On computational and combinatorial properties of the total co-independent domination number of graphs
Abel Cabrera Martinez, Frank A. Hernandez Mira, Jose M. Sigarreta, Almira, Ismael G. Yero

TL;DR
This paper investigates the properties and computational complexity of the total co-independent domination number in graphs, proving NP-completeness of related decision problems and characterizing trees with equal domination numbers.
Contribution
It introduces the concept of total co-independent domination number, proves NP-completeness of its decision problem, and characterizes trees where this number equals the total domination number.
Findings
Deciding if _{t,coi}(G) k is NP-complete.
Provides bounds on _{t,coi}(G).
Characterizes trees with equal total co-independent and total domination numbers.
Abstract
A subset of vertices of a graph is a total dominating set if every vertex of is adjacent to at least one vertex of . The total dominating set is called a total co-independent dominating set if the subgraph induced by is edgeless and has at least one vertex. The minimum cardinality of any total co-independent dominating set is the total co-independent domination number of and is denoted by . In this work we study some complexity and combinatorial properties of . Specifically, we prove that deciding whether for a given integer is an NP-complete problem and give several bounds on . Also, since any total co-independent dominating set is also a total dominating set, we characterize all the trees having equal total co-independent domination number and total domination number.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Graph Theory Research Β· Graph Labeling and Dimension Problems Β· Interconnection Networks and Systems
On computational and combinatorial properties of the total co-independent domination number of graphs
Abel Cabrera MartΓnez*(1)*, Frank A. HernΓ‘ndez Mira (1)
JosΓ© M. Sigarreta Almira*(1)* and Ismael G. Yero*(2)*
(1) Facultad de MatemΓ‘ticas, Universidad AutΓ³noma de Guerrero
Carlos E. Adame 5, Col. La Garita, Acapulco, Guerrero, Mexico
[email protected], [email protected], *[email protected]
(2)Departamento de MatemΓ‘ticas, Escuela PolitΓ©cnica Superior de Algeciras
Universidad de CΓ‘diz,* Av. RamΓ³n Puyol s/n, 11202 Algeciras, Spain.
Abstract
A subset of vertices of a graph is a total dominating set if every vertex of is adjacent to at least one vertex of . The total dominating set is called a total co-independent dominating set if the subgraph induced by is edgeless and has at least one vertex. The minimum cardinality of any total co-independent dominating set is the total co-independent domination number of and is denoted by . In this work we study some complexity and combinatorial properties of . Specifically, we prove that deciding whether for a given integer is an NP-complete problem and give several bounds on . Also, since any total co-independent dominating set is also a total dominating set, we characterize all the trees having equal total co-independent domination number and total domination number.
Keywords: total co-independent domination; total domination; vertex independence; vertex cover.
AMS Subject Classification Numbers: 05C69
1 Introduction
Problems concerning domination in graphs are one of the most popular and highly investigated ones in the area of graph theory, and a rich literature in the topic is nowadays known in the research community. Such problems run from the theoretical point of view till several practical applications in real life situations. The most interesting cases of such applications are probably regarding problems in computer science. One of the most interesting features of domination in graphs involves the existence of a very high number of variants of domination parameters. A very common combination of domination is made with vertex independence of graphs, and the way of combining both concepts groups a considerable number of possibilities. In this work, we center our attention into precisely study one of these combinations between domination and independence in graphs, namely the total co-independence domination parameter. We focus the investigation on some computational complexity aspects of this parameter as well as on combinatorial properties of it.
Given a graph with vertex set and edge set , a set is a total dominating set of if every vertex in is adjacent to at least one vertex in . The total domination number of is the minimum cardinality of any total dominating set in and is denoted by . A -set is a total dominating set of cardinality . For more information on total domination we suggest the recent and fairly complete survey [10] and the book [11]. A set of vertices is independent if induced an edgeless graph. An independent set of maximum cardinality is a maximum independent set of . The independence number of is the cardinality of a maximum independent set of and is denoted by . An independent set of cardinality is called a -set. Relationships between (total) domination and independence in graphs have attracted the attention of several researchers in the last years. Several interesting connection among these parameters include independent dominating sets [2, 14], partitions into a dominating set and an independent set [13], (total) dominating sets which intersect every maximal independent set [1, 4, 9], and some other ones more, which we prefer to not mention here, since it is not the goal of this work.
