1
$\begingroup$

Let $\mathcal{O}_{z}$ be the space of germs of holomorphic functions at $z$, defined as a direct limit of a system $$\mathcal{O}_{z} = \lim_{\rightarrow}{\mathcal{O}(U_{\frac{1}{n}}(z))}$$ where $U_{\frac{1}{n}}(z)$ stands for the unit disk centered at $z$.

Each $\mathcal{O}(U_{\frac{1}{n}}(z))$ is endowed with the topology of pointwise convergence and $\mathcal{O}_{z}$ inherits the corresponding inductive locally convex topology.

$\mathcal{O}_{z}$ can be decomposed as $$\mathcal{O}_{z} = \mathbb{C}1 \oplus \mathcal{O}_{z}^{0}$$ where the latter stands for $$ \mathcal{O}_{z}^{0} = \{ f \in \mathcal{O}_{z} | f(z) = 0 \} $$ and the decomposition is as follows $$ f(z) = a_{0} + (z-a)(a_{1}+a_{2}(z-a)+a_{3}(z-a)^{2} + \ldots)$$ where $a_{k}$ stands for the $k$-th coefficient in the Taylor series expansion at $z = a$.

How to prove that the topology on $\mathcal{O}_{z}^{0}$ induced from the $\mathcal{O}(z)$ is trivial? In fact, this result would imply that the chosen inductive topology is non-Hausdorff.

$\endgroup$
1
$\begingroup$

For each $n$, let $\rho_n \colon \mathcal{O}(U_{\frac{1}{n}}(z)) \to \mathcal{O}_z$ the natural map. A neighbourhood basis of $0$ in $\mathcal{O}_z$ is given by the family of absolutely convex (i.e. convex and balanced) $V \subset \mathcal{O}_z$ such that $\rho_n^{-1}(V)$ is, for every $n$, a neighbourhood of $0$ in $\mathcal{O}(U_{\frac{1}{n}}(z))$. Let $V$ be any such set. By definition of the topology of pointwise convergence, for every $n$ there is a finite $F_n\subset U_{\frac{1}{n}}(z)$ and a $\varepsilon_n > 0$ such that $\lvert f(w)\rvert < \varepsilon_n$ for all $w\in F_n$ implies $\rho_n(f) \in V$.

Let $\delta = \min\:\bigl\{ \lvert w-z\rvert : w \in F_1\setminus \{z\}\bigr\}$ and let $n > 1/\delta$. Consider an arbitrary $f \in \mathcal{O}(U_{\frac{1}{n}}(z))$ with $\lvert f(z)\rvert < \varepsilon_1/2$. There is an entire $g$ with $g(w) = 2f(w)$ for $w \in F_n$ and $g(w) = 0$ for $w \in F_1\setminus \{z\}$. Then $g \in \rho_1^{-1}(V)$, and therefore $\tilde{g} = g\rvert_{U_{1/n}(z)}\in \rho_n^{-1}(V)$. Also, $2f - \tilde{g}$ vanishes on $F_n$, so $2f-\tilde{g} \in \rho_n^{-1}(V)$. By convexity, it follows that

$$f = \frac{1}{2}\bigl(2 f - \tilde{g}\bigr) + \frac{1}{2}\tilde{g} \in \rho_n^{-1}(V)\,.$$

Thus

$$V \supset \bigl\{ f \in \mathcal{O}_z : \lvert f(z)\rvert < \varepsilon_1/2\bigr\}\,.$$

Conversely, for every $r > 0$ the set

$$V_r = \bigl\{ f \in \mathcal{O}_z : \lvert f(z)\rvert < r\bigr\}$$

is an absolutely convex set such that $\rho_n^{-1}(V_r)$ is a neighbourhood of $0$ in $\mathcal{O}(U_{\frac{1}{n}}(z))$ for every $n$, hence this family is a neighbourhood basis of $0$ in $\mathcal{O}_z$.

In other words, the inductive limit topology on $\mathcal{O}(z)$ is induced by the seminorm $p \colon f \mapsto \lvert f(z)\rvert$ whose kernel is $\mathcal{O}_z^0$. Hence the subspace topology on $\mathcal{O}_z^0$ is the trivial topology.

$\endgroup$
  • $\begingroup$ @DanielFisher Thanks! I do admit that i was really far from the final solution. $\endgroup$ – hyperkahler Jan 28 '18 at 17:06

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.