Limit of a Sequence of Sets

While the concept of a limit for a sequence of numbers is a cornerstone of basic calculus, the notion of a limit for a sequence of sets is far less intuitive. How can we define convergence when we are dealing with collections of points that can expand, shrink, or oscillate in complex ways? This question presents a fundamental challenge, one that requires moving beyond a single limit point and embracing a new framework to capture the dynamic behavior of sets.

This article bridges that gap by providing a comprehensive introduction to the theory of set sequence limits. It demystifies these concepts for readers, guiding them from foundational principles to powerful real-world applications. In the upcoming chapters, you will embark on a structured journey. The "Principles and Mechanisms" chapter will lay the groundwork, formally defining the limit inferior and limit superior, exploring their properties through concrete examples, and establishing the essential context of σ-algebras. Following that, the "Applications and Interdisciplinary Connections" chapter will reveal how these abstract tools become indispensable lenses for understanding problems in measure theory, probability, and topology, showcasing the unifying power of this mathematical idea.

You’re familiar with the idea of a limit for a sequence of numbers. When we say the sequence $1, \frac{1}{2}, \frac{1}{3}, \dots$ approaches a limit of $0$, we have a very precise notion of what that means. But what if we have a sequence of sets? Can collections of points also have a limit? This is not just a curious mathematical puzzle; it's a foundational concept that breathes life into fields like probability theory and real analysis. Let’s embark on a journey to understand how sets can move, shrink, and grow, and what it means for them to "settle down."

Imagine a sequence of sets, $A_1, A_2, A_3, \dots$. Unlike a sequence of numbers, which we can plot on a line, a sequence of sets is a more slippery character. A point might be in $A_1$, out of $A_2$, back in $A_3$, and so on. How can we possibly talk about a "limit" for such behavior?

The brilliant insight is to stop looking for a single limit and instead define two boundaries: a lower limit and an upper limit. These are called the ​​limit inferior​​ ($\liminf$) and the ​​limit superior​​ ($\limsup$). They give us a way to bracket the ultimate behavior of the sequence.

• The ​​limit inferior​​, or $\liminf_{n \to \infty} A_n$, is the set of points that are eventually in the sequence. What does "eventually" mean? It means a point $x$ is in the $\liminf$ if there's some stage, say $N$, after which $x$ is in every single set $A_n$ for $n \ge N$. It gets in and stays in. The points in the $\liminf$ are the loyal residents. This set can be expressed with unions and intersections as the set of elements belonging to all but a finite number of the sets $A_n$: $\liminf_{n \to \infty} A_n = \bigcup_{N=1}^{\infty} \bigcap_{n=N}^{\infty} A_n$

• The ​​limit superior​​, or $\limsup_{n \to \infty} A_n$, is the set of points that are in the sequence infinitely often. A point $x$ is in the $\limsup$ if, no matter how far you go down the sequence, you can always find a later set that contains $x$. It might pop in and out, but it never leaves for good. These points are the persistent visitors. The formal definition is: $\limsup_{n \to \infty} A_n = \bigcap_{N=1}^{\infty} \bigcup_{n=N}^{\infty} A_n$

From these definitions, it's clear that if a point eventually stays in all the sets ($\liminf$), it must certainly visit infinitely often ($\limsup$). Therefore, we always have the relationship: $\liminf_{n \to \infty} A_n \subseteq \limsup_{n \to \infty} A_n$.

When the lower and upper limits coincide—when the set of loyal residents is the same as the set of persistent visitors—we say the ​​limit​​ of the sequence of sets exists and is equal to this common set.

Let’s make this concrete. The simplest paths are the monotone ones.

Consider an ​​increasing sequence​​ of sets, where each set contains the previous one: $A_1 \subseteq A_2 \subseteq A_3 \subseteq \dots$. Imagine a sequence of intervals $A_n = [-n - \frac{1}{n}, n^2]$. Each interval is larger than the one before it. A point that gets into one of these sets stays in all the subsequent, larger sets. Here, the "infinitely often" and "eventually in" conditions become the same. Any point on the real line will eventually be swallowed by these expanding intervals. Consequently, both the $\liminf$ and $\limsup$ are the union of all the sets, which in this case is the entire real line, $\mathbb{R}$. For any increasing sequence, the limit is simply its union: $\lim_{n \to \infty} A_n = \bigcup_{n=1}^{\infty} A_n$