A total dominating set of a graph is called a total co-independent dominating set (or TC-ID set for short) if the set of vertices of the subgraph induced by is independent and not empty111Notice that the condition of to be not empty is not exactly necessary. However, if such condition is not required, then we readily seen that the only graphs containing a TC-ID set of minimum cardinality with empty complement are the union of paths .. The minimum cardinality of any TC-ID set is the total co-independent domination number of and is denoted by . A TC-ID set of cardinality is a -set. These concepts were previously introduced and barely studied in [15]. Moreover, in [12], the same parameter was introduced under the name of total outer-independent domination number. Since this article ([12]) is not published in any journal and the other one ([15]) is already published, we precisely follow the terminology and notation of the latter. Since total domination is not defined for graphs having isolated vertices, all the graphs considered herein have not isolated vertices. Moreover, in order to satisfy the total domination property and that the complement of a TC-ID set is not empty, it is required that , if is the order of . Such trivial bounds were already noted in the seminal work [15].
Throughout this work we consider as a simple graph of order and size . That is, graphs that are finite, undirected, and without loops or multiple edges. Given a vertex of , represents the open neighborhood of , i.e., the set of all neighbors of in and the degree of is . The minimum and maximum degrees of are denoted by and , respectively (or and , respectively, for short). If and are two subsets of , then we denote the set of all edges of joining a vertex of with a vertex of by . For a set , the complement of is . In this work, we represent an edgeless graph of order as . For any other graph theory terminology and notation we follow the book [11].
Let be a tree (a connected graph without cycles). A leaf or a pendant vertex of is a vertex of degree one (it is similarly defined for non tree graphs). A support vertex of is a vertex adjacent to a leaf and a semi-support vertex is a vertex adjacent to a support vertex that is not a leaf. By an isolated support vertex of we mean an isolated vertex of the subgraph induced by the support vertices of . The set of leaves of is denoted by , the set of support vertices by , and the set of semi-support vertices by . Moreover, is the set of isolated support vertices of .
We first notice that if , , β¦, with , are the connected components of a graph , then any TC-ID set of minimum cardinality in is formed by a minimum total dominating set in the subgraphs where and a minimum TC-ID set in the remaining subgraphs with . That is stated in the following result.
Remark 1**.**
Let , , β¦, with , be the connected components of a graph different from the union of copies of the path . Then
[TABLE]
Proof.
Let be a -set for such that . It is easy see that is a TC-ID set of and we have
[TABLE]
On the other hand, let be a . Firstly, we observe that for every such that , it is satisfied that . Moreover, let for every such that . We notice that every must be a TC-ID set of . In this sense,
[TABLE]
which completes the proof. β
In concordance with the result above, from now on, we only consider the study of the TC-ID sets of connected graphs and omit to refer to that fact throughout all our exposition.
2 Complexity of the decision problem
We begin our exposition by considering the problem of deciding whether the total co-independent domination number of a graph is less than a given integer. That is stated in the following decision problem.
[TABLE]
In order to deal with the complexity of the TOTAL CO-INDEPENDENT DOMINATION PROBLEM (TC-ID PROBLEM), we make a reduction from a very well known decision problem concerning the independence number of graphs.
[TABLE]
The problem above is one of the classical NP-complete problems appearing in the book [8]. Moreover, it remains NP-complete even when restricted to planar graphs.
Theorem 2**.**
[8]* MAXIMAL INDEPENDENT SET PROBLEM is NP-complete even when restricted to planar graphs of maximum degree at most 3.*
Now on, in order to present our complexity results we need to introduce a family of graphs which is next defined. Let be a tree with six vertices having two adjacent vertices of degree three and the other four vertices are leaves. Clearly, each vertex of degree three has two adjacent leaves, say and (see Figure 1 (I)). Given a graph of order and trees isomorphic to the tree , the graph is constructed by adding edges between the -vertex of and the vertex of the -tree . See Figure 1 (II) for an example.
We are now able to prove the NP-completeness of the TC-ID PROBLEM.
Theorem 3**.**
TOTAL CO-INDEPENDENT DOMINATION PROBLEM is NP-complete even when restricted to planar graphs of maximum degree at most 3.
Proof.
The problem is clearly in NP since verifying that a given set is indeed a TC-ID set can be done in polynomial time. Let us now make a reduction from the MAXIMAL INDEPENDENT SET PROBLEM. Let be a not edgeless graph of order and construct the graph as described above. Let us denote by the vertices of degree three in the copy of the tree used to generate . We shall prove that .
Let be a -set and let be the set of vertices of obtained from the complement of in together with the vertices belonging to all the copies of the tree used to generate , that is . This set is clearly a total dominating set and its complement is an independent set. Thus, is a TC-ID set in and, as a consequence,
[TABLE]
On the other hand, let be a -set. In order to totally dominate the leaves of every copy of in , it must happen that for every . Moreover, , since otherwise the complement of would not be independent. Moreover the complement of in is an independent set in . Thus, and we obtain the following.
[TABLE]
As a consequence, it follows that .
