🔬 optics

3D Photonic integration leveraging hybrid-confinement circuits

This paper demonstrates a scalable 3D photonic integration platform that monolithically combines high-confinement air-clad and low-confinement polymer-clad waveguides to achieve compact, low-loss routing, efficient mode transitions, and complex circuit components for applications like optical neural networks.

Original authors: Kanhaya Sharma, Adrià Grabulosa, Erik Jung, Daniel Brunner

Published 2026-05-26

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kanhaya Sharma, Adrià Grabulosa, Erik Jung, Daniel Brunner

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 a complex city of roads for tiny beams of light. For decades, engineers have been building these "light cities" on flat, two-dimensional maps (like a flat sheet of paper). While this works, it's getting crowded. Just like a flat city hits a limit on how many roads you can squeeze in before traffic jams occur, flat light circuits are hitting a physical wall where they can't get any smaller or more efficient.

To solve this, the researchers in this paper decided to stop thinking in 2D and start building in 3D. They created a new way to mix two different types of "roads" for light into a single, seamless structure.

Here is a breakdown of their invention using simple analogies:

1. The Two Types of Roads

The team realized that different parts of a light circuit need different types of roads. They built a hybrid system combining two distinct styles:

The "Air-Highway" (High-Confinement): Imagine a narrow, high-speed tunnel with no walls, just air around it. Because the light is squeezed tightly into this tiny tunnel, it can take extremely sharp turns without crashing or leaking out. This is great for making the circuit very compact and dense. However, because the tunnel is so tight, the light can get a bit "jittery" (scattering) if the road isn't perfectly smooth.
The "Polymer-Street" (Low-Confinement): Imagine a wider, gentler road paved with a special plastic (polymer). The light here travels more loosely. It's very stable and easy to control, making it perfect for long, straight stretches where you don't want the light to get confused. However, if you try to make a sharp turn on this wide road, the light spills over the edges and gets lost.

2. The Magic Bridge (The Transition)

The biggest challenge was connecting these two very different roads. If you try to drive a car from a wide, bumpy street directly into a tiny, high-speed tunnel, you might crash or lose speed.

The researchers built a perfectly smooth transition zone between the wide polymer road and the tight air tunnel. They carefully shaped the entrance so that the "shape" of the light beam matches perfectly as it moves from one road to the other.

The Result: They achieved a nearly invisible handoff. The light loses almost no energy (only about 0.25 dB) when switching between the two types of roads. It's like a seamless teleportation where the light doesn't even notice it changed lanes.

3. The Sharp Turns (Euler Bends)

In the old flat world, making a sharp turn for light usually meant losing a lot of signal. The researchers used a special mathematical curve called an Euler bend.

The Analogy: Think of a race car driver. If they turn the steering wheel instantly to the left, the car might spin out. But if they gradually turn the wheel, increasing the sharpness of the turn until they hit the tightest point, and then gradually straighten out, the car stays on the track.
The Result: Using this gradual curve, they managed to make light turn a full 90 degrees in a space as small as a human hair (10 micrometers) with almost no loss of signal.

4. The Splitter (The Fork in the Road)

They also built a device that takes one beam of light and splits it into three separate paths.

The Analogy: Imagine a river that needs to split into three smaller streams. In the past, these splits had to be very long and winding to work well. The researchers used their "Air-Highway" to create a splitter that is ten times shorter than previous versions. It's like shrinking a massive highway interchange down to the size of a small parking lot.

5. The Final Product: A 3D Light City

By combining these elements, they built a complete, working circuit on a single chip.

They used the Polymer-Street to bring the light in safely.
They used the Magic Bridge to move it into the Air-Highway.
They used the Sharp Turns to navigate tight corners.
They used the Splitter to direct the light to different destinations.

Why This Matters (According to the Paper)

The paper states that this approach is a major step forward for creating 3D optical neural networks (computers that think using light), photonic wire bonding (connecting chips with light instead of copper wires), and photonic lanterns (devices that manage many light signals at once).

The key takeaway is that by mixing these two types of light roads and building them in 3D using a high-precision "laser pen" (a technology called two-photon polymerization), they have created a way to pack much more light-processing power into a tiny space without losing signal quality. It's a move from building flat, crowded cities to building efficient, multi-layered skyscrapers for light.

Technical Summary: 3D Photonic Integration Leveraging Hybrid-Confinement Circuits

Problem Statement
Conventional planar (2D) photonic integration faces fundamental scaling limitations similar to those in electronics, where increased complexity leads to larger footprints and constrained routing flexibility. While 3D stacking of planar layers offers a partial solution, it relies on iterative 2D patterning, increasing fabrication complexity, cost, and limiting vertical resolution. Existing 3D lithography approaches, such as two-photon polymerization (TPP), enable true 3D freeform writing but often struggle to balance two conflicting requirements: the need for tight bending radii (requiring high refractive index contrast and strong mode confinement) and the need for robust single-mode operation with low scattering (often requiring lower contrast and smoother interfaces). Specifically, air-clad waveguides offer strong confinement for dense routing but are inherently multimodal at the resolution limits of current TPP processes, while polymer-clad waveguides support single-mode operation but require large bending radii to minimize loss.

