Image of a Function

A function is one of the most fundamental concepts in mathematics, acting as a rule that assigns a unique output to every given input. While we often focus on the rule itself, a deeper understanding comes from examining the results of this process. The crucial question is not just about the set of potential outputs, known as the codomain, but about the specific collection of values the function actually generates—its ​​image​​. This distinction addresses a core knowledge gap: how does a function's structure transform an input space into an output space, and what properties are preserved, lost, or created in this transformation?

This article explores the profound concept of a function's image. In the first section, ​​"Principles and Mechanisms,"​​ we will define the image, distinguish it from the codomain using concrete examples, and uncover how properties like continuity dictate the very shape of the output set. We will then see how these principles lead to foundational results like the Intermediate Value and Extreme Value Theorems. Following this, the ​​"Applications and Interdisciplinary Connections"​​ section will reveal how the image serves as a powerful diagnostic tool, a "fingerprint" that provides insights in fields ranging from optics and computer science to topology and complex analysis. Let's begin by dissecting the machinery of a function to understand what its image truly represents.

Imagine a function as a machine, a sort of mathematical device. You feed it an input from a specified collection of objects, called the ​​domain​​, and for each input, it produces a single, definite output. The collection of all possible outputs that this machine can produce is what we call the ​​image​​ of the function. It's a simple idea, but as we'll see, it's one of the most profound in all of mathematics, linking simple arithmetic to the very shape of space.

Let's start with a very down-to-earth example. Consider a function that takes a month of the year as its input and outputs the number of days in that month (in a non-leap year). The domain is the set of twelve months. The machine's instruction manual might say it outputs a natural number, $\mathbb{N} = \{1, 2, 3, \dots\}$. This larger set of potential outputs is called the ​​codomain​​. But what does the machine actually produce?

If you feed it "February," it outputs 28. If you feed it "April," it outputs 30. For "January," it gives 31. After trying all twelve months, you'll find that the only numbers ever produced are 28, 30, and 31. This collection, $\{28, 30, 31\}$, is the function's image. Notice how small the image is compared to the entire codomain of natural numbers! The function is picky; it doesn't produce 29, or 5, or 365. When a function's image does fill its entire codomain, we call the function ​​surjective​​, or "onto." Our month function is clearly not surjective onto the set of all natural numbers.

This distinction between the potential outputs (codomain) and the actual outputs (image) is crucial. Let's take a more abstract machine, one defined by the rule $f(n) = \gcd(n, 12)$, where $\gcd$ stands for the greatest common divisor. We feed it integers from 1 to 15. What is its image? The outputs must be divisors of 12, so the possibilities are limited to $\{1, 2, 3, 4, 6, 12\}$. A quick check shows that we can indeed produce each of these values (for example, $f(12) = \gcd(12, 12) = 12$ and $f(7) = \gcd(7, 12) = 1$). So, the image is precisely the set of divisors of 12.

Sometimes, the image can have surprising "gaps." Consider a function from the real numbers $\mathbb{R}$ to the integers $\mathbb{Z}$ defined by $f(x) = (\lfloor x \rfloor)^2 - 4\lfloor x \rfloor + 3$, where $\lfloor x \rfloor$ is the floor function (the greatest integer less than or equal to $x$). Since the value of $f(x)$ depends only on the integer part of $x$, let's see what integers it can produce. If we let $n = \lfloor x \rfloor$, the output is $n^2 - 4n + 3$. By completing the square, this is $(n-2)^2 - 1$. As $n$ runs through all integers, $m = n-2$ also runs through all integers. The image is therefore the set of numbers $\{m^2 - 1 \mid m \in \mathbb{Z}\}$. This set includes $-1$ (for $m=0$), $0$ (for $m=\pm 1$), $3$ (for $m=\pm 2$), and so on. But it never produces 1, because that would require $m^2 = 2$. It never produces 2, as that would need $m^2=3$. The image is an infinite set of integers, but it's riddled with holes, so the function is not surjective.

What happens when our function operates on a continuous domain, like the entire real number line? The image often transforms in beautiful and sometimes unexpected ways.

