More fun in 2015, Part 44

Find S = log 2 ( 1 cos ( 2 k π 2015 ) ) S=-\sum\log_2 \left(1-\cos\left(\frac{2k\pi}{2015}\right)\right)

where the sum is taken over all 1 k 2015 1\leq k\leq 2015 with gcd ( k , 2015 ) = 1 \gcd(k, 2015)=1 .


The answer is 1440.

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

Otto Bretscher
Dec 28, 2015

The solution I had in mind is essentially the same as Aareyan's, with some small variants.

Lemma: Φ 2015 ( 1 ) = ( 1 ω ) = 1 \Phi_{2015}(1)=\prod(1-\omega)=1 , where the product is taken over all primitive 2015th roots of unity ω \omega .

Proof of lemma: Note that Φ p ( 1 ) = p \Phi_p(1)=p for prime p p . Now we show by induction that Φ n ( 1 ) = 1 \Phi_n(1)=1 for square-free n = p 1 . . . p m n=p_1*...*p_m with m > 1 m>1 . We write x n 1 = d n Φ d ( x ) x^n-1=\prod_{d|n}\Phi_d(x) , divide by x 1 x-1 , and evaluate at x = 1 x=1 to see that n = p 1 . . . p m Φ n ( 1 ) n=p_1*...*p_m*\Phi_n(1) so Φ n ( 1 ) = 1 \Phi_n(1)=1 and, in particular, Φ 2015 ( 1 ) = 1 \Phi_{2015}(1)=\boxed{1} .

Thus we have ( 1 ω ) = 1 \prod(1-\omega)=1 , and taking conjugates, ( 1 ω ) ˉ = 1 \prod(1-\bar{\omega)}=1 so 1 = ( 1 ω ) ( 1 ω ) ˉ 1=\prod(1-\omega)(1-\bar{\omega)} = k 2 ( 1 cos ( 2 k π / 2015 ) ) = 2 ϕ ( 2015 ) k ( 1 cos ( 2 π k / 2015 =\prod_k2(1-\cos(2k\pi/2015))=2^{\phi(2015)}\prod_k(1-\cos(2\pi k/2015 )), where the products are taken over all 1 k 2015 1\leq k\leq 2015 with gcd ( 1 , 2015 ) = 1 \gcd(1,2015)=1 . Taking logarithms to base 2, we find that S = ϕ ( 2015 ) = 1440 S=\phi(2015)=\boxed{1440}

Aareyan Manzoor
Dec 28, 2015

write the sum as log 2 ( ( 1 cos ( 2 k π 2015 ) ) ) -\log_2\left(\prod\left(1-\cos\left(\dfrac{2k\pi}{2015}\right)\right)\right) let w be the primitive 2015th roots of unity. then by eulers identity this becomes log 2 ( ( 1 w + w 1 2 ) ) = log 2 ( ( w 1 ( 1 w ) 2 2 ) ) -\log_2\left(\prod\left(1-\frac{w+w^{-1}}{2}\right)\right)=-\log_2\left(\prod\left(w^{-1}\dfrac{(1-w)^2}{2}\right)\right) product of all the primitive roots for all number≠2 are 1. neglect that.

a polynomial p can be written as p ( x ) = ( x r o o t s ) p(x)=\prod (x-roots) . what polynomial has roots w? the cyclotomic polynomial . we are just searching for log 2 ( Φ 2015 ( 1 ) 2 1440 ) = 1440 log 2 ( Φ 2015 ( 1 ) ) -\log_2\left(\dfrac{\Phi_{2015}(1)}{2^{1440}}\right)=1440-\log_2\left(\Phi_{2015}(1)\right) the formula for Φ n ( 1 ) = e Λ ( n ) \Phi_n(1)=e^{\Lambda(n)} for n≠1, and Λ ( n ) \Lambda(n) being the von magoldt function .proof in the comments. so, 1440 log 2 ( e 2 Λ ( 2015 ) ) = 1440 log 2 ( 1 ) = 1440 1440-\log_2\left(e^{2\Lambda(2015)}\right)=1440-\log_2\left(1\right)=\boxed{1440}

consider the basic x n 1 = d n Φ d ( x ) x^n-1=\prod_{d\mid n} \Phi_d(x) divide by Φ 1 ( x ) = x 1 \Phi_1(x)=x-1 . x n 1 + x n 2 + . . . + x + 1 = d n , d > 1 Φ d ( x ) x^{n-1}+x^{n-2}+...+x+1=\prod_{d\mid n,d>1} \Phi_d(x) put x=1 and take log ln ( n ) = d n , d > 1 ln ( Φ d ( 1 ) ) \ln(n)=\sum_{d|n,d>1}\ln(\Phi_d(1)) let f ( n ) = { 0 , n 1 ln ( Φ d ( 1 ) ) , n 2 f(n)=\begin{cases} 0&&,n≤1\\\ln(\Phi_d(1))&&,n≥2\end{cases} .and p ( n ) = ln ( n ) p(n)=\ln(n) .then in dirichlet convolution form: p = f u p=f*u . convolute both sides with μ \mu to get: p μ = f I = f p*\mu=f*I=f . in other words: ln ( Φ n ( 1 ) ) = d n ln ( d ) μ ( n d ) \ln(\Phi_n(1))=\sum_{d|n} \ln(d)\mu\left(\dfrac{n}{d}\right) it is well known: ln ( n ) = d n Λ ( n ) \ln(n)=\sum_{d|n} \Lambda(n) this is easy to prove, try it yourself. in dirichlet convolution form this is: p = Λ u p=\Lambda*u . convolute bothsides with μ \mu to get: p μ = Λ I = Λ p*\mu=\Lambda*I=\Lambda . in other words d n ln ( d ) μ ( n d ) = Λ ( n ) \sum_{d|n} \ln(d)\mu\left(\dfrac{n}{d}\right)=\Lambda(n) . we get that ln ( Φ n ( 1 ) ) = Λ ( n ) \ln(\Phi_n(1))=\Lambda(n) or Φ n ( 1 ) = e Λ ( n ) \Phi_n(1)=e^{\Lambda(n)} . we are done _\square .

Aareyan Manzoor - 5 years, 5 months ago

Yes, very nice! (+1) You are good!

Small typo: the final answer is 1440.

Otto Bretscher - 5 years, 5 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...