I have big doubts whether what I did was legal or not.

Investigate for convergence: $$\int_{0}^{\infty} \frac{\sin^{10}x \ln x}{\sqrt{x}} dx$$

First evaluate: $$\int_{0}^{\epsilon} ( \ln x) dx<\int_{0}^{\epsilon} (x^{10} \ln x) dx<\int_{0}^{\epsilon} \frac{x^{10} \ln x}{\sqrt{x}} dx<\int_{0}^{\infty} \frac{\sin^{10}x \ln x}{\sqrt{x}} dx$$ Work with the smallest one: $$\int_{0}^{\epsilon} ( \ln x)dx=\int_{-\infty}^{\ln\epsilon} \ln e^u de^u= \int_{-\infty}^{\ln\epsilon} u de^u$$ $$\int ude^u=ue^u-\int e^udu=ue^u-e^u=e^u(u-1)$$ $$ \int_{-\infty}^{\ln\epsilon} u de^u=\epsilon\cdot(\ln \epsilon-1)-e^{A}\cdot(A-1) ,$$ where $A\to-\infty$. The first addendum is not interesting, since it's a finite number. In the second addendum, we have indeterminacy of type $0\cdot\infty$. If we use $e^x=1+x+\cdots$, we will see it tends to $-\infty$. So the original integral is divergent.

  • 1
    $\begingroup$ The result is correct but your first inequalities are wrong $\endgroup$ – Maximilian Janisch May 22 at 13:04
  • $\begingroup$ @MaximilianJanisch Thank you, could you please explain why (or maybe how to do the exercise differently)? I tried to see what's wrong, but couldn't find the mistake. (Epsilon is close to zero.) $\endgroup$ – fragileradius May 22 at 13:08

TLDR: Your solution is not correct, but your result is correct nonetheless.

Correct proof:

Claim. Your integral diverges to $+\infty$.

Proof. Note that $\sin^{10}(x)\geq0$ for all $x\in\mathbb R$ and that $\sin^{10}(x)>c$ for all $x\in\bigcup_{n\in\Bbb N} [\frac\pi4+2n,\frac{3\pi}4+2n]$ (where $c>0$ is some constant). Thus, by "$\sigma$-additivity" of the integral, we have (note that $\frac{\ln x}{\sqrt x}$ is positive for all $x\geq 1$) \begin{equation}\label1\tag1 \int_{1}^{\infty} \frac{\sin^{10}x \ln x}{\sqrt{x}} \,\mathrm dx \geq c \sum_{n=1}^\infty\int_{\frac\pi4+2n}^{\frac{3\pi}4+2n} \frac{\ln x}{\sqrt x} \,\mathrm dx \geq c\sum_{n=1}^\infty\int_{\frac\pi4+2n}^{\frac{3\pi}4+2n} \frac1x\,\mathrm dx. \end{equation}

The right-hand side of \eqref{1} equals $c\sum_{n=1}^\infty \ln(\frac{3\pi}4+2n)-\ln(\frac\pi4+2n)$. If we can show that the sum of the last expression diverges, we have thus shown that the integral diverges. Let me thus show that the sum diverges:

We have by this question \begin{equation} \lim_{n\to\infty} \frac{\ln(\frac34 \pi + 2n)-\ln(\frac\pi4+2n)}{\ln(2n+2)-\ln(2n)} = \frac\pi4 > 0. \end{equation} We also have \begin{equation}\sum_{n=1}^\infty \ln(2n+2)-\ln(2n) = \lim_{n\to\infty}\ln(2n)=\infty.\end{equation}

Hence, by the comparison test, the sum that should diverge actually does diverge.

This achieves a proof of the claim. $\square$

Your mistakes:

  • Proving that $\int_{0}^{\epsilon} ( \ln x) \,\mathrm dx = -\infty$ does not show that our integral diverges (you would have to bound it the other way around, i.e. you would need "our integral < $\int_{0}^{\epsilon} ( \ln x) \,\mathrm dx$" for that to be correct.)
  • The above fact is actually wrong (see also the answer by mihaild). We know that $\int_{0}^{\epsilon} ( \ln x) \,\mathrm dx = \epsilon\ln(\epsilon)-\epsilon-\lim_{x\to 0}(x\ln(x)-x)$. By writing the last limit as $\frac{\ln(x)}{\frac1x}$, we find that your integral is actually a finite number for every $\epsilon$!
  • ${x^{10} \ln x}<\frac{x^{10} \ln x}{\sqrt{x}}$ is clearly wrong for $x<1$. In fact, the opposite is the case.
  • $\frac{x^{10} \ln x}{\sqrt{x}}<\frac{\sin^{10}x \ln x}{\sqrt{x}}$ is also wrong for small $x>0$.

$\int_{0}^{\epsilon} (x^{10} \ln x) dx<\int_{0}^{\epsilon} \frac{x^{10} \ln x}{\sqrt{x}} dx$ is wrong. Also $e^A \cdot (A - 1) \to 0$ as $A \to -\infty$. And $\int_{0}^\varepsilon \ln x\, dx = (x \ln x - x)\rvert_0^\varepsilon = \varepsilon(\ln \varepsilon - 1)$, so this integral converges.

To prove that your integral diverges, you can use that $\sin^{10} x > a$ for some positive $a$ if $x \in [2 \pi k + \frac{\pi}{4}, 2 \pi k + \frac{3\pi}{4}]$ for some integer $k$, $\frac{\sin^{10} x \ln x}{\sqrt{x}} > \frac{a}{\sqrt x}$ if $x > 2\pi$ and so $\int\limits_{2\pi}^{2\pi(m + 1)} \frac{\sin^{10} x \ln x}{\sqrt{x}}\, dx > \sum\limits_{k = 2}^{m} \int\limits_{2 \pi k + \frac{\pi}{4}}^{2 \pi k + \frac{3\pi}{4}}\frac{1}{\sqrt x}\, dx > \sum\limits_{k = 2}^{m} \frac{1}{\sqrt{2\pi k + \frac{3\pi}{4}}}$

As this series diverges (by comparsion with harmonic series, for example - $\frac{1}{\sqrt{2 \pi k + \frac{3 \pi}{4}}} < \frac{1}{k}$ for large enough $k$) - so does your integral.

  • $\begingroup$ It's essentially the same as @Maximilian Janisch's solution, but with a bit tighter lower bound for integration term and a bit simpler proof for divergence (it's enough to estimate integral over segment by lower bound of function, it's not necessary to calculate it exactly). $\endgroup$ – mihaild May 22 at 18:02
  • $\begingroup$ This is a bit smarter than my solution basically (+1) 😃 $\endgroup$ – Maximilian Janisch May 22 at 18:18

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.