Sharp bounds on the half-space two-point function for high-dimensional Bernoulli percolation

This paper establishes an up-to-constant estimate for the critical two-point function restricted to a half-space in high-dimensional Bernoulli percolation ($d>6$), thereby completing prior work by Chatterjee, Hanson, and Sosoe and resolving a specific question posed by Hutchcroft, Michta, and Slade.

The Gist

Imagine a giant, infinite grid made of city blocks, stretching out in every direction. This is our world, called $\mathbb{Z}^d$ (where $d$ is the number of dimensions, like height, width, depth, and a few extra invisible ones).

In this city, every street connecting two blocks has a chance of being open or closed. We flip a coin for every street: if it's "heads," the street is open; if "tails," it's a dead end. This is Bernoulli Percolation.

The Big Question: How far can you walk?

If you stand at the center of the city (the origin) and start walking on open streets, how likely are you to reach a specific destination far away?

• In a normal, open city (the whole grid), mathematicians have known for a long time that the chance of reaching a friend drops off predictably as the distance grows. It's like a signal fading: the further you go, the weaker the connection, following a specific mathematical curve.

The Twist: The "Half-City" Wall

Now, imagine a giant, impenetrable wall cuts the city in half. You can only walk in the half-city (the half-space). You cannot cross the wall.

• The Problem: Does the wall change how the signal fades?
  • If you and your friend are both far from the wall, the signal fades normally.
  • If you are right up against the wall, the signal fades differently (faster).
  • But what if you are somewhere in between? How does the signal behave when you are close to the wall, but your friend is far away? Or when you are both near the wall but far apart?

Previous mathematicians had figured out the rules for the "far from wall" scenario and the "right on the wall" scenario. But they didn't have a single, perfect formula that worked for every situation in between. It was like having a map that showed the rules for the desert and the rules for the ocean, but no rule for the beach where they meet.

The Breakthrough: A New Map for the Beach

The authors of this paper, Romain Panis and Bruno Schapira, have finally found that missing rule. They proved a "sharp bound," which is a fancy way of saying they found a formula that is perfectly accurate (up to a constant factor) for any two points in this half-city.

Their formula is a bit like a traffic light system for your connection probability:

1. The Distance Factor: The further apart you are, the harder it is to connect (this is the standard fading).
2. The Wall Factor: The closer you are to the wall, the harder it is to connect because the wall blocks some of your possible paths.

Their magic formula multiplies these two effects together. It tells you exactly how much the wall "hurts" your chances of connecting, depending on how close you are to it.

How Did They Do It? (The Detective Work)

To solve this, they had to be very clever detectives. They used a technique called the "Lace Expansion" (think of it as untangling a giant knot of string to see the main threads).

But the real secret sauce was a concept called "Regular Points."

• Imagine you are trying to cross the city. Most paths are messy and chaotic.
• The authors identified special "clean" paths (Regular Points) that are very likely to exist and behave nicely.
• They proved that even though the city is random, there are always enough of these "clean" paths to guarantee a connection, provided you aren't too far apart.
• They used these clean paths to build a "bridge" between the known rules (far from the wall) and the unknown rules (near the wall), finally stitching the whole picture together.

Why Does This Matter?

You might ask, "Who cares about a half-city with invisible walls?"

1. Physics: This model helps physicists understand how materials conduct electricity or how fluids flow through porous rocks (like oil in sandstone) when there are boundaries.
2. Universality: This paper solves a question that has been bothering experts for years. It confirms that in high dimensions (6 or more), the math behaves in a very predictable, "mean-field" way. It's like finding the universal law of gravity for these random networks.
3. Completing the Puzzle: Before this, we had pieces of the puzzle. Now, we have the whole picture. It connects the work of several other famous mathematicians and settles a debate about how these systems behave near boundaries.

The Takeaway

In simple terms: The authors figured out exactly how the "noise" of a random network gets quieter as you move away from a source, specifically when there is a wall nearby. They found the perfect mathematical recipe that works for every single spot in the city, finally closing the book on a long-standing mystery in the world of random networks.

Technical Summary

Here is a detailed technical summary of the paper "Sharp bounds on the half-space two-point function for high-dimensional Bernoulli percolation" by Romain Panis and Bruno Schapira.

1. Problem Statement and Context

The Model:
The authors study Bernoulli percolation on the integer lattice $\mathbb{Z}^d$ (and spread-out models) in high dimensions ($d > 6$). This regime corresponds to the "mean-field" behavior of the model, where critical exponents are expected to match those of the branching process (specifically, the two-point function decays as $|x-y|^{-(d-2)}$).

The Specific Challenge:
While the behavior of the critical two-point function $\tau(x, y) = \mathbb{P}_{p_c}[x \leftrightarrow y]$ in the full space $\mathbb{Z}^d$ is well-understood (via the lace expansion), the behavior in a half-space $H = \{x \in \mathbb{Z}^d : x_1 \geq 0\}$ is more complex due to the lack of full translation invariance.

• Previous work (Chatterjee & Hanson, 2021; Chatterjee, Hanson, & Sosoe, 2023) established bounds for specific regimes:
  • When points are far from the boundary.
  • When one point is on the boundary and the other is far.
  • When both points are on the boundary.
• The Gap: These results did not provide a unified, sharp estimate for the interpolation between these regimes, particularly when both points are near the boundary but separated by a distance comparable to their distance from the boundary.
• The Question: Hutchcroft, Michta, and Slade (2023) conjectured the precise form of the two-point function $\tau_H(x, y)$ in the regime where $\max(x_1, y_1) \leq |x-y|$.

