Part 1: Challenges of Low-Loss Fiber Attachment for CV Photonic Quantum Systems
1. Introduction
As photonic integrated circuits (PICs) move from laboratory prototypes to scalable quantum computing platforms, efficient and stable coupling between optical fibers and on-chip waveguides becomes a critical engineering requirement. In quantum photonics—particularly in continuous-variable (CV) computing schemes—the ability to inject and extract photons with minimal loss directly impacts the system’s fidelity, scalability, and fault tolerance.
Edge coupling, especially at telecom wavelengths (around 1550 nm), is often favoured for its low insertion loss and broadband performance. However, reliable and low-loss fiber attachment remains challenging for photonic system integration.
2. Why Fiber Attachment is Challenging
2.1 Mode Field Mismatch
The fundamental difficulty arises from a mismatch between the mode field diameter (MFD) of standard single-mode fibers (~10 µm at 1550 nm) and the much smaller and asymmetric optical modes in submicron silicon or silicon nitride waveguides (typically ~0.5–1.5 µm). Directly coupling these modes without a transition region leads to substantial coupling loss due to reflection and radiation.
2.2 Refractive Index Discontinuities
At the chip-fibre interface, refractive index discontinuities cause Fresnel reflections and transmission inefficiencies. For example, the index contrast between silica fiber (n ≈ 1.44) and silicon (n ≈ 3.48) or silicon nitride (n ≈ 2.0) is substantial, and without index matching, reflection losses can be significant.
Index-matching gels or adhesives are often introduced to reduce this discontinuity and minimize reflections. These materials must be optically transparent at the operating wavelength (typically 1550 nm), have low absorption, and maintain stability under thermal cycling.
2.3 Physical Alignment Sensitivity
Mechanical misalignment (lateral, vertical, and angular) is a critical source of coupling loss. Even sub-micron deviations can dramatically reduce coupling efficiency. The alignment must also be stable over time and robust to thermal expansion or packaging stresses.
2.4 Surface Roughness and Facet Quality
The cleaved or polished facet of the PIC must be optically smooth and perpendicular to the optical axis to avoid scattering losses. Any irregularities on the waveguide termination can reflect or scatter photons out of the guided mode, further degrading efficiency.
2.5 Packaging and Thermo-Mechanical Stresses
PICs undergo thermal cycling, mainly when used in high-power or cryogenic environments. Expansion mismatch between materials (e.g., chip, mount, adhesive, fiber ferrule) can degrade alignment over time. Robust packaging strategies must mitigate this via compliant layers or athermal designs.
3. Special Relevance to CV Quantum Photonic Systems
In CV photonic quantum computing, delays are introduced using fiber loops or waveguides to build time-multiplexed cluster states—large entangled graphs that serve as the resource for measurement-based quantum computation [2]. Any insertion loss per interface quickly becomes critical, especially when the same signal must traverse multiple loops and go in and out of the chip multiple times.
Losses directly impact:
- • Entanglement fidelity of cluster states
- • Squeezing degradation, critical for high-fidelity gate operations
- • Overall photon throughput, reducing computational depth and scalability
The ability to achieve a sub-1 dB coupling loss per facet and ensure long-term alignment stability between the chip and fiber is vital for maintaining the quality of generated Qubits promised by CV platforms.
4. Strategies to Address Fiber Coupling Loss
- • Spot Size Converters (SSCs) are one strategy to address the mode field mismatch. These converters gradually expand the waveguide mode using adiabatic tapers, multi-layer waveguides, or a combination of both, which helps to match the fiber MFD better. This technique reduces the coupling loss by providing a smooth transition between the waveguide and the fiber, thereby improving the overall coupling efficiency.
- • Index-Matching Gels/Adhesives: Minimize reflection and enable stable optical interface.
