Solving Problems on Graphs of High RankWidth
 607 Downloads
Abstract
A modulator in a graph is a vertex set whose deletion places the considered graph into some specified graph class. The cardinality of a modulator to various graph classes has long been used as a structural parameter which can be exploited to obtain fixedparameter algorithms for a range of hard problems. Here we investigate what happens when a graph contains a modulator which is large but “wellstructured” (in the sense of having bounded rankwidth). Can such modulators still be exploited to obtain efficient algorithms? And is it even possible to find such modulators efficiently? We first show that the parameters derived from such wellstructured modulators are more powerful for fixedparameter algorithms than the cardinality of modulators and rankwidth itself. Then, we develop a fixedparameter algorithm for finding such wellstructured modulators to every graph class which can be characterized by a finite set of forbidden induced subgraphs. We proceed by showing how wellstructured modulators can be used to obtain efficient parameterized algorithms for Minimum Vertex Cover and Maximum Clique. Finally, we use the concept of wellstructured modulators to develop an algorithmic metatheorem for deciding problems expressible in monadic second order logic, and prove that this result is tight in the sense that it cannot be generalized to LinEMSO problems.
Keywords
Fixedparameter algorithms Rankwidth Monadic secondorder logic Parameterized complexity1 Introduction
Many important graph problems are known to be NPhard, and yet admit efficient solutions in practice due to the inherent structure of instances. The parameterized complexity [16, 34] paradigm allows a more refined analysis of the complexity of various problems and hence enables the design of more efficient algorithms. In particular, given an instance of size n and a numerical parameter k which captures some property of the instance, one asks whether the instance can be solved in time \(f(k)\cdot n^{\mathcal {O}(1)}\). Parameterized problems which admit such an algorithm are called fixedparameter tractable (FPT), and the algorithms themselves are often called fixedparameter algorithms.
 A:

Width measures Treewidth has become an extremely successful structural parameter with a wide range of applications in many fields of computer science. However, treewidth is not suitable for use in dense graphs. This led to the development of algorithms that use the parameter cliquewidth [11], which can be viewed as a relaxation of treewidth towards dense graphs. However, while there are efficient theoretical algorithms for computing tree decompositions, this is not the case for decompositions for cliquewidth. This shortcoming has later been overcome by the notion of rankwidth [35], which improves upon cliquewidth by allowing the efficient computation of rankdecompositions while retaining all of the positive algorithmic results previously obtained for cliquewidth.
 B:

