How to Detect Hayman Directions
Transcription
How to Detect Hayman Directions
Computational Methods and Function Theory Volume 9 (2009), No. 1, 57–64 How to Detect Hayman Directions Andreas Sauer (Communicated by Ilpo Laine) Abstract. We introduce the class of Hayman exceptional functions which seem to play a similar role for the study of Hayman directions as Julia exceptional functions do for Julia directions. Further we prove that for every b there exists n ∈ N such that transcendental meromorphic function f : C → C f (k) has a Julia direction for every k ≥ n. Keywords. Hayman direction, Julia direction, filling disks, cercles de remplissages, singular direction. 2000 MSC. 30D35. 1. Julia directions b In [5] Julia proved that for transcendental meromorphic functions f : C → C with an asymptotic value, Picard’s Theorem holds on relatively small subsets of the plane. Julia considered a quite general class of sectorial neighbourhoods of paths to ∞ in C, but most of the following research on this topic was restricted to the case of sectors. It is standard to call a ray J from 0 to ∞ a Julia direction if, in every sector ∆ containing J, Picard’s Theorem holds for f , that is, f takes b with at most two exceptions infinitely often in ∆. every value of C With the help of the theory of normal families (see [9]), in particular Montel’s Theorem and Marty’s Theorem, Julia’s proof yields the following result, involving the spherical derivative f # := |f 0 |/(1 + |f |2 ). b with a sequence zj → ∞ Theorem 1. Every meromorphic function f : C → C such that (1) |zj |f # (zj ) → ∞ possesses a Julia direction. Received October 31, 2007, in revised form January 21, 2008. Published online February 26, 2008. c 2009 Heldermann Verlag ISSN 1617-9447/$ 2.50 58 A. Sauer CMFT The existence of such a sequence zj is the most common criterion for the existence of a Julia direction. The points zj serve as centres of so called filling disks (or cercles de remplissage) for f ; see [15] for the exact definition. Hence meromorphic functions without Julia directions necessarily satisfy 1 # (2) f (z) = O |z| for z → ∞. Transcendental meromorphic functions satisfying (2) exist and are called Julia exceptional functions. A. Ostrowski considered this class of functions in his classical paper [6]. It contains a complete characterization of Julia exceptional functions in terms of the distribution of zeros and poles of f [6, Satz 5]. The following example of a Julia exceptional function is also given there [6, p. 258]: ∞ n Y q −z (3) F (z) = qn + z n=0 where q > 1. This function has indeed no Julia direction. An elementary proof of this fact can be found in [10]. Condition (2) implies that f grows very slowly: for the unintegrated AhlforsShimizu characteristic Z 1 (f # (z))2 dz A(r, f ) = π |z|≤r it follows that A(r, f ) = O(log r), and thus for the Nevanlinna characteristic that (4) T (r, f ) = O((log r)2 ). In this sense “most” transcendental meromorphic functions posses a Julia direction. Furthermore, this growth condition is sharp since the growth of T (r, F ) for the function F in (3) is of the same magnitude as (log r)2 . 2. Hayman directions The starting point of our investigations is Hayman’s classical theorem [4, Cor. to Thm. 3.5] on the value distribution of f and its derivatives: b takes every Theorem 2. Every transcendental meromorphic function f : C → C value of C infinitely often or every derivative f (k) takes every value in C \ {0} infinitely often. Now it is natural to consider the question, whether there exist directions such that Hayman’s Theorem is true in every sector containing the direction. Directions with this property are called Hayman directions and have been considered by several authors (see [14, 15, 2, 3, 13]). The basic theorem was proved by Yang Lo [14]. 