---
title: Colored Linear Orderings
url: https://www.emergentmind.com/topics/colored-linear-orderings
type: topic
---

# Colored Linear Orderings

Colored linear orderings denote several mathematically distinct structures sharing an ordered substrate and a color apparatus. In one usage, a linear order is expanded by unary predicates, yielding a colored order in the model-theoretic sense. In another, a chain is colored by weakly or strictly order-preserving maps subject to a threshold on designated positions. A third usage orders the colors themselves: an LO \(k\)-coloring of an \(r\)-uniform hypergraph assigns values in \(\{1,\dots,k\}\) so that each edge has a unique maximum. Related ordered-color frameworks also appear in graph sparsity theory, forbidden ordered-pattern characterizations of colorability, shuffle constructions over \((\mathbb Q,<)\), and the classification of homogeneous colored orders [1901.06720] [2204.05628] [1810.02324] [2604.14255].

## 1. Core meanings and formal settings

The common core is the interaction between a total order and a coloring mechanism, but the direction of interaction varies by field. Sometimes the order is primary and colors are unary predicates on points; sometimes the colors themselves are linearly ordered labels; sometimes an ordering of vertices is used to control graph-theoretic reachability or to forbid particular ordered patterns. This suggests that “colored linear ordering” is not a single standard object but a family of order-sensitive color formalisms.

| Framework | Underlying object | Order/color condition |
|---|---|---|
| Model-theoretic colored order | \((M,<,\{P_i\}_{i\in I})\) | unary predicates expand a linear order |
| Chain order polynomial | chain \(C_n\) | weakly or strictly increasing maps with a celeste threshold |
| LO hypergraph coloring | \(r\)-uniform hypergraph | every edge has a unique maximum color |
| Homogeneous colored order | countable colored order | finite partial isomorphisms extend to automorphisms |

In the model-theoretic formulation, a colored order is a structure \(\mathcal M=(M,<,\{P_i\}_{i\in I})\), where each \(P_i\) is a unary subset of \(M\). In the chain-enumerative formulation, a bicolored poset \((P,\preceq)\) has a partition \(P=C\sqcup S\) into celeste and silver elements, and one counts order-preserving maps \(\varphi:P\to[x]\) satisfying a threshold on celeste elements. In the hypergraph formulation, an LO \(k\)-coloring is a function \(c:V\to\{1,\dots,k\}\) such that the multiset of colors on each edge has a unique maximum. For \(r=3\), this is equivalently the rule that if two vertices in an edge share a color, then the third must receive a strictly larger color [1810.02324] [1901.06720] [2204.05628] [2604.14255].

## 2. Chains, thresholds, and bivariate order polynomials

For a finite poset \(P\), classical order polynomials count order-preserving or strictly order-preserving maps into a chain \([n]=\{1,2,\dots,n\}\). The bivariate refinement introduces a partition \(P=C\sqcup S\), with \(C\) the celeste elements, and counts maps satisfying both monotonicity and a threshold condition on \(C\). For integers \(x\ge 1\) and \(y\ge 0\),
\[
\Omega_{P,C}(x,y)
\]
counts order-preserving \((x,y)\)-maps \(\varphi:P\to[x]\) with \(\varphi(c)\ge y\) for all \(c\in C\), while
\[
\Omega^\circ_{P,C}(x,y)
\]
counts strictly order-preserving \((x,y)\)-maps with \(\varphi(c)>y\) for all \(c\in C\) [1901.06720].

When \(P=C_n\) is the chain \(a_1\prec a_2\prec\cdots\prec a_n\), the data collapse to the position of the minimal celeste element. If \(a_{k+1}\) is minimal celeste, then the weak case imposes \(f(a_{k+1})\ge y\), and the strict case imposes \(f(a_{k+1})>y\). The classical chain formulas are
\[
\Omega_{C_n}(x)=\binom{x+n-1}{n},\qquad \Omega^\circ_{C_n}(x)=\binom{x}{n}.
\]
Their bivariate analogues are
\[
\Omega^\circ_{C_n,C}(x,y)=\sum_{i=0}^k \binom{y}{i}\binom{x-y}{n-i},
\]
and
\[
\Omega_{C_n,C}(x,y)=\sum_{i=0}^k \binom{y-2+i}{i}\binom{x-y+n-i}{n-i}.
\]
Thus, for chains, colored linear orderings admit explicit binomial-sum enumerations depending only on the first celeste position.

