On the injective norm of random fermionic states and skew-symmetric tensors

Basic Information

• Paper ID: 2510.25474
• Title: On the injective norm of random fermionic states and skew-symmetric tensors
• Authors: Stéphane Dartois (École Polytechnique), Parham Radpay (Université Paris-Saclay, CEA)
• Classification: math.PR (Probability Theory), math-ph (Mathematical Physics), math.MP (Mathematical Physics), quant-ph (Quantum Physics)
• Publication Date: October 29, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.25474v1

Abstract

This paper investigates the injective norm of random skew-symmetric tensors and associated fermionic quantum states, which provide a natural measure of many-body entanglement in systems of indistinguishable particles. Extending recent developments on random quantum states, the authors analyze the behavior of real and complex skew-symmetric Gaussian ensembles in two asymptotic regimes: fixed particle number with single-particle Hilbert space dimension tending to infinity, and a two-scale limit maintaining fixed filling fraction. Using the Kac-Rice formula on Grassmann manifolds, the authors derive high-probability upper bounds on the injective norm and establish sharp asymptotic behavior in both regimes. Notably, they discover a duality relation under particle-hole transformation, revealing symmetry of the injective norm under the action of the Hodge star operator. Numerical simulations verify theoretical predictions for low fermionic numbers.

Research Background and Motivation

Problem Statement

This paper addresses the geometric entanglement measure of random fermionic states, specifically focusing on the injective norm of skew-symmetric tensors and its asymptotic behavior. This is a central problem in quantum information theory for measuring many-body entanglement in systems of indistinguishable particles.

Problem Significance

1. Foundations of Quantum Information Theory: The Geometric Measure of Entanglement (GME) is an important tool in quantum information for characterizing many-body entanglement, crucial for understanding quantum computation and quantum communication.
2. Specificity of Fermionic Systems: For fermions (particles obeying the Pauli exclusion principle), traditional entanglement measures require modification due to particle indistinguishability. Skew-symmetric tensors naturally characterize fermionic states.
3. Extension of Random Matrix Theory: The injective norm represents a generalization of operator norms from random matrices to tensors. Studying its behavior under different symmetry classes is an important direction in random tensor theory.
4. Connection to Disordered Systems: This problem is closely related to ground state studies in disordered systems such as spin glasses, with significance in statistical physics.

Limitations of Existing Methods

1. Incomplete Symmetry Classes: Existing random tensor research primarily focuses on symmetric or unstructured tensors, with limited investigation of skew-symmetric tensors (corresponding to sign representations of $S_p$).
2. Technical Challenges: The structural zeros in skew-symmetric tensors make direct application of existing techniques (such as Subag's TAP method) difficult.
3. Specificity of Fermionic Entanglement: Strictly separable skew-symmetric states do not exist, requiring redefinition of separable state sets (using wedge products).

Research Motivation

This paper aims to:

• Extend the analysis from DM24 on random states of distinguishable particles to the fermionic case
• Systematically study the injective norm of skew-symmetric Gaussian tensor ensembles
• Establish precise upper bounds in two asymptotic regimes (fixed $p$, $d \to \infty$ and two-scale $p, d \to \infty$)
• Reveal deep symmetries such as particle-hole duality

Core Contributions

1. Established sharp asymptotic upper bounds for injective norm of skew-symmetric tensors: For both real and complex cases, in two asymptotic regimes (fixed fermionic number $p$ with $d \to \infty$; two-scale $p = \lfloor \alpha d \rfloor$, $d \to \infty$), high-probability bounds are obtained (Theorem 1.2).
2. Introduced new random matrix ensembles: Defined the Block Hollow Gaussian Antisymmetric Ensemble (BHGAE) and its complex version (cBHGAE), proving operator norm bounds (Propositions 2.8, 2.10).
3. Discovered particle-hole duality: Proved that geometric entanglement of fermionic states is invariant under Hodge duality (particle-hole transformation) (Proposition 2.5), manifesting as $\alpha \leftrightarrow 1-\alpha$ symmetry in the two-scale limit.
4. Applied Kac-Rice formula on Grassmann manifolds: This represents a novel application of this technique, overcoming technical difficulties arising from structural zeros in skew-symmetric tensors.
5. Numerical verification: Numerical simulations for $p = 2, 3, 4$ verify theoretical predictions of upper bounds.

