Researchers report that they created, distributed and stored a three-qubit entangled state across three separate laboratories, each holding a single atom in an optical cavity. The experiment links the atoms so that their quantum states are correlated in the Greenberger–Horne–Zeilinger (GHZ) form. The team measured a three-qubit entanglement fidelity of 77(1)% and an entanglement lifetime above 200 microseconds. They also report an overall three-node entanglement generation efficiency of 0.16% per attempt and a nominal rate of about 10 entangled three-node states per second (excluding technical limits not inherent to the protocol). The observed correlations violate Mermin’s inequality while closing the so-called detection loophole, meaning the measurements give a strong test of three-way quantum correlations without relying on assumptions about unmeasured results.

The setup uses one atom at each node and optical fibers that connect a central node to two remote nodes separated by tens of metres of fibre. The central node emits two photons in sequence; each photon is entangled with that node’s atom and is sent to one remote lab. The experiment joins the atoms into a three-way entangled state by two different, heralded methods. For one remote node the team used entanglement swapping: the photons from the two nodes are interfered and measured in a Bell-state measurement, which transfers (or “swaps”) entanglement onto the two atoms. For the other node the incoming photon is stored directly in the atom by a heralded memory: the detection of a second, “herald” photon signals that the storage succeeded. “Heralded” means there is a clear detection event that tells the experimenters when an entanglement step worked.

At a technical level, atoms generate photons that are entangled with the atomic states via a laser-driven process called vSTIRAP (vacuum-stimulated Raman adiabatic passage). The photonic qubits are encoded in polarization (right or left circular). The fidelity of the A–C link was limited by how nearly identical the two photons were when they interfered; the experiment achieved up to about 80% visibility in Hong–Ou–Mandel interference, a standard measure of photon indistinguishability. The source-to-detection efficiency for single photons was around 40% for nodes A and C, and repeated readout cycles raise the chance of getting a measurement outcome close to unity. Node B uses a different cavity design that yields a state-detection (readout) fidelity of about 99(1)%. Over a 10.5-hour run the group registered roughly 8,000 GHZ states out of 9,400 heralded events.