The same paper proves a decomposition over linear extensions for general posets and a bivariate reciprocity theorem,
\[
(-1)^{|P|}\,\Omega^\circ_{P,C}(-x,-y)=\Omega_{P,C}(x,y+1).
\]
For chains this becomes
\[
(-1)^n\,\Omega^\circ_{C_n,C}(-x,-y)=\Omega_{C_n,C}(x,y+1).
\]
The reciprocity shift \(y\mapsto y+1\) distinguishes the bivariate setting from the classical one-variable case. The chain formulas therefore serve both as closed-form counts and as the simplest instance of a broader reciprocity theory for order-preserving colorings [1901.06720].

## 3. Linearly ordered colorings of hypergraphs

In hypergraph theory, a linearly ordered coloring is an LO \(k\)-coloring of an \(r\)-uniform hypergraph \(H=(V,E)\): a function \(c:V\to\{1,\dots,k\}\) such that every edge has a unique maximum color. Equivalently, for \(r=3\), if two vertices in an edge share the same color, then the third must have a strictly larger color. This is stronger than classical non-monochromatic coloring, which only excludes all-equal edges. The formalism fits naturally into Promise CSPs: if \({}^r_k\) denotes the \(r\)-ary template of tuples with a unique maximum, then an LO \(k\)-coloring is exactly a homomorphism \(H\to {}^r_k\), and the promise problem LO-\(\ell\) vs LO-\(k\) is the PCSP \(\mathrm{PCSP}({}^r_\ell,{}^r_k)\) [2204.05628].

For 3-uniform hypergraphs promised to admit an LO \(2\)-coloring, a polynomial-time, deterministic, constructive algorithm outputs an LO \(k\)-coloring with
\[
k=O\!\left(\left(\frac{n\log\log n}{\log n}\right)^{1/3}\right).
\]
The method first linearizes the instance by merging vertex pairs forced equal in every LO \(2\)-coloring, then alternates two types of progress on a linear hypergraph: a Type 1 step colors a large independent set with a new largest color, while a Type 2 step finds a large set meeting each hyperedge in \(0\) or \(2\) vertices and colors it with a new smallest color. The dense regime uses the mod-\(2\) system \(x_a+x_b+x_c\equiv 0\pmod 2\) and a constant-factor approximation for Max-Ones modulo \(2\); the sparse regime uses graph independent-set approximation. Balancing the two regimes at
\[
\Delta = n^{1/3}(\log n)^{2/3}(\log\log n)^{-2/3}
\]
yields the stated bound [2204.05628].

The complexity boundary is sharp in a different parameter range. For every fixed \(k\ge 2\) and constant \(r\ge k+2\), given an \(r\)-uniform hypergraph that admits an LO \(2\)-coloring, it is NP-hard to find an LO \(k\)-coloring. More generally, for every fixed \(2\le \ell\le k\) and \(r\ge k-\ell+4\), \(\mathrm{PCSP}({}^r_\ell,{}^r_k)\) is NP-hard. The proofs use the algebraic PCSP framework of polymorphism minions, minion homomorphisms, free structures, and pp-constructions, and they isolate a structural asymmetry between the regimes \(r<k+2\) and \(r\ge k+2\). In particular, hardness for \(r\ge k+2\) cannot be transferred to \(r<k+2\) by gadget reductions because the necessary minion homomorphisms do not exist [2204.05628].

A later development gives a logarithmic approximation for the 3-uniform promised LO \(2\)-colorable case. There is an algorithm that, given a 3-uniform hypergraph \(H\) with \(n\) vertices and \(m\) edges that admits an LO \(2\)-coloring, finds an LO \(\log_2(n)\)-coloring in time \(O(n^3+nm)\), and another that finds an LO \(\tfrac12\log_2(m)\)-coloring in the same time bound. The core subroutine solves \(Av=\mathbf{1}^m\pmod 2\), derives a subset \(T\subseteq V\) intersecting each 3-edge in \(0\) or \(2\) vertices and each 2-edge in exactly \(1\) vertex, guarantees \(|T|\ge n/2\), and then peels such sets layer by layer. A randomized rational nullspace method gives, with probability at least \(\tfrac14\), an LO-coloring with \(O(1)+2\log_2 n\) colors [2404.19556].

