Homotopy

In topology, a homotopy between two continuous maps $f, g: X \to Y$ between topological spaces $X$ and $Y$ is a continuous map $H: X \times [0, 1] \to Y$ such that $H(x, 0) = f(x)$ and $H(x, 1) = g(x)$ for all $x \in X$, representing a continuous deformation of $f$ into $g$.^[1] This relation is an equivalence relation on the set of continuous maps, partitioning them into homotopy classes that capture essential topological similarities invariant under such deformations.^[1] The concept was introduced by Henri Poincaré in his 1895 paper Analysis Situs, where it formed the basis for early algebraic topology by enabling the classification of spaces via deformation-invariant properties like the fundamental group.^[2] Homotopy extends naturally to spaces themselves through homotopy equivalence: two spaces $X$ and $Y$ are homotopy equivalent if there exist continuous maps $f: X \to Y$ and $g: Y \to X$ that are homotopy inverses, meaning $f \circ g \simeq \mathrm{id}_Y$ and $g \circ f \simeq \mathrm{id}_X$.^[3] This equivalence preserves key topological invariants, such as homology groups, which remain unchanged under homotopy and thus detect obstructions to such deformations.^[3] In particular, contractible spaces—those homotopy equivalent to a point, like Euclidean space $\mathbb{R}^n$—exhibit trivial homology in positive degrees, highlighting homotopy's role in distinguishing deformable versus rigid structures.^[3] Beyond basic definitions, homotopy theory encompasses higher-dimensional generalizations through homotopy groups $\pi_n(X, x_0)$, which classify maps from $n$-spheres into a pointed space $X$ up to homotopy and form groups for $n \geq 1$.^[1] These groups, starting with the fundamental group $\pi_1$ for loops, provide a hierarchy of invariants that grow increasingly complex, with computations often relying on fibrations and spectral sequences in modern applications.^[4] Homotopy has profoundly influenced fields like algebraic geometry, via étale homotopy,^[5] and physics, through modeling configuration spaces in quantum field theory,^[6] underscoring its enduring impact on understanding spatial continuity and deformation.

Definition and Fundamentals

Formal Definition

In algebraic topology, a homotopy between two continuous maps $f, g: X \to Y$, where $X$ and $Y$ are topological spaces, is defined as a continuous function $H: X \times I \to Y$, with $I = [0, 1]$ denoting the unit interval, such that $H(x, 0) = f(x)$ and $H(x, 1) = g(x)$ for all $x \in X$.^[7] This $H$ provides a continuous family of maps $f_t: X \to Y$ for each $t \in I$, given by $f_t(x) = H(x, t)$, interpolating between $f = f_0$ and $g = f_1$.^[7] The relation of being homotopic, denoted $f \simeq g$, is an equivalence relation on the set of all continuous maps from $X$ to $Y$.^[7] Specifically, reflexivity holds via the constant homotopy $H(x, t) = f(x)$; symmetry via the reparametrization $H'(x, t) = H(x, 1 - t)$; and transitivity by concatenating homotopies along the interval.^[7] This equivalence partitions the set of continuous maps into homotopy classes, often denoted $[f]$ for the class containing $f$, or collectively as $[X, Y]$.^[7] These homotopy classes correspond to the path-connected components in the function space of continuous maps from $X$ to $Y$, equipped with the compact-open topology, where paths in this space are precisely the homotopies.^[7] Thus, homotopy serves as a fundamental equivalence relation that captures continuous deformations between maps, motivating the study of topological invariants preserved under such deformations.^[7]

