On Entropic Characterization of Symmetry Breaking in Dynamical Systems I: Spontaneous Symmetry Breaking

This paper establishes an entropic framework for analyzing spontaneous symmetry breaking in equivariant dynamical systems, distinguishing between local mechanisms characterized by increased Shannon entropy and critical slowing down, and global mechanisms where entropy changes depend on the redistribution of probability across symmetry-related sectors.

The Gist

Imagine a perfectly balanced spinning top. As long as it spins fast, it stays upright and symmetrical. But as it slows down, it eventually wobbles and falls to one side. In the world of physics and mathematics, this is called Symmetry Breaking. The rules governing the top are perfectly symmetrical (it could fall left or right with equal chance), but the final result is not.

This paper explores a new way to measure and understand how and why this happens, using the language of information and uncertainty (entropy). The authors argue that symmetry breaking isn't just one thing; it happens in two very different ways, and we need different tools to measure each one.

Here is the breakdown of their findings using everyday analogies:

1. The Two Types of Symmetry Breaking

The paper distinguishes between Local breaking and Global breaking. Think of these as two different ways a crowd of people might lose their perfect formation.

Local Symmetry Breaking: The "Wobbly Wagon"

• The Scenario: Imagine a wagon sitting perfectly still at the top of a hill (the "symmetric equilibrium"). It's balanced, but unstable. As the ground tilts slightly (a change in a parameter), the wagon starts to wobble.
• What Happens: Before it finally rolls down to one side, it slows down its ability to correct itself. It gets "lazy" about returning to the center.
• The Information Signal:
  • Slowing Down: The wagon takes longer and longer to settle back to the center after a tiny push. This is called "critical slowing down."
  • Spreading Out: Because it's not snapping back quickly, the wagon's position becomes very uncertain. It wobbles over a wider area.
  • The Result: This spreading out means Entropy (Uncertainty) goes UP. The system becomes "messier" right before the break.
  • The Message: If you watch a system getting "wobbly" and "messy" (high entropy) while slowing down, you know a symmetry break is about to happen.

Global Symmetry Breaking: The "Reorganized Party"

• The Scenario: Imagine a party where everyone is dancing in a perfect circle (symmetry). Suddenly, the music changes. Instead of everyone staying in one big circle, the crowd splits into two smaller, distinct groups dancing on opposite sides of the room.
• What Happens: The space the people occupy hasn't changed (they are still in the same room), but the pattern of where they are standing has completely reorganized.
• The Information Signal:
  • The Split: The crowd moves from one big group to two smaller groups.
  • The Surprise: Unlike the "wobbly wagon," this doesn't always mean more mess.
    • If the two new groups are very distinct and far apart, the system gains Entropy because you now have a new piece of information: "Which group is this person in?" (Left or Right?).
    • However, if the two groups are very close together or overlap, the system might actually lose entropy because the people are more "focused" in their new spots.
  • The Result: Global breaking is a trade-off. You lose some "internal" disorder (people are more focused in their new groups) but gain "label" disorder (you have to track which group they belong to). The total change depends on how well-separated the groups are.

2. The Core Discovery: No Single Rule

The most important takeaway from the paper is that there is no single rule for how entropy changes during symmetry breaking.

• In Local Breaking: Entropy almost always increases as the system gets ready to break (it gets wobbly and spreads out).
• In Global Breaking: Entropy can increase OR decrease. It depends on whether the new "groups" are far apart (high uncertainty about which group) or close together (low uncertainty).

3. Why This Matters (According to the Paper)

The authors built a mathematical framework to measure these changes. They found that:

• Directional Information: In local breaking, you can tell which way the system is about to fall by watching how information flows between different parts of the system. It's like seeing which way the wagon's wheels are turning before it tips.
• The "Label" Concept: In global breaking, the system creates a new "label" (like "Group A" or "Group B"). The paper shows that the total uncertainty of the system is just the sum of the uncertainty inside the groups plus the uncertainty of which group you are in.

Summary Analogy

Think of a symmetric system as a perfectly round snowball.

• Local Breaking: As the snowball melts, it gets soft and starts to wobble. It becomes a big, messy puddle (high entropy) before it finally settles into a specific shape. The warning sign is the messiness.
• Global Breaking: The snowball doesn't melt; instead, it suddenly cracks into two perfect, smaller snowballs. The total amount of "snow" is the same, but now you have to decide: "Is this snowball the left one or the right one?" The change in uncertainty depends on how far apart those two new snowballs are.

The paper provides the math to measure these "wobbles" and "cracks" using information theory, proving that we need to look at the problem through two different lenses to understand the full picture.

Technical Summary

Technical Summary: Entropic Characterization of Symmetry Breaking in Equivariant Dynamical Systems I

Problem Statement
Symmetry breaking is a fundamental organizing principle in nonlinear dynamics, governing phenomena ranging from ferromagnetism to biological pattern formation. However, the information-theoretic signatures of spontaneous symmetry breaking (SSB) remain poorly understood. A central challenge is that SSB is often treated as a monolithic phenomenon, whereas it can occur at fundamentally different levels of organization:

1. Local SSB: The loss of stability of a symmetric equilibrium via bifurcation, leading to the emergence of symmetry-related branches.
2. Global SSB: A qualitative reorganization of the invariant measure (the long-time statistical state) across the state space, even if the underlying support of the attractor remains unchanged.