Modulators A modulator is a vertex set whose deletion places the considered graph into some specified graph class. A substantial amount of research has been placed into finding as well as exploiting small modulators to various graph classes [3, 18]. Popular notions such as vertex cover and feedback vertex set are also special cases of modulators (to the classes of edgeless graphs and forests, respectively). One advantage of parameterizing by the size of modulators is that it allows us to build on the vast array of research of polynomialtime algorithms on specific graph classes (see, for instance, [10, 33]). In other fields of computer science, modulators are often called backdoors and have been successfully used to obtain efficient algorithms for, e.g., satisfiability and constraint satisfaction [21, 22].
 1.
We introduce a family of “hybrid” parameters that combine approaches A and B.
 2.
We develop a fixedparameter algorithm for computing \(wsn^\mathcal {H}\).
 3.
We design fixedparameter algorithms for Minimum Vertex Cover (MinVC) and Maximum Clique (MaxClq) parameterized by \(wsn^\mathcal {H}\).
 4.
We develop a metatheorem to obtain fixedparameter algorithms for problems definable in monadic second order (MSO) logic [11] parameterized by \(wsn^\mathcal {H}\).
2 Preliminaries
The set of natural numbers (that is, positive integers) will be denoted by \(\mathbb {N}\). For \(i \in \mathbb {N}\) we write [i] to denote the set \(\{1, \ldots , i \}\). If \(\sim \) is an equivalence relation over a set A, then for \(a\in A\) we use \([a]_{\sim }\) to denote the equivalence class containing a.
2.1 Graphs
We will use standard graph theoretic terminology and notation (cf. [15]). All graphs we consider are finite, simple and undirected. The nonleaf vertices of a tree are called its internal nodes. If S is a set of leaves of a tree T, then T(S) denotes the smallest connected subtree spanning S.
Given a graph \(G=(V(G),E(G))\) and \(A\subseteq V(G)\), we denote by N(A) the set of neighbors of A in \(V(G){\setminus } A\); if A contains a single vertex v, we use N(v) instead of \(N(\{v\})\). We use V and E as shorthand for V(G) and E(G), respectively, when the graph is clear from context. Two vertex sets A, B are overlapping if \(A\cap B, A{\setminus } B\), and \(B{\setminus } A\) are all nonempty. \(GA\) denotes the subgraph of G obtained by deleting A.
Given a graph \(G=(V,E)\) and a graph class \(\mathcal {H}\), a set \(X\subseteq V\) is called a modulator to \(\mathcal {H}\) if \(GX\in \mathcal {H}\). A graph class is called hereditary if it is closed under vertex deletion. A graph H is an induced subgraph of G if H can be obtained by deleting vertices (along with all of their incident edges) from G. For \(A\subseteq V(G)\) we use G[A] to denote the subgraph of G obtained by deleting \(V(G){\setminus } A\). Let \(\mathcal {F}\) be a finite set of graphs; then the class of \(\mathcal {F}\)free graphs is the class of all graphs which do not contain any graph in \(\mathcal {F}\) as an induced subgraph. We will often refer to elements of \(\mathcal {F}\) as obstructions, and we say that the class of \(\mathcal {F}\)free graphs is characterized by \(\mathcal {F}\).
2.2 FixedParameter Tractability
We refer the reader to the standard textbooks [14, 16, 34] for an introduction to parameterized complexity. A parameterized problem \(\mathcal {P}\) is a subset of \(\varSigma ^* \times \mathbb {N}\) for some finite alphabet \(\varSigma \). For a problem instance \((x,k) \in \varSigma ^* \times \mathbb {N}\) we call x the main part and k the parameter.
A parameterized problem \(\mathcal {P}\) is fixedparameter tractable (FPT in short) if a given instance (x, k) can be solved in time \(f(k) \cdot x^{\mathcal {O}(1)}\) where f is an arbitrary computable function of k. Algorithms with running time in this form are called fixedparameter algorithms, and we also slightly abuse notaton and refer to this runtime simply as FPT time.
A parameterized problem \(\mathcal {P}\) is paraNPhard if the unparameterized problem corresponding to the restriction of \(\mathcal {P}\) to some constant value of the parameter is NPhard. For instance, the classical Minimum Vertex Cover problem parameterized by the degree of a graph is paraNPhard, since Minimum Vertex Cover remains NPhard even on graphs of degree at most 3.
2.3 Splits and Graph Labeled Trees
The notions and terminology introduced in this subsection play an important role in particular when formally defining our parameters (Sect. 3) and for computing them (Sect. 3). The contents of this subsection are based on the original work of Cunningham [13] as well as on more recent results by several authors [25, 26, 27].
A split of a connected graph \(G=(V,E)\) is a vertex bipartition \(\{A,B\}\) of V such that every vertex of \(A' = N(B)\) has the same neighborhood in \(B'=N(A)\). The sets \(A'\) and \(B'\) are called frontiers of the split.
A split is said to be nontrivial if both sides have at least two vertices. A connected graph which does not contain a nontrivial split is called prime. A bipartition is trivial if one of its parts is the empty set or a singleton. Cliques and stars are called degenerate graphs; notice that every nontrivial bipartition of their vertices is a split.
Let \(G=(V,E)\) be a graph. To simplify our exposition, we will use the notion of splitmodules instead of splits where suitable. A set \(A\subseteq V\) is called a splitmodule of G if there exists a connected component \(G'=(V',E')\) of G such that \(A\subseteq V'\) and \(\{A,V'{\setminus } A\}\) forms a split of \(G'\). Notice that if A is a splitmodule then A can be partitioned into \(A_1\) and \(A_2\) such that \(N(A_2)\subseteq A\) and for each \(v_1, v_2\in A_1\) it holds that \(N(v_1)\cap (V'{\setminus } A)=N(v_2)\cap (V'{\setminus } A)\). For technical reasons, V and \(\emptyset \) are also considered splitmodules. We say that two disjoint splitmodules \(X, Y \subseteq V\) are adjacent if there exist \(x \in X\) and \(y \in Y\) such that x and y are adjacent.
Our algorithms for computing wellstructured modulators will rely on two deep results related to splits, specifically Theorems 1 and 2. To formally state these theorems, we will need to introduce a bit of extra notation.
A graphlabeled tree (which can be viewed as a more modern approach to the previously used splitdecompositions) is a pair \((T, \mathcal {F})\), where T is a tree and \(\mathcal {F}\) is a set of graphs such that each internal node u of T is labeled by a graph \(G(u)\in \mathcal {F}\) and there is a bijection between the edges of T incident to u and vertices of G(u). When clear from the context, we may use u as a shorthand for \(G(u) \in \mathcal {F}\); for instance, we use V(u) to denote V(G(u)) and we say that an edge of T incident to u is incident to the vertex of G(u) mapped to it. Graphlabeled trees were introduced by Gioan and Paul [25, 26], and in the following paragraphs we recall some useful definitions and facts that also appeared in followup work [27].
For an internal node u of T, the vertices of V(u) are called marker vertices and the edges of E(u) are called labeledges. Edges of T incident to two internal nodes are called treeedges. Marker vertices incident to a treeedge e are called the extremities of e, and each leaf v is associated with the unique marker vertex q (in the neighbor of v) mapped to the edge incident to v. Perhaps the most important notion for graphlabeled trees with respect to split decomposition is that of accessibility.
Definition 1
 1.
Every two consecutive elements of \(\varPi \) are either the vertices of a labeledge or the extremities of a treeedge;
 2.
the sequence of edges obtained above alternates between treeedges and labeledges.
Definition 2
([27]) Let e be a treeedge incident to internal nodes u and \(u'\) in a graphlabeled tree, and let \(q\in V (u)\) and \(q' \in V (u')\) be the extremities of e. The nodejoin of \(u, u'\) replaces u and \(u'\) with a new internal node v labeled by the graph formed from the disjoint union of G(u) and \(G(u')\) as follows: all possible labeledges are added between N(q) and \(N (q' )\), and then q and \(q'\) are deleted. The new node v is made adjacent to all neighbors of u and \(u'\) in T. The nodesplit is then the inverse of the nodejoin.
Notice that the nodejoin operation and the nodesplit operation preserve the accessibility graph of the graphlabeled tree. A graphlabeled tree is reduced if all its labels are either prime or degenerate, and no nodejoin of two cliques or two stars \(S_1\) and \(S_2\) where the center of \(S_1\) is adjacent to a leaf of \(S_2\) is not possible. It is known that for every connected graph G, there exists a unique reduced graphlabeled tree \((T , \mathcal {F})\) such that \(G = Gr(T , \mathcal {F})\) [13, 25, 26, 27]; this unique reduced graphlabeled tree is called the splittree and is denoted \(\text {ST}(G)\).
2.4 RankWidth
Rankwidth was introduced by Oum and Seymour [35] and is closely related to cliquewidth. To define it, we first need to introduce the bipartite adjacency matrix \(\varvec{A}_G[U,W]\).
For a graph G and \(U,W\subseteq V(G)\), let \(\varvec{A}_G[U,W]\) denote the \(U\times W\)submatrix of the adjacency matrix over the twoelement field \(\mathrm {GF}(2)\), i.e., the entry \(a_{u,w}, u\in U\) and \(w\in W\), of \(\varvec{A}_G[U,W]\) is 1 if and only if \(\{u,w\}\) is an edge of G. The cutrank function \(\rho _G\) of a graph G is defined as follows: For a bipartition (U, W) of the vertex set \(V(G), \rho _G(U)=\rho _G(W)\) equals the rank of \(\varvec{A}_G[U,W]\) over \(\mathrm {GF}(2)\). We note that \(\rho _G\) is a symmetric function, and observe that a splitmodule X can be seen as a subgraph such that \(\varvec{A}_G[X,V(G){\setminus } X]=1\).
A rankdecomposition of a graph G is a pair \((T,\mu )\) where T is a tree of maximum degree 3 and \(\mu :V(G)\rightarrow \{t: \hbox {} t \hbox {is a leaf of} T\}\) is a bijective function. For an edge e of T, the connected components of \(T  e\) induce a bipartition (X, Y) of the set of leaves of T. The width of an edge e of a rankdecomposition \((T,\mu )\) is \(\rho _G(\mu ^{1} (X))\). The width of \((T,\mu )\) is the maximum width over all edges of T. The rankwidth of G, rw(G) in short, is the minimum width over all rankdecompositions of G. A graph class \(\mathcal {H}\) is of unbounded rankwidth if for each \(i\in \mathbb {N}\) there exists a graph \(G\in \mathcal {H}\) such that \(rw(G)> i\).
Theorem 3
([29]) Let \(k \in \mathbb {N}\) and \(n \ge 2\). For an nvertex graph G, we can output a rankdecomposition of width at most k or confirm that the rankwidth of G is larger than k in time \(f(k)\cdot n^3\), where f is a computable function.
2.5 Monadic Second Order Logic on Graphs
Here we introduce monadic second order logic, which will play a crucial role in our positive (Sect. 6) as well as negative (Sect. 7) algorithmic results.
We assume that we have an infinite supply of individual variables, denoted by lowercase letters x, y, z, and an infinite supply of set variables, denoted by uppercase letters X, Y, Z. Formulas of monadic secondorder logic (MSO) are constructed from atomic formulas E(x, y), X(x), and \(x = y\) using the connectives \(\lnot \) (negation), \(\wedge \) (conjunction) and existential quantification \(\exists x\) over individual variables as well as existential quantification \(\exists X\) over set variables. Individual variables range over vertices, and set variables range over sets of vertices. The atomic formula E(x, y) expresses adjacency, \(x = y\) expresses equality, and X(x) expresses that the vertex x is in the set X. From this, we define the semantics of monadic secondorder logic in the standard way (this logic is sometimes called \({\text {MSO}} _1\)).
Free and bound variables of a formula are defined in the usual way. A sentence is a formula without free variables. We write \(\varphi (X_1, \ldots , X_n)\) to indicate that the set of free variables of formula \(\varphi \) is \(\{X_1, \ldots , X_n\}\). If \(G = (V,E)\) is a graph and \(S_1, \ldots , S_n \subseteq V\) we write \(G \models \varphi (S_1, \ldots , S_n)\) to denote that \(\varphi \) holds in G if the variables \(X_i\) are interpreted by the sets \(S_i\), for \(i \in [n]\). For a fixed MSO sentence \(\varphi \), the MSO model checking problem (\(\textsc {MSOMC}_{\varphi }\)) asks whether an input graph G satisfies \(G\models \varphi \).
It is known that MSO formulas can be checked in uniformly polynomial time on graphs of bounded rankwidth.
Theorem 4
([19]) Let \(\varphi \) and \(\psi =\psi (X)\) be fixed MSO formulas. There exists a computable function f and an algorithm such that, given an nvertex graph G and \(S\subseteq V(G)\), decides whether \(G \models \varphi \) and whether \(G\models \psi (S)\) in time \(f(rw(G))\cdot n^3\).
We review MSOtypes roughly following the presentation in the textbook by Libkin [32]. The quantifier rank of an MSO formula \(\varphi \) is defined as the nesting depth of quantifiers in \(\varphi \). For nonnegative integers q and l, let \({\text {MSO}} _{q,l}\) consist of all MSO formulas of quantifier rank at most q having at most l free set variables.
Let \(\varphi = \varphi (X_1,\ldots ,X_l)\) and \(\psi = \psi (X_1,\ldots ,X_l)\) be MSO formulas. We say \(\varphi \) and \(\psi \) are equivalent, written \(\varphi \equiv \psi \), if for all graphs G and \(U_1, \ldots , U_l \subseteq V(G), G \models \varphi (U_1,\ldots , U_l)\) if and only if \(G \models \psi (U_1,\ldots , U_l)\). Given a set F of formulas, let \({F/\mathord \equiv }\) denote the set of equivalence classes of F with respect to \(\equiv \). A system of representatives of \(F/\mathord \equiv \) is a set \(R \subseteq F\) such that \(R \cap C \ne \emptyset \) for each equivalence class \(C \in F/\mathord \equiv \). The following statement has a straightforward proof using normal forms (see [32, Proposition 7.5 and Lemma 3.13] for details).
Fact 1
Let q and l be fixed nonnegative integers. The set \({\text {MSO}} _{q,l}/\mathord \equiv \) is finite, and one can compute a finite system of representatives of \({\text {MSO}} _{q,l}/\mathord \equiv \).
Note that the system of representatives obtained in this way need not be inclusionminimal, and we do not assume to have a mapping from this system of representatives to elements of \({{\text {MSO}} _{q,l}/\mathord \equiv }\). We will assume that for every pair of nonnegative integers q and l the system of representatives of \({\text {MSO}} _{q,l}/\mathord \equiv \) given by Fact 1 is fixed.
Definition 3
(MSO Type) Let q, l be nonnegative integers. For a graph G and an ltuple \(\mathbf {U}\) of sets of vertices of G, we define MSO\( type _q(G,\mathbf {U})\) as the set of formulas \(\varphi \in {\text {MSO}} _{q,l}\) such that \(G \models \varphi (\mathbf {U})\). We call MSO\( type _q(G,\mathbf {U})\) the MSO qtype of \(\mathbf {U}\) in G.
Since we will only be dealing with MSO logic, throughout the paper we will refer to MSOtypes simply as types. It follows from Fact 1 that up to logical equivalence, every type contains only finitely many formulas. This allows us to represent types using MSO formulas, as formalized in the next lemma. We remark that the statement of the next lemma used in previous work [20] did not specify the (“fixedparameter”) dependence of the running time on the rankwidth, and so here we give a proof of the lemma for completeness.
Lemma 1
([20]) Let q and l be nonnegative integer constants. Let G be a graph, and let \(\mathbf {U}\) be an ltuple of sets of vertices of G. One can compute a formula \(\varPhi \in {\text {MSO}} _{q,l}\) such that for any graph \(G'\) and any ltuple \(\mathbf {U}'\) of sets of vertices of \(G'\) we have \(G' \models \varPhi (\mathbf {U}')\) if and only if \( type _q(G,\mathbf {U}) = type _q(G',\mathbf {U}')\). Moreover, \(\varPhi \) can be computed in time \(f(rw(G))\cdot V^{3}\).
Proof
Let R be a system of representatives of \({\text {MSO}} _{q,l}/\mathord \equiv \) given by Fact 1. Because q and l are constant, we can consider both the cardinality of R and the time required to compute it as constants. Let \(\varPhi \in {\text {MSO}} _{q,l}\) be the formula defined as \(\varPhi = \bigwedge _{\varphi \in S} \varphi \wedge \bigwedge _{\varphi \in R {\setminus } S} \lnot \varphi \), where \(S = \{\,\varphi \in R \;{}\;G \models \varphi (\mathbf {U}) \,\}\). We can compute \(\varPhi \) by deciding \(G \models \varphi (\mathbf {U})\) for each \(\varphi \in R\). Since the number of formulas in R is a constant, this can be done in time \(f(rw(G))\cdot V^{3}\) (for some computable function f) as checking whether G satisfied \(\varphi (\mathbf {U})\) can be done in time \(q(rw(G))\cdot V^{3}\) (for some computable function q).
Let \(G'\) be an arbitrary graph and let \(\mathbf {U}'\) be an ltuple of subsets of \(V(G')\). We claim that \( type _q(G, \mathbf {U}) = type _q(G', \mathbf {U'})\) if and only if \(G' \models \varPhi (\mathbf {U}')\). Since \(\varPhi \in {\text {MSO}} _{q,l}\) the forward direction is trivial. For the converse, assume \( type _q(G, \mathbf {U}) \ne type _q(G', \mathbf {U'})\). First suppose \(\varphi \in type _q(G, \mathbf {U}) {\setminus } type _q(G', \mathbf {U'})\). The set R is a system of representatives of \({\text {MSO}} _{q,l}/\mathord \equiv \) , so there has to be a \(\psi \in R\) such that \(\psi \equiv \varphi \). But \(G' \models \varPhi (\mathbf {U}')\) implies \(G' \models \psi (\mathbf {U}')\) by construction of \(\varPhi \) and thus \(G' \models \varphi (\mathbf {U}')\), a contradiction. Now suppose \(\varphi \in type _q(G', \mathbf {U}') {\setminus } type _q(G, \mathbf {U})\). An analogous argument proves that there has to be a \(\psi \in R\) such that \(\psi \equiv \varphi \) and \(G' \models \lnot \psi (\mathbf {U}')\). It follows that \(G' \not \models \varphi (\mathbf {U}')\), which again yields a contradiction. \(\square \)
The remainder of the section introduces the classical notion of MSO games (Definition 5) and their relation to MSO types (Theorem 5). However, to formally define MSO games, we first need the notion of partial isomorphism.
Definition 4
 For every \(i,j \in [m]\),$$\begin{aligned} v_i = v_j \, \Leftrightarrow \, u_i = u_j \text { and } v_iv_j \in E(G)\, \Leftrightarrow \, u_iu_j \in E(G'). \end{aligned}$$
 For every \(i \in [m]\) and \(j \in [l]\),$$\begin{aligned} v_i \in V_j \, \Leftrightarrow u_i \in U_j. \end{aligned}$$
In the definition of MSO games given below, we denote the concatenation of tuple \(\mathbf {A}\) by tuple \(\mathbf {B}\) as \(\mathbf {A}\frown \mathbf {B}\).
Definition 5

Point move: The Spoiler picks a vertex in either G or \(G'\); the Duplicator responds by picking a vertex in the other graph.

