UST branches, martingales, and multiple SLE(2)

This paper establishes that the local scaling limit of multiple boundary-to-boundary branches in a uniform spanning tree is a local multiple SLE(2) process, a result derived by constructing a weighted martingale observable that relies on the discrete domain Markov property and the convergence of partition functions.

The Gist

Imagine you are standing in a vast, foggy city made of a perfect grid of streets. This city is a Uniform Spanning Tree (UST). In this city, there are no loops; every street connects to every other street in exactly one way, forming a giant, branching tree structure that covers the whole map.

Now, imagine you drop a leaf at a specific spot inside the city. The leaf doesn't just fall straight down; it gets caught in the wind and travels along the streets, eventually reaching the city's edge (the boundary). If you do this many times from different starting points, you get many "branches" of the tree stretching from the inside to the outside.

This paper is about what happens to these branches when you zoom out so far that the individual streets disappear, and the city looks like a smooth, continuous landscape. The author, Alex Karrila, proves that these random, jagged branches settle into a very specific, beautiful, and predictable pattern known as SLE(2) (Schramm-Loewner Evolution).

Here is the breakdown of the paper's journey, using simple analogies:

1. The Problem: Too Many Branches to Count

In the past, mathematicians knew what happened if you had just one branch growing from the inside to the outside. It was like watching a single river flow to the sea; it followed a specific, wiggly path called SLE(2).

But what if you have many branches growing at the same time? They can't cross each other (they are like separate rivers that can't merge). They have to weave around one another. The question was: Do these multiple branches still follow a predictable pattern, or do they just become a chaotic mess?

The paper says: Yes, they are predictable. They don't just follow SLE(2); they follow a special version called "Local Multiple SLE(2)."

2. The Secret Weapon: The "Martingale" (The Fair Game)

To prove this, the author uses a mathematical tool called a Martingale.

• The Analogy: Imagine you are playing a fair gambling game. You have a bag of chips. Every time you make a move, the expected number of chips you will have in the future is exactly the same as what you have right now. You can't predict if you will win or lose the next hand, but you know the "average" stays steady.
• In the Paper: The author finds a special "observable" (a way of measuring the tree) that acts like this fair game. Even as the tree branches grow step-by-step, this measurement stays "fair" (it's a martingale).
• The Twist: The author takes the known "fair game" for a single branch and uses a clever mathematical trick (a Girsanov transform, which is like a weighted filter) to turn it into a "fair game" for multiple branches. This new game depends on a "Partition Function," which is essentially a score that tells you how likely a specific arrangement of branches is.

3. The Zoom-Out: From Pixels to Paint

The paper then asks: What happens when we zoom out to the "scaling limit" (when the grid size becomes zero)?

• The Discrete World: On the grid, the branches are jagged lines made of steps.
• The Continuous World: As we zoom out, these jagged lines smooth out into fluid curves.
• The Result: The author proves that the "fair game" (the martingale) on the grid turns into a "fair game" in the smooth, continuous world. Because the rules of the game in the smooth world are very strict, the only way the game can remain "fair" is if the branches follow the specific SLE(2) path.

4. The "Partition Function" as a Compass

Think of the Partition Function as a magnetic compass for the branches.

• In a simple world with one branch, the compass just points "forward."
• In a world with multiple branches, the compass is more complex. It feels the presence of all the other branches. If a branch gets too close to another, the "magnetic pull" changes, nudging it away to keep them from colliding.
• The paper shows that this "magnetic pull" is exactly what drives the SLE(2) process. The branches aren't just wandering randomly; they are being gently guided by this invisible force to maintain a specific, elegant shape.

5. The "Boundary-Visiting" Branch (The Detour)

The paper also looks at a slightly different scenario: What if a branch is forced to touch a specific point on the boundary before reaching its final destination?

• The Analogy: Imagine a river that must flow past a specific lighthouse on the shore before it can reach the ocean.
• The author shows that even with this extra rule, the river still follows a predictable SLE path, but the "compass" (the partition function) changes to account for the lighthouse. This proves the method is flexible and can handle complex rules.

Why Does This Matter?

This isn't just about trees or rivers. These mathematical models describe the behavior of critical phenomena in physics—things like:

• How magnets align at a specific temperature.
• How electricity flows through a material right before it becomes a superconductor.
• The shapes of soap bubbles or crystal growth.

By proving that these complex, multi-branch systems settle into a specific, beautiful mathematical pattern (Multiple SLE), the author gives physicists and mathematicians a powerful new tool to understand how nature organizes itself at the smallest scales.

In a nutshell: The paper proves that even when you have a chaotic forest of many growing branches, if you zoom out far enough, they don't look like a mess. They look like a perfectly choreographed dance, guided by a hidden mathematical rhythm that the author successfully decoded.

Technical Summary

Technical Summary: UST branches, martingales, and multiple SLE(2)

1. Problem Statement
The paper addresses the scaling limit of multiple boundary-to-boundary branches in a Uniform Spanning Tree (UST) on planar graphs. Specifically, it investigates the behavior of $N$ disjoint paths connecting $2N$marked boundary edges in a UST as the mesh size$\delta \to 0$. While the scaling limit of a single UST branch is known to be Schramm-Loewner Evolution with parameter$\kappa=2$ (SLE(2)) [Sch00, LSW04, Zha08], the convergence of multiple interacting branches to a "local multiple SLE(2)" had been conjectured but not rigorously proven.

