Professional Documents
Culture Documents
Solved Prelim Problems in Analysis
Solved Prelim Problems in Analysis
Hamza Ruzayqat
Email: hruzayqa@vols.utk.edu
Real Analysis
1.1 Review
Note 1.1.1. Some of the problems here are not prelim problems, but they are good candidates.
Proposition 1.1.1. Suppose O ⊂ R is open. Then O can be written as the countable union of disjoint
open intervals {Ii }i = {(ai , bi )}i , i.e. O = ∞
S
i=1 Ii .
Definition 1.1.1. .
i . Gδ set is a countable intersection of open sets, i.e. for G ∈ Gδ , G = ∩∞
i=1 Oi where Oi is open.
Note that in both cases we can have finite intersection/union, for example in (i), by letting Oi = Si for
i = 1, ..., n and opens sets {Si }i and set Oi = ∅ for i > n.
Theorem 1.1.2. Let E ⊆ R be a given set. Then the following five statements are equivalent:
(i) E is Lebesgue measurable.
(ii) Given > 0, there exists an open set O with E ⊆ O and m∗ (O \ E) < .
(vi) Given > 0, there is a finite union U of open intervals such that m∗ (U 4E) < . (Here for two sets
A, B, we define their symmetric difference A4B = (A \ B) ∪ (B \ A))
Theorem 1.1.3. Let B be the Borel σ-algebra and M the Lebesgue σ-algebra. Then B ⊂ M. Thus if f
is Borel-measurable, then it is Lebesgue measurable.
Since continuous functions are Borel functions, then they are Lebesgue-measurable.
3
MATH 545 and MATH 546 Prelim Problems 4
Proposition 1.1.4. Suppose f ≥ 0 is a non-negative and measurable function. Then there exists a
sequence of non-negative measurable simple functions sn increasing to f .
Theorem 1.1.5. Egoroff ’s Theorem: Let (X, A, µ) be a finite measure space (i.e. µ(X) < ∞). If
{fn } is a sequence of measurable functions that converge to a function f a.e. (for each x, fn (x) → f (x)
point-wisely a.e. x) on a measurable set E, then given > 0, there is a subset A ⊂ E with µ(A) < such
that {fn } converges to f uniformly on E \ A.
These sets get smaller as n increases, meaning that En+1,k is always a subset of En,k , because the first
union involves fewer sets. A point x, for which the sequence {fm (x)} converges to f (x), cannot be in
every En,k for a fixed k. Hence by the assumption of µ-almost everywhere point-wise convergence on E,
\
µ En,k =0
n∈N
for every k. Let > 0, then since µ(E) < ∞ and µ(En,k ) → 0 as n → ∞, there exists, for each k, some
natural number nk such that
ε
µ(Enk ,k ) < .
2k
Define [
A= Enk ,k
k∈N
Appealing to the sigma additivity of µ and using the geometric series, we get
X X ε
µ(A) ≤ µ(Enk ,k ) < k
= ε.
k∈N k∈N
2
If x ∈
/ A (i.e. x ∈ E \ A), then x ∈
/ Enk ,k for all k. Since Enk +1,k ⊂ Enk ,k , then for all n ≥ nk , we have
1
|fn − f | <
k
i.e. fn → f uniformly on E \ A.
Theorem 1.1.6. Lusin’s Theorem: Let (X, A, µ) be a measure space. Let f : [a, b] → R be a
measurable function on [a, b]. Then given > 0, there is a continuous function φ on [a, b] such that
µ{x : f (x) 6= φ(x)} < . (every measurable function is a continuous function on nearly all its domain).
Definition 1.1.2. Let (X, A, µ) be a measure space. If s = ni=1 ai χEi is a non-negative measurable
P
Let f be measurable and let f + = max(f, 0) and f − = max(−f, 0). Provided f − dµ are
R R
f + dµ and
not both infinite, define Z Z Z
f dµ = f dµ − f − dµ.
+
Note 1.1.2. |f | = f + + f − .
R
Note 1.1.3. If X f dµ < ∞, i.e. f ∈ L1 , then f is finite µ-a.e..
Theorem 1.1.7. Monotone Convergence Theorem (MCT): Let (X, A, µ) be a measure space.
Suppose fn is a sequence of non-negative increasing measurable functions, i.e. 0 ≤ f1 (x) ≤ f2 (x) ≤ ...
for all x, and limn→∞ fn (x) = f (x) for all x (or a.e. x). Then
Z Z
lim fn dµ = f dµ.
n→∞
Theorem 1.1.8. Fatou’s Lemma: Let (X, A, µ) be a measure space. Suppose the fn are non-negative
and measurable. Then Z Z
lim inf fn dµ ≤ lim inf fn dµ.
n→∞ n→∞
Theorem 1.1.9. Dominated Convergence Theorem (DCT): Suppose that fn are measurable real-
valued functions and fn (x) → f (x) for each x(or a.e. x). Suppose there exists a nonnegative integrable
function g such that |fn (x)| ≤ g(x) for all x. Then
Z Z
lim fn dµ = f dµ.
n→∞
Note 1.1.4. Notice that in the previous three theorems we need point-wise convergence fn (x) → f (x)
for all x or a.e. x.
Definition 1.1.4. Sense of Vitali: Let J be a collection of intervals. Then we say that J covers a set
E in the sense of Vitali, if for each > 0 and any x ∈ E, there is an interval I ∈ J such that x ∈ I and
`(I) < .
Lemma 1.1.10. Vitali: Let E be a set of finite outer measure and J a collection of intervals that cover
E in the sense of Vitali. Then, given > 0, there is a finite disjoint collection {I1 , ..., IN } of intervals in
J such that !
[N
µ∗ E \ Ii < .
i=1
MATH 545 and MATH 546 Prelim Problems 6
Theorem 1.1.11. Let f be an increasing real-valued function on the interval [a, b]. Then f is differen-
tiable a.e., the derivative f 0 is integrable, and
Z b
f 0 (x) dm(x) ≤ f (b) − f (a).
a
Definition 1.1.5. Total Variation: Let f be a real-valued function defined on the interval [a, b]. Let the
class P be a class of all possible partitions of [a, b], i.e. P = {P = {x0 , . . . , xn }|P is a partition of [a, b]}.
Then the total variation of f on the interval [a, b] is
n
X
Tab f = sup |f (xi ) − f (xi−1 )|.
P ∈P
i=1
Definition 1.1.6. Bounded Variation (BV): A real-valued function f is said to be of bounded vari-
ation (BV function) on [a, b] if its total variation is finite, i.e.
Theorem 1.1.12. A real function f is of bounded variation in an interval [a, b], (f ∈ BV ([a, b])), if
and only if it can be written as the difference f = f1 − f2 of two non-decreasing functions. (Hence all
increasing functions are of BV and hence their derivatives are exists a.e.)
Corollary 1.1.13. If f ∈ BV ([a, b]), then f 0 exists a.e. x ∈ [a, b], i.e., f 0 is in L1 (m, [a, b]). (Hence if f
is an increasing function, then f 0 exists a.e. and it is integrable on [a, b])
Theorem 1.1.14. Fundamental Theorem of Calculus I (FTC I): Let f be an integrable function
on [a, b], and suppose that Z x
F (x) = F (a) + f (t) dm(t).
a
0
Then F (x) = f (x) a.e. x in [a, b].
Lemma 1.1.15. If f ∈ AC([a, b]), then f ∈ BV ([a, b]) (and hence f 0 ∈ L1 (m).
Definition 1.1.8. σ-Finite Measure: Let (X, A, µ) be a measure space. The measure µ is called σ-
finite if X is the countable union of measurable subsets of A with finite measure. (e.g R = ∪k∈N (−k, k))
MATH 545 and MATH 546 Prelim Problems 7
Definition 1.1.9. Complete Measure Space: A measure space (X, A, µ) is said to be complete if
B ∈ A, µ(B) = 0, and A ⊂ B, imply that A ∈ A. (e.g Lebesgue measure is complete, while Lebesgue
measure restricted to B the σ-algebra of Borel sets is not complete.)
i. If A and B in S, then A ∩ B ∈ S.
ii. If A ∈ S, then Ac = a finite disjoint union of sets in S, i.e. Ac = ∪ni=1 Ai and Ai ’s are disjoint in S.
i. ∅ ∈ C.
ii. If C ∈ C, then C = ∪ni=1 Ci , where {Ci }ni=1 is a finite collection of disjoint sets in S.
iii. C is contained in any algebra containing S, i.e. if C 0 is an algebra and S ⊂ C 0 , then C ⊂ C 0 . I.e., C is
the minimum algebra contains S.
Definition 1.1.12. Rectangle: Let (X, A, µ) and (Y, B, ν) be two complete measure spaces, and con-
sider the direct product X ×Y of X and Y . If A ⊂ X and B ⊂ Y , we call R = A×B a rectangle. If A ∈ A
and B ∈ B, we call R a measurable rectangle. The collection R of measurable rectangles is a semi-algebra.
Let R = A × B be a measurable rectangle and let λ = (µ × ν). Then λ(R) = (µ × ν)(A × B) = µ(A).ν(B).
Definition 1.1.13. Product Outer Measure (µ×ν)∗ : Let (X, A, µ) and (Y, B, ν) be complete measure
spaces. Let R0 be the algebra generated by the semi-algebra R as in Definition 1.1.11. Then λ = µ × ν
can be extended on R0 with outer measure λ∗ = (µ × ν)∗ : P(X × Y ) → [0, ∞] defined by:
(∞ ∞
)
X [
λ∗ (E) = inf λ(Ei ) : Ei ∈ R0 for i ∈ N, and E ⊆ Ei .
i=1 i=1
Definition 1.1.14. Let Ri ∈ R (set of measurable rectangles) for each i ∈ N. Then, A ∈ Rσ implies
A = ∪∞ ∞
i=1 Ri . Let Ai ∈ Rσ for each i ∈ N. Then, B ∈ Rσδ implies B = ∩i=1 Ai .
Theorem 1.1.17. Suppose (X, A, µ) and (Y, B, ν) are complete measure spaces, (µ × ν)∗ is the outer
measure on X ×Y , and (µ×ν) is the restriction of (µ×ν)∗ to the (µ×ν)∗ -measurable sets. Let E ⊆ X ×Y .
Suppose either (µ × ν)∗ (E) < ∞ or (µ × ν)∗ (E) = ∞ and X × Y is σ-finite. The following are equivalent:
(i) E is (µ × ν)-measurable.
MATH 545 and MATH 546 Prelim Problems 8
(ii) For all > 0, there exists S ∈ Rσ such that E ⊂ S and (µ × ν)∗ (S \ E) < .
(iii) There exists T ∈ Rσδ such that E ⊂ T and (µ × ν)∗ (T \ E) = 0.
Theorem 1.1.18. Fubini Theorem: Let (X, A, µ) and (Y, B, ν) be two complete measure spaces and
f an integrable function on X × Y . Then
i. For almost all x the function fx defined by fx (y) = f (x, y) is an integrable function on Y . For almost
all y the function fy defined by fy (x) = f (x, y) is an integrable function on X.
R R
ii. Y f (x, y) dν(y) and X f (x, y) dµ(x) are integrable functions on X and Y respectively.
iii. Z Z Z Z Z
f dν dµ = f d(µ × ν) = f dµ dν.
X Y X×Y Y X
Theorem 1.1.19. Tonelli Theorem: Let (X, A, µ) and (Y, B, ν) be two σ-finite measure spaces, and
let f ≥ 0 be a measurable function on X × Y . Then
i. For almost all x the function fx defined by fx (y) = f (x, y) is a measurable function on Y . For
almost all y the function fy defined by fy (x) = f (x, y) is a measurable function on X.
R R
ii. Y f (x, y) dν(y) and X f (x, y) dµ(x) are measurable functions on X and Y respectively.
iii. Z Z Z Z Z
f dν dµ = f d(µ × ν) = f dµ dν.
X Y X×Y Y X
Definition 1.1.15. Signed Measure: By a signed measure on the measurable space (X, A) we mean
an extended real-valued set function ν defined for the sets of A and satisfying the following conditions:
i. ν assumes at most one of the values +∞, −∞.
ii. ν(∅) = 0.
iii. ν ( ∞
S P∞
n=1 En ) = n=1 ν(En ) for any sequence En of disjoint measurable sets, the equality taken to
mean that the series on the right converges absolutely if ν ( ∞
S
n=1 En ) is finite and that it properly
diverges otherwise.
Definition 1.1.16. Positive and Negative Sets: We say that a set A is a positive set with respect
to a signed measure ν if A is measurable and for every measurable subset E of A we have ν(E) ≥ 0.
Similarly, a set B is called a negative set if it is measurable and every measurable subset E of it has
nonpositive ν measure. A set that is both positive and negative with respect to ν is called a null set. A
measurable set A is a null set if and only if every measurable subset E ⊂ A has ν(E) = 0.
Note 1.1.6. Every measurable subset of a positive set is again positive, and the union of a countable
collection of positive sets is also positive. Proof: 1) Let A be a positive set, then for every measurable
subset E of A we have ν(E) ≥ 0. We need to prove that E is also positive. E is measurable as
mentioned and if C ⊂ E is measurable, then ν(C) ≥ 0 since C ⊂ E ⊂ A. 2) Let A = ∪i∈N Ai , Ai
is positive. Let E ⊂ A be measurable. Set Ei = E ∩ Ai ∩ Aci−1 ∩ Aci−2 ... ∩ Ac1 , then {Ei } are disjoint
sets, Ei ⊂ Ai is measurable, and E = ∪i∈N Ei . Since Ei ⊂ Ai and Ai is positive, ν(Ei ) ≥ 0 and hence
P
ν(E) = i∈N ν(Ei ) ≥ 0.#
MATH 545 and MATH 546 Prelim Problems 9
Lemma 1.1.20. Let E be a measurable set such that 0 < ν(E) < ∞. Then there is a positive set A
contained in E with ν(A) > 0.
Theorem 1.1.21. Hahn Decomposition Theorem: Let ν be a signed measure on the measurable
space (X, A). Then there is a positive set A and a negative set B such that X = A ∪ B and A ∩ B = ∅.
Note 1.1.7. Hahn Decomposition is not unique.
Definition 1.1.17. Singular Measure ν ⊥ µ: Two measures ν and µ defined on (X, A) are said to
be mutually singular (in symbols ν ⊥ µ) if there are disjoint measurable sets A and B with X = A ∪ B
such that ν(A) = µ(B) = 0. (e.g let A and B be as in Theorem 1.1.21 and ν + (E) = ν(E ∩ A) and
ν − (E) = −ν(E ∩ B), then ν + and ν − are mutually singular.) We sometimes read ν ⊥ µ as: ν is singular
with respect to µ.
Theorem 1.1.22. Jordan Decomposition: Let ν be a signed measure on the measurable space (X, A).
Then there are two mutually singular measures ν + and ν − on (X, A) such that ν = ν + − ν − . Moreover,
there is only one such pair of mutually singular measures.
Note 1.1.8. Since ν assumes at most one of the values +∞ and −∞, either ν + or ν − must be finite.
Moreover, the measure |ν| defined by |ν|(E) = ν + (E) + ν − (A) is called the absolute value or total
variation of ν.
Definition 1.1.18. Absolute Continuity of Measure ν << µ: Let ν and µ defined on (X, A). The
measure ν is said to be absolutely continuous with respect to the measure µ ( in symbols: ν << µ) if
A ∈ A with µ(A) = 0 =⇒ ν(A) = 0.
Note 1.1.9. In the case of signed measures µ and ν, we say ν << µ if |ν| << |µ|. Also in this case,
ν ⊥ µ if |ν| ⊥ |µ|.
Theorem 1.1.23. Radon-Nikodym Theorem (R-N): Let (X, A, µ) be a σ-finite measure space (so µ
is a positive measure), and let ν be a measure defined on (X, A) with ν << µ. Then there is a nonnegative
measurable function f : X → [0, ∞) such that for any measurable set E ⊂ A we have
Z
ν(E) = f dµ.
E
The function f is unique in the sense that if g is any measurable function with this property then g = f
dν
µ-a.e.. f is usually written as .
dµ
Theorem 1.1.24. Lebesgue Decomposition: Let (X, A, µ) be a σ-finite measure space (so µ is a
positive measure), and let ν be a σ-finite measure defined on (X, A). Then we can find a measure ν0 with
ν0 ⊥ µ, and a measure ν1 with ν1 << µ, such that ν = ν0 + ν1 . The measures ν0 and ν1 are unique.
Definition 1.1.19. A normed linear space is called complete if every Cauchy sequence in the space
converges, that is, if for each Cauchy sequence {fn } in the space there is an element f in the space such
that fn → f . A complete normed linear space is called a Banach space.
Theorem 1.1.25. Riesz-Fischer Theorem: The Lp spaces are complete.
Theorem 1.1.26. The set of continuous functions with compact support, i.e f (x) is defined on a compact
set like [−n, n] and zero outside, is dense in Lp .
MATH 545 and MATH 546 Prelim Problems 10
Useful Inequalities:
kf + gkp ≤ kf kp + kgkp .
If 1 < p < ∞, then equality can hold only if there are nonnegative constants α and β such that
βf = αg.
2. Minkowski Inequality for 0 < p < 1: If f and g are two nonnegative functions in Lp with
0 < p < 1, then
kf + gkp ≥ kf kp + kgkp .
3. Holder Inequality: If p and q are nonnegative extended real numbers such that
1 1
+ = 1,
p q
and if f ∈ Lp and g ∈ Lq , then f g ∈ L1 and
kf gk1 ≤ kf kp kgkq .
Equality holds if and only if there are some constants α and β, not both zero, such that α|f |p = β|g|q
a.e..
