The realization of a "quantum internet" hinges on the ability to integrate Quantum Key Distribution (QKD) into existing telecommunications infrastructure. However, a significant hurdle remains: classical optical communications utilize relatively high-power lasers, while QKD relies on fragile, ultra-weak single photons. When these signals share the same optical waveguide, the intense classical light generates noise—through scattering and non-linear effects—that can easily overwhelm the quantum signal.
Researchers have now demonstrated a breakthrough on-chip platform that allows these two disparate signal types to coexist on a single waveguide without compromising quantum integrity. Built on an "etchless" thin-film lithium niobate (TFLN) platform, this millimeter-scale chip utilizes a combination of wavelength and polarization multiplexing to achieve the extreme isolation required for scalable, quantum-secure networks.
The Challenge of Coexistence
In traditional fiber-optic networks, coexistence is often managed using bulky, discrete components that are difficult to scale or integrate into compact photonic circuits. While integrated photonics platforms like silicon-on-insulator (SOI) offer miniaturization, they suffer from significant non-linear noise, such as two-photon absorption, when subjected to the high power levels typical of classical data streams. This noise directly degrades the quantum states, making SOI less ideal for high-power coexistence scenarios.
An Etchless TFLN Solution
To address these limitations, the research team turned to Thin-Film Lithium Niobate (TFLN). TFLN is prized for its low propagation loss and high optical damage threshold. Crucially, it lacks the specific non-linear noise issues found in silicon.
However, lithium niobate is notoriously difficult to etch precisely. Direct etching often results in rough, slanted sidewalls that increase signal loss. To circumvent this, the researchers employed an "etchless" fabrication technique. Instead of etching the TFLN layer directly, they deposited a layer of silicon nitride ($Si_3N_4$) on top and etched the waveguide patterns into the nitride. This "loaded" waveguide design guides light through the underlying TFLN while maintaining smooth interfaces and low loss.
Device Design and Architecture
The chip features a hybrid multiplexing architecture that combines Wavelength-Division Multiplexing (WDM) and Polarization-Division Multiplexing (PDM) within a compact $2.7 \times 0.68$ mm$^2$ footprint.
The core of the device is a Polarization Splitter-Rotator (PSR) connected to two 4-channel WDMs. The PSR separates the Transverse Electric (TE) and Transverse Magnetic (TM) polarization modes. Each polarization then enters a dedicated WDM built from 1D photonic crystal waveguides, creating a total of eight distinct channels. By separating signals by both wavelength (with 12 nm spacing) and polarization, the system achieves massive isolation—exceeding 100 dB when signals are both orthogonally polarized and on different wavelengths.
Experimental Results and Performance
The fabricated device demonstrated exceptional optical performance. The PSR achieved an ultra-low on-chip insertion loss (IL) of 0.23 dB for the $TE_0$ mode and 0.33 dB for the $TM_0$ mode. Polarization crosstalk was kept below -56.40 dB, and wavelength-channel crosstalk remained below -20.96 dB across all channels.
To prove the platform's viability for QKD, the team conducted coexistence experiments. They emulated classical traffic using 1.1 mW semiconductor lasers and quantum traffic using a heralded single-photon source. Even under simultaneous transmission, the quantum channel suffered only a minimal relative loss (~1.9 to 2.4 dB) in single-photon coincidence counts, confirming that the quantum state successfully survived the presence of intense classical light.
Furthermore, the design proved remarkably robust to manufacturing variations. A waveguide width variation of $\pm 20$ nm resulted in only a $\pm 0.04$ nm wavelength drift, indicating a practical process window for large-scale commercial manufacturing.
Future Directions
While the results are promising, the researchers identified areas for further optimization. Current crosstalk and loss were slightly higher than simulations predicted, likely due to sidewall roughness in the etched silicon nitride layer. Future iterations will focus on smoothing these surfaces and tightening wavelength control to meet strict telecommunications standards.
Additionally, the current chip uses static routing. Integrating thermo-optic or electro-optic tuning elements would allow for dynamic network reconfiguration. Ultimately, the goal is to integrate quantum light sources and single-photon detectors directly onto the TFLN platform, creating a fully integrated, scalable solution for quantum-secure communications.
Summary updated