The complexity of nonrepetitive edge coloring of graphs

A squarefree word is a sequence $w$ of symbols such that there are no strings $x, y$, and $z$ for which $w=xyyz$. A nonrepetitive coloring of a graph is an edge coloring in which the sequence of colors along any open path is squarefree. We show that determining whether a graph $G$ has a nonrepetitive $k$-coloring is $\Sigma_2^p$-complete. When we restrict to paths of lengths at most $n$, the problem becomes NP-complete for fixed $n$.

💡 Research Summary

The paper investigates the computational complexity of non‑repetitive edge‑coloring of graphs, a problem that lies at the intersection of graph theory and combinatorics on words. A word is called square‑free (or non‑repetitive) if it contains no factor of the form xx, i.e., no substring that is the concatenation of two identical blocks. Translating this notion to graphs, an edge‑coloring is non‑repetitive if, for every open path, the sequence of colors encountered along the path is square‑free. The central decision problem studied is: given a graph G and an integer k, does G admit a non‑repetitive edge‑coloring using at most k colors?

The authors establish two main complexity results. First, they prove that the unrestricted problem is Σ₂^p‑complete. To show Σ₂^p‑hardness, they construct a polynomial‑time many‑one reduction from the canonical Σ₂^p‑complete problem of evaluating quantified Boolean formulas of the form ∀X ∃Y φ(X,Y). The reduction encodes each universally quantified variable and each clause into specially designed gadgets (vertex and clause gadgets) whose admissible colorings correspond to truth assignments. The non‑repetitive constraint forces any feasible coloring to respect the logical dependencies encoded by the quantifiers: a repetition on a path would correspond to an illegal assignment that violates the formula. The constructed graph has size polynomial in the size of the original formula, and it uses a fixed number of colors k that depends only on the structure of the gadgets. Consequently, the graph admits a k‑color non‑repetitive coloring if and only if the original quantified formula is true. Membership in Σ₂^p follows from the observation that a nondeterministic polynomial‑time machine can guess a coloring (existential step) and then a co‑NP verifier can check that no open path yields a repeated block (universal step). Hence the problem is Σ₂^p‑complete, placing it strictly above NP unless the polynomial hierarchy collapses.

The second result concerns a natural restriction: limit the length of paths that must be checked for repetitions to a fixed constant n. Under this restriction the decision problem becomes NP‑complete for each fixed n ≥ 2. The authors argue that when n is constant, the number of possible color sequences on any path is bounded by kⁿ, which is polynomial in the size of the input for fixed n and k. Consequently, a nondeterministic algorithm can guess a coloring and verify in polynomial time that every path of length at most n is non‑repetitive, placing the problem in NP. To prove NP‑hardness, they give a reduction from 3‑SAT (or another canonical NP‑complete problem) that builds a graph where any violation of the non‑repetitive condition on a short path corresponds to a falsified clause. The reduction ensures that a satisfying assignment exists if and only if a non‑repetitive k‑coloring respecting the length bound exists. Thus, for any constant bound on path length, the problem is NP‑complete.

These two theorems together delineate a sharp complexity landscape: the general non‑repetitive edge‑coloring decision problem resides at the second level of the polynomial hierarchy, while imposing a constant bound on the length of paths to be checked collapses the problem back to NP. This contrast highlights how the global non‑repetitive requirement introduces a universal quantification over exponentially many paths, which is precisely what elevates the problem to Σ₂^p.

Beyond the core complexity results, the paper discusses implications for related combinatorial problems. Non‑repetitive constraints appear in string algorithms (e.g., constructing infinite square‑free words), in graph labeling problems, and in the design of communication protocols where repeated patterns may cause interference. The authors suggest that the Σ₂^p‑completeness result may transfer to analogous decision problems in these domains, especially where a global avoidance of repetitions is required.

Finally, the authors outline several avenues for future work. One direction is parameterized complexity: investigating whether the problem becomes fixed‑parameter tractable (FPT) when parameterized by graph invariants such as treewidth, maximum degree, or the number of colors. Another is approximation: designing algorithms that use slightly more than the optimal number of colors while guaranteeing non‑repetitiveness, or randomized algorithms that succeed with high probability. They also propose experimental studies to evaluate heuristic methods on real‑world networks, where exact solutions may be infeasible but good‑enough non‑repetitive colorings could be valuable.

In summary, the paper provides a comprehensive theoretical treatment of non‑repetitive edge‑coloring, establishing Σ₂^p‑completeness for the unrestricted problem and NP‑completeness under a constant path‑length restriction, and it opens multiple promising research directions at the intersection of graph theory, combinatorics on words, and computational complexity.