9 (2009), No. 1 How to Detect Hayman Directions 59 b be a transcendental meromorphic function with Theorem 3. Let f : C → C (5) lim sup r→∞ T (r, f ) = ∞. (log r)3 Then f possesses a Julia direction that is a Hayman direction. The fact that the Hayman direction detected in Lo’s Theorem is also a Julia direction was not explicitly stated in [14], but it follows directly from the proof. Further an example of Rossi [7] shows that the growth condition (5) for this formulation of Lo’s Theorem is sharp: (6) G(z) = ∞ Y q n=0 √ n √ q n −z +z with q > 1. For this function, T (r, G) ∼ (log r)3 /3, it has exactly two Julia directions (the positive and negative imaginary axis), but neither of them is a Hayman direction. This does not mean that G has no Hayman direction at all. The only candidate for a Hayman direction is the negative real axis, though. In small sectors around the positive real axis G is bounded. Around other directions except the negative real axis G omits 0 and ∞ and grows like a polynomial, which together with Cauchy’s formula shows that from some k on all derivatives G(k) are bounded in slightly smaller sectors. (All these properties of Rossi’s example follow from the analysis of that example given in [7] and [2].) There is also a condition for sequences similar to the one in Theorem 1 (see [3]): b with a sequence zj → ∞ Theorem 4. Every meromorphic function f : C → C such that (7) |zj |f # (zj ) →∞ log |zj | possesses a Julia direction that is a Hayman direction. The proof uses filling disks where the zj are the centres. Since (7) implies (1) the detected Hayman direction is again a Julia direction. The condition (7) has been studied in its own right. In particular it was proved by Toppila [11] that there exists a sequence for which (7) holds if f has a Nevanlinna deficient value. The following corollary seems not to have been stated before, but it follows immediately from [11] and Theorem 4. b that has a Corollary 5. Every transcendental meromorphic function f : C → C Nevanlinna deficient value has a Julia direction that is a Hayman direction. 60 A. Sauer CMFT At this point it is worth noting that a Valiron deficient value already implies the existence of a Julia direction (see [16]). Toppila [11] also considered Valiron deficiencies. His result for this case implies, again with Theorem 4: b that has a Corollary 6. Every transcendental meromorphic function f : C → C Valiron deficient value and satisfies T (r, f ) →∞ (log r)2 has a Julia direction that is a Hayman direction. Unlike the case of Julia directions, where (1) implies the proper growth condition, it seems that (7) does not imply (5); if log |z| # (8) f (z) = O |z| then standard estimations show A(r, f ) = O((log r)3 ) and T (r, f ) = O((log r)4 ). On the other hand, we do not have an example with T (r, f ) ∼ (log r)4 and (8). Now we will consider directions that are not Hayman directions and derive results for the spherical derivative of the functions fk (z) := z −k f (z) with k ∈ N, that are analogous to (2). For this we need the normal family version of Hayman’s Theorem (see [9, Cor. 4.5.9]). Theorem 7. Let F be a family of meromorphic functions on a domain G such that there exist a ∈ C, b ∈ C \ {0} and k ∈ N with f (z) 6= a and f (k) (z) 6= b for all f ∈ F and z ∈ G. Then F is a normal family. From this we can prove the following lemma. b be meromorphic such that f e be a sector and let f : ∆ e →C Lemma 8. Let ∆ (k) omits a finite value and f omits a non-zero finite value for some k ∈ N. Then the function fk (z) := z −k f (z) satisfies fk# (z) = O(1/|z|) in any smaller sector e ∆ ⊂ ∆. e is strictly less than π. Proof. We can assume that the opening angle δ of ∆ e First we will show Further let eiα be the direction of the angle bisector of ∆. that every sequence gxn in the family F of all functions of the form b gx : D → C, gx (w) := x−k f (x(eiα + sw)) where s := sin(δ/2) and x > 1 possesses a subsequence that converges locally uniformly in D. If xn has a finite accumulation point x e then choose a subsequence xnk → x e and it follows that gxnk → gxe locally uniformly because of continuity. Suppose xn → ∞. First we assume that f omits 0. Then gxn omits 0 and gx(k) (w) = sk f (k) (x(eiα + sw)) n 9 (2009), No. 1 How to Detect Hayman Directions 61 omits sk b 6= 0 where b 6= 0 is the value omitted by f (k) . By Theorem 7 it follows that {gxn } is a normal family and thus possesses a convergent subsequence. If f omits a 6= 0 then consider iα −k gexn (w) := x−k n [f (xn (e + sw)) − a] = gxn (w) − xn a. (k) Then gexn omits 0 and gexn omits sk b 6= 0. Again it follows that {e gxn } is normal. −k Because xn a → 0 independently of w the same holds for {gxn }. From this and Marty’s Theorem we get for |w| ≤ r < 1 the existence of a constant Kr independent of x > 1 such that (9) x−k+1 |f 0 (x(eiα + sw))| ≤ Kr . 