Bourbaki

Exercise 29, § 5

Integration, IV, p. 132

Bourbaki marks this one as harder than the rest.

Let μ be a measure on a locally compact space X. For every μ-measurable numerical function f ≥ 0 defined on X, the mapping f'_μ : t ↦ |\mu|^*(f([t, +∞])) of R+\mathbf{R}_+ into R‾+\overline{\mathbf{R}}_+ is decreasing and right-continuous. If ν\nu is a measure on a second locally compact space Y and g is a ν\nu-measurable numerical function ⩾0\geqslant 0 defined on Y, f and g are said to be equimeasurable (for μ\mu and ν\nu respectively) if fμ′=gν′f'_\mu = g'_\nu. One denotes by f∗f^* (or fμ∗f^*_\mu) the function defined on R+\mathbf{R}_+ that is equal, for all s∈R+s \in \mathbf{R}_+, to the supremum in R‾+\overline{\mathbf{R}}_+ of the set of numbers a such that fμ′(a)⩾sf'_\mu(a) \geqslant s (the supremum being equal to 0 if this set is empty).

a) Show that the function f∗f^* is decreasing, left-continuous in R+\mathbf{R}_+, and equimeasurable to f (for the measure μ\mu and Lebesgue measure on R+\mathbf{R}_+); f∗f^* is called the decreasing rearrangement of f.

b) If 0⩽f⩽g0 \leqslant f \leqslant g are two μ\mu-measurable functions on X, show that f∗⩽g∗f^* \leqslant g^*. If (fn)(f_n) is an increasing sequence of μ\mu-measurable functions ⩾0\geqslant 0 on X, and f=sup⁡nfnf = \sup_n f_n, show that f∗(s)=sup⁡nfn∗(s)f^*(s) = \sup_n f^*_n(s) at all the points where f∗f^* is continuous.

c) Show that for every μ\mu-measurable numerical function f⩾0f \geqslant 0, ∣μ∣∗(f)=∫∗f∗(s) ds|\mu|^*(f) = \int^* f^*(s) \, ds. (Consider first the case of a step function, then make use of b).)

d) Let w be a numerical function defined on R+\mathbf{R}_+ that is finite (except possibly at the point s=0s = 0) and decreasing. For every μ\mu-measurable numerical function f⩾0f \geqslant 0 defined on X, set

λ(f)=(∫∗w(s)(f∗(s))p ds)1/p(1⩽p<+∞). \lambda(f) = \left( \int^* w(s)(f^*(s))^p \, ds \right)^{1/p} \quad (1 \leqslant p < +\infty).

Show that this function satisfies the conditions of Exer. 28. (To prove that it is convex, consider first the case that w is a decreasing step function, using c), then pass to the limit to treat the general case.)

e) With the notations of d), write LFp,w\mathscr{L}_F^{p,w} instead of LFλ\mathscr{L}_F^\lambda, and set Np,w(f)=λ(∣f∣)N_{p,w}(f) = \lambda(|f|) for every μ\mu-measurable mapping f of X into F. Show that if f∈LFp,wf \in \mathscr{L}_F^{p,w} and if, for every n, fnf_n denotes the function equal to f if ∣f∣⩽n|f| \leqslant n and to nf/∣f∣nf/|f| if ∣f∣>n|f| > n, then the sequence (∣fn∣∗)(|f_n|^*) tends almost everywhere to ∣f∣∗|f|^* and one has lim⁡n→∞Np,w(f−fn)=0\lim_{n \to \infty} N_{p,w}(f - f_n) = 0.

If f is μ\mu-negligible, then Np,w(f)=0N_{p,w}(f) = 0.

Solution

No solution has been written yet.

Transcribed from the printed volume. Not translated.