Silicon nitride photonics

Silicon nitride ($n \approx 2.0$) clad in silica occupies the middle ground between silica PLCs and silicon-on-insulator: enough index contrast for compact circuits, little enough that sidewall scattering (the loss mechanism that dominates high-contrast platforms) is dramatically weakened.

Loss is the platform's headline. Standard LPCVD nitride runs at 0.1 dB/cm or below; thin-core, high-aspect-ratio geometries with careful anneals reach 1 dB/m and below, enabling on-chip resonators with $Q$ beyond $10^7$. Numbers like these turn circuit elements that are marginal in silicon (long delay lines, ultra-narrow filters, high-finesse cavities) into routine components.

Three further properties differentiate nitride from silicon. Its bandgap (5 eV) eliminates two-photon absorption at 1550 nm, so it tolerates optical powers that would choke a silicon waveguide. That is why Kerr frequency comb generation lives almost exclusively in Si₃N₄ microresonators, where high circulating power meets usable $n_2$. Transparency extends down through the visible, opening biophotonics and atomic/quantum applications closed to silicon. And its thermo-optic coefficient ($2.4\times10^{-5}\,\mathrm{K}^{-1}$, an order below silicon's) makes circuits more thermally stable, at the cost of weaker thermal tuning.

The tradeoffs: nitride is passive (no carrier-based modulation, no detection, no gain), so systems combine a nitride circuit with silicon, III-V, or lithium niobate functions through heterogeneous or hybrid integration. Bend radii run tens of microns rather than silicon's few microns, and thick stoichiometric films accumulate stress that constrains fabrication (addressed by crack barriers, trenching, or the photonic Damascene process).

Deposition splits by temperature budget: LPCVD (~800 °C, lowest loss, front-end only) versus PECVD (~300–400 °C, back-end compatible, historically lossier).