Research

Themes in computational and integrated photonics

My research is organized by physical and computational themes rather than by isolated outputs. The common thread is the link between photonic device physics, reproducible simulation, and design methods that remain constrained by the underlying equations.

AI, PINNs, and machine learning appear here as computational methods. The research identity remains centered on photonics, device physics, and scientific modeling.

Integrated Nanophotonics

Problem: Chip-scale photonic systems require waveguides, couplers, detectors, and multiplexing structures that preserve bandwidth and modal control while remaining compatible with realistic fabrication constraints.
Methods: Waveguide and mode analysis, FDTD simulation, drift-diffusion modeling, compact-model extraction, and layout-aware validation.
Physical relevance: The theme connects modal confinement, dispersion, absorption, carrier transport, and parasitics into one design loop rather than treating device metrics in isolation.

Nonlinear Quantum Photonics

Problem: Integrated quantum photonic systems need efficient light generation, manipulation, and detection under loss, phase-matching, and coupling constraints.
Methods: Nonlinear optics analysis, source and detector structure optimization, phase-matching reasoning, and simulation-guided design of integrated quantum photonic components.
Physical relevance: The central physical question is how geometry and material response shape photon-pair generation, bandwidth, collection efficiency, and integration density.

Physics-Informed Design

Problem: Target photonic responses can be hard to derive manually from geometry, while purely data-driven models become unreliable without physical constraints.
Methods: Physics-informed neural networks, residual-based loss terms, gradient-based adjoint optimization, parametric sweeps, and electromagnetic validation solvers.
Physical relevance: This framework ensures optimized device structures satisfy Maxwell equations and boundary conditions, rather than treating optimization as black-box fitting.