Perturbatively Stable Self-Correcting Classical Memory from Gauge Averaging

Technical Summary: Perturbatively Stable Self-Correcting Classical Memory from Gauge Averaging

Problem Statement
The reliability of information storage in physical systems is foundational to computing. While certain classical systems, such as the $Z_2$-symmetric Ising model, exhibit self-correcting memory in thermal equilibrium (retaining macroscopic magnetization for exponentially long times), this property is fragile: it relies on exact symmetry and is destroyed by explicit symmetry-breaking perturbations (e.g., an external magnetic field).

The paradigmatic candidate for a more robust system is Wegner's 3d $Z_2$ gauge theory, which possesses topological order rather than a local order parameter. In this model, information is stored in topological winding numbers (homology classes of fluctuating surfaces) that are stable for exponentially long lifetimes at low temperatures. However, it has remained an open question whether this self-correcting memory is perturbatively stable—i.e., whether it survives arbitrary small perturbations that explicitly break the local gauge symmetry. Intuitively, one might expect that perturbations like surface tension would create extensive energy penalties for non-trivial homology sectors, destroying the memory. While numerical evidence and specific proofs for magnetic field perturbations suggested stability, a rigorous proof for arbitrary small perturbations was lacking.

Methodology: Gauge Averaging
The authors introduce a novel method termed "gauge averaging" to address this problem. The core insight is that even when a local symmetry is explicitly broken by a perturbation $V$, thermal fluctuations can effectively restore this symmetry if the perturbation is sufficiently weak relative to the temperature.

1. Gauge Averaging Construction: For a Hamiltonian $H_0$ with a finite symmetry group $G$ and a perturbation $V$, the authors define a "gauge-averaged" state $\rho_G$ by averaging the Gibbs state $\rho = Z^{-1}e^{-\beta(H_0+V)}$ over the group $G$:
   $$\rho_G = \frac{1}{|G|} \sum_{g \in G} g \rho g^{-1}$$
   This state preserves all correlation functions for symmetric operators. Crucially, $\rho_G$ can be formally written as a Gibbs state of an effective Hamiltonian $H_0 + V_{\text{eff}}$, where $V_{\text{eff}}$ is a new interaction term that is exactly $G$-symmetric.

2. Cluster Expansion and Locality: The central technical challenge is determining if $V_{\text{eff}}$ admits a local expression. Using a cluster expansion, the authors prove that if the perturbation $V$ satisfies a specific norm condition (defined by an interaction norm $|V|_\eta$ involving exponential decay with distance), then $V_{\text{eff}}$ is an exponentially local, symmetric interaction.

   • Theorem 2: Establishes that for a finite abelian symmetry group with bounded generator overlap, if $\beta |V|_\eta < C_{\text{gauge}}$, then a local symmetric $V_{\text{eff}}$ exists.
   • Temperature Dependence: A key feature of this method is that the condition for stability ($\beta |V|_\eta < C$) implies that robustness increases with temperature (as $\beta$ decreases), provided the system remains below a critical temperature $T_c$.

3. Bottleneck Analysis: To prove the existence of self-correcting memory, the authors utilize the "bottleneck" argument. They define a "bottle" $A$ (a set of configurations with short syndrome loops) and a "bottleneck" $\partial A$ (configurations with exactly one long loop). The mixing time of the system is lower-bounded by the ratio $\mu(A)/\mu(\partial A)$.

   • For the unperturbed model, a Peierls bound shows this ratio is exponentially large.
   • For the perturbed model, the authors apply the bottleneck analysis to the gauge-averaged distribution $\mu_G$. Since the sets $A$ and $\partial A$ are defined by topological properties (homology classes) which are invariant under $G$, the ratio $\mu(A)/\mu(\partial A)$ is identical to $\mu_G(A)/\mu_G(\partial A)$.
   • Because $V_{\text{eff}}$ is local and symmetric, the authors can derive a perimeter law bound for the energy cost of creating long loops in $\mu_G$, ensuring the bottleneck ratio remains exponentially large.

Key Results

• Theorem 1 (Main Result): The self-correcting memory in Wegner's 3d $Z_2$ gauge theory on a 3-torus is stable to all perturbations $V$ satisfying $\beta |V|_\eta < C_{\text{stab}}(T, \eta)$. This proves the existence of a "fluctuation-stabilized" phase where the memory persists for an exponentially long time.
• Phase Diagram: The authors construct a phase diagram (Fig. 1) showing a wedge-shaped region of stability in the $(T, |V|)$ plane. The region of stability widens as temperature increases (up to a critical $T_c \approx 1/\log 5$), contrasting with the intuition that higher temperatures always destroy order.
• Quantitative Comparison: In the specific case of a magnetic field perturbation ($V = -h \sum Z_e$), the authors derive an explicit form for $V_{\text{eff}}$ using electric-magnetic duality. They estimate the critical slope of the phase boundary to be $S_c \leq 1/\text{arctanh}(1/5) \approx 4.9$. This is within 10% of the numerically estimated value ($S_c \approx 4.5$), validating the method's quantitative accuracy.

Significance and Claims
The paper claims to provide the first rigorous proof that the self-correcting memory of Wegner gauge theory is stable against arbitrary small local perturbations. The significance lies in:

1. Fluctuation Stabilization: Demonstrating that thermal fluctuations can effectively restore broken symmetries, creating a phase of matter where perturbative stability is enhanced by temperature.
2. Robustness of Local Symmetries: Showing that extensive local symmetries are robust not only against spontaneous breaking (Elitzur's theorem) but also against explicit breaking, provided the breaking is weak relative to thermal energy.
3. Methodological Advance: Introducing "gauge averaging" as a tool to map symmetry-breaking problems to symmetry-preserving effective theories, allowing the use of standard cluster expansion techniques to prove stability in topological phases.

The authors remain modest regarding broader implications, noting that while the method proves stability for Wegner gauge theory, the relationship to non-equilibrium fluctuation-stabilized memories or emergent higher-form symmetries remains an open area for future investigation.