Application of Infinite Sums and Products Entire function of finite order Munsik Kim Mathematical Science Department KAIST Nov 2021 Munsik Kim (KAIST) Entire function of finite order 11. 2021 1 / 32 Table of Contents 1 Overview 2 Main 3 Discussion Munsik Kim (KAIST) Entire function of finite order 11. 2021 2 / 32 Overview In Ch 8.2, we learn that an entire function can be written as m g (z) f (z) = z e ∞ Y En−1 n=1 z an (1) where g is entire and En (z)’s are Weierstrass elementary factors so that f vanishes to orderm at 0 and aj ’s are zeros of f listed with multiplicities. Note that the behavior of zeros of a holomorphic function is arbirary. In Ch 9, we will learn that some hypotheses are imposed about the general behavior of the function can control the behavior of zeros. Munsik Kim (KAIST) Entire function of finite order 11. 2021 3 / 32 Overview In Ch 9.1, we learned that a bounded holomorphic function f on D(0, 1) which has zero at aj counted according to their multiplicities and vanishes to order m ≥ 0 at z = 0 can be written as " ∞ # Y āj f (z) = z m · − Ba (z) · F (z) (2) |aj | j j=1 where F is a bounded holomorphic function on D(0, 1), zero free, and sup |f (z)| = z∈D(0,1) sup |F (z)|. z∈D(0,1) Here, Ba (z) is the Blaschke factor. In Ch9.3, we want to extend the domain D(0, 1) to C. Munsik Kim (KAIST) Entire function of finite order 11. 2021 4 / 32 Finite Order First, we look at the definition of the ”order” of the entire function : Definition Suppose f is an entire function. Then we say that f is of finite order if there exist a, r > 0 such that |f (z)| ≤ exp(|z|a ) for |z| > r . The finite order is one of the hypotheses. Let λ = λ(f ) be the infimum of all possible a satisfying above. We call such λ the order of f . i.e., sin z is of order 1, while exp(exp z) is not of finite order. Munsik Kim (KAIST) Entire function of finite order 11. 2021 5 / 32 Relation between the zeros and order of f Suppose {aj } are the zeros of f . Then, to see the relation between the zeros and order of f , consider X |an |−(k+1) . (3) n The infimum of all K ’s for which (3) is finite is an indicator of how many zeros of f are in D(0, r ) when r is large. For example, aj = j for j = 1, 2, √ · · ·, then for any K > 0, the sum converges, that is finite. If aj = j, then K would have to be greater than 1. Let n(r ) be the number of zeros in D(0, r ) of our entire function f . Munsik Kim (KAIST) Entire function of finite order 11. 2021 6 / 32 Main Theorem 1 Theorem 1 Let f be an entire function of finite order λ and satisfying f (0) = 1. If aj are the zeros of f listed with multiplicities, then ∞ X |an |−λ−1 < ∞ (4) n=1 To prove the Theorem 1, we need some technical Lemmas. Note that we assume that f (0) = 1 6= 0. This can be extended later. Munsik Kim (KAIST) Entire function of finite order 11. 2021 7 / 32 Lemmas Lemma 1 If f is an entire function with f (0) = 1, let M(r ) = max|f (z)| where |z|=r 0 < r < ∞. Then (log 2) · n(r ) ≤ log M(2r ) Proof of Lemma 1 By Jensen’s formula in Ch 9.1, n(2r ) 0 = log |f (0)| = − X k=1 log Z 2π 2r 1 + log |f (2re iθ )|dθ, |ak | 2π 0 (5) where ak ’s are the zeros of f in D(0, 2r ) counted with multiplicity. Let |ak−1 | ≤ |ak | for all k = 1, 2, · · · so that a1 , · · · an(r ) ∈ D(0, r ). Munsik Kim (KAIST) Entire function of finite order 11. 