Set move: The Spoiler picks a subset of V(G) or a subset of \(V(G')\); the Duplicator responds by picking a subset of the vertex set of the other graph.
Theorem 5
3 WellStructured Modulators
Recall that a modulator to a graph class \(\mathcal {H}\) is a vertexsubset of a graph G such that its deletion puts G into \(\mathcal {H}\) (see Sect. 2.1).
Definition 6
 1.
\(\mathbf {X}\le k\), and
 2.
\(\bigcup _{X_i\in \mathbf {X}} X_i\) is a modulator to \(\mathcal {H}\), and
 3.
\(rw(G[X_i])\le k\) for each \(X_i\in \mathbf {X}\).
An example of a 2wellstructured modulator is provided in Figure 3. For the sake of brevity and when it is clear from the context, we will sometimes identify \(\mathbf {X}\) with \(\bigcup _{X_i\in \mathbf {X}} X_i\) (for instance \(G\mathbf {X}\) is shorthand for \(G\bigcup _{X_i\in \mathbf {X}} X_i\)). To allow a concise description of our parameters, for any hereditary graph class \(\mathcal {H}\) we let the wellstructure number (\(wsn^{\mathcal {H}}\) in short) denote the minimum k such that G has a kwellstructured modulator to \(\mathcal {H}\). Similarly, we let \(mod^{\mathcal {H}}(G)\) denote the minimum k such that G has a modulator of cardinality k to \(\mathcal {H}\).
Proposition 1
 1.
\(rw(G)\ge wsn^\mathcal {H}(G)\) for every graph G. Furthermore, for every \(i\in \mathbb {N}\) there exists a graph \(G_i\) such that \(rw(G_i)\ge wsn^\mathcal {H}(G_i)+i\), and
 2.
\(mod^\mathcal {H}(G)\ge wsn^\mathcal {H}(G)\) for every graph G. Furthermore, for every \(i\in \mathbb {N}\) there exists a graph \(G_i\) such that \(mod^\mathcal {H}(G_i)\ge wsn^\mathcal {H}(G_i)+i\).
Proof
 1.
For \(rw(G)\ge wsn^\mathcal {H}(G)\) notice that for every graph G of rankwidth k, the set \(\{V(G)\}\) is a kwellstructured modulator to the empty graph. For the second claim, since \(\mathcal {H}\) has unbounded rankwidth, for every \(i\in \mathbb {N}\) it contains some graph \(G_i\) such that \(rw(G_i)>i\); by definition, \(wsn^\mathcal {H}(G_i)=0\).
 2.
For \(mod^\mathcal {H}(G)\ge wsn^\mathcal {H}(G)\), let G be a graph containing a modulator \(X=\{v_1,\ldots ,\) \(v_k\}\) to \(\mathcal {H}\). It is easy to check that \(\mathbf {X}=\{\{v_1\},\ldots ,\{v_k\}\}\) is a kwellstructured modulator to \(\mathcal {H}\). For the second claim, let \(G'\not \in \mathcal {H}\) and let \(k=rw(G')\). Consider the graph \(G_i\) consisting of \(i+1+k\) disjoint copies of \(G'\) and a vertex q which is adjacent to every other vertex of G. Since \(\mathcal {H}\) is hereditary, we may assume without loss of generality that it contains the singlevertex graph. It is then easy to check that \(\{V(G){\setminus } \{q\}\}\) forms a kwellstructured modulator in G to \(\mathcal {H}\). Now consider an arbitrary set \(X\subseteq V(G)\) of cardinality at most \(i+k\). Clearly, there must exist some copy of \(G'\), say \(G'_j\), such that \(X\cap V(G'_j)=\emptyset \). Since \(G'_j\not \in \mathcal {H}\), it follows from the hereditarity of \(\mathcal {H}\) that \(GX\not \in \mathcal {H}\) and hence X cannot be a modulator to \(\mathcal {H}\). We conclude \(mod^\mathcal {H}(G_i)>i+k=i+wsn^\mathcal {H}(G_i)\).
4 Finding WellStructured Modulators
The objective of this subsection is to prove the following theorem.
Theorem 6
Let \(\mathcal {H}\) be a graph class characterized by a finite obstruction set. There exists a fixedparameter algorithm parameterized by k which for every input graph G either finds a kwellstructured modulator to \(\mathcal {H}\), or detects that no such kwellstructured modulator exists.
Interestingly, the techniques we will use to prove Theorem 6 only work if the rankwidth of the graph is sufficiently large. This is not a problem though, since on graphs of small rankwidth we can always directly use rankwidth to find kwellstructured modulators.
Our first course of action is the statement of several useful properties of splits in graphs. We remark that for most of this section we will restrict ourselves to connected graphs, and show how to deal with general graphs later on; this allows us to use the following result by Cunningham.
Theorem 7
([13]) Let \(\{ A, C\}, \{ B, D\}\) be splits of a connected graph G such that \(A\cap B\ge 2\) and \(A \cup B\ne V(G)\). Then \(\{A\cap B, C\cup D\}\) is a split of G.
Lemma 2
If A and B are overlapping splitmoduleof a connected graph \(G=(V,E)\), then \(A\cup B\) is also a splitmodule. Moreover, if \(A\cup B\ne V\), then also \(A\cap B\) is a splitmodule.
Proof

if \(A\cap B=1\) and \(C\cap D=1\), then both \(A\cap B\) and \(A\cup B=V{\setminus } (C\cap D)\) are easily seen to be splitmodules;

if \(A\cap B\ge 2\) and \(C\cap D=1\), then \(A\cap B\) is a splitmodule by Theorem 7 and \(A\cup B\) is also a splitmodule because \(C\cap D\) is a splitmodule;

if \(A\cap B=1\) and \(C\cap D\ge 2\), then \(A\cap B\) is a splitmodule and \(A\cup B\) is also a splitmodule because C, D satisfy the conditions of Theorem 7 and hence \(C\cap D = V{\setminus } (A\cup B)\) forms a splitmodule;

