Rudin Ch1
Aaron Lou
May 2018
1 Chapter 1
Theorem 1.29 Full Solution
To prove ϕ is a measure on M, let E1 , E2 , . . . be disjoint members of M such
that their union is E. Note that, by definition of σ-algebra, this works. Observe
that
∞
X
χE f = χEj f
j=1
and that
Z Z
ϕ(E) = χE f dµ ϕ(Ej ) = χEj f dµ
X X
By theorem 1.27 (sum version of monotone convergence theorem) we note that
∞
X
ϕ(E) = ϕ(Ej )
j=1
and we note that, since ϕ(∅) = 0 so that means that ϕ(E) is a measure since
all individual functions are measures.
To prove the second condition, if g is a χE for some value of E, we note that
Z Z Z Z Z
gdϕ = χE dϕ = dϕ = ϕ(E) = f dµ = χE f dµ
X X E E X
Where we note that the third equality comes from the definition of the ”1”
function. We can therefore build every possible simple function using linear
combinations of individual χE . From these simple functions, this shows that
this holds for every function f that is measurable (by taking the supremum),
and every other case can be shown using the monotone convergence theorem.
Problem 1
Suppose that this set is countable. Then the σ-algebra can be divided into the
sets
1
A1 , A2 , . . . Ac1 , Ac2 , . . .
T
where we exclude ∅ and X for trivial purposes and let Ai 6= ∅. We notice
that
\
A1 ∩ A2 = Ac1 ∪ Ac2 =⇒ Ai ∈ M
i=1
We denote B1 = Ai and B2 = Aci . Note that, B1 ∪ B2 ∈ M but it is clear
T S
that these values do not appear in our countable Ai or Aci since Ai ∩B2 = ∅ and
Aci ∩ B1 = ∅. It follows, by contradiction, that the σ-algebra is uncountable.
Problem 3
Notice that this is a more specific case of the previous proof (of α ∈ R instead
of just rationals). To prove this, we just need to notice that we can define a
sequences of real αn → α as n → ∞ for all α ∈ R.
More formally, we can take αn to be the number β such that β > α, β is a
decimal to the nearest n-th place. This means that,
lim αn = α =⇒ lim {x : f (x) ≥ αn } = (α, ∞]
n→∞ n→∞
and we just apply the previous problem to show that this function is measurable.
Problem 5
a. For the case f (x) < g(x), we note that Φ(u, v) = v − u is a continuous
function which implies that g − f is a measurable function. Furthermore, it
therefore follows that (g − f )+ is a measurable function. It follows that the set
{x : f (x) ≤ g(x)} is a measurable set since if is the preimage of [−∞, ∞] under
the measurable function (g − f )+ .
For the other case f (x) = g(x), we define the function as (g − f )+− , which
is clearly {x : f (x) = g(x)}. {x : f (x) < g(x)} follows from {x : f (x) ≤
g(x)} − {x : f (x) = g(x)} and both are measurable sets.
b. We let the functions be f1 , f2 , . . . converge to f . We also note that the
set of all things that converges is lim (−α, α). We can clearly see that this is
[ [ α→∞
−1 −1
f ( (−α, α)) = lim f ((−α, α)). Clearly, this is the union of mea-
α→∞
α→∞
surable sets, so we see that the points where f converges to a finite value is a
measurable set.
Problem 7
This is a natural deduction using the dominated convergence theorem. Notably,
we note that |fi | ≤ f1 , and f1 must be L1 (µ).
Problem 9
We note that the taylor expansion of log(1 + x) is given by
2
∞
x2 x3 X (−1)i+1 x
log(1 + x) = x − + ··· =
2 3 i=1
i
As we take n → ∞, we note that the terms of degree o(x2 ) are negligible for
α ≥ 1 and this value converges for x. We calculate the value to be
fα
Z Z
lim n log(1 + (f /n)α )dµ ≤ lim dµ
n→∞ X n→∞ X nα−1
R
and these values converge. If α = 1, then this converges to X f dµ = c. If α > 1,
then weα notice that these values converge to 0 since they are the equations
fn = nfα−1 are dominated by f α ∈ L1 (µ).
Otherwise, if 0 < α < 1, we apply Fatou’s lemma to find that
Z Z
lim inffn dµ ≤ lim inf fn dµ
X n→∞ n→∞ X
where each fn is
fn = n log(1 + (f /n)α )
Clearly, we note that, as n → ∞, inf fn → ∞ since n → ∞ and log(1 +
(f /n)α ) → ∞. Therefore, it follows that
Z Z
∞ ≤ lim inf fn dµ =⇒ lim n log(1 + (f /n)α ) = ∞
n→∞ X n→∞ X
Since, if the inf (lowest value) is infinity, then the entire function is obviously
infinity.
Problem 11
∞ [
\ ∞
A⊆ Ek because, by definition, every point that is in an infinite number
n=1k=n
∞
[
of Ei will be a member of Ek for all k.
k
∞ [
\ ∞
Similarly, Ek ⊆ A since every element of the RHS can’t have a finite
n=1k=n
number of elements (or else we can take k to be the supremeum +1).
To finish the proof, note that
∞ [
\ ∞ ∞
[ ∞
[ ∞
[
µ( Ek ) = lim µ( Ek ) = µ( Ek ) − µ( Ek ) = 0
k→∞
n=1k=n k=n k=n k=n
Problem 13 This is pretty obvious based on the way we define the extended
real line and the various operations on them.
