I'm trying to evaluate the integral $$\int \sqrt{\frac{x^2-a^2}{1+x^2}}\,\mathrm{d}x$$ ($a>0$) in terms of an elliptic integral of the second kind, with modulus $\kappa =1/\sqrt{1+a^2}$. I have tried the usual changes of variable $x=\tan t$, $x=\cot t$, and they lead almost to the canonical form, but with $\sin x$, $\cos x$ in the denominator. Is there any other change of variables worth trying?
-
I'm wondering if the manipulation to $1-\frac{1+a^2}{1+x^2}$ produces anything meaningful? – abiessu Jan 01 '25 at 23:24
-
4Substitute $$x=i \sin y$$ – Roland F Jan 01 '25 at 23:25
-
This leaves me with $1+\frac{1}{a^2}\sin^2 y$ in the integrand, with the wrong sign in the sine. – Grimolatto Jan 02 '25 at 16:18
-
$1+\frac{1}{a^2}\sin^2(y)=1-\frac{1}{(ia)^2}\sin^2(y)$ – Quý Nhân Jan 02 '25 at 16:23
-
Wolfram Mathematica gives E(i ArcSinh(x), i/a)a. – Somos Jan 02 '25 at 17:27
-
Yes, the complex change of variable leads to something in the spirit of Wolfram's Mma, but for physical reasons I need the modulus to be real. – Grimolatto Jan 02 '25 at 19:05
2 Answers
Denote $k=\sqrt{1+a^2}$. For $t>a$,
$$\newcommand{\arccot}{\operatorname{arccot}} \begin{align*} I(t) &= \int_a^t \frac{\sqrt{x^2-a^2}}{\sqrt{1+x^2}} \, dx \\ &= t \frac{\sqrt{t^2-a^2}}{\sqrt{1+t^2}} - k^2 \int_a^t \frac{x^2\,dx}{\sqrt{x^2-a^2}\left(1+x^2\right)^{3/2}} & \text{by parts} \\ &= t \frac{\sqrt{t^2-a^2}}{\sqrt{1+t^2}} + k^2 \int_{\operatorname{arccot}a}^{\operatorname{arccot}t} \frac{\csc y-\sin y}{\sqrt{\cot^2y-a^2}} \, dy & x=\cot y \\ &= t \frac{\sqrt{t^2-a^2}}{\sqrt{1+t^2}} - \int_{\operatorname{arccot}t}^{\operatorname{arccot}a} \sqrt{1-k^2\sin^2y} \, dy \\ &\qquad + \left(1-k^2\right) \int_{\operatorname{arccot}t}^{\operatorname{arccot}a} \frac{dy}{\sqrt{1-k^2\sin^2y}} \\ &= t \frac{\sqrt{t^2-a^2}}{\sqrt{1+t^2}} + E\left(\arccot t,k\right) - E\left(\arccot a,k\right) \\ &\qquad + a^2 \left[F\left(\arccot t,k\right) - F\left(\arccot a,k\right)\right] \end{align*}$$
Using the identities listed here we can rewrite $I(t)$ in terms of $\kappa=\dfrac1k$:
$$\begin{align*} F(\arccot t,k) &= \frac1{\sqrt k} F\left(\arcsin\sqrt{\frac k{1+t^2}},\frac1k\right) \\[2ex] E(\arccot t,k) &= \sqrt k \, E\left(\arcsin\sqrt{\frac k{1+t^2}}, \frac1k\right) + \frac{1-k}{\sqrt k} F\left(\arcsin\sqrt{\frac k{1+t^2}}, \frac1k\right) \end{align*}$$
- 25,320
-
-
-
@user170231 It would be nice if we could transform the elliptic integrals into ones with modulus $0<k<1$. If the reason we're avoiding complex moduli is because they're "unphysical", I'm not sure a modulus greater than unity is much better. – David H Jan 28 '25 at 09:06
-
@DavidH Agreed. Fortunately, B&F cover this case, according to this answer. I'll update my answer as soon as I iron out the details. – user170231 Jan 28 '25 at 16:25
Define the function $\mathcal{J}$ via the elliptic integral,
$$\mathcal{J}{\left(a,b,z\right)}:=\int_{a}^{z}\mathrm{d}t\,\sqrt{\frac{t^{2}-a^{2}}{t^{2}+b^{2}}};~~~\small{z>a>0\land b>0}.$$
For $z>a>0\land b>0$, we find
