A tiny printed path. A new way to make photon pairs. Meet a quantum light source whose polymer guide can be erased and made again.

A tiny printed path. A new way to make photon pairs. Meet a quantum light source whose polymer guide can be erased and made again.

AI-generated conceptual illustration, not to scale. Light is guided mainly in the lithium niobate beneath the strip. Colors are symbolic; the separated output paths are schematic.

The width of the printed polymer strip

The full length of the waveguide

Erase the strip, preserve the crystal, print again

Small strip. Big difference.

The polymer does not do the photon splitting. It changes how light is confined in the lithium niobate underneath. That crystal supplies the nonlinear optical effect that creates pairs.

A laser writes a solid polymer strip. Its shape creates a guiding path for light in the crystal below.

A.Making the guide

Two-photon polymerization (2PP) uses a focused, pulsed laser to solidify a liquid photoresist. Here, it prints the strip without etching the lithium niobate.

B.Making the light

Later, a separate pump laser drives spontaneous parametric down-conversion (SPDC): one higher-energy photon can become two lower-energy photons inside the crystal.

More pairs. More accidental matches.

A stronger pump creates more pairs, but also more unrelated detection events that can look like a pair. Try the balance between brightness and clean coincidences.

A teaching model of the trend, not measured point data.

Power before the fiber-to-chip coupling loss.

More pump light raises the pair rate, while accidental matches become more important.

CAR compares excess coincidences with accidental coincidences. A larger value indicates a clearer pair signal.

What the original device measured Maximum on-chip rate: 4.357 million pairs/s. Maximum CAR: about 404, at 0.5 mW and 200,730 pairs/s. The two maxima are separate operating points.

What the reprint demonstrated Maximum on-chip rate: 14.569 million pairs/s. Maximum CAR: about 3,399, also at a different operating point from its rate maximum. One successful reprint does not guarantee an improvement for every device.

The original and reprinted rate models are 0.41666 × P and 1.28 × P million pairs/s, using the paper’s reported brightness values, with P in mW. The illustrative CAR models are 202/P and 1699.5/P. These inverse curves demonstrate the reported trend; they are not fits to measured point data. Normalizing each curve separately makes the trends comparable in the graph; the values beside it show their different scales. The underlying experimental data were not publicly available with the paper.

A pair that shares the energy.

A photon at 778.6 nm can split into a signal photon and an idler photon. Their energies must add up to the original. If one takes more, the other takes less.

The two photons have perpendicular polarizations. Paths and photon sizes are schematic.

Equal wavelengths mean equal energy shares: half the pump photon’s energy each.

A reusable chip is a promising start. Getting the light out is still hard.

The quoted pair rates are inferred inside the chip by correcting detected coincidences for losses. They are not delivered photon counts. Large coupling losses remain a key limitation. The study establishes correlated photon-pair generation and a successful erase-and-reprint cycle; it does not demonstrate entanglement or a complete quantum processor.

From the paper.

High-performance photon-pair source using a two-photon-polymerized strip-loaded lithium niobate waveguide

Muhamed A. Sewidan, Alexandra Rittmeier, Laura Bollmers, Silia Babel, Elisavet Chatzizyrli, Juan S. S. Duran Gomez, Laura Padberg, Christof Eigner, Christine Silberhorn, Douglas Bremner, Andreas Wienke, Dietmar Kracht, Moritz Hinkelmann, and Michael Kues.

Optics Express 34 (16), 29244–29253 (2026) · DOI: 10.1364/OE.596879

This experience is based on the supplied complete 10-page research article, including methods, results, conclusion and references. Device dimensions and fabrication details come from Sections 2–3; optical results and the reprint comparison come from Sections 4.1–4.3. Numbers in the pump explorer are explicitly illustrative models. Section 5 identifies coupling losses as a major limitation.

The source PDF carries a Creative Commons Attribution 4.0 notice. No source figures are reproduced here: the hero is an original AI-generated conceptual illustration and all interactive diagrams were created for this companion. The linked iPaper page is the companion, not the original publication.