* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download MINIMAL NUMBER OF PERIODIC POINTS FOR SMOOTH SELF
Survey
Document related concepts
Line (geometry) wikipedia , lookup
List of important publications in mathematics wikipedia , lookup
Dynamical system wikipedia , lookup
Non-standard calculus wikipedia , lookup
Fermat's Last Theorem wikipedia , lookup
Karhunen–Loève theorem wikipedia , lookup
Wiles's proof of Fermat's Last Theorem wikipedia , lookup
Fundamental theorem of calculus wikipedia , lookup
Four color theorem wikipedia , lookup
Georg Cantor's first set theory article wikipedia , lookup
Hyperreal number wikipedia , lookup
Collatz conjecture wikipedia , lookup
Transcript
MINIMAL NUMBER OF PERIODIC POINTS FOR SMOOTH SELF-MAPS OF S 3 GRZEGORZ GRAFF AND JERZY JEZIERSKI Abstract. Let f be a continuous self-map of a smooth compact connected and simply-connected manifold of dimension m ≥ 3, r a fixed natural number. A topological invariant Drm [f ], introduced in [5], is equal to the minimal number of r-periodic points for all smooth maps homotopic to f . In this paper we calculate Dr3 [f ] for all self-maps of S 3 . 1. Introduction The classical problem in periodic point theory is to determine or estimate the least number of r-periodic points in homotopy class of a given f , a self-map of a compact manifold M m , where r is a fixed natural number (cf. [10]). If M m is simply-connected and has the dimension m ≥ 3, then one can always find a map g homotopic to f with only one point in Fix(g r ) (cf. [9]). This is however impossible, if we demand additionally that g is smooth. In [5] the authors define the topological invariant Drm [f ] equal to the minimal number of elements in Fix(g r ) for all g which are smooth and homotopic to f . This enables to explore the new smooth branch of Nielsen periodic point theory. Let us remark that Drm [f ] may be interpreted purely in the terms of smooth category, namely we may assume that f is smooth and approximate the homotopy which joins f with g by a smooth one. Then Drm [f ] gives the minimal number of periodic points in smooth homotopy class of the given f . This invariant is obtained by decomposing Lefschetz numbers of iterations into sequences which can be locally realized as fixed point indices of iterations at an isolated periodic orbit for some C 1 map. As a result, to find the exact value of Drm [f ] we need two types of data: the information about the form of {L(f n )}n|r (more precisely of the set of so-called 2000 Mathematics Subject Classification. Primary 37C25, 55M20; Secondary 37C05. Key words and phrases. Least number of periodic points, indices of iterations, smooth maps, low dimensional dynamics. Research supported by KBN grant No 1 P03A 03929. 1 2 GRZEGORZ GRAFF AND JERZY JEZIERSKI P algebraic periods: {n ∈ N : k|n µ(n/k)L(f k ) 6= 0}, for n|r, where µ is the Möbius function) and the description of all possible sequences of local indices of iterations. The article joins the ideas of three papers: [5] in which Drm [f ] is defined, [12] where the description of algebraic periods of self-maps of S 3 is given, and finally [6], where there is a classification of sequences of local indices of iterations in dimension 3. Basing on these results we are able to determine Dr3 [f ] for all self-maps of a smooth 3manifold M which is closed connected and simply-connected, i.e., M is 3-dimensional sphere, provided that the result of G. Perelman on Poincaré conjecture is true (cf. for example [8]). It is worth pointing out that Dr3 [f ] is almost independent on f , i.e., is insensitive on a homotopy class of f , which seems to be rather unexpected conclusion. For example if r is odd and f is a self-map of S 3 of degree with the modulus greater than 1, then Dr3 [f ] ∈ {ζ(r) − 1, ζ(r)}, where ζ(r) is the number of all divisors of r (cf. Theorem 4.2). This is a result of simply-connectedness of S 3 and of the fact that the set of algebraic periods is equal to the set of all natural numbers. As a consequence, for 3dimensional sphere S 3 the value of Dr3 [f ] may be perceive as the invariant of the whole space rather than the class of homotopy determined by f . At present, our method of the calculation of the invariant Dr3 [f ] is limited to 3-dimensional manifolds, because there is no description of fixed point indices of iterations at an isolated fixed point in higher dimensions. In [5] the authors determined Dr3 [f ] for self-maps of S 2 × I. The further development in this direction, i.e., finding the invariant for 3-manifolds with boundary depends on the possibility of obtaining the overall description of algebraic periods of self-maps of these manifolds. The article is organized as follows: in Sections 2 and 3 we introduce the notation and definitions as well as the statements of [5], [6] and [12] which will be used in the further part of the paper. In Section 4 we give the main results (Theorem 4.2 and Theorem 4.6). The case of odd r is a straightforward consequence of the results established in [5], [6], [12] (in particular Theorem 2.9 from [5]), while the case of even r needs careful and detailed analysis. 