0
$\begingroup$

I have a very direct question.

Is there an explicit solution for this integral:

$$\int_a^b e^{-\frac{x^2}{2\sigma^2}}I_0\left(\frac{x\cdot z}{\sigma^2}\right)\mathrm{d}x$$

where $z,\sigma^2\in\mathbb{R}^+$ and $I_0$ is the 0-order modified Bessel function of first kind?

Thanks in advance.

$\endgroup$
1
  • $\begingroup$ Maple has no explicit solution for $$\int_0^1 e^{-x^2}I_0(x)\;dx$$ $\endgroup$
    – GEdgar
    Commented Nov 28, 2020 at 21:53

1 Answer 1

3
$\begingroup$

$$J=\int e^{-\frac{x^2}{2\sigma^2}}\,I_0\left(\frac{x\, z}{\sigma^2}\right)\,dx$$

Let $x=\frac{\sigma ^2 y}{z}$ and $k=\frac{\sigma ^2}{2 z^2}$ to make $$J=\frac {\sigma ^2 }{z} \int e^{-k y^2}\,I_0(y)\,dy$$

Now using the infinite series $$I_0(y)=\sum_{n=0}^\infty \frac{2^{-2 n}}{(n!)^2}\, y^{2n}$$

we face a series of integrals $$J_n= \int e^{-k y^2}\,y^{2n}\,dy=-\frac{1}{2} k^{-\frac{2n+1}{2}} \Gamma \left(n+\frac{1}{2},k y^2\right)$$

For example, let $z=\frac 12$ and $\sigma=\frac 1{\sqrt 2}$ to make the simple $$J=\int_0^1 e^{-x^2}\, I_0(x)\,dx=\sum_{n=0}^\infty \frac{ \Gamma \left(n+\frac{1}{2},0\right)-\Gamma \left(n+\frac{1}{2},1\right)}{2^{2 n+1}\,(n!)^2}$$ which converge very fast. For the partial sums up to $n=p$ $$\left( \begin{array}{cc} p & \text{approximation} \\ 0 & 0.7468241328 \\ 1 & 0.7941922193 \\ 2 & 0.7957589192 \\ 3 & 0.7957878829 \\ 4 & 0.7957882194 \\ 5 & 0.7957882221 \\ 6 & 0.7957882221 \end{array} \right)$$

$\endgroup$
2
  • $\begingroup$ Thank you very much. Your second formula has a $\sigma^2 / z$, instead of $1/z$, am I wrong? $\endgroup$
    – Nameless
    Commented Nov 29, 2020 at 9:50
  • 1
    $\begingroup$ @Nameless. Good catch ! I fixed it. Cheers :-) $\endgroup$ Commented Nov 29, 2020 at 10:07

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .