---
title: Cochromatic Bases in Graph Colouring & Combinatorics
url: https://www.emergentmind.com/topics/cochromatic-bases
type: topic
---

# Cochromatic Bases in Graph Colouring & Combinatorics

Searching arXiv for papers related to “cochromatic bases”, generalized chromatic functions, complete multipartite bases, and cochromatic number.
“Cochromatic bases” is not introduced as a single formal term across the cited literature, but the phrase usefully names a family of closely related constructions at the interface of graph colouring, cochromatic structure, and symmetric or quasisymmetric function theory. In graph theory, the relevant primitive is the cochromatic number $\zeta(G)$, defined as the minimum number of colours in a vertex colouring whose colour classes are each either independent sets or cliques [2408.02400; 2408.13839]. In algebraic combinatorics, the analogous perspective appears when classical bases of $\mathrm{Sym}(x)$, $\mathrm{QSym}(x)$, and their noncommutative variants are realized as generalized chromatic functions of specially designed edge-coloured digraphs, or when bases are built from graphs complementary to disjoint unions of cliques, such as the complete multipartite basis $\{r_\lambda\}$ [2208.08458; 2009.14141]. Taken together, these developments present a “cochromatic” viewpoint in which partitions into independent sets and cliques, complement symmetry, and graph-based generating functions provide a common language for both extremal graph colouring problems and basis constructions in combinatorial Hopf algebras.

## 1. Graph-theoretic origin: cochromatic colourings and $\zeta(G)$

The cochromatic number of a graph $G$, denoted $\zeta(G)$, is the minimum number of colours needed for a vertex colouring in which every colour class is either an independent set or a clique [2408.02400; 2408.13839; 2409.17614]. Equivalently, $\zeta(G)$ is the minimum number of parts in a partition of $V(G)$ into homogeneous sets, where a homogeneous set is either an independent set or a clique [2408.02400]. This extends ordinary proper colouring, since the chromatic number $\chi(G)$ permits only independent sets as colour classes; consequently,
\[
\zeta(G)\le \chi(G).
\]
This inequality is explicit in the graph-theoretic papers and is the basic reason cochromatic partitions are more flexible than proper colourings [2408.02400; 2408.13839; 2409.17614].

A central structural feature is complement symmetry. The cochromatic number is invariant under complements,
\[
\zeta(G)=\zeta(\overline G),
\]
and one also has
\[
\zeta(G)\le \chi(\overline G),
\]
because a proper colouring of $\overline G$ gives a partition of $G$ into cliques [2408.13839; 2408.02400]. This complement behaviour is one of the main reasons the adjective “cochromatic” acquires a basis-like meaning in later algebraic settings: it emphasizes decomposition into clique-type and stable-set-type pieces, rather than only stable sets.

The distinction from ordinary chromatic structure can be extreme. Complete graphs satisfy $\zeta(K_n)=1$ but $\chi(K_n)=n$ [2408.02400]. A plausible implication is that cochromatic invariants are best viewed not as minor perturbations of chromatic invariants, but as hybrid objects interpolating between chromatic number and clique cover phenomena. This hybrid character is explicit in the random-graph and bounded-clique-number results discussed below.

## 2. Extremal separation between chromatic and cochromatic structure

The graph-theoretic literature around Erdős and Gimbel studies how large the difference $\chi(G)-\zeta(G)$ can be, especially under bounded clique number. For each integer $n>2$, Erdős, Gimbel, and Straight defined $f(n)$ as the smallest integer such that every graph $G$ with $\omega(G)<n$, except for $K_{n-1}$, satisfies
\[
\chi(G)\le \zeta(G)+f(n).
\]
They proved that $f(n)$ is well-defined, that $f(n)$ grows exponentially with $n$, and that $f(3)=0$ and $f(4)=1$ [2408.02400]. The 2024 note shows that for fixed $n\ge 5$, determining the exact value of $f(n)$ reduces to a finite computation: for $n\ge 5$ and $f\ge n-2$, there is an explicit constant $N=N(n,f)$ such that
\[
f(n)\le f \quad\Longleftrightarrow\quad \chi(G)\le \zeta(G)+f
\]
for all graphs $G$ with $\omega(G)<n$ and $|V(G)|\le N$ [2408.02400]. The reduction uses Ramsey theory, together with an observation bounding $\chi(G)$ by repeatedly removing independent sets of size $n$.

