20

I am trying to evaluate$$ \lim_{n\to \infty} \int_0^1 \int_0^1...\int_0^1 \cos^2\big(\frac{\pi}{2n}(x_1+x_2+...x_n)\big)dx_1 dx_2...dx_n. $$ This is from an old Putnam mathematics competition. Either 1965 or 1987 I forget. Should we re-write the $\cos^2$ term first or how should we approach it? Thanks

Jeff Faraci
  • 10,393
  • maybe write cosine in the $e^i$ form? – rlartiga Mar 26 '14 at 21:07
  • 3
    Another idea, you could use this relation: $$\cos^2\Big( \frac{\pi}{2n}(x_1+\ldots + x_n)\Big) = \frac{1}{4}\Big( e^{\frac{i\pi}{n}(x_1 + \ldots + x_n)} + e^{-\frac{i\pi}{n}(x_1 + \ldots + x_n)} \Big) + \frac{1}{2}$$ – Integral Mar 26 '14 at 21:37
  • 5
    It is a trick question, the expression is constant for all $n:;\sum x_i\mapsto n-\sum x_i:$ gives $\int_0^1\cdots\int_0^1\cos^{2}\left[\frac{\pi}{2n}\sum x_i\right] ;\mathrm{d}x_1\cdots\mathrm{d}x_n=\int_0^1\cdots\int_0^1\sin^{2}\left[\frac{\pi}{2n}\sum x_i\right] ;\mathrm{d}x_1\cdots\mathrm{d}x_n$. Add. – ocg Mar 26 '14 at 22:08
  • @JulienGodawatta This is off the Putnam math exam, what do you mean it is a "trick"? This is standard for the exam questions. – Jeff Faraci Mar 27 '14 at 01:43
  • 13
    Integral's idea works perfectly. Also, if you want some sledgehammer method, just refer to the Strong law of large number. It tells us that $\bar{x}{n} = \frac{1}{n}(x{1} + \cdots + x_{n}) \to \frac{1}{2}$ almost surely as $n \to \infty$, where $x_{k}$ are understood as i.i.d. uniform random variables on $[0, 1]$. So by the bounded convergence, the integral converges to $$ \Bbb{E}[\cos^{2}(\pi \bar{x}_{n} / 2) ] \to \Bbb{E}[\cos^{2}(\pi/4)] = \frac{1}{2}. $$ – Sangchul Lee Mar 27 '14 at 04:08
  • @Jeff Perhaps it was a bad choice of words - the question presents a limit when in fact we needn't think about limits at all, is what I meant. – ocg Mar 27 '14 at 11:07
  • @sos440 Sledgehammer? Maybe, but a really elegant (and deep) one, if you ask me. – Did Mar 31 '14 at 21:04

5 Answers5

26

Using $$ \cos^2(x)=\frac{1+\cos(2x)}{2} $$ we get that $$ \begin{align} &\int_0^1\int_0^1\cdots\int_0^1\cos^2\left(\frac{a\pi}{2n}(x_1+x_2+\dots+x_n)\right)\,\mathrm{d}x_1\,\mathrm{d}x_2\dots\,\mathrm{d}x_n\\ &=\frac12+\frac12\mathrm{Re}\left(\int_0^1\int_0^1\cdots\int_0^1e^{\frac{ia\pi}{n}(x_1+x_2+\dots+x_n)}\,\mathrm{d}x_1\,\mathrm{d}x_2\dots\,\mathrm{d}x_n\right)\\ &=\frac12+\frac12\mathrm{Re}\left(\left[\int_0^1e^{\frac{ia\pi}{n}x}\,\mathrm{d}x\right]^n\right)\\ &=\frac12+\frac12\mathrm{Re}\left(\left[\frac{n}{ia\pi}\right]^n\left[e^{\frac{ia\pi}{n}}-1\right]^n\right)\\ &=\frac12+\frac12\mathrm{Re}\left(\left[\frac{2n}{a\pi}\sin\left(\frac{a\pi}{2n}\right)\right]^ne^{\frac{ia\pi}{2}}\right)\\ &=\frac12+\frac12\left[\frac{2n}{a\pi}\sin\left(\frac{a\pi}{2n}\right)\right]^n\color{#C00000}{\cos\left(\frac{a\pi}{2}\right)}\\ &\to\frac12+\frac12\cos\left(\frac{a\pi}{2}\right)\\ &=\cos^2\left(\frac{a\pi}{4}\right) \end{align} $$ If $a=1$, $\color{#C00000}{\cos\left(\frac{a\pi}{2}\right)}$ is $0$, so the integral is $\frac12$ for all $n$.

