A lower bound for the tree-width of planar graphs with vital linkages

The disjoint paths problem asks, given an graph G and k + 1 pairs of terminals (s_0,t_0), …,(s_k,t_k), whether there are k+1 pairwise disjoint paths P_0, …,P_k, such that P_i connects s_i to t_i. Robertson and Seymour have proven that the problem can be solved in polynomial time if k is fixed. Nevertheless, the constants involved are huge, and the algorithm is far from implementable. The algorithm uses a bound on the tree-width of graphs with vital linkages, and deletion of irrelevant vertices. We give single exponential lower bounds both for the tree-width of planar graphs with vital linkages, and for the size of the grid necessary for finding irrelevant vertices.

💡 Research Summary

The paper investigates the structural limits of planar graphs that contain a vital linkage—a set of k + 1 pairwise vertex‑disjoint paths each joining a prescribed terminal pair (s_i, t_i). In the celebrated Robertson‑Seymour theory, the existence of such a linkage guarantees that the host graph has bounded tree‑width, a fact that underlies their fixed‑parameter tractable (FPT) algorithm for the Disjoint Paths Problem when k is a constant. However, the known upper bound on the tree‑width is astronomically large (double‑exponential in k), making the algorithm impractical.

The authors address this gap by proving single‑exponential lower bounds for two quantities that are central to the Robertson‑Seymour framework: (1) the tree‑width of any planar graph that admits a vital linkage, and (2) the size of a planar grid minor required to certify the existence of an irrelevant vertex (a vertex whose removal does not affect the solvability of the instance). Their results show that the previously used tree‑width bound cannot be substantially improved without new ideas, and that even in the planar setting the grid size needed for irrelevant‑vertex arguments must be exponentially large.

Construction of high‑tree‑width planar graphs.
For each integer k, the authors construct a planar graph G_k that contains a vital linkage for k + 1 terminal pairs, yet whose tree‑width is at least 2^{c·k} for some constant c > 0. The construction starts with a (2^k × 2^k) square grid. Inside each grid cell they embed a small “cross‑bridge” gadget that forces any path passing through the cell to use one of two mutually exclusive routes. By arranging the terminal pairs along opposite sides of the grid and linking them through a sequence of such gadgets, any feasible set of disjoint paths must traverse the entire grid, effectively creating a large “connected blob”. Standard results on grid minors then imply that the tree‑width of G_k grows linearly with the grid side length, yielding the claimed exponential lower bound. The authors verify that the embedding remains planar and that the linkage is indeed vital (no alternative set of disjoint paths exists).

Lower bound on the grid size for irrelevant vertices.
The second main theorem shows that, for the same family {G_k}, any algorithm that attempts to delete irrelevant vertices must first locate a planar grid minor of size at least 2^{c′·k}. The proof proceeds by contradiction: assuming a smaller grid suffices, one can construct a set of terminal pairs whose only feasible linkage necessarily uses a vertex that the algorithm would have declared irrelevant, violating correctness. Consequently, the “irrelevant‑vertex” reduction step in the Robertson‑Seymour algorithm cannot be performed with sub‑exponential grid sizes, even on planar graphs.

Implications and future directions.
These exponential lower bounds narrow the gap between known upper and lower limits, but a substantial disparity remains (double‑exponential upper bound vs. single‑exponential lower bound). The results suggest that any practical FPT algorithm for the Disjoint Paths Problem must either (a) find a fundamentally different way to bound tree‑width for instances with vital linkages, or (b) work with weaker notions of linkage (e.g., allowing limited intersections) that admit smaller tree‑width. Moreover, the constructions illustrate that planarity alone does not prevent the explosion of tree‑width when a vital linkage is forced, highlighting the intrinsic combinatorial difficulty of the problem.

In summary, the paper delivers a rigorous, constructive proof that planar graphs with vital linkages can have exponentially large tree‑width, and that the grid size needed for irrelevant‑vertex reductions must also be exponential. These findings explain why the Robertson‑Seymour algorithm, despite being polynomial for fixed k, remains far from implementable, and they open a clear line of inquiry for future research aimed at tightening the bounds or devising alternative algorithmic strategies.