Homotopy Class

In the world of geometry, when are two shapes or two paths considered "the same"? While rigid geometry focuses on exact distances and angles, topology offers a more flexible and profound perspective. It studies the properties of spaces that are preserved under continuous stretching and bending, without tearing or gluing. The central concept that formalizes this intuitive idea is homotopy, a powerful tool for understanding the essential structure of a space. This article addresses the fundamental question of how we can classify spaces and the maps between them by considering them equivalent up to continuous deformation.

This article will guide you through the core ideas of homotopy theory. First, in "Principles and Mechanisms," we will explore the formal definition of homotopy, see how it gives rise to homotopy classes, and discover its most important application: the fundamental group, an algebraic object that captures the essence of loops in a space. We will also venture into higher dimensions to understand the surprising properties of higher homotopy groups. Following this, the "Applications and Interdisciplinary Connections" chapter will reveal how these abstract concepts provide a powerful language for classifying geometric objects and describing fundamental aspects of reality, from the structure of quantum fields to the principles of modern physics.

Imagine you have a map drawn on a sheet of rubber. You can stretch, twist, and squish the sheet as much as you like, as long as you don't tear it. The cities and roads drawn on it will move and distort, but the fundamental layout—which cities are connected to which—remains the same. This simple idea of continuous deformation is the heart of ​​homotopy​​. In topology, we don't care about rigid distances or straight lines; we care about the properties that survive this kind of transformation. Homotopy is the mathematical tool we use to make this intuitive idea precise. It allows us to classify spaces and the maps between them, asking a profound question: when are two things "the same" in the eyes of topology?

Let's formalize this a bit. Suppose we have two maps, $f$ and $g$, from a space $X$ to a space $Y$. We say $f$ and $g$ are ​​homotopic​​ if we can find a "movie" or a continuous family of maps that starts at $f$ and ends at $g$. We call this movie a ​​homotopy​​, and we can write it as a function $H: X \times [0, 1] \to Y$. Here, the parameter $t \in [0, 1]$ is like the time variable in our movie. At time $t=0$, we have the map $f$ (i.e., $H(x, 0) = f(x)$), and at time $t=1$, we have the map $g$ (i.e., $H(x, 1) = g(x)$). The condition that $H$ must be continuous ensures that the transformation is smooth, without any sudden jumps or "tears."

Homotopy defines an equivalence relation: any map is homotopic to itself (reflexivity); if $f$ is homotopic to $g$, then $g$ is homotopic to $f$ (symmetry); and if $f$ is homotopic to $g$ and $g$ is homotopic to $h$, then $f$ is homotopic to $h$ (transitivity). This relation partitions the set of all continuous maps from $X$ to $Y$ into disjoint families called ​​homotopy classes​​.

A simple, yet revealing, example helps illustrate this. Consider maps from a space $X$ consisting of just two points, say $\{a, b\}$, to some other space $Y$. A map $f$ is just a choice of two points in $Y$: $f(a)$ and $f(b)$. When is one such map, $f$, homotopic to another, $g$? The homotopy $H$ must provide a continuous path from $f(a)$ to $g(a)$ and, simultaneously, a continuous path from $f(b)$ to $g(b)$ inside $Y$. This means $f$ and $g$ are homotopic if and only if $f(a)$ and $g(a)$ are in the same path-component of $Y$, and likewise for $f(b)$ and $g(b)$. If $Y$ is path-connected (like a sphere or a plane), any point can be connected to any other. In that case, any map $f$ is homotopic to any other map $g$, and there is only a single homotopy class. The space of maps is, from a homotopy perspective, trivial. But if $Y$ is disconnected, say two separate islands, then the number of homotopy classes explodes, depending on which islands the points land on.

This already tells us something powerful: homotopy classes of maps detect the connectedness of the target space.

Our intuition about "stretching a rubber sheet" works beautifully for spaces like planes and spheres, but it can be misleading. The topological properties of the spaces are paramount. Let's consider a bizarre scenario: what are the homotopy classes of maps from the set of integers, $\mathbb{Z}$, to itself, where we give $\mathbb{Z}$ the ​​discrete topology​​? In this topology, every point is its own isolated island, completely open and separate from all others.

