Solutions Homework, MATH 515, Spring 2011

Chapter 2, 1:) Since A ⊆ B it follows B = A ∪ (B \ A) with B \ A ∈ A and

A ∩ (B \ A) = ∅. Since m is countably additive over disjoint unions m(B) =
m(A ∪ (B \ A)) = m(A) + m(B \ A) ≥ m(B) because m : A → [0, ∞]. (Note
that the inequality is trivially true if m(B) = ∞, and m(B \ A) = ∞ is only
possible if m(B) = ∞.)
Chapter 2, 2:) Since ∅ ∈ A and A ∩ ∅ = ∅ it follows m(A) = m(A ∪ ∅) =
m(A) + m(∅). It follows that m(∅) = 0 because for extended real numbers
x, y ≥ 0, x + y = x and y 6= 0 is only possible if x = ∞. (Note that if m(A) = ∞
for all sets in A then of course m(∅) = ∞ too. Such m trivially satisfies the
assumptions given. Also note that very well A = {∅} is possible.)
Chapter 2, 3:) Replace Ek by Ek0 := Ek \ (∪kj=1 Ej ) for k ≥ 2. Then the
collection of sets {Ek0 } is disjoint with the same union as ∪∞
k=1 Ek . The sets
Ek0 are in A because A is a σ-algebra and m(Ek0 ) ≤ m(Ek ) because of the
monotonicity proven in Problem 1. It follows from countably additivity with
respect to disjoint unions
∞
X ∞
X
m(∪∞ ∞ 0
k=1 Ek ) = m(∪k=1 Ek ) = m(Ek0 ) ≤ m(Ek ).
k=1 k=1

Chapter 2, 4:) c is a set function by definition. Let {Ek } be a countable

collection of disjoint sets. Note that a finite union of disjoint finite sets is finite
and the numbers of elements add up for the disjoint union. Thus additivity
holds in this case. If there is one set Ek with infinitely many elements then
also ∪k Ek has infinitely many elements, independent of whether we have a
finite or countably infinite
P collection of sets. In this case both sides of the
equation c(∪k E)k ) = k c(Ek ) are ∞ and thus holds. If all sets are finite
with infinitely many nonempty the set ∪∞ k=1 Ek hasP infinitely many elements
∞
and c(∪∞ E
k=1 k ) = ∞. But in this case in the sum k=1 c(Ek ) also infinitely
many summands c(Ek ) are ≥ 1 and thus the series diverges and has value
∞. Thus countable additivity holds in all cases. Translation invariance follows
because translation by a number y is a bijection of sets (with inverse function the
translation by −y) and thus does not change cardinalities. But the cardinality
of a subset of R determines the value of c.
Chapter 2, 5:) The outer measure of the interval [0, 1] is its length and thus
1. The outer measure of each countable set is 0. Since 1 6= 0 it follows [0, 1] is
not countable
Chapter 2, 6:) Let A respectively B be the set of irrationals respectively ratio-
nals in the interval [0, 1]. Then m∗ (B) = 0 because B is a subset of a countable
set and thus countable. Also A ∪ B = [0, 1] and by countable subadditivity
1 = m∗ ([0, 1]) = m∗ (A ∪ B) ≤ m∗ (A) + m∗ (B) = m∗ (A). Because A ⊆ [0, 1] by
monotonicity m∗ (A) ≤ m∗ ([0, 1]) = 1. Thus m∗ (A) = 1.
Chapter 2, 7:) By definition of outer measure for each positive integer k there
(k)
exists a countable collection of open bounded intervals {Ij }j covering E such
(k) (k)
that j `(Ij ) < m∗ (E) + k1 . Then ∪j Ij is a union of open sets and thus
P
(k)
open for each positive integer k and G := ∩∞
k=1 (∪j Ij ) is a Gδ -set. For each k
(k)
we have E ⊆ G ⊆ ∪j Ij . From this we get by monotonicity m∗ (E) ≤ m∗ (G) ≤
m∗ (E) + k1 for all positive integers k. It follows that m∗ (G) = m∗ (E).
Chapter 2, 8:) Note that the union of two intervals is disjoint or an interval.
Thus [0, 1] \ (∪nk=1 Ik ) is a disjoint union of a set of finitely many points C =
{c1 , . . . , cs } and intervals (To see this just write ∪nk=1 Ik as a union of disjoint
intervals (ai , bi ), i = 1, . . . , r with bi < ai+1 .) The points are irrational but
actually no intervals can occur because each interval contains rational numbers.
Then m∗ (C) = 0 and by subadditivity and monotonicity:
n
X
`(Ik ) ≥ m∗ (∪nk=1 Ik ) ≥ m∗ ([0, 1]\C) = m∗ ([0, 1]\C)+m∗ (C) ≥ m∗ ([0, 1]) = 1
k=1

Chapter 2, 14:) Given E with m∗ (E) > 0 let En := E ∩ (−n, n) for n ∈ N.

Then E = ∪n∈N En and each En is bounded. By countable subadditivity
X
0 < m∗ (E) ≤ En .
n∈N

