Exercises in Functional Analysis

Peter Sjögren∗
November 2, 2010

1. Consider the space X = C 1 [0, 1] of continuously differentiable functions

on [0,1] with ome-sided derivatives at the endpoints and equip X with the
norm
kf k = sup{|f (x)| : x ∈ [0, 1]}.
Show that X is not a Banach space. (See the exercise 5.1.9 in Folland.)

2. Let X be a normed vector space. Prove that for all x, y ∈ X, | kxk −

kyk | ≤ kx − yk. Conclude that the norm is continuous, in the sense that
kxn k → kxk whenever xn → x. (See exercise 5.1.1 in Folland.)

3. Prove that any ball B (open or closed) in a normed vector space X is convex,
i.e. that λx + (1 − λ)y ∈ B whenever x, y ∈ B and 0 ≤ λ ≤ 1.

4. Consider the real two-dimensional `p , i.e. R2 with the norm k(x1 , x2 )k =

(|x1 |p + |x2 |p )1/p . Sketch the unit ball for several values of p, including
p = 1, 3/2, 2, 3, ∞. For which p’s is the closed unit ball strictly convex?
(I.e. for which p’s does kxk, kyk ≤ 1, x 6= y and λ ∈ (0, 1) imply that
kλx + (1 − λ)yk < 1?)

5. Consider the spaces `1 , `2 and `∞ . For which of these normed vector spaces
is it true that kx + yk = kxk + kyk implies that x and y are parallel (i.e. one
is a scalar multiple of the other)?

6. Show that all norms k · k on a finite dimensional vector space X are equiv-
alent, usingP
the following Pargument. Let e1 , . . . , en be a basis and define
n n
the norm k 1 ai ei k1 = 1 |ai |. First show that there is a C such that
∗
Some translation and modifications by Johan Jonasson

1
 kxk ≤ Ckxk1 for all x. From this, conclude that k · k is continuous with
respect to k · k1 and hence that k · k is bounded away from 0 on the compact
set {x : kxk1 = 1}. Finally conclude from this that there is a constant C 0
such that kxk1 ≤ C 0 kxk for all x. (See exercise 5.6 in Folland.)

7. Let 1 ≤ p < q < r ≤ ∞. Prove the inclusions

Lp ∩ Lr ⊆ Lq ⊆ Lp + Lr .

Conclude from this that the set of p’s such that f ∈ Lp is an interval. (See
exercises 6.1.3 and 6.1.4 in Folland.)

8. Let p ∈ (0, 1). Show, using characteristic funtions of disjoint sets, that
k · kp is not a norm on Lp ([0, 1], B[0, 1], Leb). Show also that the “unit ball”,
{f : kf kp ≤ 1} is not convex.

9. Let {wk } be a sequence of positive numbers

P and let mw be the measure on
p
the natural numbers given by mw (A) = k∈A wk . Show that L (mw ) is
isometrically isomorphic to `p for any 1 ≤ p ≤ ∞.

10. (Packing of balls) Let 1 ≤ p ≤ ∞. Find, inside the unit ball of `p , infinitely
many pairwise disjoint balls of the same positive radius. Conclude from this
that the closed unit ball in `p cannot be compact.

11. Let M be a complete metric space and K a subset of M . Show that the
following are equivalent.

(i) K is compact, i.e. every open cover of K can be reduced to a finite

subcover.
(ii) K is sequentially compact, i.e. every sequence in K has a convergent
subsequence.
(iii) K is closed and totally bounded, i.e. for every r > 0, K can be covered
by finitely many balls of radius r.

12. Show that every finite-dimensional subspace, V , of a normed vector space,

X, is closed. Hint: Show that the limit of a convergent sequence of elements
of V must also be an element of V . Use the result of exercise 6.

13. (More packing of balls) Let X be an infinite-dimensional normed vector

space.

