Is it known, and if yes how does one show, that the function

$$ \psi(n):=n\int_0^{+\infty} e^{-x}I_0\left(\frac{x}{n}\right)^{n-1}I_1\left(\frac{x}{n}\right)\mathrm{d}x$$

is decreasing for all $n\ge 3$? Here $I_0$ and $I_1$ denote the modified Bessel functions of the first kind of order $0$ and $1$ respectively.

  • $\begingroup$ May be it is true without integral? Try derivative? $\endgroup$ – Sergei May 30 '16 at 15:04
  • $\begingroup$ it is a derivative under the integral sigh? $\endgroup$ – Sergei May 31 '16 at 13:57
  • $\begingroup$ @Sergei: my apologies, you are right concerning the sign of the derivative. It is indeed negative. But how can one show it (without using Mathematica or any computer aid)? The derivative function is rather ugly. $\endgroup$ – ARC May 31 '16 at 16:13

As $I_0^\prime = I_1$, we have $ I_0\left(\frac{x}{n}\right)^{n-1}I_1\left(\frac{x}{n}\right) = \frac{d}{dx} \left(I_0(\frac{x}{n})^n\right)$ , so $$ \psi(n) = - n + n\int_0^\infty I_0\left(\frac{x}{n}\right)^n\exp(-x)dx,\\ = -n +n^2 \int_0^\infty (I_0(y) \exp(-y))^n dy $$
and as $0\leq \exp(-x)I_0(x)\leq1$, expanding $(n+1)^2$, $$ \psi(n+1)-\psi(n) = - 1 + (n+1)^2 \int_0^\infty (I_0(y) \exp(-y))^{n+1} dy \\ -(n)^2 \int_0^\infty (I_0(y) \exp(-y))^{n} dy \\ \leq (2n+1)\int_0^\infty (I_0(y) \exp(-y))^{n+1} dy-1. $$ Now let us use a rough upper bound for $I_0(x) \exp(-x)$, e.g., $$ I_0(x) \exp(-x) \leq \frac{4}{3\sqrt{2\pi x+1}}. $$ Then $$ \frac{4}{3}(2n+1)\int_0^\infty(2\pi x+1)^{-(n+1)/2}=\frac{4}{3\pi}\frac{2n+1}{n-1} $$ and provided $n\geq 10$, $$ \psi(n+1)-\psi(n) \leq 0 $$


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.