Introduction

All graphs, in this paper, are finite and simple.A general reference for

graph-theoretic ideas can be seen in [4].For a graph G with vertex-set

V(G) and edge-set E(G) an edge-magic total labeling is a bijection λ :

V (G) ∪ E(G) −→ {1, 2, ...., |V (G) ∪ E(G)|}with satisfying the property

that there exist an integer k such that

λ(x) + λ(xy) + λ(y) = k,

for any edge xy in G. We call λ(x) + λ(xy) + λ(y) the edge sum of xy, and k

the magic constant of graph G. In particular, if λ(V (G)) ={1, 2, ..., |V (G)|}

then λ is called super edge-magic total labeling. A graph is called (super)

edge-magic total if it admits any (super) edge-magic total labeling.

The notation of edge-magic total graphs was introduced and studied by

Kotzig and Rosa [5] with a differnt name as graphs with magic valuations,

while the term of super edge-magic total graphs was firstly introduced by

Enomoto et al. [1]. They showed that a star Sn+1 = K1,n is the only

complete bipartite graph which is super edge-magic total. They also showed

that any odd cycle is super edge-magic total, but any wheel is not.

Since then, a number of papers have studied super edge-magic property

in graphs. For instances, Figueroa-Centeno et al. [3] and [2] derived a

necessary and sufficient condition for a graph to be super edge-magic total

and they also showed several class of graphs, such as fans fn ∼

= Pn +K1 with

n≤6, ladders Ln ∼
= Pn × P2 for odd n, and the generalized prism G ∼
=

Cm ×Pn for odd m and n ≥ 2, are super edge-magic. They also studied

1
the relationship between super edge-magic labeling with other labelings.

However, a conjecture ”Every tree is super edge-magic total” proposed by

Enomoto et al. [1] still remains open.

In this paper, we study super edge-magic total labelings. We derive

more necessary conditions to be able to know more deeply the property of

such labelings. Based on this condition we give some algorithms to con-

struct a new super edge-magic labeling from some old ones. By using these

algorithms we can provide more evidence to support the correctness of the

conjecture proposed by Enomoto et al.

Laterly, new life has been injected into the subject of edge-magic la-

belings of graphs through a paper by Ringel and Lladó [10]; the study of

which originated in 1970 in two papres by Kotzig and Rosa [7,8]. This has

led naturally to the definition of a particular type of edge-magic labelings,

called super edge-magic labelings, introduced by Enomoto, Lladó, Nakami-

gawa and Ringel [2]. These are interesting since relationships between super

edge-magic labelings and previously well studied labelings have been found

by the authors [3], e.g., evey super edge-magic (p,q) graph is cordial and

whenever it is a tree or q≥p it is harmonious and sequential as well. Thus,

the construction of classes of super edge-magic graphs enlarges the collec-

tion of graphs that are, for instance, known to be harmonious. Also,since

there are few graphs that were previously shown to be (super) edge-magic

( such as caterpillars and cycles), we have decided to initially increase our

knowledge of the problem by investigation graphs that are somehow related

2
to these. Another motivation is to develop the theoretical tools necessary to

deal with questions that pertain to ( super) edge-magic labelings of graphs.

Indeed, the obvious proof technique to show that a particular class of graph

is (super) edge-magic is to provide a labeling of it; however, providing that a

graph that satisfies the elementary necessary conditions is not (super) edge-

magic can be far harder. This accumulation of knowledge is, of course, done

with the hope that we may eventually assault the central question of the

topic, that is, are all trees (super) edge-magic?

In this paper, we present several classes of (super) edge-magic graphs

(connected and disconnected); noting that, in the past, it has been difficult

to obtain classes of disconnected graphs that are cordial or sequential.

In order to formalize this presentation, we introduce some necessary

definitions and refer the reader ro Chartrand and Lesniak [1] or Hartsfield

and Ringel [6] for all other terms and notation not provided in this paper.

For a (p,q) graphs G,a bijective function

f : V(G) ∪ E(G)−→ {1, 2, .., p + q}

is an edge-magic labeling of G if f (u) + f (v) + f (uv) = k is a constant,

which is independent on the choice of any edges uv of G. If such a labeling

exists, then k is called the valence of the labeling and G is said to be an

edge-magic graph. Furthermore, f is a super edge-magic labeling if f(V(G))=

{1, 2.., p}. Thus, a super edge-magic graph is a graph that admits a super

edge-magic labeling.

We will find the following basic results useful; see [3].

3
Chapter 1

preliminaries

Definition 1.1. A graph G consists of a pair ( V(G),X(G)), where V(G) is

non-empty finite set whose elements are called Points or vertices and X(G)

is another set of unordered pairs of distinct elements of V(G). The elements

of X(G) are called lines or edges of the graph.

If X= u,v X then the line x is said to join of u and v. The points u

and v are said to adjecent if x=uv. We say that the points u and the line x

are incident with each other.

If two distinct lines x and y are incident with a common point then they

are called adjecent lines.

A graph with p points and q lines is called a ( p,q ) graph.

Definition 1.2. A graph H is a subgraph of G (H⊆G) if V(H)⊆V(G),

E(H)⊆E(G) and ψH is the restriction of ψG to E(G) when H⊆G but H6=G,

we write H⊂G.

(ie)H is proper subgraph of G.

Definition 1.3. A spanning subgraph of G is subgraph H with V(H)=V(G).

(i.e) The vertices set of H and G are same.

4
Definition 1.4. Two graph G and H are said to be isomorphic if there

are bijection θ : V (G) →V(H) and φ : E(G) →E(H) such that ψG (e)=uv

iff ψH (φ(e))=θ(u)θ(v) such a pair (θ, φ) of mapping is called isomorphism

between G and H.

Definition 1.5. A finite graph G=(V,E) is said to be trivial if it has only

one vertex. Since such a graph has only one vertex, it does not have any

edges.

Definition 1.6. Let G=(V,E) be a graph. The graph G is non-trivial if it

contains at least one edge,i.e., E6=∅ Equivalently, G is non-trivial if G is

not an empty graph.

Definition 1.7. For any graph G, We define

δ(G)=min{d(v) /vV(G)} and

∆(G)=max{d(v)/vV(G)}. If all points of G have the same degree r

then G is called a regular graph of degree r. Hence, in a regular graph

δ(G)=∆(G).

