5th April

Calculus Level 5

cos [ ( 1 + 1 3 ) π ] 1 2 + cos [ ( 1 + 1 3 ) 2 π ] 2 2 + cos [ ( 1 + 1 3 ) 3 π ] 3 2 + = ? \dfrac{\cos\left[\left(1+\sqrt{\frac{1}{3}}\right)\cdot \pi \right ] }{1^2} +\dfrac{\cos\left[ \left(1+\sqrt{\frac{1}{3}}\right)\cdot 2\pi \right ] }{2^2} +\dfrac{\cos\left[ \left(1+\sqrt{\frac{1}{3}}\right)\cdot 3\pi \right ] }{3^2} +\cdots=\, ?


The answer is 0.

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 6, 2021

If we consider the function which is periodic of period 2 π 2\pi and defined on [ π , π ] [-\pi,\pi] by the formula F ( θ ) = θ 2 θ π F(\theta) \; = \; \theta^2 \hspace{2cm} |\theta| \le \pi then we define the Fourier coefficients a n = 1 2 π π π θ 2 e i n θ d θ = { 2 ( 1 ) n n 2 n 0 1 3 π 2 n = 0 a_n \; = \; \tfrac{1}{2\pi}\int_{-\pi}^\pi \theta^2 e^{-in\theta}\,d\theta \; = \; \left\{ \begin{array}{lll} \frac{2(-1)^n}{n^2} & \hspace{1cm} & n \neq 0 \\ \frac13\pi^2 & & n = 0 \end{array}\right. This means that we have the identity 1 3 π 2 + 4 n = 1 ( 1 ) n n 2 cos n θ = F ( θ ) θ R \tfrac13\pi^2 + 4\sum_{n=1}^\infty \frac{(-1)^n}{n^2}\cos n\theta \; = \; F(\theta) \hspace{2cm} \theta \in \mathbb{R} For this problem, we want the value of F ( 1 3 π ) = 1 3 π 2 F\big(\tfrac{1}{\sqrt{3}}\pi\big) = \tfrac13\pi^2 , which makes the answer 0 \boxed{0} .


Here is the detailed proof of the convergence. Using the periodicity of F F , F N ( θ ) = n = N N a n e i n θ = 1 2 π n = N N π π F ( ϕ ) e i n ( θ ϕ ) d ϕ = 1 2 π n = N N π π F ( θ + ϕ ) e i n ϕ d ϕ = 1 2 π π π F ( θ + ϕ ) sin ( N + 1 2 ) π sin 1 2 π d ϕ = 1 π 1 2 π 1 2 π F ( θ + 2 ϕ ) sin ( 2 N + 1 ) ϕ sin ϕ d ϕ = 1 π 0 1 2 π [ F ( θ + 2 ϕ ) + F ( θ 2 ϕ ) ] sin ( 2 N + 1 ) ϕ sin ϕ d ϕ \begin{aligned} F_N(\theta) = \sum_{n=-N}^N a_n e^{in\theta} & = \; \tfrac{1}{2\pi} \sum_{n=-N}^N\int_{-\pi}^\pi F(\phi)e^{in(\theta-\phi)}\,d\phi \; =\; \tfrac{1}{2\pi} \sum_{n=-N}^N\int_{-\pi}^\pi F(\theta + \phi)e^{-in\phi}\,d\phi \\ & = \; \frac{1}{2\pi}\int_{-\pi}^\pi F(\theta+\phi) \frac{\sin(N+\frac12)\pi}{\sin\frac12\pi}\,d\phi \; =\; \frac{1}{\pi}\int_{-\frac12\pi}^{\frac12\pi} F(\theta+2\phi)\frac{\sin(2N+1)\phi}{\sin\phi}\,d\phi \\ &=\; \frac{1}{\pi}\int_0^{\frac12\pi} \big[F(\theta+2\phi) + F(\theta-2\phi)\big]\frac{\sin(2N+1)\phi}{\sin\phi}\,d\phi \end{aligned} for any N N N \in \mathbb{N} . The above calculation is valid for any continuous periodic function, and not just F F , so we deduce that 2 π 0 1 2 π sin ( 2 N + 1 ) ϕ sin ϕ d ϕ = 1 \frac{2}{\pi}\int_0^{\frac12\pi} \frac{\sin(2N+1)\phi}{\sin\phi}\,d\phi \; = \; 1 for any N N N \in \mathbb{N} , and hence F N ( θ ) F ( θ ) = 1 π 0 1 2 π [ F ( θ + 2 ϕ ) + F ( θ 2 ϕ ) 2 F ( θ ) ] sin ( 2 N + 1 ) ϕ sin ϕ d ϕ F_N(\theta) - F(\theta) \; = \; \frac{1}{\pi}\int_0^{\frac12\pi}\big[F(\theta+2\phi) + F(\theta-2\phi) - 2F(\theta)\big] \frac{\sin(2N+1)\phi}{\sin\phi}\,d\phi for any N N N \in \mathbb{N} . Since F F is continuous on R \mathbb{R} , and since the left and right-hand derivatives F ( θ ) F'(\theta-) and F ( θ + ) F'(\theta+) exist for all θ R \theta \in \mathbb{R} , we deduce that, for any π θ π -\pi \le \theta \le \pi , we can find A θ > 0 A_\theta > 0 such that F ( θ + 2 ϕ ) + F ( θ 2 ϕ ) 2 F ( θ ) A θ ϕ 0 ϕ 1 2 π \big|F(\theta + 2\phi) + F(\theta - 2\phi) - 2F(\theta)\big| \; \le \; A_\theta \phi \hspace{2cm} 0 \le \phi \le \tfrac12\pi and this implies that the function ϕ F ( θ + 2 ϕ ) + F ( θ 2 ϕ ) 2 F ( θ ) sin ϕ \phi \; \mapsto \; \frac{F(\theta + 2\phi) + F(\theta - 2\phi) - 2F(\theta)}{\sin\phi} is continuous on ( 0 , 1 2 π ] (0,\tfrac12\pi] and bounded as ϕ 0 + \phi \to 0+ , and hence belongs to L 1 ( 0 , 1 2 π ) L^1(0,\tfrac12\pi) . The Riemann-Lebesgue Lemma tells us that lim N F N ( θ ) = F ( θ ) \lim_{N \to \infty} F_N(\theta) \; =\; F(\theta) for all π θ π -\pi \le \theta \le \pi . Thus we deduce the identity n = 1 ( 1 ) n n 2 cos n θ = 1 4 θ 2 1 12 π 2 π θ π \sum_{n=1}^\infty\frac{(-1)^n}{n^2}\cos n\theta \; = \; \tfrac14\theta^2 - \tfrac{1}{12}\pi^2 \hspace{2cm} -\pi \le \theta \le \pi with the matching identity for F ( θ ) F(\theta) valid for all real θ \theta .