if \(A\cap B\ge 2\) and \(C\cap D\ge 2\), then \(A\cap B\) is a splitmodule by Theorem 7 and \(A\cup B\) is also a splitmodule because C, D satisfy the conditions of Theorem 7, as in the previous case.
Lemma 3
Let \(G=(V,E)\) be a connected graph and A, B be overlapping splitmodules. Then \(A{\setminus } B\) is also a splitmodule.
Proof
The lemma clearly holds if \(A{\setminus } B\le 1\), so we may assume that \(A{\setminus } B\ge 2\). Let \(Z=V{\setminus } B\); since B is a split module, so is Z. Furthermore, since A and B are overlapping, it holds that \(B{\setminus } A\) is nonempty and hence \(V\ne Z\cup A\). Since \(Z\cap A=A{\setminus } B\), we have \(Z\cap A\ge 2\) and hence we conclude that \(Z\cap A=A{\setminus } B\) is a split module by Theorem 7. \(\square \)
Lemma 4
Let \(k\in \mathbb {N}, G=(V,E)\) be a graph, and A, B, C be pairwise disjoint splitmodules such that \(A\cup B\cup C=V\). Let a, b, c be arbitrary vertices such that \(a\in N(A), b\in N(B)\), and \(c\in N(C)\). If \(\max \big (rw(G[A\cup \{a\}]),rw(G[B\cup \{b\}]),rw(G[C\cup \{c\}])\big )\le k\), then \(rw(G)\le k\).
Proof
Let \(\mathcal {T}_{A} = (T_{A}, \mu _{A}), \mathcal {T}_{B} = (T_{B}, \mu _{B})\), and \(\mathcal {T}_{C} = (T_{C}, \mu _{C})\) be witnessing rank decompositions of G[A], G[B], and G[C], respectively.
We construct a rank decomposition \(\mathcal {T} = (T, \mu )\) of G as follows.
Let \(l_{a}\) be the leaf (note that \(\mu _{A}\) is bijective) of \(T_{A}\) such that \(\mu _{A}(a) = l_{a}\). Similarly, let \(l_b\) and \(l_c\) be the leaves such that \(\mu _B(b)=l_b\) and \(\mu _C(c)=l_c\), respectively. We obtain T from \(T_{A}\) by adding disjoint copies of \(T_{B}\) and \(T_{C}\) and then identifying \(l_{a}\) with the copies of \(l_{b}\) and \(l_{b}\). Since \(T_{A}, T_{B}\), and \(T_{C}\) are subcubic, so is T.
 1.
\(e' \in T_{A}\). Let \((U, W) = f(\mathcal {T}_{A}, e')\). Without loss of generality assume that \(a \in W\). Then by construction of \(\mathcal {T}\) , we have \(f(\mathcal {T}, e) = (U, W \cup B\cup C)\). Let \(u\in A\) and \(v\in B\cup C\). Since A is splitmodule either \(v\notin N(A)\) and \(\mathbf {A}_G(u, v) =0\) for all \(u\in A\), or \(v\in N(A)\) in which case \(\mathbf {A}_G(u, v) = \mathbf {A}_G(u, a)\) for all \(u\in A\). Therefore, to obtain \(\mathbf {A}_G(U, W\cup B\cup C) \) one can simply copy the column corresponding to a in \(\mathbf {A}_G(U, W)\) or add some empty columns. This does not increase the rank of the matrix.
 2.
\(e' \in T_{B}\). This case is symmetric to case 1, with A and B switching their roles and b taking the role of a.
 3.
\(e' \in T_{C}\). This case is symmetric to case 1, with A and C switching their roles and c taking the role of a.
By repeating the proof technique of Lemma 4 without the set C, we obtain the following corollary.
Corollary 1
Let \(k\in \mathbb {N}, G=(V,E)\) be a graph, and A, B pairwise disjoint splitmodules such that \(A\cup B=V\). Let \(a,b\in V\) be such that \(a\in N(A)\) and \(b\in N(B)\). If \(\max \big (rw(G[A\cup \{a\}]),rw(G[B\cup \{b\}])\big )\le k\), then \(rw(G)\le k\).
Lemma 5
Let \(k \in \mathbb {N}\) and let \(G=(V,E)\) be a connected graph having splitmodules \(M_1, M_2\) where \(M_1 \cup M_2 = V\) and \(\max (rw(G[M_1]), rw(G[M_2])) \le k\). Then \(rw(G) \le k+1\).
Proof
Let \(M_{22} = M_2{\setminus } M_1\). Clearly, \(\{M_1, M_{22}\}\) is a split. Since rankwidth is preserved by taking induced subgraphs, the graph \(G[M_{22}]\) has rankwidth at most k. Let \(v_1\in N(M_{22})\) and \(v_2\in N(M_1)\). It is easy to see that graphs \(G_{1} = G[M_{1} \cup \{v_{2}\}]\) and \(G_{2} = G[M_{22} \cup \{v_{1}\}]\) have rankwidth at most \(k+1\). We finish the proof by applying Corollary 1, using \(M_1, M_{22}\) in roles of A, B and \(v_1, v_2\) in roles of a, b, respectively. \(\square \)
The following lemma in essence shows that the relation of being in a splitmodule of small rankwidth is transitive (assuming sufficiently high rankwidth). The significance of this will become clear later on.
Lemma 6
Let \(k \in \mathbb {N}\). Let \(G=(V,E)\) be a connected graph of rankwidth at least \(k+2\) and let \(M_1, M_2\) be splitmodules of G such that \(M_1 \cap M_2 \ne \emptyset \) and \(\max (rw(G[M_1]), rw(G[M_2])) \le k\). Then \(M_1 \cup M_2\) is a splitmodule of G and \(rw(G[M_1\cup M_2]) \le k\).
Proof
If \(M_1 \subseteq M_2\) or \(M_2 \subseteq M_1\) the result is immediate, hence we may assume that they are overlapping. Lemma 5 and \(rw(G)\ge k+2\) together imply that \(M_1\cup M_2 \ne V\). Let \(M_{11} = M_1 {\setminus } M_2, M_{22} = M_2 {\setminus } M_1\), and \(M_{12} = M_1 \cap M_2\). It follows from Lemmas 2 and 3 that these sets are splitmodules of G. Let \(v_{11} \in N(V{\setminus } M_{11}), v_{22} \in N(V{\setminus } M_{22})\), and \(v_{12} \in N(V{\setminus } M_{12})\). We show that \(rw(G[M_1 \cup M_2]) \le k\). By assumption, both \(G[M_1]\) and \(G[M_2]\) have rankwidth at most k. Since rankwidth is preserved by taking induced subgraphs, the graphs \(G_{11} = G[M_{11} \cup \{v_{12}\}], G_{12} = G[M_{12} \cup \{v_{22}\}]\), and \(G_{22} = G[M_{22} \cup \{v_{12}\}]\) also have rankwidth at most k. We finish the proof by applying Lemma 4, where \(M_{11}, M_{22}, M_{12}\) take the roles of A, B, and C and \(v_{12}, v_{12}\), and \(v_{22}\) take the roles of a, b, and c, respectively. \(\square \)
Definition 7
Let G be a graph and \(k \in \mathbb {N}\). We define a relation \(\sim ^G_k\) on V(G) by letting \(v \sim ^G_k w\) if and only if there is a splitmodule M of G such that \(v,w \in M\) and \(rw(G[M]) \le k\). We drop the superscript from \(\sim ^G_k\) if the graph G is clear from context.
Using Lemma 6 to deal with transitivity, we prove the following.
Proposition 2
For every \(k \in \mathbb {N}\) and graph \(G=(V,E)\) of rankwidth at least \(k+2\), the relation \(\sim _k\) is an equivalence relation, and each equivalence class U of \(\sim _k\) is a splitmodule of G such that \(rw(G[U]) \le k\).
Proof
Let G be a graph and \(k \in \mathbb {N}\). For every \(v \in V\), the singleton \(\{v\}\) is a splitmodule of G, so \(\sim _k\) is reflexive. Symmetry of \(\sim _k\) is trivial. For transitivity, let \(u,v,w \in V\) be such that \(u \sim _k v\) and \(v \sim _k w\). Then there are splitmodules \(M_1, M_2\) of G such that \(u,v \in M_1, v, w \in M_2\), and \(rw(G[M_1]), rw(G[M_2]) \le k\); in particular, since \(rw(G)\ge k+2\) this implies that there exists a connected component \(G'\) of G containing u, v, w. By Lemma 6, \(M_1 \cup M_2\) is a splitmodule of \(G'\) (and hence also of G) such that \(rw(G[M_1 \cup M_2]) \le k\). In combination with \(u,w \in M_1 \cup M_2\) that implies \(u \sim _k w\). This concludes the proof that \(\sim _k\) is an equivalence relation.
Now let \(v \in V, G'\) be the connected component containing v, and let \(U = [v]_{\sim _k}\). For each \(u \in U\) there is a splitmodule \(W_u\) of \(G'\) (and of G) such that \(u,v \in W_u\) and \(rw(G[W_u]) \le k\). By Lemma 6, \(W = \bigcup _{u \in U} W_u\) is a splitmodule of \(G'\) (and hence also of G) and \(rw(G[W]) \le k\). Clearly, \([v]_{\sim _k} \subseteq W\). On the other hand, \(u \in W\) implies \(v \sim _k u\) by definition of \(\sim _k\), so \(W \subseteq [v]_{\sim _k}\). That is, \(W = [v]_{\sim _k}\). \(\square \)
Corollary 2
Every graph G of rankwidth at least \(k+2\) has its vertex set uniquely partitioned by the equivalence classes of \(\sim _k\) into inclusionmaximal splitmodules of rankwidth at most k.
Next, we state a simple but useful observation.
Observation 1
Let \(k\in \mathbb {N}, G\) be a disconnected graph having rankwidth at least \(k+2\), and \(\mathcal {C}(G)\) be the set of connected components of G. Then \(\sim ^G_k=\bigcup _{G' \in \mathcal {C}(G)}\sim ^{G'}_k\).
Now that we know \(\sim _k\) is an equivalence relation, we show how to compute it in FPT time. It will be useful to recall Theorems 1 and 2 from Sect. 2.3.
Proposition 3
Let \(k \in \mathbb {N}\). Given an nvertex graph G of rankwidth at least \(k+2\) and two vertices v, w, we can decide whether \(v \sim _k w\) in time \(f(k)\cdot n^3\) for some computable function f.
Proof
From Observation 1 it follows that if the proposition holds for connected graphs, then it holds for disconnected graphs as well; hence we may assume that G is connected. By Theorem 2 we can compute the unique splittree \(\text {ST}(G)=(T,\mathcal {F})\) in \(\mathcal {O}((m+n)\alpha (m+n))\) time. Due to Theorem 1, every split in G is the bipartition of leaves of T induced either by removing an internal treeedge of T or an edge created by a nodesplit of a degenerate vertex of T.
Vertices of G are leaves of T and we can find a path P between v and w in T in time linear in size of T. Since the number of nodes in a splitdecomposition is linear in the number of vertices of the original graph [23], there are at most linearly many vertices on the path. Moreover, if \(u\in T\) is a degenerate node on P, then we can split it into two nodes \(u_1, u_2\) in a way such that \(G(u_1)\) contains exactly the two marker vertices of G(u) incident to an edge of T on P and a new marker vertex connecting it to \(u_2\). We split all degenerate nodes on P in this way and denote the new tree by \(T'\). Note that now every degenerate node on a new path \(P'\) between u and v has 3 vertices.
Now every edge between \(P'\) and \(T'{\setminus } P'\) corresponds to a minimal splitmodule containing v and w. Conversely, as a consequence of Theorem 1 every minimal splitmodule containing v and w is induced by removing an edge between \(P'\) and \(T'{\setminus } P'\), and let \(M_{vw}\) be the set containing all of these at most T minimal split modules. Hence, \(v \sim _k w\) if and only if there is a splitmodule X in \(M_{vw}\) such that \(rw(G[X])\le k\). By Theorem 3 we can decide, for each such X, whether \(rw(G[X])\le k\) in time \(f(k)\cdot n^3\), where f is some computable function. \(\square \)
We will use \(\sim _k\) as the input for FindWSM \(_\mathcal {F}\), however considering general equivalence relations as inputs is useful for proving correctness. Recall that the equivalence relation \(\sim _k\) (or, more precisely, the set of its equivalence classes) can be computed in time \(n^2\cdot f(k)\cdot n^3\) for some function f thanks to Proposition 3, and this only needs to be done once before starting the algorithm. The following two lemmas show that Algorithm 1 is correct and runs in FPT time. For fixed \(\mathcal {F}\), let \(c_\mathcal {F}\) denote the maximum number of vertices of a graph in \(\mathcal {F}\).
Lemma 7
FindWSM \(_\mathcal {F}\) runs in time \(\mathcal {O}(c_\mathcal {F}^k\cdot n^{c_\mathcal {F}})\).
Proof
The time required to perform the steps on lines \(\mathbf {2}\)–\(\mathbf {6}\) is \(\mathcal {O}(n^{c_\mathcal {F}})\) since \(\mathcal {F}\) is finite. Similarly, it holds that \(V(D')\) and hence also the number of times the procedure on lines \(\mathbf {8}\)–\({\mathbf {1}}{\mathbf {3}}\) is called are bounded by \(c_\mathcal {F}\).
For the rest of the proof, we proceed by induction on k. First, if \(k=0\), then the algorithm is polynomial by the above. So assume that \(k\ge 1\) and the algorithm for \(k1\) runs in time at most \(c_\mathcal {F}^{k1}\cdot n^{c_\mathcal {F}}\). Then the algorithm for k will run in polynomial time up to lines \(\mathbf {8}\)–\(\mathbf {13}\), where it will make at most \(c_\mathcal {F}\) calls to the algorithm for \(k1\), which implies that the running time for k is bounded by \(\mathcal {O}(c_\mathcal {F}^{k}\cdot n^{c_\mathcal {F}})\). \(\square \)
Lemma 8
Let \(k\ge 0, G\) be a graph and \(\sim \) an equivalence relation over a superset of V. Then FindWSM \(_\mathcal {F}(k,G,\sim )\) outputs a set \(\mathbf {X}\) of at most k equivalence classes of \(\sim \) such that \(G\mathbf {X}\) is \(\mathcal {F}\)free.
Proof
