On disjoint union of M-graphs

Given a pair (X,σ) consisting of a finite tree X and its vertex self-map σ one can construct the corresponding Markov graph Γ(X,σ) which is a digraph that encodes σ-covering relation between edges in X. M-graphs are Markov graphs up to isomorphism. We obtain several sufficient conditions for the dis...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Algebra and Discrete Mathematics
Datum:2017
1. Verfasser: Kozerenko, S.
Format: Artikel
Sprache:English
Veröffentlicht: Інститут прикладної математики і механіки НАН України 2017
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/156635
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:On disjoint union of M-graphs / S. Kozerenko // Algebra and Discrete Mathematics. — 2017. — Vol. 24, № 2. — С. 262-273. — Бібліогр.: 10 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-156635
record_format dspace
spelling Kozerenko, S.
2019-06-18T18:10:42Z
2019-06-18T18:10:42Z
2017
On disjoint union of M-graphs / S. Kozerenko // Algebra and Discrete Mathematics. — 2017. — Vol. 24, № 2. — С. 262-273. — Бібліогр.: 10 назв. — англ.
1726-3255
2010 MSC:05C20, 37E25, 37E15.
https://nasplib.isofts.kiev.ua/handle/123456789/156635
Given a pair (X,σ) consisting of a finite tree X and its vertex self-map σ one can construct the corresponding Markov graph Γ(X,σ) which is a digraph that encodes σ-covering relation between edges in X. M-graphs are Markov graphs up to isomorphism. We obtain several sufficient conditions for the disjoint union of M-graphs to be an M-graph and prove that each weak component of M-graph is an M-graph itself.
en
Інститут прикладної математики і механіки НАН України
Algebra and Discrete Mathematics
On disjoint union of M-graphs
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title On disjoint union of M-graphs
spellingShingle On disjoint union of M-graphs
Kozerenko, S.
title_short On disjoint union of M-graphs
title_full On disjoint union of M-graphs
title_fullStr On disjoint union of M-graphs
title_full_unstemmed On disjoint union of M-graphs
title_sort on disjoint union of m-graphs
author Kozerenko, S.
author_facet Kozerenko, S.
publishDate 2017
language English
container_title Algebra and Discrete Mathematics
publisher Інститут прикладної математики і механіки НАН України
format Article
description Given a pair (X,σ) consisting of a finite tree X and its vertex self-map σ one can construct the corresponding Markov graph Γ(X,σ) which is a digraph that encodes σ-covering relation between edges in X. M-graphs are Markov graphs up to isomorphism. We obtain several sufficient conditions for the disjoint union of M-graphs to be an M-graph and prove that each weak component of M-graph is an M-graph itself.
issn 1726-3255
url https://nasplib.isofts.kiev.ua/handle/123456789/156635
citation_txt On disjoint union of M-graphs / S. Kozerenko // Algebra and Discrete Mathematics. — 2017. — Vol. 24, № 2. — С. 262-273. — Бібліогр.: 10 назв. — англ.
