🔬 optics

Transverse momentum as the counter-diabatic generator in bent waveguide couplers

This paper identifies transverse momentum as the inherent counter-diabatic generator in bent waveguide couplers, deriving a closed-form expression for the optimal bending axis that achieves near-perfect supermode fidelity without requiring numerical searches or unitary transformations.

Original authors: Yu-kai Lee, Shuo-Yen Tseng

Published 2026-09-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yu-kai Lee, Shuo-Yen Tseng

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

Light traveling through a tiny, glass-like channel is a bit like a river flowing through a valley. If the valley walls are straight and the water moves slowly, the flow stays smooth and predictable. But if the valley suddenly narrows, widens, or twists, the water can churn, splash against the banks, or even spill over into a different channel entirely. In the world of modern optics, scientists build these "valleys" using microscopic channels called waveguides to guide light through computer chips and communication devices. The challenge is to move light from one channel to another without losing any of it or letting it scatter into the wrong place. For decades, the standard way to do this has been to make the transition very long and very gradual, allowing the light to adjust itself slowly, much like a car taking a wide, gentle curve rather than a sharp turn. This method works well, but it requires devices that are physically long, taking up valuable space on a microchip.

Researchers have long sought a way to make these transitions short and fast without losing the light, a goal known as finding a "shortcut to adiabaticity." One promising idea involves adding a specific, invisible force to the system that cancels out the turbulence caused by the rapid change. In the language of physics, this requires a very specific type of interaction that the standard tools for shaping light—changing the width of the channels or the distance between them—simply cannot provide. These standard tools can only push the light in two directions, but the solution requires a push in a third, perpendicular direction that the geometry of the channels cannot naturally create. For a long time, the only way to achieve this was to use a complex mathematical trick to redesign the entire shape of the channels, making them incredibly difficult to manufacture because the spacing between them would need to change with extreme precision.

A new study by Yu-kai Lee and Shuo-Yen Tseng at National Cheng Kung University in Taiwan has discovered that nature already provides the missing piece, and it is much simpler than anyone realized. The researchers found that by simply bending the path of the waveguide channels, they could generate the exact force needed to keep the light perfectly on track, even when the transition is very short. They identified that the physical act of moving the entire structure sideways as the light travels is governed by a fundamental property called transverse momentum. In the specific language of the light's behavior, this sideways motion creates a force that points exactly in the direction required to cancel out the turbulence. It is a bit like realizing that to steer a boat perfectly through a narrow, twisting canyon, you don't need to change the shape of the canyon walls; you just need to steer the boat itself along a specific curved path.

The team showed that this curved path is not a random guess or a result of trial and error. They derived a precise formula that tells engineers exactly how much to bend the waveguide at every single point along its length. This formula depends only on the shape of the light's path in a straight, unbent version of the device, meaning the design can be calculated directly without needing to run thousands of computer simulations to find the right curve. When they tested this idea using computer models, the bent waveguide performed almost perfectly. In a simulation where a standard, straight device managed to transfer only about 89 percent of the light to the new channel, the new bent design transferred 99.4 percent. They also compared this bent design to the complex, mathematically redesigned "gauge" version mentioned earlier. Both new designs performed equally well, reaching nearly 100 percent efficiency, and both were equally robust against small errors in manufacturing.

The most striking finding is that the bent device achieves this high performance without needing to change the width of the channels or the distance between them in the complicated ways required by the older method. The older method required the spacing between the two channels to shrink and expand in a very sensitive, exponential way, which is extremely difficult to build with current technology. The bent method keeps the spacing constant and the widths changing in a simple, standard way, relying entirely on the curve of the path to do the heavy lifting. The researchers confirmed that while the light behaves differently inside the two devices—following a straight path in one frame of reference and a curved path in another—the final result is identical. The light exits both devices with the same high quality and the same ability to handle different colors of light.

This work suggests that the key to faster, smaller, and more efficient optical chips may not be in inventing new materials or complex geometries, but in recognizing a simple physical principle that was already available. By understanding that moving a structure sideways generates the exact force needed to stabilize light, engineers can now design devices that are shorter and easier to build. The study provides a clear, closed-form recipe for creating these bends, removing the need for difficult numerical searches and making the technology accessible for immediate use in real-world applications. The bent waveguide coupler stands as a practical solution that matches the theoretical perfection of the most advanced methods, but with the simplicity of a single, smooth curve.

Technical Summary: Transverse Momentum as the Counter-Diabatic Generator in Bent Waveguide Couplers

Problem Statement
Mode-evolution couplers in integrated optics transfer light via the slow adiabatic evolution of a single supermode, offering robustness against fabrication errors and broad bandwidth. However, the requirement for slow geometric variation imposes a significant length penalty, often resulting in millimeter-scale devices. Shortcuts to adiabaticity (STA), specifically counter-diabatic (CD) driving, can theoretically eliminate this length penalty by adding a term to the Hamiltonian that exactly cancels non-adiabatic coupling.

In a standard two-waveguide coupler, the Hamiltonian is governed by detuning ( $\Delta$ , controlled by waveguide widths) and coupling ( $\Omega$ , controlled by spacing). In the Pauli matrix representation, these parameters lie in the $\sigma_z$ – $\sigma_x$ plane. The required CD term, however, is proportional to $\sigma_y$ . Since geometric width and spacing variations cannot generate a $\sigma_y$ term, direct implementation of CD driving in a straight coupler is impossible. The established solution involves a unitary transformation (the "gauge realization") that maps the CD dynamics onto a realizable Hamiltonian with modified $\Delta$ and $\Omega$ . This approach, however, requires simultaneous and precise modulation of both waveguide widths and spacing, where spacing is exponentially sensitive and difficult to fabricate.

Methodology
The authors investigate the physical operator responsible for the CD protocol in bent waveguide couplers. Rather than transforming the Hamiltonian to match a target, they analyze how a rigid lateral displacement of the waveguide structure enters the system's generator.

Operator Identification: The paper establishes that rigid lateral translation is generated by the transverse momentum operator, $\hat{p}_x = -i\partial_x$ . In a real supermode basis, the diagonal matrix elements of $\hat{p}_x$ vanish, and the off-diagonal element is purely imaginary. Consequently, in a two-mode system, $\hat{p}_x$ is strictly proportional to $\sigma_y$ .
Derivation of the Bending Profile: By equating the term supplied by the lateral displacement ( $-\dot{x}_0 \hat{p}_x$ ) to the required CD term ( $\dot{\theta}/2 \sigma_y$ ), the authors derive a closed-form expression for the axis slope $\dot{x}_0(z)$ . This expression depends only on matrix elements of the unperturbed supermodes at a single cross-section:
$\dot{x}_0(z) = \frac{\langle \psi_+ | \partial_z \psi_- \rangle}{\langle \psi_+ | \partial_x \psi_- \rangle} = \frac{-\dot{\theta}/2}{\langle \psi_+ | \partial_x \psi_- \rangle}$
Using a commutator identity, this is recast into coupled-mode variables as $\dot{x}_0 = \dot{\theta}/(\gamma \Delta\beta^2)$ , where $\gamma$ is a bend coefficient.
Numerical Verification: The authors simulate a polymer mode-evolution coupler using beam propagation methods. They compare:
- The untreated coupler.
- The "gauge realization" (straight coupler with modulated width and spacing).
- The "bent realization" (straight widths/spacing with a calculated axis bend).
- A numerical offset search to verify the analytical closed-form solution.

Key Contributions

Identification of the CD Generator: The paper identifies transverse momentum $\hat{p}_x$ as the fundamental operator that makes axis bending a CD protocol. This explains why bending works: it is the only geometric parameter that naturally generates the required $\sigma_y$ term without needing to transform the Hamiltonian.
Analytical Design Formula: The authors provide a closed-form solution for the bending profile (Eq. 18) that requires no trial geometries or numerical optimization. The profile is determined solely by the ratio of two matrix elements of the unperturbed structure.
Equivalence of Realizations: The study demonstrates that the bent coupler and the gauge realization are two different pictures of the same CD protocol. While their internal trajectories differ (related by a frame transformation), they yield identical output efficiencies, bandwidths, and fabrication tolerances.

Results

Accuracy: The analytical closed-form solution for the axis slope agrees with a direct numerical offset search to within 0.3%. The analytical form is shown to be more robust than the numerical search when cross-sections are sampled at realistic spacings.
Performance: Beam propagation simulations on a reference structure (length $L=1300$ $L = 1300$ $\mu$ $μ$ m) show:
- Untreated Coupler: Supermode fidelity of 0.886.
- Gauge Realization: Supermode fidelity of 1.000.
- Bent Realization: Supermode fidelity of 0.994.
Robustness: Both the bent and gauge realizations exhibit matching bandwidth (>170 nm for >0.99 fidelity) and fabrication tolerance (robust to $\pm 200$ nm width errors). The bent device shows slightly better tolerance to reduced spacing, while the gauge device tolerates increased spacing better.
Internal Dynamics: The internal population trajectories differ between the two devices, consistent with the frame transformation relating them. In the comoving frame of the bent device, the system follows the supermode with high fidelity, whereas in the laboratory frame, apparent supermode exchange occurs due to the tilt.

Significance and Claims
The paper claims that axis bending is a CD generator "by necessity rather than by design." By identifying transverse momentum as the source of the $\sigma_y$ term, the authors provide a direct physical explanation for the efficacy of bent couplers. The primary significance lies in the derivation of a closed-form design rule that eliminates the need for numerical optimization or the difficult fabrication of modulated waveguide spacing. The bent realization achieves the benefits of counter-diabatic driving (short length, high fidelity) without requiring the complex geometric modulation (simultaneous width and spacing control) demanded by the gauge realization. The authors note that the derivation is exact within the two-level description, with limitations arising only from coupling to the radiation continuum (tilt loss) and higher-order curvature effects, which are small for the weakly guiding structures considered.

More like this