Bug in limit or integrate



"OK, this antiderivative is valid on the reals."

Here is a general antidiff of the real valued function exp(-x^n) that is 
valid for all real x, n, (n > about .05 ) (by comparison with quad_qags).  I 
don't know why it fails for 0 < n <=(about .05 ).

-(gamma_incomplete(1/n,x^n)/n-gamma(1/n)/n)*signum(x);

gammai_test(n, a, b) :=
  ( _z : _z:-(gamma_incomplete(1/n,x^n)/n-gamma(1/n)/n)*signum(x),
  print("    qags = ",  first(quad_qags(exp(-x^n),x,a,b)),
          "        formula = ", float(at(_z,[x=b])-at(_z,[x=a]))))$

display2d:false$

for i : 1 thru 10 do gammai_test(random(10.0)+.5, _h:random(1.0), 
random(2.0)+_h);

    qags =  0.70997985233956         formula =  0.70997985233956
    qags =  0.64248793774255         formula =  0.64248793774255
    qags =  0.39405032292301         formula =  0.39405032292301
    qags =  0.12849352798637         formula =  0.12849352798637
    qags =  0.2383328942369         formula =  0.2383328942369
    qags =  0.14758128797191         formula =  0.14758128797191
    qags =  0.14207995833076         formula =  0.14207995833076
    qags =  0.54953884342554         formula =  0.54953884342554
    qags =  0.14493067138495         formula =  0.14493067138495
    qags =  0.25412300011431         formula =  0.25412300011431
done



-----Original Message----- 
From: Barton Willis
Sent: Sunday, January 01, 2012 9:11 AM
To: rich.hennessy at verizon.net
Cc: maxima at math.utexas.edu
Subject: Re: [Maxima] Bug in limit or integrate

----"Richard Hennessy" <rich.hennessy at verizon.net> wrote: -----

>I should not have used the abs() function. This is a simpler answer and it
>does not blow up at zero.

>(gamma(1/4)/4-gamma_incomplete(1/4,x^4)/4)*signum(x);

OK, this antiderivative is valid on the reals. An antiderivative that is 
valid on the entire complex plane
is x |-> x*hypergeometric([1/4],[5/4],-x^4). But Maxima is unable to compute 
limits of hypergeometric
functions, you are out of luck if you used this antiderivative to try to 
compute integrate(exp(-x^4),x,minf,inf).

Some definite integrals of exp(-x^4) are broken:

(%i6) integrate(exp(-t^4),t,0,1);
(%o6) (%i*(gamma(1/4)-gamma_incomplete(1/4,1)))/4

--bw