Suppose $f : \mathbb{R} \rightarrow \mathbb{R}^n$ is a real analytic function on $(a, \infty)$. I have two questions:

  1. Suppose $||f(x)|| \rightarrow \infty$ as $x \rightarrow \infty$. I know without further conditions $f$ can be not injective (for e.g. take $f(x) = x + 2 \sin(x)$ for $n=1$). I want to know if any generic conditions are known such that $f$ is eventually injective (meaning that there exists $b > a$ s.t $f$ is injective on $(b, \infty)$.

  2. Suppose $f(x)$ converges as $x \rightarrow \infty$, and $(\partial^m f / \partial x^m)(x) \rightarrow 0$ as $x \rightarrow \infty$, for all $m \geq 1$. Again without further conditions $f$ is not eventually injective (for e.g. $f(x) = e^{-x} \sin (x)$ shows it is not true generically). So again my question is whether there are some simple conditions known that makes $f$ eventually injective.

  • 2
    $\begingroup$ In 1, with "cannot be injective" you presumably mean "can be not injective". Your counterexample also doesn't work since $|x\sin(x)|$ doesn't tend to infinity. However, $f(x)=x+2\sin x$ will work for that. $\endgroup$
    – Wojowu
    Jul 1 '20 at 16:34
  • $\begingroup$ Thanks @Wojowu. Updated the question. $\endgroup$ Jul 1 '20 at 17:13
  • $\begingroup$ If $n>1$, injectivity would be generic. For n=1, monotonicity seems to be what you are looking for. $\endgroup$ Jul 1 '20 at 17:37
  • $\begingroup$ @MichaelRenardy I myself would expect injectivity to be generic for $n>2$, but not for $n=2$, by analogy with random paths. I don't know of a way to formalize this statement though, especially if we condition that the functions tend to infinity. Do you have any reason (be it informal) to expect injectivity for $n=2$? $\endgroup$
    – Wojowu
    Jul 1 '20 at 17:47
  • $\begingroup$ It can't be generically injective for $n=2$, e.g. we can force $f$ to be non-injective by requiring $|f(x) - g(x)| < 1/3$ for $-\pi \le x \le \pi$, where $g(x) = [\sin(x),\sin(2x)]$. $\endgroup$ Jul 1 '20 at 19:38

For $n=1$, such simple (non-tautological) conditions do not exist, because real-analytic functions can mimic any $C^1$ functions in terms of their monotonicity patterns and the limit value at $\infty-$, simultaneously. So, the real-analyticity condition does not help at all; it is far from any kind of a rigidity condition.

Indeed, for any $a\in\mathbb R$, take any $C^1$ function $g\colon[a,\infty)\to\mathbb R$ such that $\exists\ g(\infty-)\in[-\infty,\infty]$ and for some increasing to $\infty$ sequence $(x_j)$ in $[a,\infty)$ we have $(-1)^j g'(x_j)>0$ for all $j$. Extend $g$ to a $C^1$ function on $\mathbb R$ and denote the extension still by $g$.

Take now any continuous function $h\colon\mathbb R\to(0,\infty)$ such that $h(x_j)<|g'(x_j)|$ for all $j$ and $h(\infty-)=0$. Then, according to a theorem by Whitney (see e.g. the first paragraph of this paper), there is a real-analytic function $f\colon\mathbb R\to\mathbb R$ such that $|f-g|+|f'-g'|\le h$ and hence $f(\infty-)=g(\infty-)$ and $(-1)^j f'(x_j)>0$ for all $j$ -- so that the real-analytic function $f$ indeed mimics the monotonicity pattern of the $C^1$ function $g$ and has the same limit value at $\infty-$.


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.