Magic Graph PDF
Magic Graph PDF
A.A.G. Ngurah
1
, E.T. Baskoro
1,3
, I. Tomescu
2,3
1
Combinatorial Mathematics Research Group
Faculty of Mathematics and Natural Science, Institut Teknologi Bandung
Jalan Ganesa 10 Bandung, Indonesia.
Email: {s304agung, ebaskoro}@dns.math.itb.ac.id
2
Faculty of Mathematics and Computer Science, University of Bucharest
Str. Academia, 14, 010014 Bucharest, Romania.
Email: [email protected]
3
School of Mathematical Sciences, GC University
68-B, New Muslim Town, Lahore, Pakistan.
Abstract
A graph G is edge-magic if there exists a bijection f from V (G)
E(G) to {1, 2, 3, , |V (G)| + |E(G)|} such that for any edge uv of
G, f(u) +f(uv) +f(v) is constant. Moreover, G is super edge-magic
if V (G) receives |V (G)| smallest labels. In this paper, we propose
methods for constructing new (super) edge-magic graphs from some
old ones by adding some new pendant edges.
1 Introduction
We consider nite and simple graphs. The vertex and edge sets of graph
G are denoted by V (G) and E(G), respectively.
Let G be a graph with p vertices and q edges. A bijective function f :
V (G)E(G) {1, 2, 3, , p+q} is called an edge-magic total labeling of G
if there exists an integer k such that f(x) +f(xy) +f(y) = k independent
of the choice of any edge xy of G. If such a labeling exists, then the
constant k is called the magic constant of f, and G is said to be edge-
magic graph. An edge-magic total labeling f is called super edge-magic if
1
Permanent address: Department of Civil Engineering, Universitas Merdeka
Malang, Jalan Taman Agung 1 Malang, Indonesia
1
f(V (G)) = {1, 2, 3, , p}. Thus, a super edge-magic graph is a graph that
admit a super edge-magic total labeling.
The edge-magic concept was rst introduced and studied by Kotzig
and Rosa [12, 13], although under a dierent name, i.e., the magic val-
uation. The super edge-magic notion was rst introduced by Enomoto,
Llad o, Nakamigawa and Ringel [3]. The (super) edge-magic graphs have
been studied in several papers, see for instance [?, 4, ?, 5, 9, 11, ?], and
more complete results on (super) edge-magic graphs can be seen in the
survey paper by Gallian [10]. However, the long-standing conjectures that
every tree is edge-magic and every tree is super edge-magic, proposed
in [12] and [3], respectively, still remain open.
The following lemma presented in [4] gives a necessary and sucient
condition for a graph to be super edge-magic.
Lemma 1 A graph G with p vertices and q edges 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(x) +f(y)|xy E(G)} consists of q consecutive integers. In
such a case, f extends to a super edge-magic total labeling of G with magic
constant k = p +q +s, where s = min(S) and
S = {f(x) +f(y)|xy E(G)}
= {k (p + 1), k (p + 2), , k (p +q)}.
In [12], Kotzig and Rosa introduced the concept of edge-magic deciency
of a graph. They dened the edge-magic deciency, (G), of a graph G as a
minimum nonnegative integer n such that GnK
1
is an edge-magic graph.
Kotzig and Rosa [12] gave an upper bound of the edge-magic deciency of
a graph G with p vertices, that is (G) F
p+2
2 p
1
2
p(p 1), where
F
p
is the p-th Fibonacci number.
Furthermore, Figueroa-Centeno et al.[7] dened the concept of the super
edge-magic deciency of a graph similarly. The super edge-magic deciency,
s
(G), of a graph G is a minimum nonnegative integer n such that GnK
1
has a super edge-magic total labeling or + if there exists no such n.
Clearly, for every graph G, (G)
s
(G).
Figueroa-Centeno et al. in two separate papers [7, 8] provided the exact
values of (super) edge-magic deciency of several classes of graphs, such
as cycles, complete graphs, some classes of forests, 2-regular graphs, and
complete bipartite graphs K
2,m
. They [8] also proposed the conjecture if
F is a forest with two components, then
s
(F) 1.
In this paper, we propose some methods for constructing new (super)
2
edge-magic graphs from the old ones. From this construction we can obtain
new classes of (super) edge-magic graphs. Some of the resulting graphs give
support to the correctness of the conjectures every tree is (super) edge-
magic, and if F is a forest with two components, then
s
(F) 1 .
2 The Results
Throughout this section, we will present a construction of new (super) edge-
magic graphs by adding pendant edges to some (not all) vertices of a (super)
edge-magic graph G which having a specic property. This construction
can be viewed as a weaker version of a corona product of a graph G and
nK
1
.
The corona product G H of two given graphs G and H is dened as
a graph obtained by taking one copy of a p-vertex graph G and p copies
H
1
, H
2
, . . . , H
p
of H, and then joining the i-th vertex of G to every vertex
in H
i
. If H
= nK
1
, G H is equal with adding n pendant edges to every
vertex of G. The corona product of graphs has been studied in several
papers, see for instance [2], [6] and [14].
