A Fundamental Bound for Robust Quantum Gate Control

Robert L. Kosut, Daniel A. Lidar, Herschel Rabitz · 2025

Read A Fundamental Bound for Robust Quantum Gate Control by Robert L. Kosut, Daniel A. Lidar, Herschel Rabitz (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-2507-01215.txt.

Introduction

Robert Kosut, Daniel Lidar, and Herschel Rabitz derived a universal limit on how well a quantum gate can hold up under error, in a 2025 paper. Real quantum hardware faces both modeled and unmodeled flaws, from shaky control signals to stray links with the surroundings. A bound that covers all sources at once was missing.

Their proof joins an averaging method in a moving frame with a classic inequality. The result says the worst-case gate score sits above a floor set by one combined number: gate time times a single frequency-like measure that lumps every bounded error source. The bound needs no split between system and bath at the start. It holds for single qubits, multi-level units, and helper-based gates alike. The authors offer it as a device-free yardstick for fault-tolerance goals.

Catalog record

CREATORRobert L. Kosut, Daniel A. Lidar, Herschel Rabitz
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE88 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-01215

Browse the full library · About the Tower of Babel

Full text

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

A Fundamental Bound for Robust Quantum Gate Control Robert L. Kosut, Daniel A. Lidar, Herschel Rabitz Abstract We derive a universal performance limit for coherent quantum control in the presence of modeled and unmodeled uncertainties. For any target unitary W W that is implementable in the absence of error, we prove that the worst-case (and hence the average) gate fidelity obeys the lower bound F ≥ F lb ​ ( T ​ Ω bnd ) F\geq F_{\rm lb}\bigl({T}\Omega_{\rm bnd}\bigr) , where T {T} is the gate duration and Ω bnd \Omega_{\rm bnd} is a single frequency-like measure that aggregates all bounded uncertainty sources, e.g., coherent control imperfections, unknown couplings, and residual environment interactions, without assuming an initially factorizable system-bath state or a completely positive map. The bound is obtained by combining an interaction-picture averaging method with a Bellman-Gronwall inequality and holds for any finite-norm Hamiltonian decomposition. Hence it applies equally to qubits, multi-level qudits, and ancilla-assisted operations. Because F lb F_{\rm lb} depends only on the dimensionless product T ​ Ω bnd {T}\Omega_{\rm bnd} , it yields a device-independent metric that certifies whether a given hardware platform can, in principle, reach a specified fault-tolerance threshold, and also sets a quantitative target for robust-control synthesis and system identification. We translate the theory into a two-objective optimization problem that minimizes both the nominal infidelity and the time-averaged error generator. As an illustrative example we consider a single-qubit Hadamard gate subject to an unknown σ z \sigma_{z} system-bath coupling; we obtain a five-pulse piecewise-constant control achieving a nominal error of 10 − 7 10^{-7} while virtually nulling the average disturbance. Monte Carlo simulations confirm that every observed infidelity lies below the predicted F lb F_{\rm lb} curve and that the bound is tight to within one order of magnitude in the relevant regime 1 − F ≲ 10 − 4 1-F\lesssim 10^{-4} . Our results provide a falsifiable benchmark for experimental characterization as well as a pathway toward error budgets compatible with scalable quantum information processing. 1 Introduction Quantum processors have progressed well beyond laboratory proofs of concept, yet they remain far from the fully fault-tolerant regime envisioned for large-scale computation Preskill 2018 . Current resource estimates indicate that the physical-to-logical qubit ratio required for fault tolerance is still prohibitive Gottesman 2022 ; Beverland et al. 2022 ; Preskill 2025 . In most architectures the dominant cost driver is the physical two-qubit gate error rate; reducing infidelities would translate directly into a corresponding reduction of overhead, although the exact savings depend on device specifics, error correcting code, and layout constraints Xu et al. 2024 ; et al. 2025 . A substantial part of this overhead can be avoided by maximizing robustness to all disturbances that ultimately trigger error correction Lidar and Brun 2013 . As Feynman presciently warned, uncontrolled interactions “may produce considerable havoc” in a quantum computer Feynman 1985 . If those