robjohn
  • 353,833
8

$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert} \newcommand{\wt}[1]{\widetilde{#1}}$ \begin{align} &\color{#c00000}{\lim_{n \to \infty}\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \cos^{2}\pars{{\pi \over 2n}\,\bracks{x_{1} + x_{2} + \cdots + x_{n}}} \,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n}} \\[3mm]&=\half\bracks{% 1 + \color{#00f}{\lim_{n \to \infty}\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \cos\pars{{\pi \over n}\,\bracks{x_{1} + x_{2} + \cdots + x_{n}}} \,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n}}}\tag{1} \end{align}

\begin{align} &\color{#00f}{\lim_{n \to \infty}\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \cos\pars{{\pi \over n}\,\bracks{x_{1} + x_{2} + \cdots + x_{n}}} \,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n}} \\[3mm]&=\lim_{n \to \infty}\Re\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \int_{-\infty}^{\infty}\expo{\ic\pi x/n}\delta\pars{x - \sum_{k = 1}^{n}x_{k}}\,\dd x \,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n} \\[3mm]&=\lim_{n \to \infty}\Re\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \int_{-\infty}^{\infty}\expo{\ic\pi x/n} \int_{-\infty}^{\infty}\exp\pars{\ic q\bracks{x - \sum_{k = 1}^{n}x_{k}}} \,{\dd q \over 2\pi}\,\dd x\,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n} \\[3mm]&=\lim_{n \to \infty}\Re\int_{-\infty}^{\infty}\dd q\ \overbrace{\int_{-\infty}^{\infty}{\dd x \over 2\pi}\, \exp\pars{\ic\bracks{q + {\pi \over n}}x}}^{\ds{=\ \delta\pars{q + {\pi \over n}}}}\ \pars{\int_{0}^{1}\expo{-\ic q\xi}\,\dd\xi}^{n} \\[3mm]&=\lim_{n \to \infty} \Re\bracks{\pars{\int_{0}^{1}\expo{\ic\pi\xi/n}\,\dd\xi}^{n}} =\lim_{n \to \infty} \Re\bracks{\pars{\expo{\ic\pi/n} - 1 \over \ic\pi/n}^{n}} \\[3mm]&=\lim_{n \to \infty} \Re\bracks{% \expo{-\ic\pi/2}\pars{\expo{\ic\pi/2n} - \expo{-\ic\pi/2n} \over \ic\pi/n}^{n}} =\lim_{n \to \infty} \Re\braces{\expo{-\ic\pi/2}\bracks{\sin\pars{\pi/2n} \over \pi/2n}^{n}} \\[3mm]&=\color{#00f}{\large 0} \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad \qquad\qquad\pars{2} \end{align}

By replacing $\pars{2}$ in expression $\pars{1}$ we find: $$ \color{#00f}{\large\lim_{n \to \infty}\int_{0}^{1}\int_{0}^{1}\ldots\int_{0}^{1} \cos^{2}\pars{{\pi \over 2n}\,\bracks{x_{1} + x_{2} + \cdots + x_{n}}} \,\dd x_{1}\,\dd x_{2}\ldots\dd x_{n} = \half} $$

