Material reference

Silicon nitrideSi₃N₄

The low-loss, high-power passive platform. Moderate index contrast, transparency from the visible into the mid-IR, negligible nonlinear absorption at telecom wavelengths. No gain, detection, or fast modulation of its own.

Silicon nitride is what you reach for when the circuit has to be quiet: ultra-low-loss delay lines, high-Q resonators, visible-light photonics, and anything carrying enough optical power to make silicon misbehave. Its ~5 eV bandgap eliminates two-photon absorption at 1550 nm entirely, so nitride resonators can store watts where silicon saturates at milliwatts.

Key properties

Property	Value	Conditions
Refractive index $n$	1.996	1550 nm, LPCVD stoichiometric
Index contrast vs. SiO₂	$\Delta n \approx 0.55$	1550 nm
Transparency window	~0.4–4 μm	practical, annealed LPCVD
Thermo-optic coefficient $dn/dT$	$+2.45 \times 10^{-5}$ K⁻¹	1550 nm
Kerr index $n_2$	$\approx 2.5 \times 10^{-19}$ m²/W	1550 nm
Two-photon absorption	negligible	at 1550 nm ( $E_g \approx 5$ eV)
Typical propagation loss	0.1–0.5 dB/cm	standard LPCVD thin core
Demonstrated loss	< 1 dB/m	thick-core Damascene process

What the numbers mean in practice

The moderate index contrast is the defining trade. Nitride waveguides are less sensitive to sidewall roughness than silicon, which is where the extraordinary loss numbers come from, but bends need tens to hundreds of micrometers of radius, so circuits are physically larger.

Two deposition families matter. LPCVD films (deposited near 800 °C) are dense, stoichiometric, and low-loss, but the temperature rules them out for back-end integration above CMOS. PECVD deposits cold enough for back-end use, at the price of hydrogen incorporation: N–H bonds absorb near 1520 nm, right in the C-band. High-temperature annealing drives the hydrogen out where the thermal budget allows.

The tenfold-smaller $dn/dT$ relative to silicon makes nitride filters far more temperature-stable, and the combination of low loss, high power handling, and usable Kerr nonlinearity is why microresonator frequency combs were first tamed on this platform. Nitride has no bandgap transition, no Pockels effect, and no detection in the telecom bands: active functions come from heaters, hybrid integration, or a neighboring silicon layer.

References

K. Luke, Y. Okawachi, M. R. E. Lamont, A. L. Gaeta, M. Lipson, "Broadband mid-infrared frequency comb generation in a Si₃N₄ microresonator," Opt. Lett. 40, 4823 (2015).
A. Arbabi, L. L. Goddard, "Measurements of the refractive indices and thermo-optic coefficients of Si₃N₄ and SiOₓ using microring resonances," Opt. Lett. 38, 3878 (2013).
D. J. Blumenthal, R. Heideman, D. Geuzebroek, A. Leinse, C. Roeloffzen, "Silicon nitride in silicon photonics," Proc. IEEE 106, 2209 (2018).
J. Liu et al., "High-yield, wafer-scale fabrication of ultralow-loss, dispersion-engineered silicon nitride photonic circuits," Nat. Commun. 12, 2236 (2021).