Received 4 January 2016; accepted 26 February 2016; published 29 February 2016

1. Introduction
We consider finite graphs and digraphs, and undefined terms and notations will follow [1] for graphs and [2] for digraphs. Throughout this paper, the notation
denotes an arc oriented from u to v. A digraph D is strict if it contains no parallel arcs nor loops; and is symmetric if for any vertices u,
, if
, then
. If two arcs of D have a common vertex, we say that these two arcs are adjacent in D. A directed path in a digraph D from a vertex u to a vertex v is called a
-dipath. To emphasize the distinction between graphs and digraphs, a directed cycle or path in a digraph is often referred as a dicycle or dipath. A dipath P is a hamiltonian dipath if
. A digraph D is hamiltonian if D contains a hamiltonian dicycle. An
-hamiltonian dipath is a hamiltonian dipath from x to y. A digraph D is hamiltonian-connected if D has an
-hamiltonian dipath for every choice of distinct vertices
.
As in [2] ,
denotes the arc-strong-connectivity of D. A digraph D is strong if and only if
. For
, we define
![]()
For a subset
, the subdigraph arc-induced by
is the digraph
, where
is the set of vertices in V which are incident with at least one arc in
.
Let
![]()
When
, we write
and
. Let
and
denote the out-neighbourhood and in-neighbourhood of v in D, respectively. Vertices in
,
are called the out-neighbours, in-neighbours of v. Thus for a digraph D and an integer
,
(1)
Boesch, Suffel, and Tindell [3] in 1977 proposed the supereulerian problem, which seeks to characterize graphs that have spanning eulerian subgraphs. They indicated that this problem would be very difficult. Pulleyblank [4] later in 1979 proved that determining whether a graph is supereulerian, even within planar graphs, is NP- complete. Catlin [5] in 1992 presented the first survey on supereulerian graphs. Chen et al. [6] surveyed the reduction method associated with the supereulerian problem and their applications. An updated survey presenting the more recent developments can be found in [7] .
It is natural to consider the supereulerian problem in digraphs. A digraph D is eulerian if it contains a closed ditrail W such that
, or, equivalently, if D is strong and for any
,
. A digraph D is supereulerian if D contains a closed ditrail W such that
, or, equivalently, if D contains a spanning eulerian subdigraph. Some recent developments on supereulerian digraphs are given in [8] -[12] .
A central problem is to determine or characterize supereulerian digraphs. In Section 2, the 2-sum
of two digraphs
and
is defined, and some basic properties of 2-sums are discussed. We will observe that a 2-sum of two supereulerian (or even hamiltonian) digraphs may not be supereulerian. Thus it is natural to seek sufficient conditions on
and
for the 2-sum of
and
to be supereulerian. In the last section, we will present several sufficient conditions for supereulerian 2-sums of digraphs. In particular, we show that if
and
are either symmetrically connected or partially symmetric (to be defined in Section 3), then
is supereulerian.
2. The 2-Sums of Digraphs
The definition and some elementary properties of the 2-sums of digraphs are presented in this section. A digraph is nontrivial if it contains at least one arc. Throughout this section, all digraphs are assumed to be nontrivial.
Definition 2.1 Let
and
be two vertex disjoint digraphs, and let
and
be two distinguished arcs. The 2-sum
of
and
with base arcs
and
is obtained from the union of
and
by identifying
with
and
with
, respectively. When the arcs
and
are not emphasized or is understood from the context, we often use
for
.
Lemma 1 Let
and
be two vertex disjoint strong digraphs. Then
![]()
Proof. Let
be an integer such that
, and let
. We shall show that
. By (1), there exists a proper nonempty vertex subset
such that
. Let
. We argue by contradiction and assume that
.
By Definition 2.1, we have
and
in
. If
and
, we obtain that
and
, then
and
. It follows by (1) that
, contrary to the assumption that
. Similarly, if
and
, then
and
, hence a contradiction to the assumption that
is obtained from
.