Now, for , it is readily seen that if and only if , which complete the reduction. We also observe that, if is a planar graph, then is also planar. Therefore, since the MAXIMAL INDEPENDENT SET PROBLEM is NP-complete even when restricted to planar graphs of maximum degree at most 3, we also deduce that the TC-ID PROBLEM is NP-complete even when restricted to planar graphs of maximum degree at most 3 and the proof is completed. β
As a consequence of the result above, we deduce the following consequence.
Corollary 4**.**
The problem of computing the total co-independent domination number of graphs is NP-hard even when restricted to planar graphs of maximum degree at most 3.
3 Bounding the total co-independent domination number
In order to present the first bounds for of any graph , we need the next concepts. A set of vertices of is a vertex cover of if every edge of is incident with at least one vertex of . The vertex cover number of , denoted by , is the smallest cardinality of a vertex cover of . We refer to an -set in as a vertex cover of cardinality . The following well-known result, due to Gallai [7], states the relationship between the independence number and the vertex cover number of a graph.
Theorem 5**.**
[7](Gallai, 1959)* For any graph of order , *
On the other hand, it was shown in [15] the following relationship between and .
Theorem 6**.**
[15]* For any graph of order , *
By using the two theorems above, we can easily deduce the lower bound of our next result. However, an upper bound for in terms of the vertex cover number can also be deduced. We first consider the case whether is a star graph for which is known that and .
Remark 7**.**
For any star graph ,
In concordance with the remark above, for our next result we exclude the case of star graphs and see that they behave in a different manner.
Theorem 8**.**
For any graph of order different from a star graph,
[TABLE]
Proof.
The lower bound follows from Theorems 5 and 6. If , then . Thus, from now on in this proof we consider . Now, let be an -set.
We choose two vertices with the minimum possible distance between them and let be a shortest path. Clearly, and the distance between and is one or two (notice this also means ). For each vertex , choose a neighbor of . Then is a TC-ID set of cardinality , which completes the proof of the upper bound. β
The bounds above are tight. For instance, a characterization of that trees achieving the equality in the lower bound was given in [3] (note that in [3] the trees of order satisfying equality in the bound were characterized, which equals the lower bound of Theorem 8, in concordance with Theorem 5). The upper bound is attained for an infinite family of graphs, as we next show. To this end, we need the following operations for edges or induced paths of a graph .
Subdivision: Given an edge , remove the edge, add a vertex and the edges , .
Inflation of size : Given an induced path of , in which has degree two, remove the vertex and the two incident edges, and replace them with vertices and edges for every .
Addition of pendant vertices: Given a vertex add new vertices and the edges for every .
Now, a graph is a graph obtained from a star graph by making the following sequence of operations, which we will call as Sequence I.
- (a)
Apply the operation βSubdivisionβ to () edges of .
- (b)
Apply the operation βInflation of size β with to () paths obtained from (a).
- (c)
Apply the operation βAddition of pendant verticesβ, , to the vertices corresponding to leaves of obtained in the step (b).
- (d)
Apply the operation βAddition of pendant verticesβ, , to the leaves belonging to the remaining paths obtained from (a), which were not βinflatedβ in (b).
- (e)
If and (notice that in this case is a tree such that the central vertex of the original star graph has no adjacent leaves), then apply the operation βAddition of pendant verticesβ, , to the vertex corresponding to the central vertex of .
- (f)
If and , then apply the operation βAddition of pendant verticesβ, , to the vertex corresponding to the central vertex of .
As an example, to obtain the cycle (which belongs to ) we begin with the star (a path ), next we apply the operation βSubdivisionβ to the unique edge of and then we apply the operation βInflation of size β to the path obtained in the previous step. Note that different sequences of operations would lead to the same graph. For instance, the graph can be obtained from the star by subdividing its unique edge and then adding a pendant vertex to the leaf corresponding to the subdivision, as well as another pendant vertex to the center of (coincidentally such center is also a leaf). Moreover, the graph is obtained from the star by subdividing one of its edges and then adding a pendant vertex to the leaf corresponding to such subdivision. On the other hand, we remark that three integers would produce different graphs depending on the addition of pendant vertices that would be done. However, since it is not significant for our work to denote them, we skip to use the notations for the addition of pendant vertices. A fairly representative graph of the family is given in Figure 2.
Remark 9**.**
For any graph , and .
Proof.
For any edge of which was subdivided in step (a), it appears either a path or a cycle and all these subgraphs have in common only one vertex (the corresponding one to the center of ). Thus, in order to cover all the edges of , at least vertices are required. Moreover, a set given by those leaves corresponding to the edges of the star which were subdivided together with the central vertex form a vertex cover of cardinality . Thus, the equality follows.
