Skip to content
All apps

No percolation at criticality (3D)

About this app

Site percolation in a cube up to 64³: the largest cluster glows, p moves around p_c ≈ 0.3116. At p_c it often reaches wall to wall, yet its share shrinks towards 0 – there is no infinite cluster. openai/math family 213 (formalised in Lean).

What it is about

In percolation, each site of the cubic lattice ℤ³ is open independently with probability p, and neighbouring open sites form clusters. Above a threshold p_c, about 0.3116 for sites, there is an infinite cluster; below it there is none. What happens exactly at p_c was open in three dimensions: in the plane it follows from Harris and Kesten, in very high dimensions from the lace expansion, but for dimensions 3 to 10 no method was available. An AI-generated preprint in github.com/openai/math (family 213) claims that at p_c there is almost surely no infinite cluster, so the transition is continuous. The catalogue lists the main theorems as formalised in Lean; the preprint is not peer-reviewed.

What the app shows

A cube of up to 64³ sites shows the largest cluster glowing, with the others faint or hidden. The random numbers stay fixed while you move p, so you can watch clusters grow. A panel counts open sites, the largest and second-largest cluster and checks whether the largest spans from wall to wall. A measured curve shows the share of the largest cluster for edge lengths from 8 to 128. A slice panel shows single layers of the cube.

What to look for

Set p exactly to p_c: the largest cluster often spans from wall to wall, yet its share of the cube shrinks as the edge length grows. Above p_c a fixed share remains. The curve fits the theorem but does not prove it: a share that crept very slowly towards a small positive value would look almost the same at these sizes. In the slice panel, sites often look separate even though they connect through other layers.

Frequently asked questions

What is percolation?
Percolation is a model of random connections: each site or edge of a lattice is open independently with probability p. The question is whether open paths reach over large distances, much like water seeping through porous rock. Above a critical probability p_c an infinite cluster appears.
What is the percolation threshold of the cubic lattice?
For site percolation on the cubic lattice ℤ³, p_c is numerically about 0.3116, and for bond percolation about 0.2488. These values come from simulations; no closed formula is known. In the plane, by contrast, the bond threshold of the square lattice is exactly ½.
Is there an infinite cluster at critical percolation in 3D?
It was expected that there is no infinite cluster at p_c, but in three dimensions this had not been proved. An AI-generated preprint in the catalogue github.com/openai/math (family 213) claims a proof for bond and site percolation on ℤ³. The catalogue lists the main theorems as formalised in Lean; peer review is still pending.
What can I try out in this app?
You move p below, at and above p_c, choose the cube size and see the largest cluster in 3D. A counter checks whether it spans from wall to wall. The measured curve shows how its share behaves as the edge length grows, and a slice panel shows individual layers.

Subject: Open problems, solved by AI

More from Open problems, solved by AI

A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps