---
title: Distinguishing Chromatic Number
url: https://www.emergentmind.com/topics/distinguishing-chromatic-number
type: topic
---

# Distinguishing Chromatic Number

The distinguishing chromatic number of a graph \(G\), denoted \(\chi_D(G)\), is the minimum number of colors in a proper vertex coloring that is preserved only by the identity automorphism. It refines the chromatic number by imposing simultaneous properness and symmetry breaking, and it sits at the intersection of graph coloring, automorphism groups, metric graph theory, homomorphism theory, and structural graph classes. The parameter has been studied on finite and infinite graphs, on highly symmetric families such as circulants and Cayley graphs, and on graphs derived from subdivision operations, with both sharp general bounds and detailed exact classifications now available [2112.03594], [1910.12107], [2411.07000].

## 1. Definition and basic framework

For a graph \(G\), a proper \(k\)-coloring is a map \(c:V(G)\to\{1,\dots,k\}\) such that adjacent vertices receive distinct colors. The distinguishing chromatic number is
\[
\chi_D(G)=\min\{k:\exists\ \text{a proper }k\text{-coloring preserved only by the identity automorphism}\}.
\]
Equivalently, \(\chi_D(G)\) is the least \(k\) for which \(G\) admits a proper coloring whose color classes are preserved only by the identity element of \(\operatorname{Aut}(G)\) [2112.03594].

Two basic inequalities are immediate. Because a distinguishing proper coloring is, in particular, a proper coloring, one always has \(\chi(G)\le \chi_D(G)\). Because a distinguishing proper coloring is also a distinguishing coloring without the properness constraint, one also has \(D(G)\le \chi_D(G)\), where \(D(G)\) is the distinguishing number [1309.0416], [0907.0691].

The parameter is sensitive to residual symmetry after proper coloring. If \(\operatorname{Aut}(G)=\{\mathrm{id}\}\), then every proper coloring is automatically distinguishing, so \(\chi_D(G)=\chi(G)\) [1203.5765]. At the opposite extreme, complete multipartite graphs are exactly the graphs with \(\chi_D(G)=|V(G)|\) [2112.03594]. This includes complete graphs and complete bipartite graphs such as \(K_n\) and \(K_{m,n}\).

A useful reformulation comes from homomorphism theory: a distinguishing proper \(n\)-coloring is exactly a distinguishing homomorphism to \(K_n\) [1309.0416]. This viewpoint connects \(\chi_D\) to broader symmetry-breaking constructions on infinite graphs and to upper bounds derived from homomorphic embeddings.

The gap between \(D(G)\) and \(\chi_D(G)\) can be large. For the star \(K_{1,n}\), the leaves can be permuted arbitrarily, so \(D(K_{1,n})=n\), while properness forces the center to receive a color different from all leaves, giving \(\chi_D(K_{1,n})=n+1\) for \(n\ge 2\) [0907.0691].

## 2. Relation to locating colorings and metric structure

A central structural comparison identifies \(\chi_D\) as the weaker of two proper coloring parameters based on symmetry breaking and metric distinguishability. For a proper coloring \(c\) with color classes \(C_1,\dots,C_k\), the color code of a vertex \(v\) is
\[
r_c(v)=(d(v,C_1),d(v,C_2),\dots,d(v,C_k)).
\]
A proper coloring is locating if distinct vertices have distinct color codes, and the minimum number of colors in such a coloring is the locating chromatic number \(\chi_L(G)\) [2112.03594].

For every connected graph, any locating coloring is distinguishing. Consequently,
\[
\chi_D(G)\le \chi_L(G).
\]
This implies that any upper bound on \(\chi_L(G)\) is also an upper bound on \(\chi_D(G)\), and any lower bound on \(\chi_D(G)\) is also a lower bound on \(\chi_L(G)\) [2112.03594].

The comparison is strict at the level of colorings. On the path \(P_7\) with vertices \(a_1,\dots,a_7\), the proper coloring with classes \(\{a_1,a_5\}\), \(\{a_2,a_4,a_6\}\), and \(\{a_3,a_7\}\) is distinguishing but not locating, since \(r_c(a_2)=r_c(a_4)=(1,0,1)\). Nevertheless, the parameters still coincide on this graph: \(\chi_D(P_7)=\chi_L(P_7)=3\) [2112.03594].

