Path Homotopy

In the world of mathematics, not all paths are created equal. While two routes between the same points might differ in length or shape, we often have an intuitive sense that they are fundamentally "equivalent" if one can be smoothly deformed into the other without being broken or leaving its space. This concept, the continuous deformation of paths, is formalized in topology as path homotopy. It addresses the core question of how the shape of a space constrains movement within it. This article demystifies path homotopy, providing a crucial tool for classifying and understanding the deep structure of topological spaces.

This article will guide you through the core ideas of path homotopy. In the first chapter, "Principles and Mechanisms," we will build the concept from the ground up, starting with an intuitive string analogy and progressing to the formal mathematical definition, the algebraic structure of the fundamental group, and the elegant framework of the fundamental groupoid. Subsequently, in "Applications and Interdisciplinary Connections," we will explore how this abstract theory yields concrete insights, revealing the secrets of diverse spaces from the punctured plane to the torus and forging surprising connections to fields like complex analysis.

Imagine you have a piece of string. You pin its two ends, say at points $x_0$ and $x_1$, on a large board. The path of the string represents, well, a ​​path​​ in the space of the board. Now, you can wiggle the string, deform it, and stretch it (as long as you don't break it), but the ends must remain pinned at $x_0$ and $x_1$. All the different shapes the string can take while satisfying these rules feel, in some deep sense, "equivalent." They all connect the same two points. Homotopy is the mathematical formalization of this very simple and powerful idea: the continuous deformation of paths.

Let's translate our string-on-a-board analogy into the precise language of mathematics. A path in a space $X$ is a continuous journey, a function $f$ from the unit interval $I = [0, 1]$ into $X$. Think of the input, a number $s$ between 0 and 1, as the time elapsed on your journey. So, $f(0)$ is your starting point and $f(1)$ is your ending point.

Now, how do we describe the "wiggling" of the string from one shape, say a path $f$, to another shape, a path $g$? Let's assume both paths connect the same two points, $x_0 = f(0) = g(0)$ and $x_1 = f(1) = g(1)$. The deformation itself needs to be continuous. We can visualize this deformation by introducing a second "time" parameter, let's call it $t$, which also runs from 0 to 1. At time $t=0$, we have our original path $f$. As $t$ increases, the path morphs, until at $t=1$, it has become the path $g$.

This entire process can be described by a single continuous map, $H$, defined on a square, $I \times I$. One axis of the square, let's say the horizontal one parametrized by $s$, represents the position along the path. The other axis, the vertical one parametrized by $t$, represents the deformation time. So, $H(s, t)$ is the point in space $X$ that corresponds to the point $s$ along the path at deformation time $t$.

For this map $H$ to be a ​​path homotopy​​, it must satisfy a few simple, common-sense boundary conditions:

1. $H(s, 0) = f(s)$: At the beginning of the deformation ($t=0$), the shape is identical to our starting path $f$.
2. $H(s, 1) = g(s)$: At the end of the deformation ($t=1$), the shape is identical to our final path $g$.
3. $H(0, t) = x_0$: For all time $t$ during the deformation, the starting point of the path remains fixed at $x_0$. This is one of our pinned ends.
4. $H(1, t) = x_1$: For all time $t$, the ending point of the path also remains fixed at $x_1$. This is our other pinned end.

These four conditions perfectly capture our intuition. The first two state that we are indeed transforming $f$ into $g$, and the last two enforce that the ends of the string stay put.

Why is this last part so important? Consider two paths in the plane from $(0,0)$ to $(1,1)$: the straight line $f_0(s) = (s,s)$ and the parabola $f_1(s) = (s,s^2)$. One might propose a "homotopy" that shrinks the first path down to the origin and then grows it back out into the second path. But such a transformation would unpin the endpoint at $(1,1)$ during the process! For much of the "deformation," the path wouldn't connect $(0,0)$ to $(1,1)$ at all. This would violate our fundamental rule. A path homotopy requires that at every intermediate stage, the deforming shape is still a valid path between the original endpoints.

