Infimum and Supremum

In mathematics, understanding the boundaries of a collection of numbers is a fundamental task. While the concepts of maximum and minimum are intuitive for simple, finite sets, they prove inadequate when dealing with more complex collections, such as the infinite set of numbers between 0 and 1. This limitation creates a conceptual gap: how do we precisely define the "edge" of a set when that edge isn't part of the set itself? This article bridges that gap by introducing the powerful concepts of infimum and supremum. First, in "Principles and Mechanisms," we will build a solid understanding of what these bounds are, why they are essential for the structure of the real numbers, and how they relate to limits and sequences. Following that, "Applications and Interdisciplinary Connections" will reveal how these seemingly abstract ideas provide a crucial language for fields ranging from optimization and computer science to the fundamental principles of physics.

Imagine you own a large, strangely-shaped piece of land. To build a fence, you need to know its boundaries. Finding the northernmost, southernmost, easternmost, and westernmost points seems straightforward. In mathematics, we often face a similar task. Given a collection, or a ​​set​​, of numbers, we want to understand its extent. What is its largest value, its smallest value? This simple question, as we shall see, opens a door to some of the most profound ideas in mathematics.

Let's start with a simple set of numbers, say $S = \{ -2, 1, 5, 8 \}$. What is the largest number in this set? Clearly, it's $8$. This is the ​​maximum​​ of the set. The smallest is $-2$, the ​​minimum​​. So far, so good.

Now, let's think about fences. We could build a fence along the line $x=10$. All our numbers in $S$ are less than $10$, so we can call $10$ an ​​upper bound​​ for the set. But $9$ is also an upper bound. So is $8.1$. And so is $8$. Of all the possible northern fences we could build, the one at $x=8$ is the tightest fit. It's the least of all the upper bounds. Similarly, $-2$ is the greatest of all the lower bounds.

For this simple finite set, the "least upper bound" is just the maximum, and the "greatest lower bound" is just the minimum. It seems we've invented fancy words—​​supremum​​ for the least upper bound and ​​infimum​​ for the greatest lower bound—for concepts we already knew. But have we?

Let's get a bit more clever. Consider the set of all real numbers strictly between $0$ and $1$. We call this the open interval $(0, 1)$. What is the maximum number in this set? You might say $0.9$. But $0.99$ is in the set and is larger. How about $0.999$? No, $0.9999$ is larger still. You can continue this forever. For any number you pick in the set, I can always find one that's a little bit bigger but still less than $1$. There is no "largest" number in the set. The set has no maximum. Likewise, it has no minimum.

Here is where our new words, supremum and infimum, show their true power. The set $(0,1)$ is certainly bounded. Numbers like $1, 2, 100$ are all upper bounds. What is the least of these upper bounds? It's $1$. So, we say the ​​supremum​​ of the set $(0,1)$ is $1$. Similarly, numbers like $0, -1, -50$ are all lower bounds. The greatest of these is $0$. The ​​infimum​​ is $0$.

The supremum and infimum are the perfect "boundary" points. They represent the ultimate edges of the set, even when those edges are not themselves members of the set. The supremum is the number that the set "reaches for" but may never touch.

Let's look at a more curious construction. Imagine the set of all numbers you can make with the fraction $\frac{m}{m+n}$, where $m$ and $n$ are any positive whole numbers, like $1, 2, 3, \dots$. Since $m$ and $n$ are positive, $m+n$ is always bigger than $m$, so all these fractions are less than $1$. And they are all greater than $0$. So, $1$ is an upper bound and $0$ is a lower bound. But can we ever reach $1$? No, because that would require $n=0$, which isn't allowed. Can we reach $0$? No, that would require $m=0$, also not allowed.

However, we can get tantalizingly close. If we fix $n=1$ and let $m$ be a huge number, say a million, our fraction becomes $\frac{1,000,000}{1,000,001}$, which is practically $1$. If we fix $m=1$ and let $n$ be a million, we get $\frac{1}{1,000,001}$, which is practically $0$. The set's values can squeeze up arbitrarily close to $1$ and down arbitrarily close to $0$. So, the supremum is $1$ and the infimum is $0$. A similar logic applies to a set like $S = \{ \frac{m-n}{m+n} \}$, which you can show is "reaching" for both $1$ and $-1$.

This idea of a boundary point not being in the set itself leads to a deep truth about the numbers we use. Let's consider a set made only of ​​rational numbers​​ (fractions). Define the set as $A = \{ q \in \mathbb{Q} \mid q^2 - 6q + 7 0 \}$. By solving the inequality, we find this is the set of all rational numbers $q$ such that $3 - \sqrt{2} q 3 + \sqrt{2}$.