Now, a homotopy is a map $H: \mathbb{Z} \times [0, 1] \to \mathbb{Z}$. For any integer $n \in \mathbb{Z}$, the little slice $\{n\} \times [0, 1]$ is a connected line segment. A continuous map must send a connected set to a connected set. But in our codomain $\mathbb{Z}$, the only connected sets are single points! This forces the image of $\{n\} \times [0, 1]$ to be a single integer. In other words, for a fixed $n$, the value of $H(n, t)$ cannot change as $t$ varies from $0$ to $1$. This means $H(n, 0)$ must equal $H(n, 1)$, which is to say $f(n) = g(n)$ for all $n$.

The astonishing conclusion is that two maps are homotopic if and only if they are the exact same map. No deformation is possible! Each function from $\mathbb{Z}$ to $\mathbb{Z}$ lives in its own solitary homotopy class. Since there are uncountably many such functions, there are uncountably many homotopy classes. This extreme example demonstrates a crucial principle: the topology of the domain can "freeze" the possibility of deformation, shattering our simple rubber-sheet analogy and forcing us to rely on the rigor of the mathematical definition.

The most fruitful application of homotopy comes when we focus on a special kind of map: a path. A path in a space $X$ is simply a map from the interval $[0, 1]$ to $X$. The homotopy of paths, with the extra condition that the endpoints remain fixed, tells us about the structure of the space itself.

Consider a space made of two disjoint open disks in a plane. If we take two points $p$ and $q$ in the same disk, any path from $p$ to $q$ can be continuously deformed into any other. The disk is convex, so we can just linearly interpolate between the two paths. There is only one ​​path homotopy class​​. But what if $p$ is in the first disk and $q$ is in the second? A path is the image of a connected set, $[0, 1]$, and must therefore be connected. Since the two disks are separate, no continuous path can possibly bridge the gap. The set of paths from $p$ to $q$ is empty. Again, path homotopy detects the path components of a space.

Things get really interesting when the space is connected but has "holes." The classic example is the plane with the origin removed, $X = \mathbb{R}^2 \setminus \{(0,0)\}$. Let's consider paths from $p=(1,0)$ to $q=(-1,0)$. One path can go over the top of the origin, through the upper half-plane. Another can go underneath, through the lower half-plane. Can you deform the "top" path into the "bottom" path without crossing the forbidden origin? No! They are in different homotopy classes. What about a path that goes around the origin once and then travels to $q$? That's yet another distinct class. In fact, there is an infinite family of homotopy classes, indexed by an integer that counts how many times the path winds around the hole.

This idea of "winding" is best captured by looking at ​​loops​​: paths that start and end at the same point, say $p=(1,0)$. We can combine two loops: first traverse loop $\gamma_1$, then traverse loop $\gamma_2$. This operation, called ​​concatenation​​, gives us a new loop, $\gamma_1 * \gamma_2$. It turns out that the set of homotopy classes of loops at a point $p$ forms a ​​group​​ under this operation. This is one of the most fundamental ideas in algebraic topology, and the resulting group is called the ​​fundamental group​​, denoted $\pi_1(X, p)$.

• ​​Closure:​​ Concatenating two loops gives another loop.
• ​​Associativity:​​ Traversing $(\gamma_1 * \gamma_2) * \gamma_3$ is not strictly the same as $\gamma_1 * (\gamma_2 * \gamma_3)$—the timing is different—but they are homotopic. It's just a matter of re-parameterizing our journey.
• ​​Identity:​​ The "do-nothing" loop that just stays at the basepoint $p$ acts as the identity element.
• ​​Inverse:​​ For any loop $\gamma$, its inverse is the loop that traverses the same path in reverse. Concatenating them, $\gamma * \gamma^{-1}$, results in a journey that goes out and immediately comes back, which can be continuously shrunk back to the starting point.

