Ps 8 Sol

d-math Analysis II: Several Variables ETH Zürich
Federico Franceschini Problem set 8 FS 2024
Problems marked with a (*) are a bit more complex and can be skipped at a first read.
If you don’t have a lot of time focus on the Problems/subquestions marked with (♡).
8.1. BONUS PROBLEM. Let X ⊂ R be a Jordan-null set (as in Definition 13.8).

(a) Show rigorously that X × X ⊂ R2 is also Jordan-null.
(b) Show rigorously that X × [0, 1] ⊂ R2 is also Jordan-null.
8.2. True or False. (♡)

1. A bounded countable set is always Jordan-null.
2. A countable set is always Lebesgue-null.
3. Let D ⊂ [0, 1] be a dense set (i.e., D = [0, 1]). Then µout (D) = 1. (µout was defined
in Definition 13.7).
4. Let X, Y ⊂ [0, 1] Jordan measurable sets such that µ(X) > 1/2 and µ(Y ) > 1/2.
Then X ∩ Y ̸= ∅.
5. Let X, Y ⊂ [0, 1] such that µout (X) > 1/2 and µout (Y ) > 1/2. Then X ∩ Y ̸= ∅.
8.3. Fat Boundary. Construct an open subset U ⊂ R for which the boundary ∂U is
not a null set.
8.4. Multiple Choice. (♡) Let U ⊂ Rn be a nonempty, open subset, f : U → Rm a

function, and N ⊂ U a Jordan null set. In which of the following cases is the image
f (N ) ⊂ Rm necessarily a null set? Attention: Only one answer is correct!
1. If f is uniformly continuous.
2. If f is uniformly continuous and m ≥ n.
3. If f is locally Lipschitz continuous.
4. If f is locally Lipschitz continuous and m ≥ n.
8.5. Change of variables and Jacobians. (♡) For each of the following domains and
change of variables find the jacobian and the appropriate transformed domain. There is
no need to actually compute the integrals!
1. A := {(x1 , x2 ) ∈ R2 : x1 > 0, x2 < x1 , 1 < x21 + x22 < 4} and x1 = r cos θ, x2 = r sin θ.
Using the change of variables formula, complete the dots in the following formula
ˆ ˆ
2
x1 sin(x2 ) dx1 dx2 = ... drdθ
A ...
assignment: April 15, 2023 due: Apr 22, 2023 1/8

2. B := {(x, y) ∈ R2 | 1 < xy < 2, x2 < y < 2x2 } and u := xy, v := x2 . Using the
change of variables formula, complete the dots in the following formula
ˆ ˆ
2 −xy
y e dxdy = ... dudv
B ...
3. C := {(x, y, z) ∈ R3 | 1 < z − 2y < 2, 0 < z < 1, −2 < 3x + y + z < 0} and

u := z, v := z − 2y, w := 3x + y + z. Using the change of variables formula, complete
the dots in the following formula
ˆ ˆ
xyz dxdy = ... dudvdw.
C ...
8.6. The Cantor set. (*) Let X ⊂ [0, 1] be the set of all real numbers whose decimal
expansion does not contain the digit 8.1 Show that:
1. X is a Lebesgue null set,
2. X is uncountable,
3. X × X ⊂ [0, 1]2 is a Lebesgue null set.
4. Show that X is compact (it is important the choice made in the footnote!).
1
The decimal expansion is not always unique. For example, 0.8 = 0.79999 . . .. Whenever x has at least
one decimal expansion not containing 8, we rule that x belongs to X, so for example 0.32579 ∈ X,
0.32589 ∈ X

Hints:
8.2.5 µout can be positive and large on very sparse sets...
8.3 Any open subset of R is an union of disjoint open intervals. Try to achieve that U
has a very small “total volume”, but still contains all rational numbers in [0, 1].

8. Solutions
Solution of 8.1: For ϵ > 0 there is a finite collection of dyadic intervals {I1 , . . . , IN }
such that X
X ⊂ I1 ∪ . . . ∪ IN , µ1 (Ii ) ≤ ϵ.
1≤1≤N
Now we must have