In the next two theorems, we construct (super) edge-magic graphs by
adding n pendant edges to every vertex of particular type of edge-magic
graph except some vertices with the largest labels.
Theorem 1 Let G be a graph of even order p 2 and size of either q =
p or p 1 for which there exists an edge-magic total labeling f with the
property that all vertices of G receive odd labels such that
{f(x) +f(y)|xy E(G)} = {3p 2q, 3p 2q + 2, , 3p 4, 3p 2}. (1)
Then, the graph H formed by adding n pendant edges to each vertex of
G except the vertex with the largest label is edge-magic for every positive
integer n.
Proof Suppose V (G) = {x
i
|1 i p}. Let f be an edge-magic total
labeling of G satisfying the conditions of Theorem 1. Then, the magic
constant of f is 3p. Since all vertices receive odd labels, we may assume
that f(x
i
) = 2i 1 for every integer 1 i p. Next, let H be a graph
dened as follows.
V (H) = V (G) {y
j
i
|1 i p 1 and 1 j n},
3
and
E(H) = E(G) {x
i
y
j
i
|1 i p 1 and 1 j n}.
Now, dene a total labeling
g : V (H) E(H) {1, 2, 3, , 2n(p 1) +p +q}
such that g(x) = f(x) for every x V (G) and
g(y
j
i
) =
T
a tree formed by applying the algorithm
in the proof of Theorem 1 to P
T
. Similarly, we denote lobsters formed
by applying the same algorithm to P
1
2k,m
, P
2
2k+1,1
and P
3
2k+1,m
by L
1
2k,m
,
L
2
2k+1,1
and L
3
2k+1,m
, respectively.
Therefore, by Theorem 1 we have the following corollary.
Corollary 1 The tree P
T
and the lobsters L
1
2k,m
, L
2
2k+1,1
and L
3
2k+1,m
are
edge-magic graphs.
Now, we refer the readers to the following result.
Theorem 2 [4] Let T be an edge-magic tree of order p with and edge-
magic total labeling f whose magic constant is k such that f(v) is odd for
any vertex v of V(T). Then, the bijective function g : V (T) E(T)
{1, 2, 3, , 2p 1} dened as
g(x) =
f(x)+1
2
, if x V (T),
f(x)
2
+p, if x E(T),
is a super edge-magic labeling. Furthermore, given a super edge-magic la-
beling of a tree, a labeling can be obtained with all vertices receiving an odd
label by reversing the above process.
Note that, Theorem 2 can be extended to graphs for which p = q.
By Theorem 2, all graphs that satisfying the conditions of Theorem 1
also super edge-magic. Especially, we have the following corollary.
Corollary 2 The tree P
T
and the lobsters L
1
2k,m
, L
2
2k+1,1
and L
3
2k+1,m
are
super edge-magic graphs.
These results provides supporting examples of the conjectures proposed
by Kotzig and Rosa [12] and by Enomoto, Llado, Nakamigawa and Ringel
[3].
If the condition all vertices of G receive odd labels in Theorem 1 is
removed, then the conclusion is not true. For example, consider graph G
in Figure 1(a). If G has an edge-magic total labeling satisfying condition
5
(1) then all vertices of G must receive even labels (since the set in (1)
consists of only even numbers and G is connected). Then, there are only
two such labelings possible (see Figure 1(b) and 1(c)). Let H be a graph
formed by adding n pendant edges to every vertex of G except to the
vertex of label 12 (the largest vertex label). Then H is not edge-magic
for any integer n. In fact, if H is edge-magic, then the magic constant
is 10n + 19
12
5n+6
. Therefore, it is not possible for all positive integer n.
Therefore, the condition that all vertices of G receive odd labels in Theorem
1 is crucial.
10
2
6
8
4
12
12
2
6
10
4
8
(a) (b) (c)
Figure 1: The graph G and its vertex labeling.
Theorem 3 Let G be a graph with odd order p ( 3) for which there exists
a super edge-magic total labeling g with the property that
max{g(x) +g(y)|xy E(G)} =
1
2
(3p 1).
Then, the graph H formed by adding n pendant edges to every vertex of
G except the vertices u and v with g(u) = p 1 and g(v) = p is super
edge-magic for every positive integer n.
Proof Let G be a graph of odd order p satisfying the conditions of
Theorem 3, and g be a super edge-magic total labeling of G with the magic
constant k. Assume g(x
i
) = i for every 1 i p, where V (G) = {x
i
|1
i p}.
Now, dene H as a graph with the vertex and edge sets
V (H) = V (G) {y
j
i
: 1 i p 2, 1 j n}
and
E(H) = E(G) {xy
j
i
: 1 i p 2, 1 j n},
respectively.
6
Next, consider the vertex labeling h : V (H) {1, 2, 3, , (n+1)p2n}
dened as follows.
h(x) = g(x), for every x V (G),
and
h(y
j
i
) =
1
2
(2i + 2 p) +j(p 1), for
p
2
+ 1 i p and 1 j n,
1
2
(2i +p) +j(p 1), for 2 i
p
2
and 1 j n.
