⚛️ quantum physics

Certification of quantum properties with imperfect measurements

This paper presents a robust certification framework for quantum states that utilizes convex optimization to bound convex functions while jointly accounting for both statistical shot noise and systematic measurement imperfections.

Original authors: Leonardo Zambrano, Teodor Parella-Dilmé, Antonio Acín, Donato Farina

Published 2026-01-26

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Leonardo Zambrano, Teodor Parella-Dilmé, Antonio Acín, Donato Farina

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a chef trying to prove that a cake you baked is a perfect chocolate sponge. To do this, you take a few bites (measurements) to check the texture and taste. However, two things can go wrong:

The "Bite" Problem: You only took a few bites, so your sample might not perfectly represent the whole cake (this is shot noise or statistical error).
The "Taste Bud" Problem: Your tongue is slightly numb, or your fork is bent, so the taste you feel isn't exactly what the cake actually tastes like (this is measurement imperfection).

Most previous methods for checking quantum "cakes" (quantum states) assumed your taste buds were perfect. They only worried about taking too few bites. If your tongue was numb, those methods would give you a confident but wrong answer, like telling you the cake is chocolate when it's actually vanilla.

This paper introduces a new, more robust way to certify quantum properties that accounts for both problems at once.

The Core Idea: A "Safety Zone"

The authors propose a method that creates a "Safety Zone" (called a confidence region) around the true state of the system.

Old Way: They drew a small circle around the data, assuming the measuring tools were perfect. If the real data was slightly off due to a broken tool, the circle might not even touch the truth.
New Way: They draw a larger, expanded circle. This circle is big enough to cover the uncertainty from taking few bites plus the uncertainty from having a bent fork.

Inside this larger circle, they use a mathematical "sieve" (convex optimization) to find the best and worst possible answers for what they are trying to measure. This guarantees that the true answer is somewhere inside that range, no matter how imperfect the tools were.

How They Measure the "Broken Fork"

The paper explains how to figure out just how "broken" or "imperfect" your measuring tools are. You don't need to know the exact physics of the error; you just need to know the maximum distance between what you intended to measure and what you actually measured.

They offer a few ways to find this distance:

Simulation: If you know your machine is noisy, you can run a computer simulation to guess the error.
Calibration: You can run specific test experiments (using special "test cakes") to measure exactly how much your tools deviate from the ideal.
Mathematical Bounds: If your machine is made of smaller parts (like a multi-qubit system), you can measure the error of each small part and add them up to get the total error.

Real-World Examples from the Paper

The authors tested their method with three scenarios to show why ignoring broken tools is dangerous:

Checking Cake Quality (Fidelity): They tried to verify if a quantum state was prepared correctly. Even with noisy tools, their method gave a reliable "worst-case" score for how good the cake was.
Measuring Magnetism (The "Spinning Top"): Imagine a system of spinning tops that should all be pointing up (fully magnetized). If their measuring tools were slightly rotated (a common error), the old method would say, "The tops are pointing in all directions!" (a false conclusion). The new method, accounting for the rotation, correctly said, "The tops are still pointing up, we just looked at them from a weird angle."
Detecting Entanglement (The "Magic Link"): They tried to prove two particles were "entangled" (linked in a spooky way). With noisy tools, the old method falsely claimed a normal, unlinked pair of particles was entangled. The new method correctly identified that the particles were not entangled, preventing a false alarm.

Why This Matters

The paper concludes that this method is versatile and robust.

It doesn't need perfect tools: You don't need to build a perfect measuring device to get a valid result.
It's flexible: You don't need to measure every single possible property of the system (which is often impossible); you just need enough data to fit within the safety zone.
It's reliable: Even when the noise is high or the number of measurements is low, the method still provides a guaranteed answer, whereas older methods would simply break down or give misleading results.

In short, this paper provides a toolkit for scientists to say, "We know our tools aren't perfect, but here is a mathematically guaranteed range where the truth lies," ensuring that quantum technology advances on solid, error-checked ground.

Technical Summary: Certification of Quantum Properties with Imperfect Measurements

Problem Statement

The precise characterization of quantum systems is a fundamental prerequisite for the advancement of quantum technologies. A critical task in this domain is certifying the values of convex functions of quantum states, which include essential quantities such as expectation values of Hamiltonians, quantum fidelities, entropies, and quantum Fisher information.

Existing methods for estimating these quantities typically rely on quantum state tomography followed by the application of the function to the reconstructed state. However, these standard approaches often fail to provide rigorous certification because they neglect realistic imperfections, specifically shot noise (arising from finite statistics) and measurement errors (systematic deviations in the measurement apparatus). Ignoring these effects can lead to large uncertainties, compromising the robustness of quantum computations or resulting in incorrect conclusions about experimental outcomes. While recent works have addressed shot noise by defining confidence regions for quantum states, they generally assume that measurement devices are perfectly characterized, thereby failing to account for calibration errors or imperfections in the measurement process.

Methodology

The authors propose a certification method that jointly accounts for both shot noise and measurement imperfections. The framework extends the confidence-region approach introduced in Ref. [11] to scenarios where experimenters have imperfect control over measurements.

