I want to find an analytic continuation of the square root along the unit circle but I am not sure whether I am doing it correctly.

Let $C_0$ be the open disk of radius $1$ around $1$, and let $f_0:C_0 \to \mathbb{C}$ be defined as $f_0(re^{i \varphi})=\sqrt{r} e^{i \frac{\varphi}{2}}$, where $\varphi \in (-\pi,\pi]$. Let $\gamma: [0,1] \to \mathbb{C}$ be the path given $\gamma(t)=e^{2 it \pi}$. Find an analytic continuation of $f_0$ along $\gamma$, i.e. a sequence $(C_k,f_k)_{k=0}^{n}$ of analytic continuations $f_k$ of $f_0$ such that the $C_k$ cover the image of $\gamma$. Show that $f_n(1)=-f_0(1)$.

I tried to do this as follows. Since $f_0$ is holomorphic on $C_0$ we have the power series expansion

$$ f_0(z)=\sum_{m=0}^{\infty} a^{(0)}_m (z-1)^m \tag{1} $$

where $a^{(0)}_m=\frac{1}{m!} \frac{\partial^m}{z^m} \sqrt{z} \big|_{z=1}$. I wanted to analyticaly continue this series by defining $C_1=\{z \in \mathbb{C} \ | \ |z-e^{i \frac{\pi}{4}}|<1 \}$ and considering the function

$$ f_1: C_1 \to \mathbb{C}, \ f_1(z)=\sum_{m=0}^{\infty} a^{(1)}_m (z-e^{i \frac{\pi}{4}})^m. $$

where $a^{(1)}_m=\frac{1}{m!} \frac{\partial^m}{z^m} \sqrt{z} \big|_{z=e^{i \frac{\pi}{4}}}$ and $arg(z) \in (-\frac{3 \pi}{4}, \frac{5 \pi}{4}]$. Let $z=re^{i\varphi} \in C_0 \cap C_1$. With $z_1=e^{i \frac{\pi}{4}}$ I have

$$ \sqrt{z}=\sqrt{z_1} \sqrt{\frac{z}{z_1}} =\sqrt{z_1} \sqrt{1+\frac{z}{z_1}-1} \underset{(1)}{=}\sqrt{z_1} \sum_{m=0}^{\infty} a^{(0)}_m (\frac{z}{z_1}-1)^m =\sum_{m=0}^{\infty} \frac{\sqrt{z_1}}{z^m_1} a^{(0)}_m (z-z_1)^m $$

Since the series representation of $f_1$ is unique and since $f_0(z)=\sqrt{z}$ the functions $f_0,f_1$ agree on $C_0 \cap C_1$. By iterating the steps above I can define disks $C_2, C_3,...C_8$ with centers $e^{i k\frac{\pi}{4}}$ and corresponding holomorphic functions $f_k$, $k=2,...,8$, each time requiring $arg(z) \in (-\pi+k \frac{\pi}{4},\pi+k \frac{\pi}{4}]$ for $z \in C_k$. Upon considering the power series expansion centered at $e^{i \frac{8 \pi}{2}}=e^{i 2 \pi}$ I should get $f_8(e^{i 2 \pi})=\sqrt{1} e^{i \pi}=-1=-\sqrt{1} e^{i \cdot 0}=-f_0(1)$. Am I on the right track here or is there an error?


This approach should work, but it's considerably more complicated than writing down such functions explicitly. For instance, take:

  • $f_0 (re^{i\theta}) = \sqrt{r} e^{i\theta/2}$ on $\Omega_0 = \{re^{i\theta} \in \mathbb{C} \mid r> 0, \theta \in (-\pi/2, 3\pi/4)\}$,
  • $f_1(re^{i\theta}) = \sqrt{r}e^{i\theta/2}$ on $\Omega_1 = \{ re^{i\theta} \in \mathbb{C} \mid r>0, \theta \in (\pi/2, 7\pi/4) \}$,
  • $f_2(re^{i\theta}) = \sqrt{r}e^{i\theta/2} $ on $\Omega_2 = \{ re^{i\theta} \in \mathbb{C} \mid r>0, \theta \in (3\pi/2, 5\pi/2)\}$.

These $f_i$ are analytic on their domains of definition, $f_i, f_{i+1}$ agree on the intersections of their domains, and $f_2(1) = e^{i\pi} = -1 = - f_0(1)$.

| 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.