When two paths $f$ and $g$ can be continuously deformed into one another in this way, we say they are ​​path-homotopic​​, and we write $f \simeq g$. This relationship is wonderfully well-behaved. In fact, it is an ​​equivalence relation​​:

• ​​Reflexive​​: Any path $f$ is homotopic to itself ($f \simeq f$). The deformation is simply to do nothing!
• ​​Symmetric​​: If $f \simeq g$, then $g \simeq f$. If you can deform $f$ into $g$, you can run the "movie" of the deformation in reverse to deform $g$ back into $f$.
• ​​Transitive​​: If $f \simeq g$ and $g \simeq h$, then $f \simeq h$. You can first play the movie of $f$ deforming into $g$, and then play the movie of $g$ deforming into $h$. The combined movie is a valid deformation of $f$ into $h$.

Because path homotopy is an equivalence relation, it carves up the (often infinite) set of all paths from $x_0$ to $x_1$ into disjoint families, called ​​homotopy classes​​. All paths within a single class are considered equivalent from the perspective of topology.

This is where things get truly interesting. In a simple space like a flat plane, any two paths between the same two points are homotopic. You can always just "flatten" one path into the other. But what if the space has a "hole"?

Imagine you are in a park, which is a flat plane $\mathbb{R}^2$, but there is a statue at the origin $(0,0)$ that you are not allowed to cross. Your task is to walk from point $p = (-1, 0)$ to point $q = (1, 0)$. You could take a path $\gamma_C$ that goes "over" the statue, through the upper half-plane. Or, you could take a path $\gamma_D$ that goes "under" it, through the lower half-plane.

Are these two paths homotopic? Try to imagine deforming the upper path into the lower one. Your string, representing the path, would have to pass through the statue at the origin. But that point is not in our space! The path would be "snagged" on the hole. There is no continuous deformation from $\gamma_C$ to $\gamma_D$ that stays within the punctured plane. Therefore, these two paths belong to different homotopy classes. The topology of the space—the presence of the hole—creates a fundamental distinction between them. In fact, we can find even more classes: a path $\gamma_E$ that loops around the statue once before heading to the destination is in yet another class!

We have sorted paths into classes. The next question a mathematician always asks is: can we do algebra with them? Yes, we can!

We can combine two paths by traveling along one and then the other. If $f$ is a path from $x$ to $y$, and $g$ is a path from $y$ to $z$, we can define their ​​concatenation​​, $f \cdot g$, which is a path from $x$ to $z$. The amazing thing is that this operation respects our homotopy classes. If you replace a segment of a long journey with a homotopic segment, the new overall journey is homotopic to the old one. This means we can define a composition on the homotopy classes themselves.

The structure becomes particularly beautiful when we consider ​​loops​​: paths that start and end at the same point, say $x_0$. The set of all homotopy classes of loops based at $x_0$ forms a ​​group​​, one of the most fundamental objects in algebra. This group is called the ​​fundamental group​​ of the space $X$ at the basepoint $x_0$, denoted $\pi_1(X, x_0)$.

• The ​​group operation​​ is path concatenation.
• The ​​identity element​​ is the class of the "do nothing" loop, which stays at the point $x_0$ for the entire time interval.
• The ​​inverse​​ of a loop class $[f]$ is the class of the path traversed in reverse, $[f^{-1}]$. Concatenating $f$ and $f^{-1}$ gives a loop that goes out and immediately comes back, which can be continuously shrunk down to the constant loop at the basepoint.

It's crucial here that we use path homotopy, which keeps the basepoint fixed. There is a looser notion of "free" homotopy, where the endpoints are allowed to move during the deformation. For instance, in a figure-eight space, a loop $a$ around one circle is not path-homotopic to the loop $b \cdot a \cdot b^{-1}$, which travels to the second circle, goes around the first, and then comes back. They are different elements in the fundamental group. However, they are freely homotopic. One can be slid into the other. In the language of group theory, they are conjugate elements. The fundamental group captures the structure of loops pinned to a specific point.