2. Preliminary results A sequence of indices of iterations at an isolated fixed (periodic) point plays crucial role in minimalization of the number of periodic points in homotopy class. Let f : U → Rm , where U is an open subset of Rm , be a map such that x0 is an isolated fixed point for each iteration of f . Then the sequence of local indices {ind(f n , x0 )}∞ n=1 is well-defined. Below we introduce useful notation of representing such sequences, which will be used in the next sections. MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 3 Definition 2.1. For a given k we define the basic sequence: ½ k if k|n, regk (n) = 0 if k 6 |n. Each regk is an elementary periodic sequence of the form: (0, . . . , 0, k, 0, . . . , 0, k, . . .), where the non-zero entries appear for indices divisible by k. A sequence of indices of iterations (as well as any integer sequence) may be written in the form of so-called periodic expansion (cf. [13]), namely: (2.1) ind(f n , x0 ) = ∞ X ak regk (n), k=1 P where an = n1 k|n µ(k)ind(f (n/k) , x0 ), µ is the classical Möbius function, i.e., µ : N → Z is defined by the following three properties: µ(1) = 1, µ(k) = (−1)s if k is a product of s different primes, µ(k) = 0 otherwise. It appears that indices of iterations must satisfy some conditions, found in [4], called Dold relations (or Dold congruences). Theorem 2.2. (Dold relations) All coefficients of the periodic expansion for a sequence of indices of iterations are integers, i.e., in the formula (2.1) ak ∈ Z. S For p ≥ 1 we define Pp (f ) = Fix(f p ) \ 0<n<p Fix(f n ). If x ∈ Pp (f ), then the orbit of x will be called p-orbit. Now we introduce the notion of a Differential Dold sequence in Rm for a p-orbit, which we will denote as DDm (p) sequence. This is a sequence which can be realized as a sequence of indices of iterations on isolated p-orbit for some smooth map in m-dimensional space. m Definition 2.3. A sequence of integers {cn }∞ n=1 is called a DD (p) sequence if there is an isolated p-orbit P , its neighborhood U ⊂ Rm and a C 1 map φ : U → Rm such that cn = ind(φn , P ). If this equality holds for n|r, where r is fixed, then the finite sequence {cn }n|r will be called DDm (p|r) sequence. ¤ There is a strict relation, established in [5], between the minimal number of r-periodic points for all smooth maps in a given homotopy class and DDm (p|r) sequences (Theorem 2.5 below). Definition 2.4. Let {ξn }n|r be a sequence of integers satisfying Dold relations, i.e., its coefficients in the periodic expansion are integers. Assume 4 GRZEGORZ GRAFF AND JERZY JEZIERSKI that we are able to decompose {ξn }n|r into the sum: ξ(n) = c1 (n) + ... + cs (n), m where ci is a DD (li |r) sequence for i = 1, ..., s. Each such decomposition determine the number l = l1 + ... + ls . We define the number Drm [ξ] as the smallest l which can be obtained in this way. ¤ Let {L(f n )}n|r be the sequence of Lefchetz number of iterations, we will denote Drm [f ] = Drm [{L(f n )}n|r ]. Theorem 2.5. ([5]) Let M be a smooth compact connected and simplyconnected manifold of dimension m ≥ 3 and r ∈ N a fixed number. For M with nonempty boundary we assume additionally that f has no periodic points on the boundary. Then, min{#Fix(g r ) : g is C1 and is homotopic to f } = Drm [f ]. ¤ By Definition 2.4 determining Drm [f ] demands the knowledge of the forms of all DDm (p) sequences. There are strong restrictions on such sequences, found by Chow, Mallet-Paret and Yorke (cf. [3]). It is not difficult to observe that in order to get any DDm (p) sequence it is enough to replace each ak regk by ak regpk in the periodic expansion of some DDm (1) sequence (we will say that this DDm (p) sequence comes from the given DDm (1) sequence). As a consequence, we will be able to calculate Drm [f ] if we know all the forms of DDm (1) sequences. This is provided in dimension 3 by the following theorem: Theorem 2.6. ([6]) There are seven kinds of DD3 (1) sequences: (A) cA (n) = a1 reg1 (n) + a2 reg2 (n), (B) cB (n) = reg1 (n) + ad regd (n), (C) cC (n) = −reg1 (n) + ad regd (n), (D) cD (n) = ad regd (n), (E) cE (n) = reg1 (n) − reg2 (n) + ad regd (n), (F) cF (n) = reg1 (n) + ad regd (n) + a2d reg2d (n), where d is odd, (G) cG (n) = reg1 (n) − reg2 (n) + ad regd (n) + a2d reg2d (n), where d is