312 CHAPTER 10. THE ABSTRACT LEBESGUE INTEGRAL

11. Let the rational numbers in [0,1] be {rk}∞

k=1 and define

fn (t) ={

1 if t ∈ {r1, · · · ,rn}0 if t /∈ {r1, · · · ,rn}

Show that limn→∞ fn (t) = f (t) where f is one on the rational numbers and 0 on theirrational numbers. Explain why each fn is Riemann integrable but f is not. How-ever, each fn is actually a simple function and its Lebesgue and Riemann integral isequal to 0. Apply the monotone convergence theorem to conclude that f is Lebesgueintegrable and in fact,

∫f dm = 0.

12. Show limn→∞n2n ∑

nk=1

2k

k = 2. This problem was shown to me by Shane Tang, a for-mer student. It is a nice exercise in dominated convergence theorem if you massageit a little. Hint:

n2n

n

∑k=1

2k

k=

n

∑k=1

2k−n nk=

n−1

∑l=0

2−l nn− l

=n−1

∑l=0

2−l(

1+l

n− l

)≤

n−1

∑l

2−l (1+ l)

13. Suppose you have a real vector space E (ω) which is a subspace of a normed linearspace V , this for each ω ∈Ω where (Ω,F ) is a measurable space. Suppose E (ω) =span(b1 (ω) , · · · ,bn (ω)) where these bi (ω) are linearly independent and each ismeasurable into V . Define θ (ω) :Rn→ E (ω) by θ (ω)(∑n

i=1 aiei)≡∑ni=1 aibi (ω) .

Show that θ (ω) maps functions measurable into Rn to functions measurable into V .Now show θ (ω)−1 also maps functions measurable into V to functions measurableinto Rn. Hint: For the second part you need to start with a function ω → h(ω)which is measurable into V with values in E (ω). Thus h(ω) = ∑

ni=1 ai (ω)bi (ω) .

You need to verify that the ai are measurable. To do this, assume first that ∥h(ω)∥ isbounded by some constant M. Then consider Sr ≡

{ω : inf|a|>r ∥∑i aibi (ω)∥> M

}.

Explain why every ω is in some Sr. Then consider Φ(a,ω) which will be defined as−∥∑i aibi (ω)−h(ω)∥for ω ∈ Sr. Thus the maximum of this functional for ω ∈ Ω

is 0. Show that for ω ∈ Sr the maximum will occur on the set |a| ≤ M + 1. Thenapply Kuratowski’s lemma. Finally consider a truncation of h called hm and applywhat was just shown to this truncation which has ∥hm (ω)∥ ≤m. Then let m→∞ andobserve that for large enough m,hm (ω) = h(ω) and so the am

i (ω) are also constantfrom this value onward. Thus limm→∞ am

i (ω) ≡ ai (ω) exists and ai is measurable,being the limit of measurable functions.

14. Give an example of a sequence of functions { fn} , fn ≥ 0 and a function f ≥ 0 suchthat f (x) = liminfn→∞ fn (x) but

∫f dm < liminfn→∞

∫fndm so you get strict in-

equality in Fatou’s lemma.

15. Let f be a nonnegative Riemann integrable function defined on [a,b] . Thus there isa unique number between all the upper sums and lower sums. First explain why,if ai ≥ 0,

∫∑

ni=1 aiX[ti,ti−1) (t)dm = ∑i ai (ti− ti−1) . Explain why there exists an in-

creasing sequence of Borel measurable functions {gn} converging to a Borel mea-surable function g, and a decreasing sequence of functions {hn} which are also Borelmeasurable converging to a Borel measurable function h such that gn ≤ f ≤ hn,∫

gndm equals a lower sum,∫

hndm equals an upper sum

31211.12.13.14.15.CHAPTER 10. THE ABSTRACT LEBESGUE INTEGRALLet the rational numbers in [0, 1] be {7;,};-_, and define_ f lifte{r,--:,rn}hi) =4 Oifr dé frie sry}Show that lim; fn (t) = f (t) where f is one on the rational numbers and 0 on theirrational numbers. Explain why each f, is Riemann integrable but f is not. How-ever, each f;, is actually a simple function and its Lebesgue and Riemann integral isequal to 0. Apply the monotone convergence theorem to conclude that f is Lebesgueintegrable and in fact, [ fdm = 0.. k .Show limy—co 37 Le—1 = = 2. This problem was shown to me by Shane Tang, a for-mer student. It is a nice exercise in dominated convergence theorem if you massageit a little. Hint:—1a yar Sot = Fo(i4 4) < V2 ‘(+1= i=0lSuppose you have a real vector space E (@) which is a subspace of a normed linearspace V, this for each @ € Q where (Q, ) is a measurable space. Suppose E (@) =span (bi (@),---,b,(@)) where these b;(@) are linearly independent and each ismeasurable into V. Define 6 (@) : R” > E(@) by 0(@) (VL, aie;) = | aib; (@).Show that 6 (@) maps functions measurable into R” to functions measurable into V.Now show 0 (a)! also maps functions measurable into V to functions measurableinto R". Hint: For the second part you need to start with a function @ > h(@)which is measurable into V with values in E(@). Thus h(@) = V7. a;(@) b;(@).You need to verify that the a; are measurable. To do this, assume first that || (@)]|| isbounded by some constant M. Then consider S, = {@ : infiq);, ||L;aibi(@)|| > M}.Explain why every @ is in some S,. Then consider ® (a,@) which will be defined as—||¥;aib;(@) —h(@)||for @ € S,. Thus the maximum of this functional for @ € Qis 0. Show that for @ € S, the maximum will occur on the set |a| <M+1. Thenapply Kuratowski’s lemma. Finally consider a truncation of h called h,, and applywhat was just shown to this truncation which has ||, (@)|| <m. Then let m — 0 andobserve that for large enough m, hm (@) = h(@) and so the aj" (@) are also constantfrom this value onward. Thus lims-,..a/" (@) = a;(@) exists and a; is measurable,being the limit of measurable functions.Give an example of a sequence of functions {f,}, f, > 0 and a function f > 0 suchthat f (x) = liminf,... fn (x) but f fdm < liminf,. f frdm so you get strict in-equality in Fatou’s lemma.Let f be a nonnegative Riemann integrable function defined on [a,b]. Thus there isa unique number between all the upper sums and lower sums. First explain why,if a; > 0, {LE ai 2,4.) (()dm = Liai (ti —ti-1). Explain why there exists an in-creasing sequence of Borel measurable functions {g,} converging to a Borel mea-surable function g, and a decreasing sequence of functions {h,,} which are also Borelmeasurable converging to a Borel measurable function h such that g, < f < hn,/ g,dm equals a lower sum, / h,dm equals an upper sum