Abstract
A total dominating set of a graph G is a dominating set S of G such that the subgraph induced by S contains no isolated vertex, where a dominating set of G is a set of vertices of G such that each vertex in \(V(G){\setminus } S\) has a neighbor in S. A (total) dominating set S is said to be minimal if \(S{\setminus } \{v\}\) is not a (total) dominating set for every \(v\in S\). The upper total domination number \(\varGamma _t(G)\) and the upper domination number \(\varGamma (G)\) are the maximum cardinalities of a minimal total dominating set and a minimal dominating set of G, respectively. For every graph G without isolated vertices, it is known that \(\varGamma _t(G)\le 2\varGamma (G)\). The case in which \(\frac{\varGamma _t(G)}{\varGamma (G)}=2\) has been studied in Cyman et al. (Graphs Comb 34:261–276, 2018), which focused on the characterization of the connected cubic graphs and proposed one problem to be solved and two questions to be answered in terms of the value of \(\frac{\varGamma _t(G)}{\varGamma (G)}\). In this paper, we solve this problem, i.e., the characterization of the subcubic graphs G that satisfy \(\frac{\varGamma _t(G)}{\varGamma (G)}=2\), by constructing a class of subcubic graphs, which we call triangle-trees. Moreover, we show that the answers to the two questions are negative by constructing connected cubic graphs G that satisfy \(\frac{\varGamma _t(G)}{\varGamma (G)}>\frac{3}{2}\) and a class of regular non-complete graphs G that satisfy \(\frac{\varGamma _t(G)}{\varGamma (G)}=2\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The graphs considered in this paper are finite, undirected and simple. Given a graph G with vertex set V(G) and edge set E(G), the neighborhood of a vertex \(v\in V(G)\), which is denoted by \(N_G(v)\), is the set of vertices adjacent to v and \(N_G[v]=N_G(v)\cup \{v\}\). We say that vdominates vertices in \(N_G(v)\). For a subset \(S\subseteq V(G)\), let \(N_G(S)=(\cup _{v\in S} N_G(v)){\setminus } S\) and \(N_G[S]=N_G(S)\cup S\). Denote by \(d_G(v)=|N_G(v)|\) and \(d_G(S)=|N_G(S)|\) the degrees of v and S in G, respectively. A vertex of degree k is called a k-vertex. A k-regular graph is a graph that contains only k-vertices. We use \(\varDelta \) and \(\delta \) to denote the maximum degree and minimum degree of G, respectively. A graph is called cubic or subcubic if it is 3-regular or has maximum degree 3, respectively. We use G[S] to denote the subgraph induced by S. We say that GcontainsH if H is an induced subgraph of G. A vertex \(v\in V(G)\) or an edge \(e\in E(G)\) is incident with H if \(v\in V(H)\) or \(V(e)\cap V(H)\ne \emptyset \). Let \(C_n=c_1c_2\cdots c_nc_1\) be a cycle of length n. We call \(C_3\) a triangle. Two triangles are adjacent if they are incident with a common edge. Let \(K_n\) be the complete graph on n vertices. For two integers i, j such that \(i<j\), we use [i, j] to denote the set \(\{i,i+1,\ldots , j\}\).
Let S be a set of vertices of a graph G. If every vertex in \(V(G){\setminus } S\) is dominated by a vertex in S, then S is called a dominating set of G, which is abbreviated as a D-set. A D-set S of G is called a total dominating set of G, which is abbreviated as a TD-set, if G[S] contains no isolated vertex. If G has a TD-set, then G contains no isolated vertex. A D-set (TD-set) S of G is minimal if \(S{\setminus } \{v\}\) is not a D-set (TD-set) for every vertex \(v\in S\). The upper domination number, which is denoted by \(\varGamma (G)\), and the upper total domination number, which is denoted by \(\varGamma _t(G)\), are the maximum cardinalities of a minimal D-set and TD-set of G, respectively. We refer to a minimal D-set and TD-set of cardinalities \(\varGamma (G)\) and \(\varGamma _t(G)\) as a \(\varGamma (G)\)-set and \(\varGamma _t(G)\)-set, respectively.
Finding a D-set of a graph with the minimum cardinality is a classical NP-complete problem [8], which has been widely studied not only for its intractability in theory [12] but also for its applications in the field of networks [3, 9, 13]. For total domination in graphs, we refer the reader to works [2, 6, 10, 11]. Let G be a graph without isolated vertices. If \(\varGamma _t(G)=2\varGamma (G)\), then we call G an upfull graph. Recently, motivated by the result \(\frac{\varGamma _t(G)}{\varGamma (G)}\le 2\) [7], scholars [5] have turned to the study of the characterization of upfull graphs. In [5], the authors clarify the structure of connected upfull cubic graphs and propose the following problems and questions:
Problem 1
[5] Characterize the subcubic isolate-free upfull graphs G.
Question 1
[5] Is it true that if \(G \not \cong K_4\) is a connected cubic graph, then \(\frac{\varGamma _t(G)}{\varGamma (G)} \le \frac{3}{2}\; ?\)
Question 2
[5] For \(k \ge 2\), if G is a connected k-regular graph, then is it true that G is upfull if and only if \(G \cong K_{k+1}?\)
In this paper, we solve Problem 1 and show that the answers to Questions 1 and 2 are negative by constructing connected cubic graphs G such that \(\frac{\varGamma _t(G)}{\varGamma (G)}>\frac{3}{2}\) and a class of regular non-complete graphs G with \(\frac{\varGamma _t(G)}{\varGamma (G)}=2\). For technical convenience, we introduce additional preliminaries as follows:
Let S be a subset of a graph G. For each \(v\in S\), pn(v, S)= \(\{u\in V(G)|N_G(u)\cap S =\{v\}\}\) is the private neighborhood of v with respect to S, and each vertex in pn(v, S) is called a private neighbor of vwith respect to S. In particular, a vertex \(u\in \) pn(v, S) is external if \(u\notin S\), and epn(v, S) is used to denote the set of all external private neighbors of v with respect to S, while a vertex \(u\in \) pn(v, S) is internal if \(u\in S\), and ipn(v, S) is used to denote the set of all internal private neighbors of v with respect to S. It follows that pn(v, S)=epn\((v, S)\cup \) ipn(v, S) and epn\((v, S)\cap \) ipn\((v, S)=\emptyset \). Based on these notations, the minimal total dominating set has the following trivial property: A TD-set S of a graph G without isolated vertices is minimal if and only if either epn\((v,S)\ne \emptyset \) or pn\((v,S)=\)ipn(v, S)\(\ne \emptyset \) [4].
