Problem Set 1

Real Analysis I
Bennett Rennier
barennier@gmail.com
January 15, 2018

Exercise 1.

a) For sequences (xn ) , (yn ) of real numbers, prove that

lim sup{xn } + lim inf{yn } ≤ lim sup{xn + yn } ≤ lim sup{xn } + lim sup{yn }

b) Give a specific example where both inequalities are strict.

Proof.

a) If α = lim sup{xn }, then for an  ≥ 0, there are an infinite number of j’s where xj ≤ α− 2 .
If there were only a finite number of such j’s, then let n be the last one. This would
mean that sup{xk } ≤ α − 2 , which would mean that lim sup{xn } ≤ α − 2 < α. This is a
k≥n
contraction. If β = lim inf{yn }, then for any  > 0, there is an N , such that for all n ≥ N

yn ≥ inf {yk } ≥ β −
k≥n 2
This means there are an infinite number of n where xn > α − 2 and yn ≥ β − 2 . Thus,
for n ≥ N , it means that xn + yn ≥ α + β −  for inifinitely many n. This means that
sup{xn + yn } ≥ α + β − . Since  was arbitrary, this means that lim sup{xn + yn } ≥
lim sup{xn } + lim inf{yn }. This proves the first inequality.
Let xj ∈ {xk }k≥n and yj ∈ {yk }k≥n . Then, this means that xj ≤ sup{xk } and that
k≥n
yj ≤ sup{xk }. Thus, xj + yj ≤ sup{xk } + sup{yk }. If you take the sup off all possible
k≥n k≥n k≥n
j’s of both sides, you get sup{xj + yj } ≤ sup{sup{xk } + sup{yk }}. The RHS was already
j≥n j≥n k≥n k≥n
”supped” and therefore was just a constant, so the additional sup doesn’t do anything.
Thus sup{xj + yj } ≤ sup{xk } + sup{yk }. Taking the limit as n → ∞, one gets the second
j≥n k≥n k≥n
inequality.

1
 ( (
n if n is odd −n if n is even
b) Let y = and xn = .
−n if n is even 1 if n is odd
Computing the the limsups and liminfs, we get
lim sup{xn } + lim inf{yn } = 1 − ∞ = ∞
(
0 if n is even
lim sup{xn + yn } = 0 as xn + yn =
1 − n if n is odd
lim sup{xn } + lim sup{yn } = 1 + ∞ = ∞
As can be seen, the inequalities are strict for these xn and yn .

Exercise 2. a) Let p > 1 be an integer and x a real number with 0 < x < 1. Show that
there is a sequence of integers (an ) with 0 ≤ an < p such that
∞
X an
x=
n=1
pn

q
b) Show that the above sequence an is unique except when x is of the form pn
for some
integer q and that, in this case, there are exactly two such sequences.
c) Show, conversely, that if an is any sequence of integers with 0 ≤ an < p, the series
∞
X an
n=1
pn

converges to a number x with 0 ≤ x ≤ 1.

Proof. a) (This proof needs to be cleaned up.) Look at the partial sum Sk = kn=1 apnn . For
P
a1 , let it be the max number such that ap1 ≤ x and that a1 < p. This definitely exists as
if a1 = 0, then ap1 = 0 ≤ x. Now let an be the max number such that Sn−1 + apnn ≤ x and
that an < p. This definitely exists as if an = 0, then Sn−1 + p0n = Sn−1 ≤ x and each Sk
for k < n has already been constructed to be less than x. By part (c) of this problem,
every such sequence must converge, and by construction, this sequence must converge to
something less than or equal to x. Suppose it converged to x −  for  > 0. Look at 1 > 0.
There must exist a pk for k ∈ Z+ , such that 1 < pk (as p > 1). Let k be the least such
integer. This means that  > p1k . We see that by Sk + p1k , we’ve altered the sequence an
by adding 1 to ak . If ak = p − 1, then add 1 to ak+1 . If aj = p − 1 for all j ≥ k, then by
part (b), this means that (an ) converges to the same thing as a sequence which is 0 for
all j ≥ k. Take this new sequence to be the new an and add 1 to its kth position. Let
l be whatever position we eventually added the 1, and (bn ) be the sequence that we’ve
obtained by adding 1 to al . Note that l ≥ k. Then (bn ) converges to x −  + p1l > x − .
Also, we know that x −  + p1l ≤ x −  + p1k < x, as  > p1k . This is a contradiction, as
(an ) was constructed to be the greatest sequence such that it converged to something less
than or equal to x. This must mean that an is converged to x all along.

