3
$\begingroup$

Inspired by this post where the value of $\int_0^{\infty}\frac{\sin(\tan(x))}{x}\,dx$ was found to be $\frac{\pi}{2}(1-e^{-1})$, I set out to do the same thing with $\int_0^{\infty}\frac{\sin(\sin(x))}{x}\,dx$. Convergence is slow, which makes numerical estimation difficult, but after coaxing Mathematica for a while, I got:

 NIntegrate[Sin[Sin[x]]/x, {x, 0, 20000 Pi}, MaxRecursion -> 20, WorkingPrecision -> 20, Method -> "DoubleExponential"]
 1.4446949333948902084

My method is largely similar and currently I have achieved a Pyrrhic victory: I got down to an integral Mathematica was able to evaluate, but I don't see how to evaluate the integral myself.


My approach was largely similar to the linked post: use periodicity and a series expansion using reciprocals to rewrite the integrand. $$ \int_0^{\infty} \frac{\sin(\sin(x))}{x}\,dx = \frac{1}{2}\int_{-\infty}^{\infty} \frac{\sin(\sin(x))}{x}\,dx $$ $$ =\frac{1}{2}\sum_{n=-\infty}^{\infty} \int_{n\pi}^{(n+1)\pi} \frac{\sin(\sin(x))}{x}\,dx $$Now substitute $x=z+n\pi$: $$ =\frac{1}{2}\sum_{n=-\infty}^{\infty} \int_{0}^{\pi} \frac{\sin(\sin(z+n\pi))}{z+n\pi}\,dz $$ $$ =\frac{1}{2}\sum_{n=-\infty}^{\infty} (-1)^n \int_{0}^{\pi} \frac{\sin(\sin(z))}{z+n\pi}\,dz $$Swap the sum and integral and use the series representation for cosecant: $$ =\frac{1}{2} \int_{0}^{\pi} \sin(\sin(z))\sum_{n=-\infty}^{\infty} \frac{(-1)^n} {z+n\pi}\,dz $$ $$ =\frac{1}{2} \int_{0}^{\pi} \sin(\sin(z))\csc(z)\,dz=\int_{0}^{\pi/2} \sin(\sin(z))\csc(z)\,dz, $$where the last inequality is by symmetry. Now I substituted $\sin(z)=y$ which leads to the integral in the title: $$ = \int_0^1 \frac{\sin(y)}{y\sqrt{1-y^2}}\,dy $$(Note: at this point in the first linked post, the substitution is much nicer because the Pythagorean identity gives us a plus instead of a minus.) Now Mathematica cooperated: it tells me this integral is equal to $$ \frac{1}{4} \pi ^2 \pmb{H}_0(1) J_1(1)-\frac{1}{4} \pi (\pi \pmb{H}_1(1)-2) J_0(1) \approx 1.4447091498105593077; $$here $J_a$ and $\pmb{H}_a$ are the Bessel and Struve functions, respectively.


My question: I would appreciate if someone could explain how this last integral was evaluated (it was 'known' in a way the original wasn't). I tried a series expansion using the Cauchy product for $\sin(y)/y$ and $(1-y^2)^{-1/2}$ but couldn't quite get a hold of the coefficients. If by some miracle the closed-form could be simplified a bit, that would be good as well.

$\endgroup$
5
$\begingroup$

$\newcommand{\bbx}[1]{\,\bbox[15px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\on}[1]{\operatorname{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ \begin{align} &\bbox[5px,#ffd]{\int_{0}^{1}{\sin\pars{y} \over y\root{1 - y^{2}}}\,\dd y} \\ = &\ \int_{0}^{1}{1 \over \root{1 - y^{2}}}\ \overbrace{\pars{\int_{0}^{1} \cos\pars{ky}\,\dd k}}^{\ds{\sin\pars{y} \over y}}\ \,\dd y \\[5mm] = &\ \int_{0}^{1}\int_{0}^{1} {\cos\pars{ky} \over \root{1 - y^{2}}}\,\dd y\,\dd k = {\pi \over 2}\int_{0}^{1}\on{J}_{0}\pars{k}\,\dd k \end{align} where I used a Bessel $\ds{\on{J}_{\nu}}$ Integral Representation. $\ds{\on{\bf H}_{\nu}}$ is a Struve Function. \begin{align} &\bbox[5px,#ffd]{\int_{0}^{1}{\sin\pars{y} \over y\root{1 - y^{2}}}\,\dd y} \\[5mm] = &\ {\pi \over 2}\,\on{J}_{0}\pars{1} + {\pi^{2} \over 4}\on{J}_{1}\pars{1} \on{\bf H}_{0}\pars{1} - {\pi^{2} \over 4} \on{J}_{0}\pars{1}\on{\bf H}_{1}\pars{1} \\[5mm] = &\ 1.44470914981055930772056106554\ldots \end{align} The last result is given by Formula $\ds{{\bf 8}.}$, page $660$, of Table of Integrals, Series and Products ( seventh edition ) by I. S. Gradshteyn and I. M. Ryzhik.

