Semitotal Dominating Set
Semitotal Dominating Set
1 Introduction
The domination set problem (DS) is one of the well-studied problems in the field
of graph theory. A set D ⊆ V (G) is a dominating set if each vertex v ∈ V (G)\D is
adjacent to at least one vertex in D. The Minimum dominating set problem
aims to compute a minimum cardinality dominating set of G. The minimum
cardinality of a dominating set is called the domination of G, denoted by γ(G).
The problem is extensively studied in the literature [1-3][?]. Motivated by the
various real-world applications, researchers introduced several generalizations
and variations of domination in graphs.
The total dominating set problem (TD) is the most natural variant of the
dominating set problem. A set D ⊆ V (G) is a total dominating set if D is a
dominating set and each vertex v ∈ D has a neighbor in D. The total domination
number of G, denoted by γt (G), is the minimum cardinality of a total dominating
set of G. The computational complexity study of TD is reported in [4-7][?].
Goddard et al. [6] have introduced a problem called the semitotal dominating
set problem (SD) which is a generalization of TD. For a connected graph G, a set
D ⊆ V (G) is a semitotal dominating set if D is a dominating set and each vertex
v ∈ D has another vertex in D which is at a distance within two. SD of G asks
for finding minimum cardinality semitotal dominating set, denoted by γt2 (G).
It is important to highlight that γ(G) ≤ γt2 (G) ≤ γt (G). On the computational
complexity front, it is known from [?] that SD is NP-complete (intractable) for
general graphs.
Many intractable computational problems have several real-world applica-
tions in various fields, specifically SD has applications in wireless networks. An
approach for dealing with intractable computational problems having some small
parameter is parameterized complexity [4]. For decision problems with input size
n, and a parameter k, the objective is to design an algorithm with running time
f (k)nO(1) where f is a function of k. Problems that admit such an algorithm
are said to be fixed-parameter tractable (FPT). Similarly, problems that do not
admit such an algorithm are said to be W [i]-hard, for some i. In this paper,
we study the parameterized complexity of SD with respect to both natural pa-
rameters and structural parameters. We show that SD which is W [2]-hard when
parameterized by a natural parameter (the solution size) on special graph classes,
and SD is fixed-parameter tractable when parameterized by a structural param-
eter (cliquewidth, and treewidth).
Related works
In [?], the semitotal dominating set problem has been introduced and proved
to be NP-complete in general graphs. It is known that SD is NP-complete for
split graphs, planar graphs, chordal bipartite graphs, and circle graphs. On the
positive side, SD is polynomial-time solvable on trees, AT-free graphs, interval
graphs, and block graphs.
Our results. In this paper, we shall study the following problem in the realm
of parameterized complexity.
The semitotal dominating set problem (SD) Input: A graph G, and an
integer k.
Parameter: k
Question: Does there exists a set S of size at most k such that V (G) \ S has
a neighbour in S and every vertex in S has a vertex in S which is at a distance
one or two.
Many dominating set variants such as DS, TD, and the connected dominating
set problem have been analyzed in graphs with respect to girth. It is shown that
many of these variants exhibit a FPT algorithm when parameterized by the
solution size on graphs with girth at least five and is W [2]-hard for graphs with
girth three or four. We explore a similar direction for SD, and obtained the
following results.
2
Theorem 1. SD is W[2]-hard on graphs with girth three when parameterized
by k.
To prove the theorem, we give a parameterized reduction from the dominating
set problem on general graphs to SD on graphs with girth three.
Theorem 2. SD is W[2]-hard on graphs with girth four when parameterized by
k.
We prove the above theorem for graphs with girth four by giving a param-
eterized reduction from the dominating set problem to SD on C3 -free graphs.
We observe that the reduction instances are bipartite graphs, in particular, the
reduction instances have a cycle of length four. Thus, we have the following two
corollaries.
Corollary 1. SD is W[2]-hard on bipartite graphs when parameterized by k.
Corollary 2. SD is W[2]-hard on C3 -free graphs when parameterized by k.
Having observed the complexity of SD on graphs with girth three and girth
four, it is an interesting direction to look into the complexity of the problem on
graphs with girth at least five (G5 graphs).
Theorem 3. SD is FPT on graphs with girth at least five when parameterized
by k.
By using a branching strategy, We give an exponential kernel for the problem
on G5 graphs. Our approach is similar to one presented in [7].
The study of SD on graphs with respect to girth gives us an interesting di-
chotomy. The next natural structure of graphs one can consider is the degeneracy
of the input graph. We study SD on d-degenerate graphs.
Theorem 4. SD is W[2]-hard on d-degenerate graphs when parameterized by
k.
We prove the theorem by a reduction from r-regular graphs to SD on d-
degenerate graphs. Further, by the construction we observe the reduction in-
stances are bipartite graphs. Thus the following corollary is true.
