⚛️ quantum physics

Third-Order Quantum Information from Distributed Randomized Measurements

This paper demonstrates that third-order quantum information from distributed randomized measurements can be efficiently reconstructed by exploiting permutation symmetries to reduce the problem to inverting a 3x3 matrix, thereby avoiding the exponential complexity of full density matrix reconstruction.

Original author: Giovanni Scala, Corrado Cosimo Mattiacci, Gniewomir Sarbicki

Published 2026-11-11

📖 4 min read🧠 Deep dive

CC BY 4.0

Original author: Giovanni Scala, Corrado Cosimo Mattiacci, Gniewomir Sarbicki

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 author. For technical accuracy, refer to the original paper. Read full disclaimer

In the quantum world, the most complete description of a system is a mathematical object called a density matrix. This object contains every possible detail about the state of particles, from their energy to their correlations with one another. However, for anything but the tiniest systems, writing down or measuring this full object is impossible. The number of details required grows so fast that even the most powerful computers would struggle to hold the information for a modest collection of particles. Because of this, physicists have long sought ways to learn about quantum states without reconstructing the entire picture. They look for specific, useful properties—like how much two particles are linked or how complex a state is—that can be found by taking smaller, simpler measurements. One powerful method involves taking many snapshots of a system while randomly changing how it is observed. By averaging the results of these random guesses, scientists can extract hidden, nonlinear information about the state, such as how the probabilities of different outcomes multiply together. This approach avoids the need to know the exact settings of every measurement, relying instead on the statistical power of randomness.

A team of researchers has now taken this idea of randomized measurements and pushed it to a new level of complexity and clarity. They investigated what happens when you take three separate copies of a quantum state, measure them all at the same time using the same random settings, and then look at the product of the three results. While previous work had shown how to extract information from pairs of measurements, this study focused on the third-order relationships that emerge when three copies are involved. The researchers asked a fundamental question: if you have a collection of distant laboratories, each performing these random measurements on their own piece of a shared quantum system, what specific information can they recover just by comparing their notes? They wanted to know exactly which combinations of the state's properties are visible through this method and which ones remain forever hidden.

The team discovered that the answer depends on the symmetries of the measurement process. When the random settings are averaged out, the data does not reveal every single detail of the quantum state. Instead, it reveals specific groupings of information that remain unchanged no matter how the local observers rotate their measurement devices. The researchers found that while the number of possible outcomes grows explosively as more parties are added to the experiment, the underlying structure of the recoverable information is surprisingly simple. They proved that the complex problem of untangling these measurements can be broken down into a series of small, independent puzzles. Each of these puzzles is no larger than a simple three-by-three grid of numbers. This means that regardless of whether there are two parties or twenty, the mathematical work required to reconstruct the available information never exceeds the difficulty of inverting one of these tiny grids.

This finding is significant because it removes a major barrier to analyzing large quantum systems. In the past, trying to reconstruct information from distributed measurements often required solving massive, unwieldy equations that became impossible to handle as the system grew. The new method shows that the information is not lost, but rather organized in a way that allows it to be retrieved efficiently. The researchers demonstrated that the only information that cannot be recovered is the difference between two specific types of three-cycle patterns in the data. These patterns correspond to subtle distinctions in how the quantum state behaves under a mathematical operation called partial transposition, which is often used to detect entanglement. The averaging process naturally blends these two patterns together, making them indistinguishable. However, the study proves that everything else—the sum of these patterns and all other third-order properties—can be perfectly reconstructed.

The protocol described is practical for real-world experiments because it does not require the laboratories to keep a record of every single random rotation they apply. They only need to know which measurements were performed at the same setting. This eliminates the need for complex calibration and storage of the random unitary matrices, making the process much more economical and easier to implement in distributed networks. The researchers showed that by simply counting how often certain patterns of outcomes appear together, one can calculate the exact values of the retained quantum properties. The work provides a complete map of what is accessible and what is not, clarifying the limits of what can be learned from these distributed, randomized measurements. It confirms that while some fine details are inevitably washed out by the averaging, the core nonlinear information of the quantum state remains intact and can be extracted with a level of mathematical simplicity that defies the apparent complexity of the system.

