5
$\begingroup$

If $f$ is entire and we assume a decay rate on $\mathbb{R}^+$, does this dictate a growth rate on $\mathbb{C}$?

For example, is there a function that decays as $e^{-x^2}$ on $\mathbb{R}^+$ and is of exponential type?

What I am actually interested in is whether there exists an entire function such that $|f(z)|\le Me^{-a z}$ for all $z>0$ and $ |f(z)| \le Me^{b |z|} $ for all $z$, with $b<a$.

$\endgroup$
1
  • $\begingroup$ It should also follow from Phragmen-Lindelof $\endgroup$ Dec 7, 2017 at 2:09

3 Answers 3

7
$\begingroup$

This is a typical walk-to-the-library problem. I used Boas, but probably other standard books would have worked too.

Boas proves the following results: (1) if $f$ is of order $1$, then $\limsup m(r)M(r)^{1+\epsilon}=\infty$ (Theorem 3.3.1); (2) if $f$ is of order $0<\rho<1$, then we obtain $m(r_n)\gtrsim M(r_n)^{-a}$ along a suitable sequence $r_n\to\infty$ from Theorem 3.2.11; (3) finally, if $f$ is of order $0$, then $\log m(r)\simeq \log M(r)$ on a suitable sequence again (Theorem 3.6.2).

So there is no such function.

(I've used the usual notations $m(r), M(r)$ for the min and max, respectively, of $|f|$ on $|z|=r$.)

$\endgroup$
5
  • $\begingroup$ Boas wrote several books. Which book are you quoting from? $\endgroup$
    – GH from MO
    Dec 7, 2017 at 2:26
  • 3
    $\begingroup$ @GHfromMO: Entire functions. $\endgroup$ Dec 7, 2017 at 2:31
  • $\begingroup$ Thank you for your valuable answer. For (2) we need some control on $a$, e.g. $a=1$ is admissible by Theorem 3.2.11. $\endgroup$
    – GH from MO
    Dec 7, 2017 at 2:48
  • 3
    $\begingroup$ @GHfromMO: It doesn't really matter, but I think any $a$ works fine, since we're assuming order $<1$ in (2), so $M(r)^{-a} \gtrsim e^{-a|z|^{1-\epsilon}}$. $\endgroup$ Dec 7, 2017 at 2:56
  • $\begingroup$ You are right, and I was wrong. $\endgroup$
    – GH from MO
    Dec 7, 2017 at 3:10
2
$\begingroup$

The answer to your question is no. More generally, for functions of exponential type $b$ (which means $|f(z)|\leq Me^{b|z|}$), define the indicator: $$h(\theta)=\limsup_{r\to\infty}\frac{\log|f(re^{i\theta})|}{r}.$$ Then there is a description of all possible indicators: they are trigonometrically convex: $$h''+h\geq 0,$$ where derivatives should be understood in the generalized sense where they do not exist. Equivalently and more intuitively: possible indicators are exactly the support functions of compact convex sets. From which follows that $\min h\geq -\max h$.

There are many deep generalizations of these things to functions which are not necessarily of exponential type, and for decrease not necessarily on a straight line. For example, there is a theorem of Hayman and Kjellberg that for every curve $\gamma$ tending to infinity, $$\limsup_{z\to\infty,\;z\in\gamma}\frac{\log|f(z)|}{M(|z|)}\geq -1.$$ This holds for every non-constant entire function (not necessarily of exponential type).

$\endgroup$
1
$\begingroup$

I don't have the answer to your question but on top of my head : consider an entire function $f$ that goes to zero on $\mathbb R^+$, the function $z\mapsto f(1/z)$ is holomorphic on $\mathbb C^*$. Since $f$ is going to zero via real values, we know this function is not meromorphic with a pole . Furthermore, it cannot be bounded otherwise it would be holomorphic on the riemann sphere thus constant. Finally, $f$ has an essential singularity at infinity. However this does not answer your question.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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