Moment bounds and exclusion processes on random Delaunay triangulations with conductances

This paper establishes moment bounds for weighted degrees on random Delaunay triangulations derived from stationary point processes, demonstrating how these integrability properties ensure the well-definedness and key characteristics of symmetric simple exclusion processes while extending construction results to non-symmetric cases under specific dependence and boundedness conditions.

The Gist

Imagine you are a city planner trying to build a road network in a brand-new, chaotic city. But there's a catch: you don't get to decide where the houses are. Instead, the houses appear randomly, like stars scattered across the night sky or seeds blown by the wind.

This paper is about understanding the traffic rules and structural integrity of a city built on such random foundations.

Here is the breakdown of the paper using simple analogies:

1. The City Layout: Voronoi and Delaunay

First, the authors imagine a city where every house has a "territory" (a Voronoi cell). This territory is the area of land that is closer to that specific house than to any other house.

• The Analogy: Think of a pizza cut into slices, but the cuts aren't straight lines from the center. Instead, the cuts are jagged, curvy lines that ensure every point on the pizza is closest to the specific pepperoni slice it sits on.
• The Connection: If two houses share a border on their territories, they are considered "neighbors." The authors draw a line between these neighbors. This web of lines connecting neighbors is called the Delaunay triangulation. It's the "skeleton" of the city's road network.

2. The Traffic Lights: Conductances

Now, imagine that every road between neighbors has a "traffic light" or a "speed limit." In math terms, this is called conductance.

• High Conductance: The road is wide, smooth, and fast. Traffic flows easily.
• Low Conductance: The road is a muddy dirt path. Traffic is slow or blocked.
• Randomness: These road qualities are random. Some days the road is a highway; other days it's a dead end.

3. The Big Question: Is the City Safe?

The authors want to know: Can we trust this random city?
Specifically, they are worried about two things:

1. The "Degree" Problem: Does any single house have too many neighbors? If a house is connected to 1,000 other houses, it becomes a chaotic hub. The paper proves that under certain conditions, no house will have an unmanageable number of neighbors.
2. The "Traffic Jam" Problem: If the roads are too slow (low conductance) or the city is too sparse, could the whole network break apart? Could a "blackout" happen where a huge chunk of the city gets cut off from the rest?

4. The "Fundamental Region" (The Magic Shield)

To solve these problems, the authors invented a clever trick called the "Fundamental Region."

• The Analogy: Imagine you are standing in your house. You want to know how far your neighbors are. Instead of looking at the whole infinite city, you build a giant, invisible "shield" or bubble around your house.
• The Trick: They proved that all your neighbors must be inside this bubble. If you can prove the bubble isn't too big, you know your neighbors aren't too far away, and you don't have too many of them. This "shield" allows them to calculate the odds of the city staying connected without getting lost in the infinite chaos.

5. The "Exclusion Process" (The Party Game)

The paper also looks at a game called the Simple Exclusion Process (SEP).

• The Game: Imagine particles (like people) moving around the city. There is a rule: No two people can be in the same house at the same time. If a person wants to move to a neighbor's house, they can only do so if the neighbor's house is empty.
• The Symmetric Case (Fair Play): If the rules are fair (you can move left or right with equal ease), the authors show that as long as the road qualities aren't too crazy, the game works perfectly. The people will eventually spread out evenly across the city.
• The Asymmetric Case (Unfair Play): What if the rules are rigged? Maybe it's easier to move East than West. This is harder to analyze. The authors found that if the city has "short-range" dependencies (meaning a house's neighbors don't affect houses miles away) and the roads aren't infinitely fast, the game still works. They proved this by showing that the "bad" roads (the ones that block movement) don't form a giant, city-wide wall.

6. The "Percolation" Test (The Flood)

To prove the city won't get cut off, they used a concept called Bond Percolation.

• The Analogy: Imagine it starts raining. Every road has a chance of flooding and becoming impassable.
• The Result: The authors proved that if the rain isn't too heavy (the probability of a road being blocked is low enough), the flood will only create small, isolated puddles. It will not create a giant ocean that cuts the city in half. This ensures that the "game" (the particle movement) can still happen everywhere.

Summary: Why Does This Matter?

This paper is a mathematical safety manual for random systems. It tells us:

• When is a random network stable? (When the "shields" around points are finite).
• When can we model traffic or particles on it? (When the "traffic lights" aren't too extreme).
• What happens if the rules are unfair? (It still works, provided the randomness isn't too "long-range").

The authors essentially built a mathematical "insurance policy" that guarantees these random, messy networks won't collapse, allowing scientists to use them to model real-world things like electricity flowing through disordered materials, traffic in a chaotic city, or how diseases spread through a population with irregular contact patterns.

Technical Summary

1. Problem Statement and Context

The paper investigates stochastic processes defined on random Delaunay triangulations generated by stationary simple point processes (SPPs) in $\mathbb{R}^d$. Specifically, the authors consider a weighted graph where:

• Vertices: Points of a stationary SPP $\xi$.
• Edges: Pairs of points whose Voronoi cells share a $(d-1)$-dimensional face.
• Weights (Conductances): Random, non-negative values $c_{x,y}$ assigned to edges, representing conductance in a resistor network or jump rates in particle systems.

The central challenge is that the geometry of the Delaunay triangulation is highly sensitive to the point configuration. Local changes in the point process can induce long-range changes in the graph structure (e.g., vertex degrees, neighbor distances). This geometric randomness, combined with probabilistic dependencies in the SPP, makes it difficult to establish integrability conditions (moment bounds) for key functionals.

These bounds are crucial for applying existing theories on:

1. Random Walks in random environments.
2. Resistor Networks (homogenization).
3. Symmetric Simple Exclusion Processes (SSEP).
4. General Simple Exclusion Processes (SEP) with non-symmetric jump rates.