The principal open problem remains the constant-color regime for 3-uniform hypergraphs. The exact complexity of \(\mathrm{PCSP}({}^3_2,{}^3_k)\) for fixed \(k\ge 3\) is open, and the specific case \(k=3\) was already highlighted in the STACS’21 line of work. A plausible implication is that the 3-uniform case lies at the point where current combinatorial and algebraic methods cease to align cleanly.

## 4. Ordered vertices in graph theory

A distinct graph-theoretic tradition studies linear orderings of vertices rather than colorings of vertices by ordered labels. For a graph \(G\) and a linear ordering \(L\) of \(V(G)\), weak \(r\)-reachability \(WReach_r(G,L,v)\) and strong \(r\)-reachability \(SReach_r(G,L,v)\) determine the generalized coloring numbers
\[
wcol_r(G,L)=\max_v |WReach_r(G,L,v)|,\qquad scol_r(G,L)=\max_v |SReach_r(G,L,v)|.
\]
Optimizing over \(L\) gives \(wcol_r(G)\) and \(scol_r(G)\). These parameters interpolate between degeneracy \(+1\), treewidth \(+1\), and treedepth, and they govern acyclic and star colorings through greedy procedures based on reachability layers [1907.12149].

A central question is whether one can choose a single ordering that is simultaneously good for all radii. The main uniform-ordering theorem gives an affirmative answer up to explicit loss: for any graph \(G\), there exists an ordering \(\sigma^*\) such that for all \(r\in\mathbb N\),
\[
scol_r(G,\sigma^*)\le (2^r+1)\cdot (scol_{2r}(G))^{4r}.
\]
A more general simultaneous-control theorem treats multiple graphs on the same vertex set. These results yield new characterizations of bounded expansion and nowhere dense classes in terms of one ordering per graph or per subgraph. They also show an inherent limitation: no single ordering can be optimal for two different radii in general, and some deterioration is unavoidable [1907.12149].

A second ordered-graph framework describes hereditary properties by forbidden ordered patterns. A graph \(G\) is \(k\)-colorable if and only if there exists a linear ordering \(\le\) on \(V(G)\) with no \(k+1\) vertices
\[
v_1\le v_2\le \cdots \le v_{k+1}
\]
such that \(v_i v_{i+1}\in E(G)\) for every \(i\in\{1,\dots,k\}\). The ordered obstruction is the straight path \(St_{k+1}\). The same paper shows that every finite forbidden circular-order description can be translated into a finite forbidden linear-order description via the operator \(L(F)=c^{-1}[F]\). For \(k\ge 3\), the following are equivalent: \(G\) admits a circular ordering \(c(\le)\) such that \(SP_{k+1}\to (G,c(\le))\); \(G\) admits a \(PH_{k+1}\)-free linear ordering; and \(\chi_c(G)<k\), where \(\chi_c\) is the circular chromatic number [2112.00154].

These two theories use different notions of “colored linear ordering,” but both convert coloring constraints into structural statements about a single vertex order. In sparse graph theory, the order controls reachability; in forbidden-pattern theory, it controls the presence of monotone obstructions.

## 5. Colored orders in model theory

In model theory, a colored order is a linearly ordered first-order structure \(\mathcal M=(M,<,\ldots)\) whose language contains unary predicates interpreted as colors. Rubin’s program asks when a linearly ordered structure is not much more complicated than such a colored order. A precise answer is obtained through a hierarchy of tameness conditions. Linear finiteness (LF) bounds the number of \(\phi\)-types realized above an initial segment. Linear binarity (LB) says that, for an increasing tuple \(a_1<\dots<a_n\), the full type is determined by consecutive pair types. Strong linear binarity (SLB) strengthens this to automorphism gluing over arbitrary initial parts. Rubin’s binarity (RB) is still stronger. The implications are strict:
\[
(\mathrm{RB})\Rightarrow(\mathrm{SLB})\Rightarrow(\mathrm{LB})\Rightarrow(\mathrm{LF}).
\]
Moreover, an \(\omega\)-saturated linearly ordered structure satisfies RB if and only if it is definitionally equivalent to a colored order [1810.02324].