The core challenge lies in characterizing the joint distribution of these multiple curves. Unlike single curves, multiple curves in a UST are conditioned on a specific "link pattern" (how the endpoints pair up). The paper aims to prove that the driving function of the Loewner evolution describing one such branch converges to a specific SLE(2) process weighted by a partition function that encodes the interactions with the other $N-1$ branches.

2. Methodology
The author employs a martingale observable approach, extending the techniques used for single-curve convergence [Zha08] to the multi-curve setting. The methodology consists of three main pillars:

• Discrete Combinatorics and Martingales:

  • The paper utilizes the discrete domain Markov property of the UST.
  • It constructs discrete martingales under conditional measures. Specifically, it starts with the known martingale for a single UST branch (related to the ratio of Poisson kernels) and applies a discrete Girsanov transform.
  • This transform weights the single-branch martingale by the ratio of discrete partition functions (connectivity probabilities) of the multi-branch model. The partition functions are expressed in terms of discrete excursion kernel determinants (Theorem 3.1), which are solved combinatorially using results from Kenyon and Wilson [KW11a, KW11b].

• Convergence of Observables:

  • The proof relies on the convergence of discrete harmonic objects (Green's functions, Poisson kernels, and excursion kernels) on isoradial graphs to their continuous counterparts.
  • Theorem 4.1 establishes the uniform convergence of ratios of these discrete kernels to continuous Poisson and excursion kernels. A key technical tool here is a Beurling-type estimate for random walk excursions (Appendix A), which ensures precompactness and controls boundary behavior.

• Identification via Continuous Martingales:

  • Using the weak convergence of the driving functions and the convergence of the discrete martingales, the author constructs a continuous martingale in the scaling limit.
  • By applying Itô's calculus to this limiting martingale, the author derives the stochastic differential equation (SDE) governing the driving function. The vanishing of the drift term in the martingale condition forces the driving function to satisfy the SDE of a local multiple SLE(2).

3. Key Contributions

• Proof of Local Multiple SLE(2) Convergence (Theorem 2.1): The primary contribution is the rigorous proof that the scaling limit of multiple UST boundary-to-boundary branches is a local multiple SLE(2). The driving function $W_t$ satisfies:
  $$dW_t = \sqrt{2} dB_t + 2 \frac{\partial_j Z_\star(X_t)}{Z_\star(X_t)} dt$$
  where $Z_\star$ is a specific partition function (either the total partition function $Z_N$ or a specific link-pattern partition function $Z_\alpha$).

• Convergence of Link Pattern Probabilities (Theorem 2.2): The proof simultaneously establishes the convergence of the probabilities that the branches form a specific link pattern $\alpha$. These probabilities converge to the ratio of the corresponding partition functions:
  $$\lim_{n \to \infty} P(\text{pattern } \alpha) = \frac{Z_\alpha(X_0)}{Z_N(X_0)}$$
  This result holds without imposing boundary regularity assumptions on the domain, a significant improvement over previous results.

• Verification of PDEs for Partition Functions (Theorem 2.3): The paper provides an alternative proof that the partition functions $Z_N$ and $Z_\alpha$ satisfy the system of partial differential equations (PDEs) associated with local multiple SLEs. These PDEs are identical to the degenerate equations for correlation functions in Conformal Field Theory (CFT).

• Generalizability to Boundary-Visiting Branches (Section 6): The author demonstrates the robustness of the method by sketching a proof for a single UST branch conditioned to visit specific boundary points. This is shown to converge to a boundary-visiting SLE(2), utilizing a bijection between boundary-visiting branches and multiple disjoint branches.

4. Results

• Main Theorem (2.1): The driving function of the Loewner evolution for a UST branch, conditioned on the pairing of $2N$ boundary edges, converges weakly to the SLE(2) driving function with a partition function weight.
• Partition Function Characterization: The partition functions $Z_\alpha$ are identified as linear combinations of excursion kernel determinants, explicitly linking combinatorial probabilities to analytic objects.
• Martingale Characterization: The paper confirms that the specific linear combinations of partition functions derived from the discrete model are indeed the unique martingale observables required to characterize the SLE limit.

5. Significance

• Bridging Discrete and Continuous: The work provides a rigorous bridge between the combinatorial structure of the Uniform Spanning Tree (specifically the Kenyon-Wilson partition functions) and the probabilistic theory of SLE.
• Completing the Conjecture: It fills a critical gap in the proof outline provided in [Kar19], completing the proof of the conjecture that multiple UST branches converge to local multiple SLEs.
• Alternative to Global SLE: While recent works have characterized global multiple SLEs, this paper offers a "local" approach. It demonstrates that one can derive the scaling limit of the full system by analyzing the local evolution of a single curve weighted by global partition functions, rather than solving the global problem directly.
• Methodological Advancement: The use of the discrete Girsanov transform to lift single-curve results to multi-curve settings is presented as a generalizable technique applicable to other lattice models (e.g., FK-Ising, percolation), provided the corresponding partition function convergence can be established.
• Connection to CFT: By deriving the PDEs satisfied by the partition functions directly from the probabilistic scaling limit, the paper reinforces the deep connection between statistical mechanics models on lattices and Conformal Field Theory.

In summary, Karrila's paper successfully identifies the scaling limit of multiple UST branches as a local multiple SLE(2) by constructing a discrete martingale observable, proving its convergence, and using Itô calculus to identify the limiting driving process, thereby unifying combinatorial partition functions with conformally invariant stochastic processes.