3

This is a challenging puzzle I heard from my little brother.

For some $n$ and $x$, $\sum_{k=1}^n \sin^{2k}(x) = 2013$.

Is it possible to deduce $$\sum_{k=1}^n \cos^{2k}(x) \text{ ?}$$

Edit: I've just noticed something which now seems obvious to me.
Choose $n = 2013$ and $x = \pi/2$ which satisfies the condtion. It follows that the cosine terms would sum to zero. I'm not sure this is a unique solution.

Mark
  • 5,696
  • 2
    Check for typos - the index $i$ doesn't seem to appear in the summand, which is usually a bad sign. Maybe it's supposed to be $\sum_{i=1}^n\sin^{2i}(x)$? – icurays1 Feb 12 '13 at 23:38
  • Oh yes that's what I meant. – Mark Feb 12 '13 at 23:38
  • 1
    I've just noticed a trivial solution - x = $\pi/2$ and n = 2013. It follows that the cosine sum is 0. Maybe the problem is to show that this is the only solution? – Mark Feb 12 '13 at 23:41
  • 4
    For a large enough of $n$ there would be an $x$ satisfying the equation, by intermediate value theorem. – Maesumi Feb 12 '13 at 23:53
  • @Maesumi could you elaborate? – Mark Feb 12 '13 at 23:55
  • 1
    At $x=\pi/2$ the sum is $n$ and at $x=0$ the sum is $0$. The function, ie, the first sum, is a continuous function of $x$, so all values between 0 and $n$ will be be attained. Hence if $n\ge 2013$ then the value $2013$ will be attained. quite possibly multiple times. It looks unlikely that all such $x$'s will yield the same answer in the second sum. How about using $2$ instead of $2013$ and doing numerical experimentation. – Maesumi Feb 13 '13 at 00:00
  • 2
    @Mark Let $n \geq 2013$. Let $f(x)=\Sigma_{i=1}^{n} \sin^{2i}(x)$. Since $f(0)=0$ and $f(\frac{\pi}{2})>2013$ by the IVT there exists some $x$ so that $f(x)=2013$. – N. S. Feb 13 '13 at 00:01

2 Answers2

7

As noted, the equation holds if $n=2013$ and $x=\pi/2$. Now let $n=2014$. By continuity, there is a value of $x$ a tiny bit smaller than $\pi/2$ for which the equation will hold, and, for this value of $x$, the cosine sum will not be zero. So one cannot deduce the cosine sum from knowing the first equation holds.

Gerry Myerson
  • 179,216
5

let $r=sin^2(x)$ we have

$$\sum_{k=1}^n r^k=\frac{r(1-r^n)}{1-r}=2013$$ Now we want: $$\sum_{k=1}^n (1-r)^k=\frac{(1-r)(1-(1-r)^n)}{r}$$

We can deduce: $$2013\sum_{k=1}^n (1-r)^k=(1-r^n)(1-(1-r)^n)$$

  • So, if you know $n$, then you can evaluate the cosine sum. – Gerry Myerson Feb 12 '13 at 23:58
  • @GerryMyerson: I have not tried yet. but the second sum may be simplified and obtained from first one. –  Feb 13 '13 at 00:01
  • 1
    If you know $n$ (and if you can solve a polynomial equation of high degree), yes. If you don't know $n$, I think my answer shows that you can't find the cosine sum just from knowing the sine sum. – Gerry Myerson Feb 13 '13 at 00:25
  • @GerryMyerson: It seems the second sum cannot be written in terms of first sum free from n. And when $n$ is given, one may not need to find an exact solution for $r$ and then put it in second one. for example if one can show the first equation is symmetric around 1/2 he may be able to find the second. However it doesn't seem to be symmetric around 1/2... –  Feb 13 '13 at 00:33
  • What would be required now is to see for which $0 \le r \le 1$ the first equation has integer solutions for $n$. If there is only one, the answer is affirmative, if there are various pairs $(r, n)$, negative. – vonbrand Feb 13 '13 at 02:34
  • @vonbrand: for sure in if $n<2013$ there's no solution. and for n>=2013 there's at most one solution for $r$. –  Feb 13 '13 at 03:18
  • @CutieKrait: Why would there be at most one solution? For some n, it seems like there could be at least two solutions for r. – Mark Feb 13 '13 at 18:18
  • 1
    @Mark: Suppose r and s are two solutions and s<r. Then $$\sum_{k=1}^n{r^k-s^k}=0$$ which is impossible. –  Feb 13 '13 at 19:11
  • Oh right! So the problem can be solved given an n. – Mark Feb 13 '13 at 19:27
  • @Mark: given an n –  Feb 13 '13 at 19:32