Given that $f$ is differentiable, what does $\lim\limits_{x \to \infty} f(x) = 1$ say about $\lim\limits_{x \to \infty} f^\prime(x)$ ? Intuitively I feel that it's $0$.

I attempted to solve this by trying to evaluate $$\lim_{x \to \infty} \lim_{h \to 0} \frac{f(x + h) - f(x)}{h} $$ by trying to interchange the limits after showing that $\frac{f(x + h) - f(x)}{h}$ converges uniformly as $x \to \infty$. But I couldn't proceed further.

Trying to work backwards, as a specific example, $f(x) = \arctan(x)$ came to my mind. It's derivative certainly goes to $0$ as $x \to \infty$. Doesn't this show that $\frac{f(x + h) - f(x)}{h}$ converges uniformly to $0$ as $x \to \infty$ ?

I'm totally confused! I would really appreciate if anyone told me what I am doing wrong.

  • $\begingroup$ to interchange the limits you need the uniform convergence and I see that you are interchanging to prove the uniform convergence... I am not clear on your proof $\endgroup$ – El Moro Apr 29 '11 at 17:37
  • $\begingroup$ @El I couldn't see how I could show uniform convergence. So I tried to work backwards using a specific example to see if uniform convergence would come up. $\endgroup$ – Balaji Rao Apr 29 '11 at 17:49

This is in response to Balaji's comment.

Lemma. Let $f\colon (a,+\infty)\to\mathbb{R}$ be a bounded $C^1$ function such that $\lim_{x\to+\infty}f'(x)=\delta$ exists. Then $\delta=0$.

Proof. We proceed by contradiction and assume that $\delta\ne0$. Suppose first that $\delta>0$. Then there exists $x_0>a$ such that $f'(x)\ge\delta/2$ for all $x\ge x_0$. Then $$ f(x)=f(x_0)+\int_{x_0}^xf'(t)dt\ge f(x_0)+\frac{\delta}{2}(x-x_0)\quad\forall x\ge x_0. $$ In particular $f$ is unbounded. The case $\delta<0$ is done similarly.

If you do not want to use integrals, you may reason as follows. Let $$g(x)=f(x)-f(x_0)-\frac{\delta}{2}(x-x_0).$$ Then $g(x_0)=0$ and $g'(x)\ge 0$ for all $x>x_0$. It follows that $g$ is increasing in $(x_0,+\infty)$ and $g(x)\ge 0$ for all $x>x_0$.


Nothing. you can take something like: $x\mapsto \frac{\sin(x^2)}{x}+1$.

  • 13
    $\begingroup$ This is a correct example; however, perhaps it is not so obvious how he arrived at this result. The idea is that, as $x$ becomes larger and larger, you can force the function to be with a very small band around $y=1$, but you still have the ability to make the function jump up and down from $1-\varepsilon$ to $1+\varepsilon$, so that amplitude of the oscillations become smaller and smaller, so that the function converges (in this case to $1$), but the oscillation never goes away, so that the derivative does not converge. Hope this helps with the intuition here! $\endgroup$ – Jonathan Gleason Apr 29 '11 at 17:30
  • 4
    $\begingroup$ The counter-example can be intuitively found by thinking of a sinusoid that, as x increases, decreases its amplitud as well as its period -in the same amount, say: f(x) = g(x) sin(x / g(x)) with g(x) decreasing to zero, in particular: g(x) = 1/x. Then, f(x) tends to zero, but (because it is always a scaled down version of itself) its derivative keeps oscilating. The $\endgroup$ – leonbloy Apr 29 '11 at 17:47
  • $\begingroup$ @GleasSpty, @leonbloy, Thanks! It seems simple in retrospect. But I got so tied up with trying to use uniform convergence that I couldn't see anything else. $\endgroup$ – Balaji Rao Apr 29 '11 at 17:54
  • 7
    $\begingroup$ It should be mentioned that if $\lim_{x\to\infty}f'(x)$ exists, then the limit must be $0$. $\endgroup$ – Julián Aguirre Apr 29 '11 at 22:38
  • $\begingroup$ @Julián Thanks! After seeing the answers above I had reached this conclusion, but was waiting for someone to say it. But how can one show it ? If you can make an answer out of it, I'll be very glad. $\endgroup$ – Balaji Rao Apr 30 '11 at 3:21

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.