Goal:
To prove a sharp (up-to-constant) estimate for the critical two-point function restricted to the half-space, $\tau_H(x, y)$, for all $x, y \in H$.

---

2. Key Contributions and Main Result

Main Theorem (Theorem 1.4):
The authors prove that for $d > 6$, there exist constants $c, C > 0$ such that for all $x, y \in H$:
$$c \frac{(1 + r_{x,y})(1 + r_{y,x})}{1 + |x-y|^d} \leq \tau_H(x, y) \leq C \frac{(1 + r_{x,y})(1 + r_{y,x})}{1 + |x-y|^d}$$
where $r_{x,y} := \min(x_1, |x-y|)$.

Significance of the Result:

1. Completes Previous Work: It unifies the three previously known regimes (bulk, boundary-to-bulk, and boundary-to-boundary) into a single, coherent formula.
2. Solves a Conjecture: It confirms the conjecture by Hutchcroft, Michta, and Slade regarding the behavior in the "near-boundary" regime.
3. Sharpness: The bounds are "up-to-constant," meaning the decay rate is exact, not just an order-of-magnitude estimate.
4. Corollary: The result provides a short, alternative proof for the uniform boundedness of the expected number of critical pioneers in half-spaces (Corollary 1.6), a property previously derived using different methods for spread-out models.

---

3. Methodology

The proof relies on a combination of existing results and two novel technical ingredients. The authors work within the framework of the Lace Expansion regime (assuming the standard two-point decay $(*)$ holds) and utilize probabilistic tools like the BK inequality (van den Berg–Kesten).

A. Decomposition Strategy

The core of the proof involves decomposing the path between $x$ and $y$ into three segments:

1. A path from $x$ to a "boundary" of a box $B_n(x)$.
2. A path between the two boxes (crossing the gap).
3. A path from the boundary of $B_n(y)$ to $y$.

They establish a lower bound (reverse inequality) for this decomposition, which is technically difficult because standard BK inequalities only provide upper bounds.

B. Key Technical Ingredients

1. Reversed Inequality via Regular Points (Proposition 2.1):

• Challenge: Standard BK inequalities give $\mathbb{P}(A \cap B) \leq \mathbb{P}(A)\mathbb{P}(B)$. To get a lower bound, one must show that the events are "almost independent" or that the intersection probability is comparable to the product.
• Solution: The authors adapt the theory of regular points (introduced by Kozma & Nachmias, 2011, and refined by Aizenman, Duminil-Copin, & Sidoravicius, 2025).
• Mechanism:
  • They define "regular points" on the boundary of a box where the cluster size is well-controlled (not too large).
  • They construct "extended clusters" by adding "line good points" (segments of open edges extending orthogonally from the boundary).
  • They prove that the $(d-4)$-capacity of the set of these extended boundary points is comparable to their cardinality (Lemma 4.4).
  • Using a second-moment method (Lemma 4.1), they relate the connection probability of two sets to the product of their capacities. This allows them to derive a lower bound for the connection probability that matches the upper bound structure.

2. Summation Estimates on Boundaries (Proposition 2.2):

• Goal: Estimate the sum of connection probabilities from a point $x$ to the boundary of a box $\partial B_n(x)$.
• Result: They prove that $\sum_{u \in \partial B_n(x)} \tau_{B_n(x)}(x, u) \asymp \frac{1 + \min(x_1, n)}{n}$.
• Significance: This quantifies how the proximity to the boundary ($x_1$) affects the "flux" of connections leaving a box. It bridges the gap between the bulk behavior (decay $n^{-(d-1)}$) and the boundary behavior.

C. Proof Structure

1. Upper Bound: Uses the standard BK inequality decomposition (Equation 2.1) combined with the summation estimate (Prop 2.2) and the known bulk decay $(*)$.
2. Lower Bound:
   • Uses the "regular point" construction to define extended clusters $C_n^e(x)$ and $C_n^e(y)$.
   • Shows that the probability of connecting these extended clusters is bounded below by the product of their capacities (via Lemma 4.1).
   • Demonstrates that the capacity of the extended boundary is proportional to the number of pioneers (regular points).
   • Combines this with the summation estimates to recover the desired lower bound form.

---

4. Significance and Impact

• Theoretical Completeness: This paper closes a significant gap in the rigorous understanding of high-dimensional percolation near boundaries. It confirms that the mean-field behavior in half-spaces follows a specific "product of distances" law, analogous to Green's function estimates for random walks in domains with boundaries.
• Methodological Advancement: The successful adaptation of the "regular points" and "extended cluster" machinery to derive lower bounds (reverse Simon-Lieb type inequalities) is a major technical achievement. Previously, these tools were primarily used for upper bounds or specific exponent derivations.
• Broader Applications: The techniques and results are immediately applicable to other models in the mean-field regime, such as the weakly self-avoiding walk (as noted in the paper's references to Duminil-Copin et al.).
• Resolution of Open Problems: By solving the Hutchcroft-Michta-Slade conjecture, it provides a definitive answer to a question that motivated recent research in the field.

In summary, Panis and Schapira provide a rigorous, sharp characterization of critical percolation in half-spaces for $d>6$, utilizing a sophisticated blend of lace expansion assumptions, geometric probability (regular points), and second-moment methods to overcome the difficulties posed by boundary effects.