Is there any formula for calculating the following definite integral, including exponential and Bessel function?

$$ \int_0^{a}x^{-1} e^{x}I_2(bx)dx$$

Thanks in advance

  • $\begingroup$ For $b=1$ Alpha returns a closed form for the indefinite integral ($x\mapsto a$) : $$\frac 12-e^x\;I_0(x)+\frac{1+x}x\;e^x\;I_1(x)$$ for $b=-1$ this becomes $$\frac 12-e^{-x}\;I_0(x)+\frac{1-x}x\;e^{-x}\;I_1(x)$$ Such kind of integrals are handled in Luke Y.L. book of 1962 'Integrals of Bessel functions) $\endgroup$ – Raymond Manzoni May 13 '13 at 13:24
  • $\begingroup$ Thank you Raymond for your kind and swift reply, yet, I am looking for the integral using a general beta parameter. $\endgroup$ – dioxen May 13 '13 at 13:27
  • $\begingroup$ In the book I saw only the cases $b=\pm 1$ from a quick look (with the parameter $b$ transposed to the exponential) so no sure that a general solution in closed form exists... $\endgroup$ – Raymond Manzoni May 13 '13 at 13:31

In fact it mainly plays tricks by those formulae in http://www.efunda.com/math/bessel/modifiedbessel.cfm.

For $\int\dfrac{e^xI_2(bx)}{x}dx$ ,


$=\int e^x~d\left(\dfrac{I_1(bx)}{bx}\right)$



$=\dfrac{e^xI_1(bx)}{bx}+\int e^x\dfrac{d}{dx}(I_1(bx))~dx-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+\int e^x~d(I_1(bx))-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\int I_1(bx)~d(e^x)-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\int e^xI_1(bx)~dx-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\dfrac{1}{b}\int e^x~d(I_0(bx))-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\dfrac{e^xI_0(bx)}{b}+\dfrac{1}{b}\int I_0(bx)~d(e^x)-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\dfrac{e^xI_0(bx)}{b}+\dfrac{1}{b}\int e^xI_0(bx)~dx-\int e^xI_0(bx)~dx$

$=\dfrac{e^xI_1(bx)}{bx}+e^xI_1(bx)-\dfrac{e^xI_0(bx)}{b}+\dfrac{1-b}{b}\int e^xI_0(bx)~dx$

$\therefore$ Wolfram Alpha can get the close-form when $b=1$ .

When $b=-1$ , $\because I_2(-x)=I_2(x)$ , $\therefore$ Wolfram Alpha can get the close-form when $b=-1$ .

When $b\neq\pm1$ , the term of $\int e^xI_0(bx)~dx$ should be left.

In fact when $b\neq\pm1$ , $\int e^xI_0(bx)~dx$ can only be solved by this approach:

$\int e^xI_0(bx)~dx$

$=\int e^x\sum\limits_{n=0}^\infty\dfrac{\left(\dfrac{bx}{2}\right)^{2n}}{(n!)^2}dx$


$=\sum\limits_{n=0}^\infty\sum\limits_{k=0}^{2n}\dfrac{(-1)^kb^{2n}(2n)!x^ke^x}{4^n(n!)^2k!}+C$ (can be obtained from http://en.wikipedia.org/wiki/List_of_integrals_of_exponential_functions)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.