290
Views
0
CrossRef citations to date
0
Altmetric
Articles

Euler dynamic H-trails in edge-colored graphs

&
Pages 48-56 | Received 11 Oct 2022, Accepted 17 Aug 2023, Published online: 04 Sep 2023

Abstract

Alternating Euler trails has been extensively studied for its diverse practical and theoretical applications. Let H be a graph possibly with loops and G be a multigraph without loops. In this paper we deal with any fixed coloration of E(G) with V(H) (H-coloring of G). A sequence W=(v0,e01,,e0k0,v1,e11,,en1kn1,vn) in G, where for each i{0,,n1},ki1 and eij=vivi+1 is an edge in G, for every j{1,,ki}, is a dynamic H-trail if W does not repeat edges and c(eiki)c(ei+11) is an edge in H, for each i{0,,n2}. In particular, a dynamic H-trail is an alternating trail when H is a complete graph without loops and ki = 1, for every i{1,,n1}. In this paper, we introduce the concept of dynamic H-trail, which arises in a natural way in the modeling of many practical problems, in particular, in theoretical computer science.

We provide necessary and sufficient conditions for the existence of closed Euler dynamic H-trail in H-colored multigraphs. Also we provide polynomial time algorithms that allows us to convert a cycle in an auxiliary graph, L2H(G), in a closed dynamic H-trail in G, and vice versa, where L2H(G) is a non-colored simple graph obtained from G in a polynomial time.

1 Introduction

For basic concepts, terminology and notation not defined here, we refer the reader to [Citation3] and [Citation5]. Throughout this work, we will consider graphs, multigraphs (graphs allowing parallel edges) and simple graphs (graphs with no parallel edges nor loops). Let V(G) and E(G) the sets of vertices and edges of a multigraph G, respectively.

A spanning circuit in a graph G is defined as a closed trail that contains each vertex of G. We will say that a graph is supereulerian if it contains a spanning circuit. Let T be a spanning circuit in G: if T visits each vertex of G exactly once, then T will be called Hamilton cycle; and if T visits each edge of G, then T will be called Euler trail. We will say that a graph is hamiltonian if it has a Hamilton cycle and eulerian if it contains an Euler trail.

In [Citation15], Harary and Nash-Williams introduced the following graphs. Let G be a graph with p vertices and q edges. For n2, let Ln(G) the graph with nq vertices that are obtained as follows: for each edge e = uv of G, we take two vertices f(u, e) and f(v, e) in Ln(G) and adding a path with n – 2 new intermediate vertices connecting f(u, e) and f(v, e); finally, for each vertex u of G, we add an edge joining f(u, e) and f(u, g), whenever e and g are distinct edges with end point u. They also proved the followings relationships between eulerian graphs and hamiltonian cycles of Ln(G).

Theorem 1

(Harary and Nash-Williams [Citation15]). Let G be a graph. The following assertions holds:

  1. If G is eulerian, then Ln(G) is hamiltonian, for every n2.

  2. If Ln(G) is hamiltonian, for some n3, then G is eulerian.

  3. G is superulerian if and only if L2(G) is hamiltonian.

An edge-coloring of a graph G is defined as a mapping c:E(G)N, where N is the set of natural numbers. We will say that G is an edge-colored graph whenever we are taking a fixed edge-coloring of G. A properly colored spanning circuit is a spanning circuit with no consecutive edges having the same color, including the first and the last in a closed spanning circuit. An edge-colored graph is supereulerian, if G contains a properly colored spanning circuit. Several authors have worked with this concept, for example, Bang-Jensen et al. [Citation2]; Guo et al. [Citation13]; Guo et al. [Citation14]. Properly colored walks are of interest as a generalization of walk in undirected and directed graphs, see [Citation3], as well as, in graph theory application, for example, in genetic and molecular biology [Citation9, Citation10, Citation20], social science [Citation6], channel assignment is wireless networks [Citation1, Citation18]. For more details on properly colored walks, we refer the reader to Chapter 16 of [Citation3].

Kotzig [Citation16] gave the following characterization of edge-colored multigraphs containing properly colored closed Euler trail.

Theorem 2

(Kotzig [Citation16]). Let G be an edge-colored eulerian multigraph. Then G has a properly colored closed Euler trail if and only if δi(x)jiδj(x), where δi(x) is the number of edges with color i incident with x, for each vertex x of G.

Different kinds of edge-coloring in undirected and directed graphs have been studied, for example, in [Citation17] the arcs of a tournament were colored with the vertices of a poset. In this paper we will consider the following edge-coloring. Let H be a graph possibly with loops and G be a graph without loops. An H-coloring of G is a function c:E(G)V(H). We will say that G is an H-colored graph, whenever we are taking a fixed H-coloring of G. A walk W=(v0,e0,v1,e1,v2,,vk1,ek1,vk) in G, where ei=vivi+1 for every i in {0,,k1}, is an H-walk if (c(e0),a0,c(e1),,c(ek2),ak2,c(ek1)) is a walk in H, with ai=c(ei)c(ei+1) for every i{0,,k2}. We will say that W is closed if v0=vk and c(ek1)c(e0)E(H). Let W be an H-walk, if W is a trail then W will be called H-trail. Notice that if H is a complete graph without loops, then an H-walk is a properly colored walk.

