1. Introduction
Every network can be represented by a graph. The identification of every unique vertex (node) is of great importance so as to maintain the security of the network. Now, the question that needs to be posed is “What should be an identifying method for a given graph?” In a connected graph, distances play a vital role in the identification of unique vertices. Let
be a simple connected unweighted and undirected graph with the vertex set
and edge set
. Two vertices
u and
v are called adjacent if there is an edge between
u and
v. An edge
e is called adjacent to a vertex
v if
e has
v as one end. The degree of a vertex
is the number of edges adjacent to
v. The distance between two vertices
u and
v, denoted by
(or simply
), is the length of the shortest path between them. Let
be a subset (ordered) of the vertices of
G. The
code of a vertex
v with respect to
S, denoted by
, is the
m-tuple
The set
is called a
resolving set if
for every pair of distinct vertices
v and
w. A resolving set
S is said to be a
fault-tolerant resolving set with the tolerance level k if
is also a resolving set of
G, where
X is any subset of
S consisting of
k elements. The minimum size of a fault-tolerant resolving set with the tolerance level
k is called a
k-level metric dimension. We denote the
k-level metric dimension for a graph
G by
A fault-tolerant resolving set with tolerance
k consisting of
elements is called the
k-level metric basis. In the literature, the
k-level metric dimension is known as the metric dimension and fault-tolerant metric dimension for
and
, respectively. The concept of metric dimensions was first put forward by Slater [
1] and Harary and Melter [
2]. The problem of finding the metric dimension for a general graph is NP-hard. Khuller et al. [
3] presented a construction that proves that the metric dimension of a graph is NP-hard. For a general
k, the
k-level metric dimension problem was introduced by Estrada-Moreno et al. [
4], and they determined the exact value of the
k-metric dimension for paths and cycle graphs. Yero et al. [
5] showed that the problem of computing the
k-level metric dimension of graphs is NP-hard. However, they solved the problem in linear time for the particular case of trees. Recently, Hakanen et al. [
6] obtained some results on this topic.
Although the applications of metric bases arise in many various platforms, such as Robot Navigation [
3], Network Optimization [
7], and sensor networks [
8], it has some limitations because, if some detectors (elements of the metric basis) are faulty, then it is not possible to identify the unique nodes. In order to improve the accuracy of the detection or the robustness of the system, Estrada et al. in [
4] took the initiative to introduce a generalized metric basis, namely, the
k-level metric basis. In the
k-level metric basis, one can tolerate up to
k detector faults.
Hernendo et al. [
9] characterized all fault-tolerant resolving sets for any tree
T. In the same article, they show the relation
for any graph
G. For a positive integer
n and any list
chosen from
, a
circulant graph, denoted by
, is a graph on the vertices
such that each
,
, is adjacent to
and
(subscripts are taken modulo
n) for every
j in the list
. The circulant graph
is the complete graph
, and the graph
is the cycle
. For the cycle
, the fault-tolerant metric dimension was determined by Javaid et al. in [
10] as
. Basak at el. [
11] determined the fault-tolerant metric dimension (
of a special type of circulant graph
when
, and Saha et al. [
12] determined the same for when
. The exact value of
(i.e., when
) was determined by Javaid et al. [
10]. In [
13], the authors determined the metric dimension of the power of finite paths. The increased connectivity of circulant networks makes them ideal for parallel computing and signal processing [
14]. The applications of circulant graphs in graph theory have appeared in coding theory, VLSI design, Ramsey theory, and many other areas [
15]. Despite the widespread application of circulant networks, there are no results on fault-tolerant metrics with tolerance
k. In this article, we determine the
k-level metric dimension for the circulant graph
for all possible values of
k and
We also delineate the optimal fault-tolerant resolving sets with the tolerance level
k. In
Table 1, we summarize the existing results and our results on
for different values of
k and
.
2. Preliminaries
Henceforth, we denote the vertex set by . As we are aware, is a graph with the vertex set , and two vertices are adjacent in if and only if , where denotes the distance between and in the cycle graph . Thus, the size of the maximum clique in is 3. Additionally, the diameter of is . More explicitly, the diameter of is if n is of the form , where . Every vertex has exactly antipodal vertices (the vertices that are at diametral distances). In particular, each vertex in has a unique antipodal vertex. We let be the set of all vertices of G that resolve and . Clearly, both are in . Throughout the article, we use instead of . We let be the set of all graphs in which there exists at least one fault-tolerant resolving set with the tolerance level k. Note that . We let (or simply ) be the maximum value of k such that G has a fault-tolerant resolving set with the tolerance level k. We call the vertex of an odd vertex if the index i is an odd number; otherwise, we call it an even vertex. Two vertices and are said to have the same parity if is even, i.e., if both i and j are either even or odd. Throughout the article, the indices of the vertices of are taken to be modulo n.
