# Is it possible to evaluate analytically the following nontrivial triple integral?

In a physical mathematical problem, I came across a nontrivial triple integral below obtained upon 3D inverse Fourier transformation. It would be great if someone here could provide with some hints that could help to evaluate analytically $$I =\frac{1}{(2\pi)^3} \int_0^{2\pi} \int_0^\infty \int_0^{\pi} \frac{\sin\theta \sin^2\phi}{a \cos^2 \theta + b \sin^2 \theta} \, e^{ikh\sin\theta\cos\phi} \, \mathrm{d} \theta \, \mathrm{d} k \, \mathrm{d} \phi \, .$$ where $a$ and $b$ are two positive real numbers. Using the change of variable $q=\cos\theta$, the latter equation can be written as $$I = \frac{1}{(2\pi)^3} \int_0^{2\pi} \int_0^\infty \int_{-1}^{1} \frac{\sin^2\phi}{(a-b)q^2+b} \, e^{ikh\sqrt{1-q^2}\cos\phi} \, \mathrm{d} q \, \mathrm{d} k \, \mathrm{d} \phi \, .$$

A guess solution using Maple for some numerical values for $a$ and $b$ is obtained as $$I = \frac{1}{4\pi h \sqrt{ab}} \, .$$

But, is there a way to prove that really? Any idea / feedback is welcome.

Thanks a lot!

Fede

Well, for the first integral we have:

$$\mathscr{I}:=\int_0^{2\pi}\int_0^\infty\int_0^\pi\frac{\sin\left(\theta\right)\cdot\sin^2\left(\phi\right)\cdot\exp\left(\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{k}\cdot\text{h}\cdot i\right)}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\space\text{d}\theta\space\text{d}\text{k}\space\text{d}\phi\tag1$$

Using Fubini's theorem:

$$\mathscr{I}:=\int_0^{2\pi}\int_0^\pi\int_0^\infty\frac{\sin\left(\theta\right)\cdot\sin^2\left(\phi\right)\cdot\exp\left(\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{k}\cdot\text{h}\cdot i\right)}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\space\text{d}\text{k}\space\text{d}\theta\space\text{d}\phi=$$ $$\int_0^{2\pi}\int_0^\pi\frac{\sin\left(\theta\right)\cdot\sin^2\left(\phi\right)}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\int_0^\infty\exp\left(\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{k}\cdot\text{h}\cdot i\right)\space\text{d}\text{k}\space\text{d}\theta\space\text{d}\phi\tag2$$

Now, we can use:

$$\int_0^\infty\exp\left(\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{k}\cdot\text{h}\cdot i\right)\space\text{d}\text{k}=\frac{i}{\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{h}}\tag3$$

when $\Im\left(\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{h}\right)>0$

So, we get:

$$\mathscr{I}=\int_0^{2\pi}\int_0^\pi\frac{\sin\left(\theta\right)\cdot\sin^2\left(\phi\right)}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\cdot\frac{i}{\sin\left(\theta\right)\cdot\cos\left(\phi\right)\cdot\text{h}}\space\text{d}\theta\space\text{d}\phi=$$ $$\int_0^{2\pi}\frac{\sin^2\left(\phi\right)\cdot i}{\cos\left(\phi\right)\cdot\text{h}}\int_0^\pi\frac{\sin\left(\theta\right)}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\cdot\frac{1}{\sin\left(\theta\right)}\space\text{d}\theta\space\text{d}\phi=$$ $$\int_0^{2\pi}\frac{\sin^2\left(\phi\right)\cdot i}{\cos\left(\phi\right)\cdot\text{h}}\int_0^\pi\frac{1}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\space\text{d}\theta\space\text{d}\phi\tag4$$

Now, use:

$$\int_0^\pi\frac{1}{\text{a}\cdot\cos^2\left(\theta\right)+\text{b}\cdot\sin^2\left(\theta\right)}\space\text{d}\theta=\frac{\pi}{\sqrt{\text{a}\cdot\text{b}}}\tag5$$

So, we get:

$$\mathscr{I}=\int_0^{2\pi}\frac{\sin^2\left(\phi\right)\cdot i}{\cos\left(\phi\right)\cdot\text{h}}\cdot\frac{\pi}{\sqrt{\text{a}\cdot\text{b}}}\space\text{d}\phi=\frac{\pi\cdot i}{\text{h}\cdot\sqrt{\text{a}\cdot\text{b}}}\int_0^{2\pi}\frac{\sin^2\left(\phi\right)}{\cos\left(\phi\right)}\space\text{d}\phi\tag6$$

• Thanks for your helpful answer. But PV of the last integral is identically null. Does it mean that $I= 0$ ? thanks – Math Student Jul 9 '17 at 13:26
• I think that it shouldn't. The integral has a physical interpretation and should be strictly positive defined. The guess solution has been proved to be true using Maple but I don't know how to prove it. Indeed, your solution shows the desired $h\sqrt{ab}$ in the denominator. I think that the correct solution should not be that far... – Math Student Jul 9 '17 at 14:10
• Yes. I should prove it. It is part of my doctoral research.. – Math Student Jul 9 '17 at 14:15
• The integral is obtained when I am working on particle motion in an anisotropic medium (physics) you may not be interested in that matter.. It would be cool if you could please be of help thanks – Math Student Jul 9 '17 at 14:19
• @Federiko Notice that you've to find the integral in the right order, from the inside one to the ouside one not the other way around! – Jan Jul 9 '17 at 20:01