We can see that labels of pendant vertices are consecutive and at greater
than p. By similar argument used in the proof of Theorem 3, it can be
shown that h extends to a super edge-magic total labeling of H with the
magic constant k + 2n(p 1).
Theorem 5 Let G be a graph of order p = (c + 1)(m + 1) + 1, where
m 2, c 1 for which there exists a super edge-magic total labeling f with
the property that
max{f(x) +f(y)|xy E(G)} = (2m+ 1)(c + 1) + 1.
Then, the graph H formed by adding n pendant edges to every vertex of G
except the vertices with labels 1, 2, 3, , m(c+1)c3 is super edge-magic
for every positive integer n.
Proof Let G be a graph satisfying the conditions of Theorem 4 with
V (G) = {x
i
: 1 i p}. Take a super edge-magic total labeling g with
the magic constant k such that g(x
i
) = i for 1 i p. Now, dene the
graph H as follows:
V (H) = V (G) {y
j
i
: m(c + 1) c 2 i p, 1 j n}
8
and
E(H) = E(G) {x
i
y
j
i
: m(c + 1) c 2 i p, 1 j n}.
It is easy to verify that the vertex labeling h : V (H) {1, 2, 3, , p +
(2c + 5)n} dened:
h(x) = g(x), for every x V (G),
and
h(y
j
i
) =
1, if y = c,
j(n + 1) +
1
2
(n + 4), if y = w
j
,
1
2
(i + 2), if y = u
i
for even i,
1
2
(n +i + 3), if y = u
i
for odd i,
1
2
(2j(n + 1) +i + 3), if y = v
j
i
for odd i and 1 j m,
1
2
(2j(n + 1) +n +i + 4), if y = v
j
i
for even i = n and 1 j m,
j(n + 1) + 1, if y = v
j
i
for i = n and 1 j m,
j(n + 1) +
1
2
(n + 4), if y = w
j
.
It is not dicult to verify that {g(x) + g(y) : xy E(H)} is a set
of consecutive integers starting from
1
2
(3n + 4). By Lemma 1, g extends
to a super edge-magic total labeling of F
2
with the magic constant k =
1
2
(4nm+ 5n + 4m+ 8).
Other results concerning forests with two components see [7].
3 Acknowledgement
The rst author wish to thank Prof. A. D. R. Choudary for his support
and hospitality during his stage at the School of Mathematical Sciences,
G.C. University, Lahore, Pakistan.
References
[1] C. Barientos, Dierence vertex labelings, Ph.D. Thesis, Universitat Po-
litecnica De Catalunya, Spain, 2004.
[2] E.T. Baskoro, I.W. Sudarsana and Y.M. Cholily, How to construct new
super edge-magic graphs from some old ones, J. Indones. Math. Soc.
(MIHMI), Vol. 11, No. 2 (2005), 155 - 162.
[3] H. Enomoto, A. Llado, T. Nakamigawa, and G. Ringel, Super edge
magic graphs, SUT J. Math., 34 (1998), 105 - 109.
10
[4] R. M. Figueroa-Centeno, R. Ichishima and F. A. Muntaner-Batle, The
place of super edge-magic labelings among other classes of labelings,
Discrete Math., 231 (2001), 153 - 168.
[5] R. M. Figueroa-Centeno, R. Ichishima and F. A. Muntaner-Batle, On
edge-magic labelings of certain disjoint union graphs, J. Graph Theory,
to appear.
[6] R. M. Figueroa-Centeno, R. Ichishima and F. A. Muntaner-Batle, Ma-
gical coronations of graphs, Australas. J. Combin., 26 (2002), 199 - 208.
[7] R. M. Figueroa-Centeno, R. Ichishima and F. A. Muntaner-Batle, On
the super edge-magic deciency of graphs, Electron. Notes Discrete
Math., 11, Elsevier, Amsterdam, 2002.
[8] R. M. Figueroa-Centeno, R. Ichishima and F. A. Muntaner-Batle, Some
new results on the super edge-magic deciency of graphs, submited.
[9] Y. Fukuchi, Edge-magic labeling of generalized Petersen graphs P(n, 2),
Ars Combin., 59 (2001), 253 - 257.
[10] J. A. Gallian, A dynamic survey of graph labeling, Electron. J. Com-
bin., (2005).
[11] J. Ivanco and I. Luckanicova, On edge-magic disconnected graphs,
SUT Journal of Math., Vol. 38, No. 2, (2002), 175 - 184.
[12] A. Kotzig and A. Rosa, Magic valuation of nite graphs, Canad. Math.
Bull., Vol. 13 (4), 1970, 451 - 461.
[13] A. Kotzig dan A. Rosa, Magic valuation of complete graphs, Publica-
tions du Centre de Recherces mathematiques Universitae de Montreal,
175 (1972).
[14] A.A. G. Ngurah, E.T. Baskoro, On the new families of (super) edge-
magic graphs, Util. Math., to appear.
[15] Slamin, M. Baca, Y. Lin, M. Miller and R. Simanjuntak, Edge magic
total labeling of wheels, fans and friendship graphs, Bull. Inst. Combin.
Appl., 35 (2002), 89 - 98.
11