Definition 1.8. A finite graph G consists of a finite set of vertices, V and

a finite set E of pairs of vertices from V, called edges.This is known as finite

graph. Notation: G={V,E}.

Definition 1.9. A graph having no self loops and no parallel edges it is

called as a simple graph.

Definition 1.10. A graph G is called a bigraph or bipartite graph if the

vertex set V can be partitioned into two disjoint subsets V1 and V2 such

5
that evey line of G joins a point of V1 to a point of V2 . (V1 ,V2 ) is called a

bipartition of G.

Definition 1.11. A graph G is called a complete bipartite graph if the vertex

set V can be portioned into two disjoint subsets V1 and V2 such that every

line joining the points of V1 to the points of V2 . If V1 contains m points V2

contains n points then the complete bigraph is denoted by Km,n .

The complete graph K1,n is called a star for n≥1.

Definition 1.12. A graph is said to be planar if it can be drawn on a plane

without intersecting edges.

A graph is said to be non-planar if it is not planar.

Definition 1.13. A graph is called maximal planar if no line can be added

toit without losing planarity.

Definition 1.14. A planar graph is called outer planar if it can be embadded

in the plane so that all its vertices lie on the same face. This face is often

chosen to be the exterior face.

Definition 1.15. The outer planar graphs is called maximal outer planar

if no line can be added without losing outer planarity.

Every maximal outer planar graph is a triangularisation of a polygon.

But, every maximal plane graph is a triangularisation of the sphere.

In the otherwords, A maximal outer planar graph is an outerplanar graph

that canot have any additional edges added to it while preserving outerpla-

narity.

6
 Every outer planar graph with n vertices has exactly 2n-3 edges, and

every bounded face of a maximal outerplanar graph is a triangle.

Definition 1.16. A section of a walk W=v0 e1 , v1 e2 , ...ek vk is a walk that is

a subsequence vi ei , vi+1 ei+1 , ...ej vj of consecutive terms of W. It is denoted

by (Vi , Vj )section of W.

Definition 1.17. A walk W is said to be trail if the edgese1 , e2 , ..ek of a

walk W are distinct.

The length of the walk W is denoted by ε(W).

Definition 1.18. A walk W is said to be a path if the vertices v0 , v1 , v2 , ..vk

are distinct.

The length of a path in which the number of lines in the path.

Definition 1.19. A closed walk in which no point except the terminal point

appear more than once is called a cycle.

A closed walk v0 ,v1 , v2 ,..,vn = v0 in which n≥3 and v0 ,v1 ,v2 ,...,vn−1 are

distinct is called a cycle of length n.

The graph consisting of cycle of length n is denoted by Cn .

A graph which contains no cycles is called an acyclic graph.

Definition 1.20. A tree is an undirected graph in which any two vertices are

connected by exactly one path, or equivalently a connected acylic undirected

graph.

Definition 1.21. A cycle in graph that contains all the vertices of the graph

would be called a spanning cycle.

7
Definition 1.22. A spanning cycle in graph is called a Hamiltonian cycle.

A graph having a Hamiltonian cycle is called a Hamiltonian graph.

Definition 1.23. A two regular graph is a regular graph for which all local

degrees are 2.

A two-regular graph consists of one or more(disconnected) cycles.

Definition 1.24. Let G be a graph with p vertices and q edges. A bijec-

tion f from V(G)∪E(G) to {l,2,..p+q} is called an edge-magic labeling of

G if there exists a constant s ( called the magic number of f ) such that

f(u)+f(v)+f(uv)=s for any edge uv of G.

Definition 1.25. An edge-magic labeling f is called super edge-magic if

f(V(G))={1,2,..,p} and f(E(G))={p+1,..,p+q}. A graph G is called edge-

magic (resp.super edge-magic) if there exists an edge-magic(resp.super edge-

magic) labeling of G.

Definition 1.26. A bijection β:V ∪ E→{1, 2..n+e} is called an (a,d)-Edge

Antimagic Total Labeling of G=G(V,E) if the set of the edge weights of all

the edges in G is equal to { a,a+d,...a+(e-1)d}, for two integers a>0 and

d≥0.

Definition 1.27. In both magic and antimagic labelings, we consider the

sum of all labels associated with a graph element. This will be called the

Weight of the element.

Definition 1.28. Given a graph G=(V,E), a vertex labelling is a function

of V to a set of labels; a graph with such a function defined is called a Vertex

8
labeled graph.

Definition 1.29. A bijection α:E→ {1,2..e} is called an (a,d)-vertex an-

timagic edge labeling.

Definition 1.30. A variation of (a,d) vertex magic total labeling that is

called’Super vertex magic total labeling’.

If α(V)={1,2..n} and α(E)={n+1,n+2,..,n+e} then the vertex magic

total labeling is called a super vertex magic total labeling.

9
Chapter 2

Duality in super edge-magic

labeling

2.0.1 2.Some further necessary condition

Several necessary conditions for a graph to be super edge-magic have been

derived by several authors. Enomoto et al.[1] showed that if a nontrivial

graph G is super edge-magic then |E(G)| ≤ 2|V (G)| - 3. Furthermore,

Figueroa-Centeno et al.[3] provide a neccessary and sufficient condition for

a graph being super edge-magic as in the following lemma.

Lemma 1. A (p,q)-graph G is super edge-magic if and only if there exists

a bijective function f : V(G) → {1, 2, .., p} such that the set

S={f(u) + f(v) : uv ∈ E(G)}

consists of q consecutive integers. In such a case, f extends to a super edge-

magic labeling of G with the magic constant k=p+q+s, where s=min(S)

and

+S ={f (u) + f (v) : uv ∈ E(G)}

={k − (p + 1), k − (p + 2), ..., k − (p + q)}

10
Further, in order to know what values of k 0 s for graph G to be super, we

add the following neccessary conditions.

Lemma 2. Let a (p,q)-graph G be super edge-magic total. Then, the magic

constant k of G satisfies p + q + 3 ≤ k ≤ 3p.