4. Jensen Inequality: Let φ be a convex function on (−∞, ∞) and f ∈ L1 [0, 1]. Then
Z Z
φ f (x) dx ≤ φ(f (x)) dx.
SOLUTION:
We do this problem by induction. For n = 1, 2 it is obvious. Suppose it holds for n − 1 and let’s
prove it holds for n. W.L.O.G let p1 ≤ p2 ≤ ... ≤ pn . Assume fk ∈ Lpk for k = 1, 2, ..., n. Then we
have two cases:
Case(1): pn = ∞.
MATH 545 and MATH 546 Prelim Problems 11
Pn−1 1 1
Hence fn ∈ L∞ , i.e. kfn k∞ = ess sup |fn | < ∞, and k=1 = . So
pk r
Yn
Z 1/r
r r r r
fk
= |f1 | |f2 | ...|fn−1 | |fn | dx
k=1 r
n−1
Y
≤
fk
kfn k∞
k=1 r
≤kf1 kp1 kf2 kp2 ...kfn−1 kpn−1 kfn k∞ By induction hyposthisis
n
Y
≤ kfk kpk .
k=1
Case(2): pn < ∞.
1 Pn−1 r 1 1 pn
Let = k=1 . Then + = 1 and thus q = . So,
q pk q pn /r pn − r
r Z
Yn
r
fk
= |f1 f2 ...fn−1 | |fn |r dx
k=1 r
=k |f1 f2 ...fn−1 |r |fn |r k1
≤k |f1 f2 ...fn−1 |r kq k |fn |r kpn /r (Hölder’s)
Z 1/q Z r/pn
qr pn
= |f1 f2 ...fn−1 | dx |fn | dx
and therefore
Yn
n
Y
f ≤ kfk kpk .
k
k=1 r k=1
SOLUTION:
Z Z Z
kf k1 = |f | dµ = f dµ − f dµ
X {x∈X:f (x)≥0} {x∈X:f (x)<0}
Z Z
≤ f dµ − f dµ
{x∈X:f (x)≥0} {x∈X:f (x)<0}
Z Z
≤ f dµ + f dµ
{x∈X:f (x)≥0} {x∈X:f (x)<0}
< 2K
1
So 0 < K ≤ .
2
3. (a) Let p ∈ [1, ∞) and suppose {fn }∞ P
n=1 ⊂ L (R) is a sequence that converges to 0 in the
p norm. Prove that we may find a subsequence {fnk }∞ k=1 such that fnk → 0 a.e.
SOLUTION:
(a) So we have kfn kp < ∞ and kfn kp → 0. Thus we can find n1 ∈ N such that for all n ≥ n1 , we
have kfn kp < 1/2. Having found this n1 we may find n2 ∈ N such that n2 > n1 and for all
n ≥ n2 , we have kfn kp < 1/22 . Continue in this manner, we inductively find nk ∈ N such that
nk > nk−1 and for all n ≥ nk , we have kfn kp < 1/2k . So we got a subsequence {fnk }∞ k=1 such
k
Pm p
that kfnk kp < 1/2 for all k ∈ N. Now let fm = k=1 |fnk | , then {fm } is a monotone increasing
sequence, hence by MCT,
Z X ∞ Z Xm
p
|fnk | dµ = lim |fnk |p dµ
R k=1 R m→∞ k=1
Z X m
= lim |fnk |p dµ
m→∞ R k=1
m
XZ
= lim |fnk |p dµ
m→∞ R
k=1
∞ ∞
X X 1 1
= kfnk kpp < pk
= p−1
< ∞.
k=1 k=1
2 2
P∞ P∞
So k=1 |fnk |p is integrable, hence k=1 |fnk |p is finite for a.e. x ∈ R, hence |fnk |p → 0 as k → 0
a.e. x ∈ R, and thus fnk → 0 as k → 0 a.e. x ∈ R.
MATH 545 and MATH 546 Prelim Problems 13
(b) As in (a),
Z ∞
1X ∞ Z 1 ∞
X X 1
fn2 dµ = fn2 dµ ≤ 3
< ∞.
0 k=1 k=1 0 k=1
n
P∞
Thus k=1 fn2 < ∞ a.e. x ∈ [0, 1], hence fn2 → 0 a.e. x ∈ [0, 1], hence fn → 0 a.e. x ∈ [0, 1].
(c) Consider the sequence {En }∞
n=1 of subsets of [0, 1] defined by:
But for all x ∈ [0, 1], lim fn (x) does not exist because if for example x ∈ E6 , then already x
n→∞
in E1 and E3 . Therefore, f1 (x) = f3 (x) = f6 (x) = 1 and fn (x) = 0 for n ∈ N, n 6= 1, 3, 6. So,
if x ∈ [0, 1], then as n → ∞ x is already in infinitely many Ei ’s, and hence the limit does not
exist.
4. For each δ > 0 and each f ∈ L1 (0, ∞), find (with proof )
Z 1
lim t xδ f (tx) dx.
t→∞ 0
SOLUTION:
MATH 545 and MATH 546 Prelim Problems 14
SOLUTION:
Let An = {x ∈ X : f (x) < −1/n} for each n ∈ N. Since f is µ-measurable, An is µ-measurable. Let
sn = − n1 χAn for each n ∈ N. Then 0 ≤ −sn < −f χAn . Hence
Z Z Z
1
0 ≤ µ(An ) = (−sn ) dµ < (−f χAn ) dµ = (−f ) dµ ≤ 0,
n X X An
R R
using the definition of the integral and the assumption that X f dµ ≥ 0 and hence X (−f ) dµ ≤ 0.
Hence µ(An ) = 0 for each n ∈ N. Note that
∞
[ ∞
[
{x ∈ X : f (x) < 0} = {x ∈ X : f (x) < −1/n} = An .
n=1 n=1
Since µ(An ) = 0 for each n ∈ N, µ({x ∈ X : f (x) < 0}) = 0 (a countable union of sets of measure 0
has measure 0). Hence, f (x) ≥ 0 a.e. x ∈ X.
6. Let f be an integrable function on a measure space (X, M, µ). Show that given > 0,
there is a δ > 0 such that for each µ−measurable set E with µ(E) < δ, we have
Z Z
f dµ ≤ |f | dµ < .
E E
SOLUTION:
(
f (x) if f (x) ≤ n
Let > 0. Define fn (x) = , Then fn (x) ≤ n for all n ∈ N and fn (x) ↑
n otherwise
f (x). By (MCT)
Z Z
f (x) dµ = lim fn (x) dµ.
X n→∞ X
Choose δ < . Then
2n
Z Z Z Z
f dµ ≤ f dµ = f − fn dµ + fn dµ
E E E E
Z
≤ f − fn dµ + n µ(E)
X
< +n δ
2
< + = .
2 2
R1
7. Suppose f : [0, 1] → R is increasing. Suppose 0
f 0 dm = f (1) − f (0). Prove that
f ∈ AC([0, 1]).
SOLUTION:
Let x ∈ (0, 1] be arbitrary. Consider the interval [0, x]. Then since f is increasing on [0, 1], it is
increasing on [0, x]. By Theorem 1.1.11, f 0 exists a.e. on [0, x], f 0 is Lebesgue integrable on [0, x]
Rx Rx
and 0 f 0 (t) dm(t) ≤ f (x) − f (0). However, 0 f 0 (t) dm(t) < f (x) − f (0) cannot happen. Proof:
Suppose it happens, then because f 0 is inntegrable, by FTC(I) differentiating both sides of the strict
inequality with respect to x gives f 0 (x) < f 0 (x) a.e. which is a contradiction. So
Z x
f 0 (t) dm(t) = f (x) − f (0).
0
Claim. If for all > 0, there is δ > 0 such that for each m-measurable set E ⊂ [0, 1] with m(E) < δ
we have Z
|f 0 | dm < ,
E
be a finite collection of disjoint open intervals in [0, 1] (ordered in a way such that bi < ai+1 for
i = 1, 2, ..., n − 1) such that E = ∪ni=1 (ai , bi ). Then one has ni=1 (bi − ai ) = m(E) < δ. Notice that
P
Z bi Z ai Z bi Z
0 0 0
f (bi ) − f (ai ) = f (t) dm(t) − f (t) dm(t) = f (t) dm(t) = f 0 (t) dm(t).
0 0 ai (ai ,bi )
And thus
n
X n Z
X Z Z
0 0
|f (bi ) − f (ai )| ≤ |f | dm = |f | dm = |f 0 | dm <
i=1 i=1 (ai ,bi ) ∪n
i=1 (ai ,bi ) E
(a) Show that the series defining f (x, y) converges for every (x, y) ∈ [0, 1] × [0, 1].
(b) Determine whether f is Lebesgue integrable on [0, 1] × [0, 1].
SOLUTION:
√
Note that n [sin(xy)]n is continuous for any n ∈ N and (x, y) ∈ [0, 1]×[0, 1], thus it is Lebesgue
√
measurable. Let fm (x, y) = m n [sin(xy)]n , then fm is Lebesgue measurable because it is
P
n=1
sum of Lebesgue measurable functions. Note that f (x, y) = limm→∞ fm which is a limit of
Lebesgue measurable functions, hence it is Lebesgue measurable.
Since ([0, 1], M, m) is finite, f (x, y) is Lebesgue measurable on [0, 1] × [0, 1] and f (x, y) ≥ 0, we
can apply Tonelli’s Theorem to have
∞ 1 ∞
1X
√ √
Z X Z Z
n
n [sin(xy)] dm(x, y) = n [sin(xy)]n dm(x)dm(y).
[0,1]×[0,1] n=1 0 0 n=1
9. Evaluate the following limit:
Z n x n x
lim 1− ln 2 + cos dm(x).
n→∞ 0 n n
SOLUTION:
x n x x n
Let fn (x) = 1 − ln 2 + cos χ[0,n] . Then |fn | = fn ≤ e−x ln(3)χ[0,∞) ≡ g, since 1 − is
n n −x n
an increasing
sequence bounded by e which actually its limit. Also, since 1 ≤ 2 + cos(x/n) ≤ 3,
x
we have ln 2 + cos ≤ ln(3). Notice that limn→∞ fn = ln(3)e−x . Since g is integrable then by
n
DCT
Z n Z
x n x x n x
lim 1− ln 2 + cos dm(x) = lim 1 − ln 2 + cos χ[0,n] dm(x)
n→∞ 0 n n n→∞ n n
Z ∞
= ln(3) e−x dm(x) = ln(3).
0
10. If q < p + 1 find (with proof )
1
nxp + xq
Z
lim dx.
n→∞ 0 xp + nxq
MATH 545 and MATH 546 Prelim Problems 18
SOLUTION:
nxp + xq
Let fn = χ[0,1] , then for all n ∈ N, fn ≥ 0 and
xp + nxq
nxp + xq nxp + xq 1
fn = p q
χ[0,1] ≤ q
= xp−q + ≤ xp−q + 1 ≡ g.
x + nx nx n
But g is integrable since
Z 1
p−q xp−q+1 1 p − q + 2
(x + 1) dx = + x = < ∞,
0 p−q+1 0 p−q+1
where this integral is valid because p − q > −1. Also, notice that the point-wise limit of fn is
p xq
p
nx + x q x +
lim p χ[0,1] = lim p n χ[0,1] = xp−q χ[0,1] .
n→∞ x + nxq n→∞ x
+ xq
n
Hence by DCT,
Z 1 p
nx + xq nxp + xq
Z
lim dx = lim χ[0,1] dx
n→∞ 0 xp + nxq n→∞ xp + nxq
Z 1
1
= xp−q dx = .
0 p−q+1
Again this is true because p − q > −1.
11. Suppose f is Lebesgue measurable on the product spce [0, ∞) × [0, ∞) and g ∈ L1 ([0, ∞)).
Prove that if |xf (x, y)| ≤ g(x) for all y ∈ [0, ∞), then
Z ∞ Z ∞
t
sup f (x, y) cos dx dt < ∞.
y>0 0 t y
SOLUTION:
Notice that
Z ∞ Z ∞ Z ∞ Z ∞
Z ∞Z ∞
t
≤
t
f (x, y) cos dx dt |f (x, y)| cos
dx dt ≤ |f (x, y)| dx dt
0 t y
0 t y 0 t
Z ∞Z ∞
= |f (x, y)|χ{x≥t} dx dt.
0 0
Since f is Lebesgue measurable and the absolute value function |.| is continuous, |f (x, y)| is Lebesgue
measurable. Since t ≤ x could be a singleton when x = t or an interval otherwise which both are
Lebesgue measurable, we have |f (x, y)|χ{x≥t} is Lebesgue measurable. Also, we know that [0, ∞) is
σ-finite. Hence we can apply Tonelli’s Theorem on |f (x, y)|χ{x≥t} as follows
Z ∞Z ∞ Z ∞Z ∞
|f (x, y)| χ{x≥t} dx dt = |f (x, y)| χ{x≥t} dt dx
0 0 | {z } 0 0 | {z }
x varies and t fixed
Z ∞ Z ∞ t varies and x fixed
= |f (x, y)| χ{x≥t} dt dx
0 0
MATH 545 and MATH 546 Prelim Problems 19
Z ∞ Z x
= |f (x, y)| dt dx
Z0 ∞ 0
Taking the supremum over y > 0 will not change the answer because it does not depend on y.
SOLUTION:
Since [a, b] ⊂ [0, 1], we can choose a = 0 and b = 1, and by the fact that f is nonnegative on [0, 1],
we have Z 1 Z 1
|f (x)| dx = f (x) dx = 1 − 0 = 1 < ∞,
0 0
1
hence f ∈ L ([0, 1]). Let 0 < x ≤ 1, then [0, x] ⊂ [0, 1]. Then by the given assumption, we have
Z x
x=0+ f (t) dt.
0
SOLUTION:
By DCT,
Z ∞ Z ∞
f (tx) f (y)
lim t sin f (tx) dx = lim sin f (y) dy
t→∞ 0 t t→∞ 0 t
Z ∞
f (y)
= lim sin f (y)dy
0 t→∞ t
Z ∞
= 0 dy = 0.
0
Let fn = 1/(1 + x2 )n χ[0,n] , then the point-wise limit is limn→∞ fn = 0 everywhere except at x = 0
we have limn→∞ fn = 1, i.e. fn → 0 a.e.. By using Bernoulli’s inequality: (1 + x)n > 1 + nx for
x > −1, we have
1 1 1
|fn | = fn = χ[0,n] ≤ χ [0,n] ≤ χ[0,∞) ≡ g(x).
(1 + x2 )n 1 + nx2 1 + nx2
(Or we can consider g = χ[0,n] since fn ≤ χ[0,n] for all n and χ[0,n] is integrable) Now to show
√
g ∈ L1 (0, ∞), we let y = nx. So
Z Z ∞ ∞
1 1 1 π
g(x) dx = √ dy = √ arctan(y) = √ < ∞.
n 0 1+y 2 n 0 2 n
By DCT,
Z n Z ∞
1 1
lim 2 n
dx = lim χ[0,n] dx
n→∞ 0 (1 + x ) n→∞ 0 (1 + x2 )n
Z ∞
1
= lim χ[0,n] dx
0 n→∞ (1 + x2 )n
Z ∞
= 0 dx = 0.
0
MATH 545 and MATH 546 Prelim Problems 21
SOLUTION:
Since −1 < −y < 0 and 0 ≤ x ≤ 1, then x−y ≤ x−1 . Also, we know 1 − e−nx ≤ nx for all n ∈ N. Let
fn (x) = x−y (1 − e−nx ), then limn→∞ fn = x−y and |fn | = fn = x−y (1 − e−nx ) ≤ x−1 (nx) = n ≡ g.
But Z Z 1 1
g dx = n dx = n < ∞.
0 0
So by DCT,
Z 1 Z 1
−y −nx
f (y) := lim x (1 − e ) dx = lim x−y (1 − e−nx ) dx
n→∞ 0 n→∞
Z0 1
−y 1
= x dx = .
0 1−y
MATH 545 and MATH 546 Prelim Problems 22
SOLUTION:
SOLUTION:
Let > 0. Since m([0, 1]) < ∞, fn are measurable (since fn in L2 ) and fn → 0 (point-wisely)
a.e., we can apply Egorov’s Theorem: For > 0, there exists a measurable set A with m(A) <
2/(2−p)
/[2M p/2 ] and fn → 0 uniformly on [0, 1] \ A. This uniform convergence implies that for all
x ∈ [0, 1] \ A, there exists an N ∈ N such that for all n ≥ N we have |fn − 0| < (/2)1/p . Then, for
n ≥ N:
Z 1 Z Z
p p
|fn | dm = |fn | dm + |fn |p dm
0 [0,1]\A A
Z
< m([0, 1] \ A) + |fn |p dm
2 A
Z
≤ m([0, 1]) + |fn |p dm
2 A
Z
= + |fn |p dm
2 A
Now we apply Holders inequality on A |fn |p dm with 1/q 0 + 1/q = 1, where q, q 0 to be chosen later.
R
Z Z 1/q0 Z 1/q
p pq 0 q
|fn | dm ≤ |fn | dm 1 dm
A A A
Z 1/q0
1/q pq 0
= m(A) |fn | dm
A
MATH 545 and MATH 546 Prelim Problems 23
Choose q, q 0 such that pq 0 = 2, then q 0 = 2/p and 1/q = 1 − 1/q 0 = 1 − p/2, thus q = 2/(2 − p). So
we have
Z Z p/2
p (1−p/2) 2
|fn | dm ≤ m(A) |fn | dm ≤ m(A)(1−p/2) M p/2 <
A A 2
Thus, for n ≥ N : Z 1
|fn |p dm <
0
Which proves that the limit goes to 0 as n goes to infinity.