interactions are suppressed before allocating error-correction resources, the number of ancilla qubits, circuit depth, and other costs can be significantly reduced. There are generally two paths to potentially achieve small infidelities in the laboratory setting with qubits: (1) Starting with a model of the system and environment, achieve a control design that is robust to simulated conceivable uncertainties for transfer to the laboratory for performance evaluation. (2) Start directly in the laboratory, likely guided by (1), physical motivation, and insights. Due to a host of uncertainties being present, the collective literature shows that neither of these approaches have proved to be fully satisfactory, especially for two qubit gates. This paper takes neither of these approaches, but rather introduces a new theoretical framework and associated mathematical analysis. Our method builds on a cornerstone of classical robust control— uncertainty modeling —in which disturbances are treated as “unknown but bounded” elements of a well-defined set Zames 1966 ; Desoer and Vidyasagar 1975 ; Boyd et al. 1994 ; Zhou et al. 1996 . This naturally raises a fundamental question: given such a model, what is the ultimate performance limit of any control strategy? Here, we lay the analytical groundwork for answering that question. Figure 1: Plot of infidelity upper bound 1-F_{\rm lb} from Eq. 30 in Theorem 1 . The bound is shown in a log scale vs. uncertainty {T}\Omega_{\rm bnd} in radians. Our main theoretical result establishes an explicit upper bound on worst-case infidelity as a function of a single, dimensionless time-bandwidth uncertainty quantity {T}\Omega_{\rm bnd} (see Theorem 1 in Section 4 ). Figure 1 plots this bound. Here {T} is the gate time and \Omega_{\rm bnd} is an aggregate frequency that upper-bounds all relevant terms in the system and system-bath Hamiltonians [ Eq. 28 ]. Theorem 1 follows from the classical Method of Averaging Hale 1980 and a specialized Bellman-Gronwall inequality Coddington and Levinson 1955 . Although the bound is not guaranteed to be tight, it delivers a quantitative measure for both analysis and synthesis: any device that can implement the target gate perfectly in the uncertainty-free model must, in the presence of bounded uncertainty, achieve an actual fidelity no worse than F_{\rm lb}({T}\Omega_{\rm bnd}) . Conversely, control and design choices that lower {T}\Omega_{\rm bnd} automatically tighten the bound. The bound provides a metric to compare with the results of either method (1) and/or (2). If the observed infidelity is too big, that outcome carries the message that additional relevant details need to be included in the model for case (1) and/or for case (2) improvements need to be made in the platform design and possibly its operational features. In contrast, if the observed infidelity is well below the bound, the situation is of course attractive; this circumstance could serve as a basis to stimulate an interchange between the laboratory and modeling efforts to improve the bound, possibly providing for a new and even better infidelity to be reached in the future. Two features distinguish our approach from traditional fidelity estimates. First, the bound is derived for arbitrary initial states, including maximally entangled system-bath inputs, and therefore does not rely on the completely-positive (CP) map framework that follows from a product state assumption M.A. Nielsen and I.L. Chuang 2010 . This is essential for realistic circuit execution, where the system becomes entangled with its environment between error-correction cycles Schindler et al. 2011 , thus precluding a CP map treatment Jordan et al. 2004 ; Carteret et al. 2008 ; Rodríguez-Rosario et al. 2008 ; Buscemi 2014 ; Dominy et al. 2015 ; Dominy and Lidar 2016 . Second, {T}\Omega_{\rm bnd} captures all error sources: the “known unknowns” included in the model and the “unknown unknowns” that inevitably remain. A design based on the former might, for example, predict {T}\Omega_{\rm bnd}\leq 0.1 rad, certifying an infidelity below 10^{-5} in Figure 1 . Even if hidden errors double the bound to 0.2 rad, the guaranteed error still stays beneath 10^{-4} . Numerical evidence in Section 7 shows that the Hamiltonian control acts in real time to steer the system while additionally providing robustness akin to the effect of feedback, in this case, without measurement. However, there is a limit to the types of unknown errors that can be included in the time-bandwidth uncertainty {T}\Omega_{\rm bnd} . For example, catastrophic errors such as qubit

Read the complete text