MathGroup Archive 1999

[Date Index] [Thread Index] [Author Index]

Search the Archive

Mathematica on linux help and Mathematica returns wrong integral result

  • To: mathgroup at
  • Subject: [mg19431] Mathematica on linux help and Mathematica returns wrong integral result
  • From: psalzman at
  • Date: Tue, 24 Aug 1999 01:29:28 -0400
  • Sender: owner-wri-mathgroup at

dear all,

i have Mathematica for students, version 3.1 running on suse linux 6.1 with
enlightenment (no gnome or kde).

very often when i want to click on something, the cursor (which is normally
a pointer) becomes a hand and won't allow me to position it where i want.

for instance, i'm looking at the "help browser" right now, and i looked at
"factor" because i want to find out how to factor a pi/2 in front of an
expression.  it says "See also FactorTerms" as a blue hyperlink.  under
windows, when i click on that, the help browser would display the help for
FactorTerms.  here, when i go to click on it, the cursor becomes a hand and
won't let me click on it.

this is kind of annoying -- can someone tell me how to avoid this?  at the
very least, i want to be able to click on hyperlink help topics that appear
in the help browser.

wrong result

i've definitely confirmed that Mathematica definitely gives a wrong result for
an integral.  i've have an integral which contains a product of 3
bessel function of half order.  i expressed this integral in 3 ways --
converting no  bessel functions as sin x/x, converting 1 bessel function as 
sin[x]/x, converting 2 besselfunctions as sin x/x and converting all 3 as
sin x/ x.  the expressions are all equal, and the result that Mathematica gives is
*almost* correct.

the correct result was given by converting all bessel functions as sin x/x,
and i verified this by doing a very difficult contour integration (see what
Mathematica made me do?)   :)    i was very careful and thorough, and my answer
agrees completely with the Mathematica result of converting all bessel functions
into sin x / x.

the other 3 answers that Mathematica gave are numerically very close -- they differ 
by numbers on the order of 10^-21 and include very, very small log terms.

what seems to be happening is Mathematica isn't recognizing terms which ought to be
pure zero.  instead, they're simply (very) small.

this may be a machine precision thing, but i'm sufficiently dumb about
Mathematica's algorithms that i simply don't know.  in any event, how do i contact
wolfram about this to give details?  should i mail them my work?  do they
really want to see hardcore math relating to my research?

like i said, they're very, very close, numerically.  does anyone care about


  • Prev by Date: Re: Evaluation with UpValues
  • Next by Date: Re: Any examples of FEA in Mathematica?
  • Previous by thread: Re: Evaluation with UpValues
  • Next by thread: Re: Mathematica on linux help and Mathematica returns wrong integral result