2

Here is the question I was working on: Let $f$ be holomorphic in the open disk $\mathbb{D}$ and suppose $|f(z)|\leq 1$ for all $z\in \mathbb{D}$. If $f(\frac{1}{2})=f(-\frac{1}{2})=0$, prove that $|f(z)|\leq \left|\frac{4z^2-1}{4-z^2}\right|$ for all $z\in \mathbb{D}$.

I defined the maps $g(z)=\frac{z-\frac{1}{2}}{1-\frac{z}{2}}$ and $h(z)=\frac{z+\frac{1}{2}}{1+\frac{z}{2}}$. Both compositions $f\circ g$ and $f\circ h$ send $0$ to $0$, so using Schwarz lemma, I concluded that $|f(z)|\leq \left|\frac{z-\frac{1}{2}}{1-\frac{z}{2}}\right|$ and $|f(z)|\leq \left|\frac{z+\frac{1}{2}}{1+\frac{z}{2}}\right|$ separately. However, the statement given in the problem is stronger, and I couldn't get it. I would appreciate any kind of help. Thanks!

Winther
  • 24,478
vgmath
  • 1,345

2 Answers2

3

It is too weak to treat one zero at a time. To get the strongest result you need to use both zeros simultaneously.

Since $f(z)$ has zeros at $1/2$ and $-1/2$ then the function (with your notations for $g$ and $h$) $$ \phi(z)=\frac{f(z)}{g(z)h(z)} $$ is holomorphic in $\mathbb{D}$. Moreover, for any $z\in\mathbb{D}$ and $|z|<r<1$ we have by the Maximum Modulus Principle that $$ |\phi(z)|\le\max_{|\zeta|=1}|\phi(r\zeta)|=\max_{|\zeta|=1}\frac{|f(r\zeta)|}{|g(r\zeta)||h(r\zeta)|}\le \max_{|\zeta|=1}\frac{1}{|g(r\zeta)||h(r\zeta)|}\to 1\text{ as }r\to 1^-. $$ The last limit is due to the fact that $|g(z)|$ and $|h(z)|$ converge to $1$ uniformly as $|z|\to 1$. Thus, $|\phi(z)|\le 1$ in $\mathbb{D}$, i.e. $$ |f(z)|\le|g(z)h(z)|. $$ This gives you the necessary estimate after a slight rewriting of the RHS.

A.Γ.
  • 29,518
  • 1
    How do you know that $|f(z)|\leq 1 $ on the unit circle ? we only know that for $\mathbb{D}={z: |z|<1}$ – the8thone Aug 13 '15 at 22:03
  • Thanks for the answer, but I have a question: $f$ might not be defined on the unit circle, so how does the inequality $\frac{|f(z)|}{1}\leq 1$ make sense on the unit circle? – vgmath Aug 13 '15 at 22:04
  • The radial limits exist almost everywhere on $\partial\mathbb{D}$ for $f$ (Fatou's theorem), so I did not think it is a problem. Ok, I can clarify this part. – A.Γ. Aug 13 '15 at 22:18
  • You can also look at the circle with radius $1-\epsilon$ centered at $0$, and then take $\epsilon\to0^+$. – Batominovski Aug 13 '15 at 22:38
  • @Batominovski Yes, it is basically the same proof as in Schwarz lemma. Fatou's theorem seems to be too powerful to use here. – A.Γ. Aug 13 '15 at 22:55
1

You can also solve this problem step-by-step, but I think A.G.'s solution is a more elegant one. It can be generalized in the same way as my solution.

By Schwarz's Lemma, $\big|f(z)\big|\leq \left|\frac{z-1/2}{1-z/2}\right|$ for all $z\in\mathbb{D}$. Define $F(z)$ to be $\left(\frac{1-z/2}{z-1/2}\right)\,f(z)$ for all $z\in \mathbb{D} \setminus\left\{\frac12\right\}$, while $F\left(\frac12\right)$ is given by $\frac{3}{4}\,f'\left(\frac12\right)$. Hence, $F:\mathbb{D}\to\mathbb{D}$ is holomorphic with $F\left(-\frac12\right)=0$, and satisfies the condition $\big|F(z)\big|\leq 1$ for all $z\in\mathbb{D}$.

By Schwarz's Lemma, $\left|F(z)\right|\leq \left|\frac{z+1/2}{1+z/2}\right|$ for all $z\in\mathbb{D}$. Thus, $$\left|f(z)\right|=\left|\frac{z-1/2}{1-z/2}\right|\,\big|F(z)\big|\leq \left|\frac{z-1/2}{1-z/2}\right|\, \left|\frac{z+1/2}{1+z/2}\right|=\left|\frac{4z^2-1}{4-z^2}\right|$$ for all $z\in\mathbb{D}$.

In general, if $f:\mathbb{D}\to\mathbb{D}$ is holomorphic such that $\big|f(z)\big|\leq 1$ for all $z\in\mathbb{D}$ and $f\left(w_i\right)=0$ for $i=1,2,\ldots,n$, where $w_1,w_2,\ldots,w_n$ are pairwise distinct points in $\mathbb{D}$, then $$\big|f(z)\big|\leq \prod_{i=1}^n\,\left|\frac{z-w_i}{1-\bar{w_i}z}\right|$$ for all $z\in\mathbb{D}$.

P.S. I just found out that the generalized version has been posted here before. See Showing that $|f(z)| \leq \prod \limits_{k=1}^n \left|\frac{z-z_k}{1-\overline{z_k}z} \right|$.

Batominovski
  • 49,629
  • Thanks a lot for the nice answer! I have a question though: How did you justify that $|\frac{3}{4} f'(\frac{1}{2})|\leq 1$, which is necessary for function to satisfy Schwarz lemma. – vgmath Aug 14 '15 at 00:46
  • OK, I think I figured out that, it again follows from Schwarz inequality. Bunch of thanks again! – vgmath Aug 14 '15 at 01:01