Co-optimization Strategy Breaks Simulation Approximation Pitfall, Overcomes Coupling Efficiency Bottleneck for Silicon Nitride Photonic Chips

Research Background: The Overlooked "Fiber End"

Integrated photonic chips, leveraging scalable CMOS manufacturing, have become a core technological platform for data centers, optical interconnects, and quantum information processing. However, efficiently and reliably transferring light from optical fibers into on-chip waveguides remains a critical bottleneck limiting system performance.

Currently, silicon nitride (Si₃N₄) is garnering significant attention due to its ultra-low loss, ultra-wide transparency window, and moderate nonlinearity. To achieve light coupling, the industry commonly employs lensed fibers interfacing with inverse taper structures on the chip. However, in past designs, researchers have often prioritized the "chip" over the "fiber" :

• Traditional Pitfall: Most studies focused solely on optimizing the design of the on-chip inverse taper, neglecting the optimization of the lensed fiber itself.
• Simulation Shortcoming: To simplify calculations, traditional simulations typically approximated the focusing mode of the lensed fiber as a "paraxial Gaussian beam".
• Practical Consequence: This approximation ignores the subwavelength structure and strong evanescent field effects at the fiber tip, leading to a long-standing discrepancy between simulation results and experimental data, thereby limiting improvements in coupling efficiency.

Research Innovations: A "Real" Model 

To address this gap, the research team proposed a novel co-optimization strategy, achieving a comprehensive breakthrough from theory to fabrication.

1. Eschewing Approximations, Precise Reconstruction: The team abandoned the idealized Gaussian model. Instead, they used scanning electron microscopy (SEM) to image the tips of lensed fibers from different vendors. By extracting the real geometric profiles and importing them into 3D finite-difference time-domain (FDTD) simulations, they accurately captured the non-Gaussian emission profiles, bringing the simulation environment as close as possible to the real physical world.
2. Comprehensive Process Coverage: The team systematically characterized lensed fibers with mode-field diameters (MFDs) ranging from 2.0 μm to 6.0 μm, revealing discrepancies between the actual MFDs and vendor-labeled values. They conducted matching tests using silicon nitride inverse tapers fabricated via both subtractive and additive processes.
3. Exceptional Experiment-Simulation Agreement: Thanks to the accurate modeling, experimental results showed maximum single-facet coupling efficiencies exceeding 80%. More importantly, the experimental data exhibited excellent agreement with simulation results in both trend and magnitude, correcting the common flaw in prior studies of matching trends but not absolute values.

Figure: Schematic diagram of light transmission in the lens fiber and inverted taper coupling structure on the photonic chip.

 ​Key Takeaways: A "Troubleshooting Guide" for Engineers 

This research not only corrects traditional misconceptions but also provides practical design guidelines for the industrial-scale design, testing, and packaging of photonic chips:

• Optimal Partnership: The study found that lensed fibers with an MFD of 4.0 μm are typically the optimal choice, achieving the highest coupling efficiency for both TE and TM polarizations simultaneously.
• Enhanced Tolerance: Larger MFDs significantly improve alignment tolerance, which is crucial for enhancing packaging stability.
• Spectral Performance: Matching appropriately large-MFD fibers with small-dimension inverse tapers can effectively suppress facet reflections. The research achieved a coupling efficiency variation of only 3% over an ultra-broad spectral range from 1480 nm to 1640 nm, effectively supporting broadband nonlinear optical applications.

​​Summary and Outlook 

This research establishes a set of design rules for silicon nitride inverse taper tips and lensed fibers compatible with modern CMOS foundries. This not only provides a scalable solution for the effective manufacturing and optoelectronic co-packaging of ultra-low-loss silicon nitride photonic chips but also holds significant implications for neural network optical computing, optical quantum computing, and next-generation optical interconnect architectures of AI data center, accelerating the transition of integrated photonic chip technology from the laboratory to commercialization.

This work received high praise from the reviewers and editorial board of Physical Review Applied and was selected as an "Editors' Suggestion", fully demonstrating its broad influence in the field of photonics applications.

Publication Information 

Dengke Chen, a joint Ph.D. student from the Shenzhen International Quantum Academy and Southern University of Science and Technology, is the first author of the paper. Researcher Junqiu Liu is the corresponding author. The research was supported by projects from the Shenzhen Municipality, Guangdong Province, the National Natural Science Foundation of China, the National Major Project on Quantum Science and Technology, and the National Key R&D Program of China.

Dengke Chen et al. Light coupling to photonic integrated circuits using optimized lensed fibers, Physical Review Applied 25, 014078 (2026). Original link:https://doi.org/10.1103/th1c-nml5