Numerical simulations of the spread from the mean of the SLE and Multiple SLE dynamics

This paper presents numerical simulations using Euler's Method to analyze the spread of Schramm-Loewner Evolution (SLE) and Multiple SLE dynamics from their mean behavior, revealing that the distribution of deviations is bimodal or bell-shaped depending on the initial position and parameter $\kappa$ in standard SLE, while remaining consistently bell-shaped for Multiple SLE driven by Dyson Brownian Motion across varying $\beta$ parameters.

The Gist

Imagine you are watching a crowd of people trying to walk through a maze. In the world of this paper, the "maze" isn't made of walls, but of invisible mathematical forces that push and pull them. The paper is essentially a report on a computer simulation that watched how these "people" (mathematical curves) move, and specifically, how much they wander away from the average path.

Here is a breakdown of what the authors did, using simple analogies:

The Two Types of Walkers

The paper studies two different types of "walkers" (mathematical models called SLE and Multiple SLE):

1. The Solo Walker (SLE): Imagine one person walking through the maze. Their path is guided by a "driver," which is like a drunk friend pushing them randomly left or right (this is called Brownian Motion). The authors wanted to see: if you asked 5,000 people to take this walk, how much would their paths differ from the "average" path?
2. The Group Walkers (Multiple SLE): Now, imagine a whole group of people walking at once. But here's the catch: they are repelled by each other, like magnets with the same pole facing one another. They can't get too close, or they push each other away violently. This is called "Dyson Brownian Motion." The authors tried to simulate a whole group of these people walking together to see how their collective path spreads out.

The Experiment: "The Spread"

The researchers wanted to measure the "spread." Think of it like this:

• If you draw the "average" path in the middle of the road, how far do the individual walkers stray from that line?
• They measured two things:
  1. How far the walker is from the average distance (the absolute spread).
  2. How far the walker is from the average position on the left-right axis (the real part).

The Starting Point Matters

The authors tested two different starting spots for the walkers:

• Starting Close to the "Wall" (z = 1.02i): Imagine starting right next to a cliff edge. When the walkers started here, the results were chaotic. The distribution of where they ended up looked like a two-humped camel (bimodal). They tended to split into two distinct groups rather than clustering in the middle.
• Starting Far Away (z = 3i): Imagine starting far out in the open field, away from the edge. Here, the walkers behaved much more predictably. They clustered tightly around the average path, forming a classic bell curve (like a normal distribution). The further they started from the chaos, the "nicer" and more orderly their movement became.

The Group Challenge

Simulating the group of walkers (Multiple SLE) was much harder. Because the "magnets" pushing them apart get stronger the closer they are, the computer had to work very hard to keep them from crashing into each other numerically.

• The Result: Unlike the solo walker who sometimes split into two groups, the group walkers always formed a nice, single bell curve, no matter where they started.
• The "Knob" (Parameters): The authors turned a "knob" (changing parameters $\kappa$ and $\beta$) to see how the noise affected the walk. They found that when the "noise" was louder (higher $\kappa$), the walkers spread out more, just as you would expect if the wind was blowing harder.

Why This Matters (According to the Paper)

The authors aren't claiming this solves a medical problem or predicts stock markets right now. Instead, they are acting like cartographers of a new mathematical landscape.

• They built a map of what these random curves look like when they move.
• They found that the shape of the "spread" changes depending on where you start and how many walkers you have.
• They are handing these "maps" to other mathematicians, saying, "Here is what our computers see; now, please go and prove why this happens using pure math."

In short, this paper is a numerical field guide. It says, "If you simulate these specific mathematical curves, here is the shape of the chaos you will see, and it depends heavily on how close you start to the edge of the world."

Technical Summary

Technical Summary: Numerical Simulations of the Spread from the Mean of SLE and Multiple SLE Dynamics

Problem Statement
The Schramm-Loewner Evolution (SLE) describes a family of fractal curves arising as scaling limits in planar Statistical Physics models, driven by a standard Brownian motion $W_t = \sqrt{\kappa}B_t$ in the Loewner differential equation. A multi-driver extension, known as Multiple SLE, replaces the single Brownian driver with a Dyson Brownian Motion (DBM) system, where particles interact via Coulombic repulsion. While the theoretical behavior of these systems is well-studied, there is a need for numerical investigation into the statistical "spread" of the dynamics from their sample average behavior at a fixed time. Specifically, this work addresses the computational challenges of simulating these dynamics—particularly the singularities arising from DBM interactions—and aims to provide numerical predictions for the distributions of the quantities $|g_t(z) - \bar{g}_t(z)|$ and $\text{Re}(g_t(z)) - \text{Re}(\bar{g}_t(z))$, where $\bar{g}_t(z)$ denotes the sample mean.

