From: Wilbert Dijkhof Newsgroups: sci.math Subject: Re: problem with lebesgue-borel measures along with fubini's theorem = lots of fun! Date: Thu, 19 Nov 1998 16:56:54 +0100 Mark Lokhorst wrote: > Hi, > > I need some help on this one from > whoever *likes* to play with Lebesgue-Borel measures. > > Question: > > Proof (inf = infinity): > > inf inf > / / > d | | > -- | cos(xt) f(t) dt = - | sin(xt) tf(t) dt > dx / / > -inf -inf > > using Fubini's theorem for each f which has a Lebesgue-Borel measure > so that > > inf > / > | > | |tf(t)|dt < inf > / > -inf Hallo, Je moet hem wel goed overschrijven he (opgave 34?, grondslagen analyse II). Als je hem wel goed overgeschreven hebt kom ik er niet uit. Anyway I think it's supposed to be the following: d/dx S_-inf^inf [ sin(xt)*f(t) ] dt = S_-inf^inf [ cos(xt)*t*f(t) ] dt Let S be the integral sign. [I hope the translation is correct.] Fubini says: Let (omega_i,U_i,mu_i), i=1,2 be two sigma-finit measure-spaces and let g be an mu_1 x mu_2- integratable numeric function on omega_1 x omega_2. Then "some stuff" and S g d(mu_1 x mu_2) = S ( S g(w_1) dmu_1 ) dmu_2 = S ( S g(w_2) dmu_2 ) dmu_1. With g(w_1) I mean g_(w_2) : w_1 -> g(w_1,w_2), thus w_2 kept constant. Since we are talking about function with a Lebesgue-Borel measure we have: omega_i = R for i=1,2 U_i = B (the Borel-sets) for i=1,2 mu_i = lamda (the Lebesgue-measure) for i=1,2 We have sin(xt) = t * S_0^x cos(st) ds and thus d/dx S_-inf^inf sin(xt)*f(t) dt = d/dx S_-inf^inf [ t * f(t) * S_0^x cos(st) ds ] dt = (Fubini) d/dx S_0^x [ S_-inf^inf cos(st)*t*f(t) dt ] ds = S_-inf^inf cos(xt)*t*f(t) dt To apply Fubini we must have that the mapping g: R^2 -> g(R^2) defined by g(s,t) = cos(st) * t * f(t) is an lamda^2-measurable numeric function on R^2. Since f is a Lebesgue-Borel-measurable function and cos and "t" are continuous we have that g is a Borel-measurable function on R^2. Exercise: How have we used that S_-inf^inf |t*f(t)| dt < oo ? Wilbert