Let T be a rooted tree with root x. We say that a vertex \(y\in V(T)\) is in the \(\ell \)-level of T if the distance from x to y is \(\ell \). In a rooted tree, the parent of a vertex v is the vertex adjacent to v on the path to the root, and a child of a vertex v is a vertex for which v is the parent; every vertex except the root has a unique parent.
For a graph G and a subgraph H of G, we use \(G-V(H)\) to denote the subgraph induced by \(V(G){\setminus } V(H)\). If there is only one edge between H and \(G-V(H)\), say \(hh'\), for which \(h'\in V(H)\) and \(h\in V(G){\setminus } V(H)\), then we refer to H as a pendant subgraph of Gattached at h and we call h the attachment of H and \(h'\) a partner of h.
Here, we present some known results that relate to upfull graphs, which will be used in the proof of our results.
Theorem 1
[7] For every graph G without isolated vertices, \(\varGamma _t(G)\le 2\varGamma (G)\).
Lemma 2
[5, 7] A connected cubic graph is upfull if and only if the graph is isomorphic to \(K_4\).
Lemma 3
[5, 7] Let G be an upfull graph without isolated vertices and D an arbitrary \(\varGamma _t(G)\)-set of G. Then,
-
(a)
\(\varGamma (G)=\alpha (G)\), where \(\alpha (G)\) is the independence number of G, i.e., the number of vertices in a maximum independent set of G;
-
(b)
the subgraph induced by D is the union of disjoint copies of \(K_2\); and
-
(c)
every vertex in \(V(G){\setminus } D\), together with two adjacent vertices in D (two endpoints of a copy of \(K_2\)), forms a triangle.
2 Connected Subcubic Upfull Graphs
This section is devoted to the characterization of connected upfull subcubic graphs. By Lemma 2, it is sufficient to consider subcubic graphs with a vertex of degree less than 3. For this, we first introduce a class of subcubic graphs, namely, triangle-trees, which will be used as a technical tool to solve Problem 1.
Let T be a rooted tree with root \(v_0\) such that \(d_{T}(v_0)\le 2\) and \(\varDelta (T)\le 3\). Let \(V(T)=\{v_0,v_1, \ldots , v_n\}\). We replace each vertex \(v_i\) in T with a triangle \(T_i=x_iy_iz_ix_i\) such that for each \(v_iv_j\in E(T)\), where \(v_i\) is the parent of \(v_j\), with \(i\ne j, i,j\in [0,n]\), we add an edge between \(\{y_i,z_i\}\) and \(\{x_j\}\) such that each \(y_i\) and \(z_i\) is incident with at most one such added edge. We refer to the resulting graph as the triangle-tree of T, which is denoted by \(T_{\triangle }\), and T as the origin-tree of \(T_{\triangle }\). In addition, we call \(T_i\), for \(i=0,1,\ldots , n\), a triangle of \(T_{\triangle }\) and define \(T_0\) as the root of \(T_{\triangle }\). We say that \(T_i\) is in level\(\ell \) of \(T_{\triangle }\) if \(v_i\) is in level \(\ell \) of T. \(T_0\) is in level 0. In Fig. 1, we present an instance to illustrate the origin-tree (on the left) and its corresponding triangle-tree (on the right).
Based on triangle-tree, we introduce three classes of subcubic graphs as follows:
-
1.
\(\mathcal {T}_1\). Let \(X_1=uv\) be a copy of \(K_2\). \(\mathcal {T}_1\) consists of \(X_1\) and pendant triangle-trees attached at u or v, where the number of triangle-trees attached at each of u and v is at most two; see Fig. 2a.
-
2.
\(\mathcal {T}_2\). Let \(X_2\) be a copy of co-twin-\(C_5\) [1] with vertex set \(V(X_2)=\{u_i|i=1,2,3,4,5,6\}\) and edge set \(E(X_2)=\{u_1u_i|i=2,3,4, u_2u_3,u_2u_6, u_3u_4,u_4u_5, u_5u_6\}\). Then, \(\mathcal {T}_2\) consists of \(X_2\) and pendant triangle-trees attached at \(u_5\) and \(u_6\), where the number of triangle-trees attached at each of \(u_5\) and \(u_6\) is at most one; see Fig. 2b.
-
3.
Let \(X_3\) be a subcubic graph with vertex set \(V(X_3)=\{u_i,v_i,w_i|i=1,2,\ldots ,k\}\) and edge set \(E(X_3)=\{u_iv_i,v_iw_i,w_iu_i |i=1,\ldots ,k\} \cup \{v_1u_k, u_iv_{i+1}|i=1,\ldots , k-1\}\). Then, \(\mathcal {T}_3\) consists of \(X_3\) and pendant triangle-trees attached at \(w_i\) for \(i\in [1,k]\), where the number of triangle-trees attached at \(w_i\) is at most one; see Fig. 2c.
Let \(\mathcal {T}=\{T_{\varDelta }\cup \mathcal {T}_1 \cup \mathcal {T}_2 \cup \mathcal {T}_3\}\). Every graph G in \(\mathcal {T}\) can be partitioned into vertex-disjoint triangles, copies of \(K_2\) and co-twin-\(C_5\)s. For convenience, we call them t-units, \(k_2\)-units and co-units of G, respectively.
To solve Problem 1, we will show that a subcubic graph G is upfull if and only if \(G\in \mathcal {T}\cup \{K_4\}\). We start with the proof of the sufficiency.
Theorem 4
(Sufficiency) Every graph \(G\in \mathcal {T}\) is upfull.