If m∗ (En ) = 0 for all n ∈ N then the series has value 0, which is a contradiction.
Thus En ⊆ E has outer measure > 0.
Chapter 2, 15:) First assume E ⊆ [−M, M ] is bounded. Then a finite number
of disjoint intervals of the form Jk := (kε/2, (k + 1)ε/2] will cover [−M, M ] (by
the Archimedian property) and thus E. Since Ek := M ∩ Jk is measurable of
measure ≤ ε we have E a disjoint union of measurable sets Ek with m(Ek ) < ε.
If E is unbounded coverP E by a countably infinite collection (Ik ) of bounded
open intervals such that k `(Ik ) < m(E) + 1. Since the series converges there
PN
exists N such that k=1 `(Ik ) < ε. Let E1 := E ∩ (∪∞ k=N +1 Ik ). Then E1 is
measurable of measure < ε and E2 := E \ E1 ⊂ ∪N k=1 k is bounded and can be
I
written by the above as a finite union of disjoint measurable sets with each of
measure ≤ ε.
Chapter 2, 19:) It follows from Theorem 11, (i) that E not measurable implies
that there exists ε > 0 such that for each open set O ⊇ E, m∗ (O \ E) ≥ ε.
There exists a countable collection of open bounded intervals {Ik } such that
) < m∗ (E) + ε. Then O := ∪k Ik ⊇ E is an open set and m∗ (O) ≤
P
Pk `(Ik
∗ ∗ ∗ ∗
k `(Ik ) < m (E) + ε, which implies m (O) − m (E) < ε. Thus m (O \ E) ≥
∗ ∗ ∗ ∗ ∗
ε > m (O) − m (E) and thus m (O \ E) > m (O) − m (E).
Chapter 2, 22:) Since O ⊇ A implies m∗ (O) ≥ m∗ (A) it follows that m∗ (A) ≤
inf{m∗ (O)|O ⊇ A, O open} = m∗∗ (A). If m∗ (A) = ∞ the other inequality is
obvious. Otherwise, for each ε > 0 there exists an open set O ⊇ A, namely
a countable union of open intervals, such that m∗ (O) < m∗ (A) + ε. Thus
m∗∗ (A) ≤ m∗ (A) + ε for all ε > 0, which implies m∗∗ (A) ≤ m∗ (A) and thus
proves m∗∗ (A) = m∗ (A).
Chapter 2, 23:) If F ⊆ A then m∗ (F ) ≤ m∗ (A) and thus m∗∗∗ (A) =
sup{m∗ (F )|F ⊆ A, F closed} ≤ m∗ (A). If m∗∗∗ (A) = ∞ or m∗ (A) = 0 then the
other inequality holds obviously. In particular this is true if there exists F ⊆ A
closed with m∗ (F ) = ∞. So we can assume that no such F exists. If A is mea-
surable then by Theorem 11, (iv) for each ε > 0 there exists F ⊆ A closed such
that m∗ (E \ F ) = m∗ (E) − m∗ (F ) < ε, or m∗ (F ) > m∗ (E) − ε (note that exci-
sion works since F is closed and thus measurable). Thus we can make the values
m∗ (F ) as close as we want to m∗ (E). It follows that m∗∗∗ (E) = m∗ (E). Now as-
sume that E is not measurable. By Theorem 11, (iv) there exists ε > 0 such that
for each closed subset F ⊆ E we have m∗ (E \ F ) = m∗ (E) − m∗ (F ) ≥ ε. Thus
m∗ (F ) ≤ m∗ (E)−ε with ε > 0 fixed. Thus also m∗∗∗ (E) ≤ m∗ (E)−ε < m∗ (E)
in this case. (For the proof of the equivalency (i) and (iv) use that if O ⊇ E c is
open then Oc ⊆ E is closed and O \ E c = E \ Oc .)
Chapter 2, 24:) If E1 , E2 are measurable then all sets E1 ∪ E2 , E1 ∩ E2 ,
E2 \ (E1 ∩ E2 ) are measurable. If m(E1 ∩ E2 ) < ∞ we get from additivity of the
measure and excision: m(E1 ∪E2 ) = m(E1 ∪(E2 \E1 )) = m(E1 )+m(E2 \E1 ) =
m(E1 ) + m(E2 \ (E1 ∩ E2 )) = m(E1 ) + m(E2 ) − m(E1 ∩ E2 ). If m(E1 ∩ E2 ) = ∞
then by monotonicity of the measure we get m(E1 ) = ∞ and m(E2 ) = ∞ and
thus m(E1 ) + m(E2 ) = ∞ = m(E1 ∪ E2 ) + m(E1 ∩ E2 ).
Chapter 2, 30:) If CE is the choice set for the rational equivalence relation on
a set E then E ⊂ ∪a∈Q (CE + a) and thus is a countable union of countable sets.
Thus E is countable and has outer measure 0 contradicting the assumption that
E has positive outer measure.
Chapter 2, 33:) The argument for Problem 7, see also Problem 18, only
required m∗ (E) < ∞ and not the stronger assumption E bounded. So there
exists a Gδ -set G ⊇ E such that m∗ (G) = m∗ (E). If m∗ (G \ E) = 0 then E
would be measurable by Theorem 11 (ii). Thus m∗ (G \ E) > 0.
Chapter 2, 38:) Let f : [a, b] → R be a Lipschitz-map with constant c > 0.
We show first that f maps a set E of measure zero to a set of measure zero. Let
ε > 0. An open bounded interval I ⊂ R is mapped to a set of diameter ≤ c`(I)
(i. e. any two points in f (I) have distance ≤ c`(I). Thus f (I) is contained in
P ≤ c`(I). Cover E by countably many open bounded
some interval of length
intervals {Ik } with k `(Ik ) < εc . For each k choose an open bounded interval
Jk ⊂ R such that Jk ⊆ f (Ik ) and `(Jk ) ≤ c`(I P k ). Then P {Jk } is a collection of
open bounded intervals covering f (A) and k `(Jk ) ≤ k c`(Ik ) < cε/c = ε.
Thus f (A) has measure zero. Next we show that f maps Fσ -sets to Fσ -sets.
Since [a, b] is compact and f is continuous, f maps closed sets to closed sets.
Since f (∪i Ai )) = ∪i f (Ai ) for all unions of sets Ai it follows that f maps Fσ -
sets to Fσ -sets. [Remark: The condition of boundedness of the domain is not
necessary to see that the continuous map f maps Fσ -sets to Fσ -sets. Each
Fσ -set can also be written as a countable union of compact sets because each
closed set is a countable union of compact sets by the Heine Borel theorem
(just intersect the set with closed intervals [−n, n]). Since f maps compact
sets to compact sets it will map Fσ -sets to Fσ -sets. But note that Lipschitz
maps do not necessarily map closed sets to closed sets: f : R → R defined by
1
f (x) = arctan(x) is Lipschitz because it has derivative 1+x 2 , which is bounded,
π π
but it maps the closed set R onto the set (− 2 , 2 ), which is not closed in R.]
Finally by Theorem 11, (iv) we know that each measurable set E contains an
Fσ -set F such that m∗ (E \ F ) = 0. Then f (E) = f (F ) ∪ f (E \ F ). But both
f (F ) and f (E \ F ) are measurable by the previous two observations, and thus
f (E) is measurable since a union of measurable sets is measurable.
Chapter 2, 45:) A strictly increasing function is one-to-one and thus a bijec-
tion onto its image. Let f : I → f (I) be strictly increasing on some interval
I. Claim: f −1 : f (I) → I is continuous. Note that continuity means continu-
ity at each point. Thus it suffices to show that f −1 |[c, d] : [c, d] → [a, b] with
a = f −1 (c) and b = f −1 (d) is continuous for all c, d ∈ f (I). But f |[a, b] maps
closed sets to closed sets (because closed subsets of [a, b] are compact and con-
tinuous maps map compact sets to compact sets, and compact sets are closed).
Thus the inverse image of a closed set under f −1 |[c, d] is closed and f −1 |[c, d]
and thus also f −1 is continuous.
Chapter 2, 46:) Let f : R → R. Let Af := {E ⊆ R|f −1 (E) is Borel} ⊆ 2R .
Claim: Af is a σ-algebra containing the open sets. (This implies that f −1 (B)
