Relative Homology Groups

While homology offers a powerful method for understanding the shape of a space by counting its "holes," its view can be broad. It struggles with more nuanced questions: when a space $X$ contains a subspace $A$, how does the presence of $A$ influence the structure of $X$? What new features arise in $X$ that are not found in $A$? This knowledge gap calls for a more precise instrument, one capable of dissecting the relationship between a space and its components.

This article introduces ​​relative homology​​, the algebraic tool designed to answer these questions. We will explore how it rigorously captures the features of a space that are "left over" after accounting for a subspace. The journey is divided into two parts. First, we will delve into the "Principles and Mechanisms," defining relative homology groups, unraveling the power of the long exact sequence, and exploring the computational shortcut provided by the Excision Theorem. Following this, the "Applications and Interdisciplinary Connections" chapter will showcase how these theoretical tools are applied to solve concrete problems, from dissecting complex geometric objects to revealing deep connections across different mathematical fields.

In our journey through the landscape of topology, we've learned to use homology to detect and count "holes" in a space, giving us a powerful, albeit blurry, picture of its shape. We can tell a donut from a sphere because one has a hole and the other doesn't. But what happens when our questions become more subtle? What if we have a space $X$ that contains a smaller subspace $A$, and we want to understand the structure of $X$ in relation to $A$? For instance, what new holes are created in $X$ that weren't already there in $A$? Or, how does $A$ "sit inside" $X$? To answer such questions, we need a sharper tool, one that can focus on the difference between the two spaces. This tool is ​​relative homology​​.

Imagine you have a complex machine, $X$. Inside it, there's a well-understood component, $A$. To understand the novel parts of $X$, you might try to mentally "subtract" the features of $A$. Relative homology does exactly this, but with the rigor of algebra.

Recall that homology is built from ​​chains​​, which are formal sums of simple building blocks like points, paths, and triangles. The homology groups $H_n(X)$ are built from the chain groups $C_n(X)$. To define the ​​relative chain groups​​ $C_n(X, A)$, we take a beautifully simple and direct approach: we take all the $n$-chains in $X$ and declare any chain that lies entirely within $A$ to be "trivial" or "zero". In the language of algebra, we take the quotient:

This means we are now working with chains in $X$, but we don't distinguish between two chains if they only differ by a chain in $A$. From these relative chain groups, we can construct the ​​relative homology groups​​ $H_n(X, A)$ in the usual way, by taking "cycles modulo boundaries". These groups now capture the features of $X$ that are not contained in $A$.

Let's test this with the most basic case imaginable. What is the homology of a space relative to itself? Let $X$ be a single point, $\{p\}$, and let the subspace $A$ also be that same point, $\{p\}$. We are asking to measure the features of $\{p\}$ that aren't in $\{p\}$. The answer should, intuitively, be "nothing". And indeed, our definition gives precisely that. Since $X=A$, the chain groups are identical, $C_n(X) = C_n(A)$. The quotient is therefore the trivial group, $C_n(X,A) = 0$, for all $n$. If the chain groups are all zero, so are the homology groups. Thus, $H_n(\{p\}, \{p\}) = 0$ for all $n$. This confirms our intuition: if you subtract a space from itself, nothing is left.

Defining a new object is one thing; understanding its behavior and connections is another. The true power of relative homology is unleashed by a remarkable structure called the ​​long exact sequence of a pair​​. This sequence is the Rosetta Stone that translates between the homologies of $A$, $X$, and the pair $(X, A)$. For any pair $(X, A)$, there exists a sequence of groups and homomorphisms, stretching infinitely in both directions, that ties everything together:

Let's not be intimidated by this long chain of symbols. Think of it as a perfectly balanced system. The "exactness" of the sequence means that at every stage, the stuff flowing out of one map is precisely the stuff that gets "zeroed out" by the next map (the image of one map is the kernel of the next). This tight interlocking relationship allows us to deduce information about one group if we know about its neighbors.

