Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I would like to find an equivalent of:

$$ u_n = \int_0^1 (\arctan x)^n \mathrm dx$$

which might be: $$ u_n \sim \frac{\pi}{2n} \left(\frac{\pi}{4} \right)^n$$

$$ 0\le u_n\le \left( \frac{\pi}{4} \right)^n$$

So $$ u_n \rightarrow 0$$

In order to get rid of $\arctan x$ I used the substitution $$x=\tan \left(\frac{\pi t}{4n} \right) $$

which gives:

$$ u_n= \left(\frac{\pi}{4n} \right)^{n+1} \int_0^n t^n\left(1+\tan\left(\frac{\pi t}{4n} \right)^2 \right) \mathrm dt$$

But this integral is not easier to study!

Or: $$ t=(\arctan x)^n $$

$$ u_n = \frac{1}{n} \int_0^{(\pi/4)^n} t^{1/n}(1+\tan( t^{1/n})^2 ) \mathrm dt $$

How could I deal with this one?

share|cite|improve this question
up vote 2 down vote accepted

$$ \begin{align} &\int_0^1\arctan^n(x)\,\mathrm{d}x\\ &=\int_0^{\pi/4}x^n\sec^2(x)\,\mathrm{d}x\\ &=\int_0^{\pi/4}\left(\frac\pi4-x\right)^n\sec^2\left(\frac\pi4-x\right)\,\mathrm{d}x\\ &=\frac{(\pi/4)^{n+1}}{n}\int_0^ne^{-x}\color{#C00000}{e^x\left(1-\frac{x}{n}\right)^n}\color{#00A000}{\sec^2\left(\frac\pi4-\frac{\pi x}{4n}\right)}\,\mathrm{d}x\\ &=\frac{(\pi/4)^{n+1}}{n}\int_0^ne^{-x}\color{#C00000}{\small\left(1-\frac{x^2}{2n}-\frac{8x^3-3x^4}{24n^2}+O\left(\frac1{n^3}\right)\right)}\color{#00A000}{\small\left(2-\frac{\pi x}{n}+\frac{\pi^2x^2}{2n^2}+O\left(\frac1{n^3}\right)\right)}\,\mathrm{d}x\\ &=\frac{(\pi/4)^{n+1}}{n}{\small\left(2-\frac{\pi+2}{n}+\frac{2+3\pi+\pi^2}{n^2}+O\left(\frac1{n^3}\right)\right)} \end{align} $$

share|cite|improve this answer
Thank you very much robjohn for this much more precise asymptotic expansion! It took me some time to get your result! – Chon Jan 18 '13 at 17:10

To get a rough asymptotic answer, you can substitute $$x=1-\frac{y}{n}.$$ Then $$ \frac{u_n}{(1/n) (\pi/4)^n}=\int_0^n \left(\frac{\arctan (1-\frac{y}{n})}{\pi/4}\right)^n \, dy. $$ For fixed $y$, as $n\to\infty$,
$$ \frac{\arctan (1-\frac{y}{n})}{\pi/4}=1-\frac{2}{\pi n} y + O(\frac{1}{n^2})= \exp(-\frac{2y}{\pi n} + O(\frac{1}{n^2})), $$ so, pointwise, the integrand converges to $e^{-2y/\pi}$. Since we have the bound $$ \arctan (1-\frac{y}{n})\le \frac{\pi}{4} e^{-2y/\pi n}, \qquad 0\le y\le n, \ \ \ (*) $$ we can apply the dominated convergence theorem to conclude that $$ u_n \sim \frac{1}{n}(\frac{\pi}{4})^n \int_0^\infty e^{-2y/\pi} \, dy=\frac{\pi}{2n} (\frac{\pi}{4})^n. $$ Also, from the integral and (*), $$ 0\le u_n\le \frac{\pi}{2n}(\frac{\pi}{4})^n, \qquad \rm for\ all\ \it n. $$

share|cite|improve this answer
How did you get: $ \arctan (1-\frac{y}{n})\le \frac{\pi}{4} e^{-2y/\pi n} $? The RHS $e^{-2y/\pi n} $ is integrable but should not depend on $n$. – Chon Jan 18 '13 at 18:15
The left- and right-hand sides are both $\pi/4$ at $y/n=0$. Then, take derivatives with respect to $y/n$. The derivative of the LHS is bounded above by $-1/2$, and the derivative of the RHS is bounded below by $-1/2$. – David Moews Jan 18 '13 at 21:55

From your simplification, $$ u_n = \left(\frac{\pi}{4n}\right)^{n+1} \int_0^n t^n \left(1 + \tan^2\left(\frac{\pi t}{4n}\right) \right) dt $$ note that the integrand is non-negative and in fact on $[0,n]$, $0 \leq \tan(\pi t/4n) \leq 1$. Applying this we get $$ \left(\frac{\pi}{4n}\right)^{n+1} \int_0^n t^n dt \leq u_n \leq \left(\frac{\pi}{4n}\right)^{n+1} \int_0^n 2t^n dt $$ which gives $$ \left(\frac{\pi}{4n}\right)^n \frac{\pi}{4(n+1)} \leq u_n \leq \left(\frac{\pi}{4n}\right)^n \frac{\pi}{2(n+1)} $$ Hope that helps.

share|cite|improve this answer
$\int_0^nt^n\,\mathrm{d}t=\frac{n^{n+1}}{n+1}$ – robjohn Jan 18 '13 at 7:23
$ \left(\frac{\pi}{4}\right)^n \frac{\pi}{4(n+1)} \leq u_n \leq \left(\frac{\pi}{4}\right)^n \frac{\pi}{2(n+1)} $. $ \left(\frac{\pi}{4}\right)^n \frac{\pi}{2(n+1)} \sim \frac{\pi}{2n} \left(\frac{\pi}{4} \right)^n $ but $ \left(\frac{\pi}{4}\right)^n \frac{\pi}{4(n+1)} \sim \frac{\pi}{4n} \left(\frac{\pi}{4} \right)^n $ so we don't get the result. – Chon Jan 18 '13 at 18:22

Another (simpler) approach is to substitute $x = \tan{y}$ and get

$$u_n = \int_0^{\frac{\pi}{4}} dy \: y^n \, \sec^2{y}$$

Now we perform an analysis not unlike Laplace's Method: as $n \rightarrow \infty$, the contribution to the integral is dominated by the value of the integrand at $y=\pi/4$. We may then say that

$$u_n \sim \sec^2{\frac{\pi}{4}} \int_0^{\frac{\pi}{4}} dy \: y^n = \frac{2}{n+1} \left ( \frac{\pi}{4} \right )^{n+1} (n \rightarrow \infty) $$

The stated result follows.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.