⚛️ quantum physics

Efficient Sampling for Many-Body Fermionic Non-Gaussianity

This paper introduces an efficient perfect-sampling method based on matrix product states to compute the second-order magic Rényi entropy for many-body fermionic systems, enabling the scalable evaluation of non-Gaussianity and higher-order correlations in large systems like the XXZ chain.

Original authors: Ryota Matsuda, Masahiro Hoshino, Yuto Ashida

Published 2026-10-05

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Ryota Matsuda, Masahiro Hoshino, Yuto Ashida

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 microscopic world of quantum physics, particles called fermions—such as electrons—can arrange themselves in two distinct ways. Sometimes they behave like a calm, predictable fluid where every part is perfectly correlated with every other part in a simple, linear fashion. Physicists call this a "Gaussian" state, and because the rules are so straightforward, even the most powerful classical computers can simulate these systems with ease. But when these particles interact in complex, chaotic ways, they enter a different realm. Here, the correlations become intricate and layered, defying simple description. This "non-Gaussian" behavior is the secret ingredient that makes quantum computers potentially more powerful than classical ones, yet it is also the most difficult thing to measure. For years, scientists have struggled to quantify this complexity in large systems, often hitting a wall where the computational cost becomes so high that it grows too fast to be practical, effectively limiting our view to only the tiniest collections of particles.

A team of researchers at the University of Tokyo has now developed a new method to peer past this wall. They created a technique that allows them to calculate a specific measure of this quantum complexity, known as the magic Rényi entropy, for systems containing up to 128 sites. This is a significant leap forward, as previous methods could only handle systems small enough to fit on a standard desktop computer. The researchers achieved this by designing a clever sampling strategy that works directly with a mathematical representation of the quantum state called a matrix product state. Instead of trying to calculate the entire system at once—a task that would require impossible amounts of memory—their method generates individual snapshots of the system's behavior. By carefully controlling the statistical fluctuations in these snapshots, they can reconstruct the overall picture of the system's complexity with high precision.

The team tested their method on a model of interacting electrons arranged in a chain, a system known as the XXZ chain. They pushed the simulation to include chains with as many as 128 sites, a scale that was previously out of reach for this specific type of calculation. At one extreme of the system's behavior, where the particles align in a specific magnetic order, their results matched perfectly with known theoretical predictions, confirming that their method works correctly. More importantly, in other regions of the system where the particles interact more strongly, their method revealed a new phenomenon. They observed that the measure of quantum complexity stops growing and levels off, or saturates, as the chain gets longer. This saturation happens at a size that is too large for older, exact calculation methods to reach.

This finding is crucial because it shows that the complexity of the system is not just a simple sum of its parts. The researchers found that this leveling off coincides with the formation of a specific pattern in how the particles are correlated across the chain, a pattern that looks like a wall separating two different magnetic regions. Traditional tools that only look at simple, two-point connections between particles would miss this entirely, as they cannot see these higher-order, multi-particle relationships. The new method, however, captures these intricate connections, proving that the quantum complexity of the system is deeply tied to these large-scale structures.

The implications of this work extend beyond just this one model. The method provides a practical tool for scientists to study how quantum resources evolve in large systems, whether they are in a stable ground state or changing over time. Because the measure they calculate can also be observed in real-world experiments using multiple copies of a quantum state, these numerical results can be directly compared with data from future quantum simulators. This bridges the gap between theoretical prediction and experimental reality, offering a way to benchmark and understand the non-Gaussian resources that will be essential for building universal quantum computers. By making it possible to quantify these complex correlations in systems of realistic size, the researchers have opened a new window into the behavior of many-body fermionic systems, turning a previously intractable problem into a manageable calculation.

Technical Summary: Efficient Sampling for Many-Body Fermionic Non-Gaussianity

Problem Statement
Fermionic non-Gaussianity is a critical resource for universal quantum computation and characterizes correlations beyond mean-field or Gaussian descriptions in interacting electronic systems. While covariance-matrix-based measures offer computationally efficient diagnostics, they fail to capture higher-order correlations essential in strongly correlated regimes. A more rigorous resource measure, the magic Rényi entropy (MRE), has been proposed based on convolution operations. However, the exact evaluation of the MRE is limited to small system sizes because the number of higher-order correlations grows exponentially with system size, and direct calculation via matrix product states (MPS) involves prohibitive costs (e.g., contracting transfer matrices of dimension $O(D^8)$ ).

