Experimental Demonstration of Lindblad Tomography on a Superconducting Quantum Device

Abstract

Information loss in experimental quantum devices is traditionally characterized using metrics such as T₁ and T₂, which are readily accessible from standard time-domain measurement. While T₁ and T₂ times provide rough heuristics for interaction between single qubits and their lossy environments, these numbers stand in as mere proxies for the full multi-qubit loss channel of interest, which can be described more fully with a Lindbladian operator in the master equation formalism. In this thesis, I outline and present the results of the first experimental demonstration of Lindblad Tomography, a novel technique for tomographically reconstructing the Hamiltonian and Lindbladian operators of an arbitrary quantum channel from an ensemble of time-domain measurements. Starting from a theoretically minimal set of assumptions, I show that this method is resilient to state-preparation and measurement (SPAM) errors and places strong bounds on the degree of non-Markovianity in the channel of interest. Comparing the results for single- and two-qubit tomography of a superconducting quantum processor, I demonstrate how Lindblad Tomography can be used to identify sources of crosstalk on large quantum processors, particularly in the presence of always-on qubit-qubit interactions.