Let $f:[0,+\infty)$ be a continuous function that satisfies:

$f(x+q)$~$f(x)$ for $x\to\infty$ (for any $q$)

Does it follow that $f$ is uniformly continuous?

I have managed to show that if there exists $\space$ $\displaystyle\lim_{x\to\infty}\space f(x)=G\in\Bbb{R}$ $\space$ then the function must be uniformly continuos by for given $\epsilon$ picking an $N$ big enough that $\forall_{x>N} |f(x)-G|<\frac{\epsilon}{2}$ and then showing that the function is uniformly continuous on $[0,N]$ and satisfies the definition of unifom continuity for that $\epsilon$ on $[N,+\infty)$, thus proving it must be uniformly continuous,since we could have chosen any $\epsilon$.

However, that approach fails when we consider the cases where $\displaystyle\lim_{x\to\infty}\space f(x)$ is infinite or non-existent.

I've also tried to find a counterexample by experimenting with functions like $\frac{1}{x}\sin(x^{3})$ (which appeared promising since its derivative is unbounded) but so far I haven't found one and my intuition does not steer me to either of the answers.

I would appreciate any hints :)

  • $\begingroup$ Idk why ur looking at $\frac{\sin(x^3)}{x}$, cause that converges to $0$. Also, is it possible for the limit to just not exist (I'm excluding $\lim = +\infty$)? I feel like it's not, by a Baire Category theorem argument. $\endgroup$ Apr 10, 2020 at 18:54
  • $\begingroup$ @mathworker21 What about $\sin x?$ $\endgroup$
    – zhw.
    Apr 10, 2020 at 19:04
  • $\begingroup$ @zhw. does that satisfy $f(x+t)-f(x) \to 0$ as $x \to \infty$ for all $t$? Not to be rude, but it's a bit insulting if you were suggesting I was suggesting that it's not possible for an arbitrary function to not have a limit at infinity. $\endgroup$ Apr 10, 2020 at 19:05
  • $\begingroup$ @mathworker21 I see, it has to work for all $t.$ I missed that. $\endgroup$
    – zhw.
    Apr 10, 2020 at 19:08
  • $\begingroup$ I have a counterexample if we have the condition for just $t=1$. $\endgroup$ Apr 10, 2020 at 19:09

2 Answers 2


Yes, it does.

Firstly, let us define the following set:

$$S_{\varepsilon,X} = \{t \in \mathbb R_{\geq 0} : \text{if }x\geq X\text{ then }|f(x+t)-f(x)| \leq \varepsilon\}.$$ By hypothesis, for any fixed $\varepsilon$, we have $$\bigcup_{X\in\mathbb R_{\geq 0}}S_{\varepsilon,X} = \mathbb R_{\geq 0}.$$ Note that each set $S_{\varepsilon,X}$ is closed because for a fixed $x$, the set of values of $t$ such that $|f(x+t)-f(x)|\leq \varepsilon$ is closed and $S_{\varepsilon,X}$ is an intersection of these closed sets over all $x\geq X$.

Note that we could also say that $$\bigcup_{X\in\mathbb Z_{\geq 0}}S_{\varepsilon,X} = \mathbb R_{\geq 0}.$$ since the sets $S_{\varepsilon,X}$ increase with $X$ - giving a countable union of closed sets whose union is the whole space.

We can then apply the Baire Category Theorem to say that since a countable union of closed sets has non-empty interior, some element of the union must have an interior! In particular, for any $\varepsilon>0$, there must be some $X$ such that some interval $(a,b)$ is a subset of $S_{\varepsilon,X}$. However, then if we have that $x,y \geq X + a$ and $|x-y| < |b-a|$ we could choose some pair $a',b'\in (a,b)$ with $x-a' = y-b'$ and then observe that $$f(x)=f(x-a' + a')$$ $$f(y)=f(y-b' + b')$$ Then the distance from $f(x)$ to $f(x-a')$ is at most $\varepsilon$ as is the distance from $f(y)$ to $f(y-b')$ since $a',b'\in \subseteq S_{\varepsilon,X}$. Thus we find that if $x,y \geq X+a$ and $|x-y| < |b-a|$ we have $|f(x)-f(y)| \leq 2\varepsilon$ - and this works out for any choice of $\varepsilon>0$. This fact suffices to establish that $f$ is uniformly continuous with a small bit of further work.

  • $\begingroup$ sigh... i need to start using this form of the Baire Category Theorem. I always use a weaker one. anyways, nice solution +1 $\endgroup$ Apr 10, 2020 at 19:56
  • 1
    $\begingroup$ @mathworker21 I always end up looking up the various forms of the Baire Category Theorem whenever I see a problem where it might apply - it's a strange theorem in that it's pretty clear when it might be helpful, and yet I can never remember exactly what it says. $\endgroup$ Apr 10, 2020 at 20:02
  • $\begingroup$ Not to diminish your solution, but it really does trivialize this problem. It immediately gives all $t$ in an interval for a particular $X$. $\endgroup$ Apr 10, 2020 at 20:05

Here's a counterexample if we only know that $\lim_{x \to +\infty} f(x+t)-f(x) = 0$ for all $t \in \mathbb{Q}$.

Let $T_n$ denote an isosceles triangle with width $\frac{1}{2n!}$ and height $n$. Let $T_n^{(j)}$ for $j=0,1,\dots,(n+1)!n!-n!n!$ be a slight morphing of $T_n$ into $T_{n+1}$ (i.e. very slightly shrinking the width and increasing the height from $j$ to $j+1$).

We let $f$ be $T_n^{(j)}$ with bottom left corner at $x=n!+\frac{j}{n!}$ ($f$ is a bunch of upward spikes).

I suspect the result is true if we require $\lim_{x \to +\infty} f(x+t)-f(x) = 0$ for all $t \in \mathbb{R}$, and that a proof will use a Baire Category Theorem argument.


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.