1
$\begingroup$

Let $C(\mathbb{R})$ be equipped with the topology of compact convergence (or equivalently the compact-open topology). Then, is the subset $\left\{f\in C(\mathbb{R}): \text{$f$ injective} \right\}$ an open subset therein?

$\endgroup$
2
  • 4
    $\begingroup$ You can easily find a sequence of continuous functions $f_n$ each $[0,1]\rightarrow [0,1]$ with $f_n(0)=0, f_n(1)=1$, each non-injective, and with $f_n(x)\rightarrow x$ uniformly. Now extend each function to $\mathbb R$ by defining it as the identity outside $[0,1]$. This gives a sequence of non-injectives which converges uniformly on compacta to an injective. So the complement of your set is not closed. $\endgroup$ Commented Jul 29, 2020 at 11:12
  • 3
    $\begingroup$ A continuous function on the reals is injective iff it is either strictly decreasing or strictly increasing. $\endgroup$ Commented Jul 29, 2020 at 15:37

3 Answers 3

12
$\begingroup$

A simpler example is $f_\epsilon(x)=x^3+\epsilon x$ which is injective iff $\epsilon\geq 0$.

$\endgroup$
11
$\begingroup$

I think that this is a kind of question (which is not research level and) which should not be answered by a counterexample but by an explanation: What would it mean that, e.g., the identity map is an interior point of the set of injective continuous functions? There should be a compact interval $[-n,n]$ and $\varepsilon>0$ such that any function $f\in C(\mathbb R)$ with $|f(x)-x|<\varepsilon$ for $x\in [-n,n]$ is injective. This is absurd because you have no condition at all outside $[-n,n]$. But also inside the interval one can easily modify the graph of the identity to get a non-injective function which $\varepsilon$-close to the identity.

$\endgroup$
3
$\begingroup$

Based on Matthew's post here we go:

Let $f_n(x)\triangleq \left|\frac{x}{n+1}\cos(\frac{x-1}{n})\right| + \left(1-\frac1{n+1}\right) x$ and $f(x)=x$ and $\sup_{x \in [0,1] }\|f_n(x)-f(x)\| \in \mathscr{O}(n^{-1})$.

This provides a counter example on $C((0,1))$ and then just use the homeomorphism: $$ \begin{aligned} C(0,1) &\rightarrow C(\mathbb{R})\\ f &\mapsto f \circ \frac{1}{1+ \exp(-x)}, \end{aligned} $$ to get the conclusion. (Note, that is preserves the class of injective maps).

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .