How could I possibly indefinite integral: $$ \int{\arcsin\left(x\right) \over x}\,{\rm d}x $$ using elementary functions? If it is impossible to integrate using elementary functions, Is there any way to find the definite integral in the range $\left[0,\pi/2\right]$ ?. If so please teach me how to find it.
-
2An obvious thing to try is to substitute $x = \sin \theta$. Who knows if it will actually help, but it's an obvious thing to try. – Mar 07 '14 at 05:25
-
0 to $\pi/2$? do you mean $0$ to $1$ Since you have a arcsin in there? – Guy Mar 07 '14 at 05:39
-
See trigonometric integral, Liouville's theorem and Risch algorithm. – Lucian Mar 07 '14 at 05:55
-
@Lucian $\arcsin$ is undefined at $\pi/2$ though? – Guy Mar 07 '14 at 05:57
-
I assume you meant to say $[0,1]$ instead of $[0,\pi/2]$ since we're dealing with $\arcsin$ instead of $\sin$. – Lucian Mar 07 '14 at 06:01
-
@Lucian yes I assumed that too. OP made a typo probably.Or he meant to say $\sin(\pi/2)$ instead of just $\pi/2$. – Guy Mar 07 '14 at 06:08
-
1Inverse tangent integral? – Sangchul Lee Mar 08 '14 at 05:10
4 Answers
As, Hurkyl suggests, substitute $x=\sin\theta$. I am also assuming that you in fact intended the limits to be $0$ and $1$ since, $\arcsin$ is undefined for $\pi/2$. With some simple manipulations, the integral becomes
$$\int_0^{\pi/2}\frac{x\cos(x)}{\sin(x)}\,dx$$
Substitute, $t=\ln\sin(x)\implies dt=\cos(x)\,dx/\sin(x)$
$$\int_{-\infty}^0 x\,dt \text{ where $x = \arcsin(e^t)$}$$
Integrating by parts, i.e, applying $\int v\,du = uv - \int udv$
$$\lbrack x\ln(\sin x)\rbrack_0^{\pi/2} - \int_0^{\pi/2}\ln(\sin(x))dx$$
Both of these functions are elementary and solvable and left as an exercise to the reader. The answer is $\large \frac{\pi\ln(2)}{2}$
EDIT: I assumed that $\int_0^{\pi/2}\ln(\sin(x)dx$ would be familiar to most readers. I have since edited the answer to show the work here.
$$\int_0^{\pi/2}\ln(\sin(x))dx = \int_0^{\pi/2}\ln(\cos(x))dx$$
as noted in my comment. Adding the LHS to both sides,
$$2\int_0^{\pi/2}\ln(\sin(x))dx = \int_0^{\pi/2}\ln\left(\frac{\sin(2x)}{2}\right)\,dx$$
$$2I = \int_0^{\pi/2}\ln\left(\sin(2x)\right) -\ln(2)\,dx$$
$$2I = \int_0^{\pi/2}\ln\left(\sin(2x)\right)\,dx -\pi\ln(2)/2$$
$$2I = \int_0^{\pi}\frac{\ln\left(\sin(u)\right)}{2}\,du -\pi\ln(2)/2$$
By symmetry arguments, $\int_0^{\pi}\ln(\sin(x))\,dx=2\int_0^{\pi/2}\ln(\sin(x))\,dx$
$$2I = \int_0^{\pi/2}\ln\left(\sin(u)\right)\,du -\pi\ln(2)/2$$
$$2I = I - \pi\ln(2)/2$$
$$I = - \pi\ln(2)/2$$
- 8,857
- 1
- 28
- 57
-
If you are saying $\int\log(\sin x),dx$ is an elementary function, I don't think that's right. – Gerry Myerson Mar 07 '14 at 06:39
-
@GerryMyerson elementary in the sense trivial. At least to me, since I guess I have seen in it before. You seem to have a LOT of rep, so feel free to edit that bit into my answer. It can be solved using the property $$\int_0^{\pi/2}\log(\sin x),dx = \int_0^{\pi/2}\log(\cos x),dx$$ since, $$\int_0^af(x),dx = \int_0^af(a-x)dx$$ – Guy Mar 07 '14 at 06:44
-
@GerryMyerson I don't mean trivial in the usual sense. I just meant that our professor usually allows us to just right that particular integral's value instead of showing our work if it comes up somewhere in our answer. – Guy Mar 07 '14 at 06:47
-
In the worst case, you can expand the integrand as an infinite Taylor series and integrate it. – Claude Leibovici Mar 07 '14 at 06:49
