Abstract
For an ordered non-empty subset \( S=\{v_1,\ldots , v_k\}\) of vertices in a connected graph G and an l-clique \(V'\) of G, the l-clique metric S-representation of \(V'\) is the vector \(r^l_G(V'|S) = (d_G(V',v_1), \ldots , d_G(V',v_k))\,\) where \(d_G(V',v_i)=\min \{d_G(v,v_i): v\in V'\}\). A non-empty subset S of V(G) is an l-clique metric generator for G if all l-cliques of G have pairwise different l-clique metric S-representations. An l-clique metric generator of smallest order is an l-clique metric basis for G, its order being the l-clique metric dimension (l-CMD for short) \(\mathrm{cdim}_l(G)\) of G. In this paper, we propose this concept as an extension of the 1-clique metric dimension which is known as the metric dimension, and also study some its properties. Moreover, l-CMD for \(\Gamma ({\mathbb {Z}}_n)\) and the corona product of two graphs is investigated. Furthermore, we prove that computing the l-CMD of connected graphs is NP-hard and present an integer linear programming model for finding this parameter.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
All graphs considered in this paper are undirected and simple.
If \(u, v\in V(G)\), then \(d_G(u,v)\) denotes the number of edges on a shortest u, v-path in G. A clique \(V'\) is a subset of vertices of a graph such that every two distinct vertices in the clique are adjacent. Also, \(V'\) is called an l-clique if \(|V'|=l\). For a vertex u and an l-clique \(V'\) of G, the distance between \(V'\) and u, denoted by \(d_G(u,V')\), is defined as \(\min \{d_G(u,v): v\in V'\}\); in other words, \(d_G(u,V')=\min \{d_G(u,v): v\in V'\}\).
For an ordered non-empty subset \(S=\{v_1,\ldots , v_k\}\) of vertices in a connected graph G and an l-clique \(V'\) of G, the l -clique metric S-representation of \(V'\) is the vector \(r^l_G(V'|S) = (d_G(V',v_1), \ldots , d_G(V',v_k))\). A non-empty subset S of V(G) is an l-clique metric generator for G if all l-cliques of G have pairwise different l-clique metric S-representations. l-Clique metric generators for special cases \(l=1\) and \(l=2\) are known as metric generator and edge metric generator, respectively. An l-clique metric generator of smallest order is an l-clique metric basis for G, its order being the l-clique metric dimension (l-CMD for short) \(\mathrm{cdim}_l(G)\) of G.
Recall that the special case 1-clique metric dimension is called the metric dimension and denoted by \(\mathrm{dim}(G)\) and also the special case 2-clique metric dimension is called the edge metric dimension and denoted by \(\mathrm{dim}_e(G)\).
The concept of metric dimension was first introduced by Slater [21]. Since then lots of work has been done on this topic because of its wide range of applications in modeling of real world problems [13, 15]. For instance, Garey and Johnson [11], and Epstein et al. [10] studied NP-hardness of computing of metric dimension. Also, this invariant was investigated over the Cartesian product of graphs in [5], over the lexicographic product of graphs in [19], over the deleted lexicographic product of graphs in [9], and over the hierarchical product of graphs in [23]. Kelenc et al. [14] introduced the concept of edge metric dimension. In the present work, we expand the concept of metric dimension as l-clique metric dimension where l is a natural number. Note that in [12] resolving sets locate up to some fixed l, \(l\ge 1\), vertices in a graph, while here resolving sets locate the l-cliques of a graph. The first section of this paper is dedicated to some properties of this parameter of graphs. In the second section, we compute l-CMD for \(\Gamma ({\mathbb {Z}}_n)\). We also obtain the exact value of l-CMD of corona product of two graphs in the third section. [11, 14] showed the NP-completeness of l-CMD problems for \(l=1\) and \(l=2\), respectively. We prove the NP-completeness of l-CMD problems for \(l\ge 3\) in the last section.
Throughout this paper, our notation is standard and taken mainly from [2].
2 Basic Results
In this section, we present some basic results on the l-clique metric dimension.
The following proposition gives the l-CMD of the complete graph \(K_n\).
Proposition 2.1
Let \(n \geqslant 2\). We have
Proof
If \(l=n\), then clearly \(\mathrm{cdim}_l(K_n)=1\). Let \(l \in \{1,2\}\). Then, by [14, Remark 1], we have \(\mathrm {edim}(K_n)=\mathrm {dim}(K_n)=n-1\). Hence, in this situation, \(\mathrm{cdim}_l(K_n)=\mathrm {edim}(K_n)=\mathrm {dim}(K_n)=n-1\). So we assume that \(3 \leqslant l \leqslant n-1\) and \(n \geqslant 4\). Let S be a minimal l-clique metric generator of \(K_n\). If \(|S|\leqslant n-2\), then there exist two distinct vertices \(x,y \in K_n {\setminus } S\). Consider two l-cliques \(L_1\) and \(L_2\) such that \(x\in L_1\), \(y \in L_2\) and \(L_1 {\setminus } \{x\}=L_2{\setminus } \{y\}\). Then one can see that the l-clique metric S-representations of \(L_1\) and \(L_2\) are the same, which is impossible. Now, let \(S\subseteq V(K_n)\) with \(|S|=n-1\). Then, in this situation, for every two distinct cliques \(L_1\) and \(L_2\), there exists \(s \in S\) such that \(s \in L_1 {\setminus } L_2\). Therefore the component which is corresponding to s in the l-clique metric S-representations of \(L_1\) and \(L_2\) is 0 and 1, respectively, which implies that S is an l-clique metric generator for \(K_n\). Hence \(\mathrm{cdim}_l(K_n)=n-1\). \(\square \)
Recall that the wheel graph \(W_{1,n}\) is the graph obtained from a cycle \(C_n\) and the graph \(K_1\) by adding all the edges between the vertex of \(K_1\) and every vertex of \(C_n\).
The least integer greater than or equal to a number m is denoted by \(\lceil m \rceil \). Also, greatest integer less than or equal to a number m is denoted by \(\lfloor m \rfloor \).
In the following proposition, we investigate the l-CMD \(\mathrm{cdim}_l(W_{1,n})\). Note that if \(l=1\), then \(\mathrm{cdim}_l(W_{1,n})=\mathrm {dim}(W_{1,n})\), which is determined in [3], as follows.
Also, if \(l=2\), then \(\mathrm{cdim}_l(W_{1,n})=\mathrm {edim}(W_{1,n})\), which is
see [14].
Proposition 2.2
Let \(W_{1,n}\) be a wheel graph. Then
Proof
By Proposition 2.1, we have \(\mathrm{cdim}_3(W_{1,3}) =\mathrm{cdim}_3(K_4)=3\) and \(\mathrm{cdim}_4(W_{1,3}) =\mathrm{cdim}_4(K_4)=1\). So assume that \(n\geqslant 4\). Let \(\{g_1,g_2, \dots , g_n\}\) be the vertices of degree 3 in \(W_{1,n}\). Clearly for each two distinct triangles \(L_1\) and \(L_2\) in \(W_{1,n}\), either there exists \(1 \leqslant i \leqslant n\) such that \(L_1\) and \(L_2\) have the common vertex \(g_i\), or \(L_1\) and \(L_2\) have no common vertices from the set \(\{g_1,g_2, \dots , g_n\}\). In both of the situations, one can easily see that \(L_1\) and \(L_2\) have the same 3-clique metric S-representations if and only if their non-common vertices do not belong to S, where \(S \subseteq V(W_{1,n})\). Now let S be a 3-clique metric basis of \(W_{1,n}\). Clearly \(S \subseteq \{g_1,g_2, \dots , g_n\}\). We consider the following cases.
