find the value: $$\int^\infty_0 \frac{\sin x}{x(1+\cos ^2 x)}dx$$ I try to integrate by parts: $$-\int^\infty_0 \frac{1}{x}d\arctan(\cos x)$$ But it's not run,help me,thank you.
-
Duplicate. – metamorphy Nov 30 '22 at 06:16
3 Answers
Observe that \begin{align} I=&\ \int^\infty_0 \frac{\sin x}{2x(1-\frac{1}{2}\sin^2x)}\ dx = \frac{1}{4}\int^\infty_{-\infty} \frac{\sin x}{x}\sum^\infty_{k=0}\frac{\sin^{2k}x}{2^k}\ dx\\ =&\ \frac{1}{4}\sum^\infty_{k=0}\frac{1}{2^k}\int^\infty_{-\infty} \frac{\sin^{2k+1}x}{x}\ dx \end{align} If you believe that \begin{align} \int^\infty_{-\infty} \frac{\sin^{2k+1}x}{x}\ dx = \binom{2k}{k}\frac{\pi}{4^k} \end{align}
then it follows that
\begin{align} I=\frac{\pi}{4}\sum^\infty_{k=0}\binom{2k}{k}\frac{1}{8^k} = \frac{\sqrt{2}\pi}{4} = \frac{\pi}{2\sqrt{2}}. \end{align}
Note that the last line follows from the series expansion \begin{align} \frac{1}{\sqrt{1-x}}=\sum^\infty_{k=0}\binom{2k}{k}\frac{x^k}{4^k}. \end{align}

- 25,739
-
1This is very nice. However, I noted that this does not agree exactly with the numerical result given by WolframAlpha which is $1.12581$, whereas the number $\pi/(2\sqrt2)=1.1107...$ – am301 Feb 05 '21 at 09:14
-
Could be just a small numerical error of WA. I tried also numerical integration with MATLAB and got a message that it does not converge. Indeed the integrand is an oscillating function. – am301 Feb 05 '21 at 09:38
-
It can be computed as a Lobachevsky-type integral (see also [1] and [2] here on MSE): $$\int_0^\infty\frac{\sin x\,dx}{x(1+\cos^2 x)}=\int_0^{\pi/2}\frac{dx}{1+\cos^2 x}\underset{\tan x=y}{\phantom{\big[}\quad=\quad\phantom{\big]}}\int_0^\infty\frac{dy}{2+y^2}=\frac{\pi}{2\sqrt2}.$$

- 39,111
Too long for a comment
Introduce a parameter $a$:
$$\bbox[5px,#ffd]{I(a)=\int \frac{\sin ax}{x(1+\cos ^2 x)}dx}$$
$$I'(a)=\int \frac{\cos ax}{(1+\cos ^2 x)}dx$$
$$I''(a)=\int \frac{-x\sin ax}{(1+\cos ^2 x)}dx=-xI(a)+\int I(a)$$
$$I'''(a)=-xI'(a)+I(a)$$
The solution of this differential equation is this.
Note that $c=-x,y=I(a)$ and $x=a$

- 4,055
-
How would you get an exact solution though, I believe your initial conditions would be hard to calculate or undefined... – Henry Lee Feb 05 '21 at 23:20
-
@HenryLee we can obtain the solution by substituting $a=1$ in that huge expression and then applying the limits from $0$ to $\infty$ – DatBoi Feb 06 '21 at 01:39
-
Would you at least check the result? (Not even talking about the deduction... Nonsense, beginning with $I''(a)=-xI(a)+\ldots$) – metamorphy Feb 06 '21 at 07:53
-
@metamorphy As I have stated before, I don't know how to solve that differential equation by hand. Wolframalpha did that, but it doesn't let me substitute the limits. So I leave it as it is. – DatBoi Feb 06 '21 at 12:44