-
@ClaudeLeibovici that isn't necessary at all. There is a much simpler version to do that here. – Guy Mar 07 '14 at 06:50
-
@Sabyasachi. I am addressing the problem of the integral between bounds which could differ from $0$ and $\pi/2$. – Claude Leibovici Mar 07 '14 at 06:52
-
@ClaudeLeibovici oh. But how? Isn't $\arcsin$ undefined for above x=1? I might not know about a wider definition though. – Guy Mar 07 '14 at 07:01
-
@Sabyasachi. Sorry, I was meaning $1$ and not $\pi/2$. Cheers. – Claude Leibovici Mar 07 '14 at 07:04
-
@ClaudeLeibovici oh. ok. :) yes that would be messy though. I am not willing to get into it. feel free to post it as a separate answer. – Guy Mar 07 '14 at 07:38
-
-
@GerryMyerson As I said, my usage of "trivial" wasn't the usual one, I meant to say "this is easy, give it a shot yourself" – Guy Mar 07 '14 at 07:53
-
I wouldn't even say it was easy, for someone who had never seen it before. – Gerry Myerson Mar 07 '14 at 10:30
-
@GerryMyerson agreed yes. I just left is as exercise, although after your suggestion, I edited the rest into my answer for the sake of a complete solution. – Guy Mar 07 '14 at 10:37
-
@Sabyasachi I got some trouble with the part [xln(sinx)]0 to pi/2 because I think 0ln(sin0) is an indeterminate form of 0infinity, or does 0*infinity = 0 ? – Mike Mar 09 '14 at 16:37
-
@Mike no $0\infty \ne 0$. You can take limits? In this* case it evaluates to 0 though. – Guy Mar 09 '14 at 16:46
-
You can use L'hospital here, but as a rule of thumb, ln(x) is asymptotically lower than x, so x becomes 0 "faster" than ln(x) becomes $-\infty$. – Guy Mar 09 '14 at 16:50
-
@Sabyasachi I don't get the point there, could it be evaluated by limits? – Mike Mar 09 '14 at 16:50
-
@Mike yes. You can evaluate the limit of the indeterminate form. L'hospital. – Guy Mar 09 '14 at 17:16
You can find a antiderivative in terms of the dilogarithm.
I'm going to use the fact that $$\int_{0}^{z} f(x) \cot(x) \ dx = 2 \sum_{n=1}^{\infty} \int_{0}^{z} f(x) \sin(2nx) \ dx$$
and the fact that for $0 < x < 2 \pi,$
$$ \sum_{k=1}^{\infty} \frac{\cos kx}{k} = - \log \left( 2 \sin \frac{x}{2} \right)$$
Then
$$\int_{0}^{z} \frac{\arcsin x}{x} \ dx = \int_{0}^{\arcsin z} u \cot u \ du =$$
$$= 2 \sum_{n=1}^{\infty}\int_{0}^{\arcsin z} x \sin (2nx) \ dx = 2 \sum_{n=1}^{\infty} \Big(\frac{\sin (2nx)}{4n^{2}} - \frac{z \cos (2nx)}{2n} \Big) \Big|^{\arcsin z}_{0}$$
$$ = \sum_{n=1}^{\infty}\frac{\sin(2n \arcsin z)}{2n^{2}} - \arcsin(z)\sum_{n=1}^{\infty}\frac{ \cos (2n \arcsin z)}{n} $$
$$ = \frac{1}{2} \text{Im} \ \text{Li}_{2} \left(e^{2i \arcsin z} \right) + \arcsin(z) \log(2z)$$
- 42,026
$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle #1 \right\rangle} \newcommand{\braces}[1]{\left\lbrace #1 \right\rbrace} \newcommand{\bracks}[1]{\left\lbrack #1 \right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil #1 \right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\equalby}[1]{{#1 \atop {= \atop \vphantom{\huge A}}}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left( #1 \right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert}$ With $\ds{\Lambda \in {\mathbb R}}$: \begin{align} \int_{0}^{\Lambda}{\arcsin\pars{x} \over x}\,\dd x&= \sgn\pars{\Lambda}\int_{0}^{\verts{\Lambda}}{\arcsin\pars{x} \over x}\,\dd x \\[3mm]&= \arcsin\pars{\Lambda}\ln\pars{\verts{\Lambda}} - \sgn\pars{\Lambda}\color{#f00}{\int_{0}^{\verts{\Lambda}}\ln\pars{x}\,{1 \over \root{1 - x^{2}}}\,\dd x}\tag{1} \end{align}
