Properties of Connected Graphs Related

to Metric and Partition Dimension

Since 1864

Name : Muhammad Imran

Year of Admission : 2005
Registration No. : 23-GCU-PHD-SMS-05

Abdus Salam School of Mathematical Sciences

GC University Lahore, Pakistan
Properties of Connected Graphs Related
to Metric and Partition Dimension
Submitted to

Abdus Salam School of Mathematical Sciences

GC University Lahore, Pakistan

in the partial fulfillment of the requirements for the award of degree of

Doctor of Philosophy
in

Mathematics
By

Name : Muhammad Imran

Year of Admission : 2005
Registration No. : 23-GCU-PHD-SMS-2005

Abdus Salam School of Mathematical Sciences

GC University Lahore, Pakistan

i
 DECLARATION

I, Mr. Muhammad Imran Registration No. 23-GCU-PHD-SMS-05 student at

Abdus Salam School of Mathematical Sciences GC University Lahore in the
subject of Mathematics admitted in 2005, hereby declare that the matter printed
in this thesis titled
“Properties of Connected Graphs Related to Metric and Partition
Dimension”

is my own work and that

(i) I am not registered for the similar degree elsewhere contemporaneously.

(ii) No direct major work had already been done by me or anybody else on
this topic; I worked on for the Ph.D. degree.
(iii) The work, I am submitting for the Ph.D. degree has not already been
submitted elsewhere and shall not in future be submitted by me for
obtaining similar degree from any other institution.

Dated: ------------------------- ------------------------------------

Signature

ii
 RESEARCH COMPLETION CERTIFICATE

Certified that the research work contained in this thesis titled

“Properties of Connected Graphs Related to Metric and Partition
Dimension”
has been carried out and completed by Mr. Muhammad Imran Registration No.
23-GCU-PHD-SMS-2005 under my supervision.

----------------------------- -------------------------------
Date Supervisor
Prof. Ioan Tomescu

Submitted Through

Prof. Dr. A. D. Raza Choudary --------------------------------

Director General Controller of Examination
Abdus Salam School of Mathematical Sciences GC University
GC University Lahore, Pakistan. Lahore, Pakistan.

iii
This dissertation is gratefully dedicated to my parents for
their prayers and support.

iv
Table of Contents

Table of Contents v

Acknowledgements vii

Introduction 1

1 Basic concepts 4
1.0.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.0.2 Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.0.3 Trees and spanning trees . . . . . . . . . . . . . . . . . . . . . 9
1.1 Distances in Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Metric dimension and partition dimension of connected graphs 13

2.1 Resolving Sets and Metric Dimension . . . . . . . . . . . . . . . . . . 13
2.2 Resolving Partitions and Partition Dimension . . . . . . . . . . . . . 15
2.3 Known Results on Metric Dimension and Partition Dimension . . . . 16

3 Families of plane graphs with constant metric dimension 21

3.1 Notation and preliminary results . . . . . . . . . . . . . . . . . . . . 21
3.2 The graph of convex polytope Dn . . . . . . . . . . . . . . . . . . . . 23
3.3 The graph of convex polytope Rn . . . . . . . . . . . . . . . . . . . . 29
3.4 The graph of convex polytope Qn . . . . . . . . . . . . . . . . . . . . 33

4 Metric dimension of generalized Petersen graphs P (n, 3) 39

4.1 Notation and auxiliary results . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Upper bounds for metric dimension of generalized Petersen graphs
P (n, 3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Metric dimension of P (n, 3) for n ≡ 0, 1(mod 6) . . . . . . . . . . . . 51

v
5 Metric dimension and d-sets of connected graphs 61
5.1 Notation and preliminary results . . . . . . . . . . . . . . . . . . . . 62
5.2 Properties of d-sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Graphs with unbounded metric dimension . . . . . . . . . . . . . . . 67
5.4 Metric dimension of necklace graph . . . . . . . . . . . . . . . . . . . 70

6 On metric and partition dimension of some infinite regular graphs 74

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.2 Main results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

7 Concluding remarks and open problems 88

Bibliography 90

vi
Acknowledgements

In the name of Allah, the most Gracious and the most Merciful. All praise and glory
to Allah for His blessing upon me to finish my thesis. I would like to thank my
supervisor Professor Ioan Tomescu. I learned so much about the process of research,
the conventions of the field, and the importance of clear concise writing from him,
and he acted as a terrific sounding board for discussing my research ideas, as well as
keeping me on track for finishing my thesis. I would also like to thank Professor A.
D. Raza Choudary who believed that I could actually finish this work even when I
doubted it myself. His creativity and breadth as a person have been a constant source
of inspiration for me.
My thanks are also for all colleagues from Abdus Salam School of Mathematical
Sciences for giving me a chance to improve my academic competence. I am especially
thankful to colleagues in Batch III for the wonderful environment they created in
ASSMS.
Of course, I am grateful to my parents for their patience and love. Without them
this work would never have come into existence (literally).
Finally, I wish to thank the following: All the foreign professors at ASSMS (for
they made ASSMS, a real place of learning); Ahtsham, Ali, Fozia, Kashif, Qadeer,
Gohar, Imran Javaid, Zahid, Ikhlaq, Aziz ... (for their friendship and for all the good
time we had together); Nauman and Awais (for so many things) and my brother
Zafran and sisters (because they always stood by me).

ASSMS, Lahore Muhammad Imran

December, 2009.

vii
Introduction

Navigation can be studied in a graph-structured framework in which the naviga-

tion agent (which we shall assume to be a point robot) moves from node to node of

a “graph space”. The robot can locate itself by the presence of distinctly labeled

“landmark” nodes in the graph space. For a robot navigating in Euclidean space,

visual detection of a distinctive landmark provides information about the direction to

the landmark, and allows the robot to determine its position by triangulation. On a

graph, however, there is neither the concept of direction nor that of visibility. Instead,

we shall assume that a robot navigating on a graph can sense the distances to a set

of landmarks.

Evidently, if the robot knows its distances to a sufficiently large set of landmarks,

its position on the graph is uniquely determined. This suggests the following prob-

lem: given a graph, what are the fewest number of landmarks needed, and where

they should be located, so that the distances to the landmarks uniquely determines

the robot’s position on the graph? This is actually a classical problem about metric

spaces. A minimum set of landmarks which uniquely determines the robot’s posi-

tion is called a “metric basis”, and the minimum number of landmarks is called the

“metric dimension” of a graph.

Motivated by the problem of uniquely determining the location of an intruder in a

1
 2

network, the concept of metric dimension was introduced by Slater in [36, 37] and

studied independently by Harary and Melter in [19]. Slater refereed to the metric

dimension of a graph as its “location number” and motivated the study of this in-

variant by its application to the placement of a minimum number of sonar/loran

detecting devices in a network so that the position of every vertex in the network can

be uniquely described in terms of its distances to the devices in the set.

For a subset S ⊂ V (G) and a vertex v of a connected graph G, the distance d(v, S)

between v and S is defined as usually, by d(v, S) = min{d(v, x) : x ∈ S}. If

Π = (S1 , S2 , ...., Sk ) is an ordered k- partition of V (G), the representation of v with

respect to Π is the k- tuple r(v|Π) = (d(v, S1 ), d(v, S2 ), ....., d(v, Sk )). If the k- tuples

r(v|Π) for v ∈ V (G) are all distinct, then the partition Π is called a resolving partition

and the minimum cardinality of a resolving partition of V (G) is called the partition

dimension of G and is denoted by pd(G).

These concepts have some applications in chemistry for representing chemical com-

pounds ([9],[25]) or to problems of pattern recognition and image processing, some of

which involve the use of hierarchical data structures [31].

This thesis is divided into six chapters. The first two chapters consist of basic con-

cepts and terminology of graphs and distances in graphs. In the third chapter, the

metric dimension of plane graphs induced by some classes of convex polytopes has

been determined and it was proved that these plane graphs have constant metric di-

mension and only three vertices appropriately chosen suffice to resolve all the vertices

of these graphs.

The fourth chapter deals with the metric dimension of generalized Petersen graphs

P (n, 3). In this chapter, we study the metric dimension of the generalized Petersen
 3

graphs P (n, 3) by giving a partial answer to an open problem raised in [8]. In the

fifth chapter, d-sets for connected graphs have been defined and it is shown that for

a connected graph G of order n and diameter 2 the number of pairs {x, y} such that

their d-sets are equal to V (G) is bounded above by bn2 /4c and it is conjectured that

this holds for any connected graph. A lower bound for the metric dimension of G,

dim(G) is proposed in terms of a family of d-sets of G having the property that ev-

ery subfamily containing at least r ≥ 2 members has an empty intersection. Three

sufficient conditions which guarantee that a family F = (Gn )n≥1 of graphs with un-

bounded order has an unbounded metric dimension are also proposed. Finally, d-sets

are used to show that dim(N en ) = 3 when n is odd and 2 otherwise, where N en is the

necklace graph of order 2n + 2. In the sixth chapter, we study the metric dimension

and partition dimension of some infinite regular graphs generated by tilings of the

plane by regular triangles and hexagons. It is shown that these graphs have no finite

metric bases but their partition dimension is finite and is evaluated in some cases and

it is proved that for every n ≥ 2 there exists finite induced subgraphs of these graphs

having metric dimension equal to n as well as infinite induced subgraphs with metric

dimension equal to three. Also some open problems are suggested in the seventh

chapter.
Chapter 1

Basic concepts

In this chapter, some basic concepts will be elaborated. It includes the concept of

distances in graphs and some relevant definitions. As the first part, some definitions,

notations and terminology in graph theory that will be used throughout this disser-

tation, are introduced. In the second section, we give some definitions related with

distances in graphs.

1.0.1 Preliminaries

A graph G consists of a nonempty set V (G) of elements, called vertices and a collection

E(G) of unordered pairs of vertices called edges. A graph is symbolically represented

as G = (V (G), E(G)), unless otherwise specified, both V (G) and E(G) are finite.

The order of a graph is the number its vertices, and its size is the number of its

edges. If u and v are two vertices of a graph and if the unordered pair {u, v} is an

edge denoted by e = uv, we say that e joins u and v or that it is an edge between u

and v. In this case, the vertices u and v are said to be incident on e and e is incident

to both u and v.

4
 5

Two or more edges that join the same pair of distinct vertices are called parallel

edges. An edge represented by an unordered pair in which the two elements are not

distinct is known as a loop. A graph with no loops is a multigraph. A simple graph

is a graph with no parallel edges and loops. Unless otherwise stated, the graphs we

consider are all simple.

Two distinct vertices u and v in G are adjacent if uv is an edge in G. In this case,

u is called the neighbor of v, and vice versa. The set of neighbors of a vertex v in

a graph G is denoted by NG (v). The number of neighbors of a vertex is called its

degree. More generally, the degree of a vertex v in a graph G is the number of edges

of G incident with v. The degree of a vertex v in graph G is denoted by dG (v). The

minimum, maximum and average degrees of the vertices of a graph G are denoted by

δ(G), ∆(G) and d(G), respectively. An isolated vertex is a vertex of degree zero. A

graph is k-regular if the degree of each vertex is k.

Let A ⊆ V (G). Then A is called a clique in G if every two vertices in A are

adjacent. On the other hand, A is called independent set in G if any two vertices

in A are not adjacent in G. The clique number of G is denoted by ω(G), and the

independent number of G is denoted by α(G); they are defined as follows:

ω(G) = max{|A| : A is a clique in G},

α(G) = max{|A| : A is an independent set in G}.

Two edges e 6= f are adjacent if they have an end vertex in common. If all the vertices

of G are pairwise adjacent, then G is complete. A complete graph on n vertices is

denoted by Kn ; a K3 is a triangle. A bipartite graph is a simple graph in which the

set of vertices can be partitioned into sets U and V such that every edge is between
 6

a vertex in U and a vertex in V . The complete bipartite graph Km,n is the graph

with m vertices in U and n vertices in V in which there is an edge between every

vertex in U and every vertex in V . The complete bipartite graph K1,n is called a

star. Two graphs G and H are isomorphic if there are bijections θ : V (G) → V (H)

Figure 1.1: The complete graph K4

U V

Figure 1.2: A bipartite graph

and ϕ : E(G) → E(H) such that vertex v and edge e are incident in G if and only

if vertex θ(v) and edge ϕ(e) are incident in H. The pair of mappings (θ, ϕ) is an

isomorphism from G to H; if G = H, it is an automorphism. Thus, we usually write

G = H rather than G ' H.

 7

The graph H = (W, F ) is a subgraph of the graph G = (V, E) if W is a subset of

V and F is a subset of E. For example, a cyclic graph of order 4 is a subgraph of K4 .

Any graph G0 for which G is a subgraph is called a supergraph of G. Any subgraph

H(V, F ) of G(V, E) is a spanning subgraph of G. A factor of a graph is a spanning

subgraph with at least one edge.

1.0.2 Connectivity

A subgraph F of a graph G is called an induced subgraph of G if whenever u and

v are vertices of F and uv is an edge of G, then uv is an edge of F as well. Let v

and w be two vertices in a graph. A walk between v and w in the graph is a finite

alternating sequence v = v0 , e1 , v1 , e2 , v2 , e3 , ..., en , vn = w of vertices and edges of the

graph such that each edge ei in the sequence joins vertices vi−1 and vi . The vertices

and the edges of the walk need not be distinct. Two walks v0 , e1 , v1 , e2 , v2 , e3 , ..., en , vn

and u0 , f1 , u1 , f2 , u2 , f3 , ...fm , um are said to be equal if n = m,vi = ui for 0 ≤ i ≤ n

and ei = fi for 1 ≤ i ≤ n. Two walks are said to be different if they are not equal.

The number of edges in a walk is the length of the walk. A walk in which no edge is

repeated is called a trail. A walk in which vertices are all distinct is a path. Obviously

every path is a trail. A closed walk in a graph is a walk between a vertex and itself. A

closed walk with no edge repeated is a circuit. A cycle is a circuit in which no vertex

is repeated. In a simple graph G, any cycle consisting of k vertices (that is passing

through k vertices) is a k-cycle in G. It is an odd cycle if k is odd and it is an even

cycle if k is even. A pair of vertices in a graph is a connected pair if there is a path

between them. A graph G is a connected graph if every pair of vertices in a graph G

is a connected pair; otherwise , it is a disconnected graph. A connected subgraph H

 8

of G is a component of G if H = H 0 whenever H 0 is a connected subgraph of G that

contains H. In other words, a component is a maximal connected subgraph.

A graph is connected if and only if the number of its components is one. If F

is a set of edges in a graph G = (V, E) , the graph obtained from G by deleting all

the edges belonging to F is denoted by G − F . If F consists of the single edge f

then G − F is written as G − f . A set F of edges in a connected graph G is called a

disconnecting set in G if G − F has more than one component. If a disconnecting set

F consists of a single edge f then edge f is called a bridge. The following theorem [8]

provides a necessary and sufficient condition for an edge of a graph G to be bridge.

Theorem 1.0.1. [8] An edge e of a graph G is a bridge if and only if e lies on no

cycle of G.

A graph is said to be k-edge connected if every disconnecting set in it has at least

k edges. The edge connectivity number λ(G) of a graph is the minimum size of a

disconnecting set of edges in G. A disconnecting set F is said to be a cut set if no

proper subset of F is a disconnecting set. Analogously, if W is a set of vertices in

G = (V, E), then the graph obtained from G by deleting all the vertices belonging

to W as well as the edges incident to the vertices in W is denoted by G − W . If W

consists of a single vertex w, the graph G − W is denoted by G − w. A set W of

vertices in a connected graph G is called a separating set (also known as the vertex

cut) in G if G − W has more than one component. If a separating set consists of a

single vertex w then w is known as cut vertex. The connectivity number κ(G) of a

graph G is the minimum size of a separating set of vertices in it. Since a complete

graph has no separating set, we adopt the convention that the connectivity number

of the complete graph of order n is n − 1 for all n. The following theorem provides us
 9

with inequalities concerning the connectivity, edge-connectivity, and minimum degree

of a graph.

Theorem 1.0.2. [8] For any graph G,

κ(G) ≤ λ(G) ≤ δ(G)

A graph G is said to be k-connected if κ(G) ≥ k. Thus Kn is (n − 1)-connected for

all n, and a graph that is not complete is k-connected if and only if every separating

set in it has at least k vertices. The connectivity number of a graph G is zero if and

only if G is either the trivial graph K1 or is a disconnected graph. A cyclic graph is

2-connected.

1.0.3 Trees and spanning trees

An acyclic graph is a graph with no cycles. A tree is a connected acyclic graph.

Theorem 1.0.3. [15] The following are equivalent in a graph G with n vertices.

(i) G is a tree.

(ii) There is a unique path between every pair of distinct vertices in G.

(iii) G is connected and every edge in G is a bridge.

(iv) G is connected, and has n − 1 edges.

(v) G is acyclic, and has n − 1 edges.

(vi) G is acyclic, and whenever any two arbitrary nonadjacent vertices in G are joined

by an edge, the resulting enlarged graph has a unique cycle.

 10

(vii) G is connected, and whenever any two arbitrary nonadjacent vertices in G are

joined by an edge, the resulting enlarged graph has a unique cycle.

An acyclic connected spanning subgraph (if it exists) of G is called a spanning

tree in G.

Theorem 1.0.4. [15] A graph is connected if and only if it has a spanning tree.

1.1 Distances in Graphs

Facility location problems deal with the task of choosing a site subject to some cri-

terion. For example, in determining where to locate an emergency facility such as

hospital or fire station, we would like to minimize the response time between the fa-

cility and the location of a possible emergency. Such situations deal with the concept

of centrality. In this section, we list the definitions of various kinds of centers and

other relevant definitions.

Let G be a connected graph. Then the distance d(u, v) between two vertices u

and v is the length of a shortest path between them. The eccentricity of a vertex

v ∈ V (G) denoted by e(v) is the distance to a vertex farthest from v. Thus e(v) =

max{d(u, v) : u ∈ V }. The radius r(G) is the minimum eccentricity of the vertices

of G whereas the diameter d(G) is the maximum eccentricity of the vertices of G.

Now v is a central vertex of G if e(v) = r(G) and the center C(G) is the set of all

central vertices. Thus the center consists of all vertices having minimum eccentricity.

A vertex v ∈ V (G) is called peripheral vertex if e(v) = d(G) and periphery P er(G) is

the set of all such vertices. For a vertex v all vertices at distances e(v) from v are called
 11

eccentric vertices. For a given vertex u in a connected graph G, we have discussed

seeking a vertex v such that d(u, v) = e(v), that is, v is a vertex that is farthest from

u. Such a vertex v is called an eccentric vertex of u. A vertex v is an eccentric vertex

of the graph G if v is an eccentric vertex of some vertex of G. A connected graph G is

an eccentric graph if every vertex of G is an eccentric vertex. The eccentric subgraph

Ecc (G) of G is the subgraph of G induced by the set of eccentric vertices of G. The

q r
3 2

3 s t
H: 3 Per(H) :
3 u w v
4 v

4 w 3 x

3 2
y z

rad(H)=2 and diam(H)=4

Figure 1.3: The eccentricities of the vertices of a graph

following theorem [8] provides a necessary and sufficient condition for a graph G to

be an eccentric subgraph of some graph.

