Received 14 January 2015; accepted 14 December 2015; published 17 December 2015

![]()
1. Introduction
In this paper we make use of the usual notation:
for the complete bipartite graph with partition sets of sizes m and n,
for the path on n + 1 vertices,
for the disjoint union of D and F,
for the union of D and F with
(set of vertices) that belong to each other (i.e. union of D and F with common vertices of the set
belong to F and D),
for the complete graph on n vertices,
for an isolated vertex. The other terminologies not defined here can be found in [1] .
A decomposition
of a graph H is a partition of the edge set of H into edge-disjoint subgraphs
. If
for all
, then
is a decomposition of H by G. Two decompositions
and
of the complete bipartite graph
are or-
thogonal if,
for all
. Orthogonality requires that ![]()
for all
. A set of decompositions
of
is a set of k mutually orthogonal graph squares (MOGS) if
and
are orthogonal for all
and
. We use the notation
for the maximum number k in a largest possible set
of MOGS of
by G, where G is a bipartite graph with n edges.
If two decompositions
and
of
by G are orthogonal, then
is an orthogonal double cover of
by G. Orthogonal decompositions of graphs and orthogonal double covers (ODC) of graphs have been studied by several authors; see the survey articles [2] [3] .
It is well-known that orthogonal Latin squares exist for every
. A family of k-orthogonal Latin squares of order n is a set of k Latin squares any two of which are orthogonal. It is customary to denote
be the maximal number of squares in the largest possible set of mutually orthogonal Latin squares MOLS of side n. A decomposition of
by
is equivalent to a Latin square of side n; two decompositions
and
of
by
are orthogonal if and only if the corresponding Latin squares of side n are orthogonal; and thus
. The computation of
is one of the most difficult problems in combinatorial designs; see the survey articles by Abel et al. [4] and Colbourn and Dinitz in [5] . Since
is a natural extension of
, the study of
for general graphs is interesting.
El-Shanawany [6] establishes the following: i)
; ii)
and
; iii) let
be a prime number, then
; iv) let p be a prime
number, then
. Based on ii), El-Shanawany [6] proposed:
Conjecturer 1. Let p be a prime number. Then ![]()
Sampathkumar et al. [7] have proved El-Shanawany conjectured. In the following section, we present another technique to prove this conjecture as in Theorem 8.
The two sets
and
denote the vertices of the partite sets of
. The
length of the edge
of
is defined to be the difference
, where
. Note that sums and differences are carried over in
(that is, sums and differences are carried modulo n). Let G be a subgraph of
without isolated vertices and let
. The a-translate of G, denoted by
, is the edge- induced subgraph of
induced by
. A subgraph G of
is half-starter
if
and the lengths of all edges in G are mutually different.
Lemma 2 (see [8] ). If G is a half-starter, then the union of all translates of G forms an edge decomposition of
(i.e.
).
In what follows, we denote a half-starter G by the vector
,
where
and
can be obtained from the unique edge
of length i in G.
Theorem 3 (see [8] ). Two half-starters
and
are orthogonal if
.
If two half-starters
and
are orthogonal, then the set of translates of G and the set of translates of F are orthogonal.
A set of decompositions
of
} is a set of k mutually or-
thogonal graph squares (MOGS) if
and
are orthogonal for all
and
.
Note that
![]()
In the following, we define a G-square over additive group
.
Definition 4 (see [6] ). Let G be a subgraph of
A square matrix
of order n is called an G-square if every element in
occur exactly n times, and the graphs
,
with
are isomorphic to graph G.
We have already from Lemma 2 and Definition 4 that every half starter vector
and its translates are equivalent to G-square. For more illustration, the first matrix
in equation (1) is equivalent to the first row in Figure 3, which represented by the half starter vector
and its translates.
Definition 5. Two squares matrices
and
of order n are said to be orthogonal if for any ordered pair
, there is exactly one position
for
and
.
Now, we shall derive a class of mutually orthogonal subgraphs of
by a given graph G as follow.
Definition 6. A set of matrices
of
is called a set of k mutually orthogonal graph squares (MOGS) if
and
are orthogonal for all
and
.
Definition 7 (see [9] ). Let F be a certain graph, the graph F-path denoted by
, is a path of a set of vertices
and a set of edges
if and only if there exists the following two bijective mappings:
1)
defined by
, where
is a collection of d graphs, each one is isomorphic to the graph F.
2)
defined by
, where
is a class of disjoint sets of vertices (i.e.,
decomposed into
disjoint sets such that no two vertices within the same set are adjacent).
As a special case if the given graph F is isomorphic to
then
, is the natural path
that is,
.
For more illustration, see Figure 1, Figure 2.
![]()
Figure 1.
, the path of 6 sets of vertices (every sethas only 2 disjoint vertices) and 5 edges of
.
![]()
Figure 2.
, the path of 6 sets of vertices and 5 edges of
.
Consider
paths of length
, all attached to the same vertex (root vertex). This tree will be called
. Clearly,
is the star with s edges and
is the path with k edges. Define
as a set of all leaves of
i.e.
and
.
In the following section, we will compute
where
such that
as in theorem 8 and
as in theorem 11.
2. Mutually Orthogonal Graph-Path Squares
The following result was shown in [7] . Here we present another technique for the proof.
Theorem 8. Let q be a prime number. Then ![]()
Proof. Let
be a subgraph of
with q edges; for fixed
and
, define the q half- starter vectors as follows,
our task is to prove the orthogonality of those q half-stater vectors in mutually. Let us define the half starter vector
as
for all
. Then for all two different elements
, we have
, then
,
are mutually orthogonal half-starter
vectors of graphs
and
of
respectively iff
. It remains to prove the isomorphism of
half starter graphs
of
for all
. Let
, and therefore
. Furthermore, if
, then
, since
(orthogonality
of
and
), and therefore
. Moreover, for any
the
graph isomorphic to
has the edges:
■
An immediate consequence of the Theorem 8 and Conjecture 1 is the following result.
Example 9. The three mutually orthogonal decompositions (MOD) of
by
given in Figure 3 are associated with the three mutually orthogonal
-squares as in Equation (1):
(1)
Note that, every row in Figure 3 represents edge decompositions of
by
.
![]()
Figure 4.
, the path of 4 vertices and 3 edges of
.
The following result is a generalization of the Theorem 8.
Theorem 10. Let n be a prime power such that
with integer power
of a prime number q and G be a subgraph of
. Then ![]()
Proof. For fixed
and
, define the q half-starter vectors as follows,
(2)
Our task is to prove the orthogonality of those q half-starter vectors in mutually. Let us define the half starter vector
as
for all
. Then for all two different elements
, we
have
, and then
,
are mutually orthogonal
half-stater vectors of graphs
and
of
respectively iff
. It remains to prove the isomorphism of
half starter graphs
of
for all
. Let
, and therefore
. Furthermore, if
, then
, since
(orthogonality
of
and
), and therefore
. Moreover, for any
the
graph
isomorphic to G has the edges:
(3)
■
Note that, in the special case
the Theorem 10 proved El-Shanawany conjecture; also, in the case
, and
, Theorem 10 constructed an orthogonal double cover of
by G.
Furthermore, we can construct the following result using Theorem 10 in case
and
.
Theorem 11. Let
be a positive integer such that
and
be a subgraph
of
. Then ![]()
Proof. The result follows from the vector in Equation (2) and its edges in Equation (3) with
such that
, imply that
which define the
number of graphs isomorphic to F. As a direct application of Theorem 11; see Figure 4. ■
Conjecture 12.
if q is a prime number with an integer power
.
Conjecture 13.
if q is a prime number with an integer power
and
are positive integers.