odd. In all cases d ≥ 3 and ai ∈ Z. In fact in dimension 3 we will need to know only the forms of some special DD3 (2) sequences, in addition to DD3 (1) sequences, for finding the value of Drm [f ] (see Lemma 2.7). Let us list below three DD3 (2) sequences which come from DD3 (1) sequences of the form (E), (F) and (G): MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 5 (E’) cE 0 (n) = reg2 (n) − reg4 (n) + a2d reg2d (n), where d ≥ 3, (F’) cF 0 (n) = reg2 (n) + a2d reg2d (n) + a4d reg4d (n), where d ≥ 3 is odd, (G’) cG0 (n) = reg2 (n) − reg4 (n) + a2d reg2d (n) + a4d reg4d (n), where d ≥ 3 is odd. Lemma 2.7. ([5]) Let f : M → M be a C 1 map, dim M = 3, then in the definition of Dr3 [f ] it is enough to consider only DD3 (1|r) sequences, i.e., sequences which for n|r are of the forms (A)-(G) and DD3 (2|r) sequences of the forms (E’), (F’) and (G’). Let r be a fixed natural number. The sequence of Lefschetz numbers {L(f n )}∞ n=1 also satisfies Dold relations, so we can write it down in the form of a periodic expansion: (2.2) n L(f ) = ∞ X bk regk (n). k=1 Definition 2.8. For a given periodic expansion (2.2) of {L(f n )}∞ n=1 , we define B(f ), the set of algebraic periods of f , as B(f ) = {k ∈ N : bk 6= 0} and Br (f ), the set of algebraic periods of f up to the level r, as Br (f ) = {k ∈ N : k|r and bk 6= 0}. Let us now rewrite the formula (2.2) for n|r into: (2.3) L(f n ) = b1 reg1 (n) + b2 reg2 (n) + b4 reg4 (n) + X bk regk (n), k∈G where b1 , b2 , b4 are arbitrary integers, G = {k ∈ N : k 6∈ {1, 2, 4} and bk 6= 0} = Br (f ) \ {1, 2, 4}. Let us define H, the subset of G: H = {k ∈ G : k is odd and bk 6= 0, b2k 6= 0}. In the next section we will use the following result proved in [5]: Theorem 2.9. If r is odd, then: ½ #G 3 (∗) Dr [f ] = #G + 1 if |L(f )| ≤ #G, otherwise. If r is even and r > 4, then: (∗∗) Dr3 [f ] ∈ [#G − #H, #G − #H + 2]. 6 GRZEGORZ GRAFF AND JERZY JEZIERSKI 3. Algebraic periods In order to calculate Dr3 [f ] we need to know the decomposition of Lefschetz numbers of iterations into the periodic expansion. Thus, we need the exact form of the set of algebraic periods of f (cf. Definition 2.8). We will base on the description of algebraic periods for self-maps of S 3 which is given in [12]. We use homology spaces with the coefficients in the field of rational numbers Q. Let us recall that ½ 3 Hi (S ) = Q for i = 0, 3; 0 otherwise. Let f : S 3 → S 3 . The homomorphism f∗3 : H3 (S 3 ; Q) → H3 (S 3 ; Q) is a multiplication by a number β ∈ Z, called a degree of f (denoted also as deg(f )). Assume r ∈ N is fixed. We consider the periodic expansion of {L(f n )}∞ n=1 given by (2.2). Let us remind to the reader that by Definition 2.8 bn 6= 0 is equivalent to the statement that n ∈ B(f ), i.e., that n is an algebraic period. Lemma 3.1. ([12], Theorem 1.2) Let bn denote the nth coefficient of the periodic expansion of {L(f n )}∞ n=1 then: (a) b1 = 1 − β. (b) b2 = 0 if and only if β ∈ {0, 1}. (c) If n > 2, then bn = 0 if and only if β ∈ {−1, 0, 1}. Let ζ(r) denote the number of all divisors of r. Assume that β 6∈ {−1, 0, 1}. Then, by Lemma 3.1 B(f ) = N and the following Proposition holds: Proposition 3.2. Br (f ) = {n ∈ N : n|r}, or equivalently #Br (f ) = ζ(r). By Proposition 3.2 we obtain: Proposition 3.3. (1) For r odd G = Br (f ) \ {1}, thus #G = ζ(r) − 1. (2) For r even there is: if 4|r, then G = Br (f ) \ {1, 2, 4}, thus #G = ζ(r) − 3, if 4 6 |r, then G = Br (f ) \ {1, 2}, thus #G = ζ(r) − 2. Lemma 3.4. If r is even, then #H= η(r) − 1, where η(r) denotes the number of all odd divisors of r. MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 7 Proof. Observe that #H is the number of pairs k, 2k, where k|r, k > 1 is odd, in G. As each natural k is an algebraic period, every odd k > 1 determines such a pair and thus an element in H. ¤ 4. Minimal number of periodic points for self-maps of S 3 The complete description of the minimal number of r-periodic points for all smooth maps homotopic to the map f : S 3 → S 3 of degree β will be given in Proposition 4.1 and Theorems 4.2 and 4.6 below. Proposition 4.1. For each self-map f of S 3 we have: L(f n ) = 1 − β n . This implies the following statements. If β = 1, then L(f n ) = 0 for all n, hence Dr3 [f ] = 0. If β = −1, then L(f n ) = 2reg1 (n) − reg2 (n), hence Dr3 [f ] = 1 for all r, because we can realize Lefschetz numbers of iterations by one sequence of the type (A). If β = 0, L(f n ) = reg1 (n), and analogously as in the previous case, 3 Dr [f ] = 1. ¤ In order to find Dr3 [f ], for f such that |β| = |deg(f )| > 1, we will use Theorem 2.9. In the case of odd r, Theorem 2.9 with Proposition 3.3 part (1) and the fact that L(f ) = 1 − β give immediately the exact value of Dr3 [f ]: Theorem 4.2. For r odd, |β| > 1 there is: ½ ζ(r) − 1 if ζ(r) ≥ β ≥ 2 − ζ(r), Dr3 [f ] = ζ(r) otherwise. 