1 + |x−k f (x(eiα + sw))|2 Note that the left hand side of (9) is equal to gx# (w)/s. Since fk0 (z) = −kz −k−1 f (z) + z −k f 0 (z) we get fk# (z) ≤ k|z|−1 −k+1 0 |z −k f (z)| |f (z)| −1 |z| + |z| −k 2 −k 1 + |z f (z)| 1 + |z f (z)|2 The first term is O(1/|z|) since c/(1 + c2 ) is bounded for c ≥ 0. It remains to be shown that |z|−k+1 |f 0 (z)|/(1 + |z −k f (z)|2 ) is bounded in ∆. For ∆ there exists r with 0 < r < 1 such that every z ∈ ∆ with |z| > 1 can be written as z = x(eiα + sw) =: xζw with x > 1 and |w| ≤ r. With (9) it follows |z|−k+1 |f 0 (z)| |ζw |−k+1 x−k+1 |f 0 (xζw )| = 1 + |z −k f (z)|2 1 + |ζw |−2k |x−k f (xζw )|2 x−k+1 |f 0 (xζw )| |ζw |−k+1 (1 + |x−k f (xζw )|2 ) = 1 + |x−k f (xζw )|2 1 + |ζw |−2k |x−k f (xζw )|2 |ζw |−k+1 (1 + |x−k f (xζw )|2 ) ≤ Kr 1 + |ζw |−2k |x−k f (xζw )|2 Since δ < π we have 0 < s < 1 and obviously 1 − s < |ζw | < 1 + s. Set C := (1 − s)−k+1 (1 + s)2k . Then it is easy to check that |ζw |−k+1 (1 + |x−k f (xζw )|2 ) ≤ C, 1 + |ζw |−2k |x−k f (xζw )|2 which proves the lemma. We deduce the following theorem. b be a meromorphic function. If Theorem 9. Let f : C → C lim sup |z|fk# (z) = ∞ z→∞ for all k ∈ N in every sector around some direction H, then H is a Hayman direction. 62 A. Sauer CMFT To illustrate the usefulness of this theorem, we consider Ostrowski’s example F (z) given in (3). This function is a Julia exceptional function, which implies that F satisfies (2). From Marty’s Theorem the condition (2) is easily seen to be equivalent to the normality of every sequence F (σn z) in C \ {0} with arbitrary σn → ∞. Similarly, it follows from Marty’s Theorem that if for some k the sequence Fk (σn z) = F (σn z)/(σn z)k is not normal in a sector for a positive sequence σn → ∞, then |z|Fk# (z) is unbounded in that sector. We set σn = q n and consider gn (z) := F (q n z) in C \ {0}. Then gn (1) = 0 and gn (−1) = ∞. It follows that every limit g of the normal sequence gn is non-constant with g(1) = 0 and g(−1) = ∞. Given any sector ∆ around the negative real axis, one can find a neighbourhood U of z = −1 contained in ∆. For each k, the sequence Fk (q n z) = gn (z)/(q n z)k is not normal in U , since a limit must be ∞ at z = −1 and 0 in a punctured neighbourhood of z = −1. This follows from the fact that we may assume that gn has a non-constant limit. It follows that Theorem 9 can be applied, so for Ostrowski’s example the negative real axis is a Hayman direction that is not a Julia direction. To the best of our knowledge this is the first example of this type. From Lemma 8 we also get with a standard compactness argument: b that does not have Theorem 10. For every meromorphic function f : C → C a Hayman direction there exists a covering of C by a finite number of sectors ∆1 , . . . , ∆n such that for every ∆l there exists kl ∈ N with 1 # (10) fkl (z) = O |z| in ∆l . We propose to call a meromorphic function with such a sectorial covering of C a Hayman exceptional function. It is tempting to hope that Hayman exceptional functions satisfy (4), especially since this would solve a problem that has remained unsolved for some time now (see [1, Problem 1.31] and [15, Problem p. 121]). But this is not true: g(z) := z · G(z), with G being Rossi’s example (6), is a Hayman exceptional function with T (r, g) ∼ (log r)3 /3. The filling disks of G accumulate to the imaginary axis, so, with the notation of Lemma 8, |z|g1# (z) is bounded in any sector avoiding the imaginary axis. Around the imaginary axis g2 is bounded and hence |z|g2# (z) is also bounded. (For these facts we again refer to [7, 2].) On the other hand, a Hayman exceptional function cannot grow faster than the above example. b Theorem 11. For every Hayman exceptional function f : C → C, T (r, f ) = O((log r)3 ). 