Let $f_c:\mathbb{C}\rightarrow\mathbb{C}$ be a holomorphic function, depending holomorphically on the parameter $c\in\mathbb{C}$. Let $\alpha_0$ be a geometrically attracting fixed point of $f_{c_0}$, i.e. $f_{c_0}(\alpha_0) = \alpha_0$ and $\lambda_0 = f_{c_0}'(\alpha_0) \in \mathbb{D}\backslash\{0\}$.

Suppose there exists a neighbourhood $\Delta\ni c_0$ where we can define, using the implicit function theorem a holomorphic function $\alpha:\Delta\rightarrow\mathbb{C}$ such that $f_c(\alpha(c)) = \alpha(c)$ and naturally $\alpha(c_0) = c_0$, and additionally $\lambda(c) \equiv f_{c}'(\alpha(c))\in \mathbb{D}\backslash\{0\}$.

Now, for each $c\in\Delta$ there are going to be neighbourhoods $V(c)\supset U(c) \ni \alpha(c)$ where $f_c : V(c) \rightarrow U(c)$ is biholomorphic (i.e. it admits a holomorphic inverse).

Can we restrict $\Delta$ in such a way that $$ \alpha(\Delta) \subset \bigcap_{c\in\Delta} V(c)$$ In other words, can we make $\Delta$ small enough so that for all $c_1,c_2\in\Delta$ we have $f_{c_2}(\alpha(c_1)) \in U(c_2)$?

Attempts. I have managed to prove (using continuity) that I can make $\Delta$ small enough so that $\alpha(\Delta) \subset V_(c_0)$, but I am failing to generalise to all parameters in $\Delta$. My intuition strongly says that this ought to be possible, but I fail to have a grasp of how "nicely" the domains $U(c)$ and $V(c)$ change. I tried to embed a disk around $\alpha(c)$ in $V(c)$, but I don't seem to get control on how fast it its radius changes. Maybe I need to use the implicit function theorem again?

(The context for this is that I want to generalise Koenig's Linearisation for a family of functions.)

Any help, advice or (partial) proof appreciated. I apologise if this is obvious, but my background is Physics, and I am still learning.

  • 1
    $\begingroup$ I think you mean you want $f_c$ to be one-to-one on $V(c)$ (just saying "$f_c: V(c) \to U(c)$ is holomorphic" doesn't imply it admits an inverse there). Of course the inverse is holomorphic if it exists. $\endgroup$ – Robert Israel Jun 20 '16 at 1:19
  • 1
    $\begingroup$ The following lemma may be useful. If $|f'(z) - f'(a)| < |f'(a)|$ for all $z$ in a disk containing $a$, then $f$ is one-to-one on that disk. $\endgroup$ – Robert Israel Jun 20 '16 at 1:26
  • $\begingroup$ @RobertIsrael, thanks, corrected. I'm gonna try working with the lemma now thanks. $\endgroup$ – Andrea Jun 21 '16 at 10:45
  • $\begingroup$ @RobertIsrael What is the name of this lemma? $\endgroup$ – Andrea Jun 21 '16 at 12:23
  • $\begingroup$ I don't know that it has a name. $\endgroup$ – Robert Israel Jun 22 '16 at 6:42

Lemma If $|f′(z)−f′(a)|<|f′(a)|$ for all $z$ in a disk containing $a$, then $f$ is invertible on that disk.

There is $r>0$ such that for all $z\in \mathbb{D}_{5r}(\alpha_0)$, $$ |f'_{c_0}(z) - f'_{c_0}(\alpha_0)| < \frac{1}{2} | f'_{c_0}(\alpha_0)| $$ We can then restrict $\Delta$ so that $\alpha(\Delta) \Subset \mathbb{D}_{r}(\alpha_0)$, which implies the following holds for all $c\in\Delta$: $$\alpha(\Delta) \Subset \mathbb{D}_{2r}\big(\alpha(c)\big) \Subset \mathbb{D}_{4r}(\alpha_0)$$

Claim We can find a $\delta>0$ so that $\mathbb{D}_\delta(\alpha_0)\subset \Delta$ and all $c\in\mathbb{D}_\delta(\alpha_0) $ the $f_c$ are invertible on $\mathbb{D}_r(\alpha_0)$.

Proof Note that the set $$V = \bigcap_{c\in\alpha(\Delta)} \mathbb{D}_{2r}(\alpha(c))$$ is a subset of $\mathbb{D}_{4r}(\alpha_0) $ and as such, it is compactly contained in $\mathbb{D}_{5r}(\alpha_0)$.

Since the functions $f'_{c}$ converge to $f_{c_0}$ uniformly on compact subsets as $c\rightarrow c_0$, we can choose, for any $\epsilon >0$, a $\delta>0$ such that for all $c\in\mathbb{D}_\delta({c_0})$, and for all $z\in V$, $$|f'_c(z) - f'_{c_0}(z)| < \epsilon/2$$ Additionally, we can make $\delta$ small enough so that we may assume $\mathbb{D}_\delta(c_0) \subset \Delta$ and $$|f'_{c_0}(\alpha_0) - f'_c(\alpha(c))| < \epsilon/2 ~~~\mbox{ and }~~~ |f'_c(\alpha(c))| > \frac{1}{2}|f'_{c_0}(\alpha_0)| + \epsilon $$

For all $z\in \mathbb{D}_{2r}\big(\alpha(c)\big)$ and $c\in\mathbb{D}_\delta(c_0)$, we then have $z\in V$ and \begin{equation} \begin{aligned} |f'_c(z) - f'_c(\alpha(c))| &\leq |f'_c(z) - f'_{c_0}(z)| + |f'_{c_0}(z) - f'_{c_0}(\alpha_0)| + |f'_{c_0}(\alpha_0) - f'_c(\alpha(c))| \\ &< \frac{1}{2} | f'_{c_0}(\alpha_0)| +\epsilon \\ &< |f'_c(\alpha(c))| \end{aligned} \end{equation} so that $f_c$ is invertible on $z\in \mathbb{D}_{2r}\big(\alpha(c)\big)$. We conclude that each of the functions $f_c$ for $c\in \mathbb{D}_\delta(c_0)$ are invertible over $V \supset \alpha\big( \mathbb{D}_\delta(c_0)\big)$. Finally, it is easy to check that $\mathbb{D}_r(\alpha_0) \subset V$ $\square$.


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.