0% found this document useful (0 votes)
7 views

math400-exercises-chapt0-co

The document contains exercises and solutions related to linear independence of functions, properties of linear maps, norms on vector spaces, and completeness of normed spaces. It discusses the implications of stronger norms, equivalence of norms, and boundedness of linear operators in finite dimensional spaces. The document also explores the properties of combined norms in product spaces and their completeness.

Uploaded by

Titus
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
7 views

math400-exercises-chapt0-co

The document contains exercises and solutions related to linear independence of functions, properties of linear maps, norms on vector spaces, and completeness of normed spaces. It discusses the implications of stronger norms, equivalence of norms, and boundedness of linear operators in finite dimensional spaces. The document also explores the properties of combined norms in product spaces and their completeness.

Uploaded by

Titus
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

LU - Department of Mathematics

Dr H. Gebran

Math 400 Exercises of chapter 0 – Statements and solutions Fall 2022-2023

1. Let E = C[0, 1] be the space of all continuous functions f : [0, 1] → R.


a) Let fn (x) = sin(nπx) for n ∈ N∗ and x ∈ [0, 1]. Show that {fn , n ∈ N∗ } is linearly
independent.
b) Let fn (x) = enx for n ∈ N and x ∈ [0, 1]. Show that {fn , n ∈ N∗ } is linearly independent.

Solution. a) We have to prove that every finite subset of {fn , n ∈ N∗ } is linearly independent.
But every finite subset of this set is contained in a set of the form P{f
N
1 , f2 , . . . , fN }. So we have
to prove that this set is linearly independent. Therefore, let k=1 αk fk = 0. This means
that N
P
α
k=1 k sin(kπx) = 0 ∀x ∈ [0, 1]. Let m ∈ {1, . . . , N }. Multiplying by sin(mπx) and
integrating we get
XN Z 1
αk sin(kπx) sin(mπx) dx = 0.
k=1 0

1
But using the trigonometric formula sin(kπx) sin(mπx) = 2 (cos(k − m)πx − cos(k + m)πx),
we see that (
Z 1 1
if k = m
sin(kπx) sin(mπx) dx = 2
0 0 ̸ m.
if k =
It follows that αm = 0. Since m was an arbitrary integer between 1 and N , we have αk =
0 ∀k = 1, . . . , N . This proves the linear independence of {f1 , f2 , . . . , fN } and therefore of the
infinite set {fn , n ∈ N∗ }.
PN kx = 0∀x ∈ [0, 1]. Consider the polynomial P (t) =
PN k
b) Suppose that k=1 αk e k=1 αk t .
x
Then, P (e ) = 0 ∀x ∈ [0, 1]. Therefore, P (y) = 0 ∀y ∈ [1, e]. Therefore, P has infinitely many
roots and so all its coefficients are zero.

2. Show that a linear map L : E → F between vector spaces is determined by its values on a basis.

Solution. Let B = {ei | i ∈ I} be a basis of E. Suppose that vi = Lei is given for all
i ∈ I.
P Now if x ∈ E, there exists a unique finite collection (αi )i∈J of real numbers such that
x = i∈J αj ej . We set
X
Lx = αj vj .
i∈J
It is easy to check that L is a linear operator.

3. Let E be an infinite dimensional normed space. Show that there is a linear functional L : E → R
which is not bounded. Hint: Show first that there is a basis {ei |i ∈ I} such that ||ei || = 1 ∀i ∈ I.
Since I is infinite, it contains an infinite countable subset J; let φ : N∗ → J be a bijection. Set
Leφ(n) = n and Lei = 0 for i ∈ / J.

Solution. Let {vi | i ∈ I} be a basis of E. By linear independence, vi ̸= 0 ∀i ∈ I. Set


ei = ||vvii || . It is easy to check that the set {ei | i ∈ I} is linearly independent and spans E.
Therefore it is a basis of E.

1
Now set Leφ(n) = n and Lei = 0 (or any other value) if i ∈ / φ(N∗ ). By the previous exercise,
this defines a linear functional L : E → R. L is however unbounded because L is not bounded
on the unit ball (since eφ(n) ∈ BE and (Leφ(n) ) is unbounded).

