33
$\begingroup$

I was wondering on how to evaluate the following indefinite integral for all $n\in\mathbb R$.

$$\int\frac1{1+x^n}dx$$

It seems to be peculiar in that we have

$$\begin{align} \int\frac1{1+x^{-1}}dx&=x-\ln(x+1)+c\\ \int\frac1{1+x^0}dx&=\frac12x+c\\ \int\frac1{1+x^{1/2}}dx&=2\sqrt x-2\ln(1+\sqrt x)+c\\ \int\frac1{1+x^1}dx&=\ln(x+1)+c\\ \int\frac1{1+x^2}dx&=\arctan(x)+c\\ \int\frac1{1+x^3}dx&=\frac13\ln(1+x)-\frac2{3\sqrt3}\arctan\left(\sqrt{\frac43}\left(x-\frac12\right)\right)+c \end{align}$$

Naturally, there appears to be some combination of $\ln$ and $\arctan$, though no simple formula arises to solve the general case.

It is, however, easy to see that

$$\int\frac1{1+x^{-n}}dx=\int1-\frac1{1+x^n}dx$$

So there is an easy enough connection between positive and negative $n$.

Also, it is easy enough to find the series expansion, taking advantage of the above connection we just made to circumvent problems concerning convergence.

$$\frac1{1+x^n}=1-x^n+x^{2n}-x^{3n}+\dots\forall\ |x|<1$$

$$\int\frac1{1+x^n}dx=c+x-\frac1{n+1}x^{n+1}+\frac1{2n+1}x^{2n+1}-\dots$$

$$=c+\sum_{k=0}^\infty\frac{(-1)^k}{kn+1}x^{kn+1}\ \forall\ |x|<1$$

Though this isn't very much along the lines of closed form.

For $n=\frac ab$, where $a$ and $b$ are whole numbers, we can use the substitution $x=u^b$ to get

$$\int\frac1{1+x^n}dx=\int\frac{bu^{b-1}}{1+u^a}du$$

though I'm unsure where that could lead. This reduces the integral down to

$$\int\frac1{1+x^n}dx=b\int P(u)+\frac{u^{b-1-ak}}{1+u^a}du,\quad k\in\mathbb N$$

for some polynomial $P(u)$. Though I'm still clueless as to how this can be advanced.

How can I evaluate $\int\frac1{1+x^n}dx\ \forall\ n\in\mathbb R$ in closed form? Can someone prove there at least exists some closed form solution for all $n\in\mathbb Q$ if the above is not possible? If possible, use real numbers.

$\endgroup$
  • 1
    $\begingroup$ I'm curious about whether anything can be done with $\frac 1{1+x^{-n}}+\frac 1{1+x^n}=1$, this seems to be a relatively powerful identity that I hadn't considered before... $\endgroup$ – abiessu Nov 5 '16 at 0:05
  • $\begingroup$ One way to connect all of these integrals is with the hypergeometric function. In particular, one may compute that $$ \int \frac{1}{1 + x^n}\,dx = x \cdot {}_2F_1(1,1/n;1 + 1/n; -x^n) + C $$ However, since the hypergeometric function is defined in terms of its power series, this is not a terribly satisfying epiphany. $\endgroup$ – Omnomnomnom Nov 5 '16 at 0:06
  • $\begingroup$ I have a 150 page table of mathematical formulas in my bookshelf where 20 of those are devoted to indeterminate integrals. It lists this integral for $n = 1, 2, 3$ and $4$ separately, and that's it for your integrals (although the integrals for $1/(1-x^n)$ are also listed for the same $n$). I think that would suggest that no nice closed form was known at the time of printing, which was 2003. $\endgroup$ – Arthur Nov 5 '16 at 0:10
  • $\begingroup$ I was looking at the same problem a while ago and posted a solution. See math.stackexchange.com/questions/1837784/… $\endgroup$ – 926reals Nov 5 '16 at 1:08
  • 1
    $\begingroup$ These two questions are also related (by not the same, since they do not request solution as combination of arctan and log): Solving this integral $\int\frac{1}{1+x^n} dx$? and How to find the indefinite integral $\int \frac{dx}{1+x^{n}}$? $\endgroup$ – Martin Sleziak Jan 14 '17 at 9:16
38
+100
$\begingroup$