Proof. Since G is super edge-magic total then the vertices of G receive

labels 1,2,..,p and the edges receive p+1,p+2,..,p+q so that by Lemma 1

S ={f(u)+f(v): uv∈ E(G)} consists of consecutive integers a, a+1,...,a+q-1

for some positive integer a.

The smallest possible magic constant of G obtained if a=3. In this case

the vertices of G with labels 1 and 2 are adjecent and the magic constant

for this case must be k=(a+q-1)+(p+1)=p+q+3.

If the vertices of labels p-1 and p are adjecent in G then we obtain the

biggest possible magic constant of G, namely k =(p-1)+ p +(p+1) = 3p.

Therefore we obtain p + q + 3 ≤ k ≤ 3p.

The lower and upper bounds in Lemma 2 are tight, since the super edge-

magic labelings λ1 andλ2 on a star Sn of n vertices shown in Fig.1 have the

magic constant 2n+2 and 3n, respectively.

Corollary 2.1. If k is the magic constant of a tree with p vertices then 2p

+ 2 ≤ k ≤ 3p. Furthermore, the magic constant of a (p,q)-graph G with c

components ranges between 2p - c + 3 to 3p.

Proof. The first statement holds, since in any tree the number of edges is

one less than the number of vertices. If a (p,q)-graph G has c components

11
then in each component Gi (i=1,2,..,c) we have E(Gi ) ≥ V (Gi ) − 1. Thus,

|E(G)| ≥ |V (G)|-c and by Lemma 2 it implies that 2p - c + 3 ≤ 3p.

2.0.2 Duality in Super edge-magic labeling

Given any edge-magic total labeling λ on a (p,q)-graph G, Wallis et al.[6]

define the dual labeling λ0 of labeling λ as follows.

λ0 (vi )=M -λ(vi ), ∀vi ∈ V(G),and

λ0 (x)=M -λ(x),∀ x∈ E(G),

Where M=p+q+1.

It is easy to see that if λ is edge-magic total with the magic constant k

then λ0 is edge-magic total with the magic constant k 0 = 3M - k

It is also easy to see that if λ is super edge-magic total then λ0 is no

longer super edge-magic total.

In the next theorem, we introduce another dual property which preserve

the superness of edge-magic total labelings.

Theorem 2.2. Let a (p,q)-graph G be super edge-magic total. Let λ be a

super edge-magic total labeling of G with the magic constant k. Then. the

labeling λ0 defined:

λ0 (vi ) = p + 1- λ(vi ), ∀ vi ∈V(G),and

λ0 (x) = 2p + q + 1 - λ(x), ∀ x ∈ E(G)

is a super edge-magic total labeling with the magic constant k 0 =4p+q+3-

k.

Proof. Let uv E(G). Then, λ0 (u) + λ0 (uv) + λ0 (v)

12
 =(p+1-λ(u)) + (2p + q + 1 -λ(uv)) + (p + 1 - λ(v))

= 4p + q + 3 - (λ(u)+λ(uv)+λ(v))

= 4p + q + 3 - k a constant.

Therefore, λ0 is a super edge-magic total labeling of G with magic con-

stant k 0 =4p + q + 3 - k.

The labeling λ0 in Theorem 1 is called the dual super labeling of λ on

G.

Lemma 3. A (p,q) graph G is super edge-magic if and only if there exist a

bijective function f:V(G) →{1,2..,p} such that the set

S ={f(u)+f(v): uv ∈ E(G)} consists of q consecutive integers. In such

a case, f extends to a super edge-magic labeling of G with valence k= p + q

+ s, where s=min(S) and

S={f(u) + f(v) : uv ∈ E(G)}

={k - (p+1), k-(p+2),...,k-(p+q)}.

In light of this result, it suffices to exhibit the vertex labeling of a super

edge-magic graph. However, we will also provide the valences to increase

the clarity of our results.

The next result is particular useful in showing that a regular graph is not

super edge-magic; see[3].

Lemma 4. Let G be an r-regular super edge-magic (p,q) graph, where r≥1,

then q is odd and the valence is (4p + q + 3)/2.

In the following lemma, Enomoto, Lladó, Nakamigawa and Ringel[2]

provide an upper bound for the size of super edge-magic graphs.

13
Lemma 5. If a (p,q) graph is super edge-magic, then q ≤ 2p - 3.

Proof. This lemma together with the first Theorem of Graph Theory implies

that the minimum degree is at most 3 for every super edge-magic graph.

Now, observe that given an edge-magic labeling f of a (p,q) graph, it is

always possible to find a complementary edge-magic labeling f¯ such that

f¯(x) = p + q + 1 - f(x) for every x∈ V(G)∪ E(G); see[7].

Notice that this operation does not preserve super edge-magic labeling

unless G∼
= K̄n .

The following two lemma provide a simple but often powerful method

to find new edge-magic graphs from known edge-magic graphs.

Lemma 6. Let G be an edge-magic graph, f an edge-magic labeling of G,

and u,v ∈ V(G) such that f(u) + f(v) = k, where k is the valence of f, then

G + uv is edge-magic.

Proof. Notice that if f(u) + f(v) = k, then uv∈

/ E(G); for otherwise f(uv)

= 0. Therefore, we immediately obtain an edge-magic labeling g of G + uv

by letting g(x) = f(x) + 1 for every x∈ (V(G)∪E(G)) - {uv} and g(uv) =

1.

Lemma 7. cannot be applied if f is a super edge-magic labeling of a con-

nected graph G. To see why, let G be a super edge-magic graph with a super

edge-magic labeling f. By Lemma 1.1, the valence of f is

k = p + q + min(f(u) + f(v) : uv ∈ E(G)).

14
 Thus, k ≥ p + q + 1 + 2 ≥ 2p + 2 since G is connected so that q ≥

p-1.

Now,

max(f(u) + f(v) : uv ∈ E(G)) ≤ p + (p-1) = 2p - 1

since f is a super edge-magic labeling. Therefore,

f(u) + f(v) ≤ 2p -1 < 2p +2 ≤ k.

Where Lemma 1.4 concerns the addition of an edge, our next lemma involves

the deletion of an edge.

Lemma 8. If G is an edge-magic graph and f is an edge-magic labeling of G

for which there exists e ∈ E(G) such that f(e) = 1, then G - e is edge-magic.

