$$f(x)=\lim_{n\to\infty}\biggl(n\int_0^{\pi/4}(\tan x)^n\,dx\biggr)$$ I try to this way, $\tan x\ge x$, when $x\in(0,\frac\pi4)$, but this turns out to be $\tan x\ge0$, which is obvious even without calculation. I think it can be solved by using the squeeze rule but I can't find a proper scaler to meet $A=g(x)\le f(x)\le h(x)=A$ when $n\to\infty$

$\color{red}{Edit}$: According to@Khallil's advice, I can solve that $f(n)+f(n+2)=\frac1{n+1}$, and it's trivial that $\frac1{n+1}=f(n)+f(n+2)\le 2f(n)\le f(n-2)+f(n)=\frac1{n-1}$. So $f(x)=\frac12$.

This is easy and quick. Btw, is there any other idea?

$\color{red}{Edit[2]}$: I find that @Byron Schmuland 's method is especially useful in a specific kind of problem. $$f(x)=\lim_{n\to\infty}\int_0^1\frac{x^n}{1+x}dx$$,for example, Let $X_,...,X_n$ be i.i.d. random variables with density$f(x)=1$ on $(0,1)$, CDF(cumulative density function) of $X$ is $F(x)=x$. Now let $M=\max(X_1,\dots,X_n)$, its density function is: $f_M(x)=nx^{n-1}$ for $x\in(0,1)$ Also, it's also not hard to see that $M\to1$ in distribution as $n\to\infty$. So,$$f(x)=\lim_{n\to\infty}\int_0^1\frac{x^n}{1+x}dx=\lim_{n\to\infty}\frac1n\int_0^1{f_M(x)\frac{x}{x+1}dx}=\lim_{n\to\infty}\frac1nE(\frac{M}{M+1})=0$$ Although this simple example can be solved by other ways more easily, but this gives us another perspective.

  • 2
    $\begingroup$ Have you tried setting up a reduction formula? $\endgroup$ – Kari Aug 10 '15 at 11:44
  • $\begingroup$ @Khallil Thank you. It works! I edit my answer by following your idea. $\endgroup$ – Rowan Aug 10 '15 at 12:02

You may perform the change of variable $u=\tan x$ to get easily $$ I_n:=\int_0^{\pi/4}(\tan x)^ndx=\int_0^1\frac{u^n}{1+u^2}du. $$ Then you may just integrate by parts, $$ \begin{align} I_n=\int_0^1\frac{u^n}{1+u^2}du&=\left. \frac{u^{n+1}}{(n+1)}\frac{1}{1+u^2}\right|_0^1+\frac{2}{(n+1)}\int_0^1\frac{u^{n+2}}{(1+u^2)^2}\:du\\\\ &=\color{blue}{\frac12}\frac1{(n+1)}+\frac{2}{(n+1)}\int_0^1\frac{u^{n+2}}{(1+u^2)^2}\:du. \tag1 \end{align} $$ Observing that $$ 0\leq \int_0^1\frac{u^{n+2}}{(1+u^2)^2}\:du\leq \int_0^1 u^n\:du=\frac1{n+1} $$ gives $$ 0\leq \frac{2}{(n+1)}\int_0^1\frac{u^n}{(1+u^2)^2}\:du\leq \frac{2}{(n+1)^2}. \tag2 $$ Then combining $(1)$ and $(2)$ leads to

$$ \lim_{n \to +\infty}n\int_0^{\pi/4}\!(\tan x)^n\:dx=\lim_{n \to +\infty}n\:I_n=\color{blue}{{\frac12}}$$

as suggested.

  • $\begingroup$ This way is straight. And the formula before$(1)$, the numerator is $u^{n+1+1}=u^{n+2}\ne u^{n+1-1}=u^n$. But the$\le$ part is still true. $\endgroup$ – Rowan Aug 10 '15 at 12:36
  • $\begingroup$ @user229922 Observe that $u^{n+2}\leq u^n$ since $u \in [0,1]$. Thanks! $\endgroup$ – Olivier Oloa Aug 10 '15 at 12:37
  • 1
    $\begingroup$ Sure. And many thanks. You have answered several questions of mine. :) $\endgroup$ – Rowan Aug 15 '15 at 23:10
  • $\begingroup$ I had accepted just yet and seems no response, I'll do it again ...Any I I always forgot to accept good answers cuz I almost forgot this function, so sorry about this. $\endgroup$ – Rowan Aug 15 '15 at 23:40

Here's another idea, like you asked:

$$ \lim_{n \to \infty} n\int_0^{\pi/4} (\tan x)^n dx = \lim_{n \to \infty} \int_0^1 \frac{n u^n}{u^2+1} du $$ using $ u= tanx $ and then

$$ \lim_{n \to \infty} \int_0^1 \frac{n u^n}{u^2+1} du = \lim_{n \to \infty } \int_0^1 \frac{y^{\frac{1}{n}}}{1+y^{\frac{2}{n}}} dy $$ using $ y= u^n$

Note that, in this form, we have "absorbed" the $n$ that tends to infinity in a way that we can easily see what happens in the limit. Namely, the exponents tend to zero, so $y$ to the exponent tends to $1$ and hence

$$ \lim_{n \to \infty } \int_0^1 \frac{y^{\frac{1}{n}}}{1+y^{\frac{2}{n}}} dy = \int_0^1 \frac{dy}{1+1} = \frac{1}{2} \int_0^1 dy = \frac{1}{2} $$ and you obtain your result after an easy integral.

