8
$\begingroup$

Compute

$$\lim_{n\to\infty}\frac{1}{n^3}\sum_{1\le i<j<k\le n} \sin\left(\frac{i+j+k}{n}\right)$$

$\endgroup$
11
$\begingroup$

\begin{align} \lim_{n \to \infty} \dfrac1{n^3} \sum_{1\le i<j<k\le n} \sin\left(\frac{i+j+k}{n}\right) & = \lim_{n \to \infty} \dfrac16 \sum_{i=1}^n \sum_{j=1}^n \sum_{k=1}^n \sin \left(\dfrac{i}n + \dfrac{j}n + \dfrac{k}n\right) \dfrac1{n^3} \end{align} Because each is a Riemann Sum, we can make this into an integral, when we take the limit. The above is true since the region $0 \leq x \leq y \leq z$ divides the unit cube into $1/6$ of the total region. The integral being symmetric should be the same in all six regions. There is a overlap on the boundaries but this is a measure zero set and hence won't affect the integral. \begin{align} \lim_{n \to \infty} \dfrac16 \sum_{i=1}^n \sum_{j=1}^n \sum_{k=1}^n \sin \left(\dfrac{i}n + \dfrac{j}n + \dfrac{k}n\right) \dfrac1{n^3} & = \dfrac16 \int_{[0,1]^3} \sin(x+y+z) dx dy dz\\ & = \dfrac43 \sin^4(1/2) (1+2 \cos(1)) \end{align} We could have also directly converted this directly into the integral shown below without the $1/6$ factor i.e. $$\lim_{n \to \infty} \dfrac1{n^3} \sum_{1\le i<j<k\le n} \sin\left(\frac{i+j+k}{n}\right) = \int_{z=0}^{z=1} \int_{y=0}^{y=z} \int_{x=0}^{x=y} \sin(x+y+z) dx dy dz$$ If you add in a factor of $\pi$ in your original expression i.e. $$\dfrac1{n^3} \sum_{1\le i<j<k\le n} \sin\left(\pi \left(\frac{i+j+k}{n} \right)\right) = - \dfrac{4}{3 \pi^3}$$

$\endgroup$
  • $\begingroup$ @Chris'ssister Have added the detail about the Riemann Sum. $\endgroup$ – user17762 Jan 20 '13 at 22:26
  • $\begingroup$ In case, the factor $1/6$ is confusing, we could have also directly converted this directly into the integral shown below without the $1/6$ factor as I have shown now. $\endgroup$ – user17762 Jan 20 '13 at 22:32
  • $\begingroup$ btw, do you know a generalization for this problem? $\endgroup$ – user 1357113 Jan 20 '13 at 22:36
  • $\begingroup$ You mean in $d$ dimensions? i.e. $\dfrac1{n^d} \sin((i_1+i_2 + \cdots i_d)/n)$ $\endgroup$ – user17762 Jan 20 '13 at 22:37
  • $\begingroup$ Yes. I mean a generalization in $d$ dimensions. $\endgroup$ – user 1357113 Jan 20 '13 at 22:38

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.