work_keys_str_mv AT kozerenkos ondisjointunionofmgraphs
first_indexed 2025-11-24T06:35:33Z
last_indexed 2025-11-24T06:35:33Z
_version_ 1850843125263630336
fulltext “adm-n4” 22:47 page #84 Algebra and Discrete Mathematics RESEARCH ARTICLE Volume 24 (2017). Number 2, pp. 262–273 © Journal “Algebra and Discrete Mathematics” On disjoint union of M-graphs Sergiy Kozerenko Communicated by V. V. Kirichenko Abstract. Given a pair (X, σ) consisting of a finite tree X and its vertex self-map σ one can construct the corresponding Markov graph Γ(X, σ) which is a digraph that encodes σ-covering relation between edges in X. M-graphs are Markov graphs up to isomorphism. We obtain several sufficient conditions for the disjoint union of M-graphs to be an M-graph and prove that each weak component of M-graph is an M-graph itself. Introduction In 1964 Sharkovsky proved the following remarkable theorem. Theorem 1. [9] If the continuous map f : [0, 1] → [0, 1] has a periodic point of period n ∈ N, then it also has a periodic point of period m ∈ N for all m ⊳ n, where 1⊳2⊳22 ⊳· · ·⊳2n ⊳· · ·⊳7·2n ⊳5·2n ⊳3·2n ⊳· · ·⊳7·2⊳5·2⊳3·2⊳· · ·⊳7⊳5⊳3 is Sharkovsky’s ordering of N. Moreover, for every number m ∈ N there exists a continuous map that has a periodic point of period m but does not have periodic points of periods n ∈ N, where m ⊳ n. In [10] Straffin proposed a strategy on how to prove Sharkovsky’s the- orem using some elegant combinatorial arguments. The cornerstone of his idea is to use directed graphs which naturally arise from orbits of periodic 2010 MSC: 05C20, 37E25, 37E15. Key words and phrases: tree maps, Markov graphs, Sharkovsky’s theorem. “adm-n4” 22:47 page #85 S. Kozerenko 263 points. Namely, let x ∈ [0, 1] be n-periodic point of a continuous map f : [0, 1] → [0, 1]. Consider the orbit orbf (x) = {x, f(x), . . . , fn−1(x)} = {x1 < · · · < xn} and its natural ordering inherited from the interval. Periodic graph Gf (x) has the vertex set {1, . . . , n − 1} and the arc set {(i, j) : min{f(xi), f(xi+1)} 6 xj < max{f(xi), f(xi+1)}}. Here each 1 6 i 6 n − 1 represents the minimal interval [xi, xi+1] and there is an arc i → j in Gf (x) if [xi, xi+1] “covers” [xj , xj+1] under f . Periodic graphs are useful in combinatorial dynamics because of the following result known as Itinerary lemma. Lemma 1. [10] Let x ∈ [0, 1] be some periodic point of a continuous map f : [0, 1] → [0, 1]. Suppose that there is a closed walk W = {i0 → · · · → im−1 → i0} of length m in Gf (x). Then there exists a periodic point y ∈ [0, 1] such that fm(y) = y and fk(y) ∈ [xik , xik+1] for all 0 6 k 6 m − 1. Moreover, if W is primitive, then the period of y equals m. Here the closed walk is called primitive if it is not entirely consists of a smaller walk traced several times. Note that Lemma 1 admits a converse statement. Namely, for any periodic point x ∈ [0, 1] of f we can consider its linearization Lx(f) : [0, 1] → [0, 1] which is a “connect- the-dots” map with respect to the orbit orbf (x). Then each m-periodic point of Lx(f) corresponds to some primitive closed walk of length m in GLx(f)(x) = Gf (x). Full proof of Sharkovsky’s theorem using periodic graphs can be found in [2]. Graph-theoretic properties of periodic graphs were studied in [6–8]. These are include calculation of the number of non-isomorphic periodic graphs with given number of vertices [6] and obtaining graph-theoretic criteria for periodic graphs [7] and their induced subgraphs [8]. Similar approach can be used for dynamics of continuous maps on finite topological trees (see [1] for the Sharkovsky-type result in this case). The corresponding digraphs are called Markov graphs. These are can be defined for combinatorial trees and their vertex maps. Thus, periodic graphs appear as a particular case of Markov graphs where underlying trees are paths and maps are cyclic permutations. M-graphs then defined as Markov graphs up to isomorphism. In [3] maps on trees were characterized for several classes of M-graphs including complete digraphs, complete bipartite digraphs, disjoint unions of cycles and digraphs in which each arc is a loop. It is also shown [4] that M-graphs satisfy Seymour’s Second Neighbourhood Conjecture as well as Caccetta–Häggkvist Conjecture. Various transformations including “adm-n4” 22:47 page #86 264 On disjoint union of M-graphs deletion and addition of vertices, doubling and reverse doubling of vertices and taking disjoint unions of M-graphs are studied in [5]. Also, it is proved that there exist exactly 11 pairwise non-isomorphic M-graphs which are tournaments as well as 86 pairwise non-isomorphic 3-vertex M-graphs (again, see [5]). In this paper we obtain several sufficient conditions for the disjoint union of M-graphs to be an M-graph and prove that each weak component of M-graph is an M-graph itself. 1. Definitions and preliminary results In what follows map is just a function. For any given map σ by Im σ and fix σ we denote its image and the set of its fixed points, respectively. A graph G is a pair (V, E), where V = V (G) is the set of its vertices and E = E(G) the set of its edges. By EG(u) we denote the set of all edges incident to the vertex u in G. A vertex u is called isolated if |EG(u)| = 0. Similarly, u is a leaf vertex provided |EG(u)| = 1. The unique edge incident to a leaf vertex is called a leaf edge. The set of all leaf vertices in G is denoted by L(G). For the set of vertices A ⊂ V (G) we put E(A) = {uv ∈ E(G) : u, v ∈ A} and ∂GA = {u ∈ A : EG(u)−E(A) 6= ∅}. By G[A] and G[E′] we denote the subgraphs of G induced by A ⊂ V (G) and E′ ⊂ E(G), respectively. A graph G is called connected if for every pair of its vertices u, v ∈ V (G) there exists a path joining them. The minimum number of edges in such a path is called the distance dG(u, v) between u and v in G. The set of vertices A ⊂ V (G) is connected if the induced subgraph G[A] is connected. Similarly, E′ ⊂ E(G) is connected if so is G[E′]. The eccentricity of a vertex u in a connected graph G is the value eccG(u) = maxv∈V (G) dG(u, v). For the pair of vertices u, v ∈ V (G) in a connected graph G we put [u, v]G = {w ∈ V (G) : dG(u, w) + dG(w, v) = dG(u, v)}. The set A ⊂ V (G) is convex provided [u, v]G ⊂ A for all u, v ∈ A. The convex hull ConvG(A) of A is defined as the smallest convex set containing A. Put dG(u, A) = minv∈A dG(u, v) and dG(A, B) = minu∈B dG(u, A) for all vertex sets {u}, A, B ⊂ V (G) in a connected graph G. The set A ⊂ V (G) is called Chebyshev if for every vertex u ∈ V (G) there exists a unique v ∈ A with dG(u, v) = dG(u, A). The corresponding map prA : V (G) → V (G), where prA(u) = v is called the projection on a Chebyshev set A. “adm-n4” 22:47 page #87 S. Kozerenko 265 A tree is a connected acyclic graph. It should be noted that in a tree each connected set of vertices is Chebyshev. A directed graph or digraph