 3 μm × 3 μm core， a 2 μm coupling gap between adjacent DC waveguides， and a 3 μm separation between the aluminum thin-film heater and the waveguide core. Beam Propagation Method （BPM） simulations using R-soft software were extensively conducted to optimize the waveguide geometry， including the DC coupling length， bending losses， and fiber-to-chip coupling loss. Finite Element Method （FEM） thermal simulations in COMSOL Multiphysics were performed to analyze heat distribution and cross-talk， confirming effective thermal confinement within individual switch units. Fabrication leveraged standard， cost-effective， and Complementary Metal-Oxide-Semiconductor （CMOS）-backend-compatible processes. The waveguide layers were defined through spin-coating， Ultraviolet （UV） lithography， and etching. Aluminum electrodes with a serpentine geometry were patterned via wet etching and precisely aligned atop the modulation arm of each DC unit. The fabricated devices were characterized using optical microscopy and SEM to verify dimensional accuracy and morphology. A dual-system testing approach was implemented for performance evaluation. A static switching performance test system was constructed using a 1 550 nm continuous-wave laser source， a polarization controller， single-mode fibers for edge coupling， and an optical power meter. This system measured key static parameters： insertion loss and the required switching power consumption by applying a controlled DC voltage to the heaters. A separate dynamic pulse response system was built， incorporating an Arbitrary Waveform Generator （AWG）， a Photodetector （PD）， and an oscilloscope. This system measured the temporal response， specifically the rise and fall times， of the switches under square-wave modulation at frequencies of 500 Hz and 1 kHz.The fabricated 16×16 polymer thermo-optic switch array exhibited compact dimensions of approximately 7.0 cm in length and 0.2 cm in width. Comprehensive testing at the 1 550 nm operational wavelength yielded a comprehensive set of performance data. The device achieved a minimum insertion loss of 11.4 dB. A detailed loss budget analysis attributed this to constituent factors： waveguide propagation loss， structural loss from cascaded bends， fiber-to-chip coupling loss， and additional scattering losses arising from fabrication imperfections such as sidewall roughness and imperfect cleavage planes. A core achievement was the demonstration of low-power operation. The minimum power required to toggle a single channel between its cross and bar states was measured to be 33.38 mW. This corresponds to applying a temperature change of approximately 21 K to the modulation arm， as predicted by thermal-optical simulations. The maximum extinction ratio achieved was 12.12 dB. While this value shows room for improvement compared to some inorganic platforms， it is attributed in part to a slight broadening of the fabricated waveguide width， which can support higher-order modes and introduce phase errors， as confirmed by supplementary simulations. Dynamic performance tests confirmed operation in the sub-millisecond regime. Under 500 Hz square-wave modulation， the measured rise time was 226.7 μs， and the fall time was 294.0 μs. When driven at a higher frequency of 1 kHz， the response times were 222.2 μs （rise） and 255.9 μs （fall）， establishing this as the practical operational limit for the current device design. Simulation results for the full array predicted an average optical transmission of 71.87% and an adjacent channel crosstalk better than -31.94 dB for a signal routed from a single input to any output， indicating good routing fidelity. Crucially， thermal simulation results validated a key material advantage： the temperature field from an active heater decayed to ambient levels within 20~30 μm from its center. Given that adjacent switch electrodes are spaced approximately 172 μm apart， this confirms effective suppression of thermal crosstalk， a major bottleneck in densely packed inorganic switch arrays. A comparative analysis with state-of-the-art switches on platforms like silicon-on-insulator （SOI）， Silica， and lithium niobate on insulator （LNOI） contextualizes the results. The polymer-based array demonstrates a compelling trade-off： its switching power is significantly lower than that of typical SOI or silica-based units， its unit-length propagation loss is advantageous， and its response speed is competitive with other polymer and hybrid systems. Most notably， it delivers this performance at a substantially larger port count， highlighting the scalability benefits of the polymer platform for large-scale photonic networks.This research successfully validates the polymer material platform as a highly viable and competitive candidate for implementing scalable， low-power reconfigurable photonic integrated circuits. The developed DC-HC architecture and the associated fabrication methodology provide a concrete pathway toward constructing large-scale optical switching networks. The work conclusively demonstrates that the inherent material properties of polymers， specifically the high thermo-optic coefficient and low thermal conductivity， can be effectively harnessed to overcome key limitations of traditional platforms， namely high consumption and thermal crosstalk in dense arrays. The fabricated device， with its integrated combination of low switching energy， acceptable optical performance， microsecond-scale reconfigurability， and inherent programmability， establishes a significant hardware prototype. It serves as a critical enabling component for next-generation systems requiring dynamic optical path management. The findings are offering a practical device solution and a clear material and structural strategy for future work aimed at achieving even larger scale， lower loss， and higher-speed intelligent photonic processing systems.