Now, let be a -set. We analyze the following situations for every edge (assume is the center of ) of the star which is initially subdivided.
Case 1: There is only one path between and in . Hence, the edge was subdivided with a vertex, say , and not inflated, which made a required addition of at least one pendant vertex, say , to the leaf . Thus, in order to totally dominate , .
Case 2: There are at least two paths between and in . Clearly, this means was subdivided and then inflated with at least two vertices, say , . Moreover, probably some pendant vertices were added to . So, in order to totally dominate (and probably other extra leaves adjacent to ), at least two vertices of are required.
We next consider the vertex separately. If , then the vertex has a least one adjacent leaf which needs to be totally dominated. Thus, must belong to . On the contrary, if , then we must consider the value . If , then no path was inflated ( is a tree) and so, by step (e), has at least one adjacent leaf which needs to be totally dominated, which means must belong to again. Finally, we assume . Thus, at least one path was inflated and there is a cycle to which belongs. Also, it may happen has no adjacent leaves. Now, note that if , then the two vertices of adjacent to must belong to , since is an independent set. Moreover, the fourth vertex of must belong to too, in order to get the vertices of totally dominated. As a consequence, at least three vertices of the cycle are in , which is equivalent to have in the vertex , one of its neighbors in and the vertex of which is not adjacent to .
Consequently, we can deduce that for any set of vertices of a subgraph of , induced by the vertices obtained in a subdivision of one of the leaves of and probably the corresponding addition of some pendant vertices, at least two of these vertices are in . Moreover, one extra vertex is required, which could mainly be the central vertex of . Thus, .
On the other hand, by using Theorem 8, we obtain that and the equality follows for . β
Now, a graph is a graph obtained from the cycle by making the following sequence of operations, which we will call as Sequence II.
- (a)
Apply the operation βAddition of pendant verticesβ, and , to the three vertices, say , of a -set, respectively.
- (b)
If there is a from the above operation and the degree of is two, then apply the operation βInflation of size β with to one of the two possible paths of order three between and the other two vertices in .
- (c)
Apply the operation βInflation of size β with and to the three possible paths of order three between .
An example of a graph of the family appears in Figure 3.
The following result concerning the values of and for graphs is straightforward to observe.
Remark 10**.**
For any graph , and .
According to the Remarks above, we can easily check that the upper bound of Theorem 8 is achieved for any graph . Moreover, we next prove that precisely the graphs of these families are the only ones achieving the upper bound of Theorem 8. To this end, we need the following two lemmas whose proofs can be made by using some similar techniques as in the proof of Theorem 8.
Lemma 11**.**
If a graph contains an -set which is not independent, then .
Proof.
Let be an -set which is not independent. We choose two adjacent vertices . For each vertex , choose a neighbor of . Then is a TC-ID set of cardinality . β
Lemma 12**.**
If an -set of a graph contains four different vertices such that a shortest path and a shortest path have length two and are vertex disjoint, then .
Proof.
Let be an -set such that . For each vertex , choose a neighbor of and let be two vertices such that and , which exist by assumption. Then is a TC-ID set of cardinality . β
Theorem 13**.**
Let be a graph of order such that . Then if and only if .
Proof.
In one hand, if , then it clearly happens that according to Remarks 9 and 10.
On the second hand, assume and let be any -set. We first notice that must induce an independent set according to Lemma 11. We shall now proceed by proving some partial claims that will further give our required conclusion.
Claim 1: has no triangles (cycles of order three).
Proof of Claim 1: If there is a triangle, then, in order to cover all its edges, at least two of its vertices must belong to . So, this cover is not an independent set. Thus we get a contradiction by using Lemma 11.
Claim 2: has no induced cycles of order five or larger than six.
Proof of Claim 2: Suppose contains a cycle with or . In order to cover all the edges of , and since , there must be two vertices in at distance one (which means is not independent), or there are four different vertices such that a shortest path and a shortest path have length two and are vertex disjoint. Thus, we obtain contradictions by using Lemmas 11 and 12.
As a consequence of the Claims above, we have that can only contain cycles of order four or six. We first analyze the case in which contains a cycle of order six. Let where ( means are adjacent). According to Lemma 11, it must happen is independent. Thus, without loss of generality we assume . We consider now several situations.
Suppose has degree larger than two. If with , then has a triangle, which is not possible. If , then for each vertex , choose a neighbor of and we observe that the set is a TC-ID set of cardinality , a contradiction. Thus, has a neighbor . Since , it must happen . Since and (otherwise there would be a triangle), we obtain a contradiction with Lemma 12 by using the vertices . As a consequence, must have degree two, and by symmetry also are of degree two too.
