The road to a large quantum computer runs through a problem that sounds mundane and is brutal: how do you control many qubits at once without the accuracy collapsing?
A team led by the RIKEN Center for Emergent Matter Science has put a concrete step on that road, in Nature Physics. The researchers ran five silicon spin qubits off a single shared microwave control line and kept fidelity high even with all five active. Getting there required finding an error first.
The error in the microwaves
Silicon spin qubits store information in the spin of individual electrons confined in silicon. That is attractive because such structures can be built with the same semiconductor manufacturing used for ordinary chips. Control comes from microwave signals — and that is where the problem lives. When several qubits are driven at once, the signals shift the effective energy levels of their neighbours. This AC Stark shift changes the phase of a qubit's quantum state and disrupts the intended operation.
The obvious fix is to calibrate every possible combination of qubits separately. That does not scale: with ten qubits it is more than a thousand combinations. The team took a different route and calibrated pairwise — between pairs of qubits rather than all combinations — and combined that with an improved device design and carefully adjusted microwave pulses.
The numbers
Tested individually, each qubit reached single-qubit gate fidelity well above 99.99 percent, with the best approaching 99.999 percent. With three qubits active at once, fidelity stayed above 99.99 percent. With all five driven in parallel, it held at roughly 99.9 percent — the order of magnitude practical fault-tolerant systems need.
"This shows that high-fidelity control can be maintained even when several qubits are operated at the same time and represents an important step toward scaling up silicon-based quantum processors," said Leon Camenzind, a RIKEN research scientist and co-first author who supervised the experiments.
What is left over
The errors did not vanish. The remaining loss traces to drive-induced decoherence: the higher microwave power needed to run several qubits at once destroys some of the delicate coherence the operation depends on. Camenzind calls it a limitation that better device design could mitigate. Dielectric loss and fabrication tolerance remain the real opponents.
And a distinction that gets lost in headlines: what was measured is single-qubit gates, executed in parallel. That is the metric for crosstalk and control quality, not for entanglement. Two-qubit gates are the harder problem, and for error correction their error rate is what counts. A result about parallel single-qubit control says nothing about whether the coupling between neighbouring qubits is clean enough.
Why the shared line matters
Give every qubit its own control line and the wiring becomes the limiting factor — at thousands of qubits it stops being a layout problem and becomes a cooling one, since every line carries heat into a millikelvin environment. Shared lines are the practical way out, and using them means mastering how qubits on the same line interfere with each other. That is what this paper demonstrates, on five qubits.
Five is not five hundred. But the finding is structural: parallel control does not have to cost a large share of your accuracy, if you compensate for the AC Stark shift instead of ignoring it. Seigo Tarucha, who leads a silicon quantum computing project under Japan's JST Moonshot Goal 6, states the goal plainly: a fault-tolerant quantum computer built on silicon spin qubits, for which reliable parallel control is an essential precondition.