Basic Properties

The homotopy relation $\simeq$ between continuous maps $f, g: X \to Y$ between topological spaces $X$ and $Y$ is an equivalence relation on the set of such maps.^[7] Reflexivity holds because each map $f$ is homotopic to itself via the constant homotopy $H(x, t) = f(x)$ for all $t \in [0, 1]$.^[7] Symmetry follows from the fact that if $f \simeq g$ via a homotopy $H$, then $g \simeq f$ via the reversed homotopy $H'(x, t) = H(x, 1 - t)$.^[7] Transitivity is established by concatenating homotopies: if $f \simeq g$ via $H_1$ and $g \simeq h$ via $H_2$, then $f \simeq h$ via the homotopy that applies $H_1$ on $[0, 1/2]$ and $H_2$ on $[1/2, 1]$, suitably reparametrized to ensure continuity.^[7] A key feature of homotopy is its independence from the choice of parametrization of the interval $[0, 1]$. Specifically, if $H: X \times [0, 1] \to Y$ is a homotopy from $f$ to $g$, and $\phi: [0, 1] \to [0, 1]$ is a continuous reparametrization with $\phi(0) = 0$ and $\phi(1) = 1$, then the composed map $H \circ (\mathrm{id}_X \times \phi)$ defines an equivalent homotopy from $f$ to $g$.^[7] This reparametrization invariance ensures that homotopy classes are well-defined without dependence on the specific timing of the deformation.^[7] Homotopies compose naturally in a manner that respects the structure of continuous maps. If $f \simeq g: X \to Y$ and $g \simeq h: Y \to Z$, then $f \simeq h: X \to Z$ via the concatenated homotopy described above, confirming the transitivity property algebraically.^[7] Moreover, the relation is functorial with respect to composition: if $f_0 \simeq f_1: X \to Y$ and $g_0 \simeq g_1: Y \to Z$, then $g_0 \circ f_0 \simeq g_1 \circ f_1: X \to Z$, via the homotopy $H(x, t) = g_t(f_t(x))$, where $f_t$ and $g_t$ are the interpolating maps from the respective homotopies.^[7] This property underscores the compatibility of homotopy with the category of topological spaces and continuous maps.^[7]

Examples and Illustrations

Continuous Deformations

Homotopy provides an intuitive notion of continuous deformation between maps, allowing one to "stretch" or "shrink" paths or embeddings within a space without tearing or passing through singularities. This concept is fundamental to understanding how shapes can be transformed while preserving topological features, such as connectivity or the presence of holes.^[7] A classic example of such a deformation occurs in the closed unit disk $D^2 = \{(x,y) \in \mathbb{R}^2 \mid x^2 + y^2 \leq 1\}$, which is contractible. Here, the identity map $\mathrm{id}: D^2 \to D^2$ is homotopic to a constant map sending every point to the center $(0,0)$. The homotopy can be explicitly constructed by radially shrinking the disk toward the center over time, with $H(x,t) = (1-t)x$ for $t \in [0,1]$, demonstrating how the entire disk continuously collapses to a point while remaining within itself.^[7] This contractibility highlights the disk's trivial homotopy type, equivalent to that of a single point.^[7] In simply connected spaces, such as the Euclidean plane $\mathbb{R}^2$, any closed loop—say, a circle centered at the origin—can be continuously deformed to a point. The deformation proceeds by filling the interior of the loop with a disk and then contracting that disk radially to the base point, ensuring the path remains embedded in the space throughout the process.^[7] This null-homotopy illustrates why simply connected spaces have no "non-trivial holes" that prevent such contractions.^[7] On the 2-sphere $S^2$, two paths connecting the same endpoints are homotopic if the closed loop they form bounds a disk on the sphere. For instance, consider the north pole and south pole as endpoints; a great circle arc from north to south can be deformed to a meridian by "sweeping" across the spherical disk it encloses, leveraging the sphere's simply connectedness to fill and shrink the bounding region continuously.^[7] These deformations are often visualized using the "rubber-sheet" analogy, where maps are imagined as drawings on an elastic sheet that can be stretched, twisted, or shrunk continuously without ripping or gluing, preserving the relative positions of points up to homotopy.^[8]

Null-Homotopy

A continuous map $f: X \to Y$ between topological spaces is null-homotopic if it is homotopic to a constant map, meaning there exists a continuous homotopy $H: X \times [0,1] \to Y$ such that $H(x,0) = f(x)$ for all $x \in X$ and $H(x,1)$ is constant for all $x \in X$.^[9][7] This property captures maps that can be continuously deformed to a point in the codomain, reflecting a form of topological triviality. One key characterization of null-homotopy is that $f$ extends continuously to the cone $CX = (X \times [0,1]) / (X \times \{1\})$, where the extension sends the apex (the collapsed end) to a fixed point in $Y$; since the cone $CX$ is contractible, this extension implies the original map deforms to a constant.^[7] These characterizations highlight null-homotopy as a foundational tool for identifying deformable structures in algebraic topology. Prominent examples include all continuous maps from Euclidean space $\mathbb{R}^n$ (for any $n \geq 0$) to an arbitrary space $Y$, as $\mathbb{R}^n$ is contractible and thus every such map deforms to a constant via a straight-line homotopy to the origin.^[7] Similarly, projections or any maps originating from contractible spaces, such as disks $D^n$ or simplices, are null-homotopic, as the domain's contractibility forces the homotopy class to be trivial.^[7] In contrast, the identity map on the circle $S^1$ is not null-homotopic, as it cannot be continuously deformed to a point without "tearing," reflecting the non-trivial generator in its fundamental group $\pi_1(S^1) \cong \mathbb{Z}$.^[7] In the context of homotopy classes, a map $f$ is null-homotopic precisely when its class $[f]$ equals the basepoint class $[*]$ in the pointed homotopy set $[X, Y]_*$, marking it as the trivial element that detects the absence of non-trivial topological obstructions.^[7] This role underscores null-homotopy's importance in classifying maps up to deformation, distinguishing contractible features from more complex homotopy structures.

