Você está na página 1de 14

MATH31011/41011/61011: FOURIER ANALYSIS &

LEBESGUE INTEGRATION - JANUARY 2011 SOLUTIONS


A1. The Fourier series associated to f is the series
a
0
2
+

n=1
(a
n
cos(nx) +b
n
sin(nx)) ,
where
a
0
=
1

f(x)dx, a
n
=
1

f(x) cos(nx)dx, b
n
=
1

f(x) sin(nx)dx, n 1.
Alternative solution:

n=
c
n
e
inx
,
where
c
n
=
1
2
_

f(x)e
inx
dx.
[3 marks]
FEEDBACK: Generally answered correctly. A small number of students made slips like
replacing the sum by an integral or forgetting to include the function f(x) in the formula
for the coecients.
A2. A set E is countably innite if there exists a bijection : N E.
[2 marks]
FEEDBACK: Lots of students lost marks here by writing that a set E was countably
innite if there was a surjection from N to E. This is the denition of a set being countable,
not countably innite. For example, the set 1 is countable but not innite and there is a
surjection : N 1 given by (n) = 1 for all n N. An acceptable alternative solution
was to say that there was a surjection from N to E and that E was innite.
Typeset by A
M
S-T
E
X
1
A3. B is a -algebras if (i) , X B; (ii) E B = XE B; (iii) if E
j

j=1
B then

j=1
E
j
B.
A -algebra containing Q[0, 1] but not [0, 1/2] is the set of E [0, 1] such that either
E or [0, 1]E is countable (or, alternatively, is a null set).
Two other examples from , [0, 1]; Borel sets in [0, 1]; measurable sets in [0, 1];
T([0, 1]).
[6 marks]
FEEDBACK: Lots of students had diculties in identifying a -algebra containing Q
[0, 1] but not containing [0, 1/2]. Some of these even listed one or both of the solutions
given above among the other two examples. An alternative and perfectly correct solution
was the -algebra , Q [0, 1], [0, 1]Q, [0, 1].
A4. m : / R
+
is countably additive if, whenever E
j

j=1
/ are disjoint sets, we
have
m
_
_

_
j=1
E
j
_
_
=

j=1
m(E
j
).
[3 marks]
FEEDBACK: Many students lost marks by forgetting to say the sets had to be disjoint.
A5. We have
_
f d =
k

i=1

i
(E
i
).
[3 marks]
FEEDBACK: Nearly everyone answered this correctly.
A6. A function f : [, ] R

is square integrable if [f[


2
is integrable, i.e.
_

[f[
2
d <
+.
L
2
([, ], , R) is the set of equivalence classes of square integrable functions f :
[, ] R

, where two functions are equivalent if they are equal -almost everywhere.
[4 marks]
FEEDBACK: This too was anwered correctly by most students. A few lost marks by
confusing square integrable with integrable.
A7. A set v
n
is an orthonormal basis in an inner product space V if
(1) v
n
, v
m
) = 0 if n ,= m;
(2) v
n
, v
n
) = 1 for all n;
(3) v
n
is linearly independent;
(4) v
n
spans V .
[4 marks]
FEEDBACK: Some students lost marks were lost by forgetting to say what a basis was.
Some other students lost marks by just writing down the orthonormal basis for L
2
instead
of giving the required denition.
2
B8.(i) E is a null set if, for evey > 0, there exists a countable collection of open intervals
I
j
such that
E
_
I
j
and

j
l(I
j
) < .
[2 marks]
(ii) If E is countable then we can write E = x
j

j=1
. [I dont expect the simpler nite case
to be written down explicitly.] Given > 0, dene open intervals I
j
= (x
j
/2
j+2
, x
j
+
/2
j+2
). Clearly E

I
j
and

j=1
l(I
j
) =

j=1

2
j+1
=

2
< .
Thus E is a null set.
[6 marks]
(iii) The Middle Third Cantor set is constructed as follows. Start with C
0
= [0, 1]. Remove
the open middle third (1/3, 2/3) to obtain
C
1
= [0, 1/3] [2/3, 1].
Continue in this way by removing the open middle third of each closed interval to obtain
a nested sequence of sets
C
1
C
2
C
n
.
C
n
will consist on 2
n
disjoint closed intervals each of length 3
n
. The Middle Third Cantor
set C is dened by
C =

n=0
C
n
.
[8 marks]
(iv) Call the 2
n
closed intervals making up C
n
C
1
n
, . . . , C
2
n
n
. Enlarge these to make open
intervals by setting
U
i
n
=
_
a
1
2 7
n
, b +
1
2 7
n
_
,
where C
i
n
= [a, b]. Note that
l(U
i
n
) = 3
n
+ 7
n
<
_
1
3
+
1
7
_
n
<
1
2
n
.
Then C

n
i=1
U
i
n
and
n

i=1
l(U
i
n
) < 2
n
_
1
3
+
1
7
_
n
.
3
Given > 0, since 2
_
1
3
+
1
7
_
< 1, we can choose n suciently large that
2
n
_
1
3
+
1
7
_
n
< .
Thus C is a null set.
[9 marks]
FEEDBACK: Nearly all students attempted this question. Many students lost marks by
reproducing the proof (from the mock exam) that a countable union of null sets was null.
This was not what the question asked. The rest of the question was generally quite well
answered.
4
B9.(i) A function f : [0, 1] R

is measurable if, for each c R,


x [0, 1] : f(x) > c
is a measurable set (or any similar equivalent criterion).
[3 marks]
(ii) Suppose that
f(x) +g(x) > c.
Then
f(x) > c g(x).
Since Q is dense in R, we can nd r Q such that
f(x) > r > c g(x),
which may be rewitten
f(x) > r and g(x) > c r.
Now suppose that there exists r Q such that
f(x) > r and g(x) > c r.
This may be rewritten
f(x) > r > c g(x),
so
f(x) > c g(x).
Rearranging,
f(x) +g(x) > c.
We have shown that f(x) + g(x) > c i there exists r Q such that f(x) > r and
g(x) > c r. Hence
x [0, 1] : f(x) +g(x) > c =
rQ
(x [0, 1] : f(x) > r x [0, 1] : g(x) > c r).
[11 marks]
(iii) Since f is measurable, for each r Q,
A
r
= x R : f(x) > r
is measurable. Since g is measurable, for each r Q,
B
r
= x R : g(x) > c r
5
is measurable. Since the measurable sets form a -algebra, A
r
B
r
is measurable and so
is
_
rQ
(A
r
B
r
).
By part (ii), the latter shows that
x R : f(x) +g(x) > c
is measurable, giving f +g measurable.
[11 marks]
FEEDBACK: Rather few students could give a clear account of the required argument
in part (ii) but part (iii) was answered better. There were a lot of vague and inaccurate
solutions to this question. A lot of students just regurgitated the solution to B9(iii) on the
mock exam, for which they received no marks.
6
B10.(i) f : [0, 1] R

is integrable if
_
f
+
d,
_
f

d < +
(f
+
, f

are the postive and negative parts of f) or, alternatively, if


_
[f[ d < +.
[3 marks]
(ii) Monotone Convergence Theorem: Suppose that f
n
is an increasing sequence of non-
negative integrable functions and that lim
n+
f
n
= f (pointwise). Then
lim
n+
_
f
n
d =
_
f d.
[3 marks]
Proof: For each n, choose an increasing sequence of non-negative simple functions f
n,k
converging to f
n
, as k +. Set
g
k
= max
nk
f
n,k
.
The g
k
are simple and form an increasing sequence. Thus we can dene
g = lim
k+
g
k
.
For 1 n k, we have
f
n,k
g
k
f
k
f. ()
Letting k + gives
f
n
g f.
Letting n + gives
f g f,
i.e. f = g.
Integrating (*) gives, for 1 n k,
_
f
n,k
d
_
g
k
d
_
f
k
d.
Letting k + gives
_
f
n
d
_
g d lim
k+
_
f
k
d.
7
Letting n + gives
lim
n+
_
f
n
d
_
g d lim
k+
_
f
k
d.
Since f = g this completes the proof.
(Bookwork)
[12 marks]
(iii) Write
g
n
=
n

i=1
f
i
.
Then g
n
is an increasing sequence of non-negative measurable functions whose limit is

n=1
f
n
. So, by the Monotone Convergence Theorem,
_
g
n
d
_
_

n=1
f
n
_
d,
as n +. But by additivity of the integral,
_
g
n
d =
n

i=1
_
f
i
d.
Therefore,

n=1
__
f
n
d
_
=
_
_

n=1
f
n
_
d.
[7 marks]
FEEDBACK: Part (i) of the question was answered correctly by nearly everyone who
attempted the question. A lot of students could give a complete or nearly complete proof
of the MCT. A common slip was to forget to mention that the functions f
n,k
, and hence
the functions g
k
, were simple. A number of students answered part (ii) by giving the proof
of the DCT (from the mock exam), for which they received no marks. Very few students
answered part (iii) correctly. As you can see, it is a straightforward application of the
MCT.
8
B11.(i) Let c
1
, . . . , c
n
be arbitrary real numbers and set a
k
= w, v
k
). Write u =

n
k=1
a
k
v
k
and v =

n
k=1
c
k
v
k
. By orthonormality,
|u|
2
=
n

k=1
a
2
k
and |v|
2
=
n

k=1
c
2
k
.
Also
w, v) =
_
w,
n

k=1
c
k
v
k
_
=
n

k=1
c
k
w, v
k
) =
n

k=1
c
k
a
k
.
Thus
|w v|
2
= w v, w v)
= |w|
2
2w, v) +|v|
2
= |w|
2
2
n

k=1
c
k
a
k
+
n

k=1
c
2
k
= |w|
2

k=1
a
2
k
+
n

k=1
(a
k
c
k
)
2
= |w|
2
|u|
2
+
n

k=1
(a
k
c
k
)
2
.
It follows that
|w v|
2
|w|
2
|u|
2
with equality if and only if

n
k=1
(a
k
c
k
)
2
= 0, i.e. if and only if c
k
= a
k
= w, v
k
) for
all k = 1, . . . , n .
[9 marks]
(ii) Riesz-Fischer Theorem: Let f L
2
([, ], , R). Then S
n
(f, ) converges to f in
L
2
([, ], , R), i.e,
|S
n
(f, ) f|
2
=
_
1

[S
n
(f, ) f[
2
d
_
1/2
0, as n +.
[4 marks]
Proof: Suppose f L
2
([, ], , R) and let > 0 be given. We can nd a continuous
function g : [, ] R such that |f g|
2
< /2.
By Fejers Theorem, we can choose N suciently large that
n N = |
n
(g, ) g|

<

2

2
.
Thus, if n N then
|
n
(g, ) g|
2

2|
n
(g, ) g|

<

2
.
9
Combing the two estimates, if n N then
|f
n
(g, )|
2
|f g|
2
+|g
n
(g, )|
2
<

2
+

2
= .
We may write

n
(g, x) =
n

k=n
c
k

k
(x),
for some c
k
R. By part (i),
_
_
_
_
_
f
n

k=n
f,
k
)
k
(x)
_
_
_
_
_
2

_
_
_
_
_
f
n

k=n
c
k

k
(x)
_
_
_
_
_
2
.
Thus, if n N then |f S
n
(f, )|
2
< , as required.
[12 marks]
FEEDBACK: This question was answered by relatively few students and spome of these
only made a serious attempt at part of it. Most students who did part (i) give a correct
solution. Also, most people who attempted part (ii) gave a correct statement of the Riesz-
Fischer Theorem. There were few serious attempts to give a proof and even fewer that
were completely correct.
10
C12.(i) If