The paper argues that these mechanisms require distinct analytical lenses and that there is no universal entropy law governing symmetry breaking. The work aims to develop an entropic framework that distinguishes between these local and global mechanisms in equivariant dynamical systems.

Methodology
The authors analyze autonomous dynamical systems $\dot{x} = f(x)$ on a smooth manifold $X$, where $f$ is $G$-equivariant under the action of a compact symmetry group $G$. The study employs a combination of dynamical systems theory, stochastic regularization, and information theory.

• Stochastic Regularization: To define entropy for deterministic systems (where invariant measures are often singular Dirac masses), the authors introduce small isotropic additive noise. This yields a regularized, smooth invariant density (e.g., Gaussian near equilibria or Gibbs measures on manifolds), allowing for the calculation of Shannon differential entropy and directional information transfer.
• Local Analysis: Near a symmetric equilibrium, the authors utilize linearization and critical slowing down theory. They analyze the growth of the stationary covariance matrix as the system approaches a bifurcation point, linking spectral properties of the Jacobian to entropy and information transfer.
• Global Analysis: For global restructuring, the authors model the post-break invariant density as a finite mixture of symmetry-related sectors. They apply the chain rule for entropy and mutual information to decompose the total entropy into internal sector entropy and symmetry-label information.
• Information Transfer: The study utilizes continuous-time directional information transfer (based on steady-state covariance) to quantify how fluctuations in critical modes propagate between subsystems.

Key Contributions and Results

1. Local Spontaneous Symmetry Breaking (Local SSB)
The paper establishes a coherent local mechanism linking stability loss to information-theoretic signatures:

• Critical Slowing Down: As a symmetric equilibrium approaches a bifurcation point ($\mu \to \mu_c^-$), the relaxation time diverges because the real part of the critical eigenvalue approaches zero.
• Entropy Growth: This slowing down causes the regularized invariant density to broaden (variance growth). Consequently, the Shannon differential entropy of the local stationary state diverges (or grows sharply) as the threshold is approached.
• Information Transfer Amplification: The same covariance broadening drives a sharp increase in steady-state directional information transfer. Specifically, if the critical mode projects onto a subsystem $x$, the transfer $T_{x \to y}$ from $x$ to a coupled subsystem $y$ exhibits a highly structured, parameter-sensitive dependence near the transition. This transfer acts as a local diagnostic for the instability pathway.
• Example: In a coupled pitchfork system, the entropy of the critical mode grows significantly, and directional transfer identifies the dominant instability direction.

2. Global Spontaneous Symmetry Breaking (Global SSB)
The paper derives a general entropy law for the restructuring of invariant densities:

• Entropy Decomposition: When the invariant density reorganizes into $m$ symmetry-related sectors, the total entropy $H(\rho^+)$ decomposes as:
  $$H(\rho^+) = \sum p_i H(\rho^{(i)}) + I(S; X)$$
  where the first term is the weighted average of internal sector entropies, and the second term is the mutual information between the state $X$ and the sector label $S$.
• Disjoint vs. Overlapping Sectors:
  • In the disjoint-sector limit (e.g., invariant set splitting), the mutual information equals the entropy of the label distribution ($H(p)$). If sectors are equally weighted, the entropy gain is $\log k$, where $k$ is the number of sectors (specifically, the index of the isotropy subgroup).
  • In the overlapping regime (finite noise on a fixed support), the mutual information is strictly less than $\log k$. The total entropy change depends on the balance between the localization within sectors (which may decrease entropy) and the multiplicity of sectors (which increases entropy).
• Non-Universality: Unlike local SSB, global SSB does not universally increase entropy. The net change depends on whether the probability redistribution into multiple sectors outweighs the potential narrowing of the density within those sectors.
• Examples:
  • Separated Sectors: In a $D_4$-equivariant four-well potential, the entropy increases by approximately $\log 4$ as the system splits into four disjoint wells.
  • Overlapping Sectors: In a reflection-symmetric pitchfork on a circle ($S^1$), the density reorganizes from one peak to two overlapping peaks. The entropy increase is governed by the mutual information $I(S; \Theta)$, which approaches $\log 2$ only as noise vanishes and the peaks separate.

Significance and Claims
The paper claims to provide a "quantitative bridge" between symmetry, bifurcation structure, invariant measures, and information theory. Its primary significance lies in:

1. Distinguishing Mechanisms: It rigorously separates local instability (bifurcation-driven) from global statistical reorganization (measure-driven), showing they possess different entropic signatures.
2. Refuting Universal Laws: It demonstrates that there is no single "entropy law" for symmetry breaking; global breaking can increase or decrease entropy depending on the specific probabilistic redistribution.
3. Diagnostic Tools: It proposes Shannon entropy and directional information transfer as measurable observables for detecting symmetry-breaking transitions in experimental, stochastic, and data-driven settings.
4. Unified Framework: It unifies the description of symmetry breaking by showing that classical invariant-set splitting is merely the disjoint-sector limit of a more general invariant-density restructuring process.

The work concludes that symmetry breaking is best understood as a hierarchy of local and global reorganizations, each requiring distinct probabilistic descriptions to be fully characterized.