Proof. We immediately in obtain an edge-magic labeling g of G - e by letting

g(x) = f(x) - 1 for every x∈ (V(G)∪ E(G)) - {e}.

2.0.3 RESULTS ON PATH-RELATED GRAPHS

Paths were shown to be super edge-magic by Ringel and Lladó[10]. Thus, in

this section, it is natural to explore some infinite classes of graphs obtained

from paths.

The following result is of interest because it shown how taking the k-th

power of a super edge-magic graph may or may not imply that the resulting

graph is super edge-magic. Observe that the graphs P22 ∼

= P2 ∼
= K2 and P32

∼
= K3 are clearly super edge-magic.

15
Theorem 2.3. For every integer n≥ 4, the graph G ∼
= Pnk is super edge-

magic if and only if k = 1 or 2.

Proof. The order of Pn3 is n and its size is 3n - 6 whenever n≥ 4; so, by

Lemma 1.3, Pnk is not super edge-magic for k≥3.

For the converse, let G be the graph defined as follows:

V(G) = {vi : i = 1,2..,n}

and

E(G) = {vi vi+1 :i = 1,2..,n - 1} ∪ {vi vi+2 : i = 1, 2.., n − 2}.

Then consider the vertex labeling f : V(G) → {1,2,..,n} such that f(vi )

= i. Consequently, f(vi ) + f(vi+1 ) = 2i + 1 for each i with 1≤ i ≤ n-1 and

f(vi ) + f(vi+2 )= 2i + 2 for every i with 1≤ i ≤ n - 2. Thus,

{f(u) + f(v): uv ∈ E(G)} = {3,4,..,2n -1}

is a set of 2n - 3 consecutive integers.

Therefore, it follows that f is a super edge-magic labeling of G with

valence 3n by Lemma 1.1, obtaining the desired result.

The following results is due to Kotzig and Rosa [7].

Theorem 2.4. The group nP2 is super edge-magic if n is odd. Conversely,

if nP2 is edge-magic, then n is odd.

Proof. The next corollary provides a necessary condition for a given labeling

of the vertices and edges of a graph to be edge-magic.

16
Corollary 2.5. Let G be a graph, and assume that H is a 1 - regular sub-

graph of G of even size. Then a labeling of G, where the vertices and edges

of H are labeled with consecutive integers, is never an edge-magic labeling.

Proof. Assume, to the contrary, that there exists an edge-magic labeling of

G and a 1 - regular subgraph H of G of even size whose vertices and edges

are labeled with consecutive integers. Now, let s be the smallest integer

assigned to any vertex or edges of H. Subtract s - 1 for each of the labels

of H. Then this produces an edges-magic labeling of nP2 2, where n is even,

contradicting the previous theorem.

The next theorem partially extends Theorem 2.2 in terms of linear forests

(forests whose components are paths)

Theorem 2.6. Let F ∼

Sl
= i=1 Pni , where ni is a positive integer for all values

of i, be a super edge-magic linear forest. Then mF is super edge-magic if m

is odd.

Proof. For m = 1, the result is trivial, so we assume that m≥3.

Let

Sl
V(F) = i=1 {vi,j : 1≤ j≤ ni }

and

Sl
E(F) = i=1 {vi,j vi,j+i : 1 ≤ j ≤ ni − 1},

and suppose then that f : V(F) → {1,2....., li=1 } is a vertex labeling

P

that extend to a super edge-magic labeling of F with valence k.

17
 Now, let mF be the linear forest with
Sm Sl t
V(mF) = t=1 i=1 {vi,j : 1≤ j ≤ ni } and
Sm Sl t t
E(mF) = t=1 i=1 {vi,j vi,j+1 : 1 ≤ j ≤ ni - 1}.

Then consider the vertex labeling g : V(mF) → {1,2,...,m li=1 ni } such

P

that

mf (vi,j ) − m + t,
 ifjisevenand1 ≤ t ≤ m;
t 1−m
g(vi,j ) = mf (vi,j ) + 2 + t, ifjisoddand1 ≤ t ≤ m−1
2
;
 1−3m m+1
mf (vi,j ) + 2 + t, ifisoddand 2 ≤ t ≤ m.


Finally, notice that g extends to a super edge-magic labeling of mF with

valence mk + 3(1-m)/2.

The converse of the previous theorem is not true. For example, the

linear forest 2P4 is super edge-magic(label consecutively the vertices of one

path 1 ,7, 2 and 5, and the ones of the other 3, 8, 4 and 6 to obtain a super

edge-magic labeling of 2P4 with valence 21).

The Previous theorem makes it worthwhile to investigate classes of super

edge-magic linear forests.

It is interesting, in light of Kotzig and Rosa’s result for nP2 , to point

out that the next result holds for all positive integers n.

Theorem 2.7. The linear forest F ∼

= P3 ∪ nPn is super edge-magic for

every positive integer n.

Proof. Let F be the linear forest with

V(F) = {x,y,z }∪ {ui vi : 1≤ i ≤ n}

and

18
 E(F) = {xy,yz } ∪ {ui vi : 1≤i≤n}.

We now consider three possible cases.

Case 1 : Suppose that n=1, and let f : V(F) → {1,2,...,5} be the vertex

labeling of F such that f(x) =2, f(y) = 3, f(z) = 4, f(u1 ) = 1 and f(v1 ) = 5.

Then f extends to a super edge-magic labeling of F with valence 13.

Case 2 : Suppose that n = 2m + 1, where m is a positive integer, and

let f : V(F) → {1,2,...,2n+3} be the vertex labeling of F defined as follows:




 3m + 3, if w =x ;

2m + 3,

 if w =y;

m + 3, if w =z ;



f (w) = i, if w =ui and 1 ≤ i ≤ m+2;

i + 3m + 3, if w =vi and 1 ≤ i ≤ m+2;





i + 1,


 if w =ui and m+3 ≤ i ≤ 2m+1;

i + m + 1, if w =vi and m+3 ≤ i ≤ 2m+1.
Then f extends to a super edge-magic labeling of F whose valence is

(9n+17)/2.

