Bourbaki

Example

Tag 08RZ, permanent. Cite it as [Bourbaki, Tag 08RZ].

Integration, chapter IV, § 4. Integrable functions and sets

— Let XX be a discrete space, μ\mu a measure on XX, and set α(x)=μ(φ{x})\alpha(x) = \mu(\varphi_{\{x\}}) for every x∈Xx \in X. The functions in FF1\mathscr{F}_F^1 are then integrable, in other words LF1=FF1\mathscr{L}_F^1 = \mathscr{F}_F^1; moreover, for every function f∈LF1f \in \mathscr{L}_F^1,

∫f dμ=∑x∈Xα(x)f(x). \int f\, d\mu = \sum_{x \in X} \alpha(x)f(x).

For, let f∈FF1f \in \mathscr{F}_F^1; we have ∣μ∣∗(∣f∣)=∑x∈X∣α(x)∣⋅∣f(x)∣<+∞|\mu|^*(|f|) = \sum_{x \in X} |\alpha(x)| \cdot |f(x)| < +\infty (§ 1, No. 3, Example); for every ε>0\varepsilon > 0, there exists a finite subset MM of XX such that

∑x∈X−M∣α(x)∣⋅∣f(x)∣≤ε. \sum_{x \in X - M} |\alpha(x)| \cdot |f(x)| \leq \varepsilon.

The function gg equal to ff at the points x∈Mx \in M where ∣f∣|f| is finite, and to 0 elsewhere, belongs to H(X;F)\mathscr{H}(X; F) and, by the conventions that have been made,

∣μ∣∗(∣f−g∣)≤∑x∈X−M∣α(x)∣⋅∣f(x)∣≤ε, |\mu|^*(|f - g|) \leq \sum_{x \in X - M} |\alpha(x)| \cdot |f(x)| \leq \varepsilon,

which proves that f∈LF1f \in \mathscr{L}_F^1. On the other hand,

∣μ(g)−∑x∈Xα(x)f(x)∣≤∑x∈X−M∣α(x)∣⋅∣f(x)∣≤ε, \left| \mu(g) - \sum_{x \in X} \alpha(x)f(x) \right| \leq \sum_{x \in X - M} |\alpha(x)| \cdot |f(x)| \leq \varepsilon,

whence the second assertion.

In other words, the μ\mu-integrable functions ff are those for which the family (α(x)f(x))x∈X(\alpha(x)f(x))_{x \in X} is absolutely summable (GT, IX, §3, No. 6), and the integral ∫f dμ\int f\, d\mu is the sum of this family.

Since μ(f)\mu(f) is continuous on LF1\mathscr{L}_F^1 by definition, and since it takes its values in a Hausdorff space, we have μ(f)=0\mu(f) = 0 for every function that belongs to the closure of 0 in LF1\mathscr{L}_F^1, that is, is negligible; if ff and gg are two equivalent integrable functions, then μ(f)=μ(g)\mu(f) = \mu(g). In other words, the value of μ(f)\mu(f) depends only on the class f~\tilde{f} of the integrable function ff; it is again denoted μ(f~)\mu(\tilde{f}), and the function f~↦μ(f~)\tilde{f} \mapsto \mu(\tilde{f}) is a continuous linear mapping of LF1L^1_F into FF. If a function ff, with values in FF and defined almost everywhere in XX, is equivalent to an integrable function, we again say that ff is integrable and we write ∫f dμ=μ(f~)\int f\, d\mu = \mu(\tilde{f}); one defines similarly an integrable function with values in R‾\overline{\mathbf{R}}, defined and finite almost everywhere, as well as its integral.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Transcribed from the printed volume. Not translated.