🔬 optics

Visible Spectral-Domain Optical Coherence Tomography for Photonic Integrated Circuits Characterization

This paper introduces a visible spectral-domain optical coherence tomography (SD-OCT) technique that enables high-resolution, nondestructive, single-port characterization of guided-mode back-reflections in visible photonic integrated circuits, achieving shot-noise-limited sensitivity and 8 µm axial resolution for applications ranging from AR/VR to quantum control.

Original authors: Yin Min Goh, Chao Li, Yunchan Hwang, Helaman Flores, Mahmoud Jalali Mehrabad, James G. Fujimoto, Dirk R. Englund

Published 2026-03-26

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yin Min Goh, Chao Li, Yunchan Hwang, Helaman Flores, Mahmoud Jalali Mehrabad, James G. Fujimoto, Dirk R. Englund

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 have built a tiny, incredibly complex city for light to travel through. This city is called a Photonic Integrated Circuit (PIC). Instead of cars on roads, it has beams of light zipping through microscopic glass tunnels (waveguides) on a silicon chip. These chips are the brains behind future technologies like augmented reality glasses, quantum computers, and ultra-fast internet.

But here's the problem: How do you check if the roads are broken without tearing the city apart?

Currently, checking these chips is like trying to diagnose a traffic jam in a city by only looking at the exit. You can see if cars are leaving, but you can't tell where the accident happened, if a pothole is causing a delay, or if a bridge is shaky. You have to cut the chip open or build special "test exits" that take up valuable space and mess up the design.

The Solution: A "Flashlight" for the Invisible

The researchers in this paper have invented a new way to look inside these light cities without breaking them. They adapted a medical imaging tool called Optical Coherence Tomography (OCT)—the same technology doctors use to take 3D pictures of your retina—and turned it into a diagnostic tool for computer chips.

Think of it like this:

1. The "Echo Location" Analogy

Imagine you are standing in a long, dark hallway (the chip). You shout a short, sharp sound (a pulse of light).

If the hallway is empty, the sound just travels away.
But if there are mirrors, cracks, or bumps in the walls, some of that sound bounces back to you as an echo.

In this experiment, the scientists shine a broad spectrum of visible light (like a super-bright, multi-colored flashlight) into one end of the chip. As the light travels, tiny imperfections, bends, and connections inside the chip act like little mirrors, sending tiny echoes back.

2. The "Rainbow Ruler"

Here is the magic trick. The scientists don't just listen for the echo; they look at what color the echo is.

They mix the returning echo with a "reference" beam of light (like a control sound).
Because the light is made of many colors (a rainbow), the interference between the echo and the reference creates a unique pattern of stripes, called fringes, on a camera.
By analyzing these stripes, they can calculate exactly how far away the reflection came from.

It's like having a ruler that can measure distance just by listening to the pitch of a sound. If an echo comes back quickly, the obstacle is close. If it takes longer, the obstacle is far away.

What They Discovered

Using this "Flashlight Echo" method, the team was able to:

Map the Journey: They could see exactly where light was getting lost or bouncing back inside a ring-shaped circuit, even though they only had access to one end of the chip.
Find the "Ghost" Interfaces: They tested a chip that had a tiny diamond "micro-chip" attached to it (used for quantum computing). They could clearly see the boundary where the light jumped from the silicon chip to the diamond and back, measuring exactly how much light was lost in that jump.
Measure Quality: They could tell how "smooth" the roads were. If the walls of the light tunnels were rough, the light scattered. Their tool could spot these tiny rough spots that other methods missed.

Why This Matters

Before this, checking a chip was like trying to fix a car engine by only looking at the exhaust pipe. You knew something was wrong, but you didn't know what or where.

Now, with this new tool:

No More Cutting: You don't need to destroy the chip to test it.
No Extra Ports: You don't need to build extra "test exits" on the chip, saving space for the actual technology.
Super Fast: It can scan the whole chip in seconds, showing a detailed 3D map of where the light is flowing and where it's getting stuck.

In short, the researchers turned a medical eye-scan into a "chip-scan." This allows engineers to build better, faster, and more reliable light-based computers and sensors, ensuring that the light flows exactly where it's supposed to, all without ever taking the chip apart.

1. Problem Statement

Photonic Integrated Circuits (PICs) operating in the visible spectrum are critical for emerging applications such as AR/VR, LiDAR, and quantum control (e.g., diamond color centers, trapped ions). However, characterizing these circuits presents significant diagnostic challenges:

Lack of High-Resolution Tools: Unlike infrared silicon photonics, which utilizes Optical Frequency-Domain Reflectometry (OFDR) for non-destructive, high-resolution diagnostics, visible PICs lack comparable tools. Tunable laser sources for visible wavelengths are scarce and expensive.
Limitations of Current Methods: Existing verification methods are often bulky, slow, or destructive.
- Cut-back/Transmission measurements: Collapse spatial information into a single number, failing to distinguish between distributed losses (e.g., sidewall roughness) and localized defects (e.g., couplers, junctions).
- Top-view imaging: Limited by field of view, depth of focus, and scattering from multi-layer stacks.
- Design-for-Test (DFT): Requires extra ports and monitor taps, consuming chip area and perturbing circuit performance.
The Gap: There is a need for a platform-agnostic, fast, non-destructive, single-port diagnostic tool capable of resolving propagation loss, backscattering, and dispersion with micrometer-scale resolution and millimeter-scale imaging depth.

