Ph.D.

Qualifying Exam, Real Analysis

Spring 2009, part I
Do all five problems. Write your solution for each problem in a separate blue book.

1 Fix a real number p ∈ (1, ∞), and define a linear operator T : Lp (0, ∞) → Lp (0, ∞) by
1 x
R
(T f )(x) = x 0 f (y) dy.
a. Show that
p
kT f kLp (0,∞) ≤ kf kLp (0,∞) .
p−1
− p−1  p−1
for all f ∈ Lp (0, ∞). (Hint: write |f (y)| = y p2 y p2 |f (y)| and apply Hölder’s inequality.)

p
b. Show that the operator norm of T is equal to p−1 .
2 Suppose that X, Y are Banach spaces, and let Ts denote the norm topology on X, Us the norm
topology on Y . Let Tw denote the weak topology on X, and Uw denote the weak topology on Y .
a. Show that (X, Tw ) is (T3), i.e. if x ∈ X then {x} is Tw -closed, and if x ∈ X, and C ⊂ X is
Tw -closed, x ∈ / C, then there exist Tw -open sets U, V such that U ∩ V = ∅, x ∈ U , C ⊂ V .
b. Show that a linear map T : X → Y is continuous as a map from (X, Ts ) to (Y, Us ) if and only
if it is continuous as a map from (X, Tw ) to (Y, Uw ).
3 Suppose that {gn } is a sequence of positive continuous Rfunctions on [0, 1], µ is a positive Borel
1
measure on [0, 1] and that (i) limn→∞ gn (x) = 0 a.e., (ii) 0 gn dx = 1 for all n and
Z 1 Z 1
(iii) lim f gn dx = f dµ
n→∞ 0 0
for every continuous function f ∈ C[0, 1]. Does it follow that µ is mutually singular with respect to
the Lebesgue measure? Prove this or give a counterexample (and prove that it is a counterexample).
4 Let s(Z) denote the vector space of rapidly decreasing bi-infinite complex-valued sequences, i.e.
sequences a = {an }n∈Z : Z → C such that ρk ({an }) = sup |nk an | < ∞ for all k ≥ 0 integers,
and let T be the weakest topology on s(Z) in which the functions
fk,b (a) = ρk (a − b) : s(Z) → [0, ∞),
k ≥ 0, b ∈ s(Z), are continuous.
a. Show that T is metrizable, and give an explicit metric giving rise to the topology T .
b. Show that there exists no norm k.k on s(Z) such that T is the topology given by the norm k.k.
5 A Banach space B is uniformly convex if for every ǫ ∈ (0, 1) there exists η < 1 such that if
x, y ∈ B, kxk = kyk = 1 and kx − yk > 2ǫ then k 12 (x + y)k < η.
a. Show that every Hilbert space is uniformly convex, and one may take η = (1 − ǫ2 )1/2 .
For 1 < p < ∞, show that Lp is uniformly convex. (Hint: For (b), write |f + g|p dµ =
R
b.
f g
|φ + ψ|p dν where φ = (|f |p +|g| p p
R
p )1/p , ψ = (|f |p +|g|p )1/p and ν = (|f | + |g| ) µ. This reduces the

general case to that where the functions are bounded by 1 and the measure space has total mass 2.
Use the convexity of the function xp in [−1, 1].)
 Ph.D. Qualifying Exam, Real Analysis
Spring 2009, part II

Do all five problems. Write your solution for each problem in a separate blue book.

1 Two quick problems.

a. Suppose that f ∈ C(T), T = R/(2πZ). Let Mf be the multiplication operator on Lp (T), p ∈
[1, ∞), (Mf φ)(x) = f (x)φ(x) for φ ∈ Lp (T). Show that the spectrum of Mf is σ(Mf ) = f (T).
b. Suppose that X is a Banach space, D is a dense subspace, An ∈ L(X), n ∈ N, A ∈ L(X).
Show that An → A in the strong operator topology if and only if {kAn kL(X) : n ∈ N} is bounded
and An x → Ax (in the topology of X) for all x ∈ D.

2 a. Suppose that Y is a normed complex vector space with norm k.k, and λ : Y → C is linear but
is not continuous. Show that N = λ−1 ({0}) is dense in Y .
b. Show that L2 ([0, 1]) has an orthonormal basis {φn : n ∈ N} such that each φn ∈ C 1 ([0, 1])
and φ′n (1/4) = φn (1/2) for each n.

3 Suppose that K ∈ Lp ([0, 1] × [0, 1]), 1 < p < ∞. Let q be the dual exponent, p−1 + q −1 = 1.
a. For f ∈ Lq ([0, 1]), let (Af )(x) = K(x, y)f (y)dy. Show that (Af )(x) indeed exists for
R

almost every x and A ∈ L(Lq ([0, 1]), Lp ([0, 1])).

b. Suppose that for every f ∈ Lq ([0, 1]), (Af )(x) = 0 for almost every x. Show that K = 0 a.e.

4 Let S ′ (Rn ) denote the set of tempered distributions, and H s (Rn ) be the subspace of S ′ (Rn )
consisting of distributions f whose Fourier transforms are functions fˆ such that kf k2H s (Rn ) =
(1 + |ξ|2 )s |fˆ(ξ)|2 dξ < +∞.
R

a. Show that if 2 ≤ p ≤ ∞ then for s > n2 − np every element of H s (Rn ) is in Lp (Rn ), and there
exists C = C(n, p, s) such that kf kLp (Rn ) ≤ Ckf kH s (Rn ) for every f ∈ H s (Rn ). (Hint: consider
p = 2, ∞ first.)
b. For p = ∞, show that this is sharp, namely if s = n2 , there is no constant C such that
kf kLp (Rn ) ≤ Ckf kH s (Rn ) for every f ∈ H s (Rn ).

5 Suppose that X is a Banach space over C, M and N are closed subspaces of X, and let M + N =
{x ∈ X : ∃m ∈ M, n ∈ N, x = m + n}.
a. Show that M + N is closed if and only if there exists C > 0 such that for all x ∈ M + N
there exist m ∈ M , n ∈ N such that x = m + n and kmk + knk ≤ Ckxk.
b. Suppose that ℓM : M → C and ℓN : N → C are continuous linear functionals and ℓM |M ∩N =
ℓN |M ∩N . Show that if M +N is closed, then there exists ℓ ∈ X ∗ such that ℓ|M = ℓM and ℓ|N = ℓN .
c. Give an example of a Banach space X and closed subspaces M, N such that M ∩ N = {0}
but M + N is not closed.