Γ is a pair (V, A), where V = V (Γ) is the set of its vertices and A = A(Γ) ⊂ V ×V is the set of its arcs. If (u, v) ∈ A(Γ), then we write u → v in Γ. The arc of the form u → u is called a loop. For the vertex u ∈ V (Γ) we put N+ Γ (u) = {v ∈ V (Γ) : u → v in Γ} and N− Γ (u) = {v ∈ V (Γ) : v → u in Γ}. The cardinalities d+ Γ (u) = |N+ Γ (u)| and d− Γ (u) = |N− Γ (u)| are called the outdegree and the indegree of u, respectively. A digraph Γ is called complete provided A(Γ) = V (Γ)×V (Γ). Similarly, Γ is empty if A(Γ) = ∅. By Kn and Kn we denote the complete and the empty digraph with n vertices, respectively. A digraph is called weakly connected if its underlying graph (which is obtained by “forgetting” orientation of the edges and ignoring loops) is connected. Weak component of a digraph is its maximal weakly connected subgraph. By Γ1 ⊔ Γ2 we denote the disjoint union of digraphs Γ1 and Γ2. A pair (X, u0) consisting of a tree X and its distinguished vertex u0 ∈ V (X) is called a rooted tree. The digraph Γ which is obtained from the rooted tree (X, u0) by orienting the edges of X towards u0 is called an in-tree. The vertex u0 is the center of an in-tree Γ. It is easy to see that for an in-tree its center is the unique vertex with zero outdegree. For every map f : X → X one can define its functional graph as a digraph with the vertex set X and the arc set {(x, y) : f(x) = y}. A digraph is called functional if it isomorphic to a functional graph for some map. It is easy to see that Γ is functional digraph if and only if d+ Γ (v) = 1 for all v ∈ V (Γ). Similarly, Γ is called partial functional if d+ Γ (v) 6 1 for all v ∈ V (Γ). Each partial functional digraph Γ corresponds to some partial map of the form f : V (Γ) → V (Γ). Definition 1. Let X be a tree and σ : V (X) → V (X) be some map. The Markov graph Γ = Γ(X, σ) has the vertex set V (Γ) = E(X) and there is an arc e1 → e2 in Γ if u2, v2 ∈ [σ(u1), σ(v1)]X for ei = uivi ∈ V (Γ), i = 1, 2. In other words, N+ Γ (uv) = E([σ(u), σ(v)]X) for all edges uv ∈ E(X). Example 1. Consider the tree X with V (X) = {1, . . . , 7}, E(X) = {12, 23, 34, 45, 26, 37} and its map σ = ( 1 2 3 4 5 6 7 4 1 3 6 2 4 2 ) which are shown in Figure 1. Then the corresponding Markov graph Γ(X, σ) is shown in Figure 2. “adm-n4” 22:47 page #88 266 On disjoint union of M-graphs 1 2 3 4 5 6 7 Figure 1. The pair (X, σ) from Example 1 (dashed arrows denote σ). 34 26 45 12 23 37 Figure 2. Markov graph Γ(X, σ) for the pair (X, σ) from Example 1. A digraph Γ is called an M-graph if there exists a pair (X, σ) such that Γ ≃ Γ(X, σ). Each such a pair is called the realization of Γ. Lemma 2. [3] Let X be a tree and σ : V (X) → V (X) be a map. Then for every pair of vertices u, v ∈ V (X) and an edge xy ∈ E([σ(u), σ(v)]X) there exists an edge wz ∈ E([u, v]X) with wz → xy in Γ(X, σ). In particular, [σ(u), σ(v)]X ⊂ ⋃ wz∈E([u,v]X) [σ(w), σ(z)]X . Lemma 3. [5] Let X be a tree, A ⊂ V (X) be some connected set of vertices, σ : V (X) → V (X) be a map and Γ = Γ(X, σ). Then Γ(X[A], prA ◦σ) = Γ[E(A)]. Proposition 1. Let X be a tree and σ : V (X) → V (X) be some map. Put E(σ) = {e ∈ E(X) : d− Γ (e) > 1}. Then E(σ) = E(ConvX(Im σ)). In particular, X[E(σ)] is the connected subgraph of X. Proof. Let V1 = V (E(σ)) and V2 = ConvX(Im σ). If u ∈ V1, then there exists an edge e = uv ∈ E(σ). By definition, d− Γ (e) > 1. This means “adm-n4” 22:47 page #89 S. Kozerenko 267 that there is an