Antimagic Labeling and Canonical

Decomposition of graphs
Presented by:
Maimuna Begum Kali
cse-0416052008

Paper Details
Author:
Michael D. Barrus,Department of

Mathematics, Black Hills State University,

Spearfish, SD 57799, United States
Published in:
Journal: Information Processing Letters
Volume 110 Issue 7, March, 2010

Pages 261-263

Presentation contents
Problem definition
Results of the paper
Contribution of the paper in respect to

previous results
Algorithms and methodologies including

proofs
Future works

Antimagic Labeling
A graph G(V,E) where|E|=n ,is called

antimagic if the n edges of G can be

distinctly labeled 1 through n in such a way
that when taking the sum of the edge labels
incident to each vertex, the sums will all be
different
A graph is called antimagic if it has an
K4

antimagic labeling.
3

1
5

4
6

13

Figure :antimagic labeling for K4.

2
1

11
5

4
6

12

Problem Definition
The concept of an antimagic graph was

introduced by Hartfield and Ringel (1990).

They conjectured that
Every connected graph, except K2 , is antimagic
This conjecture is still open
Various Classes of Graph have been shown to be
antimagic

-i.e. Lattice grids and prisms, Dense

Graph ,Regular Bipartite graph,Toroidal
grids, Regular graphs of odd degree etc.

Results of the paper

This paper intended to present an algorithm

that produces an Antimagic Labeling for any

graph containing a clique with special
neighborhood properties.
They characterized the graphs having such
cliques , these are precisely
The split graph
The graph which is canonically decomposable.

As a result they obtained a condition on the

degree sequence of a graph which ensures

that the graph is antimagic

State of art
Hartfield and Ringel (1990)
Every

connected graph, except K2 , is

antimagic

Alon et al.(2004)
Every n-vertex graphs with maximum degree

at least n 2 and graphs with high minimum

degree are antimagic.
Y. Cheng(2007)
Lattice grids and prisms are antimagic.

Summary of Antimagic
Labeling
Graph

Labeling

Notes

Pn

For n >=3

Cn

Wn

Kn

every tree except K2

A?

every connected
graph except K2

A?

n >=4 vertices
(G) >=n- 2

all complete partite

graphs except K2

For n >=3

OPEN

Terminologies
A Clique in a graph is a set of pair wise

adjacent vertices, whereas An independent

set is a set of pair wise non-adjacent
vertices.
Here, {1,2,3},
{4,5,6,7} are some
cliques of the many
cliques present in the
graph.
Highlighted portion indicate
Two independent sets ,one
containing the black colored
vertices and the other
containing the white colored
vertices.

Terminologies
Split Graph is a graph whose vertex set can

be partitioned into a clique and an

independent set.

An
independe
nt set

A Clique

P4 is a split graph

Methodology

A dominating Clique
Condition
V(G) denotes vertex set of Graph G
dG (v) is the degree of vertex v in G
Open neighborhoods of a vertex v in G to

be the set
NG(v)={u V(G):uv E(G)}

Closed neighborhood of a vertex v in G to

be
NG[v]={v} {uV(G):uv E(G)} }

Given WV(G), then G[W] denotes the

induced subgraph of G with vertetx set W.

A dominating Clique
Condition
Lemma 1:

Let G be a connected graph on at least 3

vertices . If G has a clique B such that for every
vertex v in G ,either NG(v) B or B NG[v], then
G is antimagic.

Lemma 1:Let G be a connected graph on at least 3 vertices . If G has a clique B such

that for every vertex v in G , either NG(v)B or
B NG[v], then G is antimagic.

Proof:

Let A denote the set of vertices v not in B

such that
NG(v) B . Let A={a1,a2,a|a|} and
B={b1,b2,b|b|} ,
where in each set the vertices are indexed in nondecreasing order of degrees , And the vertices of
degree 1 in A are indexed in non decreasing order
of the degree of their respective neighbors in B .
Let C = V (G) A B.

a1

b1

a2

b2

C
A is an independent
Set so each vertex of
C is adjacent to every
vertex of B

Lemma 1:Let G be a connected graph on at least 3 vertices . If G has a clique B such

that for every vertex v in G , either NG(v)B or
B NG[v], then G is antimagic.