Since, properly colored walks have been very useful for modeling and solving different problems, as we mentioned above, H-colored graphs and H-walks can also model interesting problems, for example, routing problems or in communications networks were some transitions are restricted because of natural phenomenon, external attack or failure. More applications of colored walks with restrictions in color’s succession can be found in [Citation19].

The concepts of H-coloring and H-walks were introduced, for the first time by Linek and Sands in [Citation17], and have been worked mainly in the context of kernel theory and related topics, see [Citation7, Citation8, Citation12].

Galeana-Sánchez et al. [Citation11] defined the graph Gu as follows: Let u be a vertex of G; Gu is the graph such that V(Gu)={eE(G):eisincidentwithu}, and two different vertices a and b are joining by only one edge in Gu if and only if c(a) and c(b) are adjacent in H.

They also showed necessary and sufficient conditions for the existence of closed Euler H-trails, as follows.

Theorem 3

(Galeana-Sánchez et al. [Citation11]). Let H be a graph possibly with loops and G be an H-colored multigraph without loops. Suppose that G is Eulerian and Gu is a complete ku -partite graph, for every u in V(G) and for some ku in N. Then G has a closed Euler H-trail if and only if |Ciu|ji|Cju| for every u in V(G), where {C1u,,Ckuu} is the partition of V(Gu) into independent sets.

Benítez-Bobadilla et al. [Citation4] introduced a generalization of H-walk as follows. They allowed “lane changes”, i.e., they allowed concatenation of two H-walks as long as, the last edge of the first one and the first edge of the second one satisfy that xk1=y0 and xk=y1. As a result, they defined the following new concept: Let G be an H-colored multigraph, a dynamic H-walk in G is a sequence of vertices W=(x0,x1,,xk) in G such that for each i{0,,k2} there exists an edge fi=xixi+1 and there exists an edge fi+1=xi+1xi+2 such that c(fi)c(fi+1) is an edge in H.

For the purpose of this paper, we need a definition and notation that will allows us to know the edges that belong to a dynamic H-walk. So, we will say that W=(v0,e01,,e0k0,v1,e11,,e1k1,v2,,vn1,en1kn1,,en1kn1,vn), where for each i{0,,n1},ki1 and eij=vivi+1 for every j{1,,ki}, is a dynamic H-walk if c(eiki)c(ei+11) is an edge in H, for each i{0,,n2}. If W is a dynamic H-walk that does not repeat edges, then W will be called dynamic H-trail. We will say that a dynamic H-trail, W, is an Euler dynamic H-trail if E(G)=E(W). We will say that a dynamic H-trail is a closed dynamic H-trail whenever a) v0=vn and c(en1kn1)c(e01) is an edge in H; or b) v1=vn and en1kn1 and e01 are parallel in G. Notice that if W is a closed dynamic H-walk that satisfies condition b, then W can be rewrite as W=(v1,e11,,vn1=v0,en11,,en1kn1,e01,,e0k0,vn=v1), and W satisfies condition (a) (unless n = 1, i.e., W is of the form (x,e0,,ek,y), where k1).

It follows from the definition of dynamic H-trail that every H-trail in G is a dynamic H-trail in G. Moreover, if G has no parallel edges, then every dynamic H-trail is an H-trail.

A motivation for the study of dynamic H-walks are their possibles applications. For example, suppose that it is require to send a message from point A to point B through communication network which can have faults in its links (such as damage, attack, virus or blockage), where each one of its edge has an assigned color depending on the probability that the message is delivered correctly (in time, complete, unchanged or virus-free). In this case, the vertices of H are the colors assigned to the edges of the network and the set of edges of H will depend on the transitions that are convenient (for example, if transition with the same probability of failure are forbidden, then H will have no loops). In practice, it is possible to send messages simultaneously over two or more connections. In these cases, the use of dynamic H-walks can improve message delivery.

Motivated by Theorem 3 and its proof, we think that it is possible to give conditions, similar to those of Theorem 3, on an H-colored multigraph that guarantee the existence of closed Euler dynamic H-trail. In the process, we find an auxiliary graph, we call it L3H(G), that allows us to link the closed dynamic H-trails with hamiltonian graphs. As a result, we will prove that G contains a closed Euler dynamic H-trails if and only if LnH(G) is hamiltonian, for every n3.

2 Main results

In what follows, H will be a graph possibly with loops, and G will be a multigraph without loops.

First of all, we will introduce an auxiliary graph, denoted by LnH(G), which is defined as follows.

Definition 1.

Let G be an H-colored multigraph with |E(G)|=q. For n2,LnH(G) is the graph with nq vertices, obtained as follows: for each edge e = uv of G, we take two vertices f(u, e) and f(v, e) in LnH(G), and adding a path with n – 2 new intermediate vertices connecting f(u, e) and f(v, e). And the rest of the edges of LnH(G) are defined as follows: a) f(u, e) and f(u, g) are adjacent iff eg and c(e)c(g)E(H); b) f(u, e) and f(v, g) are adjacent iff uv and e and g are parallel in G.

Notice that LnH(G) is a simple graph.

Before giving a new way to construct the graph LnH(G), we first introduce some additional notation.

Recall that Gu is the graph such that V(Gu)={eE(G):eisincidentwithu}, and two different vertices a and b are joining by only one edge in Gu if and only if c(a) and c(b) are adjacent in H. Notice that V(Gu) can also be denoted as {f(u,e):e is incident in u}.

For each e = xy in E(G), by the definition of Gx and Gy , we have that eV(Gx) and eV(Gy). So, we will say that f(x, e) and f(y, e) are the copies of e seen as a vertex in Gx and Gy , respectively; and if Exy is the set of all the edges with end vertices x and y, then we will say that Exyx={f(x,e):eExy}.

Notice that LnH(G) can be constructed as follows. First take the disjoint union of Gx , for every x in V(G). Then, for every pair of distinct vertices, x and y in V(G), such that |Exy|=m1, we will have a complete bipartite graph Km,m between Exyx and Exyy. Finally, for every e=xyE(G), we change the edge joining f(x, e) and f(y, e) by a path with n – 2 new intermediate vertices. The construction of L3H(G) is illustrated in .

Fig. 1 The sequence P=(v1,e1,v2,e6,e5,v4,e4,v3,e3,e2,v1) is a closed Euler dynamic H-trail in G but there is no closed Euler H-trail in G.

Fig. 1 The sequence P=(v1,e1,v2,e6,e5,v4,e4,v3,e3,e2,v1) is a closed Euler dynamic H-trail in G but there is no closed Euler H-trail in G.

It follows from the definition of L2H(G) that MJ={f(x,e)f(y,e)E(L2H(G)):e=xyE(G)} is a perfect matching, that we will called it joint matching of L2H(G).

The following two lemmas show how to construct a cycle in L2H(G), based on the order of the edges in a closed dynamic H-trail in G, and vice versa.

Lemma 4.

Let G be an H-colored multigraph. If P=(x0,e1,,ep0,x1,ep0+1,,ep0++p1,x2,,xn1,ep0++pn2+1,,ep0++pn1,xn) is a closed dynamic H-trail in G, then C=(f(x0,e1),f(x1,e1),,f(x0,ep0),f(x1,ep0),f(x1,ep0+1),f(x2,ep0+1),,f(xn,ep0++pn1),f(x0,e1)) is a cycle in L2H(G).

Proof.

It follows from the definition of L2H(G) that f(xi,e)f(xi+1,e) is an edge of L2H(G), for every e=xixi+1E(P).

Let ei and ei+1 be consecutive edges in P (if i=p0++pn1, then ei+1=e1). If ei and ei+1 are parallel, such that xj and xj+1 are the ends of both edges, then f(xj+1,ei) and f(xj,ei+1) are adjacent in L2H(G), by the definition of L2H(G). Otherwise, ei and ei+1 are incident with xj , for some xj in V(P). Since P is a closed dynamic H-trail, then c(ei)c(ei+1) is an edge in H and f(xj,ei)f(xj,ei+1)E(L2H(G)). Hence, C is a walk in L2H(G). Since, P is a closed dynamic H-trail, then C does not repeat vertices and, so C is a cycle. Moreover, C alternate edges between E(L2H(G))MJ and MJ . □

Lemma 5.

Let G be an H-colored multigraph and MJ be the joint matching of L2H(G). If C=(f(x1,e1),f(y1,e1),f(x2,e2),f(y2,e2),f(yq,eq),f(x1,e1)), where ei=xiyiE(G), is a cycle in L2H(G) that alternate edges between E(L2H(G))MJ and MJ , then the following steps produces a closed dynamic H-trail in G.

  1. Star with P=(x1,e1,y1) and k = 2.

  2. Let ek , with end vertices xk and yk . If ek1 and ek are parallel, xk=xk1 and yk=yk1, then we change lanes from the edge ek1 to ek in P. Otherwise, continue P by the edge ek .

  3. k=k+1.

  4. If k=q+1, finish. Otherwise, go to step 2.

Proof.

First, we will prove that P is a dynamic H-trail in G.

Let ek1 and ek , with 2kq.

Since f(yk1,ek1)f(xk,ek)E(C), we have that f(yk1,ek1)f(xk,ek) is in E(L2H(G)). Hence, yk1=xk and c(ek1)c(ek)E(H) or yk1xk and ek1 and ek are parallel in G.

If yk1=xk and c(ek1)c(ek)E(H), then ek1 and ek are incident in yk1=xk and (x1,e1,P,xk1,ek1,yk1=xk,ek,yk) is a dynamic H-walk in G.

Otherwise, yk1xk and ek1 and ek are parallel edges in G, and (x1,e1,P,xk1=xk,ek1,ek,yk1=yk) is a dynamic H-walk in G.

Hence, P is a dynamic H-walk in G.

On the other hand, since C is a cycle, then ei appears exactly once in P, for every i{1,,q}, and so P is a dynamic H-trail.

Now, we prove that P is closed.

Recall that ei=xiyi, for every i{1,,q}, and yi=xi+1 or xi=xi+1 and yi=yi+1 (if i=q+1, then xi+1=x1 and yi+1=y1). We will consider two cases.

Case 1. xi = xj and yi = yj , for every pair of distinct elements, i and j, in {1,,q}.

It follows from the construction of P that P=(x1,e1,,eq,y1). Since C is a cycle in L2H(G), then q2 and, so P is closed.

Case 2. There exists i in {1,,q} such that yi=xi+1.

If yq=x1, then c(eq)c(e1)E(H). Therefore, P is closed.

Otherwise, eq and e1 are parallel and, so P is closed.

Therefore, P is a closed dynamic H-trail in G. □

It follows from Lemmas 4 and 5 the following result.

Theorem 6.

Let G be an H-colored multigraph and MJ be the joint matching. Then, there is a bijection between the set of closed dynamic H-trails in G and the set of cycles in L2H(G) that alternate edges between E(L2H(G))MJ and MJ .

