Higher-order Erdos--Szekeres theorems

TL;DR

This paper generalizes classical Erdos--Szekeres theorems to higher-order monotonicity, establishing bounds for the size of such subsequences in point sets and connecting to hypergraph colorings and order-type homogeneous subsets.

Contribution

It introduces a new framework for higher-order monotonicity, providing bounds and constructions that unify and extend classical geometric Ramsey results.

Findings

Established an $oldsymbol{oldsymbol{oldsymbol{ ext{Omega}}}( ext{log}^{(k-1)}N)}$ lower bound for kth-order monotone subsequences.

Constructed a geometric example with an $oldsymbol{ ext{O}( ext{log} ext{log}N)}$ upper bound for k=3, tight up to a constant.

Connected higher-order monotonicity to hypergraph colorings and order-type homogeneous subsets in higher dimensions.

Abstract

Let P=(p_1,p_2,...,p_N) be a sequence of points in the plane, where p_i=(x_i,y_i) and x_1<x_2<...<x_N. A famous 1935 Erdos--Szekeres theorem asserts that every such P contains a monotone subsequence S of $\sqrt N$ points. Another, equally famous theorem from the same paper implies that every such P contains a convex or concave subsequence of $\Omega(\log N)$ points. Monotonicity is a property determined by pairs of points, and convexity concerns triples of points. We propose a generalization making both of these theorems members of an infinite family of Ramsey-type results. First we define a (k+1)-tuple $K\subseteq P$ to be positive if it lies on the graph of a function whose kth derivative is everywhere nonnegative, and similarly for a negative (k+1)-tuple. Then we say that $S\subseteq P$ is kth-order monotone if its (k+1)-tuples are all positive or all negative. We investigate quantitative bound for the corresponding Ramsey-type result (i.e., how large kth-order monotone subsequence can be guaranteed in every N-point P). We obtain an $\Omega(\log^{(k-1)}N)$ lower bound ((k-1)-times iterated logarithm). This is based on a quantitative Ramsey-type theorem for what we call transitive colorings of the complete (k+1)-uniform hypergraph; it also provides a unified view of the two classical Erdos--Szekeres results mentioned above. For k=3, we construct a geometric example providing an $O(\log\log N)$ upper bound, tight up to a multiplicative constant. As a consequence, we obtain similar upper bounds for a Ramsey-type theorem for order-type homogeneous subsets in R^3, as well as for a Ramsey-type theorem for hyperplanes in R^4 recently used by Dujmovic and Langerman.