Methodology
The authors employed Euler's Method to numerically simulate both the Dyson Brownian Motion drivers and the resulting Loewner maps $g_t(z)$.

• Driver Simulation: For the Multiple SLE case, the DBM particles $\lambda^i_t$ were simulated using a discretized scheme involving independent normal increments and a repulsion term. To mitigate numerical instabilities caused by the Coulombic singularity (where particles approach each other), the authors utilized a relatively short final time ($T=0.25$) and ensured initial particle spacing was sufficient.
• Map Simulation: The conformal maps $g_t(z)$ were evolved using a separate Euler scheme driven by the sampled driver values.
• Validation: The accuracy of the algorithm was verified by comparing finite-$N$ simulations against the known closed-form theoretical solution for the limit $N \to \infty$ (involving the Lambert $W$ function), confirming convergence as $N$ increased.
• Experimental Design: The study examined two starting points, $z_0 = 1.02i$ (near the theoretical hull boundary) and $z_0 = 3i$ (further from the origin), across various parameters: $\kappa \in \{1, 2, 8/3, 3, 4, 5\}$ for standard SLE and $\beta \in \{1, 2, 3, 4\}$ for Multiple SLE. For each configuration, thousands of independent trajectories were generated to construct histograms of the spread metrics. Distribution fitting was performed using SciPy's library, selecting the model with the lowest Sum Squared Error (excluding Levy Stable due to computational cost).

Key Results
The numerical experiments yielded distinct distributional behaviors depending on the starting point and the model type:

1. Standard SLE ($N=1$):

   • Near the Hull ($z_0 = 1.02i$): The distribution of the real part spread, $\text{Re}(g_t(z)) - \text{Re}(\bar{g}_t(z))$, was predicted to be bimodal, often well-described by a Wrapped Cauchy distribution.
   • Far from the Hull ($z_0 = 3i$): The distribution became bell-shaped (unimodal), indicating tighter clustering around the mean. This aligns with the theoretical intuition that $g_t(z) \to z$ as $|z| \to \infty$, resulting in less distortion.
   • Parameter Dependence: As $\kappa$ increases (and $\beta$ decreases), the spread around the sample mean increases, consistent with the amplification of noise.

2. Multiple SLE (DBM Drivers):

   • Across all tested parameters ($\beta$) and particle counts ($N$), the distribution of the real part spread consistently exhibited a bell-shaped profile.
   • The absolute value spread $|g_t(z) - \bar{g}_t(z)|$ displayed a left-skewed distribution.
   • Similar to the standard SLE case, a decrease in $\beta$ (corresponding to an increase in $\kappa$) resulted in a wider spread of values.

Significance and Claims
The authors position this work as a foundational step in the numerical study of SLE and Multiple SLE dynamics, specifically focusing on the statistical properties of their deviation from the mean.

• Predictive Value: The primary scope is to provide numerical predictions to guide future theoretical investigations into these specific observables. The authors explicitly state they do not claim to have solved the underlying theoretical problems but rather to have generated data that may inform them.
• Computational Challenges: The paper highlights the computational difficulties inherent in simulating Multiple SLE, particularly the singularities in the driver dynamics and the storage requirements for large datasets.
• Future Directions: The authors modestly suggest that their results could inspire further theoretical work. They propose extending the analysis to the imaginary parts of the maps, studying the spread as a time-dependent process rather than at a fixed time, and improving numerical schemes (e.g., using Tamed Euler Schemes) to handle the singularities more robustly. They also note the potential for applying these methods to coupled systems with shared noise drivers, a concept previously explored in Random Matrix Theory.

The paper concludes by acknowledging that these are initial attempts to model the dynamics and their spread, with the hope of stimulating the Scientific Computing community to develop better tools for addressing the specific challenges of Multiple SLE simulations.