Proof.

Let G be an H-colored multigraph and MJ the joint matching of L2H(G). We will denote by P the set of closed dynamic H-trails in G and by C the set of cycles in L2H(G) that alternate edges between E(L2H(G))MJ and MJ .

Let T:PC defined by T(P) = C, where P=(x0,e1,,ep0,x1,ep0+1,,ep1,x2,ep1+1,,xn1,ep0++pn2+1,,ep0++pn1,xn), and C=(f(x0,e1),f(x1,e1),,f(x0,ep0),f(x1,ep0),f(x1,ep0+1),,f(xn,ep0++pn1),f(x0,e1)).

Claim 1. T is well-defined.

It follows from Lemma 4 that T(P)C.

Let P1=(x0,e0,,ep0,x1,ep0+1,,ep0+p1,x2,,xn1,ep0++pn2+1,,ep0++pn1,xn) and P2=(y0,f0,,fq0,y1,fq0+1,,fq0+q1,y2,,yn1,fq0++qn2+1,,fq0++qn1,yn) in P such that P1 = P2.

Since P1 = P2, then E(P1)=E(P2) and there exists kN such that ei=fi+k(modp0++pn1) (since the edges are traversed in the same order but the first edge is not the same). It follows from the definition of T that V(T(P1))=V(T(P2)) and the order of the vertices are the same. Then, we have that T(P1)=T(P2). Therefore, T is well-defined.

Claim 2. T is injective.

Let P1 and P2 in P such that P1P2.

If E(P1)E(P2), then V(T(P1))V(T(P2)) and, so T(P1)T(P2). Otherwise, since P1P2, then there exists {ei,ei+1}E(P1)=E(P2) such that ei is the edge preceding ei+1 in P1 and ei is not the edge preceding ei+1 in P2. Therefore, T(P1)T(P2) and T is injective.

Claim 3. T is surjective.

Let C be a cycle in C. Since C alternate edges between E(L2H(G))MJ and MJ , C must be of the form C=(f(x1,e1),f(y1,e1),,f(xq,eq),f(yq,eq),f(x1,e1)), where ei=xiyiE(G).

It follows from the definition of T that T(P) = C, where P be the closed dynamic H-trail obtained by applying the Lemma 5 to the cycle C. Hence, T is surjective.

Therefore, T is a bijection. □

Recall the following classical lemma, which will be useful for what follows.

Lemma 7.

Let M and M be two perfect matching of a graph G. If MM=, then there exists a partition of the vertices of G into even cycles. Moreover, every cycle alternate edges between M and M.

Theorem 8.

Let G be an H-colored multigraph and MJ be the joint matching of L2H(G). There exists a perfect matching M in L2H(G)MJ if and only if there exists a partition of the edges of G into closed dynamic H-trails. (Notice that some of the closed dynamic H-trails can be of the form (x,e0,,ek,y), where k1).

Proof.

Let G be an H-colored multigraph and MJ be the joint matching of L2H(G).

Suppose that there exists a perfect matching M in L2H(G)MJ.

It follows by Lemma 7, that there exists a partition of the vertices of L2H(G) into vertex-disjoint cycles, say C={C1,,Cn}. Moreover, each cycle in C alternate edges between MJ and M.

Let P={P1,,Pn} be the set of closed dynamic H-trails in G such that T(Pi)=Ci, for every i{1,,n}.

Claim 1 E(Pi)E(Pj)=, for every {i,j}{1,,n}.

Proceeding by contradiction, suppose that there exists a pair of distinct elements, i and j in {1,,n}, such that E(Pi)E(Pj).

Let e=xyE(Pi)E(Pj), it follows from the definition of T that f(x, e) and f(y, e) are vertices in V(Ci)V(Cj), a contradiction.

Claim 2 i=1nE(Pi)=E(G).

By the definition of T, we have that i=1nE(Pi)E(G).

On the other hand, let eE(G) join x and y. Since, C is a partition of the vertices of L2H(G), then there exists a cycle in C, say Cj , such that f(x,e)f(y,e)E(Cj). It follows from the construction of Pj that eE(Pj). Therefore, ei=1nE(Pi) and E(G)i=1nE(Pi).

Therefore, P={P1,,Pn} is a partition of the edges of G into closed dynamic H-trails.

Conversely, suppose that there exists a partition of the edges of G into closed dynamic H-trails, say P={P1,,Pk}.

Let C={C1,,Ck}a set of cycles in L2H(G) such that T(Pi)=Ci, for every i{1,,k}.

Claim 3 If ij, then V(Ci)V(Cj)=.

Suppose, to the contrary, that there exists i and j in {1,,k}, such that V(Ci)V(Cj). Let f(x,e)V(Ci)V(Cj), for some eE(G) and xV(G), such that e is incident with x.

It follows from the construction of Ci and Cj that eE(Pi) and eE(Pj), a contradiction.

Claim 4 i=1kV(Ci)=V(L2H(G)).

By the construction of Ci , we have that i=1kV(Ci)V(L2H(G)).

Let f(x,e)V(L2H(G)), with eE(G) and xV(G), such that e is incident with x.

Since P is a partition of the edges of G, then there exists PiP such that eE(Pi). It follows from the construction of Ci that f(x,e)V(Ci)j=1kV(Cj). Hence, V(L2H(G))j=1kV(Cj).

