Speaker
Description
As quantum processors scale, understanding the structure of the noise affecting their operations becomes essential for error mitigation and error correction. Many widely used characterization and mitigation methods assume the noise to be effectively stochastic, yet coherent crosstalk persists in state-of-the-art devices, and testing such assumptions typically requires full process tomography. In this work, we develop a scalable certification framework to detect coherent crosstalk directly at the level of quantum channels. Our key idea is that coherent crosstalk corresponds to entanglement in the Choi state of the noise channel across the input-output pairs associated with different qubits, turning crosstalk detection into an entanglement-detection problem. This connection allows us to combine techniques from entanglement detection with tools from noise learning: we estimate channel correlators from randomized product-state preparations and local Pauli measurements, invert the ideal gate in classical post-processing, and test compatibility with a separable-channel model through a hierarchy of moment-based semidefinite-programming relaxations. The same construction yields a stricter test against a Pauli-channel model, detecting non-Pauli features of the noise. At fixed relaxation order, the method requires only a polynomial number of moment variables. We demonstrate both tests numerically on entangling unitary channels and amplitude-damping noise, certifying coherent crosstalk and non-Pauli noise even in the presence of finite statistical uncertainty.
| Topic | Quantum state and process tomography with learning-based approaches |
|---|---|
| Additional Topic | Quantum state reconstruction from data |