2021 8 / 32 Lemmas Proof of Lemma 1 - continued Then n(r ) X k=1 log 2 ≤ n(r ) X k=1 Z 2π n(2r ) X 2r 2r 1 log ≤ log = log |f (2re iθ )|dθ. ak ak 2π 0 k=1 Using the equation (5), we can conclude that 1 n(r ) · log 2 ≤ 2π Munsik Kim (KAIST) Z 2π log |f (2re iθ )|dθ ≤ log M(2r ) (6) 0 Entire function of finite order 11. 2021 9 / 32 Main Theorem 1 - Proof Proof Since f has finite order λ, M(r ) ≤ exp(r λ+/2 ) for any small > 0 and sufficiently large r > 0.. Using the Lemma 1 n(r ) ≤ Note that 1 1 log M(2r ) ≤ (2r )λ+ 2 log 2 log 2 (7) n(r ) 1 (2r )λ+ 2 ≤ log 2 r λ+ r λ+ (8) converges to 0 as r → ∞ since λ is finite. Munsik Kim (KAIST) Entire function of finite order 11. 2021 10 / 32 Main Theorem 1 - Proof Proof Therefore we can write for sufficiently large r , n(r ) ≤ r λ+ (9) By assumptions, {a1 , a2 , · · · aj } ⊆ D̄(0, |aj |) and we have j ≤ n(|aj | + δ) ≤ (|aj | + δ)λ+ (10) for large j and arbitrarily small δ > 0. Letting δ → 0, we get |aj |−(λ+1) ≤ j −(λ+1)/(λ+) (11) P Then by setting < 1, from the fact that n n1p converges when p > 1, X |an |−(λ+1) converges. (12) n Munsik Kim (KAIST) Entire function of finite order 11. 2021 11 / 32 Main Theorem 1 Using the Weierstrass factorization, we want to factorize the entire function f as follows: Y f (z) = exp(g (z)) · z m E[λ] (z/an ) (13) n with Theorem 8.2.2 and Theorem 1. That is, for an entire function of finite order λ , E[λ] (z/an ) is enough. Let pn = [λ](:= greatest integer such that ≤ λ) for all n. We can easily check that it satisfies the assumption 8.2.2 ; (i) |aj | P in Theorem −p−1 < ∞. is not bounded (ii) ∃m ∈ N such that ∞ |a | n n=m+1 Munsik Kim (KAIST) Entire function of finite order 11. 2021 12 / 32 Theorem 8.2.2 Theorem 8.2.2 Let {aj }∞ j=1 be a sequence of nonzero complex numbers with no accumulation point in C. If {pj } are positive integers that satisfy ∞ X r pn +1 <∞ |an | n=1 for every r > 0, then ∞ Y n=1 Ep n z an converges uniformly on compact subsets of C to an entire function F . Munsik Kim (KAIST) Entire function of finite order 11. 2021 13 / 32 Main Theorem 2 We have examined what λ says about the zeros of f . Now we consider the exponent g (z). Here, we will prove that an entire function f of finite order can be written as products as mentioned before. Theorem 2 If f is an entire function of finite order λ and f (0) = 1, then the Weierstrass canonical product f (z) = e g (z) P(z) (14) has the property that g is a polynomial of degree less than or equal to λ. Before prove the Theorem 2, we need some Lemmas. Munsik Kim (KAIST) Entire function of finite order 11. 2021 14 / 32 Lemmas Lemma 2 Let f be an entire function of finite order λ with f (0) = 1. Let p > λ − 1 and let {aj } be the zeros of f listed with multiplicities. Suppose that |a1 | ≤ |a2 | ≤ · · ·. Then, for any z ∈ C, lim n(r ) X r →+∞ ākp+1 · (r 2 − āk z)−p−1 = 0 (15) k=1 Proof of Lemma 2 Let z be fixed and r > 2|z|. Consider a1 , · · · an(r ) so that they are in D(0, r ). Then for those ak ’s |r 2 − āk z| ≤ r 2 − Munsik Kim (KAIST) r2 r2 = 2 2 Entire function of finite order (16) 11. 2021 15 / 32 Lemmas Proof of Lemma 2 - continued Hence 2 p+1 2 p+1 = (17) r2 r From the definition of the order and Lemma 1, for any > 0 small and r large enough, |āk |p+1 |r 2 − āk z|−p−1 ≤ r p+1 log 2 · n(r )r −p−1 ≤ (log M(2r ))r −p−1 (18) ≤ (2r )λ+ r −p−1 Then n(r ) X k=1 2 p+1 1 λ+p++1 λ+−p−1 ≤ 2 r . (19) ākp+1 (r 2 −āk z)−p−1 ≤ n(r ) r log 2 Munsik Kim (KAIST) Entire function of finite order 11. 