2
 (a) Show that for sufficiently small r > 0, one can find infinitely many
pairwise disjoint balls of radius r inside the unit ball.
(b) Make the conclusion that a ball in X cannot be compact.
14. (Extension by continuity) Let M be a dense subspace of the normed space
X. Here dense means that M = X, and the restriction to M of the norm of
X makes M into a normed space. Let Y be a Banach space. Show that any
bounded linear operator T : M → Y has a unique extension to a bounded
linear operator X → Y .
15. Let E be the subset of `p , 1 ≤ p ≤ ∞, consisting of all sequences with only
finitely many nonzero entries. Determine the closure E.
16. Prove that the normed spaces Lp (R) and Lp (0, 1) are isometrically isomor-
phic. Here 1 ≤ p ≤ ∞.
17. Consider convergence in the Lp norm with 1 ≤ p < ∞ and a.e. conver-
gence, for some measure of your choice. Prove that neither implies the
other.
18. With 1 ≤ p < ∞, assume that the function sequence (fj ) converges to f
in Lp for some measure. Prove that there is a subsequence which converges
a.e. to f .
19. Prove that `∞ and L∞ [0, 1] are not separable. Hint: Find an uncountable set
of points, all at mutual distances at least r, for some r > 0. In the first case,
cf. Problem 10.
20. Prove that translation is continuous in Lp (R) for 1 ≤ p < ∞: If ft (x) =
f (x − t) with f ∈ Lp , then ft → ft0 in Lp as t → t0 . Prove also that this is
false for p = ∞.
21. Assume that the function g is measurable and a.e. finite, with respect to
Lebesgue measure, but that g ∈/ Lq . Here 1 < q < ∞, and we let p denote
the dual exponent. Find a function f ∈ Lp for which f g ∈
/ L1 .
22. For which sequences (dk ) of positive numbers is
{(xk ) ∈ `p : |xk | ≤ dk , k = 1, 2, ...}
a compact subset of `p ? Here 1 ≤ p ≤ ∞. As for compactness, see problem
11. Use that compactness is equivalent to sequential compactness.

3
23. The map (s, 0) 7→ s is a linear functional on the subspace R × {0} of R2 .
Determine all its linear extensions to R2 with the same norm, if R2 is given
(a) the `1 norm or (b) the `∞ norm.
24. (Separable case of Hahn-Banach) Let X be a separable normed space. As-
sume that M is a subspace of X and that f is a linear functional defined on
M and bounded with respect to the norm of X. Prove that f has an exten-
sion to a bounded linear functional on X, without using anything equivalent
to the axiom of choice.
25. A linear subspace M of a vector space X is called a hyperplane if for some
(and hence all) x ∈ X \ M the subspace M and the vector x together span
X. One says that the codimension of M is 1. Assume in addition that X is
a normed space.
(a) Prove that a hyperplane is either closed or dense in X.
(b) Prove that any hyperplane is given by M = f −1 (0) (the kernel or zero
space of f ) for some linear functional on X. Also show that M is closed
precisely when f is continuous.
(c) If two linear functionals have the same kernel, prove that they are pro-
portional.
26. Let X be a Banach space of infinite dimension. Prove that X has no count-
able “basis”, if we let this mean that any vector can be written as a finite
linear combination of vectors from the “basis”. Hint: If there was a count-
able base, then X could be written as a union of nowhere dense sets.
27. Baire category is not related to Lebesgue measure. Find a meager subset of
R whose complement has measure 0.
28. Consider the boundary of an open subset U of the real line.
(a) Is ∂U always nowhere dense?
(b) Is ∂U always of Lebesgue measure 0?
29. Prove that the principle of uniform boundedness does not hold in all normed
linear spaces.
30. Let Q = (qj )∞ 1 be an enumeration of the rational numbers, and let n,j , n, j =
1, 2, ... be arbitrary positive numbers. Then the set
∞ [
\ ∞
(qj − n,j , qj + n,j )
n=1 j=1

4
 obviously contains Q. Prove that this inclusion is always strict.

31. Show that a linear isometry T from a Hilbert space into another preserves
the inner product, in the sense that hT x, T yi = hx, yi.