Core Framework

The method defines a confidence region for the experimental quantum state $\omega$ by bounding the distance between the ideal probabilities and the experimentally estimated counterparts. This is achieved through two primary components:

Statistical Error (Shot Noise): The deviation between the empirical estimate $\hat{f}$ and the true probability vector $\vec{p}$ is bounded using the Bretagnolle–Huber–Carol inequality. For $N$ copies of the state and $m$ measurement outcomes, the $l_1$ distance is bounded by $\epsilon_1 = \sqrt{\frac{2}{N} \ln \frac{2m}{\delta}}$ with probability $1-\delta$ .
Systematic Error (Measurement Imperfections): The discrepancy between the target Positive-Operator Valued Measure (POVM) $\{E_k\}$ and the actual implemented POVM $\{F_k\}$ is quantified using the operational distance $d_{op}(E, F)$ . This distance is defined as the maximum $l_1$ distance between the probability distributions generated by the two POVMs over all possible states. The authors bound this distance by $\epsilon_2$ , such that $d_{op}(E, F) \leq \epsilon_2/2$ .

The Certification Protocol

The certified bounds for a convex function $F(\omega)$ are obtained by solving a convex optimization problem. The goal is to find the minimum (or maximum) of $F(\rho)$ over a set of density matrices $\rho$ that satisfy the following constraints with probability at least $1-\delta$ :

$\text{tr}(\rho) = 1$ and $\rho \geq 0$ .
$\sum_{k=1}^m |\text{tr}(E_k \rho) - \hat{f}_k| \leq \epsilon_1 + \epsilon_2$ .

Here, $\hat{f}_k$ represents the experimental frequency obtained from measuring the imperfect POVM $\{F_k\}$ . The constraint sum $\epsilon_1 + \epsilon_2$ is derived via the triangle inequality, combining the statistical error and the operational distance between the ideal and actual measurements.

Determining the Operational Distance Bound ( $\epsilon_2$ )

The paper provides several explicit prescriptions for estimating $\epsilon_2$ depending on the available knowledge of the experimental setup:

Numerical Simulation: If noise sources are well-characterized, $\epsilon_2$ can be estimated via numerical simulations using the definition of operational distance or its upper bounds.
Tensor Product Decomposition: For $n$ -qubit systems where POVMs decompose into local tensor products, the global operational distance can be bounded by the sum of local operational distances, simplifying calculations.
Quantum Measurement Tomography: The actual POVM can be reconstructed from experimental statistics to bound the distance.
Direct Estimation: Using relations involving 2-designs or assuming rank-one projectors for the ideal POVM, $\epsilon_2$ can be bounded by estimating specific fidelity-like quantities experimentally.

A key feature of this method is that it does not require the measurements to be informationally complete (IC). This significantly reduces the experimental overhead, as IC-POVMs typically require an exponentially large number of settings.

Key Results and Applications

The authors validate the method through three numerical examples, demonstrating its necessity in avoiding incorrect conclusions:

Minimum Preparation Fidelity: The method was applied to certify the fidelity of 100 Haar-random two-qubit states under various noise models (amplitude damping, phase damping, depolarizing, and rotation shifts). The results show that the method provides non-trivial lower bounds for fidelity, with the tightness of the bounds depending on the noise intensity and the number of shots.
Magnetization in Spin Systems: In a study of a ferromagnetic system, the authors showed that neglecting measurement imperfections (specifically a rotation of the measurement basis) leads to the false conclusion that the system is not fully magnetized. While the method accounting for noise correctly identified the magnetization as 1, the method ignoring systematic errors suggested the magnetization was significantly lower or even negative, depending on the rotation angle and shot count.
Entanglement Witness: The protocol was tested on a two-qubit entanglement witness. The results demonstrated that ignoring systematic errors can lead to the false certification of entanglement for separable states. Conversely, the proposed method correctly identified the separable state as such, while still successfully certifying entanglement for maximally entangled states. Furthermore, the authors noted that methods ignoring imperfections often become infeasible (no solution exists) when noise or shot noise is large, whereas their method remains valid and yields certified bounds in all regimes.

Significance and Claims

The paper claims to provide a versatile and robust certification framework for quantum experiments that integrates both statistical and systematic errors. Its significance lies in several specific contributions:

Generalization of Confidence Regions: It extends existing confidence-region frameworks to include situations where measurement control is imperfect, a scenario common in near-term quantum experiments.
Robustness Against False Conclusions: By explicitly modeling measurement imperfections, the method prevents the erroneous conclusions (e.g., false entanglement or incorrect magnetization values) that arise when systematic errors are ignored.
Experimental Efficiency: The method does not require informationally complete measurements, allowing experimentalists to choose simpler measurement sets based on their specific setup constraints.
Computational Tractability: For convex or concave functions, the resulting optimization problem is convex and can be efficiently solved using interior-point methods.
Applicability: While demonstrated on fidelities, magnetization, and entanglement, the framework is applicable to a broad class of convex quantities, including energies, entropies, and quantum Fisher information.

The authors conclude that this approach offers a practical pathway to obtaining rigorous guarantees in realistic settings where devices are noisy and calibration is limited, which is essential for the reliable development of quantum technologies.