The numbers $3 - \sqrt{2}$ and $3 + \sqrt{2}$ are ​​irrational​​—they cannot be written as a fraction. Our set $A$ contains only fractions. So the boundary points, the infimum $3 - \sqrt{2}$ and the supremum $3 + \sqrt{2}$, are not members of the set. This is not surprising. But imagine for a moment that our number system only contained rational numbers. The set $A$ would be defined, it would have upper bounds (like $3+1.5 = 4.5$), but there would be no least upper bound within the rational number system. There would be a "hole" where $3 + \sqrt{2}$ should be.

The ​​real numbers​​ ($\mathbb{R}$) are defined to fix this very problem. The crucial property, sometimes called the ​​Completeness Axiom​​, states that every non-empty set of real numbers that has an upper bound must have a supremum that is also a real number. This guarantees that the real number line is a true continuum, with no gaps or holes. It's the very foundation that allows calculus to work.

Finding a supremum or infimum often involves a dynamic process, like watching a sequence unfold or tracing a function's graph. Consider the sequence of numbers generated by the formula $f(n) = \frac{4n - 1}{n + 2}$ for $n = 0, 1, 2, \dots$.

Let's calculate the first few terms: For $n=0$, $f(0) = \frac{-1}{2} = -0.5$. For $n=1$, $f(1) = \frac{3}{3} = 1$. For $n=2$, $f(2) = \frac{7}{4} = 1.75$. For $n=3$, $f(3) = \frac{11}{5} = 2.2$.

It looks like the sequence is always increasing. If we can prove this (and we can!), then the very first term, $f(0) = -\frac{1}{2}$, must be the smallest value. It is the infimum, and in this case, it's also the minimum because it's part of the set.

But where is it headed? Let's rewrite the formula with a little algebraic trickery: $f(n) = \frac{4n - 1}{n + 2} = \frac{4(n+2) - 8 - 1}{n + 2} = \frac{4(n+2) - 9}{n + 2} = 4 - \frac{9}{n+2}$ Now it's clear! As $n$ gets larger and larger, the term $\frac{9}{n+2}$ becomes smaller and smaller, approaching zero. So, the value of $f(n)$ gets closer and closer to $4$, but from below. It never actually reaches $4$. The sequence is reaching for $4$, making $4$ its supremum. This connection between the ​​limit​​ of a sequence and its supremum/infimum is fundamental. A similar analysis of a function like $g(x) = \frac{3x^2 + 1}{x^2 + 2}$ reveals its infimum is $\frac{1}{2}$ (which is also its minimum value, at $x=0$) and its supremum is $3$ (the limit as $|x|$ becomes large).

The real beauty of a powerful concept is how elegantly it interacts with other ideas. What happens if we take two sets of numbers, $A$ and $B$, and create a new set $S$ by adding every number in $A$ to every number in $B$? That is, $S = \{a+b \mid a \in A, b \in B\}$. What is the supremum of this new, more complex set?

The answer is beautifully simple: $\sup(A+B) = \sup(A) + \sup(B)$. The greatest lower bound behaves just as nicely: $\inf(A+B) = \inf(A) + \inf(B)$. The boundary of the sum is the sum of the boundaries!

Let's see this in action with a wonderful example that ties together trigonometry, number theory, and analysis. Consider the set $S = \{ \frac{\cos(n)}{3} + \frac{\sin(m)}{5} \}$, where $n$ and $m$ are positive integers (and the angles are in radians). This looks fearsome! But we can break it down. Let $A = \{ \frac{\cos(n)}{3} \}$ and $B = \{ \frac{\sin(m)}{5} \}$.

Now, what is the supremum of the set of all values of $\cos(n)$? Though $n$ is an integer and will never be exactly $0$ or $2\pi$, it turns out that the values of $n$ (in radians) get arbitrarily close to any point on the circle. This is a deep result, but the consequence is that the set $\{\cos(n)\}$ gets arbitrarily close to $1$ and $-1$. Therefore, its supremum is $1$ and its infimum is $-1$. So, for our set $A$, $\sup(A) = \frac{1}{3} \times \sup\{\cos(n)\} = \frac{1}{3}$ and $\inf(A) = -\frac{1}{3}$. Similarly, for set $B$, $\sup(B) = \frac{1}{5}$ and $\inf(B) = -\frac{1}{5}$.

Using our simple rule for sums, the boundaries of the combined set $S$ are: $\sup(S) = \sup(A) + \sup(B) = \frac{1}{3} + \frac{1}{5} = \frac{8}{15}$ $\inf(S) = \inf(A) + \inf(B) = -\frac{1}{3} - \frac{1}{5} = -\frac{8}{15}$ A seemingly complicated problem dissolves into simple arithmetic, thanks to a robust underlying principle. This is the kind of unity and elegance that makes mathematics so rewarding.