Theorem 1.1.1. [8] A nontrivial graph G is the eccentric subgraph of some graph if
and only if every vertex of G has eccentricity 1 or no vertex of G has eccentricity 1.

A tree with one central vertex is called a central tree and a tree with two central

vertices is called bicentral. A graph G is a unique eccentric vertex graph if each vertex

in G has exactly one eccentric vertex. A graph G is self-centered if every vertex is in

the center. Thus in a self-centered graph every vertex has the same eccentricity as

r(G) = d(G).
 12

Theorem 1.1.2. [3] The center of a tree consists of either a single vertex or a pair
of adjacent vertices.

Theorem 1.1.3. [3] A unique eccentric vertex graph is self-centered if and only if
each node of G is eccentric.

The center of a graph is important in applications involving emergency facilities

where response time (distance) to each single location (vertex) in the region (graph)

is critical. Suppose instead we consider a service facility such as post office, shopping

mall, bank, or power station . When deciding where to locate a post office, we want

to minimize the average distance that a person serviced by the post office must travel.

This is equivalent to minimizing the total distance traveled by all people within the

district. For such situations, the concept of median is described.

Let G be a connected graph. The status s(v) of a vertex v in G is the sum of the

distances from v to each other vertex in G. This concept was introduced by Harary

[20]. The median M (G) of a graph G is the set of vertices with minimum status. The

minimum status ms(G) of a graph G is the value of the minimum status; the total

status ts(G) of a graph G is the sum of all the status values.

Chapter 2

Metric dimension and partition

dimension of connected graphs

In this chapter, we give the definitions of resolving sets, resolving partitions, metric

dimension and partition dimensions which are the key notions of this thesis. We de-

scribe the some known results for these invariants and applications of these invariants

to different branches of applied sciences.

2.1 Resolving Sets and Metric Dimension

If G is a connected graph, the distance d(u, v) between two vertices u, v ∈ V (G) is

the length of a shortest path between them. Let W = {w1 , w2 , ...., wk } be an ordered

set of vertices of G and let v be a vertex of G. The representation r(v|W ) of v with

respect to W is the k-tuple (d(v, w1 ), d(v, w2 ), ....., d(v, wk )). If distinct vertices of G

have distinct representations with respect to W , then W is called a resolving set for

G[1]. A resolving set of minimum cardinality is called a basis for G and this cardi-

nality is the metric dimension of G, denoted by dim(G).

For a given ordered set of vertices W = {w1 , w2 , ...., wk } of a graph G, the i-th

13
 14

component of r(v|W ) is 0 if and only if v = wi . Thus, to show that W is a resolv-

ing set it suffices to verify that r(x|W ) 6= r(y|W ) for each pair of distinct vertices

x, y ∈ V (G)\W .

For example, consider the graph G of Fig. 2.1. The set W1 = {v1 , v3 } is not a resolv-

v1
v
2
v
3

v v
4 5

Figure 2.1: A graph with dim(G)=2

ing set of G since r(v2 |W1 ) = (1, 1) = r(v4 |W1 ). On the other hand, W2 = {v1 , v2 , v3 }

is a resolving set for G since the representation for the vertices of G with respect

to W2 are r(v1 |W2 ) = (0, 1, 1), r(v2 |W2 ) = (1, 0, 1), r(v3 |W2 ) = (1, 1, 0), r(v4 |W2 ) =

(1, 2, 1), r(v5 |W2 ) = (2, 1, 1). However, W2 is not a minimum resolving set since

W3 = {v1 , v2 } is also a resolving set. Since no single vertex constitutes a resolving set

for G, it follows that W3 is a minimum resolving set implying that dim(G) = 2.

A useful property in finding dim(G) is the following:

Lemma 2.1.1. Let W be a resolving set for a connected graph (G) and u, v ∈ V (G).

If d(u, w) = d(v, w) for all vertices w ∈ V (G) \ {u, v}, then {u, v} ∩ W 6= ∅.

Slater refereed to the metric dimension of a graph as its location number and

motivated the study of this invariant by its application to the placement of a minimum
 15

number of sonar/loran detecting devices in a network so that the position of every

vertex in the network can be uniquely described in terms of its distances to the devices

in the set ([36],[37]).

2.2 Resolving Partitions and Partition Dimension

Another kind of dimension of a connected graph, called partition dimension was

introduced in [10, 11] as follows: For a subset S ⊂ V (G) and a vertex v of a con-

nected graph G, the distance d(v, S) between v and S is defined as usually , by

d(v, S) = min{d(v, x) : x ∈ S}. If Π = (S1 , S2 , ...., Sk ) is an ordered k- parti-

tion of V (G), the representation of v with respect to Π is the k- tuple r(v|Π) =

(d(v, S1 ), d(v, S2 ), ....., d(v, Sk )). If the k- tuples r(v|Π) for v ∈ V (G) are all distinct,

then the partition Π is called a resolving partition and the minimum cardinality of a

resolving partition of V (G) is called the partition dimension of G and is denoted by

pd(G). Let Π = {S1 , S2 , ...., Sk } be an ordered partition of V (G). If u ∈ Si , v ∈ Sj

where 1 ≤ i, j ≤ k and i 6= j, then r(u|Π) 6= r(v|Π) since d(u, Si ) = 0 but d(v, Si ) 6= 0.

Thus, when determining whether a given partition Π of V (G) is a resolving partition

for V (G), we need only to verify if the vertices of G belonging to the same class

of Π have distinct representations with respect to Π. When d(u, Si ) 6= d(v, Si ) we

shall say that the class Si distinguishes vertices u and v. Another useful property in

determining pd(G) is the following lemma [11].

Lemma 2.2.1. Let Π be a resolving partition of V (G) and u, v ∈ V (G). If d(u, w) =

d(v, w) for all vertices w ∈ V (G) \ {u, v}, then u and v belong to different classes of

Π.
 16

These concepts have some applications in chemistry for representing chemical

compounds ([9], [25]) or to problems of pattern recognition and image processing,

some of which involve the use of hierarchical data structures [31].

2.3 Known Results on Metric Dimension and Par-

tition Dimension

If G is a nontrivial connected k-dimensional graph of order n, then 1 ≤ k ≤ n − 1.

Here we present some results which have been established already. The following

result was established in [9].

Theorem 2.3.1. For every pair n, k of integers with 1 ≤ k ≤ n − 1, there exists a

k-dimensional graph of order n.

The following two theorems ([9],[19],[36],[37]) characterize connected graphs of

order n with dimension 1, n − 1 or n − 2 and provide the dimensions of some well-

known classes of graphs.

For two vertex-disjoint graphs G and H, G ∪ H is disconnected graph with vertex set

V (G) ∪ V (H) and edge set E(G) ∪ E(H). The join G + H consists of G ∪ H and all

edges joining a vertex of G and a vertex of H.

Theorem 2.3.2. Let G be a connected graph of order n ≥ 2. Then

(a) dim(G) = 1 if and only if G = Pn .

(b) dim(G) = n − 1 if and only if G = Kn .

(c) for n ≥ 4, dim(G) = n − 2 if and only if G = Kr,s ,r, s ≥ 1, G = Kr + K̄s , r ≥

1, s ≥ 2 or G = Kr + (K1 ∪ Ks ), r, s ≥ 1.
 17

We will use the terminology given in [9], [34] and [35]. A vertex of degree at

least 3 will be referred to as a major vertex. An end-vertex u of a graph T is said

to be a terminal vertex of a major vertex v of T if d(u, v) < d(u, w) for every other

major vertex w of T . The terminal degree ter(v) of a major vertex v is the number

of terminal vertices of v. A major vertex v of T is an exterior major vertex of T if it

has a positive terminal degree. Let σ(T ) denote the sum of the terminal degrees of

the major vertices of T and let ex(T ) denote the number of exterior major vertices

of T .

Theorem 2.3.3. [19] If T is a tree that is not a path, then

dim(T ) = σ(T ) − ex(T ).

A fundamental question in graph theory concerns how the value of a parameter is

affected by making a small change in the graph. The dimension of a connected graph

G is also affected by the addition of a single edge. It has been proved in [9] that

the dimension of a tree is also affected by the addition of a single edge. The result

suggests that the dimension can increase by at most 1 or decrease at most by 2. We

state the theorem.

Theorem 2.3.4. [9] Let T be a tree of order at least 3 and let e be an edge in its

complement T̄ . Then

dim(T ) − 2 ≤ dim(T + e) ≤ dim(T ) + 1.

Moreover, for each integer i with −2 ≤ i ≤ 1, there exists a tree Ti and an edge ei in

T̄i such that

dim(Ti + ei ) = dim(Ti ) + i.
 18

A Cartesian product of two graphs G and H, denoted by G × H, is the graph with

vertex set V (G) × V (H), where two vertices (x, x0 ) and (y, y 0 ) are adjacent if and

only if x = y and x0 y 0 ∈ E(H) or x0 = y 0 and xy ∈ E(G). The relationship between

the dimension of cartesian product H × K2 of a connected graph H and K2 and the

dimension of H was also established in [9].

Theorem 2.3.5. For every nontrivial connected graph H,

dim(H) ≤ dim(H × K2 ) ≤ dim(H) + 1.

The metric dimension of cartesian product of graphs has been studied in [5] and

[33]. The following result in [9] gives bounds for the dimension of a graph in terms

of its order and diameter.

Theorem 2.3.6. For positive integers d and n with d < n, define f (n, d) as the least

positive integer k such that k + dk ≥ n. Then for a connected graph G of order n ≥ 2

and diameter d,

f (n, d) ≤ dim(G) ≤ n − d.

A sharp lower bound for the dimension of a connected graph G in terms of its

maximum degree ∆(G) was established in [13].

Theorem 2.3.7. Let G be a connected nontrivial connected graph. Then

dim(G) ≥ dlog3 (∆(G) + 1)e

and this bound is sharp.

In fact, for each pair k, 4 of integers such that 3k = 4+1, there exists a connected

graph Gk,4 such that dim(Gk,4 ) = k and 4(Gk,4 ) = 4.

 19

Now we present some theorems about the partition dimension of graphs. If G is a

connected graph of order n ≥ 2, then certainly 2 ≤ pd(G) ≤ n. It was also shown

in [11] that every pair k, n of integers with 2 ≤ k ≤ n is realizable as the partition

dimension and order of some connected graph.

Theorem 2.3.8. For every pair k, n of integers with 2 ≤ k ≤ n, there exists a

connected graph of order n with partition dimension k.

The partition dimension and metric dimension of a connected graph are related.

In [11], the following theorem has been proved.

Theorem 2.3.9. If G is a nontrivial connected graph, then

pd(G) ≤ dim(G) + 1.

Moreover, for every pair a, b of positive integers with d 2b e + 1 ≤ a ≤ b + 1, there exists

a connected graph G such that pd(G) = a and dim(G) = b.

In [11], an open problem was proposed concerning Theorem 2.3.9. Is it the case

that
dim(G)
pd(G) ≥ d e+1
2
for every nontrivial connected graph G? Tomescu gave a negative answer to this

question in [40].

Chartrand and Zhang obtained an improved upper bound for pd(G) in terms of

order and diameter of G in [11].

Theorem 2.3.10. If G is a connected graph of order n ≥ 3 and diameter d, then

pd(G) ≤ n − d + 1.
 20

Chartrand, Salehi and Zhang determined the partition dimension of some well

known classes of the graphs in [11] where the connected graphs of order n with

pd(G) = 2, n, n − 1 are characterized.

Theorem 2.3.11. Let G be a nontrivial connected graph of order n. Then

(a) pd(G) = 2 if and only if G = Pn

(b) pd(G) = n if and only if G = Kn and

(c) for n ≥ 3, pd(G) = n − 1 if and only if

G ∈ {K1,n−1 , Kn − e, K1 + (K1 ∪ Kn−1 )}.

Tomescu [40] characterized all graphs of order n ≥ 9 having partition dimension

n − 2 thus completing the characterization of graphs of order n having partition

dimension 2, n or n − 1 given by Chartrand, Salehi and Zhang. The list of these

graphs includes 23 members.

Chartrand, Salehi, Zhang also studied the partition dimension of a tree in [12].

Although, the partition dimension of some special type of trees, such as paths, stars,

double stars, and caterpillars, have been determined, a general formula for the parti-

tion dimension of a tree is still unknown. However it was shown that there is no tree

of order n with partition dimension n − 2.

The partition dimension as well as the connected partition dimension of the wheel

Wn with n spokes has been determined in [41]. The metric dimension and partition

dimension of Cayley digraphs have been studied in [16] and [17].

Chapter 3

Families of plane graphs with

constant metric dimension

The metric dimension of some classes of plane graphs has been determined in [4],

[5], [7], [24], [27] and [39]. In this chapter, we extend this study by considering

a class of plane graphs defined in [1] and two classes of plane graphs which are

generated by the convex polytopes defined in [2]. We show that these classes of plane

graphs have constant metric dimension and only three vertices appropriately chosen

suffice to resolve all the vertices of these plane graphs. It is natural to ask for the

characterization of families of plane graphs with constant metric dimension.

3.1 Notation and preliminary results

A graph G is said to be plane if it is drawn on the Euclidean plane such that edges

do not cross each other except at vertices of the graph.

By denoting G + H the join of G and H a wheel Wn is defined as Wn = K1 + Cn , for

n ≥ 3, a f an is fn = K1 + Pn for n ≥ 1 and Jahangir graph J2n , (n ≥ 2) (also known

as gear graph) is obtained from the wheel W2n by alternately deleting n spokes.

21
 22

Buczkowski et al. [4] determined the dimension of wheel Wn , Caceres et al. [7] the

dimension of f an fn and Tomescu and Javaid [39] the dimension of Jahangir graph

J2n .

Theorem 3.1.1. ([4], [7], [39]) Let Wn be a wheel of order n ≥ 3, fn be fan of order

n ≥ 1 and J2n be a Jahangir graph. Then

(i) For n ≥ 7, dim(Wn ) = b 2n+2

5
c;

(ii) For n ≥ 7, dim(fn ) = b 2n+2

5
c;

(iii) For n ≥ 4, dim(J2n ) = b 2n

3
c.

The metric dimension of all these plane graphs depends upon the number of ver-

tices in the graph.

On the other hand, we say that a family G of connected graphs is a family with

constant metric dimension if dim(G) is finite and does not depend upon the choice

of G in G. In [9] it was shown that a graph has metric dimension 1 if and only if

it is a path, hence paths on n vertices constitute a family of graphs with constant

metric dimension. Similarly, cycles with n(≥ 3) vertices also constitute such a family

of graphs as their metric dimension is 2 and does not depend upon on the number of

vertices n. In [5] it was proved that

(
2, if n is odd;
dim(Pm × Cn )=
3, if n is even.
Since prisms Dn are the trivalent plane graphs obtained by the cross product of path

P2 with a cycle Cn , this implies that

(
2, if n is odd;
dim(Dn )=
3, if n is even.
 23

So, prisms constitutes a family of 3-regular graphs with constant metric dimension.

Also Javaid et al. proved in [24] that the plane graph antiprism An constitute a

family of regular graphs with constant metric dimension as dim(An ) = 3 for every

n ≥ 5. The prism and the antiprism are Archimedean convex polytopes defined

e.g. in [26]. A grid Gm

n is obtained by the cartesian product of two paths Pn by

Pm . In [27], it was shown that dim(Pn × Pm ) = 2, so grids constitute a family of

plane graphs with constant metric dimension as their metric dimension is finite and

does not depend upon the number of vertices in the graph. Note that the problem of

determining whether dim(G) < k is an N P -complete problem [18].

In this chapter, we study the metric dimension of some plane graphs which are the

convex polytops defined in [1] and [2]. In the second section, we study the metric

dimension of graph of convex polytope Dn consisting of 5-sided faces and n-sided

faces. In the third section, we investigate the metric dimension of the graph of convex

polytope Rn which is obtained by the combination of the graph of a prism and graph

of an antiprism. In fourth section, metric dimension of the graph of convex polytope

Qn consisting of 3-sided faces, 4-sided faces, 5-sided faces and n-sided faces has been

determined.

3.2 The graph of convex polytope Dn

Let the graph of antiprism An [1] be given. We insert a vertex an+1 inside the n-gone

P and bn+1 inside the n-gone P 0 .

We join any vertex ai of P with the vertex an+1 and any vertex bi of P 0 with the

vertex bn+1 for i = 1, 2, ..., n. Thus we obtain the graph A0n . The dual graph to A0n

with vertices a1 , a2 , . . . , an ; b1 , b2 , ..., bn ; c1 , c2 , ..., cn ; d1 , d2 , ..., dn (Fig 3.1) is the graph

 24

d1 dn

c1
d2 cn dn−1
b1
c2
b2 bn cn−1

b3 a1 bn−1
c3 a2 an dn−2
d3
a3 an−1 cn−2

Figure 3.1: The graph of convex polytope Dn

of the convex polytope defined in [1] and will be denoted by Dn . It has 5-sided faces

and n-sided faces.

For our purpose, we call the cycle induced by {a1 , a2 , ..., an }, the inner cycle, cycle

induced by {d1 , d2 , ..., dn }, the outer cycle, set of vertices {b1 , b2 , ..., bn } which are con-

nected with the inner cycle, the set of inner vertices, the set of vertices {c1 , c2 , ..., cn }

which are connected with the outer cycle, the set of outer vertices.

In the next theorem, we show that the metric dimension of the graph of convex poly-

tope Dn is 3. Note that the choice of appropriate basis vertices (also referred to as

landmarks in [27]) is core of the problem.

Theorem 3.2.1. For n ≥ 4, let the graph of convex polytopes be Dn ; then dim(Dn ) =

3.

Proof. We will prove the above equality by double inequalities. We consider two

cases.

Case(i) When n is even.

In this case, we can write n = 2k, k ≥ 2, k ∈ Z+ . Let W = {a1 , a2 , ak+2 } ⊂ V (Dn ),

 25

we show that W is a resolving set for Dn in this case. For this we give representations

of any vertex of V (Dn ).

Representations of the vertices on inner cycle are


 (i − 1, i − 2, k − i + 2),

3 ≤ i ≤ k + 1;
r(ai |W ) =
 (2k − i + 1, 2k − i + 2, i − k − 2), k + 3 ≤ i ≤ 2k.


Representations of the set of inner vertices are





 (1, 2, k), i = 1;

r(bi |W ) = (i, i − 1, k − i + 3), 2 ≤ i ≤ k + 1;



 (2k − i + 2, 2k − i + 3, i − k − 1), k + 2 ≤ i ≤ 2k.


Representations of the set of outer vertices are




 (2, 2, k + 1), i = 1;



 (i + 1, i, k − i + 3),

2 ≤ i ≤ k;
r(ci |W ) =
(k + 1, k + 1, 2), i = k + 1;






 (2k − i + 2, 2k − i + 3, i − k), k + 2 ≤ i ≤ 2k.


Representations of the vertices on outer cycle are




 (3, 3, k + 2), i = 1;



 (i + 2, i + 1, k − i + 4),

2 ≤ i ≤ k;
r(di |W ) =
(k + 2, k + 2, 3), i = k + 1;






 (2k − i + 3, 2k − i + 4, i − k + 1), k + 2 ≤ i ≤ 2k.