2
b) (This proof is also partially incomplete.) Essentially, let (an ) and (bn ) be sequences that
are not equal, but converge to the same x. Then, let k be the first position that they
differ. If they differ by more than 1 in the kth position, then they differ by p2k so far,
which is more than the rest of the tail combined can be. Thus, they converge to different
things, a contradiction. If they only differ by 1, then look at the k + 1th place. Say that
(bn ) was the sequence that was higher in the kth position, without loss of generality. If
bk+1 6= 0 and ak+1 6= p − 1, then their difference at this position is at most pp−2 k+1 . Their
1 p−2 p−p+2 2
difference so far would be at least pk − pk+1 = pk+1 = pk+1 , and as mentioned before, this
is too great of a distance for the rest of the tail after the k + 1th position to overcome.
Repeat this process for the proceding positions. (TODO: Next, show that if this does
occur, then they converge to the same thing.) Since this only happens to x when one
of its representations is all 0’s after a certain position, this means that x = Sn for some
n ∈ N, as the rest of the sequence contributes nothing to the partial sum. We see that
k−1 a
Sk = apkk + apk−1
k−1
· · · + ap1 = ak +pak−1p···+p
k
1
. Thus, x = pqk for some integers q and k

c) Firstly, since apnn ≥ 0 for all n, we see that the partial sums Sk = kn=1 apnn are monotoni-
P
cally increasing. Thus, we need only to prove that P∞ it’s boundedPby 0 and 1. We see that
∞ an P∞
0 ≤ pn ≤ pn . Thus, by the comparison test, n=1 0 = 0 ≤ n=1 pn ≤ n=1 p−1
an p−1
pn
. We
P∞ p−1 P∞  1  P
∞
 
∞
see that n=1 pn = n=1 pn−1 − p1n = n=0 p1n − p1n + p10 = 1. Thus n=1 apnn ≤ 1.
P

Since this means that Sk is increasing and bounded, this means that Sk converges to a
number between 0 and 1.

Exercise 3. Let f be a mapping from a set X to the set of subsets of X. By considering

the set E = {x ∈ X : x 6∈ f (x)}, show that there is a subset of X which is not in the range
of f .
 
Proof. Let f : X → P (X) and let E = x ∈ X  x 6∈ f (x) . Assume that f is surjective.
Since E ∈ P (X), then there exists an x0 such that f (x0 ) = E. Now, x0 must be either in E or
not in E. Say x0 ∈ E, then x0 ∈ E = f (x0 ). However, since x0 ∈ E, this means by definition
of E that x0 6∈ f (x0 ). Thus, we have a contradiction. Now say x0 6∈ E, then x0 6∈ E = f (x0 ).
But, by definition of E, this means that x0 ∈ E. This is also a contradiction. Thus, there is
no x ∈ X such that f (x0 ) = E, which means f is not surjective. This shows there’s a subset
of X which is not in the range of f .

Exercise 4. If S ⊆ X is uncountable and A ⊆ X is countable, show that S ∩ Ac is un-

countable.

Proof. Assume that S ∩Ac is countable. Then (S ∩ Ac )∪A is the union of two countable sets,
and is thus countable. By distributivity, (S ∩ Ac ) ∪ A = (S ∪ A) ∩ (A ∪ Ac ) = (S ∪ A) ∩ X =
S ∪ A. S ∪ A is clearly uncountable, as S ⊆ S ∪ A. This is a contradiction. Thus, S ∩ AC
must be uncountable.