If G does not contain any D as an induced subgraph, then we correctly return the empty set. So, assume there exists an induced subgraph \(D'\) of G isomorphic to D. We prove the lemma by induction on k.
Clearly, if \(k=0\) but there exists some obstruction, then the algorithm outputs False and this is correct; if \(k=0\) and no obstruction exists, then the algorithm correctly outputs \(\emptyset \). Let \(k\ge 1\) and assume that the algorithm is correct for \(k1\). If G does not contain any such \(\mathbf {X}\), then for any equivalence class \([a]_\sim \), FindWSM\(_\mathcal {F}({k1}, G[a]_{\sim }, \sim )\) will correctly output False.
On the other hand, assume G does contain some \(\mathbf {X}\) with the desired properties. In particular, this implies that \(\mathbf {X}\) must intersect \(V(D')\). Let \(X_i\) be an arbitrary equivalence class of \(\mathbf {X}\) which intersects \(V(D')\). Then \(\mathbf {X}'{\setminus } \{X_i\}\) is a set of at most \(k1\) equivalence classes of \(\sim \) in \(GX_i\), and hence FindWSM\(_\mathcal {F}(k1, GX'_i, \sim )\) will output some solution \(\mathbf {X}''\) for \(GX'_i\) by our inductive assumption. Since any obstruction in G intersecting \(X'_i\) is removed by \(X'_i\) and \(GX'_i\) is made \(\mathcal {F}\)free by \(\mathbf {X}''\), we observe that \(\mathbf {X}''\cup X'_i\) intersects every obstruction in G and hence the proof is complete. \(\square \)
From Lemma 8 and Corollary 2 we obtain the following.
Corollary 3
Let \(k\in \mathbb {N}, G\) be a graph of rankwidth at least \(k+2\) and \(\sim _k\) be the equivalence relation computed by Proposition 3. Then FindWSM \(_\mathcal {F}(k,G,\sim _k)\) either outputs a kwellstructured modulator to \(\mathcal {H}\) or correctly detects that no such modulator exists.
Proof of Theorem 6
Consider the following procedure. First, we test whether G has rankwidth at most \(k+1\) [29], and if this is the case then one can find a kwellstructured modulator to \(\mathcal {H}\) by using standard techniques. For instance, one may use the extension of Courcelle’s Theorem [11] to counting monadic second order logic (CMSO); see for instance the work of Courcelle and Oum [12]. CMSO extends MSO logic by adding atomic formulas which express that the cardinality of a set is divisible by a (fixed) number. The property of “having rankwidth at most k” is known to be expressible in CMSO logic [12], the properties of “being a splitmodule” and “not containing any obstruction from a finite set” are easily verified to be expressible in CMSO and even MSO logic, and CMSO logic can be model checked in polynomial time on graphs of bounded rankwidth [12, 29].
On the other hand, assume that the input graph G has rankwidth at least \(k+2\). Then the theorem follows by using Proposition 3 and then Algorithm 1 in conjunction with Lemma 7 and Corollary 3. \(\square \)
5 Examples of Algorithmic Applications
This section contains several examples of how the notion of kwellstructured modulators can be used to design fixedparameter algorithms.
5.1 Results for Specific Problems
Establishing the following theorem is the main objective of this subsection.
Theorem 8
Let \(\mathcal {P}\in \{\textsc {MinVC},\textsc {MaxClq}\}\) and \(\mathcal {H}\) be a graph class characterized by a finite obstruction set. Then \(\mathcal {P}\) is FPT parameterized by \(wsn^\mathcal {H}\) if and only if \(\mathcal {P}\) is polynomialtime tractable on \(\mathcal {H}\).
Since \(wsn^\mathcal {H}(G)=0\) for any \(\mathcal {F}\)free graph G, the “only if” direction is immediate; in other words, being polynomialtime tractable on \(\mathcal {H}\) is clearly a necessary condition for being fixedparameter tractable when parameterized by \(wsn^\mathcal {H}(G)\). Below we prove that for the selected problems this condition is also sufficient.
Lemma 9
If MinVC is polynomialtime tractable on a graph class \(\mathcal {H}\) characterized by a finite obstruction set, then \(\textsc {MinVC}\) parameterized by \(wsn^\mathcal {H}\) is FPT.
Proof
Let \(G=(V,E)\) be a graph and let \(k=wsn^\mathcal {H}(G)\). If \(rw(G)\le k+2\), then we simply use known algorithms to solve the problem in FPT time [19]. Otherwise, we proceed by using Theorem 6 to compute a kwellstructured modulator \(\mathbf {X}=\{X_1,\ldots ,X_k\}\) in FPT time. For each \(i\in [k]\), we let \(A_i\) be the frontier of \(X_i\) and we let \(B_i=N(A_i)\).
Since for each \(i\in [k]\) the graph \(G[A_i\cup B_i]\) contains a complete bipartite graph, any vertex cover of G must be a superset of either \(A_i\) or \(B_i\). We can branch over these options for each i in \(2^k\) time; formally, we branch over all of the at most \(2^k\) functions \(f:[i]\rightarrow \{A,B\}\), and refer to these as signatures. Each vertex cover Y of G can be associated with at least one signature f, constructed in the following way: for each \(i\in [k]\) such that \(A_i\subseteq Y\), we set \(f(i)=A\), and otherwise we set \(f(i)=B\).
Our algorithm then proceeds as follows. For a graph G and a signature f, we construct a partial vertex cover \(Z=\bigcup _{i\in [k]} f(i)\). We let \(G'=GZ\). Consider any connected component C of \(G'\). If C intersects some \(X_i\), then by the construction of Z it must hold that \(C\subseteq X_i\). Hence it follows that C either has rankwidth at most k (in the case \(C\subseteq X_i\) for some i), or C is in \(\mathcal {H}\) (if C does not intersect \(\mathbf {X}\)), or both. Then we find a minimum vertex cover for each connected component of \(G'\) independently, by either calling the known fixedparameter algorithm (if C has bounded rankwidth) or the polynomial algorithm (if C is in \(\mathcal {H}\)) at most C times. Let \(Z'\) be the union of the obtained minimum vertex covers over all the components of \(G'\), and let \(Y_f=Z\cup Z'\). After branching over all possible functions f, we compare the obtained cardinalities of \(Y_f\) and choose any \(Y_f\) of minimum cardinality. Finally, we compare \(Y_f\) and the value of m provided in the input.
We argue correctness in two steps. First, assume for a contradiction that G contains an edge e which is not covered by \(Y_f\) for some f. Then e cannot have both endpoints in \(G'\), since \(Y_f\) contains a (minimum) vertex cover for each connected component of \(G'\), but e cannot have an endpoint outside of \(G'\), since \(Z\subseteq Y_f\). Hence each \(Y_f\) is a vertex cover of G.
Second, assume for a contradiction that there exists a vertex cover \(Y'\) of G which has a lower cardinality than the vertex cover found by the algorithm described above. Let f be the signature of \(Y'\). Then it follows that \(Z\subseteq Y'\), and since \(Z\subseteq Y_f\), there would exist a component C of \(G{\setminus } Z\) such that \(Y'\cap C\le Y_f\cap C\). However, this would contradict the minimality of \(Z'\cap C=Y_f\cap C\). Hence we conclude that no such \(Y'\) can exist, and the algorithm is correct. \(\square \)
We deal with the second problem below.
Lemma 10
If MaxClq is polynomialtime tractable on a graph class \(\mathcal {H}\) characterized by a finite obstruction set, then \(\textsc {MaxClq}\) parameterized by \(wsn^\mathcal {H}\) is FPT.
Proof
We begin in the same way as for MinVC: let \(G=(V,E)\) be a graph and let \(k=wsn^\mathcal {H}(G)\). If \(rw(G)\le k+2\), then we simply use known algorithms to solve the problem in FPT time [19]. Otherwise, we proceed by using Theorem 6 to compute a kwellstructured modulator \(\mathbf {X}=\{X_1,\ldots ,X_k\}\) in FPT time. For each \(i\in [k]\), we let \(A_i\) be the frontier of \(X_i\) and we let \(B_i=N(A_i)\).
Let \(X_0=G\mathbf {X}\) and let \(s\subseteq \{0\}\cup [k]\). Then any clique C in G can be uniquely associated with a signature s by letting \(i\in s\) if and only if \(X_i\cap C\ne \emptyset \). The algorithm proceeds by branching over all of the at most \(2^{k+1}\) possible nonempty signatures s. If \(s=1\), then the algorithm simply computes a maximumcardinality clique in \(X_{s}\) (by calling the respective FPT or polynomial algorithm at most a linear number of times) and stores it as \(Y_{s}\).
If \(s\ge 2\), then the algorithm makes two checks before proceeding. First, if \(0\in s\) then it constructs the set \(X'_0\) of all vertices \(x\in X_0\) such that x is adjacent to every \(A_i\) for \(i\in s{\setminus } \{0\}\). If \(X'_0=\emptyset \) then the current choice of s is discarded and the algorithm proceeds to the next choice of s. Second, for every \(a\ne b\) such that \(a,b\in s{\setminus } \{0\}\) it checks that \(X'_a=A_a\) and \(X'_b=A_b\) are adjacent; again, if this is not the case, then we discard this choice of s and proceed to the next choice of s. Finally, if the current choice of s passed both tests then for each \(i\in s\) we compute a maximum clique in each \(G[X'_i]\) and save their union as \(Y_s\). In the end, we choose a maximumcardinality set \(Y_s\) and compare its cardinality to the value of m provided in the input.
We again argue correctness in two steps. First, assume for a contradiction that \(Y_s\) is not a clique, i.e., there exist distinct nonadjacent \(a,b\in Y_s\). Since \(Y_s\) consists of a union of cliques within subsets of \(X'_{i\in s}\), it follows that there would have to exist distinct \(c,d\in s\) such that \(a\in X'_c\) and \(b\in X'_d\). This can however be ruled out for c or d equal to 0 by the construction of \(X'_0\). Similarly, if c and d are both nonzero, then this is impossible by the second check which tests adjacency of every pair of \(X'_c\) and \(X'_d\) for every \(c,d\in s\).
Second, assume for a contradiction that there exists a clique \(Y'\) in G which has a higher cardinality than the largest clique obtained by the above algorithm. Let s be the signature of \(Y'\). If \(s=1\) then \(Y_s\ge Y'\) by the correctness of the respective FPT or polynomial algorithm used for each \(X_s\). If \(s\ge 2\) then \(Y'\) may only intersect the sets \(X'\) constructed above for s. Moreover, if there exists \(i\in [k]\cup \{0\}\) such that \(Y'\cap X'_i> Y_s\cap X'_i\) then we again arrive at a contradiction with the correctness of the respective FPT or polynomial algorithms used for \(X'_i\). Hence we conclude that no such \(Y'\) can exist, and the algorithm is correct. \(\square \)
Fact 2
 1.
\((2K_2,C_4,C_5)\)free graphs (split graphs);
 2.
\(P_5\)free graphs;
 3.
forkfree graphs;
 4.
\((\text {banner}, T_{2,2,2})\)free graphs and \((\text {banner}, K_{3,3}\text {}e, \text {twinhouse})\)free graphs.
Proof
For item 1, recall that split graphs are graphs whose vertex set can be partitioned into one clique and one independent set, and such a partitioning can be found in linear time. If each vertex in the clique is adjacent to at least one independent vertex, then the clique is a minimum vertex cover, otherwise the clique without a pendantfree vertex is a minimum vertex cover. Item 2 follows from [33]. Item 3 follows from [1]. Item 4 follows from [24] and [5]. \(\square \)
Fact 3
 1.
Any complementary graph class to the classes listed in Fact 2 (such as coforkfree graphs and split graphs);
 2.
Graphs of bounded degree.
Proof
 1.
It is wellknown that each maximum clique corresponds to a maximum independent set (and viceversa) in the complement graph.
 2.
The degree bounds the size of a maximum clique, again resulting in a simple folklore branching algorithm. The class of graphs of degree at most d is exactly the class of \(\mathcal {F}\)free graphs for \(\mathcal {F}\) containing all \((d+1)\)vertex supergraphs of the star having d leaves. \(\square \)
5.2 Results for Other Graph Classes
Next, we turn our attention to computing kwellstructured modulators to examples of graph classes which are not characterized by a finite obstruction set (i.e., by a finite set of forbidden induced subgraphs). In the following lemmas, n denotes the size of the vertex set of the input graph.
Lemma 11
It is possible to compute a kwellstructured modulator to the class of forests in time \(f(k)\cdot n^5\) for some computable function f.
Proof

