Homotopy

A homotopy between maps $f,g:Y\to X$ is a continuous map $H:Y\times [0,1]\to X$ with $H(y,0)=f(y)$ and $H(y,1)=g(y)$.

Study homotopy after killing torsion, especially for simply connected spaces of finite type. The rational Whitehead theorem detects rational homotopy equivalences by rational homology, while Sullivan minimal models encode rational homotopy generators and cohomological relations.

A free graded-commutative rational differential graded algebra with decomposable differentials and a quasi-isomorphism to the rational polynomial forms of a space. For simply connected finite-type spaces, $V^i$ is dual to ${\pi }_i(X)\otimes \mathbb{Q}$. A proposed finite model must have its cohomology and representing quasi-isomorphism checked; matching a list of relations alone is insufficient.

The two quadratic relations form a regular sequence in $\mathbb{Q}[a,b]$, so their Koszul complex has only the quotient-ring cohomology. Represent the two degree-two classes by rational polynomial forms and choose primitives for the exact relations to obtain a quasi-isomorphism. There are no further generators.

The based suspension collapses both ends and the basepoint cylinder of $X\times I$ to one basepoint. The loop space and reduced suspension are adjoint in based homotopy theory. For a connected based CW complex, the James reduced product models the loop space of its reduced suspension.

An equivalence class of continuous maps under homotopy, with any endpoint or boundary restrictions specified. Unbased loop classes are free homotopy classes; periods of a closed differential form are constant on such classes.

A free homotopy between maps is an ordinary homotopy with no specified basepoint held fixed. Two loops at the same basepoint are freely homotopic precisely when their elements of the fundamental group are conjugate: a homotopy traces a connecting basepoint path, and the boundary of its parameter square gives conjugation. Conversely that connecting path can be slid around the loop to realize the conjugation.

A map is null-homotopic when it is homotopic to a constant map.

A map $f:S^{n-1}\to X$ extends to $D^n$ exactly when it is null-homotopic. An extension contracts radially through the disc; conversely, a null-homotopy descends through the cone quotient $C{S}^{n-1}\cong D^n$.

A space is contractible when its identity map is homotopic to a constant map. Every map into a contractible space is null-homotopic.

A contraction is a homotopy from the identity map of $X$ to a constant map. Its existence says that $X$ is a contractible space. For a tree, choosing a root and moving each point along its unique finite path to that root gives a contraction, with continuity understood in the weak topology of a CW complex.

Spaces $X$ and $Y$ are homotopy equivalent when there are continuous mapssuch that $g\circ f$ is homotopic to ${\mathrm{id}}_X$ and $f\circ g$ is homotopic to ${\mathrm{id}}_Y$.

Two spaces have the same homotopy type when there is a homotopy equivalence between them. Their homology and homotopy groups are then isomorphic. In Morse theory, attaching a cell records a change in the homotopy type of a sublevel set.

If $g\circ f\simeq {\mathrm{id}}_X$ and $f\circ g\simeq {\mathrm{id}}_Y$, then $g$ is a homotopy inverse of $f$ and conversely.

A subspace $A\subseteq X$ is a retract of $X$ when there is a continuous map $r:X\to A$ whose restriction to $A$ is the identity.

A deformation retraction of $X$ onto $A$ is a homotopy from ${\mathrm{id}}_X$ to a retraction $r:X\to A$ that keeps $A$ inside itself throughout. It makes the inclusion $A\hookrightarrow X$ a homotopy equivalence.

Every retract of a contractible space is contractible. If $i:A\hookrightarrow X$ is the inclusion, $r:X\to A$ is a retraction, and $H$ contracts $X$, thencontracts $A$.

A pair $(X,A)$ has the homotopy extension property when every homotopy on $A$ whose initial map extends to $X$ can itself be extended to a homotopy on $X$.

A cofibration is a map with the homotopy extension property. An inclusion of a CW complex into another as a subcomplex is a cofibration; collapsing that subcomplex gives the quotient used in the K-theory six-term exact sequence.

If $A\subseteq X$ is contractible and $(X,A)$ has the homotopy extension property, the quotient mapis a homotopy equivalence. Extend a contraction of $A$ to $X$; its endpoint is constant on $A$ and therefore factors through $q$ to supply a homotopy inverse.

The $n$th homotopy group consists of based homotopy classes of maps $S^n\to X$. It is a group for $n\ge 1$ and an abelian group for $n\ge 2$.

