Multi-qubit circuit characterization through physics-based statistical inference

We develop new physical models for realistic two-qubit gates in superconducting qubit architectures to account for specific shape of qubit control pulses and effects of noise and decoherence, including those induced by two-level defects prevalent in the fabrications of superconducting chip. We then formulate statistical inference method to efficiently estimate the physical model parameters for an ensemble of quasi-random circuits that contain non-Clifford gates. For a family of circuit ensembles, we are able to obtain analytically the circuit fidelities and their variances by averaging the log-likelihood distribution of the model parameters over Haar measure. Our inference method applies to quantum systems with arbitrary number of qubits and thus serves as valuable tools for characterizing large scale quantum circuit that will likely outperform the most powerful classical computers existing to date.