This following fact came up in a course of complex analysis I was studying, and I was wondering how to prove it.

Suppose that $f:D \rightarrow \mathbb{C}$ is continuous, and that $\oint f(z) dz=0$. $D$ is a domain, not necessarily simply connected.

Let $\Gamma$ be a curve connecting $z_0,z \in \mathbb{C}$, define $F(z)=\int_{\Gamma} f(w) dw$.

Then, $F$ is an analytic function.


It is easy to show that $F$ is well defined, and I was able to do that. I know that there is a real analysis version that is the fundamental theorem of calculus, but to prove analycity I need to show that the C-R eq's hold which is what I am struggling with.


This sort of problem is easier if you don't try to decouple the real and imaginary parts. Just try to compute the complex derivative as follows.

$F(x) -F(y) = \int_{\Gamma_1} f(w) dw - \int_{\Gamma_2} f(w) dw = \int_{\Gamma_1} f(w) dw + \int_{\Gamma_2'} f(w) dw $

where $\Gamma_1$ goes from $z_0$ to $x$ and $\Gamma_2'$ goes from $y$ to $z_0$. Using the closed-loop property we see

$F(x) -F(y) = \int_{\Gamma} f(w) dw $

where $\Gamma$ goes from $y$ to $x$. Again the closed loop property says we can take $\Gamma$ as a straight-line segment.

Now estimate $\frac{F(x) -F(y) }{x-y}$ using the fact that $f(w)$ is very close to $f(z_0)$ for $x$ close to $y$.

  • $\begingroup$ "the closed loop property says we can take $\Gamma$ as a straight-line segment" Yes, but $D$ might not be convex. $\endgroup$ – Arthur Jul 29 '19 at 11:03
  • $\begingroup$ Is the domain open? In that case just restrict attention to the $y \in D$ in a small convex disc around $x$. Since we're ultimately taking the limit as $y \to x$ this does not cause a problem. $\endgroup$ – Daron Jul 29 '19 at 11:11
  • $\begingroup$ Sure, it can be mended. But leaving it as an unqualified statement is objectively false. $\endgroup$ – Arthur Jul 29 '19 at 11:13
  • $\begingroup$ Objectively false! That's the worst kind of false champ! $\endgroup$ – Daron Jul 29 '19 at 11:22

The main reason with it holds is that analycity is a local condition, not a global one. Once you know $F$ exists, you can work locally, and assume the interior of $D$ is connected, star-shaped, convex...


I suppose that when you write that $\oint f(z)\,\mathrm dz=0$, what you mean is that the integral of $f$ over any closed path is equal to $0$.

Let $z\in D$ and consider the quotient$$\frac{F(z+h)-F(z)-hf(z)}h,$$where $h$ is such that $z+h\in D$. Now, let $\Gamma$ be a path in $D$ going from $z_0$ to $z$ and let $\Gamma^\star$ be the same path followed by a path that goes in a straight line from $z$ to $z+h$. Then$$F(z+h)-F(z)=\int_{\Gamma^\star}f(w)\,\mathrm dw-\int_\Gamma f(w)\,\mathrm dw=\int_{\eta}f(w)\,\mathrm dw,$$where $\eta(t)=(1-t)z+t(z+h)$. Therefore,$$F(z+h)-F(z)-hf(z)=\int_\eta f(w)-f(z)\,\mathrm dw\tag1$$Take $\varepsilon>0$. And now take $\delta>0$ such that $\lvert w-z\rvert<\delta\implies\bigl\lvert f(w)-f(z)\bigr\rvert<\varepsilon$. It follows from $(1)$ and from the fact that the length of $\eta$ is $\lvert h\rvert$ that $\left\lvert F(z+h)-F(z)-hf(z)\right\rvert<\lvert h\rvert\varepsilon$. So$$\lvert h\rvert<\delta\implies\left\lvert\frac{F(z+h)-F(z)-hf(z)}h\right\rvert<\frac{\lvert h\rvert\varepsilon}{\lvert h\rvert}=\varepsilon.$$And this proves that$$\lim_{h\to0}\frac{F(z+h)-F(z)-hf(z)}h=0$$which is equivalent to$$\lim_{h\to0}\frac{F(z+h)-F(z)}h=f(z).$$In other words, $F'(z)=f(z)$.

As you can see, the Cauchy-Riemann equations are not needed here.


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.