5

Prove that if $\lim\limits_{x\to\infty}f(x)$ and $\lim\limits_{x\to\infty}f''(x)$ exist, then $\lim\limits_{x\to\infty}f'(x)=0$.

I can prove that $\lim\limits_{x\to\infty}f''(x)=0$. Otherwise $f'(x)$ goes to infinity and $f(x)$ goes to infinity, contradicting the fact that $\lim\limits_{x\to\infty}f(x)$ exists. I can also prove that if $\lim\limits_{x\to\infty}f'(x)$ exists, it must be 0. So it remains to prove that $\lim\limits_{x\to\infty}f'(x)$ exists. I'm stuck at this point.

  • 1
    A related question (which 723842934 mentions he or she has already solved) was asked recently: http://math.stackexchange.com/questions/78810/prove-lim-limits-x-fx-0 – Jonas Meyer Nov 07 '11 at 04:40
  • These questions are two parts of a single question in Spivak's text. –  Nov 07 '11 at 04:44

2 Answers2

3

This is similar to a recent Putnam problem, actually. By Taylor's theorem with error term, we know that for any $x$, $$ f(x+1) = f(x) + f'(x) + \tfrac12f''(t) $$ for some $x\le t\le x+1$. Solve for $f'(x)$ and take limits....

Greg Martin
  • 78,820
  • Nice! But there should be a solution without Taylor's theorem, because it has not been introduced in this part of the book. –  Nov 07 '11 at 06:41
2

Hint $\ $ This follows easily from L'Hôpital's rule since

$$\rm \lim_{x\to\infty}\ f-f'\ =\ \lim_{x\to\infty}\frac{e^x\ (f-f')}{e^x}\ =\ \lim_{x\to\infty}\frac{e^x\ (\:f-f'+f'-f'')}{e^x}\ =\ \lim_{x\to\infty}\ f-f''\ exists$$

See also the similar classic Hardy old problem.

Bill Dubuque
  • 272,048