Co-Gem-Free Graphs

• Co-gem-free graphs are defined by excluding the induced co-gem (P4+P1), ensuring that every induced P4 is a dominating set.
• They lie strictly between P4-free cographs and P5-free graphs, allowing for clear structural decompositions and tractable algorithmic approaches.
• Their bounded twin-width and clique-width support efficient algorithms in coloring, decomposition, and various graph problems.

A co-gem-free graph is a graph containing no induced subgraph isomorphic to the "co-gem," which is the five-vertex graph formed by the disjoint union of an induced four-vertex path ($P_4$) and an isolated vertex ($P_1$), that is, $P_4 + P_1$. This hereditary graph class sits strictly between cographs (the $P_4$-free graphs) and $P_5$-free graphs, enjoying structural and algorithmic properties that make it a tractable yet nontrivial object of study in finite and parameterized graph theory.

1. Definitions and Basic Properties

Let $G = (V, E)$ denote a simple graph. The co-gem is the graph on vertices $\{v_1, v_2, v_3, v_4, u\}$ with the edge set $\{v_1v_2, v_2v_3, v_3v_4\}$; equivalently, it is $P_4 + P_1$. A graph $G$ is co-gem-free if no subset of five vertices induces a subgraph isomorphic to $P_4 + P_1$.

Key closure properties:

• Co-gem-free graphs are closed under taking induced subgraphs.
• Every cograph (i.e., $P_4$-free graph) is co-gem-free, but the converse is false; some co-gem-free graphs contain induced $P_4$s, provided every such $P_4$ dominates the entire vertex set (see below).

A central structural observation is:

• Every induced $P_4$ in a co-gem-free graph is a dominating set. That is, for $a$-$b$-$c$-$d$ forming an induced $P_4$, every vertex $x\notin\{a,b,c,d\}$ must be adjacent to at least one of $a,b,c,d$; otherwise, $\{a,b,c,d,x\}$ would induce a co-gem (Rosenke et al., 31 Jan 2026).

2. Structural Characterizations and Decomposition

Cographs are exactly the $P_4$-free graphs. Co-gem-free graphs generalize cographs by allowing $P_4$ as an induced subgraph, but never a $P_4$ plus an isolated vertex. The following simple structural decomposition holds:

• If a connected co-gem-free graph is not a cograph (i.e., contains a $P_4$), every such $P_4$ is a dominating set.

This property enables straightforward decompositions:

• If the graph admits no $P_4$, apply cograph methods.
• Otherwise, select a dominating $P_4$ as a "small hub" for algorithmic processing (Rosenke et al., 31 Jan 2026).

More generally, co-gem-free graphs are a proper subclass of $P_5$-free graphs (since a $P_5$ contains a $P_4+P_1$ as an induced subgraph), and they do not admit an induced complement of the gem (the gem being $P_4$ with a universal vertex).

3. Recognition and Algorithmic Properties

Recognition Algorithms

• The naive recognition algorithm runs in $O(n^5)$ time, testing all five-vertex subsets for co-gem subgraphs (Rosenke et al., 31 Jan 2026).
• For the broader class $(C_5, \text{bull}, \text{gem}, \text{co-gem})$-free graphs (where $C_5$ is the 5-cycle and the "bull" is a triangle plus two pending edges), efficient bi-join decomposition algorithms allow recognition in near-linear time by detecting only "complete" nodes in a decomposition tree (Chang et al., 2023).

Table: Containment Relations

Graph class	Forbidden induced subgraph(s)	Properly contains?
Cographs	$P_4$	Co-gem-free
Co-gem-free	$P_4 + P_1$ (co-gem)	$P_5$-free
$P_5$-free	$P_5$	All graphs

Width Parameters and Decomposition

• Every co-gem-free graph has twin-width at most 2, as twin-width is at most the radius-$r$ flip-width and the latter is bounded by 2 for this class (Chang et al., 2023).
• From known relations, the clique-width and rank-width are $O(1)$-bounded in this class.

Notable Structural Characterization

• The absence of a co-gem is crucial for total decomposability by bi-joins; the presence of a co-gem subgraph marks a "prime" obstruction that halts the bi-join decomposition, so co-gem-free graphs are totally decomposable in the sense of Cunningham and de Montgolfier–Rao (Chang et al., 2023).