is Borel ( ⇐= B ∈ Af ) for each Borel set B because the collection of Borel
sets is the smallest σ-algebra containing the open sets and thus Af contains all
Borel sets.) Since f is continuous the inverse images of open sets are open, and
thus Af contains all open sets. In particular R ∈ Af since f −1 (R) = R is Borel.
Let E ∈ Af and thus f −1 (E) Borel. Then R \ f −1 (E) = f −1 (R \ E) is Borel
and thus R \ E ∈ Af . Let (Ek ) be a countable sequence of sets in Af so that
f −1 (Ek ) is Borel for all k. Then f −1 (∪k Ek ) = ∪k f −1 (Ek ) is also Borel since
the Borel sets form a σ-algebra. Thus ∪k Ek ∈ Af . This proves the claim. (If f
is defined on an interval E only Borel set has to be interpreted as Borel set in
E, i. e. as the smallest σ-algebra containing all open subsets of E. The result is
true in this case too.)
Chapter 2, 47:) By Problem 45 the inverse function is continuous, and thus
the inverse images of Borel sets under the inverse functions are Borel sets. But
this precisely means that the function itself maps Borel sets to Borel sets.
Chapter 3, 6:) Note that the domain of g is the measurable set R. The
following set equalities are checked from the definition of g. For c ≥ 0 it follows
that {x ∈ R|g(x) > c} = {x ∈ D|f (x) > c}. If c < 0 then {x ∈ R|g(x) > c} =
{x ∈ D|f (x) > c} ∪ (R \ D) and {x ∈ D|f (x) > c} = {x ∈ R|g(x) > c} ∩ D. If
f is measurable then g is measurable because R \ D is measurable and unions
of measurable sets are measurable. If g is measurable then f is measurable
because D is measurable and intersections of measurable sets are measurable.
[Alternatively: By Proposition 5 (ii) we know that g is measurable if and only
if g|D and g|R\D, which is the function constant zero, are measurable. Because
a constant function is measurable the equivalence follows.]
Chapter 3, 7:) We first show that for each function f : E → R with measurable
domain E the collection Bf of sets A ⊆ R such that f −1 (A) ⊆ E is measurable
is a σ-algebra. Note that f −1 (R) = E, which is measurable by assumption. If
A ∈ Bf then f −1 (A) is measurable. Thus E \f −1 (A) = f −1 (R\A) is measurable
because complements of measurable sets in measurable sets are measurable.
Thus R \ A = Ac ∈ Bf . Finally let {Ak } be a countable collection of sets in
Bf . Then f −1 (Ak ) is measurable for each k. Thus ∪k f −1 (Ak ) = f −1 (∪k Ak ) is
measurable because unions of measurable sets are measurable. Thus ∪k Ak ∈ Bf .
Now suppose that f is measurable. Then inverse images of open sets are
measurable by Prop. 2 and thus Af contains the open sets. Because it is a
σ-algebra containing the open sets it contains all Borel sets because the Borel
algebra is the smallest σ-algebra containing the open sets. Thus Af contains the
Borel algebra. This means that the inverse image of each Borel set is measurable.
Conversely, each open set U is a Borel set. Because f −1 (U ) is measurable for
all open sets U it follows that f is measurable by Prop. 2.
Chapter 3, 11:) From the remark about the solution to Problem 38 in Chapter
2 above it follows that the result that a Lipschitz function maps measurable sets
to measurable sets also holds for Lipschitz functions with domain R. Thus for
U ⊆ R open we have f −1 (U ) measurable by Proposition 2. Thus g −1 (f −1 (U )) =
(f ◦ g)−1 (U ) is measurable. Then again by Proposition 2 it follows that f ◦ g
is measurable. [More directly without reference to the remark to the solution of
Problem 38 we can argue as follows (but are in fact using the same argument):
f −1 (U ) is measurable and thus f −1 (U ) = V ∪ W with V an Fσ -set and W
a set of measure 0 (Theorem 11 in Chapter 2). Then we know by the proof
given for Problem 38 that g −1 maps sets of measure 0 to sets of measure 0 (this
part of the argument did not require compactness of the domain). Since g −1 is
continuous it maps Fσ -sets to Fσ -sets. Thus g −1 (V ∪ W ) = g −1 (V ) ∪ g −1 (W )
is the union of a set of measure 0 and an Fσ -set and thus is measurable.]
Chapter 3, 21:) Let f := inf{fn }. Note that f (x) < c ⇐⇒ fn (x) < c for
some n. Thus for all c ∈ R, {x ∈ E|f (x) < c) = ∪n {x ∈ E|fn (x) < c} (⊇
follows from f ≤ fn for all n, and ⊆ follows because if f (x) < c then there
exists some n such that fn (x) < c by the definition of inf) is a countable union
of measurable sets and thus measurable. It follows by Proposition 1 that f is
measurable. By Theorem 6 since fn is measurable also −fn is measurable. Thus
inf{−fn } and by Theorem 6 again also − inf{−fn } = sup{fn } is measurable.
Finally lim inf{fn } = supn inf k≥n {fk } and lim sup{fn } = inf n supk≥n {fk } are
measurable. (See Definition 19 in Chapter 1 for the definition of lim inf and
lim sup and note that the sequence of sets An := {ak |k ≥ n} is a decreasing
sequence of subsets of R and thus inf Ak respectively sup Ak is an increasing
respectively decreasing sequence of real numbers.)
Chapter 3, 22:) Note that for fn increasing with fn (x) → f (x) for n → ∞ it
follows that |f (x) − fn (x)| = f (x) − fn (x). For each ε > 0 let
En := {x ∈ E|f (x)−fn (x) < ε}. Then {En } is a covering of E and the sequence
of sets (En ) is increasing: If f (x) − fn (x) < ε then also f (x) − fn+1 (x) < ε,
and if x ∈ E then because of convergence of fn (x) to f (x) there exists some
N such that f (x) − fN (x) < ε and thus x ∈ EN . Since f − fn is continuous
for each n, it follows that En = (f − fn )−1 (−∞, ε) = (f − fn )−1 (−ε, ε) is
open, being the preimage of an open set under a continuous function. Thus
the covering {En } is an open covering. By compactness of [a, b] there is a
finite subcovering {Ek1 , . . . , Ekr }. If N := max(k1 , . . . , kr ) then EN ⊇ [a, b].
Thus |f (x) − fn (x)| < ε for all x ∈ E and all n > N (where we use that
f (x) − fn (x) ≤ f (x) − fN (x) for n ≥ N ).
Chapter 3, 23:) Let f + := max(f, 0) and f − := − max(−f, 0). Then both
f ± ≥ 0 are measurable by Proposition 8, and f = f + −f − , also |f | = f + +f − =
|f + | + |f − |. By the simple approximation theorem we can find sequences ϕ± n
of simple functions such that ϕ± n → f
±
pointwise on E and 0 ≤ ϕ± ±
n ≤ f .
−
Then ϕn := ϕ+ n − ϕn is a sequence of simple functions converging pointwise to
− − −
f = f − f , and |ϕn | = |ϕ+
+ + +
n − ϕ− | ≤ |ϕn | + |ϕn | ≤ f + f = |f |. This
construction coincides with the old construction from the proof of the Simple
Approximation Theorem, and thus concludes the proof.
Chapter 3, 24:) Let g be a strictly increasing function on an interval I.
