⚛️ quantum physics

Minimal Projective Measurement Settings for Estimating Multiple Observables

This paper establishes structural bounds and employs a classical variational algorithm to determine the minimum number of orthogonal projective measurement settings required for unbiased estimation of multiple observables, demonstrating that shared measurement bases can significantly improve precision, as certified by a four-setting design for two-qubit moments that nearly achieves the theoretical variance lower bound.

Original authors: Yibin Huang, Muchun Yang, D. L. Zhou

Published 2026-10-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yibin Huang, Muchun Yang, D. L. Zhou

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

In the quantum world, scientists often need to know the average behavior of many different properties at once. Imagine trying to describe a complex object by measuring its color, weight, temperature, and texture all at the same time. In quantum physics, these properties are called observables, and measuring them is the only way to understand the state of a system, such as a group of tiny particles known as qubits. The challenge is that quantum mechanics has a strict rule: you cannot measure certain properties simultaneously with perfect precision. To get around this, researchers usually have to rotate their measuring device, take a reading, rotate it again, take another reading, and repeat this process many times. Each rotation and measurement is a "setting," and the more settings required, the more time and resources are wasted. The goal for any experimentalist is to find the most efficient way to gather all the necessary information with the fewest possible settings, ensuring that no data is thrown away and no unnecessary steps are taken.

A team of researchers has tackled this problem by developing a new method to determine the absolute minimum number of measurement settings needed to estimate multiple quantum properties without any bias. They focused on a specific type of measurement called an orthogonal projective measurement, which is a standard tool in quantum experiments. Instead of guessing or using trial and error, they combined a powerful computer search with rigorous mathematical proofs to find the exact answer for specific groups of properties. Their work shows that for certain tasks, the number of settings required is far lower than previously thought possible, and that sharing the same measurement setup across different properties can dramatically improve the precision of the results.

The researchers began by asking a fundamental question: given a specific list of properties to measure, what is the smallest number of measurement settings that can capture all of them? They discovered that simply counting the number of properties is not enough to determine the answer; the internal structure of the properties matters just as much. To solve this, they created a two-part strategy. First, they used a classical computer algorithm to search for the best possible measurement settings and to figure out how to split the available data among them. Second, they developed mathematical "witnesses"—rigorous tests that can prove it is impossible to do the job with fewer settings than a certain number. By matching the best computer-found solution with these mathematical proofs, they could certify the exact minimum number of settings required for specific tasks, moving beyond simple estimates.

One of their most striking findings involved the collective behavior of just two qubits. They wanted to measure the first and second moments of their spin, which describes how the particles are oriented and how they fluctuate. While a simple count of the properties suggested that three settings might be enough, their new mathematical proof showed that three settings are actually insufficient. They proved that four settings are strictly necessary and sufficient to measure these nine properties for any possible input state. Furthermore, they designed a specific four-setting plan that is nearly perfect. When they tested this plan, the statistical error was only about 0.0341% higher than the theoretical best possible error, a margin so small it is almost negligible. This result certifies that for this specific task, no one can do better than four settings, and their design comes incredibly close to the ideal limit.

The team also tested their method on a more complex system involving four qubits and thirty different properties. In this case, they found that exactly three settings were enough to cover all the targets, a number that was confirmed by both their computer search and their mathematical proofs. To show the power of this approach, they compared their shared three-setting design against several other common methods. They found that their method was vastly more efficient. For the same level of precision, other methods would require about 4.1 times as many copies of the quantum state. In other words, if a standard method needed 410 measurements to get a clear answer, their shared design could achieve the same clarity with only 100. This demonstrates that by carefully choosing measurement bases that exploit the structure of the properties, researchers can save a tremendous amount of experimental time and resources.

The study also looked at what happens when the measurements are not perfect, which is often the case in real-world experiments where circuits have limited depth and noise is present. They simulated a scenario with six qubits and ninety-six properties, using shallow circuits that might leave a small amount of error. Even with this imperfection, their shared design outperformed other methods for a wide range of sample sizes. The results showed that while other methods might eventually catch up if an infinite number of measurements were taken, the shared design provided a significant advantage for any practical, finite number of measurements. This suggests that the strategy of sharing measurement settings is robust and useful even when experimental conditions are not ideal.

Ultimately, this work provides a clear roadmap for quantum experiments. It establishes that the minimum number of settings is not just a matter of counting, but depends on the deep mathematical relationships between the properties being measured. By combining computer optimization with rigorous proofs, the researchers have shown that it is possible to find the exact minimum number of settings for specific tasks and to design measurement plans that are nearly as efficient as physics allows. Their findings offer a practical way to reduce the cost and time of quantum experiments, making it easier to study complex quantum systems with greater precision.

