The λ-fold Spectrum Problem for the Orientations of the 6-Cycle

Peter Adams1, Ryan C. Bunge2, Jordan Dulowski2, Saad I. El-Zanati2, Maddillon Kenney2, Ugur Odabası3, Kaitlyn Zale2
1The University of Queensland, QLD 4072, Australia
2Illinois State University, Normal, IL 61790-4520, USA
3Istanbul University-Cerrahpasa, Istanbul, 34320, Turkey

Abstract

The λ-fold complete symmetric directed graph of order v, denoted λKv*, is the directed graph on v vertices and λ directed edges in each direction between each pair of vertices. For a given directed graph D, the set of all v for which λKv* admits a D-decomposition is called the λ-fold spectrum of D. In this paper, we settle the λ-fold spectrum of each of the nine non-isomorphic orientations of a 6-cycle.

Keywords: Spectrum problem, Directed graph, Directed cycl

1. Introduction

If a and b are integers with ab, we let [a,b] denote the set {a,a+1,,b}. For a graph (or directed graph) D, we use V(D) and E(D) to denote the vertex set of D and the edge set (or arc set) of D, respectively. Furthermore, we use DλD to denote the multigraph (or directed multigraph) with vertex set V(D) and λ copies of each edge (or arc) in E(D). For a simple graph G, we use G to denote the symmetric digraph with vertex set V(G)=V(G) and arc set E(G)={u,v}E(G){(u,v),(v,u)}. Hence, λKv is the λ-fold complete symmetric directed graph of order v.

A decomposition of a directed multigraph K is a collection Δ={D1,D2,,Dt} of subgraphs of K such that each directed edge, or arc, of K appears in exactly one DiΔ. If each Di in Δ is isomorphic to a given digraph D, the decomposition is called a D-decomposition of K. A D-decomposition of K is also known as a (K,D)-design. The set of all v for which Kv admits a D-decomposition is called the spectrum of D. Similarly, the set of all v for which KλKv admits a D-decomposition is called the λ-fold spectrum of D.

The reverse orientation of D, denoted RevD, is the digraph with vertex set V(D) and arc set {(v,u):(u,v)E(D)}. We note that the existence of a D-decomposition of K necessarily implies the existence of a RevD-decomposition of RevK. Since Kv is its own reverse orientation, we note that the spectrum of D is equal to the spectrum of RevD.

The necessary conditions for a digraph D to decompose KλKv include

  • (a) ]|V(D)|v,

  • (b) ]|E(D)| divides λv(v1), and

  • (c) ]gcd{outdegree(x):xV(D)} and gcd{indegree(x):xV(D)} both divide λ(v1).

The spectrum problem for certain subgraphs (both bipartite and non-bipartite) of K4 has already been studied. When D is a cyclic orientation of K3, then a (Kv,D)-design is known as a Mendelsohn triple system. The spectrum for Mendelsohn triple systems was found independently by Mendelsohn [1]and Bermond [2]. When D is a transitive orientation of K3, then a (Kv,D)-design is known as a transitive triple system. The spectrum for transitive triple systems was found by Hung and Mendelsohn [3]. There are exactly four orientations of a 4-cycle (i.e., a quadrilateral). It was shown in [4]that if D is a cyclic orientation of a 4-cycle, then a (Kv,D)-design exists if and only if v0 or 1(mod4) and v4. The spectrum problem for the remaining three orientations of a 4-cycle were setled in [5]. In [6], Alspach et al. showed that Kv can be decomposed into each of the four orientations of a 5-cycle (i.e., a pentagon) if and only if v0 or 1(mod5). In [7], it is shown that for positive integers m and v with 2mv the directed graph Kv can be decomposed into directed cycles (i.e., with all the edges being oriented in the same direction) of length m if and only if m divides the number of arcs in Kv and (v,m){(4,4),(6,3),(6,6)}. Also recently [8], Odabaşı settled the spectrum problem for all possible orientations of a 7-cycle.

There are nine non-isomorphic orientations of a 6-cycle. We denote these with D1, D2, …, D9 as seen in Figure 1. The λ-fold spectrum problem was settled for the directed 6-cycle (i.e., D1) in [9]. In this work, we settle this problem for the remaining eight orientations. Our main result, which is proved in Section 3, is as follows.