The fundamental group $\pi_1(X, x_0)$ seems to depend on our choice of basepoint $x_0$. What is the relationship between the group at $x_0$ and the group at another point $x_1$? And what if the space isn't even path-connected, meaning you can't get from $x_0$ to $x_1$ at all?

This is where a more general and elegant structure comes into play: the ​​fundamental groupoid​​, denoted $\pi_1(X, A)$, where $A$ is a set of basepoints we care about.

• The ​​objects​​ of the groupoid are the points in our set $A$.
• The ​​morphisms​​ from an object $p$ to an object $q$ are the homotopy classes of paths from $p$ to $q$.

If we choose our set of basepoints $A$ to be just a single point, $A = \{x_0\}$, then there's only one object. The only morphisms are those from $x_0$ to $x_0$—the homotopy classes of loops! And their composition is just path concatenation. In this case, the groupoid is precisely the fundamental group $\pi_1(X, x_0)$. The group is just a groupoid with a single object.

The groupoid elegantly handles spaces that are not path-connected. If points $p$ and $q$ are in different path-components, then there are simply no paths, and thus no morphisms, between them. The groupoid naturally splits into pieces, one for each path-component of the space. This is beautifully reflected in a simple case: two constant paths, $c_x$ and $c_y$, are freely homotopic if and only if there is a path connecting $x$ and $y$. The "objects" of homotopy are connected if and only if the points themselves are.

The groupoid reveals a profound unity between paths and loops. Let's fix two points, $x_0$ and $x_1$, in a path-connected space. Let $P(x_0, x_1)$ be the set of all homotopy classes of paths from $x_0$ to $x_1$. Let $\pi_1(X, x_0)$ be the fundamental group of loops at the starting point.

It turns out that the group of loops $\pi_1(X, x_0)$ ​​acts​​ on the set of paths $P(x_0, x_1)$. The action is simple: take a path class $[\gamma]$, and a loop class $[f]$. The action of $[f]$ on $[\gamma]$ is just the class of the concatenated path, $[f \cdot \gamma]$. You first run around the loop $f$, returning to $x_0$, and then travel along the path $\gamma$ to $x_1$.

This action has two remarkable properties: it is ​​free​​ and ​​transitive​​.

• ​​Transitive​​ means that from any path class $[\gamma_1]$, you can get to any other path class $[\gamma_2]$ by applying some loop. Specifically, the loop you need is $[\gamma_2 \cdot \gamma_1^{-1}]$. This tells us that once you know a single way to get from $x_0$ to $x_1$, you can find all other ways (up to homotopy) simply by tacking on all possible loops at the start.

• ​​Free​​ means that if you take a path class $[\gamma]$ and act on it with a non-trivial loop class $[f]$, you will always get a different path class. No two distinct loops will ever lead you to the same place.

Together, these properties tell us that the set of all paths from $x_0$ to $x_1$ looks, for all intents and purposes, exactly like the fundamental group $\pi_1(X, x_0)$. It's a set that the group acts on perfectly. This is the ultimate expression of the relationship: the diversity of paths between two points is completely and precisely described by the algebraic structure of the loops at the starting point. This beautiful correspondence, where geometry is perfectly mirrored by algebra, is the central mechanism of algebraic topology, and it all begins with the simple, intuitive dance of a wiggling string. And it even gives us predictive power: if we have two ways of composing paths, say $f_1 \cdot g_1$ and $f_2 \cdot g_2$, and we know they are homotopic, then the "detour" taken at the intermediate point, represented by the loop $g_1 \cdot g_2^{-1}$, must be precisely the inverse of the "initial deviation" loop $f_2^{-1} \cdot f_1$. The algebra must balance.

In our previous discussion, we developed a rather abstract notion: path-homotopy. We learned to see paths not as individual, unique trajectories, but as members of families, or "homotopy classes." We might be tempted to ask, "So what?" Is this just a game for mathematicians, a way of organizing abstract squiggles? The answer, you might be surprised to learn, is a resounding no. This way of thinking is not just a mathematical curiosity; it is a profound lens through which we can understand the very fabric of the spaces we live in and analyze. The study of which paths are "the same" is, in essence, the study of the shape of space itself. It tells us what is possible and what is impossible within a given universe.