Take the simple-looking function $f(x) = \frac{1}{x^2 + 1}$. The domain is $\mathbb{R}$, an infinitely long line stretching from $-\infty$ to $+\infty$. What does the image look like? Let's reason it out. The term $x^2$ can never be negative; its smallest value is 0. So, the denominator, $x^2 + 1$, has a minimum value of 1. This means the function's maximum value is $\frac{1}{0^2+1} = 1$. As $x$ gets very large (either positive or negative), $x^2+1$ becomes huge, so $\frac{1}{x^2+1}$ gets incredibly close to 0. But it never actually reaches 0. The function takes the entire infinite line and squashes it into the small, half-open interval $(0, 1]$. It paints a finite picture from an infinite canvas.

This kind of analysis, looking at the structure of a formula, is a powerful tool. A more complex function like $f(x) = \frac{x^2 - 4x + 9}{x^2 - 4x + 5}$ can be tamed with a bit of algebra. Completing the square reveals it as $f(x) = \frac{(x-2)^2 + 5}{(x-2)^2 + 1}$. By letting $u = (x-2)^2$, which can be any non-negative number, the function becomes $y = \frac{u+5}{u+1}$. This is the same sort of structure we just saw! The maximum occurs when $u=0$ (giving $y=5$), and as $u$ approaches infinity, $y$ approaches 1. The seemingly complicated function has a simple image: the interval $(1, 5]$.

We can even chain these machines together. Consider the function $h(x) = \cos(\frac{\pi}{2}(x - \lfloor x \rfloor))$. This is a composite function. The inner part, $f(x) = x - \lfloor x \rfloor$, calculates the fractional part of a number. No matter what real number $x$ you feed it, the output is always in the interval $[0, 1)$. This first machine takes the infinite real line and wraps it endlessly around a segment of length one. The output of this machine then becomes the input for the second machine, $g(y) = \cos(\frac{\pi}{2} y)$. As its input $y$ varies over $[0, 1)$, the cosine function sweeps down from $\cos(0)=1$ towards $\cos(\pi/2)=0$. Because the input $y$ never quite reaches 1, the output never quite reaches 0. The final image of the composite function $h(x)$ is therefore $(0, 1]$.

So far, we've been finding images. But now we ask a deeper question: Are there general rules that govern the shape of an image? The answer is a resounding yes, and it lies in the concept of ​​continuity​​. A continuous function is, informally, one you can draw without lifting your pen. This simple property has profound consequences for the image.

Connectedness and the Intermediate Value Theorem

A set is ​​connected​​ if it's all in one piece. The interval $[0, 2]$ is connected. The set $[0, 1] \cup [2, 3]$ is ​​disconnected​​; it has a gap. The first great theorem of continuous images is this: ​​the continuous image of a connected set is connected.​​

This is just a fancy way of stating the familiar ​​Intermediate Value Theorem​​. If you draw a continuous line from a height of $y=A$ to a height of $y=B$, you must pass through every height in between. A continuous function cannot create gaps out of thin air.

To see this principle in stark relief, consider the function $f(x) = x^3 - 10x$ acting on the disconnected domain $X = [-3, -2] \cup [2, 3]$. This domain is explicitly in two separate pieces. What does the image look like? On the first piece, $[-3, -2]$, the function is increasing and its values range from $f(-3)=3$ to $f(-2)=12$. So the image of this piece is the interval $[3, 12]$. On the second piece, $[2, 3]$, the function is also increasing, and its values range from $f(2)=-12$ to $f(3)=-3$. The image of this piece is $[-12, -3]$. The total image is the union of these two results: $f(X) = [-12, -3] \cup [3, 12]$. Just like the domain, the image is in two pieces! There is a huge gap between -3 and 3 that contains no output values. The function faithfully reproduced the disconnectedness of its domain.

The logic works in reverse, too. If we have a continuous function and we find that its image is disconnected, we can be absolutely certain that its domain must have been disconnected as well.

Compactness and the Extreme Value Theorem

Another fundamental property a set can have is ​​compactness​​. In the familiar world of Euclidean space, a compact set is one that is both ​​closed​​ (it includes its boundary points) and ​​bounded​​ (it doesn't run off to infinity). The interval $[0, 1]$ is compact. The sphere $S^2 = \{(x, y, z) \mid x^2+y^2+z^2=1\}$ is compact. However, the interval $[0, 1)$ is not compact because it's missing its boundary point at 1. The entire real line $\mathbb{R}$ is not compact because it's unbounded.

The second great theorem is: ​​the continuous image of a compact set is compact.​​

This is the generalization of the ​​Extreme Value Theorem​​, which says that any continuous function on a closed, bounded interval $[a, b]$ must achieve a maximum and a minimum value. In other words, its image will be the closed, bounded interval $[m, M]$.