For c ∈ R let Ac := g −1 (c, ∞). Case 1: c ∈ / g(I). If c ≥ g(t) for all t ∈
I then Ac = ∅. Otherwise there exist t ∈ I such that c < g(t). Then for
tc := inf{t|g(t) > c} we have Ac = [tc , ∞) ∩ I respectively Ac = (tc , ∞) ∩ I if
f (gc ) > c respectively g(tc ) < c (Note that by the strictly increasing property
t > tc implies g(t) > g(tc ) and t < tc implies g(t) < g(tc ). Case 2: c ∈ g(I) then
with the uniquely determined tc such that g(tc ) = c, we have Ac = (tc , ∞) ∩ I.
Thus Ac is a measurable set in each case, and g is measurable. If f is increasing
but not necessarily strictly then the sequence of strictly increasing functions fn
defined by fn (x) = f (x) + nx is a sequence of measurable functions by the above,
which pointwise converges to f . Thus f is measurable by Proposition 9.
Chapter 3, 25:) We can assume F 6= ∅ because otherwise any constant func-
tion is a continuous extension. Define an extension g : R → R as described in
the hint. Let R \ F = ∪k Ik with countably many disjoint open intervals Ik . If
an interval Ik is unbounded then define g by the function, which is constant
with the value of f (x) in the uniquely determined endpoint. If x ∈ Ik for some
k then g(x) is defined by an affine function (the sum of a linear function and
a constant) in a neighborhood of x and thus g is continuous at x. If x ∈ F
then let ε > 0. Since g|F = f is continuous there exists δ1 > 0 such that
|g(y) − g(x)| < ε for all y ∈ F with |x − y| < δ1 . Consider some k such that
x ∈ Ik . If x ∈ Ik then x is left or right endpoint of the interval Ik . Suppose it
is the left endpoint. Then there exists some δ2 > 0 such that (x, x + δ2 ) ⊂ Ik .
Similarly if x is a right endpoint of some interval I` then there exists some
δ3 > 0 such that (x − δ3 , x) ⊂ I` . It follows that x is not in the closure of any
other interval with index 6= k, ` because the intervals are disjoint. Note that
there are three possibilities depending on whether x is in the closure of none,
one or two intervals. If x is in the closure of two intervals then we can choose
δ2 , δ3 > 0 such that |f (y) − f (x)| < ε for each x ∈ (x − δ3 , x + δ2 ) and we
have |f (y) − f (x)| < ε for x ∈ (x − δ, x + δ) with δ = min(δ2 , δ3 ). Note that
in this case no point of F different from x is in (x − δ, x + δ). The other two
cases are similar using the minimum of δ1 above and one of δ2 , δ3 . Thus f is a
continuous function with domain R. (Alternatively it is not hard to argue that
the oscillation of the extension g on a compact neighborhood of x ∈ F is ≤ to
the oscillation of f on that neighborhood. This also easily implies the claim.)
Chapter 4, 17:) f ≥ 0 is measurable because its domain is a set of measure
0. Because f is non-zeroR only on a set of measure, namely E itself, it follows
from Proposition 9 that E f = 0.
Chapter 4, 19:) For α ≥ 0 the function f : [0, 1] → R is piecewise continuous
and bounded and thus RiemannR1 integrable. By Theorem 4 of Chapter 3 the
1
function is integrable and 0 f = α+1 . For α < 0 and n ≥ 1 let fn : [0, 1] → R
be defined by fn (x) = 0 for x ∈ [0, n1 ) and let f (x) = xα for x ∈ [ n1 , 1].
The functions fn are Riemann integrable and thus again integrable on [0, 1]
with Riemann integral equal to the Lebesgue integral. The sequence (fn ) is
an increasing sequence Rof nonnegative functions converging pointwise R 1 to f . For
1 1 1
α < 0, α 6= −1 we have 0 fn = α+1 (1− nα+1 ). For α = −1 we have 0 fn = ln n.
R1 1
R1
So for −1 < α < 0 we have 0 fn → α+1 while for α < −1 we have 0 fn →
R1 1
∞. Thus by the monotone convergence theorem 0 f = α+1 for α > −1 and
R1
0
f = ∞ for α ≤ −1.
R R
Chapter 4, 22:) First note that by integrability of R f over R: E f R≤ R f <
∞. Suppose R for the sake of contradiction that E f 6= limn→∞R E fn . If
lim inf n→∞ R\E fn = ∞ we also get by monotonicity lim inf n→∞ R fn = ∞
R R R
contradicting R f = limn→∞ R f < ∞. Thus we can assume lim inf n→∞ R\E fn
< ∞. We need this below R to argue that inequalities
R remain strict if we add the
inequality lim inf n→∞ R\E fn ≤ lim supn→∞ R\E fn respectively the Fatou in-
R R
equality R\E f ≤ lim inf n→∞ R\E fn to a strict inequality. First assume that
R R R
lim inf n→∞ E fn < lim supn→∞ E fn . Because we have lim inf n→∞ R\E fn ≤
R R
lim supn→∞ R\E fn , by additivity over domains of integration lim inf n→∞ R fn
R R R
< lim supn→∞ RRfn contradicting R fR = limn→∞ R fnR. So we can assume
that lim inf n→∞ E fn = lim supn→∞ E fn = limn→∞ R E fn , and by RFatou’s
Lemma from R the assumption for contradiction that E
f R< limn→∞ E fn =
lim inf n→∞ E fn . By Fatou’s Lemma again we know that R\E f ≤
R
lim inf n→∞ R\E fn and again by the additivity over domains of integration we
R R R
get R f < lim inf n→∞ R fn ≤ limn→∞ R fn contradicting the assumption.
Chapter 4, 24:) (i) By the simple approximation theorem there exists an
increasing sequence (ϕn ) of simple nonnegative functions converging to f . Let
En := E ∩(−n, n) and ψn := ϕn χn . Then ψn is an increasing sequence of simple
functions of finite support converging to f . Note that for each x ∈ E there is N
such that ϕn = ψnR for n > N .
R sup{ E ϕ|ϕ simple of finite support and 0 ≤ ϕ ≤ f }. Then obvi-
(ii) Let s :=
ously s ≤ E f by definition. But because of Rthe monotone R convergence theo-
rem Rthere exists a sequence
R ψ n as in (i) with E
ψn → E
f , which contradicts
s < E f . Thus s = E f .
R R
Chapter 4, 25:)R By Fatou’s R Lemma E f ≤ lim inf n→∞ R fR . Because fn ≤ f
E n
by monotonicity E fn ≤ E f and thus lim supn→∞ E fn ≤ E f . Thus
Z Z Z
lim sup fn ≤ f ≤ lim inf fn
n→∞ E E n→∞ E
R R
and thus E
f = limn→∞ E
fn .
Chapter 4, 27:) Consider the sequence gn := inf k≥n fk of measurable func-
tions. ThenR0 ≤ gn ≤ fn and gn → lim inf n→∞ R fn . Applying Fatou’s
R Lemma
to gn gives E lim inf n→∞ fn ≤ lim inf n→∞ E gn ≤ lim inf n→∞ E fn because
of the monotonicity of the integral and lim inf.
Chapter 4, 30:) By possibly excising a set of measure 0 we can assume that
|fn | ≤ g holds on E. Let gn := inf k≥n fk ≤ fn . Then gn → lim inf n→∞ fn . Note
that from −g ≤ fn ≤ g for all n it also follows that −g ≤ gn ≤ g for all n and
thus R|gn | ≤ g. Thus by the LebesgueR dominated convergence
R theorem it follows
R
that E lim inf n→∞ fn = limn→∞ E gn = lim inf n→∞ E gn ≤ lim inf n→∞ E fn
by monotonicity of the integral. Thus
Z Z Z
lim inf fn ≤ lim inf fn ≤ lim sup fn
E n→∞ n→∞ E n→∞ E

