Kernelization for Treewidth-2 Vertex Deletion

The Treewidth-2 Vertex Deletion problem asks whether a set of at most $t$ vertices can be removed from a graph, such that the resulting graph has treewidth at most two. A graph has treewidth at most two if and only if it does not contain a $K4$ minor. Hence, this problem corresponds to the NP-hard $\mathcal{F}$-Minor Cover problem with $\mathcal{F} = {K4}$. For any variant of the $\mathcal{F}$-Minor Cover problem where $\mathcal{F}$ contains a planar graph, it is known that a polynomial kernel exists. I.e., a preprocessing routine that in polynomial time outputs an equivalent instance of size $t^{O(1)}$. However, this proof is non-constructive, meaning that this proof does not yield an explicit bound on the kernel size. The ${K4}$-Minor Cover problem is the simplest variant of the $\mathcal{F}$-Minor Cover problem with an unknown kernel size. To develop a constructive kernelization algorithm, we present a new method to decompose graphs into near-protrusions, such that near-protrusions in this new decomposition can be reduced using elementary reduction rules. Our method extends the `approximation and tidying' framework by van Bevern et al. [Algorithmica 2012] to provide guarantees stronger than those provided by both this framework and a regular protrusion decomposition. Furthermore, we provide extensions of the elementary reduction rules used by the ${K4, K_{2,3}}$-Minor Cover kernelization algorithm introduced by Donkers et al. [IPEC 2021]. Using the new decomposition method and reduction rules, we obtain a kernel consisting of $O(t^{41})$ vertices, which is the first constructive kernel. This kernel is a step towards more concrete kernelization bounds for the $\mathcal{F}$-Minor Cover problem where $\mathcal{F}$ contains a planar graph, and our decomposition provides a potential direction to achieve these new bounds.