32. Consider the function sequence (einx )n in Lp (−π, π). For which p, 1 ≤
p ≤ ∞, is it weakly or weakly* convergent?

33. Let (en ) denote the “standard basis” in `p . For which p with 1 ≤ p ≤ ∞
does en converge weakly or weakly* as n → ∞, and what is the limit? (Be
extra careful with the cases p = 1 and p = ∞.)

34. Assume that fn → f weakly in a normed space, or weakly∗ in its dual.

Prove that
kf k ≤ lim kfn k.
Also give examples where this inequality is strict.

35. In L∞ translation is weakly∗ continuous: With the notation from problem

20, show that ft → ft0 weakly∗ in L∞ = (L1 )∗ , as t → t0 . Hint: Use the
result of exercise 20 and Hölder’s inequality.

36. We know that `1 = c∗0 and (`1 )∗ = `∞ . Give an example of a sequence

of vectors in `1 that converges weakly∗ but not weakly, which means that
hf, xn i → hf, xi for all f ∈ c0 but not for all f ∈ `∞ .

37. Let X be a Banach space and X ∗ its dual. Assume that (xj ) is a sequence
in X that converges to x ∈ X, either in norm or weakly. Similarly, (ϕj ) is a
sequence in X ∗ with a limit ϕ, and the convergence is assumed to be either
in norm or weak∗ . This gives four different cases of convergence assump-
tions. The question is: does it follow that hϕj , xj i → hϕ, xi ? Answer this
question in each of the four cases, by means of a proof or a counterexample.
Hint: Use that a sequence that converges weakly or weakly∗ is bounded in
norm. (This is exercise 5.47(b) in Folland. Try that one too.)

38. Let µ be a Radon measure in a locally compact Hausdorff space X. (This

means µ ≥ 0.) Take 1 ≤ p < ∞. Prove that the space of simple functions
is dense in Lp (µ). By a simple function we mean a finite linear combination
of characteristic functions of Borel sets with finite measure. Also prove that
Cc (X) is dense in Lp (µ).

5
39. Let X be a locally compact Hausdorff space. Assume that the signed Radon
measures µn in X converge vaguely, i.e. weakly*, in M (X) = C0∗ , to µ.
If in addition it is assumed that X is compact and that the µn are positive,
prove that kµn k → kµk. Also prove that this need not hold if either of the
two additional assumptions is suppressed.
40. In the locally compact Hausdorff space X, assume that the (positive) finite
Radon measures µn ∈ M (X) = C0∗ converge vaguely to µ. Give an exam-
ple to show that µn (A) need not tend to µ(A) for a Borel set A ⊂ X. Prove
that, however, µ(U ) ≤ lim µn (U ) for each open set U .
41. In the unit disc, prove that the positive harmonic functions are precisely the
Poisson integrals of the (positive) Radon measures on the unit circle. Hint:
Use a modification of the example on Alaoglu’s Theorem from the lectures.
One can use the mean value property of harmonic functions to verify the
conditions of Alaoglu. It says that the mean value of the function on any
circle coincides with the value at the centre of the circle. You need not
verify that a Poisson integral is harmonic.
42. Consider the space weak Lp for some measure and some p > 0. Show that
the quantity [f ]p has all the properties of a norm, except that the triangle
inequality is violated. Also prove that instead of the triangle inequality
one has [f + g]p ≤ C([f ]p + [g]p ) for some constant C; one speaks of a
quasinorm. Cf. here Folland’s exercise 6.4.35 p 199.
43. With 1 ≤ p, r, s ≤ ∞, we want to see that the equation 1/s = 1/p +
1/r−1 is necessary for Young’s inequality. Consider the line with Lebesgue
measure, and assume that the equation does not hold.
(a) Show that there is no norm inequality
kf ∗ gks ≤ Ckf kp kgkr
for any C.
(b) Find a counterexample to the inclusion Lp ∗ Lr ⊂ Ls . Hint: One can use
linear combinations of examples from (a).
44. Let X be a locally compact Hausdorff space, or Rn if you prefer. The
space M (X) of finite signed or complex measures in X is the dual both of
C0 (X) and Cc (X). Weak* convergence in M (X) can thus be defined in
two different ways. Are they equivalent?