This is a preprint of: “On the periods of a continuous self-map on a graph”, Juan Luis Garcia Guirao, Jaume Llibre, Comput. Appl. Math., vol. 38(2), 78: 1–6, 2019. DOI: [10.1007/s40314-019-0846-0] ON THE PERIODS OF A CONTINUOUS SELF–MAP ON A GRAPH JUAN LUIS GARCÍA GUIRAO1 AND JAUME LLIBRE2 Abstract. Let G be a graph and f be a continuous self–map on G. We present new and known results (from another point of view) on the periods of the periodic orbits of f using mainly the action of f on its homology, or the shape of the graph G. 1. Introduction and statement of the main results A discrete dynamical system (G, f ) is formed by a continuous map f : G → G where G is a topological space. A point x ∈ G is periodic of period k if f k (x) = x and f i (x) 6= x if 0 < i < k. If k = 1, then x is called a fixed point. Per(f ) denotes the set of periods of all the periodic points of f . The orbit of the point x ∈ G is the set {x, f (x), f 2 (x), . . . , f n (x), . . .} where by f n we denote the composition of f with itself n times. To knowledge the behavior of all different kind of orbits of f is to study the dynamics of the map f . Many times the periodic points play an important role for understand- ing the dynamics of a discrete dynamical system. One of the best known results in this direction is the paper Period three implies chaos for con- tinuous interval maps, see [8]. Here a graph G is a compact connected space containing a finite set V such that G \ V has finitely many open connected components, each one homeomorphic to the interval (0, 1), called edges of G, and the points of V are called the vertexes of G. The edges are disjoint from the vertexes, and the vertexes are at the boundary of the edges. In this paper we shall work with a graph G. Our goal is to study the periods of the periodic points of the continuous maps f : G → G. Key words and phrases. topological graph, discrete dynamical systems, Lefschetz numbers, Lefschetz zeta function, periodic point, period. 2010 Mathematics Subject Classification: 37E25, 37C25, 37C30. 1 http://www.gsd.uab.cat 2 J.L.G. GUIRAO AND J. LLIBRE The degree of a vertex V of a graph G is the number of edges having V in its boundary, if an edge has both boundaries in V then we count this edge twice. An endpoint of a graph G is a vertex of degree one. A branching point of a graph G is a vertex of degree at least three. The homological spaces of G with coefficients in Q are denoted by Hk (G, Q). Since G is a graph k = 0, 1. A continuous map f : G → G induces linear maps f∗k : Hk (G, Q) → Hk (G, Q). We only work with graphs, so H0 (G, Q) ≈ Q and f∗0 is the identity map. A subset of G homeomorphic to a circle is a circuit. It is known that H1 (G, Q) ≈ Qm being m the number of the independent circuits of G in the sense of the homology. Here f∗1 is a m × m matrix A with integer entries. For more details on this homology see for instance [12]. If A be a m × m matrix, then a submatrix lying in the same set of k rows and columns is a k × k principal submatrix of A. The determinant of a principal submatrix is a k × k principal minor. The sum of the nk different k × k principal minors of A is denoted by Ek (A). Note that Em (A) is the determinant of A and E1 (A) is the trace of A. Of course the characteristic polynomial of A is given by (1) det(tI − A) = tm − E1 (A)tm−1 + E2 (A)tm−2 − . . . + (−1)m Em (A). The biggest modulus of the eigenvalues of the matrix A is called the spectral radius of A and it is denoted by sp(A). Our main results are the following ones. Theorem 1. Let G be a graph, f : G → G be a continuous map, and A be the integral matrix of the endomorphism f∗1 : H1 (G, Q) → H1 (G, Q) induced by f on the first homology group of G. The following statements hold. (a) If E1 (A) 6= 1, then 1 ∈ Per(f ). (b) If E1 (A) = 1 and E2 (A) 6= 0, then Per(f ) ∩ {1, 2} = 6 ∅. (c) If E1 (A) = 1, E2 (A) = ... = Ek−1 (A) = 0 and Ek (A) = 6 0 for k = 3, . . . , m then Per(f ) intersection the set of all the divisors of k is not empty. Theorem 1 is proved in section 2 using the Lefschetz fixed point theory. The next result is an immediate consequence of Theorem 1. Corollary 2. Under the assumptions of Theorem 1, if the character- istic polynomial of the matrix A is different from 1 − t, then