2
$\begingroup$

How to study large $r \gg 1$ asymptotics of $$I(r):=\int_0^{\infty} \frac{1-e^{-q}}{1+q} J_0(rq) \ dq,$$ where $J_0$ is the zeroth order Bessel function of the first kind.

I did some numerics and it seems to indicate that $I(r) \sim 1/r^3$ for large $r$, but this is not obvious to me where it comes from.

$\endgroup$

1 Answer 1

2
$\begingroup$

To obtain an asymptotic expansion for large $r$, one might proceed as follows ($H_0$ is Struve function, $Y_0$ is Bessel function): $$I(r)=\int_0^{\infty} \frac{1-e^{-q}}{1+q} J_0(rq) \ dq=\int_0^{\infty} \frac{1}{1+q} J_0(rq)-\sum_{n=0}^\infty(-1)^n\int_0^{\infty} q^ne^{-q} J_0(rq)$$ $$\qquad=\tfrac{1}{2}\pi {H}_0(r)-\tfrac{1}{2}\pi Y_0(r)-\sum_{n=0}^\infty(-1)^n\Gamma (n+1) \, _2F_1\left(\frac{n+1}{2},\frac{n+2}{2};1;-r^2\right)$$ $$\qquad=\frac{3}{2 r^3}-\frac{123}{8 r^5}+{\cal O}(r^{-7}).$$


This is confirmed by a calculation using the representation derived here of the Bessel function integral as an integral of elementary functions, $$I(r)=\int_0^\infty dt\,e^{-t} \left(\frac{1}{\sqrt{ t^2+r^2}}-\frac{1}{\sqrt{( t+1)^2+r^2}}\right)$$ $$\qquad=\frac{3}{2 r^3}-\frac{123}{8 r^5}+\frac{6185}{16 r^7}-\frac{2424835}{128 r^9}+{\cal O}(r^{-11}).$$

$\endgroup$
8
  • 1
    $\begingroup$ It is unclear to me why "the integrand will be dominated by the small-$q$ range for large $r$". By this kind of logic, one would be also able to write $I(r)\sim\int_0^{\infty} \frac q{1+0} J_0(rq) \ dq$, but the latter integral does not exist. $\endgroup$ Commented Apr 15 at 11:49
  • $\begingroup$ I think the series in your alternative approach does not converge. So, you need a bound on the remainder. Also, the result of your alternative approach is incomatible with that of your "Simple approach". $\endgroup$ Commented Apr 15 at 13:16
  • $\begingroup$ You did not mention an asymptotic expansion before my previous comment. On the other hand, I did try to understand your series as an asymptotic expansion -- which is why I said that you need a bound on the remainder, and you have still not provided such a bound. $\endgroup$ Commented Apr 15 at 13:39
  • 1
    $\begingroup$ The key is that you don't need expansions of $Y_0$ and $H_0$ separately, but of their difference. Mathematica and Wikipedia tell me that the difference $H_0-Y_0$ has powerlaw asymptotics at infinity: $$H_0(r)-Y_0(r)=\frac{2 \left(r^4-r^2+9\right)}{\pi r^5}+{\cal O}(r^{-7}).$$ $\endgroup$ Commented Apr 16 at 7:53
  • 1
    $\begingroup$ that's pretty smart, thank you! $\endgroup$ Commented Apr 16 at 8:07

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.