Programmable quantum photonic chip shows two-photon interference in orbit
A small satellite payload proved quantum light can be made, steered and measured in space, and its failures show what space-ready quantum hardware needs.
Researchers from the University of Vienna, the German Aerospace Center (DLR) and Italy's National Research Council (CNR) have operated a programmable quantum photonic processor in orbit and observed two-photon quantum interference on it, according to a report by The Quantum Insider on 29 September 2026. The results are in a preprint on arXiv and have not yet been peer reviewed. Simon Steiner and Peter Schiansky are joint first authors, with Iris Agresti and Philip Walther as corresponding researchers.
The payload launched on 23 June 2025 on SpaceX's Transporter-14 rideshare and runs aboard D-Orbit's ION orbital transfer vehicle, about 317 miles (roughly 510 km) above Earth. It weighs about 22 pounds (about 10 kg), measures roughly 6 by 6 by 18 inches, and draws about 10 watts on average. The paper covers its first eight months in orbit. A laser pumps a crystal to create pairs of photons, fibres carry them into a glass chip with six light paths, tiny heaters reprogramme the circuit, and detectors count the photons that come out.
In a programmability test across nine settings, the observed detection patterns matched predictions with an average fidelity of 0.888, rising to 0.949 when two settings with calibration problems are excluded. Ground systems typically exceed 0.99. By tuning the crystal to about 32.5°C, the team made the photons indistinguishable and recorded a Hong-Ou-Mandel interference dip with a visibility of 0.908, plus or minus 0.191, which beats the classical limit by 2.14 standard deviations. The result was reproduced on two of three operating days.
The mission was hard on the hardware. Only three of six detectors worked after launch, sunlight limited measurements to about 30 minutes per 92-minute orbit, radiation steadily degraded the detectors, and laser power fell as adhesive contaminated optical parts.
Why it matters
This does not show quantum advantage, and the researchers say so. It answers a more basic question: whether a small system can create, steer and detect quantum light after the shock of launch and months of radiation. That is a precondition for any quantum processing in space, including the team's suggested first use, filtering Earth-observation data on board. It also shows that engineering, not only physics, decides how fast quantum technology leaves the laboratory.
This does not show quantum advantage, and the researchers say so.
Make photon pairs
A laser pumps a crystal, which produces pairs of photons.
Route into the chip
Optical fibres carry the photons into a glass chip with six light-carrying paths.
Programme with heat
Tiny heaters change the chip's optical properties to set how light moves through the circuit.
Detect the outputs
Detectors register photons leaving the outputs; only three of six worked after launch.
Check for interference
Tuning the crystal to about 32.5 C gave a Hong-Ou-Mandel dip with visibility 0.908.
What you can learn from this
- Photonic qubits use light as the carrier. Photonic quantum computing encodes information in single photons, for example in which path a photon takes through a circuit. Photons barely interact with their surroundings, so they hold quantum states well and do not need the extreme cooling that superconducting qubits require. The trade-off is that photons also barely interact with each other, so computation relies on interference and measurement rather than direct gates between qubits. That makes clean sources, stable circuits and good detectors the core engineering problem.
- Programmable interferometers are the circuit. The chip on this satellite guides light through waveguides that meet and split, forming a network of beam splitters. Small heaters change the refractive index of the glass nearby, which shifts the phase of the light and changes where it ends up. Setting those heaters is how the circuit is programmed, much like setting parameters in software. Temperature swings in orbit shift the phases too, which is why calibration was difficult and why the team wants automatic recalibration.
- The Hong-Ou-Mandel dip proves the light is quantum. When two identical photons arrive at a balanced beam splitter at the same moment, quantum interference makes them leave together through the same output, so detectors on the two outputs almost never fire at once. Plotting coincidences while you make the photons more or less alike produces a dip, and the depth of that dip is the visibility. Classical light cannot produce a dip deeper than 50 per cent, so a visibility of 0.908 that clears that bound by 2.14 standard deviations is evidence of genuine quantum behaviour, though with wide error bars.
- Fidelity measures how well a device obeys its design. Fidelity compares the output distribution the device produced with the one theory predicts, on a scale where 1 is a perfect match. The gap between 0.888 in orbit and above 0.99 on the ground reflects noise, drift and damaged detectors rather than faulty physics. Reporting results with and without the badly calibrated settings, as this team did, is good practice because it shows readers where the error comes from.
- Radiation and heat are the real enemies in space. Single-photon detectors are extremely sensitive, so stray sunlight and radiation damage raise dark counts, meaning clicks with no photon behind them. That noise drowns the signal, which is why measurements were confined to the Earth's shadow. The team's list of fixes, including shielding, better temperature control and annealing to repair detectors, is a map of what space-rated quantum hardware will need.
We teach this in the AI Tutor
How to use this in practice
- Simulate a Hong-Ou-Mandel dip in Python. With NumPy and Matplotlib, model two photons as Gaussian wave packets with a variable time delay and compute the coincidence probability as one half times one minus the squared overlap of the two packets. Plot it against delay, then add a second curve where the photons have slightly different central frequencies. Done is a chart showing a dip that reaches zero for identical photons and a shallower dip when you make the packets different widths.
- Build a beam-splitter network as matrices. Represent a 50:50 beam splitter as a 2x2 unitary matrix and a heater-controlled phase shifter as a diagonal matrix. Chain two splitters with a phase shifter between them to form a Mach-Zehnder interferometer, and sweep the phase from zero to two pi. Add a small random phase error to mimic temperature drift and see how quickly the output stops matching your intended setting. Done is a plot showing all the light switching from one output to the other as the phase changes.
- Run a photonic circuit in a real library. Install an open-source photonics simulator such as Perceval with pip and follow its introductory example of two photons on a beam splitter. Compare its coincidence output with your hand-built version from the first step. Done is matching numbers and a sentence explaining any difference.
- Read the preprint's limitations section. Find the paper on arXiv using the author names and read the part on detector failures and radiation. List each problem and the fix the authors propose, and mark which fixes are mechanical engineering and which need new physics. Done is a two-column table you could use to explain why space quantum hardware is an engineering challenge.
Sources
Our reporting is an original summary. Full coverage is at the links above.
How do you score in Artificial Intelligence?
A few minutes of questions, then a report on where you stand and what to learn next.