integrate bessel_k(2,%i*y) surprise



On Oct. 26, 2011, Raymond Toy wrote:
-------------------------------
>> (%o26) 1.736293756153568-2.1367387806763253E+42*%i
>> ----------------------------------------------------------------
>> Where does the 10^42 come from??
>
>It comes from the bessel_y(1,100*%i) term that integrate produces.   I
>checked the value of this and I think it's correct.  I assume that
>integrate produces the incorrect result.  I didn't check that.
----------------------------------------------
I agree that bessel_y(1,100*%i) produces 10^42.
I also assume that integrate produces the incorrect result.

Wolfram alpha gives a numerical value
  NIntegrate[BesselK[2,y*I],{y,1,100}]  --->

    -1.42033 + 1.62942*I

and gives the indefinite integral in terms of the hypergeomentric
function: the Meijer G-function.

Ted