Similarly let hn := supk≥n fk ≥ fn such that hn → lim supn→∞ fn and note

that
R |hn | ≤ g as above. Again
R by Lebesgue dominatedR convergence we R get
E
lim sup n→∞ fn = lim n→∞ E
hn = lim sup n→∞ E h n ≥ lim supn→∞ E fn ,
so the result follows. Alternatively we could apply the first part to the sequence
−fn to get the second inequality. Also, a different proof can be given by applying
Fatou’s lemma to g + fn and g − fn , thus following the idea of the proof of the
dominated convergence theorem, and not using the result.
Chapter 4, 32:) As usual, by possibly excising a set of measure zero we can
assume convergence on the set E. Just as in the proof of the dominated con-
vergence theorem we can also assume that all functions are finite. Note that
gn + fn , gn −R fn ≥ 0.
R Since R|fn | ≤ gn for all n also |f | ≤ g and thus by com-
parison test E f ≤ E |f | ≤ E gR < ∞.RSo we Rknow fn , f are integrable. So by
Fatou’s Lemma
R using linearity:
R g − E f = E g − f R≤ lim inf n→∞ gn − fn =
E R
limn→∞
R E
gn −lim sup n→∞ E
fn = E
g−lim supn→∞ E gn and afterR subtract-
R
ing E g on both sides and multiplication by −1 we get lim supn→∞ E fn ≤ E f .
 R R
Similarly by applying Fatou’s Lemma to g + fn we get E
f ≤ lim inf n→∞ E
fn ,
and the result follows.
Chapter 4, 37:) (En ) is a decreasing sequence
R of RmeasurableRsets and ∩n∈N En =
∅. By continuity of integration limn→∞ En f = ∩n En f = ∅ f = 0. Thus for
R
each ε > 0 there is a natural number N such that | En f | < ε for n ≥ N . Note
that integrability
R of f is defined by integrability of |f | so actually the stronger
statement En |f | → 0 for n → ∞ holds.

Chapter 4, 39:) (i) Note that f integrable over E implies that f is integrable
over all measurable subsets of E. Let E0 := ∅ and for k ≥ 1 let Ck := Ek \Ek−1 .
Then ∪n≥0 En = ∪n≥1 Cn and by the countable additivity of integration
Z Z ∞ Z
X ∞ Z
X Z n Z
X Z
f= f= f= f− f = lim f− f=
∪Ek ∪Ck Ck Ek Ek−1 n→∞ Ek Ek−1
k=1 k=1 k=1
Z Z
lim f− f = lim f.
n→∞ En E0 n→∞

(ii) Define Dk := E1 \ Ek for k ≥ 1. The sequence (Dk ) is ascending and we

know by (i) Z Z
f = lim f.
∪k Dk k→∞ Dk
R R R
The result follows from ∪k Dk = E1 \ ∩k Ek and Dk f = E1 f − Ek f , which
R R R R R
implies E1 f − ∩k Ek f = E1 \∩k Ek f = limn→∞ E1 f − En f and this implies
R
the result by subtracting E1 f and multiplying by −1.

Chapter 4, 48:) Since f, g are measurable we Rknow thatR f g is measurable

R by
Ch. 3, Theorem 6. Moreover, if |g| ≤R M then E |f g| ≤ E M |f | ≤ M E f <
∞ by monotonicity and linearity of for nonnegative functions. Thus f g is
integrable.
Chapter 4, 49:) (i) =⇒ (ii): RIf f = 0 a. e. and g is bounded then f g = 0
a. e. and f g is measurable so R f gR = 0. (ii)R =⇒ (iii): g = χAR is bounded
and measurable and so by (ii) and R f χA = A f it follows that A Rf = 0 for
every measurable set A. (iii) =⇒ (iv): O is measurable and thus O f = 0
for each open set O. (iv) =⇒ (i): Suppose that f 6= 0 a. e. and consider
E := {x ∈ R|f (x) > 0} and {x ∈ R|f (x) < 0}. If both sets have measure zero
then f = 0 a. e. Thus we can assume that E has positive measure (otherwise
replace f by −f ). Since E = ∪∞ n=1 En with En := {x ∈ R|f (x) > n } we
1

also find some n ≥ 1 such that m(En ) > 0 because otherwise the countable
union Rhas measure 0. RNow find aR Gδ -set G ⊇ En such that m(G \ En ) = 0.
Then G f = En f + G\En f = En f ≥ n1 m(En ) > 0. Now G = ∩Ok for a
R