Case 1 \(n=3k\), where \(k \geqslant 2\). Let S be a 3-clique metric basis of \(W_{1,n}\). If there are two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\), say \(g_2\) and \(g_3\), such that \(g_2,g_3 \notin S\), then we should have \(g_4,g_5\in S\) and \(g_{3\lfloor \frac{n}{3}\rfloor }, g_1 \in S\). So, if \(k=2\), then \(\mathrm{cdim}_3(W_{1,6}) =4\). Let \(k>2\). Since S is a 3-clique metric basis, without loss of generality, we may assume that \(g_6 \notin S , g_7,g_8\in S, \dots ,g_{3\lfloor \frac{n}{3}\rfloor -3}\notin S, g_{3\lfloor \frac{n}{3}\rfloor -2}, g_{3\lfloor \frac{n}{3}\rfloor -1} \in S\). Therefore, in this situation, \(|S|=n-\lceil \frac{n}{3}\rceil \).
Now, assume that there exists a 3-clique metric basis of \(W_{1,n}\), say S, such that for any two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) at least one of them belongs to S. Without loss of generality, assume that \(g_3 \notin S\). Since S is a 3-clique metric basis, we may assume that
where \(1 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor \). Clearly, in this situation we again have \(|S|=n-\lceil \frac{n}{3}\rceil \).
Note that in either of the above situationes, by the structure that we obtain for a 3-clique metric basis of \(W_{1,n}\), it is easy to see that any subset of \(\{g_1,g_2, \dots , g_n\}\) with less that \(n-\lceil \frac{n}{3}\rceil \) elements is not a 3-clique metric generator of \(W_{1,n}\). Therefore, in this case the 3-CMD of \(W_{1,n}\) is equal to \(n-\lceil \frac{n}{3}\rceil \).
Case 2 \(n=3k+1\) or \(n=3k+2\), where \(k \geqslant 1\). First we show that for any 3-clique metric basis of \(W_{1,n}\), say S, there exist two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) such that they do not belong to S. Assume on the contrary that for any two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\), at least one of them belongs to S. Without loss of generality, we may assume that \(g_3 \notin S\). Since S is a 3-clique metric basis, we may assume that
where \(1 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor \). Now consider the set \(S'=S {\setminus } \{g_2\}\). One can easily see that \(S'\) is a 3-clique metric generator of \(W_{1,n}\) with \(|S'|<|S|\), which is a contradiction.
Now let S be a 3-clique metric basis of \(W_{1,n}\). Then there are two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\), say \(g_2\) and \(g_3\), such that \(g_2,g_3 \notin S\). By using a similar discussion as we used in Case 1, we obtain that
where \(1 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor \) and \(|S|=n-\lceil \frac{n}{3}\rceil \). Also, by the structure that we obtain for S, it is easy to see that any subset of \(\{g_1,g_2, \dots , g_n\}\) with less that \(n-\lceil \frac{n}{3}\rceil \) elements, is not a 3-clique metric generator of \(W_{1,n}\).
Therefore we have \(\mathrm{cdim}_3(W_{1,n}) =n-\lceil \frac{n}{3}\rceil \), when \(n \geqslant 4\). \(\square \)
Similarly to the wheel graph, the fan graph, which is denoted by \(F_{1,n}\), is the graph that is obtained from a path \(P_n\) and the graph \(K_1\) by adding all the edges between the vertex of \(K_1\) and every vertex of \(P_n\). In [4, 14], \(\mathrm {dim}(F_{1,n})\) and \(\mathrm {edim}(F_{1,n})\) are determined as follows:
and
In the following proposition, we investigate the l-CMD of \(F_{1,n}\) in the case that \(l=3\).
Proposition 2.3
For the fan graph \(F_{1,n}\) we have
Proof
Clearly if \(n\in \{1,2,3\}\), we have \(\mathrm{cdim}_3(F_{1,n}) =1\). Let \(\{g_1, g_2, \dots , g_n\}\) be the vertices of the path \(P_n\) in the structure of \(F_{1,n}\). Note that for each two distinct triangles \(L_1\) and \(L_2\) in \(F_{1,n}\), they have the same 3-clique metric S-representations if and only if their non-common vertices do not belong to S, where \(S\subseteq V(F_{1,n})\). Also clearly each 3-clique metric basis of \(F_{1,n}\) is a subset of \(\{g_1, g_2, \dots , g_n\}\). Now we have the following cases:
Case 1 \(n=3k\), where \(k \geqslant 2\). First we show that for any 3-clique metric basis of \(F_{1,n}\), say S, there exist two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) such that they do not belong to S. Assume on the contrary that for any two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\), at least one of them belongs to S. If \(g_1 \notin S\), then by using a similar method as we used in the proof of Proposition 2.2, we get that
where \(0 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor -1\). But one can easily see that the set \(S'=S {\setminus } \{g_{3\lfloor \frac{n}{3}\rfloor -1}\}\) is a 3-clique metric generator of \(F_{1,n}\) with \(|S'|<|S|\), which is a contradiction. Now, let \(g_1 \in S\). Then we may assume that
where \(0 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor -1\). Again we see that the set \(S'=S {\setminus } \{g_{3\lfloor \frac{n}{3}\rfloor }\}\) is a 3-clique metric generator of \(F_{1,n}\) with \(|S'|<|S|\), which is a contradiction. Therefore for any 3-clique metric basis of \(F_{1,n}\), say S, there exist two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) such that they do not belong to S. Now it is easy to see that
where \(0 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor -1\) is a 3-clique metric generator of \(F_{1,n}\), and any subset of \(\{g_1,g_2, \dots , g_n\}\) with cardinality less than \(|S|=n-\lceil \frac{n}{3}\rceil -1\) is not a 3-clique metric generator for \(F_{1,n}\). Hence in this case we have \(\mathrm{cdim}_3(F_{1,n}) =n-\lceil \frac{n}{3}\rceil -1 \).
Case 2 \(n=3k+1\), where \(k \geqslant 1\). Let S be a 3-clique metric basis. First assume that for any two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\), at least one of them belongs to S. If \(g_1 \in S\), then if \(g_2 \notin S\), then \(S {\setminus } \{g_1\}\) is a 3-clique metric generator with less than |S| elements which is impossible. Also if \(g_2 \in S\), then \(S {\setminus } \{g_2\}\) is a 3-clique metric generator with less than |S| elements which is again impossible. So we have \(g_1 \notin S\). In this situation, one can easily see that
where \(0 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor \) is a 3-clique metric basis for \(F_{1,n}\), with \(|S|=n-\lceil \frac{n}{3}\rceil \). Now, suppose that there exist two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) such that they do not belong to S. In this situation, we again have \(|S|=n-\lceil \frac{n}{3}\rceil \). Therefore in this case we have \(\mathrm{cdim}_3(F_{1,n}) =n-\lceil \frac{n}{3}\rceil \).
Case 3 \(n=3k+2\), where \(k \geqslant 1\). Similar to Case 1, we can see that for any 3-clique metric basis of \(F_{1,n}\), say S, there exist two adjacent vertices of the set \(\{g_1,g_2, \dots , g_n\}\) such that they do not belong to S. Now one can easily see that
where \(0 \leqslant i \leqslant \lfloor \frac{n}{3}\rfloor \), is a 3-clique metric generator of \(F_{1,n}\), and any subset of \(\{g_1,g_2, \dots , g_n\}\) with cardinality less than \(|S|=n-\lceil \frac{n}{3}\rceil -1\) is not a 3-clique metric generator for \(F_{1,n}\). Hence in this case we have \(\mathrm{cdim}_3(F_{1,n}) =n-\lceil \frac{n}{3}\rceil -1\). \(\square \)
Proposition 2.4
Let G be a graph with n vertices such that the number of its l-cliques are t. Then if \(t \geqslant 2\), we have \(\mathrm{cdim}_l(G) \leqslant \mathrm {min}\{n, \begin{pmatrix} t \\ 2 \end{pmatrix}\}.\) Otherwise \(\mathrm{cdim}_l(G) =1\).
