Beyond Basel

Algebra Level 4

There exist positive integers A , B , C , D , E , F A,B,C,D,E,F such that p = 1 n cot 6 ( p π 2 n + 1 ) = 1 F n ( 2 n 1 ) ( A n 4 + B n 3 + C n 2 D n + E ) \sum_{p=1}^n \cot^6 \left(\tfrac{p \pi}{2n+1}\right) = \tfrac{1}{F}n(2n-1)\big(An^4 + Bn^3 + Cn^2 - Dn + E\big) for all positive integers n n , where F F is as small as possible.

What is A + B + C + D + E + F ? A + B + C + D + E + F?


Bonus: Use this result to prove that p = 1 1 p 6 = 1 945 π 6 . \sum_{p=1}^\infty \tfrac{1}{p^6} = \tfrac{1}{945}\pi^6.


The answer is 1484.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Mark Hennings
Apr 4, 2018

Note that sin ( 2 n + 1 ) x = I m ( ( cos x + i sin x ) 2 n + 1 ) = m = 0 n ( 1 ) m ( 2 n + 1 2 m + 1 ) cos 2 n 2 m x sin 2 m + 1 x = sin 2 n + 1 x F n ( cot 2 x ) \begin{aligned} \sin(2n+1)x & = \; \mathrm{Im}\left((\cos x + i \sin x)^{2n+1}\right) \; = \; \sum_{m=0}^n (-1)^m {2n+1 \choose 2m+1} \cos^{2n-2m}x \sin^{2m+1}x \\ & = \; \sin^{2n+1}x F_n(\cot^2x) \end{aligned} where F n ( X ) F_n(X) is the degree n n polynomial F n ( X ) = m = 0 n ( 1 ) m ( 2 n + 1 2 m + 1 ) X n m = ( 2 n + 1 1 ) X n ( 2 n + 1 3 ) X n 1 + ( 2 n + 1 5 ) X n 2 + F_n(X) \; = \; \sum_{m=0}^n (-1)^m{2n+1 \choose 2m+1}X^{n-m} \; = \; {2n+1 \choose 1} X^n - {2n+1 \choose 3}X^{n-1} + {2n+1 \choose 5}X^{n-2} + \cdots Note that the zeros of F n ( X ) F_n(X) are the values of cot 2 x \cot^2x where sin ( 2 n + 1 ) x = 0 \sin(2n+1)x = 0 but sin x 0 \sin x \neq 0 , and so F n ( X ) F_n(X) has n n distinct real zeros, namely u p = cot 2 ( p π 2 n + 1 ) p = 1 , 2 , . . . , n u_p \; = \; \cot^2 \left(\tfrac{p\pi}{2n+1}\right) \hspace{2cm} p \; = \; 1,2,...,n Thus we know that e 1 = p = 1 n u p = ( 2 n + 1 1 ) 1 ( 2 n + 1 3 ) = 1 3 n ( 2 n 1 ) e 2 = 1 p < q n u p u q = ( 2 n + 1 1 ) 1 ( 2 n + 1 5 ) = 1 30 n ( n 1 ) ( 2 n 1 ) ( 2 n 3 ) e 3 = 1 p < q < r n u p u q u r = ( 2 n + 1 1 ) 1 ( 2 n + 1 7 ) = 1 630 n ( n 1 ) ( n 2 ) ( 2 n 1 ) ( 2 n 3 ) ( 2 n 5 ) \begin{aligned} e_1 \; = \; \sum_{p=1}^n u_p & = \; {2n+1 \choose 1}^{-1}{2n+1 \choose 3} \; = \; \tfrac13n(2n-1) \\ e_2 \; = \; \sum_{1 \le p < q \le n} u_p u_q & = \; {2n+1 \choose 1}^{-1}{2n+1 \choose 5} \; = \; \tfrac{1}{30}n(n-1)(2n-1)(2n-3)\\ e_3 \; = \; \sum_{1 \le p < q < r \le n} u_p u_q u_r & = \; {2n+1 \choose 1}^{-1}{2n+1 \choose 7} \; = \; \tfrac{1}{630}n(n-1)(n-2)(2n-1)(2n-3)(2n-5) \end{aligned} and hence p = 1 n cot 2 ( p π 2 n + 1 ) = e 1 = 1 3 n ( 2 n 1 ) p = 1 n cot 4 ( p π 2 n + 1 ) = e 1 2 2 e 2 = 1 45 n ( 2 n 1 ) ( 4 n 2 + 10 n 9 ) p = 1 n cot 6 ( p π 2 n + 1 ) = e 1 3 3 e 1 e 2 + 3 e 3 = 1 945 n ( 2 n 1 ) ( 32 n 4 + 112 n 3 + 8 n 2 252 n + 135 ) \begin{aligned} \sum_{p=1}^n \cot^2\big(\tfrac{p \pi}{2n+1}\big) & = \; e_1 \; = \; \tfrac13n(2n-1) \\ \sum_{p=1}^n \cot^4\big(\tfrac{p \pi}{2n+1}\big) & = \; e_1^2 - 2e_2 \; = \; \tfrac{1}{45}n(2n-1)(4n^2 + 10n - 9)\\ \sum_{p=1}^n \cot^6\big(\tfrac{p \pi}{2n+1}\big) & = \; e_1^3 - 3e_1e_2 + 3e_3 \; = \; \tfrac{1}{945}n(2n-1)(32n^4 + 112n^3+ 8n^2 - 252n + 135) \end{aligned} which makes the answer 32 + 112 + 8 + 252 + 135 + 945 = 1484 32 + 112 + 8 + 252 + 135 + 945 = \boxed{1484} .