Theorem 1. Let D be an orientation of a 6-cycle and let λ and v be positive integers such that v6. There exists a D-decomposition of λKv if and only if λv(v1)0(mod3) and neither of the following hold

  • (D,λ,v)=(D1,1,6) or

  • D=D9 and λ(v1) is odd.

From the necessary conditions stated earlier, we have the following.

Lemma 1. Let D{D2,D3,,D8} and let λ and v be positive integers such that v6. There exists a D-decomposition of λKv only if λv(v1)0(mod3). Furthermore, there exists a D9-decomposition of λKv only if λv(v1)0(mod3) and λ(v1)0(mod2).

In 1978, Bermond, Huang, and Sotteau [9]showed that with the exception that there is no D1-decomposition of K6, these necessary conditions are sufficient for D1.

Theorem 2. For integers v6 and λ1, there exists a D1-decomposition of KλKv if and only if λv(v1)0(mod6) and (λ,v)(1,6).

The remainder of this paper is dedicated to establishing sufficiency of the above necessary conditions. We achieve this by exhibiting constructions for the desired decompositions (see Section 3) using certain small examples (see Section 2). Henceforth, each of the graphs in Figure 1, with vertices labeled as in the figure, will be represented by Di[v1,v1,,v6].

For m2, the following result of Sotteau proves the existence of 2m-cycle decompositions of complete bipartite graphs.

Theorem 3 ([10]). Let x, y, and m be positive integers such that m2. There exists a 2m-cycle decomposition of K2x,2y if and only if m2xy and min{2x,2y}m.

Consider an orientation of a 6-cycle that is isomorphic to its own reverse, i.e. any Di in Figure 1 such that i{7,8}. By definition of reverse orientation, the set {Di,RevDi} is an obvious Di-decomposition of C6 (the symmetric digraph with a 6-cycle as the underlying simple graph). Since a G-decomposition of a graph K necessarily implies a G-decomposition of the digraph K, we get the following corollary from the case m=3 in Theorem 3.

Corollary 1. Let D{D1,D2,D3,D4,D5,D6,D9}. There exists a D-decomposition of K2x,2y if 3xy and min{x,y}2.

2. Examples of Small Designs

We first present several Di-decompositions of various graphs for i[2,9]. Beyond establishing existence of necessary base cases, these decompositions are used extensively in the general constructions seen in Section 3.

If i,v1,v2,,v6 are integers and D{D1,D2,,D9}, we define D[v1,v2,,v6]+i to indicate D[v1+i,$$v2+i,$$,$$v6+i]. Similarly, if the vertices of D are ordered pairs in Zm×Zn, then D[(u1,v1),$$(u2,v2),$$,$$(u6,v6)]+(i,0) means the digraph D[(u1+i,v1),$$(u2+i,v2),$$,$$(u6+i,v6)]. We also use the convention that both +i and +(i,0) result in simply .

Example 1. Let V(K6)=Z5{} and let Δ2={D2[0,3,4,2,1,]+i:iZ5},Δ3={D3[0,4,1,3,2,]+i:iZ5},Δ4={D4[0,1,2,4,,3]+i:iZ5},Δ5={D5[0,1,3,2,,4],D5[1,3,4,,2,0],D5[2,4,1,,3,0],D5[4,1,2,3,,0],D5[3,0,,1,2,4]},Δ6={D6[0,1,3,2,4,]+i:iZ5},Δ7={D7[0,1,3,4,2,]+i:iZ5},Δ8={D8[0,,1,3,2,4]+i:iZ5}. Then Δi is a Di-decomposition of K6 for i[2,8].

Example 2. Let V(K2K6)=Z5{} and let Δ9=iZ5{D9[0,1,2,3,4,]+i,D9[,0,2,4,1,3]+i}. Then Δ9 is a D9-decomposition of K2K6.

Example 3. Let V(K7)=Z7 and let Δ2={D2[0,1,4,6,5,2],D2[0,4,1,5,3,6],D2[0,5,4,2,6,3],D2[1,6,4,3,2,5],D2[4,0,3,1,6,2],D2[5,0,1,2,3,4],D2[6,0,2,1,3,5]},Δ3={D3[3,1,0,6,2,4],D3[4,5,1,0,3,6],D3[2,0,6,5,3,1],D3[1,6,5,2,0,4],D3[0,5,4,2,6,3],D3[5,3,2,1,4,0],D3[6,4,3,2,5,1]},Δ4={D4[0,1,3,2,6,4]+i:iZ7},Δ5={D5[0,1,3,6,5,2]+i:iZ7},Δ6={D6[0,1,2,3,4,5],D6[0,2,1,3,5,6],D6[0,3,1,6,2,4],D6[3,2,4,5,1,6],D6[3,4,6,0,2,5],D6[5,1,4,0,3,6],D6[6,2,5,0,1,4]},Δ7={D7[0,1,3,5,2,6]+i:iZ7},Δ8={D8[0,6,2,5,3,1]+i:iZ7},Δ9={D9[0,1,2,4,6,3]+i:iZ7}. Then Δi is a Di-decomposition of K7 for i[2,9].

