The price of homogeneity is polynomial

We provide explicit and polynomial bounds for the Homogeneous Wall Lemma which occurred for the first time implicitly in the $13$th entry of Robertson and Seymour’s Graph Minors Series [JCTB 1990] and has since become a cornerstone in the algorithmic theory of graph minors. A wall where each brick is assigned a set of colours is said to be homogeneous if each brick is assigned the same set of colours. The Homogeneous Wall Lemma says that there exists a function $h$ that, given non-negative integers $q$ and $k$ and an $h(q,k)$-wall $W$ where each brick is assigned a, possibly empty, subset of ${ 1, \ldots , q }$ contains a $k$-wall $W’$ as a subgraph such that, if one assigns to each brick $B$ of $W’$ the union of the sets assigned to the bricks of $W$ in its interior, then $W’$ is homogeneous. It is well-known that $h(q,k) \in k^{\mathcal{O}(q)}$. The Homogeneous Wall Lemma plays a key role in most applications of the Irrelevant Vertex Technique where an exponential dependency of $h$ on $q$ usually causes non-uniform dependencies on meta-parameters at best and additional exponential blow-ups at worst. By proving that $h(q,k) \in \mathcal{O}(q^4 \cdot k^6)$, we provide a positive answer to a problem raised by Sau, Stamoulis, and Thilikos [ICALP 2020].

💡 Research Summary

The paper tackles a long‑standing bottleneck in the algorithmic theory of graph minors: the Homogeneous Wall Lemma, a tool that underlies the Irrelevant Vertex Technique and the Flat Wall Theorem. Historically, the lemma guaranteed that for any non‑negative integers q (the number of “colours” or profiles attached to bricks of a wall) and k (the size of a desired sub‑wall), there exists a function h(q,k) with h(q,k)∈k^{O(q)} such that any h(q,k)‑wall with a q‑colouring contains a homogeneous k‑wall. This exponential dependence on q has been the source of non‑uniformity in many parameterized algorithms, often forcing an extra exponential factor in the running time or making the exponent’s degree depend on a meta‑parameter (e.g., the size of a vertex set A that may attach arbitrarily to the interior of a wall).

The authors present a dramatically improved bound: there exists a function f(q,k) ∈ O(q⁴·k⁶) with the same existential guarantee. Moreover, they provide a constructive algorithm that, given a q‑colourful graph (G,χ) and a flat f(q,k)‑wall W₀, computes in O((q+k)·|E(G)|) time a flat k‑wall W₁ ⊆ W₀ that is homogeneous. The key technical shift is that W₁ is required only to be a subgraph of W₀, not a sub‑wall. Nevertheless, the algorithm preserves two crucial properties: (1) the same flatness witness (the “almost‑embedding” that certifies planarity of the wall’s interior) works for both W₀ and W₁, and (2) the tangle associated with W₁ is a truncation of the tangle of W₀. These conditions guarantee that all downstream applications that rely on the original lemma’s structural guarantees remain valid.

To achieve the polynomial bound, the paper introduces the notion of a q‑colorful graph, where each vertex carries a (possibly empty) subset of {1,…,q}. A wall is homogeneous if there exists a bipartition I∪O of the colour set such that no vertex in the wall’s compass carries a colour from O, and for every brick B and every colour i∈I there is a B‑bridge whose interior contains i. This abstract encoding of “profiles” (e.g., neighbourhoods of a small vertex set A, or more complex attachment patterns) allows the authors to treat many concrete situations uniformly.

The constructive proof proceeds in three conceptual stages. First, the authors analyse the distribution of colours across bricks and use combinatorial arguments (based on the pigeonhole principle and properties of tangles) to locate a dense region where the colour pattern repeats uniformly. Second, they apply a refined version of the Flat Wall Theorem to extract a flat sub‑wall of size k while maintaining the same flatness witness. Third, they verify that the extracted sub‑wall satisfies the homogeneity condition with respect to the bipartition I∪O. The entire process runs in polynomial time, yielding the claimed O(q⁴·k⁶) size bound.

The impact of this result is illustrated through the “price of homogeneity” problem raised by Sau, Stamoulis, and Thilikos. Previously, to homogenise a wall with respect to a set A of size |A|, one needed a wall of size k^{O(|A|)}. The new theorem reduces this to O(|A|⁴·k⁶), eliminating the exponential blow‑up. Consequently, algorithms that rely on homogenisation—most notably those using the Irrelevant Vertex Technique—no longer suffer from non‑uniform dependence on meta‑parameters. The authors demonstrate this by revisiting the recent algorithm for k‑H‑Minor Deletion by Morelle et al. The original algorithm’s running time contained a factor k^{O(f₁.₂(h_H))}, where h_H depends on the forbidden minor family H. By substituting the new homogeneous wall lemma, the exponent becomes a fixed polynomial in k and |A|, yielding a uniformly polynomial dependence and a simpler overall runtime.

Finally, the paper discusses the necessity of allowing subgraphs rather than sub‑walls. It argues that if one insists on a sub‑wall, the exponential dependence on q might be unavoidable, suggesting that the presented weakening is essentially optimal under current techniques. The authors leave open the possibility of future work that could retain sub‑wall structure while still achieving polynomial bounds, perhaps by developing new tools for handling tangles and flatness simultaneously.

In summary, the authors deliver a concrete, polynomial‑time algorithm that replaces the exponential k^{O(q)} bound of the Homogeneous Wall Lemma with a tight O(q⁴·k⁶) bound, preserve essential structural properties, and unlock more efficient parameterized algorithms across a range of graph‑minor‑based problems. This represents a significant step toward making the graph minors theory not only theoretically elegant but also practically algorithmic.