Technical Summary: Minimal Projective Measurement Settings for Estimating Multiple Observables

Problem Statement
The paper addresses the fundamental challenge of estimating the expectation values of multiple Hermitian observables ( $O_1, \dots, O_M$ ) for arbitrary input quantum states. In quantum information experiments, each measurement setting consists of applying a fixed unitary transformation $U_i$ followed by an orthogonal projective measurement in the computational basis. While commuting observables can be estimated from a single setting, non-commuting observables typically require multiple settings. The authors seek to determine the minimum number of orthogonal projective measurement settings ( $K$ ) required to estimate a specific set of traceless target observables without bias, and to optimize the statistical precision (variance) achievable with a finite number of state copies ( $N$ ).

Methodology
The authors propose a hybrid approach combining structural theoretical bounds with classical variational optimization:

Variational Optimization Framework:
- Objective: Jointly optimize the measurement unitaries ( $U_i$ ), estimator coefficients ( $C$ ), and sample allocation ( $p_i$ ) to minimize the mean squared error (MSE) or variance.
- Algorithm: A classical variational algorithm parameterizes unitaries as $U_i = \exp(iH_i)$ (or via variational circuits) and solves for linear coefficients using singular value decomposition (SVD) at each step.
- Bias Handling: The framework accommodates both exact reconstruction (zero bias) and approximate reconstruction (finite bias), where the MSE is defined as the sum of variance and squared reconstruction bias.
Structural Lower Bounds and Witnesses:
- Dimension Counting: A basic lower bound is derived by comparing the dimension of the target space ( $m$ ) against the capacity of $K$ settings ( $K(d-1)$ ).
- Structural Bound (Theorem 1): A tighter bound is derived based on the internal structure of the target space $V$ relative to a larger space $W$ (e.g., defined by symmetries). This bound accounts for the dimension of the internal center and the maximum number of commuting directions, ruling out certain values of $K$ even if dimension counting permits them.
- Alternating-Trace Witness (Theorem 2): A specific witness based on alternating traces ( $\tau_q$ ) is constructed to distinguish between 2 and 3 settings. If a non-zero value is found for odd $q \ge 3$ , it proves that at least three settings are necessary.
Certification of Minimums:
- The authors use the variational search to find an upper bound (a feasible decomposition with $K$ settings) and the structural bounds/witnesses to establish a lower bound. When these bounds match, the exact minimum number of settings is certified.

Key Results

Two-Qubit Collective Moments: For the task of estimating the collective first and second moments of two qubits (9 observables), the authors prove that four settings are necessary and sufficient.
- The structural bound (Theorem 1) excludes any solution with three global bases.
- A construction using four mutually unbiased bases (MUBs) in the triplet sector achieves the task.
- A numerically optimized four-setting design achieves a Haar-averaged variance within 0.0341% of a theoretical lower bound valid for any number of global projective settings in the unbiased linear estimation model.
Four-Qubit Observables (30 Targets):
- For a set of 30 observables on four qubits, dimension counting suggests $K \ge 2$ , but the cubic witness (Theorem 2) proves $K \ge 3$ .
- The variational algorithm successfully finds a decomposition with three settings, certifying $K^* = 3$ .
- Precision Gain: Compared to "spectral reuse" (using every target's eigenbasis independently), the shared three-setting design reduces the variance constant by a factor of approximately 4.1. This means the shared design requires 4.1 times fewer state copies to achieve the same average variance.
Six-Qubit Finite-Depth Circuits:
- The authors examine scenarios where measurement circuits have finite depth, leading to small reconstruction biases.
- For 96 observables on six qubits, shared measurement designs with optimized shallow circuits ( $L=8$ blocks) demonstrate lower Root Mean Squared Error (RMSE) than comparison protocols (including spectral reuse and global shadows) for a range of sample budgets.
- A "crossover" point is identified ( $N \times \approx 1.75 \times 10^5$ ); below this sample count, the shared design with bias is more efficient. Above this count, protocols with zero bias (like spectral reuse) become more efficient as the shared design's error approaches a residual floor determined by circuit depth.

Significance and Claims
The paper claims to provide a rigorous method for certifying the exact minimum number of measurement settings required for specific target spaces, moving beyond simple dimension counting. By combining structural bounds with variational constructions, the authors demonstrate that shared measurement bases can significantly outperform protocols that treat observables independently or use standard shadow tomography techniques.

Key contributions include:

Certification of Minimality: Providing exact minimum setting counts (e.g., 4 for two-qubit moments, 3 for specific four-qubit sets) where previous methods might only offer loose bounds.
Precision Optimization: Demonstrating that optimizing shared bases can reduce the sample complexity by factors of 4 to 6 compared to standard spectral reuse or Pauli-group-based methods.
Robustness to Bias: Showing that shared measurement strategies remain advantageous even when constrained by finite-depth circuits, provided the sample size is within a specific regime where the reduction in variance outweighs the introduction of reconstruction bias.

The authors note that while their method works for selected target spaces, determining tighter structural bounds for general sets of observables and extending the alternating-trace witnesses to rule out higher numbers of settings remain open questions. They also acknowledge that the classical optimization does not guarantee a global minimum for the unitary parameters due to the non-convex nature of the search.

Technical Summary: Minimal Projective Measurement Settings for Estimating Multiple Observables

More like this