PMI Sequence — An Erdős-Szekeres Type Result
What is a PMI Sequence?
A Pairwise Monotone Increasing (PMI) sequence in a set X ⊆ ℤd is a sequence of points x(1), x(2), …, x(k) such that every coordinate of x(t+1) is strictly greater than the corresponding coordinate of x(t) — i.e., they form a chain in the product order on ℤd.
The classical Erdős-Szekeres theorem (1935) guarantees that any sequence of more than (r-1)(s-1) distinct integers contains a monotone increasing subsequence of length r or a monotone decreasing subsequence of length s. The PMI problem is a multi-dimensional analogue.
Main Result
Theorem (Lower Bound, d Dimensions). For every integer d ≥ 1 and every set X ⊆ ℤ
≥0d with |X| = n, the longest PMI sequence in X has length at least
⌈(d! · n)1/d⌉ − Cd
where C
d is a constant depending only on d. Together with a matching upper bound, this shows the maximum PMI length over all n-point sets in ℤ
≥0d is (d! · n)
1/d up to lower-order terms.
Abstract
We study the maximum length of a Pairwise Monotone Increasing (PMI) sequence in a set of n points in the d-dimensional non-negative integer grid ℤ≥0d. This is an Erdős-Szekeres type question in the multi-dimensional product order.
We prove tight asymptotic bounds: the answer is Θ((d! · n)1/d), resolving a question motivated by applications to network controllability (strong structural controllability via zero-forcing sets) and combinatorial geometry.
The proof introduces a layered induction on "levels" of the grid, decomposing the point set along the minimum coordinate, and applies the inductive hypothesis on the resulting lower-dimensional faces with carefully tracked intersection multiplicities.
Keywords: PMI sequence · Erdős-Szekeres · Product order · Combinatorial bounds · Network controllability
Motivation from Network Theory
PMI sequences arise naturally in the study of strong structural controllability of networks. A key quantity — the distance-based bound on zero-forcing number — can be characterized using the length of the longest PMI sequence in a set derived from pairwise distances in the network graph. This connection, established in earlier work (Shabbir, Abbas, Yazıcıoğlu, Koutsoukos, IEEE TAC 2023), motivates tight combinatorial bounds on PMI sequence length.
See Also
The companion repository pmialgo contains the algorithmic construction and distance-plot analysis.