Per(f ) ∩ {1, 2, . . . , m} = 6 ∅. ON THE PERIODS OF A CONTINUOUS SELF–MAP ON A GRAPH 3 Let k be a positive integer we denote by god(k) the greatest odd divisor of k. Let S be a subset of positive integer, the pantheon of S is the set {god(k) : k ∈ S}. Theorem 3. Let G be a graph, f : G → G be a continuous map, and A be the integral matrix of the endomorphism f∗1 : H1 (G, Q) → H1 (G, Q) induced by f on the first homology group of G. If sp(f∗1 ) > 1, then f has infinitely many periods. More precisely, there is an n ∈ N such that {kn : k ∈ N} ⊂ Per(f ) and the pantheon of Per(f ) is infinite. Theorem 4. Let G be a graph with v vertexes, e endpoints, s edges and at least one branching point. Let f : G → G be a continuous map having all the branching point of G fixed. If for some period n of f , god(n) > e + 2s − 2v + 2 then f has infinitely many periods. More precisely, there is an n ∈ N such that {kn : k ∈ N} ⊂ Per(f ) and the pantheon of Per(f ) is infinite. Figure 1. The glasses graph. From Theorem 4 we can deduce many results similar to the one given in the seminal paper Period three implies chaos for self–continuous maps on the interval in the sense of having infinitely many periods. Corollary 5. The following map f have infinitely many periods if: (a) f is a continuous self-map on the graph having the shape of the letter Y with the branching point fixed and having a period n such that god(n) > 3; (b) f is a continuous self-map on the graph having the shape of the number 8 or on the graph having the shape of the letter θ with the branching points fixed and having a period n such that god(n) > 4; (c) f is a continuous self-map on the glasses graph having the shape of the graph described in Figure 1 with the branching points fixed and having a period n such that god(n) > 2. Theorems 3, 4 and Corollary 5 are proved in section 3. We note that this paper is a kind of survey with new results. That is, Theorem 1 is completely new, but Theorems 3 and 4 essentially follow 4 J.L.G. GUIRAO AND J. LLIBRE combining known results on the continuous self-maps on graphs as we will see in their proofs. 2. Proof of Theorem 1 Let f : G → G be a continuous map on the graph G. The Lefschetz number of f is defined by L(f ) = trace(f∗0 ) − trace(f∗1 ). The Lefschetz Fixed Point Theorem states: If L(f ) 6= 0 then f has a fixed point (see for instance [5]). In order to control the whole sequence of the Lefschetz numbers of the iterates of f , i.e. {L(f n )}n≥1 , we use the formal Lefschetz zeta function of f defined by ∞ ! X L(f n ) n (2) Zf (t) = exp t . n=1 n It is known that for a continuous self–map of a graph G the Lefschetz zeta function is the rational function det(I − tf∗1 ) det(I − tA) (3) Zf (t) = = , det(I − tf∗0 ) 1−t where A is the integer matrix defined by f∗1 , for a proof see Franks [6].  Since det(I − tA) = tn det 1t I − A , from (1) we get det(I − tA) = 1 − E1 (A)t + E2 (A)t2 − . . . + (−1)m Em (A)tm . From (2) and (3) we obtain (4) P∞ L(f n ) tn = log(Zf (t)) n=1 n   det(I − tA) = log 1−t   1 − E1 (A)t + E2 (A)t2 − ... + (−1)m Em (A)tm = log 1−t 1  = (1 − E1 (A))t + 1 − E1 (A)2 + 2E2 (A) t2 2 1  = 1 − E1 (A)3 + 3E1 (A)E2 (A) − 3E3 (A) t3 + O(t4 ). 3 ON THE PERIODS OF A CONTINUOUS SELF–MAP ON A GRAPH 5 So we have that L(f ) = 1 − E1 (A), and L(f 2 ) = 1 − E1 (A)2 + 2E2 (A). Therefore if E1 (A) 6= 1 than L(f ) 6= 0 and by the Lefschetz Fixed Point Theorem statement (a) is proved. If E1 (A) = 1 and E2 (A) 6= 0, then L(f 2 ) = 2E2 (A) 6= 0, again by the Lefschetz Fixed Point Theorem state- ment (b) follows. Working with the expression (4) when E1 (A) = 1, E2 (A) = ... = Ek−1 (A) = 0 and Ek (A) 6= 0 for k = 3, . . . , m we obtain that L(f k ) = (−1)k kEk (A) 6= 0, hence by the Lefschetz Fixed Point Theorem state- ment (c) is proved. This completes the proof of Theorem 1. 