Known Results in Superior and Inferior Limits

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Known results in superior and inferior limits

Goran Huygh

Prepared for, 30/12/2023

Abstract

We prove some elementary properties concerning superior and inferior limits proposed
by Terence Tao in his book Analysis I. These property concern superior and inferior
limits and how they relate to limit points and limits. We use these results to develop
and prove the well-known squeeze theorem.
Known results in superior and inferior limits Goran Huygh

1 Preliminary notions

We quickly introduce some preliminary definitions.

Definition 1. (Superior sequence). Given a sequence (an )∞


n=m we define a new sequence


(a+
N )N =m by the formula


a+
N := sup(an )n=N

Symmetrically we define (a− ∞


N )N =m . We then define the superior limit as follows.

Definition 2. (Limit superior) We define the superior limit of a sequence by the formula


L+ = inf(a+
N )n=m

Once again, we symmetrically define L− . We now also state and prove a lemma, a corollary

and a theorem that will be used later for the proofs.


Corollary 1. Given any sequence (an )∞
n=m we have that for any N ≥ m, aN ≤ aN
+

We don’t provide a proof for this corollary as it is quite trivial.

Page 1
Known results in superior and inferior limits Goran Huygh

∞ − ∞
Lemma 1. (a+
N )N =m is a monotonically decreasing sequence and (aN )N =m is monotonically

increasing one.


K = sup(an )n=K and thus for all n ≥ K we have that
Proof. For any K ≥ m we have a+

an ≤ sup(an )∞ + ∞
n=K . Now consider aK+1 = sup(an )n=K+1 , here once again we have for all

n ≥ K + 1 that an ≤ sup(an )∞ + +
n=K+1 . Suppose for the sake of contradiction that aK < aK+1

this would mean that for all n ≥ K we have

an ≤ sup(an )∞ ∞
n=K ≤ sup(an )n=K+1

But if this statement is true for all n ≥ K then it’s also true for all n ≥ K +1, but this means
∞ ∞
that a+ +
K bounds (an )n=K whilst being smaller than aK+1 , contradicting the fact that aK+1 is

the least upper bound of (an )∞


n=K+1 . Thus we must have that aK+1 ≤ aK thus proving that
+ +

∞ − ∞
(a+
N )N =n is monotonically increasing. The proof is symmetrical for (aN )N =n


Theorem 1. Given any sequence (an )∞
n=m we have that for any M, N ≥ m, aM ≤ aN
+

Proof. The case M = N is trivial, suppose that M > N . We already know that a−
M ≤ aM .
+

Using Lemma 1 we have that

a−
M ≤ aM ≤ aN
+ +

Giving us the desired statement. The proof is symmetrical for M < N .

Page 2
Known results in superior and inferior limits Goran Huygh

2 Inequalities concerning L+ and L−

In the following sections we use the the same notation for recurring elements. We denote a

arbitrary sequence (an )∞ + −


n=m and it’s superior limit L whilst we denote the inferior one L .

Theorem 2. inf(an )∞ − ∞
n=m ≤ L ≤ L ≤ sup(an )n=m
+

Proof. We first prove L+ ≤ sup(an )∞ ∞


n=m , from which the symmetric proof of inf(an )n=m ≤ L

follows. By definition we have

L+ = inf(a+
N)

and thus

∀N, L+ ≤ a+
N

Including N = m, thus giving us


L + ≤ a+
m = sup(an )n=m

We now only need to show that L− ≤ L+ . We will proceed by contradiction. Suppose

L− > L+ , then by the definition of infimum and supremum we know there must exists a N

and M such that


− −
L + ≤ a+
N < aM ≤ L

But this contradicts Theorem 1, thus proving that L− ≤ L+ and that the whole inequality

holds.

Page 3
Known results in superior and inferior limits Goran Huygh

Theorem 3. If c is a limit point of (an )∞ −


n=m , then we have L ≤ c ≤ L
+

Proof. We will proceed by contradiction by assuming that L+ < c to show that c ≤ L+ .(The

proof for L− ≤ c is symmetrical). First of all notice that if L+ = +∞ we have an immediate

contradiction, so we may suppose L+ = −∞ or L+ = r with r ∈ R. By denseness of the

reals there exists a d such that

L+ < d < c

If L+ = −∞ we define d := c − 1. By the definition of L+ we also have there exists a M ≥ m

such that for all µ ≥ M we have

L + ≤ aµ < d < c

We also have that c is a limit point meaning that

∀ε > 0, ∀N ≥ m, ∃n0 ≥ N : |c − an0 | ≤ ε

c−d
In particular for ε = 2
> 0 and N = M we have

∃µ0 ≥ M : |c − aµ0 | ≤ ε

−ε ≤ c − aµ0 ≤ ε

−ε ≤ 2ε + d − aµ0 ≤ ε

−3ε ≤ d − aµ0 ≤ −ε < 0

But this implies that aµ0 > d, a contradiction. Thus c can not be greater then L+ and we

have that c ≤ L+ .