Therefore, i=1kV(Ci)=V(L2H(G)).

Claim 5 M=i=1kE(Ci)MJ is a perfect matching in L2H(G).

Recall MJ={f(x,e)f(y,e)E(L2H(G)):e=xyE(G)}.

It follows from the construction of Ci and claims 3 and 4 that M is a perfect matching in L2H(G).

Therefore, M is a perfect matching in L2H(G), such that MMJ=. □

Theorem 9.

Let G be an H-colored multigraph. Then:

  1. If G has a closed Euler dynamic H-trail, then LnH(G) is hamiltonian, for every n2.

  2. If LnH(G) is hamiltonian, for some n3, then G has a closed Euler dynamic H-trail.

  3. G has a closed Euler dynamic H-trail if and only if LnH(G) is hamiltonian, for every n3.

Proof.

Let G be an H-colored multigraph.

  1. It follows from Theorems 6 and 8 and definition of LnH(G).

  2. b. Suppose that there exist n3, such that LnH(G) is hamiltonian. Let C be a hamiltonian cycle in LnH(G).

Then, every path from f(x, e) to f(y, e) is in V(C) in some order, for every e = xy in E(G), by the definition of LnH(G). Hence, we replace the path for the edge f(x,e)f(y,e) to C, and we get a hamiltonian cycle in L2H(G), say C.

Since C is a hamiltonian cycle, then P=T1(C) is a closed Euler dynamic H-trail in G, by Theorem 8.

  1. It follows from a and b. □

In Theorem 9, we cannot change n3 by n2 because there are H-colored multigraphs without closed Euler dynamic H-trail and L2H(G) is hamiltonian, see .

Fig. 2 The graph L2H(G) is hamiltonian but there is no closed Euler dynamic H-trail in G.

Fig. 2 The graph L2H(G) is hamiltonian but there is no closed Euler dynamic H-trail in G.

Recall that Exy is the set of all the edges with end vertices x and y.

Definition 2.

Let G be an H-colored multigraph and M be a matching in L2H(G). For each x in V(G), we will say that Mx={f(x,e)f(x,g)M}, i.e., Mx consists of the edges in Gx that are in M. And, for every pair of vertices x and y in V(G), such that Exy, we will define the following sets: AxyM={eExy:f(x,e) is Mx -saturate and f(y, e) is not My -saturate }; BxyM={eExy:f(y,e) is My -saturate and f(x, e) is not Mx -saturate }; CxyM={eExy:f(x,e) is Mx -saturate and f(y, e) is My -saturate }; and DxyM={eExy:f(x,e) is not Mx -saturate and f(y, e) is not My -saturate }.

Theorem 10.

Let G be an H-colored multigraph. There exists a partition of the edges of G into closed dynamic H-trails, none of them in the form (x,a1,,al,y),l2, if and only if there exists a non empty matching, say M, in L2H(G), such that for each set Exy one of the following conditions hold: a) CxyM=Exy; b) |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

Proof.

Let G be an H-colored multigraph and MJ be the joint matching of L2H(G).

Suppose that P={P1,,Pk} is a partition of the edges of G into closed dynamic H-trails, such that none of them have the form (x,a1,,al,y),l2.

By Theorem 8, there exists a perfect matching in L2H(G)MJ. Let M be the perfect matching obtained as in the proof of Theorem 8, i.e., M=i=1kE(T(Pi))MJ.

Let x and y be a pair of vertices in G, such that Exy={e1,,eq}, with q1. Since P is a partition of the edges of G into closed dynamic H-trails, then there exists PiP such that ExyE(Pi), say Pi=P1.

Since P1 is not of the form (x,a1,,al,y),l2, and ExyE(P1), then P1=(,f0,x,et1,,etp1,y,f1,), where p1q and {f0,f1}E(G). (Recall that Mx consists of the edges in Gx that are in M).

If p1=1, then P1=(,f0,x,et1,y,f1,). Hence, f(x,f0)f(x,et1) and f(y,et1)f(y,f1) are in M.

Therefore, f(x,f0)f(x,et1)Mx,f(y,et1)f(y,f1)My and et1CxyM.

On the other hand, if p12, then f(x,f0)f(x,et1), f(y,etp1)f(y,f1) and f(y,eti)f(x,eti+1) are in M, for every i{1,,p11}, see . Hence, et1AxyM, etp1BxyM and etiDxyM, for every i{2,,p11}.

Fig. 3 In b, the continued edges are in M and the dashed edges are in MJ .

Fig. 3 In b, the continued edges are in M and the dashed edges are in MJ .

If Exy{et1,,etp1}=, then CxyM=Exy (when p1=1 occurs) or |CxyM|<|Exy| and 1|AxyM|=|BxyM| (when p12). So, suppose that E1=Exy{et1,,etp1}, then there exists PP, such that E(P)E1.

Since P is not of the form (x,a1,,al,y),l2, and E1E(P), then P=(,f2,x,es1,,esp2,y,f3,), where {f2,f3}E(G). We will consider two cases.

If p2=1, then P=(,f2,x,es1,y,f3,). By the construction of M we have that the edges f(x,f2)f(x,es1) and f(y,es1)f(y,f3) are in M. Thus, f(x,f2)f(x,es1)Mx,f(y,es1)f(y,s1)My and es1CxyM.