SOLUTION:
√
Let y = nx, then we have
Z ∞√ Z ∞ √
nf (x) f (y/ n)
lim dx = lim dy
n→∞ −1 1 + nx2 n→∞ −√n 1 + y 2
Z √n √ Z ∞ √
f (y/ n) f (y/ n)
= lim dy + lim √ dy.
n→∞ −√n 1 + y 2 n→∞ n 1 + y2
= 0.
So we have √ √ √
Z ∞ Z n
nf (x) f (y/ n)
lim dx = lim √
dy.
n→∞ −1 1 + nx2 n→∞ − n 1+y
2
√ √
Now notice that f is continuous, hence it is bounded on [− n, n]. So if we let
√
fn = f (y/ n)χ[−√n,√n] /(1 + y 2 ), then fn → f (0)/(1 + y 2 )χ(−∞,∞) and
√
|fn | = |f (y/ n)|χ[−√n,√n] /(1 + y 2 ) ≤ supt∈[−1,1] |f (t)|χ[−√n,√n] ≡ g, where g is clearly integrable.
Hence by DCT,
Z ∞√ Z √n √ Z ∞
nf (x) f (y/ n) dy
lim 2
dx = lim √ 2
dy = f (0) 2
= π f (0).
n→∞ −1 1 + nx n→∞ − n 1 + y −∞ 1 + y
SOLUTION:
with P (fn ) < ∞. Let N be chosen such that for all n, m > N we have
Z
P (fn − fm ) = |fn − fm |(1 + log+ |fn − fm |) dµ < .
X
Then we have Z Z
|fn − fm | dµ ≤ |fn − fm |(1 + log+ |fn − fm |) dµ < .
X X
Hence {fn } is a Cauchy sequence in L and since L1 is complete, there exists a function f ∈ L1 (X, µ)
1
such that fn → f in L1 . Now we need to prove that P (f ) < ∞. Let {fnk }k be a subsequence of
{fn } such that limk→∞ fnk (x) = f (x) a.e.. By Fatou’s Lemma
Z Z Z
|f | dµ = lim |fnk | dµ ≤ lim inf |fnk | dµ < ∞.
X X k→∞ k→∞ X
< .
21. Evaluate the following limit:
Zπ/2 √
n
lim n 1 − sin x dx.
n→∞
0
SOLUTION:
Note that
√
n
(sin x)1/n − 10
n 1 − sin x = − .
1/n
MATH 545 and MATH 546 Prelim Problems 25
Zπ/2 √ Zπ/2
n
lim n 1 − sin x dx = − ln(sin x) = π ln(2)/2.
n→∞
0 0
22. Find ∞
sin xn
Z
lim dx.
n→∞ 0 xn
SOLUTION:
We split the integral into two parts:
Z ∞ Z 1 Z ∞
sin xn sin xn sin xn
lim dx = lim dx + lim dx.
n→∞ 0 xn n→∞ 0 xn n→∞ 1 xn
The second integral on the right is zero because if we let fn = sin xn /xn , then fn → 0 and we can
use DCT to prove its limit is zero. Let u = xn , then dx = du/(nxn−1 ) = du/(nu1−1/n ) and
Z ∞ Z ∞
sin xn sin u
lim n
dx = lim du.
n→∞ 1 x n→∞ 1 nu2−1/n
Let fn = sin u/(nu2−1/n ), then fn → 0 and |fn | ≤ 1/(nu2−1/n ) ≡ g(u). But for all n ≥ 2
Z ∞ 1
u n −1 ∞ 1
g(u) du = = .
1 1−n 1 n−1
Thus by DCT
∞
sin xn
Z
lim dx = 0.
n→∞ 1 xn
n n
For the first integral, let fn = sin x /x , then fn → 1 and |fn | ≤ 1 on [0, 1], since | sin x| ≤ |x| on
[0, 1]. Hence again by DCT
Z 1 Z 1
sin xn
lim dx = 1 dx = 1.
n→∞ 0 xn 0
SOLUTION:
Ra Ra
First notice that |g(x)| ≤ x |f (t)|/t dt ≤ (1/x) x |f (t)| dt < ∞ for all x ∈ (0, a), because f is
integrable on (0, a), and hence on any subset of (0, a). That means the function f (t)/t is integrable
on (x, a) for x ∈ (0, a). So we can use Fubini’s Theorem as follows, to prove that g(x) is integrable
on (0, a):
Z aZ a Z aZ a
|f (t)| |f (t)|
dt dx = χ{0≤x≤t≤a} (t) dt dx
0 x t 0 0 t
Z a
|f (t)| a
Z
= χ{0≤x≤t≤a} (x) dx dt
0 t 0
Z a
|f (t)| t
Z
= 1 dx dt
0 t 0
Z a Z a
|f (t)|
= t dt = |f (t)| dt < ∞.
0 t 0
24. Suppose f and g are nonnegative measurable functions on the interval [0, 1], with the
properties
Z 1 Z 1 Z 1
f (x) dx = 2, g(x) dx = 1, and [f (x)]2 dx ≤ C
0 0 0
for some constant C > 4. Let E = {x ∈ [0, 1] : f (x) > g(x)}. Show that E has measure
m(E) ≥ 1/C.
SOLUTION:
≤ m(E) C + 1
1 √
<√ C + 1 = 2.
C
So we have 2 < 2 which is a contradiction. So m(E) ≥ C.
MATH 545 and MATH 546 Prelim Problems 27
25. Suppose g is increasing and differentiable on [0, 1]. For every f ∈ L2 (0, 1) define f ∗ (x),
for x ∈ [0, 1], by: Z x
∗ 0
f (x) = g (x) f (t) dt.
0
2
If fn → f in L (0, 1), then prove that fn∗ ∗
→ f in L1 (0, 1).
SOLUTION:
First notice that since g increasing, g 0 (x) ≥ 0. Now let > 0. Let
Z x
∗ 0
fn (x) = g (x) fn (t) dt.
0
We need to prove that there exists N ∈ N such that for all n ≥ N we have
Z 1
|fn∗ − f ∗ | dx < .
0
2
Since fn → f in L (0, 1), there exists N ∈ N such that for all n ≥ N we have
Z 1 1/2
2
|fn − f | dx < . (Notice that g(1) − g(0) > 0 since g increasing.)
0 g(1) − g(0)
Then for all n ≥ N , we have
Z 1 Z 1 Z x Z x
∗ ∗
0 0
|fn − f | dx = g (x)
fn (t) dt − g (x) f (t) dt dx
0 0 0 0
Z 1 Z x
≤ g 0 (x) |fn − f | dt dx notice that g 0 (x) ≥ 0
Z0 1 Z0 1
= g 0 (x) |fn − f | χ{x≥t≥0} (t) dt dx
0 0
Z 1 Z 1 1/2 Z x 1/2
0 2
≤ g (x) |fn − f | dt 1 dt dx Holder’s
0 0 0
Z 1
√ 0
= x g (x) dx
g(1) − g(0) 0
Z 1
≤ g 0 (x) dx
g(1) − g(0) 0
≤ (g(1) − g(0)) = . (By Theorem 1.1.11)
g(1) − g(0)
26. Let {fn } be a sequence of measurable functions on a measure space (X, A, µ) . Suppose
∞
X
that the infinite series µ{x ∈ X : |fn (x)| ≥ } converges for each > 0. Prove that
n=1
fn (x) → 0 a.e.
MATH 545 and MATH 546 Prelim Problems 28
SOLUTION: ∞
X
Let > 0 be given. Suppose µ{x ∈ X : |fn (x)| ≥ } = C < ∞. Notice that
n=1
∞
X m Z
X
C= µ{x ∈ X : |fn (x)| ≥ } = lim χ{|fn (x)|≥} (x) dµ
m→∞ X
n=1 n=1
m
Z X
= lim χ{|fn (x)|≥} (x) dµ.
m→∞ X n=1
Hence,
∞
Z X
χ{|fn (x)|≥} (x) dµ = C < ∞.
X n=1
That implies ∞
P
n=1 χ{|fn (x)|≥} (x) is finite [µ]-a.e. x ∈ X. That means we have finitely many n such
that |fn (x)| ≥ and infinitely many n such that |fn (x)| < , i.e. fn → 0.
27. Let (X, A, µ) be a σ-finite measure space. Let f : X → [0, ∞) be µ-measurable. Show
Z Z ∞
f dµ = µ({x ∈ X : f (x) > t}) dt.
X 0
SOLUTION:
Note that
Z∞ Z ∞ Z
µ({x ∈ X : f (x) > t}) dt = χ{f (x)>t} dµ(x) dt.
0 X
0
Since {x ∈ X : f (x) > t} is measurable, since f is measurable, χ{f (x)>t} is a nonnegative measurable
function. By Tonelli’s
Z∞ Z ∞ Z
µ({x ∈ X : f (x) > t}) dt = χ{f (x)>t} dµ(x) dt
0 X
0
Z Z ∞
= χ{f (x)>t} dt dµ(x)
X 0
Z Z f (x)
= dt dµ(x)
ZX 0
= f (x) dµ(x).
X
MATH 545 and MATH 546 Prelim Problems 29
SOLUTION:
∞
X ∞
X ∞ Z
X ∞ Z
X
∞= 1≤ σ(An ) = f dµ = f χAn dµ
n=1 n=1 n=1 An n=1 X
Z ∞
X
= f χAn dµ by MCT, since f χAn ≥ 0
X n=1
∞ ∞
Z !
X Y
= f χ An − χAn dµ
X n=1 n=1
where Z ∞ Z Z
Y
f χAn dµ = f dµ ≤ f dµ = 0
X n=1 ∩∞
n=1 An Ak
for some k > N , where N as follows: Since µ(An ) → 0, there is an N ∈ N such that for all n > N ,
µ(An ) < (Which implies µ(An ) = 0 for n > N ). Therefore,
∞ ∞
Z !
X Y
∞≤ f χ An − χAn dµ
X n=1 n=1
Z
= f χ∪∞
n=1 An
dµ
ZX
= f dµ = σ(∪∞
n=1 An ).
∪∞
n=1 An
SOLUTION:
To guess the value of the limit, notice that
ln(nx + n) 1
lim = lim π = 1.
n→∞ π + ln(nx + n) n→∞
+1
ln(nx + n)
To proof this, notice that
ln(nx + n) ln(n) + ln(x + 1) π
fn = = =1− .
π + ln(nx + n) π + ln(n) + ln(x + 1) π + ln(n) + ln(x + 1)
MATH 545 and MATH 546 Prelim Problems 30
The point-wise limit of fn is 1 since ln(∞) = ∞. Also, notice that π ≤ π + ln(n) + ln(x + 1) for all
n ∈ N and x ∈ [0, 1], hence |fn | ≤ 1 ≡ g for all n. g is integrable on [0, 1], so by DCT we have
Z 1 Z 1
ln(nx + n)
lim dx = 1 dx = 1.
n→∞ 0 π + ln(nx + n) 0
SOLUTION:
2
We expect the limit to be 0 since limn→∞ e−nx = 0 for all x > 0. Let q be conjugate to p (i.e.
1/q
π
1/p + 1/q = 1). Let > 0. Choose N ∈ N such that √ kf kpp < . Then for all n ≥ N we
2 Nq
have
Z ∞ Z ∞
−nx2 −nx 2
e f (x) dx ≤ e f (x) dx
0
Z 0∞
2
≤ e−nx |f (x)| dx
0
Z ∞ 1/q
−nqx2
≤ kf kpp e dx (by Holder’s)
0
1/q
π
= √ kf kpp
2 nq
1/q
π
≤ √ kf kpp
2 Nq
< .
Therefore Z ∞
2
lim e−nx f (x) dx = 0.
n→∞ 0
µ(A ∩ B) = µ(A)µ(B)
SOLUTION:
Step 2. Let f ≥ 0 and g ≥ 0. Then by Proposition 1.4, there exist sequences of nonnega-
tive nondecreasing measurable simple functions sfn and sgn such that limn→∞ sfn (x) = f (x) and
limn→∞ sgn (x) = g(x) for all x. Hence sfn sgn is a nondecreasing sequence and limn→∞ (sfn sgn )(x) =
(f g)(x) for all x. So
Z Z
f g dµ = lim sfn sgn dµ (by MCT)
R n→∞ R
Z Z
= lim sfn dµ lim sgn dµ (by step 1)
n→∞ R n→∞ R
Z Z
= f dµ g dµ (by MCT).
R R
Step 3. For general f, g let f = f + − f − and g = g + − g − as in Definition 1.3. Then we can apply
step 2 on f + , f − , g + and g − because they all are nonnegative to get
Z Z
f g dµ = (f + − f − )(g + − g − ) dµ
R ZR Z Z Z
= f g dµ − f g dµ − f g dµ + f − g − dµ
+ + + − − +
ZR Z R Z R
Z Z R Z Z Z
= f dµ g dµ − f dµ g dµ − f dµ g dµ + f dµ g − dµ
+ + + − − + −
ZR RZ R
Z R
Z R
Z R
Z R
R
= f + dµ g + dµ − g − dµ − f − dµ g + dµ − g − dµ
RZ R
Z RZ R
Z R R
= f + dµ − f − dµ g + dµ − g − dµ
Z R Z R R R
= f dµ g dµ.
R R
32. Let (X, M, m) be Lebesgue measure space. Suppose f ∈ L1 (X, m). Prove that
SOLUTION:
MATH 545 and MATH 546 Prelim Problems 32
Let An = {x ∈ X : |f (x)| > n}, then An+1 ⊂ An . Let fn = f χAn , then since f ∈ L1 (X, m), |f | is
finite a.e. and hence fn → 0 a.e.. Also we have |fn | ≤ |f |, so by DCT
Z Z
0 = lim f χAn dm ≥ lim nχAn dm = lim n m(An ).
n→∞ X n→∞ X n→∞
33. Let p ≥ 2 be an extended real number and > 0. Suppose (X, A, µ) is a finite measure
space and fk is a sequence of measurable functions on X such that kfk kp → 0 as k → ∞.
If φ is measurable on X and Z
φfk dµ ≥ ,
X
2
prove that φ ∈
/ L (X, µ).
SOLUTION:
SOLUTION:
Z x
Let F (x) = f (w) dw. Then by FTC(I), F 0 (x) = f (x) a.e. x.
0
Claim. Z h
1
lim f (y + t) dt = f (y) a.e. y.
h→0 2h −h
Proof:
h y+h
F (y + h) − F (y)
Z Z
1 1 1 1 1
lim f (y + t) dt = lim f (w) dw = lim = F 0 (y) = f (y) a.e..
h→0 2h 0 h→0 2h y 2 h→0 h 2 2
0 y
F (y) − F (y − h)
Z Z
1 1 1 1 1
lim f (y + t) dt = lim f (w) dw =
lim = F 0 (y) = f (y) a.e..
h→0 2h −h h→0 2h
y−h 2 h→0 h 2 2
Z h Z 0 Z h
1 1 1
=⇒ lim f (y + t) dt = lim f (y + t) dt + lim f (y + t) dt = f (y) a.e..
h→0 2h −h h→0 2h −h h→0 2h 0
MATH 545 and MATH 546 Prelim Problems 33
Notice that
Z h Z h Z h
1 1 1
lim f (y + t)g(t) dt = lim f (y + t)(g(t) − g(0)) dt + g(0) lim f (y + t) dt
h→0 2h −h h→0 2h −h h→0 2h −h
Since g(x) continuous at x = 0, as h → 0, |h − −h| = 2|h| becomes very small and by continuity
there exists δ > 0 such that if 2|h| < δ, then |g(x) − g(0)| < for all > 0. Hence sending to 0
implies g(x) − g(0) = 0. So we have
Z h Z h
1 1
lim f (y + t)g(t) dt = g(0) lim f (y + t) dt = g(0)f (y) a.e. y ∈ R.
h→0 2h −h h→0 2h −h
35. Suppose that f is integrable and absolutely continuous on [0, ∞). If f 0 ∈ L1 [0, ∞), prove
that f (x) → 0 as x → ∞
SOLUTION:
Suppose f (x) does not tend to 0 as x tends to ∞. Then there exists an N ∈ N and a sequence of
points {xn } such that for all n ≥ N we have xn+1 > xn (i.e. xn → ∞) and |f (xn )| > for some > 0.
Rx
Notice that since f ∈ AC([0, ∞)), by FTC(II) f (x) = f (0) + 0 f 0 (t) dt. And since f 0 ∈ L1 [0, ∞),
R∞
we must have limx→∞ x f 0 (t) dt = 0, thus there exists a δ > 0 such that if y > x with |y − x| < δ,
R y
then we must have x f 0 (x) dx < /2. Therefore,
Z y Z x
0 0
|f (y) − f (x)| = f (0) + f (t) dt − f (0) − f (t) dt
0 0
Z y
f 0 (x) dx < .
=
x 2
Hence if we have |y − xn | < δ for n ≥ N , then |f (xn )| > and hence |f (y)| > /2. Let xn − δ < y <
xn + δ (i.e. |y − xn | < δ), then
Z xn +δ
|f (y)| dy > δ,
xn −δ
and therefore
Z ∞ Z ∞ ∞ Z
X xn +δ ∞
X
|f (y)| dy ≥ |f (y)| dy = |f (y)| dy > δ = +∞.
0 xN −δ n=N xn −δ n=N
36. Suppose X is a normed linear space and denote its unit sphere by
S := {x ∈ X : kxk = 1}.
SOLUTION:
Suppose X is a complete normed linear space. Let {xn } be a Cauchy sequence in S, i.e. kxn k = 1
for all n, then there is x ∈ X such that xn → x, since S ⊂ X and X is complete. We need to prove
that kxk = 1. Notice that since the sequence converges, for all > 0 there exists an N ∈ N such
that for all n ≥ N we have kxn k − kxk ≤ |kxn k − kxk| ≤ kxn − xk < , hence kxk = kxn k + = 1 + .