9 (2009), No. 1 How to Detect Hayman Directions 63 Proof. We use a lemma in [12, p. 277] in each of the sectors as described in Theorem 10. With the notation of this lemma we set w := fkl , g1 (z) := z kl , g2 = g3 := 0 and g4 := 1, so that g1 (z)w(z) + g2 (z) . f (z) = g3 (z)w(z) + g4 (z) From (10) it follows for the function denoted in the lemma by S(·, r; ∆) that S(w, r; ∆l ) = O(log r). From T (r, g1 ) = O(log r) and hence Z 128r T (t, g1 ) dt = O((log r)2 ) t 1 we obtain (by decreasing the ∆l slightly, such that they still cover the whole plane) S(f, r; ∆l ) = O(log(64r)) + O((log r)2 ) = O((log r)2 ). P From A(r, f ) ≤ nl=1 S(f, r; ∆l ) + O(1), where A(r, f ) again denotes the unintegrated Ahlfors–Shimizu characteristic, we get A(r, f ) = O((log r)2 ). This proves the theorem. This reproves Yang Lo’s Theorem (at least the part on Hayman directions), but not more and also in a rather indirect manner. Maybe further research on Hayman exceptional functions will show whether Yang Lo’s growth condition (5) is sharp. Note that so far there exists no example of a transcendental meromorphic function without a Hayman direction. 3. Julia directions of derivatives We now consider a problem that has certain similarities with the study of Hayman directions, namely the existence of Julia directions for f = f (0) and its derivatives f (k) . We have already obtained some results in this direction in [8]. It seems that up to now it is not even known, whether there exist transcendental meromorphic functions such that f and all its derivatives f (k) have no Julia direction. With the results from the last section we can prove that there exist no such functions. b be a transcendental meromorphic function. Then Theorem 12. Let f : C → C there exists n in N∪{0} such that f (k) possesses a Julia direction for every k ≥ n. Proof. Suppose that f is a Julia exceptional function. As already mentioned, Ostrowski [6, Satz 5] gave a characterization of Julia exceptional functions, and one condition [6, Satz 5, Part a)] a Julia exceptional function has to satisfy is |n(r, f, 0)−n(r, f, ∞)| = O(1). It follows that N (r, f, 0)−N (r, f, ∞) ≤ c log r for some c ≥ 0. Here n and N denote the usual counting functions from Nevanlinna theory. Choose any k with k > c. If f (k) has no Julia direction, then f (k) omits a finite non-zero value around every direction and we can apply Lemma 8 to f 64 A. Sauer CMFT and f (k) in a finite number of sectors that cover the whole plane. It turns out that fk (z) = z −k f (z) has to be a Julia exceptional function with N (r, fk , 0) − N (r, fk , ∞) = N (r, f, 0) − N (r, f, ∞) − k log r ≤ (c − k) log r → −∞. But this contradicts another of Ostrovski’s conditions [6, Satz 5, Part c)]. Thus f (k) possesses a Julia direction. If none of the f (k) is Julia exceptional, we have nothing to prove. If there is one such f (k) we can apply the above reasoning to it, rather than to f . References 1. K. F. Barth, D. A. Brannan and W. K. Hayman, Research problems in complex analysis, Bull. London Math. Soc. 16 (1984), 490–517. 2. P. C. Fenton and J. Rossi, Cercles de remplissage for entire functions, Bull. London Math. Soc. 31 (1999), 59–66. , Growth of functions in cercles de remplissage, J. Aust. Math. Soc. 72 (2002), 3. 131–136. 4. W. K. Hayman, Meromorphic Functions, Oxford University Press, London, 1975. 5. G. Julia, Quelques propri´et´es des fonctions m´eromorphes g´en´erales, C. R. Acad. Sci. 168 (1919), 718–720. ¨ 6. A. Ostrowski, Uber Folgen analytischer Funktionen und einige Versch¨arfungen des Picardschen Satzes, Math. Zeitschr. 24 (1924), 215–258. 7. J. Rossi, A sharp result concerning cercles de remplissage, Ann. Acad. Sci. Fenn. Ser. A I 20 (1995), 179–185. 8. A. Sauer, Julia directions of meromorphic functions and their derivatives, Arch. Math. 79 (2002), 182–187. 9. J. L. Schiff, Normal Families, Springer-Verlag, 1993. 10. N. Toda, Sur les directions de Julia et de Borel des fonctions algebroides, Nagoya Math. J. 34 (1969), 1–23. 11. S. Toppila, On the spherical derivative of a meromorphic function with a defficient value, Complex Analysis — Fifth Romanian-Finnish Seminar, Lecture Notes in Mathematics 1013, Springer, Berlin-New York, 1983, 373–393, 12. M. Tsuji, Potential Theory in Modern Function Theory, Maruzen, Tokyo, 1959; reprinted by Chelsea, New York, 1975. 13. Z.-J. Wu, On singular direction of meromorphic function and its derivatives, Kodai Math. J. 30 (2007), 55–60. 14. L. Yang, Meromorphic functions and their derivatives, J. London Math. Soc. 25 (1982), 288–296. 15. , Value Distribution Theory, Springer-Verlag, 1993. 16. T. Zinno and N. Toda, On Julia’s exceptional functions, Proc. Japan Acad. 42 (1966), 1120–1121. Andreas Sauer E-mail: andreas.sauer@fh-dortmund.de Address: FH Dortmund, Sonnenstr. 96, 44047 Dortmund, Germany.