On the other hand, if p22, then f(x,f2)f(x,es1),f(y,esp2)f(y,f3) and f(y,esi)f(x,esi+1) are in M, for every i{1,,s21}. Hence, es1AxyM,esp2BxyM and esiDxyM, for every i{2,,p21}.

If E2=E1{es1,,esp2}=, then CxyM=Exy (when p1=p2=1) or |CxyM|<|Exy| and 1|AxyM|=|BxyM| (when p12 or p22). Otherwise, there exists PP, such that E2E(P) and we can repeat this procedure and after a finite number of steps we obtain that CxyM=Exy or |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

Now, suppose that there exists M, a matching in L2H(G), such that for every set Exy, it holds that: CxyM=Exy or |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

If M is a perfect matching in L2H(G)MJ, then there exists a partition of E(G) into closed dynamic H-trails, by Theorem 8. So, suppose that M is not a perfect matching in L2H(G)MJ. Recall that Mx=ME(Gx), for every xV(G).

Since CxyM=Exy or 1|AxyM|, then for every xV(G),Mx.

Let N=xV(G)Mx. Notice that N is a matching in L2H(G)MJ,AxyM=AxyN,BxyM=BxyN,CxyM=CxyN and DxyM=DxyN. So, for every set Exy, it holds that: CxyN=Exy or |CxyN|<|Exy| and 1|AxyN|=|BxyN|.

Let x and y be two different vertices in V(G), such that Exy.

If CxyN=Exy, then f(x, e) and f(y, e) are N-saturate, for every eExy. So, we do not add any edge to N.

On the other hand, if |CxyN|<|Exy| and 1|AxyN|=|BxyN|, then we will consider Gxy , the subgraph of L2H(G) induced by the set {f(x,e):eExyCxyN}{f(y,e):eExyCxyN}.

Since |CxyN|<|Exy|, then V(Gxy). Let AxyN={e1,,et} and BxyN={f1,,ft}, since 1|AxyN|=|BxyN| we have that t1.

If DxyN=, then we add the edges f(y,ei)f(x,fi) to N, for every i{1,,t}. By the definition of the edges of L2H(G) (point b) and the definition of Gxy , we have that f(y,ei)f(x,fi)E(Gxy). Moreover, N{f(y,ei)f(x,fi):i{1,,t}} is still a matching.

On the other hand, when DxyN, let DxyN={g1,,gs}. We add the edges f(y,e1)f(x,g1), f(x,f1)f(y,gs),f(y,ei)f(x,fi) and f(y,gj)f(x,gj+1) to N, for every i{2,,t} and j{1,,s1}.

By the definition of the edges of L2H(G) and the definition of Gxy , we have that all the edges we add to N are in E(Gxy) and N is still a matching.

Thus, we add edges to N in such a way every vertex in Gxy is N-saturate.

So, we repeat the same process for x and y in V(G) such that Exy.

Therefore, N is a perfect matching in L2H(G)MJ and there exists a partition of E(G) into closed dynamic H-trails, by Theorem 8 and, by the construction of N, none of them is of the form (x,a1,,al,y), with l2. □

The following result shows a condition on the graph Gx that guarantees the existence of closed Euler dynamic H-trail.

Theorem 11.

Let G be a connected H-colored multigraph, such that Gu is a complete ku -partite graph for every uV(G) and for some ku in N{1}. Then G has a closed Euler dynamic H-trail if and only if there exists a matching M in L2H(G), such that for every set Exy one of the following conditions hold: a) CxyM=Exy; or b) |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

Proof.

Let G be a connected H-colored multigraph, such that Gu is a complete ku -partite graph for every uV(G) and for some ku in N{1}.

Suppose that G has a closed Euler dynamic H-trail, say P.

Case 1. P is not of the form (x,e1,,ek,y), with k2.

In this case there exists a matching, say M, in L2H(G), such that for every set Exy one of the following conditions hold: CxyM=Exy or |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

Case 2. P=(x,e1,,ek,y), with k2.

In this case, since Gx is a complete kx -partite graph, for some kx2, then there exists two edges e and f in E(G)=E(P), such that c(e)c(f)E(H). Suppose, without loss of generality, that c(e1)c(e2)E(H).

If k = 2, then P=(x,e1,y,e2,x) is a closed Euler dynamic H-trail. Otherwise, since Gx is kx -partite graph, we have that c(e1)c(ek)E(H) or c(e2)c(ek)E(H).

Hence, P=(x,e1,y,e2,,ek,x) or P=(x,e1,e3,,ek,y,e2,x) is a closed Euler dynamic H-trail. By Theorem 10, there exists a matching M in L2H(G), such that for every set Exy,CxyM=Exy or |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

Conversely, suppose that there exists a matching M in L2H(G), such that for every set Exy one of the following conditions hold: CxyM=Exy or |CxyM|<|Exy| and 1|AxyM|=|BxyM|.

It follows from Theorem 10 that there exists a partition of E(G) into closed dynamic H-trails, none of them in the form (x,e1,,el,y), with l2, say P={P1,,Pk}.

If k = 1, then E(P1)=E(G) and P1 is a closed Euler dynamic H-trail. Otherwise, since G is connected it follows that there exists PiP{P1}, say P2, such that V(P1)V(P2). Since P1 and P2 are closed, we can suppose that P1=(x0,e1,,em,x0), where e1=x0x1 and em=xk1x0, and P2=(y0,f1,,fn,y0), where f1=y0y1 and fn=yk2y0, such that x0 = y0. By the definition of dynamic H-trail, we have that c(e1)c(em) and c(f1)c(fn) are in E(H). Furthermore, e1,f1,em and fn are incident with x0 = y0. Hence, e1em and f1fn are in E(Gx0).

