Six thousand qubits in a tweezer array
In the 6 November issue of Nature, a Caltech-led team reports an optical tweezer array holding more than 6,100 highly coherent atomic qubits across about 12,000 sites, with a 12.6-second coherence time, 99.99% imaging fidelity and room-temperature trapping lifetimes near 23 minutes. The platform makes thousands-to-tens-of-thousands of physical qubits a near-term prospect.
The way to build a quantum computer from neutral atoms is to trap them with light. Tiny focused beams, one per atom, hold the atoms in place like fingers pinning beads on a board, and lasers arranged so that atoms within reach of one another interact when a computation needs them to. The approach has been one of the most promising in the field, but it has shared a common limitation: arrays of a few hundred qubits, sometimes approaching a thousand, and no clear path to the thousands or tens of thousands that error correction will require. A paper in the 6 November issue of Nature reports a decisive step across that line.
The team, led by researchers at Caltech, built an array of optical tweezers trapping more than 6,100 neutral atoms in roughly 12,000 sites. The raw size is impressive, but the paper’s case rests on the numbers that matter for computing rather than on the count alone. The atoms are highly coherent: the team measured a coherence time of 12.6 seconds, a record for hyperfine qubits in an optical tweezer array, and long enough that the qubits’ states survive the time scales on which real algorithms operate. The imaging, the step that reads whether an atom is in one state or another, achieves a survival rate of 99.98952 per cent per image with a fidelity above 99.99 per cent, and the traps hold the atoms for about 23 minutes at room temperature. In every metric that determines whether a platform can become a computer, the array performs at or beyond the state of the art.
Size and coherence only matter if the machine can do something with them, and the paper addresses the missing piece explicitly: moving atoms around without destroying their states. The team demonstrates coherence-preserving transport across large spatial scales, picking atoms up from one site and dropping them into another, using an interleaved benchmarking method to verify that the operations do not corrupt the qubits. Transport is the operation that will let a large array be reorganised into working zones, a plan the authors sketch as the route to a computer built from thousands of physical qubits rather than one enormous assembly.
The significance is in the arithmetic of error correction. Quantum error correction trades many imperfect physical qubits for a few very reliable logical ones, and the overhead has always looked forbidding with arrays in the hundreds. A platform that starts from six thousand coherent qubits changes the budget: the ratio of physical to logical qubits is no longer the obstacle it was, and the path from laboratory demonstrations to fault-tolerant machines becomes a scaling question rather than a fundamental one. The paper’s own summary is direct: universal quantum computing and error correction with thousands to tens of thousands of physical qubits is a near-term prospect.
The result is a record, and records in this field tend to be followed by refinements rather than revolutions. The array is not yet a computer, and the 12,000 sites with 6,100 atoms still carry imperfections that must be understood. But the trajectory is now visible, and it is a different trajectory from the one the field was on a year ago. The numbers that will decide when quantum computers arrive are not only gate fidelities anymore. They are also counts, and the count just moved into the thousands.