The workflow of the rest of the paper is presented below. To find the exact value of
, we first present a lower bound for this in
Section 3. To find a lower bound for
, first, we identify the vertices that resolve the two consecutive pairs
and
(Lemma 5). Using this lemma, we show that any fault-tolerant resolving set with the tolerance level
k must contain at least
or
elements accordingly as
or
. In
Section 4, we explore an optimal fault-tolerant resolving set with the tolerance level
k. To construct an optimal fault-tolerant resolving set with the tolerance level
k, we first find the number of vertices of a set
S (of consecutive vertices) that may resolve the set
in Lemmas 8 and 9. We prove the results (Lemma 8 and Lemma 9) by considering two complementary cases such that
S contains at most one element of
U or
U is a proper subset of
S. Then, we construct a set
F that is the union of two sets
and
of consecutive vertices. The set
is chosen from the first
consecutive vertices, whereas the set
is chosen from
consecutive vertices starting from the middle vertex or from a vertex adjacent to the middle vertex/vertices of
.
3. Lower Bound for
Let
be the complement of the set
for two non-consecutive vertices
and
in
. Then,
, that is,
We obtain the formula for in the following lemma.
Lemma 1. Let () be two vertices such that . Then, the set is the following:
When ,
When ,
When ,
When ,
Proof. Let be an internal vertex between and in G. We consider the following three cases:
. Then, , which implies that cannot lie in the shortest path of and ; i.e., we have . Note that . Now, if the shortest path between and is via , that is, , then , and hence, . Again, if , then gives . Thus, in any case, , so resolves and when . Therefore, , a contradiction to .
. Then, . Since , the shortest path is not via , and hence, we have . Additionally, . By the aforementioned argument, , and hence, . So, resolves and . Again, , a contradiction to .
. Let
. Then,
or
according to whether
is even or odd. Now, we check whether or not the vertex
resolves
and
. For an even integer
,
,
. For an odd integer
, we calculate the distances of
and
from each element of
W in
Table 2.
Thus, the set
S of all intermediate vertices between
and
with
that do not resolve
and
is given by
Again, if
is a vertex such that
, then
will be an intermediate vertex between
and
, where
and
. Then, applying (
1) to
and
, we obtain the set
of all vertices that lie in
and do not resolve the vertices
and
, where
is given by
Putting
and
in (
2), we obtain
where
. We denote
by
for
,
. Then, from Equation (
3), we have the following:
We obtain the result by taking for . □
Corollary 1. For any two vertices and of , the following hold:
- (a)
, provided that and are non-consecutive vertices.
- (b)
If and are two consecutive vertices, then the minimum value of is .
Lemma 2. Let be a simple connected graph. A set is a fault-tolerant resolving set with the tolerance level k if and only if for every pair of vertices and of H.
Proof. Let and be two arbitrary vertices of H. First, we assume that (ordered set) is a fault-tolerant resolving set with the tolerance level k. Then, and differ by at least positions. Since and , the set consists of at least elements, which implies that for every pair of vertices and of H.
Conversely, we assume that for every pair of vertices and of H. Then, consists of at least elements, and hence, and differ by at least positions. Since and are two arbitrary vertices of H, is a fault-tolerant resolving set with the tolerance level k. □
Lemma 3. For the circulant graph , .
Proof. Let F be a fault-tolerant resolving set with the tolerance level . Then, by Lemma 2, every pair of vertices must be resolved by at least elements of F. Thus, from Corollary 1, . □
Lemma 4. For a positive integer n, exactly one of the following is true:
- (a)
- (b)
.