Based relative maps of a disk into $X$ with its boundary mapped into $A$, modulo homotopies preserving these conditions. Replacing a map by its mapping cylinder gives relative groups for a general map. They fit into the long exact sequence of relative homotopy groups and detect connectivity of the map.

Inclusion, the map sending an absolute disk class to a relative one, and restriction to the disk boundary give this exact sequence. Its low-dimensional end is interpreted with pointed sets and the usual fundamental-group action. For a mapping-cylinder pair it measures the failure of a map to induce homotopy isomorphisms.

A connected Eilenberg–MacLane space $K(G,n)$, $n\ge 1$, has ${\pi }_n\cong G$ and all other positive-degree homotopy groups trivial. The group $G$ must be abelian when $n\ge 2$, but may be nonabelian when $n=1$. For an abelian group, the Dold–Kan correspondence constructs a simplicial model from the chain complex concentrated in degree $n$. A classifying space of a discrete group gives the general $K(G,1)$ case.

For $n\ge 4$, the first four possible degrees have integral groups $\mathbb{Z}/2,0,\mathbb{Z}/2,\mathbb{Z}/2$ in degrees $n,n+1,n+2,n+3$. Mod-two Betti numbers count cyclic summands, while the nonzero first Bockstein actions on $u,{\mathrm{Sq}}^{2}u,{\mathrm{Sq}}^2{\mathrm{Sq}}^{1}u$ show that their orders are two. This prevents mistaking a nonzero mod-two group for a higher-order integral cyclic group.

For $n\ge 1$, the mod-two cohomology of $K(\mathbb{Z}/2,n)$ is polynomial on the admissible operations in the displayed strict-excess range, including the empty operation. It is obtained from path-space transgressions and compatibility with Steenrod squares. Products must still be included when listing low-degree vector-space bases.

For $n\ge 2$, the positive-degree dimensions in degrees $n,n+1,n+2,n+3$ are $1,1,1,2$. The first three classes are $u,{\mathrm{Sq}}^{1}u,{\mathrm{Sq}}^{2}u$. In degree $n+3$, use ${\mathrm{Sq}}^{3}u$ and ${\mathrm{Sq}}^2{\mathrm{Sq}}^{1}u$ for $n\ge 3$; at $n=2$ replace the unstable first class by $u{\mathrm{Sq}}^{1}u$. At $n=2$ the degree-four class is $u^2$, and at $n=3$ the degree-six class ${\mathrm{Sq}}^{3}u$ is $u^2$.

The ring is exterior on its degree-$n$ universal class for odd $n$, and polynomial for even $n$. The path-loop Serre spectral sequence alternates these two types by transgression. In the polynomial-fiber case its derivation has nonzero coefficients $k$ on powers, which become invertible over the rationals.

For a CW complex $X$, abelian $G$ and $n\ge 1$, homotopy classes of maps to the Eilenberg–MacLane space correspond to cohomology classes by pullback of the universal cohomology class of an Eilenberg–MacLane space. Pointed maps give the reduced version. This converts a cohomology class into an actual map realizing its effect on homotopy.

For abelian $G$ and $n\ge 1$, the identity of $G$ determines this class under $H^n(K(G,n);G)\cong \mathrm{Hom}(G,G)$. Pulling it back realizes representability of cohomology by Eilenberg–MacLane spaces. A map inducing this class on a fiber induces the identity of its unique positive homotopy group.

A path-connected space is aspherical when its higher homotopy groups vanish, equivalently when its universal cover is contractible. An aspherical space with fundamental group $G$ is a $K(G,1)$, so based homotopy classes of maps into it are controlled by homomorphisms of fundamental groups.

Moving the basepoint around a loop acts on a higher homotopy group. Since ${\pi }_n(X)$ is an abelian group for $n\ge 2$, this action extends linearly to its integral group ring, making ${\pi }_n(X)$ a module. For a connected CW complex, its universal cover gives an equivalent description by deck transformations, together with change of basepoint upstairs. Left and right conventions are converted by inversion of the loop. The Hurewicz theorem respects this action whenever it identifies a higher homotopy group with the corresponding homology of the cover.

