Discontinuous transition in 2D Potts: II. Order-Order Interface convergence

This paper establishes the wetting phenomenon in the $q$-state Potts model for $q>4$ at the critical temperature, proving that a disordered layer emerges between two ordered phases and that its boundaries converge to a pair of non-intersecting Brownian motions under diffusive scaling.

The Gist

The Big Picture: A Battle of Colors at a Critical Temperature

Imagine a giant chessboard where every square can be painted one of $q$ different colors (like red, blue, green, etc.). This is the Potts Model, a famous way physicists simulate how materials change states, like ice melting into water.

Usually, when you cool this board down, the squares want to agree with their neighbors. If you have a lot of red squares, they want to turn the whole board red. This is an "ordered" state.

However, this paper focuses on a very specific, chaotic scenario:

1. The Temperature is "Just Right" ($T_c$): The system is at a critical tipping point.
2. There are Many Colors ($q > 4$): There are so many options that the system behaves strangely.
3. The Setup: Imagine the top half of the board is forced to be Red, and the bottom half is forced to be Blue.

The Question: What happens in the middle where Red meets Blue?

The Old Story vs. The New Discovery

The Old Story (Cold Temperatures):
If the board is cold, Red and Blue hate each other. They form a sharp, thin line right down the middle. It's like a wall of soldiers standing shoulder-to-shoulder. The line is straight and predictable.

The New Discovery (The Critical Temperature):
The authors discovered something surprising happens at the critical temperature when there are many colors. Instead of a sharp wall, a third layer spontaneously appears between the Red and Blue.

Think of it like this:

• Red and Blue are two rival gangs.
• Usually, they fight directly at the border.
• But at this specific temperature, they realize, "Hey, fighting is expensive! Let's hire a buffer zone."
• A layer of neutral, uncolored squares (the "disordered" phase) spontaneously grows between them to keep the peace.

This is called Interfacial Wetting. The "neutral" phase "wets" the boundary, pushing the two ordered phases apart.

The Shape of the Chaos: The "Brownian Watermelon"

The most exciting part of the paper is describing the shape of this neutral layer.

In the past, scientists knew the layer existed, but they didn't know exactly how it wiggled. The authors proved that if you zoom out and look at the boundaries of this neutral layer, they don't just wiggle randomly. They wiggle in a very specific, coordinated dance.

The Analogy: The Brownian Watermelon
Imagine two drunk people (Brownian motions) walking on a tightrope.

• They start at the same point on the left.
• They end at the same point on the right.
• The Catch: They are forbidden from ever touching or crossing each other. If they get too close, a magical force pushes them apart (this is called entropic repulsion).

The paper proves that the top edge of the neutral layer and the bottom edge of the neutral layer behave exactly like these two drunk walkers. They form a shape that looks like a watermelon (a lens shape) in the middle.

In math-speak, they converge to a "Brownian Watermelon with two bridges."

How Did They Prove This? (The Detective Work)

Proving this was incredibly hard because the math gets messy when you have two interacting layers instead of one. The authors used a clever "translation" strategy:

1. The Problem: The Potts model (the colored squares) is too complicated to analyze directly.
2. The Translation: They translated the problem into a different language: Percolation (imagine the squares as islands and the connections as bridges).
3. The Secret Weapon (ATRC): They mapped the problem to something called the Ashkin-Teller Random-Cluster (ATRC) model.
   • Think of the ATRC model as a "super-model" that has a special property: Positive Correlation.
   • In plain English: In this super-model, if one part of the system pushes up, it makes it easier for the other part to push up too. This "repulsion" between the two layers is what keeps them from crashing into each other.
4. The Result: Once they translated the problem into this "super-model," they could use powerful tools from probability theory (specifically, Renewal Theory and Random Walks) to show that the layers behave exactly like the two non-touching drunk walkers described above.

Why Does This Matter?