Detailed Methodology

Task Definition

Input: Random skew-symmetric tensor $T \in \bigwedge^p \mathbb{K}^d$ ($\mathbb{K} = \mathbb{R}$ or $\mathbb{C}$), with independent elements following standard Gaussian distribution.

Output: Asymptotic behavior of injective norm $\|T\|_{inj} := \max_{x_1,\ldots,x_p \in \mathbb{K}^d, \|x_1\wedge\cdots\wedge x_p\|_2=1} |\langle T, x_1\wedge\cdots\wedge x_p\rangle|$

Key Constraints:

1. Skew-symmetry: $T_{i_{\sigma(1)},\ldots,i_{\sigma(p)}} = \text{sg}(\sigma)T_{i_1,\ldots,i_p}$
2. Normalized fermionic state: $|\psi_f\rangle = T/\|T\|_2$
3. Optimization domain is Grassmann manifold $\text{Gr}(p,d)$

Model Architecture

1. Equivalent Characterization of Injective Norm

Using the geometric structure of Grassmann manifolds, the optimization problem is transformed to: $\|T\|_{inj} = \max_{x^{(1)}\wedge\cdots\wedge x^{(p)} \in \text{Gr}(p,d)} \frac{|\langle T, x^{(1)}\wedge\cdots\wedge x^{(p)}\rangle|}{\sqrt{|\det(X^tX)|}}$ where $X$ is a $d \times p$ matrix with columns $x^{(i)}$.

2. Kac-Rice Formula Application

Define random field: $f_T(x^{(1)}\wedge\cdots\wedge x^{(p)}) = \frac{1}{\sqrt{p(d-p)}} \frac{\langle T, x^{(1)}\wedge\cdots\wedge x^{(p)}\rangle}{\|x^{(1)}\wedge\cdots\wedge x^{(p)}\|_2}$

Apply Kac-Rice formula to estimate critical points: $\mathbb{E}[\text{Crt}_{f_T}(B)] \leq K(p,d) \int_B \sqrt{p(d-p)} e^{-\frac{p(d-p)}{2}u^2} \mathbb{E}_{\text{BHGAE}}[|\det(u-W_N)|] du$

where $W_N$ is a BHGAE random matrix with $N = p(d-p)$.

3. Block Hollow Gaussian Antisymmetric Ensemble (BHGAE)

Define an $N \times N$ random matrix $W_N$ with block structure of $p \times p$ blocks, each of size $d \times d$:

• Diagonal blocks are zero
• Off-diagonal blocks $(i,j)$ are skew-symmetric Gaussian random matrices $G_{ij}$ (variance $\sigma^2 = \frac{1}{p(d-p)}$)
• Satisfy block skew-symmetry: $(W_N)_{(i,\alpha),(j,\beta)} = -(W_N)_{(j,\beta),(i,\alpha)}$

Key Properties:

• Operator norm bound (Proposition 2.8): For fixed $p$, $d \to \infty$, $\mathbb{P}\left(\|W_N\| > 2\sqrt{\frac{p-1}{p}} + \varepsilon\right) \leq e^{-N^{1-\delta}}$
• Empirical spectral measure converges to semicircle law: $\rho_p(x) = \frac{p}{2\pi(p-1)}\sqrt{4\frac{p-1}{p} - x^2}$

Technical Innovations

1. Normal Coordinates on Grassmann Manifolds

Near reference point $P_0 = \begin{pmatrix} I_p \\ 0 \end{pmatrix}$, use coordinates $B \in \mathbb{R}^{(d-p)\times p}$ from the tangent space: $\phi(B) = \tilde{B}(\tilde{B}^t\tilde{B})^{-1}\tilde{B}^t, \quad \tilde{B} = \begin{pmatrix} I_p \\ B \end{pmatrix}$

This enables covariance function computation in Euclidean space (Appendix B).

