일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | ||||
4 | 5 | 6 | 7 | 8 | 9 | 10 |
11 | 12 | 13 | 14 | 15 | 16 | 17 |
18 | 19 | 20 | 21 | 22 | 23 | 24 |
25 | 26 | 27 | 28 | 29 | 30 | 31 |
- 어학병
- infiniteproduct
- Galois
- 유수정리
- 역명논란
- STEIN
- isomorphism
- 분해정리
- 슈타인
- 방배역
- 낙성대역
- 지하철
- Algebra
- 갈루아
- 2호선
- 서울대입구역
- 전해석함수
- 아다마르
- jensenformula
- hadamard
- conjugate
- 4호선
- 총신대입구역
- 오블완
- 바이어슈트라스
- 대수학
- 복소해석학
- 갈루아이론
- 티스토리챌린지
- 특이점
- Today
- Total
Orthy
[복소해석학] 5.5. Hadamard's factorization theorem 본문
Theorem 1. (Hadamard) Suppose $f$ is entire and has growth order $\rho_0$. Let $k$ be the integer so that $k\le\rho_0<k+1$. If $a_1, a_2,\cdots$ denote the non-zero zeros of $f$, then
$$f(z)=e^{P(z)}z^m\prod_{n=1}^\infty E_k(z/a_n)$$
where $P$ is a polynomial of degree$\le k$, and $m$ is the order of the zero of $f$ at the origin.
Theorem 1은 [1]에 소개된 Hadamard's factorization theorem이다. 앞으로 Hadamard's factorization theorem을 HFT로 줄여 말하자. [3]에서는 HFT를 더 상세하게 설명하고 증명의 흐름 또한 개선되어 있어, [3]에 소개된대로 HFT를 다시 쓰고 정리를 증명한다. 읽어보면 알겠지만, Theorem 1 과 Theorem 2는 결국 같은 정리다.
Theorem 2. (Hadamard) Suppose $f$ is entire and has growth order $\rho_0$. Let $k$ be the integer so that $k\le\rho_0<k+1$ then we have
(1) If $f$ has finite zeros then there exist polynomial $f_0$ and $P$ such that $f=f_0e^P$. Also, $\operatorname{deg}P=k=\rho_0$.
(2) If $f$ has infinite zeros, the infinite product
$$\prod_{|a_n|>1}E_k(z/a_n)=\Pi(z)$$
uniformly converges on the every compact subset of $\mathbb{C}$ and there exist polynomial $f_0$ and $P$ such that $f=f_0e^P\Pi$. Also, $\operatorname{deg}P\le k$. Here, the sequence $\{a_n\}$ is a list of the zeros of $f$ with multiplicities, just as Weierstrass theorem.
Step 1. 먼저 증명에 있어 가장 핵심이 되는 다음 보조정리를 증명한다. [1]에서는 Lemma 5.5로 소개되어 있다.
Lemma 1. Suppose $g$ is entire and $s, C>0$. Assume $E\subset(0,\infty)$ is not bounded above. If we have
$$|z|\in E\;\;\Longrightarrow\;\; \big|e^{g(z)}\big|\le e^{C|z|^s}$$
then $g$ is a polynomial of $\operatorname{deg}g\le s$.