The basepoint $p$ is crucial. It acts as a common port for all our journeys. If the space is not path-connected, the group of loops you can form depends entirely on which connected "island" your basepoint lives on. Furthermore, even in a connected space, a fixed basepoint provides the necessary structure to define the group operation cleanly. If we allow loops to start and end anywhere (so-called "free homotopy"), we lose the group structure and are left with a more complicated set of conjugacy classes. However, if the space is "trivial" in a certain sense—if it can be continuously shrunk to its basepoint while keeping the basepoint fixed (a ​​based contractible​​ space)—then all loops are homotopic to the constant loop, and the fundamental group is trivial. Any map into such a space is homotopically trivial.

The fundamental group $\pi_1(X)$ measures the 1-dimensional "holes" in a space by using 1-dimensional probes (loops, which are maps from $S^1$). What if we use higher-dimensional probes? We can define the $n$-th homotopy group, $\pi_n(X, x_0)$, as the set of based homotopy classes of maps from an $n$-sphere $S^n$ into $X$.

For $n=1$, the fundamental group can be non-abelian. For instance, the fundamental group of a figure-eight space is the free group on two generators, a classic non-commutative group. A loop that goes around the first circle and then the second is not homotopic to one that goes around the second and then the first.

But for $n \ge 2$, something magical happens: all higher homotopy groups $\pi_n(X, x_0)$ are ​​abelian​​ (commutative). Why? The reason is a beautiful geometric insight known as the ​​Eckmann-Hilton argument​​. Imagine a map from a 2-dimensional square into $X$ (which is equivalent to a map from a 2-sphere). We can define a product of two such maps, $[f] \cdot_1 [g]$, by splitting the square vertically, putting $f$ on the left half and $g$ on the right. This is like concatenation for $\pi_1$, but along the first coordinate. But since we are in two dimensions, we could just as well have split the square horizontally, putting $f$ on the bottom half and $g$ on the top. This defines a second product, $[f] \cdot_2 [g]$.

The extra dimension gives us room to maneuver. It turns out that these two seemingly different ways of composing our maps are not only homotopic to each other ($[f] \cdot_1 [g] = [f] \cdot_2 [g]$), but this very fact forces the operation to be commutative! The argument involves a simple picture of a square divided into four quadrants. By filling the quadrants with maps $f$, $g$, and identity maps in a clever way, one can show that $[f] \cdot_1 [g] = [g] \cdot_1 [f]$. The geometric freedom of having more than one dimension leaves no choice but for the group to be abelian. This is a stunning example of how a simple geometric property dictates a profound algebraic law.

Homotopy theory provides not just a way to classify existing spaces, but also a way to construct "ideal" spaces that serve as building blocks. For any abelian group $G$ and any integer $n \ge 1$, one can construct a space, called an ​​Eilenberg-MacLane space​​ $K(G,n)$, with the unique property that its $n$-th homotopy group is exactly $G$, and all its other homotopy groups are trivial.

These spaces are like pure tones in the symphony of topology. They embody a single homotopy group and nothing else. And their power is immense. A cornerstone theorem of algebraic topology states that there is a natural one-to-one correspondence between the homotopy classes of maps from a space $X$ into $K(G,n)$ and a completely different algebraic object associated with $X$: its $n$-th ​​cohomology group​​ $H^n(X;G)$.

$[X, K(G,n)] \cong H^n(X;G)$

This is not just a bijection of sets; it is an isomorphism of groups. The group structure on the set of homotopy classes is induced by a kind of "addition" operation on the Eilenberg-MacLane space itself, making it what is known as an H-space.

This theorem represents a profound unification. On one side, we have $[X, K(G,n)]$, a set rooted in the geometric, visual idea of continuous deformation. On the other side, we have $H^n(X;G)$, a group derived from a purely algebraic, combinatorial construction of cochains and boundaries. The fact that they are one and the same reveals a deep and beautiful unity at the foundations of modern geometry. The journey that started with stretching a rubber sheet has led us to a bridge connecting the worlds of geometry and algebra.

We have spent some time learning the rules of a wonderful game, the game of homotopy. We’ve learned to classify shapes by studying the loops one can draw upon them, and we’ve seen that these classes of loops form a group—an algebraic structure that acts as a fingerprint for the space. But what is this all for? Is it merely a sophisticated form of stamp-collecting for mathematicians? Not at all! It turns out this abstract machinery is one of the most powerful languages we have for describing the world. It provides a bridge, a deep and beautiful connection, between the pure, visual world of geometry, the rigid, logical world of algebra, and the concrete, tangible world of physics. Now, let’s go on an adventure and see what this game can do.