We will consider the following: if {e1,f1}Ejx0 or {em,fn}Ejx0, for some j{1,,kx0}, it follows that e1fn and emf1 are in E(Gx0), because Gx0 is a complete kx0-partite graph. In other case, e1f1 and emfn are in E(Gx0).

By the definition of Gx0, we have that c(e1)c(fn) and c(em)c(f1) are edges of H or c(e1)c(f1) and c(em)c(fn) are edges of H.

If c(e1)c(fn) and c(em)c(f1) are edges of H, we have that Q1=P1P2 is a closed dynamic H-trail. And, if c(e1)c(f1) and c(em)c(fn) are edges of H, we have that Q1=P1P21 is a closed dynamic H-trail.

If E(Q1)=E(G), then Q1 is a closed Euler dynamic H-trail. Otherwise, since G is connected it follows that there exists PiP{P1,P2}, say P3, such that V(Q1)V(P3). Since Q1 and P3 are closed, we can suppose that P1=(v1,a1,,as,v0), where a1=v0v1 and as=vk3v0, and P3=(u0,g1,,gt,u0), where g1=u0u1 and gt=uk3u0, such that v0 = u0. By the definition of dynamic H-trail, we have that c(a1)c(as) and c(g1)c(gt) are in E(H). Furthermore, a1,g1,as and gt are incident with u0 = v0. Hence, a1as and g1gt are in E(Gv0).

We will consider the following: if {a1,g1}Ejv0 or {as,gt}Ejv0, for some j{1,,kv0}, it follows that a1gt and asg1 are in E(Gv0), because Gv0 is a complete kv0-partite graph. In other case, a1g1 and asgt are in E(Gv0).

By the definition of Gv0, we have that c(a1)c(gt) and c(as)c(g1) are edges of H or c(a1)c(g1) and c(as)c(gt) are edges of H.

If c(a1)c(gt) and c(as)c(g1) are edges of H, we have that Q2=Q1P3 is a closed dynamic H-trail. And, if c(a1)c(g1) and c(as)c(gt) are edges of H, we have that Q2=Q1P31 is a closed dynamic H-trail.

If E(Q2)=E(G), then Q2 is a closed Euler dynamic H-trail. Otherwise, we repeat the procedure until we get the desired closed Euler dynamic H-trail. □

In Theorem 11 we cannot change the hypothesis “Gu is a complete ku -partite for every uV(G)” to “Gu has a complete ku -partite spanning subgraph”, as shows.

Fig. 4 Gu is a complete ku -partite graph, for uV(G){x4}, and Gx4 has a bipartite spanning subgraph but there is no closed Euler dynamic H-trail in G.

Fig. 4 Gu is a complete ku -partite graph, for u∈V(G)∖{x4}, and Gx4 has a bipartite spanning subgraph but there is no closed Euler dynamic H-trail in G.

Corollary 12.

Let G be an H-colored multigraph. Then Gx has a perfect matching, for every xV(G), if and only if there exists a partition of E(G) into closed H-trails.

Proof.

Let G be an H-colored multigraph and MJ be the joint matching of L2H(G).

Suppose that Gx has a perfect matching, for every xV(G), say Mx . Let M=xV(G)Mx.

It follows from the definition of MJ that MMJ=. Hence, G has a partition of its edges into closed dynamic H-trails, by Theorem 8.

Let P be the partition that is obtained as in the proof of Theorem 8. It follows, from the construction of each dynamic H-trail in P, that there is no lane change. So, P is a partition of E(G) into closed H-trails.

Now, suppose that there exists a partition of E(G) into closed H-trails, say P. By Theorem 8, there exists a perfect matching in L2H(G)MJ. Let M be the perfect matching that is obtained as in the proof of Theorem 8.

Let f(x,e)V(L2H(G)). Then, there exists PP such that eE(P). Since P is an H-trail and by the construction of M, it follows that P=(u,g,x,e,v,) and f(x,g)f(x,e)M. Moreover, f(x,g)f(x,e)ME(Gx).

Therefore, ME(Gx) is a perfect matching of Gx . □

Recall the following classical theorem, which will be useful.

Theorem 13

(Tutte [Citation21]). G has a perfect matching if and only if o(GS)|S| for all proper subset S of V(G).

Notice that if H is a complete graph without loops and G is an H-colored multigraph. Then P is a properly colored closed trail if and only if P is a closed H-trail.

Corollary 14

(Kotzig). Let G be an edge-colored eulerian multigraph. Then G has a properly colored closed Euler trail if and only if δi(x)jiδj(x), where δi(x) is the number of edges with color i incident with x, for each vertex x of G.

Proof.

Let G be an H-colored multigraph, where H is a complete graph without loops and |V(H)|=c.

Let Gu , with uV(G). It follows from the definition of Gu and H is a complete graph that Gu is a complete ku -paritite graph with independent sets Eiu={eE(G) : e=uxforsomexV(G)andc(e)=i}, with i{1,,c}. Moreover, |Eiu|=δi(u).

Suppose that G has a closed Euler H-trail. By Corollary 12 Gu has a perfect matching, for every uV(G).

Hence, δi(u)=|Eiu|=o(GS)|S|=ji|Eju|=jiδi(u), where S=V(Gu)Eiu, by Theorem 13.