Now, consider a ​​decreasing sequence​​: $B_1 \supseteq B_2 \supseteq B_3 \supseteq \dots$. Let's take the sets $B_n = (1 - \frac{1}{n}, 3 + \frac{1}{n^2}]$. Each interval is slightly smaller than the one before. A point is in the limit only if it can survive being "squeezed" by every set in the sequence. This means it must lie in their intersection. Here, the limit is $[1, 3]$. For any decreasing sequence, the limit is its intersection: $\lim_{n \to \infty} B_n = \bigcap_{n=1}^{\infty} B_n$

But what about a more interesting, non-monotone path? Consider the sequence of sets defined by $A_n = \{\cos(\frac{n\pi}{2}), \sin(\frac{n\pi}{2})\}$. Let's write out the first few terms:

• $A_1 = \{0, 1\}$
• $A_2 = \{-1, 0\}$
• $A_3 = \{0, -1\}$
• $A_4 = \{1, 0\}$
• $A_5 = \{0, 1\}$

The sequence of sets cycles through $\{0, 1\}$ and $\{-1, 0\}$.

• Which points are in ​​infinitely often​​ ($\limsup$)? The point $0$ is in every single set. The point $1$ appears in $A_1, A_4, A_5, \dots$ (whenever $n \equiv 0,1 \pmod 4$). The point $-1$ appears in $A_2, A_3, A_6, \dots$ (whenever $n \equiv 2,3 \pmod 4$). So, all three points, $-1, 0, 1$, keep coming back. $L_s = \limsup A_n = \{-1, 0, 1\}$.
• Which points are ​​eventually in​​ and stay forever ($\liminf$)? Only the point $0$ has this property; it's in every set from the very beginning. The points $1$ and $-1$ are nomads, constantly leaving and returning. $L_i = \liminf A_n = \{0\}$. In this case, the $\liminf$ is a proper subset of the $\limsup$. The set of points that cause this difference, $L_s \setminus L_i = \{-1, 1\}$, are precisely the elements that oscillate in and out of the sequence without ever settling down.

These concepts of $\liminf$ and $\limsup$ are not just arbitrary definitions; they possess a deep and beautiful internal logic.

One of the most elegant relationships is a kind of De Morgan's Law for set limits. It connects the limit of a sequence to the limit of its complements. The statement is: $(\limsup_{n \to \infty} A_n)^c = \liminf_{n \to \infty} (A_n^c)$ Let's translate this. The left side, $(\limsup A_n)^c$, describes the points that are not in infinitely many $A_n$. This is the same as saying they are in only a finite number of $A_n$. But if a point is in only finitely many $A_n$, it must be in the complement, $A_n^c$, for all but a finite number of $n$. This is precisely the definition of being in the $\liminf$ of the complements, $\liminf (A_n^c)$! This beautiful duality shows how $\limsup$ and $\liminf$ are two sides of the same coin, perfectly mirrored through the operation of complementation.

Another way to grasp these limits is by translating them from the language of sets to the language of functions. We can define an ​​indicator function​​, $1_A(x)$, which is $1$ if $x$ is in set $A$ and $0$ otherwise. Now, our sequence of sets $\{A_n\}$ becomes a sequence of functions $\{1_{A_n}(x)\}$, where each function can only output $0$ or $1$. A point $x$ being in $A_n$ "infinitely often" is the same as the sequence of numbers $1_{A_1}(x), 1_{A_2}(x), \dots$ having the value $1$ infinitely often. The limit superior of this sequence of numbers is $1$. If $x$ is in only finitely many $A_n$, the sequence of numbers is eventually all $0$, and its limit superior is $0$. This leads to a remarkable identity: $1_{\limsup A_n}(x) = \limsup_{n \to \infty} 1_{A_n}(x)$ The indicator of the limit superior of sets is the limit superior of the indicator functions! This bridges the abstract world of sets with the more familiar territory of real-valued sequences.

To do powerful mathematics with sequences of sets, especially when we want to measure them, we need to ensure our operations don't lead us out of the world of "measurable" sets we started with. We need a stable playground. This playground is called a ​​$\sigma$-algebra​​.