Suppose has degree two. Hence, for each vertex , we choose a neighbor of and observe that the set is a TC-ID set of cardinality , a contradiction. So, must have degree at least three and, by symmetry also are of degree at least three too.
We consider now a vertex . Notice that (otherwise there would be a triangle). Also, , by using the same idea as before whether has degree larger than two and . Suppose has degree larger than one and let . Since is independent and the edge must be covered by , . If and , then we obtain a contradiction with Lemma 12 by using the vertices (notice that since is independent). As a consequence, we obtain that any neighbor of is either of degree one or has a neighbor in .
We next consider the latter situation whether . Clearly . Suppose is neighbor of and of . We choose a neighbor of for every and observe that the set is a TC-ID set of cardinality , a contradiction. Thus, is a neighbor of either or , in which case, it happens has degree two. By symmetry, we obtain similar conclusions for and . That is, for any with , is given by leaves or vertices of degree two. In the latter case, if , then where .
As a consequence, we observe that can be obtained from a cycle by the Sequence II of operations described above, or equivalently . We may now consider the case in which contains a cycle , but does not contain the cycle .
Claim 3: does not contain vertex disjoint cycles.
Proof of Claim 3: We directly obtain a contradiction by Lemma 12, since in this case there are four different vertices (two of them in one cycle, the other two in the other cycle) such that a shortest path and a shortest path have length two and are vertex disjoint.
Thus, if contains more than one cycle , then they are not vertex disjoint. Moreover, we can next see that not two adjacent vertices of a cycle can be in any other cycle.
Claim 4: If two cycles of has exactly two vertices in common, then these vertices are not adjacent.
Proof of Claim 4: Suppose there are two cycles having two adjacent vertices in common. Assume the cycles are and . Hence, we note that exactly three vertices of must belong to , otherwise there are two adjacent vertices in . Indeed, such vertices are either or , say for instance . We choose a neighbor of for every and observe that the set is a TC-ID set of cardinality , which is a contradiction.
Now, according to the Claims above, if contains more than one cycle , then only the following situations can occur.
- β’
Any two cycles have exactly one vertex in common.
- β’
Any two cycles have exactly two vertices in common which are not adjacent.
- β’
Any two cycles have exactly three vertices in common.
We note that the situation in which two cycles of have exactly three vertices in common can be understood as has three cycles with two vertices in common. We now turn our attention on the following.
Claim 5: There is a vertex such that for every .
Proof of Claim 5: We first note that there are at least two vertices such that , otherwise there would be an edge not covered by . Let be a vertex adjacent to and . Suppose there is a vertex such that and (note that and ). Thus, since there are no cycles of order larger than four in , there must happen one of the following situations.
- (a)
There is a shortest path joining and not containing nor . Also, is different from the neighbor of , say , in such path. In such case, in order to cover the edge , it must happen . Thus, we obtain a contradiction by using Lemma 12 and the vertices where is a vertex at distance two from in the path.
- (b)
Without loss of generality, there is a shortest path joining and containing . Thus, there must be a vertex belonging to this path such that (it cannot be since is independent), otherwise there should be a not covered edge. Clearly . Thus, we obtain a contradiction by using Lemma 12 and the vertices .
As a consequence, the vertex has distance two to or to . Moreover, if and , then we there is a cycle of order six, which is not possible. Thus, has distance two to exactly one vertex of and . From now on, we assume .
We next prove that for any vertex , it follows too. If , then we are done. So, me may suppose there is a vertex such that (clearly ). Consider now the shortest path between and , say . Notice that . In order to cover the edge , it must be . So, we obtain a contradiction by using Lemma 12 and the vertices . Therefore, for any vertex , we obtain that and the claim is proved.
Next step gives some result on the distances between any two vertices .
Claim 6: For any two vertices , any shortest path between and passes through .
Proof of Claim 6: From Claim 5, we know that . Thus . Clearly , since cannot be adjacent. Let such that and . If or (say ), then we choose a neighbor of for every and observe that the set is a TC-ID set of cardinality , which is a contradiction. Thus, neither nor . If there is a vertex , then is a cycle in , which is not possible. Thus . By using a similar reasoning, it can be deduced that and so, . If there is another path of length four between and not containing , then we have one of the following situations.
- β’
There is a vertex such that (note that must be in in order to cover the edges , ). In such case, we obtain a contradiction by using Lemma 12 and the vertices .
- β’
There are three vertices such that , and . In such situation, is an induced cycle of order eight in , which is not possible.
- β’
Similarly to the case above, if either or , then we obtain an induced cycle of order six in , which is also not possible.
Therefore, any shortest path between and passes throughout .
We now give several facts which are consequences of the Claims above, in order to deduce the structure of the graph .
- β’
The set is independent (otherwise there is an edge not covered by ).