- • Active alignment: Real-time feedback is provided using optical power monitoring to ensure optimal coupling during packaging and epoxy curing. Electrically controlled mechanical stages continuously adjust the alignment based on the monitored power, compensating for any changes in the system and maintaining the coupling efficiency at its peak.
- • UV-curable adhesives: These adhesives secure the fiber with minimal shrinkage and good thermal stability. By providing a stable and reliable attachment, UV-curable adhesives can help maintain the alignment and reduce the coupling loss over time, thus contributing to the overall system performance.
- • Use UHNA (ultra high numerical aperture) fiber spliced to SMF: This reduces the beam size from 10um to 3-4um, depending on the type of UHNA fiber used. Typically, splicing losses can be less than 0.5dB, but iteration is needed to perfect the splicing recipe. Splicing could be challenging in a fiber array. Due to the smaller MFD, the coupling efficiency becomes more sensitive to alignment tolerances.
- • Lensed fiber: Lensed fibers can focus the beam at the exposed waveguide. In this case, we need to leave a gap, usually termed the working distance from the waveguide, which is an additional parameter to get perfect coupling, making the alignment process even more complex. Also, the path of light is not sealed, and hence, there is always a possibility of contamination and damage, especially at high optical powers. This technique is not suitable for practical products and is limited to clean rooms and experimental setups.
5. The Quanfluence PICs
Quanfluence currently has four distinct photonic integrated circuits in their labs, each tailored to explore and optimize low-loss fiber coupling. These PICs are fabricated on silicon nitride platforms and integrate features such as inverse tapers, oxide cladding, and phase modulation. Each chip is designed with test structures to evaluate coupling efficiency, alignment robustness, thermal stability, and electrical interfacing. Experimental setups using these chips have enabled systematic studies on fiber-to-chip alignment strategies and long-term packaging reliability, forming the basis of our end-to-end integration pipeline for quantum photonic systems.
6. Test and measurement set-up
A robust testing and alignment infrastructure is essential for evaluating the optical performance of photonic chips and ensuring consistent coupling. At Quanfluence, our lab has multiple 6-axis alignment stages both manual and piezo-actuated that allow precise control over fiber-to-chip positioning in XYZ and pitch/yaw/roll. These stages are integrated with real-time optical power monitoring systems to optimize coupling during live alignment. High-resolution microscopy is used for visual inspection and coarse alignment, while infrared imaging supports fine-tuning of edge-coupled interfaces.
Our setup also includes fiber stretchers for controlling delay lines and ensuring mechanical stability during long-duration experiments. All stations are designed for quick reconfiguration between single-fiber and fiber-array coupling experiments. The figure shows a photograph of the full test station, highlighting the typical alignment process and hardware used in production characterization.
Accurately measuring coupling losses under 1 dB is a non-trivial challenge, particularly at high optical powers (10–100 mW), where nonlinear effects and scattering become prominent. Subtle back-reflections or epoxy-induced stress can cause measurable performance drifts. Moreover, achieving reproducible alignment to such precision requires iterative feedback and environmental stability, underscoring the importance of our precision automation and real-time monitoring infrastructure.
7. The spot-size conversion problem
One of the central challenges in achieving low-loss optical coupling from silicon nitride (SiN) photonic integrated circuits to standard single-mode fiber (SMF-28) lies in the significant mismatch in optical mode sizes. SiN waveguides, while advantageous for low-loss propagation and wide transparency windows, typically confine light in submicron dimensions (<1um) due to the high index contrast with surrounding materials. In comparison, SMF-28 fibers support an optical mode field diameter (MFD) of ~10.4 µm at 1550 nm. This mismatch leads to high insertion losses unless an intermediate structure known as a spot size converter (SSC) is employed. The SSC gradually transforms the tightly confined mode in the SiN waveguide into a larger, more symmetric mode profile that matches the fiber. This process involves adiabatic tapering of the waveguide and the addition of a low-index overcladding (such as SiO₂) to guide the expanding mode. Designing this transition requires a delicate balance between taper length, fabrication tolerance, and coupling efficiency, as non-adiabatic transitions or abrupt discontinuities can induce scattering and reflection losses that are particularly detrimental in quantum applications.