Proof. First, we assume that n is odd. Then, (say). Thus, and . Thus, only is true when n is an odd integer. Next, we assume that n is even. Then, (say). Thus, and . Thus, only is true when n is an even integer. □
Lemma 5. For a circulant graph , Proof. Clearly,
and
. We may write
for some positive integer
s and
. Let
be a vertex in
. For
, we have
and
. For
, we obtain
and
. Therefore, for
or
, the distances of
,
, and
from
are of the form
,
, and
or in reverse order, respectively, where
a is a positive integer. Applying Lemma 4, it is clear that
, provided that
. Now, in
Table 1, we calculate the distances of
,
, and
from
, where
(
Table 3).
From
Table 2, it is clear that if
, then
only when
and
. This completes the proof of the result. □
Example 1. For the circulant graph , . Let us find for ; i.e., we show that . Then, each vertex resolves both pairs and , as the distances are given by Thus, . One can easily show that any vertex cannot resolve both pairs and .
Remark 1. It is noted that, when , each vertex of has a unique antipodal vertex. Indeed, is the antipodal vertex of . Hence, resolves both pairs and .
Below, we give a lower bound for the k-level fault-tolerant metric dimension for the circulant graph .
Theorem 1. Every fault-tolerant resolving set with tolerance k of consists of at least elements. Moreover, if , then every fault-tolerant resolving set with tolerance k contains at least elements.
Proof. Let F be an arbitrary fault-tolerant resolving set with the tolerance level k. First, we assume . If F contains every vertex of , then the theorem holds trivially. So, we assume that there exists a vertex . Then, for some i, where . Then, using Lemma 5 with , we obtain . Again, Lemma 2 states that each of and must be at least , so (for a better understanding, refer to Example 3). Now, we assume . If F contains neither nor for some i, then, using a similar argument to that above and Lemma 5, one can easily prove that and, consequently, , as each of the sets and consists of at least elements due to Lemma 2 (for a better understanding, refer to Example 2). Again, if F contains at least one element from , then . Thus, in any case, , provided that .
We now prove that if
, then every resolving set with tolerance
k contains at least
elements. Assume to the contrary that there is a
k-tolerance resolving set
F that contains exactly
elements. Our claim is:
Let
. Note that
, and
are antipodal vertices of
. So,
cannot resolve the two pairs
and
. If
, then, by a similar argument to that above, we need at least
elements, except
, to resolve these pairs. Thus, we must have
whenever
. Without loss of generality, we assume that
. Then, from (
4), we have
. So, once
, then, again, (
4) gives
. Applying (
4) repeatedly, we obtain
, which is a contradiction. □
Example 2. For the circulant graph , . Let us take and . Then, and , i.e., and , are given byNote that . If F contains neither nor , then and, consequently, because and due to Lemma 2. Example 3. For the circulant graph , . Let us take and . Then, and , i.e., and , are given byNote that . So, , provided that . Therefore, if , then because and due to Lemma 2. 4. Optimal Fault-Tolerant Resolving Set with Tolerance
For two subsets and of vertices of G, let us define . For a set U, we denote the set of all common antipodal vertices of the elements in U by . A vertex x resolves the vertices u and v if . From here onward, by , , we mean that U is a set of any two consecutive vertices of , so a is a non-negative integer and less than n.
Lemma 6. Let S be a set of consecutive vertices of , such that , where , . If S contains at most one vertex of U, then for each , for all distinct .
Proof. For symmetries of , without loss of generality, we can assume that , i.e., . Assume to the contrary that there are such that for some . Then, , i.e., with . Thus, . Since , we may write for some positive integer s and . Then, and , where . Since , , which is a contradiction. This contradiction leads to the result. □
The following lemma can be proved using a similar argument to that in the proof of Lemma 6.
Lemma 7. Let S be a set of consecutive vertices of such that . If S contains at most one vertex of , , then for each , for all distinct .
Lemma 8. Let S be a set of consecutive vertices of , , such that , where , . Then, the elements of U are resolved by at least or at least elements of S according to whether is even or odd.
Proof. Let , where and . Then, . Since S is a set of consecutive vertices and , applying Lemma 7, we obtain for all .
contains at most one element of . Since for all , and , i.e., x and , are resolved by s in if and only if . Again, since for all , the last inequality holds only when is even. Since for distinct v and w of S, , there are or elements in S when is even, according to whether is even or odd.
is a proper subset of . If and are the first two elements or last two elements of S, then (say). Therefore, contains no elements of U and, by a similar argument to that in , the elements of U are resolved by elements of because . Since both and resolve the elements of U, there are at least elements of S that resolve U. Now, we assume that both and are internal vertices of S. We divide S into two disjoint segments and that contain and , respectively. Then, applying the same arguments as those in to both and , we see that and are resolved by elements of and elements of . Thus, and are resolved by elements of . □
Lemma 9. Let S be a set of consecutive vertices of such that . Then, the elements of , are resolved by at least or at least elements of S according to whether is even or odd.