4.1. The case of even r. Let us consider the minimal realization of Lefschetz numbers of iterations which consists P of a set A of sequences {ci }, where each ci is DD3 (pi |r) sequence and i pi = Dr3 [f ] (cf. Definition 2.4). (4.1) X X L(f n ) = b1 reg1 (n) + b2 reg2 (n) + b4 reg4 (n) + bk regk (n) = ci (n). k∈G i Note that, by Lemma 2.7, each ci has one of the forms (A)-(G), (E’)(G’) which implies that pi ≤ 2. Let us remind to the reader that by Theorem 2.9: (4.2) Dr3 [f ] ∈ [#G − #H, #G − #H + 2]. For f , a self-map S 3 , the following lemma holds. 8 GRZEGORZ GRAFF AND JERZY JEZIERSKI Lemma 4.3. If |β| = |deg(f )| > 1, 4|r and r > 4 then each set of sequences A realizing Dr3 [f ] contains a sequence of the type (B)-(E) with the basic sequence a4 reg4 . Proof. Suppose that in A there is no sequence of one of the types (B)2 4 ≤ −3, for |β| > 1, which implies the (E) with a4 reg4 . Then b4 = β −β 4 existence of three sequences: (4.3) γ1 , γ2 , γ3 of the type (G’) or (E’), since only these give the negative contribution to b4 . We will show that this leads to contradiction with minimality of A. For odd d (d > 1, 4d|r) let us consider the following triple which appears in the formula (4.1): (4.4) bd regd (n) + b2d reg2d (n) + b4d reg4d (n). By Lemma 3.1 each coefficient in (4.4) is non-zero. A sequence (G’) may be used only if it reduces a part of such triple, namely its two last terms. Then we have to use one sequence (B)-(E) to reduce bd regd , which gives the contribution of the triple (4.4) to Dr3 [f ] equal to 2 + 1 = 3. On the other hand, we may use one (F) or (G) to reduce two first terms and one sequence (B)-(E) to reduce the last. As a result we get the smaller contribution 1 + 1 = 2. Suppose that among γ1 , γ2 , γ3 in (4.3) there are three (G’)’s. They give the contribution to Dr3 [f ] equal to 9. On the other hand, independently on the values of b1 , b2 , b4 , there is a smaller realization: we may reduce these 3 triples by 6 other sequences in the way indicated above and b1 reg1 (n) + b2 reg2 (n) + b4 reg4 (n) by one sequence (A) and one (B)-(E). This gives 8 sequences in total, contradicting to minimality. Similarly, we get a contradiction when we assume that at least one of the three sequences γ1 , γ2 , γ3 is of the type (E’) (we replace the expression cE 0 = reg2 (n) − reg4 (n) + a2d reg2d (n) which counts with the multiplicity 2 with cD = a2d reg2d (n) of multiplicity 1 and repeat the same reasoning as in the case of 3 sequences (G)). ¤ Corollary 4.4. Under the assumptions of Lemma 4.3 Dr3 [f ] ∈ [#G − #H + 1, #G − #H + 2]. Proof. By Theorem 2.9 it is enough to exclude the possibility Dr3 [f ] = #G − #H. Assume that A realizes Dr3 [f ] = #G − #H. By Lemma 4.3 there is a sequence of one of the types (B)-(E) in A with the term a4 reg4 . The remaining #G − #H − 1 sequences (counting multiplicity) must reduce bk regk for k ∈ G. Only sequences (F) and (G) reduce two bk regk , k ∈ G and we may use them in such a way #H times ((F’), (G’) MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 9 also reduce two bk regk ’s but they are counted twice). Now the remaining #G − 2#H bk regk ’s must be reduced by #G − 2#H − 1 sequences (A)(E), (E’)-(G’) (counting multiplicity). This is impossible, since each of (A)-(E) reduces at most one bk regk with k ∈ G, and (E’)-(G’) is counted twice and reduces at most two bk regk with k ∈ G. ¤ Lemma 4.5. Suppose that |deg(f )| > 1 and Dr3 [f ] = #G−#H +1. Then there is a minimal set realizing Dr3 [f ] containing only sequences (B)-(G). Moreover we may assume that this set contains one sequence of type (B)-(E) with ad regd such that d = 4 and #G − 2#H sequences of the type (B)-(E) with d 6= 4 and #H sequences of the type (F)-(G). ¤ Proof. Let us fix a minimal set A. By Lemma 4.3 there is a sequence c(4) in A of one of the types (B)-(E) with a4 reg4 . We take then a4 = b4 . The remaining sequences realize bk regk for k ∈ G, hence each of them must realize at least one bk regk so none of them is of the type (A). Let us assume that A contains a sequence of the type (G’) cG0 = reg2 − reg4 + a2d reg2d + a4d reg4d . We will show that it is possible to exchange A into another minimal system A0 in which cG0 does not appear. Namely, we change 3 sequences in A: • instead of cG0 we take two sequences of the type (D): a2d reg2d , a4d reg4d ; • instead of c(4) with a4 reg4 we take c0(4) with (a4 − 1)reg4 . 