We note that there are no two vertices having the same representations implying that

dim(Dn ) ≤ 3.

On the other hand, we show that dim(Dn ) ≥ 3. Suppose on contrary that dim(Dn ) =
 26

2, then there are following possibilities to be discussed.

(1) Both vertices are in the inner cycle. Without loss of generality we can sup-

pose that one resolving vertex is a1 . Suppose that the second resolving vertex is

ai (2 ≤ i ≤ k + 1). Then for 2 ≤ i ≤ k, we have r(an |{a1 , ai }) = r(b1 |{a1 , ai }) and

for i = k + 1 we have r(a2 |{a1 , ak+1 }) = r(an |{a1 , ak+1 }), a contradiction.

(2) Both vertices belong to the set of inner vertices. Without loss of generality we

can suppose that one resolving vertex is b1 . Suppose that the second resolving vertex

is bi (2 ≤ i ≤ k + 1). Then for 2 ≤ i ≤ k, we have r(an |{b1 , bi }) = r(dn |{b1 , bi }) and

for i = k + 1 we have r(b2 |{b1 , bk+1 }) = r(bn |{b1 , bk+1 }), a contradiction.

(3) Both vertices belong to the set of outer vertices. By the symmetry of the graph,

this case is analogous to case (2).

(4) Both vertices are in the outer cycle. By the same argument this case is analogous

to case (1).

(5) One vertex is in the inner cycle and other belongs to the set of inner vertices.

Without loss of generality we can suppose that one resolving vertex is a1 . Sup-

pose that the second resolving vertex is bi (1 ≤ i ≤ k + 1). Then for i = 1, we

have r(a2 |{a1 , b1 }) = r(an |{a1 , b1 }). If i = 2, r(d1 |{a1 , b2 }) = r(b3 |{a1 , b2 }). For

3 ≤ i ≤ k, we have r(an |{a1 , bi }) = r(b1 |{a1 , bi }) and when i = k + 1, we have

r(a2 |{a1 , bk+1 }) = r(an |{a1 , bk+1 }), a contradiction.

(6) One vertex is in the inner cycle and other belongs to the set of outer vertices. With-

out loss of generality we can suppose that one resolving vertex is a1 . Suppose that the

second resolving vertex is ci (1 ≤ i ≤ k + 1). Then for i = 1, we have r(a3 |{a1 , c1 }) =

r(bn |{a1 , c1 }). If i = 2, r(a3 |{a1 , c2 }) = r(c1 |{a1 , c2 }). For 3 ≤ i ≤ k − 1, we have

r(b1 |{a1 , ci }) = r(an |{a1 , ci }). When i = k, we have r(c1 |{a1 , ck }) = r(cn |{a1 , ck })
 27

and if i = k + 1, we have r(b2 |{a1 , ck+1 }) = r(c1 |{a1 , ck+1 }), a contradiction.

(7) One vertex is in the inner cycle and other in the outer cycle. Without loss of gener-

ality we can suppose that one resolving vertex is a1 . Suppose that the second resolving

vertex is di (1 ≤ i ≤ k + 1). Then for i = 1, we have r(b2 |{a1 , d1 }) = r(cn |{a1 , d1 }).

If i = 2, r(b2 |{a1 , d2 }) = r(dn |{a1 , d2 }). For 3 ≤ i ≤ k − 1, we have r(a2 |{a1 , di }) =

r(b1 |{a1 , di }) and when i = k, k + 1 we have r(b2 |{a1 , di }) = r(c1 |{a1 , di }), a contra-

diction.

(8) One vertex belongs to the set of inner vertices and other belongs to the set of

outer vertices. Without loss of generality we can suppose that one resolving vertex

is b1 . Suppose that the second resolving vertex is ci (1 ≤ i ≤ k + 1). Then for

i = 1, we have r(a1 |{b1 , c1 }) = r(cn |{b1 , c1 }). If i = 2, r(a2 |{b1 , c2 }) = r(d1 |{b1 , c2 }).

For 3 ≤ i ≤ k, we have r(a1 |{b1 , ci }) = r(c1 |{b1 , ci }) and when i = k + 1, we have

r(a1 |{b1 , ck+1 }) = r(cn |{b1 , ck+1 }), a contradiction.

(9) One vertex belongs to the set of inner vertices and other in the outer cycle. This

case is analogous to case (6).

(10) One vertex belongs to the set of outer vertices and other is in the outer cycle.

This case is analogous to case (5).

Hence, from above it follows that there is no resolving set with two vertices for V (Dn )

implying that dim(Dn ) = 3 in this case.

Case(ii) When n is odd.

In this case, we can write n = 2k + 1, k ≥ 2, k ∈ Z+ . Again we show that

W = {a1 , a2 , ak+2 } ⊂ V (Dn ) is a resolving set for Dn in this case. For this we

give representations of any vertex of V (Dn ).

Representations of the vertices on inner cycle are

 28


 (i − 1, i − 2, k − i + 2),

3 ≤ i ≤ k + 1;
r(ai |W ) =
 (2k − i + 2, 2k − i + 3, i − k − 2), k + 3 ≤ i ≤ 2k + 1.


Representations of the set of inner vertices are




 (1, 2, k), i = 1;



 (i, i − 1, k − i + 3),

2 ≤ i ≤ k + 1;
r(bi |W ) =
(k + 1, k + 1, 1), i = k + 2;






 (2k − i + 3, 2k − i + 4, i − k − 1), k + 3 ≤ i ≤ 2k + 1.


Representations of the set of outer vertices are





 (2, 2, k + 2), i = 1;

r(ci |W ) = (i − 1, i, k − i + 3), 2 ≤ i ≤ k + 1;



 (2k − i + 3, 2k − i + 4, i − k), k + 2 ≤ i ≤ 2k + 1.


Representations of the vertices on outer cycle are





 (3, 3, k + 3), i = 1;

r(di |W ) = (i + 2, i + 1, k − i + 4), 2 ≤ i ≤ k + 1;



 (2k − i + 4, 2k − i + 5, i − k + 1), k + 2 ≤ i ≤ 2k + 1.


Again we see that there are no two vertices having the same representations which

implies that dim(Dn ) ≤ 3.

On the other hand, suppose that dim(Dn ) = 2, then there are the same possibilities as

in case (i) and contradiction can be deduced analogously. This implies that dim(Dn ) =

3 in this case, which completes the proof.

 29

3.3 The graph of convex polytope Rn

For n ≥ 5, by Rn we denote the graph of convex polytope defined in [2] which is

obtained as a combination of the graph of a prism and the graph of an antiprism (Fig

3.2). We make the convention that an+1 = a1 , bn+1 = b1 , cn+1 = c1 , dn+1 = d1 to

simplify the notation.

For our purpose, we call the cycle induced by {ai : 1 ≤ i ≤ n} the inner cycle, cycle

c1 cn

c2 cn−1
b1
bn
b2
bn−1
a1
c3 a2 an cn−2
b3
a3 an−1
bn−2

Figure 3.2: The graph of convex polytope Rn

induced by {bi : 1 ≤ i ≤ n} the middle cycle, cycle induced by {ci : 1 ≤ i ≤ n} the

outer cycle.

In the next theorem, we prove that only three vertices suffice to resolve all the vertices

of the graph of convex polytope Rn . Again, choice of appropriate landmarks is crucial.

Theorem 3.3.1. Let Rn denotes the graph of convex polytope; then dim(Rn ) = 3 for

every n ≥ 5.

Proof. We will prove the above equality by double inequalities. We consider two

cases.
 30

Case(i) When n is odd.

In this case, we can write n = 2k+1, k ≥ 2, k ∈ Z+ . Let W = {a1 , a2 , ak+2 } ⊂ V (Rn ),

we show that W is a resolving set for Rn in this case. For this we give representations

of any vertex of V (Rn ).

Representations of the vertices on inner cycle are


 (i − 1, i − 2, k − i + 2),

3 ≤ i ≤ k + 1;
r(ai |W ) =
 (2k − i + 2, 2k − i + 3, i − k − 2), k + 3 ≤ i ≤ 2k + 1.


Representations of the vertices on middle cycle are





 (1, 1, k + 1), i = 1;

r(bi |W ) = (i, i − 1, k − i + 2), 2 ≤ i ≤ k + 1;



 (2k − i + 2, 2k − i + 3, i − k − 1), k + 2 ≤ i ≤ 2k + 1.


Representations of the vertices on outer cycle are





 (2, 2, k + 2), i = 1;

r(ci |W ) = (i + 1, i, k − i + 3), 2 ≤ i ≤ k + 1;



 (2k − i + 3, 2k − i + 4, i − k), k + 2 ≤ i ≤ 2k + 1.


We note that there are no two vertices having the same representations implying that

dim(Rn ) ≤ 3.

On the other hand, we show that dim(Rn ) ≥ 3. Suppose on contrary that dim(Rn ) =

2, then there are following possibilities to be discussed.

(1) Both vertices are in the inner cycle. Without loss of generality we can sup-

pose that one resolving vertex is a1 . Suppose that the second resolving vertex is

ai (2 ≤ i ≤ k + 1). Then for 2 ≤ i ≤ k we have r(an |{a1 , ai }) = r(bn |{a1 , ai }) and

 31

for i = k + 1 we have r(an |{a1 , ak+1 }) = r(b1 |{a1 , ak+1 }), a contradiction.

(2) Both vertices are in the middle cycle. In this case, we can fix b1 as a resolv-

ing vertex. Suppose that the second resolving vertex is bi (2 ≤ i ≤ k + 1), then

r(a1 |{b1 , bi }) = r(c1 |{b1 , bi }), a contradiction.

(3) Both vertices are in the outer cycle. Here we can fix c1 as a resolving vertex.

Suppose that the second resolving vertex is ci (2 ≤ i ≤ k + 1). Then for 2 ≤ i ≤ k

we have r(b1 |{c1 , ci }) = r(cn |{c1 , ci }) and for i = k + 1 we have r(a1 |{c1 , ck+1 }) =

r(bn |{c1 , ck+1 }), a contradiction.

(4) One vertex is in the inner cycle and other in the middle cycle. Without loss of gen-

erality we can suppose that one resolving vertex is a1 . Suppose that the second resolv-

ing vertex is bi (1 ≤ i ≤ k + 1), then for i = 1 we have r(a2 |{a1 , b1 }) = r(bn |{a1 , b1 })

and for 2 ≤ i ≤ k + 1, r(a2 |{a1 , bi }) = r(b1 |{a1 , bi }), a contradiction.

(5) One vertex is in the inner cycle and other in the outer cycle. We can sup-

pose that one resolving vertex is a1 . Suppose that the second resolving vertex is

ci (1 ≤ i ≤ k +1). Then for i = 1, r(a2 |{a1 , c1 }) = r(bn |{a1 , c1 }) and for 2 ≤ i ≤ k +1

we have r(a2 |{a1 , ci }) = r(b1 |{a1 , ci }), a contradiction.

(6) One vertex is in the middle cycle and other in the outer cycle. Here we can fix b1

as a resolving vertex. Suppose that the second resolving vertex is ci (1 ≤ i ≤ k + 1),

then for i = 1 we have r(a1 |{b1 , c1 }) = r(a2 |{b1 , c1 }) and for 2 ≤ i ≤ k + 1 we have

r(b2 |{b1 , ci }) = r(c1 |{b1 , ci }), a contradiction.

Hence, from above it follows that there is no resolving set with two vertices for V (Rn )

implying that dim(Rn ) = 3 in this case.

Case(ii) When n is even.

In this case, we can write n = 2k, k ≥ 3, k ∈ Z+ . Let W = {a1 , a2 , ak+1 } ⊂ V (Rn ),

 32

we show that W is a resolving set for Rn in this case. For this we give representations

of any vertex of V (Rn ).

Representations
 of the vertices on inner cycle are
 (i − 1, i − 2, k − i + 1),

3 ≤ i ≤ k;
r(ai |W ) =
 (2k − i + 1, 2k − i + 2, i − k − 1), k + 2 ≤ i ≤ 2k.

Representations of the vertices on middle cycle are




 (1, 1, k), i = 1;



 (i, i − 1, k − i + 1),

2 ≤ i ≤ k;
r(bi |W ) =
(k, k, 1), i = k + 1;






 (2k − i + 1, 2k − i + 2, i − k), k + 2 ≤ i ≤ 2k.


Representations of the vertices on outer cycle are




 (2, 2, k + 1), i = 1;



 (i + 1, i, k − i + 2),

2 ≤ i ≤ k;
r(ci |W ) =
(k + 1, k + 1, 2), i = k + 1;






 (2k − i + 2, 2k − i + 3, i − k + 1), k + 2 ≤ i ≤ 2k.


Again we see that there are no two vertices having the same representations which

implies that dim(Rn ) ≤ 3 in this case.

On the other hand, suppose that dim(Rn ) = 2, then there are the same subcases as in

case (i) and contradiction can be deduced analogously. This implies that dim(Rn ) = 3

in this case, which completes the proof.

 33

3.4 The graph of convex polytope Qn

In this section, we shall investigate the metric dimension of the graph of convex poly-

tope Qn defined in [2] consisting of 3-sided faces, 4-sided faces, 5-sided faces and

n-sided faces (Fig 3.3). The graph of convex polytope Qn and the graph of convex

polytope Dn have the same set of vertices; moreover, E(Qn ) = E(Dn ) ∪ {bi bi+1 : 1 ≤

i ≤ n}.

For our purpose, we call the cycle induced by {ai : 1 ≤ i ≤ n} the inner cycle, cycle

d1 dn

c1
d2 cn dn−1
b1
c2
b2 bn cn−1

b3 a1 bn−1
c3 a2 an dn−2
d3
a3 an−1 cn−2

Figure 3.3: The graph of convex polytope Qn

induced by {bi : 1 ≤ i ≤ n}, the middle cycle, set of vertices {c1 , c2 , ..., cn } the set of

interior vertices and cycle induced by {di : 1 ≤ i ≤ n}, the outer cycle.

In the next theorem, we show that with only three vertices, we can resolve all the ver-

tices of the graph of convex polytope Qn . Once gain, choice of appropriate landmarks

is very important.

Theorem 3.4.1. Let the graph of convex polytopes be Qn ; then dim(Qn ) = 3 for

every n ≥ 6.
 34

Proof. We will prove the above equality by double inequalities. We consider two

cases.

Case(i) When n is even.

In this case, we can write n = 2k, k ≥ 3, k ∈ Z+ . Let W = {a1 , a2 , ak+1 } ⊂ V (Qn ),

we show that W is a resolving set for Qn in this case. For this we give representations

of any vertex of V (Qn ).

Representations of the vertices on inner cycle are


 (i − 1, i − 2, k − i + 1),

3 ≤ i ≤ k;
r(ai |W ) =
 (2k − i + 1, 2k − i + 2, i − k − 1), k + 2 ≤ i ≤ 2k.


Representations of the vertices on middle cycle are





 (1, 2, k + 1), i = 1;

r(bi |W ) = (i, i − 1, k − i + 2), 2 ≤ i ≤ k + 1;



 (2k − i + 2, 2k − i + 3, i − k), k + 2 ≤ i ≤ 2k.


Representations of the set of interior vertices are




 (2, 2, k + 1), i = 1;



 (i + 1, i, k − i + 2),

2 ≤ i ≤ k + 1;
r(ci |W ) =
(k + 1, k + 1, 2), i = k + 2;






 (2k − i + 3, 2k − i + 4, i − k), k + 3 ≤ i ≤ 2k.


Representations of vertices on outer cycle are

 35




 (3, 3, k + 2), i = 1;



 (i + 2, i + 1, k − i + 3),

2 ≤ i ≤ k;
r(di |W ) =
(k + 2, k + 2, 3), i = k + 1;






 (2k − i + 3, 2k − i + 4, i − k + 2), k + 2 ≤ i ≤ 2k.


We see that there are no two vertices having the same representations implying that

dim(Qn ) ≤ 3.

On the other hand, we show that dim(Qn ) ≥ 3. Suppose on contrary that dim(Qn ) =

2, then there are following possibilities to be discussed.

(1) Both vertices are in the inner cycle. Without loss of generality we can sup-

pose that one resolving vertex is a1 . Suppose that the second resolving vertex is

ai (2 ≤ i ≤ k + 1), then for 2 ≤ i ≤ k, r(an |{a1 , ai }) = r(b1 |{a1 , ai }) and for i = k + 1

we have r(a2 |{a1 , ak+1 }) = r(an |{a1 , ak+1 }), a contradiction.

(2) Both vertices are in the middle cycle. Without loss of generality we can sup-

pose that one resolving vertex is b1 . Suppose that the second resolving vertex is

bi (2 ≤ i ≤ k + 1), then for 2 ≤ i ≤ k, r(a1 |{b1 , bi }) = r(cn |{b1 , bi }) and for i = k + 1

we have r(b2 |{b1 , bk+1 }) = r(bn |{b1 , bk+1 }), a contradiction.

(3) Both vertices belong to the set of interior vertices. Without loss of generality we

can suppose that one resolving vertex is c1 . Suppose that the second resolving vertex

is ci (2 ≤ i ≤ k +1), then for 2 ≤ i ≤ k, r(b1 |{c1 , ci }) = r(d1 |{c1 , ci }) and for i = k +1

we have r(c2 |{c1 , ck+1 }) = r(cn |{c1 , ck+1 }), a contradiction.

(4) Both vertices are in the outer cycle. Without loss of generality we can sup-

pose that one resolving vertex is d1 . Suppose that the second resolving vertex is

di (2 ≤ i ≤ k + 1), then for 2 ≤ i ≤ k, r(c1 |{d1 , di }) = r(dn |{d1 , di }) and for i = k + 1

we have r(d2 |{d1 , dk+1 }) = r(dn |{d1 , dk+1 }), a contradiction.

 36

(5) One vertex is in the inner cycle and other in the middle cycle. Without loss of gen-

erality we can suppose that one resolving vertex is a1 . Suppose that the second resolv-

ing vertex is bi (1 ≤ i ≤ k + 1), then for i = 1 we have r(a2 |{a1 , b1 }) = r(an |{a1 , b1 })

and for 2 ≤ i ≤ k + 1, r(a2 |{a1 , bi }) = r(b1 |{a1 , bi }), a contradiction.

(6) One vertex is in the inner cycle and other in the set of interior vertices. With-

out loss of generality we can suppose that one resolving vertex is a1 . Suppose

that the second resolving vertex is ci (1 ≤ i ≤ k + 1), then for i = 1 we have

r(b2 |{a1 , c1 }) = r(cn |{a1 , c1 }). For 2 ≤ i ≤ k, r(a2 |{a1 , ci }) = r(b1 |{a1 , ci }) and when

i = k + 1 we have r(b1 |{a1 , ck+1 }) = r(an |{a1 , ck+1 }), a contradiction.