proof. Note that $\big|e^{g(z)}|=e^{\operatorname{Re}g(z)}$ then we assume
\begin{align*}
|z|\in E\;\;&\Longrightarrow\;\; e^{\operatorname{Re}g(z)}\le e^{C|z|^s}\\
&\Longrightarrow\;\; \operatorname{Re}g(z)\le C|z|^s
\end{align*}
Since $g$ is entire, we have a power series expansion of $g$, such as $g(z)=\sum_{n=0}^\infty c_nz^n$. Then our claim is $c_n=0$ whenever $n>s$. Let us denote $u=\operatorname{Re} g$ then
$$u(re^{i\theta})=\sum_{n=0}^\infty c_n r^ne^{in\theta}=\left(\;\overline{\sum_{n=0}^\infty c_n r^ne^{in\theta}}\;\right)=\sum_{n=0}^\infty\overline{c_n}r^ne^{-in\theta}$$
for $r>0$ and $\theta\in\mathbb{R}$ since $u$ is real. Then
$$2u(re^{i\theta})=\sum_{n=0}^\infty c_n r^ne^{in\theta}+\sum_{n=0}^\infty\overline{c_n}r^ne^{-in\theta}$$
From Theorem 7.1 of Chapter 3, [1] we have
$$\int_0^{2\pi}g(re^{i\theta})e^{-in\theta}d\theta=\begin{cases}
0\;\;\;\;\;\;\;\;\;\;\;\;\text{if}\;\; n<0\\
2\pi c_n r^n\;\;\;\text{if}\;\; n\ge 0
\end{cases}$$
and with the simple calculation, we also have $\int_0^{2\pi}\overline{g(re^{i\theta})}e^{-in\theta}d\theta=0$ when $n>0$. Then
$$\frac{1}{2\pi r^n}\int_0^{2\pi}g(re^{i\theta})e^{-in\theta}d\theta+\frac{1}{2\pi r^n}\int_0^{2\pi}\overline{g(re^{i\theta})}e^{-in\theta}d\theta=\frac{1}{\pi r^n}\int_0^{2\pi}u(re^{i\theta})e^{-in\theta}d\theta=c_n$$
for $n>0$ and
$$c_0=\frac{1}{2\pi}\int_0^{2\pi}g(re^{i\theta})d\theta=\frac{1}{2\pi}\int_0^{2\pi}g(re^{i\theta})d\theta+\frac{1}{2\pi}\int_0^{2\pi}\overline{g(re^{i\theta})}d\theta=\frac{1}{2\pi}\int_0^{2\pi}u(re^{i\theta})d\theta$$
Since $\int_0^{2\pi}Cr^s e^{-in\theta}d\theta=0$, we have
$$c_n=\frac{1}{\pi r^n}\int_0^{2\pi}(u(re^{i\theta})-Cr^s)e^{-in\theta}d\theta$$
For $r\in E$, by assumption we have $\operatorname{Re}g(z)\le C|z|^s\;\Rightarrow\; u(re^{i\theta})\le Cr^s$ so
$$|c_n|\le\frac{1}{\pi r^n}\int_0^{2\pi}\big|u(re^{i\theta})-Cr^s\big|d\theta=\frac{1}{\pi r^n}\int_0^{2\pi} (Cr^s-u(re^{i\theta}))d\theta$$
But since $\int_0^{2\pi}u(re^{i\theta})d\theta=2\pi c_0$ so
$$|c_n|\le\frac{1}{\pi r^n}\left(2\pi Cr^s-2\pi c_0\right)=2Cr^{s-n}-2c_0r^{-n}$$
Since $E$ is not bounded above, $c_n=0$ whenever $n>s$, proving the lemma.
이제 Theorem 2의 (1)을 증명할 수 있다. $f$의 영점을 중복도만큼 나열하면 유한개이고, 그것을 $a_1,\cdots,a_N$이라고 하자. 이에 대해, $f_0=(z-a_1)\cdots(z-a_N)$이라고 하자. $f$가 영점을 가지지 않으면 $f_0=1$로 두자. 이제 $f/f_0$를 생각하면 $f$의 각 영점에서 $f/f_0$가 제거가능 특이점을 가지므로 $f/f_0$는 전해석함수이고, 따라서 $f/f_0=e^g$인 전해석함수 $g$가 존재한다. 이 글 3절의 \textbf{Lemma 3.3}에 의해 $f=f_0e^g$일 때 $f$와 $e^g$의 증가지수가 같다. 다음으로 충분히 큰 $z$에 대해
$$\big|e^{g(z)}\big|\le|f(z)|=\big|f_0(z)e^{g(z)}\big|$$
가 성립하므로
$$E=\{|z|\in\mathbb{R}\;:\;\big|e^{g(z)}\big|\le|f(z)|\}$$
라고 하면 $E$는 위로 유계가 아니다. $f$의 증가지수가 $\rho_0$이므로 $\rho_0<s<k+1$인 $s$가 존재하여
$$|f(z)|\le e^{C|z|^s}\;\Rightarrow \; \big|e^{g(z)}\big|\le e^{C|z|^s}$$
를 만족시키고 따라서 \textbf{Lemma 6.3}에 의해 $g$는 차수가 $s$이하인 다항함수이며 따라서 $f$의 증가지수, $e^g$의 증가지수 그리고 $g$의 차수가 모두 $\rho_0$임을 알 수 있다.