countable descending sequence of open sets (If G = ∩Uk for open sets Uk let
k
R k := ∩j=1 Uj . Then the Ok are also open and form a descending sequence.).
O R R If
Ok
f = 0 for each k then also by continuity of the integral 0 = limk Ok f = G f
 R
contradicting
R G
f > 0. thus for some k and so for some open set O = Ok we
have O f 6= 0. This proves (iv) =⇒ (i) and thus the equivalence of the four
statements.
Chapter 6, 9:) P⇐=: Suppose m(E) > 0 and let (Ik ) be a covering of E by open
intervals with `(Ik ) < ∞ By the Borel Cantelli lemma x ∈ Ik for infinitely
many k can be true at most on a set of measure zero. Thus there exists at least
one x ∈ E such that x is in Ik for at most finitely many k. =⇒: Suppose that
m(E) = 0. For each natural number n find a covering of E by countably many
(n) P∞ (n)
open intervals {Ik }k such that k=1 `(Ik ) < 21n . Consider the collection of
(n) (n) P∞
all those intervals {Ik }n,k . Then n,k `(Ik ) ≤ n=1 21n < ∞, and for each
P
(n) (n)
x and positive integer n there exists some k such that x ∈ Ik , thus x ∈ Ik
for infinitely many (n, k). (Note that I interpreted belongs to infinitely many
Ik ’s as belongs to Ik for infinitely many k, which actually is a slightly different
statement. But this is what is needed for the next problem. Note also that
the intervals necessarily are not disjoint. This seems to have been a common
misunderstanding in approaches to the next problem.)
Chapter 6, P 10:) Let Ik = (ck , dk ) be the covering of E P by open intervals
such that `(Ik ) < ∞. Note that this makes the series `(Ik ∩ (−∞, x))
convergentPbecause `(Ik ∩ (−∞, x)) ≤ `(Ik ). Let fk (x) := `(Ik ∩ (−∞, x)) such
that f = fk . If y > x then fk (y) − fk (x) ≥ min(`(x, y), (x, dk )) ≥ 0 for all
k and thus f (y) ≥ f (x). This shows that f is increasing. Let x ∈ E and let
{k1 , k2 , . . .} ⊆ N be the infinite set of those k for which x ∈ Ik . Fix a number
N and consider the open interval Ik1 ∩ . . . ∩ IkN containing x. Then we can
find some tN > 0 sufficiently small such that x + tN ∈ Ik for k = 1, . . . , N .
Thus fk (x + tN ) − fk (x) = `(x, x + tN ) = tN for k = k1 , . . . , kN . Because
fk (x + tN ) − fk (x) ≥ 0 for all k it follows that f (x + tN ) − f (x) ≥ N tN and
thus Df (x) ≥ N . Because this holds for each N we have Df (x) = ∞ and thus
f is not differentiable at x.
Chapter 6, 12:) Let f = χQ . Case 1: x ∈ Q and thus f (x) = 1. Consider
0 < |t| < h. If x + t ∈ Q then f (x + t) = 1 and f (x+t)−f t
(x)
= 0. If x + t ∈
/Q
f (x+t)−f (x) 1
then f (x + t) = 0 and t = − t . Because for each h > 0 there exist
both positive and negative t with 0 < |t| < h for both cases we get Df (x) = ∞
and Df (x) = −∞. Case 2: x ∈ / Q and thus f (x) = 0. This similar with
f (x+t)−f (x) f (x+t)−f (x)
t = 0 if x + t ∈/ Q and t = 1t if x + t ∈ Q. Again because
there exist for each h > 0 both positive and negative t with x + t ∈ Q it follows
Df (x) = ∞ and Df (x) = −∞. Note that f is nowhere differentiable.
t sin( 1 )
Chapter 6, 15:) For h > 0 and 0 < |t| < h we get f (t)−f t
(0)
= t
t
= sin( 1t ).
Now for h > 0 we have g(t) = t we have g((−h, 0)∪(0, h)) = (−∞, − h )∪( h1 , ∞),
1 1