I "guess" those integrations use somehow the generating functions.

$\endgroup$
4
  • $\begingroup$ Very nice solution ! $\endgroup$ – Claude Leibovici Oct 24 '20 at 5:11
  • $\begingroup$ @ClaudeLeibovici Thanks. $\endgroup$ – Felix Marin Oct 24 '20 at 5:14
  • $\begingroup$ Unrelated: I've seen Gradshteyn & Ryzhik mentioned several times on this website. In your opinion, is it worth buying a copy? $\endgroup$ – FearfulSymmetry Oct 24 '20 at 19:07
  • 1
    $\begingroup$ @Integrand There's a $\displaystyle{\tt PDF}$ copy in Internet Archive. $\endgroup$ – Felix Marin Oct 24 '20 at 19:15
1
$\begingroup$

This is not an answer.

To me, this is one more mistery of CAS (I had a few of these in the last thirty years which I still do not understand).

May be, a part of possible explanation.

Using $$\frac{1}{y \sqrt{1-y^2}}=\sum_{n=0}^\infty (-1)^n \binom{-\frac{1}{2}}{n} y^{2 n-1}$$ we face the problem of $$I_n=\int_0^1 y^{2n-1}\sin(y)\,dy=\frac{\, _1F_2\left(n+\frac{1}{2};\frac{3}{2},n+\frac{3}{2};-\frac{1}{4}\right)}{2n+1}$$ the first expansions of the hypergeometric functions are given below as linear combinations of Bessel functions of the first kind $$\left( \begin{array}{cc} n & \sqrt{\frac 2{\pi }}\, I_n \\ 0 & \sqrt{\frac{2}{\pi }} \text{Si}(1) \\ 1 & J_{\frac{3}{2}}(1) \\ 2 & 3 J_{\frac{5}{2}}(1)-J_{\frac{7}{2}}(1) \\ 3 & 14 J_{\frac{7}{2}}(1)-J_{\frac{9}{2}}(1) \\ 4 & 97 J_{\frac{9}{2}}(1)-16 J_{\frac{11}{2}}(1) \\ 5 & 853 J_{\frac{11}{2}}(1)-45 J_{\frac{13}{2}}(1) \\ 6 & 9330 J_{\frac{13}{2}}(1)-1007 J_{\frac{15}{2}}(1) \end{array} \right)$$ that is to say $$I_n=\sqrt{\frac{\pi }{2}}\left(a_n J_{\frac{2n+1}{2}}(1)-b_n J_{\frac{2n+3}{2}}(1) \right)$$ But the $I_n$ simplify in terms of linear combinations of $\sin(1)$ and $\cos(1)$ $$\left( \begin{array}{cc} n & I_n \\ 1 & -\cos (1)+\sin (1) \\ 2 & 5 \cos (1)-3 \sin (1) \\ 3 & -101 \cos (1)+65 \sin (1) \\ 4 & 4241 \cos (1)-2723 \sin (1) \\ 5 & -305353 \cos (1)+196065 \sin (1) \\ 6 & 33588829 \cos (1)-21567139 \sin (1) \end{array} \right)$$

Considering now $$S_p=\text{Si}(1)+\sum_{n=1}^p (-1)^n \binom{-\frac{1}{2}}{n}\int_0^1 y^{2 n-1}\sin(y)\,dy$$ $$S_6=\text{Si}(1)+$$ $$\sqrt{\frac{\pi }{2}}\left(\frac{J_{\frac{3}{2}}(1)}{2}+\frac{9 J_{\frac{5}{2}}(1)}{8}+4 J_{\frac{7}{2}}(1)+\frac{3355 J_{\frac{9}{2}}(1)}{128}+\frac{52619 J_{\frac{11}{2}}(1)}{256}+\frac{1071945 J_{\frac{13}{2}}(1)}{512}-\frac{232617 J_{\frac{15}{2}}(1)}{1024} \right)$$ that is to say $$S_6=\text{Si}(1)+\frac{7 (1097603873 \cos (1)-704763287 \sin (1))}{1024}$$

$\endgroup$
2
  • $\begingroup$ The mystery (misery?) of CAS is one I know all too well... $\endgroup$ – FearfulSymmetry Oct 24 '20 at 19:08
  • 1
    $\begingroup$ @Integrand. The word I created on purpose in the combination of both misery and mystery and not a typo. Cheers :-) $\endgroup$ – Claude Leibovici Oct 26 '20 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.