If 0 < x < 1 2 π 0 < x < \tfrac12\pi then sin x < x < tan x \sin x < x < \tan x , and hence cot x < x 1 < c o s e c x \cot x < x^{-1} < \mathrm{cosec}\, x , so that cot 6 x < x 6 < c o s e c 6 x = ( cot 2 x + 1 ) 3 = cot 6 x + 3 cot 4 x + 3 cot 2 x + 1 \cot^6x \; < \; x^{-6} \;< \; \mathrm{cosec}^6 x \; = \; (\cot^2 x + 1)^3 \; = \; \cot^6x + 3\cot^4x + 3\cot^2x + 1 and hence p = 1 n cot 6 ( p π 2 n + 1 ) < ( 2 n + 1 ) 6 π 6 p = 1 n 1 p 6 < p = 1 n [ cot 2 ( p π 2 n + 1 ) + 1 ] 3 \sum_{p=1}^n \cot^6\big(\tfrac{p \pi}{2n+1}\big) \; < \; \tfrac{(2n+1)^6}{\pi^6}\sum_{p=1}^n \frac{1}{p^6} \; < \; \sum_{p=1}^n \big[\cot^2\big(\tfrac{p \pi}{2n+1}\big) + 1\big]^3 Since p = 1 n cot 2 ( p π 2 n + 1 ) 2 3 n 2 p = 1 n cot 4 ( p π 2 n + 1 ) 8 45 n 4 p = 1 n cot 6 ( p π 2 n + 1 ) 64 945 n 6 \sum_{p=1}^n \cot^2\big(\tfrac{p \pi}{2n+1}\big) \; \sim \; \tfrac23n^2 \hspace{1cm} \sum_{p=1}^n \cot^4\big(\tfrac{p \pi}{2n+1}\big) \; \sim \; \tfrac{8}{45}n^4 \hspace{1cm} \sum_{p=1}^n \cot^6\big(\tfrac{p \pi}{2n+1}\big) \; \sim \; \tfrac{64}{945}n^6 as n n \to \infty , we easily deduce that lim n 1 ( 2 n + 1 ) 6 p = 1 n cot 6 ( p π 2 n + 1 ) = lim n 1 ( 2 n + 1 ) 6 p = 1 n [ cot 2 ( p π 2 n + 1 ) + 1 ] 3 = 1 945 \lim_{n \to \infty} \tfrac{1}{(2n+1)^6}\sum_{p=1}^n \cot^6\big(\tfrac{p\pi}{2n+1}\big) \; = \; \lim_{n \to \infty} \tfrac{1}{(2n+1)^6}\sum_{p=1}^n \big[\cot^2\big(\tfrac{p\pi}{2n+1}\big) + 1\big]^3 \; = \; \tfrac{1}{945} and hence p = 1 1 p 6 = 1 945 π 6 \sum_{p=1}^\infty \tfrac{1}{p^6} \; = \; \tfrac{1}{945}\pi^6

Nice problem and nice solution. It reminds me of a problem that I wrote. :)

John Ross - 3 years, 1 month ago

Log in to reply

You can use the formula for p = 1 n cot 2 p π 2 n + 1 \sum_{p=1}^n \cot^2\tfrac{p \pi}{2n+1} for an elementary evaluation, together with estimate of the rate of convergence, of ζ ( 2 ) \zeta(2) .

Mark Hennings - 3 years, 1 month ago

I love this problem :))))

Vishruth Bharath - 3 years, 1 month ago

Can someone link me to the concepts discussed in this solution, or just say their name

A Former Brilliant Member - 3 years, 1 month ago

Log in to reply

It starts with complex number theory and, in particular, de Moivre's Theorem, to define the polynomial F n F_n . It then uses ideas about roots of polynomials, in particular Vieta's formulae, to gain information about the roots of F n F_n and, in particular, so find the values of p = 1 n cot 2 k p π 2 n + 1 \sum_{p=1}^n \cot^{2k}\tfrac{p\pi}{2n+1} for k = 1 , 2 , 3 k=1,2,3 .

