Towards Monolithically Integrated Optical Systems for Communication and Sensing

The exponential growth of global data traffic, coupled with the increasing demand for real-time, high-resolution sensing, places unprecedented pressure on information processing technologies. Conventional electronic approaches to switching and sensing face fundamental bottlenecks in scalability, energy efficiency, and latency. In this context, integrated photonics emerges as a compelling solution, offering massive bandwidth, low latency, and immunity to electromagnetic interference. Nevertheless, realizing compact, multifunctional, and reconfigurable photonic systems requires overcoming several challenges at both the device and system levels. This thesis addresses these challenges by advancing monolithically integrated photonic platforms that unify communication and sensing functions within a coherent framework. The research spans innovations in passive and hybrid material systems, novel switching architectures, and computationally enhanced sensing paradigms, culminating in a trajectory that points toward next-generation multifunctional integrated photonics. At the device level, the work begins with innovations in passive silicon nitride (SiN) circuits. Achieving monolithic light-source integration with minimal loss and reflection calls for an additional micrometre-scale passive layer beyond the III–V materials. Low-temperature PECVD SiN is a strong candidate. Given the III–V stack, thick rib-type SiN waveguides are preferred. The drawback is bend performance—radiation loss is high for this thick platform, necessitating bend radii approaching 800 µm and limiting layout density in PICs. To address this, a novel deep-etched bending structure is proposed and experimentally validated. This design reduces the bending radius from 800 μm to 37 μm while maintaining a loss as low as 0.1 dB/90°, representing a more than twenty-fold footprint reduction without sacrificing performance. Compact add-drop microring resonators are subsequently fabricated, achieving a free spectral range of 1.14 nm and a 39.1 GHz passband. These results highlight the feasibility of constructing dense and scalable photonic circuits, laying a robust passive foundation for further integration. Building upon this foundation, the hybrid α-Si/SiN platform is developed to address the limited thermo-optic tunability of SiN. By overlaying an amorphous silicon (α-Si) layer on SiN waveguides, the hybrid design enhances thermo-optic modulation efficiency while retaining the low-loss characteristics of SiN. Simulations predict robust thermo-optic tuning, enabling dynamic control over multiple resonance channels. Fabricated devices exhibit deviations from theoretical predictions—primarily due to discrepancies in the refractive index of the deposited α-Si layer—but the approach nevertheless demonstrates a promising route toward reconfigurable photonic circuits. These results underline the importance of precise material characterization and fabrication control, while simultaneously validating the hybrid strategy as a pathway for scalable tunability in integrated platforms. At the system level, attention focuses on overcoming the scalability challenges of optical switching. Conventional switch fabrics such as Beneš and Dilated Banyan networks face quadratic growth in switching elements or suffer from in-band crosstalk, making them unsuitable for large-scale deployment. To address this, a Dilated Crosspoint topology for space-and-wavelength selective switching (SWSS) is proposed. This architecture reduces the required number of switching elements by an order of magnitude and eliminates first-order crosstalk through wavelength allocation and topology dilation. A 4×4×4λ prototype fabricated on a silicon-on-insulator platform validates the design, achieving insertion losses of 2.3–8.6 dB, crosstalk suppression down to –59.7 dB, and microsecond-scale reconfiguration times. To the best of current knowledge, this represents one of the most compact and efficient SWSS prototypes to date, highlighting the architecture’s potential to serve as a practical solution for data centre interconnects and high-performance computing. Beyond communication, the thesis extends integrated photonics into sensing applications, where compactness, speed, and robustness are equally critical. A deep sensing framework is introduced in which reflectance spectra are encoded by a microring-based photonic sampler and decoded using a convolutional neural network (CNN). This approach bypasses the need for bulky interferometers, Fourier transforms, or nonlinear curve fitting, directly mapping encoded spectra into physical parameters such as thickness. Experimental demonstrations on thin polymer films and three-dimensional printed structures confirm both high accuracy and robustness, validating the framework’s potential as a compact and intelligent sensing method. In conclusion, this thesis presents a coherent trajectory from compact SiN passive devices to hybrid tunable circuits, and from scalable SWSS fabrics to intelligent sensing frameworks. Collectively, these contributions demonstrate that monolithic photonic integration is a powerful strategy for realizing energy-efficient, reconfigurable, and multifunctional systems. By addressing critical device-level limitations and extending the approach to system-level implementations, the work provides practical solutions for the urgent demands of high-capacity data communication and advanced sensing, while establishing a solid foundation for future developments in integrated photonics.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Zhang, Ziyao

Advisor dc:contributor.advisor

• Penty, Richard

• Deep sensing
• Integrated photonics
• Optical switch

Licence

• https://www.repository.cam.ac.uk/bitstreams/67c25f77-b1af-4c3c-a065-14a5283309f8/download
• http://purl.org/NET/rdflicense/allrightsreserved

Language dc:language

eng

Author Identifier

0000-0002-1161-7079

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/396272