Example 4. Let V(K3K8)=Z7{} and let D1[0,1,2,3,5,4],D1[0,2,1,3,6,],D1[0,3,1,5,2,],Δ2=iZ7{D2[0,1,2,3,5,4]+i,D2[0,2,1,3,6,]+i,D2[0,3,1,5,2,]+i,D2[0,4,2,3,5,]+i},Δ3=iZ7{D3[0,1,2,3,5,6]+i,D3[0,2,3,5,1,]+i,D3[0,3,1,6,4,]+i,D3[0,4,1,5,2,]+i},Δ4=iZ7{D4[0,1,2,3,5,6]+i,D4[0,2,1,3,6,]+i,D4[0,3,1,5,2,]+i,D4[0,5,2,6,1,]+i},Δ5=iZ7{D5[0,1,2,3,5,4]+i,D5[0,2,1,3,6,]+i,D5[0,3,5,2,6,]+i,D5[0,4,6,3,2,]+i},Δ6=iZ7{D6[0,1,2,3,4,5]+i,D6[0,2,4,3,6,]+i,D6[0,3,5,2,6,]+i,D6[0,4,2,5,1,]+i},Δ7=iZ7{D7[0,1,2,3,5,6]+i,D7[0,2,3,5,1,]+i,D7[0,3,1,2,4,]+i,D7[0,4,1,5,2,]+i},Δ8=iZ7{D8[0,1,2,3,4,5]+i,D8[0,2,6,,4,3]+i,D8[0,3,1,,5,2]+i,D8[0,3,1,,5,2]+i}. Then Δi is a Di-decomposition of K3K8 for i[2,8].

Example 5. Let V(K6K8)=Z7{} and let Δ9=iZ7{D9[0,1,2,3,4,5]+i,D9[0,1,2,3,4,5]+i,D9[0,2,6,3,,5]+i,D9[0,2,6,3,,5]+i,D9[0,2,6,3,,5]+i,D9[0,3,1,6,2,]+i,D9[0,3,1,5,6,]+i,D9[0,3,1,5,6,]+i}. Then Δ9 is a D9-decomposition of K6K8.

Example 6. Let V(K9)=(Z4×Z2){}. For brevity we use ij to denote the ordered pair (i,j)V(K9), and we (continue to) use the convention that +i0=. Let Δ2=iZ4{D2[00,30,,31,11,10]+i0,D2[00,20,31,,30,21]+i0,D2[01,31,21,10,11,20]+i0},Δ3=iZ4{D3[00,31,01,30,21,]+i0,D3[00,20,10,,31,01]+i0,D3[01,20,30,31,10,21]+i0},Δ4=iZ4{D4[00,30,20,01,11,31]+i0,D4[01,30,10,,11,00]+i0,D4[01,11,,30,21,20]+i0},Δ5=iZ4{D5[00,30,01,20,21,31]+i0,D5[00,20,10,11,21,]+i0,D5[01,21,10,31,00,]+i0},Δ6=iZ4{D6[00,11,20,,21,31]+i0,D6[,30,00,01,20,31]+i0,D6[01,00,20,10,31,11]+i0},Δ7=iZ4{D7[00,30,01,20,21,31]+i0,D7[00,20,30,31,01,]+i0,D7[01,21,10,31,00,]+i0},Δ8=iZ4{D8[00,,01,21,10,31]+i0,D8[01,30,00,31,21,20]+i0,D8[01,,00,20,30,31]+i0},Δ9=iZ4{D9[00,,21,30,01,31]+i0,D9[00,30,11,31,20,01]+i0,D9[01,00,30,10,,11]+i0}. Then Δi is a Di-decomposition of K9 for i[2,9].