- β’
If , then .
- β’
If for some , then either and has degree two, or is a vertex of degree one.
- β’
If is not a vertex of degree one, then there is exactly one vertex such that (equivalently has degree two).
As a consequence of the items above, as well as from the Claims, and all the reasoning till this point, we observe that is formed by and a set of vertices (satisfying the properties above). Clearly, for any vertex , the set of its neighbors are either leaves or vertices of degree two adjacent to . Moreover, if has only one neighbor of degree two, then it must have at least one adjacent leaf (otherwise one can find a cover set of smaller cardinality). In this sense, such set of vertices can clearly be obtained from a leaf of a star by making a subdivision of the corresponding edge, an inflation of the path obtained from the subdivision and a subsequent addition of some pendant vertices. On the other hand, if has some adjacent leaves, then they could be obtained directly from a star, if subdivisions were not done to all the leaves of the star or, by a subsequent addition of leaves to the center of the original star, if all its leaves would have been subdivided. Therefore, it is then concluded that the graph was obtained from a star by making the Sequence I of operations previously described, which means and the proof is completed. β
We close this section with two bounds for in terms of order, size and minimum and maximum degrees.
Proposition 14**.**
Let be a graph of order , minimum and maximum degrees and , respectively. Then .
Proof.
Let be a -set. Hence, the subgraph induced by is edgeless. So, . Furthermore, it follows that . β
Proposition 15**.**
Let be a graph of order , size , minimum and maximum degrees and . Then .
Proof.
Let be a -set. Hence, the subgraph induced by is edgeless. So, . Now, notice that and . Adding this inequations, we have . Therefore, it follows that . β
The two bounds above are attained for instance for the double stars (each non leaf vertex is adjacent to leaves), which has order , size , minimum degree , maximum degree and .
4 The case of trees
In order to easily proceed with our exposition, and based on the following known bound, from now on we say that a tree belongs to the family , if . Moreover, we assume in this section that , since the case ( is a and is not defined) and ( is a star graph and ) are straightforward to study.
Theorem 16**.**
[15]* For any graph , .*
It is now our goal to characterize the family of trees achieving the equality in the bound above. To this end, we observe the following basic results, which can easily be obtained by using some known properties of minimum total dominating sets.
Proposition 17**.**
[6]* If is a minimal total dominating set of a connected graph , then each has at least one of the following two properties.*
- (i)
There exists a vertex such that .
- (ii)
The subgraph induced by contains an isolated vertex.
The next remark is one useful consequence of the proposition above.
Remark 18**.**
Let be a -set of cardinality . Then, for every , at least one of the following conditions is satisfied.
- (i)
There exists a vertex such that .
- (ii)
There exists a vertex such that .
We may recall to notice that condition implies that vertex is a support, because the set is independent.
Lemma 19**.**
Let and let be a -set containing no leaves. Then for every there exist a leaf such that .
Proof.
Let . Since , we consider with . Clearly, since is independent. For every , with , by Remark 18, is adjacent to a leaf or there exist a vertex such that . Hence, as , . We assume that for every , is not adjacent to a leaf , otherwise . Now, we suppose that . Also note that, by condition above, the vertices belonging to are totally dominated by other vertices of . So, we observe that the set is a total dominating set of of cardinality smaller than , a contradiction. Furthermore, there exist such that . Thus, for any , it follows , and this completes the proof. β
From this point, the set of leaves having distance three with respect to at least one other leaf is denoted by , and given a -set , we denote by the set of vertices having distance two or three to some leaf and by the set of vertices having distance three to some vertex of .
In order to provide a constructive characterization of the trees belonging to the family , we need the following five operations , , , and on a tree (by attaching a path to a vertex of we mean adding the path and joining to a vertex of ). Moreover, through all the next results we make use of the fact that any tree always contains a -set which does not contain leaves.
Operation :
Attach a path to a vertex of , which is in some -set.
Operation :
Attach a path to a vertex of , which is in .
Operation :
Attach a path to a vertex of , which is in .
Operation :
Attach a path to a vertex of , which is in .
Operation :
Attach a path to a vertex of , which is in .
Let be the family of trees defined as is obtained from by a finite sequence of operations or . The Figure 4 contains a fairly representative example of a tree . We first show that every tree of the family belongs to the family .
Lemma 20**.**
If , then .
Proof.
We proceed by induction on the number of operations required to construct the tree . If , then and . This establishes the base case. Hence, we now assume that is an integer and that each tree with satisfies that . Let be a tree for wich . Since can be obtained from a tree with by one of the operations or , we shall prove that , by considering a -set containing no leaves and through the following situations.