(7) One vertex is in the inner cycle and other in the outer cycle. Without loss of gener-

ality we can suppose that one resolving vertex is a1 . Suppose that the second resolving

vertex is di (1 ≤ i ≤ k + 1), then for i = 1 we have r(b2 |{a1 , d1 }) = r(cn |{a1 , d1 }) and

for 2 ≤ i ≤ k + 1, r(a2 |{a1 , di }) = r(b1 |{a1 , di }), a contradiction.

(8) One vertex is in the middle cycle and other in the set of interior vertices. Without

loss of generality we can suppose that one resolving vertex is b1 . Suppose that the

second resolving vertex is ci (1 ≤ i ≤ k+1) then for 1 ≤ i ≤ k we have r(a1 |{b1 , ci }) =

r(bn |{b1 , ci }) and for i = k + 1 we have r(a1 |{b1 , ck+1 }) = r(cn |{b1 , ck+1 }), a contra-

diction.

(9) One vertex is in the middle cycle and other in the outer cycle. Without loss of

generality we can suppose that one resolving vertex is b1 . Suppose that the second

resolving vertex is di (1 ≤ i ≤ k + 1), then for 1 ≤ i ≤ k − 1 we have r(a1 |{b1 , di }) =

r(bn |{b1 , di }) and for i = k, k + 1 we have, r(bn |{b1 , di }) = r(cn |{b1 , di }), a contradic-

tion.
 37

(10) One vertex is in the set of interior vertices and other in the outer cycle. With-

out loss of generality we can suppose that one resolving vertex is c1 . Suppose that

the second resolving vertex is di (1 ≤ i ≤ k + 1). Then for i = 1, r(b1 |{c1 , d1 }) =

r(b2 |{c1 , d1 }), for i = 2, r(b3 |{c1 , d2 }) = r(dn |{c1 , d2 }) and when 3 ≤ i ≤ k + 1, we

have r(b3 |{c1 , di }) = r(c2 |{c1 , di }), a contradiction.

So from above, we conclude that there is no resolving set with two vertices for V (Qn )

implying that dim(Qn ) = 3 in this case.

Case(ii) When n is odd.

In this case, we can write n = 2k+1, k ≥ 3, k ∈ Z+ . Let W = {a1 , a2 , ak+2 } ⊂ V (Qn ),

we show that W is a resolving set for Qn in this case. For this we give representations

of any vertex of V (Qn ).

Representations of the vertices on inner cycle are


 (i − 1, i − 2, k − i + 2),

3 ≤ i ≤ k + 1;
r(ai |W ) =
 (2k − i + 2, 2k − i + 3, i − k − 2), k + 3 ≤ i ≤ 2k + 1.


Representations of the vertices on middle cycle are




 (1, 2, k + 1), i = 1;



 (i, i − 1, k − i + 3),

2 ≤ i ≤ k + 1;
r(bi |W ) =
(k + 1, k + 1, 1), i = k + 2;






 (2k − i + 3, 2k − i + 4, i − k − 1), k + 3 ≤ i ≤ 2k + 1.


Representations of the set of interior vertices are

 38





 (2, 2, k + 2), i = 1;

r(ci |W ) = (i + 1, i, k − i + 3), 2 ≤ i ≤ k + 1;



 (2k − i + 3, 2k − i + 4, i − k), k + 2 ≤ i ≤ 2k + 1.


Representations of vertices on outer cycle are





 (3, 3, k + 3), i = 1;

r(di |W ) = (i + 2, i + 1, k − i + 4), 2 ≤ i ≤ k + 1;



 (2k − i + 4, 2k − i + 5, i − k + 1), k + 2 ≤ i ≤ 2k + 1.


Again we see that there are no two vertices having the same representations implying

that dim(Qn ) ≤ 3.

On the other hand, we show that dim(Qn ) ≥ 3. Suppose on contrary that dim(Qn ) =

2, then there are the same possibilities as in case (i) and contradiction can be obtained

analogously. It follows that dim(Qn ) = 3, which completes the proof.

Chapter 4

Metric dimension of generalized

Petersen graphs P (n, 3)

In this chapter, we study the metric dimension of the generalized Petersen graphs

P (n, 3) by giving a partial answer to an open problem raised in [24]: Is P (n, m) for

n ≥ 7 and 3 ≤ m ≤ b n−1
2
c, a family of graphs with constant metric dimension? We

prove that the generalized Petersen graphs P (n, 3) have metric dimension equal to 3

for n ≡1(mod 6), n ≥ 25, and to 4 for n ≡ 0(mod 6), n ≥ 24. For the remaining cases

only 4 vertices appropriately chosen suffice to resolve all the vertices of P (n, 3), thus

implying that dim(P (n, 3)) ≤ 4, except when n ≡ 2(mod 6), when dim(P (n, 3)) ≤ 5.

4.1 Notation and auxiliary results

In [24] Javaid et al. proved that some regular graphs namely generalized Petersen

graphs P (n, 2), antiprisms An and Harary graphs H4,n are families of graphs with

constant metric dimension and raised an open problem.

Open Problem [24]: Is P (n, m) for n ≥ 7 and 3 ≤ m ≤ b n−1

2
c, a family of graphs

with constant metric dimension?

In this chapter, we give a partial answer to this open problem and we show that

39
 40

the generalized Petersen graphs P (n, 3) constitute a family of regular graphs having

bounded metric dimension and only 4 vertices appropriately chosen suffice to resolve

all vertices of the generalized Petersen graphs P (n, 3) except n ≡ 2(mod 6), when

this number equals 5. For n ≡ 1(mod 6) a minimal resolving set has cardinality equal

to 3.

In what follows all indices i which do not satisfy inequalities 1 ≤ i ≤ n will be taken

modulo n.

4.2 Upper bounds for metric dimension of gener-

alized Petersen graphs P (n, 3)

The generalized Petersen graph denoted by P (n, m), where n ≥ 3 and 1 ≤ m ≤ b n−1
2
c,

is a cubic graph having vertex set

V = {u1 , u2 , . . . , un , v1 , v2 , . . . , vn }

and edge set

E = {ui ui+1 , ui vi , vi vi+m : 1 ≤ i ≤ n}.

Generalized Petersen graphs were first defined by Watkins [44]. For m = 1 the gen-

eralized Petersen graph P (n, 1) is called prism, denoted by Dn . In [5] it was shown

that

(
2, if n is odd;
dim(Pm × Cn )=
3, if n is even.

Since the prism Dn is actually the cross product of P2 with a cycle Cn , this implies that
 41

(
2, if n is odd;
dim(Dn )=
3, if n is even.
So, prisms constitute a family of 3-regular graphs with bounded metric dimension.

In [24] it was proved that dim(P (n, 2)) = 3 for every n ≥ 5. Now we will find the

metric dimension of the generalized Petersen graphs P (n, 3) when n ≡ 0 or 1(mod 6)

and an upper bound in the remaining cases.

When m = 3, {u1 , u2 , ...., un } induces a cycle in P (n, 3) with ui ui+1 (1 ≤ i ≤ n), as

edges. If n = 3l (l ≥ 3), then {v1 , v2 , . . . , vn } induces 3 cycles of length l, otherwise

it induces a cycle of length n with vi vi+3 (1 ≤ i ≤ n), as edges. For example, P (8, 3)

is the Möbius-Kantor graph [29].

Since generalized Petersen graphs P (n, 3) form an important class of 3-regular graphs

with 2n vertices and 3n edges, it is desirable to find their metric dimensions. For our

purpose, we call the cycle induced by {u1 , u2 , . . ., un }, outer cycle and cycle(s) induced

by {v1 , v2 , . . . , vn } inner cycle(s). Note that the choice of appropriate basis vertices

(also referred to as landmarks in [27]) is core of the problem.

Theorem 4.2.1. For the generalized Petersen graph P (n, 3) we have
(a) dim(P (n, 3)) ≤ 3 for n ≡ 1(mod 6) and n ≥ 13;
(b) dim(P (n, 3)) ≤ 4 for n ≡ 0, 3, 4, 5(mod 6) and n ≥ 17.
(c) dim(P (n, 3)) ≤ 5 for n ≡ 2(mod 6) and n ≥ 8.

Proof: We denote W = {v1 , v2 , v3 , u4 } for n ≡ 0, 3, 4, 5(mod 6) and W =

{v1 , v3k−1 , v6k } for n ≡ 1(mod 6), n = 6k + 1. We show that the set W distinguishes

the vertices of P (n, 3) for n 6≡ 2(mod 6). For this purpose, we give the representation

of V (P (n, 3)) in these cases.

Case (i) n = 6k, k ∈ Z+ , k ≥ 2. For every n ≥ 12, the representations of vertices on

the outer cycle are

r(u1 |W ) = (1, 2, 3, 3), r(u2 |W ) = (2, 1, 2, 2), r(u3 |W ) = (3, 2, 1, 1).

 42

r(u2+3i |W ) =

(3, 2, 3, 1), i = 1;





(i + 2, i + 1, i + 2, i + 2), 2 ≤ i ≤ k;



(2k − i + 2, 2k − i + 1, 2k − i + 2, 2k − i + 4),

k + 1 ≤ i ≤ 2k − 1.

r(u3+3i |W ) =

(4, 3, 2, 2), i = 1;





(i + 3, i + 2, i + 1, i + 3), 2 ≤ i ≤ k − 1;



(2k − i + 1, 2k − i + 2, 2k − i + 1, 2k − i + 3),

k ≤ i ≤ 2k − 1.

r(u4+3i |W ) =

(i + 2, i + 3, i + 2, i + 2), 1 ≤ i ≤ k − 1;
(2k − i, 2k − i + 1, 2k − i + 2, 2k − i + 2), k ≤ i ≤ 2k − 2.

Representations of vertices with respect to W on the inner cycles are

r(v1+3i |W ) =

(i, i + 3, i + 2, i), 1 ≤ i ≤ k;
(2k − i, 2k − i + 3, 2k − i + 4, 2k − i + 2), k + 1 ≤ i ≤ 2k − 1.

r(v2+3i |W ) =

(i + 3, i, i + 3, i + 1), 1 ≤ i ≤ k;
(2k − i + 3, 2k − i, 2k − i + 3, 2k − i + 3), k + 1 ≤ i ≤ 2k − 1.

r(v3+3i |W ) =

(i + 4, i + 3, i, i + 2), 1 ≤ i ≤ k − 1;
(2k − i + 2, 2k − i + 3, 2k − i, 2k − i + 2), k ≤ i ≤ 2k − 1.
 43

Case (ii) n = 6k +1, k ∈ Z+ , k ≥ 2. For every n ≥ 13, the representations of vertices

with respect to W are the following:

Representations of vertices on the outer cycle are




 (i + 2, k − i + 2, i + 2), 1 ≤ i ≤ k − 1;


 (k + 2, 2, k + 1), i = k;
r(u3i |W ) =


 (2k − i + 3, i − k + 2, 2k − i + 1), k + 1 ≤ i ≤ 2k − 1;


 (3, k + 1, 1), i = 2k.
(
(i + 1, k − i + 1, i + 1), 1 ≤ i ≤ k;
r(u3i−1 |W ) =
(2k − i + 2, i − k + 1, 2k − i + 2), k + 1 ≤ i ≤ 2k.



 (i, k − i + 2, i + 2), 1 ≤ i ≤ k;


 (k + 1, 3, k + 2), i = k + 1;
r(u3i−2 |W ) =


 (2k − i + 3, i − k + 2, 2k − i + 3), k + 2 ≤ i ≤ 2k;


 (2, k + 2, 2), i = 2k + 1.
Representations of vertices with respect to W on the inner cycle are




 (i + 2, k − i, i), 1 ≤ i ≤ k − 1;


 (k + 1, 0, k), i = k;
r(v3i−1 |W ) =


 (k, 1, k + 1), i = k + 1;


 (2k − i + 1, i − k, 2k − i + 3), k + 2 ≤ i ≤ 2k.



 (i + 3, k − i + 3, i + 3), 1 ≤ i ≤ k − 2;


 (i + 3, k − i + 3, 2k − i), k − 1 ≤ i ≤ k;



r(v3i |W ) = (2k − i + 4, i − k + 3, 2k − i), k + 1 ≤ i ≤ 2k − 2;





 (5, k + 1, 1), i = 2k − 1;

 (4, k, 0),

i = 2k.
 44



 (0, k + 1, 4), i = 0;


(i, k − i + 2, i + 4), 1 ≤ i ≤ k − 1;






 (k, 4, k + 3), i = k;
r(v3i+1 |W ) =


 (k + 1, 5, k + 2), i = k + 1;





 (2k − i + 3, i − k + 4, 2k − i + 3), k + 2 ≤ i ≤ 2k − 1;

(3, k + 3, 3), i = 2k.

+
Case (iii) n = 6k + 3, k ∈ Z . For P (9, 3), the representations of the vertices

are r(u1 |W ) = (1, 2, 3, 3), r(u2 |W ) = (2, 1, 2, 2), r(u3 |W ) = (3, 2, 1, 1), r(u5 |W ) =

(3, 2, 3, 1), r(u6 |W ) = (3, 3, 2, 2), r(u7 |W ) = (2, 3, 3, 3),

r(u8 |W ) = (3, 2, 3, 4), r(u9 |W ) = (2, 3, 2, 4), r(v4 |W ) = (1, 4, 3, 1),

r(v5 |W ) = (4, 1, 4, 2), r(v6 |W ) = (4, 4, 1, 3), r(v7 |W ) = (1, 4, 4, 2),

r(v8 |W ) = (4, 1, 4, 3), r(v9 |W ) = (3, 4, 1, 3). For every n ≥ 15, representations of

vertices with respect to W are the following:

Representations of vertices on the outer cycle are

r(u1 |W ) = (1, 2, 3, 3), r(u2 |W ) = (2, 1, 2, 2), r(u3 |W ) = (3, 2, 1, 1).

r(u2+3i |W ) =




 (3, 2, 3, 1), i = 1;


(i + 2, i + 1, i + 2, i + 2), 2 ≤ i ≤ k;





 (k + 2, k + 1, k + 2, k + 3), i = k + 1;


(2k − i + 3, 2k − i + 2, 2k − i + 3, 2k − i + 5),

k + 2 ≤ i ≤ 2k.
 45

r(u3+3i |W ) =




 (4, 3, 2, 2), i = 1;


(i + 3, i + 2, i + 1, i + 3), 2 ≤ i ≤ k − 1;





 (k + 2, k + 2, k + 1, k + 3), i = k;


(2k − i + 2, 2k − i + 3, 2k − i + 2, 2k − i + 4),

k + 1 ≤ i ≤ 2k.

r(u4+3i |W ) =

(i + 2, i + 3, i + 2, i + 2), 1 ≤ i ≤ k − 1;





(k + 1, k + 2, k + 2, k + 2), i = k;



(2k − i + 1, 2k − i + 2, 2k − i + 3, 2k − i + 3),

k + 1 ≤ i ≤ 2k − 1.

Representations of the vertices with respect to W on the inner cycles are

r(v1+3i |W ) =

(i, i + 3, i + 2, i), 1 ≤ i ≤ k;





(k, k + 3, k + 3, k + 1), i = k + 1;



(2k − i + 1, 2k − i + 4, 2k − i + 5, 2k − i + 3),

k + 2 ≤ i ≤ 2k.

r(v2+3i |W ) =

(i + 3, i, i + 3, i + 1), 1 ≤ i ≤ k;





(k + 3, k, k + 3, k + 2), i = k + 1;



(2k − i + 4, 2k − i + 1, 2k − i + 4, 2k − i + 4),

k + 2 ≤ i ≤ 2k.

r(v3+3i |W ) =

(i + 4, i + 3, i, i + 2), 1 ≤ i ≤ k − 1;





(k + 3, k + 3, k, k + 2), i = k;



(2k − i + 3, 2k − i + 4, 2k − i + 1, 2k − i + 3),

k + 1 ≤ i ≤ 2k.
 46

Case (iv) n = 6k + 4, k ∈ Z+ . For every n ≥ 10, the representations of the vertices

on the outer cycle are

r(u1 |W ) = (1, 2, 3, 3), r(u2 |W ) = (2, 1, 2, 2), r(u3 |W ) = (3, 2, 1, 1).

r(u2+3i |W ) =

(3, 2, 3, 1), i = 1;





(i + 2, i + 1, i + 2, i + 2), 2 ≤ i ≤ k;



(2k − i + 2, 2k − i + 3, 2k − i + 4, 2k − i + 4),

k + 1 ≤ i ≤ 2k.

r(u3+3i |W ) =

(4, 3, 2, 2), i = 1;





(i + 3, i + 2, i + 1, i + 3), 2 ≤ i ≤ k;



(2k − i + 3, 2k − i + 2, 2k − i + 3, 2k − i + 5),

k + 1 ≤ i ≤ 2k.

r(u4+3i |W ) =

(i + 2, i + 3, i + 2, i + 2), 1 ≤ i ≤ k;
(2k − i + 2, 2k − i + 3, 2k − i + 2, 2k − i + 4), k + 1 ≤ i ≤ 2k.

Representations of vertices on inner cycle are

r(v1+3i |W ) =

(i, i + 3, i + 2, i), 1 ≤ i ≤ k;





(k + 1, k + 4, k + 1, k + 1), i = k + 1;



(2k − i + 4, 2k − i + 5, 2k − i + 2, 2k − i + 4),

k + 2 ≤ i ≤ 2k + 1.
 47

r(v2+3i |W ) =




 (i + 3, i, i + 3, i + 1), 1 ≤ i ≤ k − 1;


(k + 1, k, k + 3, k + 1),

i = k;



 (k, k + 1, k + 4, k + 2), i = k + 1;


(2k − i + 1, 2k − i + 4, 2k − i + 5, 2k − i + 3),

k + 2 ≤ i ≤ 2k.

r(v3+3i |W ) =




 (i + 4, i + 3, i, i + 2), 1 ≤ i ≤ k − 1;


(k + 4, k + 1, k, k + 2),

i = k;



 (k + 3, k, k + 1, k + 3), i = k + 1;


(2k − i + 4, 2k − i + 1, 2k − i + 4, 2k − i + 4),

k + 2 ≤ i ≤ 2k.

Case (v) n = 6k + 5, k ∈ Z+ . For every n ≥ 17, the representations of vertices on

the outer cycle are

r(u1 |W ) = (1, 2, 3, 3), r(u2 |W ) = (2, 1, 2, 2), r(u3 |W ) = (3, 2, 1, 1).

r(u2+3i |W ) =




 (3, 2, 3, 1), i = 1;


(i + 2, i + 1, i + 2, i + 2), 2 ≤ i ≤ k;





 (k + 2, k + 2, k + 2, k + 3), i = k + 1;


(2k − i + 3, 2k − i + 4, 2k − i + 3, 2k − i + 5),

k + 2 ≤ i ≤ 2k + 1.
 48

r(u3+3i |W ) =

(4, 3, 2, 2),

 i = 1;



(i + 3, i + 2, i + 1, i + 3), 2 ≤ i ≤ k − 1;





(k + 2, k + 2, k + 1, k + 3), i = k;



(k + 1, k + 2, k + 2, k + 3), i = k + 1;







(2k − i + 2, 2k − i + 3, 2k − i + 4, 2k − i + 4), k + 2 ≤ i ≤ 2k.

r(u4+3i |W ) =




 (i + 2, i + 3, i + 2, i + 2), 1 ≤ i ≤ k − 1;


(k + 2, k + 2, k + 2, k + 2),

i = k;



 (k + 2, k + 1, k + 2, k + 3), i = k + 1;


(2k − i + 3, 2k − i + 2, 2k − i + 3, 2k − i + 5),

k + 2 ≤ i ≤ 2k.

