Odd Cycle Transversal in $P_k$-Free Graphs

Abstract: The Odd Cycle Transversal (OCT) problem, which asks for a minimum subset of vertices whose removal renders a graph bipartite, is a central problem in algorithmic graph theory. It is known to be NP-complete even on $P_k$-free graphs for $k \ge 6$. Furthermore, assuming the Unique Games Conjecture (UGC), OCT does not admit a constant-factor approximation algorithm on general graphs. Motivated by these hardness results, we investigate the approximability of OCT on $P_k$-free graphs. We first establish that the problem becomes polynomial-time solvable on specific subclasses of $P_k$-free graphs, most notably $(P_6, C_3)$-free graphs, by exploiting a structural decomposition into rings of bipartite graphs. Leveraging these tractable substructures as a basis, we present a constant-factor approximation algorithm for OCT on general $P_k$-free graphs. We achieve an approximation ratio of $k-2$ when $k$ is odd and $k-3$ when $k$ is even. These results provide the first nontrivial constant-factor approximations for this class dependent on $k$, aligning with the UGC implication that no approximation factor independent of $k$ is likely to exist.

• The paper introduces an exact polynomial-time algorithm for OCT in (P6, C3)-free graphs using a novel ring decomposition that reduces the problem to minimum vertex cover computations.
• The research extends the approach to all Pk-free graphs by developing the first constant-factor approximation algorithm, with factors of k-2 for odd k and k-3 for even k.
• The study provides robust structural theorems and module partitions that elucidate odd cycle deconstruction, bridging combinatorial insights with practical algorithm design.

Odd Cycle Transversal in $P_k$-Free Graphs

Problem Setting and Motivation

The Odd Cycle Transversal (OCT) problem is a classical NP-hard problem: Given a graph $G = (V, E)$, the goal is to find a minimum-size subset $S \subseteq V$ whose removal results in a bipartite graph. OCT is equivalently the hitting-set problem for odd cycles and is fundamentally related to structural graph theory, fixed-parameter tractability, and graph modification problems.

The computational barrier for OCT is steep: The problem is NP-complete even for moderately restricted graph classes such as $P_6$-free graphs (i.e., graphs without an induced path of six vertices) and, under the Unique Games Conjecture (UGC), does not admit any constant-factor polynomial-time approximation on general graphs. This motivates systematic investigation into the behavior of OCT on hereditary graph classes, particularly $P_k$-free graphs, and the existence of approximation guarantees that depend solely on the forbidden path length $k$.

Structural Insights for $(P_6, C_3)$-Free Graphs

A principal contribution is a comprehensive structural decomposition of $(P_6, C_3)$-free graphs—a critical minimal setting where the problem switches from polynomial-time solvable ($P_5$-free) to NP-complete. The analysis is based on an explicit and fine-grained partition of the vertex set relative to an induced $C_5$, systematically forming modules that capture all possibilities for adjacency: sets $G = (V, E)$0, $G = (V, E)$1, $G = (V, E)$2, $G = (V, E)$3, $G = (V, E)$4 (corresponding to neighborhood structures with one or two nonconsecutive neighbors on the $G = (V, E)$5), and distance-2 sets. The paper proves that every vertex is at most distance-two from the $G = (V, E)$6 in such graphs.

Leveraging these partitions, the paper establishes a pivotal “ring of bipartite graphs” decomposition: Every induced subgraph corresponding to a ‘ring’ of five modules forms a cyclic structure where edges exist only between consecutive modules, with all modules being independent sets. This structure underpins the main algorithmic results and enables reduction of OCT to a sequence of minimum vertex cover computations in bipartite graphs. The detailed local analysis, including the behavior of induced $G = (V, E)$7 and chain graphs, is crucial for proving global properties.

Algorithmic Results

Polynomial-Time Exact Algorithm for $G = (V, E)$8-Free Graphs

The constructed ring decomposition yields an exact polynomial-time algorithm for OCT in $G = (V, E)$9-free graphs. The main result is the identification of bottleneck cuts—any minimal OCT is realized by a minimum vertex cover in one of the five bipartite subgraphs induced by consecutive parts of the ring. The algorithm systematically enumerates all possible ring structures (via a constant-size “coarse” auxiliary graph on the partition classes), extracts chordless rings, and recursively solves the OCT via branching on optimal vertex covers. The correctness and polynomial running time are rigorously established via induction and a suite of structural lemmas.