An ​​algebra​​ of sets is a collection closed under finite unions and complements. But this is not enough for the kinds of infinite processes we are discussing. Consider the collection $\mathcal{A}$ of all subsets of natural numbers that are either finite or have a finite complement (cofinite). This collection is an algebra. Now, take the sequence of sets $A_n = \{2n\}$, for $n=1, 2, 3, \dots$. Each $A_n$ is a singleton, so it's finite and belongs to our algebra $\mathcal{A}$. But what is their union? $\bigcup_{n=1}^{\infty} A_n = \{2, 4, 6, 8, \dots\}$ This is the set of even numbers. It is an infinite set, and its complement, the set of odd numbers, is also infinite. So the union is neither finite nor cofinite; it has escaped our algebra!

To handle limits, we need closure under countable unions. This is the defining property of a ​​$\sigma$-algebra​​. It is a collection of sets closed under complementation and countable unions (and, by De Morgan's laws, countable intersections). Uncountable unions, however, are generally not permitted.

The crucial fact is that if you take any sequence of sets $\{A_n\}$ from a $\sigma$-algebra $\mathcal{F}$, their limit-sets, $\limsup A_n$ and $\liminf A_n$, are also guaranteed to be in $\mathcal{F}$. Why? Because their definitions are built entirely from countable unions and intersections, the very operations a $\sigma$-algebra is designed to handle. This ensures our universe is complete; it contains all the limiting objects we can construct within it.

Now we arrive at the payoff. We can ask a profound question: If we know the "size" (or measure, $\mu$) of every set in a sequence, can we determine the size of the limit set? The answer is a qualified "yes," and it's called the ​​continuity of measure​​.

For an increasing sequence of measurable sets $A_1 \subseteq A_2 \subseteq \dots$, the measure of the limit is the limit of the measures: $\mu\left(\bigcup_{n=1}^{\infty} A_n\right) = \lim_{n \to \infty} \mu(A_n)$

For a decreasing sequence $B_1 \supseteq B_2 \supseteq \dots$, we have a similar result, but with a critical condition. If at least one of the sets in the sequence has a finite measure (e.g., $\mu(B_1) \infty$), then the measure of the limit is the limit of the measures: $\mu\left(\bigcap_{n=1}^{\infty} B_n\right) = \lim_{n \to \infty} \mu(B_n)$

This is an incredibly powerful tool. It allows us to calculate the measure of a complicated intersection by computing the limit of the measures of simpler sets.

But beware the fine print! The condition $\mu(B_1) \infty$ is not optional. It is the linchpin of the theorem. Consider the sequence of sets on the real line $A_n = [0, 5] \cup [n, \infty)$. This is a decreasing sequence of sets. The measure of every single set $A_n$ is infinite, because it contains an infinite ray. What is their intersection? As $n$ grows, the interval $[n, \infty)$ slides off to infinity, leaving only the common part behind. $\bigcap_{n=1}^{\infty} A_n = [0, 5]$ The measure of this intersection is $\mu([0, 5]) = 5$. However, the limit of the measures is $\lim_{n\to\infty} \mu(A_n) = \lim_{n\to\infty} \infty = \infty$. Clearly, $5 \neq \infty$. The continuity property failed spectacularly. It failed because we violated the one simple rule: for a decreasing sequence, you must start from a set of finite size. This example illustrates a deep truth in mathematics: the conditions on a theorem are not mere suggestions; they are the guardians that prevent us from falling into contradiction and paradox. They define the boundaries within which the beautiful logic holds true.

Now that we have acquainted ourselves with the machinery of set sequences—the limit superior and limit inferior—we might be tempted to ask, "What is it all for?" Is this merely an elegant game of symbolic logic, a playground for the pure mathematician? The answer, you will be delighted to hear, is a resounding no. This language is not an end in itself; it is a lens. It is a tool for asking, and rigorously answering, profound questions in fields that stretch from the heart of physics to the foundations of probability and the very structure of space. We are about to embark on a journey to see how this simple idea—a sequence of sets—blossoms into a surprisingly powerful way of understanding the world.

Let’s begin with the most tangible of ideas: measurement. How do you determine the "size" of a complicated object? A classic strategy, beloved by physicists and mathematicians alike, is to approximate. You trap your difficult shape inside a sequence of simpler shapes whose size you know, and then you watch what happens as the trap gets tighter and tighter.

