MAT 3330/4300: Solutions for problems assigned week 39 and 40.

Exercise 5.O.
∞
P ∞
P
Let E = {x : |fn (x)| = ∞}. Since |fn | is measurable, it follows that E ∈ X.
n=1 n=1
Let N be a positive integer. It follows from Monotonee Convergence Theorem that
∞ Z
X Z X ∞ Z
∞> |fn |dµ = ( |fn |)dµ ≥ N χE dµ = N µ(E).
n=1 n=1

∞
P
It follows from this inequality that µ(E) = 0. Let f (x) = fn (x) for x ∈ X − E
n=1
k
P
and f (x) = 0 for x ∈ E. Let gk (x) = fn (x) for x ∈ X − E and gk (x) = 0 for x ∈
n=1
∞
P ∞
P
E and g(x) = |fn (x)| for x ∈ X − E and g(x) = 0 for x ∈ E. Since g = |fn |
n=1 n=1
R ∞ R
P
µ-almost everywhere it follows from Corollary 4.10 that gdµ = |fn |dµ < ∞,
n=1
so g is integrable. Since |gk | ≤ g, every gk isR also integrabel.
R From Lebesgue
R
Dominated Convergence Theorem we get that f dµ = lim gk dµ = lim gk dµ.
Pk R
On the other hand we have that gk = fn µ-almost everywhere, so gk dµ =
n=1
k R
P R R ∞ R
P
fn dµ and therefore we get that f dµ = lim gk dµ = fn dµ. Moreover
n=1 n=1
∞
P
it is clear that the series fn converges to f on X −E so it converges to f µ-almost
n=1
everywhere. (Note that since an integrable function by definition only takes (finite)
real values, we cannot use Lebesgue Dominated Convergence Theorem directely to
Pk
the sequence fn since this sequence not necessarily is dominated by an integrable
n=1
function with finite values. Therefore we need all the above definitions to guarantee
that our functions only have real values.).
Exercise 5.P.
R have that lim R|fn | = |f |, so from Fatous Lemma we get that
We R
R |f |dµ ≤ lim
R inf |fn |dµ. Since |fn | ≤ |f | + |fn −R f |, we have that R|fn |dµ ≤
R |f |dµ + |fn −R f |dµ. This give usR that lim sup |fn |dµ ≤ lim sup( |f |dµ +
|fn − f |dµ) = |f |dµ (because lim |fn − f |dµ = 0). So, we consequently get
that: Z Z Z Z
|f |dµ ≤ lim inf |fn |dµ ≤ lim sup |fn |dµ ≤ |f |dµ.
R R
Which shows that lim |fn |dµ = |f |dµ.
Exercise 5.Q.
We have that
Z ∞ Z b b
1 −tx  1 1 1
tx
e dx = lim e dx = lim − e  = lim [ − e−tb ] = .
tx
0 b→∞ 0 b→∞ t 0 b→∞ t t t

Since s → es is strictly increasing e−tx ≤ e−ax for t ≥ a > 0 and x ≥ 0. Let

X = [0, ∞), B the Borel sets on X and let λ be the Lebesque measure on B.
1
2

ax
Let 1 > a > 0, andRlet g(x) = e− 2 . From exercise 4.M and the calculation
above, we have that gdλ = a2 , so g is integrable. Let gn (x) = xn e−ax . Since
lim gg(x)
n (x)
= lim xn g(x) = 0(this is a well known limit), gn is also integrable
x→∞ x→∞
for each n. For t ∈ [a, 1] let fn (x, t) = xn e−tx . Since |fn (x, a)| = gn (x) and
| ∂f
∂t (x, t)| = fn+1 (x, t) ≤ gn+1 (x),
n
R the hypothesis of Corrolary 5.9 is satisfied. We
consequently get that Fn (t) = fn (x, t)dλ(x) is differentiable and
Z Z Z ∞
dFn ∂fn n+1 −tx
= (x, t)dλ(x) = −x e dλ(x) = −xn+1 e−tx dx,
dt ∂t 0

where we have used the conclusion of 5.9 and exercise 4.M. Putting n = 0, 1, 2, . . . ,
we get from all this:
Z ∞
1
F0 (t) = e−tx dx =
t
Z 0∞
dF0 d 1 1
= −xe−tx dx = −F1 (t) = ( ) = − 2
dt dt t t
Z0 ∞
dF1 d 1 2
= −x2 e−tx dx = −F2 (t) = ( 2 ) = − 3
dt dt t t
Z0 ∞
dF2 d 2 3·2
= −x3 e−tx dx = −F3 (t) = ( 3 ) = − 4 . . .
dt 0 dt t t
R∞
n!
After n-iterations we get that tn+1 = 0 xn e−tx dx. Letting t = 1 we get the
formula Z ∞
n! = xn e−x dx.
0