2021 16 / 32 Lemmas Proof of Lemma 2 - continued Take = p+1−λ , then the exponent of r is negative so that it goes to zero 2 as r → +∞ Munsik Kim (KAIST) Entire function of finite order 11. 2021 17 / 32 Lemmas Lemma 3 If f is an entire function of finite order λ and f (0) = 1, then, for p > λ − 1, p an integer and fixed z ∈ C, Z 2π 1 2re iθ (re iθ − z)−p−2 log |f (re iθ )|dθ → 0 2π 0 (20) as r → +∞ Munsik Kim (KAIST) Entire function of finite order 11. 2021 18 / 32 Lemmas Proof of Lemma 3 Consider the function φ(w ) = 1 (w − z)p+2 which has residue 0 at z ∈ D(0, r ) and satisfies I φ(w )dw = 0. ∂D(0,r ) In parametrized form, Z 2π φ(re iθ )ire iθ dθ = 0 0 Munsik Kim (KAIST) Entire function of finite order 11. 2021 19 / 32 Lemmas Proof of Lemma 3 - continued Then, we can get 1 2π Z 2π 2re iθ (re iθ − z)−p−2 log |f (re iθ )|dθ 0 Z 2π 1 = 2re iθ (re iθ − z)−p−2 (log |f (re iθ )| − log M(r ))dθ 2π 0 Z 2π r −p−2 1 ≤ 2r · (log M(r ) − log |f (re iθ )|)dθ 2π 0 2 (21) ≤ 2p+3 r −p−1 log M(r ) ≤ 2p+3 r λ+−p−1 using the fact that log M(r ) is constant with respect to θ, Lemma 1, and Jensen’s inequality. Take = (p + 1 − λ)/2. Munsik Kim (KAIST) Entire function of finite order 11. 2021 20 / 32 Lemmas Lemma 4 Let f be a non constant entire function of finite order λ and f (0) = 1. Let {a1 , · · · } be zeros of f listed with multiplicities. Suppose that |a1 | ≤ |a2 | ≤ · · ·. If p > λ − 1 is an integer, the X 1 d p h f 0 (z) i = −p! dz p f (z) (aj − z)p+1 (22) j Munsik Kim (KAIST) Entire function of finite order 11. 2021 21 / 32 Lemmas Proof of Lemma 4 Let r > 2|z|. From Poisson-Jensen formula, we get log |f (z)| = − n(r ) X j=1 r 2 − āj z 1 log + r (z − aj ) 2π Z 2π Re 0 re iθ + z re iθ − z log |f (re iθ )|dθ. (23) By logarithmic differentiation in z n(r ) n(r ) X f 0 (z) X −1 = (z − aj ) + āj (r 2 − āj z)−1 f (z) j=1 j=1 Z 2π 1 2re iθ (re iθ − z)−2 log |f (re iθ )|dθ. + 2π 0 Munsik Kim (KAIST) Entire function of finite order (24) 11. 2021 22 / 32 Lemmas Proof of Lemma 4 Differentiate both sides of the equation p times to get n(r ) n(r ) X X d p h f 0 (z) i −p−1 = − p! (aj − z) + p! ājp+1 (r 2 − āj z)−p−1 ( ) dz f (z) j=1 j=1 (25) Z 2π 1 2re iθ (re iθ − z)−2 log |f (re iθ )|dθ. + (p + 1)! 2π 0 By Lemma 2, the second term goes to zero as r → ∞. By Lemma 3, the integral term goes to zero as r → ∞. d Note that dz log |f (z)| = q ¯ |f (z)| = f (z) · f (z). Munsik Kim (KAIST) f 0 (z) f (z) . We can check by setting f = u + iv and Entire function of finite order 11. 2021 23 / 32 Lemmas Lemma 5 If f is an entire function of finite order λ, f (0) = 1, and if P(z) = ∞ Y E[λ] n=1 z an is the associated product, then for any integer p > λ − 1 we have X d p h P 0 (z) i 1 = −p! dz p P(z) (aj − z)p+1 (26) j Munsik Kim (KAIST) Entire function of finite order 11. 2021 24 / 32 Lemmas Proof of Lemma 5 If PN is the Nth partial product, then it is clear that N X 1 d p h PN0 (z) i = −p! + (error) p dz PN (z) (an − z)p+1 (27) n=1 By Lemma 4, the first term on the RHS converges to equation (26) and the error becomes 0 when p > λ − 1, since it is the (p + 1)st derivative of a polynomial of degree p. Since PN → P normally, then PN0 → P 0 normally so that the LHS converges to the one in the equation (26) Munsik Kim (KAIST) Entire function of finite order 11. 