Consider a sequence of shrinking closed intervals on the real number line: $[-1, 1]$, then $[-\frac{1}{2}, \frac{1}{2}]$, then $[-\frac{1}{3}, \frac{1}{3}]$, and so on. This is a decreasing sequence of sets; each one is nestled inside the one before. What single, stubborn point survives inside all of them, no matter how far down the sequence we go? Only the point zero. The sequence of sets "converges" to the set $\{0\}$. Now, what about their lengths, or what we call their Lebesgue measure? The lengths are $2, 1, \frac{2}{3}, \frac{2}{4}, \dots, \frac{2}{n}, \dots$. This sequence of numbers clearly converges to zero.

It seems wonderfully, satisfyingly logical that if the sets themselves shrink to a single point, their measures should shrink to the measure of that point. This principle, known as the continuity of measure, is not just a pleasant coincidence; it is a cornerstone of modern analysis. It gives us confidence that under the right conditions (a decreasing sequence of sets, with the first one having finite measure), the limit of the measures is precisely the measure of the limit set.

This tool allows us to tackle far more bizarre objects. Imagine starting with a solid square. Now, divide it into a $3 \times 3$ grid of nine smaller squares, and throw away the five that form the central cross, keeping only the four corner squares. You're left with a shape made of four smaller, disconnected squares. Now, do the exact same thing to each of those four squares. And then again to the sixteen squares you have now, and so on, forever. You are constructing a decreasing sequence of sets, and their intersection is a beautiful, infinitely detailed pattern known as a Cantor dust. What is its two-dimensional area? At the first step, we kept $\frac{4}{9}$ of the original area. At the second, we keep $\frac{4}{9}$ of that, giving $(\frac{4}{9})^2$ of the original area. The area after $n$ steps is $(\frac{4}{9})^n$. As $n$ tends to infinity, this quantity rushes to zero. By the continuity of measure, we can declare with certainty that this intricate, endlessly complex fractal dust has a total area of exactly zero! It’s a set you can see, a set containing an uncountable infinity of points, yet its two-dimensional "footprint" is nothing. This is the kind of profound, and often counter-intuitive, result that sequences of sets allow us to handle with perfect rigor.

The limit superior of a sequence of sets, you'll recall, is the collection of all points that belong to infinitely many of the sets $A_n$. This idea has a fantastically intuitive interpretation in the world of probability. If each set $A_n$ represents some event happening at time $n$, then $\limsup A_n$ is the event that "$A_n$ happens infinitely often."

So, when can we say that something will almost certainly not happen infinitely often? The brilliant Borel-Cantelli Lemma gives us a surprisingly simple condition. Imagine a sequence of events $A_n$, and let's say their measures (or probabilities) are $m(A_n)$. If the sum of all these measures is finite, $\sum_{n=1}^\infty m(A_n) \infty$, then the measure of the set of points that fall into infinitely many of these $A_n$ is zero. Think about it this way: if you have a book with infinitely many pages, and on each page you spill a little bit of ink, but the total amount of ink you spill across all pages is finite (say, one bottle), what is the probability that a specific spot on your desk gets hit by ink from infinitely many different pages? It's zero! Although any one spill might hit it, the diminishing amounts of ink make it "infinitely unlikely" to be a perpetual target. This lemma is a workhorse in probability theory for proving that certain "bad" events almost surely happen only a finite number of times.

But we must be careful! One's intuition might leap to the conclusion that as long as the measure of the sets themselves, $m(A_n)$, goes to zero, the same result should hold. After all, if the events become smaller and smaller, shouldn't they be harder and harder to fall into? Nature, however, is more subtle.

Consider a clever construction where we lay down intervals on the line from $0$ to $1$. First, the whole interval $[0,1]$. Then, we cover it with two half-length intervals, $[0, \frac{1}{2}]$ and $[\frac{1}{2}, 1]$. Then with three third-length intervals, and so on. We can list all these intervals out to form an infinite sequence of sets $\{A_n\}$. The length of these intervals, $m(A_n)$, clearly goes to zero as we move down the sequence into blocks of smaller and smaller pieces. Yet, what is the set of points that gets covered infinitely many times? It is the entire interval $[0,1]$! Every point is caught in one of the intervals in the block of size $k$ for every $k$. The measure of the limit superior is 1, not 0. This "sweeping typewriter" example is a beautiful warning: for the Borel-Cantelli magic to work, it is not enough for the measures to just dwindle to zero; their sum must be finite.