Methodology
The authors propose and experimentally demonstrate a monolithic 3D photonic platform that integrates two distinct waveguide regimes: high-confinement air-clad waveguides and low-confinement polymer-clad waveguides. The fabrication utilizes a commercially available Direct Laser Writing (DLW) system (Nanoscribe Photonic Professional GT+) employing Two-Photon Polymerization (TPP) and Flash-TPP (combining TPP with One-Photon Polymerization, OPP).

The methodology involves:

Hybrid Fabrication Strategy:
- Air-clad sections: Fabricated as free-standing cylindrical pillars via TPP. The surrounding resin is removed during development. These sections utilize high refractive index contrast ( $\Delta n \approx 0.5$ ) to enable tight bending.
- Polymer-clad sections: Fabricated as a core within a TPP-printed "cage," which is subsequently filled and polymerized via OPP. These sections utilize low refractive index contrast ( $\Delta n \approx 6 \times 10^{-3}$ ) to ensure single-mode operation.
Mode Matching: The core radii of the air-clad and polymer-clad sections are tailored to match the effective refractive indices ( $n_{eff}$ ) and Mode Field Diameters (MFD) of their respective fundamental modes ( $LP_{01}$ ). This ensures adiabatic transition between the multimode air-clad and single-mode polymer-clad sections, minimizing excitation of higher-order modes.
Component Design:
- Bends: Euler S- and U-shaped bends are designed to provide a linearly decreasing curvature, ensuring adiabatic mode displacement from the waveguide center to the cladding interface, thereby reducing bending losses compared to circular bends.
- Splitters: Adiabatic air-clad splitters are designed by extruding the input waveguide cross-section along a parametric path, splitting the mode into multiple outputs without relying on evanescent field coupling.
Materials: The study utilizes IP-Dip photoresist with a 63× objective for high-resolution component characterization (bends, splitters) and IP-S photoresist with a 25× objective for the fabrication of complete hybrid circuits to leverage faster printing speeds and mechanical robustness.

Key Contributions and Results
The paper reports the successful fabrication and characterization of the following building blocks and hybrid circuits:

Confinement Transitions: Efficient mode transitions between polymer-clad and air-clad waveguides were demonstrated. By matching the fundamental modes, the authors achieved transition losses as low as 0.25 dB per interface.
Compact Bends:
- S-bends: Euler geometry S-bends achieved bending losses of 0.5 dB at a minimal bending radius of 10 µm. Circular bends showed higher losses and multimodal behavior at similar radii.
- U-bends: Euler geometry U-bends achieved losses as low as 0.4 dB at a bending radius of 10 µm. At radii below 10 µm (e.g., 4 µm), higher-order modes were observed, but single-mode operation was maintained at 10 µm.
Adiabatic Splitters: Compact air-clad splitters were realized with a splitting length of 52 µm, achieving a splitting loss of 0.6 dB. This represents a tenfold reduction in footprint compared to previous evanescent-field-based polymer-clad splitters.
Propagation Losses: The propagation loss of the fundamental mode ( $LP_{01}$ ) in the air-clad waveguides was measured to be 5.8 dB/mm at a radius of 3.0 µm.
Full Hybrid Circuit Integration: A complete hybrid circuit combining polymer-clad input/output sections with air-clad bends and splitters was fabricated on a single chip. While the integration using IP-S (25× objective) resulted in slightly higher losses for bends (2 dB) and splitters (1.5 dB) due to voxel asymmetry and axial resolution limitations (3.6 µm), it demonstrated the feasibility of the approach.

Significance
The authors claim that this work represents a significant step toward large-scale 3D photonic integration by leveraging the distinct advantages of both confinement regimes within a single monolithic platform. The hybrid approach allows for:

Dense Integration: Utilizing air-clad waveguides for tight bends and compact routing.
Robust Single-Mode Operation: Utilizing polymer-clad waveguides for mode control and sections requiring evanescent field manipulation.
Scalability: The approach is CMOS-compatible and utilizes additive manufacturing, offering a pathway to overcome the footprint scaling limitations of 2D planar circuits.

The paper identifies the technology as particularly relevant for applications such as optical neural networks, photonic wire bonding, and photonic lanterns. The authors note that further reduction in losses is anticipated by utilizing higher-resolution fabrication (IP-Dip with 63× objective) and operating at longer wavelengths (e.g., telecommunications band) to mitigate surface scattering and allow for larger waveguide diameters.