Technical Summary: Third-Order Quantum Information from Distributed Randomized Measurements

Problem Statement
Quantum state characterization often relies on nonlinear functions of the density matrix, such as moments of the partial transpose or purity, which are difficult to access without full state tomography. While randomized measurements offer a pathway to probe these nonlinear properties by averaging products of outcome probabilities over random bases, a critical challenge remains for distributed systems: determining exactly which third-order local-unitary invariant combinations can be reconstructed from local measurements and classical post-processing. Specifically, when three independent copies of a multipartite state are measured at the same local random settings and averaged over Haar distributions, the resulting correlators lose information regarding the relative orientation of permutation cycles. The paper addresses the identifiability problem: which invariant combinations are retained by these averages, and how can they be reconstructed without inverting exponentially large matrices that typically arise in multipartite systems?

Methodology
The authors propose a protocol involving $n$ parties, each applying a local unitary $U_k$ drawn from a Haar-random distribution (or an exact 3-design) and performing projective measurements in the computational basis on three independently prepared copies of the state $\rho$ . The core of the methodology involves analyzing the third-order correlators formed by the product of three outcome probabilities, averaged over the local unitaries.

The reconstruction process is structured through several mathematical reductions:

Equality Patterns and Local Cycle Merging: The averaged data is first categorized by "equality patterns" of the outcome triples (e.g., all distinct, one pair equal, all equal). Using Weingarten calculus, the authors map these patterns to scalar invariants of the state. A key step is the "local cycle merging" map ( $C$ ), which sums the contributions of the two distinct three-cycle permutations ( $(123)$ and $(132)$ ) because the Haar average assigns them identical coefficients. This map reduces the dimensionality of the invariant space but inherently discards the difference between these cycle orientations.
Orbit Reduction: The protocol exploits the symmetry of simultaneous relabelling of the three copies across all parties. This action partitions the space of contraction labels and equality patterns into orbits. By averaging over these orbits, the authors reduce the number of independent coordinates from an exponential size to a smaller set $N_n$ (scaling as $5^n/6$ ).
Sector Decomposition: To solve the reconstruction problem analytically, the authors decompose the invariant subspace into tensor sectors based on local invariant ( $V$ ) and zero-sum ( $W$ ) subspaces. This decomposition reveals that the global reconstruction map, despite its high dimension, is block-diagonal. Each block corresponds to a specific sector and acts independently.

Key Results
The central result is Theorem 1, which provides an exact characterization of the reconstructible information. The theorem demonstrates that the reconstruction map for an arbitrary number of parties and local dimensions can be inverted by decomposing it into independent sectors. Crucially:

Dimensionality Reduction: The inversion of the full reconstruction matrix, which would scale exponentially with the number of parties, collapses to the symbolic inversion of $3 \times 3$ matrices (specifically the one-party map $H(d)$ ) and scalar weights.
Reconstructible Invariants: The protocol recovers all combinations of third-order invariants retained by the cycle-merging map. For a bipartite system, this includes the sum of the global third moment and the partial-transpose third moment ( $Tr \rho^3 + Tr[(\rho^{\Gamma_B})^3]$ ), but explicitly not their difference.
Information Limits: The paper proves that the inability to separate the two three-cycle orientations (e.g., distinguishing $Tr \rho^3$ from $Tr[(\rho^{\Gamma_B})^3]$ in the bipartite case) is a fundamental consequence of the common-setting Haar averaging. The protocol requires only that the same unknown unitary acts on each copy; it does not require recording the specific unitary matrices, distinguishing it from classical shadow protocols that track settings.

Significance and Claims
The paper claims to provide an exact characterization of the nonlinear information available from distributed randomized measurements. Its primary significance lies in solving the reconstruction problem for arbitrary multipartite systems without the computational bottleneck of inverting exponentially large matrices. By reducing the problem to the symbolic inversion of $3 \times 3$ blocks, the method makes the extraction of third-order information feasible for systems with many parties.

The authors emphasize that their approach removes the need to characterize or store individual random rotation matrices, relying only on identifying shots taken at the same setting. However, the paper is modest regarding statistical efficiency; it notes that while the algebraic reconstruction is tractable, the statistical cost (sample complexity) and conditioning of the estimators for the retained combinations remain open tasks. Furthermore, the work establishes a clear boundary on what information is accessible: the protocol recovers exactly what survives the cycle-merging operation, and no more, as the difference between specific cycle orientations is fundamentally lost in the averaging process.

Technical Summary: Third-Order Quantum Information from Distributed Randomized Measurements

More like this