Homotopy Equivalence

Definition and Criteria

In algebraic topology, two topological spaces $X$ and $Y$ are homotopy equivalent, denoted $X \simeq Y$, if there exist continuous maps $f: X \to Y$ and $g: Y \to X$ such that the compositions $f \circ g$ is homotopic to the identity map $\mathrm{id}_Y$ and $g \circ f$ is homotopic to $\mathrm{id}_X$.^[10] This relation is an equivalence relation on the class of topological spaces, partitioning them into equivalence classes known as homotopy types, which capture the essential "shape" of spaces up to continuous deformation.^[10] A key criterion for homotopy equivalence is the existence of homotopy inverses, as embodied in the maps $f$ and $g$ above; such maps are called homotopy inverses to each other.^[10] Another important criterion involves deformation retracts: a subspace $A \subset X$ is a deformation retract of $X$ if the inclusion map $i: A \to X$ admits a retraction $r: X \to A$ (satisfying $r \circ i = \mathrm{id}_A$) such that $i \circ r \simeq \mathrm{id}_X$ via a homotopy $H: X \times I \to X$ with $H(x, 0) = x$ for all $x \in X$, $H(x, 1) \in A$ for all $x \in X$, and $H(a, t) = a$ for all $a \in A$ and $t \in I$.^[10] In this case, $X$ and $A$ are homotopy equivalent, with $i$ and $r$ serving as homotopy inverses.^[10] Weak homotopy equivalence provides a related but potentially stricter criterion in certain contexts: a continuous map $f: X \to Y$ is a weak homotopy equivalence if it induces isomorphisms on the sets of path components $\pi_0$ and on all homotopy groups $\pi_n$ for $n \geq 1$ at every basepoint.^[11] For CW-complexes, weak homotopy equivalences coincide with homotopy equivalences by Whitehead's theorem, but in general they may differ.^[12] Representative examples of homotopy equivalent pairs include a point space and any contractible space, such as the closed unit disk in $\mathbb{R}^2$, which deformation retracts to its center point via radial contraction.^[10] Similarly, the cone on any space is contractible and thus homotopy equivalent to a point.^[10]

Relation to Homeomorphism

A homeomorphism between topological spaces $X$ and $Y$ is a bijective continuous map $f: X \to Y$ whose inverse $f^{-1}: Y \to X$ is also continuous, establishing an isomorphism that preserves all topological properties, including local structure, dimension, and compactness.^[7] This relation is strictly stronger than homotopy equivalence: every homeomorphism induces a homotopy equivalence, since the identity maps serve as the required homotopies, but the reverse implication fails in general.^[7] Homotopy equivalence captures large-scale or "global" topological features, such as connectivity and the existence of holes, while disregarding finer details like exact shape or dimensionality that homeomorphisms preserve. For instance, the Euclidean spaces $\mathbb{R}^n$ and $\mathbb{R}^m$ for $n \neq m$ are both contractible—meaning each is homotopy equivalent to a point via a straight-line contraction to the origin—but they are not homeomorphic, as their topological dimensions differ.^[7] Similarly, the closed $n$-ball $B^n$ in $\mathbb{R}^n$ is homotopy equivalent to a point through radial contraction, yet it cannot be homeomorphic to a point, which has dimension 0 while $B^n$ has dimension $n > 0$.^[7] In infinite-dimensional settings, this distinction becomes even more pronounced. The Hilbert cube $Q = \prod_{i=1}^\infty [0,1]$, equipped with the product topology, is a compact, contractible absolute neighborhood retract, hence homotopy equivalent to a point via coordinate-wise contraction to 0. However, $Q$ is not homeomorphic to any finite-dimensional Euclidean space or a point, as it is infinite-dimensional and contains uncountably many points, violating the bijectivity and local Euclidean properties required for such homeomorphisms. These examples illustrate how homotopy equivalence ignores local and dimensional intricacies, focusing instead on deformable connectivity that homeomorphisms must match exactly.^[7]

