3
$\begingroup$

There's the following theorem I'm trying to understand (Lemma 10.8 Rudin, Functional Analysis).

Let $f$ be an entire function on $\mathbb{C}$ such that $f(0) = 1, f'(0) = 0, 0 < \lvert f(\lambda) \rvert \leq e^{\lvert \lambda \rvert}$ for all $\lambda \in \mathbb{C}$. Then $f(\lambda) = 1$ for all $\lambda \in \mathbb{C}$.

The maximum modulus theorem is used (Theorem 10.24 Rudin, Real and Complex analysis).

Suppose $\Omega$ is a region, $f \in H(\Omega)$, and $\overline{D}(a;r) \subset \Omega$. Then $$ \lvert f(a) \rvert \leq \max_{\theta} \lvert f(a + re^{i\theta}) \rvert $$

For the lemma 10.8 the proof goes as follows.

Because $f$ doesn't have zeroes then there's a function $g$ such that $f = \exp\left\{ g \right\}$, $g$ entire in $\mathbb{C}$ and $Re[g(\lambda)] \leq |\lambda|$ with $\lambda \in \mathbb{C}$, $g(0) = g'(0) = 0$. Let $r \geq |\lambda|$, we can verify that

$$ \lvert g(\lambda) \rvert \leq \lvert 2r - g(\lambda) \rvert $$

$$ h_{r}(\lambda) = \frac{r^2g(\lambda)}{\lambda^2(2r - g(\lambda))} $$

such a function is holomorphic in $\left\{\lambda : |\lambda| < 2r \right\}$. Clearly $\lvert h_r(\lambda) \rvert \leq 1$ if $\vert \lambda \rvert = r$. By the maximum modulus theorem we have for $\lvert \lambda \rvert \leq r$ $$ \lvert h_r(\lambda) \rvert \leq 1. $$

If we fix $\lambda$ and let $r \to \infty$, we have $g(\lambda) = 0$.

My main question is how the maximum modulus theorem is actually used in this context?

My function $h_r$ is holomorphic in $\Omega = D(0;2r)$ and also I have $\overline{D}(0,r) \subset D(0;2r)$, this should imply

$$ \lvert h_r(0) \rvert \leq \max_{\lambda \in \partial \overline{D}(0,r)}\lvert h_r(\lambda) \rvert \leq 1 $$

where with $\partial A$ I denote the boundary set of a set $A$. Why does this imply $\lvert h_r(\lambda) \rvert \leq 1$ if $| \lambda | \leq r$?. Also why if fix $\lambda$ and let $r$ diverging implies $g(\lambda) = 0$?

$\endgroup$
  • $\begingroup$ Suppose $|h_r(\lambda)|=M>1$ for some $|\lambda|<r$, and suppose this is a maximum value for $h_r$ [which exists because the closed ball is compact]. Then in a small enough neighborhood of radius $\epsilon$ of $\lambda$ [so that the neighborhood is contained in the closed ball of radius $r$], the maximum modulus principle implies there's another point $\lambda_\epsilon$ (not equal to $\lambda$) where $|h_r(\lambda_\epsilon)|=M$. Thus there are points arbitrarily close to $\lambda$ where $|h_r|=M$, and thus $|h_r|$ is constant at value $M$ in the interior of the ball. See the contradiction? $\endgroup$ – Steve D Aug 11 '17 at 18:53
1
$\begingroup$

There is an alternative formulation of the maximum modulus theorem, which apears to be more general: https://en.wikipedia.org/wiki/Maximum_modulus_principle.

Roughly, it says that if $|f|$ has a local maximum in the interior of its domain, then it is constant.

This can be applied for your case as follows: Let $\lambda_0 \in \overline D(0,r)$ be such that it maximizes $|h_r(\lambda_0)|$. If $|\lambda_0|=r$, then $|h_r(\lambda)|\leq 1$ follows. In the other case $|\lambda_0|<r$ we are in the interior of the domain, so $h_r$ is constant.

Edit: i forgot to adress "$r\to \infty \Rightarrow g(\lambda)=0$".

From $|h_r(\lambda)|\leq 1$ it folows that $$ |g(\lambda)| \leq \frac{|\lambda^2(2r-g(\lambda))|}{r^2} $$ Taking the limit $r\to\infty$ yields $g(\lambda)=0$.

$\endgroup$
  • $\begingroup$ Is this alternative formulation equivalent to the one I've mentioned? $\endgroup$ – user8469759 Aug 11 '17 at 13:06
  • $\begingroup$ I dont know if there is a simple equivalence. But i think the author was refering to a statement similar to the wikipedia version. $\endgroup$ – supinf Aug 11 '17 at 13:16
  • 1
    $\begingroup$ I believe it was referring to the one I wrote actually, given the author is the same. But I'll read through to check whether or not such formulation is proven to be equivalent. $\endgroup$ – user8469759 Aug 11 '17 at 13:19
  • $\begingroup$ @supinv Sorry, I'm still reading reading your argument. Basically you proved that for each $r$ I have $h_r$ is a constant, right? But how do I infer that $g(\lambda) = 0$ when $r$ diverges? $\endgroup$ – user8469759 Aug 11 '17 at 15:13
  • $\begingroup$ @user8469759 i edited to adress that part of your question $\endgroup$ – supinf Aug 12 '17 at 11:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.