Characterizing noisy quantum computation with imperfectly addressed errors

Riddhi S. Gupta, Salini Karuvade, Kerstin Beer, Laura J. Henderson, Sally Shrapnel · 2025

Read Characterizing noisy quantum computation with imperfectly addressed errors by Riddhi S. Gupta, Salini Karuvade, Kerstin Beer, Laura J. Henderson, Sally Shrapnel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-03261.txt.

Introduction

Gupta and four co-authors study how noise slips past quantum error handling. Their 2025 paper asks what happens when fixes rest on wrong assumptions about the noise. Error correction and mitigation both presume a tidy noise shape. Real hardware noise often breaks those presumptions quietly.

The team models a noisy computation as a crowd of random superoperators. They study the spread of its eigenvalues and singular values. A new concentration result for random matrices powers the analysis. The spectra shift in telltale ways when assumptions fail. Those shifts apply to both correction and mitigation failures. The frame may help set spectral gaps and relaxation times for quantum processes.

Catalog record

CREATORRiddhi S. Gupta, Salini Karuvade, Kerstin Beer, Laura J. Henderson, Sally Shrapnel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-03261

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Characterizing noisy quantum computation with imperfectly addressed errors Preprint: APS/123-QED Riddhi S. Gupta, Salini Karuvade, Kerstin Beer, Laura J. Henderson, Sally Shrapnel Abstract Quantum protocols on hardware are subject to noise that prohibits performance. Protocols for addressing errors, such as error correction or error mitigation, may fail to combat errors in quantum computation if noise violates critical assumptions required for these protocols to be effective. However, tools for characterizing such failures in realistic operating conditions are limited. For example, while brute force simulations may be used to characterize the impact of such failures on a handful of input states, such simulations lack a complete description for how noise transforms state-spaces in the full quantum Hilbert space. In this work, we associate quantum computation subject to realistic noise to an ensemble of random superoperators and study the eigen- and singular spectral distributions over this ensemble. We propose a new theoretical framework to characterize singular values of random complex matrices using matrix Chernoff concentration. Using our framework, we analyze imperfectly addressed errors in error mitigation and error correction. We find that distributions of singular spectra depend on how noise violates critical assumptions of these protocols. Finally, we quantitatively discuss how our work may be applied to understanding limiting behavior of quantum computation, such as establishing spectral gaps and relaxation times for specific families of quantum Markov processes. Our work paves the way for new tools to diagnose when to trust the output of noisy quantum computers. I Introduction While quantum computing experiments continue to demonstrate remarkable progress in underlying hardware capabilities, combating the effects of noise on computation is critical to trusting the output of quantum hardware. Two dominant strategies for addressing noise arise as quantum error correction Gottesman 1997 or quantum error mitigation Temme et al. 2017 ; van den Berg et al. 2023 . While error correction detects and corrects local errors in quantum states under operating regimes with sufficiently low-error, error mitigation yields unbiased estimators of noise-free expectation values while tolerating higher noise levels. Indeed both error correction and mitigation protocols assume that intrinsic noise has certain properties that may not hold true under realistic operating conditions. Examples include, e.g., that intrinsic noise is Markov; that intrinsic noise is sufficiently weak; or that intrinsic noise has restricted impact or is restricted to certain types of operations. Indeed quantum protocols for addressing noise may themselves be subject to pathological conditions which violate critical assumptions required for protocols to be effective. Tools for simulating quantum computations under these kinds of realistic operating conditions are limited. One typical approach for simulating noisy quantum computation is to use channel formalism represented by superoperators, i.e. complex matrices, acting on vectorized quantum states Watrous . Previous literature has established distance metrics to compare ideal versus noisy quantum channels Gilchrist et al. 2005 using induced Schatten norms that depend on singular values or the numerical ranges of operators Watrous 2004 ; Johnston et al. 2007 ; Rastegin 2009 ; Chełstowski et al. 2023 . Aside from such distance metrics, it is also well recognized that channel superoperators may define transition matrices for Markov chains on non-commutative probability spaces Temme et al. ; Wolf . Existing work on quantum Markov processes establishes the limiting behavior of specific types of quantum channels under repeated applications, including how quickly initial information is forgotten by a channel Terhal and DiVincenzo ; Burgarth et al. ; Albert . Indeed the spectral properties of quantum channels are important for characterizing e.g. subspaces in error correction Ippoliti et al. , mixing times, relaxation times and fixed points of mixing channels Temme et al. ; Terhal and DiVincenzo ; Burgarth et al. ; Albert ; spectral gaps of random channels Bruzda et al. ; Kukulski et al. , random Lindblad dynamics Can 2019 , and random isometries Fischmann et al. ; González-Guillén et al. ; Karuvade et al. 2018 . Despite the importance of eigen- or singular values to establishing both channel norms and limiting behavior of channels, they have not been studied to understand when noisy quantum protocols fail. In this work, we motivate the study of eigen- and singular values of noisy quantum computation subject to realistic, imperfect operating conditions. We consider an ensemble of random superoperators where each sample represents an experimental run probabilistically subject to noise that may break a protocol’s underlying assumptions. For example, for error correction, the number of single qubit errors may exceed code distance, while for error mitigation, a learned model may drift during mitigation. We call such random processes failure modes. As a first step towards operational usefulness, we propose a general theoretical framework to provide an upper bound to any ordered singular value of ensembles of random channel superoperators. Our analysis holds even if random superoperators only satisfy Kraus conditions in an idealized limit where the probability of incurring a failure mode goes to zero. To illustrate potential applications of our framework, we empirically find that both the singular spectral radius of imperfectly error-mitigated quantum gates as well as the average singular multiplicities of imperfectly error-corrected quantum codes are correlated with different failure modes for these protocols. We discuss how these insights could be useful for comparing operational robustness of different quantum protocols and/or establish their limiting behavior. The structure of this document is as follows. section II provides the groundwork for how classical eigenspectral concentration results for random Hermitian matrices are applied to random complex matrices. In section III , we present our lemma to upper bind the maximal singular value of general random matrices with concentration. We apply this lemma to superoperators associated with quantum channels for error correction and error mitigation. We characterize the affect of failure modes empirically and compare with theoretical results. In section IV we extend this lemma to a theorem that provides an upper bound for any singular value in an ordered set with concentration. We discuss potential for future operational applications before concluding in section V . II Background To understand the spectrum of general quantum channels, we will need to elevate results applicable to Hermitian matrices to the general class of complex matrices using a Hermitian dilation Tropp . In terms of notation, we denote the eigenvalues of a general complex square matrix \hat{\Lambda}\in\mathbb{C}^{d^{\prime}\times d^{\prime}} as the set of all complex numbers that satisfy \mathrm{det}(\lambda\mathbb{I}_{d^{\prime}}-\hat{\Lambda})=0 . We use \lambda(\hat{\Lambda}) to denote the full set of eigenvalues of the operator \hat{\Lambda} , as well as to denote the i -th ordered eigenvalue \lambda_{i}(\hat{\Lambda}) . The set of Hermitian operators are denoted by \chi\in\mathbb{C}^{d^{\prime}\times d^{\prime}} and the set of non-negative Hermitian operations are denoted by X\in\mathbb{C}^{d^{\prime}\times d^{\prime}} . Definition 1 . A Hermitian dilation \mathcal{M}_{d^{\prime}}\to\mathcal{M}_{2d^{\prime}} of a complex d^{\prime}\times d^{\prime} matrix \hat{\Lambda} is \displaystyle\chi:=\begin{bmatrix}0&\hat{\Lambda}\\ \hat{\Lambda}^{\dagger}&0\end{bmatrix}, (1) where 0 represents sub-blocks, \hat{\Lambda}^{\dagger} is a conjugate transpose of \hat{\Lambda} , and \chi is Hermitian. Note that the action of the dilation is to relate the results for eigenvalues of Hermitian matrices to singular values of complex matrices. Singular values \sigma_{i} of a complex matrix, \hat{\Lambda} are the square-root of eigenvalues of the products, \displaystyle\sigma_{i}(\hat{\Lambda}):=\sqrt{\lambda_{i}(\hat{\Lambda}^{\dagger}\hat{\Lambda})}=\sqrt{\lambda_{i}(\hat{\Lambda}\hat{\Lambda}^{\dagger})},\quad\forall i. (2) The right (left) singular vectors are the eigenvectors of \hat{\Lambda}^{\dagger}\hat{\Lambda} ( \hat{\Lambda}\hat{\Lambda}^{\dagger} ). We use \sigma(\hat{\Lambda}) to denote the full set of singular

Read the complete text