Exploring and benchmarking cross-resonance gates via the second excited state

Quantum computing requires many qubits combined with high-quality gate operations. Fixed-frequency transmon qubits are a leading approach, having demonstrated long coherence, fast single qubit gates, and hundreds of qubits on a chip. The cross-resonance interaction has been used to demonstrate high-fidelity two qubit gates, but scaling is challenging due to the gate-speed and fidelity being sensitive to the frequency detuning between qubits, which is hard to control accurately during fabrication. There is a trade-off between smaller detunings (leading to frequency crowding and undesired crosstalk) and larger detunings (slower gates).

Here, we investigate a variation on cross-resonance where the control qubit is moved from first to second excited state during gate operation, then returned. We show that this allows fast gates over a range of detunings where the standard gate would be inefficient, and demonstrate this experimentally on a pair of qubits on an OQC Toshiko device, reaching a ZX drive rate of 2.5MHz, while the standard case saturates at ~0.7MHz. We calibrate a ZX(𝜋/2) two-qubit gate and measure 99% gate fidelity using interleaved randomized benchmarking. This approach could help design fixed-frequency transmon devices which avoid frequency crowding while enabling fast gates.