Re: Re: Incorrect symbolic improper integral
- To: mathgroup at smc.vnet.net
- Subject: [mg103748] Re: [mg103715] Re: Incorrect symbolic improper integral
- From: Di Lullo Alberto <Alberto.DiLullo at eni.it>
- Date: Mon, 5 Oct 2009 07:36:00 -0400 (EDT)
- References: <200909301141.HAA14962@smc.vnet.net> <ha212q$n8u$1@smc.vnet.net>
Dear Bob, I agree with you. The only thing I wanted to say is that, having provided explicit assumptions and getting back specific constraints, there shouldn't be anymore the issue of "genericity" in the results. It is like the difference between Solve and Reduce: if the latter produces wrong results, this is going to be a big trouble, while the former may be wrong under special circumstances. Perhaps, I have been too synthetic! I was referring only to your Solve example. Thanks Alberto -----Messaggio originale----- Da: DrMajorBob [mailto:btreat1 at austin.rr.com] Inviato: domenica 4 ottobre 2009 07.17 A: ADL; mathgroup at smc.vnet.net Oggetto: Re: [mg103715] Re: Incorrect symbolic improper integral > I do not think that DrMajorBob comments about Solve are applicable in > the cases below: It looks to me as if you've repeatedly proved my point, which was: Check ALL Mathematica's results with another method and/or more specific conditions, before betting the farm on them. Using the very same Integrate function on the same integrand with "a" symbolic every time (though varying its range) isn't a sufficient check, as you've demonstrated. I'm thinking more along the lines of NIntegrate[1/(1 + x^(1/2 + I)) // Im, {x, 0, Infinity}] That's both a specific value of "a" and a different method, and the result strongly indicates divergence for a value that, according to some of Mathematica's results, should give convergence. So again... Mathematica isn't always right. So check it, if the result is important. Bobby On Sat, 03 Oct 2009 08:04:04 -0500, ADL <alberto.dilullo at tiscali.it> wrote: > Following Dan Dubin's comment, I made some tests, reported below with > some visual simplification, which show that there are some troubles in > assumptions management in Mathematica. > Note that, when I gave the parameter a specific value, real or > complex, I found no problems with this integral. > > My version is 7.0.1 on Windows. > I do not think that DrMajorBob comments about Solve are applicable in > the cases below: > > Assuming[a > 0, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> If[a > 1, (Pi*Csc[Pi/a])/a, (*otherwise...*)] > -OK- > > Assuming[a < 0, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> Message: Integrate::idiv:Integral does not converge on {0, > Infinity} > ==> Integrate[(1 + x^a)^(-1), {x, 0, Infinity}] > -OK- > > Assuming[a != 0, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> If[Re[a] > 0, (Pi*Csc[Pi/a])/a, (*otherwise...*)] > -WRONG- > > Assuming[Element[a, Reals], Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> If[a > 0, (Pi*Csc[Pi/a])/a, (*otherwise...*)] > -WRONG- > > Assuming[a < 0, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> Message: Integrate::idiv:Integral does not converge on {0, > Infinity} > ==> Integrate[(1 + x^a)^(-1), {x, 0, Infinity}] > -OK- > > Assuming[a < 1, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> If[a > 0, (Pi*Csc[Pi/a])/a, (*otherwise...*)] > -WRONG- > > Assuming[a < -1, Integrate[1/(1 + x^a), {x, 0, Infinity}]] > ==> Message: Integrate::idiv:Integral does not converge on {0, > Infinity} > ==> Integrate[(1 + x^a)^(-1), {x, 0, Infinity}] > -OK- > > > Regards > ADL > > On Oct 1, 12:42 pm, Dan Dubin <ddu... at ucsd.edu> wrote: >> OK, many people have replied that the given integral was in fact done >> correctly by Mathematica. Here's a related integral that is not done >> correctly: >> >> Integrate[1/(1 + x^a),{x,0,Infinity}] >> ... >> This result is incorrect in the range 0<Re[a]<1. In this range the >> integral diverges, and is not given by the above cosecant expression. >> | Professor Dan Dubin >> | Dept of Physics , Mayer Hall Rm 3531, >> | UC San Diego La Jolla CA 92093-0319 >> | phone (858) - 534-4174 fax: (858)-534-0173 >> | ddu... at ucsd.edu > > -- DrMajorBob at yahoo.com ---------------------------------------------------------------------------= ---------------------------------------------------------------------------= ------------------ Message for the recipient only, if received in error, please notify the sen= der and read http://www.eni.it/disclaimer/