Canonical Van der Waerden Theorem

• The canonical van der Waerden theorem is a Ramsey-theoretic result that asserts every sufficiently large coloring of integers contains a k-term arithmetic progression that is either monochromatic or rainbow.
• It extends classical arithmetic progression results to include polynomial configurations and sparse random subsets using combinatorial, analytic, and hypergraph-container techniques.
• Quantitative studies reveal that canonical van der Waerden numbers grow super-exponentially, emphasizing the complexity of forcing canonical color patterns in arbitrary colorings.

Searching arXiv for papers on canonical van der Waerden theorem and related developments. The canonical van der Waerden theorem is the canonical Ramsey-theoretic strengthening of van der Waerden’s theorem on arithmetic progressions. For each fixed $k \ge 3$, it asserts that for all sufficiently large $n$, every coloring of $[n]=\{1,\dots,n\}$ contains a $k$-term arithmetic progression that is either monochromatic or rainbow, where “rainbow” means that all $k$ terms receive pairwise distinct colors (Alvarado et al., 3 Oct 2025, Fox et al., 2020). In contrast with the classical theorem, which assumes a bounded number of colors and guarantees monochromatic progressions, the canonical form allows arbitrary color sets and replaces a single forced outcome by a short list of unavoidable color patterns.

1. Definition and basic formulations

A $k$-term arithmetic progression ($k$-AP) is a sequence

$a, a+d, \dots, a+(k-1)d$

with $a,d \in \mathbb{Z}$ and $d \ne 0$ in the finitary canonical setting (Alvarado et al., 3 Oct 2025). Given a coloring $n$0, such a progression is monochromatic if $n$1 is constant on it, and rainbow if its $n$2 points receive $n$3 distinct colors (Alvarado et al., 3 Oct 2025).

The deterministic canonical van der Waerden property may be stated as follows: for every $n$4, there exists $n$5 such that every coloring of $n$6 with $n$7 contains a monochromatic or rainbow $n$8-AP (Alvarado et al., 3 Oct 2025). A closely related numerical invariant is the canonical van der Waerden number

$n$9

introduced in the quantitative study of the canonical problem (Fox et al., 1 Jun 2026).

A basic constraint is that a rainbow $[n]=\{1,\dots,n\}$0-AP requires at least $[n]=\{1,\dots,n\}$1 distinct colors. Consequently, if a coloring uses fewer than $[n]=\{1,\dots,n\}$2 colors, the canonical conclusion must be realized by the monochromatic alternative (Alvarado et al., 3 Oct 2025). This observation immediately yields the lower bound $[n]=\{1,\dots,n\}$3, where $[n]=\{1,\dots,n\}$4 denotes the classical $[n]=\{1,\dots,n\}$5-color van der Waerden number in the notation of the quantitative literature (Fox et al., 1 Jun 2026).

2. Historical origin and scope of the term

Van der Waerden’s original 1927 theorem established that for every fixed number of colors $[n]=\{1,\dots,n\}$6 and progression length $[n]=\{1,\dots,n\}$7, sufficiently long initial intervals contain a monochromatic $[n]=\{1,\dots,n\}$8-AP under every $[n]=\{1,\dots,n\}$9-coloring (Bergelson et al., 26 Mar 2026). The canonical strengthening is attributed to Erdős and Graham, who showed that if the number of colors is unrestricted, then sufficiently large intervals still force a $k$0-AP whose color pattern is canonical in the restricted sense of being either monochromatic or rainbow (Alvarado et al., 3 Oct 2025, Fox et al., 2020).

Within canonical Ramsey theory, this monochromatic-or-rainbow dichotomy is a particularly economical canonical classification. It does not attempt the full taxonomy of color patterns familiar from the Erdős–Rado canonical Ramsey theorem. Bergelson and Richter explicitly emphasize that their survey of van der Waerden theory does not present a “canonical van der Waerden theorem” in that full classification sense; instead, it develops structural refinements that are canonical in spirit because they force rigid, recurrent, or diagonalized outcomes in every finite coloring (Bergelson et al., 26 Mar 2026).

This distinction matters conceptually. In one usage, the canonical van der Waerden theorem refers specifically to the Erdős–Graham monochromatic-or-rainbow theorem. In another, broader usage, it refers to structural strengthenings of van der Waerden’s theorem involving syndeticity, piecewise syndeticity, IP-structure, topological recurrence, or ultrafilter equalization. The recent literature uses both perspectives, and they are complementary rather than contradictory (Bergelson et al., 26 Mar 2026).