Representations of vertices with respect to W on inner cycle are

r(v1+3i |W ) =



 (i, i + 3, i + 2, i), 1 ≤ i ≤ k − 1;



(k, k + 2, k + 2, k), i = k;





(k + 1, k + 1, k + 3, k + 1), i = k + 1;



(k + 2, k, k + 3, k + 2), i = k + 2;







(2k − i + 5, 2k − i + 2, 2k − i + 5, 2k − i + 5), k + 3 ≤ i ≤ 2k + 1.
 49

r(v2+3i |W ) =



 (i + 3, i, i + 3, i + 1), 1 ≤ i ≤ k − 1;



(k + 3, k, k + 2, k + 1), i = k;





(k + 3, k + 1, k + 1, k + 2), i = k + 1;



(k + 2, k + 2, k, k + 2), i = k + 2;







(2k − i + 4, 2k − i + 5, 2k − i + 2, 2k − i + 4), k + 3 ≤ i ≤ 2k + 1.

r(v3+3i |W ) =

(i + 4, i + 3, i, i + 2), 1 ≤ i ≤ k − 2;










 (k + 2, k + 2, k − 1, k + 1), i = k − 1;


(k + 1, k + 3, k, k + 2),

i = k;



 (k, k + 3, k + 1, k + 2), i = k + 1;


(k − 1, k + 2, k + 2, k + 1),



 i = k + 2;



(2k − i + 1, 2k − i + 4, 2k − i + 5, 2k − i + 3),

k + 3 ≤ i ≤ 2k.

We note that there are no two vertices in the inner cycle(s) with same representations.

Also, there are no two vertices in the inner cycle(s) and outer cycle having the same

representations and no two vertices on outer cycle having the same representations.

This implies that W = {v1 , v2 , v3 , u4 } is a resolving set for V (P (n, 3)) when n ≡ 0, 3,

4, 5(mod 6) implying that in these cases dim(P (n, 3)) ≤ 4. Also W = {v1 , v3k−1 , v6k }

is a resolving set for n = 6k + 1, when dim(P (n, 3)) ≤ 3.

Case (vi) n = 6k+2, k ∈ Z+ . It is straightforward to verify that W1 = {v1 , v2 , v3 , u4 }

and W2 = {v1 , v2 , u7 , u10 } are resolving sets for P (8, 3) and P (14, 3), respectively. For

every n ≥ 20, we show that W = {v1 , v2 , v3 , u4 , u3k+5 } is a resolving set. For this,

we first give representations of vertices with respect to W 0 = {v1 , v2 , v3 , u4 }. The

 50

representations of the vertices on the outer cycle are

r(u1 |W 0 ) = (1, 2, 3, 3), r(u2 |W 0 ) = (2, 1, 2, 2), r(u3 |W 0 ) = (3, 2, 1, 1).

r(u2+3i |W 0 ) =

(3, 2, 3, 1), i = 1;





(i + 2, i + 1, i + 2, i + 2), 2 ≤ i ≤ k;



(2k − i + 2, 2k − i + 3, 2k − i + 2, 2k − i + 4),

k + 1 ≤ i ≤ 2k.

r(u3+3i |W 0 ) =




 (4, 3, 2, 2), i = 1;


(i + 3, i + 2, i + 1, i + 3), 2 ≤ i ≤ k − 1;





 (k + 1, k + 2, k + 1, k + 3), i = k;


(2k − i + 1, 2k − i + 2, 2k − i + 3, 2k − i + 3),

k + 1 ≤ i ≤ 2k − 1.

r(u4+3i |W 0 ) =

(i + 2, i + 3, i + 2, i + 2), 1 ≤ i ≤ k − 1;




(k + 2, k + 1, k + 2, k + 2), i = k;



(2k − i + 2, 2k − i + 1, 2k − i + 2, 2k − i + 4),

k + 1 ≤ i ≤ 2k − 1.

Representations of vertices with respect to W 0 on inner cycle are

r(v1+3i |W 0 ) =




 (i, i + 3, i + 2, i), 1 ≤ i ≤ k − 1;


(k, k + 1, k + 2, k),

i = k;



 (k + 1, k, k + 3, k + 1), i = k + 1;


(2k − i + 4, 2k − i + 1, 2k − i + 4, 2k − i + 4),

k + 2 ≤ i ≤ 2k.
 51

r(v2+3i |W 0 ) =




 (i + 3, i, i + 3, i + 1), 1 ≤ i ≤ k − 1;


(k + 3, k, k + 1, k + 1),

i = k;



 (k + 2, k + 1, k, k + 2), i = k + 1;


(2k − i + 3, 2k − i + 4, 2k − i + 1, 2k − i + 3),

k + 2 ≤ i ≤ 2k.

r(v3+3i |W 0 ) =



 (i + 4, i + 3, i, i + 2), 1 ≤ i ≤ k − 2;



(k + 1, k + 2, k − 1, k + 1), i = k − 1;





(k, k + 3, k, k + 2), i = k;



(k − 1, k + 2, k + 1, k + 1), i = k + 1;







(2k − i, 2k − i + 3, 2k − i + 4, 2k − i + 2), k + 2 ≤ i ≤ 2k − 1.

Consequently, r(u2+3k |W 0 ) = r(u4+3k |W 0 ) = (k + 2, k + 1, k + 2, k + 2);

r(u5+3k |W 0 ) = r(u3+3k |W 0 ) = (k + 1, k + 2, k + 1, k + 3);

r(u3k |W 0 ) = r(v5+3k |W 0 ) = (k + 2, k + 1, k, k + 2);

r(v8+3k |W 0 ) = r(v3k |W 0 ) = (k + 1, k + 2, k − 1, k + 1).

The vertex u3k+5 distinguishes these pairs of vertices with same representations as

d(u3k+5 , u3k+2 ) = 3, d(u3k+5 , u3k+4 ) = 1, d(u3k+5 , u3k ) = 5, d(u3k+5 , v3k+5 ) = 1,

d(u3k+5 , v3k+8 ) = 2 and d(u3k+5 , v3k ) = 4. This suggests that W = {v1 , v2 , v3 , u4 , u3k+5 }

is a resolving set for V (P (n, 3)) in this case implying that dim(P (n, 3)) ≤ 5. 2

4.3 Metric dimension of P (n, 3) for n ≡ 0, 1(mod 6)

In this section we will prove that dim(P (n, 3)) ≥ 3 for n ≡ 1(mod 6) and n ≥ 25 and

dim(P (n, 3)) ≥ 4 for n ≡ 0(mod 6) and n ≥ 24, yielding exact values of dim(P (n, 3))

in these cases by Theorem 4.2.1. For this purpose we need some more notations and
 52

definitions. Without loss of generality we can suppose that the vertices of the outer

cycle are u1 , u2 , . . . , un in the clockwise direction. For two vertices ui and uj (i 6= j)

we shall define the ”clockwise distance” from ui to uj , denoted by d∗ (ui , uj ) the dis-

tance, measured in clockwise direction, from ui to uj , in the subgraph induced by

the outer cycle. For example, d∗ (u1 , un ) = n − 1 and d∗ (un , u1 ) = 1; in general we

have d∗ (ui , uj ) + d∗ (uj , ui ) = n. This definition can be extended to any two vertices

of P (n, 3) for i 6= j by: d∗ (ui , vj ) = d∗ (vi , uj ) = d∗ (vi , vj ) = d∗ (ui , uj ).

Consider a vertex on the outer cycle, say u1 . A vertex ui is called a good vertex for

u1 if ui and ui+2 have equal distances to u1 , i. e., d(u1 , ui ) = d(u1 , ui+2 ); otherwise ui

is called a bad vertex for u1 . This definition can be extended to vertices of the inner

cycle: ui is a good vertex for v1 if d(v1 , ui ) = d(v1 , ui+2 ) and bad otherwise.

In figure 4.1 we have represented by black dots all good vertices for u1 when n =

6k + 1 ≥ 25.

It is important to note that the set of good vertices for v1 is deduced from the set of

good vertices for u1 by adding 4 new vertices, namely u2 , u3 , u6k−1 and u6k−2 . Simi-

larly, a vertex uj is said to be good for the pair {u1 , ui } if d(u1 , uj ) = d(u1 , uj+2 ) and

d(ui , uj ) = d(ui , uj+2 ). If uk is a good vertex for the pairs {u1 , ui } and {u1 , uj } then uk

is also a good vertex for the triplet {u1 , ui , uj }, i. e., d(u1 , uk ) = d(u1 , uk+2 ), d(ui , uk ) =

d(ui , uk+2 ) and d(uj , uk ) = d(uj , uk+2 ).

Due to the rotational symmetry of P (n, 3) we deduce

Lemma 4.3.1. For any two vertices ui and uj on the outer cycle of P (n, 3) we have
d(ui , uj ) = d(ui+r , uj+r ) for any 1 ≤ r ≤ n − 1.

In order to find good vertices for pairs of vertices belonging to the outer cycle the

following lemma will be useful.

Lemma 4.3.2. Let 1 ≤ i ≤ n − 2. If uj is good for u1 and uj−i is also good for u1 ,
 53

u6k-1 u6k u 6k+1 u

u6k-2 1
u6k-3 u2
u6k-4 u3
v1 u4
u6k-5
u5

u6

u7

u3k+6 u8

u3k+5 u9

u3k+4 u10
u3k+3
u3k+2
u3k+1 u u3k-1
3k

Figure 4.1: Good vertices for u1 (n = 6k + 1)

then uj is good for the pair {u1 , ui+1 }.

Proof: By hypothesis we can write d(u1 , uj ) = d(u1 , uj+2 ) and

d(u1 , uj−i ) = d(u1 , uj−i+2 ). By Lemma 4.3.1 the last equality is equivalent to d(ui+1 , uj ) =

d(ui+1 , uj+2 ). 2
Theorem 4.3.3. dim(P (n, 3)) = 3 if n = 6k + 1 and n ≥ 25.

Proof: We shall prove that dim(P (n, 3)) ≥ 3 in this case, by showing that there

is no resolving set of V (P (n, 3)) consisting of two vertices, X and Y . If both vertices

X and Y belong to the outer cycle, we can suppose that X = u1 . We distinguish

three cases:

1)d∗ (u1 , Y ) ≡ 0(mod 3). We choose vertex u6k−4 . Since vertices u6k−7 ,

u6k−10 , . . . , u3k+5 , u3k+2 , u3k−1 , u3k−4 , . . . , u8 , u5 are good vertices for u1 , but u2 , u6k
 54

and u6k−3 are bad vertices for u1 (see fig. 4.1), applying Lemma 4.3.2 we find that

u6k−4 is a good vertex for any pair {u1 , Y } such that Y ∈

/ {u6k−5 , u6k−2 , u6k+1 }.

But in this case u5 is a good vertex for the pairs {u1 , u6k−2 } and {u1 , u6k+1 } if k ≥ 4

and for {u1 , u6k−5 } if k ≥ 5; for k = 4 we have that u3k = u12 is not a good vertex

for u1 . It remains to consider the pair {u1 , u6k−5 } = {u1 , u19 } when k = 4. In this

case u9 is a good vertex for this pair. It follows that any pair {u1 , Y } cannot be a

resolving set having two vertices.

2)d∗ (u1 , Y ) ≡ 1(mod 3). We consider vertex u5 . By starting from u5 and going in

the counter-clockwise direction at distances 1,4,7,... the only bad vertices encountered

are u4 , u1 and u6k−1 . By Lemma 4.3.2 we deduce that u5 is a good vertex for any

pair {u1 , Y } such that Y ∈

/ {u2 , u5 , u8 }. In a similar manner we get that u6k−4 is a

good vertex for the pairs {u1 , u2 } and {u1 , u5 } if k ≥ 4 and for {u1 , u8 } if k ≥ 5. For

k = 4 the vertex u16 is a good vertex for {u1 , u8 }.

3)d∗ (u1 , Y ) ≡ 2(mod 3). In this case in order to minimize the number of bad

vertices for the pairs {u1 , Y } we choose vertex u6 . This vertex is a good vertex for

any pair {u1 , Y } such that Y ∈

/ {u3 , u6 , u9 }. Vertex u6k−5 is a good vertex for all

pairs {u1 , u3 }, {u1 , u6 } and {u1 , u9 } for any k ≥ 4.

If both X and Y belong to the inner cycle we can consider that X = v1 and Y = vi

(i > 1); this case can be reduced to the case when X = u1 and Y = ui since the set

of good vertices for vp includes the set of good vertices for up for every 1 ≤ p ≤ n.

If X = ui and Y = vi then any good vertex for ui is also a good vertex for vi , hence

for the pair {X, Y }. The remaining case when X = ui , Y = vj and i 6= j can also

be reduced to the case X = ui , Y = uj . It follows that there is no resolving set

containing two vertices, which concludes the proof. 2

Theorem 4.3.4. If n = 6k and n ≥ 24 then dim(P (n, 3)) = 4.
 55

Proof: By Theorem 4.2.1 it is necessary only to show that dim(P (n, 3)) ≥ 4, or

that there is no resolving set of V (P (n, 3)) consisting of three vertices, X, Y and Z.

By the same reasoning as in the proof of Theorem 4.3.1 it is sufficient to consider

only the case when X, Y, Z belong to the outer cycle since the set of good vertices for

v1 can be deduced from the set of good vertices for u1 (represented in figure 4.2) by

adding vertices u2 , u3 , u6k−2 and u6k−3 . As in the case n = 6k + 1 we shall see that for

any three vertices X, Y, Z such that d∗ (X, Y ) < d∗ (X, Z) it is possible to find a pair

of vertices at distance 2 on the outer cycle, {ui , ui+2 } having equal distances to X, Y

and Z, respectively. If n = 6k and X, Y, Z are on the outer cycle, we can suppose

that X = u1 . By denoting (x, y) ≡ (a, b)(mod 3) if x ≡ a(mod 3) and y ≡ b(mod 3),

the following 9 cases can occur: (d∗ (u1 , Y ), d∗ (u1 , Z)) is congruent modulo 3 to:

1)(0,0); 2)(1,1); 3)(2,2); 4)(0,1); 5)(0,2); 6)(1,0); 7)(1,2); 8)(2,0); 9)(2,1).

Some of these cases can be reduced to another cases. For example, from case

2 by permutation X → Y, Y → Z, Z → X we obtain case 5 and by permutation

X → Z, Y → X, Z → Y we get case 8.

The graph of reducibility between cases is illustrated in figure 4.3.

It follows that it is sufficient to consider only cases 1,2,3,7,9.

Case 1. If we choose good vertex u6 and we go in the counter-clockwise direction,

reaching vertices u3 , u6k , u6k−6 , . . . , u9 we encounter only two bad vertices, u3 and

u6k−3 (see figure 4.2). By Lemma 3.2 it follows that u6 is a good vertex for all pairs

{u1 , Y } where d∗ (u1 , Y ) ≡ 0(mod 3) and Y ∈

/ {u4 , u10 }. This implies that u6 is a

good vertex for all triplets {u1 , Y, Z}, unless Y = u4 and Z ∈ {u7 , u10 , u13 , . . . , u6k−2 };

Y = u10 and Z ∈ {u7 , u13 , u16 , . . . , u6k−2 }. For these triplets we must find other good
 56

u6k
u6k-1 u1
u6k-2 u2
u3
u6k-3
v1 u4

u6k-4 u5

u6k-5 u6

u7

u8

u9

u3k+3 u 3k-2
u3k+2 u3k-1
u3k+1 u3k

Figure 4.2: Good vertices for u1 (n = 6k)

vertices on the outer cycle. Similarly, u3k+6 is a good vertex for u1 since 6k − 6 ≥

3k + 6 and for all pairs {u1 , Y } where d∗ (u1 , Y ) ≡ 0(mod 3) and Y ∈

/ {u3k+4 , u3k+10 }.

Consequently, we have found a good vertex (u6 or u3k+6 ) for all triplets {u1 , Y, Z}

such that {Y, Z} =

6 {u4 , u3k+4 }, {u4 , u3k+10 },

{u10 , u3k+4 }, {u10 , u3k+10 }. Finally, u9 is a good vertex for all pairs {u1 , Y }, where

d∗ (u1 , Y ) ≡ 0(mod 3) and Y ∈

/ {u7 , u13 }. Since k ≥ 4 we have 3k + 4 > 13 and

u9 is a good vertex for the remaining triplets {u1 , Y, Z}, where Y ∈ {u4 , u10 } and

Z ∈ {u3k+4 , u3k+10 }.

Case 2. In a similar way we get that u6k−5 is a good vertex for all pairs {u1 , Y },

where d∗ (u1 , Y ) ≡ 1(mod 3) and Y ∈

/ {u6k−7 , u6k−1 }, therefore u6k−5 is a good vertex

for all triplets {u1 , Y, Z}, unless Y = u6k−7 and Z ∈ {u2 , u5 , . . . , u6k−10 , u6k−4 , u6k−1 };
 57

1 7 9

2 8 3 6

5 4

Figure 4.3: Reducibility between cases

Y = u6k−1 and Z ∈ {u2 , u5 , . . . , u6k−10 , u6k−4 }. Since 6k − 8 > 3k − 2 it follows that

u6k−8 is a good vertex for u1 and for all pairs {u1 , Y }, where d∗ (u1 , Y ) ≡ 1(mod 3)

and Y ∈
/ {u6k−10 , u6k−4 }. We have found a good vertex (u6k−5 or u6k−8 ) for all triplets

{u1 , Y, Z} such that {Y, Z} =

6 {u6k−7 , u6k−10 }, {u6k−7 , u6k−4 },

{u6k−1 , u6k−10 }, {u6k−1 , u6k−4 }.

Since k ≥ 4 we find for triplets {u1 , u6k−7 , u6k−10 }, {u1 , u6k−7 , u6k−4 },

{u1 , u6k−1 , u6k−10 } and {u1 , u6k−1 , u6k−4 } good vertices u3k−4 , u3k−4 ,

u6k−11 and u3k−1 , respectively (e. g., using Lemma 4.3.2).

Case 3. We deduce that u5 is a good vertex for all pairs {u1 , Y }, where d∗ (u1 , Y ) ≡

2(mod 3) and Y ∈
/ {u3 , u9 }. It follows that u5 is a good vertex for all triplets

{u1 , Y, Z}, unless Y = u3 and Z ∈ {u6 , u9 , . . . , u6k }; Y = u9 and Z ∈ {u6 , u12 , u15 , . . . ,

u6k }. Also u3k−1 is a good vertex for u1 and for all pairs {u1 , Y }, where d∗ (u1 , Y ) ≡

2(mod 3) and Y ∈
/ {u3k−3 , u3k+3 }. It follows that there exists a good vertex (u5 or

u3k−1 ) for all triplets {u1 , Y, Z} such that {Y, Z} =

6 {u3 , u3k−3 }, {u3 , u3k+3 }, {u9 , u3k−3 },
 58

{u9 , u3k+3 } (note that for k = 4 the third triplet must be eliminated from the list).

