0
$\begingroup$

While trying to compute the following integrals, Mathematica is not able to do that. Any suggestion will be helpful.

Assuming[{R ∈ Reals, α ∈ Reals}, 
 Integrate[q^4 BesselJ[1, q ]/(1 - α^2 q^2)^2, {q, 0, ∞}]
 ]

Assuming[{R ∈ Reals, α ∈ Reals}, 
 Integrate[q^4 BesselJ[1, q ]/(1 - α^2 q^2)^3, {q, 0, ∞}]
 ]
$\endgroup$
3
  • 1
    $\begingroup$ What makes you expect that the integrals can be written down in closed-form? $\endgroup$ Commented Sep 12, 2018 at 7:30
  • $\begingroup$ In the "Table of Integrals, Series, and Products" by Gradshteyn and Ryzhik, (seventh edition), there is a similar integral given in page no 679, which reads $\endgroup$
    – ark
    Commented Sep 12, 2018 at 8:30
  • 1
    $\begingroup$ @ark I do not have the seventh edition, but in the fourth edition, which I have, this integral does not exist. Can you publish a copy of the page or write out the integral itself here? $\endgroup$ Commented Sep 12, 2018 at 15:06

3 Answers 3

2
$\begingroup$

Don't know about the non-convergence messages for NIntegrate, but this seems to work.

$Assumptions = \[Alpha] \[Element] Reals

int = Integrate[
  q^4 BesselJ[1, q]/(1 - \[Alpha]^2 q^2)^n, {q, 0, \[Infinity]}];

ConditionalExpression[((1/(\[Alpha]^4*Gamma[n - 1]*Gamma[n]))*Pi*Csc[Pi*n]*
    Abs[\[Alpha]]^(-n - 3)*(Abs[\[Alpha]]^3*Gamma[n - 1]*
      ((1 - 4*\[Alpha]^2*(n^2 - 3*n + 2))*BesselI[n - 2, -(I/Abs[\[Alpha]])] - 
       BesselI[4 - n, -(I/Abs[\[Alpha]])]) - 
     2*I*\[Alpha]^4*(2*Gamma[n - 1]*BesselI[3 - n, -(I/Abs[\[Alpha]])] - 
       Gamma[n]*BesselI[n - 1, -(I/Abs[\[Alpha]])])))/
   (2^n*E^((1/2)*I*Pi*n)), Re[n] > 7/4]

Limit[int, n -> 2] // FunctionExpand

(*Piecewise[{{(I*Pi*BesselJ[0, 1/\[Alpha]])/(4*\[Alpha]^6) - (Pi*BesselY[0, 1/\[Alpha]])/
      (4*\[Alpha]^6) + (I*Pi*BesselJ[1, 1/\[Alpha]])/(2*\[Alpha]^5) - 
     (Pi*BesselY[1, 1/\[Alpha]])/(2*\[Alpha]^5), \[Alpha] >= 0}}, 
  -((Pi*BesselY[0, 1/\[Alpha]])/(4*\[Alpha]^6)) + 
   ((Log[1/\[Alpha]] - Log[I/\[Alpha]])*BesselJ[0, 1/\[Alpha]])/(2*\[Alpha]^6) - 
   (Pi*BesselY[1, 1/\[Alpha]])/(2*\[Alpha]^5) + 
   ((Log[1/\[Alpha]] - Log[I/\[Alpha]])*BesselJ[1, 1/\[Alpha]])/\[Alpha]^5]*)

Limit[int, n -> 3] // FunctionExpand

(*Piecewise[{{(-4*I*Pi*\[Alpha]*BesselJ[0, 1/\[Alpha]] + I*Pi*BesselJ[1, 1/\[Alpha]] + 
      4*Pi*\[Alpha]*BesselY[0, 1/\[Alpha]] - Pi*BesselY[1, 1/\[Alpha]])/(16*\[Alpha]^7), 
    \[Alpha] >= 0}}, (-12*I*Pi*\[Alpha]*BesselJ[0, 1/\[Alpha]] + 
    3*I*Pi*BesselJ[1, 1/\[Alpha]] + 4*Pi*\[Alpha]*BesselY[0, 1/\[Alpha]] - 
    Pi*BesselY[1, 1/\[Alpha]])/(16*\[Alpha]^7)]*)

Without taking limits, the results are ComplexInfinity, which may explain why NIntegrate has trouble.

$\endgroup$
1
  • $\begingroup$ Thank You. @Bill $\endgroup$
    – ark
    Commented Sep 13, 2018 at 9:12
1
$\begingroup$
          Quiet[Integrate[
         q^4 BesselJ[1, q]/(1 - \[Alpha]^2 q^2)^n, {q, 0, \[Infinity]}]]

     (*ConditionalExpression[(1/(\[Alpha]^4 Gamma[n]))
     4^-n (-\[Alpha]^2)^(-(1/2) - 
      n) (2^n \[Pi] (-\[Alpha]^2)^(
      n/2) (4 \[Alpha]^2 BesselI[3 - n, 1/Sqrt[-\[Alpha]^2]] - 
        Sqrt[-\[Alpha]^2] BesselI[4 - n, 1/Sqrt[-\[Alpha]^2]]) Csc[
        n \[Pi]] + 
      16 (-1 + n) \[Alpha]^4 Sqrt[-\[Alpha]^2]
      Gamma[3 - n] HypergeometricPFQ[{n}, {-2 + n, -1 + n}, -(1/(
      4 \[Alpha]^2))]), 
       4 Re[n] > 
       7 && (Re[\[Alpha]^2] <= 0 || \[Alpha]^2 \[NotElement] Reals)]*)
      for real \[Alpha]^2 intégral is divergent.
$\endgroup$
2
  • 1
    $\begingroup$ I was looking for a compact analytical form of the mentioned integral not the Numerical integration, if possible. Thanks $\endgroup$
    – ark
    Commented Sep 12, 2018 at 9:05
  • 4
    $\begingroup$ It is dangerous to switch of the MMA error messages, especially in the case of not converging integrals! $\endgroup$ Commented Sep 12, 2018 at 9:52
0
$\begingroup$

If you try to solve the integrals numerically

int[\[Alpha]_?NumericQ, n_?NumericQ] :=NIntegrate[q^4 BesselJ[1, q]/(1 - \[Alpha]^2 q^2)^n, {q, 0, \[Infinity]} ]

the evaluation of for example

int[  1, 2]

gives error messages NIntegrate::ncvb: NIntegrate failed to converge to prescribed accuracy after 9 recursive bisections in q near {q} = {1.00277}. NIntegrate obtained 6648.614253396791 and 6097.681026759085for the integral and error estimates.

Elaborate the comment Henrik Schumacher gave...

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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