here's where I am stuck:

$$\lim_{n \to \infty} \sum_{k=0}^{n-1} \cos(kx/n)\frac{1}{n}$$

so...it looks like at this point I could convert to a Riemann integral, but to which one?

Maybe $$\int_0^{\infty} \cos(t) dt?$$

That wouldn't converge, so it wouldn't be good.

How could I proceed? It's easy to accept that $\large \frac{1}{n}$ maps to $dt$, when going from a Riemann sum to a Riemann integral, but I particularly have trouble figuring out how to convert the discrete variable to the continuous, integration variable, t.

Any ideas are welcome.


  • $\begingroup$ The limit is $1/x \cdot \int_0^x \cos(t)dt $ $\endgroup$ – Joel Dec 29 '15 at 2:00
  • 1
    $\begingroup$ The limit corresponds to the integral $\int_0^1 \cos(tx)\, dt$. $\endgroup$ – levap Dec 29 '15 at 2:00
  • $\begingroup$ @levap I doubt it since for $x=0$ it's the harmonic series, which diverges. $\endgroup$ – Gregory Grant Dec 29 '15 at 2:02
  • 1
    $\begingroup$ @GregoryGrant For $x = 0$, the sum is identically one (note that the summation is over $k$, not $n$). $\endgroup$ – levap Dec 29 '15 at 2:08
  • 1
    $\begingroup$ Hi @Joel, my answer agrees with yours: $\large \frac{sin(x)}{x}$. Thanks so much for your comment. I feel a bit better now with converting the discrete variable into the continuous variable :-) $\endgroup$ – user301446 Dec 29 '15 at 2:32

Hint: How would you write $$\int_0^1 \cos(xt) \,dt$$ as a Riemann sum?

  • $\begingroup$ Hmmm @Snow, something like $\sum_1^{N} cos ( ?) \frac{1}{N}$ , so that the total interval length is 1. What do you think? I am confused about the input for cosine, though. Would it just be... $xn$? Seems like such an...arbitrary choice...and that I'm not following any hard rules to make that conversion. (the only rule I am aware of in such conversions is that mesh(p) must go to zero, which we clearly have from 1/n...) Thanks, $\endgroup$ – user301446 Dec 29 '15 at 2:08
  • $\begingroup$ @user301446, Almost. You have $\sum_{k=0}^{N-1} \cos(x t_k) \frac{1}{N}$. The first time most students encounter Riemann sums, they see it not in the generality of partitions and mesh, but with equally sized subintervals. Can you write an expression for $t_k$ if you want subintervals to be of equal length and $t_k$ is the left end point of the $k$-th subinterval? $\endgroup$ – Snow Dec 29 '15 at 2:12
  • $\begingroup$ Hi @Snow ooh...hmm...$\frac {k}{N}$? so that $\frac{k}{N}$ belongs in the k-th sub-interval where it is to be the input of cosine - to give the height of the rectangle...I think... $\endgroup$ – user301446 Dec 29 '15 at 2:20
  • $\begingroup$ ....so that...letting mesh(p) go to zero, such an input becomes the continuous variable, $t$. so, it's the mapping $\frac{k}{N} \to t$, so that in total the mapping is from $\frac{k}{N}x \to tx$ $\endgroup$ – user301446 Dec 29 '15 at 2:23
  • $\begingroup$ What do you think @Snow? :-) $\endgroup$ – user301446 Dec 29 '15 at 2:25

Consider the function $f(t) = \cos(tx)$ on $[0,1]$. Let

$$ 0 = t_0 < t_1 = \frac{1}{n} < t_2 = \frac{2}{n} < \ldots < t_{n-1} = \frac{n-1}{n} < t_n = 1$$ be a partition of $[0,1]$ into $n$ sub-intervals of length $\frac{1}{n}$. In the interval $[t_i, t_{i+1}]$ choose the point $\xi_i = t_i$. Then the Riemann sum corresponding to the partition $\{t_i\}_{i=0}^{n-1}$ and the choice of points $\{\xi_i\}_{i=0}^{n-1}$ is

$$ \sum_{k=0}^{n-1} f(\xi_k) (t_{k+1} - t_k) = \sum_{k=0}^{n-1} f \left( \frac{k}{n} \right) \frac{1}{n} = \sum_{k=0}^{n-1} \cos \left( \frac{k}{n} x \right) \frac{1}{n}. $$

Hence, the limit of the expression (as $n \to \infty$) converges to

$$ \int_{0}^1 \cos(tx) \, dt = \begin{cases} [\frac{\sin(tx)}{x}]^{t = 1}_{t=0} & x \neq 0, \\ 1 & x = 0. \end{cases} = \begin{cases} \frac{\sin(x)}{x} & x \neq 0, \\ 1 & x = 0. \end{cases} $$

  • $\begingroup$ Thanks for your answer, @levap :-) $\endgroup$ – user301446 Dec 29 '15 at 2:34

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.