Introduction to Probability Theory

K. Suresh Kumar
Department of Mathematics
Indian Institute of Technology Bombay

August 28, 2017

2

LECTURES 11

0.0.1 Borel-Cantelli Lemma

In many situations which involves independent trials, one may interested in
calculating the probabilities of the events in which some specific scenario
happens repeatedly, for example in denumerable independent tosses of a
coin, getting infintely many H. Borel-Cantelli Lemma which will prove next
will help us in calculating probabilities of such events. We need set theoretic
model for events of these form, which are given by the notion of limsup of
sets which is defined below. Borel-Cantelli lemma is all about the probabil-
ity of limsup of events.

Definition 4.12 (lim sup of sets) For A1 , A2 , · · · , subsets of Ω , define

lim sup An := ∩∞ ∞
n=1 ∪k=n Ak .
n

Similarly
lim inf An := ∪∞ ∞
n=1 ∩k=n .
n
In the following theorem, we list some useful properties of limsup and liminf
with an intention of making the objects lim inf and lim sup of sets much
clearer.

Theorem 0.1 1. Let

{An i.o} := {ω ∈ Ω | ω ∈ An for infinitely many n}.

Then
lim sup An = {An i.o} .
n

2. Let

{An a.a. } := {ω ∈ Ω | ω ∈ Ak for all k ≥ n for some n ≥ 1} .

Here a.a. stands for ’all most all’. Then

lim inf An = {An a.a. } .

n

3. The following inclusion holds.

lim sup An ⊇ lim inf An .

n n
 3

4. The following identity holds.

[lim sup An ]c = lim inf (An )c .
n n

5. If A1 ⊆ A2 ⊆ · · · , then
lim sup An = ∪∞
n=1 An = lim inf An
n n

6. If A1 ⊇ A2 ⊇ · · · , then
lim sup An = ∩∞
n=1 An = lim inf An .
n n

Proof.
1. Consider
ω ∈ lim sup An iff ω ∈ ∪∞
k=n Ak for all n ≥ 1
n
iff There exists n1 , n2 , . . . , such that ω ∈ Ank , for all k ≥ 1
iff ω ∈ An i.o.
This proves (1).
2. Consider
ω ∈ lim inf An iff ω ∈ ∩∞
k=n Ak for some n ≥ 1
n
iff There exists n0 ≥ 1 such that ω ∈ An , for all n ≥ n0
iff ω ∈ An all but finitely many
Thus (2).
3. Proof of (3) follows from (1) and (2).
4. Proof of (4) follows from De Morgen’s laws.
5. Note
lim sup An = ∩∞ ∞ ∞
n=1 ∪k=n Ak ⊆ ∪k=1 Ak .
n
Now using the fact that An ’s are increasing, we get
lim inf An = ∪∞ ∞ ∞
n=1 ∩k=n Ak = ∪n=1 An .

Hence
lim inf An = ∪∞
n=1 An ⊆ lim sup An ⊆ lim inf An .
n

From this (5) follows.

4

6. The proof of (6) follows from (4) and (5).

Remark 0.1 The properties (1)-(6) are analogous to the corresponding prop-
erties of lim sup1 and lim inf of real numbers.

Remark 0.2 Analogous to the notion of limit of sequence of numbers, one

can say that limn→∞ An exists if

lim inf An = lim sup An .

n→∞ n→∞

Hence from Theorem 3.0.11 (5), if {An } is an increasing sequence of sets,

then limn→∞ An exists and is ∪∞ n=1 An . Similarly using Theorem 3.0.11 (6),
if {An } is an decreasing sequence of sets, then limn→∞ An exists and is
∩∞n=1 An . Now student can see why property (6) in Theorem 1.0.1 is called
continuity property of probability.

Theorem 0.2 (Borel - Cantelli Lemma) Let (Ω, F, P ) be a probability space

and A1 , A2 , · · · ∈ F.
(i) If
X∞
P (An ) < ∞
n=1

then
P (lim sup An ) = 0 .
n

(ii) If
∞
X
P (An ) = ∞
n=1

and
{An | n = 1, 2, · · · }
are independent, then

P (lim sup An ) = 1 .
1
lim sup of a sequence of real numbers {an } is defined as

lim sup{an , an+1 , · · · }

n→∞

and is denoted by lim supn→∞ an .

 5

Proof: (i) Consider

P (lim supn An ) = P (∩∞ ∞

n=1 ∪k=n Ak )

= lim P (∪∞
k=n Ak )
n→∞

∞
X
≤ lim P (Ak ) .
n→∞
k=n

Note that the r.h.s → 0 as n → ∞, since

∞
X
P (An ) < ∞ .
n=1

Therefore
P (lim sup An ) = 0 .
n

(ii) Note that

(lim sup An )c = ∪∞ ∞ c
n=1 ∩k=n Ak .
n

Therefore
P (lim sup An ) = 1 − P (∪∞ ∞ c
n=1 ∩k=n Ak ) .
n

Hence it is enough to show that

P (lim inf Acn ) = 0.

n→∞

The proof the above uses the following two results.

{A1 , A2 , · · · } are independent ⇒ {Ac1 , Ac2 , · · · } are independent .

• For each x ∈ R,
1 − x ≤ e−x .

(The proof of this inequlaity follows from the fact that f (x) = e−x +
x − 1, x ∈ R is convex and has a minimum at x = 0 which is unique.)
6

Now
n+m c
P (∩∞ c
k=n Ak ) ≤ P (∩k=n Ak ) for all m = 1, 2, · · ·

n+m
= Πk=n P (Ack ) .

The last equality follows from

{A1 , A2 , · · · } are independent ⇒ {Ac1 , Ac2 , · · · } are independent .

Using 1 − x ≤ e−x , we get

P (∩∞ c n+m n+m −P (Ak )

k=n Ak ) ≤ Πk=n (1 − P (Ak ) ≤ Πk=n e .

Since
∞
X
P (An ) = ∞
n=1
we get Pn+m
lim e− k=n P (Ak )
= 0.
n→∞

Therefore
P (∩∞ c
k=n Ak ) = 0 ∀ n = 1, 2, . . . .

Thus
∞
X
P (∪∞ ∞ c
n=1 ∩k=n Ak ) ≤ P (∩∞ c
k=n Ak ) = 0 .
n=1

Therefore
P (lim sup An ) = 1 .
n

This completes the proof.