Length-resolved Operator Growth and Path-Entropy… — Plain-Language Explanation

Imagine you have a long line of people (a "spin chain") holding hands. Some are holding hands tightly (strong connections), and others are holding hands loosely or not at all (disorder). In physics, we often ask: If you push the person at the very start of the line, does that "push" stay localized near them, or does it spread out to shake everyone else?

In the world of quantum physics, this question is about Many-Body Localization (MBL). For a long time, physicists hoped that if the disorder (the "looseness" of the connections) was strong enough, the push would get stuck, and the system would never forget its initial state. This would be like a traffic jam that never clears.

However, this paper argues that for certain types of quantum systems, that traffic jam is an illusion. Here is the breakdown of their findings using simple analogies:

1. The "Explosion" of Possibilities

The authors study what happens when you repeatedly "poke" the system (mathematically, taking commutators). They found that the "push" doesn't just spread; it explodes in complexity.

Imagine you are trying to walk from your house to a friend's house.

The Old View (Localization): You take one path, maybe a few detours, but you stay on a specific street.
The New View (This Paper): Every time you take a step, the number of possible paths you could have taken multiplies wildly. It's not just a few detours; it's as if every step you take splits into a branching tree of possibilities that grows almost factorially (a number that grows faster than you can count, like $100!$ ).

The paper proves that in these disordered systems, the "weight" of the push isn't just spreading out; it is being distributed across a massive, almost infinite number of different paths simultaneously.

2. The "Path Entropy" Obstacle

The authors introduce a concept called Path Entropy. Think of this as the sheer "noise" or "confusion" caused by having too many options.

The Analogy: Imagine trying to hear a whisper in a room. If the room is quiet (low entropy), you can hear it. But if the room is filled with millions of people all shouting different random things (high path entropy), the whisper gets drowned out.
The Result: In these quantum systems, the "noise" of the billions of possible paths is so loud that it overpowers any attempt to keep the information localized. The paper argues that for the system to remain localized, all these billions of random paths would have to magically cancel each other out perfectly (like a choir singing so perfectly that the sound disappears). The authors say this is statistically impossible without some special, hidden rule that we haven't found yet.

3. The "Finite-Size" Illusion

One of the most practical findings is about why computer simulations have been confusing.

The Analogy: Imagine you are studying how fast a forest fire spreads. If you only look at a tiny patch of grass (a small computer simulation), the fire might seem to die out quickly because it runs out of grass to burn. It looks like the fire is "localized."
The Reality: But if you look at the whole forest, the fire spreads everywhere.
The Paper's Claim: The authors prove that current computer simulations are looking at "tiny patches." They derived a specific scale: $L \sim (W/J)^2$ . As long as the system size ( $L$ ) is smaller than this scale, the system looks like it's localized. But once the system gets big enough (larger than this scale), the "fire" (the operator growth) inevitably spreads out. The "localization" seen in small simulations is just a pre-asymptotic regime—a temporary illusion before the true, spreading behavior takes over.

4. The Failure of the "Fix-It" Tool

Physicists have a mathematical tool (called a Schrieffer-Wolff transformation) used to "fix" a messy system and turn it into a neat, localized one. They hoped this tool would work for these disordered chains.

The Analogy: Imagine trying to organize a messy room by moving items one by one.
The Problem: The authors show that as you try to organize the room, the "mess" (the number of possible ways to arrange things) grows so fast that your organizing tool breaks down. The "path entropy" (the sheer number of ways the mess can happen) overwhelms the tool's ability to keep things tidy.
The Conclusion: You cannot mathematically construct a "localized" version of this system using standard methods because the complexity of the paths is too high.

The Bottom Line

The paper concludes that true, permanent localization (where the system never forgets its start) is likely impossible in these specific quantum chains, no matter how strong the disorder is.

Short term/Small systems: It looks like the system is stuck (localized).
Long term/Large systems: The "path entropy" wins. The system eventually spreads out, forgets its initial state, and becomes "ergodic" (fully mixed).

The authors suggest that if localization does exist, it would require a miraculous, hidden mechanism where billions of random paths cancel each other out perfectly—a scenario they consider highly unlikely. Therefore, in the real, infinite world, these systems are likely always chaotic and spreading, not stuck.

Technical Summary: Length-Resolved Operator Growth and Path-Entropy Obstructions to Many-Body Localization

Problem Statement
The existence and stability of Many-Body Localization (MBL) in the thermodynamic limit remains a subject of intense debate. While rigorous proofs exist for Anderson localization in non-interacting systems, the stability of MBL in interacting disordered systems is contested. A primary challenge is distinguishing a true phase transition from finite-size crossovers in numerical studies, which are limited to small system sizes. Furthermore, there is a structural tension between the hypothesis that ergodic systems exhibit maximal (almost factorial) operator growth and the possibility that interacting disordered systems could remain localized. The paper addresses whether maximal operator growth is compatible with the strongest forms of localization (dynamical locality) and the existence of Local Integrals of Motion (LIOMs) constructed via perturbative schemes.

