Prove the multifunction $[z^{1/2}]$ has no holomorphic branch on $B(0,r)$.
Here is my attempt:
Let $f: B(0,r) /(-\infty,0] \to \mathbb{C}$ be a holomorphic branch of $[z^{1/2}]$. Note $-f$ defines another holomorphic branch on the same cutplane.
For sake of contradiction assume $h: B(0,r) \to \mathbb{C}$ is a holomorphic branch of $[z^{1/2}]$. We want $h(z) \in [z^{1/2}]$, which requires $|z|^{1/2} = |h(z)|$ and $\arg(h(z)) \in [\arg(z^{1/2})]$, so $\arg(h(z)) = \frac{\theta}{2} + 2\pi n(z)$, where $\theta \in [\arg(z)]$ is the principal argument and $n: \mathbb{C} \to \mathbb{Z}$ is some integer valued function. Therefore $h(z)=|z^{1/2}|e^{\frac{i( \theta + 2 \pi n(z))}{2}}$.
Consider $g(z) :=\frac{h(z)}{f(z)}$, where we assume $f(z) \neq 0$. Then, $g(z)=e^{i\pi n(z)} \in \{-1,1\}$. Since $g$ is continuous on a connected domain $g$, is constant, so that $h= \pm f $, which is impossible because $ \pm f$ cannot be holomorphically extended to $B(0,r)$*. That completes my argument, as we have hopefully arrived at a contradiction.
*[$ \lim_{z \to -1} h(z)= \pm f(z)$ approaching above the cut $\neq \lim_{z \to -1} h(z)= \pm f(z)$ approaching below the cut.]
If correct, does this method generalize?
Is my working correct? Thanks!