7

Let $f: \mathbb{C}\mapsto \mathbb{C}$ be an entire function such that for each $\theta$, $|f(re^{i\theta})|\rightarrow \infty$ as $r\rightarrow \infty$.
a) Does this imply that $|f(z)|\rightarrow \infty$ as $|z|\rightarrow\infty$?
b) Does this imply that $f(z)$ is a polynomial?

I know that if a) is true, then $b)$ is true since a) implies that $\infty$ is a pole for $f(z)$ which is equivalent to $f(z)$ is a polynomial.
However, I don't know how to show that a) is true or construct a counter example.

Edited 5/11:
I am also considering a closely related question: Let $f: \mathbb{R}^2\mapsto \mathbb{R}$ be a continuous function. Suppose for all $k\in \mathbb{R}$, we have $|f(x,kx)|\rightarrow \infty$ as $x\rightarrow \infty$.
a) Does this imply that $|f(x,y)|\rightarrow \infty$ as $|(x,y)|\rightarrow\infty$
b) If we require $f$ to be differentiable, does this imply that $|f(x,y)|\rightarrow \infty$ as $|(x,y)|\rightarrow\infty$

TH Wang
  • 369
  • 2
  • 12

2 Answers2

4

Here is a simpler example of a transcendental entire function $f : \mathbb{C} \rightarrow \mathbb{C}$ such that $f(r e^{i \theta}) \underset{r \rightarrow +\infty}{\longrightarrow} \infty$ for every $\theta \in \mathbb{R}$.

Consider the entire function $f : z \mapsto e^{z} +e^{-z} +z^{2}$.

It is obvious that $f$ is not polynomial - one can easily write its power series expansion or check that none of its derivatives is identically zero.

Furthermore, one can check that for all $\theta \in \mathbb{R}$, $f(r e^{i \theta}) \underset{r \rightarrow +\infty}{\longrightarrow} \infty$.
Indeed, let $\theta \in \mathbb{R}$.
Notice that since for all $z \in \mathbb{C}$, $f(z) = f(-z)$, we can assume that $\theta \in \left( -\frac{\pi}{2}, \frac{\pi}{2} \right]$.

  • If $\theta \in \left( -\frac{\pi}{2}, \frac{\pi}{2} \right)$, then $$\left|f(r e^{i \theta})\right| \geq e^{r \cos(\theta)} - e^{-r \cos(\theta)} -r^{2} \underset{r \rightarrow +\infty}{\longrightarrow} +\infty$$

  • If $\theta = \frac{\pi}{2}$, then $$\left|f(r e^{i \theta})\right| = |f(i r)| = \left|2 \cos(r) -r^{2}\right| \geq r^{2} -2 \underset{r \rightarrow +\infty}{\longrightarrow} +\infty$$

v_lentin
  • 1,233
  • 1
  • 9
  • 16
  • Very nice! I stopped looking at elementary functions after seeing $\cos z + \cosh z$ failed. I should have continued searching for those. – Sungjin Kim May 14 '18 at 04:44
  • Brilliant answer by v_lentin. – Kavi Rama Murthy May 14 '18 at 06:12
  • @v_lentin Could you please tell us if you can give a simple example to show that the hypothesis of Louiville's Theorem cannot be weakened to: ${f(re^{i\theta})}$ bounded for each fixed $\theta$? – Kavi Rama Murthy May 14 '18 at 06:15
  • @KaviRamaMurthy I tried to find one but couldn't. However, I found a related topic on MathOverflow - https://mathoverflow.net/questions/29734/must-the-set-of-lines-through-the-origin-on-which-a-nonconstant-entire-function - in which one talks about an article of D. J. Newman entitled "An Entire Function Bounded in Every Direction". But I can't get access to it... If you can, please tell me if it gives a "simple" counterexample. – v_lentin May 14 '18 at 11:04
  • The function $H$ in my answer is Newman's example. The paper by Armitage mentions this. – Sungjin Kim May 14 '18 at 14:07
  • @i707107 Ah ok. Thank you. – v_lentin May 14 '18 at 14:17
  • A slight modification $H(z+i\pi)$. That one approaches zero on every line through zero except for the real line, and bounded on the real line. – Sungjin Kim May 14 '18 at 14:25
0

You are right about (a) implies (b). If we show that (b) is false, then we can show that (a) is also false.

Indeed, (b) is false.

Consider the same function $H(z)$ in the link: Radial Limits for Holomorphic Functions

$$ H(z)=\int_0^{\infty} t^{-t} e^{tz} dt, $$ $$F(z)=H(z+i\pi)H(iz+i\pi).$$ Then $F(z)\rightarrow 0$ as $z\rightarrow \infty$ along every line through $0$.

Now, let $f(z)=F(z)+1$, then $f(z)\rightarrow 1$ as $z\rightarrow \infty$ along every line through $0$. We consider $$ G(z)=\int_0^z f(w)dw. $$ The function $G$ is well-defined since $f$ is entire and integral is path independent, and then $G$ is also entire with $G'=f$.

Let $z\rightarrow\infty$ along a line through $0$. We take the path of the integral to be also a line through $0$, from $0$ to $z$. Then, we can see that $G(z)\rightarrow\infty$ as $z\rightarrow\infty$ along every line through $0$.

To clarify this, let $L_{\theta}(a,b)$ be the line segment from $a$ to $b$ on the line $L_{\theta}=\{re^{i\theta}|r\geq 0\}$. Then we have $$ f(w)\rightarrow 1 \ \textrm{as} \ \ w\rightarrow\infty \ \textrm{along} \ \ L_{\theta}. $$ $$ \int_{L_{\theta}(0,z)} f(w)dw = z +o(|z|) \ \textrm{as} \ \ z\rightarrow\infty \ \textrm{along} \ \ L_{\theta}. $$ However, $G$ is not a polynomial since $f$ is not a polynomial.

By this argument with repeated constant multiplication and integration, we have further:

Let $P(z)$ be a nonconstant polynomial. Then there is an entire function $G_P(z)$ which is not a polynomial such that $$ \frac{G_P(z)}{P(z)} \rightarrow 1 \ \textrm{as} \ \ z\rightarrow\infty \ \textrm{along each} \ \ L_{\theta}. $$

Sungjin Kim
  • 20,102