4. Let E be a vector space and let || · ||1 and || · ||2 be two norms on E. We say that || · ||2 is
stronger than || · ||1 if there is a positive constant C such that ||x||1 ≤ C||x||2 for all x ∈ E.

a) Show that a stronger norm generates a stronger (or finer) topology.


b) Two norms || · ||1 and || · ||2 are called equivalent if R|| · ||1 is stronger than || · ||2 and || · ||2 is
1
stronger than ||·||1 . Let E = C([0, 1]). Let ||u||1 = 0 |u(t)|dt and ||u||∞ = max0≤t≤1 |u(t)|.
Show that || · ||∞ is strictly stronger than || · ||1 (the two norms are not equivalent). Hint.
Find a sequence (un ) such that ||un ||∞ is unbounded but ||un ||1 is bounded.
c) Show that all norms on a finite dimensional Pvector space are equivalent. Hint. Let {eP 1 , . . . , en }
be a basis of E. Then if x ∈ E, then x = ni=1 ai ei in a unique way. Set ||x||1 = ni=1 |ai |.
Show that || · ||1 is a norm on E. Let || · || be an arbitrary norm on E. Show that || · ||1 is
stronger than || · ||. Deduce that x 7→ ||x|| is continuous in (E, || · ||1 ). Use the compactness
of the unit sphere ||x||1 = 1 to minimize ||x||. Conclude.
d) Let || · ||1 and || · ||2 be two equivalent norms on a vector space E. Show that (E, || · ||1 ) is
complete if and only if (E, || · ||2 ) is complete.

Solution. a) Let B1 (x, r) = {y ∈ E | ||y − x||1 < r} and B2 (x, r) = {y ∈ E | ||y − x||2 < r}.
The inequality ||x||1 ≤ C||x||2 implies that B2 (x, Cr ) ⊂ B1 (x, r).
Now let O be open for || · ||1 and let x ∈ O. Then there exists r > 0 such that B1 (x, r) ⊂ O.
By the previous remark, B2 (x, Cr ) ⊂ O. Therefore O is open for || · ∥2 .

b) Observe first that ||u||1 ≤ ||u||∞ . On the other hand consider the sequence (un ) defined by
(
−n2 t + n if 0 ≤ t ≤ n1
un (t) =
0 otherwise.

Draw a figure. Then ||un ||1 = 1 but ||un ||∞ = n → ∞ as n → 0 so we cannot have ||u||∞ ≤
C||u||1 .

c) Let us check the four properties of a norm.

(i) ||x||1 = ni=1 |αi | ≥ 0.


P

(ii) ||x||1 = 0 ⇔ ni=1 |αi | = 0 ⇔ αi = 0 ∀i ⇔ x = 0.


P

(iii) ||αx||1 = ni=1 |α||ai | = |α| ni=1 |ai | = |α||x||1 .


P P

(iv) Let x = ni=1 ai ei and y = ni=1 bi ei . Then, x + y = ni=1 (ai + bi )ei and so
P P P

n
X n
X n
X
||x + y||1 = |ai + bi | ≤ |ai | + |bi | = ||x||1 + ||y||1 .
i=1 i=1 i=1
Pn Pn
Now let x = i=1 ai ei . Then, ||x|| ≤ i=1 |ai |||ei || ≤ M ||x||1 where M = max1≤i≤n ||ei ||.
Next we have |||x|| − ||y||| ≤ ∥x − y|| ≤ M ||x − y||1 . It follows that the map x 7→ ||x|| is
uniformly continuous on (E, || · ||1 ). Let S1 = {x ∈ E | ||x||1 = 1} be the unit sphere in
(E, || · ||1 ). Since S1 is compact, the continuous map x 7→ ||x|| has a minimum value at a point
x0 ∈ S1 . So we have ||x|| ≥ ||x0 || for all x ∈ S1 . If we set m = ||x0 ||, then m > 0 because

2
x x
otherwise x0 = 0 ∈ / S1 . Now let x ∈ E satisfy x ̸= 0. Then, ||x|| 1
∈ S1 and so ||x||1 ≥ m. It
follows that ||x|| ≥ m||x||1 . Since this inequality still holds for x = 0. We get

||x|| ≥ m||x||1 ∀x ∈ E.

Thus, the two norms are equivalent. It follows that all norms on E are equivalent.

d) By assumption, there exist two positive constants c and C such that

c||x||1 ≤ ||x||2 ≤ C||x||1 .