Let’s start with something familiar, the surface of a donut, or what a mathematician calls a torus, $T^2$. We learned that its fundamental group is $\pi_1(T^2) \cong \mathbb{Z} \times \mathbb{Z}$. What does this really mean? It’s a perfect "catalogue" of all possible round-trips one can make on the donut's surface. Imagine you are a tiny ant crawling on it. Any journey you take that starts and ends at the same spot can be described by just two numbers: how many times you went around the "long way" (longitudinally) and how many times you went through the hole (meridionally). An element like $(3, 0)$ in this group doesn't just represent one specific path; it represents the entire family of paths that wrap three times around the long way and have no net wrapping through the hole. All such paths can be smoothly deformed into one another. The algebra (m, n) is a direct, intuitive description of a geometric action.

Now, let's contrast this with a more bizarre surface: the Klein bottle. A Klein bottle is famous for being "non-orientable"—it has no distinct inside or outside. This single topological twist has a dramatic consequence for the loops one can draw on it. While you can still identify longitudinal and meridional directions, the twist imposes a strange rule on them, captured by the algebraic relation $a b a^{-1} b = 1$ in its fundamental group. If you try to map a loop from a Klein bottle onto a simple circle, this relation suddenly matters. A map to the circle corresponds to a homomorphism into the integers $\mathbb{Z}$, which is an abelian group. The Klein bottle's relation $a b a^{-1} b = 1$ becomes $a+b-a+b=2b=0$ in the abelian world of the integers. This equation forces the image of the loop $b$ to be zero! The physical twist in the space translates directly into a strict algebraic constraint on the possible maps out of it. The algebra isn't arbitrary; it faithfully reports on the geometry.

This idea of a space's structure constraining things extends further. Consider a simple loop, like a rubber band stretched around the handle of a coffee mug. As long as the band stays on the surface of the mug ($A$), you can't shrink it to a point. It's "stuck" on the handle. But the mug sits in a 3D room ($X$). If you are allowed to lift the rubber band off the mug into the room, you can, of course, shrink it to a point with no trouble. The loop is trivial in the room but non-trivial on the mug's surface. This is precisely the concept captured by the kernel of an induced homomorphism. The set of loops in a subspace $A$ that become shrinkable in the larger space $X$ forms this kernel. It tells us exactly which "holes" in $A$ are "filled in" by the ambient space $X$.

So far, we have used homotopy to classify loops within a space. Can we use it to classify maps between different spaces? The answer is a resounding yes, and it leads to some astonishing conclusions.

Consider a simple question: in how many fundamentally different ways can you map a 2-sphere, $S^2$, onto a circle, $S^1$? Your intuition might suggest many ways. You could crush the sphere along its north-south axis onto the equator. Or you could map each point to the circle based on its longitude. It seems easy. Yet, homotopy theory delivers a stunning verdict: there is only one way, and it is the trivial one. Any continuous map from a sphere to a circle can be continuously deformed into a constant map, where the entire sphere is sent to a single point. You cannot "wrap" a sphere around a circle in any meaningful way.

Why is this? The deep reason lies in a magical correspondence. For certain well-behaved spaces called Eilenberg-MacLane spaces, questions about homotopy classes can be translated into questions in a completely different area of mathematics: cohomology. The circle $S^1$ happens to be such a space, a $K(\mathbb{Z}, 1)$. The problem of classifying maps from $S^2$ to $S^1$ becomes the problem of calculating the "first cohomology group of $S^2$ with integer coefficients," which turns out to be zero. A zero result means there's only one possibility, the trivial one. This is a beautiful example of the unity of mathematics, where a difficult question in one field becomes simple when viewed through the lens of another.

