You are on page 1of 2

MATH 605, HW 1 SOLUTIONS

Folland’s Real Analysis; Chapter 1:


4.) This follows since any countable union can be written as an increasing countable union:
h i
j
[1 1
j=1 Ej = [j=1 [k=1 Ek ;

note that [jk=1 Ek is a finite union of sets in the algebra and is hence in the algebra.
5.) Let K ⇢ P(X ) be the set of all countable subsets and show that N = [F 2K NF is a
-algebra where NF is the -algebra generated by F . Suppose that Ei 2 NFi . Obviously,
Eic ⇢ NFi ⇢ N since NFi is an algebra. Furthermore, [1j=1 Ej ⇢ N[1
j=1 Fj
(note that [1
j=1 Fj is
also countable).
7). This follows in two steps: (a) If µ is a measure and a > 0, then it is obvious that aµ is
a measure; (b) If µ1 , µ2 are measures, it is also obvious that µ1 + µ2 is a measure. (I would
expect you to write down the equations for this.)
8). Recall that lim inf Ek = [1 1
k=1 \n=k En and lim inf ak = limk!1 inf n k ak . Define
1
Fk = \n=k En so that F1 ⇢ F2 ⇢ · · · . By continuity from below,
µ(lim inf Ek ) = µ([1
k=1 Fk ) = lim µ(Fk )  lim inf µ(En ) = lim inf µ(Ek ) .
k!1 k!1 n k

The inequality used monotonicity since Fk ⇢ En if n k. (I’ll leave µ(lim sup Ek ) for you.)
9). Use the splitting property twice: µ(E [ F ) = µ([E [ F ] \ E) + µ([E [ F ] \ E c ) and
µ(F ) = µ(F \ E) + µ(F \ E c ). Rewrite the first as µ(E [ F ) = µ(E) + µ(F \ E c ) and then
substitute in µ(F \ E c ) = µ(F ) µ(F \ E) to get µ(E [ F ) = µ(E) + µ(F ) µ(F \ E).
10). Set µE (A) = µ(A \ E); we’ll show that µE is a measure. Obviously µE (;) = µ(;) = 0,
so we just have to check -additivity. If Ai in the -algebra are disjoint, then so are E \ Ai .
Since E \ [[i Ai ] = [i [E \ Ai ], the -additivity of µ gives
X X
µE ([i Ai ) = µ([i [E \ Ai ]) = µ(E \ Ai ) = µE (Ai ) .
i i

14.) We’ll argue by contradiction: suppose µ is semi-finite, µ(E) = 1, and


sup{µ(F ) | F ⇢ E and µ(F ) < 1} = C < 1 .
Choose Fi ⇢ E with C 2 i < µ(Fi ) < 1. Set Gi = [ik=1 Fk so G1 ⇢ G2 ⇢ ·, Fi ⇢ Gi ⇢ E,
P
and C 2 i < µ(Fi )  µ(Gi ) (by monotonicity). By sub-additivity, µ(Gi )  ik=1 µ(Fk ) <
1 so the definition of C gives µ(Gi )  C. Set G = [1 i=1 Gi ; note that G ⇢ E. Continuity
from below implies that µ(G) = limi!1 µ(Gi ) = C. Since µ(E) = 1, additivity gives
µ(E \ C) = 1. Since µ is semi-finite, there exists F ⇢ (E \ G) with µ(F ) > 0. By additivity,
µ(F [ G) = µ(F ) + µ(G) > C. This contradicts that C was the supremum - we conclude
that the supremum had to be 1.

Professor Minicozzi, Fall 2002.


1
MATH 605, HW 2 SOLUTIONS

Folland’s Real Analysis; Chapter 1:

P
18.) (a) By definition, given ✏ > 0, there exist Aj 2 A (j = 1, 2, . . . ) with E ⇢ [1
1 Aj and
µ0 (Aj )  µ⇤ (E) + ✏. Subadditivity of µ⇤ and Proposition 1.13 give
X X
µ⇤ ([Aj )  µ⇤ (Aj ) = µ0 (Aj ) .
n
(b) To get one way, use part (a) for ✏ = 2 and take countable intersection. The other way
just uses subadditivity and monotonicity of outer measures.
(c) Write E as a countable union of sets of finite measure and use properties of -algebras.
19.) Let µ⇤ be an outer measure from a finite premeasure µ0 . Define the inner measure
µ⇤ (E) = µ0 (X) µ⇤ (X \ E). Show that E is measurable i↵ µ⇤ (E) = µ⇤ (E):
a). E measurable implies this immediately since µ(X) = µ(E) + µ(E \ X).
b). If µ⇤ (E) = µ⇤ (E), then µ⇤ (X \ E) + µ⇤ (E) = µ0 (X). By 18a), there are A sets B1 , B2
with E ⇢ B1 , E c ⇢ B2 , µ(B1 ) = µ⇤ (E), and µ(B2 ) = µ⇤ (E c ). It follows that µ(B1 \ B2 ) = 0
(since µ(X) = µ(B1 ) + µ(B2 ) = µ(X) + µ(B1 \ B2 )). Since B1 \ E = B1 \ E c ⇢ B1 \ B2 ,
monotonicity of outer measures implies that µ⇤ (B1 \ E) = 0. E is measurable by 18b).
P P
27). Prove 1.22 (a): First, use that aj 3 j < bj 3 j i↵ there exists n so an < bn and
aj = bj for j < n to prove (the hint) that x, y 2 C with x < y implies that there are z1 2 C
and z2 2/ C with x < zi < y. Follows immediately that C is totally disconnected and has
no isolated points. C is compact since it is an intersection of compact sets. Compact and
totally disconnected together imply that C is nowhere dense.
30). Given ✏ > 0, Proposition 1.20 gives a finite union A = [Ij of disjoint open intervals
so µ(A \ E) + µ(E \ A) < ✏ µ(E) (since µ(E) > 0). We use this twice. First,
X
µ(E) = µ(E \ A) + µ(E \ A)  µ(Ij ) + ✏ µ(E) ,
P
so that (1 ✏) µ(E)  µ(Ij ). Second,
X X X X
µ(Ij ) = µ(Ij \ E) + µ(Ij \ E) < µ(Ij \ E) + ✏ µ(E) .
Using the first estimate on µ, this gives
X X ✏ X
µ(Ij ) < µ(Ij \ E) + µ(Ij ) .
1 ✏

It follows easily from here if we choose ✏ > 0 small so 1 1 ✏
↵.
31). If E is Lebesgue measurable and m(E) > 0, show that the set E E contains an
interval about 0: Use (30) to find an interval I so that m(E \ I) > 3/4 m(I). Set F = E \ I
so µ(F ) > 3/2 µ(I). If | | < m(I)/2 and 2 / F F , then F \ (F + ) = ; and so (since m
is translation invariant)
2µ(F ) = µ(F ) + µ(F + ) = µ [F [ (F + )]  µ(I [ (I + )) = m(I) + < 3/2 m(I) .
Professor Minicozzi, Fall 2002.
1

You might also like