Let $f:[0,\infty) \rightarrow \mathbb R$ be a function such that for any positive $a$ the sequence $f(an)$ converges to zero.Does the limit $lim_{x\rightarrow \infty} f(x)$ exist?

My problem is should I look for a counter example or try to prove it in which I have not succeeded till now.

  • $\begingroup$ Generally, if you are unsure if something is true, you should spend half your time trying to prove it, and the other half trying to disprove it. Eventually you will succeed. $\endgroup$ – Mike Earnest Oct 25 '14 at 3:02
  • 2
    $\begingroup$ Did you forget to include the condition that $f$ is continuous? If this is not a requirement, then the statement is false (define $f(\sqrt{p})=1$ for any prime number $p$, $f(x)=0$ otherwise). $\endgroup$ – Mike Earnest Oct 25 '14 at 3:36

When you add the condition that $f$ also be continuous, it turns out to that $f(x)\to0$. Here is a sketch of the proof.

Suppose $\lim_{x\to\infty}f(x)\neq 0$. Then there will be some $\epsilon>0$ and a sequence of points $x_k\to\infty$ for which $f(x_k)>2\epsilon$. Since $f$ is continuous, there is a closed interval $I_k$ around each point where $f(x)>\epsilon$ for $x\in I_k$.

For any $E\subset \mathbb{R}$, $c\in\mathbb{R}$, let $c\cdot E$ denote $\{cx\,|\,x\in E\}$.

Let $E_0=[0,1]$. Define a sequence of closed intervals $E_n$ for $n\ge1$ recursively as follows: choose some integer $m$ large enough so that, for some $k$, $(m\cdot E_{n-1})\cap I_k$ is a nontrivial closed interval (not just a point). You can show such an $m$ exists. Then, for that $k$, define $E_n=\frac1m\cdot( (m\cdot E_{n-1})\cap I_k)$. You may have to also insist if you use $I_k$ you choose for $E_n$ was not used for any previous $E_m$.

Since the $E_n$ are nested closed intervals, their intersection is nonempty. You can then show that, for any point $e$ in their intersection, the sequence $f(e),f(2e),f(3e)\dots$ will not approach zero, since the sequence $e,2e,3e\dots$ will hit infinitely many intervals $I_k$.


I don't believe this is true.

Let $\{x_n\}$ be a sequence of distinct real numbers greater than 1 which are not multiples of each other and let $f(x_i*n) = 1$ if $i \le n$ and $f(x_i*n) = 0$ if $i \gt n$. Now suppose there was an $L$ such that $|f(x)| < \epsilon$ if $x > L$, this is not possible because for any $L$ we would have some $x_i*n \gt L$ where $f(x) = 1$.

  • $\begingroup$ err, is this even continuous? $\endgroup$ – quasicompactscheme Oct 25 '14 at 3:46
  • $\begingroup$ Does it have to be? $\endgroup$ – integrator Oct 25 '14 at 4:01
  • $\begingroup$ for some reason I thought it had to be, I guess not $\endgroup$ – quasicompactscheme Oct 25 '14 at 12:36

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.