Invariants and Classification

Homotopy Invariants

In algebraic topology, the relation of homotopy equivalence partitions continuous maps between topological spaces into equivalence classes, denoted $[X, Y]$, where $X$ and $Y$ are pointed topological spaces. A continuous map $\phi: Y \to Z$ induces a well-defined function on these homotopy classes, $[\phi]_*: [X, Y] \to [X, Z]$, given by post-composition $[f] \mapsto [\phi \circ f]$; this is well-defined because if two maps $f \sim g$ are homotopic, then $\phi \circ f \sim \phi \circ g$.^[7] Among the basic homotopy invariants, the zeroth homotopy set $\pi_0(X)$ consists of the path-connected components of $X$, classifying the path-connectedness of the space: $X$ is path-connected if and only if $\pi_0(X)$ has a single element. The Euler characteristic $\chi(X) = \sum_{k \geq 0} (-1)^k \rank H_k(X; \mathbb{Z})$, defined via the alternating sum of ranks of singular homology groups, is a coarser homotopy invariant that detects certain structural features but is not exclusively homotopy-theoretic, as it also remains unchanged under homeomorphisms.^[7] For continuous maps $f: S^n \to S^n$ between $n$-spheres, the topological degree $\deg(f) \in \mathbb{Z}$ provides a precise homotopy invariant; homotopic maps share the same degree, and for oriented spheres, this integer completely classifies the homotopy classes.^[7] Whitehead's theorem asserts that if $X$ and $Y$ are CW-complexes and $f: X \to Y$ induces isomorphisms $\pi_n(f): \pi_n(X, x_0) \to \pi_n(Y, f(x_0))$ on all homotopy groups for $n \geq 0$, then $f$ is a homotopy equivalence.^[13]

Homotopy Groups

The nth homotopy group of a pointed topological space (X, x_0), denoted $\pi_n(X, x_0)$, is defined as the set of homotopy classes of basepoint-preserving continuous maps $f: (S^n, *) \to (X, x_0)$, where $S^n$ is the $n$-sphere with basepoint $*$ at the north pole, and two such maps are equivalent if they are connected by a basepoint-preserving homotopy. This construction, introduced by Hurewicz, endows $\pi_n(X, x_0)$ with an abelian group structure for $n \geq 2$ via the pinch map on the equator of $S^n$, which induces a well-defined addition of classes; the identity element corresponds to the class of constant maps, which are precisely the null-homotopic ones. For $n=1$, $\pi_1(X, x_0)$ recovers the fundamental group, which need not be abelian in general. These groups serve as powerful invariants for classifying spaces up to homotopy, as a homotopy equivalence induces isomorphisms on all homotopy groups. A canonical example arises with spheres: $\pi_n(S^k) \cong \mathbb{Z}$ for $n=k \geq 1$, generated by the degree map, while $\pi_n(S^k)$ is trivial for $n < k$; however, exceptions occur in higher dimensions, such as $\pi_3(S^2) \cong \mathbb{Z}$, generated by the attaching map of the Hopf fibration $S^3 \to S^2$. The Hopf fibration provides the first non-trivial example of a higher homotopy group, demonstrating that spheres do not classify simply by their dimension alone. The Freudenthal suspension theorem states that if X is an (n-1)-connected CW-complex for some n ≥ 2, then the suspension homomorphism $\pi_k(X) \to \pi_{k+1}(\Sigma X)$ is an isomorphism for k < 2n - 1 and a surjection for k = 2n - 1. This result, proved by Freudenthal, establishes a stability pattern in homotopy groups: for a simply connected space, suspensions yield isomorphisms in a range of low dimensions, allowing computations of unstable groups to inform stable ones via iterated suspensions.^[7] Despite their utility, computing homotopy groups remains highly challenging, with no general algorithm available; for instance, the fundamental groups of compact orientable surfaces of genus $g \geq 1$ are presented as $\langle a_1, b_1, \dots, a_g, b_g \mid \prod_{i=1}^g [a_i, b_i] = 1 \rangle$, which is free abelian of rank $2g$ for the torus ($g=1$), but higher groups like those of spheres require spectral sequences and are known explicitly only up to dimension around 100 through computational efforts building on seminal work by Serre and Toda.