Proof
Let \(\ell \) be the number of t-units of G. To prove \(\varGamma _t(G)=2\varGamma (G)\), it is sufficient to show that \(\varGamma _t(G)\ge 2\varGamma (G)\) by Theorem 1. We address the following two cases:
Case 1G contains no pendant triangle. It follows that either G is isomorphic to a triangle (which satisfies \(\varGamma _t(G)\ge 2\varGamma (G)\)) or \(G\in \mathcal {T}_1\cup \mathcal {T}_2\cup \mathcal {T}_3\). If \(G\in \mathcal {T}_1\cup \mathcal {T}_2\), then G is isomorphic to \(K_2\) or co-twin-\(C_5\), for which \(\varGamma _t(G)=2\varGamma (G)\). If \(G\in \mathcal {T}_3\), we denote \(V(G)=\cup _{i=1}^{\ell } \{u_i,v_i,w_i\}\) and \(E(G)=\{u_iv_i,v_iw_i,w_iu_i|i=1,\ldots , \ell \}\cup \{v_1u_{\ell }, u_iv_{i+1}|i=1,\ldots , \ell -1\}\). Let S be a \(\varGamma (G)\)-set of G. Since S is a minimal D-set, it follows that \(1\le |S\cap \{u_i,v_i,w_i\}|\le 2\) for every \(i\in [1,\ell ]\), and \(w_i\notin S\) when \(|S\cap \{u_i,v_i,w_i\}|=2\). Suppose that there exists \(i\in [1,\ell ]\), say \(i=1\), that satisfies \(|S\cap \{u_1,v_1,w_1\}|=2\), i.e., \(u_1,v_1\in S\). By the minimality of S, \(v_2\in \) epn\((u_1, S)\), which implies that \(\{w_2,u_2\}\cap S=\emptyset \) and \(w_2\) is not dominated by any vertex in S, which is a contradiction. Therefore, \(|S\cap \{u_i,v_i,w_i\}|=1\) for every \(i\in [1,\ell ]\), i.e., \(|S|=\ell \). Observe that \(S'=\{w_i,v_i|i=1,\ldots , \ell \}\) is a minimal TD-set of cardinality \(2\ell \). It follows that \(\varGamma _t(G)\ge |S'|=2\varGamma (G)\).
Case 2G contains pendant triangles.
Case 2.1 Every pendant triangle-tree of G is a pendant triangle. In this case, if G is a triangle-tree, then G contains two vertex-disjoint triangles and one edge between them. If \(G\in \mathcal {T}_1\), then G consists of a copy of \(K_2(=uv)\) and pendant triangles attached at u or v such that each u and v is the attachment of at most two pendant triangles. If \(G\in \mathcal {T}_2\), then G consists of a copy of co-twin-\(C_5\) and pendant triangles attached at \(u_5\) or \(u_6\) such that each \(u_5\) and \(u_6\) is the attachment of at most one pendant triangle, where \(u_5\) and \(u_6\) are the two 2-vertices of the copy of co-twin-\(C_5\) (see Fig. 2b). For all these trivial cases, we can readily check that \(\varGamma _t(G)=2\varGamma (G)\). In the following, we consider the case of \(G\in \mathcal {T}_3\). Let \(T_i=u_iv_iw_iu_i\), \(i=1,2,\ldots , k\) be the k non-pendant t-units of G, where \(\{v_1u_k, u_iv_{i+1}|i=1,2,\ldots , k-1\}\subseteq E(G)\).
Let S be a \(\varGamma (G)\)-set S of G. Then, S contains exactly one vertex of each pendant triangle (since S is minimal). Hence, any non-pendant t-unit contains at most 2 vertices of S (otherwise, if {\(u_i,v_i,w_i\)} \(\subseteq S\) for some \(i\in [1,k]\), then \(S{\setminus } \{w_i\}\) is also a D-set, which is a contradiction). If \(|S|>\ell \), then there is a non-pendant t-unit, say \(T_1\), that contains two vertices of S. Since S is minimal, it follows that \(u_1,v_1\in S\), which implies that \(v_2\in \) epn\((u_1, S)\) and \(u_k\in \) epn\((v_1, S)\). Hence, \(V(T_2)\cap S=\emptyset \) and \(V(T_k)\cap S=\emptyset \). Thus, by the connectivity among \(T_1,\ldots , T_k\), we deduce that every \(T_i\) that satisfies \(|V(T_i)\cap S|=2\) corresponds uniquely to a \(T_j\) that satisfies \(V(T_j)\cap S=\emptyset \), \(i,j\in [1,k]\) and \(i\ne j\). Therefore, \(|S\cap (\bigcup _{i=1}^k V(T_i))|\le k\). Hence, \(|S|\le k+(\ell -k)=\ell \), i.e., \(\varGamma (G)=|S|\le \ell \). In addition, let \(S'\) be the set of 2-vertices of G. Observe that each pendant triangle contains two 2-vertices and \(T_i\) contains no 2-vertices for all \(i\in [1,k]\); we have \(|S'|=2|\ell -k|\). It follows that \(S'\cup \{v_i,w_i|i=1,2,\ldots , k\}\) is a minimal TD-set of G. Therefore, \(\varGamma _t(G)\ge 2\ell \ge 2\varGamma (G)\).
Case 2.2G contains a pendant triangle-tree \(T_{\varDelta }\) that includes at least two triangles. Let a be the attachment of \(T_{\varDelta }\) and \(a'\) the partner of a. We set the triangle of \(T_{\varDelta }\) incident with \(a'\) as the root of \(T_{\varDelta }\). Let \(T'\) be a triangle of \(T_{\varDelta }\) in the maximum level. Then, \(T'\) is a pendant triangle of \(T_{\varDelta }\) (also a pendant triangle of G). Let h be the attachment of \(T'\) and \(h'\in V(T')\) the partner of h. It follows that the vertices in \(V(T'){\setminus } \{h'\}\) are 2-vertices in G and h is incident with a triangle, say \(hh_1h_2h\).
We proceed by induction on \(\ell \). Clearly, \(\ell \ge 2\). Let \(G'=G-T'\). Then, \(G'\in \mathcal {T}\), and the number of t-units in \(G'\) is \(\ell -1\). If \(G'\) contains no pendant triangle or \(G'\) contains no pendant triangle-tree with at least two triangles, then by Case 1 and Case 2.1, we have \(\varGamma _t(G')\ge 2\varGamma (G')\). If \(G'\) contains a pendant triangle-tree that includes at least two triangles, then by the hypothesis, \(\varGamma _t(G')\ge 2\varGamma (G')\). Observe that any \(\varGamma _t(G')\)-set (and \(\varGamma (G')\)-set) together with \(V(T'){\setminus } \{h'\}\) (and one vertex in \(V(T'){\setminus } \{h'\}\)) is a minimal TD-set (and a minimal D-set) of G. Therefore, \(\varGamma _t(G)\ge \varGamma _t(G')+2\) and \(\varGamma (G)\ge \varGamma (G')+1\). In the following, we will prove that \(\varGamma (G)=\varGamma (G')+1\) and hence \(\varGamma _t(G)\ge \varGamma _t(G')+2\ge 2\varGamma (G')+2=2\varGamma (G)\). It is sufficient to prove that \(\varGamma (G)\le \varGamma (G')+1\).
