INTEGRATED THIN-FILM LITHIUM NIOBATE PULSE SHAPERS FOR FREQUENCY-BIN QUANTUM INFORMATION PROCESSING
Ruihan Chen
Abstract
Integrated quantum photonics requires material platforms that can combine photon generation, spectral control, and electro-optic reconfiguration with low loss and operational stability. Thin-film lithium niobate (TFLN) is a promising platform for this purpose because its second-order nonlinearity and Pockels effect support nonlinear optical processes and active phase control within compact photonic circuits. This dissertation develops TFLN integrated photonic devices and, through collaborative experiments, demonstrates methods toward a frequency-bin quantum information processing platform in which quantum states are represented by discrete optical frequency modes. The work focuses on the design and validation of chip-scale building blocks for frequency-bin operation, including spectral selection, electro-optic tuning, and programmable phase control. Device design, process development, packaging, and experimental characterization led to functional TFLN hardware with resonator-based frequency selectivity, electrically controlled channel tuning, and stable optical and electrical access. Measurements demonstrate stable and programmable frequency-bin control: drift-compensated bias control mitigates electro-optic relaxation and stabilizes multiple frequency channels, while dual-comb spectroscopy verifies repeatable spectral phase programming and waveform reconstruction. These results establish a foundation for future integration of a complete frequency-bin quantum photonic system on TFLN. To support artificial-intelligence-assisted design of photonic and electromagnetic systems, this dissertation further develops a data-driven framework for electromagnetic wave modeling. The EM-OmniVerse dataset and foundation-model framework organize simulated field dynamics together with material, source, geometry, and language information. Across seven out-of-distribution transfer domains, the jointly trained foundation model gives the lowest prediction error in five cases and the second-lowest in two. A representative autoregressive test also shows improved long-horizon stability relative to a single-domain baseline. These results connect experimental photonic-platform development with scalable computational tools for electromagnetic design.