Variants and Extensions

Relative Homotopy

In algebraic topology, relative homotopy refers to a homotopy between two continuous maps $f, g: X \to Y$ that remains fixed on a subspace $A \subset X$. Specifically, a homotopy $H: X \times I \to Y$, where $I = [0,1]$ is the unit interval, is relative to $A$ if $H(x, t) = f(x)$ for all $x \in A$ and $t \in I$, ensuring that the deformation does not move points in $A$.^[7] This concept extends the notion of homotopy to pairs of spaces $(X, A)$, where maps preserve the subspace structure, and is fundamental for analyzing how deformations behave when constrained by a fixed subset.^[7] Relative homotopy groups generalize absolute homotopy groups to pairs $(X, A, x_0)$ with basepoint $x_0 \in A$. The $n$th relative homotopy group $\pi_n(X, A, x_0)$ consists of homotopy classes of continuous maps $(D^n, S^{n-1}, s_0) \to (X, A, x_0)$, where $D^n$ is the $n$-dimensional disk, $S^{n-1}$ its boundary sphere, and $s_0 \in S^{n-1}$. These classes form a group under concatenation of maps, abelian for $n \geq 2$, capturing "spheres" attached to $A$ within $X$. When $A$ is a single point, $\pi_n(X, A, x_0)$ reduces to the standard homotopy group $\pi_n(X, x_0)$.^[7] For a pair $(X, A)$ with $A \subset X$, there exists a long exact sequence in homotopy groups:

$$\cdots \to \pi_n(A, x_0) \to \pi_n(X, x_0) \to \pi_n(X, A, x_0) \to \pi_{n-1}(A, x_0) \to \cdots \to \pi_0(X, A, x_0) \to 0,$$

where the maps are induced by inclusions and boundary operators, with exactness meaning the image of each map equals the kernel of the next. This sequence arises from viewing the pair as a fibration and provides a tool to relate the topology of $X$ and $A$ through their relative structure.^[7] Applications of relative homotopy are prominent in the study of cell complexes, such as CW-complexes, where cellular approximation simplifies computations. The cellular approximation theorem states that any map $(X, A) \to (Y, B)$ between CW-pairs is homotopic, relative to $A$, to a cellular map, allowing homotopy groups to be calculated using skeletal filtrations and chain complexes of cells.^[7] Additionally, the excision theorem ensures that if $U \subset \operatorname{int} A$ with $\overline{U} \subset A$, then the inclusion $(X - U, A - U) \to (X, A)$ induces an isomorphism $\pi_n(X - U, A - U) \cong \pi_n(X, A)$ for all $n$, facilitating local-to-global analysis by removing small open sets without altering homotopy types.^[7]

Isotopy

In topology, an isotopy is a homotopy between two maps that remains invertible at every stage of the deformation. Specifically, for two homeomorphisms $f_0, f_1: X \to Y$ between topological spaces, an isotopy is a continuous family of homeomorphisms $H_t: X \to Y$ for $t \in [0,1]$ such that $H_0 = f_0$ and $H_1 = f_1$.^[14] Similarly, for embeddings of a manifold $M$ into another manifold $N$, an isotopy is a homotopy between two embeddings $f, g: M \to N$ such that each intermediate map $H_t: M \to N$ is also an embedding.^[15] This levelwise invertibility distinguishes isotopy from general homotopy, ensuring the deformation preserves the topological structure without self-intersections or collapses.^[16] Ambient isotopy refines this concept by considering deformations within a fixed ambient space. Given two embeddings $f, g: M \to N$, they are ambient isotopic if there exists an isotopy of self-homeomorphisms of $N$ (starting from the identity) that carries the image $f(M)$ to $g(M)$ while fixing the complement $N \setminus f(M)$ setwise at each stage.^[17] This extends the deformation to the entire ambient manifold, providing a global perspective on equivalence under continuous distortions that respect the surrounding space.^[15] Ambient isotopy is particularly useful for studying embedded objects, as it models "ambient-preserving" reconfigurations without altering the topology of the exterior.^[18] A classic example arises in knot theory, where two embeddings of the circle $S^1$ into $\mathbb{R}^3$ (knots) are equivalent if they are ambient isotopic; in particular, any unknotted circle is ambient isotopic to the standard round circle via a deformation that avoids intersections.^[17] Reidemeister moves provide a combinatorial realization of this equivalence: two knot diagrams represent ambient isotopic knots if and only if one can be transformed into the other through a finite sequence of these local moves (type I: twist/untwist; type II: create/annihilate crossing pair; type III: slide over crossing), as established by Reidemeister's theorem.^[19] For higher-dimensional spheres, the Schönflies theorem implies that every locally flat embedding of the 2-sphere $S^2$ into $\mathbb{R}^3$ is ambient isotopic to the standard equatorial sphere, confirming the uniqueness of such embeddings up to deformation.^[20] In the smooth category, the topological notion of isotopy generalizes to diffeotopy, where each stage of the homotopy consists of diffeomorphisms rather than mere homeomorphisms.^[21] This smooth variant is essential for studying manifolds with differentiable structure, ensuring the deformation preserves not only topology but also local differentiability, though the two concepts coincide in low dimensions due to the density of smooth approximations.