8. The latest generation spot size converters
Quanfluence’s photonic chips integrate highly optimized spot size converters (SSCs) at the chip facets to bridge the modal mismatch between high-confinement silicon nitride waveguides and standard single-mode fibers. These SSCs employ inverse tapers and oxide overcladding to adiabatically expand the mode field diameter from submicron dimensions to over 10 µm, achieving excellent overlap with SMF-28 fibers. Recent fabrication runs have demonstrated coupling losses of <1 dB per facet, representing a significant improvement over earlier iterations. This level of performance ensures that fiber attachment no longer constitutes a dominant source of loss in system-level experiments and supports the fidelity requirements of quantum photonic computing platforms.
9. Working with fiber arrays and electrical contacts.
Quanfluence’s photonic packaging process is designed around the industry-standard 127 µm-pitch, single-mode fiber arrays, enabling compatibility with off-the-shelf v-groove assemblies and minimizing custom alignment overheads. Precise integration with these arrays still presents several engineering challenges. The parallel placement of multiple fibers introduces stringent yaw and roll tolerances, as even slight angular misalignments can lead to significant insertion loss across the array. To address this, our PICs are specifically designed with mirrored coupler placements and SSC tapers aligned to 127 µm spacing, and our alignment stages allow fine angular correction during bonding. Additionally, Quanfluence’s fully packaged chips support dual-side fiber array attachment, allowing simultaneous optical I/O on both facets of the chip a configuration critical for interferometric or bidirectional architectures in quantum photonics. Active alignment and low-shrinkage adhesives lock both arrays in place while maintaining sub-micron lateral and angular precision across the array.
In the packaging of photonic integrated circuits (PICs), establishing reliable electrical connections is as critical as ensuring low-loss optical coupling. Quanfluence devices integrate both DC and RF electrical connections, routed through bond pads distributed along the chip periphery. Care must be taken during fiber array attachment to prevent any encroachment of UV-curable adhesive (or index-matching encapsulants) onto exposed electrical pads. The physical footprint of standard fiber arrays often 127 µm pitch for 8- or 16-channel configurations necessitates maintaining a safe clearance zone between optical and electrical interfaces. This design constraint imposes a trade-off between fiber density and bond pad density.
To accommodate larger interferometric photonic circuits, such as 20×20 unitary meshes required for certain quantum gate operations, Quanfluence employs staggered multi-row wire bonding strategies (2 or even 3 rows). Though less common, this layout allows for higher I/O density without compromising optical attachment or risking short circuits. In continuous-variable quantum photonics, such high-density unitary networks are used to generate cluster states and implement Gaussian operations. Thus, this integration strategy is essential for scaling functional devices.
References
[1] Bogaerts, W., et al. “Silicon photonic circuit design and performance: The role of spot-size converters.” IEEE Journal of Selected Topics in Quantum Electronics (2010).
[2] Yokoyama, S., et al. “Ultra-large-scale continuous-variable cluster states multiplexed in the time domain.” Nature Photonics 7, 982–986 (2013).
[3] Menicucci, N. C., et al. “Universal quantum computation with continuous-variable cluster states.” Physical Review Letters 97.11 (2006): 110501.
[4] Seok, T. J., et al. “Wafer-scale silicon photonic edge couplers fabricated using standard CMOS processing.” Optics Express 23.5 (2015): 6284-6296.
[5] Aghaee Rad, H., Ainsworth, T., Alexander, R.N. et al. “Scaling and networking a modular photonic quantum computer.” Nature 638, 912–919 (2025).
[6] PsiQuantum Team. “A manufacturable platform for photonic quantum computing.” Nature (2025). https://doi.org/10.1038/s41586-025-08820-7