Now u8 is a good vertex for all pairs {u1 , Y }, where d∗ (u1 , Y ) ≡ 2(mod 3) and

Y ∈
/ {u6 , u12 }. It follows that u8 is a good vertex for all four remaining triplets,

except {u1 , u3 , u12 } for k = 5 and n = 30. For this last triplet u3k+1 = u16 is a good

vertex.

Cases 7 and 9 can be reduced to the case 7 without imposing any inequality be-

tween the distances d∗ (X, Y ) and d∗ (X, Z).

Let A = {u2 , u5 , u8 , . . . , u6k−1 } and B = {u3 , u6 , u9 , . . . , u6k }. It is necessary to prove

that for any triplet {u1 , Y, Z}, where Y ∈ A and Z ∈ B there is a good vertex on the

outer cycle. From the previous case we have seen that u5 is a bad vertex for pairs

{u1 , Z}, where Z ∈ B if and only if Z ∈ {u3 , u9 } and a bad vertex for pairs {u1 , Y },

where Y ∈ A if and only if Y ∈ {u2 , u5 , u8 , u3k+8 , u3k+11 , . . . , u6k−1 }. It follows that

u5 is a good vertex for all triplets {u1 , Y, Z}, where Y ∈ A and Z ∈ B, unless: Y ∈ A

and Z ∈ {u3 , u9 }; Y ∈ {u2 , u5 , u8 , u3k+8 , u3k+11 , . . . , u6k−1 } and Z ∈ B (these sets of

pairs {Y, Z} will be denoted by α and β, respectively). For the remaining triplets

{u1 , Y, Z}, where {Y, Z} ∈ α ∪ β we must find other good vertices on the outer cycle.

Consider now vertex u3k+1 . This vertex is a good vertex for all pairs {u1 , Y }, where

Y ∈ A\{u3k−1 , u3k+5 }. Since 3k − 7 ≥ 5 it follows that u3k+1 is also a good vertex

for all pairs {u1 , Z}, where Z ∈ {u3 , u9 }. Therefore the set α is reduced to the set

α1 of pairs {Y, Z} such that Y ∈ {u3k−1 , u3k+5 } and Z ∈ {u3 , u9 }. Now u8 is a good

vertex for all pairs {u1 , Y }, where Y ∈ {u14 , u17 , . . . , u3k+8 } and all pairs {u1 , Z},

where Z ∈ {u3 , u9 , u15 , u18 , . . . , u6k }. Since 3k − 1 ≥ 14 for k ≥ 5 it follows that u8

is a good vertex for all pairs in α1 if k ≥ 5 and for k = 4 we must consider only the

pairs {u11 , u3 } and {u11 , u9 }. From figure 2 we deduce that u3k+3 = u15 is a good
 59

vertex for {u1 , u11 , u3 } and u3k+4 = u16 is a good vertex for {u1 , u11 , u9 }.

It remains to find good vertices for pairs in β. Since u6k−5 is a good vertex for all pairs

{u1 , Y }, where Y ∈ A\{u6k−7 , u6k−1 } and a good vertex for all pairs {u1 , Z}, where

Z ∈ {u3k−3 , u3k , . . . , u6k−9 }, β is reduced to γ ∪ δ, where γ consists of all pairs {Y, Z}

with Y ∈ {u6k−7 , u6k−1 } and Z ∈ B and δ of {Y, Z} with Y ∈ {u2 , u5 , u8 , u3k+8 , u3k+11 ,

. . . , u6k−10 , u6k−4 } and Z ∈ {u3 , u6 , . . . , u3k−6 , u6k−6 , u6k−3 , u6k }.

u3k−1 is a good vertex for all pairs {u1 , Y }, where Y ∈ {u3k+5 , u3k+8 , . . . ,

u6k−1 } and all pairs {u1 , Z}, where Z ∈ B\{u3k−3 , u3k+3 }. Therefore γ is reduced to

ε, consisting of all pairs {Y, Z} such that Y ∈ {u6k−7 , u6k−1 } and Z ∈ {u3k−3 , u3k+3 }

and δ is reduced to µ, which consists of {Y, Z} with Y ∈ {u2 , u5 , u8 } and Z ∈

{u3 , u6 , u3k−6 , u6k−6 , u6k−3 , u6k }.

u6 is a good vertex for all pairs {u1 , Y }, where Y ∈ {u2 , u3k+8 , u3k+11 , . . . ,

u6k−1 } and all pairs {u1 , Z}, where Z ∈ {u12 , u15 , . . . , u3k+6 }.

If k ≥ 5 then 6k−7 ≥ 3k+8 and 3k−3 ≥ 12; therefore u6 is a good vertex for all triples

{u1 , Y, Z}, where {Y, Z} ∈ ε. If k = 4 the pairs in ε are {u17 , u9 }, {u17 , u15 }, {u23 , u9 }

and {u23 , u15 }; good vertices for these pairs are vertices u24 , u5 , u16 and u9 , respec-

tively, which also are good vertices for u1 .

It remains to study the pairs from µ. u3k is a good vertex for all pairs {u1 , Y }, where

Y ∈ {u2 , u5 , . . . , u3k−4 } and all pairs {u1 , Z}, where Z ∈ {u3k+6 , u3k+9 , . . . , u6k }. Since

3k − 4 ≥ 8 and 6k − 6 ≥ 3k + 6 it follows that µ is reduced to the set ν of pairs

{Y, Z}, where Y ∈ {u2 , u5 , u8 } and Z ∈ {u3 , u6 , u3k−6 }.

Finally, u3k+1 is a good vertex for all pairs {u1 , Y }, where

Y ∈ A\{u3k−1 , u3k+5 } and for all pairs {u1 , Z}, where

Z ∈ {u3 , u6 , . . . , u3k−3 }.
 60

Consequently, u3k+1 is a good vertex for u1 and all pairs from ν. 2

Chapter 5

Metric dimension and d-sets of

connected graphs

A d-set relatively to a pair of distinct vertices of a connected graph G is the set of

vertices having different distances from these vertices.

In this chapter, it is shown that for a connected graph G of order n and diameter 2

the number of pairs such that their d-sets are equal to V (G) is bounded above by

bn2 /4c and it is conjectured that this holds for any connected graph.

A lower bound for the metric dimension of G, dim(G) is proposed in terms of a

family of d-sets of G having the property that every subfamily containing at least

r ≥ 2 members has an empty intersection.

Three sufficient conditions which guarantee that a family F = (Gn )n≥1 of graphs with

unbounded order has an unbounded metric dimension are also proposed.

Finally, d-sets are used to show that dim(N en ) = 3 when n is odd and 2 otherwise,

where N en is the necklace graph of order 2n + 2.

61
 62

5.1 Notation and preliminary results

Let G be a connected graph. The distance d(u, v) between two vertices u, v ∈ V (G)

is the length of a shortest path between them and the diameter of G, denoted by

diam(G) is max d(u, v).

u,v∈V (G)

For a pair of distinct vertices of G, p = {x, y} we shall denote by D(p) or D(x, y)

the set of vertices z ∈ V (G) such that d(z, x) 6= d(z, y). Such a set will be called a

distinguishing set (or a d-set) relatively to the pair {x, y}.

It is clear that {x, y} ⊂ D(x, y) ⊂ V (G) for any pair {x, y}. Some properties of

d-pairs will be shown in the next section.

The metric dimension of a connected graph G has been defined in [19]. An equivalent

definition is the following [9]: Let W = {w1 , w2 , ..., wk } be an ordered set of vertices

of G and let v be a vertex of G. The representation r(v|W ) of v with respect to W

is the k-tuple (d(v, w1 ), d(v, w2 ), ..., d(v, wk )). If distinct vertices of G have distinct

representations with respect to W , then W is called a resolving set for G. It is clear

that for any pair of distinct vertices {x, y} of G there exists a vertex wi ∈ W such

that d(x, wi ) 6= d(y, wi ), hence D(x, y) ∩ W 6= ∅ for any resolving set W .

A resolving set of minimum cardinality is called a basis for G and the number of

elements in a basis is the metric dimension of G, denoted by dim(G). The problem

of determining whether dim(G) < k is an N P -complete problem [18].

The property of a given set W ⊂ V (G) to be resolving set of G can be verified only

for vertices from V (G)\W , since every vertex w ∈ W is the only vertex of G whose

distance from W is 0.

Let F be a family of connected graphs Gn : F = (Gn )n≥1 depending on n as follows:

the order |V (G)| = ϕ(n) and lim ϕ(n) = ∞. If there exists a constant C > 0 such
n→∞
 63

that dim(G) ≤ C for every n ≥ 1 then we shall say that F has bounded metric

dimension; otherwise F has unbounded metric dimension.

If all graphs in F have the same metric dimension (which does not depend on n),

F is called a family with constant metric dimension [24]. A connected graph G has

dim(G) = 1 if and only if G is a path [9]; cycles Cn have metric dimension 2 for every

n ≥ 3. Also generalized Petersen graphs P (n, 2), antiprisms An and Harary graphs

H4,n are families of graphs with constant metric dimension [24].

Other families of graphs have unbounded metric dimension: if Wn denotes a wheel

with n spokes and J2n the graph deduced from the wheel W2n by alternately deleting

n spokes, then dim(Wn ) = b 2n+2

5
c for every n ≥ 7 [4] and dim(J2n ) = b 2n
3
c [39] for

every n ≥ 4.

An example of a family which has bounded metric dimension is the family of necklaces

(N en )n≥1 . The necklace graph, denoted by N en [38] is a cubic Halin graph [30]

obtained by joining by a cycle all vertices of degree 1 of a caterpillar (also called comb)

having n vertices of degree 3 and n + 2 vertices of degree 1, denoted by u1 , u2 , ..., un

and v0 , v1 , ..., vn+1 , respectively (see fig. 5.2).

The metric dimension dim(N en ) is bounded but not constant and it depends on the

parity of n. This will be shown in the last section.

The clique number of a graph G, denoted by ω(G), is the maximum number of vertices

in a complete subgraph of G and N (v) denotes the set of vertices adjacent with v.

5.2 Properties of d-sets

The following lemma is based on a simple observation: if for a pair of distinct vertices

{x, y} the distance d(x, y) is even, the middle vertex v of a shortest path between x
 64

and y has equal distances to x and to y, hence v 6∈ D(x, y).

Lemma 5.2.1. If D(x, y) = V (G) for a pair {x, y} of distinct vertices of a connected

graph G then d(x, y) is odd.

Lemma 5.2.2. If D(x, y) = {x, y} for any pair of distinct vertices of a connected

graph G then G is a complete graph.

Proof. Let G be a graph of order n satisfying this property and u, v ∈ V (G), u 6= v.

If d(u, v) = k ≥ 2, let u, x1 , ..., xk = v be a shortest path between them. We have

d(v, u) = k and d(v, x1 ) = k − 1, which implies v ∈ D(u, x1 ) and contradicts the

hypothesis. It follows that d(u, v) = 1 for any vertex v 6= u, or d(u) = n − 1. Since u

was an arbitrary vertex it follows that G is a complete graph.

Theorem 5.2.3. If G has n vertices and diam(G) = 2 then the number of pairs

{x, y} such that D(x, y) = V (G) is bounded above by bn2 /4c. This bound is reached

only for Kbn/2c,dn/2e .

Proof. Let G be a diameter 2 graph of order n and a, b be a pair such that D(a, b) =

V (G).

By Lemma 5.2.1 it follows that ab ∈ E(G). Also N (a)∩N (b) = ∅ since otherwise every

vertex c ∈ N (a) ∩ N (b) does not belong to D(a, b) which contradicts the hypothesis.

By denoting A = N (a) and B = N (b), A ∪ B is a partition of V (G) since otherwise

there exists a vertex c such that d(a, c) = d(b, c) = 2, hence D(a, b) 6= V (G), a

contradiction.

Since every two distinct vertices u, v ∈ A have a common neighbor which is a, we

have D(u, v) 6= V (G); a similar situation occurs for B. It follows that the number

of pairs {x, y} such that D(x, y) = V (G) is bounded above by |A||B| ≤ bn2 /4c since
 65

|A| + |B| = n.

Equality can hold when A∪B is an equipartition of V (G), i.e. −1 ≤ |A|−|B| ≤ 1 and

vertex sets A and B induce subgraphs consisting of isolated vertices. Indeed, if an edge

uv ∈ E(G) has u, v ∈ A then we have d(v, u) = d(v, a) = 1 and d(u, v) = d(u, a) = 1,

hence both pairs {u, a} and {v, a} do not satisfy the required property and the number

of pairs is strictly less than bn2 /4c.

Because diam(G) = 2 we deduce that every vertex of A is adjacent to every vertex

of B, hence only Kbn/2c,dn/2e can have bn2 /4c pairs {x, y} such that D(x, y) = V (G).

But Kbn/2c,dn/2e has this property since every pair of adjacent vertices {x, y} has

D(x, y) = V (Kbn/2c,dn/2e ).

If a graph G of order n has diam(G) = n − 1 then G is a path Pn and every pair

of vertices {x, y} of Pn such that d(x, y) is odd satisfies D(x, y) = V (Pn ); the number

of such pairs equals bn2 /4c.

By joining by an edge the centers of the stars K1,bn/2c−1 and K1,dn/2e−1 for any n ≥ 4

the resulting graph G has diameter 3 and the number of pairs such that D(x, y) =

V (G) also equals bn2 /4c.

These facts lead to the following conjecture:

Conjecture. For any connected graph G of order n ≥ 2 the number of pairs {x, y}

such that D(x, y) = V (G) is bounded above by bn2 /4c.

We can use d-sets for obtaining bounds for the metric dimension of a graph G.

Theorem 5.2.4. Let m, r ≥ 2 be two integers. Suppose that a connected graph G

T
has m pairs of vertices p1 , ..., pm such that D(pi ) = ∅ for any I ⊂ {1, ..., m} and
i∈I
|I| ≥ r.

Then dim(G) ≥ dm/(r − 1)e and this inequality is tight.

 66

Proof. Let W be a resolving set of G such that |W | = dim(G). It follows that

W ∩D(pi ) 6= ∅ for any i = 1, ..., m. Let p be a natural number such that p(r −1)+1 ≤

m. We shall prove that

pr−p+1
[
|W ∩ D(pi )| ≥ p + 1 (5.2.1)
i=1
pr−p+1
S
Suppose that W ∩ D(pi ) = {x1 , ..., xt } and t ≤ p. Let Aj = {i : 1 ≤ i ≤
i=1
pr − p + 1, xj ∈ D(pi )} for any 1 ≤ j ≤ t. Since |W ∩ D(pi )| ≥ 1 for 1 ≤ i ≤ m and

|Aj | ≤ r − 1 for any 1 ≤ j ≤ t by the hypothesis, we can write

t
[ t
X
pr − p + 1 = | Aj | ≤ |Aj | ≤ t(r − 1),
j=1 j=1

a contradiction since t ≤ p.
pr−p+1
S
Because W ⊃ W ∩ D(pi ) from (5.2.1) it follows that |W | ≥ p + 1. Since
i=1
p ≤ (m − 1)/(r − 1), we can choose p = b(m − 1)/(r − 1)c and we deduce that

dim(G) = |W | ≥ b(m − 1)/(r − 1)c + 1 = dm/(r − 1)e.

To see that this bound be reached consider the graph denoted by ZHn which con-

sists of a zig-zag sequence of hexagons having common edges as in fig. 5.1 for

n = 5. We get |V (ZHn )| = 4n + 2 and diam(ZHn ) = 2n + 1 for any n ≥

x 2 u2 x4 u4
x 1 u1 x3 u3 x5 u5

u0 y
y2 v2 y v4 6
4
y1 v1 y3 v3 y v5
5

Figure 5.1: ZH5

 67

1. Also D(x1 , y1 ) = D(u1 , v1 ) = {x1 , y1 , u1 , v1 , x2 , u2 }; D(x2 , y2 ) = D(u2 , v2 ) =

{x2 , y2 , u2 , v2 , y1 , v1 , y3 , v3 }; D(x3 , y3 ) = D(u3 , v3 ) = {x3 , y3 , u3 , v3 , x2 , u2 , x4 , u4 } and

so on. For any 2 ≤ i ≤ n − 1 we have |D(xi , yi )| = |D(ui , vi )| = 8. Considering

m = 2n pairs (xi , yi ) and (ui , vi ) for 1 ≤ i ≤ n the hypothesis of the theorem is

satisfied for r = 7. It follows that dim(ZHn ) ≥ d n3 e.

The opposite inequality holds for every n ≥ 7. Suppose that the hexagons of ZHn

are denoted by H1 , ..., Hn from left to right. For any n ≥ 7 we can construct a metric

basis W of ZHn of cardinality d n3 e as follows (see Theorem 6.2.2b):

We choose any vertex of degree 2 in the hexagons numbered by 2, 5, 8, ..., n−4, n−1 for

n ≡ 0(mod 3); 2, 5, 8, ...., n − 5, n − 2, n for n ≡ 1(mod 3) and 2, 5, 8, ...., n − 6, n − 3, n

for n ≡ 2(mod 3). It can be easily verified that any two vertices of ZHn having equal

distances to any vertex of W can be distinguished by other vertices of W .

5.3 Graphs with unbounded metric dimension

In this section three sufficient conditions that guarantee that a family F = (Gn )n≥1

have unbounded metric dimension are presented.

The observation that the distances from a given vertex to all vertices of a clique can

take only 2 distinct values leads to the following property.

Lemma 5.3.1. Suppose that there exists integers r, d, m ≥ 1 and C ⊂ V (G), |C| = m

where m ≥ (d + 1)r + 1 such that the distance between any two vertices in C is less

than or equal to d. Then for any subset A ⊂ V (G), |A| = r there exist x, y ∈ C, x 6= y

such that D(x, y) ∩ A = ∅.

Proof. Let A = {x1 , ..., xr }. We apply induction on r. For r = 1 we deduce m ≥ d+2;

 68

all vertices of C being pairwise at a distance at most d apart, the multiset {d(x1 , w) :

w ∈ C} contains at most d + 1 distinct distances. It follows that there exist x, y ∈ C,

x 6= y such that d(x, x1 ) = d(y, x1 ) and the property is verified. Let r ≥ 2 and suppose

that the property is true for r − 1. If C is such that |C| = m ≥ (d + 1)r + 1 and A

is any set containing r vertices x1 , ..., xr , the multiset {d(x1 , w) : w ∈ C} contains at

most d + 1 distinct distances.

It follows that there exists a subset C1 ⊂ C such that the multiset {d(x1 , w) : w ∈ C1 }

contains only equal numbers and |C1 | ≥ d((d + 1)r + 1)/(d + 1)e = (d + 1)r−1 + 1.

Applying the induction hypothesis for C1 and subset {x2 , ..., xr } we find a pair {x, y},

where x, y ∈ C1 ⊂ C, x 6= y such that D(x, y) ∩ {x2 , ..., xr } = ∅. By construction

x1 6∈ D(x, y), hence D(x, y) ∩ A = ∅.