Case 1. is obtained from by operation . Let be the vertex added to in order to obtain . Since is a leaf of and is adjacent to a vertex of , the set remains to be a total dominating set in . Moreover, is a -set, since otherwise we would find a total dominating set in of cardinality smaller than . On the other hand, since is independent, we deduce is a TC-ID set in . Thus, (by also using the inductive hypothesis). Thus, by Theorem 16, we get the equality , which means .
Case 2. is obtained from by operation . Assume is obtained from by adding the vertex and the edge where . As , there exist a path in where is a leaf and are support vertices adjacent to , respectively. Now, in , the vertices are supports and belong to any TC-ID set in . Hence, the set is a TC-ID set in , and so
[TABLE]
(by also using Theorem 16 and the inductive hypothesis). Now, let be a -set containing no leaves. Notice that the vertex is a support and so, it belong to , also the vertex belongs to too, because has degree two. Moreover, note that the set is a total dominating set in , which leads to . By using this, it follows that all the inequalities in (1) must be equalities. Thus , and .
Case 3. is obtained from by operation . Assume is obtained from by adding the path to a vertex through the edge . By using some similar reasons as in the case above (now we must use instead of ), it is observed that .
Case 4. is obtained from by operation . Assume is obtained from by adding the path to a vertex through the edge . We notice that belong to any TC-ID set containing no leaves of . Hence, the set is a TC-ID set in . Thus (by also using Theorem 16 and the inductive hypothesis). Now, let be a -set. Since the vertex is a support, it belongs to and so, . Moreover, note that . Hence, . Again, as in Case 2, we deduce , which means .
Case 5. is obtained from by operation . Assume is obtained from by adding the path to a vertex through the edge . By using some similar reasons as in the case above, it can be deduced that , which gives .
β
We now turn our attention to the opposite direction concerning the lemma above. In this sense, from now on we shall need the following terminology and notation in our results. Given a tree and a set , by we denote a tree obtained from by removing from all the vertices in and all its incident edges (if for some vertex , then we simply write ). For an integer , by we mean a graph which is obtained from a path by attaching a path to every vertex of . In Figure 5 we show the example of .
We next show that every tree of the family belongs to the family .
Lemma 21**.**
If , then .
Proof.
We proceed by induction on the order of the trees . If is a double star, then can be obtained from by repeatedly applying operation . This establishes the base case. We assume next that is an integer and that each tree with satisfies .
Let be a tree such that and . Let be a -set containing no leaves and let . We analyze the following situations.
Case 1: . We consider a support vertex that is adjacent to at least two leaves. Let and . Thus, the set is a -set too, and by inductive hypothesis, . Therefore, since can be obtained from by operation , it follows .
Case 2: and . In this case we note that and clearly, is a -set (moreover since otherwise is a double star). Let such that (note that such always exists) and let be the leaf adjacent to . We first notice that there exists a leaf having distance three to the support . Thus, we deduce that is a -set, where . By induction hypothesis and, since can be obtained from by operation , we get .
Case 3: and . Herein we denote by the set of vertices of one shortest path between and , including and . Let be two leaves at the maximum possible distance in such that there is with or . Without loss of generality assume that and let be the support adjacent to . Since and by the maximality of the path between and , we observe that and also, that every support vertex is adjacent to exactly one leaf. We have now some possible scenarios.
Case 3.1 . Hence, by the maximality of the path , it must happen that has an induced subgraph isomorphic to a graph , as previously described, obtained from the vertices and some supports, say , with the leaves , adjacent to the supports , respectively, and such that .
Assume . Note that and that . Let . Notice that is also a TC-ID set in , and so
[TABLE]
(by using Theorem 16 and hypothesis). On the other hand, let be a -set containing no leaves. We observe that because is a support in , and because . Thus, clearly is also a total dominating set in . Hence . Thus, all the inequalities in the relation (2) must be equalities, from which follows and by the inductive hypothesis . Since can be obtained from by operation , we obtain .
Assume now . Note that and that there is a leaf at distance three from . Let . Hence, is a TC-ID set in , and so (by using Theorem 16 and hypothesis). Moreover, the set is a -set, otherwise we would find a total dominating set of of cardinality smaller than , which is not possible. So, which leads to , as in the previous case. Now, by the inductive hypothesis , and since can be obtained from by operation , we deduce .
Case 3.2 . An analogous procedure to the one above (Case 3.1) leads to our desired conclusion, based on the fact that must have at least two neighbors and there are at least two induced subgraphs isomorphic to the graphs and , which can be used instead of of Case 3.1.
Case 3.3: . Clearly, has degree two since it has one leaf neighbor, no support neighbors and cannot have more than one (it has exactly one) semi-support neighbor due to the maximality of . Also, it must happen , and . Assume the subgraph induced by is , where and . Note that . We consider again some possible scenarios.