Page 4
Known results in superior and inferior limits Goran Huygh

3 L+ or L− as limit points or limits

Theorem 4. If L+ or L− is finite, then it is a limit point of (an )∞


n=m

Proof. Once again we only prove the statement for L+ as the proof is symmetrical for L− .

Given an arbitrary ε > 0 we obviously have

L+ − ε < L + < L + + ε

Implying that for every K we have that there exists some k ≥ K for which we have L+ − ε <

ak , fix some arbitrary K, combining this with the fact that there exists some N for which

all n ≥ N , an < L+ + ε. Now putting M = max(K, N ) implies that

∃m ≥ M, |am − L+ | ≤ ε

But since the choice of ε and K is totally arbitrary we have that

∀ε ∀N ∃n ≥ N : |an − L+ | ≤ ε

Theorem 5. Given a sequence (an )∞ −


n=m and a real number c, limn→∞ an = c ⇐⇒ L =

c = L+

Proof. We split the implication. ( =⇒ ) Since the sequence is convergent we know that there

is a upper bound M for which −M ≤ an ≤ M for all n ≥ m. We then have

an ≤ M =⇒ sup(an )∞
n=m ≤ M =⇒ L ≤ M
+

Page 5
Known results in superior and inferior limits Goran Huygh

We do the same to show that −M ≤ L− giving us

−M ≤ L− ≤ L+ ≤ M

Proving that both must be finite. But since they are finite they are both limit points of the

sequence. But notice that a convergent sequence has only 1 limit point which is always it’s

limit. Thus we must have L− = c = L+ . (⇐=) Notice that we have for any arbitrary ε > 0

that L− − ε < c < L+ + ε. We thus know there exists a K and a M such that for all k ≥ K

and m ≥ M

L − − ε < a k ∧ am < L + + ε

If we then put N := max(K, M ) we get that for all n ≥ N

|an − c| ≤ ε

Since the choice of N is fully dependant on the choice of epsilon, which was arbitrary, we

have that limn→∞ an = c.

Page 6
Known results in superior and inferior limits Goran Huygh

4 Comparison principles of sequences

Theorem 6. (Comparison principle) Given two sequences (an )∞ ∞


n=m and (bn )n=m for which

for all n ≥ m, an ≤ bn , the following holds

sup(an )∞ ∞
n=m ≤ sup(bn )n=m

inf(an )∞ ∞
n=m ≤ inf(an )n=m

lim sup(an )∞ ∞
n=m ≤ lim sup(bn )n=m
n→∞ n→∞

lim inf(an )∞ ∞
n=m ≤ lim inf(bn )n=m
n→∞ n→∞

Proof. Notice that

an ≤ bn ≤ sup(bn )∞ ∞ ∞
n=m =⇒ sup(an )n=m ≤ sup(bn )n=m

because otherwise we would have a contradiction against the definition of the supremum of

an . Similarly we have

lim sup an ≤ sup(an )∞ ∞


n=m ≤ sup(bn )n=m =⇒ lim sup an ≤ lim sup bn
n→∞ n→∞ n→∞

to once again avoid a contradiction. The proofs is symmetric for the limit inferior and

infimum.

Page 7
Known results in superior and inferior limits Goran Huygh

Theorem 7. (Squeeze theorem) Let (an )∞ ∞ ∞


n=m , (bn )n=m and (cn )n=m be sequences of real

numbers such that

an ≤ b n ≤ c n

for all n ≥ m. Suppose also that (an )∞ ∞


n=m and (cn )n=m both converge to the same limit L.

Then (bn )∞
n=m converges to L

Proof. Since a sequence converging to L implies that L− = L = L+ we have that by the

comparison principle

L ≤ lim sup bn ≤ L
n→∞

L ≤ lim inf bn ≤ L
n→∞

But since the L+ and L− of (bn )∞ ∞


n=m are equal this implies that (bn )n=m itself converges to

L.

Page 8

You might also like