In THIS ANSWER, I showed that that the indefinite integral of interest is given by

$$\bbox[5px,border:2px solid #C0A000]{\int\frac{1}{x^n+1}dx=-\frac1n\sum_{k=1}^n\left(\frac12 x_{kr}\log(x^2-2x_{kr}x+1)-x_{ki}\arctan\left(\frac{x-x_{kr}}{x_{ki}}\right)\right)+C} $$

for $n\ge 1$, where $x_{kr}$ and $x_{ki}$ can be written, respectively, as

$$x_{kr}=\cos \left(\frac{(2k-1)\pi}{n}\right)$$

$$x_{ki}=\sin \left(\frac{(2k-1)\pi}{n}\right)$$


For $n<0$, we simply write

$$\int \frac{1}{1+x^{-|n|}}\,dx=x-\int \frac{1}{1+x^{|n|}}\,dx$$

and use the aforementioned result with $n$ replaced with $|n|$.

$\endgroup$
  • 2
    $\begingroup$ Wow, that is quite beautiful. Mind me asking if you can do anything for $n\in\mathbb Q$ or $\mathbb R$? $\endgroup$ – Simply Beautiful Art Nov 5 '16 at 13:03
  • $\begingroup$ @SimpleArt Thank you for the nice comment! Much appreciated. Now, unfortunately, I don't know how to generalize this result to the case for non-integer values of $n$ without appealing to the hypergeometric function. But that is unlikely the goal here. -Mark $\endgroup$ – Mark Viola Nov 5 '16 at 15:05
  • $\begingroup$ @marcocantarini Thanks Marco. Much appreciative. $\endgroup$ – Mark Viola Nov 15 '16 at 14:19
  • $\begingroup$ Would you be so kind as to look at $\displaystyle\int\frac{x^b}{1+x^a}dx$ for $a,b\in\mathbb N$? $\endgroup$ – Simply Beautiful Art Nov 19 '16 at 0:23
  • 2
    $\begingroup$ Very nice! How in the world could I have missed this precious answer to this question? +1 $\endgroup$ – DonAntonio Jan 13 '17 at 21:25
10
$\begingroup$

For positive integers $n$ you can write $$ \dfrac{1}{1+x^n} = \sum_{\omega} \dfrac{r(\omega)}{x - \omega}$$ where the sum is over the $n$'th roots of $-1$ and $r(\omega)$ is the residue of $1/(1+x^n)$ at $x = \omega$, so that your integral is $$ c + \sum_\omega r(\omega) \log(x - \omega)$$

You can also express the power series solution in terms of the Lerch Phi function: $$ c + \dfrac{x}{n} {\rm LerchPhi}(-x^n,1,1/n) $$

$\endgroup$
  • $\begingroup$ Ah, interesting idea! Basically equivalent to partial fraction decomposition. Though it suggests I should use complex logarithms to write my $\arctan$. If possible, keeping everything under real numbers is recommended. $\endgroup$ – Simply Beautiful Art Nov 5 '16 at 0:18
  • $\begingroup$ Robert, SEE THIS, in which I developed a closed-form solution. ;-)) $\endgroup$ – Mark Viola Nov 5 '16 at 2:12
0
$\begingroup$

Slightly different approach to Robert Israel's. Using $$\frac{1}{1+x}=1-x+x^2-x^3+x^4-\cdots,\tag{1}$$ where $|x|<1$ you can substitute $x^n$ for $x$ to obtain $$\int\frac{1}{1+x^n}dx=\int 1-x^n+x^{2n}-x^{3n}+x^{4n}-\cdots dx$$

So we end up with

$$\int\sum_{k=0}^\infty (-1)^kx^{kn}dx=\sum_{k=0}^\infty(-1)^k\frac{x^{kn+1}}{kn+1}+c=\frac{x}{n} \Phi \left(-x^n,1,\frac{1}{n}\right)+c.$$ This actually works for $x,n\in\mathbb{R}$, but you'd have to prove that for $x$ since it does not follow from (1).

$\endgroup$

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.