## Problem with NIntegrate when WorkingPrecision is specified

7

0

I am trying to evaluate this integral numerically: $$\int_0^{\infty } m \exp (-m) J_1(m){}^2 \, dm$$ Everything is OK when only the integration method is specified:

NIntegrate[-m Exp[-m] BesselJ[1, m]^2, {m, 0, Infinity}, Method -> "ClenshawCurtisRule"]


but when I specify the WorkingPrecision, the integral remains unevaluated:

NIntegrate[-m Exp[-m] BesselJ[1, m]^2, {m, 0, Infinity}, Method -> "ClenshawCurtisRule",
WorkingPrecision -> 10]


What is wrong with this code?

I am using Mathematica v9.0.1

UPDATE

This bug is still present in version 10.0.0.0.

You can make it easier for others to check your code when you copy it straight from the Mathematica cell (copy as plain text) and paste it in your question with an indentation of 4 spaces. – Thies Heidecke – 2013-04-29T08:57:24.677

@ThiesHeidecke Codes are replaced with plain text. – M6299 – 2013-04-29T09:20:48.877

5

This is a bug. As a workaround for this specific integral you could use a symbolic solution:

Integrate[-m*Exp[-m]*BesselJ[1, m]^2, {m, 0, Infinity}]

(* (-3*EllipticE[-4] + 5*EllipticK[-4])/(5*Pi) *)


3

"LevinRule" should work splendidly here, I think:

NIntegrate[-m Exp[-m] BesselJ[1, m]^2, {m, 0, Infinity},
Method -> "LevinRule", WorkingPrecision -> 20]
-0.18196415067209554877


ruebenko's answer has given a closed form for this particular definite integral. Personally, I prefer it when the parameters of the elliptic integrals are within $[0,1)$, so I apply the imaginary modulus transformations here to yield

N[(EllipticK[4/5] - 3 EllipticE[4/5])/(Sqrt[5] π), 20]
-0.18196415067209708741


yes of course, good point. Using a different method is certainly an option. – None – 2013-04-29T12:03:16.133

2This bug is apparently not present in version 7 so I added a version-9 tag; can you determine if this is in version 8? – Mr.Wizard – 2013-04-30T11:29:30.800

2@Mr. Wizard, Yes, it's busted in version 8. – J. M.'s ennui – 2013-04-30T11:37:03.543