Let $\mathcal O(\Omega)$ be the algebra of functions holomorphic on the open set $\Omega\subset\mathbb C$. For $\gamma$ a simple compact curve in $\mathbb C$ consider the linear operator given by path-integrating against the Cauchy kernel $$ \Gamma : \mathcal O(\mathbb C) \longrightarrow \mathcal O(\mathbb C\setminus \gamma) \\ f\longmapsto \left(z\mapsto \int_\gamma \frac{f(w)}{w-z}\mathrm{d}w\right)$$

In the trivial case where $\gamma$ is closed then $\Gamma$ is injective (apply Cauchy formula to all $z$ in the bounded connected component of $\mathbb C\setminus \gamma$) and its range consists in those functions that are zero outside $\gamma$ and the restriction of an entire function inside.

My question: what happens if $\gamma$ is not closed (i.e. $\mathbb C\setminus \gamma$ connected)? Is it possible to characterize the range of $\Gamma$, besides the obvious $O(|z|^{-1})$ towards infinity?

(In this case $f$ may only be a germ near $\gamma$ of a holomorphic function, but keep it simple if necessary.)

I'd find it hard to believe that injectivity would fail (although I don't quite have an argument).

If this question has already a well-documented answer (which is fairly probable but I didn't know which keywords to look for), please accept my apologies. A pointer to a reference would be great!


Such an integral over a simple (non-closed) curve is called the Cauchy type integral.

The curve $\gamma$ is oriented, so for a function $F$ defined in $C\backslash\gamma$ and $z\in\gamma$ different from an endpoint, we can talk of the right limit $F^+(z)$ and left limit $F^-(z)$. Then we have Sokhotski formula $$F^-(z)-F^+(z)=f(z),\quad z\in\Gamma$$ As in addition $F(z)\to0,\; z\to\infty$, this shows that your map is injective. The image consists of those analytic functions in $C\backslash\gamma$ which tend to $0$ as $z\to\infty$, have limits on both sides of $\gamma$ and the jump between these limits is an entire function (analytic in $C$). (At the end-points, the limits do not exist, and we have logarithmic singularities.)

With some regularity of $\gamma$ and $f$ (non necessary analytic) we have $$F^+(z)=F(z)-\frac{1}{2}f(z),\quad F^-(z)=F^*(z)+\frac{1}{2}f(z),$$ where $F^*(z)$ is the principal value of your integral. The above formula is obtained by subtracting these two relations. But when $f$ is entire, the curve can be slightly deformed (preserving its ends), and you get rid of the regularity condition by approximating it with a smooth (or piecewise-linear) curve with the same ends.

If the curve happens to be closed then $F^+=0$ and we recover the result for the closed curve which you cite.

It is interesting that the curve $\gamma$ plays little role in the description of the image: functions from the image have analytic continuation along any curve starting and ending at $\infty$ and not passing through the endpoints of $\gamma$.

References. This material is standard and is included to every Russian undergraduate Complex Variables text, but for some strange reasons none of them is translated into English. A good reference in English which comes to my mind is a translation from the Japanese:

Kaneko, A. Introduction to hyperfunctions. Kluwer, Dordrecht, Tokyo, 1988.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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