$$\begin{align} \mathcal{J}{\left(a,b,z\right)} &=\int_{a}^{z}\mathrm{d}t\,\sqrt{\frac{t^{2}-a^{2}}{t^{2}+b^{2}}}\\ &=\int_{\frac{1}{z}}^{\frac{1}{a}}\mathrm{d}u\,\frac{1}{u^{2}}\sqrt{\frac{1-a^{2}u^{2}}{1+b^{2}u^{2}}};~~~\small{\left[t=\frac{1}{u}\right]}\\ &=\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{a^{2}}{v^{2}}\sqrt{\frac{1-v^{2}}{a^{2}+b^{2}v^{2}}};~~~\small{\left[au=v\right]}\\ &=\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{a^{2}-a^{2}v^{2}}{v^{2}\sqrt{(1-v^{2})(a^{2}+b^{2}v^{2})}}\\ &=\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{a^{2}+b^{2}v^{4}-b^{2}v^{4}-a^{2}v^{2}}{v^{2}\sqrt{(1-v^{2})(a^{2}+b^{2}v^{2})}}\\ &=\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{a^{2}+b^{2}v^{4}}{v^{2}\sqrt{(1-v^{2})(a^{2}+b^{2}v^{2})}}-\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{a^{2}+b^{2}v^{2}}{\sqrt{(1-v^{2})(a^{2}+b^{2}v^{2})}}\\ &=\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{d}{dv}\left[-\frac{\sqrt{(1-v^{2})(a^{2}+b^{2}v^{2})}}{v}\right]-\int_{\frac{a}{z}}^{1}\mathrm{d}v\,\frac{\sqrt{a^{2}+b^{2}v^{2}}}{\sqrt{1-v^{2}}}\\ &=\frac{z}{a}\sqrt{\left(1-\frac{a^{2}}{z^{2}}\right)\left(a^{2}+\frac{a^{2}b^{2}}{z^{2}}\right)}\\ &~~~~~-\int_{\arcsin{\left(\frac{a}{z}\right)}}^{\arcsin{\left(1\right)}}\mathrm{d}\theta\,\frac{\cos{\left(\theta\right)}\sqrt{a^{2}+b^{2}\sin^{2}{\left(\theta\right)}}}{\sqrt{1-\sin^{2}{\left(\theta\right)}}};~~~\small{\left[\arcsin{\left(v\right)}=\theta\right]}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}-\int_{\arcsin{\left(\frac{a}{z}\right)}}^{\frac{\pi}{2}}\mathrm{d}\theta\,\sqrt{a^{2}+b^{2}\sin^{2}{\left(\theta\right)}}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}\\ &~~~~~-\int_{0}^{\frac{\pi}{2}-\arcsin{\left(\frac{a}{z}\right)}}\mathrm{d}\varphi\,\sqrt{a^{2}+b^{2}\sin^{2}{\left(\frac{\pi}{2}-\varphi\right)}};~~~\small{\left[\theta=\frac{\pi}{2}-\varphi\right]}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}-\int_{0}^{\arccos{\left(\frac{a}{z}\right)}}\mathrm{d}\varphi\,\sqrt{a^{2}+b^{2}\cos^{2}{\left(\varphi\right)}}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}-\int_{0}^{\arccos{\left(\frac{a}{z}\right)}}\mathrm{d}\varphi\,\sqrt{a^{2}+b^{2}-b^{2}\sin^{2}{\left(\varphi\right)}}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}-\sqrt{a^{2}+b^{2}}\int_{0}^{\arccos{\left(\frac{a}{z}\right)}}\mathrm{d}\varphi\,\sqrt{1-\frac{b^{2}}{a^{2}+b^{2}}\sin^{2}{\left(\varphi\right)}}\\ &=\frac{\sqrt{\left(z^{2}-a^{2}\right)\left(z^{2}+b^{2}\right)}}{z}-\sqrt{a^{2}+b^{2}} E{\left(\arccos{\left(\frac{a}{z}\right)},\frac{b}{\sqrt{a^{2}+b^{2}}}\right)},\\ \end{align}$$
where the incomplete elliptic integral of the second kind $E$ is given here by
$$E{\left(\alpha,\kappa\right)}:=\int_{0}^{\alpha}\mathrm{d}\varphi\,\sqrt{1-\kappa^{2}\sin^{2}{\left(\varphi\right)}};~~~\small{0<\kappa<1\land0<\alpha\le\frac{\pi}{2}}.$$
- 32,536
-
(+1) Nice and compact! I guess the connection between our results comes down to some version of the identity listed at the bottom of the "incomplete EI of second kind" section but I'm not immediately seeing it. – user170231 Jan 29 '25 at 17:58