(E
n
) = + for some n then the result is immediate.
Suppose

(E
n
) < + for all n. Choose > 0 and, for each n, a cover I
(n)
j
of E
n
by
intervals such that

j
l(I
(n)
j
) <

(E
n
) +

2
n
.
Then

n=1
I
(n)
j
is a cover of

n=1
E
n
by intervals, so

_

_
n=1
E
n
_

n=1

j
l(I
(n)
j
)

n=1
_

(E
n
) +

2
n
_
=

n=1

(E
n
) +.
Since > 0 is arbitrary,

_

_
n=1
E
n
_

n=1

(E
n
).
[8 marks]
(ii) First we show that (a, ) is measurable. Take A R and set
A
1
= (a, ) A, A
2
= (, a] A.
If

(A) = + the criterion for measurability is automatically satised. Suppose

(A) <
+ and choose > 0. We can nd a cover I
j
of A by intervals such that

j
l(I
j
)

(A) +.
Write
I

j
= I
j
(a, ), I

j
= I
j
(, a]
(intervals or empty). Then I

j
(resp.I

j
) is a cover of A
1
(resp. A
2
) by intervals. Thus

(A
1
) +

(A
2
)

j
_
l(I

j
) +l(I

j
)
_
=

j
l(I
j
)

(A) +.
Since > 0 is arbitrary,

(A
1
) +

(A
2
)

(A),
11
which gives (a, ) measurable.
Since measurable sets form a -algebra,
(, a] = R(a, )
is measurable and so is
(, a) =

_
n=1
(, a
1
n
].
Hence
(a, b) = (a, ) (, b)
is measurable.
[13 marks]
(iii) The Borel sets B are the smallest -algebra containing all open intervals in R. By
part (ii), / (measurable sets) is a -algebra containing all open intervals. Thus B /,
i.e. all Borel sets are measurable.
[4 marks]
FEEDBACK: This question was generally quite well answered by everyone who made a
serious attempt at it.
12
C13.(i) Since E
1
, E
2
measurable, we may use the criterion with E = E
1
and A = E
1
E
2
.
We have E A = E
1
and E
c
A = E
2
and so
(E
1
) +(E
2
) =

(E
1
) +

(E
2
) =

(E
1
E
2
) = (E
1
E
2
).
[5 marks]
(ii) Let
E
1
= E [0, 1 x), E
2
= E [1 x, 1).
These are measurable and disjoint. By part (i),
(E) = (E
1
) +(E
2
).
We have
x E
1
= x +E
1
so x E
1
is measurable and
(x E
1
) = (x +E
1
) = (E
1
).
Also
x E
2
= (x 1) +E
2
so x E
2
is measurable and
(x E
2
) = ((x 1) +E
2
) = (E
2
).
Thus
x E = (x E
1
) (x E
2
)
(disjoint union) is measurable and
(x E) = (x E
1
) +((x E
2
) = (E
1
) +(E
2
) = (E).
[7 marks]
(iii) Dene an equivalence relation on [0, 1) by
x y x y Q.
By the Axiom of Choice, we can choose a set E consisting of exactly one element from
each equivalence class.
Claim: E is not measurable.
Justication: Since Q is countable, we can write
Q [0, 1) = r
n

n=1
.
13
Consider x r
n
E, for some n. Then either
x = r
n
+e
n
or x = r
n
+e
n
1 ()
for some e
n
E.
Suppose
x (r
n
E) (r
m
E).
By (*),
e
n
e
m
Q.
Since E has one element from each equivalence class, e
n
= e
m
. By (*) again, r
n
= r
m
(since both are in [0, 1)). Hence
(r
n
E) (r
m
E) ,= = r
n
E = r
m
E
(disjointness).
Since E contains an element from every equivalence class
[0, 1) =

_
n=1
(r
n
E)
(disjoint union). If E is measurable then so is every r
n
E, so
1 = ([0, 1)) =
_

_
n=1
(r
n
E)
_
=

n=1
(r
n
E).
By part (ii), each summand is equal to (E). If (E) = 0, the sum is zero; if (E) > 0
the sum is innite. Either case gives a contradiction. Therefore E is not measurable.
[13 marks]
FEEDBACK: Some students had problems with giving accurate solutions to parts (i)
and (ii) but I was pleased to see that part (iii) was generally well answered.
14

Você também pode gostar