A more decisive result concerns the conjecture that every graph $G$ with $\omega(G)<5$ and $\zeta(G)>3$ satisfies
\[
\chi(G)\le \zeta(G)+2.
\]
This conjecture is disproved by infinitely many explicit counterexamples [2408.02400]. The paper proves that there are infinitely many graphs $G$ such that
\[
\omega(G)<5,\qquad \zeta(G)=4,\qquad \chi(G)=7,
\]
so in particular
\[
\chi(G)-\zeta(G)=3
\]
for infinitely many $K_5$-free graphs [2408.02400]. The construction is organized around an 11-vertex auxiliary graph $H$ with $\omega(H)<5$, whose vertex set can be partitioned into 3 cliques, and such that every proper 6-colouring of $H$ exposes a subset $X$ with $\omega(H[X])<4$ on which all six colours appear [2408.02400]. Enlarging $H$ by adding independent sets $V_X$ adjacent exactly to such subsets forces $\chi(G)\ge 7$, while a cochromatic partition of $V(H)$ into 3 cliques together with the new vertices as an independent set yields $\zeta(G)\le 4$; the paper then shows $\zeta(G)\neq 3$, hence $\zeta(G)=4$ [2408.02400].

This use of a small “trigger” gadget gives a concrete model of cochromatic structure. The cited summary states that the paper “does not use the phrase ‘cochromatic bases’ as a formal term,” but that the gadget $H$ forces any proper colouring to expose a “basis-like” subset $X$ on which all colours appear while clique number remains small [2408.02400]. This suggests an interpretation of cochromatic bases, in the graph-theoretic sense, as structured families of homogeneous pieces or witness subsets through which low cochromatic number fails to control full chromatic complexity.

## 3. Random graphs and the asymptotic gap $\chi(G)-\zeta(G)$

For $G\sim G_{n,1/2}$, Erdős and Gimbel asked whether
\[
\chi(G)-\zeta(G)\to\infty \quad \text{whp}.
\]
This question appears in multiple 2024 papers and is treated through progressively stronger results [2408.02400; 2408.13839; 2409.17614].

One note gives positive evidence by proving that for every $\varepsilon>0$, there exists an absolute constant $c>0$ such that for infinitely many $n$,
\[
\mathbb{P}\big(\chi(G_n)-\zeta(G_n)\ge n^{1/2-\varepsilon}\big)\ge c,
\]
and therefore
\[
\mathbb{E}\big(\chi(G_n)-\zeta(G_n)\big)\ge \Omega(n^{1/2-\varepsilon})
\]
[2408.02400]. The argument combines anti-concentration for $\chi(G_{n,p})$ with the Harris–FKG inequality and the bound $\zeta(G)\le \chi(\overline G)$ [2408.02400].

A companion note establishes a non-concentration obstruction. Its main theorem states that if $G\sim G_{n,1/2}$ and $g(n)$ satisfies
\[
\mathbb{P}\big(\chi(G)-\zeta(G)\le g(n)\big)>0.999,
\]
then there is a sequence of integers $n^*$ such that
\[
g(n^*) > c\,\frac{\log^3 n^*}{\sqrt{n^*}\,\log\log n^*}
\]
for some constant $c>0$ [2408.13839]. The mechanism is to show that a high-probability bound on $\chi(G)-\zeta(G)$ would force concentration of $\chi(G_{n,1/2})$ into intervals of length $g(n)$, contradicting a prior lower bound on the concentration width of the chromatic number [2408.13839]. The proof relies on complement invariance of $\zeta(G)$, the fact that $\chi(G)$ is decreasing in the edge set, the fact that $\chi(\overline G)$ is increasing in the edge set of $G$, and Harris’s Lemma [2408.13839].

