# If $\lim_{x \to \infty} f(x) - xf'(x)$ exists, does $\lim_{x \to\infty} f'(x)$ exist as well?

Let $$f(x)$$ be a differentiable function on $$(0, \infty)$$ with $$\lim_{x\to \infty} f(x) - xf'(x) = L\in \mathbb{R}$$. I'm trying to prove or disprove that $$\lim_{x\to\infty} f'(x)$$ exists as well. Here's what I have so far: if the limit does exist, it must also equal $$\lim_{x\to\infty} \frac{f(x)}{x}$$ (divide the limit condition by $$x$$). Then I rewrote the condition as $$x^2 \frac{d}{dx} \frac{f(x)}{x} \to L$$, and from this I expect (using some $$1/x^2$$ asymptotic argument) my statement to be true - but I'm not sure how to rigorously proceed with this argument.. Can someone provide some next steps, or a counter example?

• Try the case, whenever $\frac{f(x)}{x}=f'(x)$ , but $lim_{x\to\infty} \frac{f(x)}{x}$ doesn't exist! Is there any such case???? – A learner Aug 1 at 4:30
• @FarazMasroor what do you precisely mean when you say $\lim_{x\to \infty} f'(x)$ exists,do you mean exists in $\mathbb R$ of exists in $\mathbb R\cup \{\pm \infty\}$ ?Please make it clear? – Kishalay Sarkar Aug 1 at 5:39

Yes, $$\lim_{x \to\infty} f'(x)$$ exists under those conditions. In fact it suffices to require that the function $$r(x) = f(x) - xf'(x)$$ is bounded for $$x \to \infty$$, i.e. that $$M = \sup \{ |r(x)| : x \ge 1 \} < \infty \, .$$

Proof: $$f'(x) = \frac{f(x)}{x} - \frac{r(x)}{x}$$ so that the claim is equivalent to the existence of $$\lim_{x \to\infty} \frac{f(x)}{x}$$. Here we can apply the Cauchy criterion to $$F(x) = \frac{f(x)}{x} \, .$$ For $$x \ge 1$$ is $$|F'(x)| = \left|\frac{xf'(x) - f(x)}{x^2}\right| = \frac{|r(x)|}{x^2} \le \frac{M}{x^2} \, .$$ It follows that for $$1 \le x < y$$ $$|F(x) - F(y)| \le \int_x^y \frac{M}{t^2} \, dt = M \left( \frac 1x - \frac 1y \right) < \frac Mx \, .$$ (This is a straight-forward estimate if the fundamental theorem of calculus can be applied to $$F$$, e.g. if $$F'$$ is continuous or Riemann integrable. For the general case see $|f'(x)| \le g(x)$ implies $|f(b) - f(a)| \le \int_a^b g(x) dx$, without assuming $f'$ to be integrable.).

So for every $$\epsilon > 0$$ $$x, y > \frac{M}{\epsilon} \implies |F(x) - F(y) | < \epsilon$$ and that completes the proof.

• Nice proof indeed! I am wondering if one can replace integrals with the use of mean value theorem. – Paramanand Singh Aug 1 at 10:35
• @ParamanandSingh: I have added a proof using only the MVT which does not require the derivative to be continuous (or integrable). – Martin R Aug 1 at 15:00
• Well, I had already given my +1, can't do it again. – Paramanand Singh Aug 1 at 15:06
• Your proof via MVT is similar to the proof of Fundamental Theorem of Calculus (evaluation of integral via anti-derivative) and should be used as an example to illustrate its power. – Paramanand Singh Aug 1 at 15:10
• @ParamanandSingh: But the FTC requires the derivative to be integrable. What I needed here is essentially that $|F'(x)| \le h(x)$ implies $|F(y) - F(x)| \le \int_x^y h(t) dt$, and apparently that holds without additional assumptions on $F'$. I wonder if there is a simpler argument, without considering a sequence of partitions. – Perhaps I'll post a question about this. – Martin R Aug 1 at 15:18

Here is a path to a proof. Denote $$f(x) = x\cdot g(x)$$ Then equivalently what we are asking is that if the limit

$$\lim_{x\to\infty}-x^2g'(x) = L$$

exists, then does the limit

$$\lim_{x\to\infty}g(x)+xg'(x)$$

exist? Further we have by squeeze theorem that

$$\lim_{x\to\infty} xg'(x) = 0$$

so we can simplify this to the question does the limit

$$\lim_{x\to\infty}g(x)$$

exist given the condition above?