🔬 optics

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

By characterizing 2,233 resonances in monolithic thin-film lithium niobate ring resonators using an adiabatic coupling architecture, this study reveals that intrinsic loss follows a statistical distribution with a baseline and discrete event tail, demonstrating a most probable loss rate comparable to state-of-the-art silicon nitride platforms.

Original authors: Xinrui Zhu, Hana K. Warner, Yunxiang Song, Donald Witt, Marko Loncar

Published 2026-04-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Xinrui Zhu, Hana K. Warner, Yunxiang Song, Donald Witt, Marko Loncar

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build the world's most perfect echo chamber. You want sound (or in this case, light) to bounce around inside a circular track for as long as possible without losing any energy. This is the goal of ring resonators in the world of integrated photonics (chips that use light instead of electricity).

The material the researchers are using is Thin-Film Lithium Niobate (TFLN). Think of this material as a "super-material" that is incredibly good at manipulating light, making it a top contender for future super-fast computers and quantum computers.

However, there's a problem: Loss. Just like a real echo chamber has cracks in the walls or dusty corners that swallow the sound, these light rings have tiny imperfections that steal the light energy. The researchers wanted to know exactly how much light is being stolen and why.

Here is the story of their discovery, broken down simply:

1. The Problem: The "Noisy" Coupler

To get light into the ring, you need a "bus" (a straight road) to connect to the "ring" (the circular track).

The Old Way: Previous methods were like trying to merge a single-lane highway into a massive, multi-lane superhighway. It was messy. The light would get confused, jump into the wrong lanes (higher-order modes), or crash into the walls, creating a lot of "noise" and losing energy.
The New Way (Adiabatic Coupling): The team invented a clever "ramp." Imagine a highway that slowly, gently curves and widens to merge perfectly with the superhighway. This is called adiabatic coupling. It ensures the light enters the ring smoothly, staying in the perfect "lane" (the fundamental mode) without crashing or getting lost. This gave them a very clean, quiet signal to measure.

2. The Big Data Experiment

Instead of testing just one or two rings, they tested 2,233 different light resonances (basically, 2,233 different "notes" the rings could sing). This is like listening to a choir of 2,000 singers to understand the average voice quality, rather than just listening to one person.

3. The Discovery: The "Base" and the "Tail"

When they looked at the data, they found something fascinating about how the light is lost. They expected the loss to be a simple, average number. Instead, they found a statistical distribution that looks like a hill with a long tail.

The Hill (The Baseline): Most of the time, the light loses a small, consistent amount of energy. This is the "background noise" of the universe—tiny scratches on the walls, dust on the lens, or the material itself absorbing a tiny bit of heat. This is predictable and happens to everyone.
The Tail (The Discrete Events): Occasionally, a specific ring or a specific "note" would lose a huge amount of energy. This is the "tail" of the graph. The researchers realized these aren't just random errors; they are rare, specific accidents. Maybe a tiny defect in the material happened to be right where the light wave was strongest at that exact moment, or a specific imperfection caused a "short circuit" for the light.

The Analogy:
Think of driving a car.

The Baseline: You always lose a little gas just by idling and rolling on the road (friction). This is normal.
The Tail: Sometimes, you hit a pothole, or a bird flies into your windshield, or you accidentally hit a patch of ice. These are rare, discrete events that cause a sudden, massive loss of control (or in this case, light).

4. The Result: A World-Class Performance

By using their new "ramp" (adiabatic coupling) and analyzing this massive amount of data, they found that the most common (or "most probable") loss rate is incredibly low.

They calculated that their rings can keep light bouncing around for a very long time, achieving a quality factor (a score for how good the ring is) of about 19 million.

This puts them on par with the best "thick silicon nitride" chips currently in existence, which are considered the gold standard.
The best part? They achieved this with a relatively simple, standard manufacturing process, not a super-complicated, expensive one.

Why Does This Matter?

This paper changes how we think about building these chips.

It's not just about the average: You can't just say, "Our chips have an average loss of X." You have to understand that there is a "baseline" you can engineer, and a "tail" of rare disasters you have to hunt down.
Better Design: By using the smooth "ramp" coupling, they eliminated the confusion of light entering the ring, allowing them to see the true performance of the material.
Future Tech: Because these rings are so good at holding light without losing it, they are perfect for building the next generation of quantum computers, ultra-fast internet modems, and sensors that can detect the tiniest changes in the world.

In a nutshell: The researchers built a smoother on-ramp for light to enter a circular track, listened to thousands of "notes" to map out exactly where the light was leaking, and discovered that while there are always tiny leaks (the baseline), the big leaks are rare accidents. With this knowledge, they proved their new design is one of the best in the world at keeping light trapped and useful.

1. Problem Statement

Thin-film lithium niobate (TFLN) is a premier platform for integrated photonics due to its strong electro-optic and nonlinear properties. Ring resonators are critical components in this platform for filtering, modulation, and nonlinear processes. However, the performance of these resonators is fundamentally limited by intrinsic loss ( $\kappa_i$ ).

Previous studies have characterized TFLN loss using single-valued metrics (e.g., a single $Q_i$ value per device), which fails to capture the statistical variability of loss across different resonances within a single device or across a wafer. Furthermore, traditional coupling schemes used to excite resonators often introduce trade-offs:

Coupling a single-mode bus to a wide, low-loss ring often excites multiple mode families, creating spectral clutter.
Narrowing the waveguide at the coupling region to enforce single-mode operation reintroduces scattering loss due to increased mode overlap with sidewall roughness.