With $\ds{x \equiv t^{1/2}\quad\imp\quad t = x^{2}}$ \begin{align} &\color{#f00}{\int_{0}^{\verts{\Lambda}}\ln\pars{x}\,{1 \over \root{1 - x^{2}}} \,\dd x}= \int_{0}^{\Lambda^{2}}\ln\pars{t^{1/2}}\, {1 \over \root{1 - \pars{t^{1/2}}^{2}}}\,\half\,t^{-1/2}\dd t \\[3mm]&= {1 \over 4}\lim_{\mu \to -1/2}\totald{}{\mu}\int_{0}^{\Lambda^{2}}t^{\mu} \pars{1 - t}^{-1/2}\,\dd t ={1 \over 4}\lim_{\mu \to -1/2}\totald{{\rm B}_{\Lambda^{2}}\pars{\mu + 1,1/2}}{\mu} \end{align} where $\ds{{\rm B}_{z}\pars{a,b}}$ is the Incomplete Beta Function. By replacing in $\pars{1}$: $$ \color{#00f}{\int_{0}^{\Lambda}{\arcsin\pars{x} \over x}\,\dd x =\arcsin\pars{\Lambda}\ln\pars{\verts{\Lambda}} - {1 \over 4}\,\sgn\pars{\Lambda} \lim_{\mu \to -1/2}\totald{{\rm B}_{\Lambda^{2}}\pars{\mu + 1,1/2}}{\mu}} $$
For the particular cases $\ds{\Lambda = \pm 1}$, $\ds{{\rm B}_{\Lambda^{2}}\pars{\mu + 1,1/2} = {\rm B}_{1}\pars{\mu + 1,1/2} ={\rm B}\pars{\mu + 1,1/2}}$ where $\ds{{\rm B}\pars{a,b}}$ is the Beta Function. Also, $\ds{{\rm B}\pars{a,b} = \Gamma\pars{a}\Gamma\pars{b}/\Gamma\pars{a + b}}$. $\ds{\Gamma\pars{z}}$ is the Gamma Function.
Then, \begin{align} \int_{0}^{\pm 1}{\arcsin\pars{x} \over x}\,\dd x &=\mp\,{1 \over 4}\lim_{\mu \to -1/2}\totald{}{\mu}\bracks{% \Gamma\pars{\mu + 1}\Gamma\pars{1/2} \over \Gamma\pars{\mu + 3/2}} \\[3mm]&=\mp\,{1 \over 4}\,\Gamma\pars{\half}\bracks{% {\Gamma\pars{1/2}\Psi\pars{1/2} \over \Gamma\pars{1}} -{\Gamma\pars{1/2}\Psi\pars{1} \over \Gamma\pars{1}}} \\[3mm]&=\mp\,{1 \over 4}\,\Gamma^{2}\pars{\half}\bracks{\Psi\pars{\half} - \Psi\pars{1}} \end{align} $\ds{\Gamma\pars{1/2} = \root{\pi}}$. $\ds{\Psi\pars{1/2} = -\gamma - 2\ln\pars{2}.\ \Psi\pars{1} = -\gamma}$. $\ds{\Psi\pars{z}}$ and $\ds{\gamma}$ are the Digamma Function and the Euler-Mascheroni Constant, respectively. See this page. Then, $$ \color{#00f}{\large\int_{0}^{\pm 1}{\arcsin\pars{x} \over x}\,\dd x = \pm\,\half\,\pi\ln\pars{2}} $$
- 89,464
-
@SohamChowdhury Just to point out clearly the main results. Thanks. – Felix Marin Mar 08 '14 at 06:35
In a comment, I suggested to develop the integrand as an infinite Taylor expansion to be integrated later. Doing so, I obtained as a result $$\int \frac{\arcsin x}{x}\,dx=\sum _{n=0}^{\infty } \frac{4^{-n} (2 n)!}{(2 n+1)^2 (n!)^2} x^{2 n+1}$$ which has been later found to be the development of an hypergeometric function. So, as a final result, $$\int \frac{\arcsin x}{x}\,dx=x \, _3F_2\left(\frac{1}{2},\frac{1}{2},\frac{1}{2};\frac{3}{2},\frac{3}{2};x^2\right)$$ For sure, this result is not expressed on the basis of elementary functions but it is analytical.
If the hypergeometric function cannot be used for any reason, the Taylor expansion can be efficiently used since it converges very quickly. As an example $$\int_{0}^{1} \frac{\arcsin x}{x} \ dx$$ is computed iwth a relative error of less than $0.1$% adding $20$ terms and this is probably more than sufficient for most engineering calculations.
What is interesting is that this approach could be used for the calculation of
$$\int \frac{\arcsin x^a}{x^b}\,dx $$ where $a$ and $b$ could be any numbers.
- 260,315