Here the sum is n = 1 cos ( 1 + 1 3 ) n π n 2 \displaystyle \sum_{n=1}^∞ \dfrac{\cos\left(1+\sqrt{\dfrac{1}{3}}\right)nπ}{n^2}

In previous I have generalised that previous problem

n = 1 sin ( n θ ) n = π θ 2 \displaystyle\sum_{n=1}^∞ \dfrac{\sin(n\theta)}{n} = \dfrac{π-\theta}{2} ( 0 < θ π ) (0<\theta≤π)

Now taking integral in both sides

n = 1 0 θ sin ( n θ ) n d θ = 0 θ π θ 2 d θ \displaystyle\sum_{n=1}^∞ \int_0^{\theta} \dfrac{\sin(n\theta)}{n} d\theta = \int_0^{\theta} \dfrac{π-\theta}{2} d\theta

n = 1 cos ( n θ ) n 2 n = 1 1 n 2 = θ 2 4 π θ 4 \implies\displaystyle\sum_{n=1}^∞ \dfrac{\cos(n\theta)}{n^2} -\sum_{n=1}^∞ \dfrac{1}{n^2} = \dfrac{\theta^2}{4} -\dfrac{π\theta}{4} So

n = 1 cos ( n θ ) n 2 = π 2 6 + θ 2 4 π θ 2 \displaystyle\sum_{n=1}^∞ \dfrac{\cos(n\theta)}{n^2} = \dfrac{π^2}{6} +\dfrac{\theta^2}{4} -\dfrac{π\theta}{2} ( 0 θ π ) \left(0≤\theta≤π\right)

Here θ = ( 1 + 1 3 ) π \theta=\left( 1+\sqrt{\dfrac{1}{3}} \right)π

put that and you will get n = 1 cos ( 1 + 1 3 ) n π n 2 = 0 \displaystyle \sum_{n=1}^∞ \dfrac{\cos\left(1+\sqrt{\dfrac{1}{3}}\right)nπ}{n^2} = 0

Very nice - I wonder if this leads to nice ways of evaluating the Riemann zeta function.

One question - why isn't this result periodic? If we call the function f ( θ ) f(\theta) , we'd expect f ( θ + 2 π ) = f ( θ ) f(\theta+2\pi)=f(\theta) , but this clearly isn't the case in your formula. Also, since it's not, which set of θ \theta values is it correct for?

Chris Lewis - 2 months, 1 week ago

Log in to reply

@Chris Lewis 0 θ π 0≤\theta≤π

Yes sir! I have found a way to find the values of Riemann zeta functions . I will write an discussion about that

Thanks sir

Dwaipayan Shikari - 2 months, 1 week ago

Log in to reply

@Dwaipayan Shikari . The function satisfies Dirichlet condition in [ 0 , π ] [0,\pi] .

Arghyadeep Chatterjee - 2 months, 1 week ago

Log in to reply

@Arghyadeep Chatterjee Yes - that's exactly it, but perhaps this should be mentioned/explained in the solution. In fact the definition of the function means it's correct on [ 0 , 2 π ] [0,2\pi] , since it's symmetric in the line θ = π \theta=\pi .

Chris Lewis - 2 months, 1 week ago

Log in to reply

@Chris Lewis Thanks for corrections

Dwaipayan Shikari - 2 months, 1 week ago

The convergence results from the function sarisfying dirichlet's condition in the given interval. So the fourier series converges. It does not guarantee periodicty. Ofcourse the function itself is not periodic but can be redefined and extended so that it forms a periodic function. @Chris Lewis

Nicholas Flammel - 2 months, 1 week ago

Log in to reply

Just a clarification. Nicholas Flammel and me are the same. One of the accounts are linked to my google account in my phone and the other one on my pc.

Arghyadeep Chatterjee - 2 months, 1 week ago

Log in to reply

@Arghyadeep Chatterjee "Nicholas Flammel"seems familiar . May be from Harry Potter . Alchemist may be

Dwaipayan Shikari - 2 months ago

That discussion

Dwaipayan Shikari - 2 months, 1 week ago

Although your formula is correct, your justification is wrong. You establish your formula for 0 < θ < π 0 < \theta < \pi , and then try to plug in a value of θ \theta that is greater than π \pi . It is much easier to separate out the sign changes... My proof goes into all the details of the Fourier analysis. @Chris Lewis @Nicholas Flammel @Dwaipayan Shikari

Mark Hennings - 2 months, 1 week ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...