Example 7. Let V(K10)=Z5×Z2. For brevity we use ij to denote the ordered pair (i,j)V(K10). Let Δ2=iZ5{D2[00,11,10,01,41,21]+i0,D2[00,10,41,11,21,30]+i0,D2[01,30,10,21,40,00]+i0},Δ3=iZ5{D3[00,41,11,21,10,40]+i0,D3[01,00,10,30,41,11]+i0,D3[01,10,31,00,21,20]+i0},Δ4=iZ5{D4[00,41,01,20,11,30]+i0,D4[00,20,31,30,01,21]+i0,D4[00,10,20,31,21,01]+i0},Δ5=iZ5{D5[00,40,20,31,41,21]+i0,D5[00,30,21,31,11,10]+i0,D5[00,41,30,11,10,31]+i0},Δ6=iZ5{D6[00,40,21,10,30,01]+i0,D6[00,11,10,01,41,21]+i0,D6[01,41,21,40,30,10]+i0},Δ7=iZ5{D7[00,10,01,31,41,30]+i0,D7[00,21,30,01,20,31]+i0,D7[00,01,41,11,10,40]+i0},Δ8=iZ5{D8[00,10,11,01,21,20]+i0,D8[00,41,11,21,10,40]+i0,D8[01,40,21,00,41,20]+i0}. Then Δi is a Di-decomposition of K10 for i[2,8].

Example 8. Let V(K2K10)=Z10 and let Δ9=iZ10{D9[0,4,8,7,9,1]+i,D9[0,5,2,3,1,4]+i,D9[0,9,2,7,1,8]+i}. Then Δ9 is a D9-decomposition of K2K10.

Example 9. Let V(K3K11)=Z11 and let Δ2=iZ11{D2[0,5,1,2,4,7]+i,D2[0,5,1,2,4,7]+i,D2[0,6,1,2,4,7]+i,D2[0,2,1,10,9,3]+i,D2[0,4,1,6,3,2]+i},Δ3=iZ11{D3[0,2,1,3,6,5]+i,D3[0,2,1,3,6,5]+i,D3[0,2,1,5,9,3]+i,D3[0,1,7,2,10,6]+i,D3[0,3,1,8,4,7]+i},Δ4=iZ11{D4[0,1,7,10,2,6]+i,D4[0,1,7,10,2,6]+i,D4[0,1,7,10,2,6]+i,D4[0,4,6,10,1,2]+i,D4[0,9,10,3,1,2]+i},Δ5=iZ11{D5[0,1,2,5,10,4]+i,D5[0,1,2,5,10,4]+i,D5[0,1,2,5,10,4]+i,D5[0,2,4,1,3,7]+i,D5[0,2,4,1,10,3]+i},Δ6=iZ11{D6[0,1,3,6,10,4]+i,D6[0,1,3,6,10,4]+i,D6[0,1,3,6,10,4]+i,D6[0,5,3,2,4,10]+i,D6[0,10,4,2,5,8]+i},Δ7=iZ11{D7[0,1,3,6,10,7]+i,D7[0,1,3,6,10,7]+i,D7[0,1,3,6,10,7]+i,D7[0,6,5,10,4,9]+i,D7[0,10,4,5,3,9]+i},Δ8=iZ11{D8[0,1,3,6,2,5]+i,D8[0,1,3,6,2,5]+i,D8[0,1,3,6,2,5]+i,D8[0,6,4,8,10,9]+i,D8[0,6,5,9,3,10]+i},Δ9=iZ11{D9[0,1,2,4,6,3]+i,D9[0,1,2,4,6,3]+i,D9[0,1,2,4,6,3]+i,D9[0,5,1,8,2,6]+i,D9[0,5,1,8,2,7]+i}. Then Δi is a Di-decomposition of K3K11 for i[2,9].

Example 10. Let V(K3,4)=Z7 with vertex partition {{0,1,2},$${3,4,5,6}} and let Δ7={D7[0,3,1,4,2,6],D7[3,0,5,1,6,2],D7[2,5,1,6,0,4],D7[5,2,3,1,4,0]},Δ8={D8[0,4,2,5,1,6],D8[3,2,5,0,4,1],D8[1,3,0,6,2,4],D8[6,2,3,0,5,1]}. Then Δi is a Di-decomposition of K3,4 for i{7,8}.

Example 11. Let V(K6,6)=Z6×Z2 with the obvious vertex bipartition. For brevity we use ij to denote the ordered pair (i,j)V(K6,6). Let Δ7=iZ6{D7[00,51,10,11,50,21]+i0,D7[01,50,31,00,11,10]+i0},Δ8=iZ6{D8[00,41,50,11,40,01]+i0,D8[01,00,51,40,21,50]+i0}. Then Δi is a Di-decomposition of K6,6 for i{7,8}.

