Understanding Intrinsic Loss in Thin-Film Lithium Niobate Ring Resonators via Adiabatic Coupling

Abstract

Thin-film lithium niobate (TFLN) has emerged as a versatile integrated photonics platform, combining strong electro-optic and nonlinear effects. Among TFLN devices, ring resonators play a central role in filtering, modulation, and nonlinear optical processes. However, intrinsic loss, which ultimately limits ring performance, is most often summarized by single-valued metrics, and its statistical variability across resonances has received limited attention. Here, we show that intrinsic loss rates in monolithic TFLN ring resonators follow a statistical distribution, comprising a baseline loss and a tail arising from discrete loss events. This behavior is revealed by characterizing 2233 resonances, using an adiabatic waveguide-ring coupling architecture that selectively excites the fundamental mode and yields clean spectra in the ultra-high-Qi regime. We find the most probable intrinsic loss rate ki = 2 pi x 10.4 MHz, indicating operation in a low-loss regime comparable to state-of-the-art thick silicon nitride platforms.

Figures (7)

• Figure 1: Mode-selective coupling in thin-film lithium niobate (TFLN) ring resonators.(a) Schematic of a bus-coupled racetrack resonator. The straight-section length ($L$) and bend radius ($R$) define the resonator geometry. The intrinsic loss and external coupling rates are denoted by $\kappa_i$ and $\kappa_e$, respectively. (b) Three representative coupling strategies. In conventional approaches, light from a single-mode bus waveguide is either (i) coupled into a wide, multimode ring, thereby exciting multiple modes, or (ii) the ring is tapered to a single-mode width in the coupling region, which can introduce additional scattering loss. In this work (iii), we use a wide, low-loss bus waveguide and ring resonator, but introduce an adiabatic coupler that gradually bends the bus waveguide toward the ring, enabling excitation of only the fundamental mode while preserving the wide ring geometry. (c) Optical microscope image of a fabricated racetrack resonator incorporating an adiabatic coupling region for selective fundamental-mode excitation. (d) False-colored scanning electron microscope (SEM) image of the coupling region. (e) Zoomed-in SEM image of the coupling gap.
• Figure 2: Measurement setup. A tunable laser is split using a 1/99 splitter, where the 99% arm is sent through a polarization controller (PC) to the device under test, and the 1% arm is routed to a Mach--Zehnder interferometer (MZI) to generate calibration fringes for wavelength referencing.
• Figure 3: Mode-selective coupling enabled by an adiabatic coupling geometry.(a) Schematic of the coupling region. The parameter $w$ denotes the lateral offset of the bus-waveguide center between the entrance and the midpoint of the coupling region, and $L$ is the total coupler length. (b) Simulated guided modes supported in a $3~\mu\mathrm{m}$-wide TFLN waveguide (TE$_0$–TE$_2$, TM$_0$–TM$_2$). (c) Simulated modal power fractions in the ring waveguide for coupling geometries with different $(w, L)$. The fundamental mode is excited at the bus-waveguide entrance (point A in panel (a)), while the mode composition is evaluated at the exit of the ring-waveguide coupling region (point B in panel (a)). (d--g) Normalized transmission spectra measured for the corresponding fabricated devices. (h--i) On-resonance transmission versus total linewidth $\kappa/2\pi$ for measured resonances. The solid curve indicates the theoretical linewidth–extinction relation for ideal coupling, assuming an intrinsic loss rate of $\kappa_i/2\pi = 10.5~\mathrm{MHz}$. Scattered points represent experimental data acquired for different coupling gaps, in the wavelength range of 1520-1580 nm.
• Figure 4: Statistical distribution of intrinsic loss rates $\kappa_{i}$ across multiple resonators.(a--l) Normalized histograms of the intrinsic linewidth $\kappa_{i}/2\pi$ for twelve ring resonators fabricated with the same coupling design $(w = 5~\mu\mathrm{m}, L = 300~\mu\mathrm{m})$. Resonances within the 1520–1580 nm wavelength range are extracted and fitted to obtain $\kappa_{i}$. Vertical guides indicate the mean (light blue, long-dashed), median (teal, dotted), and mode (light green, dash–dot) of each device's distribution. The x-axis range is fixed from 5 to 26 MHz for consistent comparison. Devices (a--j) are fabricated on the same chip, while (k) and (l) are fabricated on two additional chips. The systematic separation between mean, median, and mode across all devices highlights the intrinsically asymmetric nature of loss statistics.
• Figure 5: Statistical characterization and modeling of intrinsic loss.(a) Combined normalized histogram of intrinsic linewidth $\kappa_{i}/2\pi$ from ten devices (2,233 resonances). The distribution exhibits a Gaussian-like core (pink) together with a truncated power-law-like tail (purple). (b) Complementary cumulative distribution function (CCDF) of the same dataset, showing a deviation from Gaussian behavior in the high-loss regime, with approximately power-law-like decay over an intermediate range followed by a gradual cutoff at large linewidths. (c) Intrinsic linewidth $\kappa_{i}/2\pi$ versus wavelength for three representative devices (corresponding to devices in Fig. 4(a--c)). Each point corresponds to a single resonance. The shaded band denotes the Gaussian core ($\mu \pm 2\sigma$) extracted from the global fit in (a), illustrating that most resonances cluster around the baseline loss, while high-loss events occur sporadically across the wavelength range. For clarity, the histogram is displayed up to $\kappa_i/2\pi = 45~\mathrm{MHz}$, and the CCDF is displayed up to $\kappa_i/2\pi = 70~\mathrm{MHz}$, focusing on the regime with robust statistical support.