Case 3.3.1: . In this case, the vertex is also totally dominated by another support different from . Let be the leaf adjacent to the support . Notice that is a TC-ID set of . Moreover, we note that the vertex is a leaf in having distance three to the leaf . So, by using a similar procedure as above (Case 3.1 and ) we obtain . Therefore, due to that can be obtained from by operation , it follows .
Case 3.3.2: and . Clearly have degree two and belong to . We firstly consider the case whether . By Remark 18 we note that or that there is a vertex with . If , then is a TC-ID in set, and by using a similar procedure as above (Case 3.1 and ) we obtain . Since can be obtained from by operation , we are able to claim .
On the other hand, assume that and there is a vertex such that . We note that, by Remark 18, . Hence, is a TC-ID set in . Thus, (by using Theorem 16 and hypothesis). Again, by using a similar procedure as above (Case 3.1 and ) we get that . So, , and by inductive hypothesis, . Also, it can relatively clearly be seen that is having distance three to a leaf. This means can be obtained from by operation , and so .
Now, consider the case in which . By the maximality of and by the fact that , there is a leaf distinct to at distance two or three from . Hence, the set is a TC-ID set in . Again, by using a similar procedure as above (Case 3.1 and ) we obtain and, due to that can be obtained from by operation , we get .
Case 3.3.3: and . Clearly have degree two and belong to . We only consider the case whether , otherwise implies that is a leaf and is , which can be obtained by operation from . As , we get . Notice that, as has to be totally dominated, there exist a vertex such that . So, by Remark 18 and Lemma 19, it follows .
If , then this case is analogous to the Case 3.3.2 and . If , then we see that . So, the set is a TC-ID set in and again, by using a similar procedure as above (Case 3.1 and ) we obtain . Finally, due to that can be obtained from by operation , we have , which completes the proof.
β
As an immediate consequence of Lemma 20 and Lemma 21 we have the following characterization.
Theorem 22**.**
Let be a tree. Then if and only if .
We next see that all the operations to are required in the characterization above. First, we see that operation is required to obtain a double star from the path . The operations are required to obtain the paths , respectively, from the path , and the path can only be obtained from by a sequence of operations .
Concluding remarks
We have study several combinatorial and complexity properties of the total co-independent domination number of graphs. As a consequence of the study a couple of questions could be remarked as a possible future research lines.
- β’
We have proved that computing the total co-independent domination number of graphs is NP-hard even when restricted to planar graphs of maximum degree at most 3. However, it would be interesting to find some non trivial families of graphs in which the problem above can be solved in polynomial time. On the other hand, the bounds of Theorem 8 together with the fact that the problem of computing the vertex cover number can be approximated within a factor of 2, allow to claim that the problem of computing the total co-independent domination number can be approximated within a constant factor. In this sense, it would be interesting to give some other approximation (or inapproximation) results on this parameter.
- β’
We have characterized the family of graphs achieving the upper bound of Theorem 8. According to the construction of such family, it seems one could also characterize the graphs for which for some values of like for instance or . Moreover, it would be of interest to characterize the family of graphs attaining the lower bound of Theorem 8 (note that for instance the trees satisfying such bound were characterized in [3]).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] H. Abdollahzadeh Ahangar, V. Samodivkin, and I. G. Yero, Independent transversal dominating sets in graphs: complexity and structural properties, FILOMAT 30 (2) (2016) 293β303.
- 2[2] C. Berge, Theory of graphs and its applications , Methuen, London, 1962.
- 3[3] A. Cabrera-MartΓnez, F. A. HernΓ‘ndez-Mira, J. M. Sigarreta Almira, and I. G. Yero, A note on total co-independent domination in trees. Manuscript (2017).
- 4[4] A. Cabrera MartΓnez, J. M. Sigarreta, and I. G. Yero, On the independence transversal total domination number of graphs, Discrete Applied Mathematics 219 (2017) 65β73.
- 5[5] M. Chellali and T. W. Haynes, A note on the total domination of a tree, Journal of Combinatorial Mathematics and Combinatorial Computing 58 (2006) 189β193.
- 6[6] E. J. Cockayne, R.M. Dawes, and S.T. Hedetniemi, Total domination in graphs, Networks 10 (1980) 211β219.
- 7[7] T. Gallai, Γber extreme Punkt- und Kantenmengen, Annales Universitatis Scientarium Budapestinensis de Rolando EΓΆtvΓΆs Nominatae, Sectio Mathematica 2 (1959) 133β138.
- 8[8] M. R. Garey and D. S. Johnson, Computers and Intractability: A Guide to the Theory of NP-Completeness , W. H. Freeman & Co., New York, USA, 1979.