Felix Marin
  • 94,079
5

$\displaystyle \lim_{n\to \infty} \int_0^1 \int_0^1...\int_0^1 \cos^2\big(\frac{\pi}{2n}(x_1+x_2+...x_n)\big)dx_1 dx_2...dx_n $

$\displaystyle = \lim_{n\to \infty} \int_0^1 \int_0^1...\int_0^1 \cos^2\big(\frac{\pi}{2n}(1-x_1+1-x_2+...+1-x_n)\big)dx_1 dx_2...dx_n $

$\displaystyle = \lim_{n\to \infty} \int_0^1 \int_0^1...\int_0^1 \sin^2\big(\frac{\pi}{2n}(x_1+x_2+...x_n)\big)dx_1 dx_2...dx_n$

$\displaystyle = \frac12 \lim_{n\to \infty} \int_0^1 \int_0^1...\int_0^1 \sin^2\big(\frac{\pi}{2n}(x_1+x_2+...x_n)\big) + \cos^2\big(\frac{\pi}{2n}(x_1+x_2+...+x_n)\big)dx_1 dx_2...dx_n$

$=\dfrac12$

r9m
  • 18,208
  • Hi, I have known about this integral for more than 5 years now, and while I have been able to solve by other means, this solution stumps me. Please if anyone reads this, point me to the right direction. – SSBASE Nov 19 '22 at 21:10
  • 1
    @SSBASE The first line is just making the change of variable $x_i \mapsto 1-x_i$ (for each $i = 1, 2, \cdots, n$). Then use the fact $\cos \theta = \sin \left(\frac{\pi}{2} - \theta\right)$ .. finally add the integrals in line $(1)$ and $(3)$ and divide by $2$ to get the result. – r9m Nov 20 '22 at 06:33
  • I see... I was considering that it was due to something $P(X) - P'(X) = 1$ from elementary probability theory. I was also thinking it was due to $cos(x)$ being "orthogonal" to $sin(x)$. Also, how does this solution deal with the integration interval? (I think this is the cleanest and algebraically simplest solution on this thread, which is why I ask so much). – SSBASE Nov 21 '22 at 09:55
  • 1
    @SSBASE Finally we are integrating 1/2 over the unit cube $[0,1]^n$ .. the final result is independent of $n$. There's no point in writing the $\lim\limits_{n \to \infty}$ I guess .. – r9m Nov 21 '22 at 18:04
  • Apologies, there's a typo. It should be $P(X) + P'(X)=1$. Sorry! – SSBASE Nov 21 '22 at 21:00
  • thank you for responding even 8 years after your original answer! – SSBASE Nov 21 '22 at 21:01
3

Hint: $\displaystyle\frac1n\sum_1^nx_k$ is the mean value of $\bar x$, which, for $n\to\infty$, tends to $\dfrac{a+b}2$ , for $x_k\in(a,b)$.

Lucian
  • 49,312
0

I have checked the book of Putnam mathematics competition and your question is a little bit different from the original topic.

Since the following relation holds

$$\cos^{2}(π/2n(x_1+\dots+x_n))=\frac14 (e^{iπ/n}(x_1+\dots+x_n)+e^{−iπ/n}(x_1+\dots+x_n))+\frac12$$

it seems that it is not so hard to integrate your function:

$$ \lim_{n\to \infty} \int_0^1 \int_0^1\cdots\int_0^1 \cos^2\big(\frac{\pi}{2n}(x_1+x_2+...x_n)\big)\,dx_1\, dx_2\dots\,dx_n. $$

you can use the following formulas:

$$\int{dx_{1}} \int{dx_{2}}\dots\int{f(x_{1})f(x_{2})...f(x_{n})}{dx_{n}} = \frac{1}{n!} (\int{f(\tau)}d\tau)^{'}$$

where, $f(x)=e^{i\pi/n}x$.

It seems that this method will be easy to use.