The reason why I posed the question is that these techniques give us an elementary method of evaluating the infinite sums p = 1 p 2 = 1 6 π 2 p = 1 p 4 = 1 90 π 4 p = 1 p 6 = 1 945 π 6 \sum_{p=1}^\infty p^{-2} = \tfrac16\pi^2 \hspace{1cm} \sum_{p=1}^\infty p^{-4} = \tfrac{1}{90}\pi^4 \hspace{1cm} \sum_{p=1}^\infty p^{-6} = \tfrac{1}{945}\pi^6 The bonus part of the question proved the last one of these; the first two are handled using the inequality sin x < x < tan x \sin x < x < \tan x in just the same way.

Mark Hennings - 3 years, 1 month ago

Log in to reply

Beautiful problem and solution

John Price II - 3 years, 1 month ago
Jeremy Galvagni
Apr 16, 2018

My skills do not run this deep. This problem was so far beyond me that I knew I would have to approach it more like a detective than a mathematician.

First I directly evaluated the first 6 terms on my calculator: 1 27 , 34 5 , 563 7 , 11692 27 , 1575 , 4510 \frac{1}{27}, \frac{34}{5}, \frac{563}{7}, \frac{11692}{27}, 1575, 4510

The common denominator is 945. That's just like the bonus. Let's assume the continued terms are either integers or don't require any other factors in their denominators. OK then F = 945 F=945

Now to tackle the polynomial. I have 6 terms but I only need 5 for a fourth-degree polynomial. Good. I can check my result with it.

1 27 = 1 945 ( 1 ) ( 1 ) ( 35 ) \frac{1}{27}=\frac{1}{945}(1)(1)(35)

34 5 = 1 945 ( 2 ) ( 3 ) ( 1071 ) \frac{34}{5}=\frac{1}{945}(2)(3)(1071)

563 7 = 1 945 ( 3 ) ( 5 ) ( 5067 ) \frac{563}{7}=\frac{1}{945}(3)(5)(5067)

11692 27 = 1 945 ( 4 ) ( 7 ) ( 14615 ) \frac{11692}{27}=\frac{1}{945}(4)(7)(14615)

1575 = 1 945 ( 5 ) ( 9 ) ( 33075 ) 1575=\frac{1}{945}(5)(9)(33075)

4510 = 1 945 ( 6 ) ( 11 ) ( 64575 ) 4510=\frac{1}{945}(6)(11)(64575)

I used my calculator to fit a polynomial to the first 5 terms and it nicely fit the 6th. Looking good.

The polynomial is 32 x 4 + 112 x 3 + 8 x 2 252 x + 135 32x^{4}+112x^{3}+8x^{2}-252x+135

The x x term is the only negative! Another clue that I am on the right track.

Then the moment of truth: 32 + 112 + 8 + 252 + 135 + 945 = 1484 32+112+8+252+135+945=1484

Try it and... Advanced Problem #5, you are MINE!

F = 945 would be an excellent guess because of the bonus problem posted.

Michael Mendrin - 3 years, 1 month ago

Log in to reply

Given that the answer is true, finding the values of A F \tfrac{A}{F} , B F \tfrac{B}{F} , C F \tfrac{C}{F} , D F \tfrac{D}{F} and E F \tfrac{E}{F} is a matter of solving a 5 × 5 5 \times 5 matrix equation. Given the bonus part, it is indeed a reasonable first guess to suppose that the solutions to the matrix equation are all rationals with denominator F = 945 F=945

Mark Hennings - 3 years, 1 month ago

Log in to reply

945 = 3 3 5 7 945 = 3^3 \cdot 5 \cdot 7 is a number that actually crops up in other places, but why F should be 945 in this problem isn't immediately or even intuititvely obvious. It takes an analysis like yours to bring that out.

Michael Mendrin - 3 years, 1 month ago

Log in to reply

@Michael Mendrin I agree. without the clue that the value of ζ ( 6 ) \zeta(6) is relevant to this question, the reason why F = 945 F=945 is not at all obvious.

Mark Hennings - 3 years, 1 month ago

Log in to reply

@Mark Hennings I might have still guessed F=945 since it is the common denominator of the terms I had. That terms 5 and 6 are whole numbers makes me suspect all of the rest are as well.

As I said, I was just making my best guess.

Jeremy Galvagni - 3 years, 1 month ago

Log in to reply

@Jeremy Galvagni I just wonder where else we could uncover mathematical truths just by "detective experimentation" that you have done? Kind of like doing mathematics in the Thomas Edison way.

Michael Mendrin - 3 years, 1 month ago

Log in to reply

@Michael Mendrin It isn't uncommon. Follow a numerical pattern until you notice some kind of formula. Then try to prove the formula. It happens to me a lot with probability.

Important note: In this case I proved nothing, I just got the right number.

Jeremy Galvagni - 3 years, 1 month ago

Awesome solution!

Krishnaraj Sambath - 3 years, 1 month ago

Good solve. I guess, approaching this problem more like a math calculation than a rigorous proof has proved helpful.

Aditya Kondamudi - 3 years, 1 month ago

Did more or less the same - guessed F, evaluated the sum for n=1,2,...,5 and solved 5x5 system with wolfram. What can you do when math gets hard...

lovro cupic - 3 years, 1 month ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...