Excursion decomposition of the XOR-Ising model

✨

This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are looking at a vast, chaotic ocean of tiny magnets (spins) on a grid. This is the Ising model, a classic way physicists study how things like iron become magnetic. Now, imagine you take two of these oceans and multiply them together. The result is the XOR-Ising model. It's a bit like taking two noisy radio signals and mixing them to create a new, complex sound.

For decades, physicists knew that this "mixed" signal looked a lot like the ripples on a Gaussian Free Field (GFF). You can think of the GFF as a perfectly smooth, random landscape of hills and valleys (like a very wavy, invisible terrain). The XOR-Ising model, they suspected, was just the sine or cosine of this wavy terrain.

But there was a missing piece of the puzzle: How do we break this complex signal down into its simplest, independent building blocks?

This paper, by Tomás Alcalde López and Avelio Sepúlveda, solves that puzzle. They show how to take this chaotic, mixed-up field and decompose it into a collection of distinct, non-overlapping "excursions" (like separate islands or bubbles) that float on the landscape.

Here is the breakdown of their discovery using simple analogies:

1. The "Excursion Decomposition": Breaking the Ocean into Islands

Imagine the XOR-Ising field as a stormy sea. The authors show that you can describe this entire storm not as one big mess, but as a collection of independent islands (called excursions).

The Islands: These are connected regions where the field is "active."
The Signs: Each island is painted either White (+) or Black (-). These colors are chosen randomly, like flipping a coin for each island.
The Magic: If you add up all the White islands and subtract all the Black islands, you get the exact same stormy sea you started with.

The paper proves that this works for the "continuous" version of the model (the smooth, mathematical limit) and that it emerges naturally from the "discrete" version (the actual grid of magnets).

2. The "Height Function": The Hidden Map

How did they find these islands? They used a secret map called the Height Function.

Imagine the grid of magnets is actually a topographic map. The "height" at any point tells you the state of the magnets.
The XOR-Ising model is essentially the cosine of this height map (for one type of boundary) or the sine (for another).
The "islands" (excursions) correspond to specific contour lines on this map. When the height crosses a certain threshold, it creates a loop. The authors show that if you trace these loops, you find the boundaries of your islands.

3. The "Two-Valued Set": The Skeleton of the Storm

To find the islands, the authors looked at something called a Two-Valued Set (TVS).

Think of the GFF landscape as having a "skeleton" of loops drawn on it. These loops are special because the "height" of the terrain inside them is always either $+H$ or $-H$ .
The paper proves that the XOR-Ising field is zero everywhere except on these loops and the areas they enclose.
The "excursions" are essentially the areas inside these loops. The authors developed a recursive algorithm:
1. Find the big loops (the outer boundaries).
2. Inside those loops, find smaller loops.
3. Inside those, find even smaller ones.
4. Repeat until you have mapped the whole structure.

4. The "Scaling Limit": From Pixels to Paint

The paper has two main parts:

Part 1 (The Theory): They built the decomposition directly using the smooth, continuous math of the GFF. They showed that for any parameter $\alpha$ (which controls the "strength" of the mixing), you can break the field down into these independent islands.
Part 2 (The Proof): They went back to the original grid of magnets (the discrete model). They showed that as you make the grid finer and finer (zooming out until the pixels disappear), the "islands" formed by the grid magnets perfectly match the "islands" they constructed in the continuous theory.

Why is this a Big Deal?

In physics, when you have a system where everything is connected (dependent), it's hard to understand.

Analogy: Imagine trying to understand a crowded party where everyone is talking to everyone else. It's chaotic.
The Breakthrough: This paper shows that if you look at the party through the right lens (the GFF height map), you can see that the crowd is actually made up of distinct, non-interacting groups (the islands). Each group is independent, and the only thing connecting them is a random coin flip (the sign).

This allows physicists to:

Simplify calculations: Instead of dealing with the whole messy system, they can study the islands one by one.
Connect different worlds: It bridges the gap between the discrete world of computer simulations (pixels) and the continuous world of calculus (smooth curves).
Predict the future: It suggests that other complex models (like the Ashkin-Teller model) might also have this hidden "island" structure, opening the door to solving even harder problems.

In a Nutshell

The authors took a complex, noisy signal (XOR-Ising), realized it was just a wave pattern (GFF), and proved that this wave pattern is actually made of a collection of independent, randomly colored bubbles. They showed that these bubbles exist in the mathematical "limit" and that they emerge naturally from the microscopic grid of magnets. It's like discovering that a chaotic storm is actually just a collection of independent, floating clouds.

1. Problem Statement

The paper addresses the construction and scaling limits of excursion decompositions for the two-dimensional critical XOR-Ising model.

Context: The XOR-Ising model is the product of two independent Ising models ( $\tau = \sigma \tilde{\sigma}$ ). It is a specific instance of the Ashkin-Teller model.
Background: While excursion decompositions (representing a field as a sum of independent, disjoint "excursions" with random signs) are well-understood for the Ising model (via FK clusters) and the Gaussian Free Field (GFF), the XOR-Ising model presents unique challenges. It lacks a simple Markov property, and its continuum limit is identified with the real and imaginary parts of the imaginary multiplicative chaos of a GFF, specifically $\cos(\alpha \phi)$ and $\sin(\alpha \phi)$ with $\alpha = 1/\sqrt{2}$ .
Core Question: Can the discrete excursion decomposition of the XOR-Ising model (based on Double Random Currents) be shown to converge to a continuum decomposition, and can this continuum decomposition be constructed directly using the geometry of the GFF?

2. Methodology

The authors employ a two-pronged approach, bridging discrete statistical mechanics and continuum probability theory.

