Here's something similar to what Farsight has been doing here: the Koide formulaWP for the leptons' masses. It works remarkably well, but is it a lucky accident?
I took a look at what's in ctamblyn's spoiler tags:
I looked at the five "identities" on the bottom, and the first and fourth ones essentially state that pi is an algebraic number. Since that is not the case, those "identities" are thus incorrect, even if they are rather close numerically. The third and fifth ones are also close numerically, but like the first and fourth ones, their errors are well above Mathematica's default floating-point precision.
The second identity is, however, correct, and it's fairly easy to prove. It's equivalent to
cos(pi/7) + cos(3pi/7) + cos(5pi/7) + cos(7pi/7) + cos(9pi/7) + cos(11pi/7) + cos(13pi/7) = 0
= Re( sum of exp(k*pi*i/7) for k from 1 to 13 by 2 )
= Re( exp(pi*i/7) * sum of exp(2k*pi*i/7) for k from 0 to 6 )
= Re( exp(pi*i/7) * (1-exp(14pi*i/7))/(1 - exp(2pi*i/7)) )
= 0
Mathematica doesn't handle trig functions as well as I might want, so I worked with Chebyshev polynomials in x = cos(pi/7).
1st kind: Tn(cos(a)) = cos(n*a)
2nd kind: Un(cos(a)) = sin((n+1)*a)/sin(a)
I took a look at what's in ctamblyn's spoiler tags:
I looked at the five "identities" on the bottom, and the first and fourth ones essentially state that pi is an algebraic number. Since that is not the case, those "identities" are thus incorrect, even if they are rather close numerically. The third and fifth ones are also close numerically, but like the first and fourth ones, their errors are well above Mathematica's default floating-point precision.
The second identity is, however, correct, and it's fairly easy to prove. It's equivalent to
cos(pi/7) + cos(3pi/7) + cos(5pi/7) + cos(7pi/7) + cos(9pi/7) + cos(11pi/7) + cos(13pi/7) = 0
= Re( sum of exp(k*pi*i/7) for k from 1 to 13 by 2 )
= Re( exp(pi*i/7) * sum of exp(2k*pi*i/7) for k from 0 to 6 )
= Re( exp(pi*i/7) * (1-exp(14pi*i/7))/(1 - exp(2pi*i/7)) )
= 0
Mathematica doesn't handle trig functions as well as I might want, so I worked with Chebyshev polynomials in x = cos(pi/7).
1st kind: Tn(cos(a)) = cos(n*a)
2nd kind: Un(cos(a)) = sin((n+1)*a)/sin(a)
!