2 • Let us notice that b2 = β−β < 0 implies the existence of a 2 sequence ψ(n) of the type (E) or (G), since only these have negative contribution to b2 . Instead of such sequences we take ψ 0 (n) = reg2 + ψ(n), which is an expression of the type (B) or (F) respectively. This gives a system A0 with the same multiplicity as A (so also minimal), with one expression of the type (G’) less. Similarly we may remove expressions of type (E’) and (F’). ¤ Theorem 4.6. Let r be even, |β| > 1, then Dr3 [f ] ∈ {ζ(r)−η(r)−1, ζ(r)− η(r)}. What is more, Dr3 [f ] = ζ(r) − η(r) − 1 if and only if one of the two following conjunctions holds: (4.5) (i) ζ(r) − η(r) ≥ β 2 (ii) ζ(r) − 3η(r) ≥ β − 2 and and β 2 −β 2 β 2 −β 2 − η(r) ≥ 0, − η(r) ≤ −1. 10 GRZEGORZ GRAFF AND JERZY JEZIERSKI Before we give the proof of the above theorem, we illustrate it by the following example: Example 4.7. Let f : S 3 → S 3 have the degree β = 2 and let r = 12. The Lefschetz numbers L(f n ) are equal to 1 − 2n . We represent this sequence (for n|12) in the form of the periodic expansion. We get that 2 2 4 b1 = 1 − β = −1, b2 = β−β = −1, b4 = β −β = −3, thus: 2 4 (4.6) L(f n ) = −reg1 (n)−reg2 (n)−3reg4 (n)+b3 reg3 (n)+b6 reg6 (n)+b12 reg12 (n). There is: ζ(12) = 6, η(12) = 2. Because ζ(r) − 3η(r) = β − 2 = 0 and β 2 −β − η(r) = −1, we see that the second condition of (4.5) is satisfied 2 and thus Dr3 [f ] = ζ(r) − η(r) − 1 = 3. Indeed, we may take the 3 following sequence, which in sum realize {L(f n )}n|12 : cG (n) = reg1 (n) − reg2 (n) + b3 reg3 (n) + b6 reg6 (n), cC1 (n) = −reg1 (n) − 3reg4 (n), cC2 (n) = −reg1 (n) + b12 reg12 (n). Proof of Theorem 4.6. First notice that if |β| > 1, then D23 (f ) = 1 and D43 (f ) = 2, thus we verify straightforwardly that for r = 2 and r = 4 the theorem holds. For r > 4 we divide the proof into two cases. PART (I): 4|r. By Corollary 4.4 Dr3 [f ] ∈ {#G−#H +1, #G−#H +2}, which is equal, by item (2) of Proposition 3.3 for 4|r, to {ζ(r) − η(r) − 1, ζ(r) − η(r)}. We will now find the conditions equivalent to the statement that Dr3 [f ] = #G − #H + 1 = ζ(r) − η(r) − 1. By Lemma 4.5 we may assume that there is a minimal set of sequences A realizing Dr3 [f ] with #H sequences (F) or (G) and #G−2#H sequences (B)-(F), which reduce bk regk for k 6= 4, and one extra expression c(4) of the type (B), (C), (D) or (E), which reduces b4 reg4 . Thus there are #G − #H sequences plus one extra in A. Let us denote the contribution of the single sequence c(4) to the first two terms of the formula (4.1) by ²1 reg1 (n) + ²2 reg2 (n), where (²1 , ²2 ) ∈ {(1, 0), (−1, 0), (0, 0)(1, −1)}. By mX , where X ∈ {B, C, D, E, F, G} we denote the number of sequences (not counting c(4) ) of the given type in the minimal realization A. Then Dr3 [f ] = #G − #H + 1 if and only if there are integers MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 11 (²1 , ²2 ) ∈ {(1, 0), (−1, 0), (0, 0)(1, −1)} such that the following system of equations has an integer solution in non-negative unknowns mX : (4.7) mB mB +mC −mC +mD +mE +mE −mE +mF +mF +mF +mG +mG −mG +mG = #G − #H, = b1 − ²1 , = b2 − ²2 , = #H, where the first and last equations describe the number of sequences (not counting c(4) ), the second and third - their contribution to the first two terms of periodic expansion. The above system is equivalent to: mD (4.8) +mB +mB +mF +mF +mF +mG +mG +mG +mG = #G − #H − mC − mE , = b1 + mC − mE − ²1 , = #H, = −b2 − mE + ²2 . Notice that: • for any fixed mC , mE (4.8) is a Cramer system with the determinant +1, thus mD , mB , mF , mG are uniquely determined. • If mC , mE are integers, then the other unknowns must be integers. • For any fixed values of mC ≥ 0, mE ≥ 0 the solutions of (4.8) are non-negative if and only if the following system of inequalities holds: (4.9) 0 ≤ −b2 − mE + ²2 ≤ #H ≤ b1 + mC − mE − ²1 ≤ #G − #H − mC − mE . As a consequence of the above facts, in order to find the solution of the system (4.7), it is enough to solve (4.9) with the unknowns mC , mE , such that mC ≥ 0, mE ≥ 0. We rewrite (4.9) in the system of four inequalities: (4.10) mE ≤ −b2 + ²2 , (4.11) mE ≥ −b2 + ²2 − #H, (4.12) mC − mE ≥ #H − b1 + ²1 , 1 (−b1 + #G − #H + ²1 ). 2 The problem of finding the needed solution of (4.7) reduces to the question if there is a point in the plane (mC , mE ) ∈ Z2 , mC ≥ 0, mE ≥ 0 for which inequalities (4.10)-(4.13) are satisfied. (4.13) mC ≤ 12 GRZEGORZ GRAFF AND JERZY JEZIERSKI We substitute the values of #G and #H using Lemma 3.3 part (2) and Lemma 3.4 and the values of b1 and b2 calculated straightforwardly: 1 #G = ζ(r) − 3; #H = η(r) − 1; b1 = 1 − β; b2 = (β − β 2 ). 