A comb that fits the atom pitch
Takashi Yamamoto's group at the University of Osaka, working with the National Institute of Information and Communications Technology and Hamamatsu Photonics, lined up 10 neutral atoms at micrometre spacing — the pitch a working atom-array quantum computer already uses. Photons from each atom were coupled into 10 neighbouring guides of a 32-channel integrated waveguide array formed in a glass plate, then carried through 10 optical fibres into a 10-channel superconducting nanostrip photon detector. The detector stack itself is a 32-channel system the NICT and Hamamatsu Photonics teams built for this experiment. Yuya Maeda is the first author of the peer-reviewed Optica paper that reports the interface, published on 31 August 2026.[1]
Takashi Yamamoto's announcement treats one-at-a-time handling as the bottleneck for linking many atomic qubits into a network. Earlier multiplexing sat in discrete optical fibres run in parallel: only a few channels, and an atom spacing too wide for the micrometre array a neutral-atom machine uses. The dense 32-channel waveguide array is the part that can sit at that native pitch; a loose fibre bundle cannot. The construction that lets 10 parallel paths share the same array is this integrated waveguide. Taping a handful of fibres side by side leaves the atom spacing too wide.[1]
What arrived at the detectors
The quantum state of each atomic qubit stands in an entanglement relation with the polarisation of its photon. The Osaka announcement says the experiment confirmed part of that relation, and the university summary states that the expected correlations between each atom's quantum state and its photon's polarisation were observed. Optica published the peer-reviewed account on 31 August 2026. Parallel photon collection with those state correlations preserved is what the experiment demonstrates. Each check sits on one atom and the photon collected from that same site.[1]
The piece that has to carry the architecture is the match between waveguide pitch and atom pitch: once the guides sit at the same few micrometres as the atoms, 10 channels can be filled without spreading the array. An equally plausible reading is that the newly built 32-channel superconducting detector is what makes 10-way collection look clean, and that a fibre bundle aimed at the same 32-channel detector would have looked similar. The announcement does not separate those two contributions with a baseline that holds the detector fixed.[1]
100 channels, still on paper
Takashi Yamamoto's group says the waveguide approach should scale to roughly 100 parallel channels. The same announcement estimates that a single neutral-atom machine can hold about 10,000 atoms in a 100 by 100 array, and that a fault-tolerant general-purpose machine is thought to need more than 1,000,000 qubits. Several such processors would then have to be linked. Those figures describe the motive for a multiplexed optical interface; they are not measurements from this 10-atom experiment.[1]
The experiment demonstrates parallel photon collection with the state correlations preserved. Entanglement between remote atoms, error rates and repetition rates have not yet been shown at this scale. The architecture text describes a later step in which a photon carries a qubit to another machine by quantum teleportation; that step was not part of the 10-channel experiment. The correlation that was checked sits between an atom and its own photon. A network link still has to share entanglement between two distant atoms.[1]
If Takashi Yamamoto's group or another laboratory publishes entanglement between remote atoms through this waveguide-array interface by the end of 2027, the 10-channel collection becomes a link that can be priced in error rate and repetition rate. Until that paper appears, the inspectable object remains 10 atoms, 10 of 32 waveguide channels, and a polarisation correlation on the bench.[1]