Since that true for all > 0, kxk = 1, so S is complete.
Now suppose S is complete. Let {xn } be a Cauchy sequence in X. Since kxn k ∈ R for all n and
every Cauchy sequence in R is convergent, if we consider the sequence {kxn k} and we prove it is
Cauchy in R, then there is α ∈ R such that kxn k → α. Notice that {xn } is Cauchy in X implies for
all > 0 there is N ∈ N such that for all n, m ≥ N we have |kxn k − kxm k| ≤ kxn − xm k < which
implies that {kxn k} is Cauchy in R and hence there is α ∈ R such that kxn k → α.
Notice that if the limit of {xn } we are looking for equals 0, then we are done, because 0 ∈ X (since
X is vector space). Suppose the limit is not 0, and hence α 6= 0, then there is an N ∈ N such that
for all n ≥ N , xn 6= 0. Let n ≥ N and consider the sequence yn = xn /kxn k. Then {yn } ⊂ S since
kyn k = 1. Since S is complete, if we prove that {yn } is Cauchy, then there is y ∈ S, yn → y. Let
> 0, and choose M ∈ N such that kkxn k − αk < α/3 and kxn − xm k < α/3 for all n, m ≥ M ,
then for all n, m ≥ M we have
xn x
xn x
xm x
m n m
+ 1 kxn − xm k
kyn − ym k =
−
≤
kxn k kxm k
kxn k −
+
−
α
kxm k α
|α|
1 1 1
= kkxn k − αk + kkxm k − αk + kxn − xm k
α α α
< + + = .
3 3 3
So {yn } is Cauchy in S and hence there is y ∈ S, yn → y. Since xn = kxn kyn , xn → αy.
SOLUTION:
Notice that
∞ ∞ t
e−x − e−xt
Z Z Z
dx = e−xy dy dx.
0 x 0 1
MATH 545 and MATH 546 Prelim Problems 35
Let (R+ , M, m) be Lebesgue measure space, where R+ = (0, ∞). Then this space is σ-finite. Let
f (x, y) = e−xy , then f > 0, and it is m × m-measurable, since it is continuous. By Tonelli’s Theorem
Z ∞Z t Z tZ ∞ Z t −xy x=∞ Z t
−xy −xy e 1
e dy dx = e dx dy = dy = dy = ln t.
1 −y x=0 1 y
0 1 1 0
SOLUTION:
x2 + nx − 1
Let fn = tan , then fn → tan x and |fn | ≤ 2x ≡ g on the interval [0, π/3]. g clearly
n
is integrable, hence by DCT we have
Z π/3 2 Z π/3
x + nx − 1 π/3 2
lim tan dx = tan x dx = ln(sec x) = ln √ .
n→∞ 0 n 0 0 3
SOLUTION:
1 ex 1 1 ex √
Let fn = √ sin χ[1/n,1] , then fn → 0 and |fn | ≤ √ χ[1/n,1] = n ex χ[1/n,1] ≡ g. So
nx x n (1/n)
Z 1
√ √
Z
g dx = n ex dx = n(e1 − e1/n ) < ∞.
1/n
(a) Show that the collection of Borel sets in R is closed under translation.
MATH 545 and MATH 546 Prelim Problems 36
µ(E + y) = µ(E)
SOLUTION:
(a) For a fixed y, let C = {B − y : B is Borel}. We need to prove that C is a σ-algebra that contains
all open sets and hence all Borel sets. If we do that, then if D ∈ C, then D is Borel and D = B−y
where B is also Borel, and hence B = D + y is Borel. Notice that C is not empty, since for exam-
ple [−y, 0] ∈ C because [−y, 0] = [0, y] − y and [0, y] is Borel. Let D ∈ C, then there is B Borel
such that D = B − y and hence Dc = B c − y (Proof: Let a ∈ Dc ⇐⇒ a ∈ / B − y ⇐⇒ a + y ∈ /
B ⇐⇒ a + y ∈ B c ⇐⇒ a ∈ B c − y) and B c is Borel, so Dc ∈ C. Let Di ∈ C for all i ∈ N.
Then there exists {Bi }i , Bi is Borel for each i, such that Di = Bi − y. ∪i Di = ∪i Bi − y (Proof:
Let a ∈ ∪i Di ⇐⇒ ∃i0 3 a ∈ Bi0 − y ⇐⇒ a + y ∈ Bi0 ⇐⇒ a + y ∈ ∪i Bi ⇐⇒ a ∈ ∪i Bi + y.)
and ∪i Bi is Borel so ∪i Di ∈ C.
Now Let O ∈ R be an open set. Then O+y is open (Proof: Let x ∈ O+y then x−y ∈ O and since
O is open there exists an > 0 such that (x−y −, x−y +) ⊂ O and thus (x−, x+) ⊂ O +y.)
Then O + y being open implies it is Borel which implies that O = (O + y − y) ∈ C. C contains
all open sets and hence all Borel sets and thus if D ∈ C, then D is Borel and D = B − y where
B is also Borel, and hence B = D + y is Borel.
(b) We will solve this part in three steps.
Step 1. Let y ∈ R be fixed and let f (x) = nk=1 ck χEk (x) be an integrable nonnegative simple
P
Xn
= ck µ(Ek − y)
k=1
n
X
= ck µ(Ek )
k=1
n
Z X Z
= ck χEk dµ = f dµ.
R k=1 R
MATH 545 and MATH 546 Prelim Problems 37
Step 2. Let f ≥ 0 be an integrable function, hence is measurable, and therefore there exists a
sequence of nonnegative measurable simple functions {sn } increasing to f . Then by MCT
Z Z Z Z
f (x + y) dµ(x) = lim sn (x + y) dµ(x) = lim sn (x) dµ = f dµ.
R n→∞ R n→∞ R R
Step 3. For general f let f = f + − f − as in Definition 1.3. Then we can apply step 2 on f +
and f − because they all are nonnegative to get
Z Z Z Z Z
−
f (x + y) dµ(x) = +
f (x + y) dµ(x) − f (x + y) dµ(x) = f (x) dµ − f − (x) dµ
+
R R R ZR R
= f dµ.
R
41. Suppose (X, A, µ) is a finite measure space and fk a sequence of integrable function on
X. Suppose further that to each > 0 there corresponds an integer N > 0 such that
Z
|fk | dµ <
E
for all measurable sets E which satisfy N µ(E) < 1 and all integers k which satisfy k > N .
If fk → f a.e. µ-on X, prove that f ∈ L1 (X, µ) and fk → f in L1 norm.
SOLUTION:
First notice since fk → f a.e. µ on X (i.e. point-wise limit a.e.), and hence |fk | → |f | a.e., by
Fatou’s Lemma we have
Z Z Z
|f | dµ = lim |fk | dµ ≤ lim inf |fk | dµ < ∞ (Since fk is integrable for all k).
X X k→∞ k→∞ X
Thus f ∈ L1 (X, µ). Now let > 0. Since (X, A, µ) is a finite measure space and fk → f a.e. µ-on
X, by Egoroff’s Theorem for any δ > 0 there exists a set A with µ(X \ A) < δ such that fk → f
uniformly on A. This uniform convergence implies that, there exists an integer M ∈ N such that
|fk − f | < for all k > M and for all x ∈ A. Let N be chosen such that N < 1/δ, then
3µ(A) Z
N µ(X \ A) < 1 and |fk | dµ < /3 for all k > N . Let L = max{N, M }, then for all k > L we
X\A
have
Z Z Z
|fk − f | dµ = |fk − f | dµ +
|fk − f | dµ
X A X\A
Z Z Z
≤ |fk − f | dµ + |f | dµ + |fk | dµ
A X\A X\A
Z Z Z
≤ |fk − f | dµ + lim inf |fn | dµ + |fk | dµ
A n→∞ n>L X\A X\A
< µ(A) + lim inf + = .
3µ(A) n→∞ n>L 3 3
MATH 545 and MATH 546 Prelim Problems 38
42. (a) Suppose a, b ∈ R, a < b, and f ∈ L1 ([a, b], m), where m is Lebesgue measure. Prove
that m-a.e. x ∈ [a, b] has the following property: For any sequence {(an , bn )}∞ n=1 of
intervals such that x ∈ (an , bn ) for every n ∈ N and limn→∞ (bn − an ) = 0, we have
Z bn
1
lim f dm = f (x).
n→∞ bn − an a
n
(That is, a locally integrable function is a.e. equal to the limit of its averages.)
(b) Let E ⊂ R be Lebesgue measurable. A point x ∈ R is called a point of density of E
if
m(E ∩ (x − h, x + h))
lim+ = 1.
h→0 2h
Prove that a.e. x ∈ E is a point of density of E. And prove that if x ∈
/ E, then
m(E ∩ (x − h, x + h))
lim+ = 0.
h→0 2h
1 x+h
Z Z x+h Z x
1
lim f (t) dt = lim+ f (t) dt − f (t) dt
h→0+ h x h→0 h a a
F (x + h) − F (x)
= lim+ = F 0 (x) = f (x).
h→0 h
and
Z x Z x Z x−h
1 1
lim+ f (t) dt = lim+ f (t) dt − f (t) dt
h→0 h x−h h→0 h a a
F (x) − F (x − h)
= lim+ = F 0 (x) = f (x).
h→0 h
Let > 0. Then since an < x < bn for all n and limn→∞ (bn − an ) = 0, limn→∞ (bn − x) = 0 and
limn→∞ (x − an ) = 0. So there exists N ∈ N such that for all n > N we have
Z x Z bn
1 1
x − an (f (t) − f (x)) dt < and
bn − x (f (t) − f (x)) dt < .
an x
MATH 545 and MATH 546 Prelim Problems 39
we have
m(E c ∩ (x − h, x + h))
lim = 1.
h→0+ 2h
But since
m(E ∩ (x − h, x + h)) m(E c ∩ (x − h, x + h))
m((x − h, x + h))
1 = lim+ = lim+ + ,
h→0 2h h→0 2h 2h
we have
m(E ∩ (x − h, x + h)) m(E c ∩ (x − h, x + h))
lim+ = 1 − lim+ = 1 − 1 = 0.
h→0 2h h→0 2h
MATH 545 and MATH 546 Prelim Problems 40
(c) Suppose for contradiction that m(E) > 0, then there must exist a point of density x of E. By
assumption we have m(E ∩ (x − h, x + h)) ≤ 0.99 m((x − h, x + h)) = 0.99(2h). By part b
m(E ∩ (x − h, x + h)) 0.99(2h)
1 = lim+ ≤ lim+ = 0.99.
h→0 2h h→0 2h
Which is a contradiction.
43. Let L1Loc (R, m) be the set of Lebesgue measurable functions f : R → [−∞, ∞] such that
Rn
−n
|f | dm < ∞ for each n ∈ N.
SOLUTION:
(a) Since f ∈ L1Loc (R, m), f ∈ L1 (R, m). So by Problem 42, we have for m-a.e. x0 ∈ R,
Z x0 +h Z x0 Z x0 +h
1 1 1
lim f (x) dm(x) = lim+ f (x) dm(x) + lim+ f (x) dm(x) = f (x0 ).
h→0+ 2h x0 −h h→0 2h x0 −h h→0 2h x0
since f = fn on (x0 − h, x0 + h) ⊂ (−n, n) for h sufficiently small. Since fn ∈ L1 ((−n, n), m),
Z x0 +h
1
lim fn (x) dm(x) = fn (x0 )
h→0+ 2h x0 −h
for m-a.e. x0 ∈ (−n, n) and hence m(An ) = 0 for all n. Therefore m(∪n An ) = 0 and hence for
all x0 ∈ R \ ∪n An ) (i.e., a.e. x0 ∈ R) we have
Z x0 +h
1
lim f (x) dm(x) = f (x0 ).
h→0+ 2h x0 −h
MATH 545 and MATH 546 Prelim Problems 41
Then since |f (x) − ri | ∈ L1 (R, m) for each i, m(Ei ) = 0, by part a, and hence m(∪i Ei ) = 0.
/ ∪∞
Claim. For all x0 ∈ i=1 Ei , x0 is a Lebesgue point of f .
Proof: Let > 0. Since rationals are dense in R, there exists i ∈ N such that |f (x0 ) − ri | < /2.
Then
Z x0 +h
1
lim |f (x) − f (x0 )| dm(x)
h→0+ 2h x0 −h
Z x0 +h Z x0 +h
1 1
≤ lim+ |f (x) − ri )| dm(x) + lim+ |f (x0 ) − ri | dm(x)
h→0 2h x0 −h h→0 2h x0 −h
44. Let f ∈ L1 (Rk , m) where m is Lebesgue measure. Define g(x) = f (x) if |f (x)| ≤ 1 and
g(x) = 0 otherwise. Suppose x0 is a Lebesgue point for f such that f (x0 ) < 1. Show that
x0 is a Lebesgue point for g.
SOLUTION:
Let A = {x ∈ Rk : |f (x)| ≤ 1}. Then g(x) = f (x)χA (x). Since f ∈ L1 (Rk , m), f is Lebesgue
measurable and hence A is Lebesgue measurable. By Problem 42 (b), a.e. x ∈ A is a point of
density of A, and since |f (x0 )| < 1, x0 ∈ A and therefore it may or may not be a density point of
A, but for sure
m(Ac ∩ B(x0 , r))
lim+ = 0,
r→0 m(B(x0 , r))
MATH 545 and MATH 546 Prelim Problems 42
SOLUTION:
|fn (x)| = |f (x) − f (x + 1/n)|p ≤ (|f (x)| + |f (x + 1/n)|)p ≤ (2 max{|f (x)|, |f (x + 1/n)|})p
≤ 2p max{|f (x)|p , |f (x + 1/n)|p }
≤ 2p (|f (x)|p + |f (x + 1/n)|p )
≡ g.
But
Z Z Z
p p p
g(x) dx = 2 |f (x)| + |f (x + 1/n)| < ∞. (Since f ∈ Lp )
R R R
Then by DCT,
Z Z
p
lim |f (x) − f (x + 1/n)| dx = lim |f (x) − f (x + 1/n)|p dx = 0.
n→∞ R R n→∞
Another Solution
First suppose f is continuous and there exists n ∈ N such that f (x) = 0 for x ∈
/ [−n, n], i.e. f is
continuous with compact support. Then f is uniformly continuous on [−n − 1, n + 1] (A continuous
MATH 545 and MATH 546 Prelim Problems 43
function on a compact set is uniformly continuous). Note that if |h| < 1, then for x ∈ / [−n − 1, n + 1]
we have f (x + h) = 0. Let > 0. By uniform continuity of f , there exists δ ∈ (0, 1) such that
|f (x) − f (x + h)| < /(2n + 2)p for all h such that |h| < δ and for all x ∈ [−n − 1, n + 1]. Hence for
all h such that |h| < δ,
Z 1/p Z n+1 1/p
p p
|f (x) − f (x + h)| dx = |f (x) − f (x + h)| dx
R −n−1
n+1 1/p
p
Z
< 1 dx = .
2n + 2 −n−1
This establishes the result in the case that f is continuous with compact support.
Now let f ∈ Lp . Let > 0. Then by Theorem 1.1.26 there exists a continuous function g vanishing
outside [−n, n] for some n ∈ N such that kf − gkp < /3. By the case above there exists δ > 0
such that for all h with |h| < δ, kg(x) − g(x + h)| < /3. Note that by the translation-invariance of
Lebesgue integral, kg(x + h) − f (x + h)kp = kg(x) − f (x)kp < /3. Hence by Minkowski’s inequality
kf (x) − f (x + h)kp ≤ kf (x) − g(x)kp + kg(x) − g(x + h)kp + kg(x + h) − f (x + h)kp
< + + = .
3 3 3
46. If h, g ∈ L2 (R, m), prove that
Z 1
F (x) := g(y − x)h(y) dy
0
is a continuous function on R.
SOLUTION:
Let c > 0, we need to prove that
lim |F (x + c) − F (x)| = 0.
c→0
To do that we are going to use Holder’s inequality. For each x ∈ R and y ∈ [0, 1], let G(x) := g(y−x),
then G ∈ L2 (R, m) since g is. So
Z 1 Z 1
lim |F (x + c) − F (x)| = lim g(y − (x + c))h(y) dy − g(y − x)h(y) dy
c→0 c→0 0 0
Z 1
≤ lim |h(y)||g(y − (x + c)) − g(y − x)| dy
c→0 0
Z 1 1/2 Z 1 1/2
2 2
≤ lim |h(y)| dy |g(y − (x + c)) − g(y − x)| dy
c→0 0 0
= 0.
Where last limit is zero by the previous problem.
MATH 545 and MATH 546 Prelim Problems 44
47. The Laplace transform of real measurable functions on [0, ∞) is defined by L[f ] =
R∞
0
f (t) e−st dt. If f ∈ L1 [0, ∞), prove that L[f ] exists and is bounded for all s ≥ 0. If
fn , f ∈ L1 [0, ∞) for all n, and fn → f in L1 [0, ∞), show that L[fn ] → L[f ] uniformly on
[0, ∞).
SOLUTION:
For all s, t ≥ 0, Z ∞ Z ∞
|f (t)| e−st dt ≤
|L[f ]| ≤ |f (t)| dt = kf k1 < ∞.
0 0
Hence if f ∈ L1 [0, ∞), L[f ] exists and is bounded for all s ≥ 0. Now let > 0. Then since fn → f
R∞
in L1 , there exists an N ∈ N such that for all n > N , 0 |fn (t) − f (t)| dt < . For all s ≥ 0 and
n > N , we have
Z ∞
−st
|L[fn ] − L[f ]| = e (fn (t) − f (t)) dt
Z 0∞
≤ |fn (t) − f (t)| dt < .
