3

I am working on a problem stating as below:

Consider a holomorphic function $f$ defined on the puncture disc $D(0,1)\setminus\{0\}$. Show that $0$ is a removable singularity of $f$ if $f$ is square integrable.

This question is similar to the post here: Singularities in the punctured unit disc and square integrability

In fact, I've solved it following the idea in above post.

Below is my proof:

We can write $f(z)$ as Laurent Expansion around $z_{0}=0$, such that $$f(z)=\sum_{n=-\infty}^{\infty}a_{n}z^{n}.$$

Then, we have $$f(re^{i\theta})=\sum_{n=-\infty}^{\infty}a_{n}r^{n}e^{in\theta},\ \overline{f(re^{i\theta})}=\sum_{n=-\infty}^{\infty}\overline{a_{n}}r^{n}e^{-in\theta}.$$

Note that for the integral $$\int_{0}^{2\pi}e^{in\theta}e^{im\theta}d\theta,$$ if $n=-m$, then the above integral is $2\pi$, but if $n\neq -m$, then the above integral is a complex integral of a holomorphic function along a circle and thus by Cauchy's Theorem, the above integral is $0$.

Now, with this in mind, we have \begin{align*} \int_{0}^{2\pi}|f(re^{i\theta}|^{2}d\theta&=\int_{0}^{2\pi}\Big(\sum_{n=-\infty}^{\infty}a_{n}r^{n}e^{in\theta}\Big)\Big(\sum_{n=-\infty}^{\infty}\overline{a_{n}}r^{n}e^{-in\theta}\Big)d\theta \\ &=2\pi\sum_{n=-\infty}^{\infty}|a_{n}|^{2}r^{2n}.\\ \end{align*}

On the other hand, since $\|f\|_{L_{2}}<\infty$, for any disc $D_{z_{0}}(R)$ centered at $z_{0}=0$ with radius $R$, we have \begin{align*} \infty>\int_{D}|f(z)|^{2}dz&=\int_{0}^{R}\int_{0}^{2\pi}|f(re^{i\theta})|^{2}4d\theta dr\\ &=2\pi\int_{0}^{R}\sum_{n=-\infty}^{\infty}|a_{n}|^{2}r^{2n+1}dr\\ &=2\pi\sum_{n=-\infty}^{\infty}|a_{n}|^{2}\int_{0}^{R}r^{2n+1}dr\\ \end{align*}

Now, for all $2n+1\geq 0$, $\int_{0}^{R}r^{2n+1}dr<\infty$, but for all $2n+1<0$, $\int_{0}^{R}r^{2n+1}dr=\infty$.

Thus, the only way to make the above inequality hold is that $2n+1\geq 0$, which means that $n\geq 0$ since $n\in\mathbb{Z}$.

This implies that in the Laurent series, $a_{n}=0$ for all $n\leq -1$. This implies that $z_{0}=0$ is a removable singularity.

However, this question is the part (c) of a problem, and I am wondering if there is another way to prove it, by using part (a) and (b).

Here is the part (a) and part (b):

(a) Show that $0$ is a removable singularity if $|f(z)|\leq C|z|^{-\alpha}$, with $\alpha<1$.

(b) Show that, for any holomorphic function $g$ on the disc of center $b$, radius $\epsilon$, we have $$|g(b)|\leq\dfrac{C}{\epsilon}\Big(\int_{D(b,\epsilon)}|g(x+iy)|^{2}dxdy\Big)^{1/2}.$$

I have proven those two parts and they both of a generalization in Stein Chapter 3 Exercise 13 and 20, respectively.

However, I have no idea about how to apply those two to part (c). Perhaps they are really not connected to each other.

1 Answers1

3

You can use (b) and the following version of Riemann's theorem on removable singularities:

(a') $0$ is a removable singularity if $\lim_{z \to 0} z f(z) = 0$.

For $0 < r < 1/2$ and $|z|= r$ apply (b) to $f$ on the disk $D(z, r)$: $$ |f(z)| \le \frac {C}{r}\left(\int_{D(z,r)}|f(x+iy)|^{2}dxdy\right)^{1/2} \\ \implies |zf(z)| \le C \left(\int_{\dot D(0, 2r)}|f(x+iy)|^{2}dxdy\right)^{1/2} $$ where $\dot D(0, 2r) = D(0, 2r) \setminus \{ 0 \}$ shall denote the punctured disk.

The right-hand side converges to zero for $r \to 0$ (this follows e.g. from the Lebesgue theorem on dominated convergence), and then (a') implies that the singularity is removable.

Martin R
  • 113,040
  • To apply Lebesgue Dominated Convergence Theorem, which sequence should I use? – JacobsonRadical Jul 14 '19 at 00:01
  • @JacobsonRadical: $g_r(x, y) = |f(x+iy)|^2$ if $x^2+y^2 < r^2$, and $g_r(x,y) = 0$ otherwiese. – Martin R Jul 14 '19 at 04:54
  • Yes... I was being dumb. Thank you so much! – JacobsonRadical Jul 14 '19 at 06:26
  • ah i am sorry for reopening such an old question but today I re-did this question and I could not figure it out. So by part (b), we have $$|zf(z)|\leq \dfrac{C}{r}|z|\Bigg(\int_{D(z, r)}|f(z)|^{2}dz\Bigg)^{\frac{1}{2}}=C\Bigg(\int_{D(z, r)}|f(z)|^{2}dz\Bigg)^{\frac{1}{2}},$$ but how did you translate the integral domain from $D(z,r)$ to $D(0,2r)\setminus{0}$? Somehow you shift $D(z,r)$ to $D(0,r)$ and expand it? I understand that after translation the integral is the same since Lebesgue measure is invariant under this, but I am not sure if we can shift $D(z,r)$ to $D(0,r)\setminus{0}$. – JacobsonRadical Dec 05 '20 at 23:49
  • @JacobsonRadical: $D(z, r)$ is a subset of $ D(0, 2r) \setminus { 0 }$. So the integration domain is just increased. – Martin R Dec 06 '20 at 02:46
  • Sorry for the late reply. Okay I get it, then I have $$|zf(z)|\leq C\Bigg(\int_{D(0,2r)\setminus{0}}|f(x+iy)|^{2}dxdy\Bigg),$$ so we take $r\rightarrow 0$, by DCT we have RHS=0, but how could I conclude $$\lim_{z\rightarrow 0}|zf(z)|=0?$$ Like, we only have $$\lim_{|z|=r\rightarrow 0}|zf(z)|=0,$$ right? – JacobsonRadical Dec 06 '20 at 05:22
  • @JacobsonRadical: Where is the difference? $z \to 0$ is equivalent to $|z| \to 0$. – Martin R Dec 06 '20 at 08:14
  • Right. Sorry, I was being confused. Thanks for the great clarification! – JacobsonRadical Dec 06 '20 at 21:12