2. Matrix Dyson Equation

Exploit Kronecker structure: $W_N = \frac{1}{\sqrt{2p(d-p)}} \sum_{1\leq i<j\leq p} \beta_{ij}^{(p)} \otimes G_{ij} + (\beta_{ij}^{(p)})^t \otimes G_{ij}^t$ where $\beta_{ij}^{(p)} = E_{ij}^{(p)} - E_{ji}^{(p)}$.

Apply local law from AEKN19 to obtain Stieltjes transform: $m_p(z) = \frac{-z + \sqrt{z^2 - 4(p-1)/p}}{2(p-1)/p}$

3. Laplace Method and Complexity Function

Define complexity function (fixed $p$ case): $\Sigma_R^{(p)}(u) = \frac{1+\log p}{2} - \frac{u^2}{2} + \Omega\left(u\sqrt{\frac{p-1}{p}}\right) + \frac{1}{2}\log\frac{p-1}{p}$

where $\Omega$ is the logarithmic potential of the semicircle law. Solving $\Sigma_R^{(p)}(E_0(p))=0$ gives the asymptotic upper bound.

Two-scale case ($p = \lfloor \alpha d \rfloor$): $\gamma_\alpha(d) = \sqrt{\log d} + \frac{\log\log d}{2\sqrt{\log d}} + \frac{\beta(\alpha)}{\sqrt{\log d}} + o\left(\frac{1}{\sqrt{\log d}}\right)$ where $\beta(\alpha) = \frac{3}{4} + \frac{1}{4}\frac{\alpha}{1-\alpha}\log\alpha + \frac{1}{4}\frac{1-\alpha}{\alpha}\log(1-\alpha) + \frac{1}{2}\log(\alpha(1-\alpha))$

4. Manifestation of Particle-Hole Duality

Hodge duality $*: \bigwedge^p \mathbb{R}^d \to \bigwedge^{d-p} \mathbb{R}^d$ preserves injective norm (Proposition 2.5). In the two-scale limit, $\beta(\alpha)$ is symmetric about $\alpha = 1/2$ (Figure 1), embodying the $\alpha \leftrightarrow 1-\alpha$ symmetry.

Experimental Setup

Numerical Methods

Use methods from FLN22 for numerical verification:

Case $p=2$: Injective norm equals the maximum singular value of the corresponding skew-symmetric matrix, computed exactly via SVD.

Case $p \geq 3$: Use gradient descent to approximate optimization: $\max_{x_1,\ldots,x_p} \frac{|\langle T, x_1\wedge\cdots\wedge x_p\rangle|}{\|x_1\wedge\cdots\wedge x_p\|_2}$

Parameter Settings

• Tensor Order: $p = 2, 3, 4$
• Dimension Range:
  • $p=2$: $d \in [10, 500]$
  • $p=3$: $d \in [10, 350]$ (real), $d \in [10, 300]$ (complex)
  • $p=4$: $d \in [10, 120]$
• Sample Sizes:
  • $p=2$: 200 samples (all $d$)
  • $p=3$: 100 samples for $d \leq 250$, 20 samples for $d > 250$
  • $p=4$: 100 samples for $d \leq 70$, 10 samples for $d > 70$

Observables

Normalized injective norm: $\frac{\|T\|_{inj}}{\sqrt{p(d-p)}}$

Theoretical predicted upper bounds:

• $p=2$: $\alpha(2) = 2$
• $p=3$: $\alpha(3) \approx 2.870$
• $p=4$: $\alpha(4) \approx 3.588$

Experimental Results

Main Results

1. Convergence for Fixed $p$, $d \to \infty$

Figures 2(a-c) show results for real case with $p=2,3,4$:

