ProofComplete

Proof of Hadamard's Factorization Theorem

Hadamard's factorization theorem refines the Weierstrass factorization for entire functions of finite order, constraining the exponential factor to be a polynomial.


Statement

Theorem9.11Hadamard factorization theorem (restated)

Let ff be entire of finite order ρ\rho with zeros a1,a2,…a_1, a_2, \ldots (none at the origin) and a zero of order mm at the origin. Let p=βŒŠΟβŒ‹p = \lfloor\rho\rfloor. Then

f(z)=zmeQ(z)∏n=1∞Ep(z/an)f(z) = z^m e^{Q(z)}\prod_{n=1}^{\infty}E_p(z/a_n)

where Q(z)Q(z) is a polynomial of degree at most pp, and the product converges absolutely.


Key Ingredients

Theorem9.12Jensen's formula

If ff is holomorphic on ∣zβˆ£β‰€R|z| \leq R, f(0)β‰ 0f(0) \neq 0, and a1,…,aka_1, \ldots, a_k are the zeros of ff in ∣z∣<R|z| < R (counted with multiplicity), then

log⁑∣f(0)∣=βˆ’βˆ‘j=1klog⁑R∣aj∣+12Ο€βˆ«02Ο€log⁑∣f(ReiΞΈ)βˆ£β€‰dΞΈ.\log|f(0)| = -\sum_{j=1}^k \log\frac{R}{|a_j|} + \frac{1}{2\pi}\int_0^{2\pi}\log|f(Re^{i\theta})|\,d\theta.

RemarkJensen's formula and zero counting

Jensen's formula relates the growth of ff (measured by log⁑∣f∣\log|f| on circles) to the number and distribution of zeros. Specifically, the counting function n(r)=#{∣akβˆ£β‰€r}n(r) = \#\{|a_k| \leq r\} satisfies

N(r)=∫0rn(t)t dt≀log⁑M(r)+O(1)N(r) = \int_0^r \frac{n(t)}{t}\,dt \leq \log M(r) + O(1)

where M(r)=max⁑∣z∣=r∣f(z)∣M(r) = \max_{|z|=r}|f(z)|. For ff of order ρ\rho: N(r)=O(rρ+Ξ΅)N(r) = O(r^{\rho+\varepsilon}), which implies βˆ‘βˆ£anβˆ£βˆ’(ρ+Ξ΅)<∞\sum |a_n|^{-(\rho+\varepsilon)} < \infty for every Ξ΅>0\varepsilon > 0.


Proof Outline

Proof

Step 1: The canonical product converges.

Since βˆ‘βˆ£anβˆ£βˆ’(ρ+Ξ΅)<∞\sum |a_n|^{-(\rho+\varepsilon)} < \infty for Ξ΅>0\varepsilon > 0 and p=βŒŠΟβŒ‹β‰₯Οβˆ’1p = \lfloor\rho\rfloor \geq \rho - 1, we have p+1>ρp + 1 > \rho, so βˆ‘βˆ£anβˆ£βˆ’(p+1)<∞\sum |a_n|^{-(p+1)} < \infty. This ensures the canonical product P(z)=∏Ep(z/an)P(z) = \prod E_p(z/a_n) converges absolutely and uniformly on compact sets.

Step 2: P(z)P(z) has the same zeros as ff.

By construction, PP is entire with zeros exactly at {an}\{a_n\} (with correct multiplicities). Therefore h(z)=f(z)/(zmP(z))h(z) = f(z)/(z^m P(z)) is entire and nonvanishing, so h(z)=eg(z)h(z) = e^{g(z)} for some entire function gg.

Step 3: Growth estimate on gg.

The key step is showing gg is a polynomial of degree ≀p\leq p. The growth of P(z)P(z) can be estimated: log⁑∣P(z)∣=O(rρ+Ξ΅)\log|P(z)| = O(r^{\rho+\varepsilon}) using properties of the elementary factors. Therefore

Re(g(z))=log⁑∣h(z)∣=log⁑∣f(z)βˆ£βˆ’mlog⁑∣zβˆ£βˆ’log⁑∣P(z)∣=O(rρ+Ξ΅).\text{Re}(g(z)) = \log|h(z)| = \log|f(z)| - m\log|z| - \log|P(z)| = O(r^{\rho+\varepsilon}).

Step 4: A real part bound implies polynomial.

If g(z)=βˆ‘cnzng(z) = \sum c_n z^n is entire with Re(g(z))≀Crρ+Ξ΅\text{Re}(g(z)) \leq C r^{\rho+\varepsilon}, then by Cauchy's estimates and the Borel-Caratheodory theorem (which bounds ∣g∣|g| in terms of Re(g)\text{Re}(g)):

∣cn∣Rn≀2max⁑∣z∣=RRe(g(z))+2∣g(0)βˆ£β‰€Cβ€²Rρ+Ξ΅|c_n| R^n \leq 2\max_{|z|=R}\text{Re}(g(z)) + 2|g(0)| \leq C'R^{\rho+\varepsilon}

so ∣cnβˆ£β‰€Cβ€²Rρ+Ξ΅βˆ’n|c_n| \leq C'R^{\rho+\varepsilon-n}. For n>ρn > \rho, letting Rβ†’βˆžR \to \infty gives cn=0c_n = 0. Therefore g=Qg = Q is a polynomial of degree at most βŒŠΟβŒ‹=p\lfloor\rho\rfloor = p. β– \blacksquare

β– 

Applications

ExampleEntire functions of order 1

Every entire function of order 11 with zeros {an}\{a_n\} satisfying βˆ‘1/∣an∣1+Ξ΅<∞\sum 1/|a_n|^{1+\varepsilon} < \infty has the form

f(z)=zmeΞ±z+β∏n(1βˆ’zan)ez/an.f(z) = z^m e^{\alpha z + \beta}\prod_n\left(1 - \frac{z}{a_n}\right)e^{z/a_n}.

This applies to sin⁑(Ο€z)\sin(\pi z), 1/Ξ“(z)1/\Gamma(z), and the Riemann xi function ΞΎ(s)\xi(s).

RemarkSignificance

Hadamard's theorem bridges analysis and algebra: it shows entire functions of finite order are essentially determined by their zeros plus a polynomial, much like polynomials themselves. This has profound applications in number theory (the prime number theorem uses the Hadamard factorization of the Riemann zeta function).