Let S be an arbitrary \(\varGamma (G)\)-set of G, \(S_1=S\cap V(T')\) and \(S_2=S{\setminus } V(T')\). Since S is a minimal D-set of G, \(|S_1|=1\). We suppose that \(|S_2|>\varGamma (G')\). In this case, \(S_2\) is not a D-set of \(G'\) (if \(S_2\) were a D-set of \(G'\), then \(S_2\) would be minimal since S is minimal, a contradiction). This implies that \(S_1=\{h'\}\), \(h\notin S_2\), and \(h\in \)epn\((h', S)\), i.e., \(\{h_1,h_2\}\cap S_2=\emptyset \). Observe that \(S_2\cup \{h\}\) is a D-set of \(G'\). Since \(|S_2|>\varGamma (G')\), \(S_2\cup \{h\}\) is not minimal. Therefore, there exists an \(h_3 (\ne h)\in N_{G'}(\{h_1,h_2\})\) such that \((S_2\cup \{h\}){\setminus } \{h_3\}\) is also a D-set of \(G'\). Observe that \(T'\) is in the maximum level of \(T_{\varDelta }\), which implies that \(\{h_1,h_2\}\) contains a 2-vertex or a 3-vertex that is the attachment of a pendant triangle. Since S contains exactly one vertex of each pendant triangle, we deduce that S contains only one such vertex \(h_3\), i.e., \((S_2\cup \{h\}){\setminus } \{h_3\}\) is minimal. This indicates that \(|S_2|\le \varGamma (G')\), a contradiction. Therefore, \(|S_2|\le \varGamma (G')\) and \(\varGamma (G)=|S|=|S_2|+1\le \varGamma (G')+1\). This completes the proof. \(\square \)
Now, we turn to the proof of the necessity. First, we present two lemmas, based on which we obtain Theorem 7.
Lemma 5
Let G be a connected upfull subcubic graph, \(G\ne K_4\). If G contains no co-twin-\(C_5\), then G contains no adjacent triangles.
Proof
Let D be a \(\varGamma _t(G)\)-set of G. To the contrary, we assume that G contains two adjacent triangles, say \(a_1b_1c_1a_1\) and \(a_1b_1c_2a_1\). Since \(\varDelta (G)\le 3\), by Lemma 3 (b) and (c) we have that \(\{a_1,b_1\}\subseteq D\) and \(\{c_1,c_2\}\cap D=\emptyset \). According to Lemma 3 (b), we let \(D=\{a_i, b_i|i=1,\ldots , \frac{|D|}{2}\}\), where \(a_ib_i\in E(G)\) and there is no edge between \(\{a_i,b_i\}\) and \(\{a_j,b_j\}\) for any \(i,j\in [1,\frac{|D|}{2}]\), \(i\ne j\).
We claim that both \(c_1\) and \(c_2\) are 3-vertices. Otherwise, suppose that \(c_1\) is a 2-vertex. Then, \(D'=(D{\setminus } \{a_1\})\cup \{c_1\}\) is a \(\varGamma _t(G)\)-set of G, for which \(c_2 (\notin D')\) is adjacent to only one vertex in \(D'\), which contradicts Lemma 3 (c). Now, let \(\{c'_1\}=N_G(c_1){\setminus } \{a_1,b_1\}\) and \(\{c'_2\}=N_G(c_2){\setminus } \{a_1,b_1\}\). If \(D{\setminus } \{a_1,b_1\}\) contains a subset S such that \(|S\cap \{a_i,b_i\}|=1\) for \(i=2,3, \ldots , \frac{|D|}{2}\) and \(S\cap \{c'_1,c'_2\}=\emptyset \), then \(S\cup \{c_1,c_2\}\) is an independent set of G that contains \(\frac{|D|}{2}+1\) vertices, which contradicts Lemma 3 (a). Therefore, such an S does not exist. This implies that \(c'_1\ne c'_2\), \(\{c'_1,c'_2\}\subseteq D\) and \(c'_1c'_2\in E(G)\). Then, \(G[\{a_1,b_1,c_1,c_2,c'_1, c'_2\}]\) is isomorphic to co-twin-\(C_5\), which contradicts the assumption on G. \(\square \)
Lemma 6
Let G be a connected upfull subcubic graph that contains a co-twin-\(C_5\) with vertex set \(\{u_i|i=1,2,\ldots , 6\}\), where \(u_1,u_2,u_3,u_4\) are the four 3-vertices and \(u_2u_4\notin E(G)\) (see Fig. 2b). Then, \(G'=G-\{u_1,u_2,u_3,u_4\}\) is upfull, and the restriction of any \(\varGamma _t(G)\)-set of G to \(G'\) is a \(\varGamma _t(G')\)-set.
Proof
Let D be an arbitrary \(\varGamma _t(G)\)-set of G. Then, since G is subcubic, by Lemma 3 (b) and (c), we easily deduce that \(u_1,u_3\in D\), \(u_5,u_6\in D\) and \(u_2,u_4\notin D\). \(D{\setminus } \{u_1,u_3\}\) is a minimal TD-set of \(G'\). Therefore, \(\varGamma _t(G')\ge |D|-2\).