If $X$ is an $(n-1)$-connected based CW complex, $n\ge 2$, suspension induces an isomorphism ${\pi }_k(X)\to {\pi }_{k+1}(\Sigma X)$ for $k\le 2n-2$ and a surjection for $k=2n-1$. For a sphere, $\Sigma S^n\cong S^{n+1}$, giving the stable range ${\pi }_k(S^n)\cong {\pi }_{k+1}(S^{n+1})$.

For $\varphi :S^{2q-1}\to S^q$, $q>1$, its mapping cone has integral cohomology generators $a$ in degree $q$ and $b$ in degree $2q$. The integer defined by $a⌣a=H(\varphi )b$, with fixed cell orientations, is the Hopf invariant. It measures how an attaching map changes multiplication without changing the additive groups. The complex Hopf fibration has invariant one because its mapping cone is ${\mathbb{CP}}^2$; a constant attaching map has invariant zero.

For the Hopf map $\eta$, the indicated four-cell attachment has cellular boundary multiplication by $f$ from dimension four to three. Its integral cohomology is $\mathbb{Z}$ in degrees zero and two, $\ker (f:\mathbb{Z}\to \mathbb{Z})$ in degree three, and $\mathbb{Z}/f\mathbb{Z}$ in degree four. If $x$ generates degree two and $z$ is the four-cell cochain class, then $x^2=dz$, with every other positive-degree product zero. Collapsing the three-sphere reduces the cup-square calculation to the Hopf invariant; the induced map on degree-four cohomology reduces its integer coefficient modulo $f$. When $f=0$ there is an extra free degree-three class, whose products still vanish by dimension.

Map $X_{d,f}$ to $K(\mathbb{Z},2)={\mathbb{CP}}^{\infty }$ using its degree-two generator. The homotopy fibre is the total space of the corresponding circle bundle, is 2-connected, and has ${\pi }_3$ equal to that of $X_{d,f}$. Over $S^2\vee S^3$ its total space is $P=S^3{\cup }_{S^1}(S^1\times S^3)$, with $H_3(P)=\mathbb{Z}A\oplus \mathbb{Z}B$. The four-cell bundle is trivial; its relative degree-four generator has boundary $dA+fB$, the lifted attaching map. The long exact sequence in relative homology and Hurewicz theorem therefore give ${\pi }_3(X_{d,f})={\mathbb{Z}}^2/⟨(d,f)⟩$. Smith normal form gives the displayed expression when $d>0$, including $f=0$.

Let $\varphi :S^{2q-1}\to S^q$ and precompose with a sphere self-map of topological degree $d$. The resulting map of mapping cones is the identity on the bottom $q$-cell and has degree $d$ on the top $2q$-cell. Pulling back the defining cup product relation for the Hopf invariant multiplies its coefficient by $d$. In particular the element $d\eta \in {\pi }_3(S^2)$ has Hopf invariant $d$. Postcomposition by a degree-$d$ map of the target sphere instead multiplies the invariant by $d^2$.

A map is a weak homotopy equivalence when it induces a bijection on path components and isomorphisms on every based homotopy group.

A map is $n$-connected when it induces isomorphisms on homotopy groups below degree $n$ and a surjection in degree $n$. Equivalently, each of its homotopy fibers is $(n-1)$-connected.

For a based space $X$, adjunction gives

The rational homotopy group is obtained by tensoring a homotopy group with the rational numbers. It retains the free part and discards torsion.

The Sullivan model of the connected sum of two complex projective planes has two generators in each of degrees two and three, and none elsewhere. Its rational homotopy groups are therefore exactly the displayed two copies of ${\mathbb{Q}}^2$, with all other positive degrees zero.

For an odd-dimensional sphere $S^{2m+1}$, only ${\pi }_{2m+1}(S^{2m+1})\otimes \mathbb{Q}$ is nonzero. For an even-dimensional sphere $S^{2m}$, the nonzero rational homotopy groups occur in degrees $2m$ and $4m-1$, and both are isomorphic to $\mathbb{Q}$.

If a path-connected space is $(n-1)$-connected for $n\ge 2$, then its first potentially nonzero homotopy group maps isomorphically to homology:

If $X$ and $A$ are simply connected and the pair is $(r-1)$-connected, with $r\ge 3$, its first possible nonzero relative homotopy group maps isomorphically to relative integral homology. This lets a basis of a free relative homology group be realized by attaching disks.