Proof. By a similar argument to that in the proof of Lemma 8, we can prove the result. □
Lemma 10. Suppose that and are two subsets of , where or and . Then, any two consecutive vertices in are resolved by at least elements of .
Proof. We consider the following two cases:
. For symmetries of and the properties of the sets and , it is sufficient to prove the result for two consecutive vertices and of . Then, for some t with Let . Then, . We now consider the following cases.
. First, we assume that . Then, applying Lemma 8 to both and and combining them, the vertices and are resolved by at least elements. Again, and are also resolved by . Thus, there are at least elements of that resolve and when . Next, we assume that . Then, applying Lemma 8 to the sets and and a applying similar argument, we obtain the result.
. In this case, and , where and }. Note that and . If t is even, then applying Lemma 8 to both and , we have the following:
- (a)
The elements of U are resolved by at least elements of ;
- (b)
The elements of U are resolved by at least or at least elements of according to whether k is even or odd.
Thus, when t is even, then the elements of U are resolved by at least elements or at least elements of according to whether k is even or odd. Again, applying the same argument as is used in the proof in of Lemma 8, the elements of U are resolved by at least elements of . If t is even, then the value of is or according to whether k is even or odd. Thus, combining all of the above facts, there are at least elements in that resolve the elements of U when t is even. Using a similar argument, we can prove the result for an odd integer t.
. Let and . Since , , gives either or . We prove the result for . The proof for will be similar. Since , . For , we divide into two parts, and . Then, applying Lemma 9 twice for and for , we see that and are resolved by at least elements of . For , we partition as , where and . Since every vertex in has a unique antipodal vertex, and and are antipodal vertices of and , respectively, each vertex in resolves and . Again, applying Lemma 9 to and combining them, we see that and are resolved by at least elements from . Thus, and are resolved by at least elements of when . Now, , which shows that the result is true when . For the remaining values of i, i.e., , we can prove the result by a similar argument. □
Example 4. We consider the circulant graph (i.e., ) and . For this graph, and . The codes of the vertices in are given in the matrix below. In this matrix, the columns corresponding to the vertex represent the code of with respect to .It is easy to observe that any two consecutive columns differ by at least four places. Lemma 11. For and an odd integer , let and be two sets of vertices of . Then, any two consecutive vertices of are resolved by at least elements of .
Proof. Let and be two consecutive vertices of and . Recall that , where the indices are taken to be modulo .
. In this case, or . If , then for . Since k is odd and both and contain elements, applying Lemma 8 twice for and combining them, we see that the vertices of U are resolved by at least . Otherwise, . Then, . Let . Then, , and also and . Thus, both and are even, so applying Lemma 8, we obtain the result when .
. As and , implies either or . First, we assume that . Then, . Then, by a similar argument to that in the proof in of Lemma 8, we see that the vertices and are resolved by at least elements of . Now, for , we divide into two segment and . Then, for , and and . Applying Lemma 8 to and , the vertices are resolved by or elements of according to whether i is odd or even. Again, and are resolved by at least elements of , i.e., by or elements of according to whether i is odd or even. Hence, we obtain the result when . For , it is easy to observe that the elements of U are resolved by at least elements of . Next, we assume that . Similarly, we prove the result. □
Example 5. In this example, we consider the circulant graph and . For this graph, and . The codes of the vertices in are given in the matrix below. In this matrix, the columns corresponding to the vertex represent the code of with respect to .It is easy to observe that any two consecutive columns differ by at least six places. Remark 2. If we drop the condition on k, i.e., if we consider k to be an even integer, then the result of Lemma 11 may not hold. For example, we take the circulant graph and . Then, and . The codes of the vertices and (two consecutive vertices outside of ) with respect to are given by Note that and differ by at exactly four places not in places, where .
By a similar argument to that described in Lemma 11, we can prove the following result.
Lemma 12. For an even integer k, let and be two sets of vertices of . Then, any two consecutive vertices of are resolved by at least elements of .