Proof
If \(l=1\) or \(t \leqslant 1\), then clearly we are done. So assume that \(l \geqslant 2\). Let \(L_1,L_2, \dots ,L_t\) be the l-cliques of G. For each \(1 \leqslant i < j \leqslant n\), consider a vertex \(x_{i,j}\) which belongs to \(L_i {\setminus } L_j\). Let \(S=\{x_{i,j} \mid 1 \leqslant i < j \leqslant n\}\). Now one can see that S is an l-clique metric generator for G and \(|S| \leqslant \begin{pmatrix} t \\ 2 \end{pmatrix}\). Hence the result holds. \(\square \)
The next corollary follows from Proposition 2.4.
Corollary 2.5
Let G be a graph with at most two l-cliques. Then \(\mathrm{cdim}_l(G) =1\).
Proposition 2.6
Let G be a graph with n vertices and \(L_1,L_2, \dots ,L_t\) be the l-cliques of G such that \(L_i \nsubseteq \bigcup _{i \ne j, j=1}^t L_j\), for \(1 \leqslant i \leqslant t-1\). Then \(\mathrm{cdim}_l(G) \leqslant t-1\).
Proof
Let \(x_i \in L_i {\setminus } \bigcup _{i \ne j, j=1}^t L_j\), for \(1 \leqslant i \leqslant t-1\). Set \(S=\{x_i \mid 1 \leqslant i \leqslant t-1\}\). Then the ith component of the l-clique metric S-representation of \(L_j\) is zero if and only if \(i=j\), for \(1 \leqslant i \leqslant t-1\). Moreover, none of the components of the l-clique metric S-representation of \(L_t\) is zero. Hence S is an l-clique metric generator of G, and so \(\mathrm{cdim}_l(G) \leqslant t-1\). \(\square \)
If we consider disconnected graphs, then l-CMD could be easily defined by considering the distance between two vertices in two different components as infinite. In fact we have the following result.
Remark 2.7
Let G be a disconnected graph with components \(G_1, \dots , G_r\). If \(I=\{i \mid G_i \, \mathrm {has} \, \mathrm {one} \, l\mathrm {-clique}\}\) and \(J=\{i \mid G_i \, \mathrm {has} \, \mathrm {at} \, \mathrm {least} \, \mathrm {two} \, l\mathrm {-cliques}\}\), then
Recall that for two graphs \(H_1\) and \(H_2\) with disjoint vertex sets, the join \(H_1 \vee H_2\) of the graphs \(H_1\) and \(H_2\) is the graph obtained from the union of \(H_1\) and \(H_2\) by adding new edges from each vertex of \(H_1\) to every vertex of \(H_2\). The concept of join graph is generalized (in [17], it is called as a generalized composition graph). Assume that G is a graph on k vertices with \(V(G)=\{v_1,v_2,\dots ,v_k\}\), and let \(H_1,H_2,\dots ,H_k\) be k pairwise disjoint graphs. The G-generalized join graph \(G[H_1,H_2,\dots ,H_k]\) of \(H_1,H_2, \dots ,H_k\) is the graph formed by replacing each vertex \(v_i\) of G by the graph \(H_i\) and then joining each vertex of \(H_i\) to each vertex of \(H_j\) whenever \(v_i\sim v_j\) in the graph G. Now, if the graph G consists of two adjacent vertices, then the G-generalized join graph \(G[H_1,H_2]\) coincides with the join \(H_1 \vee H_2\) of the graphs \(H_1\) and \(H_2\).
Note that in the rest of this section, we assume that there exists at least a nontrivial \(H_i\), with \(1 \leqslant i \leqslant k\), in \(G[H_1,H_2,\dots ,H_k]\).
In the following proposition, we study the l-CMD of the G-generalized join graph \(G[H_1,H_2,\dots ,H_k]\), in the case that \(H_i\)’s are empty graphs.
Proposition 2.8
Assume that G is a connected graph on k vertices with \(V(G)=\{v_1,v_2,\dots ,v_k\}\), and let \(H_1,H_2,\dots ,H_k\) be k pairwise disjoint empty graphs. If \(\{v_1,v_2,\dots ,v_t\}\), where \(0 \leqslant t \leqslant k\) are the vertices in G such that each of them belongs to an l-clique, then
Proof
Let \(\{v_{i_1},v_{i_2},\dots ,v_{i_t}\}\), where \(0 \leqslant t \leqslant k\) be the vertices in G such that each of them belongs to at least one l-clique. If \(t=0\), then \(\mathrm{cdim}_l(G[H_1,H_2,\dots ,H_k]) =\mathrm{cdim}_l(G)=1\). So assume that \(t>0\). Let \(h_1, \dots , h_t\) be arbitrary vertices in \(H_1,\dots ,H_t\), respectively. Assume that S is an l-clique metric generator of the graph \(G[H_1,H_2,\dots ,H_k]\). For each \(1 \leqslant i \leqslant t\), we show that \(V(H_i) {\setminus } \{h_i\} \subseteq S\). Suppose on the contrary that there exists \(h'_i \in V(H_i)\) with \(h'_i\ne h_i\) such that \(h'_i \notin S\). Now consider two l-cliques \(L_1\) and \(L_2\) such that \(h_i\) is a vertex of \(L_1\), \(h'_i\) is a vertex of \(L_2\) and \(L_1 {\setminus } \{h_i\}=L_2 {\setminus } \{h'_i\}\). Now, one can see that the l-clique metric S-representations of \(L_1\) and \(L_2\) are the same, which is a contradiction. Hence \(V(H_i) {\setminus } \{h_i\} \subseteq S\), for each \(1 \leqslant i \leqslant t\). Therefore we have
Let \(G'\) be the induced subgraph on vertex set \(\{h_1,\dots ,h_t,v_{t+1},\dots ,v_k\}\). Clearly \(G'\) is isomorphic to G. Now, let \(S'\) be an l-clique metric basis for \(G'\). Since, for each \(h_j,h'_j \in V(H_j)\), where \(t+1 \leqslant j \leqslant k\), we have \(d(L,h_j)=d(L,h'_j)\), where L is an l-clique, \(S' \cup \bigcup _{i=1}^t (V(H_i) {\setminus } \{h_i\}) \) is an l-clique metric generator for \(G[H_1,H_2,\dots ,H_k]\). So
\(\square \)
In the following theorem, we determine the l-CMD of the G-generalized join graph \(G[H_1,H_2,\dots ,H_n]\), in the case that \(H_i\)’s are empty graphs and G is a path \(P_n\). In fact the following theorem shows examples where the bounds in Proposition 2.8 are reached.