Methodology
The authors develop a perfect-sampling method to calculate the second-order MRE ( $M_2$ ) from matrix product states (MPS) without explicitly constructing the convolved state or the replicated tensor network. The core of the method involves two key innovations:

Bounded Estimator via Rao-Blackwellization:
The purity $\nu_2$ (from which $M_2 = -\ln \nu_2$ is derived) is recast as an expectation value over Fock-basis samples of the output state. The authors define an amplitude ratio $r(X)$ associated with exchanging output configurations. Since the unbiased estimator $r(X)$ is unbounded and prone to large fluctuations, they apply Rao-Blackwellization. By averaging $r(X)$ over exchange pairs with conditional weights, they construct a bounded estimator $h(X)$ such that $-1 \le h(X) \le 1$ . This eliminates large fluctuations caused by small denominators while preserving the mean and reducing variance.
Recursive Perfect Sampling from MPS:
To sample from the Born distribution $|\Phi|^2$ and evaluate the necessary amplitudes without constructing the full output MPS, the authors employ a recursive contraction scheme.
- They represent the input state as a normalized right-canonical MPS.
- They introduce two parity-resolved boundary matrices that encode the sampled prefix.
- Using a sweep algorithm (Algorithm 1), they compute conditional probabilities for the next site's output configuration by contracting local MPS tensors with the beam-splitter unitary and the boundary matrices.
- This process generates exact Born samples and evaluates prescribed amplitudes (including "crossed" amplitudes needed for the estimator) with a computational cost of $O(nD^3)$ per sample, where $n$ is the number of modes and $D$ is the MPS bond dimension.
- The method can be further optimized by exploiting particle-number conservation (U(1) symmetry), restricting calculations to allowed charge sectors.

Key Results
The method was benchmarked on a spinful fermion XXZ chain with up to 128 sites:

Ferromagnetic Point ( $\Delta = -1$ ): The sampled $M_2$ agrees perfectly with an analytically derived asymptotic formula for the Dicke state across all system sizes up to $L=128$ .
Beyond Ferromagnetic Point ( $\Delta < -1$ ): The method reveals a saturation of the MRE at system sizes ( $L > 32$ ) inaccessible to exact diagonalization. This saturation coincides with the emergence of a domain-wall profile in spin-spin correlation functions.
Comparison with Covariance Measures: The study demonstrates that while covariance-matrix-based measures (like fermionic antiflatness) yield identical values for the entire family of ground states (due to spin-flip symmetry and particle-number conservation), the MRE successfully distinguishes these states and captures the transition to domain-wall formation.
Computational Efficiency: The sampling cost scales as $O(N n D^3)$ , where $N$ is the number of samples. For extensive MRE, the required sample count scales as $e^{M_2}$ , which, while exponential in the resource value, remains feasible for system sizes up to $\sim 100$ sites. For $L=128$ at $\Delta=-1$ , approximately $1.4 \times 10^7$ samples were sufficient for a standard error of 0.01.

Significance and Claims
The paper claims to provide a practical computational tool that quantitatively evaluates many-body fermionic non-Gaussianity by accounting for higher-order correlations. The significance of this work lies in:

Scalability: Enabling the study of non-Gaussianity in systems with up to 128 sites, a scale previously unreachable for MRE evaluation.
Physical Insight: Demonstrating that the MRE can resolve intricate many-body correlations (such as domain-wall formation) that are invisible to Gaussian-based diagnostics.
Experimental Relevance: Since the MRE is defined via purity and can be measured experimentally (e.g., via SWAP tests or interference of two copies), the numerical results provide a direct benchmark for fermionic quantum simulators.
General Applicability: The method is applicable to both equilibrium and non-equilibrium states represented by MPS, allowing for the tracking of non-Gaussian resource evolution after quantum quenches.

The authors emphasize that their approach avoids the explicit construction of the beam-splitter output MPS, thereby keeping the sampling cost to a low-order polynomial in system size and bond dimension, making it a viable alternative to exact methods for large-scale fermionic systems.

More like this