1. It's a First: This is the first time anyone has rigorously described the geometry of this "wetting" phenomenon in a lattice model. Before, we knew it happened, but we didn't know the exact shape of the boundary.
2. It Breaks the Rules: Usually, in physics, when things get big, they smooth out into simple curves (like a single Brownian bridge). Here, because of the interaction between the two layers, the result is more complex and "non-Gaussian" (it doesn't follow the standard bell curve rules).
3. Universal Truths: This helps us understand how interfaces behave in many different physical systems, from magnets to fluid dynamics, whenever they are at a critical tipping point.

Summary in One Sentence

At a critical temperature with many colors, the boundary between two solid colors doesn't just touch; it sprouts a thick, wiggly, neutral layer that pushes the colors apart, and the edges of this layer dance in a perfectly coordinated, non-touching pattern known as a "Brownian Watermelon."

Technical Summary

1. Problem Statement and Context

The paper investigates the 2D $q$-state Potts model on a square lattice at the critical temperature $T_c(q)$ in the regime of discontinuous phase transitions ($q > 4$).

• Phase Coexistence: At $T_c(q)$ for $q > 4$, there are $q+1$ extremal Gibbs measures: $q$ ordered (monochromatic) phases and one disordered (free) phase.
• Dobrushin Boundary Conditions: The authors study the geometry of interfaces separating different phases under specific boundary conditions:
  • Order-Disorder: One half of the boundary is fixed to a color, the other is free. (Studied in their companion work [DGO25]).
  • Order-Order: One half of the boundary is fixed to color 1, the other to color 2. This is the focus of the current paper.
• The Phenomenon: Unlike the subcritical regime ($T < T_c$) where the interface between two ordered phases is of width $O(1)$, the order-order interface at $T_c(q)$ exhibits interfacial wetting. A mesoscopic layer of the disordered phase spontaneously emerges between the two ordered phases.
• The Goal: To provide a rigorous geometric description of this wetting layer and prove the scaling limit of the interfaces separating the ordered phases.

2. Methodology and Technical Framework

The authors employ a sophisticated chain of couplings to map the Potts model to a more tractable percolation model, leveraging recent advances in the analysis of the Ashkin-Teller Random-Cluster (ATRC) model.

A. The Chain of Couplings

The core strategy involves mapping the Potts model to the ATRC model via the six-vertex model:

1. Potts $\to$ FK Percolation: The Potts model is represented via the Fortuin-Kasteleyn (FK) random-cluster model.
2. FK $\to$ Six-Vertex: Using the Baxter-Kelland-Wu (BKW) coupling, the FK model is mapped to the six-vertex model (specifically, its spin representation).
3. Six-Vertex $\to$ ATRC: The six-vertex model is coupled to the ATRC model (a graphical representation of the Ashkin-Teller model).
   • The ATRC model consists of two coupled percolation configurations, $\omega_\tau$ and $\omega_{\tau\tau'}$, satisfying $\omega_\tau \subseteq \omega_{\tau\tau'}$.
   • Crucially, the ATRC model possesses strong positive association (FKG property) and exponential decay of correlations in the infinite volume limit, properties that are difficult to establish directly for the Potts model at $T_c$.

B. Interface Definitions

In the coupled ATRC framework, the authors define the interfaces separating the ordered phases:

• Upper Interface ($\Gamma^1_{FK}$): The boundary of the cluster of color 1 (primal connection).
• Lower Interface ($\Gamma^2_{FK}$): The boundary of the cluster of color 2 (dual connection).
• In the ATRC representation, these correspond to the boundaries of specific clusters $C$ (in $\omega_\tau$) and $C'$ (in $\omega^*_{\tau\tau'}$) connecting the left and right boundaries of the domain.

