On September 4, 2026, a terse four-page preprint appeared on arXiv: *A photonic source with half-a-GHz single-photon flux*. Three days later, a company press release followed, and the quantum community took notice: for the first time, a deterministic single-photon source has pushed more than half a billion photons per second—stably—into a single-mode optical fiber. The best previously reported figure was 76 million per second.
What this device is for
Photonic quantum computing, quantum key distribution (QKD), and multi-photon interference experiments all share one fundamental component: a machine that emits photons one at a time. Literally one at a time.
Ordinary light sources emit light "in bulk": a laser pointer sprays trillions of photons per second, with countless photons packed into each pulse. Quantum protocols demand singletons—exactly one photon per pulse, because multi-photon contamination corrupts the mathematics of quantum interference.
The field has spent decades chasing one number: how many "clean" single photons can be produced per second. Until that number rises, multi-photon experiments never have enough raw material. The frequently cited goal is ten- to twenty-photon interference experiments—each added channel raises the difficulty another notch, and every channel needs feeding.
Fast vs. efficient: a thirty-year trade-off
The record-setting approach uses a quantum dot: a nanoscale "artificial atom" grown in gallium arsenide, which emits one photon each time it is excited. The challenge has always been balancing two metrics:
- Fast: how densely you can pulse. This dot needs 220 picoseconds to recover after emitting a photon—theory allows up to ~4 billion pulses per second.
- Efficient: how many of those photons actually get captured into the fiber. Photons fly out in all directions from a nanoscale dot; capturing even half is expert-level work.
Rate times efficiency equals usable photon count. For thirty years, the two factors never stood up at the same time.
What the Danish team actually did
Sparrow Quantum, spun out of the Niels Bohr Institute in 2016, was founded by Peter Lodahl, professor of quantum optics and the company's Chief Quantum Officer. The paper has 13 authors: 11 from the company, 2 from Ruhr University Bochum, which supplied the epitaxial material—a twenty-year collaboration. First author Patrik Zahálka is an industrial PhD at the company. This is not a campus paper; it is a chip company showing its chops.
The technology stacks three layers of engineering:
1. Excitation. Two cascaded electro-optic modulators carve a continuous laser into pulses—40 ps wide, extinction ratio above 40 dB, FPGA-programmable clock rates up to 1 GHz. Each ultrashort pulse kicks the quantum dot exactly once, producing one photon.
2. Collection. The quantum dot sits inside a photonic-crystal waveguide. This structure, etched with periodic holes, acts as a funnel, directing nearly all emission into the waveguide mode and on into single-mode fiber. At the full 1 GHz rate, measured fiber-coupling efficiency is 51.5%—read directly off an ordinary commercial power meter.
3. The paper's cleverest trick: no spectral filtering. Filters usually purify the output at the cost of discarding photons. The team did the opposite—collect everything and report a lower bound, even counting phonon-sideband light in the denominator. The measured fiber power was 110 picowatts; dividing by the known photon energy yields the efficiency directly, bypassing the entire calibration chain of superconducting single-photon detectors.
For perspective: 110 picowatts is about two billion times less power than a button cell battery—invisible to eyes and cameras alike. But in the single-photon world, this brightness moves the source from "countable" to "measurable"—previously you needed expensive single-photon counters; now you can simply measure. As Lodahl put it in the press release: "Every bit of improvement comes from solid engineering, not redefining the technology. We drive the source to the limit the emitter physics allows, capture every photon it produces, and the quality holds."
Reading the fine print at the record point
"The quality holds" is true at 500 MHz—but at the record-setting 1 GHz full rate, it needs an asterisk. The paper's own data is honest:
| Clock rate | Multi-photon contamination (g2) | |---|---| | 80 MHz | 3.98% | | 500 MHz | 3.63% | | 1 GHz (full rate) | 11.13% |
Photon indistinguishability also drops from 82% to 55% at 1 GHz. The physics is clear: the dot needs 220 ps to recover, and at 1 GHz consecutive photon wavepackets overlap. The paper's comfort criterion is a pulse spacing of ten times the lifetime—below roughly 450 MHz for this dot. Running a truly clean 1 GHz will require a dot with a shorter lifetime.
The fair summary: quantity and quality are simultaneously achieved at the 500 MHz setting—about 269 million photons per second in fiber with >96% purity. Even at this setting alone, that's three and a half times the old record. At full speed, the flux reaches 515 million per second.
What half a billion photons per second can do
The paper claims two applications: serving as an optical power standard for metrology, and time-multiplexing into many channels to feed ten- to twenty-photon quantum simulation experiments. The arithmetic on the second is vivid: split a 515-million-per-second conveyor into ten channels and each gets roughly 50 million per second—each channel approaching two-thirds of the best single source in the world as of 2024. A day's worth of data acquisition now takes under an hour.
The press release sketches a broader horizon—linear optical quantum computing, detector calibration, quantum networks—but those are company vision, not claims in the paper. And one thing must be stated plainly: this is a preprint posted September 4, not yet peer-reviewed—the company's own release dutifully labels it "the preprint."
A small company, a small mountain
Zooming out: Sparrow Quantum spun out of the Bohr Institute in 2016, raised €4.1 million in seed funding in 2023, and €21.5 million in a 2025 Series A (expanded to €27.5 million by year-end), led by a Danish pension fund. The company lists 55+ staff and sells its Sparrow Core single-photon-source chip and Sparrow Nest integration platform openly. The new record is twenty-five times the company's own 2023 product specification—the chip-company rhythm of welding paper results into product catalogs year after year.
Lodahl himself named the next mountain in the release: entangling multiple such sources to work together, "with the same level of control and reliability." The single-photon source is only the first building block of photonic quantum computing; with a half-billion-per-second supply line laid, the bottleneck shifts to entanglement distribution and multiplexed integration—a slower fight.
The meaning of a record was never the record itself. It is that the boundary of "impossible" gets pushed one notch outward—and then everyone can see the new terrain.
---
Sources: arXiv:2609.05387, *A photonic source with half-a-GHz single-photon flux* (2026-09-04, v1 preprint, 13 authors: Sparrow Quantum ×11, Ruhr University Bochum ×2); Sparrow Quantum press release *Enough light, at last* (2026-09-07); The Quantum Insider coverage (2026-09-07); Sparrow Quantum website team and product pages.