The maps $i_*$ and $j_*$ are quite natural. The map $i_*: H_n(A) \to H_n(X)$ is simply induced by the inclusion of $A$ into $X$. The map $j_*$ is induced by the projection from chains in $X$ to relative chains. The real star of the show is the ​​connecting homomorphism​​, $\partial: H_n(X, A) \to H_{n-1}(A)$. It's the magical link that drops the dimension by one. An element in $H_n(X, A)$ is represented by a chain in $X$ whose boundary lies in $A$. The map $\partial$ simply says, "Okay, take that boundary." This boundary is an $(n-1)$-cycle in $A$, and it represents an element of $H_{n-1}(A)$. This map is what makes the whole structure tick.

What can we do with this engine? For one, we can give a profound meaning to the statement "$H_n(X, A) = 0$". If all the relative homology groups are trivial, the long exact sequence breaks apart into a series of short segments that force the map $i_*: H_n(A) \to H_n(X)$ to be an isomorphism for all $n$. In other words, if there is no "relative homology", it means that from homology's perspective, the spaces $A$ and $X$ are identical! The inclusion of $A$ into $X$ adds no new holes and destroys no old ones.

Let's see this engine in action. Consider a simple path, like the interval $X = [0, 1]$, and its two endpoints, $A = \{0, 1\}$. The interval $X$ is contractible, so its only non-trivial homology is $H_0(X) \cong \mathbb{Z}$. The subspace $A$ consists of two points, so $H_0(A) \cong \mathbb{Z} \oplus \mathbb{Z}$ and its higher homology is trivial. What is the first relative homology group, $H_1(X, A)$? Neither $X$ nor $A$ has a 1-dimensional hole. But plugging what we know into the long exact sequence reveals a surprise:

Since $H_1(A)=0$ and $H_1(X)=0$, the sequence tells us that the map $\partial$ is injective. Its image is the kernel of the next map, $i_*: H_0(A) \to H_0(X)$. This map takes the two points in $A$ and sees them both as part of the single path-component of $X$. Algebraically, it sends a pair of integers $(m,n)$ to their sum $m+n$. The kernel of this map is the set of pairs $(m, -m)$, a group isomorphic to $\mathbb{Z}$. Because $\partial$ is injective, we find that $H_1(X, A) \cong \mathbb{Z}$. A one-dimensional hole has appeared out of nowhere! The generator of this group is the path from 0 to 1, which is a chain in $X$ whose boundary, $\{1\} - \{0\}$, lies in $A$.

It is fascinating to contrast this with the corresponding homotopy group. The relative homotopy set $\pi_1(X, A, 0)$ (with basepoint at 0) asks for paths in $X$ that start in $A$ and end at 0. There are two distinct types of such paths that cannot be deformed into one another: paths that start at 0 and end at 0, and paths that start at 1 and end at 0. So, $\pi_1(X, A, 0)$ is just a set with two elements. Homology, being an abelian theory, can take the "difference" of the two endpoints, creating a cycle. Homotopy just sees two distinct situations. This highlights the unique algebraic perspective that homology brings.

The long exact sequence is powerful, but calculations can still be tedious. Fortunately, there is a wonderful shortcut that often works, based on a simple and intuitive idea: "squashing". If the subspace $A$ is "well-behaved" inside $X$ (forming what's called a ​​good pair​​), we can compute the relative homology $H_n(X, A)$ by taking the whole space $X$, collapsing the subspace $A$ down to a single point, and then computing the homology of the resulting quotient space $X/A$. More precisely, for $n \ge 1$:

Here, $\tilde{H}_n$ denotes ​​reduced homology​​, a slight modification of standard homology that ensures a point has trivial homology in all dimensions. This result is a consequence of the ​​Excision Theorem​​, a deeper theorem which states that you can cut out parts of the interior of $A$ without changing the relative homology groups. Collapsing $A$ is the ultimate excision. This principle allows us to transform a problem about a pair of spaces into a problem about a single, often simpler, space.

