A qubit redesign targets the noise left behind by built-in protection
New simulations extend a March experiment, outlining a possible redesign rather than demonstrating longer-lived hardware.
Reading tools for text size, copying, printing and your reading list

A Paris-led team simulated a redesign of a that is already built to limit dielectric loss, the leakage of energy into insulating materials. The proposed design would also make the qubit less sensitive to changes in the magnetic field threading its circuit. The study tested the redesign in simulations.[1][2]
The November 12 preprint revision builds on an experiment first posted in March. That earlier version already reported control and single-shot readout of a qubit whose circuit symmetry suppresses coupling to one loss channel. Its energy-relaxation time, measuring how long its excitation energy survives, was about 70 microseconds. Its echo coherence time, measuring phase-information survival after a refocusing sequence, was about 2.5 microseconds. Magnetic-flux noise remained a bottleneck.[1][2]
What is new is a simulated route beyond that bottleneck, not a new experimental lifetime. The researchers use an existing circuit model to explore larger inductances, adjusting other circuit parameters to preserve important transition frequencies and protection against dielectric loss. Their calculations identify a regime in which a trade-off suggested by a simpler model need not apply.[1][2]
Using the present device’s noise level, the proposed redesign has a predicted flux-noise-limited relaxation time of 2.7 milliseconds and an echo time of 99 microseconds. Those are modeled limits associated with flux noise, not measured total performance. A separate, more optimistic scenario assumes lower noise as well as the redesign.[1]
Other effects could intervene first. Readout interactions and other circuit losses are not fully accounted for; controlling and reading a protected qubit also remain challenges. Suppressing one failure channel does not establish protection against every error or a working error-corrected processor.[1]
The next useful hardware test would check whether the redesigned circuit retains its charge protection, then compare measured relaxation and coherence with predictions based on its actual noise. That would reveal whether the proposed improvement survives in a real device, or simply uncovers another limiting mechanism.[1]