3. Polynomial and structural extensions

A major extension replaces linear progressions by polynomial configurations. Girão proved that if $k$1 with $k$2 for all $k$3, then for any coloring $k$4, where $k$5 may be finite or infinite, there exist $k$6 such that

$k$7

is either monochromatic or rainbow (Girão, 2020). Fox, Wigderson, and Zhao gave a short finitary proof of the corresponding statement on $k$8 for sufficiently large $k$9, under the hypothesis that the polynomials are distinct and vanish at $k$0; taking $k$1 recovers the classical canonical van der Waerden theorem for arithmetic progressions (Fox et al., 2020).

The polynomial theorem clarifies what “canonical” means in this setting. The unavoidable color patterns are precisely the two extreme possibilities: complete uniformity and complete diversity. The result thus extends the classical progression case while remaining aligned with the canonical Ramsey-theoretic philosophy that arbitrary colorings must exhibit one of a short list of structurally determined behaviors (Girão, 2020, Fox et al., 2020).

The structural literature develops a different, though related, family of canonical phenomena. Bergelson and Richter survey several such refinements. One is the Brauer–Schur strengthening, which forces not only a monochromatic progression $k$2 but also the common difference $k$3 in the same color class. Another is the amplified piecewise-syndetic form: for every piecewise syndetic set $k$4, there exist a piecewise syndetic set of starters $k$5 and a fixed difference $k$6 such that

$k$7

Here the canonical rigidity lies in the fixed common difference and the large structured set of admissible starting points (Bergelson et al., 26 Mar 2026).

The same survey places these results inside a broader hierarchy. Syndetic and piecewise syndetic sets must contain progressions of every length; multiplicatively syndetic and multiplicatively piecewise syndetic sets satisfy analogous dilation-invariant statements; minimal dynamical systems satisfy topological multiple recurrence; ultrafilter formulations produce $k$8-tuple ultrafilters on the space of arithmetic progressions with equal coordinate projections; IP van der Waerden theorems force IP-simplex configurations; Hales–Jewett yields monochromatic combinatorial lines; and polynomial van der Waerden theorems force monochromatic polynomial translates anchored at $k$9 (Bergelson et al., 26 Mar 2026). These are not canonical classifications in the Erdős–Rado sense, but they are canonical structural outcomes.

4. Quantitative theory and canonical van der Waerden numbers

The recent quantitative theory centers on the growth of $k$0. Erdős and Graham proved that $k$1 is finite via Szemerédi’s theorem and asked whether

$k$2

equivalently whether $k$3 grows faster than $k$4 (Fox et al., 1 Jun 2026). This problem is now resolved affirmatively.

For sufficiently large $k$5, it is proved that

$k$6

so the three-color van der Waerden number grows faster than any exponential in $k$7 (Fox et al., 1 Jun 2026). In a many-color regime, the same work shows that for each $k$8, if $k$9 is large enough and $k$0, then

$k$1

Specializing to $k$2 and using $k$3 gives the canonical lower bound

$k$4

which settles the Erdős–Graham problem (Fox et al., 1 Jun 2026).

These bounds show that allowing a rainbow escape route does not make the finite extremal problem easy. A plausible implication is that canonicality is quantitatively expensive: very long intervals are still required before every coloring must reveal either a monochromatic or a rainbow $k$5-AP. The same paper notes an independent bound of Bae,

$k$6

which already implies $k$7, but this is weaker than the $k$8 lower bound (Fox et al., 1 Jun 2026).

By contrast, the Bergelson–Richter survey stresses that it provides no “canonical number” and no bounds specific to the structural refinements such as IP or topological variants; those results are qualitative existence theorems rather than effective estimates (Bergelson et al., 26 Mar 2026).

5. Sparse and random canonical van der Waerden theory

The canonical theorem has a sharp sparse analogue in binomial random subsets. For fixed $k$9, the random set $a, a+d, \dots, a+(k-1)d$0 is formed by retaining each element of $a, a+d, \dots, a+(k-1)d$1 independently with probability $a, a+d, \dots, a+(k-1)d$2. The threshold for $a, a+d, \dots, a+(k-1)d$3 to inherit the canonical van der Waerden property is

$a, a+d, \dots, a+(k-1)d$4

More precisely, there exist constants $a, a+d, \dots, a+(k-1)d$5 such that if $a, a+d, \dots, a+(k-1)d$6, then $a, a+d, \dots, a+(k-1)d$7 asymptotically almost surely fails the canonical property, while if $a, a+d, \dots, a+(k-1)d$8, then $a, a+d, \dots, a+(k-1)d$9 asymptotically almost surely has it (Alvarado et al., 3 Oct 2025).