• $p=2$: Numerical results rapidly converge to theoretical value 2, nearly perfect agreement at $d=500$
• $p=3$: Converges to 2.87, with minor fluctuations at $d=350$ but clear trend
• $p=4$: Converges to 3.588, limited to $d=120$ due to computational constraints, but trend is clear

Key Observations:

• Normalized injective norm remains bounded as $d$ increases in all cases
• Convergence speed decreases with increasing $p$ (larger $d$ needed to reach asymptotic regime)
• Sample variance significantly decreases with $d$ (self-averaging effect)

2. Verification for Complex Case

Figure 2(d) shows $p=3$ complex case:

• Converges to same upper bound 2.87 as real case
• Confirms Theorem 4.3: real and complex cases have identical asymptotic behavior after normalization
• Higher computational complexity, only to $d=300$

3. Comparison with Other Symmetry Classes

Figure 3 compares different symmetries for $p=3$:

• Skew-symmetric: Converges to 2.87
• Symmetric: Significantly larger values (around 3.2)
• Unstructured: Between the two

Figure 3(b) shows ratio evolution with $d$:

• Skew-symmetric/unstructured ratio approximately 0.85-0.90
• Skew-symmetric/symmetric ratio approximately 0.75-0.80
• Demonstrates significant impact of symmetry on injective norm

Experimental Findings

1. Tightness of Theoretical Bounds: Numerical results indicate theoretical bounds are asymptotically tight, with no systematic gaps.
2. Convergence Speed:
   • $p=2$ fastest ($d \sim 100$ reaches asymptotic regime)
   • $p=3$ requires $d \sim 300$
   • $p=4$ requires larger $d$ (>120)
3. Self-Averaging Phenomenon: In high dimensions, variance between different samples significantly decreases, indicating strong self-averaging property of injective norm.
4. Computational Bottleneck: For $p \geq 3$, computational cost of gradient descent grows rapidly with $d$ and $p$, limiting large-scale verification.

Related Work

Quantum Information and Entanglement Measures

• Shi95, WG03: Early work on geometric measure of entanglement
• AMM10a, AMM10b: Geometric entanglement and Majorana representation for symmetric states
• AS17: Quantum entanglement from Banach space perspective
• FLN22, SG24: Entanglement estimation for random many-body states

Fermionic Entanglement

• GKM12: Segre maps and entanglement for multipartite systems of indistinguishable particles
• BHK21: Page curve for fermionic Gaussian states
• DNT22: Entanglement criteria for bosonic and fermionic induced ensembles
• HW23, PS24: Entropy fluctuations in fermionic systems

Random Tensors and Injective Norm

• DM24: Injective norm of random states for distinguishable particles (directly extended in this work)
• BGJ+24: Geometric perspective on random tensor injective norm
• Boe24: Random tensors with variance profile
• BS25: Balanced multi-species spin glasses (matched lower bounds of DM24)

Random Matrix Theory

• AEKN19: Spectral localization for Kronecker random matrices (local law)
• BABM22, BBvH23: Random determinant growth beyond invariance
• McK24: Complexity of bipartite spin glasses

Kac-Rice Formula and Landscape Complexity

• ABAČ13: Kac-Rice formula application in spin glasses
• Sub23: TAP method for multi-species spherical spin glasses
• Sto25: Ground state energy for many-body $p$-spin models

Physical Methods

• Sas24, DLSST25: Using quantum field theory methods to study (sign) eigenvalue distribution of skew-symmetric tensors

Unique Contributions of This Work

Compared to existing work:

1. First systematic study of skew-symmetric tensors: Fills gap in symmetry classes
2. Kac-Rice on Grassmann manifolds: Technical innovation in application
3. Particle-hole duality: Reveals new symmetry
4. Exact asymptotics in two-scale regime: Complete characterization of both limits