This power extends to higher dimensions. We've talked about $\pi_1$, the fundamental group, which studies loops ($S^1$). What about $\pi_2$, which studies maps from a 2-sphere ($S^2$)? What are these "higher homotopy groups" for? They answer questions about extending maps and filling in holes. Imagine you have a map defined on the boundary of a disk, say, the equator of a ball. Can you extend this map to the entire interior of the ball? Obstruction theory, built on higher homotopy groups, tells you when you can and cannot.

Let's say we start with the simplest possible map on the boundary: we take the entire boundary circle $S^1$ and map it to a single point on a sphere $S^2$. Now we ask: in how many ways can we fill this in with a map from the disk $D^2$? This is like asking for all the fundamentally different ways to wrap a 2-sphere, given that its equator is held fixed at a point. The set of all possible ways to do this corresponds precisely to the elements of the second homotopy group, $\pi_2(S^2)$. It is a known, and remarkable, fact that $\pi_2(S^2)$ is isomorphic to the integers $\mathbb{Z}$. This means there are infinitely many distinct ways to "wrap" a sphere onto itself, each classified by an integer "wrapping number".

Perhaps the most profound application of homotopy theory is in modern physics. Many fundamental theories, from electromagnetism to the Standard Model of particle physics, are formulated in the language of differential geometry, specifically using "principal bundles" and "gauge fields." At its heart, a principal $G$-bundle over a space $M$ is a way of attaching a "fiber" space (which has the structure of a group $G$) to every point of $M$ in a consistent way. The group $G$ represents some symmetry of the system—like rotations in space, or more abstract internal symmetries of particles.

Isomorphism classes of these bundles—that is, the different global configurations the fields can take—are classified by homotopy classes of maps! Specifically, the set of $G$-bundles over a space $M$ is in one-to-one correspondence with $[M, BG]$, the set of homotopy classes of maps from $M$ into a special "classifying space" $BG$.

Let's make this concrete. Consider principal bundles with the rotation group $G = SO(3)$ over the 2-sphere $M = S^2$. This is like asking: how many ways can we assign a 3D orientation (a reference frame) to every point on a sphere? The set of such bundles is classified by $[S^2, BSO(3)]$, which is isomorphic to $\pi_1(SO(3))$. A wonderful fact is that $\pi_1(SO(3)) \cong \mathbb{Z}_2$, the group with two elements. This means there are exactly two types of $SO(3)$ bundles over a sphere: a trivial one and a twisted one. The trivial one is easy to imagine. The twisted one is related to the famous "hairy ball theorem": you cannot comb the hair on a sphere without creating a cowlick. This twisted configuration of vector fields corresponds to a non-trivial bundle, and it has real physical consequences.

The fact that $\pi_1(SO(3)) \cong \mathbb{Z}_2$ is one of the deepest secrets of our universe. A loop in $SO(3)$ represents a continuous rotation that ends up back where it started. A $360^{\circ}$ rotation is such a loop. But this loop is not shrinkable to a point! You can visualize this with the famous "belt trick" or Dirac's string trick. You must perform another full $360^{\circ}$ turn (for a total of $720^{\circ}$) to get a loop that can be undone. This topological property is the mathematical root of electron spin. Particles like electrons, called fermions, must be rotated by $720^{\circ}$ to return to their original quantum state. Their existence is a physical manifestation of a non-trivial homotopy group. The classification of such configurations using homotopy theory is not just an academic exercise; it's a description of reality.

Finally, the reach of homotopy extends even to the heart of calculus on manifolds. Consider a loop $\gamma$ on a surface. When does the integral of every closed 1-form $\omega$ over this loop vanish? The answer, provided by de Rham's theorem, is breathtaking. This condition holds if and only if the loop's homotopy class belongs to the commutator subgroup of the fundamental group. This means that the loops that are "invisible" to this type of integration are precisely those that can be expressed as a product of commutators, like $a b a^{-1} b^{-1}$. A deep analytical property is perfectly mirrored by a purely algebraic one.

From a simple trip around a donut to the strange nature of quantum spin and the very structure of the fields that govern the universe, homotopy classes provide a language of startling power and clarity. They reveal the hidden harmonies between the worlds of shape, algebra, and physical law, standing as a testament to the profound and often surprising unity of science.