Borel-Cantelli

Q: What does $ \limsup_{n\to\infty} A_n $ mean?

$ \limsup_{n\to\infty} A_n $ represents the event that infinitely many of the events $ A_n $ occur. Formally, it is the set of outcomes $ \omega \in \Omega $ such that $ \omega \in A_n $ for infinitely many $ n $.

Q: Can we have $ P(\limsup_{n\to\infty} A_n) $ strictly between 0 and 1?

Yes, if the events are not independent and the sum $ \sum P(A_n) $ diverges. In such cases, neither lemma directly applies, and the probability of the limit superior can be any value in $ [0, 1] $.

Infinite predictors.

The Formal Theorem

Let

(\Omega, \mathcal{F}, P)

be a probability space, and let

\{A_n\}_{n=1}^\infty

be a sequence of events. \\ First Borel-Cantelli Lemma: If

\sum_{n=1}^\infty P(A_n) < \infty

, then

P(\limsup_{n\to\infty} A_n) = 0

. \\ Second Borel-Cantelli Lemma: If the events

\{A_n\}_{n=1}^\infty

are independent and

\sum_{n=1}^\infty P(A_n) = \infty

, then

P(\limsup_{n\to\infty} A_n) = 1

Analytical Intuition.

Imagine a cosmic projector, its beam sweeping across the vast expanse of

\Omega

, highlighting events

A_n

. The first lemma is like a whisper: if the 'importance' (probability) of these events dwindles fast enough – like grains of sand slipping through fingers – such that the total importance is finite

\sum P(A_n) < \infty

, then eventually, the projector will pass by *all* these significant events infinitely often and you'll never see them again. The second lemma, however, shouts: if these events are independent, like lottery draws, and their importance *never* truly wanes – the sum of probabilities is infinite

\sum P(A_n) = \infty

– then eventually, you are guaranteed to see *all* of them occur infinitely many times. It's the guarantee of an infinite cascade.

CAUTION

Institutional Warning.

Mixing up the condition for the first lemma (convergent sum) with the condition for the second lemma (divergent sum), or forgetting the independence requirement for the second lemma.

Institutional Deep Dive.

The Borel-Cantelli lemmas represent a foundational truth regarding the persistence of stochastic events, revealing a brutal calculus of convergence and divergence in the realm of probability. They articulate, with unnerving precision, the conditions under which an event will recur infinitely often, or conversely, cease to occur after some finite epoch. This is not a matter of subjective intuition, but of rigorous analytical deduction.

Core Analytical Logic:

The first Borel-Cantelli lemma provides a criterion for the cessation of events. It states that if we have a sequence of events

\{A_n\}_{n=1}^{\infty}

in a probability space

(\Omega, \mathcal{F}, P)

, and the sum of their probabilities converges, i.e.,

\sum_{n=1}^{\infty} P(A_n) < \infty

, then the probability that infinitely many of these events occur is zero. Formally,

P(A_n \text{ i.o.}) = P(\limsup_{n \to \infty} A_n) = 0

. The intuition derives from the convergence of the expected number of occurrences. If

\sum P(A_n)

converges, this sum can be interpreted as the expected number of times *any* event

A_n

occurs, provided the events are simple indicators. More formally, by the union bound, for any

k \in \mathbb{N}

, we have

P(\bigcup_{n=k}^{\infty} A_n) \le \sum_{n=k}^{\infty} P(A_n)

. As

k \to \infty

, the tail sum

\sum_{n=k}^{\infty} P(A_n)

must tend to zero due to the convergence of the infinite series. The event

\{A_n \text{ i.o.}\}

is precisely

\bigcap_{k=1}^{\infty} \bigcup_{n=k}^{\infty} A_n

. Thus,

P(A_n \text{ i.o.}) \le P(\bigcup_{n=k}^{\infty} A_n)

for any

k

, and since the right-hand side can be made arbitrarily small by choosing

k

sufficiently large, the probability of infinitely many occurrences must be zero. This lemma predicts that even if events are always possible, if their individual likelihoods diminish rapidly enough, their collective persistence is doomed. The monkeys typing Shakespeare eventually cease their productive spasms.

The second Borel-Cantelli lemma addresses the complementary scenario. If the events

\{A_n\}_{n=1}^{\infty}

are independent, and the sum of their probabilities diverges, i.e.,

\sum_{n=1}^{\infty} P(A_n) = \infty

, then the probability that infinitely many of these events occur is one. Formally,

P(A_n \text{ i.o.}) = P(\limsup_{n \to \infty} A_n) = 1

. This is contingent upon independence. To grasp this, consider the probability that

A_n

occurs *finitely often*. This is the event

\bigcup_{k=1}^{\infty} \bigcap_{n=k}^{\infty} A_n^c

, where

A_n^c

is the complement of

A_n

. If

\sum P(A_n)

diverges, then we consider the product

\prod_{n=k}^{\infty} P(A_n^c) = \prod_{n=k}^{\infty} (1 - P(A_n))

. Using the inequality

1-x \le e^{-x}

for

x \ge 0

, we have

\prod_{n=k}^{\infty} (1 - P(A_n)) \le \prod_{n=k}^{\infty} e^{-P(A_n)} = e^{-\sum_{n=k}^{\infty} P(A_n)}

. Since

\sum P(A_n)

diverges, the tail sum

\sum_{n=k}^{\infty} P(A_n)

also diverges to

\infty

. Consequently,

e^{-\sum_{n=k}^{\infty} P(A_n)}

tends to

0

k \to \infty

. This means the probability of *not* observing any

A_n

for all

n \ge k

approaches zero as

k

increases. Therefore, the probability of observing

A_n

infinitely often must be one.