For a simply connected space, if lower homotopy groups lie in an admissible Serre class, the Hurewicz map in the next degree is an isomorphism modulo that class, and the lower homology groups lie in the class. This applies to finitely generated groups and to finite groups supported at prescribed primes.

A weak homotopy equivalence between connected CW complexes is a homotopy equivalence. The analogous statement holds for Kan complexes.

For maps between simply connected spaces, a rational homology equivalence is a rational homotopy equivalence. It follows from the relative Hurewicz theorem modulo torsion, or from rationalization and the homological Whitehead theorem.

An integral homology equivalence between simply connected spaces is a weak homotopy equivalence. A first nonzero relative homotopy group would contradict the Relative Hurewicz theorem. Between CW complexes the ordinary Whitehead theorem then gives a homotopy equivalence.

The homotopy fiber of $f:X\to Y$ over $y_0$ consists of pairs $(x,\gamma )$ where $\gamma$ is a path from $f(x)$ to $y_0$. It replaces an ordinary fiber by a homotopy-invariant construction.

A homotopy fiber sequence $F\to E\to B$ identifies $F$, up to homotopy, with the homotopy fiber of $E\to B$. It induces a long exact sequence of homotopy groups.

A homotopy pullback is the derived form of a pullback. A strict pullback square of fibrant objects computes a homotopy pullback whenever one of the two maps into the lower-right object is a fibration.

The loop space $\Omega X$ is the space of based loops in $X$. It is the homotopy fiber of the inclusion of the base point into $X$.

For $n\ge 1$, there is one integral cohomology generator $a_k$ in every degree $2nk$, and its products are $a_i{a}_j=\left(\genfrac{}{}{0ex}{}{i+j}{i}\right)a_{i+j}$. The Bott–Samelson theorem supplies free polynomial homology; its primitive generator gives the binomial diagonal dual to this divided power algebra.

The homology ring formed from loop concatenation and the homology cross product. Homotopy associativity is enough for associativity in homology, and the constant loop is the unit. More generally the same construction applies to a homotopy-associative H-space.

For $n\ge 1$, the Bott–Samelson theorem gives a single homology generator of degree $2n$ with no word relations. Its Pontryagin product satisfies $x^i\ast x^j=x^{i+j}$. Unlike this polynomial homology product, the cohomology cup product has divided-power coefficients.

Apply the homology cross product and then the map induced by loop concatenation. This degree-additive bilinear product supplies the Pontryagin ring. The Bott–Samelson theorem computes it for looped suspensions with free homology.

Finite words in a based space, with occurrences of its basepoint deleted. Word concatenation defines a multiplication. For connected based CW spaces, the natural map to $\Omega \Sigma X$ is a weak homotopy equivalence. With free integral homology, the Bott–Samelson theorem identifies its homology algebra with the tensor algebra on reduced homology.

For a connected based CW complex with free integral homology, the Pontryagin ring of its looped reduced suspension is the indicated tensor algebra. The generator inclusion comes from $X\to \Omega \Sigma X$, and multiplication corresponds to concatenation of words in the James reduced product.

Loop-space homology is the homology of the based loop space. It also carries the Pontryagin product induced by concatenation of loops. For $n>2$, a Morse theory cell structure on $\Omega S^n$ has one cell in each dimension $j(n-1)$, so its integer homology is $\mathbb{Z}$ in those dimensions and zero elsewhere.

The evaluation map from the based path space gives a fiber sequenceSince $PX$ is contractible, it relates invariants of $X$ to those of its loop space.

Pull back along loop evaluation $S^1\times \Omega B\to B$ and take the slant product with the circle's fundamental class. This degree-lowering operation kills positive-degree cup products. It commutes with stable cohomology operations such as Steenrod squares and identifies a transgressing base class with its fiber class in the path-loop fibration.

Simplicial homotopy theory models spaces and higher categories using simplicial sets.

The simplex category has objects the finite nonempty ordered sets $[n]=\{0<\cdots <n\}$ and morphisms the order-preserving maps.

A simplicial set is a functor from the opposite of the simplex category to sets. Its face and degeneracy maps satisfy the simplicial identities.

The geometric realization of a simplicial set glues one topological simplex for each simplicial simplex, using the face and degeneracy maps as the gluing relations.

The singular simplicial set of a topological space $X$ has the continuous maps ${\Delta }^n\to X$ as its $n$-simplices. Evaluation gives a natural weak homotopy equivalence $|\mathrm{Sing}X|\to X$.

