QCNEWS
In a 31,000 atom quantum gas, a Tsinghua led team has detected a quantum correlation no collection of two state systems, or qubits, could match, the first such result in the many body regime.
About 31,000 rubidium-87 atoms, cooled until they share a single quantum state, have produced a form of correlation that no equivalent group of two-state systems can match. The result, reported this month by a Tsinghua-led team on arXiv, is the first many-body platform to push Bell-nonlocality certification past the limit that collections of two-level particles cannot cross, into territory that demands a three-level quantum description.
The test traces to a 1935 argument by Einstein and two collaborators that quantum mechanics could not be complete, because particles separated in space seemed to coordinate in ways the theory could not explain without "spooky action at a distance." Decades of single-atom and single-photon experiments have since closed the loopholes that bothered Einstein, certifying the correlations are real. The new experiment extends the same logic to a different setting: a cloud of atoms acting as one entity, measured as one entity, has been shown to carry a correlation that an ensemble of the same number of two-state atoms could never reach.
The metric that captures the cross-over is a Bell witness. It is a number built from a set of joint measurements: a value below 2 means the system cannot be described by any local classical model, while a value below 1 means it cannot even be described by local collections of two-level quantum systems, the building blocks known as qubits. The team, which includes researchers at Tsinghua, Nanjing, Jagiellonian, ETH Zürich, and the Beijing Academy of Quantum Information Sciences, drove the witness below 1 in a spin-1 Bose-Einstein condensate, a quantum gas whose internal spin can point in more than two directions. The strongest violation, at a measurement angle of 0.79 radians, cannot be reproduced by any number of qubits, only by a qutrit, a three-state quantum switch.
Reaching that threshold required two ingredients. The first is a special kind of quantum noise reduction known as spin-nematic squeezing, in this case 11.8 decibels of squeezing, generated by letting the atoms collide and exchange spin. Squeezing alone is not enough. It is what lets the team resolve witness values that would otherwise drown in noise. The second is the choice of measurement. Instead of looking at atoms one at a time, the team built probabilities from measurements that look at the whole cloud at once. The resulting Bell witness is permutationally invariant, meaning it does not depend on which atom is which, and it is built from collective one- and two-body observables rather than atom-resolved detection. The full preprint reports the witness crossing the qubit bound after roughly 50 shots per data point, with cross-checks against truncated Wigner simulations.
The result is not a loophole-free Bell test in the photonic sense. In a Bose-Einstein condensate, atoms sit micrometers apart in a single trap, and there is no analogue of the spatial separation or random setting choice that earlier single-atom and single-photon experiments used to rule out classical explanations. The experiment certifies a many-body, coarse-grained counterpart: the same correlation structure, observed in a macroscopic quantum gas, with measurements any lab with a similar apparatus could in principle repeat.
The source identifies the next step the team intends to chase. The coarse-grained measurement recipe is portable, and the authors point to other macroscopic platforms, including trapped ions and cavity QED systems, where the same witness could in principle be evaluated. If the recipe holds up, Bell correlations that require higher-dimensional local Hilbert spaces could be catalogued in systems large enough to sit on an optical table, not just in the single-atom and single-photon experiments that have dominated the field.