Skip to content
PaperFren

Can a chip untangle two overlapping light beams of the same colour?

Open paper intelligence

A self-tuning optical chip split two overlapping laser beams of identical wavelength and polarisation into separate outputs with almost no leakage, even after the beams had been scrambled.

Source

Separating arbitrary free-space beams with an integrated photonic processor

Milanizadeh M, SeyedinNavadeh S, Zanetto F, et al. · Light, science & applications · 2022

doi.org/10.1038/s41377-022-00884-8Read the full paper ↗18 citationscc by

Study at a glance

Design
Other — Optical bench experiment on a silicon photonic chip (3x3 grating-coupler array feeding two rows of Mach-Zehnder interferometers) receiving pairs of free-space beams at 1550 nm.
N
No sample size; one fabricated chip tested with several beam pairs (direction-diverse, mode-diverse, and mode-mixed).
Population
A single programmable silicon photonic processor with 9 optical antennas and 15 interferometers
Outcome
Crosstalk suppression between separated beams, eye diagrams and bit error rate versus optical signal-to-noise ratio for 10 Gbit/s data channels

Structured fields used in claim comparison tables when every cited study has a complete layer.

What they did

The authors built a silicon chip in which 9 tiny antennas sample an incoming light beam and feed a mesh of 15 tunable interferometers arranged in two rows. Each interferometer is tuned by a local feedback loop that simply minimises light at a monitor detector, so the chip configures itself without knowing the beam shapes. They sent in pairs of beams that differed only in arrival direction, only in spatial mode shape (Hermite-Gaussian modes), or that had been mixed by a phase mask, and each beam carried a 10 Gbit/s data signal.

What they found

Beams arriving 1.25 degrees apart were separated with more than 25 dB crosstalk suppression, and beams differing only in mode shape with more than 30 dB. Data channels showed open eye diagrams and no measurable signal-to-noise penalty in bit error rate compared with each beam sent alone. Mode rejection stayed above 20 dB over a 35 nm wavelength range, and even arbitrarily mixed beams with unfamiliar shapes were separated.

The limits

What it doesn't show

Only two beams were separated at once, using one chip with just 9 antennas, so scaling to many channels is argued rather than demonstrated. The end-to-end loss was about 28 dB, largely from geometric coupling into the sparse antenna array, which would matter in a real link. The mixing tested was a static lab phase mask, not real time-varying atmospheric turbulence, and separation only works when the sampled beams remain orthogonal on the antenna array.

Key terms

Mach-Zehnder interferometer (MZI)
A device that splits light into two paths and recombines it; tuning the phase between paths sets how much light goes to each output.
Orthogonal beams
Beams whose sampled field patterns have zero overlap, so a linear optical system can route them to different outputs without loss.
Crosstalk suppression
How much weaker the unwanted beam is than the wanted one at an output port, usually given in decibels.
Space-division multiplexing
Sending several independent data channels at once by using different spatial paths, directions or mode shapes of light.
Hermite-Gaussian modes
A family of laser beam shapes with a set number of intensity lobes, used as a standard orthogonal basis.

Flashcards

1 / 11

0 of 11 answers reviewed

Research intelligence for this paper

See its role on concept claims, tensions it is part of, placement history, and related discoveries.

Open paper intelligence

Quiz yourself

1 / 5

What distinguished the two beams in the mode-diversity experiment?

Common questions

Why can't you just use a colour filter to separate the beams?

Both beams had the same wavelength and polarisation, so only their spatial pattern or direction distinguished them; the chip exploits that spatial difference.

Does the chip need to know what shape the beams are?

No. Each interferometer is tuned by minimising power at a detector, so the mesh finds the right settings automatically.

What limits how many beams it can separate?

A processor with M antennas and N rows can separate up to N beams from a set of M orthogonal ones, so more antennas and rows are needed for more channels.

More on Structured light