Suppose that |Eiu|=δi(x)jiδj(x)=ji|Eju|, for each vertex x of G.

Claim 1. Gu has a perfect matching.

Let S be a proper subset of V(Gu). Consider GuS.

Case 1. GuS is connected.

Then o(GuS)1|S|.

Case 2. GuS is disconnected.

Since Gu is a complete ku -partite graph, then V(GuS)Eiu, for some i{1,,c}. Hence o(GuS)=|V(GuS)||Eiu|ji|Eju||S|.

Therefore, it follows from Theorem 13 that Gu has a perfect matching.

Let M=uV(G)Mu, where Mu is a perfect matching in Gu . So, M is a perfect matching in L2H(G) and CxyM=Exy, for every Exy.

It follows from Theorem 11 and Corollary 12 that G has a closed Euler H-trail. □

Corollary 15

(Galeana-Sánchez et al.). Let H be a graph possibly with loops and G be an H-colored multigraph without loops. Suppose that G is Eulerian and Gu is a complete ku -partite graph, for every u in V(G) and for some ku in N. Then G has a closed Euler H-trail if and only if |Ciu|ji|Cju| for every u in V(G), where {C1u,,Ckuu} is the partition of V(Gu) into independent sets.

Acknowledgments

The authors wish to thank the anonymous referees for many suggestions which improved the rewriting of this paper.

Disclosure statement

The authors report there are no competing interests to declare

Additional information

Funding

This research was supported by CONACYT FORDECYT-PRONACES/39570/2020; UNAM DGAPA-PAPIIT IN102320. The second author received a fellowship from CONACYT under Grant 782239.

References

  • Ahuja, S. K. (2010). Algorithms for routing and channel assignment in wireless infrastructure networks. PhD thesis. University of Arizona, Arizona, USA.
  • Bang-Jensen, J., Bellitto, T., Yeo, A. (2021). Supereulerian 2-edge-coloured graphs. Graphs Comb. 37(6): 2601–2620.
  • Bang-Jensen, J., Gutin, G. Z. (2008). Digraphs: Theory, Algorithms and Applications. London: Springer.
  • Benítez-Bobadilla, G.,Galeana-Sánchez, H., Hernández-Cruz, C. (2019). Characterization of color patterns by dynamic H-paths. Discrete Appl. Math. 267:41–51.
  • Bondy, J. A., Murty, U. S. R. (1976). Graph Theory with Applications. London, UK: Macmillan.
  • Chou, W., Manoussakis, Y., Megalakaki, O., Spyratos, M., Tuza, Z. (1994). Paths through fixed vertices in edge-colored graphs. Math. Inform. Sci. Humaines 127:49–58.
  • Delgado-Escalante, P., Galeana-Sánchez, H. (2014). Restricted domination in arc-colored digraphs. AKCE Int. J. Graphs Comb. 11(1): 95–104.
  • Delgado-Escalante, P., Galeana-Sánchez, H., Ramírez, L. P. (2012). Independent restricted domination and the line digraph. AKCE Int. J. Graphs Comb. 9(1): 31–42.
  • Dorninger, D. (1994). Hamiltonian circuits determining the order of chromosomes. Discrete Appl. Math. 50(2): 159–168.
  • Dorninger, D., Timischl, W. (1987). Geometrical constraints on Bennett’s predictions of chromosome order. Heredity 58: 321–325.
  • Galeana-Sánchez, H., Rojas-Monroy, R., Sánchez-López, R., Villarreal-Valdés, J. I. (2019). Some conditions for the existence of Euler H-trails. Graphs Comb. 35(5): 1197–1208.
  • Galeana-Sánchez, H., Sánchez-López, R. (2013). H-kernels in infinite digraphs. Graphs Comb. 29(4): 913–920.
  • Guo, Z., Broersma, H., Li, B., Zhang, S. (2020). Almost Eulerian compatible spanning circuits in edge-colored graphs. Discrete Math. 344(1): 112174.
  • Guo, Z., Li, B., Li, X., Zhang, S. (2020). Compatible spanning circuits in edge-colored graphs. Discrete Math. 343(7): 111908.
  • Harary, F., Nash-Williams, C. S. J. (1965). On Eulerian and Hamiltonian graphs and line graphs. Can. Math. Bull. 8(6): 701–709.
  • Kotzig, A. (1968). Moves without forbidden transitions in a graph. Math. Slovaca 18(1): 76–80.
  • Linek, V., Sands, B. (1996). A note on paths in edge-coloured tournaments. Ars Combin. 44: 225–228.
  • Sankararaman, S., Efrat, A., Ramasubramanian, S., Agarwal, P. K. (2014). On channel-discontinuity-constraint routing in wireless networks. Ad hoc Networks 13: 153–169.
  • Szachniuk, M., De Cola, M. C., Felici, G., Blazewicz, J. (2014). The orderly colored longest path problem–a survey of applications and new algorithms. RAIRO Oper. Res. 48(1): 25–51.
  • Szachniuk, M., Popenda, M., Adamiak, R. W., Blazewicz, J. (2009). An assignment walk through 3D NMR spectrum. Proceedings of the 2009 IEEE Symposium on Computational Intelligence in Bioinformatics and Computational Biology. Nashville, TN, USA: IEEE, pp. 215–219.
  • Tutte, W. T. (1947). The factorization of linear graphs. J. London Math. Soc. 1(2): 107–111.