Table of Contents
Fetching ...

Dependence of functions on their variables

Ilijas Farah

TL;DR

The paper analyzes when a function $f$ on a combinatorial cube $f: \prod_{i<d} X_i \to Y$ essentially depends on a single variable. It establishes a dichotomy: either $f$ is determined, up to partitions and coordinate projections, by at most one coordinate, or there exist infinite sequences $(x_m)$ and $(y_m)$ with $d=u\sqcup v$ that exhibit a nontrivial cross-coordinate dependence, as derived via Ramsey theory and ultrafilter/compactness arguments. The proof proceeds in the equal-sets case, uses a Ramsey construction on triples to produce homogeneous sets, and relies on the Čech–Stone compactification $\beta X$ and AC to carry the inductive argument from $d=2$ to higher arities. The work also discusses the necessity of Choice, model-theoretic aspects in ZF, and connections to Čech–Stone remainder phenomena and dimension-type results in topological combinatorics.

Abstract

In this note I present a readable version of the proof of my 2001 result, giving a sufficient and necessary condition for a function on a combinatorial cube to essentially (locally) depend on at most one variable (see the end of the paper for the motivation), as well as some limiting results.

Dependence of functions on their variables

TL;DR

The paper analyzes when a function on a combinatorial cube essentially depends on a single variable. It establishes a dichotomy: either is determined, up to partitions and coordinate projections, by at most one coordinate, or there exist infinite sequences and with that exhibit a nontrivial cross-coordinate dependence, as derived via Ramsey theory and ultrafilter/compactness arguments. The proof proceeds in the equal-sets case, uses a Ramsey construction on triples to produce homogeneous sets, and relies on the Čech–Stone compactification and AC to carry the inductive argument from to higher arities. The work also discusses the necessity of Choice, model-theoretic aspects in ZF, and connections to Čech–Stone remainder phenomena and dimension-type results in topological combinatorics.

Abstract

In this note I present a readable version of the proof of my 2001 result, giving a sufficient and necessary condition for a function on a combinatorial cube to essentially (locally) depend on at most one variable (see the end of the paper for the motivation), as well as some limiting results.

Paper Structure

This paper contains 2 sections, 7 theorems, 2 equations.

Key Result

Theorem 1

For all $d\geq 1$, sets $X_i$, for $i<d$ and $Y$, every $f\colon \prod_{i<d} X_i\to Y$ satisfies exactly one of the following.

Theorems & Definitions (7)

  • Theorem 1
  • Lemma 1.1
  • Lemma 1.2
  • Lemma 1.3
  • Proposition 2.1
  • Proposition 2.2
  • Proposition 2.3: ZF