A later paper proves a much stronger statement for most $n$. With
\[
a_0(n)=2\log_2 n-2\log_2\log_2 n+2\log_2(e/2)+1,\qquad a=\lfloor a_0\rfloor,
\]
and
\[
H_a=\binom{n}{a}2^{-\binom{a}{2}},
\]
the theorem states: if $\varepsilon>0$ is fixed and
\[
n^{0.05}<H_a<n^{1-\varepsilon},
\]
then for $G\sim G_{n,1/2}$,
\[
\chi(G)-\zeta(G)\ge n^{1-\varepsilon}\qquad \text{whp}
\]
[2409.17614]. The paper summarizes this as a positive answer for roughly $95\%$ of all values $n$ [2409.17614]. The proof combines a lower bound on $\chi(G)$ derived from a result of Heckel–Panagiotou with an upper bound on $\zeta(G)$ obtained by a second moment argument for cochromatic colourings, followed by Azuma–Hoeffding concentration and a “shift by a little” trick from Frieze [2409.17614].

The structural message of these random-graph results is that cochromatic flexibility is asymptotically significant even when $\chi(G)$ and $\zeta(G)$ have the same first-order scale. Indeed, the same paper records the heuristic that cochromatic colourings at a given profile are more numerous than ordinary colourings by a factor of $2^k$, because each colour class may be either a clique or an independent set [2409.17614]. This is formalized in the identity
\[
\mathbb{E}_{1/2}[X^{\mathrm{co}}_k] = 2^k\,\mathbb{E}_{1/2}[X_k]
\]
for appropriate profiles with $k_1=0$ [2409.17614]. A plausible implication is that “cochromatic basis” phenomena in random graphs are driven by the multiplicative choice of clique-type versus stable-set-type classes.

## 4. Generalized chromatic functions as a unifying algebraic framework

The most explicit basis-theoretic realization of a cochromatic viewpoint appears in the theory of generalized chromatic functions for edge-partitioned digraphs [2208.08458]. Here edges come in three types:
- dashed: $a \dashrightarrow b$,
- solid: $a \to b$,
- double: $a \Rightarrow b$.

A proper vertex-colouring of such a digraph $G$ is a map $K:V(G)\to \mathbb{P}$ satisfying
\[
\text{if } a \dashrightarrow b,\quad K(a)\neq K(b),
\]
\[
\text{if } a \to b,\quad K(a)<K(b),
\]
\[
\text{if } a \Rightarrow b,\quad K(a)\le K(b).
\]
With
\[
x^K=\prod_{a\in V(G)} x_{K(a)},
\qquad
\operatorname{asc}(K)=\bigl|\{(a,b)\in E(G): K(a)<K(b)\}\bigr|,
\]
the generalized chromatic function is
\[
X_G(x,t)=\sum_{K\in C(G)} t^{\operatorname{asc}(K)} x^K,
\qquad
X_G(x)=X_G(x,1).
\]
This single definition simultaneously recovers proper graph colourings, $P$-partitions, chromatic symmetric functions, chromatic quasisymmetric functions, and related objects [2208.08458].

The graph-colouring case arises when all edges are dashed, so $X_G(x)$ becomes Stanley’s chromatic symmetric function [2208.08458]. The $P$-partition case arises from the Hasse diagram of a poset with edge types determined by the poset-labeling; Proposition 3.3 gives a bijection between $P$-partitions and proper colourings of the corresponding edge-coloured digraph [2208.08458]. The framework also recovers the Shareshian–Wachs chromatic quasisymmetric function and Ellzey’s chromatic quasisymmetric function by using only dashed edges with suitable orientations [2208.08458].

Within this formalism, many classical bases of $\mathrm{Sym}(x)$ and $\mathrm{QSym}(x)$ are realized as generalized chromatic functions of specially chosen digraphs [2208.08458]. The cited summary explicitly lists the monomial symmetric basis $m_\lambda$, augmented monomial basis, elementary basis $e_\lambda$, complete homogeneous basis $h_\lambda$, power sum basis $p_\lambda$, and Schur basis $s_\lambda$ as realizable special cases, together with the monomial quasisymmetric basis $M_\alpha$, the fundamental basis $F_\alpha$, and the upper-fundamental basis [2208.08458]. It also states that if one generalizes the graph for a partition $\lambda$ to a composition $\alpha$, one recovers the dual immaculate and row-strict dual immaculate functions as well [2208.08458].

The basis-theoretic significance is not merely representational. The functions satisfy a product formula
\[
X_{G_1}(x,t)\,X_{G_2}(x,t)=X_{G_1\sqcup G_2}(x,t),
\]
and a coproduct formula
\[
\Delta(X_G(x))=\sum_F X_{G|_{V(G)\setminus V(F)}}(x)\otimes X_F(x),
\]
where $F$ ranges over the $\{\to,\Rightarrow\}$-induced subdigraphs of $G$ [2208.08458]. This is the algebraic mechanism through which the graph-theoretic data of edge-coloured digraphs becomes a source of bases and Hopf structures. The summary explicitly describes this as a conceptual contribution: a single chromatic-generating-function formalism encodes graph colourings, poset $P$-partitions, symmetric functions, quasisymmetric functions, and their noncommutative analogues [2208.08458].

In this sense, “cochromatic bases” can be understood as basis families whose defining data consist of graph- or digraph-based colouring rules broad enough to interpolate between independent-set and clique-type behaviour. This interpretation is consistent with the summary’s statement that generalized chromatic functions provide a unifying “cochromatic” language in which many classical bases are “literally the same objects arising from different edge-coloured digraphs” [2208.08458].

## 5. Complete multipartite bases and the complement-of-cliques viewpoint

A more specific basis construction is the complete multipartite basis $\{r_\lambda\}$ of symmetric functions [2009.14141]. For an integer partition $\lambda=(\lambda_1,\dots,\lambda_k)$ of $n$, define the complete multipartite graph $G_\lambda$ with vertex set
\[
V(G_{\lambda}) = \{v_{11},\dots,v_{1 \lambda_1},v_{21},\dots,v_{2\lambda_2},v_{31},\dots,v_{k\lambda_k}\},
\]
and edge set
\[
E(G_{\lambda}) = \{v_{ij}v_{ab} \mid i \neq a\}.
\]
Then
\[
r_{\lambda} = X_{G_{\lambda}}
\]
[2009.14141]. Since $G_\lambda$ consists of disjoint stable sets of sizes $\lambda_1,\dots,\lambda_k$ with all possible edges between distinct parts, it is the complement of a disjoint union of cliques. The paper notes that
\[
X_{\overline{G_{\lambda}}} = \left(\prod_{i=1}^{k} \lambda_i!\right)e_{\lambda},
\]
so the $r$-basis is the graph-complement analogue of the elementary basis [2009.14141].

This complement relation is why the basis can naturally be described as “complete multipartite” and, in the paper’s own summarized wording, “cochromatic” [2009.14141]. The terminology reflects two facts: each $r_\lambda$ is the chromatic symmetric function of a complete multipartite graph, and the complement of that graph is a disjoint union of cliques, the structure governing the elementary basis [2009.14141]. The paper also states that Penaguiao had already shown that $\{r_\lambda:\lambda\vdash d\}$ is a basis for $\Lambda^d$ [2009.14141].

The basis admits explicit change-of-basis formulas with the monomial basis. For partitions $\lambda$ and $\mu$, the coefficient $[m_\mu]r_\lambda$ is interpreted by “puzzles of $\mu$ into $\lambda$,” and the transition matrix from $m$ to $r$ is upper triangular with $1$’s on the diagonal [2009.14141]. The inverse coefficients $[r_\mu]m_\lambda$ admit a necklace interpretation; in particular,
\[
m_n = \sum_{\mu \vdash n} (-1)^{l(\mu)-1} c_{\mu}r_{\mu},
\qquad
c_{\mu} = \frac{n!(l(\mu)-1)!}{\prod_i \mu_i!\prod_i n_i(\mu)!},
\]
where $c_\mu$ counts cyclically ordered set partitions of $[n]$ of type $\mu$ [2009.14141].