First, it uses Proposition 3 to partition V(G) into equivalence classes of \(\sim _k\) in time at most \(f(k)\cdot n^5\) for some computable function f, and sets \(j:=k; S:=\emptyset ; \sim :=\sim _k\).

Second, for each tuple (X, Y, Z) of equivalence classes of \(\sim , {\mathcal {A}}\) checks whether \(G[X\cup Y\cup Z]\) is acyclic. If this is not the case, then \({\mathcal {A}}\) chooses (by branching) one class out of \(\{X,Y,Z\}\) to delete from \(\sim \), saves the deleted equivalence class in S, and restarts the second step with \(j:=j1\). If \(j=1\), then the algorithm terminates the given branch.

Third, \({\mathcal {A}}\) constructs an auxiliary graph \(G'=(V',E')\) by setting \(V'\) to be the set of equivalence classes of \(\sim \) and \(E'\) to contain an edge between \(A,B\in V'\) iff there exist vertices \(a\in A, b\in B\) such that \(ab\in E(G)\).

Finally, \({\mathcal {A}}\) tries to find a feedback vertex set in \(G'\) of size at most j in time \(\mathcal {O}(3.83^j\cdot jV'^{2})\) [7]. If no such feedback vertex set exists, then \({\mathcal {A}}\) terminates the given branch; otherwise it adds the feedback vertex set to \(\mathbf {S}\) and outputs \(\mathbf {S}\).
Moreover, assume \(C\cap X>1\) for some equivalence class X of \(\sim _k\). Since C spans at least 4 equivalence classes, H must contain at least two neighbors of X in \(C{\setminus } X\) which are adjacent to at least two vertices in X (indeed, recall that X is a splitmodule and hence all vertices of X with a neighbor outside X have the same neighborhood outside X); let us denote these vertices \(y,z,x_1,x_2\), respectively. Since \(x_1, x_2\) are in the frontier of X, the vertices \(y,x_1,z,x_2\) must form a cycle in H which spans at most 3 equivalence classes, contradicting our previous conclusion that no such cycles are present in H. Hence we may conclude that \(C\cap X\le 1\) for every equivalence class X.
The only case we are left with now is that C intersects each equivalence class at most once. But then it must be the case that C also forms a cycle in \(G'\), which would have necessarily been removed in step 4 of \({\mathcal {A}}\), a contradiction. So H must indeed be acyclic.
For the other direction, assume that G contains a minimal kwellstructured modulator \(\mathbf {X}=\{X_1,\ldots , X_j\}\) to the class of forests. Then consider the branch of step 2 of \({\mathcal {A}}\) which hits a maximal number of elements of \(\mathbf {X}\), and let us denote the elements removed by \({\mathcal {A}}\) in this way \(\mathbf {Y}\). By the same argument as above, each cycle remaining in G after deleting \(\mathbf {Y}\) intersects each equivalence class at most once and hence corresponds to a cycle in the graph \(G'\) constructed by \({\mathcal {A}}\). In particular, the equivalence classes in \(\mathbf {X}{\setminus } \mathbf {Y}\) form a feedback vertex set of size \(\ell =\mathbf {X}{\setminus } \mathbf {Y}\) in \(G'\). By the correctness of the feedback vertex set algorithm used in step 4, at least one branch of \({\mathcal {A}}\) is guaranteed to output a solution \(\mathbf {S}\supset \mathbf {Y}\) of size at most j. \(\square \)
For the next result, recall that a cycle is chordless if it is also an induced cycle of length at least 4, and a graph is chordal if it contains no chordless cycles.
Lemma 12
It is possible to compute a kwellstructured modulator to the class of chordal graphs in time \(f(k)\cdot n^{\mathcal {O}(1)}\).
Proof

First, it uses Proposition 3 to partition V(G) into equivalence classes of \(\sim _k\) in time at most \(f(k)\cdot n^5\), and sets \(j:=k; S:=\emptyset ; \sim :=\sim _k\).