Finally, what about sets that are not bounded at all? Consider the set of positive integers $\mathbb{N} = \{1, 2, 3, \dots\}$. It has an infimum of $1$. But it has no upper bound; you can't name a real number that's bigger than all the integers.

To speak about such sets, mathematicians find it convenient to use the ​​extended real number system​​, $\bar{\mathbb{R}}$, which is just the real numbers with two extra points added: $\infty$ (infinity) and $-\infty$ (negative infinity). In this system, we can say that if a set is not bounded above, its supremum is $\infty$. If it is not bounded below, its infimum is $-\infty$.

This allows us to describe any set. For instance, the set of all values of the tangent function, $\tan(x)$, includes all real numbers. As $x$ approaches $\frac{\pi}{2}$, $\tan(x)$ flies off to positive infinity. As $x$ approaches $-\frac{\pi}{2}$, it plunges to negative infinity. The function's range covers the entire real line. Thus, in the extended real number system, the supremum of the set of tangent values is $\infty$, and its infimum is $-\infty$.

From finding a simple maximum to defining the very structure of the number line and taming infinity, the concepts of infimum and supremum are fundamental tools. They provide the precise language needed to talk about boundaries, limits, and continuity, forming the bedrock upon which the entire edifice of mathematical analysis is built. They are a perfect example of how pursuing a simple, intuitive question can lead us to a far richer and more powerful understanding of the world.

Now that we’ve taken the time to carefully build and understand the machinery of the infimum and the supremum, we might be tempted to put these tools back in the mathematician’s toolbox and walk away. After all, they seem a bit abstract, a bit pedantic—what are they really for? To do so would be to miss the whole point. We would be like someone who has learned the grammar of a new language but never listens to its poetry or reads its stories.

The truth is, the concepts of infimum and supremum are not just curiosities of real analysis. They are a fundamental part of the language nature speaks. They appear, sometimes in disguise, in an astonishing array of fields. They provide the precise vocabulary needed to describe structure, to find optimal solutions, to define the very notion of an "area," and even to state the limits of what is possible in the quantum world. So, let's take a journey and see where these ideas lead us. We will find that what at first seemed like a fine distinction is, in fact, a golden thread connecting a vast tapestry of scientific thought.

Perhaps the most intuitive place to see our new tools at work is in the world of discrete structures, where things can be put into a specific order. You might be surprised to learn that you’ve been using infima and suprema since elementary school.

Consider the set of all positive integers that divide the number 360. We can order them by the relation of "divisibility": we say $a \preceq b$ if $a$ divides $b$. In this ordered world, what is the greatest lower bound, or infimum, of the set $\{12, 30, 45\}$? We are looking for the largest number that divides all three. Of course, that's just the ​​greatest common divisor (GCD)​​! And what is their least upper bound, or supremum? It’s the smallest number that is a multiple of all three—the ​​least common multiple (LCM)​​. Here, the abstract concepts of inf and sup reveal themselves to be old, familiar friends. This is no accident. In any system where elements are ordered by divisibility, the infimum is the GCD and the supremum is the LCM.

This idea of finding structure through bounds is far more general. Think about the file system on your computer. We can define an order where one directory is "greater" than another if it's an ancestor. For instance, /home/user/ is "greater" than /home/user/documents/. Let's consider a set of directories, say S = {/usr/bin, /usr/local/bin, /home/admin/bin}. What is their least upper bound (supremum) in this hierarchy? We're looking for the "smallest" directory that contains all of them. Tracing back their paths, we see the only common ancestor for all three is the root directory, /. Thus, sup(S) = /. What about their infimum? We would need a directory that is a subdirectory of all three simultaneously, a logical impossibility. In this system, the infimum of this particular set simply does not exist. This teaches us a valuable lesson: while suprema and infima provide a powerful way to describe structure, their existence depends on the specific rules of the world we are exploring.

Mathematicians, in their spirit of boundless curiosity, even explore what happens in bizarre number systems, like the arithmetic on a 24-hour clock ($\mathbb{Z}_{24}$). If we define divisibility in this finite world, the concepts of infimum and supremum still apply, but they can lead to wonderfully counter-intuitive results that reveal the strange, hidden structure of these systems.

From the discrete world of integers and data structures, we turn to the continuous world of shapes and space. Here, infimum and supremum are the tools we use to describe the boundaries of objects and to solve problems of optimization—the search for the "best" possible outcome.

A simple, beautiful example is a hemisphere of radius $R$. Imagine it sitting on a table. What are the possible heights ($z$-coordinates) of points on its surface? A point can be at height 0 (on the equator) or at height $R$ (at the North Pole), or anywhere in between. The set of all possible heights is the interval $[0, R]$. The infimum of this set is $0$, and its supremum is $R$. Here, inf and sup simply and precisely describe the vertical extent of our geometric object.

But we can ask more subtle questions. Suppose we inscribe a rectangle inside a circle of radius $R$. There are infinitely many such rectangles, from tall and skinny ones to short and wide ones. What are the supremum and infimum of the set of all their possible perimeters? This is an optimization problem in disguise. As you might guess, the perimeter is maximized when the rectangle is a square. This maximum value is the supremum, which turns out to be $4\sqrt{2}R$. But what about the infimum? We can make the rectangle flatter and flatter, with one side's length approaching zero and the other approaching the circle's diameter, $2R$. The perimeter in this case approaches $2(0 + 2R) = 4R$. We can get arbitrarily close to a perimeter of $4R$, but we can never actually have a rectangle with side zero. So, the infimum of the set of perimeters is $4R$, but this value is never attained. There is no inscribed rectangle with the "minimum" perimeter. This is a perfect physical illustration of the crucial difference between an infimum and a minimum!

The true power of infimum and supremum is most profoundly felt in the fields of mathematical analysis and modern physics. Here, they are not just descriptive tools; they are foundational concepts upon which entire theories are built.

Have you ever wondered what it really means to find the "area under a curve"? The method of integration, invented by Newton and Leibniz, was put on a truly rigorous footing by Bernhard Riemann using infima and suprema. The idea is to "trap" the area. First, we fit a series of rectangles entirely under the curve. We can sum up their areas. We can do this in infinitely many ways, creating a set of "under-estimates." The true area must be greater than or equal to all of them. The best possible lower bound for the area, then, is the ​​supremum​​ of the set of all these "under-sums." Then, we do the opposite: we draw rectangles that completely contain the curve. This gives us a set of "over-estimates." The best possible upper bound for the area is the ​​infimum​​ of this set of "over-sums." If the supremum of the under-estimates equals the infimum of the over-estimates, we say the function is integrable, and this common value is the area. Every definite integral you have ever computed rests on this elegant dance of a meeting supremum and infimum.

This idea can be extended from numbers to functions themselves. Imagine you have an infinite sequence of functions, say $f_n(x) = (1 + \frac{(-1)^n}{n})x^2$. For any given $x$, the values $f_n(x)$ bounce up and down as $n$ changes. Can we define a smooth "upper boundary" and "lower boundary" for this entire family of functions? Yes! We can define a new function, $g(x)$, which for each $x$ is the supremum of all the values $\{f_1(x), f_2(x), \dots\}$. We can also define $h(x)$ as the infimum of that set of values. These new functions, $g(x)$ and $h(x)$, act as perfect envelopes for the entire infinite sequence. This leap—from finding the bounds of a set of numbers to finding the bounding functions for a set of functions—is a cornerstone of the field of functional analysis, which provides the mathematical language for quantum mechanics.

In physics, many phenomena, from the flow of heat to the behavior of electrostatic fields, are described by so-called "harmonic functions." A key property of these functions, the Maximum Principle, states that the maximum and minimum values of the function must occur on the boundary of the domain. But what if the domain is infinite, like the entire right-half of a 2D plane? The notions of "maximum" and "minimum" can break down. Once again, infimum and supremum come to the rescue. A physical field in this half-plane might be bounded, say its value never exceeds 1. Yet, by traveling farther and farther out towards infinity, we might get arbitrarily close to 1. In this case, the supremum of the field is 1, even though this value is never attained at any finite point. The infimum, however, might be a true minimum value that is attained at a specific point on the boundary, say at the origin. This is how physics describes bounded fields in infinite spaces.

Finally, we take the ultimate leap into the quantum realm. In the strange world of quantum mechanics, quantities like position and momentum are not simple numbers; they are represented by mathematical objects called operators. When we perform a measurement, the possible outcomes are given by the "spectrum" of the operator. The spectrum is a set of numbers. For the momentum operator $P$ of a particle free to move anywhere, the spectrum is the entire real line, $\mathbb{R}$. Now, what if we decide to measure not the momentum itself, but a quantity represented by the operator $C = \cos(\alpha P)$? According to the rules of quantum mechanics, the spectrum of possible measurement outcomes for $C$ will be the set of all values of $\cos(\alpha x)$ where $x$ is a possible outcome for momentum. Since the momentum spectrum is all of $\mathbb{R}$, the spectrum for $C$ is the set $\cos(\mathbb{R})$, which is simply the closed interval $[-1, 1]$. The infimum of the possible measurement results is -1, and the supremum is 1. We are using the language of bounds to describe the fundamental limits of what we can possibly observe in our universe.

From GCDs to quantum fields, the journey of the infimum and supremum is breathtaking. They are not merely an exercise in pedantry. They are a unifying concept, a precise and powerful language that allows us to find order in chaos, to identify the best and worst of all possibilities, and to delineate the very boundaries of reality.