Thus, we may assume that
and
. Let
. Then
is a proper nonempty subset of
, and
. It follows by (1) that
contrary to the assumption that
.
Example 2.1 The converse of Lemma 1 may not always stand, as indicated by the example below, depicted in Figure 1. Let
and
. Let
and
. Let
and
.
Then, it is routine to verify that
. While
is strong, the digraph
contains a vertex
with
, and so
.
Lemma 2 A digraph D is not supereulerian if for some integer
,
has vertex disjoint subsets
satisfying both of the following:
i) ![]()
ii)
.
Proof. By contradiction, we assume that both i) and ii) hold and D is supereulerian. Let S be a spanning eulerian subdigraph of D, then
and
. Since S is eulerian, for any subset
, it follows that
. Thus, by ii), we conclude that
(2)
By i) and by (2), there must be a
with
such that
, contrary to the assumption that
.
Lemma 2 can be applied to find examples of hamiltonian digraphs whose 2-sum is not supereulerian, as shown in Example 2.2 below.
Example 2.2 Let
be integers and
and
be two vertex disjoint dicycles with length
and
, respectively. We claim that
is not supereulerian. To justify this claim, we denote
, and
. Without loss of generality, we assume that
and
, and
. Let
and
be subdigraphs of
with
,
and
, respectively. By Lemma 2, we conclude that
is not supereulerian (see Figure 2).
3. Sufficient Conditions for Supereulerian 2-Sums of Digraphs
In this section, we will show several sufficient conditions on
and
to assure that the 2-sum ![]()
is supereulerian.
Proposition 1 Let
and
be two vertex disjoint supereulerian digraphs with
and
, and let
denote
. Each of the following holds.
i) For some
, if
has a spanning eulerian subdigraph
such that
, then
is supereulerian.
ii) If for some
,
is hamiltonian-connected, then
is supereulerian.
Proof. i) Since
and
are supereulerian digraphs,
and
are strongly connected, and so by Lemma 1,
is also strongly connected. Without loss of generality, we assume that
and
has a spanning eulerian subdigraph
such that
. Since
is supereulerian, we can pick a spanning eulerian subdigraph
in
. Then
and
. It follows that
is a spanning eulerian subdigraph in
.
ii) Without loss of generality, we assume that
and
is hamiltonian-connected, and so
has a
-hamiltonian dipath
and a
-hamiltonian dipath
. Since
is supereulerian,
contains a spanning eulerian subdigraph
. Define
![]()
As in any case, S is strongly connected and every vertex
satisfies
, and so S is eulerian. Since
, for
, we conclude that S is a spanning eulerian subdigraph of
, and so
is supereulerian.
Theorem 2 [13] If a strict digraph on
vertices has
or more arcs, then it is hamiltonian- connected.
Corollary 1 Let
be a strict digraph on
vertices and with
. If
is a supereulerian digraph, then
is supereulerian.
Proof. By Theorem 2,
is hamiltonian-connected. Then by Proposition 1 (ii),
is supereulerian.
Two classes of supereulerian digraphs seem to be of particular interests in studying supereulerian digraph 2- sums. We first present their definitions.
Definition 3.2 Let D be a digraph such that either
or
. If for any
, D contains a symmetric dipath from u to v, then D is called a symmetrically connected digraph.
Given a digraph D, define a relation ~ on
such that
if and only if
or D has a symmetrically connected subdigraph H with
. By definition, one can routinely verify that ~ is an equivalence relation. Each equivalence class induces a symmetrically connected component of D. Hence D is symmetrically connected if and only if D has only one symmetrically connected component. A symmetrically connected component of D is also called a maximal symmetrically connected subdigraph of D. When D has more than one symmetrically connected components, we have the following definition.
Definition 3.3 Let D be a weakly connected digraph and
be the set of maximal symmetrically connected subdigraphs of D with
. If for any proper nonempty subset
,
(3)
then D is partially symmetric.
It is known that both symmetrically connected digraphs and partially symmetric digraphs are supereulerian.
