6

Consider the two disjoint semicircles $A = \{-e^{i \phi} | -\pi/2 \leq \phi \leq \pi / 2\}$ and $B = \{1 + e^{i \phi} | -\pi/2 \leq \phi \leq \pi / 2\}$ in the complex plane: plot of the two semicircles

Let $f$ be defined and analytic on $\mathbb{C} \setminus A \setminus B$.

Can we construct $f$ so that $g(z) = f(z) - f(z + 1)$ fulfills $g(z) = 0$ for $|z| < 1$, but $g$ does not vanish identically? That is, $f$ is "periodic", but only inside a subset of the plane.

The reason I believe this is possible is because $g$ is only defined on $\mathbb{C} \setminus A \setminus B \setminus (A - 1) \setminus (B - 1)$, which is not connected:

plot of the domain of g

Since $g$ is not defined for $|z| = 1$, the identity theorem does not say anything about its behavior for $|z| > 1$.

If we can find a single solution for $f$, we can also generate new ones of the form $a(f(z))$. However, actually constructing such a function has been challenging to me.

My attempt is to take an arbitrary entire function $e$ (possibly $e(z) = 0$) and then write $$f(z) = e(z) + \int_{L} \mathrm{d}t \frac{u(t)}{z-t}$$ where $L = A + B$ and $u$ is an unknown weight function. From $f(z) = f(z + 1)$ we can then derive a singular integral equation for $u$, for which there seem to be existence theorems. However, this seems very circuitous since $u$ itself is not all that interesting to me.

Is there a simpler formulation that lets us construct $f$ directly? Obviously, any literature on similar problems would be appreciated, since I have a whole host of possible follow-up questions.

This question relates to a previous one. After reorganizing my thoughts I hopefully have extracted the actual mathematical content in a more understandable way.

4177318477
  • 109
  • 4
  • 1
    The domain of $g$ is not connected, since it isn't defined for $|z| = 1$. Unless I am very wrong, this means that we cannot conclude anything about $|z| > 1$. – 4177318477 Dec 31 '25 at 13:45
  • Yes, you are very wrong. Your function $g$ will be analytic on a part of the circle $|z|=1$, namely, where $Re(z)>0$. This suffices. – Moishe Kohan Dec 31 '25 at 14:40
  • $f$ is analytic on that part of the circle, but $g$ is not (because this is where $z + 1 \in B$ and $f(z + 1)$ is not analytic). It may still be impossible to construct $f$, but wouldn't this need a different argument? – 4177318477 Dec 31 '25 at 14:44
  • Oh, sorry, did not notice that $B$ is centered at $1$. – Moishe Kohan Dec 31 '25 at 14:51
  • I believe you may have translated the blue semicircle the wrong direction. I have added a plot of what should be the domain of $g$. – 4177318477 Dec 31 '25 at 16:04
  • yes you are right - my bad - it is definitely an interestng question and will think about it – Conrad Dec 31 '25 at 16:18
  • must $f$ be analytic? if $f$ is not analytic we can use ceiling functions. – Tveltzel Dec 31 '25 at 17:03
  • Yes, $f$ should be analytic (or at least meromorphic) except on $A$ and $B$. Otherwise there are many options. For example, we could set $f(z) = 1$ in the "stadium" bounded by $A$ and $B$ and then link this to an arbitrary function on the outside using either discontinuities or bump functions (https://en.wikipedia.org/wiki/Bump_function). – 4177318477 Dec 31 '25 at 17:15
  • This is very interesting because $f$ must be analytic. (+1) – Tveltzel Dec 31 '25 at 17:26

0 Answers0