Polarization-independent deterministic mode localization in a photonic lantern

TL;DR

This work addresses deterministic, polarization-insensitive coherent control of multimode light in photonic lanterns (PLs) without bulky adaptive optics. It introduces an all-fiber scheme using phase-only control via piezoelectric phase shifters and a reciprocal Faraday-mirror loop to coherently recombine PL outputs, achieving Gaussian-like spots at multiple MM facet locations corresponding to $LP_{01}$, $LP_{11a}$, and $LP_{11b}$ with near-unity conversion efficiency and sub-micron stability. The experimental demonstration in a three-mode PL is complemented by six-mode simulations that confirm scalability, and single-mode coupling reaches up to $55\%$ for well-localized Gaussian-like profiles. The approach promises compact, turbulence-resilient beam forming and deterministic MM routing with wide applicability to communications, biomedical imaging, and quantum photonics, while reducing system complexity relative to adaptive-optics-based solutions.

Abstract

Coherent interference in multimode photonic systems underpins scalable, high-fidelity control for beam shaping, power delivery, and signal processing, yet most existing approaches rely on bulky adaptive optics or polarization-sensitive waveguides. Here, we demonstrate an all-fiber, polarization-independent coherent mode-recombination scheme that deterministically localizes Gaussian-like spots with a Gaussian similarity index up to 0.95 at three distinct positions on the multimode facet of a commercial three-mode graded-index photonic lantern (PL). The device coherently combines the lantern's individual outputs using piezoelectric phase shifters and a reciprocal Faraday-mirror feedback loop, which enforces polarization reciprocity and eliminates alignment sensitivity. This configuration achieves near-unity (100%) relative mode-conversion efficiency, three-spot switching, and long-term stability with sub-micron centroid drift ($0.55μm$) without active feedback. The phase-locked profiles maintain high Gaussian correspondence, strong spatial confinement, and high single-mode coupling efficiency, demonstrating robustness under laboratory-scale perturbations. Numerical simulations quantitatively reproduce the experimental recombination dynamics and further establish scalability through six-mode commercial-lantern modeling. The polarization-insensitive, compact, and low-loss architecture establishes PLs as practical engines for coherent beam forming and deterministic spatial localization, enabling turbulence-resilient beam delivery, reconfigurable mode-division multiplexing, biomedical imaging and sensing, and quantum photonics, while reducing system complexity and preserving efficiency.

Figures (8)

• Figure 1: Experimental system used for the coherent mode recombination system. The red arrow denotes the forward-propagating optical signal, whereas the black arrow indicates the trajectory of the back-reflected light. Inset - PL mode profiles obtained at the MM input when selectively coupling light from the PL output fibers, (i) LP$_{01}$ mode, (ii) LP$_{11a}$ mode, (iii) LP$_{11b}$ mode. Scale bar, 0.5 mm. P – polarizer, PBS – polarizing beamsplitter, BS- beamsplitter, QWP – quarter waveplate, PLO- photonic lantern output, FDL- fiber delay line, PES- piezoelectric stack, FRM- Faraday rotating mirror.
• Figure 2: Forward and backward light propagation through the PL and path-length calibration. (a) A Gaussian input excites the supported LP modes in controlled proportions. (b) After reflection from the Faraday reflectors, the three returned modal fields recombine coherently in the MMF core, producing a distinct interference pattern at the MMF output facet based on the relative modal delays/phases (see Supplementary Material). The terms $a_{\ell m}$ and $\phi_{\ell m}$ denote the amplitude (the power weighting of each LP mode after propagation) and phase (accumulated along the path and upon reflection from the FRM) of mode $LP_{\ell m}$, respectively. (c) When the optical path-lengths of the PL arms are not equalized, the modal delays of the returned modes are mismatched, resulting in modes arriving at different axial locations along the MMF with different arrival times (t$_{1}$-t$_{3}$). (d) After path-length calibration of the PL arms, the reflected modes arrive at the same time and axial location in MMF.
• Figure 3: (a) 20 random mode interference profiles obtained at the MM end of the PL for various piezo voltage combinations. The frames are extracted from the on-the-fly video file during the voltage scan. As seen, the mode energy is concentrated in Gaussian-like profiles on various voltage combinations. The dotted circle outlines the approximate core region of the MM end. All images cover the same spatial area, scale bar, 0.5 mm. (b) Respective mode energies in the image frames, quantified by summing the pixels within the core region. See Supplementary material (Figure S7) for simulated recombined profiles and efficiencies. (c) 3D visualization of 6 random images with smoothed surface rendering and corresponding images projected on the YZ wall, capturing the diverse spatial evolution of the light field upon PE scan. The final frame exhibits a Gaussian-like intensity distribution, indicative of effective MM beam shaping and high-fidelity conversion to a desired SM spatial profile.
• Figure 4: Representative mode-interference outcomes and efficiency metrics. (a-f) Mode intensity profiles extracted from the recorded video sequence, illustrating key interference outcomes. Subfigure (a) presents the image frame with the minimum photon flux, subfigure (b) highlights a highly non-Gaussian interference state, and subfigures (c-f) present image frames with distinct instances of Gaussian-like mode interference patterns (see Supplementary material for simulated 3-spot generation). Subfigure (f) is the mode profile associated with the maximum photon flux. Scale bar, 0.5 mm. (g) Total photon counts and efficiencies of the generated mode profiles. Bar plots represent total photon counts for each of the six frames (subfigures (a–f), overlaid on the bars), calculated by integrating pixel counts within the defined MM core region. The line plot (blue) shows the energy conversion efficiencies of the mode profiles, calculated by normalizing their photon fluxes to that of subfigure (f), which exhibits the highest photon flux in the entire video.
• Figure 5: Spatial mode confinement, energy distribution, and Gaussian similarity of the mode profiles presented in Figure \ref{['fig4']}. (a) Measured intensity overlays with 1/e$^{2}$ contours showing occupied modal areas of the images within the MM fiber core. Scale bar, 0.5 mm. (b) Comparison of mode area (blue) and normalized energy density (red), highlighting confinement. (c) Major and minor 1/e$^{2}$ radii representing Gaussian similarity for each mode. (d) SSIM between measured and ideal Gaussian profiles with representative beam thumbnails.