The locating-coloring comparison yields additional bounds. If \(\dim(G)\) is the metric dimension, then \(\chi_L(G)\le \chi(G)+\dim(G)\), hence
\[
\chi_D(G)\le \chi(G)+\dim(G).
\]
If \(G\) has order \(n\ge 3\) and \(\operatorname{diam}(G)\ge 2\), then
\[
\chi_D(G)\le n-\operatorname{diam}(G)+2.
\]
For complete multipartite graphs, the two parameters coincide maximally:
\[
\chi_D(G)=\chi_L(G)=|V(G)|.
\]
More generally, for every pair of integers \(2\le n\le m\le 2n-1\), there exists a graph \(G\) with \(\chi_D(G)=n\) and \(\chi_L(G)=m\), so the gap \(\chi_L(G)-\chi_D(G)\) can realize every value from \(0\) to \(n-1\) [2112.03594].

The equality problem remains open in general. The class of graphs satisfying \(\chi_D(G)=\chi_L(G)\) is not characterized, although there are complete descriptions when both parameters equal \(3\). For trees, \(\chi_D(T)=\chi_L(T)=3\) exactly for those \(T\) in the Baskoro–Asmiati family \(\mathcal T\) with \(|\operatorname{Aut}(T)|\ge 2\). For graphs containing cycles, the same equality holds for the Asmiati–Baskoro family \(\mathcal G\) when either the graph is bipartite with \(|\operatorname{Aut}(G)|\ge 2\), or it is non-bipartite [2112.03594].

## 3. General bounds and hereditary graph classes

The universal finite-graph bound due to Collins and Trenk is
\[
\chi_D(G)\le 2\Delta(G)
\]
for every connected graph \(G\), with equality only for \(K_{\Delta,\Delta}\) and \(C_6\). Much of the later literature shows that this upper bound can be reduced substantially under structural restrictions such as large girth or forbidden induced subgraphs [1707.05439], [2505.17193].

The following bounds are explicitly established for connected finite graphs.

| Graph class | Bound on \(\chi_D(G)\) | Equality information |
|---|---:|---|
| Girth at least \(5\), \(G\neq C_6\) | \(\le \Delta(G)+1\) | Stars show sharpness |
| Bipartite, girth at least \(6\), \(\Delta(G)\ge 3\) | \(\le \Delta(G)+1\) | Includes trees and unicyclic extensions |
| \(C_4\)-free | \(\le \Delta(G)+2\) | Equality iff \(G\cong C_6\) |
| Chordal | \(\le \Delta(G)+1\) | Equality iff \(G\) is symmetric or \(G=\alpha(G)K_1+K_{\omega(G)-1}\) |
| \((C_4,2K_2)\)-free | \(\le \Delta(G)+1\) | Equality iff \(G\cong \alpha(G)K_1+K_{\omega(G)-1}\) or \(G\cong C_5\) |
| \(2K_2\)-free | \(\le 2\Delta(G)-\omega(G)+2\) | Equality iff \(G\) is complete or balanced complete bipartite |
| Claw-free | \(\le \chi(G)+p(G)\) | Also \(\le \Delta(G)+2\), equality iff \(C_6\) or \(K_{n/2}[2K_1]\) |
| \((\text{claw},\text{diamond})\)-free | \(\le \Delta(G)+1\) unless \(C_4,C_6\) | If also \(K_k\)-free, then \(\le k\) |

These results combine BFS-based fixing arguments, simplicial-vertex reductions, dominating-clique structure, module decompositions, and line-graph translations [1707.05439], [1709.10021], [2505.17193].

Large girth is especially effective. For connected graphs of girth at least \(5\), a BFS-order greedy coloring yields \(\chi_D(G)\le \Delta(G)+1\) except for \(C_6\) [1707.05439]. In the bipartite girth-at-least-\(6\) case, the absence of \(3\)-, \(4\)-, and \(5\)-cycles gives strong neighborhood separation across BFS layers, which again forces \(\chi_D(G)\le \Delta(G)+1\) [1709.10021].

Forbidden induced subgraphs lead to sharper hereditary-class theorems. In particular, connected \(C_4\)-free graphs satisfy \(\chi_D(G)\le \Delta(G)+2\), with \(C_6\) as the unique equality case. Chordal graphs satisfy \(\chi_D(G)\le \Delta(G)+1\), and the equality cases are completely characterized by symmetric constructions and joins of the form \(\alpha(G)K_1+K_{\omega(G)-1}\) [2505.17193].

A distinct line of work studies list versions. If \(G\) is \((n-2r+1)\)-connected, \(n>2r\), \(4\le r\le \Delta(G)/2\), and \(G\) contains no induced \(K_{r,r+1}\) or \(K_{r+1,r+1}\), then
\[
\chi_D^{\ell}(G)\le 2\Delta(G)-\bigl(3\lfloor(\Delta(G)+2)/(r+1)\rfloor-4\bigr).
\]
In particular, when \(r>6\) and \(\Delta(G)\ge 3r+1\), one obtains
\[
\chi_D^{\ell}(G)\le 2\Delta(G)-5,
\]
and since \(\chi_D(G)\le \chi_D^{\ell}(G)\), the same upper bounds transfer to \(\chi_D(G)\) [2509.11992].

