The Wulff crystal of self-dual FK-percolation becomes round when approaching criticality

This paper proves that as the parameter $q$ of planar FK-percolation approaches the critical threshold of 4 from above in the discontinuous regime, the model's correlation length becomes isotropic, causing its Wulff crystal to become round.

The Gist

Imagine a vast, infinite floor covered in a grid of tiles. On this floor, we are playing a game of "connect the dots." We randomly decide whether to paint the lines between the dots (making them "open") or leave them blank ("closed").

The rules of our game depend on a single number, let's call it $q$. This number represents how much the game "likes" to form big, connected clusters.

• If $q$ is small (between 1 and 4), the game is gentle. As we tune the probability of painting lines, the clusters grow slowly and smoothly. At the exact tipping point, the pattern looks the same no matter how you rotate the floor. It's perfectly symmetrical, like a snowflake.
• If $q$ is large (greater than 4), the game is chaotic and abrupt. The clusters don't grow smoothly; they suddenly snap into existence. In this "discontinuous" regime, the shape of these clusters usually depends on the direction you look. If you stretch the floor horizontally, the clusters stretch with it. They aren't round; they are lopsided.

The Big Question
The authors of this paper asked a fascinating question: What happens if we are in the chaotic regime ($q > 4$) but we slowly turn the dial down, getting closer and closer to the smooth tipping point ($q = 4$)?

Do the clusters remain lopsided and chaotic right up until the very last second? Or do they suddenly "remember" how to be round and symmetrical as they approach the calm zone?

The Discovery: The Crystal Becomes Round
The answer, surprisingly, is that they become round.

The paper proves that as the parameter $q$ approaches 4 from above, the "correlation length" (a fancy way of saying how far the influence of one dot reaches before fading away) becomes isotropic.

The Analogy of the Wulff Crystal
To visualize this, imagine a drop of water freezing on a table.

• In the chaotic regime ($q > 4$), the ice doesn't form a perfect sphere. It forms a jagged, diamond-like shape (a "Wulff crystal") because the underlying grid of the table makes it easier to grow in some directions than others. It's like a crystal growing on a wooden floor with a strong grain; it follows the grain.
• As you get closer to the critical point ($q \to 4$), the paper shows that this jagged crystal slowly loses its sharp edges. It starts to look more and more like a perfect, smooth circle.
• By the time you hit $q=4$, the crystal is a perfect circle (the unit disk). The "grain" of the floor no longer matters; the ice grows equally in all directions.

How Did They Prove It? (The Magic Trick)
Proving this was hard because the chaotic regime ($q > 4$) is very sensitive. If you change the rules even slightly, the whole system changes.

The authors used a clever mathematical "magic trick" involving track exchanges.
Imagine the floor is made of two sets of parallel train tracks running in different directions.

1. The Setup: They started with a floor where the tracks were tilted at a weird angle (representing a distorted grid).
2. The Swap: They imagined a machine that could swap the tracks back and forth, one by one, until the floor looked like a standard square grid.
3. The Insight: In the smooth regime ($q \le 4$), mathematicians already knew that swapping these tracks didn't change the big picture of the clusters. The clusters were "universal"—they looked the same regardless of the grid's distortion.
4. The Breakthrough: The authors showed that even in the chaotic regime ($q > 4$), if you are very close to the tipping point ($q=4$), this "swapping" trick still works almost perfectly. The clusters are so close to the critical point that they start behaving like the smooth ones. The "drift" or bias caused by the grid's distortion vanishes.

Why Does This Matter?
This is a beautiful example of Universality in physics. It tells us that even in systems that are usually messy and direction-dependent, there is a "sweet spot" where nature forces everything to become perfectly symmetrical.

It's like watching a chaotic crowd of people running in different directions. Usually, they might bunch up on the left side of the room. But as they get closer to a specific moment of calm (the critical point), they suddenly stop favoring the left or right and start spreading out in a perfect circle.

