Analysis and Suppression of Errors in Quantum Random Access Memory under Extended Noise Models

TL;DR

This paper extends the understanding of quantum random access memory's (QRAM) robustness against various realistic noise models, showing it maintains polylogarithmic error scaling even with complex errors, guiding future QRAM design.

Contribution

It generalizes previous resilience results of bucket-brigade QRAM to include initialization, correlated, and coherent errors, and discusses error mitigation strategies.

Findings

QRAM maintains polylogarithmic error growth under generalized noise models.

Initialization errors may not require reset protocols between queries.

Coherent errors can be mitigated using randomized compiling schemes.

Abstract

Quantum random access memory (QRAM) is required for numerous quantum algorithms and network architectures. Previous work has shown that the ubiquitous bucket-brigade QRAM is highly resilient to arbitrary local incoherent noise channels occurring during the operation of the QRAM [PRX Quantum 2, 020311 (2021)], with query infidelities growing only polylogarithmically with memory width when errors are assumed to only occur on individual routers. We extend this result to a large class of generalized settings that arise in realistic situations, including arbitrary initialization errors, spatially correlated errors, as well as coherent errors, maintaining the polylogarithmic scaling in all instances. Fully quantifying the extent to which QRAM's noise resilience holds may provide a guide for the design of QRAM architectures - for instance, the resilience to initialization errors indicates that a reset protocol between successive queries may not be necessary. In the case of coherent errors, we find an up-to-quadratic increase in the infidelity bound, and therefore discuss generalizations to randomized compiling schemes, which usually are rendered inapplicable in the QRAM setting, to tailor these errors into more favorable stochastic noise.