## 4. Exact values on classical and derived families

Several classical families admit exact formulas. For cycles,
\[
\chi_D(C_n)=
\begin{cases}
3,& n\in\{3,5\}\ \text{or}\ n\ge 7,\\
4,& n\in\{4,6\}.
\end{cases}
\]
For odd paths \(P_{2k+1}\), one has \(\chi(P_{2k+1})=2\), \(\dim(P_{2k+1})=1\), \(\chi_L(P_{2k+1})=3\), and consequently \(\chi_D(P_{2k+1})=3\) [2112.03594], [2303.13759].

Complete multipartite graphs form the maximal regime:
\[
\chi_D(G)=|V(G)|
\quad\Longleftrightarrow\quad
G\ \text{is complete multipartite}.
\]
Thus \(K_n\) and \(K_{m,n}\) require \(|V(G)|\) colors in distinguishing proper colorings [2112.03594].

Hamiltonian circulant graphs of maximum degree at most \(4\) provide a large exact classification. For Möbius ladders,
\[
\chi_D\bigl(C_n(1,n/2)\bigr)=
\begin{cases}
4,& n=4,\\
6,& n=6,\\
3,& n\ge 8.
\end{cases}
\]
For all \(n\ge 7\),
\[
\chi_D\bigl(C_n(1,2)\bigr)=4,
\]
and by isomorphism the same holds for \(k=(n-1)/2\). For wreath graphs,
\[
\chi_D\bigl(C_n(1,n/2-1)\bigr)=5\quad\text{for }n\ge 10.
\]
Outside a controlled list of arithmetic and small-order exceptions, most tetravalent \(C_n(1,k)\) satisfy \(\chi_D=3\), while exceptional cases include \(\chi_D(C_{10}(1,3))=5\), \(\chi_D(C_{13}(1,5))=4\), and \(\chi_D(C_{15}(1,4))=4\) [2303.13759].

Subdivision-derived constructions also admit exact or near-exact formulas. For the middle graph \(M(G)\) of a connected graph \(G\) of order at least \(3\),
\[
\chi_D(M(G))=
\begin{cases}
\Delta(G)+1,& G\notin\{C_4,K_4,C_6,K_{3,3}\},\\
\Delta(G)+2,& G\in\{C_4,K_4,C_6,K_{3,3}\}.
\end{cases}
\]
The proof uses the identity \(M(G)=L(G^+)\), where \(G^+\) is the endline graph, together with results on distinguishing chromatic indices of line graphs [2411.07000].

For subdivision graphs \(S(G)\), the sharp statements are conditional on the base graph. If \(G\) is not a cycle and \(D(G)\ge 3\), then
\[
\chi_D(S(G))\le D(G),
\]
and this bound is sharp. If \(G\) is not a cycle and \(D(G)=2\), then \(\chi_D(S(G))=3\). If \(D(G)=1\), then \(\chi_D(S(G))=2\). For cycles,
\[
\chi_D(S(C_n))=
\begin{cases}
3,& n=3\ \text{or}\ n\ge 6,\\
3,& n=4\ \text{or}\ 5,
\end{cases}
\]
with the distinction that in the first case this equals \(D(C_n)+1\), and in the second it equals \(D(C_n)\) [2411.07000].

## 5. Infinite graphs, homomorphisms, and probabilistic models

On infinite graphs, the homomorphism perspective becomes especially powerful. A homomorphism \(f:G\to H\) is distinguishing if the only automorphism of \(G\) preserving every fiber \(f^{-1}(x)\) is the identity. Distinguishing proper colorings are exactly distinguishing homomorphisms to complete graphs [1309.0416].

For connected existentially closed graphs, Bonato and Delić prove a strong existence theorem: if \(G\) is c.e.c. and \(G\to H\), then there are \(2^{\aleph_0}\) distinct distinguishing homomorphisms
\[
G\to H\vee K_2.
\]
As a corollary,
\[
\chi_D(G)\le \chi(H\vee K_2)=\chi(H)+2,
\]
and in particular \(\chi_D(G)\le \chi(G)+2\) under the same hypotheses [1309.0416].

A separate degree-based theory applies to connected infinite graphs with finite maximum degree. One has
\[
\chi_D(G)\le 2\Delta(G)-1.
\]
This bound is sharp for \(\Delta=2\), since the double ray has \(\chi_D=3=2\Delta-1\). For infinite trees with finite maximum degree \(\Delta\), the stronger bound \(\chi_D(T)\le \Delta\) holds, and if the tree is locally finite with infinite motion, then \(\chi_D(T)\le 3\). More generally, connected subcubic graphs with infinite motion satisfy \(\chi_D(G)\le 4\) [1910.12107].