Advanced Structures

Lifting and Extension

In algebraic topology, the homotopy lifting property (HLP) is a key characteristic of fibrations, enabling the extension of lifts from maps to homotopies. Specifically, for a map $ p: E \to B $ defined as a fibration, given any space $ X $, a homotopy $ G: X \times I \to B $, and a lift $ f: X \to E $ of the initial map $ G_0: X \to B $ (satisfying $ p \circ f = G_0 $), there exists a homotopy $ \tilde{G}: X \times I \to E $ such that $ p \circ \tilde{G} = G $ and $ \tilde{G}_0 = f $.^[7] This property ensures that fibrations behave well under homotopy, preserving structural information from the base space $ B $ to the total space $ E $.^[7] A classic example of a fibration exhibiting the HLP is the path-loop fibration $ p: P B \to B $, where $ P B $ denotes the space of paths in $ B $ starting at a fixed basepoint, and the fiber over the basepoint is the based loop space $ \Omega B $.^[7] This fibration lifts homotopies of based loops in $ B $ to paths in $ P B $, which is essential for computing homotopy groups.^[7] Dually, the homotopy extension property (HEP) applies to cofibrations, facilitating the extension of homotopies from subspaces. For a cofibration $ i: A \hookrightarrow X $, given a map $ f: X \to Y $ and any homotopy $ H: A \times I \to Y $ such that $ H(a, 0) = f(a) $ for all $ a \in A $, there exists a homotopy $ \tilde{H}: X \times I \to Y $ such that $ \tilde{H}|_{A \times I} = H $ and $ \tilde{H}(x, 0) = f(x) $ for all $ x \in X $.^[7] This property is crucial for constructions involving cell attachments, as it allows homotopies defined on subcomplexes to propagate to the full space without obstruction.^[7] Cell inclusions in CW complexes provide a concrete illustration of the HEP; for instance, the inclusion of the boundary sphere $ S^{n-1} \hookrightarrow D^n $ (or skeleta inclusions $ X^{k-1} \hookrightarrow X^k $) admits extensions of homotopies from the boundary or lower skeleton to the disk or full complex.^[7] These examples underpin the deformation properties of CW pairs, enabling approximations and retractions in homotopy theory.^[7] The HLP and HEP play foundational roles in model category theory, where fibrations are defined via the right lifting property against acyclic cofibrations (incorporating HLP), and cofibrations via the left lifting property against acyclic fibrations (incorporating HEP).^[22] In the category of simplicial sets, these properties manifest in the Kan fibrations (which satisfy HLP) and cofibrations (which satisfy HEP), providing a combinatorial framework for homotopy theory that mirrors topological constructions.^[7]

