2024 SPRING 27200 HOMEWORK 2, DUE 04/08 23:59

1. Exercises from Textbook, 2 Pt Total

Chapter 5: Exercise 2 on page 38, Exercise 3 on page 39, Exercise 5 and 6 on page 44.

2. Basic Definitions, 3 Pt Total

1. This is an exercise about quotient spaces in order to give you more intuition towards the definition of it.
Consider X = R3 , and let Y be the two dimensional subspace x1 + x2 + x3 = 0. Recall that the quotient
space X/Y is the set of all translations x + Y , where x runs through R3 .
(1) Determine whether the following pair of points are in the same translation or not: (1,1,1) and (3,0,0);
(1, 0, −1) and (2,0,0); (2, 1, −1) and (0,1,1).
(2) How many parameters do you need to parameterize the quotient space X/Y ? Explain the reason.
(3) There is some general pattern hidden in (2). In general, if X = Rn and Y is an m-dimensional
subspace, then what is the dimension of the quotient X/Y ?

2. Let Cc be the vector space of all real sequences in which there are only finitely many non-zero terms.
Prove that for 1 ≤ p < ∞, the closure of Cc under the ℓp norm
∞
!1/p
X
p
∥(an )∥ℓp := |an |
n=1
p p
is exactly the space ℓ . Use this to prove that ℓ is separable.
What is the closure of Cc under the norm

∥(an )∥ℓ∞ := sup |an |,

n
∞
and what is the relation of this closure with ℓ , the space all bounded sequences?

3. Let 1 < p ≤ ∞. Let M ⊂ Lp [0, 1] be a bounded subset. Show that the set of functions of the form
Z x
f (t)dt, f ∈ M
0
0
has compact closure in C [0, 1].
Show by counterexample that this cannot be true for p = 1. Hint. Approximate a step function by broken
lines.

4. Fix constants A > 0, s > 0. Let M ⊂ ℓ2 be a bounded subset, such that for every element (an ) ∈ M,
there holds
X∞
n2s |an |2 ≤ A.
n=1
Prove that M has compact closure in ℓ2 . This is an example of tightness: the value of the sequences (an )
localizes away from infinity.
We conclude once again that some additional information must be known in order to derive compactness
in function space.
1
2 2024 SPRING 27200 HOMEWORK 2, DUE 04/08 23:59

5. We know from analysis that a continuous function defined on a compact metric space must be bounded,
and will attain its maximum and minimum. Construct a continuous function defined on the closed unit ball
of ℓ2 that is unbounded. This again shows that the closed unit ball of ℓ2 is not compact, although it is a
bounded closed set.

6. (1) Prove that the sequence space ℓ∞ is not separable. Hint. The cardinality of the set of all subsets of
N is the same as R.
(2) Prove that the space L∞ (Rn ) is not separable.
Thus it is always necessary to keep in mind that L∞ on a measure space is exotic.

3. Facts About Approximation in Lp Spaces, 2 Pt Total

1. (1) Suppose 1 ≤ p < ∞. Show that every function f ∈ Lp (Rn ) can be approximated by continuous
functions in the Lp norm. You may directly use the fact that f can be approximated by simple functions.
(2) Use the result of (1) to prove that a function f ∈ Lp (Rn ) with 1 ≤ p < ∞ is “continuous in the Lp
mean”, that is, as |h| → 0, Z
|f (x + h) − f (x)|p dx → 0.
Rn
(3) Using the fact that continuous functions on a compact subset can be uniformly approximated by
polynomials, prove that the space Lp (Rn ) is separable when 1 ≤ p < ∞.
(4) How does the proof of (1) fail for L∞ ?

2. This is a constructive proof that Lp functions on Rn can be approximated by smooth functions of compact
support.
(1) Define the convolution of two functions f, g on Rn by the following equality:
Z
(f ∗ g)(x) = f (x − y)g(y)dy,
Rn
provided that the right-hand-side is a well-defined Lebesgue integral.
Show that f ∗ g = g ∗ f , and prove Young’s inequality: for 1 ≤ p ≤ +∞, if f ∈ Lp (Rn ), g ∈ L1 (Rn ), then

∥f ∗ g∥Lp (Rn ) ≤ ∥f ∥Lp (Rn ) ∥g∥L1 (Rn ) .

You may directly use the Hölder inequality and Minkowski inequality without proof.
(2) Let φ(x) be a smooth function such that φ(0) = 1, φ(x) = 0 for |x| > 1, and
Z
φ(x)dx = 1.
Rn
−n
Set φt (x) = t φ(x/t), which is a scaling of φ. Given f ∈ Lp (Rn ), show that for every t > 0, the convolution
(f ∗ φt )(x) is a smooth function of x.
(3) Let 1 ≤ p < ∞. Prove that as t → 0, the convolution f ∗ φt converges to f in the Lp norm. Use this
to show that as t → 0, the smooth functions of compact support

φ1/t (x) · (f ∗ φt )(x)

converges to f in the Lp norm. Hint. Use well the dominated convergence theorem of Lebesgue.
(4) Which step of the proof of (3) fails for L∞ ?

4. Initial Application of Compactness, 1 Pt

In this exercise, you will see an initial application of the idea of “extracting convergent subsequence” in
proving existence of solution to differential equations.
 2024 SPRING 27200 HOMEWORK 2, DUE 04/08 23:59 3

Consider a continuous function f (t, y) defined for

(t, y) ∈ [0, a] × B̄(y0 , b).

Here B̄(y0 , b) is the closed ball centered at y0 of radius b in Rn . We are interested in the initial value problem
of ordinary differential equation
y ′ (t) = f t, y(t) , y(0) = y0 .

Fix L = |f |C 0 , and the interval  

h b i
I = 0, min a, .
L
We aim to prove that the initial value problem has at least one solution in the interval I.
(1) Show that the initial value problem is equivalent to an integral equation
Z t


y(t) = y0 + f τ, y(τ ) dτ.
0

(2) Let K be the set of all mappings y : I → B̄(y0 , b), such that y(0) = b, and
|y(t) − y(t′ )|
sup ≤ M.
t,t′ ∈I |t − t′ |
Show that K is a compact subset of the Banach space of Rn -valued continuous functions on I.
(3) Define a functional Φ : K → R by
Z t


Φ(y) = max y(t) − y0 − f τ, y(τ ) dτ .
t∈I 0

Show that Φ must attain its minimum on K, and every minimizing sequence must contain a convergent
subsequence converging to K.
(4) Prove minK Φ = 0 by showing that Φ can attain arbitrarily small value on piecewise linear functions
in K. These piecewise linear functions are nothing but the broken line approximation of the solution.