234 CHAPTER 9. BASIC FUNCTION SPACES

Proof: Since f is bounded and µ (X) ,λ (Y ) < ∞, (∫

Y (∫

X | f (x,y)|dµ)pdλ )1p < ∞. Let

J(y) =∫

X | f (x,y)|dµ . Note there is no problem in writing this for a.e. y because f isproduct measurable. Then by Fubini’s theorem,∫

Y

(∫X| f (x,y)|dµ

)p

dλ =∫

YJ(y)p−1

∫X| f (x,y)|dµ dλ

=∫

X

∫Y

J(y)p−1| f (x,y)|dλ dµ

Now apply Holder’s inequality in the last integral above and recall p−1 = pq . This yields

∫Y

(∫X| f (x,y)|dµ

)p

dλ ≤∫

X

(∫Y

J(y)pdλ

) 1q(∫

Y| f (x,y)|pdλ

) 1p

=

(∫Y

J(y)pdλ

) 1q ∫

X

(∫Y| f (x,y)|pdλ

) 1p

=

(∫Y(∫

X| f (x,y)|dµ)pdλ

) 1q ∫

X

(∫Y| f (x,y)|pdλ

) 1p

dµ . (9.12)

Therefore, dividing both sides by the first factor in the above expression,(∫Y

(∫X| f (x,y)|dµ

)p

) 1p

≤∫

X

(∫Y| f (x,y)|pdλ

) 1p

dµ . (9.13)

Note that 9.13 holds even if the first factor of 9.12 equals zero. ■Now consider the case where f is not assumed to be bounded and where the measure

spaces are σ finite.

Theorem 9.6.2 Let (X ,S ,µ) and (Y,F ,λ ) be σ -finite measure spaces and let fbe product measurable. Then the following inequality is valid for p≥ 1.

∫X

(∫Y| f (x,y)|p dλ

) 1p

dµ ≥(∫

Y(∫

X| f (x,y)|dµ)pdλ

) 1p

. (9.14)

Proof: Since the two measure spaces are σ finite, there exist measurable sets, Xm andYk such that Xm ⊆ Xm+1 for all m, Yk ⊆ Yk+1 for all k, and also µ (Xm) ,λ (Yk) < ∞. Nowdefine

fn (x,y)≡{

f (x,y) if | f (x,y)| ≤ nn if | f (x,y)|> n.

Thus fn is uniformly bounded and product measurable. By the above lemma,

∫Xm

(∫Yk

| fn(x,y)|p dλ

) 1p

dµ ≥(∫

Yk

(∫

Xm

| fn(x,y)|dµ)pdλ

) 1p

. (9.15)

Now observe that | fn (x,y)| increases in n and the pointwise limit is | f (x,y)|. Therefore,using the monotone convergence theorem in 9.15 yields the same inequality with f replac-ing fn. Next let k→∞ and use the monotone convergence theorem again to replace Yk withY . Finally let m→ ∞ in what is left to obtain 9.14. ■

234 CHAPTER 9. BASIC FUNCTION SPACESProof: Since f is bounded and p(X) ,A (Y) < », (yy f(x,y) |dw)Pd A) P < co, LetJ(y) = fy |f(,y¥)|du. Note there is no problem in writing this for a.e. y because f isproduct measurable. Then by Fubini’s theorem,[(fireoian)’aa = [soy [irteonlanan= [ [soy tireesylaa duNow apply Holder’s inequality in the last integral above and recall p— 1 = E. This yieldsLG. flosyildu) a2 <[ (Lure): (/ sy) Pan)" dpL= (faora)" [Cf /tss)Pan) "a1=([fresoyiamrar)” [Cf sy) Pa2)" dp (9.12)Therefore, dividing both sides by the first factor in the above expression,(/ (/ ess)an)" a2)’ <[(/ tes) Pan)" dp (9.13)Note that 9.13 holds even if the first factor of 9.12 equals zero.Now consider the case where f is not assumed to be bounded and where the measurespaces are o finite.Theorem 9.6.2 Let (X,.%,u) and (Y,F,A) be o-finite measure spaces and let fbe product measurable. Then the following inequality is valid for p > 1.LU, itsotaa) a = (¢/, ees)lanyran) (9.14)Proof: Since the two measure spaces are o finite, there exist measurable sets, X,, andY, such that X, © X41 for all m, Y~ C Yy41 for all k, and also U (Xm),A (Yq) < 00. Nowseme foxy) if Lf (9)_ x,y) i X,y)| <n| nif |f (x,y)| > an.Thus f, is uniformly bounded and product measurable. By the above lemma,[ (f. Jalsy) Pa ) j du > (fof Jabsy)lau)?ar ) z (9.15)Now observe that | fn (x,y)| increases in n and the pointwise limit is | f (x,y)|. Therefore,using the monotone convergence theorem in 9.15 yields the same inequality with f replac-ing fn. Next let k — - and use the monotone convergence theorem again to replace Y; withY. Finally let m — in what is left to obtain 9.14. Mi