Statistics of Matrix Elements of Operators in a Disorder-Free SYK model

This paper investigates the disorder-free Sachdev-Ye-Kitaev (SYK) model and finds that, unlike the Fréchet distributions observed in the Lieb-Liniger model, the off-diagonal matrix elements of operators composed of four or more Majorana fermions are well-described by a generalized inverse Gaussian distribution.

The Gist

Imagine you are trying to understand how a chaotic, complex system (like a crowded room of people or a pot of boiling water) eventually settles down into a calm, predictable state. In physics, this is called thermalization.

For decades, physicists have used a rulebook called the Eigenstate Thermalization Hypothesis (ETH) to explain this. Think of ETH as a recipe that predicts how different parts of a quantum system "talk" to each other. Specifically, it looks at the "volume" of these conversations, which are mathematically represented as matrix elements (numbers that describe how one state transforms into another).

The Previous Discovery: The "Fréchet" Rule

Recently, scientists studied a specific model called the Lieb-Liniger model (think of it as a line of bosons, like a queue of people on a 1D track). They discovered something surprising: when they looked at the "volume" of the conversations between different states, the numbers followed a specific statistical pattern called the Fréchet distribution.

You can imagine the Fréchet distribution like a tsunami. Most waves are small and manageable, but occasionally, there are massive, rare "rogue waves" that are huge. This distribution is famous for describing extreme events.

The New Discovery: The "GIG" Rule

In this new paper, the authors, Tingfei Li and Shuanghong Li, asked: "Does this 'tsunami' rule apply to all quantum systems?"

They decided to test a different model: the Disorder-Free SYK model.

• The Analogy: If the Lieb-Liniger model is a 1D line of people, the SYK model is a massive, chaotic mosh pit where everyone is connected to everyone else (all-to-all interactions).
• The Twist: Usually, these models rely on "randomness" (disorder) to work. But this specific version is "disorder-free," meaning the connections are fixed and precise, yet it still behaves chaotically.

What They Found

They looked at the "conversations" (matrix elements) between different energy states in this mosh pit. They expected to see the same "tsunami" (Fréchet) pattern they saw in the line of people.

They were wrong.

Instead of a tsunami, they found a Goldilocks distribution (specifically, a Generalized Inverse Gaussian or GIG distribution).

• The Metaphor: Imagine the Fréchet distribution is like a lottery where you mostly get small prizes, but once in a blue moon, you win the jackpot. The GIG distribution they found is more like a well-balanced bell curve with a twist. It doesn't have those extreme, massive outliers. The "conversations" are more moderate and predictable in their extremes.

Why Does This Matter?

The authors found that this new pattern depends mostly on how many particles are involved in the conversation (the size of the operator), rather than the specific details of who is talking to whom or the temperature of the room.

• The "Insensitivity" Surprise: In the SYK model, it doesn't matter if you pick specific people to talk to; the statistical pattern remains the same. It's as if the mosh pit is so well-mixed that the specific arrangement of people doesn't change the overall "vibe" of the noise.
• The "Size" Rule: The pattern only changes if you change the number of people involved in the interaction (e.g., a 4-person conversation vs. an 8-person conversation).

The Big Picture

This paper is a crucial piece of the puzzle for understanding Quantum Chaos.

1. Different Rules for Different Worlds: It proves that the "tsunami" rule (Fréchet) isn't universal. One-dimensional systems (like the line of people) behave differently than zero-dimensional, all-to-all systems (like the mosh pit).
2. A New Signature: The authors propose that the GIG distribution is a new "fingerprint" for this specific type of quantum chaos. If you see this pattern in a system, you know you are dealing with a solvable, all-to-all interacting system.
3. The Bridge: This helps us understand how quantum systems, which are usually weird and unpredictable, eventually settle down to follow the laws of thermodynamics (heat and energy).

In summary: The authors took a famous rulebook about how quantum systems settle down, tested it in a new, chaotic "mosh pit" environment, and discovered that the old rulebook was wrong for this specific crowd. They wrote a new chapter in the rulebook, showing that in this chaotic mosh pit, the "extreme events" aren't as extreme as we thought, and the system follows a different, more balanced statistical law.

Technical Summary

1. Problem Statement and Motivation

The paper addresses a fundamental challenge in theoretical physics: characterizing how equilibrium statistical mechanics emerges from the nonequilibrium evolution of isolated quantum many-body systems. The Eigenstate Thermalization Hypothesis (ETH) is the prevailing framework, positing that thermalization in non-integrable systems arises from the statistical properties of matrix elements of local operators in energy eigenstates.

Recent work on the Lieb-Liniger model (a 1D integrable system of bosons) revealed that the probability distribution function (PDF) of off-diagonal matrix elements $\langle \mu | O | \lambda \rangle$ within the same macro-state follows a Fréchet distribution. This finding suggests that integrable systems possess distinct statistical signatures compared to chaotic ones.

The authors investigate whether this statistical behavior holds for other solvable systems, specifically focusing on the disorder-free Sachdev-Ye-Kitaev (SYK) model. Unlike the standard SYK model (which relies on random couplings and exhibits quantum chaos), the disorder-free variant features deterministic, all-to-all interactions and is integrable. The central question is: What is the statistical distribution of off-diagonal matrix elements in this zero-dimensional, integrable, all-to-all interacting model?