Steps for labeling the graph:

Order the edges of the form aibj

lexicographically on the index pair(i,j) and

assign them the first numbers in the order.
Label all the edges in G[C] arbitrarily with
the next smallest numbers . A function g
on C is defined by letting g(v) to be the sum
of labels on all edges incident with v in
G[C]. Now denote the vertices c1,c2. c|C|
where the vertices are now indexed in non
decreasing order of their values under g.

a1

b1
3

b2

a2

4
2

Lemma 1:Let G be a connected graph on at least 3 vertices . If G has a clique B such

that for every vertex v in G , either NG(v)B or
B NG[v], then G is antimagic.

Steps for labeling the graph:

Order the edges of the form aibj

lexicographically on the index pair(i,j) and

assign them the first numbers in the order.
Label all the edges in G[C] arbitrarily with
the next smallest numbers . A function g
on C is defined by letting g(v) to be the sum
of labels on all edges incident with v in
G[C]. Now denote the vertices c1,c2. c|C|
where the vertices are now indexed in non
decreasing order of their values under g.

a1

b1

c1 G(v)=
3

b2

a2
2

c2

G(v)=
4

G(v)=
7
c3

Lemma 1:Let G be a connected graph on at least 3 vertices . If G has a clique B such

that for every vertex v in G , either NG(v)B or
B NG[v], then G is antimagic.

Steps for labeling the graph:

Order the edges of the form cibj

lexicographically on the index pair(i,j) and

label them with next first numbers in the
order.
For every vertices v in B, a function is defined
g`(v) to be the sum of labels on all edges
joining v with a vertex outside B. Now the
vertices in B set is indexed according to their
value under g`.
Label the remaining edges b ibj in lexicographic order

on the pairs (i,j)

a1

b1

c1

3
6

b2

a2

9
10
4

c2

c3

Lemma 1:Let G be a connected graph on at least 3 vertices . If G has a clique B such

that for every vertex v in G , either NG(v)B or
B NG[v], then G is antimagic.

Steps for labeling the graph:

Order the edges of the form cibj

lexicographically on the index pair(i,j) and

label them with next first numbers in the
order.
For every vertices v in B, a function is defined
g`(v) to be the sum of labels on all edges
joining v with a vertex outside B. Now the
vertices in B set is indexed according to their
value under g`.
Label the remaining edges b ibj in lexicographic order

on the pairs (i,j)

a1

g(v)=1+5+9+7
=22
1
b1
5

c1
3

1
b12

a2
2

6
8

g(v)=2+6+10+8
=26

9
10
4

c2

c3

33

14

3
1
1

2
2

37

9
10

8
19

26

Well labeling is done , whats next ?

Is the labeling antimagic labeling ?
For each vertex v in G, let f(v) denote

the sum of the labels on edges

containing v.
For c C and ai, aj A with 1
i<
j,< 3
< 2
dG(ai) dG(aj) < dG(c),

are labels on
edges incident
with ai,aj c

1 < 2 < 3
Hence,f(ai) < f(aj) < f(c), so f is
injective on A,

a1

b1

c1

3
1
1
b2

a2

6
8

9
10
4

c2

c3

Is the labeling antimagic labeling ?

For ci, cj C with i < j, we have

g(ci) g(cj) => f(ci) < f(cj), so f is

injective on C.
For u (A C) and b B, if v is a neighbor of
u other than b, then v is also a neighbor of b,
and edge uv receives a label less than the label
on bv; hence f(u) < f(b).
If u has no neighbors other than b, then b has
a neighbor other than u (since G is connected
and not K2) and again we see that f(u) < f(b).

Is the labeling antimagic labeling ?

For ci, cj C with i < j, we have

g(ci) g(cj) => f(ci) < f(cj), so f is

injective on C.
For u (A C) and b B, if v is a neighbor of
u other than b, then v is also a neighbor of b,
and edge uv receives a label less than the label
on bv; hence f(u) < f(b).
If u has no neighbors other than b, then b has
a neighbor other than u (since G is connected
and not K2) and again we see that f(u) < f(b).

Is the labeling antimagic labeling ?

Finally, it is shown that f takes on different

values for all vertices of B.