0
48. Let f be a real everywhere differentiable function on (−∞, ∞) show that f 0 is Borel
measurable. Further if |f 0 | ≤ M , show that f is absolutely continuous on (−∞, ∞).
SOLUTION:
Since f is differentiable everywhere, f is continuous everywhere and hence is Borel. Now notice that
f (x + 1/n) − f (x)
f 0 (x) = lim = lim (nf (x + 1/n) − nf (x)).
n→∞ 1/n n→∞
But nf (x) and nf (x + 1/n) are continuous everywhere, since f is, and hence Borel. A point-wise
limit of Borel functions is also Borel as well. Therefore, f 0 is Borel measurable. Now let > 0 and
let {(xi−1 , xi )}ni=1 be a collection of disjoint open intervals with ni=1 (xi − xi−1 ) < /M . Then
P
Xn n Z xi
X X n Z xi Xn
0 0
|f (xi ) − f (xi−1 )| =
f (x) dx ≤
|f | dx ≤ M (xi − xi−1 ) < .
i=1 i=1 xi−1 i=1 xi−1 i=1
49. Let {fk } be a non-decreasing sequence of real valued measurable functions on (X, F)
and let µ be a non-negative measure on (X, F). Assume X f1− dµ < ∞, show that
R
Z Z
lim fk dµ = lim fk dµ.
k→∞ X X k→∞
MATH 545 and MATH 546 Prelim Problems 45
SOLUTION:
Notice that if f is non-decreasing, then its positive and negative parts are still non-decreasing. Now
consider the sequence gk = fk +f1− , then {gk } is non-decreasing and gk ≥ g1 = f1+ −f1− +f1− = f1+ ≥ 0
for all k. Then we can apply MCT on gk to have
Z Z
−
lim (fk + f1 ) dµ = lim (fk + f1− ) d
k→∞ X k→∞
X Z
Z Z Z
−
lim fk dµ + f1 dµ = lim fk d + f1− dµ
k→∞ X k→∞
Z ZX X Z X
=⇒ lim fk dµ = lim fk dµ (Since f1− dµ < ∞).
k→∞ X X k→∞ X
50. Let f, g be non-negative real measurable functions on [0, ∞). Let 1 < p ≤ 2 and q =
R∞ 1
p/(p − 1). If g ∈ Lp ([0, ∞), m) and 0 f (t) t q −1 dt < ∞, show that
Z ∞Z x Z ∞
g(x) 1
f (t) dt dx ≤ kgkp (q − 1)−1/q
f (t) t q −1 dt < ∞.
0 0 x 0
SOLUTION:
g(x)
Since f, g are measurable and 1/x is continuous on [0, ∞), and hence measurable, f (t) is m×m-
x
g(x)
measurable on [0, ∞). Moreover, f (t) ≥ 0 for all t, x in [0, ∞). With these information and the
x
fact ([0, ∞), M, m) is a σ-finite measure space, Tonelli’s Theorem can be applied. Also note that
q > 1. So we have
Z ∞Z x Z ∞Z ∞
g(x) g(x)
f (t) dt dx = f (t)χ{x≥t≥0} dt dx
0 0 x 0 0 x
Z ∞Z ∞
g(x)
= f (t)χ{x≥t≥0} dx dt (Tonelli’s)
0 0 x
Z ∞ Z ∞
g(x)
= f (t) dx dt
0 t x
Z ∞ Z ∞ 1/p Z ∞ 1/q
p 1
≤ f (t) g (x) dx dx dt (Holder’s)
0 t t xq
Z ∞ Z ∞ 1/q
1
≤ kgkp f (t) dx dt
0 t xq
Z ∞ ∞ !1/q
x1−q
= kgkp f (t) dt
1 − q
0
Z ∞ t
1
= kgkp (q − 1)−1/q f (t)t q −1 dt < ∞.
0
MATH 545 and MATH 546 Prelim Problems 46
51. Let (X, A, µ) be a measure and let {Ek }k be a sequence of measurable sets in X, such
that
X∞
µ(Ek ) < ∞.
k=1
SOLUTION:
Let A = {x ∈ X : x ∈ Ek for infinitely many k}. Then we need to prove that µ(A) = 0.
Claim.
∞ [
\ ∞
A= Ek .
n=1 k=n
Proof: Let x ∈ A, then x ∈ Ek for infinitely many k, so for any n ∈ N, there is k ∈ N, k ≥ n such
that x ∈ Ek . That is x ∈ ∪∞ ∞ ∞
k=n for every n ∈ N and so x ∈ ∩n=1 ∪k=n Ek . For the other side, let
x ∈ ∩∞ ∞ ∞
n=1 ∪k=n Ek , then x ∈ ∪k=n Ek for some n ∈ N and hence x ∈ Ek for all k ≥ n, i.e. infinitely
many k, and thus x ∈ A. #
By the claim A ⊂ ∪∞
k=n Ek , and therefor
∞
X
µ(A) ≤ µ(Ek ) for each n ∈ N. (∗)
k=n
P∞ P∞
Since k=1 µ(Ek ) < ∞, limn→∞ k=n µ(Ek ) = 0. Take the limit as n → ∞ on both sides of (∗) we
have µ(A) = 0.
52. Let µ, ν, and λ be σ-finite, nonnegative and nontrivial measures on the measure space
(X, M) such that for every measurable set A ∈ M,
Show:
Hence µ(A) = 0.
SOLUTION:
Notice that
| sin πx|
Z Z
1
dm(x × y) = dm(x × y)
[0,∞)×[0,∞) (y + ex | sin πx|)2 [0,∞)×[0,∞)
y ex
( 2 + )2
sin πx | sin πx|
Z
1
≤ x 2
dm(x × y)
[0,∞)×[0,∞) (y + e )
Notice that f (x, y) = 1/(y+ex )2 is continuous for all x, y ∈ [0, ∞), hence m×m-Lebesgue measurable,
f (x, y) ≥ 0 and ([0, ∞), M, m) is σ-finite measure space. Thus by Tonelli’s Theorem
Z Z ∞ Z ∞
1 1
dm(x × y) = dy dx
[0,∞)×[0,∞) (y + ex )2 0 0 (y + ex )2
∞ Z ∞
−1 ∞
Z
= dx = e−x dx = 1 < ∞.
0 y + ex 0 0
MATH 545 and MATH 546 Prelim Problems 48
Hence sin πx/(y + ex | sin πx|)2 is integrable and hence by Fubini theorem
Z ∞Z ∞ Z ∞Z ∞
sin πx sin πx
x 2
dx dy = dy dx
0 0 (y + e | sin πx|) 0 0 (y + ex | sin πx|)2
Z ∞ ∞
sin(πx)
= − dx
0 y + ex | sin(πx)| y=0
Z ∞
sin(πx) −x
= e dx
0 | sin(πx)|
X∞ Z n+1
= (−1)n e−x dx
n=0 n
X∞
= (e−n − e−(n+1) )(−1)n
n=0
Note 1.2.2. We can do the following to check that f (x, y) = sin πx/(y + ex | sin πx|)2 is integrable
and hence Fubini can be applied: Notice that f (x, y) is measurable since it is continuous, and
|f (x, y)| ≥ 0, then by Tonelli’s
Z ∞Z ∞
| sin πx| | sin πx|
Z
x 2
dm(x × y) = dy dx
[0,∞)×[0,∞) (y + e | sin πx|) 0 0 (y + ex | sin πx|)2
Z ∞ Z ∞
1
= | sin πx| dy dx
0 0 (y + e | sin πx|)2
x
Z ∞
−1 ∞
= | sin πx| dx
y + ex | sin πx| 0
0
Z ∞
1
= | sin πx| x dx
0 e | sin πx|
Z ∞
= e−x dx = 1 < ∞.
0
Z
1
54. Let f ∈ L1 (R). For s > 0 define g(s) := (s − r) f (r) dr.
s (0,s)
SOLUTION:
MATH 545 and MATH 546 Prelim Problems 49
Claim. Z s+h
1
lim+ (s + h − r) f (r) dr = 0.
h→0 sh + h2 s
MATH 545 and MATH 546 Prelim Problems 50
Proof: Again since f ∈ L1 (R), for all > 0, there is δ > 0 with m((s, s + h)) = h < δ and
Z
|f (r)| dr < (s + h).
(s,s+h)
Therefore,
s+h Z s+h
Z
1 1
sh + h2 (s + h − r) f (r) dr − 0 ≤ (s + h − s) |f (r)| dr
s sh + h2 s
Z s+h
1
= |f (r)| dr < .#
s+h s
1 R
Hence g 0 (s) = r f (r) dr.
s2 (0,s)
(c) Clearly we can define g(0) = 0 because lims→0+ g(s) = 0 as shown in part a. Recall from
Problem 7, the first step we need to do is proving that g 0 is integrable on an arbitrary interval
[0, x] ⊂ [0, 1] and
Z x
g 0 (s) ds = g(x) − g(0) = g(x). (Note that 0 ≤ s ≤ x)
0
0
So, to prove g is integrable we have to show
Z x Z xZ s Z xZ x
0 r r
|g (s)| ds = 2
|f (r)| dr ds = 2
|f (r)|χ{x≥s≥r≥0} dr ds < ∞.
0 0 0 s 0 0 s
Notice that (r/s2 )|f (r)| is measurable since f is measurable and r/s2 is continuous and hence
measurable. Also (r/s2 )|f (r)| ≥ 0, hence by Tonelli’s
Z x Z xZ x
0 r
|g (s)| ds = 2
|f (r)|χ{x≥s≥r≥0} ds dr
0 0 0 s
Z x Z x
1
= r |f (r)| 2
ds dr
r s
Z0 x Z x
r
= 1− |f (r)| dr ≤ |f (r)| dr < ∞. (Since f ∈ L1 )
0 x 0
x
So g 0 is integrable. Now we need to show 0 g 0 (s) ds = g(x). Since g 0 is integrable, by Fubini’s
R
Theorem
Z x Z xZ x
0 r
g (s) ds = f (r)χ{x≥s≥r≥0} ds dr
0 s2
Z0 x 0 Z x
1
= r f (r) 2
ds
r s
Z0 x
1 x
Z
r
= 1− f (r) dr = (x − r) f (r) dr = g(x).
0 x x 0
Since g 0 is integrable, for all > 0, there is a δ > 0 such that if E is a measurable set with
m(E) < δ, then Z
|g 0 (s)| ds < .
E
MATH 545 and MATH 546 Prelim Problems 51
Now let E = ∪ni=1 (ai , bi ) with m(E) = ni=1 (bi − ai ) < δ, where {(ai , bi )}ni=1 is a collection of
P
finite disjoint intervals in [0, 1] ordered in such a way bi < ai+1 for i = 1, 2, ..., n − 1. Then
Xn n Z b i
X Z ai n Z bi
0 0
X
|g 0 (s)| ds
|g(bi ) − g(ai )| =
g (s) ds − g (s) ds ≤
i=1 i=1 0 0 ai
Zi=1
= |g 0 (s)| ds
∪n
i=1 (ai ,bi )
Z
= |g 0 (s)| ds < .
E
SOLUTION:
(We can find this limit by L’Hopital’s rule, also notice that if n(1 − e−x/n ) = y, then (1 − ny )n = e−x ,
thus as we know the limit on the left will give us e−y . Compare it to e−x , we have y = x.)
Also, |fn | ≤ n ≡ g, which is clearly integrable on [0, 1]. Hence by DCT
Z 1 Z 1 Z 1
−x/n
−x/n
1
lim n 1−e dx = lim n 1 − e dx = x dx = .
n→∞ 0 0 n→∞ 0 2
56. Let µn be a sequence of nonnegative measures on a measurable space (X, M) with
µn (X) = 1, for n ∈ N ∪ {0}. Construct a measure ν such that ν(X) = 1 and µn ν, for all
n ∈ N ∪ {0}.
SOLUTION:
P∞ µn
Let ν = n=0 . We need to prove that ν is a measure (not a signed measure), and it satisfies
2n+1
ν(X) = 1 and µn ν, for all n ∈ N ∪ {0}.
P∞ µn (∅)
ν is nonnegative since µn is nonnegative for all n = 0, 1, .... Also, ν(∅) = n=0 = 0 since µn
2n+1
is measure for all n = 0, 1, .... Finally, let {Ei } be a countable collection of disjoint sets in M, then
∞ ∞ X ∞
X µn (∪∞i=1 Ei )
X µn (Ei )
ν(∪∞
i=1 Ei ) = n+1
= .
n=0
2 n=0 i=0
2n+1
But since µn is snonnegative for all n, we can switch the order of summations (We can do that also
if we consider the counting measure and then apply Tonelli’s). So we get
∞ X
∞ ∞
X µn (Ei ) X
ν(∪∞
i=1 Ei ) = ν(Ei ).
i=0 n=0
2n+1 i=0
Note that ∞ ∞
X µn (X) X 1
ν(X) = = = 1. (Geometric series)
n=0
2n+1 n=0
2n+1
but this is possible if and only if µn (A) = 0 for all n = 0, 1, .... Hence µn ν for all n.
Chapter 2
Complex Analysis
2.1 Review
Definition 2.1.1. Let f (z) = u + iv. Define
∂ 1 ∂ 1 ∂ ∂ 1 ∂ 1 ∂
= + and = − .
∂z 2 ∂x i ∂y ∂z 2 ∂x i ∂y
Also, we can define the Laplacian ∆ as
∂2 ∂2 ∂ ∂ ∂ ∂
∆= 2
+ 2
=4 =4 .
∂x ∂y ∂z ∂z ∂z ∂z
Proposition 2.1.1. If f is holomorphic at z0 , then
∂f ∂f ∂u
(z0 ) = 0 and f 0 (z0 ) = (z0 ) = 2 (z0 ) = ux + ivx = ux − ivy .
∂z ∂z ∂z
In other words, if f is holomorphic, then it satisfies Cauchy-Riemann equations.
Theorem 2.1.2. Suppose f = u + iv is a complex-valued function defined on an open set Ω. If u and
v are continuously differentiable and satisfy the Cauchy-Riemann equations on Ω, then f is holomorphic
on Ω and f 0 (z) = ∂f /∂z.
Theorem 2.1.3. Given a power series ∞ n
P
n=0 an (z − z0 ) , there exists 0 ≤ R ≤ ∞ such that :
53
MATH 545 and MATH 546 Prelim Problems 54
Corollary 2.1.5. A power series is infinitely complex differentiable in its disc of convergence, and the
higher derivatives are also power series obtained by term-wise differentiation.
Definition 2.1.2. Integral of f Along Curve γ: Given a smooth curve γ parametrized by z(t) :
[a, b] → C (e.g z(t) = z0 + reit , t ∈ [0, 2π]) and f a continuous function on γ, we define the integral of f
along γ by Z Z b
f (z) dz = f (z(t))z 0 (t) dt.
γ a
Definition 2.1.3. Primitive: Suppose f is a function on the open set Ω. A primitive for f on Ω is a
function F that is holomorphic on Ω and such that F 0 (z) = f (z) for all z ∈ Ω.
Theorem 2.1.6. If a continuous function f has a primitive F in Ω, and γ is a curve in Ω that begins
at w1 and ends at w2 , then Z
f (z) dz = F (w2 ) − F (w1 ).
γ
Theorem 2.1.7. Goursat’s Theorem: If f is holomorphic in open set Ω and T ⊂ Ω a triangle whose
interior is also contained in Ω. Then Z
f (z) dz = 0.
T
Proof: Let T denote the original T oriented counterclockwise. Let d0 and p0 denote the diameter
0
and perimeter of T 0 , respectively. Bisect each side of the triangle and connecting the midpoints. This
creates four new smaller triangles, denoted T11 , T21 , T31 and T41 with the same orientation as T 0 . Hence we
have Z Z Z Z Z
f (z) dz = f (z) dz + f (z) dz + f (z) dz + f (z) dz.
T0 T11 T21 T31 T41
And therefore Z Z
f (z) dz ≤ 4
f (z) dz ,
T 0 T 1
MATH 545 and MATH 546 Prelim Problems 55
Then d1 = d0 /2 and p1 = p0 /2. We now repeat this process for the triangle T 1 , bisecting it into four
smaller triangles. Continuing this process, we obtain a sequence of triangles T 0 , T 1 , ..., T n with the
properties that Z Z
n
0 f (z) dz ≤ 4 n f (z) dz ,
T T
and d = d /2 and p = p /2 . We also denote by T the solid closed triangle with boundary T n , and
n 0 n n 0 n n
observe that our construction yields a sequence of nested compact sets T n ⊂ T n−1 ⊂ ... ⊂ T 1 whose
diameter goes to 0. Hence there exists a unique point z0 that belongs to all the solid triangles T n . Since
f is holomorphic at z0 we can write
where ψ(z) → 0 as z → z0 . Now both the constant and linear functions f (z0 ) and (z − z0 )f 0 (z0 ) have
primitive in T 1 and hence in T n . So we can integrate to get
Z Z Z
f (z) dz = n
ψ(z)(z − z0 ) dz ≤ d sup |ψ(z)| dz = dn pn sup |ψ(z)| = 4−n d0 p0 sup |ψ(z)|.
Tn Tn z∈T n Tn z∈T n z∈T n
Therefore, Z
f (z) dz ≤ 4n 4−n d0 p0 sup |ψ(z)| = d0 p0 sup |ψ(z)| → 0 as z → 0.
z∈T n z∈T n
0
T
Theorem 2.1.8. A holomorphic function in an open disc has a primitive in that disc.
Theorem 2.1.10. Cauchy’s Integral Formulas: Suppose f is holomorphic in an open set Ω that
contains the closure of a disc D. If C denotes the boundary circle of this disc with the positive orientation,
then Z
1 f (ζ)
f (z) = dζ for any point z ∈ D.
