0
$\begingroup$

Find the set of values of $a$ such that

$$\displaystyle\int_{0}^{\infty}\frac{t^a}{1+t}\mathrm dt$$

converges. I have managed to prove that the integral diverges for $a>0$.

  • 0
    You want that $$\int_0t^adt$$ and $$\int^\infty t^{a-1}dt$$ both converge. The former does iff $a>-1$ while the latter does iff $a-1<-1$, hence...2017-02-27
  • 0
    Converges for $-1 $\int_0^\infty t^a /(1+t) = \pi/\sin(\pi a)$. The integral is also sometimes known as the mellin transform of $1/(1+t)$, (with $a = -a$). – 2017-02-27
  • 0
    @N3buchadnezzar Your comment is wrong (sorry).2017-02-27
  • 0
    @Did, fixed I think. I tried to give some broader references to the problem. I am so used to seeing this integral with a negative exponent I forgot to flip the signs =)2017-02-27
  • 0
    @N3buchadnezzar I made similar (but stupider) errors in my answer. I think problems like this really sift out backgrounds: analysts get it right away, [analytic] number theorists get the infinity parts, but royally mess up near $0$, students use the neat tricks nearest them in their temporal education, and algebraists just throw their hands up and study more groups. :)2017-02-27

1 Answers 1

0

Let $-b=a<0$. Then note

$$0\le \int_1^x{dt\over t^b+t^{1+b}}\,dt\le \int_1^x{dt\over t^{1+b}}\le {1\over b}$$

So the monotone sequence $\displaystyle\int_1^n{t^a\over 1+t}$ is bounded, hence converges, and of course this is exactly $\displaystyle\lim_{x\to\infty}\int_1^x{t^a\over 1+t}$, so this part of the improper integral converges. Now we turn to the other part to see if the whole thing converges or not:

$$\int_\epsilon^1{t^a\over 1+t}\,dt\ge {1\over 2}\int_\epsilon^1t^a\,dt={1\over 2}\left({1\over {a+1}}-{\epsilon^{a+1}\over a+1}\right)$$

so when $a<-1$ we get divergence, since then $a+1<0$ and we are dividing by something going to $0$.

For $-1< a < 0$ we see

$$\int_{\epsilon}^1{t^a\over 1+t}\,dt \le \int_{\epsilon}^1 t^a = {1\over a+1}-{\epsilon^{a+1}\over a+1}$$

which is finite. So for certain we get convergence for $-1

Finally the edge cases:

For $a=0$ we see

$$\log x=\int_1^x {dt\over t}<\int_0^x {dt\over 1+t}$$

so again taking limits we get divergence for $a=0$, without even having to consider the other integral, $\displaystyle\int_0^1{dt\over 1+t}$ we would need to converge to conclude convergence of the whole integral!

And for $a=-1$ we consider

$$\int_{\epsilon}^1{dt\over t+t^2}>\int_{\epsilon}^1{dt\over 2t}=-{1\over 2}\log\epsilon$$

which goes to infinity as $\epsilon\to 0$, showing divergence.

  • 1
    ?? $\int_0^x\frac{dt}{t^{1+b}}$ diverges for every positive $x$ and $b$.2017-02-27
  • 0
    It seems like we need to make special considerations near $t = 0$ as well. If $a\le -1$ then near $t = 0$ we have $$\frac{t^a}{1+t} \sim t^a$$ and for such $a$, the integral will diverge.2017-02-27
  • 0
    @Did fixed, thanks again!2017-02-27
  • 0
    @User8128 yeah, Did made a great point on that, I did the necessary edits, take a look at how the story changes.2017-02-27
  • 0
    The part $\int_\epsilon^1$ is still quite wrong. Please **think** before posting.2017-02-27
  • 0
    @Did I'm too casual on basic posts to probably ever achieve all the details every time, but thanks again for the catch. :)2017-02-27
  • 0
    I agree with @Did. This post still needs to be fixed. Introducing $b$ seems like a bad choice here as it just muddies that waters. You need both $$\int_1^\infty \frac{1}{t^{1-a}} dt, \,\,\,\,\, \text{ and } \,\,\,\,\, \int_0^1 t^a dt $$ to converge. The former requires that $a < 0$, the latter that $a > -1$ so it converges when $-1 < a < 0$.2017-02-27
  • 0
    @User8128 I introduce $b$ to end up with something positive, I fixed the issue with the integral, the term $-{2\over b}$ was missing.2017-02-27
  • 0
    Again quite wrong. Downvoted.2017-02-27
  • 0
    The integral $\int_0^1{t^a\over 1+t}\,dt$ does not diverge for every $a<0$ and the lower bound you propose is wrong. (You know what, we should not have this exchange at all about such a beginner's exercise...)2017-02-27
  • 0
    @Did Thanks! I figured out what you were referring to before: I thought your issue was with forgetting the bottom end of the integral, but it seems I correctly was able to decipher your comment the second go-around. In particular you are quite right, it was quite a silly mistake to make.2017-02-27