Researchers Keep Five Silicon Qubits Near 99.9% Accuracy by Taming a Hidden Phase-Shift Error

A row of five silicon qubits, all fed by a single shared microwave control line, has been run at the same time while holding accuracy near 99.9%. That result, reported in Nature Physics by researchers at the RIKEN Center for Emergent Matter Science and their collaborators, addresses a problem that has stood in the way of building bigger silicon quantum processors.

The qubits in question store information in the spin of individual electrons confined within silicon. Unlike the bits in a conventional computer, which are either 0 or 1, a qubit can exist in combinations of both states. Reliable calculation depends on extremely precise control of those spins, and microwave signals provide that control. The trouble is that driving several qubits at once can produce unwanted effects that degrade accuracy.

The team traced much of the problem to the AC Stark shift. When a microwave control signal alters a qubit's effective energy levels, the phase of its quantum state changes. Those phase shifts can disrupt the intended operation of individual qubits, and the effect matters most when several are being controlled together.

Their answer was a calibration method that compensates for the shifts. Calibrating every possible combination of qubits separately would quickly become unmanageable as devices grow. Instead, the technique relies on calibrations between pairs of qubits. Yi-Hsien Wu, the first author, who carried out the experiments during his PhD research, said the method avoids needing an entirely separate procedure for each combination.

The researchers paired the calibration with an improved device design and carefully adjusted microwave pulses. Tested one at a time, each of the five qubits achieved single-qubit gate fidelity well above 99.99%, and the best approached 99.999%. Fidelity describes how closely a quantum operation matches its intended result. With up to three qubits running simultaneously, fidelity stayed above 99.99%. With all five controlled at once, it held at roughly 99.9%, which the researchers describe as approaching the levels needed for practical fault-tolerant quantum computing.

The method did not remove errors entirely. The remaining loss of accuracy was linked to drive-induced decoherence, caused by the greater microwave power needed to operate multiple qubits at once. This differs from the phase shifts the new technique corrects: decoherence is the loss of the delicate quantum coherence that reliable operations require. Leon Camenzind, a RIKEN research scientist and co-first author who supervised the experiments, called it a limitation but said further improvements to device design could help mitigate it.

The stakes are about architecture. Giving every qubit its own independent control system would make increasingly complex devices hard to manage, whereas shared control lines could offer a more practical route to expansion. Camenzind said the study supports shared lines as a practical approach to scaling, and that, combined with silicon's compatibility with established semiconductor manufacturing, it could help pave the way toward larger processors capable of more complex computations.

The longer goal is a fault-tolerant quantum computer, one that calculates reliably despite the errors and disturbances that affect quantum systems. Seigo Tarucha, project manager of a silicon quantum computing project under JST's Moonshot Goal 6, said that goal is to build such a machine from silicon spin qubits, and that reliable parallel control is an essential step. The next thing to watch is whether device design changes can bring down drive-induced decoherence, the error that remains.