Theorem 2.9
Assume that G is a path on \(n\geqslant 2\) vertices with \(V(G)=\{v_1,v_2,\dots ,v_n\}\), and let \(H_1,H_2,\dots ,H_n\) be n pairwise disjoint empty graphs. Then \(\sum _{i=1}^n|V(H_i)|-n \leqslant \mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) \leqslant \sum _{i=1}^n|V(H_i)|-n+1\), when \(l \in \{1,2\}\). Also if \(|V(H_i)|>1\), for each \(1 \leqslant i \leqslant n\), then we have
Proof
If \(l \geqslant 3\), then clearly \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) =1\). So let \(l \in \{1,2\}\). Let \(h_1, \dots , h_n\) be arbitrary vertices in \(H_1,\dots ,H_n\), respectively. Set \(S=\bigcup _{i=1}^n(V(H_i) {\setminus } \{h_i\})\), where \(h_i\) is an arbitrary vertex in \(H_i\). By Proposition 2.8, every l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\) contains S. Also \(S \cup \{h_1\}\) is an l-clique metric generator for \(G[H_1,H_2,\dots ,H_n]\). Hence we have
If \(n=3\), then we have \(r_{G[H_1,H_2,H_3]}^1(h_1|S)=r_{G[H_1,H_2,H_3]}^1(h_3|S)\) and also we have \(r_{G[H_1,H_2,H_3]}^2(h_1h_2|S)=r_{G[H_1,H_2,H_3]}^2(h_2h_3|S)\), which means that S is not an l-clique metric generator of \(G[H_1,H_2,H_3]\), and as a consequence, \(\mathrm{cdim}_l(G[H_1,H_2,H_3])> |S|=\sum _{i=1}^3|V(H_i)|-3\). Set \(S'=S \cup \{h_1\}\). Now, one can see that \(S'\) is an l-clique metric basis of \(G[H_1,H_2,H_3]\), and so \(\mathrm{cdim}_l(G[H_1,H_2,H_3])=\sum _{i=1}^3|V(H_i)|-2\). Now, let \(|V(H_i)|>1\), for each \(1 \leqslant i \leqslant n\) and, assume that \(n \ne 3\). Then it is easy to see that S is an l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\), which implies that \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) =\sum _{i=1}^n|V(H_i)|-n\). \(\square \)
In the following theorem, we determine the l-CMD of the G-generalized join graph \(G[H_1,H_2,\dots ,H_n]\), in the case that \(H_i\)’s are empty graphs and G is the complete graph \(K_n\).
Theorem 2.10
Assume that \(G \cong K_n\) with \(V(G)=\{v_1,v_2,\dots ,v_n\}\), \(n >2\), and let \(H_1,H_2,\dots ,H_n\) be n pairwise disjoint empty graphs such that the number of trivial \(H_i\)’s is \(r<n\). Then we have
Proof
Assume that \(h_1, \dots , h_n\) are arbitrary vertices in \(H_1,\dots ,H_n\), respectively. Let \(S=\bigcup _{i=1}^n(V(H_i) {\setminus } \{h_i\})\). By Proposition 2.8, every l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\) contains S, which implies that \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) \geqslant \sum _{i=1}^n|V(H_i)|-n\). First assume that \(l=1\). Since the places in which there is a 2, if exists, appears in the l-clique metric S-representation of each two distinct \(h_i\) and \(h_j\), with \(1 \leqslant i\ne j \leqslant n\), are different from each other, their l-clique metric S-representations are not equal. Without loss of generality, assume that \(|V(H_1)|=\dots =|V(H_r)|=1\). Hence the l-clique metric S-representation of all \(h_i\)’s, for \(1 \leqslant i \leqslant r\) is equal. So, in this situation, any l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\) is of the form \(S \cup \bigcup _{i=1,i \ne j}^r\{h_i\}\), for some \(1 \leqslant j \leqslant r\). Hence we have \(\mathrm{cdim}_1(G[H_1,H_2,\dots , H_n])= \sum _{i=1}^n|V(H_i)|-n+r-1\), for \(0<r<n\). Clearly if \(r=0\), then S is a 1-clique metric basis of \(G[H_1,H_2,\dots ,H_n]\), and so \(\mathrm{cdim}_1(G[H_1,H_2,\dots , H_n]) = \sum _{i=1}^n|V(H_i)|-n\).
Now, assume that \(l \geqslant 2\). Let \(S'\) be an l-clique metric generator and L be an arbitrary l-clique of \(G[H_1,H_2,\dots ,H_n]\). For each \(x \in S'\), we have
So, for each two distinct l-cliques \(L_1\) and \(L_2\), \(L_1 \cap S'=L_2 \cap S'\) if and only if \(L_1\) and \(L_2\) have the same l-clique metric \(S'\)-representations. If \(l=n\), then, for each two distinct l-cliques \(L_1\) and \(L_2\), \(L_1 \cap S=L_2 \cap S\) implies that \(L_1=L_2\). This implies that S is an l-clique metric basis, and so \(\mathrm{cdim}_n(G[H_1,H_2,\dots , H_n])= \sum _{i=1}^n|V(H_i)|-n\). Now, assume that \(2 \leqslant l \leqslant n-1\). If there are \(h_i\) and \(h_j\) with \(1 \leqslant i\ne j \leqslant n\) such that they do not belong to an l-clique metric generator \(S'\), then consider two l-cliques \(L_1\) and \(L_2\) with \(h_i \in L_1\), \(h_j \in L_2\) and \(L_1 {\setminus } \{h_i\}=L_2 {\setminus } \{h_j\}\). Since \(L_1 \cap S'=L_2 \cap S'\), they have the same l-clique metric \(S'\)-representations, which is impossible. So in this situation, any l-clique metric generator is of the form \(S \cup \bigcup _{i=1, i \ne j}^nV(H_i)\), for some \(1 \leqslant j \leqslant n\). Thus we have \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n])= \sum _{i=1}^n|V(H_i)|-1\). \(\square \)
In the following theorem, we determine the l-CMD of the G-generalized join graph \(G[H_1,H_2,\dots ,H_n]\), in the case that \(H_i\)’s are empty graphs and G is isomorphic to the cycle \(C_n\), where \(n>3\). Note that the case \(n=3\) is obtained by Theorem 3.3.
Theorem 2.11
Assume that G is a cycle \(C_n\) with vertex set \(V(G)=\{v_1,v_2,\dots ,v_n\}\), \(n>3\), and let \(H_1,H_2,\dots ,H_n\) be n pairwise disjoint empty graphs. Then
when \(l \in \{1,2\}\), and \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n])=1\), for \( l \geqslant 3\). Also, for \(n=4\) and \( l \in \{1,2\}\), we have \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n])=\sum _{i=1}^n|V(H_i)|-n+2 \), and if \(|V(H_i)|>1\), for each \(1 \leqslant i \leqslant n\), then we have \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) =\sum _{i=1}^n|V(H_i)|-n\), when \( n> 4\) and \( l \in \{1,2\}\).
Proof
Clearly if \(l \geqslant 3\), then \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) =1\). So assume that \(l \in \{1,2\}\). Let \(h_1, \dots , h_n\) be arbitrary vertices in \(H_1,\dots ,H_n\), respectively, and \(S=\bigcup _{i=1}^n(V(H_i) {\setminus } \{h_i\})\). By Proposition 2.8, every l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\), contains S. Also \(S\cup \{h_1,h_2\}\) is an l-clique metric generator of \(G[H_1,H_2,\dots ,H_n]\). Hence \(\sum _{i=1}^n|V(H_i)|-n \leqslant \mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) \leqslant \sum _{i=1}^n|V(H_i)|-n+2\). If \(n=4\), then one can see that \(S\cup \{h_1,h_2\}\) is an l-clique metric basis of \(G[H_1,H_2,\dots ,H_n]\). So \(\mathrm{cdim}_l(G[H_1,H_2,H_3, H_4]) =\sum _{i=1}^4|V(H_i)|-2\).