Corollary 3. SD is W[2]-hard on d-degenerate bipartite graphs when parame-
terized by k.
The structural parameters expand the boundary of FPT for SD. We analyze
the complexity of SD with respect to well-known structural parameters such as
treewidth and cliquewidth.
Theorem 5. SD admits an FPT algorithm, when parameterized by cliquewidth.
By giving a monadic second-order logic formula for SD, we shall prove that
SD parameterized by cliquewidth is FPT. Since any graph of treewidth at most
k admits clique-width at most k[3], SD parameterized by treewidth is also in
FPT. Further, we also provide an explicit algorithm for SD parameterized by
the treewidth of the input graph.
3
Theorem 6. SD admits an FPT algorithm, when parameterized by treewidth
of the input graph.
To prove the above theorem, we give a dynamic programming algorithm.
It is known that treewidth of a d-degenerate graph is at most d [?]. Thus,
SD on d-degenerate graphs admit FPT algorithm when parameterized by d.
4
that (NH [xi ] ∪ NH [yi ]) ∩ D = ∅, which is a contradiction that D is a semitotal
dominating set. Thus, it follows that SD is W[2]-hard on graphs with girth
three.
Proof ( Proof of Theorem 2). Given an instance (G, k) of DS , we construct a
corresponding instance of (H, k+1) in polynomial time such that H is a bipartite
graph. The vertex set of H is: V (H) = X ∪ Y ∪ {w, z}, let X = {xi | vi ∈ V (G)},
and let Y = {yi | vi ∈ V (G)}. The edge set of H is: E(G) = {{xi , yj }, {yi , xj } |
{vi , vj } ∈ E(G)} ∪ {{xi , yi } | 1 ≤ i ≤ n} ∪ {{w, yi } ∪ 1 ≤ i ≤ n} ∪ {{w, z}}.
The girth of the reduced instance H is four because H is bipartite and the
presence of a single edge in G creates a cycle of length four.
(⇒) If G has a dominating set S of size at most k, then H has a semitotal
dominating set of size at most k + 1. Let D = {yi | vi ∈ S} ∪ {w}. By our
construction, any vertex in Y is at a distance two from each other. Thus D is a
semitotal dominating set.
(⇐) If H has a semitotal dominating set D of size at most k + 1, then
G has a dominating S of size at most k. Observe that D ∩ {w, z} ̸= ∅. Let
S = {vi | vi ∈ V (G), {xi , yi } ∩ D ̸= ∅}. Since |D| ≤ k + 1, |S| ≤ k. Now
we need to show that S is a dominating set of G. Suppose that there exists
a vertex vi ∈ V (G) such that NG [vi ] ∩ S = ∅. Then it must be the case that
(NH [xi ] ∪ NH [yi ]) ∩ D = ∅, which is a contradiction that D is a semitotal
dominating set. Thus, it follows that SD is W[2]-hard on graphs with girth
four.
5
We use a coloring scheme to identify the vertices that are in D and have a
neighbor in D which is at a distance within two, D, and do not have a neighbor
in D which is at a distance within two, N [D], and vertices that are not yet
dominated. At any instance of our algorithm, the vertices are colored with one
of the following colors;
Observe that initially all vertices are colored black. Now we define a semitotal
dominating set on G5 graphs with vertices colored with orange, red, white, or
black. We call the colored G5 graph as an orwb-graph.
ORWB-Semitotal dominating set: Let G be an orwb-graph satisfying
the following conditions, and let k be a positive integer parameter. Let O, R, W ,
and B be the set of vertices colored orange, red, white, and black respectively.
can we write Does G have at most k − |O ∪ R| vertices that dominate all the black vertices?
|R| = ∅ Finally, we want our orwb-graph to have |O| ≤ k, R = ∅, and B = ∅. The
following lemma holds true for semitotal dominating set on G5 graphs.
Lemma 3. Let G which is an orwb-graph be an instance of ORWB-semitotal
dominating set with a positive integer k as the parameter. Let v be the black
or white vertex with more than k − |O ∪ R| black neighbors. Then if G has a set
of size at most k − |O ∪ R| that dominates all black vertices, then v must be part
of every such set.
The proof of Lemma 3 is similar to one presented in [?].
From Lemma 3, we can observe the following reduction rules.
6
Observe that now we have five partitions of vertices, that is vertices that are
colored orange, red, black, and white vertices are partitioned into W and W ′ ,
respectively.
Lemma 4. Let G be an instance of ORWB- Semitotal dominating set, and
let G′ be the instance after applying rules R1 to R4 once. Let k be the integer
parameter. Then G has a set of size at most k − |R ∪ O| dominating all vertices
in B and satisfying the constraint that |R| = ∅ if and only if G′ has a set of size
at most k − |R′ | dominating all vertices in B ′ and satisfying the constraint that
|R′ | = ∅.