Since every resolving set W of G must have W ∩D(x, y) 6= ∅ we get dim(G) ≥ r+1.

If ωd (G) denotes the maximum number of vertices of a subset C ⊂ V (G) such that the

distance between any two vertices in C is not greater than d, we deduce the following

consequence.

Corollary 5.3.2. We have dim(G) ≥ dlogd+1 ωd (G)e for any connected graph G.

If d = 1 then ω1 (G) = ω(G), the clique number of G and the following corollary

holds.

Corollary 5.3.3. If G is a connected graph with clique number ω(G), then dim(G) ≥

dlog2 ω(G)e.

This inequality is tight for ω(G) ≤ 5. For example, take a copy of K5 having ver-

tices {x, y, z, t, w}. We shall consider three new vertices a, b, c and define a new graph

G such that V (G) = V (K5 ) ∪ {a, b, c} and E(G) = E(K5 ) ∪ {ax, ay, by, bz, cz, ct}. We
 69

have ω(G) = 5 and dim(G) = 3, a metric basis being {a, b, c}. Similar examples of

graphs G for which dim(G) ≥ dlog2 ω(G)e can be found for ω(G) ≤ 4, but for greater

values of ω(G) the inequality may be strict. We deduce the following consequence for

families F having unbounded clique number.

Corollary 5.3.4. If a family F = (Gn )n≥1 has unbounded clique number then it also

has unbounded metric dimension.

Theorem 5.2.4 also has the following consequence concerning families with un-

bounded metric dimension.

Corollary 5.3.5. Let r ≥ 2be a fixed natural number. Suppose that family F =

(Gn )n≥1 has the following property: for each n graph Gn contains m = ψ(n) pairs of
T
vertices p1 , ..., pm such that D(pi ) = ∅ for any I ⊂ {1, ..., m} and |I| ≥ r.
i∈1
If lim ψ(n) = ∞ then lim dim(Gn ) = ∞, which implies that F has unbounded
n→∞ n→∞

metric dimension.

An example of such a family is F1 = (ZHn )n≥1 considered above. Another example

is F2 = (Kn,n )n≥1 since dim(Kn,n ) = 2n − 2 for every n ≥ 2 and there exist n2 − n

pairs {z, t} contained in the same partite set of Kn,n such that D(x, y) = {z, t}.

Lemma 5.3.6. If for the family F = (Gn )n≥1 there exists a constant C > 0 such that

dim(Gn ) ≤ C for any n ≥ 1 then F has unbounded metric dimension.

Proof. Let diam(G) = d = d(n) ≤ C and Wn = {w1 , ..., wk } be a minimum resolving

set for Gn containing k = k(n) vertices. The representation (d(v, w1 ), ..., d(v, wk )) of

a vertex v ∈ V (Gn ) can contain a component equal to 0 only if v ∈ W .

It follows that the number of distinct representations is bounded above by kdk−1 +

 70

dk . Because the sequence of metric dimensions (k(n))n≥1 is bounded, there exists a

constant C1 > 0 such that k(n) = dim(Gn ) ≤ C1 for every n ≥ 1. Since all vertices of

Gn have distinct representations we deduce that kdk−1 + dk ≥ |V (Gn )| = ϕ(n) → ∞

as n → ∞.

On the other hand kdk−1 + dk ≤ C1 C C1 −1 + C C1 , a contradiction. It follows that F

has unbounded metric dimension.

5.4 Metric dimension of necklace graph

Since necklace graph N en is not a path we have dim(N en ) ≥ 2 for any n ≥ 1. N e1 is

K4 , so dim(N e1 ) = 3 and also dim(N e2 ) = 3.

Theorem 5.4.1. For every n ≥ 3 we have


 2, if n is even ;

dim(N eh )=
 3, if n is odd


Proof. a) Let n be even. In this case a resolving set of N en is W = {v0 , vn/2 } since

the representations of the vertices with respect to W are the following:


 (i, n − i),

for 0 ≤ i ≤ n2 ;
2
r(vi |W ) =
 (n − i + 2, i − n ), for
 n
+ 1 ≤ i ≤ n + 1.
2 2

and


 (i, n − i + 1),

for 1 ≤ i ≤ n2 ;
2
r(ui |W ) =
 (n − i + 2, i − n n
+ 1), for + 1 ≤ i ≤ n.

2 2
 71

Since all vertices have distinct representations we obtain dim(N en ) = 2 in this case.

b) When n is odd we show that W = {v0 , v(n+1)/2 , u(n−1)/2 } is a resolving set. The

representations of the vertices of N en with respect to U = {v0 , v(n+1)/2 } are the

following:

 (i, n−1 − i + 1),

for 0 ≤ i ≤ n+1
;
2 2
r(vi |U ) =
 (n − i + 2, i − n+ n+3
), for + 1 ≤ i ≤ n + 1.

2 2

and


 (i, n−1 − i + 2),

for 1 ≤ i ≤ n+1
;
2 2
r(ui |U ) =
 (n − i + 2, i − n+ n+3
+ 1), for ≤ i ≤ n.

2 2

n+1
U distinguishes all vertices of N en unless ui and vn+2−i for 1 ≤ i ≤ 2
. This can be

done by u(n−1)/2 , hence W is a resolving set, thus implying dim(N en ) ≤ 3.

We will show that dim(N en ) ≥ 3, by proving that any resolving set has at least three

vertices. Suppose that there exists a resolving set W of N en such that |W | = 2. We

shall prove that this leads to a contradiction.

By taking into account the action of the automorphism group of N en , it is suf-

ficient to consider only the cases when vk−1 , vk , ... or vn+1 belongs to W , where

k = diam(N en ) = b(n + 3)/2c.

A. If vn+1 ∈ W we have d(vk−2 , vn+1 ) = d(uk−2 , vn+1 ) = d(vk−1 , un+1 ) = d(uk−1 , un+1 ) =

k − 1 (see fig. 5.2).

Also D(vk−2 , uk−1 ) = {v0 }∪{uk , uk+1 , ..., un , v1 , v2 ..., vk−3 }∪{vk−2 , uk−1 }; D(vk−1 , uk−2 ) =

{v0 } ∪ {u1 , u2 , ..., uk−3 , vk , vk+1 ..., vn } ∪ {vk−1 , uk−2 }; D(vk−2 , uk−1 ) ∩ D(vk−1 , uk−2 ) =

{v0 } but d(v0 , vk−2 ) = d(v0 , uk−2 ) = k − 2. It follows that there is no resolving set

having two vertices including vn+1 .

 72

v1 v2 vk-2 vk-1 vk vk+1 vn

v0
vn+1

u1 u2 uk-2 uk-1 uk uk+1 un

Figure 5.2: N en

B. If vk−1 ∈ W we deduce d(vk−2 , vk−1 ) = d(uk−1 , vk−1 ) = d(uk , vk−1 ) = 1. We get

(see fig. 5.2)

D(vk−2 , uk−1 ) = {v0 } ∪ {v1 , v2 , ..., vk−3 , uk , uk+1 , ..., un } ∪ {vk−2 , uk−1 }; D(vk , uk−1 ) =

{vn+1 }∪{u1 , u2 , ..., uk−2 , vk+1 , ..., vn }∪{vk , uk−1 }; D(vk−2 , uk−1 )∩D(vk , uk−1 ) = {uk−1 },

but d(uk−1 , vk ) = d(uk−1 , vk ) = 2 and vk−2 , vk have unit distances from vk−1 . It fol-

lows that vk−1 6∈ W , a contradiction.

C. vk ∈ W . In this case D(vk−1 , uk ) = {vn+1 }∪{v1 , , ..., vk−2 , uk+1 , ..., un }∪{vk−1 , uk };

D(vk+1 , uk ) = {v0 , vn+1 } ∪ {vk+2 , , ..., vn , u2 , ..., uk−1 } ∪ {vk+1 , uk } (see fig. 5.2).

We have D(vk+1 , uk ) ∩ D(vk−1 , uk ) = {vn+1 , uk }. But vn+1 has equal distances apart

from vk−2 and uk−1 , which have equal distances to vk . Also uk does not distinguish

vertices vk−1 and vk+1 , hence vk 6∈ W .

D. If vi ∈ W and k+1 ≤ i ≤ n we shall consider its antipodal vertex ui−k+1 (for which

d(vi , ui−k+1 ) = k) and its neighbor (see fig. 5.3). We get d(vi , vi−k+1 ) = d(vi , ui−k ) =

d(vi , ui−k+2 ) = k + 1 and d(vi , v1 ) = d(vi , u1 ) = n − i + 3.

Since vn+1 and v0 have equal distances to u1 and v1 , respectively, it follows that

v0 , vn+1 6∈ W .

Also d(vj , v1 ) = d(vj , u1 ) and d(uj , v1 ) = d(uj , u1 ) for every k + 1 ≤ j ≤ n, which

 73

v1 vi-k+1 vk+1 vi vn
v0
vn+1

u1 u i-k u u uk+1 un
i-k+1 i-k+2

Figure 5.3: N en with two antipodal vertices

implies that

(W \{vi }) ∩ {v0 , vk+1 , ..., vn+1 , uk+1 , ..., un } = ∅

For any x, y ∈ V (N en ) and x 6= y denote by D0 (x, y) the set D(x, y)\{v0 , vk+1 , ..., vn+1 , uk+1 , ..., un }.

We deduce D0 (ui−k , vi−k+1 ) = {vi−k+2 , ..., vk , u1 , ..., ui−k+1 }∪{ui−k , vi−k+1 }; D0 (ui−k+2 , vi−k+1 ) =

{v1 , ..., vi−k , ui−k+3 , ..., uk }∪{ui−k+2 , vi−k+1 } hence D0 (ui−k , vi−k+1 )∩D0 (ui−k+2 , vi−k+1 ) =

{vi−k+1 }. It follows that W = {vi , vi−k+1 }. But vertices ui−k and ui−k+2 have equal

distances from both vertices of W , a contradiction. Consequently, every resolving set

of V (N en ) has at least three vertices for n odd.

Chapter 6

On metric and partition dimension

of some infinite regular graphs

In this chapter some infinite regular graphs generated by tilings of the plane by

regular triangles and hexagons are considered. These graphs have no finite metric

bases but their partition dimension is finite and is evaluated in some cases. Also, it

is proved that for every n ≥ 2 there exist finite induced subgraphs of these graphs

having metric dimension equal to n as well as infinite induced subgraphs with metric

dimension equal to three.

6.1 Introduction

It is natural to think that the partition dimension and metric dimension are related;

in [4] it was shown that for any nontrivial connected graph G we have pd(G) ≤

dim(G) + 1.

However, the partition dimension may be much smaller than the metric dimension.

Let (i, j) and (i0 , j 0 ) be two points with integral coordinates in Z2 . It is well known

that the following definitions yield metrics for Z2 : d4 ((i, j), (i0 , j 0 )) = |i − i0 | + |j − j 0 |

74
 75

(city block distance) and d8 ((i, j), (i0 , j 0 )) = max(|i−i0 |, |j −j 0 |) (chessboard distance).

The indices 4 and 8 are appropriate because they represent the number of points at

distance one(the neighbors) from a given point with respect to these two metrics.

These two metrics on Z2 generate two infinite graphs (Z2 , E4 ) and (Z2 , E8 ) having

the same vertex set Z2 and the set of edges consisting of all pairs of vertices whose

city block and chessboard distances are 1. (Z2 , E4 ) is a planar 4-regular graph whose

regions are unit squares and it is also known as the square lattice graph. (Z2 , E8 ) is

8-regular and can be obtained from (Z2 , E4 ) by drawing all diagonals of unit squares.

In [31] it was proved that these two graphs have no finite metric bases and for any

natural number n ≥ 3, there exist induced subgraphs of (Z2 , E4 ) and (Z2 , E8 ), respec-

tively having metric dimension equal to n and partition dimension equal to three.

Also, in [40] it was proved that pd(Z2 , E4 ) = 3 and pd(Z2 , E8 ) = 4.

In what follows we shall consider some infinite regular graphs generated by tilings of

the plane by regular hexagons or equilateral triangles.

The graph G3 is a planar 3-regular infinite graph whose regions are regular hexagons

of unit side. V (G3 ) consists of the vertices of these hexagons, two vertices being adja-

cent if they are the extremities of a unit side of a hexagon in the tiling. Similarly, G6

is a planar 6-regular infinite graph whose regions are equilateral triangles of unit side.

It is also known as the triangular lattice graph. Three kinds of rhombuses of unit side

appear in G6 . By drawing all diagonals of these rhombuses we get a 12-regular infinite

graph denoted by G12 ; if all vertical diagonals are deleted, the resulting 10-regular

infinite graph is denoted by G10 (see Fig.6.1). The indices 3, 6, 10 and 12 represent

the number of vertices at distance 1 from a given vertex. For sake of simplicity for

G10 and G12 or their subgraphs we shall not draw the corresponding diagonals of the
 76

G6
G3

G12 G10

Figure 6.1: Graphs G3 , G6 , G10 and G12

rhombuses; this will be clear from the context.

6.2 Main results

Lemma 6.2.1. The graphs G3 , G6 , G10 and G12 have no finite metric basis, i.e.,

dim(G3 ) = dim(G6 ) = dim(G10 ) = dim(G12 ) = ∞.

Proof. Figs. 6.2a) − d) represent two vertices x, y in G3 , G6 , G10 , G12 having their
√
Euclidean distances equal to 1 and to 3, respectively and subgraphs Gi (x, y)(i =

3, 6, 10, 12) consisting of vertices z such that d(x, z) = d(z, y). Suppose that G3 has

a finite metric basis S. We can find two vertices x, y and a subset T ⊂ G3 (x, y)
 77

consisting of all vertices z ∈ G3 (x, y) such that d(z, x) = d(z, y) ≤ k for k large

enough, such that S ⊂ T . This implies that d(x, z) = d(y, z) for all z ∈ S, a

contradiction. The proof is similar for other three graphs.

x y

x y

a) b)

x y y

c) d)

Figure 6.2: Subgraphs of vertices having equal distances to x and y.

Fig. 6.3 represents some induced subgraphs of G3 and G6 : Fn has n hexagons,

ZZn is a zig-zag sequence of n hexagons, H1 and H2 are infinite subgraphs of G6 and

Ln is an induced subgraph of G6 consisting of n rhombuses with one diagonal.

We have dim(H1 ) = dim(H2 ) = 3 since vertices having the same distance to B are

distinguished by A and C, respectively.

Theorem 6.2.2. a) For every n ≥ 2 we have: dim(Fn ) = n and Fn has n!4n metric

bases; dim(Ln ) = n and Ln has n!2n metric bases.

 78

u v

...........
w t
Fn

e g u v

...........
f h w t
a b c d

Z Z ( n odd)
n

A B

..... .....
C

H
1
A

..... .....
B

H
2
u

.....
v
Ln

Figure 6.3: Induced subgraphs of G3 and G6 .

b) For every n ≥ 7 dim(ZZn ) = d n3 e.

Proof. a) u and w have equal distances to all vertices of Fn different from v and t and

v and t may be distinguished only by u or w if v and t do not belong to any basis of

Fn . It follows that at least one of u, v, w, t must belong to any basis of Fn .

On the other hand, by choosing exactly one vertex of degree two in each hexagon of

Fn ( in 4n ways), the set of these vertices is a metric basis of Fn . These vertices can

be ordered in n! ways and the result follows. The situation of Ln is similar to Fn .

b) Suppose that the hexagons of ZZn are denoted by H1 , . . . , Hn from left to right.
 79

The vertices e and f of H1 (see Fig. 6.3) can be distinguished only by g, h, a or b

and vertices g and h only by e, f, a or b. We deduce that any resolving set W of ZZn

must contain at least one vertex from unavoidable set {e, f, g, h, a, b} of the pair of

hexagons (H1 , H2 ). This set will be denoted by U (H1 , H2 ). A similar situation occurs

for the pair (Hn−1 , Hn ). We shall say that W satisfies the 2 consecutive hexagons

(shortly 2CH) property.

Consider now a triplet of consecutive hexagons (H2 , H3 , H4 ). The vertices u and w of

H3 can be distinguished only by v, t, a, b, c or d and a similar situation also occurs for

v and t. It follows that any resolving set W of ZZn must contain at least one vertex

from unavoidable set {u, v, w, t, a, b, c, d} of (H2 , H3 , H4 ) and a similar property holds

for any triplet (Hi , Hi+1 , Hi+2 ), where 2 ≤ i ≤ n − 3. This set will be denoted by

U (Hi , Hi+1 , Hi+2 ) and we shall say that W satisfies the 3 consecutive hexagons (3CH)

property.

Let B be a metric basis of ZZn . We shall transform B into a set B 0 such that

|B| ≥ |B 0 | , B 0 contains only vertices of degree 2 and B 0 also satisfies 2CH and 3CH

properties by the following algorithm:

Initially B 0 = ∅. We have U (H1 , H2 ) ∩ U (H2 , H3 , H4 ) = {a, b}. If there exists

x ∈ {a, b} such that x ∈ B then label x, assign B 0 ← B 0 ∪ {x} and consider the next

triplet (H2 , H3 , H4 ). Otherwise, if there exists y ∈ {e, f, g} such that y ∈ B then label

y, define B 0 ← B 0 ∪{y} and consider the next triplet. Otherwise, {h} = U (H1 , H2 )∩B.

In this case label g and do B 0 ← B 0 ∪ {g}. Consequently, only one of the vertices

e, f, g, a, b were labeled and included in B 0 .

Consider now the triplet (H2 , H3 , H4 ). If U (H2 , H3 , H4 ) contains a labeled vertex

(or equivalently, a or b has been selected in B 0 ), then consider the next triplet
 80

(H3 , H4 , H5 ). Otherwise a and b are not labeled. We deduce U (H2 , H3 , H4 )∩U (H3 , H4 , H5 ) =

{u, v, c, d}, U (H2 , H3 , H4 )∩U (H4 , H5 , H6 ) = {c, d} and U (H2 , H3 , H4 )∩U (Hi , Hi+1 , Hi+2 ) =

∅ for i ≥ 5. If there exists x ∈ {c, d} such that x ∈ B then label it, do B 0 ← B 0 ∪ {x}

and consider the next triplet. Otherwise, if there exists y ∈ {u, v} such that y ∈ B,

then label it, assign B 0 ← B 0 ∪ {y} and consider the next triplet. If u, v ∈
/ B but there

exists z ∈ {w, t} such that z ∈ B then label u (or v), B 0 ← B 0 ∪ {u} and consider the

next triplet. Otherwise, there exists s ∈ {a, b} such that s ∈ B. In this case label s,

define B 0 ← B 0 ∪ {s} and consider the next triplet (H3 , H4 , H5 ).

Each time when the unavoidable set of a triplet has a vertex which has been labeled

in the previous steps we shall consider the next triplet (or the last pair). We shall pro-

ceed in a similar way for all triplets (Hi , Hi+1 , Hi+2 ) in the order i = 3, . . . , n − 4 and

also for the triplet (Hn−3 , Hn−2 , Hn−1 ) and the pair (Hn−1 , Hn ) (like as for (H1 , H2 )

and (H2 , H3 , H4 )).

