Values of some integrals



>>>>> "Raymond" == Raymond Toy <raymond.toy at ericsson.com> writes:

>>>>> "Richard" == Richard Fateman <fateman at cs.berkeley.edu> writes:
    Richard> in commercial macsyma
    Richard> After setting intanalysis:false,

    Raymond> What happens if intanalysis has the default value of true?

    Richard> and saying that k is an integer and positive,
    Richard> I got these answers...

    Richard> [(( - ((sqrt(3) * %i)/2) - (1/2)) * 2^((2 * %i + 4)/3) * G(((2 * %i +
    Richard> 4)/3))/3),
    Richard> ((2 * 2^(1/3) * (((sqrt(3) * %i)/2) - (1/2)) * G((1/3)))/9),
    Richard> ((2 * %i)/3),

    Raymond> Assuming G here means the gamma function, then these 3 seem to be
    Raymond> wrong, based on the derivation in Wang's thesis, p 84-85.

    Raymond> Wang says integrate(x^m*exp(%i*k*x^n),x,0,inf) is

    Raymond> exp(signum(k)*%i*%pi*(m+1)/(2*n))*gamma(s/n/(abs(k))^s)

    Raymond> where s = (m+1)/n.  I've gone over the derivation, and it looks
    Raymond> right. 

I'm mistaken.  The text in Wang's thesis is wrong.  He writes the
above formula for

  J = exp(-%pi*%i*(m+1)/(2*n))*integrate(R^m*exp(k*R^n),R,0,inf)

but the last integral is a gamma function and its value is
gamma((m+1)/n)/(-k)^((m+1)/n)/n.  (Because k < 0 in this formula.)

With this corrected formula, maxima and Macsyma return the expected
values.

Ray