Let \(S'\) be a \(\varGamma (G')\)-set of \(G'\). We have that \(|S'|\le \frac{|D|}{2}-1\). Otherwise, if \(|S'|\ge \frac{|D|}{2}\), then \(S'\cup \{u_1\}\) is a minimal D-set of G that contains at least \(\frac{|D|}{2}+1\) vertices. This shows that \(\varGamma (G)\ge \frac{|D|}{2}+1\). However, since G is upfull, it follows that \(\varGamma (G)=\frac{\varGamma _t(G)}{2}=\frac{|D|}{2}\), which is a contradiction. Thus, \(\varGamma (G')=|S'|\le \frac{|D|}{2}-1\), i.e., \(|D|\ge 2\varGamma (G')+2\). This shows that \(\varGamma _t(G')\ge |D|-2\ge 2\varGamma (G')\). In addition, by Theorem 1, \(\varGamma _t(G')\le 2\varGamma (G')\). Thus, \(\varGamma _t(G')= 2\varGamma (G')=|D|-2\). Hence, \(D{\setminus } \{u_1,u_3\}\) is a \(\varGamma _t\)-set of \(G'\). \(\square \)
Theorem 7
Let G be a connected upfull subcubic graph and D a \(\varGamma _t(G)\)-set of G. If D contains two adjacent vertices \(x\in D\) and \(x_0\notin D\) such that \(xx_0\) is neither an edge of a triangle nor an edge of the co-twin-\(C_5\) of G, then the component of \(G-xx_0\) that contains vertex \(x_0\) is a triangle-tree.
Proof
Let G be a counterexample with |V(G)| as small as possible. That is, the component of \(G-xx_0\) that contains vertex \(x_0\) is not a triangle-tree. Moreover, for any upfull subcubic graph H with \(|V(H)|<|V(G)|\) and any \(\varGamma _t(H)\)-set \(D'\) of H, if there are \(x\in D'\) and \(x_0\notin D'\) that satisfy \(xx_0\in E(H)\) and \(xx_0\) is neither an edge of a triangle nor an edge of the co-twin-\(C_5\) of H, then the component of \(H-xx_0\) that contains vertex \(x_0\) is a triangle-tree.
If G contains a co-twin-\(C_5\), let H be the graph obtained from G by deleting the four 3-vertices of the co-twin-\(C_5\) and their incident edges. Then, by Lemma 6, H is upfull, and \(D'=D\cap V(H)\) is a \(\varGamma _t(H)\)-set of H. Since \(x\in D'\) and \(x_0\notin D'\) satisfy the conditions of the theorem for H, by the minimality, the conclusion holds for H. Since \(xx_0\) is not an edge of the co-twin-\(C_5\) of G, it follows that the component of \(G-xx_0\) that contains vertex \(x_0\) is also a triangle-tree, which contradicts the assumption on G. In the following, we assume that G contains no co-twin-\(C_5\). Then, by Lemma 5, G contains no adjacent triangles.
Let \(G'\) be the component of \(G-xx_0\) that contains \(x_0\). By Lemma 3 (c), \(x_0\) is incident with a triangle \(T_0\) in G. Since \(xx_0\) is not an edge of a triangle of G, it follows that \(x\notin V(T_0)\). Let \(T_0=x_0y_0z_0x_0\), where \(y_0,z_0\in D\). Starting with \(T_0\) and applying Lemma 3 (b) and (c) repeatedly, we can depict the structure of \(G'\) by labeling its vertices as follows:
-
1.
Define \(T_0\) as the zero level of \(G'\).
-
2.
Let \(N_{G'}(V(T_0))\)=\(\{x_{1,1}, \ldots , x_{1,\ell _1}\}\). Since G is subcubic, we have \(\ell _1\le 2\). When \(N_{G'}(V(T_0))\)\(=\emptyset \), it follows that \(G'\) is isomorphic to a triangle, which contradicts \(G'\). When \(N_{G'}(V(T_0))\)\(\ne \emptyset \), \(N_{G'}(V(T_0))\cap D=\emptyset \). Then, every vertex \(x_{1,j}\) for \(j\in [1,\ell _1]\) is incident with a triangle \(T_{1,j}\). Let \(T_{1,j}=x_{1,j}y_{1,j}z_{1,j}x_{1,j}\), where \(y_{1,j}, z_{1,j}\in D\). Observe that since G is a subcubic graph that contains no adjacent triangle, there does not exist a \(j\in [1,\ell _1]\) such that \(\{y_{1,j}, z_{1,j}\}\cap \{y_0,z_0\}\ne \emptyset \) (i.e., \(\{y_{1,j}, z_{1,j}\}=\{y_0,z_0\})\), and \(\{y_{1,1}, z_{1,1}\} \cap \{y_{1,2}, z_{1,2}\}=\emptyset \) when \(\ell _1=2\). Let \(T_{1,j}\), \(j=1,\ldots , \ell _1\) be the first level of \(G'\).
-
3.
For \(i\ge 1\), define \( U_{i+1}=N_{G'}(\cup _{j=1}^{\ell _i} \{y_{i,j}, z_{i,j}\}) {\setminus } \cup _{j=1}^{\ell _i} \{x_{i,j}\}, \) and let \(U_{i+1} = \{ x_{i+1,1}, \ldots , x_{i+1,\ell _{i+1}} \}\). If \(U_{i+1}=\emptyset \), then we stop labeling \(V(G')\). If \(U_{i+1}\ne \emptyset \), then \(U_{i+1}\cap D=\emptyset \). Therefore, every vertex \(x_{i+1,j}\in U_{i+1}\), \(j\in [1, \ell _{i+1}]\), is incident with a triangle, which is denoted by \(T_{i+1, j}=x_{i+1,j}y_{i+1,j}z_{i+1,j}x_{i+1,j}\), where \(y_{i+1,j}, z_{i+1,j}\in D\). Let \(T_{i+1,j}\), \(j=1,\ldots , \ell _{i+1}\), be the \((i+1)\)th level of \(G'\). Observe that G is finite: There always exists an integer i such that \(U_{i+1}\) is empty.
Since \(\varDelta (G)\le 3\) and G contains no adjacent triangles, by Lemma 3 (c), every vertex \(x_{i+1,j'}\in U_{i+1}\) is adjacent to exactly one vertex in \(\{y_{i,j},z_{i,j} | j=1,2,\ldots , \ell _i\}\). This implies that each vertex in \(V(G'){\setminus } D\) uniquely corresponds to a pair of adjacent vertices in D, and they form a triangle. Therefore, every vertex in \(V(G')\) is incident with a triangle, and every vertex in \(V(G'){\setminus } D\) is adjacent to exactly three vertices of D; i.e., \(V(G'){\setminus } D\) is an independent set. In addition, \(d_{G'}(V(T_{i,j}))\le 3\) for each \(j\in [1, \ell _i]\) and \(d_{G'}(V(T_0))=2\). Therefore, we deduce that \(G'\) is a triangle-tree rooted at \(T_0\). This contradicts the assumption on G. \(\square \)
Based on the above result, we are ready to prove the necessity; prior to this, we show a corollary as follows:
Corollary 8
Let G be a connected upfull subcubic graph and D an arbitrary \(\varGamma _t(G)\)-set of G. If \(V(G){\setminus } D\) contains a 2-vertex, then G is a triangle-tree.