Exercise 5.R.
Firstly, assume that t0 6= t1 and let t 6= t0 and also let t 6= t1 . Then we have that

|f (x, t)| ≤ |f (x, t0 )| + |f (x, t) − f (x, t0 )|

≤ |f (x, t0 )| + |f (x, t) − f (x, t1 )| + |f (x, t1 ) − f (x, t0 )|
≤ |f (x, t0 )| + |t − t1 |g(x) + |t1 − t0 |g(x).

Since |f (x, t0 )| + |t − t1 |g(x) + |t1 − t0 |g(x) obviously is integrable, we have that

x → f (x, t) is integrable for every such t. In the same manner, for t = t1 we have
that |f (x, t1 )| ≤ |f (x, t0 )| + |t1 − t0 |g(x). If t0 = t1 then |f (x, t)| ≤ |f (x, t0 )| +
|t − t0 |g(x). So for all t the function x → R f (x, t) is bounded by integrable functions
and is therefore integrable. Let F (t) = f (x, t)dµ(x). Let (sn ) be a sequence such
that sn → t1 and assume that sn 6= t1 . Then F (ssnn)−F (t1 )
= f (ssnn)−f (t1 )
R
−t1 −t1 dµ(x), and
since | f (ssnn)−f
−t1
(t1 )
| ≤ g(x) Lebesgue Dominated Convergence Theorem will imply
that
F (sn ) − F (t1 ) f (sn ) − f (t1 )
Z Z
∂f
lim = lim dµ(x) = (x, t1 )dµ(x).
n→∞ sn − t1 n→∞ sn − t1 ∂t

Here, the right hand side is independent of the choice of sequence

R ∂f (sn ) so F (t) is
dF
differentiable at t1 , and we consequently get that dt (t1 ) = ∂t (x, t1 )dµ(x).
 3

Exercise 5.S.
R∞
Since the improper Riemann integral |f (x, t)|dt exists for each x, the integral
−∞
R∞ R∞ Rn
f (x, t)dt will also exist, and we have that f (x, t)dt = lim f (x, t)dt.
−∞ −∞ n→∞ −n
Rn Rn
Put gn (x) = f (x, t)dt. Since every Riemann sum for the integral f (x, t)dt only
−n −n
is a linear combination of measurable functions x → f (x, t), such Riemann sums
become measurable functions on X, and since gn is a limit of such Riemann sums,
R∞
each gn is a measurable function. Then, it also follows that lim gn (x) = f (x, t)dt
−∞
is a measurable function on X. Moreover it is clear thar |gn (x)| ≤ h(x) for all n.
Lebesgue Dominated Convergence Theorem therefore implies that

Z Z Z∞
lim gn (x)dµ(x) = [ f (x, t)dt]dµ(x)
−∞
that is
Z Zn Z Z∞
(*) lim [ f (x, t)dt]dµ(x) = [ f (x, t)dt]dµ(x).
n→∞
−n −∞

R∞
(∗) shows that x → f (x, t)dt is integrable. Letting [a, b] = [−n, n] the ass-
−∞
sumptions of Corollary 5.10 are satisfied, so from the conclusion of 5.10 we get
that Z n Z Z Z n
[ f (x, t)dµ(x)]dt = [ f (x, t)dt]dµ(x).
−n −n

letting n → ∞ and using (∗), we get that

Z n Z Z Z∞
lim [ f (x, t)dµ(x)]dt = [ f (x, t)dt]dµ(x).
−n
−∞

If
R ∞f (x,
R t) was non-negativthen this will show that the improper Riemann integral
−∞
[ f (x, t)dµ(x)]dt exsits and we have that