3. General Constructions

For two edge-disjoint graphs (or digraphs) G and H, we use GH to denote the graph (or digraph) with vertex set V(G)V(H) and edge (or arc) set E(G)E(H). Furthermore, given a positive integer x, we use xG to denote the edge-disjoint union of x copies of G, which are not necessarily vertex-joint. If G and H are vertex-disjoint, then we use GH to denote the join of G and H, which has vertex set V(G)V(H) and edge (or arc) set E(G)E(H){{u,v}:uV(G),vV(H)}. To illustrate the different types of notation described here, consider that K13 can be viewed as (K6K6)K1K6,6=K7K7K6,6. (Note that the join precedes the union in the order of operations.)

We first prove a result about decompositions of K4,6, K6,6, and K6,8.

Lemma 2. For D{D2,D3,,D9}, then there exists a D-decomposition of K4,6, K6,6 and K6,8.

Proof. Let D{D2,D3,,D9}. The result follows from Corollary 1 for D{D7,D8}. For i{7,8}, a Di-decomposition of K3,4 (and hence of K6,4 and K6,8) exists by Example 10. Moreover, D7– and D8-decompositions of K6,6 are given in Example 11. ◻

We now give our constructions for decompositions of λKv in the following lemmas, which cover values of v working modulo 6. The main result is summarized in Theorem 4.

Lemma 3. Let λ and v be positive integers such that v0(mod6). If D{D2,D3,,D8}, then there exists a D-decomposition of λKv. Furthermore, if λ is even, then there exists a D9-decomposition of λKv.

Proof. Let D{D2,D3,,D9}. If v=6 and DD9, then the result follows from λ copies of a D-decomposition of K6 (see Example 1). If v=6, λ is even, and D=D9, then the result follows from λ/2 copies of a D9-decomposition of K2K6 (see Example 2). For the remainder of the proof, we let v=6x for some integer x2, and we assume λ is even whenever D=D9. Finally,

we note that K6x=xK6(x2)K6,6. Thus KλK6x=x(KλK6)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK6 and KλK6,6, where the latter decomposition follows from λ copies of a D-decomposition of K6,6 (see Lemma 2). ◻

Lemma 4. Let λ and v be positive integers such that v1(mod6) and v7. If D{D2,D3,,D9}, then there exists a D-decomposition of λKv.

Proof. If v=7, then the result follows from λ copies of a D-decomposition of K7 (see Example 3). For the remainder of the proof, we let v=6x+1 for some integer x2. We note that K6x+1=(xK6)K1(x2)K6,6=xK7(x2)K6,6. Thus KλK6x+1=x(KλK7)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK7 and KλK6,6◻

Lemma 5. Let λ and v be positive integers such that λ0(mod3), v2(mod6), and v8. If D{D2,D3,,D8}, then there exists a D-decomposition of λKv. Furthermore, if λ0(mod6), then there exists a D9-decomposition of λKv.

Proof. Let D{D2,D3,,D9}. If v=8 and DD9, then the result follows from λ/3 copies of a D-decomposition of K3K8 (see Example 4). If v=8, λ0(mod6), and D=D9, then the result follows from λ/6 copies of a D9-decomposition of K6K8 (see Example 5).

Next, for v=14, we note that KλK14=KλK8KλK6KλK8,6. Thus the result follows from the existence of D-decompositions of KλK8, KλK6 and and KλK8,6.

For the remainder of the proof, we let v=6x+8 for some integer x2 and λ=3y for some integer y1, and we assume y is even whenever D=D9. Finally, we note that K6x+8=K8xK6xK8,6(x2)K6,6. Thus KλK6x+8=KλK8x(KλK6)x(KλK8,6)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK8, KλK6, KλK8,6, and KλK6,6◻

Lemma 6. Let λ and v be positive integers such that v3(mod6) and v9. If D{D2,D3,,D9}, then there exists a D-decomposition of λKv.

Proof. If v=9, then the result follows from λ copies of a D-decomposition of K9 (see Example 6). For v=15, we note that KλK15=(KλK8KλK6)KλK1KλK8,6=KλK9KλK7KλK8,6, and the result follows from the existence of D-decompositions of KλK9, KλK7 (see Lemma 4), and KλK8,6.