Let's embark on a journey through a menagerie of strange and wonderful worlds. By trying to navigate them, we will discover how the abstract idea of path-homotopy reveals their deepest secrets and even solves problems in seemingly unrelated fields.

The character of a space, its "topology," has a dramatic effect on the paths within it. Consider two extreme examples. Imagine a world that is a "barren desert" of disconnected points—what topologists call a ​​discrete space​​. In this world, every point is an isolated island. A path is a continuous journey. But since the space between any two distinct points is a void, the only way for a journey to be continuous is for it not to happen at all! The only possible paths are those that start and end at the same point without ever moving. Travel between different islands is impossible. The topology is so restrictive that it chokes off almost all possibility of movement.

Now, imagine the opposite extreme: a "cosmic blob" or an ​​indiscrete space​​. In this universe, the only recognizable regions are "nothing" and "everything." There are no local features, no landmarks, no texture. Any function you can possibly imagine into this space is automatically continuous. As a result, any path you draw from point $p$ to $q$ can be continuously deformed into any other path. It's as if there are no obstacles at all; you can morph one route into another with perfect freedom. In this world, there is only one "way" to get from $p$ to $q$.

These examples, while abstract, frame the conversation. Most spaces we care about lie somewhere in between. A simple, but crucial, case is a world made of several separate pieces, like an archipelago of disjoint islands in an ocean. If we start and end our journey on the same island, and that island is a simple, "un-holey" shape like a disk (a convex set), then just like in the cosmic blob, all paths between two points are equivalent. You can always "straighten out" a meandering path into a direct one. But what if you want to travel from a point $p$ on one island to a point $q$ on another? It's impossible. A continuous path cannot leap across the empty void between the islands. The set of paths from $p$ to $q$ is empty, and so is the set of homotopy classes. This elementary observation—that paths are confined to their connected components—is the first and most fundamental rule of navigation in any space.

What happens when a space is connected, but not as simple as a solid disk? What if it has holes? This is where path-homotopy truly begins to shine.

Consider the plane with a single point removed—a ​​punctured plane​​, $\mathbb{R}^2 \setminus \{(0,0)\}$. Imagine this as a vast prairie with an infinitely tall, uncrossable pillar at the origin. You want to travel from a point $p=(1,0)$ on one side to $q=(-1,0)$ on the other. You could take a path that goes over the top, or a path that goes under the bottom. Are these two paths the same in the homotopic sense? Try as you might, you can never deform the "top" path into the "bottom" path without breaking it or crossing the forbidden pillar. They represent two fundamentally different ways to make the journey.

But we don't have to stop there. What about a path that goes over the top, circles the pillar completely, and then proceeds to $q$? This is a new, third type of journey! What about circling twice? Or circling in the opposite direction? We quickly discover that there are infinitely many distinct classes of paths from $p$ to $q$. We can label each class with an integer, $n \in \mathbb{Z}$, which we can call the "winding number"—it counts how many net turns the path makes around the central pillar. This integer is an algebraic invariant that perfectly classifies the geometric possibilities.

This idea extends to other "loopy" spaces. Take a ​​Möbius strip​​, that famous one-sided surface. It has a "core" that is itself a loop. When we travel between two points on the strip, our path can be classified by how many times it effectively traverses this central loop. Just like the punctured plane, we find a countably infinite number of distinct path homotopy classes, one for each integer in $\mathbb{Z}$.

When spaces become more complex, our inventory of paths becomes richer. The key to understanding them is often to find a simpler, "unfolded" version of the space—its ​​covering space​​.

Think of the surface of a ​​torus​​, the shape of a donut. You can imagine it as the screen of a classic video game like Asteroids, where moving off the right edge makes you reappear on the left, and moving off the top brings you to the bottom. Now, how many ways are there to get from a point $p$ to a point $q$? There's the "direct" route. But you could also go "the long way around," wrapping once around the torus horizontally before arriving at $q$. Or you could wrap twice vertically. Or you could wrap once horizontally and once vertically.

