Search will be available as the English edition grows.
Researchers have demonstrated single-qubit gates down to 6.8 nanoseconds with fidelity above 99.98% and leakage below two errors per 100,000 operations. The key is suppressing multi-photon transitions that become important under strong driving.
A quantum computer must do more than keep its qubits stable. It also needs to control them quickly and precisely. Shorter control pulses can allow more operations before noise and decoherence destroy the quantum state, but stronger and faster driving creates a problem of its own.
Superconducting transmon qubits are not perfect two-level systems. They contain additional energy levels, which means a control pulse can push part of the quantum state outside the two levels used for computation. This type of error is known as leakage.
In a peer-reviewed study in Nature Communications, researchers now show that some of the leakage produced by very fast gates comes from transitions that have often been overlooked. By shaping the control pulse to suppress these transitions as well, they demonstrate single-qubit gates as short as 6.8 nanoseconds while maintaining very low leakage.
A common technique for reducing leakage in superconducting qubits is DRAG, or Derivative Removal by Adiabatic Gate. It shapes the microwave control pulse so that it is less likely to drive unwanted transitions between energy levels.
Traditionally, much of the attention has focused on the transition from the first excited state to the level immediately above it. The researchers show, however, that when control pulses become short and strong, direct transitions to still higher levels become important. These higher-order processes can involve multiple photons and form a significant leakage channel in the strongly driven regime.
The team therefore developed a double-recursive version of DRAG that they call R2D. Rather than suppressing only the nearest unwanted transition, the pulse spectrum is shaped to suppress the identified multi-photon transitions as well.
The method was tested on a superconducting transmon qubit. For both X and X/2 gates, the researchers report fidelities above 99.98% at pulse durations as short as 6.8 nanoseconds.
At the same time, total leakage error remained consistently below 2.0 × 10^-5. That corresponds to fewer than roughly two leakage events per 100,000 gates for the measured single-qubit operation.
The figures describe different aspects of performance. Fidelity measures how closely the actual operation matches the intended quantum operation, while leakage specifically measures how often the state leaves the computational subspace. A gate can therefore still contain other errors even when its leakage rate is extremely low.
The researchers tested the gates across multiple pulse durations and used several benchmarking methods to distinguish different error channels. Their results indicate that higher-order transitions become the dominant source of leakage as the gates are pushed toward the shortest durations.
Extremely small error rates are difficult to measure directly. The researchers therefore also introduced a sequence that amplifies the effect of leakage across repeated operations.
This method allowed them to quantify leakage rates below 10^-6, meaning below one leakage event per million operations in the rate estimated by the measurement protocol. It provides a tool for separating tiny leakage contributions from other errors as control pulses are optimized further.
That matters because leakage can be particularly troublesome for quantum error correction. A qubit that leaves the computational subspace can remain there across multiple operations, creating errors that do not behave like the simpler error models on which many protocols are built.
The result concerns single-qubit gates on a transmon. The study does not demonstrate the same improvement for two-qubit gates, a complete processor or a long fault-tolerant computation. Nor does it show that an algorithm becomes faster by the same factor as the gate duration is reduced.
Its significance lies in the control problem. Quantum processors need gates that are fast enough to limit exposure to noise and precise enough to prevent errors from accumulating. The study shows that a previously underappreciated leakage channel can be identified and actively suppressed in the strongly driven regime.
R2D was developed and demonstrated on a widely used type of superconducting qubit, and the researchers say the principle can be adapted to other qubit platforms. How well it transfers across hardware and scales to larger processors remains to be demonstrated.
For superconducting quantum computers, this is therefore not a leap to fault tolerance by itself. It is a narrower improvement: fast control does not necessarily have to mean uncontrolled leakage out of the computational subspace.
KEEP EXPLORING
Quantinuum reports 99.9995% logical memory fidelity on its Helios system. The result shows strong error suppression with a color code, while scaling to useful fault-tolerant computation remains unproven.
Google has demonstrated a new way to produce a critical resource for error-corrected quantum computers. On its Willow processor, the method reduced errors by about 40 times and achieved a reported fidelity of 99.99 percent.
Microalgae embedded with a hydrogel make a soft actuator bend as light switches on and off. Researchers used the principle to build a simple gripper that lifts small objects without a conventional motor or battery.