Implementing and benchmarking dynamically corrected gates on superconducting devices using space curve quantum control

We use Space Curve Quantum Control (SCQC) to design, experimentally demonstrate, and benchmark dynamically corrected single-qubit gates on IBM hardware, comparing their performance to that of the standard gates provided by IBM. Our gates are designed to dynamically suppress both detuning and pulse-amplitude noise, with gate times as short as 88 ns. We compare our gates against those of IBM on two separate IBM devices and across sets of up to 18 qubits. Randomized benchmarking is done utilizing our detuning- and amplitude-robust gates in randomized Clifford circuits containing up to 4000 gates. Our gates achieve error-per-Clifford rates that reach as low as 7$\times10^{-5}$ ($\pm10^{-6}$) and which remain nearly constant as the compound noise is increased up to 4% amplitude noise and up to a detuning noise of 342 kHz; this is in contrast to the IBM gates, which exhibit rates that drop to order $10^{-3}$ across this range. This range is consistent with the commonly reported frequency fluctuations and with the upper bound of the statistical uncertainty in gate calibration. In addition, we investigate the performance across larger noise ranges of up to 20% amplitude and 3.5 MHz detuning noise using quantum process tomography. Finally, we experimentally demonstrate how SCQC can be tailored to different practical use cases by trading off amplitude-robustness for ultrafast 60 ns dephasing-only robust pulses. Our work establishes experimental guidelines for implementing SCQC-designed dynamically corrected gates on a broad range of qubit hardware to limit the effect of noise-induced errors and decoherence.