The same work introduces the ccel-reduct \(\mathcal M^{ccel}\), which names all unary definable sets and all definable convex equivalence relations. Its main structural theorem states that \(T=\operatorname{Th}(\mathcal M)\) satisfies SLB if and only if \(\mathcal M\) is definitionally equivalent to \(\mathcal M^{ccel}\); equivalently, \(T\) is, up to definitional equivalence, a theory of colored orders expanded by definable convex equivalence relations. Under SLB, every formula is equivalent modulo \(T\) to a Boolean combination of \(u\)-convex formulas. These are built from unary conditions and order constraints expressed through successor and predecessor classes of definable convex equivalence relations.

At the LF level one already gets strong geometric consequences. Every \(A\)-definable convex subset of \(M\) has an explicit successor-class form such as
\[
S^{-m}_{E_1}(a)\le x<S^{-n}_{E_2}(b)
\quad\text{or}\quad
S^{m}_{E_1}(a)<x\le S^{n}_{E_2}(b),
\]
and monotone binary relations decompose into finite disjunctions of such pieces. Definable unary functions are finite unions of successor-class maps. After naming all unary definable sets, convex equivalence relations, and the basic relations
\[
R_{n,E}(x,y): x\le S^n_E(y)\quad (n\ge 0),\qquad
R_{n,E}(x,y): x<S^n_E(y)\quad (n<0),
\]
the complete theory eliminates quantifiers. The resulting picture is that colored orders and their convex-equivalence expansions have a highly stratified yet explicitly describable geometry of definable sets [1810.02324].

## 6. Finite colored linear orders and Ehrenfeucht–Fraïssé classification

For finite colored linear orders, the natural equivalence notion is \(n\)-equivalence under the \(n\)-move Ehrenfeucht–Fraïssé game. If \(A=(L_A,<,F_A)\) and \(B=(L_B,<,F_B)\) are colored linear orders over the same color set, then \(A\equiv_n B\) means Duplicator wins the \(n\)-move game preserving both order and colors. The key invariant is the \(n\)-character of a point \(a\), namely
\[
\{[A<a]_n,[A>a]_n\}/c,
\]
where \(c\) is the color of \(a\). The fundamental theorem states that \(A\equiv_{n+1} B\) if and only if the multisets \(p_c(A)\) and \(p_c(B)\) agree for every color \(c\) [1705.04632].

At the level of \(\equiv_2\), the classification is completely explicit. For \(m\) colors, one associates to a finite colored linear order a \(T\)-configuration consisting of points \(x_1<\cdots <x_m\) and \(y_1>\cdots >y_m\), where the \(x_i\) record the first appearance of the \(i\)-th new color from the left and the \(y_i\) the last appearance of the \(i\)-th new color from the right. A realizability criterion is
\[
i+j<m+1 \Rightarrow x_i\le y_j.
\]
Together with the sets \(g(u,v)\) of colors occurring in each interval between consecutive \(T\)-points, this completely determines the \(\equiv_2\)-class. The families \(C_{T,g}\) classify the \(\equiv_2\)-classes of finite \(m\)-colored linear orders.

This classification yields an algorithmic reduction to canonical form. Any finite colored string has a \(2\)-equivalent \(\equiv_2\)-optimal substring obtained by computing the colored \(T\)-configuration, computing each \(g(u,v)\), and then replacing each interval \((u,v)\) by a minimal interval in which each color from \(g(u,v)\) appears exactly once. The result is a substring of the original structure and is optimal in its \(\equiv_2\)-class. The exact least upper bound on lengths of optimal representatives of \(\equiv_2\)-classes with \(m\) colors is \(m^2+2m\).

At the level of \(\equiv_3\), the structure becomes substantially more intricate. A 2-colored string of length \(70\) is constructed that is \(\equiv_3\)-optimal because all its 2-characters are distinct, and a longer example of length \(74\) is \(\equiv_3\)-optimal despite repeated 2-characters. The paper uses De Bruijn-type constructions and directed graphs of character transitions to control these phenomena. This shows that, beyond \(\equiv_2\), local repetition of characters no longer suffices to determine reducibility; global transition constraints matter [1705.04632].