Observe that the vertices in W ′ might be useful in ensuring the semitotal
domination property for vertices in R. From Lemma 3, and Lemma 4, we shall
prove the following lemma.
Lemma 5. Let (G, k) be a yes instance of ORWB- Semitotal dominating
set and (G′ , k ′ ) be the reduced instance of (G, k) after applying rules R1-R4
until no longer possible. Then the number of vertices in G′ is O(k 3 + 2k ).
Proof. to be added
Thus ORWB-Semitotal dominating set can be solved in time O((k 3 +
2k )k nO(1) ) time. Since the resultant graph after applying reduction rules R1-R4
recursively, until no longer possible results in a graph with at most k 3 + 2k + k
vertices, we can just try all possible subsets of size at most k for k 3 + 2k vertices.
4 SD on d-degenerate graphs
4.1 W[2]-hardness on d-degenerate graphs
Proof. (proof of Theorem 4) It is known from [?] that DS on r-regular graphs is
W[2]-hard and this can be reduced in polynomial time to SD in d-degenerate
graphs using the following reduction algorithm. We map an instance (G, k) of
DS on r-regular graphs to the corresponding instance (G′ , k) of SD as follows:
V (G′ ) = W ∪ X ∪ Y ∪ Z,
W = {wi | vi ∈ V (G)}, X = {xi | vi ∈ V (G)},
Y = {yi | vi ∈ V (G)},
Z = {zi | vi ∈ V (G)}.
We shall now describe the edges of G′ ,
E(G∗ ) = E1 ∪ E2 ,
E1 = {wi , xj , {xi , wj }, {yi , zj }, {zi , yj }, {xi , yj }, {xj , yi } | {vi , vj } ∈ E(G)},
and E2 = {{wi , xi }, {xi , yi }, {yi , zi } | 1 ≤ i ≤ n}.
Observe that the maximum degree of G′ is 2r + 2. Thus, G′ is a 2r + 2-
degenerate graph.
(⇒) If G has a dominating set S of size at most k, then G′ has a semitotal
dominating set D of size at most 2k. Let D = {yi , xi | vi ∈ S}. By our
construction, for each xi , yi ∈ D, they are at a distance one. Thus D is a semitotal
dominating set.
7
(⇐) If G′ has a semitotal dominating set D of size at most 2k, then G has
a dominating set S of size at most k. Suppose that |(W ∪ X) ∩ D| ≥ k. Then
|(Y ∪Z)∩D| ≤ k. Let S = {vi | yi ∈ D∨zi ∈ D, 1 ≤ i ≤ n}. Now we need to show
that S is a dominating set of G. Suppose that there exist a vertex vi ∈ V (G) such
that NG [vi ] ∩ S = ∅. Then it must be the case that (NG′ [xi ] ∪ NG′ [yi ]) ∩ D = ∅,
which is a contradiction that D is a semitotal dominating set. Thus it follows
that SD W[2]-hard on d-degenerate graphs.
5 Structural parameters
1. Leaf node.
2. Introduce vertex node.
3. Introduce edge node.
4. Forget node.
5. Join node.
8
Our approach is similar to the one presented in [1]. Since we need to distin-
guish between the vertices in the dominating set and those that do not have a
distance within two vertex in the dominating set, we use three colors. For the
remaining vertices, we use two colors to differentiate whether or not they have
been dominated by at least one vertex in the dominating set. Thus, a coloring
of bag Xt is a mapping f : Xt → 1, 1′ , 2, 0, 0′ .
- 1, meaning that the vertex is in the dominating set and satisfies semitotal
dominating set property or has a neighbor in 2.
- 1′ , meaning that the vertex is in the dominating set and does not satisfy
semitotal dominating set property.
- 2, meaning that the vertex is already dominated, not in D, and it is adjacent
to a vertex in 1. These are helper vertices for vertices in 1 to satisfy the
semitotal dominating set.
- 0, meaning that the vertex is not in the dominating set and is already dom-
inated
- 0′ , meaning that the vertex is not in the dominating set and is yet to be
dominated.
We use colors to refer to a function mapping of a vertex.
Observe that at a node t, there are at most 5tw colorings of Xt . For a coloring
f of Xt , let c[t, f ] denotes the minimum size of a semitotal dominating set D of
Vt such that D ∩ Xt = f −1 (1) ∪ f −1 (1′ ), and every vertex in f −1 (2) ∪ f −1 (0)
is adjacent to at least one vertex in f −1 (1) ∪ f −1 (1′ ). Further every vertex in
f −1 (1′ ) does not have a neighbor in f −1 (2) or f −1 (1). We call such a set D a
minimum compatible set for t and f . We have c[t, f ] = +∞, if t and f are not
minimum compatible in D.