Beyond size and probability, sequences of sets help us understand something even more fundamental: shape and structure. This is the domain of topology. In topology, we care less about "how big" a set is and more about properties like whether it's "all in one piece" (connected) or if it "contains its own boundary" (closed).

Let’s consider the property of being "closed." A sequence of nested, non-empty, closed and bounded sets in $\mathbb{R}$ (like a sequence of shrinking closed intervals) can never have an empty intersection. This is the famous Cantor Intersection Theorem. But what if we relax just one condition? What if the sets are not closed? Consider the sequence of open intervals $S_n = (0, \frac{1}{n})$. This is a nested sequence of non-empty, bounded sets. Each one looks almost like a closed interval. But for any number $x > 0$, no matter how small, we can always find an integer $n$ large enough so that $\frac{1}{n} x$, meaning $x$ is not in $S_n$. So no positive number is in the intersection. And zero is in none of them. The intersection is empty! The requirement of being "closed" is not a mere technicality; it is the very glue that holds the intersection together.

Limiting operations also interact with these topological properties in beautiful ways. One might ask if the limit superior of a sequence of closed sets is also guaranteed to be closed. While this is not true in general, the structure of the limit sets (as countable unions and intersections) ensures they belong to a well-behaved class of sets (specifically, Borel sets), and the property of being closed is preserved under specific conditions often studied in topology.

What about connectedness? If you have a chain of connected sets in the real line—a sequence of intervals where each one overlaps with the next—is their union also connected? Intuition suggests it should be, like linking together paper clips to form a single chain. And indeed, this is true. This simple theorem about sequences of sets forms the basis for how we prove more complex spaces are connected, a concept vital in everything from network analysis to understanding the domains of functions.

Finally, the theory of set sequences allows us to build bridges to even more abstract realms. We can rephrase questions about sets as questions about functions. For any set $A$, we can define its characteristic function, $\chi_A(x)$, which is 1 if $x$ is in $A$ and 0 otherwise. What happens to these functions when we have a sequence of sets? If we have a decreasing sequence of sets $A_n$ converging to an intersection $A$, the corresponding sequence of functions $\chi_{A_n}(x)$ is a decreasing sequence of numbers for each $x$, and its pointwise limit is exactly the characteristic function of the intersection, $\chi_A(x)$. This simple observation is the seed for some of the most powerful theorems in analysis, like the Monotone Convergence Theorem, which tells us when we can interchange the operations of limit and integration.

This perspective also reveals that "convergence" is a slippery concept. Consider a sequence of functions $f_n$ that "converges in measure" to the zero function, a type of convergence important in advanced analysis. This means the regions where $f_n$ is large are shrinking away to nothing. You might guess that the sets $A_n = \{x : f_n(x) > 0\}$ must also be shrinking away in some sense. But it's possible to construct a sequence of functions that converges to zero in measure, while the corresponding sets $A_n$ oscillate wildly and fail to converge to anything at all. It’s another beautiful reminder that our intuition must be guided by rigorous definitions.

Perhaps the most mind-bending application is to turn the tables entirely. Instead of thinking of sequences of sets that live inside a space, what if we imagine the collection of all measurable sets as a space in its own right? We can define a "distance" between two sets, $A$ and $B$, as the measure of their symmetric difference: $d(A, B) = m(A \Delta B)$. This turns the collection of sets into a genuine metric space! A "Cauchy sequence" of sets is one where the symmetric differences between sets far down the line become vanishingly small. A profound result is that this space is complete: every Cauchy sequence of sets converges to a well-defined limit set within the space. This gives us ultimate confidence in our limiting processes. It means that if we have a sequence of sets that seems to be settling down, there really is a bona fide set waiting for it at the end. The world of sets is not a chaotic mess; it has a beautiful, complete geometric structure of its own.

From measuring simple lines to calculating the area of fractal dust, from guaranteeing events in probability to understanding the topological structure of space, the humble sequence of sets proves to be a key that unlocks a remarkable number of doors. It is a testament to the unifying power of mathematics, where a single, elegant idea can ripple outwards, connecting and clarifying a vast landscape of scientific thought.