Homotopy Category

The homotopy category of topological spaces, denoted $\mathrm{Ho}(\mathrm{Top})$, is a category whose objects are topological spaces and whose morphisms are homotopy classes of continuous maps between them, denoted $[f]$ for a continuous map $f: X \to Y$. This construction formalizes homotopy theory by treating homotopies as isomorphisms, allowing the study of spaces up to homotopy equivalence. Specifically, the composition of morphisms $[f] \circ [g]$ is induced by the usual composition of representatives, modulo homotopy. $\mathrm{Ho}(\mathrm{Top})$ arises as the localization of the category $\mathrm{Top}$ of topological spaces and continuous maps at the class of weak homotopy equivalences (or homotopy equivalences for well-behaved spaces like CW-complexes), inverting these maps to make them isomorphisms in the localized category.^[23] This localization process, in the sense of model categories, yields a category where the homotopy type of a space determines its essential properties, enabling the application of categorical tools to homotopy theory.^[24] The resulting structure is equivalent to the homotopy category of simplicial sets under the standard model structure.^[23] A key adjunction underpinning this equivalence is between the singular simplicial set functor $\mathrm{Sing}: \mathrm{Top} \to \mathrm{sSet}$, which assigns to each space $X$ the simplicial set of its singular simplices, and the geometric realization functor $|-|: \mathrm{sSet} \to \mathrm{Top}$, which constructs a topological space from a simplicial set.^[25] The functor $\mathrm{Sing}$ preserves homotopies, mapping continuous homotopies in $\mathrm{Top}$ to simplicial homotopies in $\mathrm{sSet}$, and thus induces a functor on the homotopy categories $\mathrm{Ho}(\mathrm{Top}) \to \mathrm{Ho}(\mathrm{sSet})$.^[25] Conversely, geometric realization preserves weak equivalences and fibrations, ensuring the adjunction descends to an equivalence of homotopy categories. The stable homotopy category extends this framework by considering the homotopy category of spectra, which stabilizes $\mathrm{Ho}(\mathrm{Top}_*)$ (the pointed version) under infinite suspension; it captures the stable homotopy groups of spaces and serves as the foundation for stable homotopy theory.^[26] The suspension spectrum functor $\Sigma^\infty: \mathrm{Ho}(\mathrm{Top}_*) \to \mathrm{Ho}(\mathrm{Sp})$ embeds unstable homotopy into this stable setting, where homotopy groups become independent of dimension after sufficient suspension.^[27]

Applications

In Algebraic Topology

In algebraic topology, homotopy theory plays a central role in classifying topological spaces, particularly through the use of homotopy groups to distinguish non-homeomorphic manifolds with the same homotopy type. A prominent example is the family of lens spaces, which are 3-dimensional manifolds constructed as quotients of the 3-sphere by cyclic group actions. All lens spaces $L(p,q)$ with coprime integers $p$ and $q$ share the same fundamental group $\pi_1(L(p,q)) \cong \mathbb{Z}/p\mathbb{Z}$, making $\pi_1$ insufficient for full classification. The topological classification requires additional invariants, such as the Reidemeister torsion, to determine when two lens spaces are homeomorphic: $L(p,q) \cong L(p,q')$ if and only if $q \equiv q' \pmod{p}$ or $q \equiv -q'^{-1} \pmod{p}$, up to orientation. This demonstrates how homotopy invariants like $\pi_1$ provide initial distinctions but necessitate complementary tools for precise classification.^[28] The Poincaré conjecture exemplifies the power of homotopy in manifold classification, positing that every closed, simply connected 3-manifold is homeomorphic to the 3-sphere $S^3$. Simply connectedness means $\pi_1 = 0$, and the conjecture asserts that this, combined with the homotopy type of $S^3$, implies homeomorphism. Grigori Perelman resolved this in 2002–2003 using Ricci flow with surgery, showing that any such manifold evolves under the flow to a metric of constant sectional curvature, thereby confirming it is diffeomorphic to $S^3$ and establishing the conjecture. This resolution not only verified the topological invariance of homotopy type in dimension 3 but also advanced the broader geometrization conjecture, highlighting homotopy's role in bridging topology and differential geometry. CW-complexes further enable homotopy-based classification by providing combinatorial models for spaces. These structures, built by attaching cells of increasing dimension, approximate arbitrary topological spaces up to weak homotopy equivalence. The cellular approximation theorem states that for CW-complexes $X$ and $Y$, any continuous map $f: X \to Y$ is homotopic to a cellular map, which sends the $n$-skeleton of $X$ into the $n$-skeleton of $Y$. This theorem simplifies the study of homotopy classes $[X,Y]$, as cellular maps reduce computations to algebraic data on cells, facilitating the classification of maps and spaces via chain complexes and attachment maps.^[7] Spectral sequences provide a systematic method to compute homotopy groups from fibration structures. For a Serre fibration $F \to E \to B$, the associated spectral sequence converges to the homology groups $H_*(E)$, with $E_2^{p,q} = H_p(B; H_q(F))$, allowing indirect computation of homotopy groups via the Hurewicz homomorphism, which relates $\pi_n$ to $H_n$ in simply connected spaces. In more general settings, such as towers of principal fibrations from Postnikov truncations, a spectral sequence arises from the exact couple of long exact homotopy sequences, converging to $\pi_*(E)$ and enabling inductive calculations of higher homotopy groups from lower ones and $k$-invariants. These tools are essential for classifying spaces by unraveling their fibration decompositions.^[4] Surgery theory extends homotopy classification to manifolds by addressing when homotopy equivalences can be realized as homeomorphisms or diffeomorphisms. Developed in the 1960s, it involves excising embedded spheres and attaching handles to modify manifolds while preserving homotopy type. For a homotopy equivalence $f: M \to N$ between closed $n$-manifolds ($n \geq 5$), the surgery obstruction groups $L_n(\pi_1(N))$ measure whether $f$ is homotopic to a homeomorphism; vanishing obstructions imply the existence of such a surgery sequence leading to diffeomorphism in the smooth case under stable range conditions. This framework classifies manifolds up to homotopy equivalence, particularly in high dimensions, by relating geometric structures to algebraic $L$-theory.