The brilliant way to classify these paths is to "unroll" the torus into its covering space: the infinite Euclidean plane, $\mathbb{R}^2$. A path on the torus "lifts" to a path on this plane. A journey from $p$ to $q$ on the torus becomes a journey from a point $\tilde{p}$ to some corresponding point $\tilde{q}'$ on the plane. The key is that different paths on the torus, distinguished by how they wrap around, will lift to paths on the plane that start at the same point $\tilde{p}$ but end at different points. If the direct path ends at $\tilde{q}$, a path that wraps once horizontally will end at $\tilde{q} + (1,0)$. A path that wraps once vertically will end at $\tilde{q} + (0,1)$. Thus, the homotopy classes of paths from $p$ to $q$ are in a one-to-one correspondence with the integer lattice $\mathbb{Z}^2$. Each class is uniquely identified by a pair of integers $(m, n)$ representing the net number of horizontal and vertical wraps.

The structure can be even more elaborate. For a ​​figure-eight​​ space—two circles joined at their basepoint—the structure of possible loops becomes non-commutative. A loop that goes around circle A then circle B is fundamentally different from one around B then A. The set of homotopy classes of these loops, the fundamental group, is the free group on two generators, $\mathbb{Z} * \mathbb{Z}$. The topology of the space dictates the algebra of its paths.

Sometimes, topology throws us a curveball that shatters our intuition. Consider the ​​topologist's sine curve​​. This strange space consists of the graph of $y = \sin(1/x)$ for $x > 0$, plus the vertical line segment on the $y$-axis where the curve accumulates. The space certainly looks connected; the wiggly part gets arbitrarily close to the vertical line. So, can we travel from a point on the wiggly part to a point on the line segment?

The astonishing answer is no. A continuous path, a finite-time journey, simply cannot be made. To reach the line segment, the path would have to oscillate up and down faster and faster, traversing an infinite amount of vertical distance in a finite amount of time. This is a physical and mathematical impossibility for a continuous path. This bizarre space is connected, but it is not path-connected. It serves as a stark warning that our everyday geometric intuition must be sharpened by the precise definitions of topology.

Perhaps the most beautiful application is one that bridges a gap between two seemingly distant fields: topology and complex analysis. A fundamental problem in complex numbers is defining the logarithm. Because $\exp(w) = \exp(w + 2\pi i k)$ for any integer $k$, the logarithm is inherently multi-valued. How can we select a single, consistent, continuous value for $\ln(z)$?

The answer, it turns out, is a problem of path-homotopy in disguise. Suppose we have a holomorphic (complex differentiable) function $f(z)$ that is never zero. We want to define its logarithm, $g(z)$. We can start by picking a point $z_0$, choosing one of the many possible values for $g(z_0)$ such that $\exp(g(z_0))=f(z_0)$, and then defining $g(z)$ for any other $z$ by "dragging" this value along a path from $z_0$ to $z$. The process of "dragging" is made precise by the ​​path-lifting property​​ of the exponential map, which is a covering map from the complex plane $\mathbb{C}$ to the punctured plane $\mathbb{C} \setminus \{0\}$.

The crucial question is: is this process well-defined? If we choose a different path from $z_0$ to $z$, do we arrive at the same value for $g(z)$? The answer is yes if and only if the two paths are homotopic. If we can continuously deform one path into the other, the resulting value of the logarithm will be the same.

So, when can we guarantee that any two paths between two points are homotopic? Precisely when the domain $D$ of our function $f$ is ​​simply connected​​—a space with no "holes" that can be snagged by a loop. The purely topological property of simple connectivity is the exact condition required to ensure the existence of a single-valued, holomorphic logarithm. This remarkable connection shows that the abstract classification of paths is not just a game; it is a deep principle that brings structure and sense to other branches of science and mathematics, revealing the profound unity of our intellectual landscape.