The authors aim to resolve the trade-off between low scattering and mode selectivity while providing a rigorous statistical framework to understand the distribution of intrinsic losses in monolithic TFLN devices.

2. Methodology

Device Design: Adiabatic Coupling Architecture

To achieve single-mode excitation without sacrificing the low-scattering benefits of wide waveguides, the authors designed a mode-selective adiabatic coupler:

Geometry: A racetrack resonator with a wide waveguide ( $3 \, \mu\text{m}$ ) and Euler bends ( $R=120 \, \mu\text{m}$ ).
Coupling Strategy: Instead of tapering the ring to a narrow width, the bus waveguide is adiabatically bent toward the wide ring. The bus waveguide transitions from a single-mode width ( $0.8 \, \mu\text{m}$ ) to the ring width ( $3 \, \mu\text{m}$ ) via a squared-sinusoidal trajectory.
Principle: This gradual evolution ensures the optical mode remains in the fundamental eigenstate ( $TE_0$ ) throughout the coupling region, suppressing higher-order modes while maintaining the wide geometry that minimizes sidewall scattering.

Fabrication and Characterization

Materials: Devices were fabricated on X-cut TFLN-on-insulator wafers (600 nm device layer, 4.7 $\mu$ m buried oxide).
Process: Electron-beam lithography (HSQ resist), Ar plasma etching, SC-1/HF cleaning, and oxygen annealing ( $520^\circ\text{C}$ ) to mitigate plasma damage.
Measurement: A tunable laser system with a Mach–Zehnder interferometer (MZI) for wavelength calibration was used to measure transmission spectra across 12 nominally identical devices (spanning 1480–1600 nm).
Data Extraction: A statistical analysis was performed on 2,233 resonances extracted from these devices. The intrinsic loss rate $\kappa_i$ was isolated from the total decay rate $\kappa$ by fitting the transmission spectra using a model that accounts for backscattering-induced mode splitting.

3. Key Contributions

Novel Coupling Architecture: The demonstration of an adiabatic coupling scheme that enables fundamental-mode-only excitation in wide-waveguide TFLN rings. This eliminates the need for local waveguide narrowing, thereby preserving the ultra-low scattering characteristics of the wide geometry.
Statistical Framework for Loss: Moving beyond single-value metrics, the authors characterize intrinsic loss as a statistical distribution. They reveal that loss is not uniform but comprises:
- A Gaussian-like baseline representing cumulative small, uncorrelated imperfections (roughness, absorption).
- A truncated power-law tail representing discrete, probabilistic high-loss events (e.g., specific defect configurations or mode interactions).
High-Performance Benchmark: Establishing that TFLN, when optimized with this architecture, operates in a low-loss regime comparable to state-of-the-art thick silicon nitride (SiN) platforms.

4. Results

Mode Purity: Simulations and experimental spectra confirmed that the adiabatic coupler ( $L=300 \, \mu\text{m}$ , lateral offset $w=5 \, \mu\text{m}$ ) suppresses higher-order modes. Devices with shorter or non-adiabatic couplers exhibited multiple resonance families, whereas the optimized design showed clean, single-family spectra.
Loss Statistics:
- Most Probable Loss Rate: The mode (most probable value) of the intrinsic loss distribution is $\kappa_i = 2\pi \times 10.4 \, \text{MHz}$ .
- Quality Factor: This corresponds to an intrinsic quality factor of $Q_i \approx 1.9 \times 10^7$ .
- Distribution Shape: The aggregated data (2,233 resonances) showed a mean $\kappa_i/2\pi$ of $13.5 \, \text{MHz}$ and a median of $11.4 \, \text{MHz}$ . The significant difference between the mean and the mode highlights the impact of the "tail" events (rare high-loss resonances).
- Baseline vs. Tail: The Gaussian core ( $\mu \approx 10.48 \, \text{MHz}$ ) represents the fundamental material and fabrication limit. The tail arises from sporadic, conditionally activated loss channels (e.g., specific defect alignments or backscattering).
Comparison: The achieved $Q_i$ ( $\sim 2 \times 10^7$ ) is comparable to the best thick SiN platforms ( $3-4 \times 10^7$ ), but achieved via a straightforward monolithic fabrication process without complex reflow or mechanical polishing.

5. Significance

Paradigm Shift in Loss Analysis: The paper argues that optimizing TFLN devices requires addressing two distinct challenges: refining the baseline loss (via material/geometry optimization) and mitigating rare tail events (via defect control and statistical yield improvement). This provides a more nuanced roadmap for loss engineering than simply maximizing the average $Q$ .
Platform Maturity: By demonstrating $Q_i \sim 10^7$ with clean single-mode spectra, the work confirms TFLN as a mature platform for both classical and quantum photonic applications, particularly where high nonlinearity and electro-optic efficiency are required alongside low loss.
Scalability: The adiabatic coupling design is compatible with standard fabrication processes, offering a scalable route to high-performance photonic integrated circuits without the complexity of hybrid bonding or post-fabrication polishing.

In summary, this work provides a comprehensive physical and statistical understanding of loss in TFLN resonators, introducing a coupling architecture that unlocks the full potential of the material's low-scattering capabilities.