Divide-and-conquer verification method for noisy intermediate-scale quantum computation

TL;DR

This paper introduces an efficient divide-and-conquer verification method for noisy intermediate-scale quantum computations, enabling fidelity estimation with polynomial resources for certain circuit classes, demonstrated through experiments on IBM's 5-qubit chip.

Contribution

The paper presents a novel verification approach that reduces resource requirements for fidelity estimation in noisy quantum circuits, especially for logarithmic-depth circuits on sparse chips.

Findings

Efficient fidelity estimation with polynomial resource scaling for specific quantum circuits.

Characterization of small-scale quantum operations using the diamond norm.

Successful proof-of-principle experiment on IBM's 5-qubit quantum chip.

Abstract

Several noisy intermediate-scale quantum computations can be regarded as logarithmic-depth quantum circuits on a sparse quantum computing chip, where two-qubit gates can be directly applied on only some pairs of qubits. In this paper, we propose a method to efficiently verify such noisy intermediate-scale quantum computation. To this end, we first characterize small-scale quantum operations with respect to the diamond norm. Then by using these characterized quantum operations, we estimate the fidelity $\langle\psi_t|\hat{\rho}_{\rm out}|\psi_t\rangle$ between an actual $n$-qubit output state $\hat{\rho}_{\rm out}$ obtained from the noisy intermediate-scale quantum computation and the ideal output state (i.e., the target state) $|\psi_t\rangle$. Although the direct fidelity estimation method requires $O(2^n)$ copies of $\hat{\rho}_{\rm out}$ on average, our method requires only $O(D^32^{12D})$ copies even in the worst case, where $D$ is the denseness of $|\psi_t\rangle$. For logarithmic-depth quantum circuits on a sparse chip, $D$ is at most $O(\log{n})$, and thus $O(D^32^{12D})$ is a polynomial in $n$. By using the IBM Manila 5-qubit chip, we also perform a proof-of-principle experiment to observe the practical performance of our method.