In a Nutshell:
The paper proves that as a complex, chaotic physical system approaches its critical tipping point, it forgets the specific shape of the grid it sits on. The jagged, lopsided shapes of the clusters smooth out and become perfectly round, just like a crystal melting into a sphere. It's a mathematical proof that symmetry wins right at the edge of chaos.

Technical Summary

1. Problem Statement and Context

The Model:
The paper investigates the Fortuin-Kasteleyn (FK) random-cluster model on the two-dimensional square lattice $\mathbb{Z}^2$. This model is a unifying framework in statistical mechanics that generalizes Bernoulli percolation ($q=1$) and the Ising model ($q=2$). The model is parameterized by the edge probability $p$ and the cluster weight $q$.

The Phase Transition Regime:
A critical threshold $p_c(q) = \frac{\sqrt{q}}{1+\sqrt{q}}$ exists. The nature of the phase transition at $p_c$ depends on $q$:

• Continuous Regime ($1 \le q \le 4$): The transition is continuous. The model exhibits scale invariance, and at $q=4$, it was recently proven to possess rotational invariance (asymptotic conformal invariance) at large scales.
• Discontinuous Regime ($q > 4$): The transition is first-order (discontinuous). At the critical point $p_c$, there is no unique critical measure; instead, the system exhibits either sub-critical or super-critical behavior. Crucially, the correlation length $\xi_{p,q}(\theta)$ remains finite at $p_c$ for $q > 4$, unlike the divergent behavior in the continuous regime.

The Core Question:
While the correlation length is finite for $q > 4$, it is anisotropic (direction-dependent) for fixed $q$. The paper addresses the behavior of the model as the parameter $q$ approaches the critical value $4$ from above ($q \searrow 4$) while maintaining $p = p_c(q)$. Specifically, the authors ask: Does the correlation length become isotropic (direction-independent) as $q \to 4$?

This is equivalent to asking whether the Wulff shape (the asymptotic shape of a large cluster conditioned on volume) becomes a perfect circle (disk) as $q \to 4$.

2. Methodology

The proof relies on a sophisticated combination of integrability, coupling techniques, and universality arguments adapted from the study of the critical regime ($q \le 4$).

A. Isoradial Graphs and Track-Exchanges

The authors generalize the model from the standard square lattice to isoradial rectangular lattices $L(\alpha)$. These are embeddings of $\mathbb{Z}^2$ where edges have lengths determined by an angle parameter $\alpha$.

• Key Tool: The Star-Triangle Transformation (or Track-Exchange). This is an exact local transformation that allows one to map a configuration on a lattice with a sequence of angles $\alpha$ to a lattice with a permuted sequence of angles, preserving the measure (for specific boundary conditions).
• Strategy: By applying a sequence of track-exchanges, one can continuously deform a lattice $L(\alpha)$ into a lattice $L(\beta)$. If a quantity is invariant (or nearly invariant) under these exchanges, it implies universality across different lattice geometries.

B. The Challenge of $q > 4$

In the continuous regime ($q \le 4$), the uniqueness of the infinite-volume measure allows for straightforward application of track-exchanges. However, for $q > 4$, the measure is not unique, and boundary conditions influence the system at infinite distances.

• Solution: The authors introduce Half-Plane Measures ($\phi^0_{L(\alpha), q}$) with free boundary conditions on the bottom edge. They prove the uniqueness of this half-plane measure for periodic angle sequences, circumventing the non-uniqueness issue of the full plane.

C. Coupling and Drift Analysis

To compare correlation lengths on different lattices $L(\alpha)$ and $L(\beta)$, the authors construct a coupling of configurations $(\omega_t)$ evolving through a sequence of track-exchanges.