Case 3: Suppose that n=2m, where m is a positive integer, and let f:

v(F) → {1,2..,2n+3} be the vertex labeling of F obtained as follows,




 2m + 2, if w =x ;



 m + 1, if w =y;

2m + 3, ifw =z ;



f (w) = i, if w =ui and 1 ≤ i ≤ m;

i + 3m + 3, if w =vi and 1 ≤ i ≤ m;







 i + 1, if w =ui and m+1 ≤ i ≤ 2m;

i + m + 3, if w =vi and m+1 ≤ i ≤ 2m.
Then f extend t a super edge-magic labeling of F whose valence is (

9n+16)/2.

Therefore, we conclude that F is super edge-magic.

19
 Another infinite class of linear forest is shown to be super edge-magic in

the following theorem.

Theorem 2.8. The linear forest F ∼

= P2 ∪ Pn is super edge-magic for every

integer n ≥ 3.

Proof. Let F ∼
= P2 ∪ Pn be the linear forest with

V(F) = {u1 ,u2 } ∪ {vi : 1 ≤ i ≤ n}

and

E(F) = {u1 u2 } ∪ {vi vi+1 : 1 ≤ i ≤ n-1},

and the consider a vertex labeling f:V(F) → {1,2...,n+2}.

We now proceed by cases. Case 1: Assume that n ≡ 0 (mod 4), and let

f(u1 )=1 ; f(u2 )= 12 n+3 ;

and

n

 2
+ 2, if j=1;
n

+ 4, if j=3;



 2
j=4i and 1 ≤ i ≤ n4 ;

2i, if
f (vj ) = n


 2
+ 2i + 4, if j=4i+1 and 1 ≤ i ≤ n−4 4
;
j=4i+2 and 0 ≤ i ≤ n−4



 2i + 3, if 4
;

n n−4
2
+ 2i + 3, if j=4i+3 and 1≤ i ≤ 4 .
Then f extends to a super edge-magic labeling of F with valence (5n+12)/2.

Case 2: Assume that n ≡ 1 (mod4),and let f(u1 )=1; f(u2 )=n+2; and


2j+n+5
 4 ,
 if j is odd;
f (vj ) = 2j+3n+5
4
, if j is even and 2 ≤ j ≤ n−1
2
;
 4j−n+5
 n+3
4
, if j is even and 2 ≤ j ≤ n-1.
Then f extend to a super edge-magic labeling of F with valence (5n+11)/2.

Case 2: Assume that n ≡ 2 (mod4), and let f(u1 )=1; f(u2 )= 21 n + 2;

and

20
 
n + 2,

 if j=1;



 n, if j=3;

n + 1, if j=n;



f (vj ) = n2 − 2i + 2, if j=4i and 1 ≤ i ≤ n−2
4
;
 n−2
n − 2i, if j=4i+1 and 1 ≤ i ≤ 4 ;




n
j=4i+2 and 0 ≤ i ≤ n−6




 2
− 2i − 1, if 4
;
n−6

n − 2i + 1, if j=4i+3 and 1 ≤ i ≤ 4 .
Then f extends to a super edge-magic labeling of F with valence (5n+10)/2.

Case 4: Assume that n ≡ 3 (mod4), and let f(u1 )=1 ; f(u2 )= n+2 ; and

 j+3

 2
, if j is odd and 1 ≤ j ≤ n−1
2
;
 j+n+2 , if j

is n+3
odd and 2 ≤ j ≤ n;
2
f (vj ) = j+n+3
 2 , if j

 is even and 2 ≤ j ≤ n−3
2
;
 j+4 n+1
2
, if j is even and 2 ≤ j ≤ n-1.
Then f extends to a super edge-magic labeling of F with valence (5n+11)/2.

Therefore, we conclude that F is super edge-magic.

21
Chapter 3

Construction of new labelings

In this section, we give algorithms to construct new super edge-magic graphs

by extending the old ones.

Theorem 3.1. From any super edge-magic (p,q)-graph G with the magic

constant k, we can construct a new super edge-magic total graph from G by

adding one pendant incident to vertex x of G whose label l - 2p -1. The

magic constant of the new graph is k 0 = k + 2.

Proof. In the new graph, define a labeling in the following.Preserve all ver-

tices labels of G.

Increase the labels of all edges ( except the new one ) by 2.

Label the new vertex and edge by p + 1 and p + 2 respectively.

It can be verified that the resulting labeling on the new graph is super

edge-magic total labeling with magic constant k = k + 2.

Since 2p + 2 ≤ k ≤ 3p ( by Lemma 2 ), the proposition holds for any

value of k.

Theorem 3.2. Let a ( p,q )-graph G be super edge-magic total with the

magic constant k and k ≥ 2p + 3. Then, a new graph formed from G ny

22
adding exactly two pendants incident to two distinct vertices x any y of G

whose labels k - 2p and k - 2p - 2 respectively is super edge-magic total with

the magic constant k 0 = k + 4.

Proof. In the new graph denoted by u and v the new vertices adjacent to

x and y, respectively. Then, define a labeling in the new graph as follows.

Preserve labels of all the vertices of G.

Add all edge labels ( except the new ones ) by 4. Label vertices x and

y by p + 1 and p + 2, respectively and label two new edges xu and yu by

p + 3 and p +4, respectively.

For the new edges, clearly we have the edge sum of each is k + 4.

Since each label of old edge increased by 4 then we get the edge sum of

each old edge is also k + 4.

Therefore, the new graph is super edge-magic total labeling.

This process works only if k - 2p - 2 ≥ 1.

This implies that k ≥ 2p + 3.

Theorem 3.3. Let a ( p, q )-graph G be super edge-magic total with the

magic constant k and k ≥ 2p + 3. Then, a new graph formed from G by

adding exactly three pendants incident to three distinct vertices x, y and z of

G whose labels k - 2p, k-2p -1 and k- 2p -2 respectively is super edge-magic

total with the magic constant k 0 = k + 6.

Proof. In the new graph, define a labeling as follows. Preserve all vertex

labels of G in the new graph.

Increase all edge labels ( except the new ones ) by 6 in the new graph.

23
 Label the three new vertices u which adjacent to x,y and z by using the

second row of either matrix A or B.

Labelthe corresponding new edge

 e byusing the third row fromA or B.
x y z x y z
A = u : p + 1 p + 3 p + 2, B = u : p + 2 p + 1 p + 3
e:p+5 p+4 p+6 e:p+4 p+6 p+5
For the new edges, clearly we have the edge sum of each is k + 6 (from

the above matrix ).