Now, assume that \(n \geqslant 5\). Let \(|V(H_i)|>1\), for each \(1 \leqslant i \leqslant n\). Since \(n \geqslant 5\) and \(|V(H_i)| \geqslant 2\), for any two vertices \(h_i,h_j \notin S\), the distance between \(h_i\) and any vertex belonging to \(S \cap (V(H_{i-1})\cup V(H_{i+1}))\) is one, while the distance between \(h_j\) and any vertex belonging to at least one of these two sets \(S\cap V(H_{i-1})\) or \(S \cap V(H_{i+1})\) is different than one. Thus, S is an 1-clique metric generator for \(G[H_1,H_2,\dots , H_n]\). Now, let \(L_1\) and \(L_2\) be two distinct 2-cliques. If \(L_1 \cap S=\phi =L_2 \cap S\), then the places that 1 appears in their 2-clique metric S-representations are different. So, without loss of generality, assume that \(s \in L_1 \cap S\). If \(s \notin L_2\), then the corresponding components to s in the 2-clique metric S-representations of \(L_1\) and \(L_2\) are zero and nonzero, respectively. Thus, let \(s \in L_2\). If \(L_1 \subseteq S\) or \(L_2 \subseteq S\), then clearly their 2-clique metric S-representations are different. Now, assume that \(L_1 \nsubseteq S\) and \(L_2 \nsubseteq S\). Then one can see that the places of 1 in their 2-clique metric S-representations are different. So S is an 2-clique metric generator for \(G[H_1,H_2,\dots , H_n]\). Hence we have \(\mathrm{cdim}_l(G[H_1,H_2,\dots , H_n]) = \sum _{i=1}^n|V(H_i)|-n\). \(\square \)
3 l-Clique Metric Dimension of \(\Gamma ({\mathbb {Z}}_n)\)
Let R be a commutative ring with nonzero identity. We denote the set of all unit elements and zero divisors of R by U(R) and Z(R), respectively. Also by \(Z^*(R)\) we denote the set \(Z(R) {\setminus } \{0\}\). Sharma and Bhatwadekar [20] defined the comaximal graph of a commutative ring R. The comaximal graph of R is a simple graph whose vertices consists of all elements of R, and two distinct vertices a and b are adjacent if and only if \(aR+bR=R\), where cR is the ideal generated by c, for \(c \in R\). Let \(\Gamma (R)\) be an induced subgraph of the comaximal graph with nonunit elements of R as vertices. The properties of the graph \(\Gamma (R)\) were studied in [16, 22, 25].
For two integers r and s, the notation (r, s) stands for the greatest common divisor of r and s. Also we denote the elements of the ring \({\mathbb {Z}}_n\), where \(n>1\), by \(0,1,2, \dots ,n-1\). For every nonzero element a in \({\mathbb {Z}}_n\), if \((a,n)=1\), then a is a unit element; otherwise, \((a,n)\ne 1\), and so a is a zerodivisor. Therefore, \(|U({\mathbb {Z}}_n)|=\phi (n)\) and \(|Z({\mathbb {Z}}_n)|=n - \phi (n)\), where \(\phi \) is the Euler’s totient function.
An integer d is said to be a proper divisor of n if \(1<d<n\) and \(d \mid n\). Now let \(d_1,d_2, \dots ,d_k\) be the distinct proper divisors of n. For \(1 \leqslant i \leqslant k\), set
Clearly, the sets \(A_{d_1}, A_{d_2}, \dots ,A_{d_k}\) are pairwise disjoint and we have
and
The following lemma is stated from [27].
Lemma 3.1
[27, Proposition 2.1] Let \(1\leqslant i \leqslant k\). Then \(|A_{d_i}|=\phi (\frac{n}{d_i})\).
In this section, the induced subgraph of \(\Gamma ({\mathbb {Z}}_n)\) on the set \(A_{d_i}\) is denoted by \(\Gamma (A_{d_i})\), where \(1\leqslant i \leqslant k\).
The following lemma states some adjacencies in \(\Gamma ({\mathbb {Z}}_n)\).
Lemma 3.2
The following statements hold:
-
(i)
Two distinct vertices x and y are adjacent in \(\Gamma ({\mathbb {Z}}_n)\) if and only if \((x,y) \in U({\mathbb {Z}}_n)\).
-
(ii)
For \(1\leqslant i \leqslant k\), \(\Gamma (A_{d_i})\) is isomorphic to \({\overline{K}}_{\phi (\frac{n}{d_i})}\).
-
(iii)
For \(1\leqslant i\ne j \leqslant k\), a vertex of \(A_{d_i}\) is adjacent to a vertex of \(A_{d_j}\) if and only if \((d_i,d_j)=1\).
Proof
-
(i)
First suppose that x and y are adjacent vertices in \(\Gamma ({\mathbb {Z}}_n)\). Assume on the contrary that \(d=(x,y)\notin U({\mathbb {Z}}_n)\). So we have \(x{\mathbb {Z}}_n \subseteq d{\mathbb {Z}}_n\) and \(y{\mathbb {Z}}_n \subseteq d{\mathbb {Z}}_n\). Thus \(x{\mathbb {Z}}_n + y{\mathbb {Z}}_n \subseteq d{\mathbb {Z}}_n \ne {\mathbb {Z}}_n\), and this means that x and y are not adjacent, which is a contradiction. Now, let \(u=(x,y)\in U({\mathbb {Z}}_n)\). So there exist \(r,s \in {\mathbb {Z}}\) such that \(u=rx+sy \in x{\mathbb {Z}}_n+y{\mathbb {Z}}_n\). Therefore we have \(x{\mathbb {Z}}_n+y{\mathbb {Z}}_n={\mathbb {Z}}_n\), which implies that x and y are adjacent.
-
(ii)
For each two distinct elements \(x,y \in A_{d_i}\), we have \((x,n)=d_i=(y,n)\). So \(d_i \mid (x,y)\), which implies that \((x,y)\notin U({\mathbb {Z}}_n)\). Hence by (i), we have that x and y are not adjacent. Therefore by Lemma 3.1, we have \(\Gamma (A_{d_i}) \cong {\overline{K}}_{\phi (\frac{n}{d_i})}\).
-
(iii)
Let \(i,j \in \{1,2,\dots ,k\}\) with \(i \ne j\). First assume that \(x\in A_{d_i}\) and \(y \in A_{d_j}\) are adjacent vertices. If \((d_i,d_j)=d\ne 1\), then \((n,d)=d\). Since \((x,n)=d_i\) and \((y,n)=d_j\), we have that \(d \mid x,y\). Hence \(Rx+Ry \subseteq Rd \ne R\), which is impossible. Now suppose that \((d_i,d_j)=1\). Let \(x\in A_{d_i}\) and \(y\in A_{d_j}\) be arbitrary vertices. If \(d=(x,y)\notin U({\mathbb {Z}}_n)\), then \(t=(d,n)\ne 1\). Since \(t \mid x,y,n\), we have \(t \mid (d_i,d_j)\) and this is impossible. Hence \((x,y)\in U({\mathbb {Z}}_n)\) which means that x and y are adjacent. \(\square \)
Now, we introduce a simple graph \(G_n\), which plays an important role in the structure of \(\Gamma ({\mathbb {Z}}_n)\). The graph \(G_n\) is the simple graph with vertex set \(\{d_1,d_2,\dots ,d_k\}\), where \(d_i\)’s, \(1\leqslant i \leqslant k\), are the proper divisors of n, and two distinct vertices \(d_i\) and \(d_j\) are adjacent if and only if \((d_i,d_j)=1\).
Let \(n=p_1^{\alpha _1}p_2^{\alpha _2}\dots p_t^{\alpha _t}\) be the factorization of n to its prime powers, where \(t,\alpha _1,\dots ,\alpha _t\) are positive integers and \(p_1,\dots ,p_t\) are distinct prime numbers. Every divisor of n is of the form \(p_1^{\beta _1}p_2^{\beta _2}\dots p_t^{\beta _t}\), for some integers \(\beta _1,\dots ,\beta _t\), where \(0 \leqslant \beta _i \leqslant \alpha _i\) for each \(i\in \{1,2,\dots ,t\}\). Hence the number of proper divisors of n is equal to \(\prod _{i=1}^t(n_i+1)-2\). Therefore we have \(k=|V(G_n)|=\prod _{i=1}^t(n_i+1)-2\).
