The Signal
TRIO Pulses Keep X Gates Above 99.5% Under Control Errors
A team has tested TRIO microwave pulses for single-qubit gates on a superconducting processor. Its 90-nanosecond X gate stayed above 99.5% fidelity under simultaneous drive-amplitude and qubit-frequency offsets.
Quantum.world Team · Written with AI · fact-checked ·
A 90-nanosecond gate faces two errors at once
A 90-nanosecond X gate held fidelity above 99.5% while researchers deliberately changed both the microwave drive amplitude and the qubit frequency. The result comes from a superconducting processor test of a pulse design the team calls TRIO, described in a paper posted to arXiv on October 16, 2026.
The authors write, “For the X gate, we maintain fidelities above 99.5% for simultaneous drive-amplitude variations up to 20% and qubit detuning up to 4 MHz.” Tangyou Huang et al., arXiv paper
The test targets a familiar control problem: a gate calibrated to work precisely can lose accuracy when drive settings or qubit frequencies shift. Rather than relying only on recalibrating after changes occur, the researchers shape the pulse to make its result less sensitive to those offsets.
Smoother waveforms preserve the designed trajectory
TRIO combines a pulse trajectory optimized for energy and tolerance to control errors with a time-rescaling method that produces smoother waveforms. The smoothing matters because sharp changes in an idealized pulse can be distorted by real microwave hardware. In a transmon, those distortions can also push the system out of the two states used to encode a qubit.
The researchers implement the pulses using in-phase and quadrature microwave control, with a compensating virtual-Z rotation. They also apply calibrated DRAG corrections, a standard pulse adjustment intended to limit unwanted excitation and compensate for a frequency shift caused by the transmon’s imperfectly spaced energy levels.
For the X gate, interleaved randomized benchmarking measured fidelity at 99.90% with TRIO, compared with 99.94% for a conventional Gaussian pulse when the controls were calibrated. Under simultaneous amplitude and detuning offsets, however, the paper reports about 99.5% fidelity for TRIO versus about 92% for the same-duration Gaussian pulse. The advantage is therefore tolerance to the tested errors, not a higher best-case score.
The test is controlled—not a long-term drift trial
The team tested a single qubit in a processor containing two fixed-frequency transmons connected through a flux-tunable coupler. Researchers introduced frequency offsets by adjusting the coupler’s direct-current flux bias and amplitude offsets by changing the pulse generator’s output. Those controlled perturbations show how the gate responds to specified errors; they are not a demonstration of a processor maintaining accuracy through an extended period of natural drift.
The paper also extends the method to Hadamard-type gates. At 90 nanoseconds, the reported TRIO fidelities are about 99.63% for the Hadamard operation, 99.69% for a Y rotation by minus pi over two, and 99.62% for an X rotation by pi over two. The authors say the set’s measured error distributions also favor TRIO over Gaussian pulses across sampled amplitude and frequency offsets.
Shorter gates bring a leakage trade-off
The result does not mean the pulse can simply be shortened without cost. For gate times at or below 90 nanoseconds, the X-gate TRIO pulse’s larger area requires a stronger envelope, and the paper reports increased leakage—about 10-3, compared with about 10-5 for the Gaussian pi pulse. That leakage contributes to lower fidelity at shorter durations.
The experiments establish a control technique on one superconducting processor, not a system-wide performance improvement across a large processor. The paper identifies higher-bandwidth waveform generation and additional leakage mitigation as possible routes to faster gates. Its measured demonstration remains a 90-nanosecond single-qubit gate, with fidelity above 99.5% under the specified simultaneous offsets.