edge e′ = u′v′ ∈ E(X) with e′ → e in Γ(X, σ), i.e. u, v ∈ [σ(u′), σ(v′)]X . Therefore, u ∈ V2. Conversely, suppose u ∈ V2. Then there exists a pair of vertices x, y ∈ V (X) such that u ∈ [σ(x), σ(y)]X . At first, suppose that σ(x) 6= σ(y). Then we can fix an edge e = uv ∈ E([σ(x), σ(y)]X). From Lemma 2 it follows that there is an edge e′ ∈ E([x, y]X) with e′ → e in Γ. Thus d− Γ (e) > 1 and u ∈ V1. Otherwise, let σ(x) = σ(y). Then u ∈ Im σ. If σ is a constant map, then E(σ) = E(ConvX(Im σ)) = ∅. Thus, suppose that σ is non-constant. This means that there exists a vertex v ∈ Im σ − {u}. Let σ(z) = v. Since u 6= v, we can fix an edge e = uw ∈ E([u, v]X). Again, by Lemma 2, d− Γ (e) > 1 which implies u ∈ V1. 2. Main results From Lemma 3 it strictly follows that each nontrivial M-graph Γ contains a vertex v ∈ V (Γ) such that Γ − {v} is also an M-graph. In [5] it was proved that any digraph obtained from an M-graph by deletion of a vertex with zero outdegree (in particular, an isolated vertex) is an M-graph itself. We generalize this result using the following theorem. Theorem 2. Let X be a tree and σ : V (X) → V (X) be some map. Suppose that we have a collection Ai ⊂ V (X), 1 6 i 6 m of pairwise disjoint connected sets such that for every 1 6 i 6 m either |σ(∂XAi)| = 1 or there exists 1 6 j 6 m with σ(∂XAi) ⊂ Aj. Then Γ(X, σ)− ⋃m i=1 E(Ai) is an M-graph. Proof. Consider the set of indices I1 = {1 6 i 6 m : |σ(∂XAi)| = 1} and the corresponding map g : I1 → V (X), where σ(∂XAi) = {g(i)} for all i ∈ I1. Similarly, the set of indices I2 = {1, . . . , m} − I1 defines the map f : I2 → {1, . . . , m}, where σ(∂XAi) ⊂ Af(i) for all i ∈ I2. Take a graph X − ⋃m i=1 Ai and add to it m new vertices zi for each 1 6 i 6 m with new edges ziyi for all yi ∈ ∂X(V (X) − Ai) to obtain a new graph X ′. It is easy to see that X ′ is a tree (one can think of X ′ as of tree which is obtained from X by “contracting” sets Ai into points). Put σ′(x) =              zi if σ(x) ∈ Ai, g(i) if x = zi and i ∈ I1, zf(i) if x = zi and i ∈ I2, σ(x) otherwise, for all x ∈ V (X ′). Then Γ(X, σ) − ⋃m i=1 E(Ai) ≃ Γ(X ′, σ′). “adm-n4” 22:47 page #90 268 On disjoint union of M-graphs Corollary 1. Let Γ be an M-graph and v ∈ V (Γ) be its vertex with N+ Γ (v) ⊂ {v}. Then Γ − {v} is an M-graph. Moreover, if N+ Γ (v) = {v}, then there exists a realization (X, σ) of Γ − {v} such that fix σ 6= ∅. Proof. Fix some realization (X, σ) of Γ. Let the edge e = ux ∈ E(X) corresponds to the vertex v ∈ V (Γ). If N+ Γ (v) = ∅, then σ(u) = σ(x). In this case for the connected set of vertices A = {u, x} we have |σ(∂XA)| = 1. By Theorem 2, Γ − {v} is an M-graph. Otherwise, let N+ Γ (v) = {v}. Then σ(u) = u and σ(x) = x, or σ(u) = x and σ(x) = u. In both cases σ(∂XA) ⊂ A. Again, by Theorem 2, Γ − {v} is an M-graph. Moreover, with the notation of Theorem 2, σ′(z1) = z1 (here A = A1). Therefore, in this case fix σ′ 6= ∅. Example 2. Consider the tree X with V (X) = {1, . . . , 7}, E(X) = {12, 23, 34, 16, 25, 67} and its map σ = ( 1 2 3 4 5 6 7 4 1 5 4 5 2 3 ) . Then d− Γ(X,σ)(16) = 0, however Γ(X, σ) − {16} is not an M-graph (see Figure 3). 