5

find the sum $$\sum_{k=1}^{\infty} \frac {\sin^2(kx)}{k^2},$$ for $x\in [-\pi,\pi]$.

Also i know that $$\sum_{k=1}^{\infty} \frac {\sin(kx)}{k}=\frac {\pi-x}{2}$$

Any help would be much appreciated.

4 Answers4

2

Write $$ \sin^2(k\,x)=\frac{1-\cos(2\,k\,x)}{2}. $$ Can you find a function whose Fourier series is $$ \sum_{k=1}^\infty\frac{\cos(2\,k\,x)}{k^2}\text{ ?} $$

  • Thanks for your answer. Also i thought of this step,but i couldn't find such a function.any hint? – user113576 Feb 02 '14 at 19:20
  • I think i found it. It the function $$f(x)=\frac {(\pi-2x)^2}{4}-\frac {\pi^2}{12}$$,right? – user113576 Feb 02 '14 at 19:32
  • No: $$\int_{-\pi}^\pi f(x)\cos((2,k+1),x)dx\ne0.$$ I think that my idea wasn't so good after all. Mathematica gives the sum in terms of the polylogarithm function. – Julián Aguirre Feb 02 '14 at 19:53
  • @JuliánAguirre: Mathematica in this case returns a polylog when in fact the result is much simpler. The numerical results are identical. It is one of those times when a brain-dead use of computer algebra leads to catastrophic results. – Ron Gordon Feb 03 '14 at 00:52
  • Maple also produces the polylogs. I would not have been as severe in my choice of words, though, commenting on this. BTW I have seen several of your computations with Parseval and it does seem to be the method of choice here. I would describe the polylogs as a re-write rather than an evaluation. – Marko Riedel Feb 03 '14 at 00:57
  • @MarkoRiedel: I really don't know what other words to use. How else does one explain allowing an explanation of a problem in terms of the polylogs output by computer algebra when the answer is a quadratic (or linear)? – Ron Gordon Feb 03 '14 at 01:46
  • It depends what algorithm was used to obtain these formulas. They look like what one gets from Euler's formula for the sine term, substituting that into the sum, recognizing $\zeta(2)$ and collecting the other terms. If that's how they were obtained there is not much they can teach us. If however they are the product of some other, more sophisticated algorithm, then they would have some value, supposing we can trace what that algorithm might have been. – Marko Riedel Feb 03 '14 at 02:09
  • @MarkoRiedel: I think we are speaking of two completely different things. I am not arguing that computer algorithms are worthless; that would be foolish. I am arguing, however, that the acceptance of computer output without any quality control is worthless. In this case, less than worthless. – Ron Gordon Feb 03 '14 at 13:40
1

You can have a closed form in terms of the polylogarithm function

$$\sum_{k=1}^{\infty} \frac {\sin^2(kx)}{k^2}= \frac{\pi^2}{12}-\frac{1}{4}\,{Li_2} \left( {{\rm e}^{2\,ix}} \right) -\frac{1}{4}\,{Li_2} \left( {{\rm e}^{-2ix}} \right). $$

You need to represent $\sin(kx)^2$ in terms of $e^{ikx}$ and then use the polylog function. See here.

  • I think i found it. It's the function $$f(x)=\frac {(\pi-2x)^2}{4}-\frac {\pi^2}{12}$$, that has fourier series this $$\sum_{k=1}^\infty\frac{\cos(2,k,x)}{k^2}$$ right? – user113576 Feb 02 '14 at 19:33
  • @user113576: I think you are missing the coefficients of $\sin(2kx)$, they are not $0$. – Mhenni Benghorbal Feb 02 '14 at 19:44
1

Consider the sum $$S(x) = \sum_{k\ge 1} \frac{\sin^2 (kx)}{(kx)^2}.$$

The sum term is harmonic and may be evaluated by inverting its Mellin transform.