Geometric Mechanics:

Visualize the probability space as a unit interval

[0,1]

. Each event

A_n

corresponds to a measurable subset of this interval with measure

P(A_n)

. For the first lemma, imagine covering

[0,1]

with a sequence of increasingly smaller 'patches' or 'regions'. If the sum of the lengths of these patches,

\sum P(A_n)

, is finite, then even if we try to place an infinite number of them, their total 'mass' or 'coverage' is limited. Any point

x \in [0,1]

that is covered by infinitely many

A_n

must belong to a region that has an infinitely diminishing density of coverage. The measure of points covered infinitely often becomes negligible, a set of measure zero. It is akin to attempting to pave an infinitely long road with a finite amount of material; the paving must eventually cease. The boundary is the point where the cumulative 'area' of events fails to grow beyond a finite limit.

For the second lemma, with the crucial independence assumption, the situation is different. Imagine these events as independent 'hits' on targets scattered across the unit interval. If the sum of the probabilities

\sum P(A_n)

is infinite, it implies that the 'budget' for these hits is inexhaustible. Each event

A_n

"fails" with probability

1-P(A_n)

. If these failures are independent, the product of their probabilities

\prod (1-P(A_n))

gauges the chance of indefinite evasion. When

\sum P(A_n)

diverges, this product collapses to zero, signifying that indefinite evasion becomes impossible. The 'boundary' is the critical divergence of the sum, at which point the cumulative effect of independent chances dictates certainty of recurrence.

Institutional Pitfalls:

Students frequently falter in their understanding of Borel-Cantelli due to several ingrained misconceptions. Firstly, the concept of "almost surely" is often conflated with "definitely" or "certainly." A probability of zero does not preclude the logical possibility of an event; it merely indicates its statistical irrelevance in the measure-theoretic sense. The set of outcomes where infinitely many events occur may be non-empty, but it is a null set.

Secondly, and most critically for the second lemma, the indispensable requirement of *independence* is frequently overlooked or misapplied. Without independence, the argument based on product probabilities collapses, and the conclusion no longer holds. Consider the trivial case where

A_n = A

for all

n

, with

P(A) > 0

. Then

\sum P(A_n) = \infty

, but

A

occurs either once or zero times, not infinitely often. This fundamental misunderstanding of independence versus dependence often renders the lemma inert in practical application for the uninitiated.

Finally, the abstract nature of "infinitely often" itself presents a cognitive hurdle. It is not merely "a great many times," but rather the absolute absence of a "last time." The lemmas force a precise distinction between transient phenomena and truly recurrent ones, a distinction that demands a disciplined approach to the infinite. Failure to internalize these distinctions represents a failure to grasp the true power and limitations of measure-theoretic probability.

Academic Inquiries.

What does $\limsup_{n\to\infty} A_n$ mean?

$\limsup_{n\to\infty} A_n$ represents the event that infinitely many of the events $A_n$ occur. Formally, it is the set of outcomes $\omega \in \Omega$ such that $\omega \in A_n$ for infinitely many $n$ .

Is the independence assumption in the Second Borel-Cantelli Lemma essential?

Yes, it is crucial. The proof relies on the fact that the probability of the intersection of independent events is the product of their probabilities, which is used to show that the tails of the series for $P(\liminf A_n^c)$ tend to zero when $\sum P(A_n) = \infty$ .

Can we have $P(\limsup_{n\to\infty} A_n)$ strictly between 0 and 1?

Yes, if the events are not independent and the sum $\sum P(A_n)$ diverges. In such cases, neither lemma directly applies, and the probability of the limit superior can be any value in $[0, 1]$ .

Standardized References.

Definitive Institutional SourceBillingsley, Probability and Measure

Advanced

Proof: Borel-Cantelli Lemma 2 (Independence, Divergent Sum)

The second Borel-Cantelli Lemma for independent events with a divergent sum of probabilities.

Advanced

Martingale Convergence

Fair game fate.

Advanced

Ergodic Theorem

Time-space unity.

Institutional Citation

Reference this proof in your academic research or publications.

NICEFA Visual Mathematics. (2026). Borel-Cantelli: Visual Proof & Intuition. Retrieved from https://nicefa.org/library/advanced-probability-theory/borel-cantelli-lemma-1-convergence

Dominate the Logic.

"Abstract theory is just a movement we haven't seen yet."

Master the Proof Early Access

The Formal Theorem

Analytical Intuition.

Institutional Warning.

Institutional Deep Dive.

Academic Inquiries.

What does lim sup⁡n→∞An \limsup_{n\to\infty} A_n limsupn→∞​An​ mean?

Is the independence assumption in the Second Borel-Cantelli Lemma essential?

Can we have P(lim sup⁡n→∞An) P(\limsup_{n\to\infty} A_n) P(limsupn→∞​An​) strictly between 0 and 1?

Standardized References.

Related Proofs Cluster.

Proof: Borel-Cantelli Lemma 2 (Independence, Divergent Sum)

Martingale Convergence

Ergodic Theorem

Institutional Citation

Dominate the Logic.

What does $\limsup_{n\to\infty} A_n$ mean?

Can we have $P(\limsup_{n\to\infty} A_n)$ strictly between 0 and 1?