Let's see this in action on the unit sphere, $S^2$. The sphere is a beautiful compact object. Let's define a function on it, say $f(x, y, z) = x^2 + 2y^2 + 3z^2$. Without doing any hard calculations, our theorem guarantees that the image must be a closed interval $[m, M]$. The function must attain a minimum value $m$ and a maximum value $M$ somewhere on the sphere. A more detailed analysis (using a technique called Lagrange multipliers) shows that the minimum value is 1 (occurring at $(\pm 1, 0, 0)$) and the maximum is 3 (occurring at $(0, 0, \pm 1)$). Because the domain (the sphere) is connected, the image must be the entire interval between these extremes. So, the image is $[1, 3]$. It is indeed a compact set.

What happens if the domain is not compact? Then all bets are off. Let's take the non-compact interval $I = [0, 1)$. It's bounded, but not closed. Can we find a continuous function on $I$ whose image is not bounded? Absolutely. The function $f(x) = \frac{1}{1-x}$ is continuous on this interval. As $x$ gets closer and closer to the missing endpoint 1, the denominator $1-x$ gets closer to 0, and the function's value shoots off to infinity. The image is $[1, \infty)$, which is an unbounded set. The "hole" at the edge of the domain allowed the function to escape to infinity.

The study of a function's image, then, is not just a matter of calculation. It is a window into the deep relationship between a function and its domain. By understanding the image, we understand how a function transforms, preserves, or breaks the fundamental geometric properties of the spaces it acts upon, revealing a hidden and beautiful unity across all of mathematics.

After dissecting the machinery of functions, we might be tempted to put these ideas in a box labeled "abstract math" and move on. But that would be like learning the rules of grammar without ever reading a poem. The true beauty of a concept like the image of a function reveals itself not in its definition, but in what it does. The image is where the abstract rubber meets the road of reality. It's the tangible output, the observable phenomenon, the solution to the puzzle. It's not just a set of values; it's a story told by the function.

Let's begin with a wonderfully concrete example from a field that has "image" in its very name: optics. When you look through a camera lens, the lens acts as a function. Its input is the light coming from an object in the world (the "object plane"), and its output is the pattern of light focused onto the camera's sensor (the "image plane"). If your input is an idealized, infinitely small point of light, the output isn't another perfect point. Due to the physics of diffraction, the light spreads out into a small, blurry pattern. This pattern, the observed distribution of light in the image plane, is called the Point Spread Function (PSF). It is, quite literally, the image of a point source. The reason we describe the PSF using the coordinates of the image plane is because that's where it is—it's the output, the result of the function's action. The image of a function is, by its very nature, a description of the output space.

This input-output way of thinking extends far beyond the physical world. The image acts as a "fingerprint," revealing the specific character of the function or the system it models.

Consider the software that compiles computer code. In a simplified model of this process, a function might take a data type (like "char" or "int") as input and output the number of bytes of memory it requires. While a computer could theoretically use any positive integer number of bytes, a specific system might only use 1, 4, and 8 bytes for its fundamental types. The set of actual output values, {1, 4, 8}, is the function's image. This image is a concise summary of the system's design choices, a fingerprint distinguishing it from a different system that might use {1, 2, 4, 8} bytes instead.

This idea even finds a home in the abstract realm of number theory and algebra. Imagine a function defined on the numbers in a finite field, like the integers modulo 23. A seemingly simple polynomial function like $f(x) = x - x^2$ doesn't necessarily produce every number from 0 to 22 as an output. When we map all 23 possible inputs, we find that the image contains only 12 distinct values. Why 12? The answer is tied to the deep structure of the field—specifically, which numbers are "perfect squares." The image, once again, is a fingerprint, revealing hidden algebraic properties of the space it acts upon.

In some cases, the fingerprint can be surprisingly large. Consider the set of all possible square matrices of a certain size. There's a simple function called the trace, which just sums up the numbers on the main diagonal. What is the image of the trace function? Can it produce any number? For any real number $c$ you can dream of, it's trivial to construct a matrix whose diagonal entries sum to $c$. You could just put $c$ in the top-left corner and zeros everywhere else. This means the image of the trace function is the entire set of real numbers. The function is powerful enough to "reach" any point in its codomain; we call such a function surjective.