Since each label of old edge increased by 6 then we get the edge sum of

each old edge in the new graph is k + 6.

Therefore, the new graph is super edge-magic total. Note that this

process works only if k -2p- 2 ≥ 1.

This implies that k ≥ 2p + 3.

Alternatively, we have the following theorem for adding three pendants.

Theorem 3.4. Let a (p,q)-graph G be super edge-magic total with the magic

constant k and k ≥ 2p + 4. Then, a new graph formed from G by adding

exactly three pendants incident to three distinct vertices x , y and z of G

whose labels k -2p +1, k -2 -1 and k -2p -3 respectively is super edge-magic

total with the magic constant k 0 = k + 6.

Proof. Take Proof is similar with the one of Theorem 4 by using the follow-

ing matrix
 C. 
x y z
C = u : p + 1 p + 2 p + 3
e:p+4 p+5 p+6
Theorem 3.5. Let p be an odd integer. Let a (p,q)-graph G be super edge-

magic total with the magic constant k = (5p+3)/2. Then, a new graph

24
formed from G by adding exactly p pendants incident to all vertices of G is

also super edge-magic total with the magic constant k = (9p+3)/2.

Proof. In the new graph, define a labeling as follows. Preserve all vertex

labels of G in the new graph.

Increase all edge labels ( except the new ones) by 2p in the new graph.

Label each new vertex u which adjacent to the old vertex v by using the

second row of the following matrix.

Label the corresponding new edge e= vu by using the third row. 

p−1 p+1 p+3

v:1 2 ... 2 2 2
... p − 1 p
 u : 3p+3 3p+5
... 2p p + 1 p + 2 ... 3p−1 3p+1 
2 2 2 2
e : 3p − 1 3p − 3 ... 2p + 2 3p 3p − 2 ... 2p + 3 2p + 1
For the new edges, clearly we have the edge sum of each is 9p+3
2
( from

the above matrix ). Since each label of old edge increased by 2p then we

get the edge sum of each old edge in the new graph is k+2p= 9p+3
2
.

Therefore, the new graph is super edge-magic total.

Note that the ( extension ) construction method in Theorem 6 only works

for a super edge-magic total graph with the magic constant k = (5p+3)/2.

There are several graphs of p vertices known to have the magic constant

(5p+3)/2, such as odd cycles and paths with odd number of vertices.

Let C(n,s) be a graph constructed from path Pn of n vertices by adding

s pendants to each vertex of Pn . Let us call C(n,s) by a caterpillar with s

legs.

We know that C(n,s) is super edge-magic total[5]. The following corol-

lary shows one way how to label this graph so that super edge-magic to-

tal.

25
Corollary 3.6. For odd n and s≥ 1, the graph C(n,s) is super edge-magic

total.

Proof. Take a super edge-magic total labeling for path Pn , for odd n, with

the magic constant (5n+3)/2, namely label the vertices in the odd positions

of Pn from left to right consecutively by 1,2,..,(n+1)/2; And then label the

even position from left to right consecutively by (n+3)/2 , (n+5)/2,...,n ;

Next, it is easy to label all the edges of Pn so that we have a super

edge-magic total labeling.

Apply Theorem 6 to Pn . Denote the resulting graph by C(n,1).

For i=1,2,..,s-1 , apply Theorem 6 repeatedly to graph C(n,i) have a

super edge-magic labeling for C(n,i+1). In the final result, we have a super

edge-magic labeling for C(n,s).

Let Tp be a tree of p vertices, for p≥ 3. For h ≥ 1 we denote by Tn +

Ah a graph which is obtained by adding h pendants to one vertex of tree

Tp . Then, we have;

Theorem 3.7. From super edge-magic total tree Tp with the magic constant

k = 2p + 1 + s, for some s ∈ {1,2,..,p}, we can constant a new super edge-

magic total tree Tp + Ah .

Proof. Apply Theorem 2 to Tp h times by attaching a new pendant each to

vertex whose label k - 2p - 1.

By using Theorem 7 we can obtain a class of trees which is super edge-

magic total from just one super edge-magic total tree.

26
 For instances, the tree in Fig. 2(b) is obtained by applying Theorem 7

to the tree in (a).

This theorem provides more facts to support the correctness of the Con-

jecture proposed by Enomoto et al [1].

3.0.1 RESULTS ON STAR-RELATED GRAPHS

In [2], it was shown that the completely bipartite graph Km, n is super

edge-magic if and only if m = 1 or n =1. Then next theorem partially

extends their result by determining all edge-magic and super edge-magic

labeling of the star K1,n .

Theorem 3.8. Every star K1,n is super edge-magic. Moreover, there are

exactly 3.2n distinct edge-magic labeling of K1,n of which only two are super

edge-magic labeling up to isomorphisms.

Proof. First, notice that the order of K1,n is n+1 and its size is n. Next,

define the star G∼

= K1,n as follows;

V(G) = {u} ∪ {vi : 1 ≤ i ≤ n} and E(G) = {ei = uvi : 1 ≤ i ≤ n}.

Assume that there exists an edge-magic labeling f of G, and let k be its

valence. Then
n
P
( (f (vi ) + f (ei ) ) + f(ei )) + nf(u) = nk.
i=1
n
P
Thus, n divides (f (vi ) + f (ei )).
i=1

Now,
n
P
( (f (vi ) + f (ei ))) + f(u) = 1 +...+(2n+1)
i=n

= 2n2 + 3n + 1,

27
 so
n
(f (vi ) + f (ei ))) = 2n2 + 3n + (1-f(u)).
P
(
i=1

Hence, n divides f(u)-1 , but 1 ≤ f(u) ≤ 2n+1, which implies that f(u)

is 1 , n+1 or 2n+1. Since

nk = 2n2 + 3n + (n-1) g(u),

it follows that k = 2n + 4 , 3n + 3 or 4n + 2 l, which correspond to f(u) =

1, n+1,2n+1, respectively.

It suffices now to exhibit labeling with each of the three possible valences,

and then describe how to obtain all of the other labeling from them.