  • 2
    $\begingroup$ Please, how do you justify permuting limit and integral here? Thanks. $\endgroup$ – Olivier Oloa Aug 10 '15 at 12:38
  • $\begingroup$ Wow! Nice method. $\endgroup$ – Rowan Aug 10 '15 at 12:43
  • 3
    $\begingroup$ Dominated convergence or monotonic convergence would both work @OlivierOloa. $\endgroup$ – Daniel Fischer Aug 10 '15 at 12:43
  • 3
    $\begingroup$ @DanielFischer Ok, thanks for him :) It might be better to mention it. $\endgroup$ – Olivier Oloa Aug 10 '15 at 12:44

Here's a solution based on order statistics, similar to my answer here.

Let $X_1,\dots, X_n$ be i.i.d. random variables with density $f(x)=1+\tan(x)^2$ on $(0,\pi/4)$. The distribution function of $X$ is $F(x)=\tan(x)$ for $0\leq x\leq \pi/4$. Now let $M=\max(X_1,\dots, X_n)$; its density function is $$f_M(x)=n F(x)^{n-1}f_X(x)=n\,(\tan(x))^{n-1}\,(1+\tan(x)^2)\text{ for }0\leq x\leq \pi/4.$$ Also, it is not hard to see that $M\to \pi/4$ in distribution as $n\to\infty$. Now $$\int_0^{\pi/4} n \tan(x)^n \,dx =\int_0^{\pi/4} {\tan(x)\over 1+\tan(x)^2}\, f_M(x) \,dx =\mathbb{E}\left({\tan(M)\over 1+\tan(M)^2}\right).$$

Since $\tan(\pi/4)=1$, this converges to ${1\over 1+1}={1\over 2}$ as $n\to\infty$.

  • $\begingroup$ Thanks you for providing such a wonderful method. For me, it's easy to understand your answer but really hard to think of this method. Is there any general rule to find a problem can be transferred to a statistic one? Or, is it empirical that you can tell there is a pdf $f(x)=1+(tanx)^2$ qualify this question. $\endgroup$ – Rowan Aug 10 '15 at 13:09
  • $\begingroup$ I first noticed that $F(x)=\tan(x)$ on $(0,\pi/4)$ increases from 0 to 1, so it is a distribution function. Then I just differentiated it to find the density. $\endgroup$ – user940 Aug 10 '15 at 13:42
  • $\begingroup$ I tried a similar problem using your method.Could you please check it for me to see whether it's correct? I edit my question and wrote the problem in the bottom. $\endgroup$ – Rowan Aug 10 '15 at 14:23
  • $\begingroup$ @user229922 What you've done is OK, but a bit complicated for the simple problem where you don't multiply by $n$. $\endgroup$ – user940 Aug 10 '15 at 14:34

After letting $\tan(x)\mapsto x$ and using the elementary limit $\lim_{n\to\infty} n \int_0^1 x^n f(x) \ dx = f(1)$, where $f(x)$ is continuous, we conclude that $$\lim_{n\to\infty}\biggl(n\int_0^{\pi/4}(\tan x)^n\,dx\biggr)=\frac{1}{2}.$$

  • $\begingroup$ It's concise; Byron Schmuland's answer provide a good way to prove this elementary limit. It's so great to learn so many different methods! Thank you. $\endgroup$ – Rowan Aug 10 '15 at 16:51

$\newcommand{\bbx}[1]{\,\bbox[8px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$

Laplace Method:

\begin{align} \lim_{n \to \infty}\bracks{n\int_{0}^{\pi/4}\tan^{n}\pars{x}\,\dd x} & = \lim_{n \to \infty}\bracks{n\int_{0}^{\pi/4} \tan^{n}\pars{{\pi \over 4} - x}\,\dd x} \\[5mm] & = \lim_{n \to \infty}\bracks{n\int_{0}^{\pi/4} \exp\pars{n\ln\pars{\tan\pars{{\pi \over 4} - x}}}\,\dd x} \\[5mm] & = \lim_{n \to \infty}\bracks{n\int_{0}^{\infty} \exp\pars{-2nx}\,\dd x} = \lim_{n \to \infty}\pars{n\,{1 \over 2n}} = \bbx{\ds{1 \over 2}} \end{align}


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.