Ultrathin lensless endoscopes built from multicore fibers form a focus by co-phasing the beams from many widely spaced single-mode cores, but because the cores fill only a small part of the fiber face, even perfect co-phasing leaves most light in a speckle halo (a focal-spot power ratio of about 0.01 reported for earlier bare-fiber systems). The authors design and 3D print, by two-photon lithography directly on a 180 µm, 120-core Fermat-spiral multicore fiber, a miniature 'hyper-telescope': one microlens per core expands and collimates each beamlet to fill the pupil, and a single top lens refo

Optics Letters · 3D-printed coherent beam combiner

A hair-thin multicore fiber can only focus light by making 120 separate beamlets arrive in step. Line them up, see where the light really goes, then print the lenses that let the spot keep more of it.

of all light leaving the 120 cores

Simulated scan of illustrative 2 µm beads at the same laser power. Two-photon signal grows with the square of the power in the spot.

1 Co-phase

2 Look closer

Log reveals the faint halo.

3 Rearrange

Same cores, same footprint.

4 Fill the pupil

One lenslet per core + one top lens.

5 Steer

Or drag the spot in the focal plane.

The bench runs an idealised scalar model after Eq. (2) of the paper (perfect phasing, equal core powers, one polarization, ideal thin lenses; 120 cores, 3.6 µm mode diameter, 11.8 µm mean spacing, 39 µm lenslets, 800 µm top lens, 1030 nm). Its percentages illustrate trends; they are not the authors' non-paraxial design calculation.

The principle

The spot is a sum of 120 arrows

Each core of the fiber carries its own tiny beam, and each beam reaches the target with its own phase : where its wave is in its up-and-down cycle. The field at any point ahead of the fiber is the sum of 120 such contributions, which you can draw as arrows laid head to tail. The spot's brightness is the squared length of the chain.

Light travelling through a fiber picks up a different, scrambled phase in every core, so the arrows point every which way and mostly cancel: speckle. In the experiment, a spatial light modulator at the input end first measures each core's phase (by phase-stepping interferometry), then applies the opposite phase. Now every arrow points the same way at the chosen point, and a focus appears. Adding a gentle phase ramp moves that meeting point sideways, so the endoscope scans its spot with no moving parts at the tip .

But phasing only decides where the beams meet, not how much of their light can get there. The fiber face is mostly dark glass dotted with small bright cores. Such a broken-up field contains fine detail, and fine detail sends light out at wide angles, into a halo around the spot. Even perfectly phased, most of the light misses.

Eq. (1): the field on the fiber face is one core's mode shape A copied onto each of the N core positions r n. Eq. (2): far from the fiber, the intensity is the broad glow of a single core, |A( k )|², times the array sum. The focus is the k = 0 term, where all exponentials equal 1 and the sum reaches N once the cores are co-phased. Everywhere else the sum leaves a speckle halo, spread across the whole width of the single-core glow.

The idea for the cure comes from astronomy's hyper-telescope (combining light from arrays of telescopes) and from tiled-aperture laser beam combining: make each beamlet wider before refocusing, so that together they fill the gaps. In the paper this raises the fill factor by wcoll/wfiber, the ratio of beam width after each lenslet to beam width at the core.

The limits

Four things co-phasing cannot fix on its own

Perfect phase is not enough

With bare, widely spaced cores, a perfectly phased spot still holds a small share of the light; the rest fills the halo.

Paper: η ≈ 0.01 for earlier bare multicore fibers (reported).

Rearranging trades satellites for speckle

A periodic grid throws bright copies of the spot (satellite peaks). The aperiodic Fermat spiral spreads that light thinly: better contrast, but no more light in the spot.

Each obvious fix breaks something else

More laser power : self-phase modulation and pulse broadening, which counteract the signal.

More cores : a drastically larger device.

Cores closer together : cross-talk and mode coupling, hurting bending insensitivity, flexibility, field of view, fast scanning and calibration.

Wider beamlets, smaller field

A wider beamlet casts a narrower glow, so the spot can only be steered within a smaller field. Widened too far, beams are clipped by the lenslet edges (vignetting).

Design chosen as a trade-off: lenslets f = 39 µm, top lens F = 800 µm → ideal η 0.5, spot 4.8 µm, field ≈ 60 µm.

The device

A telescope printed on the tip of a fiber

The authors printed the combiner directly onto a custom 180 µm, 120-core fiber whose cores follow an aperiodic Fermat golden spiral (mode-field diameter 3.6 µm, average spacing 11.8 µm). Each core gets its own lenslet (focal length 39 µm, radius of curvature 14 µm), which widens and straightens its beam. A single top lens 100 µm above the array (focal length 800 µm, radius of curvature 400 µm) brings all 120 beamlets together. The whole optic is a cylinder 180 µm across and 200 µm tall.

They used two-photon lithography (a Nanoscribe Photonic Professional GT with IP-S resist and a 25× 0.8 NA objective). A camera and a computer-vision routine find each core and print one lenslet at a time on top of it. Instead of slicing each lens into flat layers, which can leave staircase steps or overexpose nearly flat regions, they wrote every lens as a single layer of concentric circles placed on different planes.

Not to scale; four of the 120 beamlets shown. Redrawn after Fig. 1b of the paper.

What was measured

From about 1 % to 35 % of the light in the focus

Source

Miniature 120-beam coherent combiner with 3D printed optics for multicore fiber based endoscopy

Siddharth Sivankutty, Andrea Bertoncini, Victor Tsvirkun, Naveen Gajendra Kumar, Gaelle Brévalle, Géraud Bouwmans, Esben Ravn Andresen, Carlo Liberale, Hervé Rigneault. Preprint arXiv:2203.01636v1 (2022); published in Optics Letters 46(19), 4968 (2021). doi:10.1364/OL.435063 · arXiv:2203.01636

Figures 1a, 2a, 2b, 3b, 3d and 4 are reproduced from the arXiv preprint, © the authors, under CC BY-SA 4.0; they were cropped and resized (Fig. 2 converted to greyscale), and the adapted figures are shared under the same license. The side-view schematic is redrawn after Fig. 1b. The bench is an idealised scalar wave-optics model after Eq. (2) of the paper; its core layout, mode size, spacing and lens focal lengths follow the paper, while the lenslet clear aperture (5.9 µm radius) and the bead sample are illustrative choices. All numbers labelled “paper”, “measured” or “design” are quoted from the paper.