Integration, IV, p. 121
Bourbaki marks this one as harder than the rest.
a) Let f be a numerical function such that μ∗(f) (Exer. 5) is finite. Show that there exists an integrable function f1≥f such that μ(f1)=μ∗(f); if f2 is a second integrable function such that f2≥f and μ(f2)=μ∗(f), then f1 and f2 are equivalent.
b) For a numerical function f to be integrable, it is necessary and sufficient that μ∗(f) and μ∗(f) be finite and equal.
c) Let f be a numerical function such that μ∗(f) and μ∗(f) are both finite; let g and h be two integrable functions such that g≤f≤h and μ(g)=μ∗(f), μ(h)=μ∗(f). Show that
μ∗(f−g)=μ∗(h−f)=0andμ∗(f−g)=μ∗(h−f)=μ∗(f)−μ∗(f).
d) Let f1,f2 be two numerical functions such that the numbers μ∗(f1),μ∗(f2),μ∗(f1) and μ∗(f2) are all finite; show that
μ∗(f1+f2)≤μ∗(f1)+μ∗(f2)≤μ∗(f1+f2)
(if g2 is an integrable function such that f2≤g2 and μ∗(f2)=μ(g2), note that, for every integrable function h such that h≤f1+f2, one has h−g2≤f1). From this, deduce that if f1 is integrable then
μ∗(f1+f2)=μ(f1)+μ∗(f2),μ∗(f1+f2)=μ(f1)+μ∗(f2),
and μ∗(f1+f2)=μ∗(sup(f1,f2))+μ∗(inf(f1,f2)) (for the latter relation, make use of Exer. 1 of §1).
e) Let f be an integrable function. In order that a function g, such that μ∗(g) is finite, be integrable, it is necessary and sufficient that μ(f)=μ∗(g)+μ∗(f−g) (if g1 is an integrable function such that g≤g1 and μ∗(g)=μ(g1), note that f−g1≤f−g).