The lower bound is inherited from the classical two-color threshold of Rödl–Ruciński. Below the threshold, $a,d \in \mathbb{Z}$0 asymptotically almost surely admits a 2-coloring with no monochromatic $a,d \in \mathbb{Z}$1-AP, and with only two colors available, rainbow $a,d \in \mathbb{Z}$2-APs are impossible. The upper bound is quantitative: $a,d \in \mathbb{Z}$3 for $a,d \in \mathbb{Z}$4 (Alvarado et al., 3 Oct 2025).

The proof separates two regimes of a coloring $a,d \in \mathbb{Z}$5 of $a,d \in \mathbb{Z}$6. If some color class has positive density, a sparse transference version of Szemerédi’s theorem forces a monochromatic $a,d \in \mathbb{Z}$7-AP. If all color classes are sparse, then a hypergraph-container analysis together with a rainbow supersaturation lemma forces a rainbow $a,d \in \mathbb{Z}$8-AP for every $a,d \in \mathbb{Z}$9-bounded coloring (Alvarado et al., 3 Oct 2025). This establishes that the canonical threshold has the same exponent as the ordinary $d \ne 0$0-color van der Waerden threshold, with no additional logarithmic factor.

An application concerns high-girth arithmetic-progression hypergraphs. For any $d \ne 0$1 and any target girth $d \ne 0$2, there exists $d \ne 0$3 such that the $d \ne 0$4-AP hypergraph on $d \ne 0$5 has girth at least $d \ne 0$6, yet every coloring of $d \ne 0$7 produces a monochromatic or rainbow $d \ne 0$8-AP (Alvarado et al., 3 Oct 2025). This shows that canonical Ramsey forcing can coexist with strong local sparsity in the underlying progression hypergraph.

6. Proof methods, strong abundance, and open directions

Several distinct proof paradigms now coexist. Girão’s proof of the canonical polynomial theorem is purely combinatorial and uses type colorings, fully-rainbow focused families, a weight vector on polynomial sets, and a two-layer induction that either finds a monochromatic polynomial configuration or builds enough disjoint rainbow structure to force the rainbow alternative (Girão, 2020). The short proof of Fox, Wigderson, and Zhao follows a different route: it combines the Bergelson–Leibman polynomial extension of Szemerédi’s theorem with analytic collision-counting estimates derived from Linnik’s exponential-sum bound, showing that if monochromatic configurations are absent, then non-rainbow configurations are too sparse to exhaust all polynomial patterns (Fox et al., 2020).

The structural literature provides stronger abundance statements than mere existence. Di Nasso proves that in every finite coloring $d \ne 0$9, some color class $n$00 is piecewise syndetic, and for every $n$01, the set

$n$02

is itself piecewise syndetic (Nasso, 2018). Thus the canonical content is not only that one color class contains arbitrarily long monochromatic progressions, but that the starting points of such progressions form a large structured set. The proof uses translation-invariant filters maximal with respect to inclusion, constructed recursively on countable algebras of sets without invoking the axiom of choice or Zorn’s lemma (Nasso, 2018).

Bergelson and Richter organize a complementary arsenal: intersectivity lemmas for piecewise syndetic and multiplicative largeness, Brown’s lemma, minimal systems and topological correspondence, a topological van der Corput difference lemma, Stone–Čech compactification, minimal left ideals, idempotents via Ellis’s lemma, IP systems, and a semigroup lifting theorem that equalizes coordinate projections on minimal ultrafilters (Bergelson et al., 26 Mar 2026). The quantitative paper on $n$03 adds a rather different toolkit: sparse hitting sets with robust intersection properties, Lovász-local-lemma set-colorings far from monochromaticity, random shifted product constructions, and iterative Chinese-remainder-theoretic amplification (Fox et al., 1 Jun 2026). The sparse random theorem adds hypergraph containers and rainbow supersaturation (Alvarado et al., 3 Oct 2025).

Several open problems remain explicit in the literature. Bergelson and Richter highlight an amenable-group IP van der Waerden conjecture and a Density Polynomial Hales–Jewett conjecture (Bergelson et al., 26 Mar 2026). The quantitative lower-bound paper isolates the two-color problem—whether $n$04 has super-exponential growth—as a major challenge, and also asks whether the three-color lower bound can be strengthened beyond the current $n$05 factor in the exponent (Fox et al., 1 Jun 2026). In the sparse setting, the threshold location is known, but sharpening the constants or proving a sharp threshold in the strongest sense remains open (Alvarado et al., 3 Oct 2025).

Taken together, these developments show that the canonical van der Waerden phenomenon has at least three distinct but interacting meanings: the Erdős–Graham monochromatic-or-rainbow theorem, its polynomial and sparse generalizations, and a broader family of structural refinements in which large sets, recurrent systems, or ultrafilters force rigid arithmetic patterns. The modern theory treats all three as facets of the same Ramsey-theoretic principle: arbitrary colorings cannot avoid highly organized arithmetic behavior.