2. Methodology

The authors develop a rigorous framework to derive moment bounds with respect to the Palm distribution $P_0$ (the probability measure conditioned on a point existing at the origin).

A. The Fundamental Region Approach

Inspired by previous works on Poisson processes, the authors introduce the Fundamental Region $D(x|\xi)$.

• Definition: For a vertex $x$, $D(x|\xi)$ is the union of closed balls centered at the vertices of the Voronoi cell $Vor(x|\xi)$ with radii equal to the distance from $x$ to those vertices.
• Key Geometric Lemma: All neighbors of $x$ in the Delaunay triangulation lie within $D(x|\xi)$.
• Bounding Strategy: By analyzing the "shield" of boxes surrounding the origin, the authors prove that if the point process has points in specific surrounding regions, the degree of the origin and the distance to its neighbors are bounded by the number of points in a large ball. This reduces geometric control to counting points in boxes.

B. Probabilistic Assumptions

The analysis covers three distinct classes of SPPs regarding correlation decay:

1. Finite Range of Dependence: Points in distant regions are independent.
2. Positive Association: Correlations are non-negative (e.g., Gibbsian processes with attractive interactions).
3. General Stationary/Ergodic: No specific correlation structure, relying on Void Probability decay conditions (Condition $C(\alpha)$).

C. The $\mathbb{Z}^d$-Process for Percolation

To handle non-symmetric exclusion processes, the authors employ a Bernoulli bond percolation analysis. They construct a coarse-grained $\mathbb{Z}^d$-process (inspired by works on Voronoi percolation) to map the continuous Delaunay graph to a discrete lattice. This allows them to use standard percolation arguments to prove that the graph has only finite connected components under specific conditions.

3. Key Contributions and Results

A. Moment Bounds for Geometric and Conductance Functionals

The paper provides sufficient conditions for the finiteness of moments of several critical quantities under the Palm distribution $P_0$:

1. Weighted Degrees: $E_0[(\text{deg}_{DT}(0))^p] < \infty$.
2. Spatial Moments: $E_0[\sum_{x \sim 0} |x|^\zeta] < \infty$.
3. Conductance Sums:
   • $\mu_\omega(0) = \sum_{x \sim 0} c_{0,x}$ (Total conductance).
   • $\nu_\omega(0) = \sum_{x \sim 0} c_{0,x}^{-1}$ (Total resistance).
   • $\lambda_k = \sum_{x \sim 0} c_{0,x} |x|^k$.

Main Theorems (4.8, 4.9, 4.13):

• Finite Range Dependence: If the SPP has finite range of dependence and the intensity moments $\rho_\gamma$ are finite, the bounds hold.
• Positive Association: If the SPP has positive association and satisfies a void probability decay condition ($P(\xi(\Lambda_\ell)=0) \le \kappa \ell^{-\alpha}$), the bounds hold.
• General Case: If the SPP satisfies Condition $C(\alpha)$ (polynomial decay of void probabilities) and has finite intensity moments, the bounds hold provided $\alpha$ is sufficiently large relative to the dimension $d$ and the moment order.

B. Application to Symmetric Simple Exclusion Process (SSEP)

The authors show that the derived moment bounds ($\lambda_0, \lambda_2 \in L^1(P_0)$) are sufficient to apply existing results (from [11, 12, 14, 15]) to the SSEP on Delaunay triangulations.

• Result: Under these integrability conditions, the SSEP admits a rigorous graphical construction, is a Feller process, and satisfies a hydrodynamic limit (macroscopic density evolution) described by a diffusion equation $\partial_t \rho = \nabla \cdot (D \nabla \rho)$ with an effective homogenized matrix $D$.

C. Construction of Non-Symmetric Exclusion Processes (SEP)

For generic exclusion processes with non-symmetric jump rates, well-definedness is more delicate.

• Assumption SEP: The paper introduces a condition requiring that a "thinned" Delaunay graph (where edges are kept with probability $1-e^{-t_0 c_{x,y}}$) has only finite connected components almost surely.
• Theorem 6.2: If the SPP has finite range of dependence and the conductances are uniformly bounded, Assumption SEP is satisfied.
• Method: This is proven via a subcritical Bernoulli bond percolation argument on the Delaunay triangulation, utilizing the $\mathbb{Z}^d$-process to show that infinite clusters do not form when the edge retention probability is low enough.

D. Examples

The paper validates these conditions for specific point processes:

• Poisson Point Processes: Satisfy all conditions (exponential void decay).
• Determinantal Point Processes: Have negative association, satisfying the conditions.
• Gibbsian Point Processes: Satisfy conditions if the potential is locally stable and has finite range (or decays sufficiently fast).

4. Significance

1. Generalization: The results significantly extend previous work (which was largely restricted to Poisson processes or stronger assumptions) to a broad class of stationary and ergodic point processes, including those with long-range correlations (provided they satisfy specific decay or association properties).
2. Bridging Geometry and Probability: The paper provides a robust toolkit (Fundamental Region + Void Probability analysis) to control geometric properties of random graphs generated by point processes, which is a non-trivial task due to the non-local nature of Delaunay triangulations.
3. Enabling Non-Symmetric Models: By establishing the validity of Assumption SEP for bounded conductances and finite-range dependence, the paper enables the rigorous study of non-equilibrium particle systems (non-symmetric SEP) on random geometric graphs, a topic previously lacking general existence results.
4. Homogenization: The moment bounds are the necessary prerequisites for proving homogenization (diffusive scaling limits) for random walks and exclusion processes in these complex, disordered media.

In summary, this paper establishes the foundational probabilistic estimates required to treat random Delaunay triangulations as a valid and tractable model for disordered media in statistical mechanics and probability theory.