• The Observable: They track the extremal coordinate $E_\theta(C)$ of the cluster containing the origin in a specific direction $\theta$.
• The Increment: They analyze the expected change (drift) of this extremal coordinate after a full cycle of track-exchanges.
• Connection to $q=4$: The core insight is that as $q \to 4$, the local environment of a large cluster extremum in the $q > 4$ regime converges to the Incipient Infinite Cluster (IIC) measure of the $q=4$ regime.
• Vanishing Drift: It was previously established (in [DCKK+20]) that for $q=4$, the expected drift of the extremum under track-exchanges is exactly zero due to rotational symmetry. The authors prove that for $q$ sufficiently close to 4, this drift becomes arbitrarily small.

3. Key Contributions and Results

Main Theorem (Theorem 1.1)

For any $\epsilon > 0$, there exists a $q_0 > 4$ such that for all $q \in (4, q_0]$ and any two directions $\theta_1, \theta_2$:
$$\left| \frac{\xi_{p_c, q}(\theta_1)}{\xi_{p_c, q}(\theta_2)} - 1 \right| < \epsilon$$
and similarly for the point-to-hyperplane decay rate $\zeta$.
Interpretation: As $q \searrow 4$, the correlation length becomes asymptotically isotropic.

Corollary 1.2 (Roundness of the Wulff Shape)

Let $W_q$ be the Wulff shape associated with the inverse correlation length at $q$. As $q \searrow 4$:
$$\frac{W_q}{\sqrt{\text{Vol}(W_q)}} \to U$$
where $U$ is the unit disk.
Interpretation: The typical shape of a large cluster conditioned to have a large volume becomes a perfect circle as the system approaches the continuous transition point.

Universality Result (Theorem 1.3)

The authors prove that the decay rates $\xi_{\alpha, q}(\theta)$ and $\zeta_{\alpha, q}(\theta)$ on any isoradial rectangular lattice $L(\alpha)$ are asymptotically equivalent to those on the standard square lattice (or rotated square lattice) as $q \to 4$. This establishes a form of universality for the near-critical behavior in the discontinuous regime.

4. Technical Highlights of the Proof

1. Half-Plane Uniqueness: Proving that the free half-plane measure is unique for $q > 4$ on periodic lattices is a non-trivial step required to define the coupling without boundary condition ambiguity.
2. Flower Domains and Mixing: The proof utilizes "flower domains" (regions revealed by exploring interfaces) and mixing properties derived from the $q=4$ case. They show that for $q$ close to 4, the probability of "good" flower domains (well-separated petals) remains bounded away from zero.
3. Continuity in $q$: The argument relies heavily on the continuity of the random-cluster measure with respect to $q$. By showing that the measure for $q > 4$ is close in Total Variation distance to the measure at $q=4$ within local boxes (conditioned on the extremum), they transfer the "zero drift" property of the $q=4$ IIC to the $q > 4$ regime.
4. Rounding of Extremes: To avoid degeneracy issues with exact lattice positions, the authors analyze a "rounded" version of the extremal coordinate, allowing them to use martingale arguments (sub-martingales) to bound the probability of the cluster shrinking significantly during the transformation.

5. Significance and Impact

• Bridging Regimes: This work provides a rigorous mathematical bridge between the discontinuous ($q > 4$) and continuous ($q \le 4$) phases of the FK-percolation model. It confirms that the "rounding" of the Wulff crystal is a universal feature of the transition point.
• Validation of Conformal Invariance: The result supports the conjecture that the critical point $q=4$ is a conformally invariant theory. The fact that anisotropy vanishes as $q \to 4$ suggests that the conformal symmetry is restored exactly at the transition.
• Methodological Advance: The adaptation of track-exchange and IIC techniques to the non-uniqueness regime ($q > 4$) via half-plane measures opens new avenues for studying first-order phase transitions in 2D statistical mechanics.
• Physical Implication: In physical terms, it implies that as a system approaches a critical point from a first-order phase, the anisotropy of the correlation length (and thus the shape of critical droplets) disappears, becoming isotropic.

In summary, the paper proves that the "roundness" of the Wulff crystal is not just a property of the critical point itself, but a limiting behavior of the discontinuous regime as it approaches criticality, driven by the underlying rotational invariance of the $q=4$ model.