Lemma 13. For an odd integer k, let and be two sets of vertices . Then, any two consecutive vertices of are resolved by at least elements of .
Proof. Let
and
be two consecutive vertices of
and
. Recall that
, where the indices are taken to be modulo
. Thus,
We consider the following two cases depending on the positions of
and
.
Neither nor is in . From the above, it is clear that for . Thus, applying Lemma 8 to both and , we see that the elements of U are resolved by at least elements from each of and .
Exactly one of and is in . Without loss of generality, we can assume that either or is in . Since is a set of consecutive vertices and exactly one of and is in , either or . For the first one, , and for the second, .
. In this case, and . Let . Then, and , an even integer. Thus, applying Lemma 8 to both and S, we see that the vertices of U are resolved by at least elements from each of and S.
. In this case, and {,,}. By applying the same argument as is used in and applying Lemma 8 to both and , we obtain the result.
Both and are in . Since and , in this case, either or . We assume that . The proof will be similar to when . Here, . Divide the set into two segments, and . Note that and . Then, using Lemma 8 and a similar argument to that used in of Lemma 10, we obtain the result. □
Example 6. For the circulant graph and , and . The codes of the vertices in are given in the matrix below. In this matrix, the columns corresponding to the vertex represent the code of with respect to . It is easy to observe that any two consecutive columns differ by at least six places. Remark 3. If we drop the condition on k, i.e., if we consider k to be an even integer, then the result of Lemma 13 may not hold. For example, we consider the circulant graph and . Then, and . The codes of the vertices and (two consecutive vertices in ) with respect to are given by Note that and differ by exactly four places not in places, where .
Lemma 14. Let be an even integer and and be two subsets of . Then, any two consecutive vertices in are resolved by at least elements of .
Proof. By a similar argument to that stated in the proof of Lemma 13, we can prove the result. □
Example 7. For the circulant graph and , and . The codes of the vertices in are given in the matrix below. In this matrix, the columns corresponding to the vertex represent the code of with respect to . It is easy to observe that any two consecutive columns differ by at least five places. Remark 4. If we drop the condition on k, i.e., if we consider k to be an odd integer, then the result of Lemma 14 may not hold. For example, we consider the circulant graph and . Then, and . The codes of the vertices and (two consecutive vertices in ) with respect to are given by Note that and differ by exactly five places not in places, where .
Theorem 2. Let n and k be integers such that . Then, the fault-tolerant metric dimension with the tolerance level k of the circulant graph is or according to whether or
Proof. In Theorem 1, it is shown that every fault-tolerant resolving set F with the tolerance level k of must contain at least or at least elements according to whether or . Thus, to prove the result, we need to construct a fault-tolerant resolving set F with the tolerance level k that contains and elements accordingly as and . Our claim is that the set F forms a fault-tolerant resolving set with the tolerance level k, where F is defined in the following manner.
- (a)
For or , .
- (b)
For and odd integer k, .
- (c)
For and even integer k, .
- (d)
For and odd integer k,
- (e)
For and even integer k,
Let and be any two vertices of . Since F contains at least elements, applying Lemma 1, the vertices and are resolved by at least element of F, provided that . So, we assume that , i.e., we take two consecutive vertices and . Our claim is that and are resolved by at least elements of F. When , by Lemmas 11 and 12, our claim is true. Again, by Lemmas 13 and 14, our claim is also true when n is of the form . Thus, we consider the following cases.
. Here, , where and . Let . If , then applying Lemma 7, the vertices of U are resolved by at least elements of . So, we now assume that , i.e., either or .
. Since , applying Lemma 8 twice for and , the vertices and are resolved by at least , i.e., at least elements of F. Thus, the proof is complete when k is odd. For an even integer k, we calculate and . Then, , an odd integer when k is even. So, at least one of and is even, and hence, applying Lemma 7, we see that and are resolved by at least elements.
. If k is odd, then by a similar argument to that in , our claim is true. So, we assume that k is even integer. Then, and , so , an odd integer. Thus, exactly one of and is odd, so applying Lemma 7 to both and , we see that our claim is true.
. In this case, . Here, . Now, , provided that . Again, Lemma 10 states that our claim is true when . This completes the proof of the theorem. □
Example 8. In this example, we take and the circulant graphs , where . Then, for , or {, , , , , , , } according to whether or . When ,Finally, for ,