2. Methodology

The authors adapted Spectral-Domain Optical Coherence Tomography (SD-OCT), a technique traditionally used in biomedical imaging, for on-chip guided-mode characterization.

System Architecture:
- Interferometer: A modified fiber-based Michelson interferometer.
- Sample Arm: Uses microscope objectives (e.g., 100×, NA 0.95 for cleaved edges; 10×, NA 0.25 for grating couplers) to efficiently couple broadband light into tightly confined visible waveguide modes.
- Reference Arm: Replicates sample optics but uses a metallic mirror in a "cat's-eye" retroreflector configuration to ensure stability and alignment tolerance.
- Detection: A high-resolution spectrometer records the interference spectrum ( $I(k)$ ) in a single shot.
Operation Principle:
- Broadband light is injected into the PIC. Weak back-reflections and back-scatter from imperfections, interfaces, and facets return to the input port.
- These return signals interfere with the reference arm field.
- A Fourier transform of the spectral fringes yields an axial reflectivity profile (A-scan), mapping reflectivity versus depth along the optical path.
Key Technical Adaptations:
- Single-Port Access: The system only requires access to the input coupler, eliminating the need for output ports or DFT structures.
- Dispersion Compensation: A two-step numerical procedure is employed:
  1. Correction of system dispersion (fibers, optics) to maximize the sharpness of the initial reflection.
  2. Correction of waveguide-induced Group Velocity Dispersion (GVD) using Finite-Difference Time-Domain (FDTD) simulations to restore intrinsic pulse width.
- Stitching: For circuits longer than the spectrometer's imaging depth (~2 mm), the reference arm is scanned to stitch multiple A-scans into a continuous profile.

3. Key Contributions

First Visible-PIC SD-OCT Platform: Demonstrated the adaptation of SD-OCT specifically for visible-wavelength PICs (400–800 nm range), overcoming the scarcity of tunable visible sources.
Single-Port Reflectometry: Proved that complex circuit parameters (loss, coupling, dispersion) can be extracted using only a single input/output port, removing the need for redesigning chips for testing.
Hybrid Quantum Diagnostics: Successfully applied the technique to heterogeneous systems, specifically PICs integrated with diamond quantum micro-chiplets (QMC), resolving interfaces between the silicon nitride waveguide and the diamond chip.
Quantitative Parameter Extraction: Developed a model-based inversion method to extract coupling coefficients and intrinsic loss directly from "echo trains" in the reflectogram.

4. Results

The system was validated on two distinct platforms:

A. Ring Resonator Characterization (SiN)

Setup: A Si $_3$ N $_4$ ring resonator coupled to a bus waveguide.
Observation: The OCT reflectogram revealed a sequence of "echoes" corresponding to round-trip times in the ring.
Metrics Achieved:
- Axial Resolution: ~8 µm in silicon nitride (16 µm in air).
- Imaging Depth: 2 mm at 6 dB roll-off (extendable via scanning).
- Dynamic Range: 50 dB.
- Sensitivity: Shot-noise-limited.
Quantitative Analysis: By fitting the decay envelope of the echo train, the authors extracted a loaded quality factor ( $Q_L$ ) of $4.5 \times 10^4$ and a propagation loss of 4.0 dB/cm. These values matched independent transmission-based measurements within 10%, validating the single-port approach.

B. Hybrid Quantum Photonics (PIC + Diamond QMC)

Setup: A PIC with on-chip polarizers, beam splitters, and a socket interface to a diamond waveguide array hosting quantum memories.
Imaging: The system resolved distinct features including the input/output facets, the diamond QMC interface, and polarizer perturbations.
Quantitative Analysis:
- Measured the splitting ratio of a beam splitter channel (~1:3), matching design specifications.
- Calculated the transmission coefficient between the diamond QMC and the PIC, inferring a loss of 0.8 dB (transmission coefficient of 0.8).
- Successfully distinguished the QMC-PIC transition loss from splitter loss, which is impossible with standard transmission measurements.

5. Significance and Impact

Rapid Debugging & Yield: Provides a fast, non-destructive method to localize defects (e.g., sidewall roughness, bad couplers) without destroying the chip or requiring complex test fixtures.
Physics Insights: Converts "nuisance" backscattering into a spatially localized metric. This is crucial for high-Q resonators where coherent backscattering scales quadratically with group index, and for topological photonics where scattering distribution matters more than total loss.
Scalability: The method is compatible with 3D and hybrid architectures (MEMS/NEMS, quantum composites) and scales with circuit complexity.
Future Potential: The framework opens the door for spectroscopic OCT in photonics, potentially enabling wavelength-dependent loss mapping and functional diagnostics (e.g., identifying material absorption vs. scattering) by adapting techniques from OCT angiography.

In summary, this work establishes a practical metrology blueprint for the visible photonic era, enabling designers to visualize and quantify how light actually flows, reflects, and disperses within complex integrated circuits using a single optical port.