Proof
By Lemma 3 (b), let D=\(\{b_i,c_i|i=1,2,\ldots , \frac{|D|}{2}\}\), where \(b_ic_i\in E(G)\) for every \(i\in [1, \frac{|D|}{2}]\). Let \(a_1\) be a 2-vertex in \(V(G){\setminus } D\). Then, by Lemma 3 (b) and (c), \(a_1\) is incident with a triangle \(T'\), say \(T'=a_1b_1c_1a_1\). If \(N_G(V(T'))=\emptyset \), then G is isomorphic to a triangle, which is a triangle-tree. Suppose that \(N_G(V(T'))\ne \emptyset \). Since \(d_G(b_1)\le 3\) and \(d_G(c_1)\le 3\), it follows that \(|N_G(V(T'))|\le 2\). By Lemma 3 (b), \(N_G(V(T'))\cap D=\emptyset \).
If \(|N_G(V(T'))|=1\), let \(N_G(V(T'))=\{b'_1\}\) and assume \(b_1b'_1\in E(G)\). Then, we have that \(b'_1c_1\notin E(G)\). Otherwise, there exists a set of vertices \(S=\{s_i| s_i\in \{b_i,c_i\}, i=2,3,\ldots , \frac{|D|}{2}\}\) such that \(S\cap N_G(b'_1)=\emptyset \); hence, \(S\cup \{a_1, b'_1\}\) is an independent set of G that contains \(\frac{|D|}{2}+1\) vertices, which contradicts Lemma 3 (a). Therefore, \(bb'_1\) is not an edge of a triangle or a co-twin-\(C_5\) of G. By Theorem 7, \(G-T'\) is a triangle-tree. Hence, G is a triangle-tree.
If \(|N_G(V(T'))|=2\), let \(N_G(V(T'))=\{b'_1, c'_1\}\), where \(b_1b'_1\in E(G)\) and \(c_1c'_1\in E(G)\). In this case, neither \(b_1b'_1\) nor \(c_1c'_1\) is an edge of a triangle or a co-twin-\(C_5\) of G. Therefore, by Theorem 7, \(G-T'\) is the union of two vertex-disjoint triangle-trees. Hence, G is a triangle-tree. \(\square \)
Theorem 9
(Necessity) Let G be a connected upfull subcubic graph. If \(G\ne K_4\), then \(G\in \mathcal {T}\).
Proof
By Lemma 2, \(\delta (G)\le 2\). Let D be a \(\varGamma _t(G)\)-set of G. Then, by Lemma 3 (b), we let \(D=\{b_i, c_i| i=1,2,\ldots , p\}\), where \(b_ic_i\in E(G)\), \(p=\frac{|D|}{2}\) and there does not exist any edge between \(\{b_{i_1}, c_{i_1}\}\) and \(\{b_{i_2}, c_{i_2}\}\) for any \(i_1,i_2\in [1,p], i_1\ne i_2\). Let \((V(G){\setminus } D)=\{a_j|j=1,2,\ldots , |V(G)|-2p\}\). By Lemma 3 (c), for every \(j\in [1, |V(G)|-2p]\), there is an \(i\in [1,p]\) such that \(a_jb_ic_ia_j\) is a triangle. For any distinct \(j_1,j_2\in [1, |V(G)|-2p]\), \(a_{j_1}\) and \(a_{j_2}\) are not incident with the identical triangle.
If D contains a vertex, say \(b_1\), that is not incident with a triangle, then \(N_G(\{b_1, c_1\})\cap D=\emptyset \). By Theorem 7, for every vertex \(a\in N_G(\{b_1, c_1\})\), say \(ab_1\in E(G)\), if \(ab_1\) is not an edge of a co-twin-\(C_5\) of G, then the component of \(G-ab_1\) that contains a is a triangle-tree. Otherwise, \(b_1c_1\) is also an edge of a co-twin-\(C_5\) of G. Hence, \(G-\{b_1,c_1\}\) is the union of vertex-disjoint triangle-trees and two adjacent triangles. This implies that either \(G\in \mathcal {T}_1\) or \(G\in \mathcal {T}_2\) (when \(b_1c_1\) is an edge of a co-twin-\(C_5\) of G).
In what follows, we assume that all vertices in D are incident with triangles. Then, \(\delta (G)\ge 2\). If G has a 2-vertex not in D, then by Corollary 8, G is a triangle-tree. Now, we address the case in which all 2-vertices are in D, that is, \(d_G(a_i)=3\) for all \(i\in [1,|V(G)|-2p]\).
Since \(\delta (G)\le 2\), without loss of generality, we let \(c_1\) be a 2-vertex. Since \(d_{G}(a_1)=3\), \(a_1\) is adjacent to a triangle, say \(T_2=a_2b_2c_2a_2\). If \(a_1a_2\in E(G)\), then \((D{\setminus } \{c_1\})\cup \{a_1\}\) is also a \(\varGamma _t(G)\)-set that does not contain \(c_1\) and G is a triangle-tree by Corollary 8. Therefore, we assume that \(a_1b_2\in E(G)\). Then, \(a_2\) is also adjacent to a triangle of G, say \(T_3\). In this way, we obtain that each \(a_i\), for \(i\ge 1\), is adjacent to a triangle \(T_{i+1}\) of G such that \(a_ia_{i+1}\notin E(G)\) (otherwise, \((D{\setminus } \{c_1,b_2, b_3,\ldots , b_{i}\})\cup \{a_1,a_2, \ldots , a_i\}\) is a \(\varGamma _t(G)\)-set of G that does not contain \(c_1\) and G is a triangle-tree by Corollary 8) and \(a_ib_{i+1}\in E(G)\).