## 7. Countable shuffles, homogeneous colored orders, and enumeration

A different countable theory begins with shuffles over the rationals. For a non-empty countable set \(S\) of linear orders, a dense \(S\)-coloring \(\chi:\mathbb Q\to S\) assigns to each rational a color \(s\in S\) so that every color class is dense in \((\mathbb Q,<)\). Replacing each \(q\in\mathbb Q\) by a copy of the order \(\chi(q)\) yields the shuffle
\[
\Xi(S)=\sum_{q\in\mathbb Q} L_{\chi(q)}.
\]
Skolem’s theorem gives existence and uniqueness of dense colorings up to an automorphism of \((\mathbb Q,<)\), so \(\Xi(S)\) depends only on \(S\) up to isomorphism. For this class, a Cantor–Schröder–Bernstein phenomenon holds: if two countable shuffles embed as convex subsets into each other, then they are order-isomorphic [2411.02297].

Homogeneous colored linear orders admit a finite-skeleton classification. A countable colored linear ordering \((L,<,c_1,\dots,c_k)\) is homogeneous if every isomorphism between finitely generated colored substructures extends to an automorphism. Such an order is homogeneous if and only if there exist pairwise disjoint subsets \(A\subseteq C\) and \(\{A_v:v\in V\}\subseteq \mathcal P(C)\) such that each open interval \(I_v\) is isomorphic to the color shuffle \(Sh_c(A_v)\), and each \(a\in A\) occurs as a single isolated point of color \(a\). Thus homogeneous colored orders are finite unions of color-dense open intervals together with finitely many isolated colored points, with colors not repeated across different components. The paper proves that the number \(L(k)\) of isomorphism types of countable homogeneous colored linear orders in \(k\) colors has exponential generating function
\[
H(x)=\sum_{k\ge 0} L(k)\frac{x^k}{k!}=\frac{e^x}{2-x-e^x}.
\]
The first values are
\[
L(0)=1,\ L(1)=3,\ L(2)=14,\ L(3)=95,\ L(4)=858,\ L(5)=9687,\dots
\]
and, with
\[
Z=2-W(e^2)\approx 0.442854,\qquad
R=\frac{-e^2}{e^{W(e^2)}+e^2}\approx -0.6089389,
\]
one has
\[
L(k)\sim -k!\cdot R\cdot (1/Z)^{k+1}.
\]
The proportion of homogeneous colored orders using all \(k\) colors tends to \(1/W(e^2)\approx 0.6422007\) [2604.14255].

The same finite-skeleton method approximates \(sp\)-homogeneity through the relational notions \(C_{n,m}\)-homogeneity. Here the language records exact numbers of predecessors and successors and exact finite distances by predicates \(S_i\), \(P_j\), and \(Adj_k\). A linear ordering is \(C_{n,m}\)-homogeneous if this expansion is homogeneous, and \(C_{\infty,\infty}\)-homogeneity is exactly \(sp\)-homogeneity. For finite \(n,m\), the class is finite up to isomorphism, and its members are very closely related to homogeneous colored linear orders with finitely many colors. If \(k=n+m+1\), the number \(I(k)\) of \(C_{n,m}\)-homogeneous orders satisfies
\[
I(k)=O(k!\cdot 2.123^k),
\]
and explicit recurrence and Stirling-number formulas are given. A later paper places these classes in a computability-theoretic hierarchy: \(sp\)-homogeneous orders are always relatively \(\Delta^0_4\)-categorical, the set of \(sp\)-homogeneous orderings is \(\Pi^0_5\)-complete, and the set of weakly \(sp\)-homogeneous orderings is \(\Sigma^0_6\)-complete [2509.25005].

Taken together, these results show that countable colored linear orderings can be both rigid and enumerable. Dense shuffles over \((\mathbb Q,<)\) exhibit a strong convex-embedding rigidity, while homogeneous colored orders reduce to finite combinatorics despite being generally infinite structures. An open enumerative problem is whether the bound \(I(k)=O(k!\cdot 2^k)\) holds for the \(C_{n,m}\)-homogeneous counts [2604.14255].

Source: https://www.emergentmind.com/topics/colored-linear-orderings