10
$\begingroup$

In complex analysis, Hurwitz's  theorem  roughly states that, under certain conditions, if a sequence of holomorphic functions converges uniformly to a holomorphic function on compact sets, then after a while  (above a certain rank ) those functions and the limit function have the same number of zeros  in any open disk.

I read somewhere that the result is valid more generally on bounded convex sets , from what I remember, but no references were given.

Question. Where can I find this more general statement of Hurwitz's theorem, and where can I find some useful references in this direction? Thank you.

$\endgroup$

1 Answer 1

12
$\begingroup$

This has nothing to do with convexity. The exact formulation is this:

Let $\Omega$ be an arbitrary region, and $f_n\to f$ is a sequence of holomorphic functions converging uniformly on compacts in $\Omega$. If $f\neq 0$ (this is an important condition!), and $D$ is a region, such that $\overline{D}\subset \Omega$, and $f(z)\neq 0$ on $\partial D_1$, then there exists $N$ (dependng on $D$) such that for $n\geq N$, $f$ and $f_n$ have the same number of zeros in $D$.

For the proof, you find an intermediate region $D_1$ such that $D\subset D_1\subset\overline{D_1}\subset\Omega$, and $\partial D_1$ is piecewise smooth, and the number of zeros of $f$ in $\overline{D_1}$ is the same as the number of zeros in $D$. Such region exists since $\overline{D}\subset\Omega$, and since the zeros of $f$ are isolated and $f(z)\neq 0,\; z\in\partial D$. Then the number of zeros $f$ in $D$ is the same as in $D_1$ and is equal to $$\frac{1}{2\pi i}\int_{\partial D_1} \frac{df}{f},$$ and the number of zeros of $f_n$ in $D_1$ is $$\frac{1}{2\pi i}\int_{\partial D_1} \frac{df_n}{f_n}.$$ Since $f_n\to f$ uniformly on the compact set $K:=\overline{D_1}\backslash D$, and $f(z)\neq 0$ on $K$, we conclude that for $n\geq N$, $f_n$ has the same number of zeros in $D$ and $D_1$. Since the integrals for $f_n^\prime/f_n$ converge to the integral for $f'/f$, and all these integrals are integers, we conclude that they are equal for $n\geq N$.

Refs. This is stated in full generality in the book

A. I. Markushevich, Theory of functions of a complex variable, vol. I, Ch. IV, Sect 3 (p. 426 of the Russian original).

Special cases are in:

Ahlfors, Complex Analysis, p. 178, Theorem 2,

Titchmarsh, The theory of functions, sect. 3.45,

Marshall, Complex Analysis, Theorem 8.8,

But all these special cases have the same proof as the general theorem whose complete proof I wrote.

$\endgroup$
6
  • $\begingroup$ The condition that $f$ has no zeros on $\delta D$ is necessary, right? $\endgroup$ Jan 29, 2022 at 17:11
  • $\begingroup$ @AlexandreEremenko I suspect the condition is that f doesn't vanish on the boundary of D1, and it's implied by the condition that f doesn't vanish on K. $\endgroup$ Jan 29, 2022 at 19:08
  • 1
    $\begingroup$ There is a simple counterexample: $f_n(z)=z+(-1)^n\frac{1}{n}$ and $D$ is the disk $|z+1|<1$ $\endgroup$ Jan 29, 2022 at 21:16
  • 1
    $\begingroup$ The mistake is in the assumption that $f$ has the same number of zeros in $D$ and in $D_1$. That's not true, if there are zeros on the boundary. $\endgroup$ Jan 29, 2022 at 21:20
  • 2
    $\begingroup$ @Oleg Eroshkin: Thanks, I corrected the mistake. $\endgroup$ Jan 29, 2022 at 21:41

Your Answer

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

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