Let \(\Gamma ^*({\mathbb {Z}}_n)=\Gamma ({\mathbb {Z}}_n) {\setminus } \{0\}\). Consider the graph \(G_n\) and replace each vertex \(d_i\) of \(G_n\) by \(\Gamma [A_{d_i}]\). In view of Lemma 3.1, we have
Now, since the zero element is adjacent to none of the vertices of \(\Gamma ^*({\mathbb {Z}}_n)\), we have
In the following theorem, we study the l-CMD of \(\Gamma ({\mathbb {Z}}_n)\).
Theorem 3.3
Assume that \(\{d_1,d_2,\dots ,d_t\}\), where \(1\leqslant t \leqslant k\), are those vertices of \(G_n\) that each of them belongs to an l-clique. Then for \(l=1\) we have
and for \(l>1\),
where r is the number of isolated vertices of \(G_n\).
Proof
Note that the graph \(G_n\) is not connected in general. Let r be the number of isolated vertices of \(G_n\). Since 0 is the isolated vertex of \(\Gamma ({\mathbb {Z}}_n)\), we assume that \(0,a_1, \dots , a_r\) are the isolated vertices of \(\Gamma ({\mathbb {Z}}_n)\). By Remark 2.7, we have
Now, the results follow from Proposition 2.8 and Remark 2.7. \(\square \)
Example 3.4
Consider the ring \({\mathbb {Z}}_{12}\). We have \(d_1=2, d_2=3,d_3=4\), and \(d_4=6\). Then \(G_{12}\) is the graph \(2\sim 3 \sim 4 \cup \{6\}\), which is isomorphic to \( P_3 \cup K_1\). Hence we have
and, by Theorems 2.9 and 3.3 , we have
In the rest of this section, we discuss the CMD of \(\Gamma ({\mathbb {Z}}_n)\), for (i) \(n=p^t\), (ii) \(n=pq\) and (iii) \(n=p^2q\), where p and q are distinct prime numbers and t is a positive integer.
-
(i)
Let \(n=p^t\). Then \(\Gamma ({\mathbb {Z}}_{p^t})\) is an empty graph with \(p^t - \phi (p^t)=p^{t-1}\) vertices, and so \(\Gamma ({\mathbb {Z}}_{p^t})=\overline{K_{p^{t-1}}}\). Now, by Remark 2.7 we have
$$\begin{aligned} \mathrm{cdim}_l(\Gamma ({\mathbb {Z}}_{p^t})) =\left\{ \begin{array}{ll} p^{t-1}-1 &{}\quad l=1 \\ 1 &{}\quad l \geqslant 2. \end{array} \right. \end{aligned}$$ -
(ii)
Let \(n=pq\), where p and q are distinct prime numbers. Since the only proper divisors of n are p and q, the graph \(G_{pq}\) is \(p \sim q\). So we have
$$\begin{aligned} \Gamma ({\mathbb {Z}}_{pq})=K_1\cup G_{pq}[{\overline{K}}_{\phi (q)},{\overline{K}}_{\phi (p)}]. \end{aligned}$$
Now, by Theorem 2.9, we have
-
(iii)
Let \(n=p^2q\), where p and q are distinct prime numbers. Since p, q, and pq are the proper divisors of n, the graph \(G_{p^2q}\) is \(p \sim q \sim p^2 \cup \{pq\}\). Hence we have
$$\begin{aligned} \Gamma ({\mathbb {Z}}_{p^2q})=K_1\cup G_{p^2q}[{\overline{K}}_{\phi (pq)}, {\overline{K}}_{\phi (p^2)},{\overline{K}}_{\phi (q)},{\overline{K}}_{\phi (p)}]. \end{aligned}$$
Since \(\phi (pq)=pq-p-q+1\) and \(\phi (p^2)=p^2-p\), by Theorem 2.9 and Remark 2.7,
4 l-Clique Metric Dimension Over Corona Product
Let G and H be two graphs with the vertex sets \(\{g_1,\ldots , g_n\}\) and \(\{h_1,\ldots ,h_m\}\), respectively. The corona of G and H, denoted by \(G\circ H\), is the graph whose vertex and edge sets are defined as below:
The metric dimension (1-CMD) of corona product graphs was investigated in [26]. After that Peterin and Yero studied the edge metric dimension (2-CMD) over corona product in [18]. In this section, we give a formula for the l-CMD of corona product of two graphs G and H for \(l\ge 3\). In what follows, we say the vertex v distinguishes two l-cliques U and W if \(d(v,U)\ne d(v,W)\).
Theorem 4.1
Let G and H be two connected graphs of order n and m, respectively, and \(l\ge 3\) be an integer number. If \(\{V_1(H),\ldots ,V_k(H)\}\) is the \((l-1)\)-clique set of H, then
where \(\omega (G)\) and \(\omega (H)\) are the clique numbers of G and H, respectively.
Proof
Let \(V(G)=\{g_1, \ldots , g_n\}\) and \(H_i\) be the i-th copy of H in \(G\circ H\), \(1\le i\le n\). Then \(G\circ H\) is obtained by joining each vertex of the i-copy of H to the i-th vertex, \(g_i\), of G.
Let \(S_G\) be an l-clique metric basis of G and \(\{V_1(G),\ldots , V_t(G)\}\) be the l-clique set of G. Also, let \(V_{j_i}(H)\) denote the i-the copy of \(V_j(H)\) in \(G\circ H\), for \(1\le i\le n\) and \(1\le j\le k\). Thus, it is clear that \(V'_{j_i}(H)=V_{j_i}(H)\cup \{g_i\}\), \(1\le i\le n\), is an l-clique in \(G\circ H\).
First, we prove that if \(\omega (H)<l-1\) (or \(k=0\)), then \(\mathrm{cdim}_l(G\circ H)=\mathrm{cdim}_l(G)\). To do this, we prove that \(S_G\) is also an l-clique metric basis of \(G\circ H\). Clearly \(S_G\) is an l-clique metric generator for \(G\circ H\) and so \(\mathrm{cdim}_l(G\circ H)\le \mathrm{cdim}_l(G)\). Suppose that S is an l-clique metric basis of \(G\circ H\). We claim that \(|S\cap V(H_i)|\le 1\) for \(1\le i\le n\). To prove this claim, suppose, on the contrary that there exist \(u,z\in S\cap V(H_i)\). Then \(S'=S{\setminus } \{u\}\) is not an l-clique metric generator for \(G\circ H\). Thus there exist two l-cliques U and W in \(G\circ H\) such that \(d_{G\circ H}(v,U)=d_{G\circ H}(v,W)\) for each \(v\in S'\). Hence \(d_{G\circ H}(z,U)=d_{G\circ H}(z,W)\). On the other hand, since \(\omega (H)<l-1\), then \(d_{G\circ H}(z,U)=d_{G\circ H}(z,W)=d_G(g_i,U)+1=d_G(g_i,W)+1\). Also, since \(\omega (H)<l-1\), then \(d_{G\circ H}(u,U)=d_{G\circ H}(u,W)=d_G(g_i,U)+1=d_G(g_i,W)+1\). Therefore S is not an l-clique metric generator for \(G\circ H\) which is a contradiction.
Now suppose that \(u\in S\cap V(H_i)\). Then \(S'=(S-\{u\})\cup \{g_i\}\) is also an l-clique metric basis of \(G\circ H\). Because \(d_{G\circ H}(u, V_j(G))=d_G(g_i,V_j(G))+1\) for each \(1\le j\le t\). By repeating this technique, we reach an l-clique metric basis \(S''\) of \(G\circ H\) with this property that all vertices of \(S''\) are in G. Therefore, \(\mathrm{cdim}_l(G\circ H)\ge \mathrm{cdim}_l(G)\).