By the property of nice tree decomposition, we know that for root node r,
Xr = ∅ and Gr = G. Thus the root node has only one possible coloring which is
empty coloring. Thus the size of the minimum semitotal dominating set in G is
exactly the value of c[r, ∅].
For a coloring f of X, Y ⊆ X, f |Y denotes the mapping of f restricted to
Y . Next, we shall define the recursive formulas for the values of c at each node
of the tree decomposition.
1. Leaf node. For a leaf node Xt = ∅. Hence there is only one coloring possible.
c[t, ∅] = 0
9
3. Introduce edge node. Let t be an introduce edge node labelled with an
edge {u, v}. Let t′ be the child of t. Then V (Gt ) = V (Gt′ ) and E(Gt ) =
E(Gt′ ) ∪ {{u, v}}. The edge {u, v} is useful in the following ways; if f (u) = 1
or f (u) = 1′ and f (v) = 0 or f (v) = 0′ , then taking precomputed solution
for t′ can help the color of v from 0′ to 0 (or vice versa).
c[t′ , fv→0′ ], if f (u) ∈ {1, 1′ } and f (v) = 0
c[t′ , fu→0′ ], if f (u) = 0 and f (v) ∈ {1, 1′ }
′
c[t , fu→1′ ,v→1′ ], if f (u) = 1 and f (v) = 1
c[t′ , f
u→1′ ,v→0′ ], if f (u) = 1 and f (v) = 2
c[t, f ] = ′
c[t , fu→0′ ,v→1′ ], if f (u) = 2 and f (v) = 1
c[t′ , fu→1′ ,v→2 ], if f (u) = 1 and f (v) = 0
c[t′ , fu→2,v→1′ ], if f (u) = 0 and f (v) = 1
′
c[t , f ], otherwise
4. Forget node. Let t be a forget node with a child t′ . Then for some w ∈ Xt′ ,
Xt = Xt′ \ {w}. Every coloring of t′ that maps w to 1 or 0 is a valid coloring.
Thus, we obtain the following recursion.
5. Join node. Let t be a join node with children t1 and t2 . Then we know that
Xt = Xt1 = Xt2 . We say that colorings f1 of Xt1 or f2 of Xt2 are consistent
with a coloring f of Xt if for every v ∈ Xt the following conditions hold;
(a) f (v) = 1 if and only if f1 (v) = f2 (v) = 1.
(b) f (v) = 1′ if and only if f1 (v) = f2 (v) = 1′ .
(c) f (v) = 2 if and only if f1 (v) = f2 (v) = 2
(d) f (v) = 0 if and only if (f1 (v), f2 (v)) ∈ {(0′ , 0), (0, 0′ )}
(e) f (v) = 0′ if and only if f1 (v) = f2 (v) = 0′
Observe that D is a compatible set for t and f if and only if D must be a
compatible set for t1 and f1 , and t2 and f2 . Let D1 in Gt1 and D2 in Gt2 .
Since D1 ⊆ D and D2 ⊆ D, D∩Xt = D1 ∩Xt1 = D2 ∩Xt2 = f −1 (1)∪f −1 (1′ ).
Thus, |D| = |D1 | + |D2 | − |f −1 (1) ∪ f −1 (1′ )|. Hence, we obtain the following
recursion.
Running time.
References
1. Alber, Bodlaender, Fernau, Kloks, and Niedermeier. Fixed parameter algorithms for
dominating set and related problems on planar graphs. Algorithmica, 33:461–493,
2002.
10
2. Bruno Courcelle. The monadic second-order logic of graphs xiv: uniformly sparse
graphs and edge set quantifications. Theoretical Computer Science, 299(1):1–36,
2003.
3. Bruno Courcelle and Stephan Olariu. Upper bounds to the clique width of graphs.
Discrete Applied Mathematics, 101(1):77–114, 2000.
4. Marek Cygan, Fedor V Fomin, Lukasz Kowalik, Daniel Lokshtanov, Dániel Marx,
Marcin Pilipczuk, Michal Pilipczuk, and Saket Saurabh. Parameterized algorithms,
volume 4. Springer, 2015.
5. Marek Cygan, Fedor V. Fomin, Lukasz Kowalik, Daniel Lokshtanov, Daniel Marx,
Marcin Pilipczuk, Michal Pilipczuk, and Saket Saurabh. Parameterized Algorithms.
Springer Publishing Company, Incorporated, 1st edition, 2015.
6. Wayne Goddard, Michael A Henning, and Charles A McPillan. Semitotal domina-
tion in graphs. Utilitas Mathematica, 94, 2014.
7. Venkatesh Raman and Saket Saurabh. Short cycles make W-hard problems hard:
FPT algorithms for W-hard problems in graphs with no short cycles. Algorithmica,
52:203–225, 2008.
11