Conclusions and Discussion

Main Conclusions

1. Establishment of Sharp Upper Bounds:
   • Fixed $p$, $d \to \infty$: $\|T\|_{inj}/\sqrt{p(d-p)} \lesssim E_0(p)$ (solution of equation 2.11)
   • Two-scale $p = \lfloor \alpha d \rfloor$, $d \to \infty$: $\|T\|_{inj}/\sqrt{d^2\alpha(1-\alpha)} \lesssim \gamma_\alpha(d) \sim \sqrt{\log d}$
2. Discovery of Duality: Geometric entanglement is invariant under particle-hole transformation, $\text{GME}(|\psi_f\rangle) = \text{GME}(|\psi_h\rangle)$
3. Unification of Real and Complex Cases: After normalization, both exhibit identical asymptotic behavior
4. Numerical Verification: Simulations for $p=2,3,4$ confirm theoretical predictions

Limitations

1. Only Upper Bounds: Unlike DM24, matching lower bounds have not been established. Techniques from BS25 do not directly apply to skew-symmetric case.
2. Challenge of Structural Zeros: BHGAE matrix diagonal blocks being zero means spectral gap property (Property 3) only holds for sufficiently large $|u|$, requiring conditional event $\mathcal{E}_K$.
3. Technical Limitations in Two-Scale Regime:
   • Only proved $\gamma_\alpha(d) = o(\exp(d^{1-\delta}))$ case
   • Applicability of Laplace method to more general sequences not fully established
4. Numerical Computation Limitations:
   • For $p \geq 3$ relies on gradient descent, may get stuck in local extrema
   • Memory and computational cost of higher-order tensors limits large-scale verification
5. Incompleteness of Symmetry Classes: Only studies completely skew-symmetric case; other irreducible representations of $S_p$ corresponding to tensor symmetry classes not addressed

Future Directions

1. Establishment of Lower Bounds:
   • Use second moment method or other techniques to prove matching lower bounds
   • Establish optimality of $E_0(p)$ and $\gamma_\alpha(d)$
2. Other Symmetry Classes:
   • Study tensors corresponding to general Young tableaux of $S_p$
   • Establish systematic relationship between symmetry classes and injective norm
3. Physical Applications:
   • Apply results to practical entanglement measurement in fermionic systems
   • Connect to topological order and other condensed matter physics problems
4. Computational Method Improvements:
   • Develop more efficient algorithms for injective norm computation
   • Extend numerical verification to higher orders and larger dimensions
5. Fine Structure of Spectral Measures:
   • Inspired by DLSST25, study global properties of "eigenvalue" distribution of skew-symmetric tensors
   • Establish deeper connections with free probability theory
6. Applications to Quantum Error Correction:
   • As shown in DZ25, injective norm relates to CSS code performance
   • Explore fermionic code design

In-Depth Evaluation

Strengths

1. Mathematical Rigor

• Complete proof structure: From Kac-Rice formula to Laplace method, logical chain is clear
• Sufficient technical detail: Appendices thoroughly present Grassmann geometry, covariance computation, etc.
• Appropriate use of authoritative results: Properly employs recent random matrix theory results from BABM22, BBvH23

2. Novelty

• Introduction of new ensembles: BHGAE and complex version fill gap in random matrix theory
• Geometric insights: Analysis on Grassmann manifolds reveals intrinsic geometric structure
• Discovery of duality: Particle-hole symmetry is physically profound and mathematically natural

3. Interdisciplinary Value

• Connects quantum information, random matrices, statistical physics, algebraic geometry
• Provides solid mathematical foundation for entanglement research in fermionic systems
• Extends methodology from spin glass landscape complexity

4. Experimental Verification

• Numerical simulations highly consistent with theoretical predictions
• Comparison of different symmetry classes provides additional insights
• Real/complex case comparison confirms theoretical unity

Weaknesses

1. Mismatch Between Upper and Lower Bounds

• Currently only upper bounds; cannot exclude possibility of smaller true values
• While numerical results support tightness, lacks rigorous proof
• Compared to DM24 (later completed by BS25), incomplete