In Other Fields

In differential geometry, Morse theory provides a powerful connection between the critical points of smooth functions on manifolds and the homotopy type of those manifolds. Specifically, for a Morse function on a compact manifold, the homotopy type changes at critical points through the attachment of handles, where the dimension of each handle corresponds to the index of the critical point, allowing the manifold to be reconstructed up to homotopy equivalence via a handlebody decomposition. This framework, developed by Marston Morse and refined by John Milnor, enables the computation of homotopy invariants from the Morse data, such as the number and indices of critical points, which determine the Betti numbers via the Morse inequalities.^[29] In algebraic geometry, the étale homotopy type offers an analogue of singular homotopy theory adapted to schemes, replacing continuous maps with étale morphisms to capture the "topological" structure over fields of arbitrary characteristic. Introduced by Michael Artin and Barry Mazur, this construction assigns to a scheme a pro-object in the homotopy category of spaces, computed via étale hypercovers, which approximates the classical homotopy type when the scheme is over the complex numbers by profinite completion. For varieties over algebraically closed fields, the étale homotopy groups provide obstructions to lifting properties from characteristic zero to positive characteristic, and they are particularly useful in studying moduli spaces of algebraic curves where the étale fundamental group detects non-trivial coverings. In physics, homotopy theory plays a crucial role in string theory through the moduli spaces of Riemann surfaces, which parametrize the worldsheets of propagating strings and carry homotopy-invariant structures that govern scattering amplitudes. The compactification of these moduli spaces reveals operad structures, leading to homotopy Lie algebras on the string state space, as shown by Ezra Getzler, where the homotopy operations arise from gluing surfaces and encode the algebraic relations in open string field theory. Similarly, in topological quantum field theory (TQFT), homotopy enters via the cobordism hypothesis, which equates fully extended (n+1)-dimensional TQFTs with fully dualizable objects in a symmetric monoidal (∞,n)-category, with bordisms modeled as homotopy equivalences preserving the topological invariants of manifolds. This perspective, formalized by John Baez and James Dolan, unifies the axiomatic framework of Michael Atiyah with higher category theory, enabling computations of partition functions from the homotopy type of the classifying space of the structure group.^[30] In computer science and mathematical foundations, homotopy type theory (HoTT) reinterprets identity types in Martin-Löf dependent type theory as paths in a space, providing a foundation for mathematics where proofs of equality are higher-dimensional homotopies, thus bridging logic with algebraic topology. Developed by the Univalent Foundations Program, HoTT incorporates the univalence axiom, which states that equivalences of types induce equalities in the type of types, allowing synthetic reasoning about homotopy-theoretic constructions like the fundamental groupoid directly in the type theory. This approach facilitates formal verification of mathematical proofs in proof assistants like Coq, with applications to computational topology where types model homotopy types, and identity proofs correspond to paths, enabling the definition of higher inductive types that capture colimits and free constructions up to homotopy.