Scalable error mitigation for noisy quantum circuits produces competitive expectation values

Abstract

The measurement of relevant physical observables on near-term quantum processors that are classically intractable is a key milestone for demonstrating a useful quantum advantage. Existing quantum computers are noisy and require error mitigation techniques to achieve an accurate estimation of the observables. Zero-noise extrapolation (ZNE) is a popular error mitigation technique that has been adopted in several small-scale experiments, however, the practical scaling of this method to a larger system size remains unknown. Here, we establish the scalability of ZNE and use it to enhance the accuracy of the quench dynamics of a 2D Ising spin-lattice using up to 26 qubits of a fully programmable superconducting processor. We discuss several additional experiment strategies to extend the reach of ZNE. Finally, we show that the measured observables can surpass an established classical approximate method in the limit of increasing entanglement.