2πi C ζ − z
Moreover, f has infinitely many complex derivatives in Ω (Regularity Theorem), and
Z
(n) n! f (ζ)
f (z) = dζ for any point z ∈ D.
2πi C (ζ − z)n+1
The above integrals vanish when z is outside D.
MATH 545 and MATH 546 Prelim Problems 56
Corollary 2.1.11. Cauchy Inequalities: If f is holomorphic in an open set that contains the closure
of a disc D centered at z0 and of radius R, then
f (z0 ) ≤ n!kf kC ,
(n)
Rn
where kf kC = supz∈C |f (z)| denotes the supremum of |f | on the boundary circle C.
Theorem 2.1.12. f Holomorphic =⇒ f Analytic: Suppose f is holomorphic in an open set Ω. If
D is a disc centered at z0 and whose closure is contained in Ω, then f has a power series expansion at z0
∞
X f (n) (z0 )
f (z) = (z − z0 )n .
n=0
n!
Proof: Fix z ∈ D. By Cauchy’s integral formula:
Z
1 f (ζ)
f (z) = dζ. (∗)
2πi C ζ − z
Write 1/(ζ − z) as
1 1 1 1
= = .
ζ −z ζ − z0 − (z − z0 ) ζ − z0 1 − − z0
z
ζ − z0
Since z ∈ D and ζ ∈ C, the boundary of D, we have
z − z0
ζ − z0 < 1.
In other words: If the zeros of a holomorphic function f in the connected open set Ω accumulate in
Ω, then f is identically 0.
Therefore, if we suppose f and g are holomorphic in a region Ω and f (z) = g(z) for all z in some
non-empty open subset of Ω (or more generally for z in some sequence of distinct points with limit point
in Ω). Then f (z) = g(z) throughout Ω.
Note 2.1.2. By this theorem, if f is analytic in Ω, then we can possibly extend f to outside Ω to get
a new holomorphic function F on a new domain Ω0 with F = f on Ω. Such F is called an analytic
continuation of f into the region Ω0 .
Theorem 2.1.15 (Morera’s Theorem). Suppose f is a continuous function in the open disc D such
that for any triangle T ⊂ D, we have Z
f (z) dz = 0.
T
Then f is holomorphic.
Proof: If we prove that f has a primitive F , then since F is holomorphic by Definition 2.1.3, it has
infinitely many holomorphic derivatives by regularity theorem Theorem 2.1.10 and hence F 00 = f 0 exists,
so f is holomorphic. After a translation, we may assume W.L.O.G that the disc D is centered at the
origin. Given a point z ∈ D, consider the piecewise-smooth curve that joins 0 to z first by moving in the
horizontal direction from 0 to z̃ where z̃ =Re(z), and then in the vertical direction from z̃ to z. We choose
the orientation from 0 to z, and denote this polygonal line, which consists of at most two segments, by
γz . Define Z
F (z) = f (w) dw.
γz
Then
"Z # "Z #
F (z + h) − F (z)
Z Z
1 1
= f (w) dw − f (w) dw = f (w) dw + f (w) dw
h h γz+h γz h γz+h −γz
Z
1
= f (w) dw,
h η
where η is the straight line segment between z and z + h. Since f is continuous at z we can write
f (w) = f (z) + ψ(w),
where ψ(w) → 0 as w → z. (Since f is continuous.) Hence
F (z + h) − F (z)
Z Z
f (z) 1
= dw + ψ(w) dw
h h η h η
f (z) supw∈η |ψ(w)|
≤ [(z + h) − (z)] + [(z + h) − (z)]
h h
→ f (z) as h → 0.
MATH 545 and MATH 546 Prelim Problems 58
Proof: First recall that a uniform limit of a sequence of continuous functions is continuous. Moreover,
if these functions {fn } are integrable on a compact set D ⊂ Ω, then f is integrable and
Z Z
lim fn (z) dz = f (z) dz.
n→∞ D D
Let D be any disk whose closure is contained in Ω. Let T be a triangle contained in Ω. Then since {fn }
are holomorphic on Ω and hence on T , by Goursat’s Theorem Theorem 2.1.7,
Z
fn (z) dz = 0
T
for all n. Since {fn } converges uniformly to f , f is continuous and because D is a closed set contained
in Ω, then it is compact, and thus by the above note f is integrable on D and hence any subset of it and
Z Z
0 = lim fn (z) dz = f (z) dz.
n→∞ T T
Thus, by Morera’s Theorem Theorem 2.1.15, f is holomorphic on D. Since that true for all D whose
closure contained in Ω, f is holomorphic on Ω.
Theorem 2.1.17 (fn0 converges uniformly to f 0 ). If {fn } is a sequence of holomorphic functions that
converges uniformly to a function f in every compact subset of Ω, then the sequence of derivatives {fn0 }
converges uniformly to f 0 on every compact subset of Ω.
Proof: W.L.O.G. suppose {fn } converges uniformly on all of Ω. For each δ > 0 define
Ωδ = {z ∈ Ω : D(z, δ) ⊂ Ω}.
Then Ωδ ⊂ Ω and hence by Cauchy’s integral formula Theorem 2.1.10 for any z ∈ Dδ (z) ⊂ Ω
Z
0 0 0 1 fn (ζ) − f (ζ)
|(fn − f ) (z)| = |fn (z) − f (z)| = dζ
2π Cδ (ζ − z)2
supζ∈Ω |fn (ζ) − f (ζ)| 2π δieiθ
Z
≤ dθ
2π
0 δ 2 e2iθ
1
= sup |fn (ζ) − f (ζ)| → 0 (since fn converges uniformly to f ).
δ ζ∈Ω
Theorem 2.1.18. Let F (z, s) be defined for (z, s) ∈ Ω × [0, 1] where Ω open set in C. Suppose F satisfies
the following:
is holomorphic on all of Ω.
F (z) = f (z).
Theorem 2.1.21 (Runge Theorem). Any function that is holomorphic in a neighborhood of compact
set K can be approximated uniformly on K by rational functions whose singularities are in K c . Moreover,
if K c is connected, then f can be approximated by polynomials on K.
i.) Removable singularity Let f be a function holomorphic in an open set Ω except possibly at one
point z0 in Ω. If we can define f at z0 in such a way that f becomes holomorphic in all of Ω, we
say that z0 is a removable singularity for f . E.g.: f (z) = z on C \ {0}, then z = 0 is a removable
singularity, because we can define f (z) to be 0 at z = 0 and we notice that f (z) = z is holomorphic
on all C.
ii.) Pole singularity: We say that a function f defined in a deleted neighborhood of z0 has a pole at
z0 , if the function 1/f , defined to be zero at z0 , is holomorphic in a full neighborhood of z0 . E.g. :
f (z) = 1/z defined on C \ {0} has a pole at z = 0, because g(z) = 1/f (z) = z has a zero at z = 0
and is defined on all C.
iii.) Essential singularity: An essential singularity is not a removable singularity or a pole. E.g.:
f (z) = e1/z defined on C \ {0} behaves crazy near z = 0. If we reach z = 0 on the real axis from
right, f goes to ∞, while if we reach z = 0 on the real axis from left f goes to 0. Also, it oscillates
rapidly, yet remains bounded, as z approaches the origin on the imaginary axis.
1 dn−1
resz0 f = lim n−1
(z − z0 )n f (z).
z→z0 (n − 1)! dz
Theorem 2.1.24. Suppose that f is holomorphic in an open set containing a circle C and its interior,
except for a pole at z0 inside C. Then
Z
f (z) dz = 2πi resz0 f.
C
Corollary 2.1.25 (Residue Formula). Suppose that f is holomorphic in an open set containing a
circle C or a toy contour γ and its interior, except for poles at the points z1 , ..., zN inside C or γ . Then
for the circle C
Z XN
f (z) dz = 2πi reszk f.
C k=1
Proof: Let D be a disc centered at z0 whose closure is contained in Ω. Let C denote its boundary
circle. We need to prove that for z 6= z0
Z
1 f (ζ)
f (z) = dζ. (∗)
2πi C ζ −z
Why? Because we can consider the parametrization of C defined by z(t) : [0, 1] → C and then rewrite
the above integral as
Z 1
f (ζ(t))ζ 0 (t)
Z
f (ζ)
dζ = dt.
C ζ −z 0 ζ(t) − z
And this integral is holomorphic in z for each t and continuous in both z and t, then by Theorem 2.1.18,
this integral is holomorphic for all D. Thus for z 6= z0 this integral agrees with f (z), and it is holomorphic
on all of D. Then by the definition of removable singularity, this is the continuation we looking for and
hence z0 is a removable singularity. To prove (∗), fix z ∈ D, z 6= z0 and consider the toy contour Γ shown
in the figure
MATH 545 and MATH 546 Prelim Problems 62
Figure 2.1: The multiple keyhole contour Γ in the proof of Riemann’s theorem
Let γ and γ0 be small circles of radius with negative orientation and centered at z and z0 respectively.
Then since f (ζ)/(ζ − z) is holomorphic on Γ and in its interior, then
Z Z Z Z
f (ζ) f (ζ) f (ζ) f (ζ)
dζ + dζ + dζ = dζ = 0.
C ζ −z γ ζ − z γ0 ζ − z Γ ζ −z
But f (z) is holomorphic on γ and its interior. Hence by Cauchy’s integral formula,
Z
f (ζ)
dζ = −2πif (z),
γ ζ − z
Z
f (ζ)
and the negative sign comes from negative orientation of γ . For the second integral dζ, as
γ0 ζ − z
→ 0, there exists δ > 0 such that |ζ − z| > δ, where ζ ∈ γ0 (because z 6= z0 and as → 0, ζ → z0 ). And
from assumption, |f (z)| ≤ M for all z ∈ D \ {z0 }. Hence
Z Z
f (ζ) |f (ζ)| 2πM
0 ζ − z ≤ 0 |ζ − z| dζ ≤ δ → 0 as → 0.
dζ
γ γ
Then g(z) is holomorphic in Dr (z0 ) \ {z0 } since f (z) − w is, and we have |g(z)| < 1/δ, i.e. bounded.
Then by Riemann theorem of removable singularity, g(z) has a removable singularity at z0 , i.e. g can be
redefined on all Dr (z0 ) in such a way that makes it holomorphic on all Dr (z0 ). So we let g(z0 ) = c. If
c 6= 0, then f (z0 ) − w is finite and f (z) − w is holomorphic at z0 , because g, and hence 1/g, is holomorphic
at z0 . But this contradicts the fact that z0 is an essential singularity of f (which means f cannot be
defined at z = z0 . If c = 0, then by the definition of the pole, z = z0 is a pole for f (z) − w, i.e. a pole
for f (z). But this is also contradicts the fact that z0 is an essential singularity.
Definition 2.1.6. Suppose that f is holomorphic for all large values of z. Let F (z) = f (1/z), then F
is holomorphic in a deleted neighborhood of the origin. We say that f has a pole at infinity if F has
a pole at z = 0, i.e. if 1/f (1/z) has a zero at z = 0. And we say f has a removable singularity at
infinity if F has a removable singularity at z = 0, i.e. f (1/z) is holomorphic at z = 0. And we say that
f has an essential singularity at infinity if F has an essential singularity at z = 0.
A meromorphic function in the complex plane that is either holomorphic at infinity or has a pole
at infinity is said to be meromorphic in the extended complex plane.
1. A topological space X is called compact if each of its open covers has a finite subcover.
2. The complex plane C is not compact. For instance, here is an open cover that does not admit any
finite subcover:
The reason why U does not admit any finite subcover is easy: if you pick just a finite number of
Un ’s, then their union would be equal to the biggest one of them, which is not the whole C.
To understand it more deeply: A sequence like {1, 2, 3, ...} which tends to infinity does not have
a convergent subsequence in C, whereas any sequence in C has a convergent subsequence (because
convergence here accepts infinity as a limit).
The addition of infinity of called the one- point compactification (Riemann sphere)
4. Every closed discrete subset of a compact space is finite. Hence every closed discrete subset of C is
finite.
MATH 545 and MATH 546 Prelim Problems 64
Theorem 2.1.29. If f is meromorphic in the extended complex plane, then f has finitely many poles.
Proof:
By definition a pole is an isolated singularity, and so each pole has a neighborhood containing no other
poles besides itself. Thus the set of all poles is discrete.
Also, The set of poles is closed, since its complement, the set of points at which f is holomorphic, is
open, since any point where f is holomorphic has a neighborhood where f is holomorphic on it.
Also notice from the definition of meromorphic functions, if we suppose we have a sequence of poles
{z0 , z1 , z2 , ...} (sequence means: infinite number of points), then by definition of meromorphic functions,
this sequence has no limit points in the domain of f . But since f is defined on C, then the sequence will
accumulate in C (no other place to accumulate in), hence f is not meromorphic (against the definition),
which is a contradiction.
Theorem 2.1.30. The meromorphic functions in the extended complex plane are the rational functions.
Proof:
From the previous theorem, f has finite number of poles: {z0 , z1 , ..., zn } and by Definition 2.1.7 f could
be holomorphic at infinity or has a pole there. The idea is to write g(z) = f (z) − P∞ (z) − nk=0 Pk (z),
P
at ∞, then in both cases f (z) → P∞ (z) as z → ∞. Also, notice that Pk (z) → 0 as z → ∞ for all such
k. By the definition of g, g is entire. And since limz→∞ g(z) = P∞ (z) − P∞ (z) − 0 = 0. Hence, there
exists an M > 0 such that |g(z)| ≤ M for all z. Hence by Liouville’s theorem Theorem 2.1.13 g is constant.
Another proof: As in the previous proof there are a finite number of poles for f , namely {z0 , z1 , ..., zN }.
Let n0 , n1 , ..., nN be the orders of these poles, respectively. Then
(1) (1) (2) (2) (N ) (N )
a−n0 a−1 a−n1 a−1 a−nN a−1
f (z) = n
+ ... + + n
+ ... + + ... + n
+ ... +
(z − z0 ) 0 (z − z0 ) (z − z0 ) 1 (z − z0 ) (z − z0 ) N (z − z0 )
+ G1 (z) + G2 (z) + ... + GN (z),
MATH 545 and MATH 546 Prelim Problems 66
Theorem 2.1.32 (Rouché’s Theorem). Suppose f and g are holomorphic functions in an open set
containing a circle C and its interior. If
Definition 2.1.8 (Open Map). A mapping is said to be open if it maps open sets to open sets.
Proof:
Let w0 belong to the image of f , say w0 = f (z0 ), where z0 ∈ Ω. We need to prove that if there exists
> 0 such that for all w with |w − w0 | < , then we must have w belongs to the image of f , i.e. there
exists z1 ∈ Ω such that f (z1 ) = w. The idea is to show that the function g(z) = f (z) − w has a zero at
z1 . Rewrite g as
g(z) = (f (z) − w0 ) + (w0 − w) = F (z) + G(z).
Choose δ > 0 such that the disc |z − z0 | ≤ δ is contained in Ω and f (z) 6= w0 on |z − z0 | = δ. Select
such that |f (z) − w0 | ≥ on the circle |z − z0 | = δ. So if |w − w0 | < , then on the circle |z − z0 | = δ,
F (z) = |f (z) − w0 | ≥ > |w − w0 | = G(z). Thus by Rouché’s theorem F and F + G = g have the same
number of zeros inside the circle |z − z0 | = δ. Since F (z) has one zero at z0 , then there exists z1 such
that |z1 − z0 | < δ and g(z1 ) = f (z1 ) − w = 0, thus w in the image of f . So f is an open mapping.
Definition 2.1.9 (homotopy). Let γ0 and γ1 be two curves in an open set Ω with common end-points.
So if γ0 (t) and γ1 (t) are two parametrizations defined on [a, b], we have
These two curves are said to be homotopic in Ω if for each 0 ≤ s ≤ 1 there exists a curve γs ⊂ Ω,
parametrized by γs (t) defined on [a, b], such that for every s
Definition 2.1.10 (Simply Connected Domain). A region Ω in the complex plane is simply connected
if any two pair of curves in Ω with the same end-points are homotopic.
Theorem 2.1.36. Any holomorphic function in a simply connected domain has a primitive.
Corollary 2.1.37 (General Cauchy’s Theorem). If f is holomorphic in the simply connected region
Ω, then Z
f (z) dz = 0
γ
Theorem 2.1.38. If f is a nowhere vanishing holomorphic function in a simply connected region Ω, then
there exists a holomorphic function g on Ω such that
f (z) = eg(z) .
Theorem 2.1.39. If u is harmonic function on a simply connected domain Ω, then there exists a holo-
morphic function f on Ω such that u =Re(f ). Moreover, Im(f ) is uniquely defined up to an additive real
constant.
∂ 1 ∂
Proof: Let g be the complex function defined by g(z) = 2∂u/∂z = + u(x, y) = ux − iuy .
∂x i ∂y
Let û be the real part of g and v̂ be the imaginary part of g. Then, by harmonicity ûx = uxx = −uyy = v̂y .
Also, ûy = uxy = uyx = −v̂x . Hence, since û and v̂ are continuously differentiable and satisfy the Cauchy-
Riemann equations on Ω, g(z) is holomorphic on Ω. Then by Theorem 2.1.36 g(z) has a primitive F in
Ω such that F 0 = g.
Proof of claim: Let F = Re(F ) + iIm(F ). Since F is a primitive of g, F 0 = (Re(F ))x + i(Im(F ))x =
(Re(F ))x − i(Re(F ))y = g = û + iv̂ = ux − iuy . Hence, (Re(F ))x = ux , so u(x, y) = Re(F )(x, y) + c(y) for
some functions c which only depends on y. Likewise, by comparing the imaginary parts, (Re(F ))y = uy ,
so u(x, y) = Re(F )(x, y) + c(x), where c is a function depends only on x. Hence, c has to be constant,
and therefore Re(F ) − u = c(constant). End of the proof of claim.