These infinite-motion results reflect a recurring principle: low-color symmetry breaking becomes possible when nontrivial automorphisms move infinitely many vertices. The ray has \(\chi_D=2\), the double ray has \(\chi_D=3\), and infinite \(d\)-regular trees with \(d\ge 3\) satisfy \(\chi_D\le 3\) [1910.12107].

Random Cayley graphs provide a probabilistic counterpart. For random inverse-closed Cayley graphs \(\Gamma_p(A)\) over finite abelian groups of Type I, where \(\gcd(|A|,6)=1\), and of Type II, where \(A\cong \mathbb Z_5\times N\) with \(N\) odd-order and non-cyclic, the paper proves that in the stated \(p\)-ranges,
\[
\chi_D(\Gamma_p(A))\le \chi(\Gamma_p(A))+1
\]
with probability at least \(1-n^{-2(\log n)}\) [1406.5358]. The argument combines asymptotically minimal automorphism groups with either an independent triple not stabilized by any nontrivial automorphism or a motion-lemma refinement inside a largest color class.

## 6. Algorithmic, list, and cross-domain extensions

The computational complexity of \(\chi_D\) depends sharply on the target number of colors. For \(k\ge 3\), deciding whether \(\chi_D(G)\le k\) is NP-hard. For \(k=2\), the problem is at least as hard as Graph Automorphism and no harder than Graph Isomorphism: the connected-case decision problem is many-one equivalent to Graph Automorphism, and the general decision problem is polynomial-time Turing reducible to Graph Isomorphism [0907.0691].

Trees form a tractable and unusually rigid class. For every tree \(T\),
\[
D_\ell(T)=D(T),\qquad \chi_D(T)=\chi_{D_\ell}(T),\qquad \chi_D(T)\le D(T)+1.
\]
The proofs rely on recursive counts of inequivalent distinguishing colorings over isomorphism classes of rooted child subtrees, extending Cheng’s enumerative technique. The resulting dynamic programs compute both \(D(T)\) and \(\chi_D(T)\) in \(O(n\log n)\) time, and they characterize exactly when the upper bound \(\chi_D(T)=D(T)+1\) is attained [1111.4989].

The parameter also satisfies a Nordhaus–Gaddum-type inequality. For every graph \(G\) of order \(n\),
\[
\chi_D(G)+\chi_D(\overline G)\le n+D(G).
\]
This yields the product bound
\[
\chi_D(G)\chi_D(\overline G)\le \bigl((n+D(G))/2\bigr)^2.
\]
Equality cases are studied through NGD-graphs, and the classical NG-graph structure can be recognized in polynomial time via a degree-based partition into the sets \(A_G\), \(B_G\), and \(C_G\) defined by degrees relative to \(\chi(G)-1\) [1203.5765].

Several extensions preserve the same symmetry-breaking philosophy while changing the underlying combinatorial object. For oriented graphs, the extremal oriented distinguishing chromatic number \(D\chi^{\min}\) or \(D\chi^{\max}\) is known for paths, cycles, complete graphs, and complete bipartite graphs. For example,
\[
D\chi^{\max}(P_n)=
\begin{cases}
2,& n\ \text{even},\\
3,& n\ \text{odd},
\end{cases}
\qquad
D\chi^{\max}(C_4)=4,
\qquad
D\chi^{\max}(K_n)=n,
\]
and for balanced complete bipartite graphs,
\[
D\chi^{\min}(K_{n,n})=2,\qquad D\chi^{\max}(K_{n,n})=2n
\]
[1910.12738].

For posets, properness is reinterpreted by requiring comparable points to receive distinct colors. The resulting distinguishing chromatic number \(\chi_D(P)\) satisfies \(\chi_D(P)\le \chi_D(G_P)\), where \(G_P\) is the comparability graph, but the gap can be arbitrarily large. In distributive lattices \(L\), one has
\[
\chi_D(L)\le |Q_L|+\chi_D(Q_L)-1
\quad\text{when }\chi_D(Q_L)\ge 3,
\]
and for Boolean lattices,
\[
\chi_D(B_n)\le n+3
\]
[1905.09858].

Taken together, these results show that \(\chi_D\) is not merely a chromatic invariant with an automorphism constraint. It is a structural measure of how much proper coloring must be refined to destroy symmetry, and its behavior is governed by the interplay among local degree, global automorphism structure, metric distinguishability, homomorphic codings, and class-specific decomposition theorems.

Source: https://www.emergentmind.com/topics/distinguishing-chromatic-number