3. Proof of Theorems 3, 4 and Corollary 5 Let G be a graph, and let f : G → G bea continuous map. One way of measuring the complexity of the dynamics of the map f is through the notion of topological entropy. Here we introduce the topological entropy using the definition of Bowen [4]. Since a graph G is a subset of R3 , we consider the distance between two points of G as the distance of these two points in R3 . Now, we define the distance dn on G by dn (x, y) = max d(f i (x), f i (y)), ∀x, y ∈ G. 0≤i≤n A finite set S is called (n, ε)–separated with respect to f if for different points x, y ∈ S we have dn (x, y) > ε. We denote by Sn the maximal cardinality of an (n, ε)–separated set. Define 1 h(f, ε) = lim sup log Sn . n→∞ n Then h(f ) = lim h(f, ε). ε→0 is the topological entropy of f . We have given the definition of Bowen because probably is the shorter one, the classical definition was due to Adler, Konheim and Mc Andrew [1]. See for instance the book of Hasselblatt and Katok [7] and [2] for other equivalent definitions and properties of the topological entropy. The next result is due to Manning [11]. Theorem 6. Let f : G → G a continuous map on the graph G, then log max{1, sp(f∗,1 )} ≤ h(f ). 6 J.L.G. GUIRAO AND J. LLIBRE There are two different proofs for the next result, see [9] and [3]: Theorem 7. Let f : G → G a continuous map on the graph G. Then the following statements are equivalent: (a) There is an m ∈ N such that {mn : n ∈ N} ⊂ Per(f ). (b) h(f ) > 0. (c) The pantheon of Per(f ) is infinite. Proof of Theorem 3. Since sp(f∗,1 ) > 1 by Theorem 6 we have that h(f ) > 0. Then by Theorem 7 Theorem 3 follows.  The following result can be found in [10]. Theorem 8. Let f : G → G a continuous map on the graph G having e endpoints, s edges, v vertexes and at least one branching point. Assume that f has all branching points fixed. Then god(n) > e + 2s − 2v + 2 for some period n of f if and only if h(f ) > 0. Proof of Theorem 4. Under the assumptions of Theorem 4 we have that god(n) > e + 2s − 2v + 2 for some period n of f , so h(f ) > 0 by Theorem 8. Again by Theorem 7 Theorem 4 is proved.  Proof of Corollary 5. The proof is a direct consequence of the application of Theorem 4 taking account that e + 2s − 2v + 2 is respectively equal to 3 (e = 3, v = 4 and s = 3) for the graph Y ; equal to 4 for the graph 8 (e = 0, v = 1 and s = 2) and for the graph θ (e = 0, v = 2 and s = 3), and 2 for the glasses graph (e = 2, v = 6 and s = 7).  Acknowledgements The first author of this work was partially supported by MINECO grant number MTM2014-51891-P and Fundación Séneca de la Región de Murcia grant number 19219/PI/14.. The second author is partially supported by a FEDER-MINECO grant MTM2016-77278-P, a MINECO grant MTM2013-40998-P, and an AGAUR grant number 2014SGR-568. References [1] R.L. Adler, A.G. Konheim and M.H. McAndrew, Topological entropy, Trans. Amer. Math. Soc. 114 (1965), 309–319. [2] F. Balibrea, On problems of Topological Dynamics in non-autonomous discrete systems, Applied Mathematics and Nonlinear Sciences 1(2) (2016), 391–404. [3] A.M. Blokh, The Spectral Decomposition, Periods of Cycles and Misiurewicz Conjecture for Graph Maps, 24–31, Lecture Notes in Math., Springer 1514, 1991. ON THE PERIODS OF A CONTINUOUS SELF–MAP ON A GRAPH 7 [4] R. Bowen, Entropy for group endomorphisms and homogeneous spaces Trans. Amer. Math. Soc 153 (1971), 401–414; erratum: Trans. Amer. Math. Soc. 181 (1973), 509–510. [5] R.F. Brown, The Lefschetz fixed point theorem, Scott, Foresman and Company, Glenview, IL, 1971. [6] J. Franks, Homology and dynamical systems, CBSM Regional Conf. Ser. in Math. 49, Amer. Math. Soc., Providence, R.I. 1982. [7] B. Hasselblatt and A. Katok, Handbook of dynamical systems, Vol. 1A. North–Holland, Amsterdam, 2002. [8] T.Y. Li and J. Yorke, Period three implies chaos, Amer. Math. Monthly 82 (1975), 985–992. [9] J. Llibre and M. Misiurewicz, Horseshoes, entropy and periods for graph maps, Topology 52 (1993), l49–164. [10] J. Llibre and R. Saghin, Topological entropy and periods of graph maps, J. Difference Equ. Appl. 18 (2012), 589–598. [11] A. Manning, Topological entropy and the first homology group, in Dynamical systems - Warwick 1974, Lecture Notes in Math 468, Springer-Verlag, Berlin, 1975, 185–190. [12] E.H. Spanier, Algebraic Topology, Springer–Berlag, New York (1981). 1 Departamento de Matemática Aplicada y Estadística. Universidad Politécnica de Cartagena, Hospital de Marina, 30203-Cartagena, Región de Murcia, Spain–Corresponding Author– E-mail address: juan.garcia@upct.es 2 Departament de Matemàtiques. Universitat Autònoma de Barcelona, Bellaterra, 08193-Barcelona, Catalonia, Spain E-mail address: jllibre@mat.uab.cat