Short answer: A tunable coupler is an extra circuit element placed between two superconducting qubits so their interaction can be switched on for a gate and suppressed otherwise. Experiments report controlled-Z gates of 30 to 60 nanoseconds with fidelities from above 98 percent to 99.55 percent, and residual coupling as low as a few kilohertz. Fidelity in these papers is mostly limited by qubit coherence. They are laboratory results on two-qubit devices.
What the coupler is for
A 2020 Physical Review Letters paper says two ingredients are highly desirable for scalable two-qubit gates. The first is a tunable coupler, which it describes as proven effective at resolving frequency crowding, suppressing residual coupling and enabling fast, high-fidelity gates. The second is compatibility with fixed-frequency qubits. A 2022 npj Quantum Information paper adds that crosstalk suppression is a major issue for large devices, and that among transmon qubits one of the most critical kinds of crosstalk is static ZZ interaction. It says tunable couplers have recently been widely used to mitigate ZZ crosstalk in a scalable and effective way.
A fixed-frequency scheme with a fast gate
The 2020 paper demonstrates a two-qubit gate scheme using fixed-frequency qubits and a tunable coupler. The effective longitudinal coupling between the qubits is adjusted by a single control parameter of the coupler. With a proper idle point the residual coupling strength is as small as 20 kHz. Strong qubit-coupler coupling, above 100 MHz, gives enhanced adiabaticity, and the authors realise an adiabatic controlled-Z gate of 30 ns with a fidelity of 99.5 percent, derived from interleaved randomised benchmarking. They say the scheme needs fewer control lines, reduces crosstalk and simplifies calibration, and that errors from separate measurements are mostly due to decoherence.
A three-mode coupler
A Physical Review Applied paper proposes a building block of two transmons and a tunable three-mode coupler that allows control of the ZZ interaction. It demonstrates a native controlled-Z gate with a 60 ns pulse and a fidelity above 98 percent, limited mostly by qubit coherence time. Numerical simulations in the paper suggest that optimising the gate duration could push the fidelity above 99.97 percent. This is a simulation result and not a measured one.
Couplers with fluxonium qubits
The 2022 paper notes that fluxonium qubits offer better coherence and larger anharmonicity than transmons, but that a scalable, crosstalk-free multi-qubit architecture with high-fidelity gates, single-shot readout and state initialisation was still a major challenge. The authors present a two-qubit fluxonium processor with a tunable coupler element and demonstrate fSim-type and controlled-Z gates with fidelities of 99.55 and 99.23 percent. The residual ZZ interaction is suppressed to a few kilohertz. A galvanically coupled flux control line allows high-fidelity single-qubit gates and ground state initialisation with one arbitrary waveform generator channel per qubit.
What this does not tell you
Each experiment uses its own qubit type, benchmark method and device, so the fidelities are not directly comparable. All are two-qubit systems, so they do not show how the coupler performs in a large processor with many neighbours.
FAQ
What does a tunable coupler do?
It switches the interaction between two qubits on for a gate and suppresses it otherwise.
How fast are the gates?
The papers report 30 ns and 60 ns controlled-Z gates.
How good is the fidelity?
From above 98 percent up to 99.55 percent in the measured results.
What limits fidelity?
Mostly qubit coherence, according to the 2020 and Physical Review Applied papers.
Does it reduce crosstalk?
Yes. The papers report residual coupling of 20 kHz or a few kilohertz at the idle point.
All field notes