This is trivial if |B| = 1.
Suppose that |B| = 2. In this case ( AC) is
nonempty, since G is not K2. If A is nonempty,
then it consists of pendant vertices adjacent to
either b1 or b2. Since dG(b1) dG(b2), the
contribution to f(b1) from edges joining b1 to
vertices in A is strictly less than the
corresponding contribution to f(b2), by
construction. Each vertex in C is adjacent to
both b1 and b2, and the label on the edge
joining it to b1 is smaller than the label on the

Is the labeling antimagic labeling ?

Finally, suppose that |B| 3. Let bi and bj be

vertices of B with i < j;

by definition, g(bi) g(bj).
Since every other vertex bk in B is adjacent to

both bi and bj, with edge bibk receiving a

lesser label than bjbk, it is clear that f(bi) < f(b
j).

Dense graph with dominating vertex is

antimagic, likewise Lemma 1 extends the

result to graph having a special
dominating clique.

Canonically decomposable
Graph
Tyshkevich defines a canonically decomposable

graph as followsFor a split graph S with a given partition of its

vertex set into an independent set A and a
clique B (denoted by S(A,B)), and an arbitrary
graph H the composition S(A,B) H is the
graph obtained by taking the disjoint union of
S(A,B) and H .
If G contains nonempty induced sub graphs H
and S and vertex subsets A and B such that G =
S(A,B) H, then G is canonically decomposable;
otherwise G is canonically indecomposable.

The composition of P4 and

P3.

Tyshkevich also showed that each

graph can be expressed as a

composition
S1(A1, B1). Sk(Ak, Bk) H
of indecomposable induced sub graphs (note
that is associative);
indecomposable graphs are those for which k
= 0. This representation is known as the
canonical decomposition of the graph

Lemma 2. The following are equivalent for a

graph G:
(1) G has a clique B such that for all v V (G) NG(v)

B or B NG[v], ;
(2) G is split or canonically decomposable.
Proof.
(1) (2):
Let A be the set of vertices v not in B such that NG(v)

B, and let C = V (G) A B. Note that A is an

independent set.
If C =, then G is a split graph. Otherwise, we may
write G as the composition G(A, B) G[C], where G =
G[A B].

Lemma 2. The following are equivalent for a

graph G:
(1) G has a clique B such that for all v V (G) NG(v)

B or B NG[v], ;
(2) G is split or canonically decomposable.
Proof.
(2)(1):
If G is split, then we may partition V (G) into an

independent set A and a clique B. If G is

decomposable in the canonical decomposition, then
we may write G = S(A, B) H for vertex subsets A and
B and induced sub graphs S and H of G. In both cases
either NG(v) B or B NG[v], for each vertex v in G.

The previous two lemmas immediately establish

our main result.

Theorem 3. Connected graphs on at

least 3 vertices that are
split or canonically decomposable are
antimagic.
It is observable that canonically decomposable
graphs on at least three vertices are connected if
and only if they have no isolated vertices.

 As a consequence of known degree sequence

characterizations of split graphs and

decomposable graphs a sufficient degree
sequence condition for antimagic graphs can be
found :
Corollary 4.
Let G be an n-vertex graph (n 3) with degree
sequence (d1, . . . ,dn) in nonincreasing order such
that dn > 0. If there exist integers p and q such that 0
< p +q <=n and
Then G is
antimagic

Conclusion
To settle the conjecture of Hartsfield and Ringel, it suffices

to consider graphs that are indecomposable under the

canonical decomposition .
Barrus and West gave a characterization of
indecomposable graphs in terms of alternating 4-cycles, A4
structure, configurations on four vertices a, b, c, and d such
that ab and cd are edges and ad and bc are not. They
showed the following-- Theorem 5. A graph G is indecomposable under the canonical

decomposition if and only if for every pair u , v of vertices there is a

sequence V1Vk of 4-element subsets of V (G) such that u and v
belong to V1 and Vk respectively,
1

Vi Vi+1= for 1 <=i <= k 1, and each Vi is the vertex set of

an alternating 4-cycle.

Open question
The conjecture is still open, Many more

work to be done in this field.

The author left a question to wonder
Question.
Are there large families of connected graphs for

which it is possible to use the structure of the

alternating 4-cycles in the graph to produce an
antimagic labeling?

Thank You