Second, for each tuple (X, Y, Z) of equivalence classes of \(\sim , {\mathcal {A}}\) checks whether \(G[X\cup Y\cup Z]\) is chordal in linear time [36]. If this is not the case, then \({\mathcal {A}}\) chooses (by branching) one class out of \(\{X,Y,Z\}\) to delete from \(\sim \), saves the deleted equivalence class in S, and restarts the second step with \(j:=j1\). If \(j=1\), then the algorithm terminates the given branch.

Third, \({\mathcal {A}}\) constructs an auxiliary graph \(G'=(V',E')\) by setting \(V'\) to be the set of equivalence classes of \(\sim \) and \(E'\) to contain an edge between \(A,B\in V'\) iff there exist vertices \(a\in A, b\in B\) such that \(ab\in E(G)\).

Finally, \({\mathcal {A}}\) tries to find a modulator to chordal graphs of size at most j in \(G'\) using the algorithm by Marx [8], which takes time at most \(2^{\mathcal {O}(k\log k)}\cdot V'^{\mathcal {O}(1)}\). If no such modulator exists, then \({\mathcal {A}}\) terminates the given branch; otherwise it adds the modulator to \(\mathbf {S}\) and outputs \(\mathbf {S}\).
We now claim that C contains at most one vertex from each equivalence class of \(\sim _k\). To see this, assume for a contradiction that C contains two vertices in some equivalence class Z. Since C must also intersect other equivalence classes, it follows that C must in fact contain at least two vertices, say x, y, in the frontier of Z which have distinct neighbors, say \(x', y'\) respectively, in \(C{\setminus } Z\). First, observe that x, y cannot occur consecutively along C, as that would violate the assumption that C is a minimumlength chordless cycle. Hence by the chordality of C we also see that xy cannot be an edge of G, and the same also applies for the nonedge of \(x'y'\). But then \(x',x,y',y\) forms a chordless cycle which intersects at most 3 equivalence classes, contradicting our previous assumptions.
Let us now consider the set \(C'\) of equivalence classes which intersect C; recall that \(C'>3\). By the above claim, it follows that \(C'\) would also be a chordless cycle in \(V(G')\), contradicting the correctness of the chordal vertex deletion algorithm [8]. Hence we conclude that H must in fact be chordal.
For the other direction, assume that G contains a minimal kwellstructured modulator \(\mathbf {X}=\{X_1,\ldots , X_j\}\) to the class of chordal graphs. Then consider the branch of step 2 of \({\mathcal {A}}\) which hits a maximal number of elements of \(\mathbf {X}\), and let us denote the elements removed by \({\mathcal {A}}\) in this way \(\mathbf {Y}\). By the same argument as above, each chordless cycle remaining in G after deleting \(\mathbf {Y}\) intersects each equivalence class at most once and hence corresponds to a chordless cycle in the graph \(G'\) constructed by \({\mathcal {A}}\). In particular, the equivalence classes in \(\mathbf {X}{\setminus } \mathbf {Y}\) form a modulator to chordal graphs of size \(\ell =\mathbf {X}{\setminus } \mathbf {Y}\) in \(G'\). By the correctness of the chordal vertex deletion algorithm used in step 4, at least one branch of \({\mathcal {A}}\) is guaranteed to output a solution \(\mathbf {S}\supset \mathbf {Y}\) of size at most j. \(\square \)
6 MSO Model Checking with WellStructured Modulators
Here we show how wellstructured modulators can be used to solve the MSO model checking problem, as formalized in Theorem 9 below. Note that our metatheorem captures not only the generality of MSO model checking problems, but also applies to a potentially unbounded number of choices of the graph class \(\mathcal {H}\). Thus, the metatheorem supports two dimensions of generality.
Theorem 9
Let \(\phi \) be a \({\text {MSO}} \) sentence and \(\mathcal {H}\) be a graph class characterized by a finite obstruction set. If \(\textsc {MSOMC}_{\phi }\) is FPT parameterized by \(mod^\mathcal {H}(G)\), then \(\textsc {MSOMC}_{\phi }\) is also FPT parameterized by \(wsn^\mathcal {H}(G)\).
The condition that \(\textsc {MSOMC}_{\phi }\) is FPT parameterized by \(mod^\mathcal {H}(G)\) is a necessary condition for the theorem to hold by Proposition 1. However, it is natural to ask whether it is possible to use a weaker necessary condition instead, specifically that \(\textsc {MSOMC}_{\phi }\) is polynomialtime tractable on \(\mathcal {H}\) (as was done for specific problems in Sect. 5). Before proceeding towards a proof of Theorem 9, we make a digression and show that the weaker condition used in Theorem 8 is in fact not sufficient for the general case of MSO model checking.
Lemma 13
There exists an MSO sentence \(\phi \) and a graph class \(\mathcal {H}\) characterized by a finite obstruction set such that \(\textsc {MSOMC}_{\phi }\) is polynomialtime tractable on \(\mathcal {H}\) but NPhard on the class of graphs having \(wsn^\mathcal {H}(G)\le 2\) and even \(mod^\mathcal {H}(G)\le 2\).
Proof
Consider the sentence \(\phi \) which describes the existence of a proper 5coloring of the vertices of G, and let \(\mathcal {H}\) be the class of graphs of degree at most 4 (in other words, let \(\mathcal {F}\) contain all 6vertex supergraphs of the star having 5 leaves). There exists a trivial greedy algorithm to obtain a proper 5coloring of any graph of degree at most 4, hence \(\textsc {MSOMC}_{\phi }\) is polynomialtime tractable on \(\mathcal {H}\). Now consider the class of graphs obtained from \(\mathcal {H}\) by adding, to any graph in \(\mathcal {H}\), two adjacent vertices y, z which are both adjacent to every other vertex in the graph. By construction, any graph \(G'\) from this new class satisfies \(mod^\mathcal {H}(G')\le 2\) and hence also \(wsn^\mathcal {H}(G')\le 2\). However, \(G'\) admits a proper 5coloring if and only if \(G'\{y,z\}\) admits a proper 3coloring. Testing 3colorability on graphs of degree at most 4 is known to be NPhard [30], and hence the proof is complete. \(\square \)
Our strategy for proving Theorem 9 relies on a replacement technique, where each splitmodule in the wellstructured modulator is replaced by a small representative. We use the notion of similarity defined below to prove that this procedure does not change the outcome of \(\textsc {MSOMC}_{\varphi }\).
Definition 8
 1.
There exists an isomorphism \(\tau \) between \(G\mathbf {X}\) and \(G'\mathbf {X}'\).
 2.
For every \(v\in V(G){\setminus } \mathbf {X}\) and \(i\in [k]\), it holds that v is adjacent to \(S_i\) if and only if \(\tau (v)\) is adjacent to \(S'_i\).
 3.
If \(k\ge 2\), then for every \(1\le i<j\le k\) it holds that every \(s_i\in S_i\) is adjacent to every \(s_j\in S_j\) if and only if every \(s'_i\in S'_i\) is adjacent to every \(s'_j\in S'_j\).
 4.
For each \(i\in [k]\), it holds that \( type _q(G[X_i],S_i)= type _q(G'[X'_i],S'_i)\).
Lemma 14
Let q and k be nonnegative integers, \(\mathcal {H}\) be a graph class, and let G and \(G'\) be graphs having kwellstructured modulators \(\mathbf {X}=\{X_1,\ldots ,X_k\}\) and \(\mathbf {X'}=\{X_1',\ldots ,X_k'\}\) to \(\mathcal {H}\), respectively. If \((G, \mathbf {X})\) and \((G', \mathbf {X}')\) are qsimilar, then it holds that \( type _q(G, \emptyset ) = type _q(G', \emptyset )\).
Proof
 1.
Suppose Spoiler makes a set move W and assume without loss of generality that \(W \subseteq V(G)\). For \(i \in [k]\), let \(W_i = X_i \cap W\), and let \(W_i'\) be Duplicator’s response to \(W_i\) according to \(\pi _i\). Furthermore, let \(W'_0=\{\,\tau (v)\;{}\;v\in W\cap X_0 \,\}\). Then Duplicator responds with \(W' = W'_0 \cup \bigcup _{i=1}^k W_i'\).
 2.
Suppose Spoiler makes a point move s and again assume without loss of generality that \(s \in V(G)\). If \(s \in X_i\) for some \(i \in [k]\), then Duplicator responds with \(s' \in X_i'\) according to \(\pi _i\); otherwise, Duplicator responds with \(\tau (s)\) as per Definition 8 item 1.

Let \(j_1,j_2 \in [m]\) and let \(v_{j_1}, v_{j_2}\in X_0\). Since \(\tau \) is an isomorphism as per Definition 8 item 1, it follows that \(v_{j_1}=v_{j_2}\) if and only if \(u_{j_1} = u_{j_2}\) and \(v_{j_1}v_{j_2}\in E(G)\) if and only if \(u_{j_1}u_{j_2}\in E(G')\).

Let \(j_1,j_2 \in [m]\) and let \(i\in [k]\) be such that \(v_{j_1}\in X_0\) and \(v_{j_2}\in X_i\). Then clearly \(v_{j_1}\ne v_{j_2}\) and \(u_{j_1}\ne u_{j_2}\). Consider the case \(v_{j_1}v_{j_2}\in E(G)\). Then \(v_{j_2}\) must lie in the frontier of \(X_i\), and hence \(v_{j_2}\in S_i\). Since Duplicator’s strategy \(\pi _i\) is winning for \((G_i,S_i)\) and \((G'_i,S'_i)\), it must hold that \(u_{j_2}\in S'_i\). By Definition 8 item 2, it then follows that \(\tau (v_{j_1})u_{j_2}\in E(G')\). So, consider the case \(v_{j_1}v_{j_2}\not \in E(G)\). Then either \(v_{j_2}\not \in S_i\), in which case it holds that \(u_{j_2}\not \in S'_i\) because of the choice of \(\pi _i\) and hence there cannot be an edge \(u_{j_2}u_{j_1}\) in \(G'\), or \(v_{j_2}\in S_i\), in which case it holds once again that \(u_{j_2}u_{j_1}\not \in E(G')\) by Definition 8 item 2.

Let \(j_1,j_2 \in [m]\) and let \(i \in [k]\) be such that \(v_{j_1},v_{j_2} \in X_{i}\). Since Duplicator plays according to a winning strategy \(\pi _i\) in the game on \(G_i\) and \(G_i'\), the restriction \((\mathbf {v}_i, \mathbf {u}_i)\) defines a partial isomorphism between \((G_i, (\mathbf {V})_i)\) and \((G_i', (\mathbf {U})_i)\). It follows that \((v_{j_1},v_{j_2}) \in E(G)\) if and only if \((u_{j_1},u_{j_2}) \in E(G')\) and \(v_{j_1} = v_{j_2}\) if and only if \(u_{j_1} = u_{j_2}\).

Let \(j_1,j_2 \in [m]\) and let \(i_1,i_2 \in [k]\) be pairwise distinct numbers such that \(v_{j_1}\in X_{i_1}\) and \(v_{j_2}\in X_{i_2}\). Then \(v_{j_1} \ne v_{j_2}\) and also \(u_{j_1} \ne u_{j_2}\) since \(u_{j_1} \in X_{i_1}'\) and \(u_{j_2} \in X_{i_2}'\) by the Duplicator’s strategy. Suppose \(v_{j_1}v_{j_2}\in E(G)\). Then \(v_{j_1}\in S_{i_1}\), and \(v_{j_2}\in S_{i_2}\), and \(S_{i_1}\) and \(S_{i_2}\) are adjacent in G. From the correctness of \(\pi _{i_1}\) and \(\pi _{i_2}\) it follows that \(u_{j_1}\in S'_{i_1}\) and \(u_{j_2}\in S'_{i_2}\), and from Definition 8 item 3 it follows that \(S'_{i_1}\) and \(S'_{i_2}\) are adjacent in \(G'\), which together implies \(u_{j_1}u_{j_2}\in E(G')\). On the other hand, suppose \(v_{j_1}v_{j_2}\not \in E(G)\). Then either \(v_{j_1}\not \in S_{i_1}\), or \(v_{j_2}\not \in S_{i_2}\), or \(S_{i_1}\) and \(S_{i_2}\) are not adjacent in G. In the first case we have \(u_{j_1}\not \in S'_{i_1}\), in the second case we have \(u_{j_2}\not \in S'_{i_2}\), and in the third case it holds that \(S'_1\) and \(S'_2\) are not adjacent in \(G'\); any of these three cases imply \(u_{j_1}u_{j_2}\not \in E(G')\).