Theorem 3 ( [14] and [15] ) Each of the following holds.
i) Every symmetrically connected digraph is supereulerian.
ii) Every partially symmetric digraph is supereulerian.
A main result of this section is to show that the digraph 2-sums of symmetrically connected or partially symmetric digraphs are supereulerian.
Lemma 3 Let
and
be two vertex disjoint digraphs with
and
, and let
denote
. Each of the following holds.
i) If
and
are symmetrically connected, then
is symmetrically connected.
ii) If
and
are partially symmetric, then
is partially symmetric.
iii) If
is symmetric and
is partially symmetric, then
is partially symmetric.
Proof. i) For any vertices
, we shall show that
always has a symmetric
dipath. If for some
, we have
, then as
is symmetrically connected,
contains a symmetric
-dipath P. Since
is a subdigraph of
, P is also a symmetric
-dipath of
. Hence we may assume that
and
. Since
and
are symmetrically connected,
contains a symmetric
-dipath
and
contains a symmetric
-dipath
. By Definition 2.1,
and
represent the same vertex in
, and so
is a symmetric
-dipath in
.
ii) Fix
. Since
is partially symmetric, for some integer
, let
be the set of all maximal symmetrically connected subdigraphs of
. Without loss of generality, we assume that
and
; and for some
with
and
,
and
. (We allow the possibility that
and/or
). Define, for
and
,
![]()
Then,
is the set of all maximal symmetrically connected sub- digraphs of
. Note that
and
. We shall show by definition that
is partially symmetric. To do that, let
be a nonempty proper subset of
. We shall show that (3) holds.
Since
, we either have
or
. By symmetry, we may assume that
.
Suppose first that
. Let
. Then
. Since
is partially symmetric, there exist an
, a vertex
,
and an
such that
![]()
This implies that the vertex
,
, and
such that
![]()
Thus (3) holds in this case.
Hence we may assume that
. Since
is a proper subset, we must have
. Since
, we also have
. With a similar argument, we conclude that (3) must also hold in this case.
iii) Let
and let
be the set of all maximal symmetrically connected subdigraphs of
with
and for some
,
. (We allow the possibility that
). Define
![]()
Then
is the set of all maximal symmetrically connected subdigraphs of
. Note that
with this notation. Let
be a nonempty proper subset of
. We shall show that (3) holds.
Let
. Since
is proper,
is a nonempty proper subset of
. Since
is partially symmetric, by Definition 3.2, there exist an
, a vertex
, and an
such that
![]()
This implies that vertex
,
and
such that
![]()
Thus (3) holds, and so by definition,
is partially symmetric.
Theorem 4 Let
and
be two digraphs. Each of the following holds.
i) If
and
are symmetrically connected, then
is supereulerian.
ii) If
and
are partially symmetric, then
is supereulerian.
iii) If
is symmetric and
is partially symmetric, then
is supereulerian.
Proof. This follows from Theorem 3 and Lemma 3.
It is also natural to consider sufficient conditions on
and
for
to be hamiltonian.
Theorem 5 If
is hamiltonian and
is hamiltonian-connected digraphs, then
is hamiltonian.
Proof. Let
with
be a hamiltonian dicycle of
and
. Let
and
, and
. Since
is hamiltonian-connected,
contains a
-hamiltonian dipath P. Thus
is a hamiltonian dicycle in
.
Theorem 6 (Thomassen [16] ) If a semicomplete digraph D is 4-strong, then D is hamiltonian-connected.
By Theorem 5 and 6, we have the following corollary.
Corollary 2 Let
and
be two 4-strong semicomplete digraphs, then
is hamiltonian.
Acknowledgements
The research of Juan Liu was partially supported by grants NSFC (No. 61363020, 11301450) and China Scholarship Council, and the research of Xindong Zhang was supported in part by grants NSFC (No. 11461072) and the Youth Science and Technology Education Project of Xinjiang (No. 2014731003).