2 Then the inequalities (4.10)-(4.13) can be rewritten as: (4.14) (4.15) (4.16) (4.17) 1 (ζ(r) − η(r) + β − 3 + ²1 ), 2 mC ≤ mE ≥ 1 2 (β − β) − η(r) + 1 + ²2 , 2 mE ≤ 1 2 (β − β) + ²2 , 2 mE ≤ mC − η(r) − β + 2 − ²1 . Now the problem becomes: for which numbers: r ∈ N, β ∈ Z (4|r , |β| ≥ 2) one can choose parameters (²1 , ²2 ) ∈ {(0, 0), (1, 0), (−1, 0), (1, −1)} so that the inequalities (4.14) - (4.17) have a non-negative integer solution (mC , mE ). To simplify the notations we denote: 1 b = (ζ(r) − η(r) + β − 3 + ²1 ), 2 1 c = (β 2 − β) − η(r) + 1 + ²2 , 2 1 d = (β 2 − β) + ²2 , 2 e = −η(r) − β + 2 − ²1 . Now the system of inequalities (4.14) - (4.17) takes the form: (4.18) mC ≤ b; c ≤ mE ; mE ≤ d; mE ≤ mC + e. Lemma 4.8. Let b, c, d, e ∈ R satisfy: c, e ∈ Z and c ≤ d. Then the inequalities (4.18) have a non-negative integer solution (mC , mE ) ⇐⇒ b ≥ 0, d ≥ 0, max{c, 0} ≤ [b] + e, where [b] denotes the integer part of b. In order to proof the above lemma it is enough to notice that the first three inequalities give (−∞, b > × < c, d > while the last defines the half-plane under the line mE = mC + e in (mC , mE ) coordinates. Finally by Lemma 4.8 the problem becomes: for which β and r one can choose such parameters (²1 , ²2 ) that the following inequalities hold. (4.19) ζ(r) − η(r) + β − 3 + ²1 ≥ 0, MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 (4.20) 13 1 2 (β − β) + ²2 ≥ 0, 2 (4.21) 1 1 max{ (β 2 −β)−η(r)+1+²2 , 0} ≤ [ (ζ(r)−η(r)+β−3+²1 )]−η(r)−β+2−²1 . 2 2 We notice that the inequality (4.20) always hold. In fact |β| ≥ 2 implies β 2 − β ≥ 2 and 21 (β 2 − β) + ²2 ≥ 1 − 1 ≥ 0. Now we study (4.19). We will consider four cases. Case (1) ζ(r) − η(r) + β ≤ 1. Then the inequality (4.19) never holds, hence the system has no solution. Case (2) ζ(r) − η(r) + β = 2 and Case (3) ζ(r) − η(r) + β = 3 will be discussed individually. Case (4) ζ(r) − η(r) + β ≥ 4. Then the inequality (4.19) holds for each ²1 . We will consider these cases (in reverse order: starting from Case 4 to Case 1) as the assumptions in the next subcases. We will look for the solutions of the inequality (4.21). Case (4) We assume that ζ(r)−η(r)+β ≥ 4 (the inequality (4.19) holds for each ²1 ). To release from maximum and the integer part in (4.21) we consider several subcases. Subcase (4.≥) 21 (β 2 − β) − η(r) ≥ 0. Now c = 12 (β 2 − β) − η(r) + 1 − ²2 ≥ 0, since ²2 ≥ −1. The inequality (4.21) takes the form: 1 2 1 (β − β) − η(r) + 1 + ²2 ≤ [ (ζ(r) − η(r) + β − 3 + ²1 )] − η(r) − β + 2 − ²1 , 2 2 or 1 2 1 (β + β) − 1 + ²1 + ²2 ≤ [ (ζ(r) − η(r) + β − 3 + ²1 )]. 2 2 We notice that if the above inequality holds for a (²1 , ²2 ) ∈ {(0, 0), (1, 0), (−1, 0), (1, −1)} then it also holds for (−1, 0), thus it is enough to solve 1 2 1 (β + β) − 2 ≤ [ (ζ(r) − η(r) + β − 4)], 2 2 hence 1 2 1 (β + β) ≤ [ (ζ(r) − η(r) + β)]. 2 2 Subsubcase (4.≥.even) ζ(r) − η(r) + β is even. Now we may leave the integer part: 1 1 2 (β + β) ≤ (ζ(r) − η(r) + β), 2 2 which implies β 2 ≤ ζ(r) − η(r). 14 GRZEGORZ GRAFF AND JERZY JEZIERSKI Subsubcase (4.≥.odd) ζ(r) − η(r) + β is odd. Now we get: 1 1 1 2 (β + β) ≤ (ζ(r) − η(r) + β) − , 2 2 2 which implies β 2 ≤ ζ(r) − η(r) − 1. Moreover we notice that in this subsubcase the above inequality is equivalent to β 2 ≤ ζ(r) − η(r). In fact, by the parity assumption in this subsubcase we have that ζ(r) − η(r) + β ≡ 1 (mod 2), and thus ζ(r) − η(r) − β 2 ≡ 1 (mod 2). As a consequence the equality β 2 = ζ(r) − η(r) can not hold. Thus the assumptions of Case 4, Subcase (4.≥) and the obtained inequality give the following system of conditions: (4.22) 2 ζ(r) − η(r) + β ≥ 4 and β 2−β − η(r) ≥ 0 and ζ(r) − η(r) ≥ β 2 . Subcase (4.<) 21 (β 2 − β) − η(r) < 0. Now c = 12 (β 2 − β) − η(r) + 1 + ²2 ≤ 0 (for any ²2 ) hence inequality (4.21) becomes 1 0 ≤ [ (ζ(r) − η(r) + β − 3 + ²1 )] − η(r) − β + 2 − ²1 . 2 Let us notice that if the inequality holds for an −1 ≤ ²1 ≤ +1, then it also holds for ²1 = −1, hence we get 1 η(r) + β − 3 ≤ [ (ζ(r) − η(r) + β) − 2]. 2 Subsubcase (4.<.even) ζ(r) − η(r) + β is even. Now η(r) + β − 3 ≤ 21 (ζ(r) − η(r) + β) − 2 or ζ(r) − 3η(r) ≥ β − 2. Subsubcase (4.<.odd) ζ(r) − η(r) + β is odd. Now η(r) + β − 3 ≤ 21 (ζ(r) − η(r) + β) − 2 − 21 hence we get ζ(r) − 3η(r) ≥ β − 1. Moreover we notice that in this subsubcase the above inequality is equivalent to ζ(r) − 3η(r) ≥ β − 2. In fact the equality ζ(r)−3η(r) = β −1 can not hold because of the parity assumptions in (4.<.odd). MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 15 Thus, Case 4, Subcase (4.<) and the obtained inequality give the following system of conditions: (4.23) 2 ζ(r) − η(r) + β ≥ 4 and β 2−β − η(r) ≤ −1 and ζ(r) − 3η(r) ≥ β − 2. Case (3) ζ(r) − η(r) + β = 3. The assumption of Case 3 implies that 1 1 b = (ζ(r) − η(r) + β − 3 + ²1 ) = ²1 . 2 2 Now b ≥ 0 for ²1 = 0 or +1. Subcase (3.0) ²1 = 0. Since ²1 = 0 implies ²2 = 0, the inequality (4.21) takes the form 1 max{ (β 2 − β) − η(r) + 1, 0} ≤ −η(r) − β + 2. 2 Subsubcase (3.0.