Paper project · Combinatorics and probability

When do the Kahn–Kalai bounds provide nontrivial information?

StatusPublished journal article
Boundpc(F) ≤ K q(F) log ℓ(F)
QuestionWhen is the right side actually below 1?
GeometryMinimal elements must become dispersed

Overview

A powerful bound can still be vacuous.

The Park–Pham theorem bounds the critical probability pc(F) of a nontrivial upper set using its expectation threshold q(F) and the size ℓ(F) of its largest minimal element. Because pc(F)<1 is always known, the theorem is informative only when Kq(F)logℓ(F)<1.

The paper studies when that happens. In the regime ℓ(Fn)→∞, it proves a strong necessary dispersal condition: after deleting any fixed number of minimal elements, the remaining collection can have a common intersection only finitely often. It also identifies sufficient conditions under which the bound becomes asymptotically exact.

01

Test nontriviality

Compare with the universal bound pc<1

The question is not whether the theorem is true, but whether its numerical upper bound contributes any new information.

02

Inspect minimal elements

Shared cores obstruct useful bounds

If almost all minimal elements keep a persistent common intersection, the Park–Pham estimate cannot become nontrivial in the growing regime.

03

Look for a widening wedge

Growth must be accompanied by spread

The minimal elements must proliferate and disperse through the Boolean lattice rather than climbing inside a narrow principal region.

Boolean-lattice geometry

Useful bounds require more than height.

The paper’s geometric intuition is that a growing upper set must spread down and across the lattice as it climbs; a persistent shared core makes the bound vacuous.

Citation

Use or reference this work.

The article is published open access; the arXiv version is also linked above.

BibTeX citation
@article{christopherson2025kahnkalai,
  title   = {When do the Kahn--Kalai bounds provide nontrivial information?},
  author  = {Christopherson, Bryce Alan and Baretz, Jack},
  journal = {Journal of Inequalities and Applications},
  volume  = {2025},
  number  = {1},
  eid     = {23},
  pages   = {1--7},
  year    = {2025},
  doi     = {10.1186/s13660-025-03272-z}
}