6

Let $f$ be an entire function on the complex plane.

Is the radius of convergence of $f$ around any point $z_0$ infinite? If so, why?

Thank you.

u1571372
  • 3,345

1 Answers1

10

A function $f: \mathbb{C} \to \mathbb{C}$ is said to be entire if it is holomorphic in the all complex plane, but using the Cauchy's Integral Formula, we have that a holomorphic function in a region is also analytic in this same region. Therefore, an entire function $f$ is analytic in the all complex plane. Furthermore, we also have that the radius of convergence of a series centred at a point $z_0$ is the distance between $z_0$ and the nearest singularity of the series, in that case, it will be infinite.

EDIT (10 years later, time really flies) If I remember correctly, this is how the argument goes:

The radius of convergence is precisely defined as $$ R(f,z_0):= \sup\left\{r \ge 0: \left\|\sum_{n=0}^\infty \frac{f^(n)(z_0)}{n!} (z-z_0)^n\right\| < \infty, \forall z\in B(z_0,r)\right\}. $$ Let us also define another value $$ \tilde{R}(f,z_0):= \sup\left\{r \ge 0: f'(z) \text{ exists for all } \forall z\in B(z_0,r)\right\}. $$

Showing that $R(f,z_0)\le \tilde{R}(f,z_0)$ is an mere exercise of differentiating power series. We will now show that $R(f,z_0)=\tilde{R}(f,z_0)$. In particular, for entire functions, we know that $\tilde{R}(f,z_0)$ is infinity, and we want to show that $R(f,z_0)$ is also infinity.

For all $z \in B(z_0,r)$ for any $r \in (0,\tilde{R}(f,z_0))$, we can use the Cauchy integral formula to get $$ (\star)\qquad f(z)=\frac{1}{2\pi i}\oint_\gamma \frac{f(w)}{z-w}dw $$ where $\gamma(t):= z+r e^{it}$.

We can then use the Taylor series $$ \frac{1}{w-z}=\frac{1}{(w-z_0)-(z-z_0)}=\frac{1+\frac{z-z_0}{w-z_0}+\left(\frac{z-z_0}{w-z_0}\right)^2+\cdots}{w-z_0}.$$ And show that the series $$\sum_{n=0}^\infty \frac{f(w)(z-z_0)^n}{(w-z_0)^{n+1}}$$ converges absolutely (also a simple exercise).

This allows us to exchange the order of the series and integral to rewrite the right-hand side of $(\star)$ as $$ \sum_{n=0}^\infty \frac{1}{2 \pi i} (z-z_0)^n \oint_\gamma f(w) \frac{1}{(w-z_0)^{n+1}} dw. $$

Well, but because of $(\star)$, this coincides with $f(z)$. But now we have a power series representation of $f(z)$ that works anywhere in $B(z_0,r)$. Once again, by differentiating this power series, it is easy to show that this must be the Taylor series of $f$. As $r \le \tilde{R}(f,z_0)$ was arbitrary, we get $R(f,z_0)\ge\tilde{R}(f,z_0)$, therefore, $R(f,z_0)=\tilde{R}(f,z_0)$.

Kernel
  • 1,951
  • 1
    Here, there is a link which explains that in detail: http://en.wikipedia.org/wiki/Analyticity_of_holomorphic_functions . – Kernel Jun 22 '14 at 18:51
  • 1
    Thank you, that's what I wanted. – u1571372 Jun 22 '14 at 18:52
  • We know from the definition of an analytic function that in some neighborhood of every point in it's domain it can be represented by power series. So for an entire function it means that for every point $a \in \mathbb{C}$ we have the equality $f(z)=\sum_{k \ge 0} c_k \cdot (z-a)^k$ in some neighborhood of $a$. I don't understand why do we get the radius is $\infty$. Can you please elaborate on that a little bit more? The link you provided doesn't answer the question actually :(( – Levon Minasian Apr 04 '22 at 16:48
  • 1
    Notice that the proof did not depend on the radius of the disk $D$, so it can be taken arbitrarily large. Does that make sense? – Kernel Apr 06 '22 at 08:57
  • I don't really get where the claim over the distance of $z_0$ from singularities comes from. – Andreas Compagnoni Dec 18 '24 at 00:58
  • Hi there, hopefully this clarifies – Kernel Dec 18 '24 at 10:44
  • I think it should be uniform convergence that allows interchanging the series and the integral instead of absolute convergence. – 19021605 Apr 20 '25 at 06:24