≥) 12 (β 2 − β) − η(r) + 1 ≥ 0. Now we get 1 2 (β − β) − η(r) + 1 ≤ −η(r) − β + 2, 2 which is equivalent to β2 + β ≤ 2 and the last holds only for β = −2. Then the conditions (3.0.≥) and of Case (3) take the form: 4 ≥ η(r) and ζ(r) = η(r) + 5 respectively. Since η(r)|ζ(r), η(r) = 1 and ζ(r) = 6. This implies r = 25 . Subsubcase (3.0.<) 12 (β 2 − β) − η(r) + 1 < 0. We get: 0 ≤ −η(r) − β + 2. In other words η(r) ≤ −β + 2. On the other hand the assumption (3.0.<) gives β 2 − β < 2η(r) − 2. The above inequalities imply β 2 + β − 2 < 0, which is never true for |β| > 1. Subcase (3.1) ²1 = 1. 1 max{ (β 2 − β) − η(r) + 1 + ²2 , 0} ≤ −η(r) − β + 1. 2 Since ²2 may be 0 or −1, here we may put ²2 = −1 which implies 1 max{ (β 2 − β) − η(r), 0} ≤ −η(r) − β + 1. 2 Subsubcase (3.1.≥) 12 (β 2 − β) − η(r) ≥ 0. Now 1 2 (β − β) − η(r) ≤ −η(r) − β + 1 2 16 GRZEGORZ GRAFF AND JERZY JEZIERSKI implies β 2 + β − 2 ≤ 0, hence β = −2. Then the assumption of Case 3: ζ(r) − η(r) + β = 3 gives ζ(r) = η(r) + 5. On the other hand the condition (3.1.≥) takes the form: 3 ≥ η(r). Again η(r)|ζ(r) implies η(r) = 1, ζ(r) = 6 and thus r = 25 . Subsubcase (3.1.<) 12 (β 2 − β) − η(r) < 0. The inequality (4.21) takes the form 0 ≤ −η(r) − β + 1, hence implies η(r) ≤ −β + 1. We join this inequality with the condition (3.1.<) and get 1 2 (β − β) < η(r) ≤ −β + 1. 2 This implies the inequality 1 2 (β − β) < −β + 1, 2 which is not valid for any |β| > 12. Case (2) ζ(r) − η(r) + β = 2. Here we get: 1 1 (ζ(r) − η(r) + β − 3 + ²1 ) = (²1 − 1). 2 2 Notice that b ≥ 0 only for ²1 = +1 and then b = 0. This yields (4.21) has the form 1 max{ (β 2 − β) − η(r) + 1 + ²2 , 0} ≤ −η(r) − β + 1. 2 Since ²2 may be 0 or −1, it is enough to consider ²2 = −1. We get then: b= 1 max{ (β 2 − β) − η(r), 0} ≤ −η(r) − β + 1. 2 Subcase (2.≥) 21 (β 2 − β) − η(r) ≥ 0. The inequality (4.21) takes the form 1 2 (β − β) − η(r) ≤ −η(r) − β + 1, 2 which implies β 2 + β ≤ 2 hence β = −2. Moreover the conditions (2.≥) and of Case (2) become: 3 ≥ η(r) and ζ(r) = η(r) + 4 respectively. Since η(r)|ζ(r), there are two possibilities: η(r) = 1 or η(r) = 2. In the first case ζ(r) = 5 and r = 24 . In the second case ζ(r) = 6 and r = 22 p where p is a prime number greater than 2. Now we get either • β = −2 and r = 24 or • β = −2 and r = 4p for an odd prime number p. MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 17 Subcase (2.<) 21 (β 2 − β) − η(r) < 0. Now the inequality (4.21) takes the form 0 ≤ −η(r) − β + 1, which implies η(r) ≤ −β + 1. On the other hand the condition (2.<) gives β 2 − β < 2η(r). This implies β 2 + β < 2 which has no solution for |β| > 1. Summing up the previous considerations, we get that Dr3 [f ] = ζ(r) − η(r) − 1 if and only if one of the five following conjunctions hold (cf. (4.22), (4.23), Cases (3) and (2)): (4.24) β 2 −β (1) ζ(r) − η(r) + β ≥ 4 and ζ(r) − η(r) ≥ β 2 and − η(r) ≥ 0, 2 β 2 −β (2) ζ(r) − η(r) + β ≥ 4 and ζ(r) − 3η(r) ≥ β − 2 and − η(r) ≤ −1, 2 (3) β = −2 and r = 16, (4) β = −2 and r = 32, (5) β = −2 and r = 4p for an odd prime p. We show now that the system of conditions (4.24) is equivalent to the conditions (i) and (ii) in Theorem 4.5, which we remind below: (4.25) (i) ζ(r) − η(r) ≥ β 2 (ii) ζ(r) − 3η(r) ≥ β − 2 and and β 2 −β 2 β 2 −β 2 − η(r) ≥ 0, − η(r) ≤ −1. We have to show that [(i) or (ii)] ⇐⇒ [(1) or (2) or (3) or (4) or (5)]. First we prove that (ii) ⇐⇒ (2). ⇐ is trivial. To prove ⇒ it remains to show that the second at the third inequalities in (2) imply the first. The second and third inequality give 2 respectively ζ(r) − η(r) ≥ 2η(r) + β − 2 and η(r) ≥ β 2−β + 1. This implies that ζ(r) − η(r) ≥ β 2 , and as β 2 ≥ 4 − β for β 6= −2, we get that the first inequality results from the second and the third for β 6= −2. If β = −2, we get η(r) ≥ 4 and we should check whether ζ(r) − η(r) ≥ 2η(r) − 4 implies ζ(r) − η(r) ≥ 6. This is obviously satisfied if η(r) > 4, and for η(r) = 4 (because 4|r) we get that ζ(r) must be greater than 10 so ζ(r) − η(r) ≥ 6 is in this case also satisfied. We obtained that (2) ⇐⇒ (ii). Now, to prove the Theorem, it remains to show that (i) ⇐⇒ [(1) or (3) or (4) or (5)]. ⇐ (1) implies (i) in a trivial way. Then we check case by case that each of (3), (4), (5) implies (i). ⇒ We show that (i) and the negation of (1) imply the alternative [(3) or (4) or (5)]. 18 GRZEGORZ GRAFF AND JERZY JEZIERSKI Recall that (i) means β2 − β − η(r) ≥ 0. 