Suppose that (E, || · ||1 ) is complete. Let (xn ) be a Cauchy sequence in (E, || · ||2 ). The
inequality ||xn − xm ||1 ≤ 1c ||xn − xm ||2 implies that (xn ) is a Cauchy sequence in (E, || · ||1 ).
Completeness of (E, || · ||1 ) implies that (xn ) converges to some x in (E, || · ||1 ). The inequality
||xn −x||2 ≤ C||xn −x||1 implies that (xn ) converges in (E, ||·||2 ). Thus, (E, ||·||2 ) is complete.
Since || · ||1 and || · ||1 play the same role, completeness of (E, || · ||2 ) implies completeness of
(E, || · ||1 ).

5. Let E and F be normed spaces and let L : E → F be a linear operator. Show that if E is finite
dimensional then L is bounded.

Solution. Let {e
P1 , . . . , en } be a basis ofP
E. Let x ∈ E. Then, x can be written in a
unique way P as x = ni=1 αi ei . Then Lx = n
i=1 αi Lei . Let M = max1≤i≤n ||Lei ||. Then
n Pn
||Lx|| ≤ M i=1 |αi |. If we set ||x||1 = i=1 |αi |, then || · ||1 is a norm equivalent to any other
norm on E (see exercise 4). Moreover ||Lx|| ≤ M ||x||1 . Since there exists a constant C such
that ||x||1 ≤ C||x||, we get ||Lx|| ≤ M C||x||. This proves the boundedness of L.

6. Let (E, || · ||E ) and (F, || · ||F ) be two normed spaces. Set

||(x, y)||1 = ||x||E + ||y||F .


||(x, y)||∞ = max(||x||E , ||y||F ).
1/2
||(x, y)||2 = ||x||2E + ||y||2F .

(a) Show that || · ||1 , || · ||∞ , and || · ||2 , are norms on E × F .


(b) Show that all the above norms are equivalent.
(c) Show that if (E, || · ||E ) and (F, || · ||F ) are complete, then (E × F, || · ||1 ) is complete.

Solution. a) We only prove that that || · ||2 is a norm.

(i) ||(x, y)||2 ≥ 0.


(ii) ||(x, y)||2 = 0 ⇔ ||x||2E + ||y||2F = 0 ⇔ ||x||E = 0 and ||y||F = 0 ⇔ x = 0 and y = 0 ⇔
(x, y) = 0.
(iii) ||α(x, y)||22 = ||(αx, αy)||22 = ||αx||2E + ||αy||2F = |α|2 ||x||2E + ||y||2F = |α|2 |||(x, y)||22 .

(iv) To simplify the writing we drop the subscripts. We have to prove that

||(x, y) + (x′ , y ′ )|| ≤ ||(x, y)|| + ||(x′ , y ′ )||.

This inequality is equivalent to the inequality


1/2 1/2
||x+x′ ||2 +||y+y ′ ||2 ≤ ||x||2 +||y||2 +||x′ ||2 +||y ′ ||2 +2 ||x||2 + ||y||2 ||x′ ||2 + ||y ′ ||2 .

3
Now by the triangle inequality for || · ||E and || · ||F we have

||x + x′ ||2 + ||y + y ′ ||2 ≤ ||x||2 + ||x′ ||2 + 2||x||||x′ || + ||y||2 + ||y ′ ||2 + 2||y||||y ′ ||.

Therefore it is enough to prove that


1/2 1/2
||x||||x′ || + ||y||||y ′ || ≤ ||x||2 + ||y||2 ||x′ ||2 + ||y ′ ||2 .

This inequality is equivalent to the inequality

2||x||||x′ ||||y||||y ′ || ≤ ||x||2 ||y ′ ||2 + ||y||2 ||x′ ||2 .

This inequality is indeed true since 2ab ≤ a2 + b2 .

b) We have
||(x, y)||∞ ≤ ∥(x, y)||1 ≤ 2||(x, y)||∞
and √
||(x, y)||∞ ≤ ∥(x, y)||2 ≤ 2||(x, y)||∞ .

c) Let (xn , yn ) be a Cauchy sequence in (E × F, || · ||1 ). Then, for any ε > 0 there exists n0
such that
||xn − xm || + ||yn − ym || ≤ ε
for n, m ≥ n0 . It follows that (xn ) and (yn ) are Cauchy sequences in E and F respectively.
Completeness of E and F implies that (xn ) converges to some x in E and (yn ) converges to
some y in F . It follows that

||(xn , yn ) − (x, y)||1 = ||xn − x|| + ||yn − y|| → 0

and so ((xn , yn )) converges to (x, y) in E × F .