Note that G is finite. There exists an integer k such that \(a_k\) is adjacent to a vertex in \(\{b_{i},c_{i}|i=1,2,\ldots , k-1\}\). Here, if necessary, renaming the triangle \(T_1, T_2,\ldots , T_k\), we may assume that \(a_k\) is adjacent to \(b_1\). Thus, by Theorem 7, if \(d_G(c_i)=3\), then \(c_i\) is adjacent to a triangle-tree for every \(i\in [1,k]\); see Fig. 2c.This shows that \(G\in \mathcal {T}_3\). \(\square \)
By Theorems 4 and 9 and Lemma 2, we characterize the upfull subcubic graphs by the following theorem:
Theorem 10
Let G be a connected subcubic graph. Then, G is upfull if and only if \(G\in \{K_4\}\cup \mathcal {T}\).
Observe that when a graph G is disconnected, we have that \(\varGamma _t(G)=\sum _{i=1}^{\ell } \varGamma _t(G_i)\) and \(\varGamma (G)=\sum _{i=1}^{\ell } \varGamma (G_i)\), where \(G_1,\ldots , G_{\ell }\) are the \(\ell \) components of G. By Theorem 1, \(\varGamma _t(G_i)\le 2\varGamma (G_i)\) for any \(i\in [1,\ell ]\). Therefore, \(\varGamma _t(G)=2\varGamma (G)\) if and only if \(\varGamma _t(G_i)=2\varGamma (G_i)\) for any \(i\in [1,\ell ]\). Thus, according to Theorem 10, we have the following conclusion:
Theorem 11
Let G be a subcubic graph. Then, G is upfull if and only if each component of G belongs to \(\{K_4\}\cup \mathcal {T}\).
3 Answers to Questions 1 and 2
3.1 Question 1
We consider the graph \(G_{12}\) in Fig. 3a, which is a cubic graph of order 12 with vertex set \(V(G_{12})=\{v_i|i=1,2,\ldots , 12\}\) and edge set \(E(G_{12})=\{v_iv_{i+1}|i\in [1,10], v_1v_5, v_1v_{11}, v_2v_{12},v_3v_{12}, v_4v_6,\)\(v_7v_{12}, v_8v_{10},\)\( v_9v_{11}\}\).
Let S be a \(\varGamma (G_{12})\)-set of \(G_{12}\). Let \(S_1= S\cap \{v_1, v_7, v_8,v_9,v_{10},v_{11}\}\), \(S_2=S\cap \{v_4,v_5,v_6\}\), and \(S_3=S\cap \{v_2,v_3,v_{12}\}\). One can readily check that \(1\le |S_1|\le 3\) and \(|S_i|\le 2\) for \(i=2,3\). (Otherwise, by symmetry, we assume \(|S_2|=3\). Then, \(v_3\in \) epn(\(v_4\), S) and \(v_7\in \) epn(\(v_6\), S). This implies that \(\{v_2,v_3,v_{12},v_7\}\cap S=\emptyset \) and \(v_{12}\) is not dominated by any vertex in S, which is a contradiction.) In particular, \(|S_2|+|S_3|\le 3\). Otherwise, if \(|S_2|+|S_3|=4\), then \(|S_2|=|S_3|=2\), \(S_2=\{v_5,v_6\}\) and \(S_3=\{v_2,v_{12}\}\) (observe that if \(v_4\in S_2\) or \(v_3\in S_3\), then \(v_3\in \) epn(\(v_4\), S) or \(v_4\in \) epn(\(v_3\), S), which implies that \(|S_3|=0\) or \(|S_2|=0\)). Then, \(S{\setminus } \{v_{12}, v_5\}\) is also a D-set of \(G_{12}\), which is a contradiction.
Now, by symmetry, if \(|S_2|+|S_3|=3\), we assume that \(S_2\cup S_3=\{ v_5,v_6, v_3\}\). Then, \(v_1\in \) epn \((v_5, S)\) and \(v_7\in \) epn \((v_6, S)\), which shows that \(|S_1\cap S|=1\), i.e., \(v_9\in S_1\) or \(v_{10}\in S_1\), and \(|S|=4\). If \(|S_2|+|S_3|=2\), then \(|S_1|\le 2\) (since when \(|S_1|=3\), \(\{v_1,v_7\}\subset S_1\)). If \(v_5\in S_2\), then epn\((v_1, S)\ne \emptyset \) and epn\((v_5, S)\ne \emptyset \). Clearly, \(v_4,v_6\notin S_2\). If \(v_2\in \) epn\((v_1, S)\), then \(|S_3|=0\) and \(|S_2|+|S_3|=1\); if \(v_{11}\in \) epn\((v_1, S)\), then \(v_8\in S_1\) and \(v_{12}\in \) epn \((v_7,S)\), which also implies that \(|S_3|=0\) and \(|S_2|+|S_3|=1\). Now, we suppose that \(\{v_5,v_6,v_2,v_{12}\}\cap S=\emptyset \) (the cases in which \(v_2\in S_3\), \(v_6\in S_2\) or \(v_{12}\in S_3\) are analogous to the case of \(v_5\in S_2\)). Since only \(v_3\) and \(v_4\) are not dominated by vertices in \(S_1\), we observe that \(|S\cap \{v_3,v_4\}|=1\), i.e., \(|S_2|+|S_3|=1\). Thus, \(|S|\le 4\). If \(|S_2|+|S_3|=1\), then \(|S|\le 4\). Therefore, \(\varGamma (G_{12})\le 4\) (\(\varGamma (G_{12})=4\) since \(\{v_1,v_9,v_{7}, v_4\}\) is a minimal D-set of \(G_{12}\)).
In addition, \(\{v_1,v_4,v_5,v_6,v_7, v_9,v_{10}\}\) is a minimal TD-set of \(G_{12}\). Therefore, \(\varGamma _t(G)\ge 7\) and \(\frac{\varGamma _t(G_{12})}{\varGamma (G_{12})}\ge \frac{7}{4}>\frac{3}{2}\). This indicates that the answer to Question 1 is negative.