The central combinatorial application is the $r$-basis expansion of chromatic and Tutte symmetric functions. If $M_1,\dots,M_k$ are the maximal stable partitions of a graph $G$, then
\[
X_G = \sum_{\textnormal{nonempty}\, S \subseteq [k]} (-1)^{|S|-1}r_{(\wedge_{i \in S} M_i)},
\]
where $\wedge$ is the meet of partitions [2009.14141]. Thus the coefficient of $r_\mu$ is controlled by inclusion–exclusion over intersections of maximal stable partitions [2009.14141]. This is precisely a basis description in terms of overlap patterns of stable-set decompositions, which is why the $r$-basis is especially apt for a cochromatic interpretation.

## 6. Noncommutative and $\Gamma$-analogues of chromatic-basis constructions

The generalized chromatic framework extends to noncommuting variables. For a labelled edge-coloured digraph $G$ with vertex set $[n]$, and a colouring $K$, define
\[
X_K = x_{K(1)}x_{K(2)}\cdots x_{K(n)}.
\]
Then
\[
\mathbf{X}_G(x,t)=\sum_{K\in C(G)} t^{\operatorname{asc}(K)} X_K,
\qquad
\mathbf{X}_G(x)=\mathbf{X}_G(x,1),
\]
and there is a commutation map
\[
p:\mathbb{Q}\langle x_1,x_2,\dots\rangle\to \mathbb{Q}[x_1,x_2,\dots]
\]
such that
\[
p(\mathbf{X}_G(x,t))=X_G(x,t)
\]
[2208.08458]. The same paper introduces the Hopf algebra
\[
\mathrm{NCQSym}^r(x)=\mathrm{NCQSym}_1(x)\supset \mathrm{NCQSym}_2(x)\supset \cdots \supset \mathrm{NCQSym}_r(x)=\mathrm{NCSym}(x),
\]
and proves in Theorem 13.1 that $\mathrm{NCQSym}^r(x)$ is a Hopf algebra [2208.08458]. The natural bases include the $r$-dominant monomial basis $\{M(\beta,\mu)\}$ and the $r$-fundamental basis $\{F(\beta,\mu)\}$ [2208.08458]. These are again obtained directly from generalized chromatic functions of labelled edge-coloured digraphs.

A different analogue arises in the algebra
\[
\Gamma=\mathbb{Q}[p_1,p_3,p_5,\dots]
\]
of symmetric functions generated by the odd power sums [1907.09722]. The obstacle is explicit: ordinary chromatic symmetric functions almost never lie in $\Gamma$. In fact, the paper proves that a finite simple graph $G$ has no edges if and only if
\[
X_G\in\Gamma
\]
[1907.09722]. To obtain an analogue of chromatic bases inside $\Gamma$, the authors define the near chromatic symmetric function
\[
Y_G=\frac{X_G+\omega(X_G)}{2},
\]
where $\omega$ is the involution on $\Lambda$ [1907.09722].

The classification is quite restrictive. If $Y_G\in\Gamma$, then $G$ has no pair of disjoint edges; if $G$ is connected, then $G$ is either $C_3$ or a star $S_n$ [1907.09722]. More precisely,
\[
Y_G\in\Gamma \quad\Longleftrightarrow\quad G\text{ is a disjoint union of a null graph and one connected component }H,
\]
where $H$ is either $C_3$ or $S_n$ for some $n\ge1$ [1907.09722]. The paper then identifies two algebraically independent generating sets of near chromatic symmetric functions,
\[
\{Y_{S_1}, Y_{C_3}, Y_{S_5}, Y_{S_7}, \dots\},
\qquad
\{Y_{S_1}, Y_{S_3}, Y_{S_5}, Y_{S_7}, \dots\},
\]
and states that these are the only algebraically independent generator sets of $\Gamma$ consisting of near chromatic symmetric functions [1907.09722]. For the corresponding graph sets $\mathbf B_1$ and $\mathbf B_2$, the families $\mathcal Y(\mathbf B_1)$ and $\mathcal Y(\mathbf B_2)$ form bases of $\Gamma^n$ [1907.09722].

