Approaching the chaotic regime with a fully connected superconducting quantum processor

Abstract

The task of sampling the output distribution of random quantum circuits will drive a multiqubit system into chaos, reaching the Porter-Thomas distribution where small multiqubit probabilities appear most often. Such a task is hard to be computed classically and has been proposed as an important measure for quantum supremacy. An equivalent route to reach chaos is the controlled dynamics uniformly exploring the large Hilbert space of coupled qubits, and a proof-of-concept experiment was recently demonstrated by Neill et al. [Science 360, 195 (2018)] in a chain of nine qubits featuring only neighboring couplings. Here, we show numerically that a faster convergence to chaos is possible when full connectivity, instead of partial connectivity, is present among the qubits, i.e., cycles or time required to reach chaotic regime increase only slightly with the number of qubits. We validate the numerical result by carrying out an experiment using up to 12 qubits in a fully connected superconducting quantum processor. Based on a machine learning model, we quantify the evolution error rate to be less than (0.5%/cycle)/qubit, which is mostly determined by qubit dephasing.

Similar papers