Now, suppose that \(\omega (G)<l\), \(k=1\) and \(V_1(H)\) is the \((l-1)\)-clique of H. Let \(S_G\) be a 1-clique metric basis of G. We claim that \(S_G\) is an l-clique metric generator for \(G\circ H\). Then, since \(d_{G\circ H}(V_{1_i}'(H),v)=d_G(g_i,v)\) for each \(v\in S_G\), then every pair of l-cliques \(V_{1_i}(H)\)’s, \(1\le i\le n\), is distinguished by a vertex of \(S_G\). Therefore, \(S_G\) is an l-clique metric generator for \(G\circ H\) and so \(\mathrm{cdim}_l(G\circ H)\le |S_G|=\mathrm{dim}(G)\). Then, it is sufficient to show that \(\mathrm{cdim}_l(G\circ H)\ge \mathrm{dim}(G)\). To do this, suppose that \(S'\) is an l-clique metric basis of \(G\circ H\). By the above argument, if \(|S'\cap V(G)|=|S'|\), then we have nothing to prove. Otherwise, there exists \(v\in S'\) such that \(v\in V_{1_i}\) for an \(i\in \{1,\ldots , n\}\). Since \(d_{G}(v,V'_{1_j})=d_{G}(g_i,V'_{1_j})+1\) for \(i\ne j\in \{1,\ldots , n\}\), then \(S''=(S-v)\cup \{g_i\}\) is also an l-clique metric basis of \(G\circ H\). We use this technique to reach an l-clique metric basis \(S'''\) of \(G\circ H\) with this property that \(|S'''\cap V(G)|=|S'''|\). Therefore, \(\mathrm{cdim}_l(G\circ H)\ge \mathrm{dim}(G)\). \(\square \)
The concept of global forcing sets for maximal matchings was presented in [24]. Here we need to introduce an extension of the idea of global forcing sets for l-cliques of a graph.
A global forcing set for l-cliques of a graph G is a subset S of V(G) with this property that \(V_1\cap S\ne V_2\cap S\) for any two l-cliques \(V_1\) and \(V_2\) of G. A global forcing set for l-cliques of G with minimum cardinality is called a minimum global forcing set for l-cliques of G, and its cardinality, denoted by \(\varphi _l\), is the global forcing number for l-cliques of G.
We can find a global forcing set for l-cliques of G by the following ILP.
Let G be a graph with \(V(G)=\{v_1, \ldots v_n\}\) and let \(\{V_1, \ldots , V_k\}\) be the set of all l-cliques of G. Let \(D_G=[d_{ij}]\) be a \(k\times n\) matrix, where \(d_{ij}= 1\) if \(v_j\in V_i\), and \(d_{ij}= 0\) otherwise. Let \(F: \{0,1\}^{n} \rightarrow {{\mathbb {N}}}_0\) be defined by
Then our goal is to determine \(\min F\) subject to the constraints
Note that if \(x'_1, \ldots , x'_n\) is a set of values for which F attains its minimum, then \(S = \{v_i: x'_i =1\}\) is a minimum global forcing set for l-cliques of G.
Theorem 4.2
Let G and H be two connected graphs with \(|V(G)|=n\), and \(l\ge 3\) be an integer number. If \(\{V_1(H),\ldots ,V_k(H)\}\) is the \((l-1)\)-clique set of H and \(\omega (H)=l-1\), then for \(k\ge 2\) we have
Proof
Let S be an l-clique metric generator for \(G\circ H\). Suppose, on the contrary that there exists \(H_i\), a copy of H in \(G\circ H\), that \(|S\cap V(H_i)|<\varphi _{l-1}(H)\). Then there exist two \((l-1)\)-cliques \(V_{j_i}(H)\) and \(V_{q_i}(H)\) in \(H_i\) such that \(S\cap V_{j_i}(H)=S\cap V_{q_i}(H)\). Hence \(d_{G\circ H}(u, V_{j_i}(H))=d_{G\circ H}(u, V_{q_i}(H))=0\) for each \(u\in S\cap V_{j_i}(H)\), and \(d_{G\circ H}(u, V_{j_i}(H))=d_{G\circ H}(u, V_{q_i}(H))=1\) for each \(u\in S\cap (V(H_i){\setminus } V_{j_i}(H))\). On the other hand, it is not difficult to check that \(d_{G\circ H}(u, V_{j_i}(H))=d_{G\circ H}(u, V_{q_i}(H))\) for each \(u\in S{\setminus } V(H_i)\). Thus, \(d_{G\circ H}(u,V_{j_i}(H))=d_{G\circ H}(u,V_{q_i}(H))\) for each \(u\in S\), which is contrary to our assumption. Therefore, \(\mathrm{cdim}_l(G\circ H)\ge n\cdot \varphi _{l-1}(H)\).
It remains to prove that \(\mathrm{cdim}_l(G\circ H)\le n\cdot \varphi _{l-1}(H)\). Let \(S_H\) be a minimum global forcing set for (\(l-1\))-cliques of H, and let \(S_{H_i}\) be the i-th copy of \(S_H\) in \(G\circ H\). Then, it is easy to check that \(S'=\bigcup _{i=1}^nS_{H_i}\) is an l-clique metric generator for \(G\circ H\). Therefore, \(\mathrm{cdim}_l(G\circ H)\le n\cdot \varphi _{l-1}(H)\). \(\square \)
5 Complexity Issues
The clique problem is the optimization problem of finding a clique of maximum size in a graph. As a decision problem, we ask simply whether a clique of a given size k exists in the graph.
Theorem 5.1
[8] The clique problem is \(\textit{NP}\)-complete.
Therefore, the problem of finding all l-cliques in a graph is NP-hard. Hence, throughout this section we are assuming that all the l-cliques of the graph are given.
In this section, we prove the l-CMD problem is NP-complete. Recall that for \(l=1,2\), l-CMD problems are the metric dimension and the edge metric dimension problems, respectively. On the other hand, Garey and Johnson [11] proved that the decision version of the metric dimension problem is NP-complete on connected graphs. Also, NP-completeness of computing the edge metric dimension of connected graphs was proved in [14]. Moreover, Epstein, Levin, and Woeginger showed that for split graphs, bipartite graphs, co-bipartite graphs, and line graphs of bipartite graphs, the problem of computing the metric dimension of the graph is NP-hard [10]. Then, we prove NP-completeness of computing the l-CMD of connected graphs for \(L\ge 3\). Let us start with the below decision problem.
l-CMD problem: For a given positive integer l. Let G be a connected graph with n where \(n \ge 3\), X be the set of all distinct l-cliques of G, and let r be a positive integer such that \(1 \le r \le n-1\). Is \(\mathrm{cdim}_l(G) \le r\)?
Note that the l-CMD problem is the decision version of the problem of computing \(\mathrm{cdim}_l(G)\) for a given connected graph G.
Our proof for showing that the NP-completeness of l-CMD problem is based on a reduction from the metric dimension problem on connected bipartite graphs. We recommend [7] for more details on the reduction technique. Now, we are ready to prove that the l-CMD problem is NP-complete.
Theorem 5.2
The l-CMD problem, for \(l\ge 3\), is NP-complete.
Proof
Note that the l-CMD problem is clearly in NP because we can check its feasibility as a l-clique metric generator in polynomial time.
For showing NP-hardness of this problem, we present a reduction from the metric dimension for connected bipartite graphs.