Step 2. 이제 (2)를 살펴보자. $f$의 영점을 중복도만큼 나열한 수열을 $a_1, a_2,\cdots$라고 하자. 무한곱
$$\prod_{|a_n|>1}E_k(z/a_n)=\Pi(z)$$
을 살펴보자. Lemma 4.2, Chapter 5, [1]에 의해 $|z/a_n|\le 1/2$이면 $|1-E_k(z)|\le c|z|^{k+1}$이 어떤 양수 $c$에 대해 성립한다. 그런데 Theorem 2.1, Chapter 5, [1]에 의해
$$\sum_{|a_n|>1}|a_n|^{-(k+1)}<\infty$$
이므로 $n\to\infty$이면 $|a_n|\to\infty$이고 따라서 $\mathbb{C}$의 어느 컴팩트 집합 $K$를 골라도 $\forall z\in K,\;|z/a_m|\le 1/2$이도록 하는 $m\in\mathbb{N}$이 존재한다. 그러면 Theorme 3.2, Chapter 5, [1]에 의해 $\Pi(z)$는 $\mathbb{C}$의 임의의 컴팩트 집합 $L$에서 고르게 수렴하고 따라서 $\Pi$는 전해석함수이다. 이제
$$f_0(z)=\prod_{|a_n|\le 1}(z-a_n)$$
으로 두면 $f$는 다항함수이고 그러면 Step 1에서와 같은 논리로 $f=f_0e^g\Pi$인 전해석함수 $g$가 존재함을 알 수 있다. 역시 비슷하게 충분히 큰 모든 $z$에 대해
\begin{align*}
\Big|e^{g(z)}\Pi(z)\Big|\le e^{C|z|^s}
\end{align*}
가 성립한다. 그러면 증명의 편의를 위해 위의 식이 모든 $z\in\mathbb{C}$에서 성립하고 $|a_1|, |a_2|,\cdots>1$이라 해도 문제없다. 따라서 $\Pi(z)=\prod_{n=1}^\infty E_k(z/a_n)$으로 둔다. 그런데 Theorem 2의 (1)에서와는 달리 (2)에서는 $\Pi(z)$ 때문에 Lemma 1을 바로 적용할 수 없다. 이 문제를 해결하기 위해 다음 몇 가지 보조정리를 추가로 도입한다.
Lemma 2. There exist positive real number $c_1, c_2$ such that
(1) $|z|\le 1/2\;\;\;\Longrightarrow\;\;\; |E_k(z)|\ge e^{-c_1|z|^{k+1}}$
(2) $|z|\ge 1/2\;\;\;\Longrightarrow\;\;\; |E_k(z)|\ge |1-z|e^{-c_2|z|^k}$
proof. Suppose $|z|\le 1/2$. Then
$$\operatorname{Log}(1-z)=-\sum_{n=1}^\infty\frac{z^n}{n}$$
is well defined so
$$E_k(z)=(1-z)\exp(z+\cdots+z^k/k)=\exp(\operatorname{Log}(1-z)+z+\cdots+z^k/k)=e^w$$
for $w=\sum_{n=k+1}^\infty z^n/n$. Then we have
$$\big|E_k(z)\big|=|e^w|\ge e^{-|w|}$$
and since
$$|w|\le\sum_{n=k+1}^\infty\Bigg|\frac{z^n}{n}\Bigg|\le|z|^{k+1}\sum_{n=k+1}^\infty\frac{1}{n\cdot 2^{n-k-1}} \le 2|z|^{k+1}$$
we have
$$\big|E_k(z)\big|\ge e^{-2|z|^{k+1}}$$
Now suppose $|z|\ge 1/2$ then since
$$E_k(z)=(1-z)\exp\left(\sum_{n=1}^k\frac{z^n}{n}\right)$$
we have $|e^h|=\operatorname{Re} h$ so from
$$\operatorname{Re}\left(\sum_{n=1}^k\frac{z^n}{n}\right)\ge -\Bigg|\sum_{n=1}^k\frac{z^n}{n}\Bigg|\ge -|z|^k\sum_{n=1}^k\frac{1}{n\cdot 2^{n-k}}$$
then
$$\big|E_k(z)\big|\ge |1-z|e^{-c_2|z|^k}$$
This ends the proof.
