Inverse-Designed Lithium Niobate Nanophotonics and Applications

Abstract

Integrated photonics has emerged as a cornerstone of modern technology, providing revolutionary solutions for next-generation 6G networks, inter-chip interconnects and quantum information processing. By miniaturizing optical functions into a single chip, the field has surpassed the physical limitations of discrete components to offer significant improvements in speed, energy efficiency, and scalability. Among various material platforms, thin-film lithium niobate (TFLN) has recently surfaced as a competitive candidate for next-generation integrated photonics due to its exceptional material properties, including high transparency from visible to mid-infrared (MIR) wavelength, significant electro-optic and nonlinear coefficients, and great material stability. The transition from traditional bulk crystals to a thin-film-on-insulator configuration has enabled better optical confinement and stronger light-matter interaction, drastically reducing the device footprints while boosting the photonic system performance.

However, as the complexity of photonic integrated circuits (PICs) grows, the limitations of traditional design methodologies—which rely heavily on human intuition and exhaustive, inefficient parameter sweeps—have become increasingly apparent. Conventional approaches are often restricted to simple geometries and existing physical models, which struggle to achieve optimal performance as demand for device functionality increases. To unlock the full potential of the TFLN platform, the transition toward a more efficient, multi-functional and universal design framework is essential. Unlike traditional methods, the inverse design method, which is based on “objective first” principle, could leverage advanced optimization algorithms to automatically synthesize non-intuitive nanophotonic structures based on user-defined performance metrics. This performance-driven approach allows for the exploration of a vast design space to discover topologies that would be almost impossible to conceive through manual modeling.

In Chapter 1, we establish the research landscape by reviewing the physical properties and intrinsic advantages of the TFLN platform. We discuss the evolution from bulk crystals to thin-film waveguides and highlight the necessity of adopting inverse design to overcome the bottlenecks of current design paradigms.

In Chapter 2, we present the theoretical and mathematical foundations of our work. We detail the implementation of a 3D gradient-based inverse design framework using the adjoint-variable method and Finite-Difference Time-Domain (FDTD) solvers. We further discuss density-based topology optimization (TO) and the incorporation of geometric constraints to ensure that optimized designs adhere to practical fabrication limits and minimum feature sizes.

In Chapter 3, we apply this framework to develop high-performance passive TFLN components. We experimentally demonstrate a spatial-mode multiplexer, a waveguide crossing, and a compact waveguide bend, all of which exhibit low insertion loss and footprints orders of magnitude smaller than traditional designs, showing excellent agreement between simulation and experiment.

In Chapter 4, we address the challenge of chip-to-fiber coupling by designing ultra-compact photonic interfaces. By transcending the limits of traditional eigenmode expansion (EME), these inverse-designed couplers achieve efficient mode field diameter (MFD) expansion for lensed and ultra-high-numerical-aperture (UHNA) fibers while remaining compatible with single-step etching processes.

In Chapter 5, we extend our inverse design methodology to the visible spectrum. We demonstrate key building blocks, including a broadband power splitter, an anisotropic waveguide crossing, and a spatial mode converter, showcasing the platform’s versatility across different wavelength regimes.

Finally, in Chapter 6, we summarize the contributions of this thesis and provide an outlook on the future convergence of multi-physics optimization and intelligent design tools, which could transition TFLN photonics toward a new era of high-performance, system-level integrated photonics.

Date of Award	16 Jun 2026
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Cheng WANG (Supervisor)