Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I am looking for nice ways of proving the divergence of the sequence $(x_n)_{n=1}^\infty$ defined by $$x_n:=\sin(n).$$ One (not so nice) way is to construct two subsequences: one where the indexes are picked such that they lie in the intervals $$I_k:=\left(\frac{\pi}{6}+2\pi(k-1),\frac{5\pi}{6}+2\pi(k-1)\right)$$ and one where they lie in $$J_k:=\left(\frac{7\pi}{6}+2\pi(k-1),\frac{11\pi}{6}+2\pi(k-1)\right).$$ If $(x_n)$ converges, then all its subsequences must converge to the same number, but here the first subsequence has all its values in the interval $\left[\frac{1}{2},1\right]$ while the second has its values in $\left[-1,-\frac{1}{2}\right]$. Contradiction.

Filling the details of this proof is rather tedious and not very elegant. Anyone has a better idea?

share|improve this question

6 Answers 6

up vote 11 down vote accepted

Equidistribution argument is very elegant, but it becomes a sledgehammer method when it comes to a question of mere convergence.

A simple argument can reveal the divergence of $(\sin n, n \geq 1)$. Let

$$ (x_n, y_n) = (\cos n, \sin n)$$

(I changed the notation for the sake of consistency of notation.) Then the application of the addition formula for trigonometric function or the rotation matrix gives

$$ \begin{align*} x_{n+1} &= x_n \cos 1 - y_n \sin 1 \\ y_{n+1} &= x_n \sin 1 + y_n \cos 1. \end{align*} $$

Now assume $(y_n)$ converges. Then since $\sin 1 \neq 0$, we have

$$ x_{n+1} = (y_{n+1} - y_n \cos 1) \cot 1 - y_n \sin 1$$

and hence $(x_n)$ also converges. Now let $(x_n, y_n) \to (\alpha, \beta)$. Then taking limit to the recursive formula we have

$$ \begin{align*} \alpha &= \alpha \cos 1 - \beta \sin 1 \\ \beta &= \alpha \sin 1 + \beta \cos 1. \end{align*} $$

Solving this system of linear equations give $(\alpha, \beta) = (0, 0)$. On the other hand, since

$$ x_n^2 + y_n^2 = 1, $$

we must have

$$ \alpha^2 + \beta^2 = 1,$$

a contradiction! Therefore $(y_n)$ cannot converge. ////

Of course, we can say much more on $(y_n)$. For example, we can show that the set of limit points of $(y_n)$ is exactly $[-1, 1]$, and the Cesaro mean of $(y_n)$ is 0 from Weyl's criterion.

(This proof is from the book Problems in Real Analysis Advanced Calculus on the Real Axis.)

share|improve this answer
This is very nice! –  Jason DeVito Nov 17 '12 at 2:42
Very nice indeed! Thanks! –  Spenser Nov 17 '12 at 2:45
Thanks! Anyway I forgot to comment that this is not my original idea, but the idea that I learned from the book Problems in Real Analysis Advanced Calculus on the Real Axis. –  sos440 Nov 17 '12 at 2:49
Very nice argument! So this also proves that $(x_n)=(\cos n)$ cannot converge. –  Paul Apr 16 '13 at 21:21
Another quick way to see that, if $\lim_{n \to \infty} \sin(n)$ and $\lim_{n \to \infty} \cos(n)$ exist, then both limits are zero is to note that $\sum_{n=0} \sin(n)$ and $\sum_{n=0}^\infty \cos(n)$ have bounded partial sums. Indeed, $\sum_{k=0}^n \sin(k)$ and $\sum_{k=0}^n \cos(k)$ are the real and imaginary parts of $\sum_{k=0}^n e^{ik} = \frac{e^{i(n+1)} - 1}{e^i - 1}$ which is bounded. –  Mike F May 26 at 8:34

If $\alpha$ is an irrational number, the numbers $n\alpha$, considered mod $1$, are dense in $[0,1]$. (A stronger result is that they are equidistributed.) Thus the numbers $2n/\pi$ are dense in $[0,1]$, and therefore the integers are dense mod $\pi/2$. It follows that $\sin n$ diverges.

share|improve this answer

This question was posed today (April 16, 2013) as http://mathoverflow.net/questions/127726/integer-multiples-of-a-irrational-dense-in-r-z and Douglas Zare gave a nice trig-identity-based proof in comments there.

share|improve this answer

Essentially, every point in the interval $[-1, 1]$ is a limit point for the sequence $\left\{ {\sin n} \right\}$. Since there is more than one limit point, the sequence diverges.

share|improve this answer


[x] is a rounding function if [x] is the biggest integer such that for a real number x, ⌊x⌋

Theorem 1:

Define x_n=[〖10〗^n x]/〖10〗^n for a real number x. Then lim(n→∞)〖x_n=x〗. Proof:

|x_n-x|=|([〖10〗^n x]-〖10〗^n x)/〖10〗^n |≤1/〖10〗^n

Theorem 2:

|[〖10〗^n π]-〖10〗^n π| diverges.


Suppose |[〖10〗^n π]-〖10〗^n π| converges to a real number. This implies that there are integers k between 0 and 10 and K such that for n≥K, the nth decimal place of π is k. This is absurd since π is irrational.

Theorem 3:

sin(n) diverges.


If sin[〖10〗^n x] has a limit s, then for any ϵ>0 there exist an integer K such that for n≥K, |s-sin[〖10〗^n x]|< ϵ. However, this implies that |[〖10〗^n π]-〖10〗^n π| converges which is proven to be false.

share|improve this answer

A known resultis that if {x_n} and {y_n} are two sequences such that y_n is divergent and x_n\y_n converges to 0 then x_n also diverges. Take x_n=sin (n) and y_n=n then the result follows.

share|improve this answer
This is false. (What happens if you take $x_n = 1$ and $y_n = n$ with the same argument?) –  mrf May 26 at 7:59

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.