The canonical example is the pair $(D^n, S^{n-1})$, an $n$-dimensional disk and its $(n-1)$-dimensional boundary sphere. What is $H_k(D^n, S^{n-1})$? Let's use the power of collapse. Imagine a cloth disk, $D^2$. If you grab the entire circular boundary, $S^1$, and pinch it together to a single point, what do you get? You get a little pouch, which is topologically a 2-sphere, $S^2$. In general, $D^n / S^{n-1}$ is homeomorphic to $S^n$. Therefore:

Since we know the homology of a sphere is just $\mathbb{Z}$ in dimension $n$ and 0 otherwise, we immediately find that $H_k(D^n, S^{n-1})$ is $\mathbb{Z}$ for $k=n$ and trivial for all other $k$. This single, elegant result is a cornerstone of algebraic topology, forming the basis for proving deep theorems like the Brouwer fixed-point theorem.

This "collapsing" idea also provides a lovely interpretation for the relative homology with respect to a single point, $A=\{x_0\}$. The quotient space $X/\{x_0\}$ is just $X$ itself, but with a special "basepoint". The theorem for good pairs then tells us that $H_n(X, \{x_0\}) \cong \tilde{H}_n(X)$ for all $n \ge 0$. This gives a concrete meaning to reduced homology: it is simply the homology of a space relative to a point inside it.

Let's try one more example. Let $X$ be a 2-sphere with a circle attached at one point ($S^2 \vee S^1$), and let $A$ be the circle. What is $H_2(X, A)$? If we collapse the circle $A$ to a point, the sphere is all that's left. So we expect $H_2(X, A) \cong \tilde{H}_2(S^2) \cong \mathbb{Z}$. And indeed, a careful calculation with the long exact sequence confirms this prediction. The shortcut works beautifully.

Armed with the long exact sequence and the collapsing principle, we can start to analyze more complicated structures by breaking them down into parts we understand. For instance, relative homology respects disjoint unions in the most straightforward way imaginable. If you have a pair $(X, A)$ which is the disjoint union of two other pairs, $(X_1, A_1)$ and $(X_2, A_2)$, then the homology simply adds up:

This means we can compute the relative homology of a complex, disconnected space by analyzing each connected piece separately and then combining the results. This is a powerful computational principle, allowing us to apply our knowledge of simple pairs to build up an understanding of more elaborate ones. For instance, computing the Betti numbers $\beta_0, \beta_1, \beta_2$ for a space made of a sphere and a torus, relative to a subspace made of two points on the sphere and a circle on the torus, might seem daunting. But using additivity, we can calculate the homology for the sphere pair and the torus pair independently and just sum their Betti numbers to get the final answer, which results in Betti numbers $(\beta_0, \beta_1, \beta_2)$ of $(0, 2, 1)$.

Finally, these tools reveal connections that are as surprising as they are profound. Consider a continuous map $f: X \to Y$. We can construct a new space called the ​​mapping cone​​, $C_f$, by taking a cylinder over $X$ and "gluing" one end of it to $Y$ according to the map $f$. The space $Y$ sits inside this new cone. The relative homology groups of the pair $(C_f, Y)$ might seem hopelessly complicated, depending on $X$, $Y$, and the map $f$. Yet, an elegant application of the principles we've discussed reveals an astonishingly simple relationship:

The relative homology of the cone pair is completely determined by the homology of the original space $X$, just shifted down by one dimension! The space $Y$ and the specific map $f$ seem to vanish from the final formula, their influence entirely absorbed into the structure of the long exact sequence that produces this result.

This is the beauty of relative homology. It starts with a simple question—"what is left over?"—and provides a framework of such power and elegance that it not only solves the initial problem but also uncovers a deep, hidden unity in the world of shapes, connecting spaces and maps in ways we never would have anticipated.

