3
$\begingroup$

Let $f$ be a function regular in $\Re z \geq 0$ whose indicator function satisfies: $$h(\theta) = \limsup_{r \to \infty} \frac{\log |f(r e^{i \theta})|}{r} \leq c < \pi$$ in this half plane. Let $(\lambda_n)_{n \geq 0}$ be an increasing sequence of positive real numbers satisfying $|\lambda_n - n| < L$ and $\lambda_{n + 1} - \lambda_{n} > \delta$ > 0. My question is about the equivalence of the sum of $|f|$ over this sequence and its integral over the positive real axis, more precisely, whether: $$ \sum_{n = 0}^{\infty} |f(\lambda_n)| < \infty\tag{1}\label{eq:series} $$ implies $$ \int_0^{\infty} |f(x)| \, dx < \infty,\tag{2}\label{eq:integral} $$ possibly under any extra hypotheses.

For context: results by Boas and Kjellberg imply that the converse is true, namely \eqref{eq:integral} implies \eqref{eq:series} (they actually prove a result about more general convex functions $\phi$ of $|f|$). Another result by Boas says that \eqref{eq:series} implies \eqref{eq:integral} if we consider instead the entire real axis and a sequence $(\lambda_n)_{n \in \mathbb{Z}}$, i.e.: $$ \sum_{n = -\infty}^{\infty} |f(\lambda_n)| < \infty. $$ implies $$ \int_{-\infty}^{\infty} |f(x)| \, dx < \infty $$ Can this result be extended to the real positive half-axis?

$\endgroup$
2
  • $\begingroup$ The result of Kjellberg has an extra exponential factor in the sum, and he writes that Boas showed that one cannot dispense of this factor. (Boas paper is not available through your link). For your question, it is easy to construct a counterexample of a function analytic in the open half-plane, continuous in the closed one, so probably the answer is negative. $\endgroup$ Commented May 24 at 11:58
  • $\begingroup$ @AlexandreEremenko I think since the exponential $e^{-c \delta}$ is a constant, it factors out for the particular case of the convex function $\phi(x) = x$ considered here. The counterexample Boas puts forward is $\phi(x) = \log^+(x)$ and $f(z) = (1 + z^{-1}) \sin(\pi z / 2)$. $\endgroup$ Commented May 24 at 19:36

1 Answer 1

3
$\begingroup$

The function $$f(z)=e^{i\pi z}\sin \pi z=\frac{1}{2i}(e^{2\pi iz}-1)$$ is entire, bounded in the upper half-plane, $f(n)=0$, while $$\int_{-\infty}^\infty |f(x)|dx=\infty,$$ since the function is periodic.

$\endgroup$
2
  • $\begingroup$ Thanks for your answer. Maybe I'm misunderstanding the argument, might there be a typo? I would believe that $f(n) \neq 0$ since $e^{2 i n} \neq 1$. If we add a factor $\pi$ to the exponent, then indeed $f(n) = 0$ but $h(\theta = -\pi / 2) = 2 \pi \geq \pi$ against the hypothesis. $\endgroup$ Commented May 24 at 19:48
  • $\begingroup$ @Esteban Martinez: Thanks, I edited. $\endgroup$ Commented May 25 at 12:05

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.