Recall the harmonic sum identity $$\mathfrak{M}\left(\sum_{k\ge 1} \lambda_k g(\mu_k x);s\right) = \left(\sum_{k\ge 1} \frac{\lambda_k}{\mu_k^s} \right) g^*(s)$$ where $g^*(s)$ is the Mellin transform of $g(x).$

In the present case we have $$\lambda_k = 1, \quad \mu_k = k \quad \text{and} \quad g(x) = \frac{\sin^2 x}{x^2}.$$ We need the Mellin transform $g^*(s)$ of $g(x)$ which can be obtained by starting with the transform of $\sin^2(x):$ $$\int_0^\infty \sin^2(x) x^{s-1} dx = \frac{1}{2} \int_0^\infty (1-\cos(2x)) x^{s-1} dx \\= -\frac{1}{2} \int_0^\infty \left(\sum_{q\ge 1} (-1)^q \frac{2^{2q}}{(2q)!} x^{2q} \right) x^{s-1} dx = -\frac{1}{2} \int_0^\infty \left(\sum_{q\ge 1} i^{2q} \frac{2^{2q}}{(2q)!} (-x)^{2q} \right) x^{s-1} dx \\ = -\frac{1}{4} \int_0^\infty \left(\sum_{q\ge 1} i^q (1+(-1)^q) \frac{2^q}{q!} (-x)^q \right) x^{s-1} dx.$$ Now apply the Ramanujan Master Theorem to obtain the transform $$-\frac{1}{4}\Gamma(s) i^{-s} (1+(-1)^s) 2^{-s} = -\frac{1}{4} (e^{-i\pi/2 s} + e^{i\pi/2 s}) 2^{-s} = -\frac{1}{2^{s+1}} \Gamma(s) \cos(\pi/2 s).$$ This transform can also be computed directly as shown at this MSE link. Here we have used the fact that when we omit an initial segment of the expansion about zero of a function whose Mellin transform we want to compute then the effect is merely to shift the fundamental strip.

The Mellin transform $g^*(s)$ of $g(x)$ is thus given by $$-\frac{1}{2^{s-1}} \Gamma(s-2) \cos(\pi/2 s-\pi) = \frac{1}{2^{s-1}} \Gamma(s-2) \cos(\pi/2 s).$$

It follows that the Mellin transform $Q(s)$ of the harmonic sum $S(x)$ is given by $$Q(s) = \frac{1}{2^{s-1}} \Gamma(s-2) \cos(\pi/2 s) \zeta(s).$$ The Mellin inversion integral here is $$\frac{1}{2\pi i} \int_{3/2-i\infty}^{3/2+i\infty} Q(s)/x^s ds$$ which we evaluate by shifting it to the left for an expansion about zero. There is a lot of cancellation among the poles from the gamma function since the cosine term cancels the ones at the odd negative integers and the zeta term with its trivial zeros the ones at the even negative integers. This leaves just two poles with their residues to the left of $\Re(s) = 3/2$: $$\mathrm{Res}(Q(s)/x^s; s=1) = \frac{\pi}{2x} \quad\text{and}\quad \mathrm{Res}(Q(s)/x^s; s=0) = -\frac{1}{2}.$$ This gives $$S(x) = \frac{\pi}{2x} -\frac{1}{2}$$ on $(0,\pi)$ and for the original quantity we were looking for $$x^2 S(x) = \sum_{k\ge 1} \frac{\sin^2 (kx)}{k^2} = \frac{\pi}{2} x - \frac{1}{2} x^2.$$ this time for $[0,\pi).$ That was quite a computation for a simple answer but it allowed us to showcase Mellin transforms and the Ramanujan Master Theorem.

Remark. Use $$ \frac{\pi}{2} |x| - \frac{1}{2} x^2 $$ to extend this to $(-\pi,\pi).$

Marko Riedel
  • 61,317
1

Unless I am missing something, Parseval's theorem tells us that

$$\sum_{k=-\infty}^{\infty} \frac{\sin^2{k x}}{k^2} = \pi |x| \quad x \in [-\pi,\pi)$$

Then

$$\sum_{k=1}^{\infty} \frac{\sin^2{k x}}{k^2} = \frac12 |x| \, (\pi -|x|) \quad x \in [-\pi,\pi)$$

Ron Gordon
  • 138,521