3

Given $\alpha=e^{i\pi/2017}$. We need to find $$\prod_{r=1}^{2016}(\alpha^{r}+\alpha^{-r})$$

With De moivre's, the problem can be modified to $$\prod_{r=1}^{2016}2\cos\left(\frac{r\pi}{2017}\right)$$ at which I paired up the front and back terms to make it$$\prod_{r=1}^{1008}4\cos^2\left(\frac{r\pi}{2017}\right)$$ where I couldn't proceed further. Please help with this.

Also (as this problem was present with a few of the kind) is there any geometrical interpretation on the argand plane for this expression?

Ted Shifrin
  • 115,160
  • 1
    Spelling: https://en.m.wikipedia.org/wiki/Abraham_de_Moivre – Paul Frost May 04 '22 at 11:04
  • 2
    General method in this situation: find a polynomial whose roots are exactly the $\cos(r\pi/2017)$ for $1 \le r \le 1008$ and use relations between coefficients and roots. – Christophe Leuridan May 04 '22 at 11:11
  • @PaulFrost corrected – Sung Jin-Woo May 04 '22 at 12:58
  • @ChristopheLeuridan thank you ! After some more trying, I also got https://www.google.com/url?sa=t&source=web&rct=j&url=https://math.stackexchange.com/questions/1351337/product-of-cosines-prod-r-17-cos-left-fracr-pi15-right&ved=2ahUKEwjA36at9MX3AhWE63MBHQu6CDwQFnoECAsQAQ&usg=AOvVaw2PJd8Qbq7VpPZ73lZISkFY also seems good,though didn't come under suggestions .I will leave my question as well in case a geometrical input comes along – Sung Jin-Woo May 04 '22 at 13:03
  • 2
    Is $2017$ a prime? If it is, see if you can prove that $\alpha^r+\alpha^{-r}$ is a unit, because that (with a little Galois Theory) will show the product is either one or minus one. Then, you have to figure out which. – Gerry Myerson May 04 '22 at 13:24
  • yes 2017 is prime, and the product is one (unless I bombed), @GerryMyerson, but I hesitated about posting a number theoretic argument of the fact as I did not think it would help the OP. – peter a g May 04 '22 at 13:30
  • @peter, OP has to learn Number Theory sometime, might as well be now. – Gerry Myerson May 04 '22 at 13:34
  • 1
    if no one writes up what I would expect to be more a useful answer (crossing fingers), I'll try to post tonight. – peter a g May 04 '22 at 13:46
  • 1
    Alt. hint: $,\alpha^{2r} \big|{r=0,1,\dots,2016},$ are the roots of $,p(x)=x^{2017}-1,$ so $,p(x)= \prod{r=0}^{2016}(x-\alpha^{2r}),$. Then see how $,p(-1),$ relates to the product in the problem. – dxiv May 04 '22 at 19:17

1 Answers1

6

Write $a = \alpha^{-2}$. and $p = 2017$, which is an (odd!) prime. Then your product can be written as $$ \alpha ^{p\cdot (p-1)/2 } N,\text{ where } N= \prod_{r=1}^{p-1} (1 + a^r).$$ The first term equals $+1$, since $p(p-1)/2= 2017\cdot 1008$ is even.

For the second term... Set $$\phi(x) ={x^p-1 \over x-1}= x^{p-1}+\cdots + 1.$$ (Consider the first equality as 'definition', and the second as arising from long division. FWIW, $\phi$ is called the $p$th 'cyclotomic' polynomial.)

Evaluating $\phi(x)= x^{p-1}+\cdots + 1$ at $x=1$, one gets $\phi(1) = p$.

On the other hand, If $b^p=1$, but $b^r\not =1 $, for any $1\le r<p$ ( e.g., $b=e^{2\pi i /p}$ - any such $b$ is called a 'primitive' $p$th root of unity), all of the zeros/roots of $x^p-1$ are of the form $b^r$, where $r=0,\cdots, p-1$, therefore $$ \phi(x) = {\prod_{r=0}^{p-1} (x- b^r) \over (x-1)} = \prod_{r=1}^{p-1} (x- b^r).$$ Therefore, (from the earlier evaluation), $$\phi(1) = \prod_{r=1}^{p-1} (1- b^r) = p.$$

So, using the fact that $a$ (defined in the first line of this post) is a $p$th primitive root of unity, one has that
$$ P = \prod_{r=1}^{p-1} ( 1-a^r) = p.$$ Likewise, $$NP = \prod_{r=1}^{p-1} (1- (a^2)^r) =p,$$ as $a^2$ is also a primitive pth root of unity. Hence, (earning me a million dollars of Clay prize money and possibly even a job somewhere), $NP=p$. Canceling the $p$'s, one gets $N =1$. Therefore, the original product is also equal to one.

BTW, the above evaluating of $N$ only really seems to use only that $p$ is odd; one does not need that $p$ be a prime. On the other hand, $\phi(x)$ would not be 'the' cyclotomic polynomial.

peter a g
  • 4,923
  • 1
    Wow! Until now I was quite sure that $P \neq NP$ :) – Zubin Mukerjee May 04 '22 at 19:32
  • 1
    For a shortcut at the end, you could use that $,1 = \phi(-1)=\prod(-1-b^r)=\prod(1+a^r),$. And you're right, it is enough for $,p,$ to be odd, not necessarily prime (+1). – dxiv May 04 '22 at 19:39
  • @dxiv - you're right, that would have been smarter. – peter a g May 04 '22 at 19:48
  • Really helpful! And yes you were right there as 2017 just happens to be the year of the problem. By the way, I didn't get one line in your answer, the clay prize one – Sung Jin-Woo May 05 '22 at 02:23
  • 1
    Thank you @Sungjin-woo... but just to make sure you got dxiv's argument $\phi(-1)= (-1)^{p-1}+\cdots+1 = 1-1+\cdots+ 1 =1$ is a quicker way of making the argument. Best wishes! – peter a g May 05 '22 at 14:11