which contains an interval of length 2π for all h > 0. Thus by the intermediate
value theorem {sin( 1t )|0 < |t| < h} = [−1, 1] and thus Df (0) = −1 and Df (0) =
1.
Chapter 6, 26:) If f were of bounded variation then it would be differentiable
almost everywhere by Corollary 6 in Chapter 6. But Problem 12 above shows
that f is not differentiable on (0, 1). Alternatively it is easy to construct par-
titions Pn such that V (f, Pn ) = n to see T V (f ) = ∞ to see that f is not of
bounded variation.
Chapter 6, 27:) The easiest solution is just f = g − h with g(x) = sin(x) + x
and h(x) = x. Then g and h are increasing because g 0 (x) = cos(x) + 1 ≥ 0
and h0 (x) = 1 > 0 for all x ∈ [0, 2π]. The standard Jordan decomposition
with g(x) = f (x) + T V (f[0,x] ) and h(x) = T V (f[0,x] ) is also easy to deduce by
calculating the total variation function using additivity: T V (f[0,x] ) = sin(x) for
0 ≤ x ≤ π/2, T V (f[π/2,x] ) = 1 − sin(x) for π/2 ≤ x ≤ 3π/2, T V (f[3π/2,x] ) =
sin(x) + 1 for 3π/2 ≤ x ≤ 2π. Note that in particular T V (f[0,π/2] ) = 1 and
T V (f[π/2,3π/2] ) = 2. Then T V (f[0,x] is sin(x) on [0, π/2], 2−sin(x) on [π/2, 3π/2]
and 4 + sin(x) on [3π/2, 2π].
Chapter 6, 38:) Claim: f is absolutely continuous on [a, b] if and only if for
each ε > 0 there is a δ > 0 such that for every countable P∞ disjoint collection
∞
{(a
P∞ k , bk )}k=1 of open disjoint intervals of (a, b), if k=1 (b k − ak ) < δ then
k=1 |f (bk )−f (ak )| < ε. Proof. ⇐=: Since a finite collection is also a countable
collection the definition of absolutely continuous is implied. =⇒: Let ε > 0 and
let δ > 0 be such that for each finite collection {(ck , dk )}N k=1 of open disjoint
PN PN
intervals it follows that if k=1 (dk − ck ) < δ then k=1 |f (dk ) − f (ck )| < ε/2.
∞
Now let P∞{(ak , bk )}k=1 be a countably infinite collection of open Pdisjoint
n
intervals
with k=1 (b
Pn k − a k ) < δ. Then for each n ≥ 1 it follows that k=1 (b k − ak ) < δ
and thus
P∞ k=1 |f (b k ) − f (ak )| < ε/2.
PnBecause this holds for each n it follows
that k=1 |f (bk ) − f (ak )| = limn→∞ k=1 |f (bk ) − f (ak )| ≤ ε/2 < ε.
Chapter 6, 39:) Suppose that f is increasing. Then f ((a, b)) ⊂ (f (a), f (b))
and so m∗ (f ((a, b))) ≤ f (b) − f (a). Moreover if O P is a countable disjoint
∗ ∞ ∗
collection
P∞ of open intervals (ak , b k ) then m (f (O)) ≤ k=1 m (f ((ak , bk ))) ≤
k=1 (f (bk ) − f (ak )). Note that we can assume E ⊆ (a, b) because changing
E to E \ {a, b} and correspondingly changing f (E) to f (E \ {a, b}) will not
change the measure of these sets. Next suppose that f is moreover absolutely
continuous. Let ε > 0 and let δ > 0 be suchP that for each countable disjoint
∞ ∞
collection
P∞ of open intervals {(a k , b k )} k=1 with k=1 (bk − ak ) < δ it follows that
k=1 (f (bk ) − f (ak )) < ε. Let E be any measurable set with m(E) < δ. We
know that E = ∩∞ n=1 On for a descending sequence of open sets On ⊂ (a, b).
From the continuity of measure it follows that for n sufficiently large m(On ) <
δ. Because On is a countable collection of open disjoint intervals it follows
that m∗ (f (E)) ≤ m∗ (f (On )) < ε. Conversely suppose that f is increasing
and continuous, and for each ε > 0 there exists δ > 0 such that m(E) < δ
implies m∗ (f (E)) < ε. Let {(ak , bk )}nk=1 be a finite collection of open disjoint
intervals in (a, b). Then m∗ (f ((ak , bk ))) = m((f (ak ), fP (bk )). We can apply
n n
the
Pn assumption to E = ∪k=1 (a k , b k ) and conclude that k=1 f (bk ) − f (ak ) =
k=1 m((f (ak ), f (bk )) < ε. Remark: We have to assume that f is continuous
for this direction because otherwise consider the increasing function: f (x) = 0
for x ∈ [0, 12 ) and f (x) = 1for x ∈ [ 21 , 1]. Then f (x) satisfies the condition on
measure because f ([0, 1]) = {0, 1} has measure zero. But f (x) is not continuous
and thus also not absolutely continuous.
Chapter 6, 40:) Let f be increasing and absolutely continuous. If E has
measure zero then m(E) < δ for each δ > 0. Thus for each ε > 0 we conclude
that m∗ (f (E)) < ε and thus m∗ (f (E)) = 0. If the Cantor-Lebesgue function
ϕ were absolutely continuous then also ψ defined by ψ(x) := x + ϕ(x) because
the set of absolutely continuous functions is a real vector space. But ψ maps
a set of measure 0 onto a set of positive measure. This contradicts that ψ is
abolutely continuous.
Chapter 6, 52:) The product of two absolutely continuous functions f, g on
[a, b] is absolutely continuous on [a, b]. The proof is essentially the same as
for continuity. In fact because f, g are continuous on [a, b] it follows |f |, |g| ≤
M for a constant M > 0. Let ε > 0 be given. We know that there exists
δ > 0P such that if {(ak , bk )}nk=1Pis a finite collection of open intervals
Pn in (a, b)
n n ε
with k=1 (bk − ak ) < δ then k=1 |f (bk ) −P f (ak )| < 2M and k=1 |g(bk ) −
ε n
g(a
Pnk )| < 2M . Thus by the triangle Pn inequality k=1 |f (bk )g(bk ) −P f (ak )g(ak )| ≤
n
k=1 |f (bk )P
− f (ak )||g(bk )| + k=1 |g(bk ) − g(ak )||f (ak )| ≤ M k=1 |f (bk ) −
n
f (ak )| + M k=1 |g(bk ) − g(ak )| < ε/2 + ε/2 = ε. Thus we can apply Theorem
10 to f g and use the product rule on the set of points in [a, b] where f g is
Rb Rb
differentiable to deduce: a (f g)0 = a f 0 g + f g 0 = f (b)g(b) − f (a)g(a), which is
the claim. (Note that the proof of the product rule requires the function to be
defined in a neighborhood of a point but is a pointwise statement.)
Chapter 6, 53:) Suppose that f is absolutely continuous and Lipschitz. Let
x ∈ (a, b) and h sufficiently small such that x + t ∈ (a, b) for 0 < |t| ≤ h. Then
1
|f (x + t) − f (x)| ≤ c|t| for 0 < |t| ≤ h. It follows that |t| |f (x + t) − f (x)| ≤ c
for those x, t. If f exists at x this implies |f | ≤ c. Since f 0 exists a.e. by
0 0