MATH 545 and MATH 546 Prelim Problems 68
Now let f = F − (Re(F ) − u) = u + iIm(F ), then Re(f ) = u and f is holomorphic on D, since F and
Re(F ) − u = (constant) are holomorphic.
Now suppose h is another holomorphic function on Ω such that Re(f ) =Re(h) = u, then Re(f −
h) = 0. Therefore, (f − h)0 = 2∂[Re(f − h)]/∂z = 0 and hence f = h + c where c ∈ C, but since
Re(f ) =Re(h)+Re(c) = u =Re(h), Re(c) = 0 and hence c is purely imaginary. Thus Im(f ) is uniquely
defined up to an additive real constant.
Proposition 2.1.40. If f : U → V is holomorphic and injective (one-to-one), then f 0 (z) 6= 0 for all
z ∈ U . In particular, the inverse of f defined on its range is holomorphic, and thus the inverse of a
conformal map is also holomorphic.
ii) If for some z0 6= 0 we have |f (z0 )| = |z0 |, then f is a rotation, i.e. f (z) = eiθ z, θ ∈ R.
Proof:
i) Let f (z) = a0 + a1 z + a2 z 2 + ... be the power series expansion of f in D. Since f (0) = 0, we have
a0 = 0. Consider the function f (z)/z. Then f (z)/z is holomorphic on D \ {0}, and since f (0) = 0,
there exists a natural integer n and a non-vanishing holomorphic function h in a neighborhood U
of z = 0 such that f (z) = z n h(z). Hence f (z)/z = z n−1 h(z) is holomorphic on U , since f and z n−1
holomorphic on D. Therefore it is holomorphic at z = 0 and so it is holomorphic throughout D. (In
fact z = 0 is a removable singularity) If |z| = r < 1, then since |f (z)| ≤ 1, we have
f (z) 1 1
z ≤ |z| = r ,
and by the maximum modulus principle, we can conclude that this is true whenever |z| ≤ r. Let
r → 1 we conclude (i).
ii) f (z)/z attains its maximum at z when |f (z)/z| = 1, since |f (z)| ≤ |z| for all z ∈ D. By assumption
|f (z0 )/z0 | = 1, hence f /z attains a maximum at z0 . But since z0 ∈ D, by maximum modulus
principle converse ??, f /z is constant, i.e. f (z) = cz. But since |z0 | = |f (z0 )| = |c||z0 |, c = eiθ for
some θ ∈ R. Thus f is a rotation.
MATH 545 and MATH 546 Prelim Problems 69
Corollary 2.1.45. Any two proper simply connected open subsets in C are conformally equivalent.
Definition 2.1.13 (Normal Family). Let Ω be an open subset of C. A family F of holomorphic
functions on Ω is said to be normal if every sequence in F has a subsequence that converges uniformly
on every compact subset of Ω (the limit need not be in F).
Definition 2.1.14 (Uniform Boundedness on Compact Sets). The family F is said to be uniformly
bounded on compact subsets of Ω if for each compact set K ⊂ Ω there exists B > 0, such that
Definition 2.1.15 (Equicontinuity). The family F is equicontinuous on a compact set K if for every
> 0 there exists δ > 0 such that whenever z, w ∈ K and |z − w| < δ, then
az + b
f (z) =
cz + d
of one complex variable z; here the coefficients a, b, c, d are complex numbers satisfying ad − bc 6= 0.
The main using of a fractional linear transformation is when we have a region that is bounded with
two curves intersected at one or two points z0 , z1 to send one to ∞, and the other to 0, by
z − z0
f (z) =
z − z1
f sends z0 to 0 and z1 to infinity.
that f is holomorphic on the unit disk |z| < 1. If there is an r ∈ (0, 1) so that
1. Suppose
f 1 ≤ rn for n ∈ N, prove that f is constant on |z| < 1.
n
SOLUTION:
For contradiction suppose f is not constant. Let n → ∞, then |f (0)| ≤ 0 and hence f (0) = 0. Since
f (z) is holomorphic and non-constant, f (z) = z m g(z), where m is the order of the zero, z = 0, and
g(z) is a non-vanishing
holomorphic function in a neighborhood of 0. Re plug f (z) in the inequality
1
above to get g ≤ nm rn . Take the limit as n → ∞ again to get g(0) = 0, but this is a
n
contradiction, g(z) can’t be zero at 0. f is constant and in fact f (z) ≡ 0.
SOLUTION:
f f (z) f (z) f (a)
resa = lim (z − a) = lim = 0
g z→a g(z) z→a g(z) − g(a) g (a)
z−a
since both f and g holomorphic at a.
MATH 545 and MATH 546 Prelim Problems 73
3. Suppose that f is entire and if |f (z) − 1| ≥ 0.01 for all complex z . Prove that f must be
constant.
SOLUTION:
1
Let g(z) = , then g(z) is entire since f (z) − 1 never zero. Also we have
f (z) − 1
1
|g(z)| = ≤ 100.
|f (z) − 1|
By Liouville’s Theorem, g(z) is constant c. Since f is holomorphic and hence doesn’t take infinity,
c can’t be zero and therefore f (z) = c−1 + 1.
4. Suppose that f : C → C is holomorphic, m and n integers, and (5 + |z|m )−1 f (n) (z) is
bounded on C. Prove that f is a polynomial and find it is possible degree.
SOLUTION:
|f (n) (z)| ≤ M (5 + |z|m ) for all z ∈ C and and some positive integer M.
We know that f is entire. Let g(z) = f (n) (z) then and |g(z)| ≤ M (5 + |z|m ). Consider the disc D
centered at 0 and has radius R, then by Cauchy’s inequalities
k!kgk∂D k!M (5 + Rm )
|g(0)(k) | ≤ ≤ .
Rk Rk
Let R → ∞, then the last term goes to 0 when k > m, i.e. g(0)(k) = f (0)(k+n) = 0 when k > m.
Being entire, f has a power series expansion around 0, i.e.
∞
X
f (z) = al z l
l=0
5. Let f be a holomorphic function in the unit disk D and |f (z)| ≤ M for all z ∈ D. Prove
that |f 0 (z)| ≤ M (1 − |z|)−1 for all z ∈ D.
SOLUTION:
Fix z0 ∈ D and consider the disc D centered at z0 and has radius R = δ − |z0 | where |z0 | < δ < 1
(to make sure D inside D.) By Cauchy’s inequalities,
kf k∂D M
|f 0 (z0 )| ≤ ≤ .
R δ − |z0 |
MATH 545 and MATH 546 Prelim Problems 74
Since the last inequality true for all δ such that |z0 | < δ < 1, let δ → 1 to get the desired result.
For contradiction, suppose that f (C) is not dense in C. Then there exist some > 0 and a complex
1
number w ∈ C such that |f (z) − w| > . Let g(z) = , then g is entire and
f (z) − w
1 1
|g(z)| = < .
|f (z) − w|
By Liouville’s Theorem g is constant. Since f is entire, and hence doesn’t take infinity, and w 6= ∞,
g is not zero and therefore f is constant, which is a contradiction.
100
7. Determine the number of zeros (counted with multiplicity) of f (z) = ez − 5z 2 in the
open unit disk D.
SOLUTION:
100
Let g(z) = ez and h(z) = −5z 2 , then on the circle ∂D, we have |h(z)| = | − 5z 2 | = 5 and
|g(z)| = ez = eRe(z ) ≤ e|z| = e. We note that |h(z)| > |g(z)| on the unit circle and hence by
100 100 100
Rouche’s Theorem h and h + g have the same number of zeros in D, but h has only z = 0 as a zero
of order 2. So f has 2 zeros in D.
8. Let g be a non-constant holomorphic function in the unit disk D and suppose that
g(0) = 0 and |g(z)| ≤ 1. Prove that for any integer n ∈ N, g(z) − 2n z n has exactly n zeros
(counted with multiplicity) in D0.5 (0).
SOLUTION:
9. Let f (z) be a holomorphic function on C which takes values in the upper half plane H.
Show that f (z) is constant.
MATH 545 and MATH 546 Prelim Problems 75
SOLUTION:
i−z
Consider the conformal map g : H → D defined by g(z) = . Then h = g ◦ f : C → D is entire,
i+z
because f is, and bounded (|h(z)| < 1). By Liouville’s Theorem h is constant and thus f = g −1 ◦ h
is constant.
10. Let f, g be holomorphic and 1-1 on D with f (0) = g(0) and f (D) ⊂ g(D). Show that
|f 0 (0)| ≤ |g 0 (0)|.
SOLUTION:
Notice that f 0 , g 0 6= 0 and f −1 , g −1 exist and they are defined on f and g ranges, respectively, by
Proposition 2.1.40. Now consider the function F (z) = (g −1 ◦ f )(z). Note that F is from D to D, it
is defined because f (D) in range of g. We have F (0) = g −1 (f (0)) = g −1 (g(0)) = 0. Hence, we can
apply Schwarz Lemma on F to get
0
−1 0
−1 0
0
|F (0)| = (g ◦ f ) (0) = g
(f (0)) f (0) ≤ 1.
But
−1 0 1 1 1
g (f (0)) = = = .
g 0 [g −1 (f (0))] g 0 [g −1 (0)] g 0 (0)
Therefore,
0
−1 0
0
f (0)
g (f (0)) f (0) = ≤ 1.
0
g (0)
11. Find an explicit conformal mapping from the unit disc D onto the plane with the real
intervals (−∞, −1] and [1, ∞) removed.
SOLUTION:
1−z
Consider the function f1 (z) = i which maps conformally the unit disc D onto the upper half
1+z
plane H. Then let f2 (z) = −iz, then f2 takes H onto the right half-plane (the set of points z ∈ C such
that Rez > 0). Next consider f3 (z) = log z which maps the right half-plane onto the horizontal strip
{z ∈ C : −π/2 < Imz < π/2}. Next consider the map f4 (z) = iz which maps the horizontal strip
onto the vertical strip {z ∈ C : −π/2 < Rez < π/2}. Finally, consider the function f5 (z) = sin z
which maps that vertical strip onto the plane with the real intervals (−∞, −1] and [1, ∞) removed.
So, our desired map is
1−z
F (z) = (f5 ◦ f4 ◦ f3 ◦ f2 ◦ f1 )(z) = sin i log .
1+z
MATH 545 and MATH 546 Prelim Problems 76
√ √
12. Find a conformal map of U := {z ∈ C : |z − i| < 2 and |z + i| < 2} onto the right half
plane.
SOLUTION:
-1 0 1 2
−i
z+1
Intersection points are z = ±1. Let f1 (z) = . Then f1 (−1+ ) = 0− , f1 (0) = −1, f1 (1− ) = −∞,
z−1
√ 1 i √ 1 i
f1 (i( 2 − 1)− ) = − √ − √ and f1 (−i( 2 − 1)+ ) = − √ + √ . So f1 maps U onto the sector
2 2 2 2
3π 5π
{z = reiθ : r > 0 and <θ< }
4 4
π/2
0 1
π π
Let f2 (z) = −z. This will map the sector to another sector {z = reiθ : r > 0 and − <θ< }
4 4
MATH 545 and MATH 546 Prelim Problems 77
π/2
0 1
Let f3 (z) = z 2 . This will map the previous sector to the right half plane. So the conformal map is
F = f3 ◦ f2 ◦ f1 .
13. Prove that there is no function f that is analytic on the punctured disk D \ {0} so that
f 0 has a simple pole at 0.
SOLUTION:
Suppose, for contradiction, that there is such a function, call it f . Then there exist Dδ (0) \ {0}, a
deleted neighborhood of 0 in which f 0 (z) = a−1 /z + a0 + a1 z + a2 z 2 ... where a−1 6= 0 and g(z) =
a0 + a1 z + a2 z 2 + ... is a holomorphic function in Dδ (0). Since Dδ (0) is simply connected, g has a
primitive G on it. Hence, f (z) = a−1 log z + G(z) + c, where c is a constant. Since f is holomorphic
on D \ {0}, and hence it is holomorphic on Dδ (0) \ {0}, log z must be holomorphic on Dδ (0) \ {0}.
But this is impossible, there is no branch of log z that is holomorphic in Dδ (0) \ {0}, which is a
contradiction.
14. Suppose f is a non-constant holomorphic function on C such that for all z ∈ C it satisfies
the equation
f (1 − z) + f (z) = 1.
SOLUTION:
We can write the above equation as f (1 − z) = 1 − f (z). For contradiction, suppose f omits at
least one point w ∈ C, i.e. w ∈ / f (C). From the above equation we notice also 1 − w ∈ / f (C).
If w = 1 − w, then w = 1/2 which is not in f (C) as was supposed. But for z = 1/2, we have
f (1 − 21 ) + f ( 12 ) = 2f ( 12 ) = 1, and thus f (1/2) = 1/2, therefore 1/2 ∈ f (C). So, w 6= 1 − w, but that
means f omits two points, namely, w and 1 − w. Then, by Little Picard’s Theorem f is constant
which is the contradiction we wanted. Hence, f omits no points in C.
MATH 545 and MATH 546 Prelim Problems 78
15. Write the explicit formula for the conformal map from U = {z : |z| < 1 and |z − 1| < 1}
to the unit disc D.
SOLUTION:
0 1 2
√
1+i 3
√
√ z− 2 2z − 1 − i 3
The intersection points are z = 1/2 ± i 3/2. Let f1 (z) = √ = √ . Then
1−i 3 2z − 1 + i 3
z− 2
√ √ √ √
1−i 3 1 i 3 −(1 + i 3) 1 i 3
f1 (1) = √ =− − and f1 (0) = √ =− +
1+i 3 2 2 −1 + i 3 2 2
2π 4π
and f1 (1/2) = −1. Hence f1 maps U to the sector {z = reiθ : r > 0 and <θ< }
3 3
2π/3
0 1
Let f2 (z) = z 3/4 . This will map the previous region into the second quadrant
π
{z = reiθ : r > 0 and < θ < π}
2
MATH 545 and MATH 546 Prelim Problems 79
-1 0 1
Let f3 (z) = −iz. This will map the second quadrant into the first quadrant
π
{z = reiθ : r > 0 and 0 < θ < }.
2
0 1
Let f4 (z) = z 2 . This will map the first quadrant into the upper half plane H. Finally, let f5 (z) =
(i − z)/(i + z) which maps H into D. Let the required map be F = f5 ◦ f4 ◦ f3 ◦ f2 ◦ f1 .
16. Prove the converse of Montel’s Theorem, i.e. prove that if F is a normal family of
holomorphic functions on Ω, then F is uniformly bounded on compact subsets of Ω.
SOLUTION:
For contrary suppose that F is normal but not uniformly bounded. Then for a compact subset K
in Ω and for each n ∈ N there is a holomorphic function fn ∈ F and a point zn ∈ K such that
|fn (zn )| > n. Since F is normal, the sequence of holomorphic functions {fn }∞
n=1 has a subsequence
∞
{fnk }k=1 that converges uniformly on K to a holomorphic function f . That is for some N ∈ N, we
have
|fnk (z) − f (z)| < 1, ∀k ≥ N, and z ∈ K.
Since f is holomorphic (hence continuous) on K, f has a maximum, i.e. there is a positive constant
M such that M = maxz∈K |f (z)|. Therefore, for each z ∈ K and k ≥ N , we have |fnk (z)| <
1 + |f (z)| ≤ 1 + M but this is impossible, because |fnl (z)| > nk and by choosing k very large we get
a contradiction.
MATH 545 and MATH 546 Prelim Problems 80
17. Find a conformal map of the unit disk onto itself with f (1/4) = −1/3 and f 0 (1/4) > 0.
SOLUTION:
z − 1/4 z + 1/3
Let h(z) = and g(z) = . Then both h and g are from D to D and h(1/4) =
1 − z/4 1 + z/3
13z − 7
g(−1/3) = 0. Let f (z) = (g −1 ◦ h)(z) = . Then f is from D to D, f (1/4) = g −1 (h(1/4)) =
13 − 7z
g −1 (0) = −1/3, and f 0 (1/4) = 128/135 > 0.
18. Let Ω = {z ∈ D : Im(z) > 0}. Evaluate sup{Ref 0 (i/2) | f : Ω → D}.
SOLUTION:
The idea is to use Schwarz Lemma. That is, we want g : D → Ω so that we have f ◦ g : D → D and
(f ◦ g)(0) = 0. Since f : Ω → D takes its values in D, amongst all these holomorphic functions f ,
1+z √
supf :Ω→D Ref 0 (i/2) = supf :Ω→D |f 0 (i/2)|. The map g1 (z) = i takes D into H, g2 (z) = z maps
1−z
z−1
H into the first quadrant, and g3 (z) = maps the first quadrant to Ω. Let g : D → Ω be
z+1
r
1+z
i −1
1−z
g(z) = (g3 ◦ g2 ◦ g1 )(z) = r .
1+z
i +1
1−z
√ √ √
Note that g(0) = ( i − 1)/( i + 1) = i( 2 − 1). Let f1 be the automorphism of the unit disc taking
g(0) to i/2 , and f2 be the automorphism of the unit disc taking f (i/2) to 0. These functions can
be written out explicitly as
z − g(0) z − 2i
f1 (z) = (T2−1 ◦ T1 )(z) where T1 (z) = and T2 (z) = . Hence,
1 − g(0)z 1 + 2i z
(2 − ig(0))z + (i − 2g(0))
f1 (z) = .
(2 + ig(0)) − (i + 2g(0))z
z − f (i/2)
f2 (z) = .