Let G be a connected bipartite graph where \(V(G)=\{g_1,\ldots ,g_n\}\). Now, we construct graph \(G'\) from G by taking one copy of G and n copies of the complete graph \(K_{l-1}\) and by joining each vertex of the i-th copy of \(K_{l-1}\) to the i-th vertex of G, \(i=1,\ldots ,n\). In other words, \(G'=G\circ K_{l-1}\). For more illustration, see an example of G and \(G'\) in Fig. 1. Since G is bipartite, then \(\omega (G)<3\). Thus by Theorem 4.1, \(\mathrm{cdim}_l(G')=\mathrm{cdim}_l(G\circ H)=\mathrm{dim}(G)\). Moreover, it is easy to see that constructing \(G'\) from G can be done in polynomial time. Therefore, if there exists a polynomial-time algorithm for computing \(\mathrm{cdim}_l(G')\), then there exists a polynomial-time algorithm for computing \(\mathrm{dim}(G)\). \(\square \)
An integer linear programming (ILP) model for the classical metric dimension problem was presented in [6]. Motivated by this work and using its notations, we consider here an IPL model for computing \({cdim}_l(G)\) for a given connected graph G and its l-cliques. Let \(G=(V,E)\) be a connected graphs with \(V=\{u_1,\ldots , u_n\}\). Let \(V_1,\ldots , V_k\) be the l-cliques of G. Also, suppose that \(D_G=[d_{ij}]\) is a \(k\times n\) matrix such that \(d_{ij}=d_G(V_i,u_j)\) for \(i\in \{1,\ldots , k\}\) and \(j\in \{1,\ldots , n\}\). Consider the binary decision variables \(x_i\) for \(i\in \{1,\ldots , n\}\) where \(x_i\in \{0,1\}\). By \(x_i\), we mean the vertex \(u_i\) is a member of an l-clique metric generator of G and \(x_i=0\) for otherwise. we define the objective function F by
Minimize F subject to the following constraints
is equivalent to finding a basis in the sense that if \(x_1',\ldots ,x_n'\) is a set of values for which F attains its minimum, then \(W=\{u_i\; | \; x_i'=1\}\) is a basis for G.
For example, consider graph G shown in Fig. 2 with 3-cliques \(V_1=\{u_1,u_2,u_3\}\) and \(V_2=\{u_3,u_4,u_5\}\). Then, \(D_G=\begin{pmatrix} 0 &{} 0&{} 0&{} 1&{} 1\\ 1 &{} 1 &{} 0&{} 0&{} 0 \end{pmatrix}\). Therefore, minimize \(F(x_1,x_2,x_3)=x_1+x_2+x_3+x_4+x_5\) subject to the constraints \(x_1+x_2+x_4+x_5 > 0\), \(x_1, x_2, x_3, x_4, x_5 \in \{0, 1\}\). Thus F attains its minimum for \(x_1=1\), \(x_2=x_3=x_4=x_5=0\), hence \(W=\{u_1\}\) is a 3-clique metric basis for G.
References
Afkhami, M., Barati, Z., Khashyarmanesh, K.: On the Laplacian spectrum of the comaximal graphs (submitted)
Bondy, J.A., Murty, U.S.R.: Graph theory, Graduate Texts in Mathematics, vol. 244. Springer, New York (2008)
Buszkowski, P.S., Chartrand, G., Poisson, C., Zhang, P.: On \(K\)-dimensional graphs and their bases. Periodico Mathematica Hungarica 46, 9–15 (2003)
Caceres, J., Hernando, C., Mora, M., Pelayo, I.M., Puertas, M.L., Seara, C., Wood, D.R.: On the metric dimension of some families of graphs. Electron. Notes Discret. Math. 22, 129–133 (2005)
Caceres, J., Hernando, C., Mora, M., Pelayo, I.M., Puertas, M.L., Seara, C., Wood, D.R.: On the metric dimension of cartesian products of graphs. SIAM J. Discret. Math. 21(2), 423–441 (2007)
Chartrand, G., Eroh, L., Johnson, M.A., Oellermann, O.R.: Resolvability in graphs and the metric dimension of a graph. Discret. Appl. Math. 105, 99–113 (2000)
Cormen, T.H., Leiserson, C.E., Rivest, R.L., Stein, C.: Introduction to Algorithms, 2nd edn. McGraw-Hill book company, The MIT Press (2003)
Cormen, T.H., Leiserson, C.E., Rivest, R.L., Stein, C.: Introduction to Algorithms, 3rd edn. The MIT Press, Cambridge (2009)
Das, K.C., Tavakoli, M.: Bounds for metric dimension and defensive \(k\)-alliance of graphs under deleted lexicographic product. Trans. Comb. 9(1), 31–39 (2020)
Epstein, L.L., Levin, A., Woeginger, G.J.: The (weighted) metric dimension of graphs: hard and easy cases. Algorithmica 72, 1130–1171 (2015)
Garey, M.R., Johnson, D.S.: Computers and Intractability: A Guide to the Theory of NP-Completeness. W.H. Freeman, New York (1979)
Hakanen, A., Laihonen, T.: On {l}-metric dimensions in graphs. Fund. Inform. 162, 143–160 (2018)
Johnson, M.: Structure-activity maps for visualizing the graph variables arising in drug design. J. Biopharm. Stat. 3, 203–236 (1993)
Kelenc, A., Tratnik, N., Yero, I.G.: Uniquely identifying the edges of a graph: the edge metric dimension. Discret. Appl. Math. 256, 204–220 (2018)
Khuller, S., Raghavachari, B., Rosenfeld, A.: Landmarks in graphs. Discret. Appl. Math. 70, 217–229 (1996)
Maimani, H.R., Salimi, M., Sattari, A., Yassemi, S.: Comaximal graph of commutative rings. J. Algebra 319, 1801–1808 (2008)
Schwenk, A.J.: Computing the characteristic polynomial of a graph. In: Graphs and Combinatorics. Lecture Notes in Math., vol. 406, pp. 153–172. Springer, Berlin (1974)
Peterin, I., Yero, I.G.: Edge metric dimension of some graph operations. Bull. Malays. Math. Sci. Soc. (2019). https://doi.org/10.1007/s40840-019-00816-7
Saputro, S.W., Simanjuntak, R., Uttunggadewa, S., Assiyatun, H., Baskoro, E.T., Salman, A.N.M., Bača, M.: The metric dimension of the lexicographic product of graphs. Discret. Math. 313, 1045–1051 (2013)
Sharma, P.K., Bhatwadekar, S.M.: A note on graphical representation of rings. J. Algebra 176, 124–127 (1995)
Slater, P.J.: Leaves of trees. Congr. Numer. 14, 549–559 (1975)
Slavko, M.M., Petrovic, Z.Z.: On the structure of comaximal graphs of commutative rings with identity. Bull. Aust. Math. Soc. 83, 11–21 (2011)
Tavakoli, M., Rahbarnia, F., Ashrafi, A.R.: Distribution of some graph invariants over hierarchical product of graphs. Appl. Math. Comput. 220, 405–413 (2013)
Vukičević, D., Zhao, S., Sedlar, J., Xu, S.-J., Došlić, T.: Global forcing number for maximal matchings. Discret. Math. 341, 801–809 (2018)
Wang, H.J.: Graphs associated to co-maximal ideals of commutative rings. J. Algebra 320, 2917–2933 (2008)
Yero, I.G., Kuziak, D., Rodríguez-Velázquez, J.A.: On the metric dimension of corona product graphs. Comput. Math. Appl. 61, 2793–2798 (2011)
Young, M.: Adjacency matrices of zero-divisor graphs of integers modulo \(n\). Involve 8, 753–761 (2015)
Acknowledgements
The authors are very grateful to the referees for their comments and suggestions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Sanming Zhou.
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
Afkhami, M., Khashyarmanesh, K. & Tavakoli, M. l-Clique Metric Dimension of Graphs. Bull. Malays. Math. Sci. Soc. 45, 2865–2883 (2022). https://doi.org/10.1007/s40840-022-01299-9
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40840-022-01299-9