In an old exercise sheet I have found the following question:
For which values $p\in \mathbb{R_{\geq 0}}$ does the following integral converge
$$\int_0^\infty x\sin(x^p) dx $$
In order to solve it, the paper suggests one should use the substitution $u = x^p$ (for $p > 0$) and use the Leibnitz-Criterium.
Applying this trick gives me $\int_0 ^\infty\frac{u^{\frac{1-p}{p}}}{p} \sin(u) u ^{\frac{1}{p}} $. ($\rightarrow$ using the substitution from above)
How does one proceed from here?
Another idea I have had was to use the taylor polynomial for $\sin(x) = x- \frac{x^2}{2!} \cdots$ and then approximate the integral from above and below $\rightarrow$ this however, seems to get very messy.
I am open for other ideas for solving this exercise.
You can determine what the answer is non-rigorously as follows. First of all:
$$I:=\int_0^\infty x \sin(x^p) dx = \sum_{k=1}^\infty \int_{r_{k-1}}^{r_k} x \sin(x^p) dx$$
where $r_k=(k\pi)^{1/p}$. This is of course rigorous (if we interpret a divergent integral and a divergent sum as "the same"). The point of doing this decomposition is to take each integrand to be of one sign.
Now the non-rigorous part starts. Think of the sine as if it just contributes a sign and a $\Theta(1)$ positive constant factor to the integral in between each pair of roots of $\sin(x^p)$. Similarly, think of the $x$ factor as not changing much between each pair of roots. This means that we are claiming
$$I=\sum_k (-1)^k a_k r_k (r_{k+1}-r_k)$$
for some positive numbers $a_k$. Now we deal with the difference:
$$(k+1)^{1/p}-k^{1/p}=k^{1/p}((1+1/k)^{1/p}-1) =b_k k^{1/p-1}$$
where $b_k>0,b_k=\Theta(1)$. So with $c_k=a_k b_k$ you claim
$$I=\sum_k c_k (-1)^k k^{2/p-1}.$$
If the $c_k$ doesn't matter for the convergence, then you get convergence iff $p>2$.
Cleaning this up to make it rigorous is probably more trouble than it's worth; you'd be better off following the hint or using a different technique entirely. Nevertheless this heuristic gives the correct answer.