Let f1 , f2 , and f3 be edge-mafgic labeling of G defined as follows:

f1 (u) = 1, f1 (vi ) = i+1, f1 (uvi ) = 2n + 2 - i,

f2 (u) = n+1, f2 (vi ) = i, f2 (uvi ) = 2n + 2 - i,

f3 (u) = 2n+1, f3 (vi ) = i, and f3 (uvi ) = 2n+ 1 - i,

where 1 ≤ i ≤ n. Then the valences of f1 , f2 and f3 be edge-magic

labelings if G can be obtained by permuting the labels of uvi and vi for any

i with 1 ≤ i ≤ n, and that of these 3.2n possible permutations, only f1 and

f2 are super edge-magic labeling of G.

The following corollary is an immediate consequence of the proof of the

preceding theorem.

It is interesting since Godbold and Slater [5] have conjectured that for

sufficiently large cycles, there are no gaps between the possible valences.

Corollary 3.9. For every integer n ≥ 2, there exists a super edge-magic

28
graph G such that |k1 − k2 | ≥ n-1, where k1 and k2 are two possible distinct

valances of G.

Proof. The next corollary describes how new super edge-magic graphs can

be found from known super edge-magic graphs.

Corollary 3.10. For every positive integer n, the graph K2 + K̄n is super

edge-magic.

Proof. Let G∼
= K1,n be defined as in the proof of the previous theorem, and

consider the following edge-magic labeling g of G: g(u) = n + 1, g(vi ) =

2(n+1) - i and g(uvi ) = i for 1 ≤ i ≤ n.

Now, notice that the valence k is 3n +3 and g(v1 ) + g(vi ) = 4n + 3 - i

for 2 ≤ i ≤ n.

Then define the graph H ∼

= K2 + K̄n as follows: V(H) = V(G) and

E(H) = E(G) ∪ {v1 vi : 2 ≤ i ≤ n}; and consider the following edge-magic

labeling og H with valence 6n: f(v) = g(v) + n - 1 for any vertex v of H,

f(uvi ) = g(uvi ) + n - 1 for 1 ≤ i ≤ n and f(v1 vi ) = i - 1 for 2 ≤ i ≤ n.

Finally, observe that f¯ is a super edge-magic labeling of this graph since

f(u) > f(e) for any vertex v and edge e of H.

Notice that the above corollary establishes the sharpness of Lemma 1.3.

We remark that from the preceding proof, we can obtain a sequence of

super edge-magic graphs as follows. Take the labeling f employed for K2 +

K̄n in the proof and then remove the edge labeled 1 from it and decrease

all labels by 1.

Continue in this fashion until arriving to K1,n .

29
 Every such labeling of each graph is edge-magic by Lemma 1.5 and its

complementary labeling is super edge-magic.

The above corollary also allows us to characterize all the super edge-

magic complete m-partite graphs.

Theorem 3.11. The only super edge-magic complete m-partite graphs are

K1,n and K1,1,n , where n≥ 1.

Proof. Recall that Enomoto, Lladó, Nakamigawa and Ringel[2] have already

shown that the star K1,n , n ≥ 1, is the only super edge-magic complete

bipartite graph. Furthermore, the complete 3-partite graph K1,1,n is the

unique super edge-magic complete 3-partite graph and that there are no

complete 4-partite graphs with this property.

For the uniqueness of K1,1,n , Let G ∼

= Kn1 ,n2 ,n3 be a complete 3-partite

graph with n1 ≥ n2 ≥ n3 ≥ 1.

Then, assume, to the contrary, that n2 ≥ 2 and G is super edge-magic.

Now, the order of G is n1 + n2 + n3 and its size is n1 n2 + n1 n3 + n2 n3 ; so,

by Lemma 1.3,

n1 n2 + n1 n3 + n2 n3 ≤ 2n1 + 2n2 + 2n3 − 3,

which, i turn implies that n1 n3 ≤ 2n2 − 3 since n2 ≥ 2 and n3 ≥ 2.

Hence, n2 n3 ≤ 2n2 − 3, so 2 - n3 > 0 from which we conclude that

n3 = 1.

Therefore, if we apply Lemma 1.3 again, we get that n1 ≤ 1, producing

a contradiction.

30
 To show that there are no super edge-magic complete 4-partite graphs,

observe that K1,1,1,n is not super edge-magic by Lemma 1.3 and all remaining

graph have minimum degrees greater that 3, completing the proof.

The next two results show that some classes of galaxies (forests whose

components are stars ) are super edge-magic.

Theorem 3.12. The galaxy G ∼

= K1,n ∪ K1,n+1 , n≥ 1, is super edge-magic.

Proof. Define the galaxy G as follows:

V(G)= {u,v} ∪ {xi : 1 ≤ i ≤ n} ∪ {yi : 1 ≤ i ≤ n + 1}

and

E(G) = {uxi : 1 ≤ i ≤ n} ∪ {vyi : 1 ≤ i ≤ n + 1},

and then consider the vertex labeling f : V(G) → {1,2,..2n+3} such that

f(u) = 1 , f(v) = 3 , f(xi ) = 2i + 3 (1 ≤ i ≤ n), and f(yi ) =2i (1 ≤ i ≤ n−1).

In order to show that f extends to a super edge-magic labeling of G. it

suffices to verify the following by Lemma 1.1 :

(a) f(V(G)) = {1 , 2 ,.., 2n+3}

(b) S = {f(x) + f(y) : xy ∈ E(G)} is a set of 2n+1 consecutive integers.

For (a), observe that f(u) = 1 and f(xn ) = 2n + 3. Also, if i < n, then

f(xi ) < f(xn ) = 2n + 3.

Hence, the maximum possible integer that can be used for a vertex label

is 2n + 3. Now, f(xi ) 6= f(xj ) if and only if i 6= j; and f(u) + f(xi ) is even.

In addition, f(v) + f(yi ) 6= f(v) if and only if i 6= j.

31
 Further, notice that f(xi ) is odd for every i with 1 ≤ i ≤ n and f(yi ) is

even for every i with 1 ≤ i ≤ n+1.

Thus, the set of vertex labels is {1 ,2,..,2n+3}.