2. Necessity of Technical Assumptions

• Property 3 (spectral gap) requires restriction to $u \in K \subset A$ (closed set away from $E_0(p)$)
• This technical compromise stems from structural zeros but weakens generality of results
• Future work needs more refined techniques to handle this

3. Limitations in Two-Scale Regime

• Growth rate of $\gamma_\alpha(d)$ restricted to $o(\exp(d^{1-\delta}))$
• Applicability of Laplace method to faster-growing sequences not established
• Limits understanding of certain extreme cases

4. Scope of Numerical Experiments

• Only verified for $p \leq 4$
• Numerical verification of two-scale regime (fixed $\alpha$) missing
• Computational cost limits depth of verification in high dimensions

5. Depth of Physical Interpretation

• Physical meaning of $\beta(\alpha)$ function insufficiently discussed
• Specific connections between particle-hole symmetry and physical systems could be more detailed
• Connections to actual fermionic systems (e.g., electron gas, quantum dots) not explicit

Impact

Academic Contribution

1. Theoretical Advancement: Significantly extends understanding of random tensor injective norm, particularly for non-symmetric cases
2. Methodology: Kac-Rice formula application on Grassmann manifolds may inspire other problems
3. Symmetry Class Classification: Lays foundation for systematic study of all Young tableau-corresponding tensor symmetry classes

Practical Value

1. Quantum Information: Provides theoretical benchmark for experimental measurement of fermionic entanglement
2. Quantum Computing: May apply to resource estimation for fermionic quantum algorithms
3. Condensed Matter Physics: Helps understand entanglement structure in many-body fermionic systems

Reproducibility

• High: Methods described in detail, numerical experiment parameters explicit
• While code not publicly released, FLN22 methods are reproducible
• Mathematical derivations can be verified (based on published lemmas)

Applicable Scenarios

Direct Application

1. Entanglement measurement of fermionic quantum states: $p$ fermions in $d$-dimensional single-particle space
2. Statistical properties of skew-symmetric tensors: Skew-symmetric structures in machine learning
3. Random theory of exterior differential forms: Applications in differential geometry

Potential Extensions

1. Topological quantum computing: Entanglement of Majorana fermions
2. Quantum chemistry: Entanglement analysis of molecular orbitals
3. Statistical learning: Theory of skew-symmetric kernel methods
4. Algebraic topology: Properties of random simplicial complexes

Limiting Conditions

• Requires Gaussian assumption (i.i.d. standard normal distribution)
• Asymptotic results need sufficiently large $d$
• Fixed $p$ case requires $p$ not too large ($p \ll d$)

Selected References

Core References

1. DM24 Dartois & McKenna: Injective norm of real and complex random tensors (directly extended in this work)
2. BABM22 Ben Arous, Bourgade & McKenna: Exponential growth of random determinants beyond invariance
3. BBvH23 Bandeira, Boedihardjo & van Handel: Matrix concentration inequalities and free probability

Technical Foundations

4. AEKN19 Alt, Erdős, Krüger & Nemish: Location of the spectrum of Kronecker random matrices
5. ABAČ13 Auffinger, Ben Arous & Černý: Random matrices and complexity of spin glasses

Quantum Information

6. AS17 Aubrun & Szarek: Alice and Bob meet Banach
7. FLN22 Fitter, Lancien & Nechita: Estimating the entanglement of random multipartite quantum states

Fermionic Systems

8. BHK21 Bianchi, Hackl & Kieburg: Page curve for fermionic Gaussian states
9. GKM12 Grabowski, Kuś & Marmo: Segre maps and entanglement for multipartite systems of indistinguishable particles

Overall Assessment: This is a high-quality mathematical physics paper making substantive contributions at the intersection of random tensor theory and quantum information. Technically rigorous, highly innovative, with clear physical significance. While some problems remain (such as matching upper and lower bounds), it provides a solid foundation for subsequent research. Recommended for researchers interested in random matrices, quantum entanglement, or statistical physics.