Having acquainted ourselves with the principles and mechanisms of relative homology, we might be tempted to view it as a mere technical generalization—a clever bit of algebraic machinery. But that would be like looking at a master artist's brush set and seeing only wood and hair. The real magic lies not in the tools themselves, but in what they allow us to see and create. Relative homology is a lens that sharpens our vision, allowing us to perceive the subtle relationships, hidden structures, and dynamic changes within the world of shapes. It is the art of understanding a space by considering what it is relative to something else. Let us now embark on a journey to see this art in action.

One of the most intuitive applications of relative homology is its ability to simplify a complex space by "ignoring" a part of it. Imagine you have a detailed map of a city, and you're only interested in the main highway network. You might mentally "collapse" all the local streets and buildings, treating entire neighborhoods as single points, to see the large-scale structure more clearly. Relative homology does this with mathematical precision.

When we study the homology of a pair $(X, A)$, we are, in a sense, asking what the homology of $X$ would look like if the subspace $A$ were crushed down to a single point. Consider the surface of a cube, which is topologically a sphere $S^2$. Let's call this space $X$. Now, pick one of its faces, a solid square which is topologically a disk $D^2$, and call it $A$. The absolute homology of the face $A$ is trivial (it has no holes). The homology of the sphere $X$ tells us it contains a single two-dimensional "void". What does the relative homology $H_*(X, A)$ tell us? The calculation reveals that the only non-trivial group is $H_2(X, A) \cong \mathbb{Z}$. This is precisely the homology of a 2-sphere! By looking at the sphere relative to the disk, we have effectively cancelled out the disk and recovered the essential "sphereness" of the space. This powerful idea—that for a well-behaved pair, $H_n(X, A)$ is the same as the reduced homology of the quotient space $\tilde{H}_n(X/A)$—is a cornerstone of many computations.

This principle becomes a formidable construction tool when dealing with spaces built in stages, like CW complexes. Imagine building a torus, $T^2$, by starting with a point, adding two circular loops ($a$ and $b$) to form a figure-eight, and then stretching a 2-dimensional sheet over it. The figure-eight forms the 1-skeleton, let's call it $A$. By computing the homology of the torus relative to its skeleton, $H_*(T^2, A)$, we isolate the contribution of the final step. The calculation becomes astonishingly simple and shows that the only new feature is a single generator in the second relative homology group, $H_2(T^2, A) \cong \mathbb{Z}$. Relative homology allows us to see exactly how each new piece we attach creates or fills in holes.

Some of the most profound laws of nature describe the relationship between a region and its boundary. Relative homology provides the perfect language for this dialogue. Consider a solid torus $X = D^2 \times S^1$ (think of a donut) and its boundary surface $A = S^1 \times S^1$. The boundary is a hollow torus with a 1D hole (the circle running through the center) and a 2D hole (the interior void). The solid torus, however, is "filled in". How does algebra capture this?

The long exact sequence for the pair $(X, A)$ acts as a bridge. It tells us that the second relative homology group, $H_2(X, A)$, is isomorphic to $\mathbb{Z}$. This non-trivial group is a witness to the fact that the 2-dimensional void of the boundary $A$ gets filled by the solid interior $X$. It precisely detects that the "meridian" circle on the torus surface (a small loop around the tube) can be shrunk to a point inside the solid torus, but the "longitudinal" circle (the large loop around the donut's hole) cannot. Relative homology doesn't just count holes; it understands their context and fate.

This principle extends to more complex situations. Take a cylinder built on a sphere, $X = S^2 \times I$, with its boundary $A$ being the two spheres at each end. Neither the cylinder (which is just a thick sphere) nor its boundary (two separate spheres) has any 3-dimensional features. Yet, the relative homology group $H_3(X, A)$ is found to be $\mathbb{Z}$! This group doesn't belong to the interior or the boundary alone; it belongs to the relationship between them. It represents the 3-dimensional "connection" spanning from one boundary component to the other. It is the algebraic embodiment of the cylinder's volume.