For (b), observe first that the minimum element in S is 3 + f(y1 ) = 5

and the maximum element is 3 + f(yn+1 ) = 2n +5. Now, f(u) + f(xi ) 6=

f(v) + f(xj ) if and only if i 6= j; and f(u) + f(xi ) is even.

In addition, Therefore, all elements of S are distinct and |S| is the size

of G.

Finally, notices that the valence of the labeling f is 4n + 9, which com-

pletes the proof.

The following theorem is a partial generalization Theorem 3.1

Theorem 3.13. For positive integer m and n, where m is odd, the galaxy

g∼
= mK1,n is super edge-magic.

Proof. Let G be the galaxy with

V(G) = {ui : 1 ≤ i ≤ m} ∪ {xi,j : 1 ≤ i ≤ m, 1 ≤ j ≤ n}

and

E(G) = {ui xi,j : 1 ≤ i ≤ m, 1 ≤ j ≤ n}.

Then consider the vertex labeling f : V(G)→ {1, 2, ...,m(n+1)} such

that


i, if w = ui for 1≤ i ≤ m;
i + 3m+1 ,

if w = xi,1 for 1 ≤ i ≤ m+1 ;
2 2
f (w) = m+1 m+1


i + 2
, if w = xi,1 for 2 ≤ i ≤ m;
f (xi,1 ) + m(j − 1), =xi,j for 1 ≤ i ≤ m and 2 ≤ j ≤ n.

if w
hence, f extends to a super edge-magic labeling of G with valence 2mn+

2m + 3.

32
3.0.2 RESULTS OF 2-REGULAR GRAPHS

In this section, we consider 2-regular graphs. These are of interest to us

since, in their seminal paper, Kotzig and Rosa[7] pondered on whether one

may find necessary and sufficient conditions to determine if 2-regular gr-

daphs are edge-magic.

Consider the following results by Kotzig and Rosa [7].

Theorem 3.14. Every n-cycle Cn is edge-magic.

Proof. Also, recall the analogous result for super-edge magic graphs by En-

imoto, Lladó, Nakamigawa and Ringel[2].

Theorem 3.15. The n-cycle Cn is super edge-magic if and only if n is odd.

Proof. The following is a natural generalization of Theorem 4.1 and 4.2.

Theorem 3.16. The 2-regular graph G ∼

= mCn is super edge-magic if and

only if m ≥ 1 and n≥ 3 are odd.

Proof. The case where m = 1 has already been handled by Kotzig and

Rosa[7], so we assume that m ≥ 3.

Let G be the 2-regular graph with

V(G)= {vi,j : 1 ≤ i ≤ m, 1 ≤ j ≤ n} and

E(G) = {vi,j vi,j+1 : 1 ≤ i ≤ m, 1 ≤ j ≤ n − 1} ∪ {vi,n vi,1 : 1 ≤ i ≤ m}.

The consider a vertex labeling f : V(G) → {1, 2, ..., mn} such that

33
 

 i, if 1 ≤ i ≤ m and j=1

n j−2 2i+1+m m−1
m([ 2 ] + 2 ) + 2 , if 1 ≤ i≤ 2 and j is even;



f (vi,j ) = m([ n2 ] + j−2
2
) + 2i+1−m
2
, if m+12
≤ i ≤ m and j is even;
j−1

m( 2 + 1) + 1 − 2i, if 1 ≤ i≤ m−1 and j 6= 1 is odd;



 2
m( j−1 + 2) + 1 − 2i,
 m+1
if 2 ≤ i ≤ m and j 6= 1 is odd.
2

Therefore, f extends to a super edge-magic labeling of G with valence

(5n+2)m+3
2
.

The converse follows immediately from Lemma 1.2.

The 2-regular graph 2C4 is edge-magic ( label the vertices of one 4-cycle

clockwise 1,14,9 and 13, and the ones of the other 4,6,12 and 5, and let the

valence be 25 ). Therefore, the previous theorem cannot be strengthened

by considering edge-magic graphs instead of super edge-magic graphs. The

problem determining whether mCn is edge-magic or not when either m or

n is even is open.

3.0.3 RESULTS ON EDGE-ANTIMAGIC GRAPHS

Ringel [9] has also provided the definition for edge-antimagic graphs.

For a (p,q) graph G, a bijection function f : V(G) → {1,2,...,p} is an

edge-antimagic labeling of G if

|{f (u) + f (v) : uv ∈ E(G)}| = q

If such a labeling exists, then G is called an edge-antimagic graphs.

In this section, we present some relationship between super edge-magic

graphs and edge-antimagic grpahs.

The following is an immediate consequence of Lemma 1.1.

34
Theorem 3.17. Every super edge-magic graph is edge-antimagic.

Proof. We then note that Lemma 1.3 follows from Theorem 5.1 and a com-

ment by Ringel [9] to the effect that the inequality q ≤ 2p - 3 holds for

edge-antimagic (p,q) graphs.

Ringle [9] also mentioned that if a graph G of order p is edge-antimagic

with an edge-antimagic labeling f, then

{f(u) + f(v) : uv ∈ E(G)} ⊆ {3,4,...,2p-1}.

This remark implies the following partial converse of Theorem 5.1.

Theorem 3.18. If G is an edge-antimagic (p,q) graph with q = 2p - 3, then

G is super edge-magic.

Proof. Let G be an edge-antimagic (p,q) gdraph such that q = 2p - 3 with

an edge-antimagic labeling f. Then

{f(u) + f(v) : uv ∈ E(G)} + {3,4,...,2p - 1}

so the result follows from Lemma 1.2

Theorem 3.19. If G is a maximal outerplanar graph of order p with exactly

two vertices a, b of degree 2 and whose distance dH (a, b) on the Hamilton

cycle H in G is

[ p2 ] or [ p2 ] - 1,

then G is edge-antimagic.

Since all maximal outerplanar (p,q) graphs satisfy q = 2p - 3, we have

the following result from Theorems 5.2 and 5.3.

Corollary 3.20. If G is a maximal outerplanar graph of order p with exactly

35
two vertices a, b of degree 2 and whose distance dH (a, b) on the Hamilton

cycle H is G is

[ p2 ] or [ p2 - 1, then G is super edge-magic.

The previous corollary implies that the upper bound in Lemma 1.3 is

also sharp for maximal outerplanar graphs.

36