I would really appreciate feedback/guidance. This is from a past year exam and my professor didn't cover much on improper integrals. I'm trying to prove that $\int_0^{\infty} \frac{\sin x}{x^p}\, dx$ converges for $0<p<2$ using the comparison test for improper integrals. I know that from the test, if $|f(x)| \leq g(x) \forall x\geq a$ then $\int_a^{\infty} g(x)\, dx$ converges.

EDIT: Here's my second attempt at the proof after much discussion in the comments below.

We know that $\frac{\sin x}{x^p}$ will be continuous on 1 because $\lim_{x\rightarrow 1} \frac{\sin x}{x^p} = lim_{x\rightarrow 1} x^{-p} \sin(x) = 1^{-p}\sin(1) = \sin(1)$.

However, for 0 we have that:

$\lim_{x \rightarrow 0} \frac{\sin x}{x^p} = \frac{\sin 0} {0^{p}} = \frac{0}{0^{-p}} = \frac{1}{0^{p-1}}$. Since we can't have 0 in the denominator, we only have continuity on 0 if p-1 < 1, which implies p<2.

And so, by a theorem, we know that $\frac{\sin(x)}{x^p}$ is Riemann integrable on [0,1] for p<2. Furthermore, by the improper limit comparison test, if we prove that $\int_1^{\infty} f(x)\, dx$, then we can show that $\int_0^{\infty} f(x)\, dx$ converges. Hence we have that

$\lim_{b \rightarrow \infty} \int_1^{b} \frac{\sin(x)}{x^p}\,dx = \lim_{b \rightarrow \infty} \int_1^{b} \frac{\sin(x)}{x^p}\,dx =\lim_{b\rightarrow \infty} -\frac{cos x}{{p}} |_p^{b} - \int_1^{b} \frac{\cos x}{(p-1)x^{p+1}}\,dx$

And so I know that since $\int \frac{cos x}{x^{p+1}} < \int \frac{1}{x^{p+1}}$, and since the RHS will converge if and only if p+1 > 1, we have that p>0.

And so, we conclude that the integral converges for $0<p<2$.

  • 1
    $\begingroup$ Your integration by parts is wrong. $$\int_1^b \frac{\sin x}{x^p}\,dx = \frac{-\cos x}{x^p}\biggr\rvert_1^b - p\int_1^b \frac{\cos x}{x^{p+1}}\,dx$$ $\endgroup$ – Daniel Fischer Dec 18 '16 at 23:32
  • $\begingroup$ I don't see how you can ignore $x=0$ like this. In fact, about half of the difficulty in showing convergence is due to the blowup of $\sin(x)/x^p$ at $x=0$ for $p>1$. Certainly there are many functions such that $\int_1^\infty f <\infty$ but $\int_0^\infty f = \infty$; say, $f(x) = 1/x^2$. $\endgroup$ – Gyu Eun Lee Dec 18 '16 at 23:35
  • $\begingroup$ @Gyu Eun Lee Hi there! I apologize, I didn't intend to ignore x=0. It slipped my mind that we would have an issue at x=0. Thanks ofr pointing it out! $\endgroup$ – Nikitau Dec 18 '16 at 23:40
  • $\begingroup$ @DanielFischer Ah, that was careless of me. Thank you! $\endgroup$ – Nikitau Dec 18 '16 at 23:42

Integrate by parts with $u=x^{-p}$ and $v=-\cos(x)$. Then, we can write

$$\int_1^L \frac{\sin(x)}{x^p}\,dx=\left.\left(-\frac{\cos(x)}{x^p}\right)\right|_{1}^{L}-p\int_1^L \frac{\cos(x)}{x^{p+1}}\,dx$$

  • $\begingroup$ Thanks! It seems like I got my integration by parts completely mixed up. Judging from what we get from integration by parts, we can write that $\int_1^{L} \frac{cos(x)}{x^{p+1}} < \int_1^{L} \frac{1}{x^{p+1}}$, and we know that the RHS converges only if p+1 > 1, which implies p>0. But how do we show that we also need p<2? $\endgroup$ – Nikitau Dec 18 '16 at 23:44
  • $\begingroup$ @Nikitau That constraint comes from the behaviour at $0$. $\endgroup$ – Daniel Fischer Dec 18 '16 at 23:44
  • $\begingroup$ @DanielFischer This is probably a dumb question, but is it the behaviour of x at zero? Oh then I guess what I wrote earlier in my original post didn't make sense, where I said if $\int_1^{\infty} f(x) dx$ converges then $\int_0^{\infty} f(x) dx$ converges? $\endgroup$ – Nikitau Dec 18 '16 at 23:48
  • $\begingroup$ @Nikitau It makes sense, but it's only true if the behaviour on $[0,1]$ is good enough. Look at $\frac{\sin x}{x^p}$ as $x \to 0^+$. For what $p$ is that bounded? For what $p$ does it tend to $+\infty$? How fast? How fast may it grow and still remain (improperly) integrable? $\endgroup$ – Daniel Fischer Dec 18 '16 at 23:52
  • $\begingroup$ @DanielFischer Hmm. I'm guessing we would get $0^{1-p}$ as x goes to $0^+$? Since we can't have 0 in a denominator, we would need 1-p>0? and so 1>p? And $\frac{sin x}{x}$ goes to infinity for all values of p>0? $\endgroup$ – Nikitau Dec 18 '16 at 23:57

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.