Suppressing ZZ crosstalk with tunable couplers

Short answer: ZZ crosstalk is an unwanted always-on interaction between neighboring superconducting qubits, and a tunable coupler can be used to cancel it. A 2019 Physical Review Applied paper reports static ZZ suppression using two couplers whose contributions interfere destructively, with a Bell-state fidelity of 98.5% and an iSWAP-gate fidelity of 94.8%. A 2023 arXiv preprint proposes a coupler model that in theory suppresses ZZ completely. One is an experiment and one is a theory preprint.

The problem

The Physical Review Applied abstract states that parasitic crosstalk in superconducting quantum devices is a leading limitation for quantum gates. The 2023 preprint adds that better hardware needs higher gate fidelities and lower crosstalk, particularly in larger devices where qubits are coupled to several neighbors. It says progress on both would benefit from being able to fully control all interactions between pairs of qubits.

2019: two couplers that cancel

The authors demonstrate suppression of static ZZ crosstalk in a two-qubit, two-coupler circuit. The frequency of a tunable coupler can be adjusted so that the ZZ interactions from each coupler interfere destructively. They verify the elimination of crosstalk with simultaneous randomized benchmarking. They then use a parametrically activated iSWAP interaction to reach a Bell-state preparation fidelity of 98.5% and an iSWAP-gate fidelity of 94.8%, the latter obtained via quantum process tomography.

2023: a proposed coupler

The arXiv preprint, which is not peer reviewed in the version read here, proposes a new coupler model that fully decouples dispersively detuned transmon qubits, so ZZ crosstalk is completely suppressed while the computational basis states stay maximally localized. The authors reason that, for a dispersively detuned transmon system, this can only be the case if the anharmonicity of the coupler is positive at the idle point. A simulation of a 40 ns CZ gate for a lumped-element model suggests that the achievable process infidelity could be pushed below the limit set by state-of-the-art transmon coherence times. They say idle gates between qubits would no longer be limited by parasitic interactions, and that the scheme could apply to large integrated qubit grids, where it would isolate a pair of qubits during a gate from the rest of the chip. These are simulation-based claims.

How the two approaches differ

The 2019 approach cancels ZZ interaction by combining two couplers, so cancellation relies on tuning one coupler frequency against the other. The 2023 proposal is about the coupler's own properties, namely a positive anharmonicity at the idle point, which the authors argue is required for complete suppression in the dispersive regime. The first is something measured with randomized benchmarking, the second is a design rule derived by the authors and tested in simulation. They are not competing results on the same device.

Limits

The experiment involves two qubits and two couplers, and the gate fidelities there are below the simulated targets in the preprint. The preprint's results come from a lumped-element simulation and a model, with no hardware data in the abstract. Neither abstract says how the approaches would behave on a large device with fabrication variation.

FAQ

What is ZZ crosstalk?

A parasitic static interaction between qubits that limits gate quality, per the 2019 abstract.

How can a coupler cancel it?

In 2019, two couplers were tuned so their ZZ interactions interfere destructively.

How good were the gates?

A Bell-state fidelity of 98.5% and an iSWAP-gate fidelity of 94.8%.

What does the 2023 proposal claim?

Complete ZZ suppression with a coupler whose anharmonicity is positive at the idle point, shown in simulation.

Has the 2023 scheme been built?

The abstract describes simulation, not a hardware demonstration.