The power of relative homology is not confined to abstract geometric shapes. It provides surprisingly effective tools for concrete problems, such as distinguishing knots and links. Let's consider one of the simplest and most elegant links: the Hopf link, which consists of two disjoint circles, $A$, linked once in 3-dimensional space, $\mathbb{R}^3$.

Our space $X$ is $\mathbb{R}^3$, which is contractible—from a topological standpoint, it's as trivial as a point. Its absolute homology is zero in all interesting dimensions. So, how can it tell us anything? By looking at the space relative to the link, $H_*(\mathbb{R}^3, A)$, the link's structure is thrown into sharp relief. The long exact sequence performs a beautiful trick: because the homology of $\mathbb{R}^3$ is trivial, it provides a direct isomorphism between the relative homology of the pair and the homology of the link itself (shifted in dimension). We find that $H_2(\mathbb{R}^3, A) \cong H_1(A) \cong \mathbb{Z} \oplus \mathbb{Z}$. The rank of this group is two, precisely the number of circles in our link! In essence, we are using the vast, featureless expanse of $\mathbb{R}^3$ as a backdrop to illuminate the structure of the object it contains.

Relative homology truly shines when it reveals the deep, unifying themes that run through different branches of mathematics and physics. It acts as a universal language, translating ideas between seemingly disparate fields.

​​Fiber Bundles and Spacetime:​​ Many fundamental structures in physics and geometry are described as fiber bundles, where a complex space (the "total space") is built by assembling simpler spaces (the "fibers") over a "base space." A stunning example is the quaternionic Hopf fibration, which describes the 7-sphere, $S^7$, as a bundle of 3-spheres, $S^3$, over a 4-sphere base, $S^4$. By studying the pair $(S^7, S^3)$, where we look at the total space relative to a single fiber, relative homology elegantly dissects this intricate structure. The resulting relative homology groups, $H_4(S^7, S^3) \cong \mathbb{Z}$ and $H_7(S^7, S^3) \cong \mathbb{Z}$, turn out to be precisely the homology groups of the base space $S^4$. The long exact sequence of the pair effectively untangles the fiber from the total space, leaving behind the pure structure of the base.

​​Duality and Negative Space:​​ There is a profound symmetry in topology known as duality, which relates the properties of a space to the properties of its complement. Relative homology is a key player in this principle. For instance, Alexander Duality provides a dictionary that translates the homology of a subspace $V \subset \mathbb{R}^n$ into the homology of its complement, $\mathbb{R}^n \setminus V$. In its relative form, it allows us to compute the relative homology of complements, which can be much more complex, by relating them back to the simpler homology of the original objects. It's like understanding a sculpture by studying the shape of the air around it.

​​The Art of Embedding:​​ Finally, relative homology is exquisitely sensitive not just to what a space is, but how it sits inside another. A circle is always a circle, but a circle drawn on a sphere is different from a circle drawn on a torus. Consider a torus $T^2$ and the diagonal circle $\Delta$ that wraps around it once in each direction. The relative homology group $H_1(T^2, \Delta)$ turns out to be $\mathbb{Z}$. It captures the fact that the two fundamental loops of the torus become identified when we quotient by the diagonal. Now contrast this with a torus $A$ embedded inside a larger space $X = S^2 \times S^1$. Here, the relative homology group $H_2(X, A)$ is non-trivial, revealing subtle geometric facts about how the torus boundary is "trivial" in one sense but carves out a complex region in another. These algebraic invariants serve as a sophisticated fingerprint for the embedding itself. Similarly, we can understand how more complex surfaces are constructed, like a genus-2 surface built from gluing two punctured tori, by analyzing one part relative to the whole.

From collapsing disks to dissecting fiber bundles, from unknotting links to understanding boundaries, the applications of relative homology are as diverse as they are profound. It is a testament to the power of a simple idea: that to truly understand an object, you must also understand its relationship to the world around it. It is a mathematical lens that, once you learn how to use it, changes the way you see everything.