How to approximate this integral? I wonder how to find an expression for approximating:
$$
f(x)=\int_{x}^\infty  \frac{x}{\sigma a^2 {\sqrt {2\pi (a^2-x^2)}}  } e^{-\frac{(\log(a)- \mu )^2}{2\sigma^2}}\mathrm{d}a
$$ 
I don't know if it can be expressed in terms elementary functions exactly. In what direction or study topics do I need to approximate it?
My further aim is to deal with the variable $x$ with the probability distribution function $a(x-b)f(x)$ where $a$, $b$ are positive constants. I would then estimate $\sigma$ and $\mu$ from measurements (of some real geological objects).I am in the middle of undergraduate mathematical studies.  
 A: I would first make the change of variables $a=x\tau$, resulting in
$$
f(x)=\int_{x}^\infty  \frac{x}{\sigma a^2 {\sqrt {2\pi (a^2-x^2)}}  } e^{-\frac{(\log(a)- \mu )^2}{2\sigma^2}}\mathrm{d}a=
$$
$$\frac{1}{\sigma |x|\sqrt{2\pi}}\int_1^\infty\frac{\mathrm{d}\tau}{\tau^2\sqrt{\tau^2-1}}\exp\left[-\frac{(\log\tau+\log x-\mu)^2}{2\sigma^2}\right]\ .
$$
Next, I would call $\log\tau=z$, resulting in
$$
\frac{1}{\sigma |x|\sqrt{2\pi}}\int_0^\infty\frac{\mathrm{d}z}{e^z\sqrt{e^{2z}-1}}\exp\left[-\frac{(z+\log x-\mu)^2}{2\sigma^2}\right]\ .
$$
This integral still cannot be expressed in closed form. However, a systematic approximation scheme can be constructed by a series expansion
$$
\frac{1}{\sqrt{X-1}}\simeq \frac{1}{\sqrt{X}}+\frac{1}{2}\frac{1}{X^{3/2}}+\frac{3}{8}\frac{1}{X^{5/2}}+\ldots\ ,
$$
applied to $X=e^{2z}$. Therefore the integral (with $A=\mu-\log x$)
$$
\int_0^\infty\frac{\mathrm{d}z}{e^z\sqrt{e^{2z}-1}}\exp\left[-\frac{(z-A)^2}{2\sigma^2}\right]\simeq \int_0^\infty\frac{\mathrm{d}z}{e^z\times e^z}\exp\left[-\frac{(z-A)^2}{2\sigma^2}\right]+$$
$$\frac{1}{2}\int_0^\infty\frac{\mathrm{d}z}{e^z\times e^{3z}}\exp\left[-\frac{(z-A)^2}{2\sigma^2}\right]+\frac{3}{8}\int_0^\infty\frac{\mathrm{d}z}{e^z\times e^{5z}}\exp\left[-\frac{(z-A)^2}{2\sigma^2}\right]+\ldots\ ,
$$
Each of these integrals can now be expressed in closed form, and the approximation can be made as good as you wish by including more and more terms. For example, including the first three terms as above, the approximate expression reads
$$
f(x)\simeq \frac{1}{\sigma |x|\sqrt{2\pi}}\left[\frac{1}{8} \sqrt{\frac{\pi }{2}} \sigma  e^{2 \sigma ^2-6 A} \left(8 e^{4 A} \left(\text{erf}\left(\frac{A-2 \sigma ^2}{\sqrt{2} \sigma }\right)+1\right)+4 e^{2 A+6 \sigma ^2} \left(\text{erf}\left(\frac{A-4 \sigma ^2}{\sqrt{2} \sigma }\right)+1\right)+3 e^{16 \sigma ^2} \left(\text{erf}\left(\frac{A-6 \sigma ^2}{\sqrt{2} \sigma }\right)+1\right)\right)\right]\ ,
$$
where $\mathrm{erf}$ is the Error function.