2. Methodology

A. The Model

The study utilizes the disorder-free SYK model defined by a Hamiltonian involving 4-body interactions of Majorana fermions ($\chi_j$):
$$H_4 = i^2 \sum_{1 \le j_1 < j_2 < j_3 < j_4 \le N} \chi_{j_1}\chi_{j_2}\chi_{j_3}\chi_{j_4}$$
The coupling strengths are fixed constants rather than random variables. The model is exactly solvable via a transformation into complex fermions ($f_k$) using a specific unitary transformation. The spectrum is determined by single-particle energies $\epsilon_k = 2 \cot(\theta_k/2)$.

B. Microstates and Macrostates

• Microstates: Defined by occupation numbers $|n\rangle = |n_1, \dots, n_{N_F}\rangle$ of $N_F = N/2$ single-particle levels.
• Macrostates: Characterized by a continuous density of states $\rho(\theta)$ in the thermodynamic limit ($N \to \infty$). The equilibrium macrostate is determined by solving coupled integral equations derived from maximizing entropy subject to fixed energy and particle number constraints (inverse temperature $\beta$ and occupation ratio $\lambda$).

C. Operator and Matrix Element Definition

The authors analyze operators constructed from products of $k$ Majorana fermions:
$$O = \chi_{a_1}\chi_{a_2}\dots\chi_{a_k}$$
They define a normalized quantity $M_{\{a\}}$ to study the statistics:
$$M_{\{a\}} = -\frac{1}{|a|} \ln |\langle n | \chi_{\{a\}} | m \rangle|^2 + \ln \frac{2}{N}$$
where $|m\rangle$ is a reference state and $|n\rangle$ is sampled from the same macrostate.

D. Numerical and Analytical Approach

1. Sampling: Microstates are generated by discretizing the $\theta$-interval into bins and sampling occupation numbers according to the equilibrium density $\rho(\theta)$.
2. Calculation: The matrix element $\langle n | \chi_{\{a\}} | m \rangle$ is computed efficiently. The authors show that for the dominant case where the Hamming distance between states equals the operator size ($d_{nm} = |a|$), the matrix element reduces to the determinant of a specific matrix $A$:
   $$\langle n | \chi_{\{a\}} | m \rangle = (-1)^P \det(A)$$
   where $A_{ij} \propto \exp[i s_i (a_j - 1)\theta_{k_i}]$.
3. Statistical Fitting: The authors generate $10^7$ samples for various operator sizes ($|a|$) and temperatures, then fit the resulting distributions of $M_{\{a\}}$ to candidate probability distributions (Fréchet vs. Generalized Inverse Gaussian).

3. Key Contributions and Results

A. Discovery of the Generalized Inverse Gaussian (GIG) Distribution

The primary finding is that for operators with size $|a| \ge 4$, the statistics of the off-diagonal matrix elements in the disorder-free SYK model are not described by the Fréchet distribution (as seen in Lieb-Liniger). Instead, they are well-fitted by the Generalized Inverse Gaussian (GIG) distribution with parameter $p=1$.

The GIG density function (for $x > \nu$) is:
$$P(x) \propto (x-\nu)^{p-1} \exp\left( -\eta(x-\nu) - \frac{\sigma}{x-\nu} \right)$$
For $p=1$, the log-density is a hyperbola. The authors demonstrate that the GIG fit is significantly more accurate than the Fréchet fit, particularly in capturing the tails of the distribution for larger $|a|$.

B. Insensitivity to Specifics

The study reveals that the distribution of matrix elements depends primarily on the operator size $|a|$ and is largely insensitive to:

• The specific choice of indices $\{a\}$ (due to the all-to-all nature of the interactions).
• The temperature $\beta$ (except in extreme limits like $T \to 0$ or for very small index separations).
  This universality arises because the phases in the determinant elements effectively "wind" around the unit circle, averaging out specific temperature-dependent distributions of $\theta$.

C. Special Cases

• $|a| = 2$: The distribution follows a specific form derived from $-\log|e^{i\theta}-1|$.
• $|a| = 3$: Also fits the GIG distribution, though with a slight mismatch near the peak.
• Odd vs. Even: The results hold for both even and odd $|a|$ (though particle number conservation usually restricts $|a|$ to be even in the studied setup).

4. Significance and Implications

• Differentiation of Integrable Models: The paper establishes that "integrability" does not imply a universal statistical signature for matrix elements. The disorder-free SYK model (zero-dimensional, all-to-all) exhibits GIG statistics, whereas the Lieb-Liniger model (one-dimensional, local) exhibits Fréchet statistics. This highlights the critical role of dimensionality and interaction structure (local vs. all-to-all) in determining thermalization behavior.
• Refinement of ETH: The results challenge the assumption that integrable systems share a single statistical profile. The GIG distribution serves as a new statistical signature for off-diagonal matrix elements in specific classes of solvable quantum many-body systems.
• Quantum Chaos vs. Integrability: Despite the disorder-free SYK model exhibiting "weak quantum chaos" (diagnosed by OTOCs), its matrix element statistics align with an integrable classification (distinct from chaotic random matrix theory predictions), yet distinct from other integrable models.
• Future Directions: The authors suggest exploring GIG distributions in other zero-dimensional models (e.g., colored SYK, tensor models) and investigating the relationship between GIG statistics and quantum chaos measures like OTOCs and spectral statistics.

In summary, this work provides a rigorous statistical characterization of matrix elements in a disorder-free SYK model, identifying the Generalized Inverse Gaussian distribution as the governing law for off-diagonal elements, thereby enriching the understanding of the Eigenstate Thermalization Hypothesis in solvable quantum systems.