7. Let E and F be normed spaces and let L ∈ L(E, F ). Recall that the norm of L is defined by
||L|| = sup||x||≤1 ||Lx||. Show that

||Lx||
||L|| = sup ||Lx|| = sup ||Lx|| = sup = inf{C > 0 ; ||Lx|| ≤ C||x||}.
||x||=1 ||x||<1 x̸=0 ||x||

Solution. First, we indeed have

sup ||Lx|| ≤ sup ||Lx|| = ||L||.


||x||=1 ||x||≤1

x
Set M = sup||x||=1 ||Lx|| = supx∈SE ||Lx||. Let now x ∈ BE with x ̸= 0. Then ||x|| ∈ SE .
Therefore,
x
L ≤ M.
||x||
This means that ||Lx|| ≤ M ||x|| ≤ M . Since ||L0|| = 0 ≤ M we have

||Lx|| ≤ M ∀x ∈ BE .

It follows that ||L|| = supx∈BE ||Lx|| ≤ M . Hence the first equality.

Let m = sup||x||<1 ||Lx||. Then, m ≤ ||L|| since {x; ||x|| < 1} ⊂ {x; ||x|| ≤ 1}. Let A = {x ∈
BE ; ||Lx|| ≤ m}. By continuity of x 7→ ||Lx||, the set A is closed. On the other hand, we have
by definition that {x; ||x|| < 1} ⊂ A. It follows that {x; ||x|| < 1} ⊂ A since A is closed. But

4
{x; ||x|| < 1} = BE . Therefore, BE ⊂ A. This means that ||Lx|| ≤ m, ∀x ∈ BE . It follows
that ||L|| ≤ m. Hence equality. Actually A = BE since by definition A ⊂ BE .

The third equality follows from that fact that



||Lx|| x
; x ̸= 0 = |L ; x ̸= 0 = {||Lu|| ; u ∈ SE }.
||x|| ||x||

Next, note that ||Lx|| ≤ ||L||||x|| for all x ∈ E. Setting α = inf{C > 0 | ||Lx|| ≤ C||x||}, we
see that α ≤ ||L||. Conversely, let ε > 0. Then, by a property of the infimum, there exists
a positive constant Cε such that ||Lx|| ≤ Cε ||x|| ∀x ∈ E and Cε < α + ε. It follows that
||Lx|| ≤ (α + ε)||x|| ∀x ∈ E. Taking the sup over BE we get ||L|| ≤ α + ε. Since ε was
arbitrary, we get ||L|| ≤ α. Hence the last equality.

8. Let E be a normed space and let f ∈ E ∗ = L(E, R). Show that

(a) ||f || = sup f (x) = sup f (x).


||x||≤1 ||x||=1

(b) −||f || = inf f (x) = inf f (x).


||x||≤1 ||x||=1

Solution. (a) Since f (x) ≤ |f (x)| for all x ∈ E, we have on the one hand

sup f (x) ≤ sup |f (x)|.


x∈BE x∈BE

On the other hand, let x ∈ BE . Then either |f (x)| = f (x) or |f (x)| = −f (x) = f (−x). Since
−x ∈ BE we have in both cases |f (x)| ≤ supx∈BE f (x). Since x was arbitrary, it follows that

sup |f (x)| ≤ sup f (x).


x∈BE x∈BE

Hence the first equality. Repeating the above reasoning with SE instead of BE , proves the
second equality.

(b) Using the facts that BE is symmetric with respect to the origin and f (−x) = −f (x), we
can write
inf f (x) = − sup (−f (x)) = − sup f (x) = −||f ||.
x∈BE x∈BE x∈BE

Repeating the above reasoning with SE , we get the second equality.

9. Let E be a finite dimensional vector space and let {e1 , . . . , en }Pbe a basis of E. Then every
element x ∈ E can be written in a unique way in the form x = ni=1 xi ei . We set then

n n
!1/p
X X
p
||x||1 = |xi |, ||x||∞ = max |xi |, ||x||p = |xi | (1 < p < ∞).
1≤i≤n
i=1 i=1

(a) We equip E with the norm || · ||1 . Let f ∈ E ∗ and let fi = f (ei ). Show that

||f ||E ∗ = max |fi |.


1≤i≤n

(b) We equip E with the norm || · ||∞ . Let f ∈ E ∗ and let fi = f (ei ). Show that
n
X
||f ||E ∗ = |fi |.
i=1

5
(c) We equip E with the norm || · ||p where 1 < p < ∞. Let f ∈ E ∗ and let fi = f (ei ). Show
that !1/p′
n

X
||f ||E ∗ = |fi |p
i=1
1 1
where p + p′ = 1.

(a) Observe first that f (x) = f ( ni=1 xi ei ) = ni=1 xi f (ei ) = ni=1 xi fi . It


P P P
Solution.
follows that |f (x)| ≤ max1≤i≤n |fi |||x||1 and so |f (x)| ≤ max1≤i≤n |fi |.
Let k be an index such that |fk | = max1≤i≤n |fi |. Observe that fk = f (ek ) ≤ ||f || since
||ek ||1 = 1. Observe also that −fk = f (−ek ) ≤ ||f ||. It follows that |fk | ≤ ||f ||. Therefore,
max1≤i≤n |fi | ≤ ||f ||. Hence equality.
Pn Pn
(b)
Pn As before, we have
Pn f (x) = i=1 x i fi . Therefore, |f (x)| ≤ ( i=1 |fi |) ||x||∞ and so ||f || ≤
i=1 |fi |. Let x = Psign (fi )ei where
i=1 Psign (t) = 1 if t ≥ 0 and sign (t) = −1 if t < 0. Then
||x||∞ = 1 and f (x) = ni=1 |fi |. Thus ni=1 |fi | ≤ ||f ||. Hence equality.
P 1/p′ P P 1/p′
n p′ n p 1/p = n p′

(c) By Hölder’s inequality, |f (x)| ≤ i=1 |fi | i=1 |xi | i=1 |fi | ||x||p .
P 1/p ′
n p′ ′ Pn
It follows that ||f || ≤ i=1 |fi | . Take xi = fi |fi |p −2 and x = i=1 xi ei . Then
P 1/p
n p′ ′
(because (p′ − 1)p = p′ ) and f (x) = ni=1 |fi |p . Therefore
P
||x||p = i=1 |fi |

n n
!1/p
p′ p′
X X
|fi | ≤ ||f || |fi | .
i=1 i=1

Therefore,
n
!1−1/p
p′
X
|fi | ≤ ||f ||.
i=1
1 1
But 1 − p = p′ . Hence the reverse inequality and therefore the equality.

10. Let E = C[0, 1]. Equip E with the norm of uniform convergence ||x|| = sup0≤t≤1 |x(t)|. Let
g ∈ E. Define the linear operator f : E → R by
Z 1
f (x) = g(t)x(t) dt.
0

a) Show that f is bounded.


Z 1
ng(t)
b) Show that ||f || = |g(t)| dx. Hint: consider the sequence xn (t) = and compute
0 1 + n|g(t)|
lim f (xn ).
c) Suppose g ≥ 0. Show that there exists an element x ∈ BE such that f (x) = ||f ||.
d) Let g(t) = cos(πt). Show that there is no element x ∈ BE such that f (x) = ||f || (i.e. the
sup is not achieved).
Z 1
Solution. a) Let ||g||1 = |g(t)| dt. We have for all x ∈ E
0
Z 1
|f (x)| ≤ |g(t)||x(t)| dt ≤ ||x||||g||1 .
0

6
It follows that f is bounded and ||f || ≤ ||g||1 .
Z 1
ng 2 (t)
b) We have f (xn ) = dt. Observe that ||xn || ≤ 1. Now
0 1 + n|g(t)|

ng 2 (t) |g(t)| 1
− |g(t)| = ≤
1 + n|g(t)| 1 + n|g(t)| n

ng 2 (t)
This means that the sequence of functions converges uniformly to |g(t)|. Therefore
1 + n|g(t)|
Z 1 1
ng 2 (t)
Z
converges to |g(t)| dt. This means that f (xn ) → ||g||1 . But
0 1 + n|g(t)| 0

|f (xn )| ≤ ||f ||||xn || ≤ ||f ||.

Letting n → ∞ in the last inequality, we get ||g||1 ≤ ||f ||. Hence equality.

c) Let x0 be the constant function equal to 1. Then x0 ∈ BE and


Z 1 Z 1
f (x0 ) = g(t) = |g(t)| dt = ||f ||.
0 0

d) Suppose that there exists x0 ∈ BE such that


Z 1 Z 1
cos(πt)x0 (t) dt = | cos πt| dt.
0 0

Then −1 ≤ x0 (t) ≤ 1 ∀t ∈ [0, 1] and


1 1
Z
2
Z 1 Z
2
Z 1
cos(πt)x0 (t) dt + cos(πt)x0 (t) dt = cos πt dt − cos πt dt.
1 1
0 2
0 2

It follows that 1
Z 1 Z
2
cos(πt)[x0 (t) + 1] dt = cos(πt)[1 − x0 (t)] dt.
1
2
0

The integral on the left is ≤ 0 because cos(πt) ≤ 0 and x0 (t) + 1 ≥ 0, whereas the integral on
the right is ≥ 0. Therefore the two integrals are zero. Since the integrands are continuous and
are of constant sign, they are zero. Thus,
(
1 if t < 21
x0 (t) =
−1 if t > 12 .

But this contradicts the continuity of x0 . Therefore the sup is not achieved.

11. Let E be a normed space. Let a, b ∈ E, r, s, λ > 0. Show that

(i) λB(a, r) = B(λa, λr).


(ii) a + B(b, r) = B(a + b, r).
(iii) B(a, r) + B(b, s) = B(a + b, r + s).

7
Solution. (i)
x x
x ∈ λB(a, r) ⇔ ∈ B(a, r) ⇔ || − a|| < r ⇔ ||x − λa|| < λr ⇔ x ∈ B(λa, λr).
λ λ
(ii)
x ∈ a + B(b, r) ⇔ x − a ∈ B(b, r) ⇔ ||x − a − b|| < r ⇔ x ∈ B(a + b, r).
(iii) The inclusion B(a, r) + B(b, s) ⊂ B(a + b, r + s) follows from the triangle inequality.
r s
Conversely, let x ∈ B(a + b, r + s). Set x1 = a + r+s (x − a − b) and x2 = b + r+s (x − a − b).
Then x1 ∈ B(a, r); x2 ∈ B(b, s) and x = x1 + x2 .

Remark. The open balls can be replaced by the corresponding closed balls.

12. Let A and B be two subsets of a normed space E. Show that if A is closed and B is compact
then A + B is closed.

Solution. Let x belong to the closure of A + B. Then, there exists a sequence (xn ) of A + B
that converges to x. We can write xn = an + bn where an ∈ A and bn ∈ B. Compactness of B
implies that (bn ) has a subsequence (bnk ) that converges to some b ∈ B. Since (xnk ) converges
to x, the sequence (ank ) converges to a := x − b. Since A is closed, a ∈ A. Therefore,
x = a + b ∈ A + B.

13. Let E be a normed space and let F be a subspace of E with nonempty interior. Show that
F = E.

Solution. Let a be an interior point of F . Then, there exists r > 0 such that B ′ (a, r) ⊂ F .
r ||x − a||
Now let x ∈ E. Let y = a + (x − a) so that x = a + (y − a). Observe
||x − a|| r
that ||y − a|| = r. Therefore y ∈ B ′ (a, r) ⊂ F . Since F is a subspace and a ∈ F , we have
||x − a||
x=a+ (y − a) ∈ F . Thus, E ⊂ F ⊂ E and so F = E.
r

14. Let F be a subspace of a vector space E. Let f : E → R be linear. Show that if f is bounded
from above or from below on F , then f = 0 on F .

Solution. Suppose that f is bounded from above. Then there exists a number α such that
f (x) ≤ α for all x ∈ F . Therefore for every t ∈ R and every x ∈ F we have

tf (x) ≤ α.

Suppose that f ̸= 0 on F . Then there exists x0 ∈ F such that f (x0 ) ̸= 0. We may assume
α
that f (x0 ) > 0 because if f (x0 ) < 0, then f (−x0 ) > 0. Take in particular a number t > f (x 0)
.
Then, we get α < tf (x0 ) ≤ α, a contradiction. Therefore f vanishes on F .
Now if f is bounded from below, then −f is bounded from above and so according to what we
said, −f = 0 on F . Therefore f = 0 on F .

15. Let E be a normed space and let C ⊂ E be convex.



(a) Prove that C and C are convex.
◦ ◦
(b) Given x ∈ C and y ∈ C, show that (1 − t)x + ty ∈ C for every t ∈]0, 1[.

8
◦ ◦
(c) Deduce that C = C whenever C ̸= ∅.

Solution. (a) Let x, y ∈ C and let t ∈ [0, 1]. Then, there exist a sequence (xn ) in C
converging to x and a sequence (yn ) of C converging to y. Since C is convex, (1−t)xn +tyn ∈ C.
But (1 − t)xn + tyn converges to (1 − t)x + ty. Therefore (1 − t)x + ty ∈ C.

Let x, y ∈ C and let t ∈]0, 1[. Then, there exists r > 0 such that B(x, r) ⊂ C and B(y, r) ⊂ C.
Now, (1 − t)B(x, r) + tB(y, r) ⊂ (1 − t)C + tC. Since C is convex, (1 − t)C + tC ⊂ C
and so (1 − t)B(x, r) + tB(y, r) ⊂ C. By the previous exercise, (1 − t)B(x, r) + tB(y, r) =

B((1 − t)x + ty, r). Therefore B((1 − t)x + ty, r) ⊂ C. This means (1 − t)x + ty ∈ C.

(b) There exists r > 0 such that B(y, r) ⊂ C. Since C is convex, (1 − t)x + tB(y, r) ⊂ C.
But it is not difficult to see that (1 − t)x + tB(y, r) = B((1 − t)x + ty, tr). It follows that

(1 − t)x + ty ∈ C.

(c) One inclusion is clear. Conversely, let x ∈ C and fix y0 ∈ C. By the previous question,
◦ ◦
(1 − n1 )x + n1 y0 ∈ C. But this sequence converges to x. It follows that x ∈ C. This proves that
◦ ◦
C ⊂ C and therefore C ⊂ C.

16. Let f : X → Y be a function between two metric spaces. The graph of f is the set G(f ) =
{(x, f (x)) | x ∈ X}.

(a) Show that if f is continuous, then G(f ) is closed.


(b) Show by an example that the converse is not true.
(c) Show that if Y is compact and G(f ) is closed, then f is continuous.

Solution. (a) Let (x, y) belong to the closure of G(f ). Then, there exists a sequence
((xn , yn )) in G(f ) that converges to (x, y). Therefore xn → x and yn = f (xn ) → y. Continuity
of f implies that f (xn ) → f (x). By uniqueness of limits in metric spaces, we get y = f (x).
Therefore (x, y) ∈ G(f ).

(b) Consider the function f : R → R defined by


(
1
x if x ̸= 0
f (x) =
0 if x = 0.

Then G(f ) is a union of the three sets F1 = {(x, x1 ) | x > 0}; F2 = {0} and F3 = {(x, x1 ) | x < 0}.
It is not difficult to see that F1 , F2 and F3 are closed subsets of R2 . Therefore, G(f ) is closed
but f is clearly not continuous.

(c) Let x ∈ X and let xn → x in X. Compactness of Y implies that the sequence (f (xn ))
has a subsequence (f (xnk )) that converges to some y ∈ Y . Now the sequence (xnk , f (xnk ))
belongs to G(f ) and converges to (x, y). Since G(f ) is closed, (x, y) ∈ G(f ) and so y = f (x).
Therefore f (xnk ) → f (x). This means that f is continuous at x. Since x was arbitrary, f is
continuous on X.

17. (a) Show that in a normed space all open balls are homeomorphic and all closed balls are
homeomorphic.

9
(b) Let E be an infinite dimensional normed space and let K ⊂ E be compact. Show that the
interior of K is empty, i.e., K cannot contain any open ball.

Solution. (a) By a previous exercise, we can write

B(a, r) = a + rB(0, 1) = φ(B(0, 1))

where φ(x) = a+rx. Indeed φ is a homeomorphism (it is a homothecy followed by a translation).


Similarly we can write B ′ (a, r) = φ(BE ).

(b) Suppose that K contains an open ball. Then it contains a closed ball B ′ (a, r). Being a closed
subset of a compact space, B ′ (a, r) is itself compact. It follows from the previous question that
BE is compact. But this contradicts the fact that E is infinite dimensional.

10

You might also like