X × [0, 1] ⊂ I1 × [0, 1] ∪ . . . ∪ IN × [0, 1],
and so X X
µout (X × [0, 1]) ≤ µ2 (Ii × [0, 1]) = µ1 (Ii ) ≤ ϵ.
1≤i≤N 1≤1≤N
Since ϵ was arbitrary, µout (X × [0, 1]) = 0. Since without loss of generality we may have
assumed that X ⊂ [0, 1] then
µout (X × X) ≤ µout (X × [0, 1]) = 0.
Solution of 8.2:
1. False, Q is acounterexample.
2. True, number it’s elements {ak }k∈N and cover it with {ak + ϵ2−k [−1, 1]}k∈N .
3. True, pick any finite dyadic partition of [0, 1]. Then each interval must contain at
least one element of D by density. Hence if we want to cover D we must keep all
these intervals, whose length sums to 1.
4. True. By contradiction, assume that X ∩ Y = ∅. By Theorem 13.18, we must have
µ(X ∪ Y ) = µ(X) + µ(Y ) > 1. On the other hand, since X ∪ Y ⊆ [0, 1], it holds
µ(X ∪ Y ) ≤ µ([0, 1]) = 1, which gives the desired contradiction.
√
5. False, take X = Q ∩ [0, 1] and Y := ( 2 + Q) ∩ [0, 1].
Solution of 8.3: Let 0 < ϵ < 1/2 and (qn )n∈N be a enumeration of [0, 1] ∩ Q. We define
In := (qn − ϵ2−n , qn + ϵ2−n ) and
[
U := In .
n∈N
By construction, we then have [0, 1] ∩ Q ⊂ U and the “total volume”2 is small, in the
sense that
∞ ∞
2−n = 2ϵ.
X X
vol(In ) = 2ϵ (1)
n=1 n=1
We now show that this set U satisfies the requirements in the problem statement. Being a
union of open intervals, U is certainly open in R. Next, we notice that due to [0, 1]∩Q ⊂ U ,
2
We always write this in quotation marks because the point of this exercise is that this set U is not
Jordan-measurable.

the entire interval [0, 1] is contained in the closure U of U . Using the boundary of U , this
can be written as
∂U ∪ U = (U \ U ) ∪ U = U ⊃ [0, 1].
In view of (1), it is therefore natural to conjecture that for a covering of ∂U by open

boxes (Ql )l in R (that is, by intervals), it is necessary that
∞
X
vol(Ql ) ≥ 1 − 2ϵ (2)
l=1
must hold. To prove this lower bound, we first notice that for such a covering of the
boundary
[ [
[0, 1] ⊂ ∂U ∪ U ⊂ In ∪ Ql
n∈N l∈N
is an open covering of the compact interval [0, 1]. Therefore, finitely many of these sets
suffice to cover it; thus, there exist N, L ∈ N such that
N
[ L
[
[0, 1] ⊂ In ∪ Ql .
n=1 l=1
Both the In and the Ql are open intervals, whose volume is computed as the difference
of the right and left endpoints. Elementary considerations3 show that therefore for the
volumes, it holds that
N
X L
X
1≤ vol(In ) + vol(Ql ).
n=1 l=1
From this, using (1), we now obtain

∞
X L
X N
X ∞
X
vol(Ql ) ≥ vol(Ql ) ≥ 1 − vol(In ) ≥ 1 − vol(In ) = 1 − 2ϵ,
l=1 l=1 n=1 n=1
exactly as in (2). Since ϵ was chosen such that 1 − 2ϵ > 0, this shows that ∂U cannot be
a null set.
Solution of 8.4: The only correct one is number 4 which is Lemma 13.6.
Solution of 8.5:
3
If an interval [a, b] is covered by finitely many intervals J1 , . . . , Jr , ordered increasingly by their left
Pk
endpoints, then the right endpoint of Jk is at most a + j=1 vol(Jj ). Therefore, for the right endpoint
Pr
b to also lie in one of the intervals Jj , it must hold that j=1 vol(Jj ) ≥ b − a.

1. If x1 = r cos θ, x2 = r sin θ then we have
(x1 , x2 ) ∈ A ⇐⇒ (r, θ) ∈ {−π < θ < π, sin θ < cos θ, 1 < r < 2}
solving the inequality
−π < θ < π, sin θ < cos θ ⇐⇒ −π < θ < π/4.
So, recalling that the jacobian is r, we find

ˆ ˆ 2 n ˆ π/4 o
2
x1 sin(x2 ) dx1 dx2 = cos2 θ sin(r sin θ)dθ r3 dr.
A 1 −π
2. Notice that B ⊂ (0, ∞)2 so (x, y) are always positive and so are (u, v). We write
the inverse mappings
√ u
x = v, y = √ .
v
So we compute the determinant of
" 1 #
∂(x, y) 0 √
2 v
det = det √1
∂(u, v) v
∗
and so we find
du dv
dx dy = .
2v
The set B changes into
√ √
(x, y) ∈ B ⇐⇒ (u, v) ∈ {1 < u < 2, v v < u < 2v v},
⇐⇒ (u, v) ∈ {1 < u < 2, u2/3 < v < 2−2/3 u2/3 },
and the integrand y 2 e−xy changes into

u2 u
e
v
so summing up we find
ˆ ˆ 2 n ˆ 2−2/3 u2/3 o
2 −xy
y e dxdy = u 2
v −2 ev dv du.
B 1 u2/3
3. We compute the determinant

 
0 0 1
∂(u, v, w)
det = det 0 −2 1 = 6,
 
∂(x, y, z)
3 1 1
so dx dy dz = 61 du dv dw. And the set becomes
(x, y, z) ∈ C ⇐⇒ 0 < u < 1, 1 < v < 2, −2 < w < 0.

