4
$\begingroup$

Question. Is there an entire function $F$ satisfying first two or all three of the following assertions:

  • $F(z)\neq 0$ for all $z\in \mathbb{C}$;
  • $1/F - 1\in H^2(\mathbb{C}_+)$ -- the classical Hardy space in the upper half-plane;
  • $F$ is bounded in every horizontal half-plane $\{z\colon \text{Im}(z) > \delta\}$?

Thoughts. Let $G= 1/F$. Then we have $G(z) = 1 + \int_0^{\infty} h(x)e^{izx}\, dx$ for some $h\in L^2[0,\infty)$ and all $z\in \mathbb{C}_+$. For nice functions $h$ (e.g., for super-exponentially decreasing) this integral representation can be extended to the whole complex plane and probably the example can be constructed in terms of $h$. However, I don't know if it is possible to find $h$ such that $G$ is non-zero for every $z$.

$\endgroup$
2
  • 2
    $\begingroup$ Can you explain, what is a "horizontal half-plane" ? Are half-planes $\{ z=x+iy: y\geq 0\}$ and $\{ x+iy: y\leq 0\}$ both horizontal? $\endgroup$ Feb 10 at 14:40
  • $\begingroup$ @AlexandreEremenko, thank you for your correction, I meant to consider only the upper half-plane. It is fixed now $\endgroup$ Feb 10 at 15:28

2 Answers 2

5
$\begingroup$

There is a zero-free entire function bounded in every left half-plane, and such that $f-1$ is in $H^2$ in every left half-plane.

Let $\gamma$ be the boundary of the region $$D=\left\{ x+iy: |y|<2\pi/3, x>0\right\} .$$ Consider the function $$g(z)=\int_\gamma \frac{\exp e^\zeta}{\zeta-z}d\zeta,\quad z\in {\mathbf{C}}\backslash D.$$ The integral evidently converges and $g(z)=O(1/z)$ in ${\mathbf{C}}\backslash D.$ Now, $g$ has an analytic continuation to an entire function: deforming the contour to $\partial D_t$, where $D_t=\{ x+iy:|y|<2\pi/3, x>t\},\; t>0$ does not change $g$ in $D$, and shows that $g$ has an analytic continuation to ${\mathbf{C}}\backslash D_t$, and this is for every $t>0$, so $g$ is entire. Now $f(z)=e^{g(z)}$ is the desired function. If you want upper half-planes take $f(iz)$.

Remark. You can improve the estimate $g(z)=O(1/z)$. Evidently, $g$ has infinitely many zeros $z_1,z_2,\ldots$. Then $g_k(z)=g(z)/((z-z_1)\ldots(z-z_k))$ satisfies $g_k(z)=O(z^{-k-1})$ as $z\to\infty$ outside $D_t$.

Remark 2. This construction is standard in the theory of entire functions, see, for example, Entire function bounded at every line Sometimes this $g$ is called the Mittag-Leffler function.

$\endgroup$
1
  • $\begingroup$ Thanks, this is a very useful example! Will follow the links $\endgroup$ Feb 10 at 16:19
0
$\begingroup$

Here is a slightly different construction, from this answer by @reuns , the function built this way has a $\exp(\exp(|z|))$ growth compared to triple exponent in the wonderful example by @AlexandreEremenko.

For $z\in\mathbb{C}$, consider $$ g(z) = \int_{i\infty}^z\frac{e^{is}(e^{is} - 1)}{s}ds = - \frac{e^{iz}(e^{iz} - 2)}{2iz} + \int_{i\infty}^z\frac{e^{is}(e^{is} - 2)}{2is^2}ds, $$ where the equality is from integration by parts. For $z\neq 0$, let the contour in the last integral consist of two components: the vertical half-line and the arc of the circle with center in $0$ (as in the picture).contour

After integration by parts the integral over the half-line is exponentially small and the integral over the arc is $O(1/|z|)$ in every upper half-plane.

It follows that $|g(z)| = O(1/|z|)$ as $z\to\infty$ uniformly in every upper half-plane and $f(z) = e^{g(z)}$ will be the answer to the initial question.

$\endgroup$

Your Answer

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

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