9.16. FADDEYEV’S LEMMA 231

Do∫

Ωdµ to both sides and use µ (Ω) = 1. Thus

φ

(∫Ω

f dµ

)≤∫

φ ( f )dµ +λ

(∫Ω

f dµ−∫

f dµ

)=∫

φ ( f )dµ.

There are no difficulties with measurability because φ is continuous. ■

Corollary 9.15.2 In the situation of Lemma 9.15.1 where µ(Ω) = 1, suppose f hasvalues in [0,∞) and is measurable. Also suppose φ is convex and increasing on [0,∞).Then φ(

∫Ω

f du)≤∫

Ωφ( f )dµ .

Proof: Let fn (ω) = f (ω) if f (ω) ≤ n and let fn (ω) = n if f (ω) ≥ n. Then bothfn,φ ( fn) are in L1 (Ω) . Therefore, the above holds and φ(

∫Ω

fndu) ≤∫

Ωφ( fn)dµ. Let

n→ ∞ and use the monotone convergence theorem. ■

9.16 Faddeyev’s LemmaThis next lemma is due to Faddeyev. I found it in [35].

Lemma 9.16.1 Let f ,g be nonnegative measurable nonnegative functions on a measurespace (Ω,µ). Then

∫f gdµ =

∫∞

0∫[g>t] f dµdt =

∫∞

0∫

0 µ ([ f > s]∩ [g > t])dsdt.

Proof: First suppose g = aXE where E is measurable, a > 0. Now [g > t] = /0 ift ≥ a and it equals XE if t < a. Thus the right side equals

∫ a0∫

E f dµdt =∫ a

0∫

XE f dµ =∫aXE f dµ which equals the left side. Thus the first equation is true if g = aXE . Similar

reasoning shows that when you have g a nonnegative simple function, g = ∑ni=1 aiXEi

where we can arrange to have {ai} increasing, the first equation still holds. Now by themonotone convergence theorem, this yields the desired result for the first equation.

To get the second equal sign, note that∫∞

0

∫[g>t]

f dµdt =∫

0

∫X[g>t] f dµdt =

∫∞

0

∫∞

0µ([

X[g>t] f > s])

dsdt

=∫

0

∫∞

0µ ([ f > s]∩ [g > t])dsdt ■

9.17 Exercises1. Let Ω = N={1,2, · · ·}. Let F = P(N), the set of all subsets of N, and let µ(S) =

number of elements in S. Thus µ({1}) = 1 = µ({2}), µ({1,2}) = 2, etc. In thiscase, all functions are measurable. For a nonnegative function, f defined on N, show∫N f dµ = ∑

∞k=1 f (k) . What do the monotone convergence and dominated conver-

gence theorems say about this example?

2. For the measure space of Problem 1, give an example of a sequence of nonnegativemeasurable functions { fn} converging pointwise to a function f , such that inequalityis obtained in Fatou’s lemma.

3. If (Ω,F ,µ) is a measure space and f ≥ 0 is measurable, show that if g(ω) = f (ω)a.e. ω and g≥ 0, then

∫gdµ =

∫f dµ. Show that if f ,g ∈ L1 (Ω) and g(ω) = f (ω)

a.e. then∫

gdµ =∫

f dµ .

9.16. FADDEYEV’S LEMMA 231Do Jo dp to both sides and use pf (Q) = 1. Thus0 ( [rau < [onan (frau [ rau) = [o(fdu.There are no difficulties with measurability because @ is continuous.Corollary 9.15.2 In the situation of Lemma 9.15.1 where u(Q) = 1, suppose f hasvalues in [0,°°) and is measurable. Also suppose @ is convex and increasing on |0,°°).Then 9(Jof du) < Joo(fdu.Proof: Let f,(@) = f(@) if f(@) <n and let f,(@) =n if f(@) >n. Then bothfas? (fn) are in L'(Q). Therefore, the above holds and $(Jg frdu) < Jo @(fn)du-. Letn — co and use the monotone convergence theorem.9.16 Faddeyev’s LemmaThis next lemma is due to Faddeyev. I found it in [35].Lemma 9.16.1 Let f,¢ be nonnegative measurable nonnegative functions on a measurespace (Q,). Then f fedut = Je" figoy fdudt = f° fe w([f >] lg > #]) dsdr.Proof: First suppose g = a2 where E is measurable, a > 0. Now |g >t] = 0 ift >a and it equals 2p if t < a. Thus the right side equals fj fj, fdudt = fo [ 2efdu =f[a2zfdp which equals the left side. Thus the first equation is true if g = a2z. Similarreasoning shows that when you have g a nonnegative simple function, g = Y_) a; 2z,where we can arrange to have {a;} increasing, the first equation still holds. Now by themonotone convergence theorem, this yields the desired result for the first equation.To get the second equal sign, note that[ eon = [ [ %e-nfaua = | [ L(|Zigsaf > s]) dsdt- [Cf wire sinte>djasae m9.17 Exercises1. Let QQ =N={1,2,---}. Let ¥ = A(N), the set of all subsets of N, and let p(S) =number of elements in S. Thus p({1}) = 1 = u({2}), u({1,2}) = 2, etc. In thiscase, all functions are measurable. For a nonnegative function, f defined on N, showJn fdu = Le, f (kK). What do the monotone convergence and dominated conver-gence theorems say about this example?2. For the measure space of Problem 1, give an example of a sequence of nonnegativemeasurable functions { f, } converging pointwise to a function f, such that inequalityis obtained in Fatou’s lemma.3. If (Q,-F, 1) is a measure space and f > 0 is measurable, show that if g(@) = f(@)ae. @ and g > 0, then { gdu = f{ fd. Show that if f,g € L!(Q) and g(@) = f (@)a.e. then fgdu = f fdu.