Methodology
The authors focus on the one-dimensional disordered Ising chain with transverse and longitudinal fields, where couplings and fields are drawn from strictly positive distributions. The study employs a rigorous operator-theoretic approach rather than numerical diagonalization.

Operator Formalism: The authors utilize the Pauli string basis to resolve operator norms by support length $\ell$ . They analyze the $k$ -fold iterated commutator $L_H^k(A) = [H, A]^{(k)}$ of the Hamiltonian $H$ with a local operator $A$ (specifically $\sigma^z_0$ ).
Gauge Transformation: To establish rigorous lower bounds, the authors introduce a specific gauge transformation (changing the basis of Pauli operators) that ensures all matrix elements of the Liouvillian superoperator are strictly positive. This eliminates the possibility of destructive interference between different commutator paths, allowing for a strict lower bound on operator growth.
Path Entropy Analysis: The core of the analysis involves counting the number of "scrambling paths" in the commutator expansion. The authors distinguish between the exponential decay of individual path amplitudes (due to disorder) and the combinatorial explosion of the number of paths (path entropy).
Perturbative Construction: The paper investigates the convergence of iterative Schrieffer-Wolff transformations used to construct LIOMs. It analyzes the generator $T$ of the unitary transformation mapping the microscopic Hamiltonian to a LIOM Hamiltonian, focusing on the growth of the support length of $T^{(n)}$ at order $n$ .

Key Contributions and Results

Length-Resolved Almost Factorial Growth:
The authors generalize previous results (Cao [1]) by resolving the operator norm by support length. They prove that for the disordered Ising chain with positive couplings, the operator weight at a specific length scale $\ell_k \sim k / \ln k$ grows almost factorially:
$\|[H, \sigma^z_0]^{(k)}_{\ell_k}\|_2 \gtrsim \left(\frac{k}{\ln k}\right)^k$
This implies that the operator weight is not confined to short ranges but spreads spatially, with the dominant weight located at lengths that diverge with the commutator order $k$ .
Rigorous Exclusion of Dynamical Locality:
Based on the length-resolved lower bound, the paper proves that "dynamical locality"—the condition that local operators remain exponentially localized under time evolution—is rigorously ruled out for this model. The entropic contribution from the branching of commutator paths overwhelms the exponential localization factor expected in localized phases.
Finite-Size Crossover Scale:
The competition between the path-entropy term (which drives delocalization) and the exponential localization term (driven by disorder strength $W$ and coupling $J$ ) establishes a rigorous finite-size crossover scale:
$L \sim \left(\frac{W}{J}\right)^2$
For system sizes $L \lesssim (W/J)^2$ , the system resides in a "pre-asymptotic" regime where path entropy has not yet overwhelmed localization. In this regime, numerical studies may observe apparent localization, but this is a transient effect that vanishes in the thermodynamic limit ( $L \to \infty$ ).
Path-Entropy Obstruction to LIOM Construction:
The paper identifies a structural obstruction to the perturbative construction of LIOMs. Even if one assumes no many-body resonances or avalanches, the iterative Schrieffer-Wolff scheme generates a number of scrambling paths that grows with the support length. The authors argue that without a specific, unidentified mechanism for destructive interference among these almost factorially many paths with random amplitudes, the generator $T$ of the unitary transformation will inevitably become non-local. This suggests that the perturbative construction of LIOMs fails in the thermodynamic limit due to "path entropy."
Consistency with Abstract LIOMs:
The authors clarify that while maximal operator growth rules out dynamical locality, it does not strictly rule out the existence of abstract LIOM Hamiltonians. They demonstrate that abstract LIOM Hamiltonians can support operators with even faster-than-factorial growth. However, the obstruction lies in the construction of the quasi-local unitary mapping the microscopic model to such an abstract Hamiltonian.

Significance and Claims
The paper claims to resolve the tension between maximal operator growth and localization by demonstrating that maximal growth is a generic feature of disordered interacting spin chains that precludes the strongest form of localization (dynamical locality).

On Numerical Studies: The work provides a rigorous explanation for why numerical studies on small systems often suggest stable MBL phases at strong disorder. It posits that these observations are artifacts of the pre-asymptotic regime defined by the crossover scale $L \sim (W/J)^2$ . Consequently, numerical simulations of small systems are ill-suited to determine the stability of MBL in the thermodynamic limit.
On MBL Stability: The authors conclude that the "path-entropy obstruction" is a fundamental mechanism that likely renders MBL unstable in the thermodynamic limit for generic disordered spin chains. They argue that for MBL to survive, a highly fine-tuned mechanism for the systematic cancellation of almost factorially many disorder-dependent paths would be required, which is considered non-generic.
On Real-Time Dynamics: The results suggest that real-time operator spreading in these systems is ballistic (saturating the Lieb-Robinson cone) up to logarithmic corrections, rather than sub-ballistic or localized, as the systematic cancellation required for sub-ballistic spreading is statistically improbable.

In summary, the paper argues that the "path-entropy" arising from the combinatorial branching of operator content is a structural barrier to localization that operates independently of resonance effects, challenging the existence of a stable MBL phase in the thermodynamic limit.

Length-resolved Operator Growth and Path-Entropy Obstructions to Many-Body Localization