To change variables in the function we need to compute the inverse mapping:
x = 31 w + 16 v − 12 u, y = 12 u − 21 v, z = u.
So we find
xyz = − 16 uvw + 16 u2 w + 13 u2 v − 1
12
uv 2 − 14 u3 .
So putting everything together
ˆ ˆ 1ˆ 2ˆ 0
1
xyz dx dy dz = v (−2uvw + 2u2 w + 4u2 v − uv 2 − 3u3 )du dv dw.
C 72 0 1 −2
Solution of 8.6:
1. Let An be the set that does not contain 8 up to the n-th decimal place. For example,
A1 = [0, 0.8] ∪ [0.9, 1].
The intervals that decompose A1 have total volume 0.9. The second set has form
A2 = [0, 0.08] ∪ [0.09, 0.18] ∪ · · · ∪ [0.69, 0.78] ∪ [0.79, 0.8] ∪ [0.9, 0.98] ∪ [0.99, 1].
The intervals that decompose A2 have total volume 9 · 0.09 = 0.81 (there are a total
of 11 intervals: 7 have length 0.09; and 4 intervals are shorter, but one finds 2 pairs
of 2 intervals each, which together also have length 0.09). We try to find a pattern
and consider A3 . Treating the 4 shorter intervals in A2 as if we had 2 intervals of
length 0.09, we see that each interval in A2 is divided into 9 intervals of length 0.009
(again by combining shorter intervals). Thus we have vol(A3 ) = 0.009 · 81 = 0.93 .
Analogously, we find that An consists of finitely many intervals whose total volume
vol(An ) = 0.9vol(An−1 )
which implies that vol(An ) = 0.9n .

To show that X is a Lebesgue null set, let ϵ > 0. Choose n ∈ N large enough such
that 0.9n < ϵ. Since An consists of finitely many intervals covering X and the sum
of the lengths of these intervals is 0.9n < ϵ, the claim follows. As a complement for
those who prefer explicit formulas: Let (Ink )sk=1
n
be the intervals defining An . Then
we have s s
n n
Ink and vol(In ) = 0.9n < ϵ.

[ X
X ⊂ An =
k=1 k=1
This shows that in fact X is also Jordan-null set.

2. If X were countable, we would write all numbers in a countably infinite list. As
in the Cantor diagonal argument, we take a number defined by the diagonal and
change each digit to a digit that is still not 8. This number is still in X, but does
not coincide with any number in our list. Contradiction! So X is uncountable.

3. We can cover X ×X by the boxes in An ×[0, 1]. We have vol(An ×[0, 1]) = vol(An ) =
0.9n and the argument is analogous to part (a).
4. Each of the {Ai } of step 1 is compact, being the union of finitely many closed
disjoint intervals in [0, 1]. Our set by definition is
[
X := Ai ,
i≥1
which is closed (being an intersection of closed sets) and bounded (since A1 is

bounded). Hence X is compact.

Ps 8 Sol

Uploaded by

Copyright:

Available Formats

Ps 8 Sol

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ps 8 Sol

Uploaded by

Copyright:

Available Formats

d-math Analysis II: Several Variables ETH Zürich

Federico Franceschini Problem set 8 FS 2024

8.1. BONUS PROBLEM. Let X ⊂ R be a Jordan-null set (as in Definition 13.8).

8.2. True or False. (♡)

8.4. Multiple Choice. (♡) Let U ⊂ Rn be a nonempty, open subset, f : U → Rm a

assignment: April 15, 2023 due: Apr 22, 2023 1/8

3. C := {(x, y, z) ∈ R3 | 1 < z − 2y < 2, 0 < z < 1, −2 < 3x + y + z < 0} and

assignment: April 15, 2023 due: Apr 22, 2023 2/8

assignment: April 15, 2023 due: Apr 22, 2023 3/8

Now we must have

µout (X × X) ≤ µout (X × [0, 1]) = 0.

assignment: April 15, 2023 due: Apr 22, 2023 4/8

In view of (1), it is therefore natural to conjecture that for a covering of ∂U by open

From this, using (1), we now obtain

assignment: April 15, 2023 due: Apr 22, 2023 5/8

1. If x1 = r cos θ, x2 = r sin θ then we have

solving the inequality

−π < θ < π, sin θ < cos θ ⇐⇒ −π < θ < π/4.

So, recalling that the jacobian is r, we find

and the integrand y 2 e−xy changes into

3. We compute the determinant

so dx dy dz = 61 du dv dw. And the set becomes

(x, y, z) ∈ C ⇐⇒ 0 < u < 1, 1 < v < 2, −2 < w < 0.

assignment: April 15, 2023 due: Apr 22, 2023 6/8

To change variables in the function we need to compute the inverse mapping:

A1 = [0, 0.8] ∪ [0.9, 1].

which implies that vol(An ) = 0.9n .

Ink and vol(In ) = 0.9n < ϵ.

This shows that in fact X is also Jordan-null set.

assignment: April 15, 2023 due: Apr 22, 2023 7/8

which is closed (being an intersection of closed sets) and bounded (since A1 is

assignment: April 15, 2023 due: Apr 22, 2023 8/8

You might also like