I found a double integral involving a Dirac distribution of a sine function,
$\int_{-1}^{1} \Big( \int_{0}^{2\pi} g(\theta,\epsilon)\delta(\epsilon-\frac{1}{3}\sin\theta)d\theta\Big)f(\epsilon)d\epsilon$
(both $g$ and $f$ are continuous and differentiable in the integration domain)
I have doubts about:
- the order in which the variables can be integrated
- the Dirac delta having more than one root
Integrating $\epsilon$ first: If one 'moves' $f(\epsilon)d\epsilon$ inside the $\theta$ integral and then considers $\delta$ as a function of $\epsilon$ and solves the $\epsilon$ integral just by setting $\epsilon= \frac{1}{3}\sin\theta$ then this results in the $\theta$ integral of both $g$ and $f$ as functions of theta only. Notice $f$ was outside the $\theta$ integral at the bigenning (as a function of $\epsilon$ only), but now we have $f(\frac{1}{3}\sin\theta)$ and it has to be integrated respect to $\theta$ too.
Integrating $\theta$ first: What would be the correct way to do this given that the argument of $\delta$ (seen as a function of $\theta$) has more than one root? (when $-\frac{1}{3}<\epsilon<\frac{1}{3}$)
For a Dirac delta of the form $\delta(g(x))$ with $g$ having simple roots $x_i$ in the integration domain:
$\delta(g(x))= \sum_i \frac{\delta(x-x_i)}{|g'(x_i)|}$
But it is not very clear to me how to use this and get to the same answer one gets by doing the $\epsilon$ integral first
Aside: The specific forms of $f$ and $g$ are $g(\theta,\epsilon)=\cos^2\theta \frac{1-\frac{1}{3}\epsilon\sin \theta}{1-(\frac{1}{3}\sin \theta)^2}$ and $f(\epsilon)=\frac{1}{a+\cosh\epsilon}$