C. Key Technical Tools

1. Entropic Repulsion: The authors prove that the two interfaces experience an effective repulsive force due to entropy. While the ATRC model has positive correlations, the conditioning on non-intersection (or the specific boundary conditions) forces the interfaces apart.
2. Ornstein-Zernike (OZ) Theory: They utilize a "renewal picture" for the ATRC clusters. Long connections are decomposed into a sequence of "irreducible" microscopic clusters (cones) that behave like random walk steps.
3. Mixing Properties: They establish strong mixing estimates for the ATRC model to decouple the behavior of the two interfaces when they are far apart.
4. Coupling to Random Walks: The geometry of the interfaces is coupled to a pair of random walk bridges conditioned not to intersect.

3. Key Contributions and Results

A. Main Theorem (Theorem 1.1 & 1.2)

The primary result establishes the convergence of the order-order interfaces under diffusive scaling.

• Scaling: Consider a box of size $N \times N$. The interfaces are rescaled by $1/\sqrt{N}$.
• Convergence: As $N \to \infty$, the pair of interfaces $(\tilde{\Gamma}^1, \tilde{\Gamma}^2)$ converges in law to a Brownian Watermelon (a pair of Brownian bridges conditioned not to intersect).
• Wetting Layer: The distance between the two interfaces scales as $\sqrt{N}$. Specifically, the probability that the interfaces are closer than $C \ln^{111}(N)$ tends to zero. This confirms the emergence of a macroscopic disordered layer of width $\sqrt{N}$ between the ordered phases.

B. Rigorous Proof of Wetting

This is the first rigorous geometric description of interfacial wetting in a lattice spin model. Previous results (e.g., Bricmont-Lebowitz '87, Messager et al. '91) were limited to surface tension estimates or large $q$ expansions. This work provides a full scaling limit.

C. Extension to FK Percolation

The result is first proven for the ATRC model and then transferred back to the FK percolation model (and thus the Potts model) via the coupling.

• Theorem 1.2: The interfaces in the FK percolation model at $p_c(q)$ for $q > 4$ also converge to the Brownian watermelon.

D. Technical Innovations

• Non-Deterministic Wall: Unlike previous works on interfaces above a fixed wall, here the "lower" interface acts as a non-deterministic, fluctuating wall for the "upper" interface. The authors overcome this by proving that the lower interface stays sufficiently far from the boundary and behaves like a random walk itself.
• Total Variation Bounds: They establish precise total variation bounds (Theorem 5.5) between the law of the coupled clusters and a pair of independent random walk bridges conditioned on non-intersection.
• Handling the ATRC: They navigate the weaker domain Markov property of the ATRC model (compared to standard FK) by using the specific structure of the coupling and the "good cluster" decomposition.

4. Significance and Impact

1. Resolution of a Long-Standing Problem: The paper completes the picture of Dobrushin interfaces in the 2D Potts model for $q > 4$. It bridges the gap between the continuous transition regime ($q \le 4$, where interfaces converge to a single Brownian bridge) and the discontinuous regime.
2. Universality of Wetting: It demonstrates that the "Brownian Watermelon" limit is the universal scaling limit for order-order interfaces at the critical point of discontinuous transitions in 2D, distinct from the single Brownian bridge seen in subcritical or continuous regimes.
3. Methodological Advancement: The work establishes a powerful framework for analyzing interacting interfaces in spin systems by mapping them to the ATRC model. This approach overcomes the lack of direct tools (like the Ornstein-Zernike theory) for the Potts model at criticality.
4. Implications for Other Models: The results have immediate implications for the Six-Vertex model (specifically the height function with step boundary conditions) and the Ashkin-Teller model, showing that their level lines also converge to the Brownian watermelon in the relevant regimes.

Summary

In essence, Dober, Glazman, and Ott prove that at the critical temperature of the 2D Potts model with $q > 4$, the interface separating two ordered phases is not a single line but a wetting layer of width $\sqrt{N}$. The boundaries of this layer fluctuate and, upon diffusive scaling, converge to a pair of non-intersecting Brownian bridges. This is achieved by mapping the problem to the ATRC model, proving entropic repulsion between the interfaces, and coupling them to conditioned random walks.