But most functions are more constrained. Think of the function $f(x) = \frac{x^2}{1+x^2}$. No matter what real number $x$ you plug in, the output is always trapped between 0 (inclusive) and 1 (exclusive). The image is the interval $[0, 1)$. The function gets tantalizingly close to 1 as $x$ gets very large, but it never actually reaches it. The points $0$ and $1$ form the "boundary" of the image, and they tell us about the function's minimum value and its limiting behavior. The shape of the image is a geometric portrait of the function's global properties.

Here we arrive at one of the most profound ideas in all of mathematics. The true power of studying a function's image comes from a simple question: What properties of the input space (the domain) are preserved in the output space (the image)? The answer reveals a beautiful unity between seemingly disparate fields like analysis and topology.

Let's start with a property we can all intuit: connectedness. An interval like $[0, 1]$ is connected; it's a single, unbroken piece. What happens if we apply a continuous function—one whose graph you can draw without lifting your pen—to this interval? The magic is that the image must also be a single, unbroken piece. The function can stretch, squash, or fold the interval, but it cannot tear it into separate parts. If you take two continuous functions, $f$ and $g$, on $[0,1]$ and look at their difference, $h(x) = f(x) - g(x)$, the function $h$ is also continuous. Its domain, $[0,1]$, is connected and "closed and bounded" (a property called compactness). The consequence is astonishingly specific: the image of $h$ must be a closed and bounded interval, like $[a, b]$. This is the very soul of the Intermediate Value Theorem, which guarantees that if you have a continuous function that starts below a certain value and ends above it, it must cross that value somewhere in between.

This principle of preserving connectedness is not just a parlor trick; it's a powerful tool. Let's generalize. Any connected space, when viewed through the lens of a continuous function, produces a connected image. Now, what if the output space is inherently disconnected? Consider the set of integers, $\mathbb{Z}$, where each integer is an isolated island. If we have a continuous function from a connected space—say, the surprisingly intricate space of all $2 \times 2$ matrices with determinant 1—to the integers, what can its image be? Since the image must be connected, and the only connected pieces of the integers are single points, the image must be a single integer! The function has to be constant. The demand that the image be connected collapses all possibilities down to one.

We can wield this principle to prove the existence of things we cannot see. Imagine a connected landscape $X$ containing two disjoint, closed regions, $A$ and $B$. Is there a point in the landscape that is exactly the same distance from $A$ as it is from $B$? We can define a function that, for any point $x$ in the landscape, outputs a pair of numbers: its distance to $A$ and its distance to $B$. This function, $f(x) = (d(x,A), d(x,B))$, maps our landscape into a 2D plane. Because the distance functions are continuous and our landscape $X$ is connected, the image $f(X)$ must be a connected set in the plane. Now, for any point in region $A$, its distance to $A$ is 0 and its distance to $B$ is positive. For any point in region $B$, the reverse is true. By analyzing an auxiliary function $g(x) = d(x,A) - d(x,B)$, we use the preservation of connectedness to show that there must be some point $x_0$ where $g(x_0)=0$. This means $d(x_0, A) = d(x_0, B)$. The image of our original function $f$ must cross the line where the two coordinates are equal. We have proven a point of equilibrium exists, just by insisting that a continuous function cannot tear space apart.

Connectedness is not the only property that is preserved. Another is compactness, a sort of generalization of being "closed and bounded." If we map a compact space continuously, its image is also compact. Now, let's revisit our "island" space, the integers (or any space with the discrete topology). In such a space, the only compact sets are the finite ones. Therefore, any continuous function from a compact space (like our interval $[0,1]$) to a discrete space must have a finite image. Again, the properties of the domain and codomain severely constrain the possible outputs.

In the shimmering world of complex analysis, the rules are even stricter. Functions that are "analytic" (differentiable in the complex sense) are incredibly rigid. The Open Mapping Theorem states that if you take a non-constant analytic function and apply it to an open set (a region, not including its boundary), the image is also an open set. This is not true for general real functions! This theorem tells us that an analytic function can't map a 2D domain onto a 1D line, for example, because a line is not an open set in the complex plane. The topology of the image becomes a powerful diagnostic tool for the nature of the function itself.

From the practicalities of compiler design and optics to the deepest theorems of topology and analysis, the concept of an image is the bridge. It connects the world of inputs to the world of outputs, the abstract rule to the concrete result. It allows us to understand a function's behavior, to classify it, and to use its properties to deduce profound truths. So the next time you encounter a function, don't just ask what it does. Ask: What does it create? What is its image? For in that image, you may find a picture of the universe.