Observe that there are many graphs G for which \(\frac{\varGamma _t(G)}{\varGamma (G)}>\frac{3}{2}\). Here, we present two additional examples, which are denoted by \(G_{14}\) and \(G_{18}\), as shown in Fig. 3b, c. With a similar proof (we omit the details here), we can show that \(\varGamma (G_{14})=5\) and \(\varGamma _t(G_{14})\ge 8\) and that \(\varGamma (G_{18})=6\) and \(\varGamma _t(G_{18})\ge 10\) (the set of white vertices in each graph is a minimal TD-set of the graphs). Thus, \(\frac{\varGamma _t(G_{14})}{\varGamma (G_{14})}\ge \frac{8}{5}>\frac{3}{2}\) and \(\frac{\varGamma _t(G_{18})}{\varGamma (G_{18})}\ge \frac{10}{6}>\frac{3}{2}\). Based on these observations, we propose the following conjecture:
Conjecture 1
Let \(G \not \cong K_4\) be a connected cubic graph. Then, \(\frac{\varGamma _t(G)}{\varGamma (G)} \le \frac{7}{4}.\)
3.2 Question 2
We construct a class of graphs as follows: Let n, k be two positive integers that satisfy \(k\le \frac{n}{2}\). Let D(n, k) be the graph with vertex set \(\{v_0,v_1,\ldots ,v_{n-1}\}\) and edge set \(\{v_iv_{i+j}, v_iv_{i-j}|i=0,1,\ldots , n,j=1,2,\ldots , k\}\), where the subscripts read module n. D(n, k) is a 2k-regular graph. The graphs in Fig. 4a, b are D(8, 2) and D(10, 3), respectively.
Theorem 12
Let n be a positive integer such that \(n\ge 8\). Then, \(\varGamma _t(D(n, \frac{n}{2}-2))=4\) and \(\varGamma (D(n, \frac{n}{2}-2))=2\).
Proof
Let \(k=\frac{n}{2}\). By the definition of D(n, k), for \(i=0,1,\ldots , n-1\), we have that
Then, \(V(D(n, k)){\setminus } N_{D(n, k)}[v_i]= \{v_{i+k-1}, v_{i+k}, v_{i+k+1}\}\) since \((i+k+2)\pmod n =(i-k+2)\pmod n\). For each \([0,k-1]\), let \(M_i=\{v_iv_{i+1}, v_{i+k}v_{i+1+k}\}\). Each vertex of D(n, k) is adjacent to a vertex in \(V(M_i)\). Therefore, for every \(i\in [0,n-1]\), \(V(M_i)\) is a minimal TD-set of \(D(n, \frac{n}{2}-2)\) and \(v_i, v_{i+k}\) is a minimal dominating set of \(D(n, \frac{n}{2}-2)\), which shows that \(\varGamma _t(D(n, \frac{n}{2}-2))\ge 4\) and \(\varGamma (D(n, \frac{n}{2}-2))\ge 2\).
Conversely, suppose that \(D(n, \frac{n}{2}-2)\) has a minimal dominating set D such that \(|D|=3\). Let \(D=\{v_i,v_j,v_p\}\), where \(i<j<p\). Then, the three vertices not dominated by \(v_i\) are dominated by \(v_j\) or \(v_p\). If \(j>i+2 \pmod n\) or \(p<i-2\pmod n\), then \(\{v_i,v_j\}\) or \(\{v_i,v_p\}\) is a dominating set of \(D(n, \frac{n}{2}-2)\)j; if \(j\le i+2 \pmod n\) or \(p\ge i-2\pmod n\), then \(\{v_j,v_p\}\) is a dominating set. This contradicts the assumption that D is minimal. Therefore, \(\varGamma (D(n, \frac{n}{2}-2))\le 2\). Hence, \(\varGamma (D(n, \frac{n}{2}-2))=2\). By Lemma 1, \(\varGamma _t(D(n, \frac{n}{2}-2))\le 2\varGamma (D(n, \frac{n}{2}-2))=4\). Thus,\(\varGamma _t(D(n, \frac{n}{2}-2))=4\). \(\square \)
By Theorem 12, \(D(n, \frac{n}{2}-2)\), for \(n\ge 8\), is an upfull \((n-4)\)-regular graph, which is not complete. This shows that the answer to Question 2 is also negative.
References
Brandstädt, A., Le, V.B., Spinrad, J.P.: Graph Classes: A Survey. Society for Industrial and Applied Mathematics, Philadelphia (1999)
Brešar, B., Henning, M.A., Rall, D.F.: Total domination in graphs. Discrete Math. 339, 1665–1676 (2016)
Chen, J., He, K., Du, R., Zheng, M., Xiang, Y., Yuan, Q.: Dominating set and network coding-based routing in wireless mesh networks. IEEE Trans. Parallel Distrib. Syst. 26(2), 423–433 (2015)
Cockayne, E.J., Dawes, R.M., Hedetniemi, S.T.: Total domination in graphs. Networks 10, 211–219 (1980)
Cyman, J., Dettlaff, M., Henning, M.A., Lemańska, M., Raczek, J.: Total domination versus domination in cubic graphs. Graphs Comb. 34, 261–276 (2018)
Desormeaux, W.J., Haynes, T.W., Henning, M.A., Yeo, A.: Total domination numbers of graphs with diameter two. J. Graph Theory 75, 91–103 (2014)
Dorbec, P., Henning, M.A., Rall, D.F.: On the upper total domination number of cartesian products of graphs. J. Comb. Optim. 16, 68–80 (2008)
Garey, M.R., Johnson, D.S.: Computers and Intractability, A Guide to the Theory of NP Completeness. W H Freeman and Company, San Francisco (1979)
Haynes, T.W., Hedetniemi, S.T., Slater, P.J.: Domination in Graphs: Advanced Topics. Marcel Dekker, New York (1998)
Henning, M.A.: A survey of selected recent results on total domination in graphs. Discrete Math. 309(1), 32–63 (2009)
Henning, M.A., Yeo, A.: A new lower bound for the total domination number in graphs proving a graffiti conjecture. Discrete Appl. Math. 173, 45–52 (2014)
Molnár, F., Sreenivasan, S., Szymanski, B.K., Korniss, G.: Minimum dominating sets in scale-free network ensembles. Sci. Rep. 3, 1736 (2013)
Wuchty, S.: Controllability in protein interaction networks. Proc. Natl. Acad. Sci. USA PNAS 111(19), 7156–7160 (2014)
Acknowledgements
This work was supported by the National Natural Science Foundation of China (61872101, 61672051, 61702075), the China Postdoctoral Science Foundation under Grant (2017M611223).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Zhu, E., Liu, C., Deng, F. et al. On Upper Total Domination Versus Upper Domination in Graphs. Graphs and Combinatorics 35, 767–778 (2019). https://doi.org/10.1007/s00373-019-02029-y
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00373-019-02029-y