16 34 12 67 23 25 Figure 3. Markov graph Γ(X, σ) for which Γ(X, σ) − {16} is not an M-graph. Denote by V0(Γ) = {v ∈ V (Γ) : d− Γ (v) = 0} the set of vertices with zero indegree in Γ. Proposition 2. For every M-graph Γ and 0 6 k 6 |V0(Γ)| there exists V ′ ⊂ V0(Γ) with |V ′| = k such that Γ − V ′ is an M-graph. In particular, Γ − V0(Γ) is an M-graph. Proof. Fix a realization (X, σ) of Γ. Let the edge set E′ ⊂ E(X) cor- responds to V0(Γ). By Proposition 1, the set E(X) − E′ = E(σ) is connected. Since X is a connected graph, for any 0 6 k 6 |V0(Γ)| there exists a connected set of edges E′′ ⊂ E(X) with E(X) − E′ ⊂ E′′ “adm-n4” 22:47 page #91 S. Kozerenko 269 and |E′| = |E(X)| − k. Let V ′ ⊂ V (Γ) corresponds to E(X) − E′′. Then |V ′| = k and by Lemma 3, Γ − V ′ ≃ Γ(X, σ) − (E(X) − E′′) = Γ(X, σ)[E′′] = Γ(X[E′′], prV (E′′) ◦σ) is an M-graph. Note that any digraph obtained from an M-graph by addition of an isolated vertex is also an M-graph. Using this fact one can conclude that Γ is an M-graph if and only if so is Γ ⊔ K1. However, not every disjoint union of two M-graphs is an M-graph itself. Example 3. Suppose that Γ is obtained from the complete digraph with two vertices K2 by deletion of a loop. Then Γ is an M-graph, but Γ ⊔ K1 is not (see Figure 4). Figure 4. Disjoint union of two M-graphs which is not an M-graph. Remark 1. [5] If we have a pair of trees Xi, i = 1, 2 and a pair of their maps σi : V (Xi) → V (Xi) with fix σi 6= ∅, i = 1, 2, then the disjoint union Γ(X1, σ1) ⊔ Γ(X2, σ2) is an M-graph. Indeed, “gluing” realizations (X1, σ1) and (X2, σ2) together along some pair of fixed vertices we obtain the realization of Γ(X1, σ1) ⊔ Γ(X2, σ2). As a corollary of the construction in Remark 1 one can obtain a sufficient condition for the disjoint union of two M-graphs to be an M- graph. Corollary 2. [5] Let Γ1 and Γ2 be a pair of M-graphs with even numbers of loops in each. Then Γ1 ⊔ Γ2 is an M-graph. In particular, any disjoint union of two M-graphs without loops is an M-graph itself. It turns out that for any given M-graph we can provide a graph- theoretic criterion for the existence of its realization (X, σ) with fix σ 6= ∅. Proposition 3. Let Γ be a digraph. Then Γ⊔K1 is an M-graph if and only if Γ is an M-graph and there exists its realization (X, σ) with fix σ 6= ∅. Proof. Sufficiency of this condition follows from Remark 1, since for K1 there obviously exists its realization (X, σ) with fix σ 6= ∅. Thus, we must prove only the necessity of this condition. To do so fix a realization (X ′, σ′) of Γ ⊔ K1. Let the vertex v ∈ V (Γ ⊔ K1) corresponds to a unique vertex from K1. Then N+ Γ⊔K1 (v) = {v} implying that by Corollary 1, Γ is an M-graph and there exists its realization (X, σ) with fix σ 6= ∅. “adm-n4” 22:47 page #92 270 On disjoint union of M-graphs Corollary 3. If Γ ⊔ K1 is an M-graph, then there exists its realization (X, σ) with fix σ 6= ∅. Combining Remark 1 and Proposition 3, we obtain the following result. Proposition 4. If for a pair of digraphs Γ1 and Γ2 the digraphs Γ1 ⊔ K1 and Γ2 ⊔ K1 are M-graphs, then Γ1 ⊔ Γ2 is an M-graph. Theorem 3. Let Γ1 be an M -graph and Γ2 be acyclic partial functional digraph. Then Γ1 ⊔ Γ2 is also an M -graph. Proof. Without loss of generality, we can assume that Γ2 is weakly con- nected. Since Γ2 is acyclic and partially functional, Γ2 is an in-tree. Let x0 ∈ V (Γ2) be its center (thus, d+ Γ2 (x0) = 0). Denote by X ′ the under- lying tree of Γ2. For every 0 6 i 6 eccX′(x0) put ai = |N i X′(x0)| for the cardinality of the sphere with radius i centered at x0 in X ′. Now fix a realization (X, σ0) of Γ1. Since V (X) is finite, σ0 has a periodic point u0 ∈ V (X) with period m > 1. Consider the restriction σ = σ0|orbσ0 (u0) of σ0 to orbσ0 (u0). Clearly, σ is a cyclic permutation of orbσ0 (u0). For every 0 6 i 6 eccX′(x0) add ai new vertices yi 1, . . . yi ai to X with the new edges yi jσ−i mod m(u0) for all 1 6 j 6 ai (of course, σ0(u0) = u0). Denote the obtained tree as X ′′. For all u ∈ V (X ′′) put σ′(u) =        σ(u) if u ∈ V (X), yi−1 k if u = yi j , i > 1 and N+ Γ2 (xi j) = {xi−1 k }, σ(u0) if u = y0 1, where N i X′(x0) = {xi j : 1 6 j 6 ai} (for example, N0 X′(x0) = {x0 1} = {x0}). Then Γ(X ′′, σ′) ≃ Γ1 ⊔ Γ2 (the edges from E(X) correspond to the vertices of Γ1 and edges of the form yi jσ−i mod m(u0) correspond to the vertices xi j). Note that the acyclicity condition in Theorem 3 is essential as can be seen from the digraph in Example 3. Example 4. Consider the pair (X, σ) from Example 1 and the corre- sponding Markov graph Γ1 = Γ(X, σ). Also, let Γ2 be the in-tree depicted in Figure 5 (the vertices of Γ2 are labeled according to the notation in the proof of Theorem 3). Thus, x0 is the center of Γ2, eccX′(x0) = 2, N1 X′(x0) = {x1 1, x1 2}, N2 X′(x0) = {x2 1, x2 2} and a1 = a2 = 2. Put u0 = 4. “adm-n4” 22:47 page #93 S. Kozerenko 271 Then orbσ(u0) = {4, 6} and therefore m = 2. The corresponding tree X ′′ is shown in Figure 6. We also have σ′(y0 1) = 6, σ′(y1 1) = σ′(y1 2) = y0 1 and σ′(y2 1) = σ′(y2 2) = y1 2. x2 1 x1 2 x0 x1 1 x2 2 Figure 5. The in-tree Γ2. 1 2 3 4 5 6 7 y1 1 y1 2 y0 1 y2 1 y2 2 Figure 6. The tree X ′′ from Example 4. Theorem 4. The disjoint union of any collection of weak components (in particular, each weak component) in an M-graph is an M-graph itself. Proof. It is sufficient to prove that for any M-graph Γ and its weak component Γ′ the digraph Γ − V (Γ′) is an M-graph. To do so fix a realization (X, σ) of Γ. Let the set of edges E′ ⊂ E(X) corresponds to the vertex set of Γ′. Consider the components X1, . . . , Xm of the induced subgraph X[E′] in X and put Ai = V (Xi) for all 1 6 i 6 m. Suppose that for some 1 6 i 6 m there exists a vertex x ∈ Ai with σ(x) /∈ ∪m j=1Aj = V (X[E′]). Then EX(σ(x)) ∩ E′ = ∅. If for some y ∈ Ai we have σ(x) 6= σ(y), then dX(σ(x), σ(y)) > 2. This means that there is a vertex u ∈ [σ(x), σ(y)]X such that e = uσ(x) ∈ E(X). By Lemma 2, there exists an edge e′ ∈ E([x, y]X) ⊂ E(Ai) with e′ → e in Γ(X, σ). Since “adm-n4” 22:47 page #94 272 On disjoint union of M-graphs e /∈ E′, Γ′ is not a weak component of Γ. The obtained contradiction implies that in this case we have σ(x) = σ(y) for every y ∈ Ai. In other words, |σ(Ai)| = 1. Now let 1 6 i 6 m is fixed and for every vertex x ∈ Ai there exists 1 6 jx 6 m with σ(x) ∈ Ajx . We want to prove that in this case jx = jy for each pair of vertices x, y ∈ Ai. To the contrary, suppose jx 6= jy for some x, y ∈ Ai. Then dX(Ajx , Ajy ) > 1. This implies the existence of an edge e ∈ E([σ(x), σ(y)]X) − ∪m k=1E(Ak). Again, from Lemma 2 it follows that there exists e′ ∈ E([x, y]X) ⊂ E(Ai) with e′ → e in Γ(X, σ) which is a contradiction. Thus, in this case σ(Ai) ⊂ Aj for some 1 6 j 6 m. Theorem 2 now asserts that Γ − V (Γ′) is an M-graph. Corollary 4. If for a pair of digraphs Γ1 and Γ2 their disjoint union Γ1 ⊔ Γ2 is an M-graph, then both Γ1 and Γ2 are M-graphs. Proof. Clearly, Γ1 and Γ2 are both disjoint unions of weak components in Γ1 ⊔ Γ2. Proposition 5. Let Γ1 and Γ2 be a pair of nontrivial digraphs having loops at each of their vertices. Then Γ1 ⊔ Γ2 is an M-graph if and only if Γ1 ⊔ K1 and Γ2 ⊔ K1 are both