2 Now the negation of (1) means that the first inequality in (1) is not true, i.e., ζ(r) − η(r) + β < 4. The two above inequalities imply: β 2 ≤ ζ(r) + η(r) < 4 − β hence β 2 + β − 4 < 0 which holds only for β = −2. This in turn implies ζ(r) − η(r) ≥ β 2 and ζ(r) − η(r) ≥ 4 ; η(r) ≤ 3 ; ζ(r) − η(r) < 6. Now we get (ζ(r) − η(r) = 4 or 5) and η(r) ≤ 3. If ζ(r) − η(r) = 5 then η(r) = 1 hence r = 25 = 32, so we get the case (4). If ζ(r) − η(r) = 4 then η(r) = 1 or 2. For η(r) = 1 we get r = 24 = 16 so we obtain the case (3). Finally for η(r) = 2, r = 4p, where p is an odd prime, which gives the case (5). This ends the proof of the equivalency of (4.24) and (4.25) and the proof of Part I. PART (II): 2|r but 4 6 |r. Because b4 = 0, we see that Dr3 [f ] ∈ [#G − #H, #G − #H + 1] = [ζ(r) − η(r) − 1, ζ(r) − η(r)], where the last equality results from Lemma 3.4 and item (2) of Proposition 3.3 for 4 6 |r. Searching for the conditions equivalent to Dr3 [f ] = #G−#H, we repeat the same reasoning as in the case 4|r, with the only difference which is due to the fact that we do not need additional sequence to reduce b4 reg4 , so we have to find non-negative integer solutions of (4.7) without parameters ²1 , ²2 . Now we follow the previous part and taking into account that #G = ζ(r) − 2, we get instead of the inequalities (4.19)-(4.21): (4.26) ζ(r) − η(r) + β − 2 ≥ 0, (4.27) 1 2 (β − β) ≥ 0, 2 1 1 (4.28) max{ (β 2 −β)−η(r)+1, 0} ≤ [ (ζ(r)−η(r)+β−2)]−η(r)−β+2. 2 2 We notice that the inequality (4.27) is always satisfied for |β| > 1. Since the inequality (4.26) is an explicit condition, it remains to solve (4.28). Again we consider some subcases. Case (>) 12 (β 2 − β) − η(r) + 1 > 0. MINIMAL NUMBER OF PERIODIC POINTS FOR SELF-MAPS OF S 3 19 Subcase (>.even) ζ(r) − η(r) + β − 2 is even. Then the inequality (4.28) becomes 1 1 2 (β − β) − η(r) + 1 ≤ (ζ(r) − η(r) + β − 2) − η(r) − β + 2, 2 2 which is equivalent to β 2 ≤ ζ(r) − η(r). Subcase (>.odd) ζ(r) − η(r) + β − 2 is odd. Then we get β 2 ≤ ζ(r) − η(r) − 1. Since by the parity assumption the equality β 2 = ζ(r)−η(r) can not hold, the last inequality is equivalent to β 2 ≤ ζ(r) − η(r). (4.29) Finally, as the inequality (4.26) follows from (4.29) we get the conjunction (i) of the conditions (4.5). Case (≤) 12 (β 2 − β) − η(r) + 1 ≤ 0. Subcase (≤.even) ζ(r) − η(r) + β − 2 is even. The inequality (4.28) becomes 1 (ζ(r) − η(r) + β − 2) − η(r) − β + 2, 2 which is equivalent to 0≤ ζ(r) − 3η(r) ≥ β − 2. Subcase (≤.odd) ζ(r) − η(r) + β is odd. Then we get ζ(r) − 3η(r) ≥ β − 1. Again by the parity assumption we may weaken the inequality putting (4.30) ζ(r) − 3η(r) ≥ β − 2. As (4.30) and 12 β 2 − β) − η(r) + 1 ≤ 0 imply the inequality (4.26) we get the conjunction (ii) of the conditions (4.5). This ends the proof of the Part II and the proof of the whole theorem. References 1. I. K. Babenko, S. A. Bogatyi, The behavior of the index of periodic points under iterations of a mapping, Math. USSR Izv. 38 (1992), 1–26. 2. R. F. Brown, The Lefschetz Fixed Point Theorem, Glenview, New York, 1971. 3. S. N. Chow, J. Mallet-Paret, J. A. Yorke, A periodic point index which is a bifurcation invariant, Geometric dynamics (Rio de Janeiro, 1981), 109–131, Springer Lecture Notes in Math. 1007, Berlin 1983. 4. A. Dold, Fixed point indices of iterated maps, Invent. Math. 74 (1983), 419–435. 20 GRZEGORZ GRAFF AND JERZY JEZIERSKI 5. G. Graff and J. Jezierski, Minimal number of periodic points for C 1 self-maps of compact simply-connected manifolds, preprint. 6. G. Graff and P. Nowak-Przygodzki, Fixed point indices of iterations of C 1 maps in R3 , Discrete Cont. Dyn. Systems 16 (2006), no. 4, 843–856. 7. M. Greenberg, Lectures on algebraic topology, Benjamin, Menlo Park, CA, 1971. 8. Huai-Dong Cao and Xi-Ping Zhu, A Complete Proof of the Poincaré and Geometrization Conjectures - application of the Hamilton-Perelman theory of the Ricci flow, Asian J. Math. 10 (2006), no. 2, 165–492. 9. J. Jezierski, Wecken’s theorem for periodic points in dimension at least 3, Topology Appl. 153 (2006), no. 11, 1825–1837. 10. B. J. Jiang, Lectures on the Nielsen Fixed Point Theory, Contemp. Math. 14, Amer. Math. Soc., Providence 1983. 11. J. Jezierski and W. Marzantowicz, Homotopy methods in topological fixed and periodic points theory, Topological Fixed Point Theory and Its Applications, 3. Springer, Dordrecht, 2005. 12. J. Llibre, J. Paranõs, J. A. Rodriguez, Periods for transversal maps on compact manifolds with a given homology, Houston J. Math. 24 (1998), No. 3, 397–407. 13. W. Marzantowicz and P. Przygodzki, Finding periodic points of a map by use of a k-adic expansion, Discrete Cont. Dyn. Systems 5 (1999), 495–514. 14. M. Shub, P. Sullivan, A remark on the Lefschetz fixed point formula for differentiable maps, Topology 13 (1974), 189–191. Faculty of Applied Physics and Mathematics, Gdansk University of Technology, Narutowicza 11/12, 80-952 Gdansk, Poland E-mail address: [email protected] Institute of Applications of Mathematics, Warsaw University of Life Sciences (SGGW), Nowoursynowska 159, 00-757 Warsaw, Poland E-mail address: [email protected]