2021 25 / 32 Main Theorem 2 - Proof Proof Differentiate f (z) = e g (z) P(z) so that P 0 (z) f 0 (z) = g 0 (z) + . f (z) P(z) Let p > λ − 1. Differentiating both sides p times and applying Lemmas 4 and Lemma 5 gives −p! X n X 1 1 d p+1 = g (z) − p! p+1 p+1 (an − z) dz (an − z)p+1 n (28) so that d p+1 g (z) = 0 (29) dz p+1 and hence g is a holomorphic polynomial of degree that does not exceed p. Munsik Kim (KAIST) Entire function of finite order 11. 2021 26 / 32 Main Theorem 2 Note that the hypothesis f (0) = 1 in Theorems and Lemmas is in effect superfluous since if f vanishes to order m at 0, then f (z) f˜(z) = m z still have finite order λ and it can be suitably normalized so that f˜(0) = 1. Theorem 1 and 2 are called Hadamard factorication theorem If f is entire of finite order and {aj } the zeros of f , then we define the rank p if f to be the least non negative integer such that X |an |−p−1 < ∞ an 6=0 Using the Theorem 1, such p exists. Munsik Kim (KAIST) Entire function of finite order 11. 2021 27 / 32 Main Theorem 3 From Theorem 2, we can write f (z) = c · z m · e g (z) · P(z) (30) where g is a polynomial. Let q be the degree of g . Then we define the genus µ of f to be the maximum of p and q. Main Theorem 3 An entire function of finite order λ is also of finite genus µ and µ ≤ λ. The proof of Theorem 3 is trivial by definition of the genus. Now we derive a few interesting consequences. Munsik Kim (KAIST) Entire function of finite order 11. 2021 28 / 32 Main Theorem 4 Main Theorem 4 Let f be an entire function of finite order λ ∈ / Z. Then there are infinitely many distinct zj ∈ C such that f (zj ) = c for any c ∈ C. Proof Without loss of generalizy, assume c = 0. Now suppose that the statement is false. Then f −1 ({0}) = {a1 , · · · , aN } for some N ∈ Z and f (z) = e g (z) · (z − a1 ) · · · (z − aN ). From the previous Theorem, g is a polynomial of degree not exceeding λ. However, if > 0 and |z| |aj | for all j, as |z| → ∞, e |z| Munsik Kim (KAIST) λ− |f (z)| λ+ ≤ |e g (z) | = QN ≤ e |z| | j=1 (z − aj )| Entire function of finite order 11. 2021 29 / 32 Main Theorem 4- Proof Proof - continued RHS is from the definition of finite order and LHS is trivial. Therefore the order of e g (z) is λ. However, since the order of e g is the degree q of g and q = λ, contradicting with λ ∈ / Z. Munsik Kim (KAIST) Entire function of finite order 11. 2021 30 / 32 Main Theorem 5 Theorem 5 Let f be a non constant entire function of finite order. Then the image of f contains all complex numbers except possibly one. If, in addition, f has non integer order, then f assumes each of these values infinitely many times. Proof If the image of f omits distinct α1 and α2 , then consider f − α1 . Since this function never vanishes, we can write f (z) − α1 = e g (z) for some entire function g (z) and Hadamard’s theorem guarantees that g is polynomial. It is the same for α2 . Since f (z) − α1 omits the value α2 − α1 , g omit the value log(α2 − α1 ). So if β ∈ C with e β = α1 − α2 , then g (z) − β never vanishes, contradicts the Fundamental theorem of algebra unless g is constant. Munsik Kim (KAIST) Entire function of finite order 11. 2021 31 / 32 Discussion In Ch 9.3, we check that, if f is an entire function of finite order, we can write f as f (z) = c · e g (z) z m P(z) (31) where P(z) ∞ Y n=1 E[λ] z an where ai ’s are the zeros of f and f vanishes at 0 in order m and g (z) is a polynomial of degree less than or equal to λ. For a finite order entire functions, the image of the function contains all complex number except possibly one. Munsik Kim (KAIST) Entire function of finite order 11. 2021 32 / 32