For the remainder of the proof, we let v=6x+9 for some integer x2. Finally, we note that K6x+9=(K8xK6)K1xK8,6(x2)K6,6=K9xK7xK8,6(x2)K6,6. Thus KλK6x+9=KλK9x(KλK7)x(KλK8,6)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK9, KλK7, KλK8,6, and KλK6,6◻

Lemma 7. Let λ and v be positive integers such that v4(mod6) and v10. If D{D2,D3,,D8}, then there exists a D-decomposition of λKv. Furthermore, if λ is even, then there exists a D9-decomposition of λKv.

Proof. Let D{D2,D3,,D9}. If v=10 and DD9, then the result follows from λ copies of a D-decomposition of K10 (see Example 7). If v=10, λ is even, and D=D9, then the result follows from λ/2 copies of a D9-decomposition of K2K10 (see Example 8). For the remainder of the proof, we let v=6x+4 for some integer x2, and we assume λ is even whenever D=D9.

Next, we note that K6x+4=K4xK6xK4,6(x2)K6,6=K10(x1)K6(x1)K4,6(x2)K6,6. Thus KλK6x+4=KλK10(x1)(KλK6)(x1)(KλK4,6)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK10, KλK6 (see Lemma 3), KλK4,6, and KλK6,6◻

Lemma 8. Let λ and v be positive integers such that λ0(mod3), v5(mod6), and v11. If D{D2,D3,,D9}, then there exists a D-decomposition of λKv.

Proof. If v=11, then the result follows from λ/3 copies of a D-decomposition of K3K11 (see Example 9). For the remainder of the proof, we let v=6x+5 for some integer x2. Finally, we note that K6x+5=(K4xK6)K1xK4,6(x2)K6,6=K11(x1)K7(x1)K4,6(x2)K6,6. Thus KλK6x+5=KλK11(x1)(KλK7)(x1)(KλK4,6)(x2)(KλK6,6), and the result follows from the existence of D-decompositions of KλK11, KλK7 (see Lemma 4), KλK4,6, and KλK6,6◻

Combining the previous results from Lemmas 3 through 8 with Theorem 2 and Lemma 1, we obtain our main theorem, which we restate here.

Theorem 4. Let D be an orientation of a 6-cycle and let λ and v be positive integers such that v6. There exists a D-decomposition of λKv if and only if λv(v1)0(mod3) and neither of the following hold

  • (D,λ,v)=(D1,1,6) or

  • D=D9 and λ(v1) is odd.

References:

  1. Mendelsohn, N.S., 1971. A generalization of Steiner triple systems. In A.O.L. Atkin & J.B. Birch (Eds.), Computers in Number Theory (pp. 323–339). Academic Press.

  2. Bermond, J.C., 1974. An application of the solution of Kirkman’s schoolgirl problem: The decomposition of the symmetric oriented complete graph into 3-circuits. Discrete Mathematics, 8, pp.301–304.

  3. Hung, H.Y. and Mendelsohn, N.S., 1973. Directed triple systems. Journal of Combinatorial Theory, Series B, 14, pp.310–318.

  4. Schönheim, J., 1975. Partition of the edges of the directed complete graph into 4-cycles. Discrete Mathematics, 11, pp.67–70.

  5. Harary, F., Heinrich, K. and Wallis, W.D., 1978. Decomposition of complete symmetric digraph into the four oriented quadrilaterals. In Springer Lecture Notes in Mathematics (Vol. 686, pp. 165–173).
  6. Alspach, B., Heinrich, K. and Varma, B.N., 1979. Decompositions of complete symmetric digraphs into the oriented pentagons. Journal of the Australian Mathematical Society, Series A, 28, pp.353–361.

  7. Alspach, B., Gavlas, H., Šajna, M. and Verrall, H., 2003. Cycle decompositions IV: Complete directed graphs and fixed length directed cycles. Journal of Combinatorial Theory, Series A, 103, pp.165–208.

  8. Odabaşı, U. The spectrum problem for the orientations of the 7-cycle. Preprint.

  9. Bermond, J.C., Huang, C. and Sotteau, D., 1978. Balanced cycle and circuit design: Even cases. Ars Combinatoria, 5, pp.293–318.

  10. Sotteau, D., 1981. Decomposition of Km,n into cycles (circuits) of length 2k. Journal of Combinatorial Theory, Series B, 30, pp.75–81.