Mitigating shot noise in local overlapping quantum tomography with semidefinite programming

TL;DR

This paper introduces a semidefinite programming approach to mitigate shot noise in quantum tomography, improving the accuracy of reduced density matrices and aiding in the preparation of low-energy states in near-term quantum computers.

Contribution

It presents a novel method to enforce physicality constraints on RDMs via semidefinite programming, enhancing measurement efficiency and accuracy in quantum state estimation.

Findings

Tighter bounds for RDMs with fewer measurements.

Improved accuracy in low-energy state preparation.

Enhanced resource efficiency in quantum tomography.

Abstract

Reduced density matrices (RDMs) are fundamental in quantum information processing, allowing the computation of local observables, such as energy and correlation functions, without the exponential complexity of fully characterizing quantum states. In the context of near-term quantum computing, RDMs provide sufficient information to effectively design variational quantum algorithms. However, their experimental estimation is challenging, as it involves preparing and measuring quantum states in multiple bases--a resource-intensive process susceptible to producing non-physical RDMs due to shot noise from limited measurements. To address this, we propose a method to mitigate shot noise by re-enforcing certain physicality constraints on RDMs. While verifying RDM compatibility with a global state is quantum Merlin-Arthur complete, we relax this condition by enforcing compatibility constraints up to a certain level using a polynomial-size semidefinite program to reconstruct overlapping RDMs from simulated data. Our approach yields, on average, tighter bounds for the same number of measurements compared to tomography without compatibility constraints. We demonstrate the versatility and efficacy of our method by integrating it into an algorithmic cooling procedure to prepare low-energy states of local Hamiltonians. Simulations on frustrated Hamiltonians reveal notable improvements in accuracy and resource efficiency, highlighting the potential of our approach for practical applications in near-term quantum computing.