In this paper, page 252, there is an equality (D.69), the numerical results integrals on the left is of interest:

$$\int_{-\infty}^{\infty}\dfrac{x\cdot J_0(x\cdot r)}{x^2-k^2}dx=\left\{ \begin{array}{cc} \dfrac{i \cdot\pi \cdot H_0^{(1)}(k\cdot r)}{2}&\text{Im}(k)>0 \\ \\ \dfrac{i \cdot\pi \cdot H_0^{(2)}(k\cdot r)}{2} &\text{Im}(k)<0\\ \end{array} \right. \tag{D.69}$$

Where $J_0$ is the BesselJ, $H_0$'s are the Hankel functions respectively.

I don't know how to prove the integral equality or whether it holds only for the non-real $k$'s .

But I am interested in the numerical (or analytical when possible) result when $k$ is a real positive number.

But how to obtain a numerical solution of such integrals? e.g., I tried to let $k=10$ and $r=6$, but found most of the available integral algorithms in NIntegrate didnot lead to easy convergence probably due to the existence of a discontinuity point $x=k$ of the integrand and the improper upper limit.


NIntegrate[(x BesselJ[0, 6 x])/(x^2 - 100), {x, 0, Infinity}, 
 Method -> {"OscillatorySelection", "TermwiseOscillatory" -> True, 
   Method -> {"GlobalAdaptive", "MaxErrorIncreases" -> 10000}}, 
 MaxRecursion -> 100, PrecisionGoal -> 8, AccuracyGoal -> 5]



and warning messages:

NIntegrate::slwcon: Numerical integration converging too slowly; suspect one of the following: singularity, value of the integration is 0, highly oscillatory integrand, or WorkingPrecision too small.

NIntegrate::ncvb: NIntegrate failed to converge to prescribed accuracy after 9 recursive bisections in x near {x} = {10.0009}. NIntegrate obtained 0.16591870616051807and 0.2705956661691797 for the integral and error estimates.

  • $\begingroup$ The integration doesn't converge: Integrate[(x BesselJ[0, 6 x])/(x^2 - 100), {x, -Infinity, Infinity}] $\endgroup$ – xzczd Jul 30 '17 at 5:22
Integrate[(x BesselJ[0, r x])/(x^2 - k^2), {x, 0, Infinity}, 
    PrincipalValue -> True, Assumptions -> k > 0 && r > 0]
(* 1/2 π BesselY[0, k r] *)

N[% /. k -> 10 /. r -> 6]
(* -0.0743913 *)
| improve this answer | |
  • 1
    $\begingroup$ You only took half of the domain in the OP's integral. $\endgroup$ – J. M.'s technical difficulties Jul 30 '17 at 6:35
  • 1
    $\begingroup$ @J.M. OP's actual code shows the integral from zero to Infinity, so I presume that is what he meant. Furthermore, the integrand is antisymmetric, so the integral from -Infinity to Infinity is zero. $\endgroup$ – bbgodfrey Jul 30 '17 at 6:37
  • 1
    $\begingroup$ Indeed, so we'll have to wait for OP to clarify. The only problem now is $k=0$, where the integral is genuinely divergent. $\endgroup$ – J. M.'s technical difficulties Jul 30 '17 at 6:40
  • $\begingroup$ Thank you. This already answers my question $\endgroup$ – LCFactorization Jul 30 '17 at 8:09

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.