A. Continuum Construction (Direct Construction)

Framework: They utilize the identification of the XOR-Ising field with the Wick-ordered trigonometric functions of a GFF: $:\cos(\alpha \phi):$ and $:\sin(\alpha \phi):$ where $\alpha \in (0, 1)$ .
Key Tool: The construction relies on Two-Valued Sets (TVS) of the GFF, denoted $A_{-a, b}$ . These are random closed sets where the GFF, conditioned on the set, takes values in $\{-a, b\}$ in the complementary components.
Iterative Exploration: The decomposition is built via an iterative algorithm:
1. Sample a critical TVS ( $A_{-2\lambda, 2\lambda}$ , related to CLE $_4$ ).
2. Use the labels of the loops of this TVS to assign random signs ( $\pm 1$ ).
3. Condition on the TVS to "peel off" a measure supported on the set, leaving a "renewed" field inside the loops.
4. Repeat the process recursively inside the loops.
Technical Challenges:
- Thinness: Proving that the field does not charge the TVS (i.e., the restriction of the field to the set is zero) for subcritical parameters, or yields a specific measure for critical parameters.
- Correlation Inequalities: Establishing monotonicity of correlation functions with respect to domain size (using Hadamard's variational formula) to control boundary effects.
- Strong Thinness: Proving $L^2(P)$ convergence of infinite sums of signed measures, allowing for rearrangement of terms independent of the signs.

B. Discrete-to-Continuum Convergence

Discrete Model: The discrete XOR-Ising model is represented via Double Random Currents (DRC). The excursions correspond to the clusters of the DRC trace.
Master Coupling: The authors exploit the "master coupling" from previous works (DCLQ25, ALHL26), which couples the XOR-Ising model, the Ising model, and a discrete height function $h_\delta$ $h_{δ}$ .
- Crucially, $\tau_\delta = \cos(\pi h_\delta)$ and $\tau^\dagger_\delta = \sin(\pi h_\delta)$ .
Convergence Strategy:
1. Prove the joint convergence of the discrete height function $h_\delta$ and the fields $\tau_\delta, \tau^\dagger_\delta$ to the continuum GFF and its chaos counterparts.
2. Establish that the local set structure (the TVS) is preserved in the limit.
3. Use a result from [AJ21] stating that if a local set for the imaginary chaos is measurable with respect to the underlying GFF, the decomposition is unique. This allows them to identify the limit of the discrete decomposition with the continuum construction without computing complex joint moments.

3. Key Contributions and Results

Theoretical Contributions

Continuum Excursion Decomposition (Theorem 1.1 & 1.2):
- Constructed a rigorous excursion decomposition for $:\sin(\alpha \phi):$ and $:\cos(\alpha \phi):$ for any $\alpha \in (0, 1)$ .
- The decomposition expresses the field as $\sum \xi_k \mu_k$ , where $\xi_k$ are i.i.d. symmetric signs and $\mu_k$ are measures supported on disjoint connected sets (excursions).
- The supports of these measures are identified with the boundaries of the loops in an iteratively sampled Two-Valued Set of the GFF.
- Proved that the sum converges almost surely in the Sobolev space $H^s$ ( $s < -1$ ) and that the order of summation does not affect the limit (provided the order is independent of the signs).
Scaling Limit Convergence (Theorem 1.4):
- Proved that the discrete excursion decomposition of the critical XOR-Ising model (based on DRC clusters) converges in distribution to the continuum decomposition constructed above.
- This establishes a rigorous link between the discrete DRC clusters and the continuum CLE $_4$ -based TVS.
Joint Convergence and Wick Products (Corollary 1.5):
- Demonstrated that the joint law of the height function, the Ising fields, and the XOR-Ising fields converges to the joint law of the GFF and its Wick products.
- Specifically, identified the Wick product of two independent Ising fields as the cosine chaos: $:\sigma \tilde{\sigma}: = :\cos(h/\sqrt{2}):$ .
New Technical Tools:
- Correlation Inequalities: Derived new domain monotonicity results for the two-point functions of $:\cos(\alpha \phi):$ and $:\sin(\alpha \phi):$ using Hadamard's variational formula.
- Strong Thinness: Developed a technique (Lemma 4.16) to prove $L^2$ convergence of signed sums for $\alpha \in [1/2, 1)$ , overcoming limitations of previous methods that relied on classical thinness arguments.

4. Significance

Unification of Models: The paper unifies the geometric understanding of the XOR-Ising model with the theory of Gaussian Free Fields and Imaginary Multiplicative Chaos. It confirms that the "bosonization" correspondence (Ising $\to$ GFF) holds at the level of excursion decompositions, not just correlation functions.
Non-Markovian Decompositions: It provides a rare example of an excursion decomposition for a model (XOR-Ising) that does not possess a simple Markov property, showing that such decompositions can arise from the underlying GFF structure even when the field itself is non-Markovian.
Ashkin-Teller Generalization: The results lay the groundwork for understanding the scaling limits of the broader Ashkin-Teller model. The authors conjecture that similar decompositions exist for the Ashkin-Teller polarization field along its critical line, parameterized by $\alpha(U)$ .
Methodological Advance: The technique of using local set couplings and measurability arguments (via [AJ21]) to prove convergence of decompositions offers a powerful alternative to the traditional method of computing joint moments, which is often intractable for complex fields.

5. Conclusion

Alcalde López and Sepúlveda successfully construct the excursion decomposition for the continuum XOR-Ising model and prove its emergence as the scaling limit of the discrete model. By leveraging the deep connection between the XOR-Ising model and the GFF, they provide a geometric description of the field's structure in terms of nested loops (CLE $_4$ ) and random signs, resolving a long-standing question about the geometric nature of the critical XOR-Ising model in the continuum.