Note that h does not belong to U (H2 , H3 , H4 ), so its possible replacement by g

does not affect the property of B 0 to satisfy 2CH and 3CH properties. Similarly,

w, t ∈
/ U (H3 , H4 , H5 ) ∪ U (H4 , H5 , H6 ), so their possible replacement by vertices of

degree 2 u or v, respectively, cannot affect the property of B 0 to satisfy the 3CH

property and this holds for all triplets of hexagons of ZZn . Consequently, B 0 satisfies

2CH and 3CH properties; moreover, B 0 has the property that it contains only ver-

tices of degree 2, i. e., it does not contain vertices common to neighboring hexagons

Hi , Hi+1 for 1 ≤ i ≤ n − 1 and each hexagon has at most one vertex in B 0 .

For i = 1, . . . , n let zi denote a binary variable representing the number of vertices of

Hi included in B 0 . Since B 0 satisfies 2CH and 3CH properties we can write:

z1 + z2 ≥ 1
 81

z2 + z3 + z4 ≥ 1

z3 + z4 + z5 ≥ 1

. . . . . .

zn−3 + zn−2 + zn−1 ≥ 1

zn−1 + zn ≥ 1

By summing up these inequalities we get:

S = z1 + 2z2 + 2z3 + 3(z4 + . . . + zn−3 ) + 2zn−2 + 2zn−1 + zn ≥ n − 2.

We have
n
X
0
3|B| ≥ 3|B | = 3 zi = S + 2z1 + z2 + z3 + zn−2 + zn−1 + 2zn ≥ n
i=1

since 2z1 + z2 ≥ z1 + z2 ≥ 1; 2zn + zn−1 ≥ zn + zn−1 ≥ 1 and z3 + zn−2 ≥ 0. It follows

that dim(ZZn ) = |B| ≥ |B 0 | ≥ d n3 e.

By direct evaluation we find that dim(ZZn ) is equal to: 2 for n = 1 or n = 2; 3 for

3 ≤ n ≤ 9; 4 for 10 ≤ n ≤ 12 and so on. For every n ≥ 7, we can construct a metric

basis B of ZZn of cardinality d n3 e as follows: we choose any vertex of degree two in the

hexagons numbered by: 2, 5, 8, ...., n−4, n−1 for n ≡ 0(mod 3); 2, 5, 8, ...., n−5, n−2, n

for n ≡ 1(mod 3) and 2, 5, 8, ...., n − 6, n − 3, n for n ≡ 2(mod 3). It can be easily

verified that any two vertices of ZZn having equal distances to any vertex of B can

be distinguished by other vertices of B, which concludes the proof.

Lemma 6.2.3. We have pd(G3 ) = pd(G6 ) = 3.

Proof. In [4] it was shown that pd(G) = 2 if and only if G is a path and this property

also holds for infinite graphs. It follows that pd(G3 ) ≥ 3 and pd(G6 ) ≥ 3.
 82

II
I II

III
III

Figure 6.4: Resolving 3-partitions for V (G3 ) and V (G6 ).

Fig. 6.4 provides resolving 3-partitions of G3 and G6 , respectively. It follows that

pd(G3 ) = pd(G6 ) = 3.

The problem of determining partition dimension of G10 and G12 is much more

difficult. We are only able to find some bounds in the next theorems.

Consider now the Euclidean plane containing plane realizations of G10 and G12 en-

dowed with a rectangular system of axes xOy.

Theorem 6.2.4. We have 4 ≤ pd(G10 ) ≤ 5.

Proof. The same argument used in Lemma 6.2.3 implies that pd(G10 ) ≥ 3. Suppose

that there exists a resolving partition Π = (S1 , S2 , S3 ) with three classes of V (G10 ).

We will show that this leads to a contradiction, which implies pd(G10 ) ≥ 4.

Claim 1. In any unit rhombus P QRS which is an induced subgraph of G10 , having

two sides parallel to Ox there are not three vertices belonging to different classes of

Π.

Suppose that P ∈ S1 , Q ∈ S2 and R ∈ S3 . S can belong to S1 , S2 or S3 . In any

case there exist two vertices of P QRS in Si having unit distances to Sj and Sk
 83

respectively(1 ≤ i, j, k ≤ 3, i, j and k pairwise distinct indices), a contradiction.

Claim 2. In any unit rhombus P QRS, induced subgraph of G10 , no three vertices

belong to a class and the fourth to another class of Π.

Suppose that P, Q, R ∈ S1 and S ∈ S2 . Since P, Q, R have equal unit distances to

the class S2 , we deduce that d(P, S3 ) 6= d(Q, S3 ), d(P, S3 ) 6= d(R, S3 ) and d(Q, S3 ) 6=

d(R, S3 ). Since any distance between vertices P, Q and R is 1, one obtains that

d(P, S3 ) and d(Q, S3 ) and d(R, S3 ) may differ by at most one, which contradicts these

inequalities. A consequence of Claims 1 and 2 follows.

Claim 3. Suppose that P, Q ∈ V (G10 ) and d(P, Q) = 1 such that P and Q belong to
√ √
the same line having a slope equal to 3 or − 3 and P ∈ S1 , Q ∈ S2 . Then all the

vertices of G10 lying on the horizontal lines lP and lQ passing through P and Q belong

to the classes S1 and S2 and each unit rhombus, induced subgraph of G10 with vertices

on lP and lQ has two vertices in S1 and two in S2 . A similar conclusion holds if P

√
and Q belong to the same horizontal line by considering the lines having slope 3 or
√
− 3 passing through P and Q.

Without loss of generality, we can assume that there exist two vertices of G10 ,
√ √
P (x, y), Q(x − 21 , y − 23 ) on the same line having slope 3 such that P ∈ S1 , Q ∈ S2

and d(P, Q) = 1. Let P, Q, R, S, P be a 4-cycle containing P Q such that S(x + 1, y).

By Claims 1 and 2 vertices R and S belong to different classes S1 , S2 of Π.

Without loss of generality we can suppose that R ∈ S2 and S ∈ S1 . Since d(P, S2 ) =

d(S, S2 ) = 1 it follows that d(P, S3 ) 6= d(S, S3 ). Because d(P, S) = 1 we deduce

that d(P, S3 ) and d(S, S3 ) differ by 1, e.g., d(P, S3 ) = d(S, S3 ) + 1. We obtain that

there exists a vertex T ∈ S3 such that d(P, S3 ) = d(P, T ), d(S, S3 ) = d(S, T ) and

d(P, T ) = d(S, T ) + 1. By Claim 3 all vertices of G10 lying on lines QR and SP

 84

belong to S1 ∪ S2 .

If d(S, T ) = 1, since all vertices of G10 on P S belong to S1 ∪ S2 , we deduce that the

√
slope of ST is ±1/ 3. Without loss of generality we can suppose that this slope is
√
equal to 1/ 3. Let J be on the line P S such that d(S, J) = d(T, J) = 1. Because

S ∈ S1 and T ∈ S3 , by Claims 1 and 3 it follows that J ∈ S1 . In this case S, J ∈ S1 ,

but d(S, S2 ) = d(J, S2 ) = 1 and d(S, S3 ) = d(J, S3 ) = 1, a contradiction.

Suppose that d(S, T ) ≥ 2; then there exists a shortest path between S and T such
√
that the last edge U T of this path incident to T has the slope equal to 1/ 3. Consider

the unit rhombus T V U W having diagonal T U and such that U V is parallel to P S.

Since U ∈ S1 ∪ S2 , we can suppose that U ∈ S1 . By Claim 1 we have V ∈ S1 ∪ S3 .

If V ∈ S1 then W ∈ S3 , but in this case either d(S, W ) = d(S, T ) − 1 or d(P, W ) =

d(S, W ) = d(S, T ) = d(S, S3 ), a contradiction.

It follows that V ∈ S3 . Consider the unit rhombus determined by the intersection

of the lines P S, QR and W U, T V . By Claim 3 two vertices of this rhombus are in

S1 and two in S3 and in the same time, two belong to S1 and two belong to S2 , a

contradiction.

A similar conclusion holds if U ∈ S2 . It follows that pd(G10 ) ≥ 4.

Figure 6.5 shows that pd(G10 ) ≤ 5, which concludes the proof.

 85

I II

IV III

Figure 6.5: A resolving 5-partition of V (G10 ).

Theorem 6.2.5. The following inequalities hold:

4 ≤ pd(G12 ) ≤ 6.

Proof. As for Theorem 6.2.3 we deduce that pd(G12 ) ≥ 3. Suppose that there exists

a resolving partition Π = (S1 , S2 , S3 ) of V (G12 ) and we will deduce a contradiction.

Since V (G12 ) = V (G10 ) and all distances equal to 1 in G10 are also equal to 1 in

G12 , we deduce that Claims 1, 2 and 3 hold for G12 . In G12 all vertical diagonals of

unit rhombus are edges; this leads to a supplementary property of G12 expressed as

follows:

Claim 4. Let P, Q ∈ V (G12 ) such that d(P, Q) = 1, line P Q is horizontal and P, Q

belong to the same class of Π, say S1 . Then all vertices of G12 lying on vertical lines

passing through P and Q also belong to S1 .

Consider the subgraph of G12 having vertices P, Q, A, B, C, D, E, F, G, drawn in Fig.

5.6. Suppose that F 6∈ S1 . Without loss of generality we can suppose that F ∈ S2 . By

Claim 2 we have C 6∈ S1 . If C ∈ S2 then by Claims 1 and 2 we deduce B ∈ S2 ; thus in

BCGF three vertices belong to a class, which contradicts Claim 2. If C ∈ S3 it follows

 86

D E F G

A B C

P Q

Figure 6.6: A subgraph of G12

that B ∈ S3 and also G ∈ S2 . In this case B, C ∈ S3 and d(B, S1 ) = d(C, S1 ) = 1 and

d(B, S2 ) = d(C, S2 ) = 1, which contradicts the fact that Π is a resolving partition of

G12 .

In a similar way, by considering the parallelogram P QED we deduce that E ∈ S1 .

By repeating this argument it follows that all vertices of G12 lying on lines EP and

F Q belong to S1 . This concludes the proof of this claim.

Now the proof of the theorem follows the same lines as the proof of Theorem 6.2.3.

In the case when T ∈ S3 is such that d(P, T ) = d(S, T ) + 1, by Claim 4 vertex T

cannot belong to the vertical line passing through S and we have the same subcases

as for Theorem 2.2.

Fig. 6.7 represents a resolving 6- partition of V (G12 ), hence pd(G12 ) ≤ 6.

We propose the following conjecture:

Conjecture. The following equalities hold:

pd(G10 ) = 5 and pd(G12 ) = 6.

 87

VI

I II

IV III

Figure 6.7: A resolving 6-partition of V (G12 ).

Chapter 7

Concluding remarks and open

problems

We have seen that the metric dimension of the graphs of convex polytopes Dn , Rn , Qn

is constant and only three vertices appropriately chosen suffice to resolve all the ver-

tices of these classes of convex polytopes. Also the metric dimension of the graphs

of prisms and antiprisms (they are Archimedean convex polytopes) is finite and does

not depend the number of vertices in the graph.

We proved that the metric dimension of the generalized Petersen graphs P (n, 3) is

bounded and determined exact value of the metric dimension when n ≡ 0 or 1(mod

6). In the remaining cases we showed that dim(P (n, 3)) ≤ 4 when n ≡ 3, 4, 5(mod

6), n ≥ 9 and dim(P (n, 3)) ≤ 5 when n ≡ 2(mod 6) and n ≥ 26. In the same way as

for the case n ≡ 1(mod 6) it is not difficult to show that dim(P (n, 3)) ≥ 3 when n ≡

2, 3, 4, 5(mod 6). In this way, some questions naturally arise from the text.

1. Characterize family of plane graphs having constant metric dimension.

2. Find the exact value of the metric dimension of P (n, 3) when n ≡ 2, 3, 4, 5(mod

6).

3. Prove that for any connected graph G of order n ≥ 2 the number of pairs {x, y}

88
 89

such that D(x, y) = V (G) is bounded above by bn2 /4c.

4. Prove that pd(G10 ) = 5 and pd(G12 ) = 6.

Bibliography

[1] M. Bača, Labellings of two classes of convex polytopes, Utilitas Math.

34(1988), 24 − 31.

[2] M. Bača, On magic labellings of convex polytopes, Annals Disc. Math.

51(1992), 13 − 16.

[3] F. Buckley, F. Harary, Distances in graphs, Addison-Wesley Publishing Com-

pany, 1990.

[4] P. S. Buczkowski, G. Chartrand, C. Poisson , P. Zhang, On k-dimensional

graphs and their bases, Periodica Math. Hung., 46(1)(2003), 9 − 15.

[5] J. Caceres, C. Hernando, M. Mora, I. M. Pelayo, M. L. Puertas, C. Seara, D. R.

Wood, On the metric dimension of cartesian product of graphs, SIAM J. Disc.
Math., 2(21), (2007), 423 − 441.

[6] J. Caceres, C. Hernando, M. Mora, M. L. Puertas, I. M. Pelayo,

On the metric dimension of infinite graphs. Preprint. Available at
http://www.arxiv.org/math/0904.4826.

[7] J. Caceres, C. Hernando, M. Mora, I. M. Pelayo, M. L. Puertas, C. Seara, D.

R. Wood, On the metric dimension of some families of graphs. Electronic Notes
in Disc. Math., 22(2005), 129 − 133.

90
 91

[8] G. Chartrand, P. Zhang, Introduction to graph theory, McGraw-Hill, 2005.

[9] G. Chartrand, L. Eroh, M. A. Johnson, O. R. Oellermann, Resolvability in

graphs and metric dimension of a graph, Disc. Appl. Math., 105(2000), 99−113.

[10] G. Chartrand, E. Salehi, P. Zhang, On the partition dimension of a graph,

Congress. Numer., 131(1998), 55 − 66.

[11] G. Chartrand, E. Salehi, P. Zhang, The partition dimension of a graph, Aequa-

tiones Math., 59(2000), 45 − 54.

[12] G. Chartrand, P. Zhang, The theory and applications of resolvability in graphs,

Congress. Numer., 160(2003), 47 − 68.

[13] G. Chartrand, C. Poisson, P. Zhang, Resolvability and upper dimension of a

graph, J. Comput. Math. Appl., 39(2000), 19 − 28.

[14] G. Chappell, J. Gimbell, C. Hartman, Bounds on the metric and partition

dimensions of a graph, Preprint.

[15] R. Diestel, Graph Theory, second edition. Springer-Verlag, 2000.

[16] M. Fehr, S. Gosselin, O. R. Oellermann, The metric dimension of Cayley di-

graphs, Discrete Math., 306(2006), 31 − 41.

[17] M. Fehr, S. Gosselin, O. R. Oellermann, The partition dimension of Cayley

digraphs, Aequationes Math., 71(2006), 1 − 18.

[18] M. R. Garey, D. S. Johnson, Computers and Intractability: A Guide to the

Theory of NP- Completeness, Freeman, New York, 1979.

[19] F. Harary, R. A. Melter, On the metric dimension of a graph, Ars Combin.,

2(1976), 191 − 195.

[20] F. Harary, Status and contrastatus, Sociometry, 22(1959), 23 − 43.

 92

[21] C. Hernando, M. Mora, I. M. Pelayo, C. Seara, D. R. Wood, Extremal

graph theory for metric dimension and diameter. Preprint. Available at
http://www.arxiv.org/math/0705.0938.

[22] M. Imran, A. Q. Baig, A. Ahmad, Families of plane graphs with constant metric
dimension, to appear in Utilitas Math.

[23] M. Imran, A. Q. Baig, M. K. Shafiq, I. Tomescu, On metric dimension of

generalized Petersen graphs P (n, 3), to appear in Ars Combin.

[24] I. Javaid, M. T. Rahim, K. Ali, Families of regular graphs with constant metric
dimension, Utilitas Math., 75(2008), 21 − 33.

[25] M. A. Johnson, Structure-activity maps for visualizing the graph variables aris-
ing in drug design, J. Biopharm. Statist., 3(1993), 203 − 236.

[26] E. Jucovič, Convex polyhedra, Veda, Bratislava, 1981. (in Slovak)

[27] S. Khuller, B. Raghavachari, A. Rosenfeld, Landmarks in graphs, Disc. Appl.

M ath., 70(1996), 217 − 229.

[28] S. Khuller, B. Raghavachari, A. Rosenfeld, Localization in graphs, Technical

Report CS-TR-3326, University of Maryland at College Park, 1994.

[29] E. Lijnen, A. Celemans, Oriented 2-cell embeddings of a graph and their symme-
try classification: Generating algorithms and case study of the Möbius-Kantor
graph, J. Chem. Inf. Comput. Sci., (2004), 44(5), 1552-1564.

[30] L. Lovász, M. D. Plummer, On a family of planar bicritical graphs, Proc. London

Math. Soc., 30(1975), 160 − 176.

[31] R. A. Melter, I. Tomescu, Metric bases in digital geometry, Computer Vision,

Graphics, and Image Processing, 25(1984), 113 − 121.
 93

[32] O. R. Oellermann, J. Peters-Fransen, The strong metric dimension of graphs

and digraphs, Discrete Appl. Math., 155(2007), 356 − 364.

[33] O. R. Oellermann, J. Peters-Fransen, Metric dimension of Cartesian products

of graphs, Utilitas Math., 69(2006), 33 − 41.

[34] C. Poisson, P. Zhang, The metric dimension of unicyclic graphs, J. Combin.

Math. Combin. Comput., 40(2002), 17 − 32.

[35] V. Saenpholphat, P. Zhang, Connected partition dimension of graphs, Discus-

sions Mathematicae Graph Theory, 22(2002), 305 − 323).

[36] P. J. Slater, Leaves of trees, Congress. Numer., 14(1975), 549 − 559.

[37] P. J. Slater, Dominating and reference sets in graphs, J. Math. Phys. Sci.,
22(1998), 445 − 455.

[38] M. M. Syslo, A. Proskurowski, On Halin graphs, in Graph theory, Proc.

Conf., Lagow (Poland), Lecture Notes Math., Springer, New York, vol.
1018(1983), 248 − 256.

[39] I. Tomescu, I. Javaid, On the metric dimension of the Jahangir graph, Bull.
Math. Soc. Sci. Math. Roumanie, 50(98), 4(2007), 371 − 376.

[40] I. Tomescu, Discrepancies between metric dimension and partition dimension

of a connected graph, Discrete Math., 22(308)(2008), 5026 − 5031.

[41] I. Tomescu, I. Javaid, Slamin, On the partition dimension and connected par-
tition dimension of wheels, Ars Combin., 84(2007), 311 − 317.

[42] I. Tomescu, M. Imran, On metric and partition dimensions of some infinite

regular graphs, Bull. Math. Soc. Sci. Math. Roumanie, 52(100), 4(2009), 461 −
472.
 94

[43] I. Tomescu, M. Imran, Metric dimension and d-sets of connected graphs, Sub-
mited

[44] M. E. Watkins, A theorem on Tait colorings with an application to the gener-

alized Petersen graphs, J. Combin. Theory 6(1969), 152 − 164.

[45] D. B. West, An introduction to graph theory, Printice Hall, 1996.