One Extra Measurement Cuts Qubit Errors by More Than Half in New Regensburg Experiment

Researchers at the University of Regensburg and international partners say a single extra measurement before a series of quantum operations can cut a qubit's error rate by more than half. The experiment, published in Applied Physics Reviews, used a superconducting circuit and real-time feedback to correct for a problem that quietly degrades quantum computers.

Quantum computers are widely seen as a path to problems beyond the reach of conventional machines, but their building blocks, qubits, are extremely sensitive to disturbance. A qubit plays a role like a bit in an ordinary computer, but it follows quantum mechanics rather than being fixed at 0 or 1. Even tiny changes in the environment can alter a qubit's properties and introduce errors.

The team studied a superconducting qubit whose resonance frequency would unpredictably switch between two known values. The resonance frequency determines which microwave signals can control the qubit precisely. If it shifts, the control signals no longer match. The researchers compared it to a musical instrument that has drifted out of tune, where the same movements stop producing the intended notes. Importantly, the two frequencies do not correspond to different computational states. They are two properties of the same device, and the qubit can still occupy the quantum states needed for computation at either one.

The jumps are caused by interactions between the qubit and tiny defects in the material. Scientists knew the two possible frequencies but could not easily tell which was active at a given moment, or adjust for it. The new method does that. Before running a sequence of operations, the team prepares the qubit in a quantum state and lets it evolve briefly. How it evolves depends on which frequency is present, and a final measurement reveals which one it is. The result is immediately used to adjust the control signals that follow, a process known as feedback.

The experiments ran on a superconducting quantum circuit at the Israeli Quantum Computing Center, with measurement data processed in real time so control signals could be updated immediately. Without adjustment, the quality of the operations fluctuated significantly over time. With feedback on, the qubit performed consistently with high fidelity. The average error rate fell from about 0.14 percent to 0.065 percent, a reduction of roughly 54 percent. Compared with the worst error seen without feedback, the reduction reached 77 percent.

The method has limits. The possible frequencies must be known ahead of time and be sufficiently distinct from one another, and the frequency must stay stable long enough for the updated signals to take effect. Ferdinand Kuemmeth, a professor of experimental physics at Regensburg, acknowledged that the work might look at first like a proof of principle with limited practical value.

He argued the opposite over the long run. As quantum materials get cleaner, millions of nearly perfect qubits will have to be operated at once, and some will likely still interact with a few remaining defects. "I am convinced that in that emerging era of quantum computing scaling, our methods will become very valuable and make a tangible impact on the actual computational power of quantum computers," Kuemmeth said. He added that the team has already developed methods to automatically and efficiently detect unknown qubit frequencies.