Exercise 5. a) Is the set of rationals open or closed in the set of real numbers?

3
b) Which sets of real numbers are both open and closed?

Proof. a) Every open interval on the real line contains both rationals and irrationals. There-
fore, every open neighborhood of q ∈ Q contains irrationals. This shows that Q is not
open. Similarly, if Q were closed, then R \ Q would be open. But, every open neighbor-
hood of i ∈ R \ Q contains a rational number. Therefore, R \ Q is not open, and Q is not
closed. This means that Q is neither open nor closed.

b) ∅ is trivally open and closed. Similarly, it’s complement, R is then open and closed.
Assume that there exists an additional open and closed set ∅ 6= A 6= R. Then the
complement Ac would also be open and closed, and not equal to ∅ or R. Now fix
a ∈ A and b ∈ Ac . Without loss of generality (since (Ac )c = A), let a < b. Now let
C = {x ∈ R  [a, x] ⊆ A}. C is nonempty, as a ∈ C. Also, C is bound above by b, as
b 6∈ A. Thus, C has a least upper bound. Let’s call it α.
Say α ∈ A, then, since A is open, there’d be a ball of radius r >
 0, suchthat (α−r, α+r) ⊆
A. Then [α, α + 2r ] ⊆ (α − r, α + r) ⊆ A. This means [a, α] ∪ α, α + 2r = a, α + 2r ⊆ A.


This means that α + 2r ∈ C, a contradiction, as α was the least upper bound of C. This
means that α 6∈ A.
Say α ∈ Ac , then, since Ac is open, there’d be a ball of radius r > 0, such that (α −
r, α + r) ⊆ Ac . Then [α − 2r , α] ⊆ (α − r, α + r) ⊆ Ac . This means that α − 2r 6∈ A, which
means α − 2r 6∈ C. However α − 2r < α and α is supposed to be the least upper bound.
This is a contradiction, thus α 6∈ Ac . This means that α ∈ A. From the last paragraph,
we proved that α 6∈ A. This contradiction shows that there is no such set A. This means
that ∅ and R are the only two open and closed sets in R.

Exercise 6. Prove that a set X is infinite if and only if there is a proper subset of X of the
same cardinality as X.

Proof. Suppose X is infinite. Then create an injective function φ : N → X. It’s easy to see
such a function exists, as you can choose an element of X for each n ∈ N, and you’ll never
run out of elements, as that would mean that there was a bijection between {1 . . . n} and
X. From this injective function let ψ : X → X where
(
φ (n + 1) x = φ (n) for some n ∈ N
ψ=
x x 6∈ Im(φ)

One can see that φ (0) 6∈ Im(ψ), so Im(ψ) ( X, and also that ψ is injective, as φ was
injective. Combined with the inclusion map from Im(ψ) to X, which is necessarily injective,
we see that there’s a bijection between X and Im(ψ) ( X. This assumes the Axiom of
Choice.
Let X be a set with a proper subset A. This means there exists an x ∈ X, such that
x 6∈ A. Let’s say they have the same cardinality. This means that there’s a bijection
φ : X → A. Construct the function ψ : N → X as follows: ψ (0) = a and ψ (i + 1) = φ(ψ(i))
for all 0 6= n ∈ N. (Tip: It might be more intuitive to view it as a sequence.) Since φ is

4
injective and a 6∈ Im(φ), we see that ψ is injective. We also see that through the inclusion
map, {1 . . . n} is injective to N. If the cardinality of X were finite, then there’d be a bijection
f : X → {1 . . . n} for some n ∈ N. Thus, the composition, f ◦ ψ : N → {1 . . . n} would
be injective, and thus there’s a bijection between N and {1 . . . n}. This is a contradiction,
which means X must be infinite.