Let \(j \in [m]\) such that \(v_j\in X_0\). Then by the Duplicator’s strategy on \(X_0\) it follows that for any \(V_q\) such that \(v_j\in V_q\) it holds that \(u_j\in U_q\) and for any \(V_q\) such that \(v_j\not \in V_q\) it holds that \(u_j\not \in U_q\).

Let \(j\in [m]\) and \(i\in [k]\) such that \(v_j\in X_k\). Let \(V_q\) be such that \(v_j\in V_q\). Since \(\pi _i\) is a winning strategy for Duplicator, it must be the case that \(u_j\in U_q\). Similarly, if \(v_j\not \in V_q\) then the correctness of \(\pi _i\) guarantetes that \(u_j\not \in U_q\). \(\square \)
Next, we show that small representatives can be computed efficiently.
Lemma 15
Let \(q\in \mathbb {N}_0\). There exist functions f, g such that one can compute, for an input graph G of rankwidth at most k and \(S\subseteq V(G)\), in time \(f(k)\cdot V(G)^{\mathcal {O}(1)}\) a graph \(G'\) and a set \(S'\subseteq V(G')\) such that \(V(G')\le g(q)\) and \( type _q(G,S) = type _q(G',S')\).
Proof
By Lemma 1 we can compute a formula \(\varPhi (Q)\) capturing the type T of (G, S) in time \(f(k)\cdot V(G)^{\mathcal {O}(1)}\). Given \(\varPhi (Q)\), a constantsize model \((G',S')\) satisfying \(\varPhi (Q)\) can be computed as follows. We start enumerating all graphs (by brute force and in any order with a nondecreasing number of vertices), and check for each graph \(G^*\) and every vertexsubset \(S^*\subseteq V(G^*)\) whether \(G^* \models \varPhi (S^*)\). If this is the case, we stop and output \((G^*,S^*)\). Since \(G \models \varPhi (S)\) this procedure must terminate eventually. Fixing the order in which graphs are enumerated, the number of graphs we have to check depends only on T. By Fact 1 the number of qtypes is finite for each q, so we can think of the total number of checks and the size of each checked graph \(G^*\) as bounded by a constant. Moreover the time spent on each check depends only on T and the size of the graph \(G^*\). Consequently, after we compute \(\varPhi (Q)\) it is possible to find a model for \(\varPhi (Q)\) in constant time. \(\square \)
Finally, in Lemma 16 below we use Lemma 15 to replace any wellstructured modulator by a small but “equivalent” modulator.
Lemma 16
Let q be a nonnegative integer constant and \(\mathcal {H}\) be a graph class. Then given a graph G and a kwellstructured modulator \(\mathbf {X}=\{X_1,\ldots X_k\}\) of G into \(\mathcal {H}\), there exists a function f such that one can in time \(f(k)\cdot V(G)^{\mathcal {O}(1)}\) compute a graph \(G'\) with a kwellstructured modulator \(\mathbf {X'}=\{X'_1,\ldots X'_k\}\) into \(\mathcal {H}\) such that \((G,\mathbf {X})\) and \((G',\mathbf {X'})\) are qsimilar and for each \(i \in [k]\) it holds that \(X_i'\) is bounded by a constant.
Proof
 1.
We construct a disjoint union of \(G_0\) and \(G'_i\) for each \(i\in [k]\);
 2.
If \(k\ge 2\) then for each \(1\le i<j \le k\) such that \(S_i\) and \(S_k\) are adjacent in G, we add edges between every \(v\in S'_i\) and \(w\in S'_j\).
 3.
for every \(v\in V(G_0)\) and \(i\in [k]\) such that \(S_i\) and \(\{v\}\) are adjacent, we add edges between v and every \(w\in S'_i\).
Proof of Theorem 9
Let G be a graph, \(k=wsn^\mathcal {H}(G)\) and q be the nesting depth of quantifiers in \(\phi \). By Theorem 6 it is possible to find a kwellstructured modulator to \(\mathcal {H}\) in time \(f(k)\cdot V^{\mathcal {O}(1)}\). We proceed by constructing \((G',\mathbf {X}')\) by Lemma 16. Since each \(X'_i\in \mathbf {X}'\) has size bounded by a constant and \(\mathbf {X}'\le k\), it follows that \(\bigcup \mathbf {X}'\) is a modulator to the class of \(\mathcal {F}\)free graphs of cardinality \(\mathcal {O}(k)\). Hence \(\textsc {MSOMC}_{\phi }\) can be decided in FPT time on \(G'\). Finally, since G and \(G'\) are qsimilar, it follows from Lemma 14 that \(G\models \phi \) if and only if \(G'\models \phi \). \(\square \)
For completeness, we remark that the same proof can be used to obtain analogues of Theorem 9 for any graph class \(\mathcal {H}\) which admits a fixedparameter algorithm for finding wellstructured modulators (i.e., even if it is not characterized by a finite obstruction set; see for instance Lemma 11 and 12).
We conclude the section by showcasing an example application of Theorem 9. cColoring asks whether the vertices of an input graph G can be colored by c colors so that each pair of neighbors have distinct colors. From the connection between cColoring, its generalization List c Coloring and modulators [6, Theorem 3.3] and tractability results for List c Coloring [28, Page 5], we obtain the following.
Corollary 4
For each \(c\in \mathbb {N}, c\)Coloring parameterized by \(wsn^{P_5\text {free}}\) is FPT.
7 Hardness of MSO Optimization
We say that \(S\subseteq V(G)\) is a dominating set if every vertex in G either is in S or has a neighbor in S. We will need the following lemma before we proceed to the main result of this secton.\(\textsc {MSOOpt}^{\le }_{\varphi }\)
Instance: A graph G and an integer \(r\in \mathbb {N}\).
Question: Is there a set \(S \subseteq V(G)\) such that \(G \models \varphi (S)\) and \(S \le r\)?
Lemma 17
The problem of finding a pcardinality dominating set in a graph G having a kcardinality modulator \(X\subseteq V(G)\) to the class of graphs of degree at most 3 is FPT when parameterized by \(p+k\).
Proof
Let \(L=V(G){\setminus } X\) and consider the following algorithm. We begin with \(D=\emptyset \), and choose an arbitrary vertex \(v\in L\) which is not yet dominated by D. We branch over the at most \(k+4\) vertices q in \(\{v\}\cup N(v)\), and add q to D. If \(D=p\) and there still exists an undominated vertex in G, we discard the current branch; hence this procedure produces a total of at most \((k+4)^p\) branches.
Now consider a branch where \(D<p\) but the only vertices left to dominate lie in X. For \(a,b\in L\), we let \(a\equiv b\) if and only if \(N(a)\cap X=N(b)\cap X\). Notice that \(\equiv \) has at most \(2^k\) equivalence classes and that these may be computed in polynomial time. For each nonempty equivalence class of \(\equiv \), we choose an arbitrary representative and construct the set P of all such chosen representatives. We then branch over all subsets Q of \(P\cup X\) of cardinality at most \(pD\), and add Q into D. Since \(P\cup X\le 2^k+k\), this can be done in time bounded by \(\mathcal {O}(2^{p\cdot k})\). Finally, we test whether this D is a dominating set, and output the minimum dominating set obtained in this manner.
It is easily observed from the description that the running time is FPT. For correctness, from the final check it follows that any set outputed by the algorithm will be a dominating set. It remains to show that if there exists a dominating set of cardinality p, then the algorithm will find such a set. So, assume there exists a pcardinality dominating set \(D'\) in G. Consider the branch arising from the first branching rule obtained as follows. Let \(v_1\) be the first undominated vertex in L chosen by the algorithm, and consider the branch where an arbitrary \(q\in D'\cap N(v_1)\) is placed into D. Hence, after the first branching, there is a branch where \(D\subseteq D'\). Similarly, there exists a branch where \(D\subseteq D'\) for each \(v_i\) chosen in the ith step of the first branching. If \(D'=D\) after the first branching, then we are done; so, let \(D'_1=D'{\setminus } D\) be nonempty. Let \(D_1\) be obtained from \(D'_1\) by replacing each \(w\in D'_1\) by the representative of \([w]_\equiv \) chosen to lie in P. Since \(D'\) dominates all vertices in L and \(D_1\) dominates the same vertices in X as \(D'_1\), it follows that \(D^*=(D'{\setminus } D'_1)\cup D_1\) is also a dominating set of G. Furthermore, \(D^*=D'\). However, since \(D_1\subseteq P\) and \(D_1\le pD\), there must exist a branch in the second branching which sets \(Q=D_1\). Hence there exists a branch in the algorithm which obtains and outputs the set \(D^*=D\cup D_1\). \(\square \)
Theorem 10
There exists an MSO formula \(\varphi \) and a graph class \(\mathcal {H}\) characterized by a finite obstruction set such that \(\textsc {MSOOpt}^{\le }_{\varphi }\) is FPT parameterized by \(mod^\mathcal {H}\) but paraNPhard parameterized by \(wsn^\mathcal {H}\).
Proof
To prove Theorem 10, we let dom(S) express that S is a dominating set in G, and let cyc(S) express that S intersects every \(C_4\) (cycle of length 4). Then we set \(\varphi (S)=dom(S)\vee cyc(S)\) and let \(\mathcal {H}\) be the class of \(C_4\)free graphs of degree at most 3 (obtained by letting the obstrucion set \(\mathcal {F}\) contain \(C_4\) and all 5vertex supergraphs of \(K_{1,4}\)).
Claim
\(\textsc {MSOOpt}^{\le }_{\varphi }\) is FPT parameterized by the cardinality of a modulator to \(\mathcal {H}\).
To argue that the above claim holds, let \((G=(V,E),r)\) be the input of \(\textsc {MSOOpt}^{\le }_{\varphi }\) and k be the cardinality of a modulator in G to \(\mathcal {H}\). We begin by computing some modulator \(X\subseteq V\) of cardinality k in G to \(\mathcal {H}\); this can be done in FPT time by a simple branching algorithm on any of the obstructions from \(\mathcal {F}\) located in G. Let \(L=V{\setminus } X\). Next, we compare r and k, and if \(r\ge k\) then we output YES. This is correct, since each \(C_4\) in G must intersect X and hence setting \(S=X\) satisfies \(\varphi (S)\).
So, assume \(r<k\). Then we check whether there exists a set A of cardinality at most r which intersects every \(C_4\); this can be done in time \(O^*(4^r)\) by a simple FPT branching algorithm. Next, we check whether there exists a dominating set B in G of cardinality at most r; this can also be done in FPT time by Lemma 17.
Finally, if A or B exists, then we output YES and otherwise we output NO. Hence the claim is indeed true.
Claim
\(\textsc {MSOOpt}^{\le }_{\varphi }\) is paraNPhard parameterized by \(wsn^\mathcal {H}(G)\).
We proceed by arguing that this claim is also correct. It is known that the Dominating Set problem, which takes as input a graph G and an integer j and asks to find a dominating set of size at most j, is NPhard on \(C_4\)free graphs of degree at most 3 [31] (see also subsequent work [2, Theorem 8]). We use this fact as the basis of our reduction. Let (G, j) be a \(C_4\)free instance of Dominating Set of degree at most 3. Then we construct \(G'\) from G by adding \((G+2)\)many copies of \(C_4\), a single vertex q adjacent to every vertex of every such \(C_4\), and a single vertex \(q'\) adjacent to q and an arbitrary vertex of G. It is easy to check that \(wsn^\mathcal {H}(G')\le 2\).
We claim that (G, j) is a YESinstance of Dominating Set if and only if \((G',j+1)\) is a yesinstance of \(\textsc {MSOOpt}^{\le }_{\varphi }\). For the forward direction, assume there exists a dominating set D in G of cardinality j. Then the set \(D\cup \{q\}\) is a dominating set in \(G'\), and hence satisfies \(\varphi \).
On the other hand, assume there exists a set \(D'\) of cardinality at most \(j+1\) which satisfies \(\varphi \). If \(j+1\ge G+2\) then clearly (G, j) is a YESinstance of Dominating Set, so assume this is not the case. But then \(D'\) cannot intersect every \(C_4\), and hence \(D'\) must be a dominating set of \(G'\) of cardinality at most \(j+1\). But this is only possible if \(q\in D'\). Furthermore, if \(q'\in D'\), then replacing \(q'\) with the neighbor of \(q'\) in G is also a dominating set of \(G'\). Hence we may assume, w.l.o.g., that \(D'\cap V(G)\) is a dominating set of cardinality at most j in V(G). Consequently, (G, j) is a YESinstance of Dominating Set and the claim holds.
The theorem now follows from the two claims proved above.\(\square \)
8 Conclusion
We have introduced a family of structural parameters which push the frontiers of fixedparameter tractability beyond rankwidth and modulator size for a wide range of problems. In particular, the wellstructure number can be computed efficiently (Theorem 6) and used to design fixedparameter algorithms for Minimum Vertex Cover and Maximum Clique (Theorem 8) as well as any problem which can be described by a sentence in MSO logic (Theorem 9). We remark that while our results are of a theoretical nature, there is hope that some of the ideas behind the presented algorithms may be useful in practice once faster algorithms for computing rankwidth become available.
For future work, it would be interesting to see whether the notion of splitmodules introduced in this work can be naturally generalized. In particular, a splitmodule X can be seen as a subgraph such that \(\varvec{A}_G[X,GX]=1\), and in this sense splitdecompositions naturally correspond to rankwidth 1. It is easy to define corresponding decompositions also for higher values of rankwidth, however it is not at all clear how such decompositions could be computed. We believe this is an interesting question on its own; furthermore, obtaining such decompositions would allow an immediate extension of our framework to the arising more general notions of splitmodules.
Finally, we remark that wellstructured modulators have also found applications in the area of data reduction and kernelization [16]. In particular, since \(wsn^{\mathcal {H}}\) lowerbounds rankwidth and rankwidth is known not to admit polynomial kernels for nearly any NPhard problems, one cannot hope to use \(wsn^{\mathcal {H}}\) for polynomial kernelization. However, a number of polynomial kernels have been developed for a restriction of \(wsn^{\mathcal {H}}\) where each split module induces a graph whose rankwidth is bounded by a constant [17].
Notes
Acknowledgements
Open access funding provided by Austrian Science Fund (FWF). This work was partially supported by the Austrian Science Fund (FWF), Projects P26696 and W1255N23. Robert Ganian is also affiliated with FI MU, Brno, Czech Republic.
References
 1.Alekseev, V.E.: Polynomial algorithm for finding the largest independent sets in graphs without forks. Discrete Appl. Math. 135(1–3), 3–16 (2004)MathSciNetGoogle Scholar
 2.Alekseev, V.E., Korobitsyn, D.V., Lozin, V.V.: Boundary classes of graphs for the dominating set problem. Discrete Math. 285(1–3), 1–6 (2004)MathSciNetCrossRefMATHGoogle Scholar
 3.Bodlaender, H.L., Jansen, B.M.P., Kratsch, S.: Kernel bounds for path and cycle problems. In: IPEC, pp. 145–158 (2011)Google Scholar
 4.Bouchet, A.: Digraph decompositions and Eulerian systems. SIAM J. Algebra Discr. 8, 323–337 (1987)MathSciNetCrossRefMATHGoogle Scholar
 5.Brandstädt, A., Lozin, V.V.: A note on alpharedundant vertices in graphs. Discrete Appl. Math. 108(3), 301–308 (2001)MathSciNetCrossRefMATHGoogle Scholar
 6.Cai, L.: Parameterized complexity of vertex colouring. Discrete Appl. Math. 127(3), 415–429 (2003)MathSciNetCrossRefMATHGoogle Scholar
 7.Cao, Y., Chen, J., Liu, Y.: On feedback vertex set: new measure and new structures. Algorithmica 73(1), 63–86 (2015)MathSciNetCrossRefMATHGoogle Scholar
 8.Cao, Y., Marx, D.: Chordal editing is fixedparameter tractable. Algorithmica 75(1), 118–137 (2016)MathSciNetCrossRefMATHGoogle Scholar
 9.Charbit, P., de Montgolfier, F., Raffinot, M.: Linear time split decomposition revisited. SIAM J. Discrete Math. 26(2), 499–514 (2012)MathSciNetCrossRefMATHGoogle Scholar
 10.Corneil, D.G., Lerchs, H., Burlingham, L.S.: Complement reducible graphs. Discrete Appl. Math. 3, 163–174 (1981)MathSciNetCrossRefMATHGoogle Scholar
 11.Courcelle, B., Makowsky, J.A., Rotics, U.: Linear time solvable optimization problems on graphs of bounded cliquewidth. Theory Comput. Syst. 33(2), 125–150 (2000)MathSciNetCrossRefMATHGoogle Scholar
 12.Courcelle, B., Oum, S.: Vertexminors, monadic secondorder logic and a conjecture by Seese. J. Comb. Theory Ser. B 97(1), 91–126 (2007)MathSciNetCrossRefMATHGoogle Scholar
 13.Cunningham, W.H.: Decomposition of directed graphs. SIAM J. Algebra Discr. 3(2), 214–228 (1982)MathSciNetCrossRefMATHGoogle Scholar
 14.Cygan, M., Fomin, F.V., Kowalik, L., Lokshtanov, D., Marx, D., Pilipczuk, M., Pilipczuk, M., Saurabh, S.: Parameterized Algorithms. Springer, New York (2015)CrossRefMATHGoogle Scholar
 15.Diestel, R.: Graph Theory, Volume 173 of Graduate Texts in Mathematics, 2nd edn. Springer, New York (2000)Google Scholar
 16.Downey, R.G., Fellows, M.R.: Fundamentals of Parameterized Complexity. Texts in Computer Science. Springer, Berlin (2013)CrossRefMATHGoogle Scholar
 17.Eiben, E., Ganian, R., Szeider, S.: Metakernelization using wellstructured modulators. In: Husfeldt, T., Kanj, I.A. (eds.) 10th International Symposium on Parameterized and Exact Computation, IPEC 2015, September 16–18, 2015, Patras, Greece, volume 43 of LIPIcs, pp 114–126. Schloss Dagstuhl  LeibnizZentrum fuer Informatik (2015)Google Scholar
 18.Gajarský, J., Hliněný, P., Obdržálek, J., Ordyniak, S., Reidl, F., Rossmanith, P., Villaamil, F.S., Sikdar, S.: Kernelization using structural parameters on sparse graph classes. In: Algorithms—ESA 2013. 21st Annual European Symposium, volume 8125 of Lecture Notes in Computer Science, pp 529–540. Springer (2013)Google Scholar
 19.Ganian, R., Hliněný, P.: On parse trees and Myhill–Nerodetype tools for handling graphs of bounded rankwidth. Discrete Appl. Math. 158(7), 851–867 (2010)MathSciNetCrossRefMATHGoogle Scholar
 20.Ganian, R., Slivovsky, F., Szeider, S.: Metakernelization with structural parameters. In: MFCS, pp. 457–468 (2013)Google Scholar
 21.Gaspers, S., Misra, N., Ordyniak, S., Szeider, S., Živný, S.: Backdoors into heterogeneous classes of SAT and CSP. In: Brodley, C.E., Stone, P. (eds.) Proceedings of the TwentyEighth AAAI Conference on Artificial Intelligence, pp. 2652–2658. AAAI Press (2014)Google Scholar
 22.Gaspers, S., Szeider, S.: Backdoors to satisfaction. In: Bodlaender, H.L., Downey, R., Fomin, F.V., Marx, D. (eds.) The Multivariate Algorithmic Revolution and Beyond—Essays Dedicated to Michael R. Fellows on the Occasion of His 60th Birthday, Volume 7370 of Lecture Notes in Computer Science, pp. 287–317. Springer (2012)Google Scholar
 23.Gavoille, C., Paul, C.: Distance labeling scheme and split decomposition. Discrete Math. 273(1–3), 115–130 (2003)MathSciNetCrossRefMATHGoogle Scholar
 24.Gerber, M.U., Lozin, V.V.: Robust algorithms for the stable set problem. Graphs Comb. 19(3), 347–356 (2003)MathSciNetCrossRefMATHGoogle Scholar
 25.Gioan, E., Paul, C.: Dynamic distance hereditary graphs using split decomposition. In: Algorithms and Computation, Volume 4835 of LNCS, pp. 41–51. Springer (2007)Google Scholar
 26.Gioan, E., Paul, C.: Split decomposition and graphlabelled trees: characterizations and fully dynamic algorithms for totally decomposable graphs. Discrete Appl. Math. 160(6), 708–733 (2012)MathSciNetCrossRefMATHGoogle Scholar
 27.Gioan, E., Paul, C., Tedder, M., Corneil, D.: Practical and efficient split decomposition via graphlabelled trees. Algorithmica 69(4), 789–843 (2014)MathSciNetCrossRefMATHGoogle Scholar
 28.Golovach, P.A., Paulusma, D., Song, J.: Closing complexity gaps for coloring problems on hfree graphs. Inf. Comput. 237, 204–214 (2014)MathSciNetCrossRefMATHGoogle Scholar
 29.Hliněný, P., Oum, S.: Finding branchdecompositions and rankdecompositions. SIAM J. Comput. 38(3), 1012–1032 (2008)MathSciNetCrossRefMATHGoogle Scholar
 30.Kochol, M., Lozin, V.V., Randerath, B.: The 3colorability problem on graphs with maximum degree four. SIAM J. Comput. 32(5), 1128–1139 (2003)MathSciNetCrossRefMATHGoogle Scholar
 31.Korobitsyn, D.: On the complexity of determining the domination number in monogenic classes of graphs. Diskret. Mat. 2(3), 90–96 (1990). In Russian, translation in. Discrete Math. Appl. 2(2), 191–199 (1992)Google Scholar
 32.Libkin, L.: Elements of Finite Model Theory. Springer, Berlin (2004)CrossRefMATHGoogle Scholar
 33.Lokshantov, D., Vatshelle, M., Villanger, Y.: Independent set in \({p}_{5}\)free graphs in polynomial time. In: Proceedings of the TwentyFifth Annual ACMSIAM Symposium on Discrete Algorithms, SODA 2014, pp. 570–581. SIAM (2014)Google Scholar
 34.Niedermeier, R.: Invitation to FixedParameter Algorithms. Oxford Lecture Series in Mathematics and its Applications. Oxford University Press, Oxford (2006)CrossRefGoogle Scholar
 35.Oum, S., Seymour, P.: Approximating cliquewidth and branchwidth. J. Combin. Theory Ser. B 96(4), 514–528 (2006)MathSciNetCrossRefMATHGoogle Scholar
 36.Rose, D.J., Tarjan, R.E., Lueker, G.S.: Algorithmic aspects of vertex elimination on graphs. SIAM J. Comput. 5(2), 266–283 (1976)MathSciNetCrossRefMATHGoogle Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.