M-graphs. Proof. The sufficiency of this condition strictly follows from Proposition 4. To prove its necessity fix a realization (X, σ) of Γ = Γ1 ⊔ Γ2. Let Γ′ be a weak component in Γ and E′ ⊂ E(X) be the corresponding set of edges in X. We want to prove that E′ is connected. To the contrary, suppose that there is a partition E′ = E1 ⊔ E2 with dX(V (E1), V (E2)) > 1. Since Γ′ is weakly connected, there is a pair of edges ei = uivi ∈ Ei, i = 1, 2 with e1 → e2 or e2 → e1 in Γ′. Without loss of generality, assume e1 → e2 in Γ′. We have e1, e2 ∈ N+ Γ(X,σ)(e1) which implies [u1, u2]X ⊂ [σ(u1), σ(v1)]X . However, the inequality dX(V (E1), V (E2)) > 1 asserts E([u1, u2]X) − E′ 6= ∅. In other words, there exists an edge e′ /∈ E′ such that e1 → e′ in Γ. Therefore, Γ′ is not a weak component in Γ. The obtained contradiction proves that E′ is connected. By Lemma 3, Γ′ ≃ Γ(X, σ)[E′] = Γ(X[E′], prV (E′) ◦σ). Furthermore, since Γ1 and Γ2 are nontrivial digraphs, we have Γ 6= Γ′. This implies ∂XV (E′) 6= ∅. Fix a vertex w ∈ ∂XV (E′) and an edge e ∈ EX(w) − E′. Let E′′ be the vertex set of the weak component in Γ(X, σ) which con- tains e. Similarly, we can prove that E′′ is connected. Also, note that w ∈ ∂XV (E′′). Finally, the proof of Theorem 4 implies that σ(∂XV (E′)) ⊂ “adm-n4” 22:47 page #95 S. Kozerenko 273 ∂XV (E′) as well as σ(∂XV (E′′)) ⊂ ∂XV (E′′). Hence, we can conclude that σ(w) = w. Thus, for every weak component Γ′ of Γ there exists its realization (X ′, σ′) with fix σ′ 6= ∅. But Γ1 (as well as Γ2) is a disjoint union of weak components in Γ. Combining this fact with Remark 1 and Proposition 4, we obtain that Γ1 ⊔ K1 as well as Γ2 ⊔ K1 are M-graphs. Example 5. Consider the path X ≃ P4 with V (X) = {1, 2, 3, 4}, E(X) = {12, 23, 34} and its vertex map σ = ( 1 2 3 4 3 1 4 2 ) . Then Γ(X, σ) has a loop at each vertex, but Γ(X, σ) ⊔ K1 is not an M-graph. References [1] C. Bernhardt, Vertex maps for trees: algebra and periods of periodic orbits, Discrete Contin. Dyn. Syst. 14 (2006), 399–408. [2] C-W. Ho and C. Morris, A graph-theoretic proof of Sharkovsky’s theorem on the periodic points of continuous functions, Pacific J. Math. 96 (1981), 361–370. [3] S. Kozerenko, Markov graphs of one-dimensional dynamical systems and their discrete analogues, Rom. J. Math. Comput. Sci. 6 (2016), 16–24. [4] S. Kozerenko, Discrete Markov graphs: loops, fixed points and maps preordering, J. Adv. Math. Stud. 9 (2016), 99–109. [5] S. Kozerenko, On the abstract properties of Markov graphs for maps on trees, Mat. Bilten 41 (2017), 5–21. [6] V. A. Pavlenko, Number of digraphs of periodic points of a continuous mapping of an interval into itself, Ukrainian Math. J. 39 (1987), 481–486 (in russian). [7] V. A. Pavlenko, Periodic digraphs and their properties, Ukrainian Math. J. 40 (1988), 455–458 (in russian). [8] V. A. Pavlenko, On characterization of periodic digraphs, Cybernet. Systems Anal. 25 (1989), 49–54 (in russian). [9] A. N. Sharkovsky, Coexistence of the cycles of a continuous mapping of the line into itself, Ukrainian Math. J. 16 (1964), 61–71 (in russian). [10] P. D. Straffin, Periodic points of continuous functions, Math. Mag. 51 (1978), 99–105. Contact information S. Kozerenko Faculty of Mechanics and Mathematics, Taras Shevchenko National University of Kyiv, Volodymyrska str., 64, 01033 Kyiv, Ukraine E-Mail(s): kozerenkosergiy@ukr.net Received by the editors: 12.03.2017 and in final form 02.11.2017.