These constructions are not phrased as cochromatic bases in the graph-theoretic sense, but they belong to the same chromatic-basis lineage. A plausible implication is that the “cochromatic” theme survives beyond ordinary chromatic symmetric functions only after suitable symmetrization or graph-class restriction.

## 7. Terminological scope, unifying themes, and open directions

Across the cited papers, “cochromatic” has two distinct but connected meanings. In graph theory, it refers directly to partitions into independent sets and cliques and to the invariant $\zeta(G)$ [2408.02400; 2408.13839; 2409.17614]. In algebraic combinatorics, it refers more broadly to basis constructions and generating functions built from graph-complement structure, multipartite graphs, or generalized colouring rules that unify clique-like and stable-set-like data [2208.08458; 2009.14141]. The literature does not standardize “cochromatic bases” as a formal umbrella term, but the summaries explicitly describe the generalized chromatic-function viewpoint as a unifying “cochromatic” language and the complete multipartite basis as both “complete multipartite” and “cochromatic” [2208.08458; 2009.14141].

Several recurrent themes organize the subject.

| Theme | Graph-theoretic manifestation | Algebraic manifestation |
|---|---|---|
| Complement symmetry | $\zeta(G)=\zeta(\overline G)$ and $\zeta(G)\le \chi(\overline G)$ | $r_\lambda$ is complementary to $e_\lambda$ via $G_\lambda$ and $\overline{G_\lambda}$ |
| Mixed homogeneous classes | Colour classes may be independent sets or cliques | Edge-coloured digraph rules interpolate between graph colourings and $P$-partitions |
| Partition structure | Cochromatic partitions and witness subsets $X$ control $\chi-\zeta$ gaps | Basis coefficients encode stable partitions and their intersections |

One open direction is explicit in the random-graph literature. The question whether
\[
\chi(G)-\zeta(G)\to\infty \quad\text{whp}
\]
for $G\sim G_{n,1/2}$ is now answered positively for roughly $95\%$ of all $n$ [2409.17614], but not yet for all $n$. The same paper conjectures that whp
\[
\chi(G)-\zeta(G)=O\!\left(\frac{n}{\log^3 n}\right)
\]
[2409.17614], while the earlier note gives the heuristic that the first-moment threshold for cochromatic number should be smaller than that for chromatic number by about
\[
\Theta\!\left(\frac{n}{\log^3 n}\right)
\]
[2408.13839]. This places the proven lower bounds and conjectured upper bounds on widely separated scales.

Another open direction concerns the algebraic side. The generalized chromatic-function formalism already realizes many classical bases of $\mathrm{Sym}(x)$, $\mathrm{QSym}(x)$, and their noncommutative analogues as graph-based objects [2208.08458]. This suggests that further “cochromatic basis” phenomena may emerge whenever one can encode basis-defining inequalities or incompatibilities through edge-coloured digraphs. That implication is interpretive rather than explicit, but it is closely aligned with the cited paper’s claim that many classical bases are realized as specialized generalized chromatic functions [2208.08458].

In summary, cochromatic bases are best understood as basis constructions and structural decompositions organized by the same principle that defines cochromatic number: graph partitions are allowed to mix stable-set and clique behaviour. In graph theory this principle governs the invariant $\zeta(G)$ and its separation from $\chi(G)$; in algebraic combinatorics it governs graph-complement bases such as $\{r_\lambda\}$ and unifying chromatic formalisms in which numerous classical bases appear as special cases [2408.02400; 2009.14141; 2208.08458].

Source: https://www.emergentmind.com/topics/cochromatic-bases