Lemma 3. There exists positive $c$ such that
$$z\notin\bigcup_{n=1}^\infty D(a_n, |a_n|^{-k-1})\;\;\;\Longrightarrow\;\;\; \big|\Pi(z)\big|\ge e^{-c|z|^s}$$
where $D(a_n, |a_n|^{-k-1})$ is a disc of radius $|a_n|^{-k-1}$, centered at $a_n$.
proof. Here we present only main ideas. We first write the product as
$$\prod_{n=1}^\infty E_k(z/a_n)=\left(\prod_{|a_n|\le 2|z|}E_k(z/a_n)\right)\cdot\left(\prod_{|a_n|> 2|z|}E_k(z/a_n)\right)$$
We then show that each product is estimated by
$$\Bigg|\prod_{|a_n|\le 2|z|}E_k(z/a_n)\Bigg|\ge e^{-c|z|^s}\;\;\;\text{and}\;\;\;\Bigg|\prod_{|a_n|> 2|z|}E_k(z/a_n)\Bigg|\ge e^{-c'|z|^s}$$
for some positive constant $c, c'$ by the previous lemma, Lemma 2. The restriction
$$z\notin\bigcup_{n=1}^\infty D(a_n, |a_n|^{-k-1})$$
is needed to estimate $\prod_{|a_n|\le 2|z|}|1-z/a_n|$ which naturally appears when estimating the first product. We need Theorem 2.1, Chapter 5, [1] in the estimation of this product. For the detailed explanation, see p.148, 149 of [1].
Lemma 4. Let $E$ be
$$E=(0,\infty)\setminus\bigcup_{n=1}^\infty I_n$$
where $I_n=(|a_n|-|a_n|^{-k-1},|a_n|+|a_n|^{-k-1})$. Then $E$ is not bounded above.
proof. Since $\sum|a_n|^{-k-1}$ converges, we have $N\in\mathbb{N}$ such that $\sum_{n=N}^\infty |a_n|^{-k-1}<1/2$. Given any two consecutive large integer $M, M+1$ we have $r>0$ such that $M\le r\le M+1$ and the circle $C(0, r)$ of radius $r$, centered at the origin does not intersect the forbidden disc $D(a_n, |a_n|^{-k-1})$ for any $n$. Otherwise the union of the intervals $I_n$ would cover $[M, M+1]$ and this implies
$$2\sum_{n=N}^\infty |a_n|^{-k-1}>1$$
which leads to the contradiction. This ends the proof.
다시 (2)의 증명으로 돌아가자. Lemma 4의 $E$에 대해
$$z\in E\;\;\;\Longrightarrow\;\;\; \big|\Pi(z)\big|\ge e^{-c|z|^s}$$
이 성립하므로 $z\in E$이면
\begin{align*}
\Big|e^{g(z)}\Pi(z)\Big|\le e^{C|z|^s}\;\;\;&\Longrightarrow\;\;\;\Big|e^{g(z)}\Big|e^{-c|z|^s}\le\Big|e^{g(z)}\Pi(z)\Big|\le e^{C|z|^s}\\
&\Longrightarrow\;\;\; \Big|e^{g(z)}\Big|\le e^{(c+C)|z|^s}
\end{align*}
이므로 $g$는 차수가 $s$이하인 다항식이고 따라서 $\deg g\le k$이다. 따라서 모든 증명이 끝났다.
[1], Stein, Elias M., Shakarachi, Rami 2003. Complex Analysis. Princeton University Press.
[2], 김영원, 계승혁. 2014. 기초복소해석. 서울대학교출판문화원.
[3], 김영원, 2023. 복소해석학 강의노트. 한빛아카데미.
[4], Rudin, Walter 1976. Principles of Mathematical Analysis 3ed. McGraw-Hill Press.
[5], Munkres, J.R., 2000. Topology 2ed. Pearson Education.
아다마르 분해정리는 성장 속도가 유한한 전해석함수는 영점에 의해 완전하게 결정될 수 있다는 정리이다. 5장에서 공부한 젠센 공식, 함수의 성장 속도, 바이어슈트라스 정리 그리고 아다마르 분해정리를 통해 전해석함수의 영점의 위치와 분포가 함수의 구성 그리고 고유성을 결정하는데 중요한 역할을 한다는 것을 알 수 있다.
지금까지 5장에서는 영점이라는 국소적(local) 정보가 함수 전체의 대역적(global) 성질을 어떻게 결정하는지 공부했다. 다음 8장에서는 한 열린 집합에서 다른 열린 집합으로의 전단사 정칙함수에 대해 공부하게 된다.
'학부 수학 > 복소해석학' 카테고리의 다른 글
[Complex Analysis], Stein, Chapter5 (0) | 2025.05.17 |
---|---|
[복소해석학] 5.4. Weierstrass infinite product (0) | 2025.05.16 |
[복소해석학] 3.1. Zeros and poles (0) | 2025.05.11 |
[복소해석학] 3.0. Preface (0) | 2025.05.11 |
[복소해석학] 5.3. Infinite product (0) | 2025.05.11 |