Lebesgue’s theorem we conclude |f 0 | ≤ c a.e. Conversely

Ry 0
0
R yif |f0 | ≤Rcy a.e. on [a, b]
then for x < y in [a, b] we get |f (y) − f (x)| = | x f | ≤ x |f | ≤ x c = c(y − x)
and we conclude that f is Lipschitz with Lipschitz constant c.
Chapter 13, 8:) First we show that ∼ = is an equivalence relation: ||x − x|| = 0
so x ∼= x for all x and ∼
= is reflexive. If ||x − y|| = 0 then also ||y − x|| = 0 so
x∼ = y implies y ∼= x and thge relation is symmetric. If x ∼ = y and y ∼ = z then
||x − y|| = 0 and ||y − z|| = 0 and thus ||x − z|| ≤ ||x − y|| + ||y − z|| = 0 thus
x∼ = z and ∼= is transitive. Let [x] be the equivalence class of x and X/ ∼ = be
the set of equivalence classes. Then for x, y ∈ X and α, β ∈ R define α[x] +
β[y] := [αx + βy] is well-defined because ||x − x0 || = 0 and ||y − y 0 || = 0 implies
||αx+βy−(αx0 +βy 0 )|| ≤ ||α(x−x0 )||+||β(y−y 0 )|| = |α|||x−x0 ||+|β|||y−y 0 || = 0.
The operation of addition and multiplication by scalars defined in this way on
X/ ∼ = make X/ ∼ = a vector space. The axioms of vector space are directly
induced from the corresponding on X. Now define ||[x]|| := ||x||. This is well-
defined because x ∼ = x0 implies ||x0 || = ||x − (x − x0 )|| ≤ ||x|| + ||x − x0 || = ||x||
and similarly ||x|| = ||x0 − (x0 − x)|| ≤ ||x0 || and thus ||x0 || = ||x||. Then ||.||
defines a norm on X/ =. ∼ It suffices to show that ||[x]|| = 0 implies [x] = 0,
the other properties of a norm are induced directly from the properties of the
pseudonorm on X. But ||[x]|| = 0 implies ||x|| = 0, which implies x ∼ = 0 and
Rb p
thus [x] = [0]. This procedure applies to X := {f : [a, b] → R| a |f | < ∞}
Rb
with the pseudonorm ||f || := ( a |f |p )1/p . Note that for X to be a vector space
functions have to be real-valued. Thus we have to restrict to representatives
with values in R. Then the quotient X/ ∼ = is Lp [a, b] with norm ||.||p . Note
Rb p
that functions with a |f | < ∞ are finite a.e. and thus we are not losing any
equivalence class in comparison with respect to the equivalence relation defined
Rb
on F in 7.1. Also note that if a |f − g|p = 0 then f = g a.e., so the equivalence
classes in 7.1. and above coincide.
Chapter 13, 9:) (a): If T is continuous at u0 then for each ε > 0 exists δ > 0
such that ||u − u0 || < δ implies ||T (u − u0 )|| = ||T u − T u0 || < ε. Then if x ∈ X
and ||x|| < δ we have x = (x + u0 ) − u0 and thus with u := x + u0 we have
||u − u0 || < δ and thus ||T x|| = ||T u − T u0 || < ε. Thus T is continuous at 0. It
follows from the argument in the proof of Theorem 1 that T is bounded and thus
by Theorem 1 that T is continuous. Since X 6= ∅ continuity implies continuity
at some point. (b): If T is Lipschitz with constant c then T is bounded with
||T || ≤ c. Thus T is continuous. If T is continuous then T is bounded. Then
T is Lipschitz with constant ||T ||. (c): Let f : R → R be a jump function
like f (x) = 0 for x < 0 and 1 for x ≥ 0. Then f is continuous at 1 but not
continuous. Let f (x) = x2 defined on R. Then |f (x) − f (y)| = |x + y||x − y| and
for e.g.y = 2x we can make y + x = 3x as large as we want, i. e. for each N there
exist x, y such that |f (y) − f (x)| > N |y − x| showing that f is not Lipschitz.
Chapter 13, 20:) ||x||1 = inf{||x − y|| | y ∈ Y } ≥ 0 is a pseudonorm: If
λ = 0 then ||λx|| = ||0|| = inf{|| − y|| | y ∈ Y } = 0 = |λ| ||x||1 because
0 ∈ Y . If λ 6= 0 then y ∈ Y ⇐⇒ λy ∈ Y . Thus ||λx||1 = inf{||λx − y|| | y ∈
Y } = inf{||λ(x − y)|| | y ∈ Y } = |λ| inf{||x − y|| | y ∈ Y } = |λ| ||x||1 . Also
||x1 + x2 ||1 = inf{||(x1 + x2 − y|| | y ∈ Y } = inf{||x1 + y1 − (x2 + y2 )|| | y1 , y2 ∈
Y } ≤ inf{||x1 + y1 || | y1 ∈ Y } + inf{||x2 + y2 || | y2 ∈ Y } = ||x1 ||1 + ||x2 ||1 using
that Y = {y1 − y2 |y1 , y2 ∈ Y }, the triangle inequality for ||.|| and properties
of the infimum. Thus ||.||1 is a pseudonorm on X. Let X/Y be the normed
linear space defined by the equivalence relation x1 ∼ = x2 ⇐⇒ ||x1 − x2 ||1 =
inf{||x1 − x2 − y|| | y ∈ Y } = 0 ⇐⇒ x1 − x2 = y ∈ Y for some y ∈ Y . Let
U ⊂ X open. Then ϕ(U ) = {[x] | x ∈ U } ⊂ X/Y and let [x] ∈ ϕ(U ). If
||[x] − [x0 ]||1 = ||x − x0 ||1 = inf{||x − x0 − y|| | y ∈ Y } < ε then there exists
yn ∈ Y such that ||x − x0 − yn || → ||x − x0 ||1 . Let y 0 = lim yn ∈ Y be the limit,
which is in Y because Y is closed. Note that [x0 ] = [x0 + y 0 ] such x0 + y 0 is in the
ε-ball at x. If ε > 0 is small enough then x0 + y 0 ∈ U since U is open and thus
[x0 ] = [x0 + y 0 ] ∈ ϕ(U ). This shows that the ε-ball at [x] is contained in ϕ(U ).
Chapter 13, 21:) Note that the projection ϕ : X → X/Y is continuous
because xn → x means ||xn − x|| → 0 but ||[xn ] − [x]||1 ≤ ||xn − x|| → 0 and
thus also converges to 0. Let X0 be a complement of Y in the Banach space X,
i. e. X = X0 ⊕ Y (uses Zorn’s lemma). Then ϕ|X0 is an isomorphism of normed
spaces. In fact, it is an algebraic isomorphism. Moreover since ϕ is continuous
on X its restriction to X0 is continuous. Also, since ϕ maps open sets to open
sets, its restriction does the same, and thus ϕ−1 is also continuous. Now let
yn be a Cauchy sequence in X/Y . Then the sequence ϕ−1 (yn ) =: xn ∈ X0 is
a Cauchy sequence in X0 . Note that ϕ−1 is linear and thus we can conclude
by Problem 9 (ii) that ||xn − xm || ≤ c||yn − ym || for some c > 0. The Cauchy
sequence (xn ) in X0 converges to a limit x ∈ X (note that we don’t whether X0
is closed in X). Then by continuity of ϕ it follows that yn converges to ϕ(x) in
X/Y . This proves that X/Y is complete.
Chapter 13, 23:) Let ϕ be a normed linear space isomorphism from X to
Rn for some n. Then ϕ maps bounded sets to bounded sets and closed sets to
closed sets. Thus C ⊂ X is closed and bounded if and only if ϕ(C) is closed
and bounded in Rn if and only if this set is compact in Rn . But the continuous
map ϕ−1 maps compact sets to compact sets, and thus C is compact set in X.
Chapter 13, 24:) Claim: Let Y ⊆ X be a closed linear subspace, Y 6= X.
Then for each ε > 0 there exists a vector x0 ∈ X, ||x0 || = 1 and ||x0 − y|| > 1 − ε
for all y ∈ Y . Proof. If ε ≥ 1 then 1 − ε ≥ 0 and the claim holds for each vector
x0 ∈ X. So we can assume that 0 < ε < 1. Let x ∈ X \ Y . Since Y is closed,
d := inf{||x − y|| | y ∈ Y } > 0, because otherwise there would be a sequence
d
(yn ) in Y with ||yn − x|| → 0, and x ∈ Ȳ = Y . Thus d < 1−ε and there exists
d x−y0
y0 ∈ Y such that ||x − y0 || < 1−ε . Let x0 := ||x−y0 || , so ||x0 || = 1. Now let
y ∈ Y arbitrarily. Then
 
 x y0  1
||x0 − y|| =  − − y  = ||x − (y0 + ||x − y0 ||y)||
||x − y0 || ||x − y0 || ||x − y0 ||

d
≥ >1−ε
||x − y0 ||
Here we use y0 + ||x − y0 ||y ∈ Y .