,
Ashutosh Rai
,
Navneet Raj
Creative Commons Attribution 4.0 International license
Matching Cut can be seen as a blend of two fundamental concepts in graph theory, namely, a matching and a cut. In the Matching Cut problem, given a graph G, the objective is to determine whether there exists a cut (A, B) such that the set of edges crossing the cut forms a matching. While the classical problems of finding a maximum sized matching and a cut with minimum number of crossing edges are solvable in polynomial time, Matching Cut is known to be NP-complete. A natural generalization of Matching Cut is the d-Cut problem, where for a given integer d ≥ 1 and a graph G, we ask if there exists a cut (A, B) in which each vertex has at most d neighbors on the opposite side.
In this paper, we improve upon the fastest known FPT algorithm for the d-Cut problem parameterized by treewidth, previously proposed by Gomes and Sau [Algorithmica, 2021], which runs in 𝒪^*((2(d+1)²)^{tw}) time. We begin by refining their approach with a tighter analysis of the join node case in the dynamic programming over tree decompositions. In particular, we obtain a better bound on the number of splittings of a characteristic vector associated with a subset of vertices in a bag. This leads to a running time of {𝒪}^*(((d+1)(d+2))^{tw}). This is further improved to 𝒪^*((2(d+1))^{tw}) by employing the technique of fast subset convolution.
Further, we present several single-exponential time algorithms parameterized by the size s of a given modulator (such as vertex cover, distance to disjoint paths, cluster vertex deletion, and twin cover) together with the maximum number of crossing edges k in the cut. The running time of these algorithms is of the form 𝒪^*(2^s c_d^k), where c_d is a constant depending on d and the specific modulator. For the vertex cover modulator, we improve this to 𝒪^*((h_d)^s), where h_d = 2(c_d)^d < 2(d+1) for all d ≥ 2 and grows sublinearly in d. Furthermore, while the only known algorithm parameterized by the number of crossing edges k alone runs in time 2^𝒪(k log k), our algorithms are single-exponential in k, at the cost of an additional factor of 2^s.
@InProceedings{bansal_et_al:LIPIcs.IPEC.2026.13,
author = {Bansal, Ruchir and Jacob, Dalu and Mandal, Soumen and Rai, Ashutosh and Raj, Navneet},
title = {{Fast Parameterized Algorithms for Matching Cut and d-Cut}},
booktitle = {21st International Symposium on Parameterized and Exact Computation (IPEC 2026)},
pages = {13:1--13:17},
series = {Leibniz International Proceedings in Informatics (LIPIcs)},
ISBN = {978-3-95977-456-7},
ISSN = {1868-8969},
year = {2026},
volume = {398},
editor = {Korhonen, Tuukka and Lampis, Michael},
publisher = {Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
address = {Dagstuhl, Germany},
URL = {https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.IPEC.2026.13},
URN = {urn:nbn:de:0030-drops-281860},
doi = {10.4230/LIPIcs.IPEC.2026.13},
annote = {Keywords: Matching-Cut, d-Cut, Treewidth, Structural Parameterization, Vertex Cover}
}