Constant-Factor Approximation for $S \subseteq V$0-Free Graphs

The paper generalizes its techniques to all $S \subseteq V$1-free graphs for $S \subseteq V$2, establishing the first constant-factor approximation algorithm for OCT in this regime, with the approximation factor depending solely on $S \subseteq V$3:

• For odd $S \subseteq V$4, the approximation factor is $S \subseteq V$5.
• For even $S \subseteq V$6, the factor is $S \subseteq V$7.

The method is based on greedy packing of short odd cycles (cycles of length $S \subseteq V$8, where $S \subseteq V$9 is $P_6$0 or $P_6$1 depending on the parity), removing all vertices of each packed cycle, and then reducing the residue to an exact OCT computation on a structurally manageable instance. The argument is tight to UGC-based lower bounds—no $P_6$2-independent approximation is expected for general $P_6$3-free graphs.

Notably, the running time is polynomial and independent of $P_6$4.

Structural Theorems for $P_6$5-Free Graphs

A second central component is a robust extension of the ring decomposition for $P_6$6-free graphs (for odd $P_6$7), and analogously for $P_6$8-free graphs (for even $P_6$9). The paper generalizes the ring structure, proving that every vertex in such graphs can be assigned to one of $P_k$0 (or $P_k$1) parts, with edges only between consecutive parts.

For each case, the minimum odd cycle transversal is given by the minimal-size part, or by solving minimum vertex cover in a bipartite link induced by two consecutive parts. All proofs are combinatorial and rely on careful forbidden subgraph and adjacency analyses.

Main Theorems and Structural Lemmas

Significant technical results include:

• Bottleneck Cut Theorem: For every ring constructed as above, the size of a minimum OCT equals the minimal vertex cover size among the ring’s bipartite links.
• Chain Graph and $P_k$2 Lemmas: The presence of certain local structures forces global decomposability, allowing replacement arguments for solutions.
• Generalization to Arbitrary $P_k$3: The recursive decomposition works for arbitrary $P_k$4, under suitable large-cycle forbiddenness.

Implications and Theoretical Significance

This work establishes a strong dichotomy for the OCT problem under induced path and cycle constraints, connecting the complexity of the problem with detailed subgraph structure. The paper strengthens the understanding of hereditary graph classes for which transversals to bipartite graphs admit efficient algorithms or meaningful approximation guarantees. By aligning the tractability boundary with known complexity-theoretic obstacles (such as UGC-hardness), the paper answers a long-standing open question on the parameterized approximability of OCT in $P_k$5-free graphs.

From a practical perspective, the dependence on $P_k$6 in the approximation ratio is qualitatively optimal, and the ring decomposition gives a powerful tool for designing exact algorithms in highly structured graph families.

Directions for Future Development

The primary open question is whether the $P_k$7 (odd $P_k$8) or $P_k$9 (even $k$0) approximation factors can be improved further, either by constant improvements or lower bounds under complexity assumptions. Of particular interest is whether $k$1-free graphs admit a 2-approximation algorithm—current results yield a 3-approximation.

Another direction is tightening kernelization and parameterized complexity for OCT in these classes and identifying forbidden subgraph regimes where stronger structural decompositions or fixed-parameter tractable algorithms arise.

Potential extension to related transversal problems (e.g., to other hereditary classes, odd-minor-free graphs, or $k$2-bounded settings) is also compelling, and the structural lemmas developed here may be useful in broader graph modification theory.

Conclusion

This work presents a rigorous structural and algorithmic theory for the Odd Cycle Transversal problem in $k$3-free graphs, providing the first parameterized constant-factor approximation for $k$4 and an exact polynomial-time algorithm for $k$5-free graphs. The convergence of combinatorial structure and algorithmic tractability illustrated in this framework is likely to inform further advances in vertex-deletion and transversal problems for structurally restricted graphs.

Reference: "Odd Cycle Transversal in $k$6-Free Graphs" (2606.07453).