4. Chromatic and Vertex-Critical Graph Structure

Vertex-critical graphs play a central role in testing $k$-colorability: a $k$-vertex-critical graph has chromatic number $k$, but deleting any vertex lowers the chromatic number below $k$.

Dichotomy for Vertex-Critical (Gem, Co-gem)-free Graphs

For the subclass forbidding both the gem and the co-gem, every $k$-vertex-critical graph is:

• Either a complete graph $K_k$
• Or a "clique-expansion" of the 5-cycle $C_5[K_{n_1}, K_{n_2}, K_{n_3}, K_{n_4}, K_{n_5}]$, where each vertex of a $C_5$ is replaced by a clique and adjacent in the cycle structure (Abuadas et al., 2022).

Enumeration of $k$-vertex-critical (gem, co-gem)-free graphs is possible for all $k \le 16$ (e.g., there are 1, 1, 2, 2, 4, 6, 11, ..., 253 such graphs for $k=1,2,3,4,5,6,7,...,16$). For each fixed $k$ the list is finite and computable. The only infinite/finiteness cases remaining open for general $H$-free vertex-critical graphs are when $H = P_4 + \ell P_1$ for $\ell \ge 1$ (Abuadas et al., 2022).

Finiteness for $(\text{co-gem}, H)$-Free Vertex-Critical Graphs

For any graph $H$ on four vertices, there are only finitely many $k$-vertex-critical $(\text{co-gem}, H)$-free graphs for all $k$. The proofs use combinatorial arguments, including Sperner's Theorem to bound antichains, and computational enumeration in the hardest cases (e.g., $(\text{co-gem}, K_4)$-free implies 4-colorability) (Beaton et al., 2024).

5. Algorithmic Applications and Complexity

The structure of co-gem-free graphs enables polynomial-time algorithms for several problems traditionally hard on general graphs.

• Coloring: For every $k$ and $H$ of order 4, $k$-colorability is polynomial-time solvable and certifying in $(\text{co-gem}, H)$-free graphs by finding a $k$-coloring or checking for a finite list of vertex-critical forbidden subgraphs (Beaton et al., 2024).
• Free Flood-It / Miniature Painting: The equivalence between the Miniature Painting problem and Free Flood-It allows polynomial-time algorithms for these problems on co-gem-free graphs. The algorithm leverages the existence of a dominating $P_4$ and enumerates canonical painting plans using the structural properties outlined above. It operates in polynomial time, with a high but constant exponent (Rosenke et al., 31 Jan 2026).
• Width Parameter Algorithms: The bounded twin-width and total decomposability by bi-joins enable tractable model checking for first-order logic and polynomial-kernel algorithms for Ramsey-type regularity properties (Chang et al., 2023).

6. Open Problems and Research Directions

Key open problems and avenues for future research include:

• Improving the $O(n^5)$-time recognition of co-gem-free graphs to sub-$n^5$ complexity (Rosenke et al., 31 Jan 2026).
• Extending polynomial-time certifying coloring algorithms to the purely co-gem-free case for $k \ge 5$, where only computational evidence up to 12 vertices is presently known (Beaton et al., 2024).
• Characterizing which graphs $H$ of higher order (e.g., $|V(H)| = 5$) force finiteness of $k$-vertex-critical $(\text{co-gem}, H)$-free graphs.
• Determining if other small hub subgraphs beyond $P_4$ yield new tractable subclasses for coloring or reconfiguration problems.
• Reducing the exponent in the polynomial-time algorithms for Free Flood-It and related problems.

7. Connections to Broader Structural Graph Theory

Co-gem-free graphs are distinguished by their dominance properties of $P_4$ and their position within cograph extensions. They are not only relevant in chromatic theory (as minimal obstructions in coloring) but also as a structural backbone in decomposition-based width parameterizations, including flip-width, clique-width, and twin-width. In particular, the precise forbidden subgraph criteria and their impact on tractability place co-gem-free graphs at a vital intersection of structural theory and algorithmic application (Abuadas et al., 2022, Chang et al., 2023, Beaton et al., 2024, Rosenke et al., 31 Jan 2026).