The standard simplicial $n$-simplex is the representable simplicial set

The $i$th horn ${\Lambda }_i^n\subseteq {\Delta }^n$ is the union of all codimension-one faces except the face opposite vertex $i$.

A horn ${\Lambda }_i^n$ is inner when $0<i<n$ and outer when $i=0$ or $i=n$.

A quasicategory is a simplicial set in which every inner horn has a filler.

A Kan complex is a simplicial set in which every simplicial horn, including every outer horn, has a filler.

The nerve of a category $\mathcal{C}$ hasso an $n$-simplex is a string of $n$ composable morphisms. Every inner horn in a nerve has a unique filler.

The simplicial mapping space is defined byIf $A\hookrightarrow B$ is a monomorphism and $X$ is a Kan complex, restriction $\underline{\mathrm{Hom}}(B,X)\to \underline{\mathrm{Hom}}(A,X)$ is a Kan fibration.

A simplicial abelian group is a simplicial object in abelian groups. Its underlying simplicial set is always a Kan complex.

For a simplicial abelian group $A$, one convention for normalized chains isEquivalently, it is the quotient of the unnormalized chain complex by its degenerate subcomplex.

The normalized-chain functor gives an equivalence between simplicial abelian groups and nonnegatively graded chain complexes of abelian groups. Its inverse is the Eilenberg–MacLane denormalization functor $K$.

For a fibration $F\to E\to B$ with suitable connectivity, the cohomological Serre spectral sequence haswith differential $d_r:E_r^{p,q}\to E_r^{p+r,q-r+1}$.

For a Serre fibration, this first-quadrant spectral sequence has differential of bidegree $(r,1-r)$ and a fiber-cohomology local coefficient system on the base. A simply connected base makes that system constant. Its edge map is restriction to the fiber; a surviving fiber class is the restriction of a total-space class. For commutative ring coefficients it is multiplicative and its differentials are graded derivations.

A Serre fibration is a map having the homotopy lifting property with respect to every disc, equivalently every CW complex.

For a Serre fibration with this fiber over a simply connected CW base, the fiber identity class has only one possible nonzero outgoing differential in the cohomological Serre spectral sequence, landing in $H^{n+1}(B;G)$. If that group vanishes, the class extends to the CW total space. Its representing map, paired with projection to $B$, is an isomorphism on fiber and base homotopy groups and hence a weak homotopy equivalence to the product.

For a map $f:X\to B$, let $E_f$ consist of points of $X$ and paths starting at their images, with projection $p(x,\gamma )=\gamma (1)$. Appending segments of a base homotopy proves the homotopy lifting property, so $p$ is a Serre fibration. The constant-path inclusion $s:X\to E_f$ satisfies $ps=f$ and is a homotopy equivalence: forget the path for an inverse, and shorten paths to prove the homotopy. The homotopy need not preserve the endpoint in $B$, so the statement concerns an equivalence of total spaces over a commuting triangle, not a fibrewise inverse. The fibre over $b_0$ is the homotopy fibre of $f$.

A map $p:E\to B$ has the homotopy lifting property with respect to $X$ when every homotopy $X\times I\to B$ whose initial map lifts to $E$ has a lift extending that initial lift.

For a based fibration $F\to E\to B$, lifting representatives defines connecting maps and an exact sequence

For a fibration over $S^n$, the only possible differential in the homological Serre spectral sequence is $d_n$. Identifying its two nonzero columns with the homology of the fiber gives the Wang homomorphism $\Delta$.

For a fibration $F\to E\to S^n$, the Wang homomorphism belongs to the exact sequence

A transgression is a spectral-sequence differential from a fiber class on the vertical edge to a base class on the horizontal edge.

For a fibration $F\to E\to B$, classes $x\in H^r(F)$ and $y\in H^{r+1}(B)$ form a transgressive pair when their images in relative cohomology satisfy $\delta x={\pi }^{\ast }y$. In the Serre spectral sequence this relation is represented by $d_{r+1}x=y$.

Steenrod operations preserve transgressive pairs. With mod-two coefficients, if $(x,y)$ is transgressive, then so is $({\mathrm{Sq}}^{j}x,{\mathrm{Sq}}^{j}y)$ whenever the square is defined by the instability range. The corresponding statement holds for odd-primary reduced powers and the Bockstein, with the appropriate page shift.