1 − f (i/2)z
Now let S(z) = (f2 ◦ f ◦ f1 ◦ g)(z) : D → D and S(0) = 0. By Schwarz Lemma |S 0 (0)| ≤ 1. But we
note that
" #
0 0
f f1 g(0) f1 g(0) g 0 (0)
0
0
S (0) = f2 f f1 g(0)
Where
√
0 2 i 3 1
f10 [g(0)] = and f20 f (i/2) =
g (0) = √ , , .
( i + 1)2 4(1 − |g(0)|2 ) 1 − |f (i/2)|2
MATH 545 and MATH 546 Prelim Problems 81
|S 0 (0)| ≤ 1 implies
√ √
2| i + 1|2 1 − |g(0)|2 1 − |f (i/2)|2 2(2 + 2)1 − |g(0)|2 1 − |f (i/2)|2
|f 0 (i/2)| ≤ √ =
3| i| 3
The equality holds when f is a rotation. Moreover, |f 0 (i/2)| is maximized when f such a rotation
with f (i/2) = 0. Hence,
√
√
2(2 + 2)1 − |g(0)|2
0 4 2
sup |f (i/2)| = = .
f :Ω→D 3 3
SOLUTION:
This region has two boundaries: The real axis and the circle |z − 2i | = 12 . The intersection between
them is at z = 0. Let f1 (z) = 1/z. Then f1 (0) = ∞, f1 (±∞) = 0, so it maps the real axis onto the
real axis, and f1 (1/2 + i/2) = 1 − i, f1 (−1/2 + i/2) = −1 − i, and f1 (i) = −i. Thus f1 maps the
circle onto the line Im(z) = −1. Hence f1 maps Ω onto the horizontal strip {z : −1 < Im(z) < 0}.
Let f2 (z) = πz, then f3 maps the previous strip onto {z : −π < Im(z) < 0}.
−iπ
Let f3 (z) = −z, then f3 maps the previous strip to the strip {z : 0 < Im(z) < π}.
MATH 545 and MATH 546 Prelim Problems 82
iπ
Now let f4 (z) = ez , then ez = ex+iy = ex ei y, hence 0 < ex = r < ∞ and 0 < y = θ < π. So f4 maps
the previous strip onto H. Let F = f4 ◦ f3 ◦ f2 ◦ f1 , then F : Ω → H is a conformal map.
20. Find a conformal map from the slit unit disc D \ (−1, 0] onto the unit disk such that
√
z0 = 2 − 1 is mapped to 0.
SOLUTION:
The slit disc is {z = reiθ : 0 < r < 1 and − π < θ < π}. Let f1 be a branch of the square root
√ iθ
function, f1 (z) = z. Then f1 takes the p√slit disc onto the right half disc {z = re : 0 < r <
1 and − π/2 < θ < π/2} and z0 to z1 = p 2 − 1. Now let f2 (z) = iz, this takes the right half disc
√
onto the upper half disc and z1 to z2 = i 2 − 1. Let f3 (z) = (1 + z)/(1 − z), this takes the upper
half disc onto the first quadrant and z2 to
p√
2−1 √
q √
1+i
z3 = p√ = 2 − 1 + i 2( 2 − 1).
1−i 2−1
√ p √
Let f4 (z) = z 2 , this takes the first quadrant onto H and z3 to z4 = 5 − 4 2 + 2i 10 2 − 14. Let
f5 (z) = (i − z)/(i + z), then f5 takes H onto D and z4 to z5 = (i − z4 )/(i + z4 ). Let f6 = ψz5 (z) =
(z5 − z)/(1 − z 5 z). Then f6 maps D onto D and z5 to z6 = 0. Now let F = f6 ◦ f5 ◦ f4 ◦ f3 ◦ f2 ◦ f1 ,
√
then F : D → D and F ( 2 − 1) = 0.
21. Let G be a simply connected region which is not the whole plane, and suppose that
z̄ ∈ G whenever z ∈ G. Let a ∈ G ∩ R and suppose that f : G → D is a conformal map
with f (a) = 0, f 0 (a) > 0. Let G+ = {z ∈ G : Im(z) > 0}. Show that f (G+ ) must lie entirely
above or below the real axis.
SOLUTION:
Notice that f (G+ ) does not lie entirely above or bellow the real axis if and only if there exists z ∈ G+
such that f (z) ∈ D ∩ R. Thus if we prove that f (G ∩ R) = D ∩ R, then f being 1-1 implies that
f (G+ ) is either above or below the real axis and hence we are done. What we need to prove is: If
z ∈ G ∩ R, then f (z) = f (z) = f (z). If we prove that f (z) = f (z) for all z ∈ G, then we are done.
Let g(z) = f (z), then:
MATH 545 and MATH 546 Prelim Problems 83
and hence ∞
X
g(z) = f (z) = ak (z − z0 )k
k=0
Claim. g(z) = f (z). Proof: Let a ∈ G ∩ R be as in the question with f (a) = 0 and f 0 (a) > 0. If
we prove that g(a) = 0 and g 0 (a) > 0, then by Riemann mapping theorem Theorem 2.1.44 f = g,
by uniqueness.
Proving g(a) = 0: g(a) = f (a) = f (a) = 0, since a ∈ G ∩ R and f (a) = 0.
Proving g 0 (a) > 0: g is holomorphic on G and has a power series expansion, hence
h i0 ∞
X ∞
X
0 k−1
g (z) = f (z) = k ak (z − z0 ) = k ak (z − z 0 )k−1 = f 0 (z).
k=1 k=1
SOLUTION:
Let f1 (z) = i(1 − z)(1 + z), this maps D onto H. Let f2 (z) = −iz, this maps H onto the right half-
plane {z = reiθ : r > 0 and − π/2 < θ < π/2}. Let f3 (z) = z 2 , this maps the right half-plane onto
√
C \ (−∞, 0]. Let f4 (z) = z + 9, then this maps C \ (−∞, 0] onto C \ (−∞, 9]. Let f5 (z) = z, then
this maps C\(−∞, 9] onto the right half-plane with the slit (0, 3] removed. Let f6 (z) = iz, this maps
the right half-plane with the slit (0, 3] removed to H \ {0 < Im(z) ≤ 3}. Let f7 (z) = (i − z)/(i + z),
this maps H \ {0 < Im(z) ≤ 3} to D \ [−1/2, 1). Let F = f7 ◦ f6 ◦ f5 ◦ f4 ◦ f3 ◦ f2 ◦ f1 , then F is the
required conformal mapping.
23. Find a conformal mapping from Ω = {z : |z| < 1 and Re(z) > 0} \ [0, 1/2] to H.
MATH 545 and MATH 546 Prelim Problems 84
SOLUTION:
24. Let S denote the class of functions f which are 1-1 and analytic on the open disc D
satisfying f (0) = 0 and f 0 (0) = 1. Prove that if {fn } ⊂ S and fn converges to f uniformly
on compact sets of D, then f ∈ S.
SOLUTION:
Suppose {fn } ⊂ S and fn converges to f uniformly on compact sets of D, then by Theorem 2.1.16,
f is holomorphic on D. Let K ∈ D be a compact subset contains 0, then fn → f uniformly on K
implies fn → f point-wisely on K and hence 0 = fn (0) → f (0) so f (0) = 0. Also, fn and f being
holomorphic implies fn0 and f 0 are holomorphic, and by Theorem 2.1.17 fn0 → f 0 uniformly on every
compact subset of D and hence point-wisely on every compact subset of D. Let K ∈ D such that K
contains 0, then 1 = fn0 (0) → f 0 (0) and thus f 0 (0) = 1. It remains to prove that f is 1-1. Suppose
f is not 1-1 on D and hence not 1-1 on some compact set K, then there exists z0 , z1 ∈ K such that
z0 6= z1 , fn (z0 ) 6= fn (z1 ) for all n and f (z0 ) = f (z1 ). Since fn (z0 ) 6= fn (z1 ) for all n, then there exists
> 0 and N ∈ N such that for all n > N
But fn → f point-wisely on K implies that there exists N 0 ∈ N such that for all n > N 0
|fn (z0 ) − f (z0 )| < and |fn (z1 ) − f (z1 )| < .
2 2
Let N 00 = max{N, N 0 }, then for all n > N 00 ,
|fn (z0 ) − fn (z1 )| ≤ |fn (z0 ) − f (z0 )| + |fn (z1 ) − f (z1 )| <
25. Let Ω ⊂ C be open and a, b ∈ R, a < b and F : Ω × [a, b] → C be continuous such that for
Z b
every t ∈ [a, b], the function ft = F (z, t) is analytic on Ω. Show that g(z) = F (z, t) dt
a
defines an analytic function on Ω.
MATH 545 and MATH 546 Prelim Problems 85
SOLUTION:
It is enough to prove it for [a, b] = [0, 1], I did that in the proof of Theorem 2.1.18. Or if you want
to do it for [a, b], then it is enough to do it for a disc D whose closure in Ω. By changing variables
t = x(b − a) + a, we get
Z b Z 1
g(z) = F (z, t) dt = (b − a) F (z, x(b − a) + a) dx,
a 0
Then gn is holomorphic on D for all n. Since F is continuous in t on [a, b], it is uniformly continuous
on [a, b], i.e. for all > 0, there is δ > 0 such that for all t1 , t2 ∈ [a, b]
|t1 − t2 | < δ =⇒ |F (z, t1 ) − F (z, t2 )| < .
b−a
Claim. {gn } converges uniformly to g on D.
Proof: Choose N ∈ N such that N > (b − a)/δ, then for all n > N
b − a X n Z 1
k
|gn − g| = F z, (b − a) + a − (b − a) F (z, x(b − a) + a) dx
n n 0
k=1
b − a X n n Z k/n
k X
= F z, (b − a) + a − (b − a) F (z, x(b − a) + a) dx
n n k=1 (k−1)/n
k=1
n Z k/n
X k
= (b − a) [F z, (b − a) + a − F (z, x(b − a) + a)] dx
k=1 (k−1)/n
n
n Z k/n
X k
≤ (b − a) F z, (b − a) + a − F (z, x(b − a) + a) dx
k=1 (k−1)/n
n
Xn Z k/n
< dx
k=1 (k−1)/n
n
X 1
= = .
k=1
n
k−1 k
Where in line (5) we used the fact that F is uniformly continuous on [a, b], since x ∈ , ,
n n
(k − 1)(b − a) k(b − a) k(b − a) (k − 1)(b − a)
hence x(b − a) + a ∈ + a, + a and
+a− − a =
n n n n
(b − a)/n < δ.
Therefore, by Theorem 2.1.16 g is holomorphic on D. Since that true for all D in Ω, then g is
holomorphic on Ω.
MATH 545 and MATH 546 Prelim Problems 86
( ∞ (n) 2 )
X |f (0)|
26. Let F = f holomorphic on D : ≤ 1 and f (1/2) = 0 . Show that there is
n!
n=0
a g ∈ F such that g(0) = sup{Re(f (0)) : f ∈ F}.
SOLUTION:
We first prove that F is a normal family. To do that we prove that F is uniformly bounded on
compact subsets of D, and hence by Montel’s theorem Theorem 2.1.46 F is a normal family. For
contrary suppose that F is not uniformly bounded on compact subsets of D. Then for some compact
disc K with radius d in D containing the origin and for each k ∈ N there is a holomorphic function
fk ∈ F and a point zk ∈ K such that |fk (zk )| > k. But for all k > 1/(1 − d2 ) we have
1
< k < |fk (zk )| ≤ |fk (zk )|2
1 − d2
2
∞ (n)
X f (0)
= k
zkn (since fk is holomorphic on D)
n=0 n!
∞ (n) 2 ∞
X |f (0)| X
≤ |zk |2n (By Cauchy-Shwarz inequality)
n=0
n! n=0
∞
X
≤ |zk |2n
n=0
1
=
1 − |zk |2
1
≤
1 − d2
which is a contradiction. Hence F is a normal.
Let E = sup{Re(f (0)) : f ∈ F} and let {fn } ⊂ F such that Re(fn ) → E. Then since F is normal,
{fn } has a subsequence {fnk } such that {Re(fnk )} converges uniformly on every compact subset
of D, say to g (may not be in F) but it is holomorphic by Theorem 2.1.16. Hence Re(fnk ) → g
point-wisely, so Re(fnk (0)) → g(0). But since Re(fn ) → E and {fnk } a subsequence of {fn } such
that Re(fnk (0)) → g(0), g(0) = E. It remains to prove that g ∈ F. First, notice since {fnk } in F,
fn (1/2) = 0 for all n. Then since Re(fnk ) → g point-wisely, we have 0 = limk→∞ Re(fnk (0)) = g(1/2).
Thus g(1/2) = 0. Second, notice that since {fnk } a subsequence such that {Re(fnk )} converges
(n)
uniformly on every compact subset of D to g, by Theorem 2.1.17 {Re(fnk )} converges uniformly on
every compact subset of D to g (n) and hence point-wisely, so
(n) 2 (n) 2
∞
X |g (n) (0)|
2 X∞ Re fnk (0) ∞
X Re fnk (0)
= lim =≤ lim
n=0
n! n=0
k→∞ n! k→∞
n=0
n!
∞ (n) 2
X |f (0)|
≤ lim ≤ lim 1 = 1.
k→∞
n=0
n! k→∞
MATH 545 and MATH 546 Prelim Problems 87
Thus g ∈ F.
27. Show that F ⊂ Hol(D) is normal if and only if there is a sequence {Mn } of positive
√
constants such that lim sup n Mn ≤ 1 and if f (z) = ∞ n
P
n=0 an z is in F, then |an | ≤ Mn for
all n.
SOLUTION:
√
Now suppose lim sup n Mn ≤ 1 and let g(z) = ∞ n
P
n=0 Mn |z| . Then the radius of convergence r of
√
g(z) is given by 1/r = lim sup n Mn ≤ 1, hence r ≥ 1 and thus g(z) converges on every compact disc
D in D, say to LD . Therefore, fol all f ∈ F, |f (z)| ≤ ∞
P n
P∞ n
n=0 |an ||z| ≤ n=0 Mn |z| = LD . Thus F
is uniformly bounded on compact subsets of D, and hence by Montel’s theorem Theorem 2.1.46 F
is normal.
28. Suppose f is analytic on D and satisfies |f (z)| < 1 for |z| = 1. Find the number of solu-
tions, counting multiplicities of the equation f (z) = z n where n is an integer at least 1.
SOLUTION:
Let g(z) = f (z) − z n , then on |z| = 1, | − z n | = |z|n = 1 > |f (z)| thus by Rouche’s theorem
Theorem 2.1.32 −z n and f − z n have the same number of zeros. But −z n has n zeros counted with
multiplicities.
SOLUTION:
V is open, since if a ∈ V , then a ∈ Ω and u vanishes in some neighborhood of a, call it Dδ (a), where
δ some positive real number. Let D (a) be another neighborhood of a such that D (a)
subsetneqDδ (a), then for any z ∈ D (a), there is a neighborhood Dδ0 (z) ⊂ Dδ (a). Since u vanishes
on Dδ (a), u vanishes on Dδ0 (z), thus z ∈ V .
V also closed. Let {zn } ⊂ V such that zn → z ∈ Ω, we need to prove that z ∈ V . Choose > 0
such that D (z) ⊂ Ω. Then there exists a holomorphic function f on D (z) such that u =Re(f ) on
D (z). But since zn gets closer to z as n → ∞, then there is an n ∈ N such that zn ∈ D (z). But
since zn ∈ V , then u, and hence f , vanish in a neighborhood of zn . But since f is holomorphic on
D (z) and it vanishes on a subset of it, then by Theorem 2.1.14 f , and hence u, vanish on D (z),
but that means z ∈ V .
Let U be the complement of V in Ω, then both V and U are disjoint open sets and Ω = V ∪ U . But
since Ω is connected, either U = ∅ and V = Ω or U = Ω and V = ∅.
30. Let f be analytic in a neighborhood of the closed unit disc D with |f (z)| ≤ 1 for all z
with |z| = 1. Show that if there are z1 , z2 ∈ D with f (z1 ) = z1 and f (z2 ) = z2 , then either
z1 = z2 or f (z) = z for all z ∈ D.
SOLUTION:
If z1 = z2 , then there is nothing to prove. Let z1 6= z2 , then f (z1 ) = z1 6= z2 = f (z2 ) and hence f is
not constant. Since f is analytic and hence continuous on D, f attains a maximum in D, but since
f is non constant in D, then by maximum modulus principle Theorem 2.1.34 f does not attain a
maximum in D, thus it attains its maximum on ∂D, i.e. when |z| = 1, but since |f (z)| ≤ 1 when
|z| = 1, by maximum modulus principle
Hence by Schwarz lemma since z3 6= 0, ψz1 ◦ f ◦ ψz1 = eiθ z for some θ ∈ R. But by (∗) θ = 0 or 2π
and hence ψz1 ◦ f ◦ ψz1 = z, which means f (ψz1 (z)) = ψz1 (z) or f (w) = w for all w ∈ D.
MATH 545 and MATH 546 Prelim Problems 89
SOLUTION:
But since 0 ∈
/ O and O is symmetric with respect to |z| = 1, if z ∈ O, then 1/z ∈ O. Hence
∞
X
f (1/z) = an (1/z − z0 )n .
n=0
Hence ∞
X
f (1/z) = an (1/z − z0 )n , (∗)
n=0
but since z0 ∈ ∂D, (∗) defines a convergent series in O, thus f (1/z) is holomorphic in O. Now let
O0 ⊂ O such that f (1/z) 6= 0 for all z ∈ O0 (a such neighborhood exists by continuity of f around
|z| = 1, since |f (z)| = 1 there), then 1/f (1/z) is holomorphic on O0 and hence g(z) is holomorphic
on O0 , then by Theorem 2.1.14 g is identically zero in O0 and hence f (z) = 1/f (1/z) in O0 .