Evaluate: $$\int\limits_{x=0}^\infty e^{-\alpha x^2}I_0(x) \ln(I_0(x))x \, dx$$

where $I_0(x)$ is the modified Bessel function of the first kind and zeroth order, and $\alpha>0$.

I can find upper bounds and lower bounds using the expansion of $I_0(x)$; however, they are not enough. I need an exact value, no bounds.

  • 2
    $\begingroup$ A not-so-terrible upper bound can be derived from $\log I_0(x)\leq x$. We have $$\int_{0}^{+\infty} x\cdot I_0(x) e^{-ax^2}\,dx = \frac{1}{2a}e^{\frac{1}{4a}}.$$ $\endgroup$ – Jack D'Aurizio Jul 25 '18 at 20:39
  • $\begingroup$ @JackD'Aurizio Thanks. The diamond means you are one of the moderators? $\endgroup$ – Susan_Math123 Jul 25 '18 at 21:05
  • $\begingroup$ You're welcome, and... yup. $\endgroup$ – Jack D'Aurizio Jul 25 '18 at 21:16
  • $\begingroup$ @JackD'Aurizio To get another bound I can use $\ln(x)\leq x-1$. Then I will need to evaluate $\int_{0}^{\infty}x I_0(x)^2 e^{-a x^2} dx$. Do you have an answer for this as well? $\endgroup$ – Susan_Math123 Jul 26 '18 at 16:30
  • $\begingroup$ Such bound is much worse than the previous one, but in any case $$\int_{0}^{+\infty}x I_0(x)^2 e^{-ax^2}\,dx = \frac{1}{2a} e^{1/(2a)} I_0(1/(2a)).$$ $\endgroup$ – Jack D'Aurizio Jul 26 '18 at 16:42

This is more a comment than an answer, but too long to be edited in the comments section. $$S(\alpha)=\int\limits_{x=0}^\infty e^{-\alpha x^2}I_0(x) \ln(I_0(x))x \, dx$$ $$S(\alpha)=\frac12\int\limits_{x=0}^\infty e^{-\alpha x}I_0(\sqrt{x}) \ln(I_0(\sqrt{x})) \, dx$$ In fact $S(\alpha)$ is the Laplace transform of the function $I_0(\sqrt{x}) \ln(I_0(\sqrt{x}))$. $$S(\alpha)=\mathcal{L}_x\Big(I_0(\sqrt{x}) \ln(I_0(\sqrt{x}))\Big)(\alpha) $$ In the extended tables of Laplace transform (for example from H.Bateman) one find the Laplace transforms of a lot of functions involving combinations of Bessel functions, such as $I_\nu(c_1\sqrt{x}) I_\nu(c_2\sqrt{x}))$ for example. But not the function with logarithm $I_0(\sqrt{x}) \ln(I_0(\sqrt{x}))$.

Also math software such as WolframAlpha find no result in terms of standard mathematical function.

This draw to think that nowadays $S(\alpha)$ cannot be expressed in terms of standard mathematical functions. Of course, this is not a definitive proof. Moreover, new special functions might be defined in the future and among them one allowing to express $S(\alpha)$ in closed form.


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.