Z ∞ Z Z Z∞
[ f (x, t)dµ(x)]dt = [ f (x, t)dt]dµ(x).
−∞
−∞

The improper Riemann integral is however defined by letting the lower and upper
limit in the integral approach −∞ and ∞ independently, so dealing with functions
of
R 0 both
R positive and negative
R ∞ values,
R we actually must show that both the limits
−∞
[ f (x, t)dµ(x)]dt and 0 [ f (x, t)dµ(x)]dt exist. In our case this is however
obvious, because we can in the above arguments replace f (x, t) by |f (x, t)| and
R0 R
prove the conclusion for this function, and the integrals −∞ [ f (x, t)dµ(x)]dt and
4
R∞ R
0
[ f (x, t)dµ(x)]dt must then exist since they are dominated of similar integrals
of |f (x, t)| which then must be finite.
Problem 5.T.
For a measurable function f we define fn by

fn (x) = f (x), hvis |f (x)| ≤ n

= n, if |f (x)| > n
= −n, if |f (x)| < −n.

We have |fn | ≤ |f | and lim fn = f ,so if f is integrable then |f | is also integrable,

R
andR we will get from Lebesgue Dominated Convergence Theorem that f dµ =
lim fn dµ. Moreover we have that lim |fn | = |f |, so from Fatous Lemma we get
that Z Z Z
|f |dµ ≤ lim inf |fn |dµ ≤ sup |fn |dµ.
R
So if sup |fn |dµ < ∞, e |f |and therefore f is integrable.
Problem 6.A.
. If f ∈ C[0, 1 and α ∈ R, then |f (o)| ≥ 0 and |(αf )(0)| = |αf (0)| = |α||f (0)|, and
if f, g ∈ C[0, 1] then |(f + g)(0)| = |f (0) + g(0)| ≤ |f (o)| + |g(0)| and from all this
we get that N0 is a semi-norm.
Problem 6.B.
R1
Let f ∈ C[0, 1]. Since |f | ≥ 0, we get that N1 (f ) = |f |dx ≥ 0. If f (x) = 0 for
0
all x then it is clear that N1 (f ) = 0. If f ∈ C[0, 1] og and there exists x ∈ [0, 1]
such that f (x) 6= 0, then there exists an interval I ⊂ [0, 1] with length l(I) > 0
shuch that x ∈ I and |f (t)| > |f (x)| 2 for t ∈ I. Since |f | ≥ |f (x)|
2 χI , we will get
1
that N1 (f ) ≥ |f (x)| |f (x)|
R
2 χI dx = 2 l(I) 6= 0. Let α ∈ R. since the Riemannintegral
0
is linear we get that N1 (αf ) = |α|N1 (f ). Let f, g ∈ C[0, 1]. Then we have that
|f + g| ≤ |f |+|g|. Integrating this inequality we get that N1 (f +g) ≤ N1 (f )+N1 (g).
Alltogether this shows that N1 is a norm on C[0, 1].
Let m > n. Then we have that fn = fm on [0, (1 − n1 )/2] ∪ [ 21 , 1], so

1 1
Z 1 Z Z
2 2 1
N1 (fn − fm ) = |fn − fm |dx = |fn − fm |dx ≤ 2dx = .
0 1
(1− n )/2 1
(1− n )/2 n

Since n1 → 0 when n → ∞, it follows that (fn ) is a Cauchy sequence.

Assume now that there exists f ∈ C such that lim fn = f in the norm N1 , that is
lim N1 (f − fn ) = 0. Let m be a positive integer. Then
1
Z 1 Z (1− m )/2 Z 1
0 = lim |f − fn |dx ≥ lim |f − fn |dx + lim |f − fn |dx =
n→∞ 0 n→∞ 0 n→∞ 1
2
1 1
Z (1− m )/2 Z 1 Z (1− m )/2 Z 1
lim |f |dx + lim |f − 1|dx = |f |dx + |f − 1|dx.
n→∞ 0 n→∞ 1
0 1
2 2
 5

1
This shows that for every m we must have f (x) = 0 on [0, (1 − m )/2] and f (x) = 1
1 1 1
on [ 2 , 1]. So f becomes equal 0 on [0, 2 ) and equal 1 on [ 2 , 1], and such a function
f is not continous. So we have got a contradiction.
Problem 6.C.
From property (i) of a norm it follows that d(u, v) = N (u − v) ≥ 0. From property
(ii) it follows that d(u, v) = N (u − v) = 0 ⇔ u − v = 0 ⇔ u = v. From property
(iv) we get that if u, v and w are vectors in V , then

d(u, v) = N (u−v) = N ((u−w)+(w−v)) ≤ N (u−w)+N (w−v) = d(u, w)+d(w, v).

This shows that d is a metric.