Higher Homotopy Groups and the Hurewicz Theorem

⚡ Key Takeaways

Higher homotopy groups $\pi_n(X,x_0)$ generalize the fundamental group to maps of $n$ -spheres into a space; they are abelian for $n\geq 2$ . The Hurewicz theorem states that for an $(n-1)$ -connected space ( $n\geq 2$ ), the Hurewicz map $h:\pi_n(X)\to H_n(X)$ is an isomorphism, linking homotopy and homology. Applies throughout algebraic topology, differential topology, and homotopy theory. Higher groups introduced by Eduard Čech (1932) and developed by Witold Hurewicz (1935).

💡 The Core Insight

Homotopy groups remember how spheres wrap, while homology only counts cycles up to boundaries. Below the first nonzero degree they agree perfectly (Hurewicz), giving us a computable foothold; above it they diverge wildly. The whole game of homotopy theory is leveraging this foothold — through fibrations, Postnikov towers, and spectral sequences — to climb past the degree where homology stops helping.

DefinitionHigher Homotopy Groups

Let $(X,x_0)$ be a pointed topological space and $n\geq 1$ . Let $I^n$ be the $n$ -cube with boundary $\partial I^n$ . The $n$ -th homotopy group is

1

\quad \pi_n(X,x_0)=\big\{\,f:(I^n,\partial I^n)\to (X,x_0)\,\big\}\big/\simeq

Derivation hint: maps collapsing the cube boundary to the basepoint factor through $I^n/\partial I^n\cong S^n$ ; homotopy classes rel boundary form the set. The group operation is concatenation in any one coordinate, say the first:

2

\quad (f+g)(t_1,\dots,t_n)=\begin{cases}f(2t_1,t_2,\dots,t_n),& t_1\in[0,\tfrac12]\\ g(2t_1-1,t_2,\dots,t_n),& t_1\in[\tfrac12,1]\end{cases}

Derivation hint: glue two cube-maps along the face $t_1=\tfrac12$ ; both agree (at $x_0$ ) on the shared boundary so the result is continuous. Here $\pi_0(X,x_0)$ is the pointed set of path components, $\pi_1$ is the fundamental group, and $\pi_n$ for $n\geq 2$ is abelian.

TheoremAbelianness for

n\geq 2

— Eckmann–Hilton

For $n\geq 2$ , $\pi_n(X,x_0)$ is an abelian group. More generally, a set with two unital binary operations satisfying the interchange law $(a*b)\cdot(c*d)=(a\cdot c)*(b\cdot d)$ has a single commutative, associative operation.

ProofEckmann–Hilton argument

1. For $n\geq 2$ there are at least two independent cube coordinates, giving two compositions $+_1$ (in $t_1$ ) and $+_2$ (in $t_2$ ), each with the constant map $c_{x_0}$ as unit. 2. The two operations interchange: subdividing $I^n$ into a $2\times2$ block in the $(t_1,t_2)$ -plane, one verifies

3

\quad (a+_1 b)+_2(c+_1 d)=(a+_2 c)+_1(b+_2 d)

Derivation hint: both sides describe the same four-quadrant map of the unit square into $X$ , so they are literally equal, not just homotopic. 3. Set $b=c=c_{x_0}$ : then $a+_2 d=a+_1 d$ , so the two operations coincide. 4. Set $a=d=c_{x_0}$ in [3]: $b+_2 c = c +_1 b = c+_2 b$ , giving commutativity. 5. Associativity follows likewise (or from the single-operation interchange identity). $\blacksquare$

∎

DefinitionRelative Homotopy Groups

For a pair $(X,A,x_0)$ with $x_0\in A$ and $n\geq 1$ , let $J^{n-1}=\overline{\partial I^n\setminus I^{n-1}}$ . Then

4

\quad \pi_n(X,A,x_0)=\big\{\,f:(I^n,\partial I^n,J^{n-1})\to(X,A,x_0)\,\big\}\big/\simeq

Derivation hint: the face $I^{n-1}$ maps into $A$ , the rest of the boundary $J^{n-1}$ to $x_0$ ; quotienting gives based maps $(D^n,S^{n-1})\to(X,A)$ . This is a group for $n\geq 2$ , abelian for $n\geq 3$ , and there is a long exact sequence of the pair:

5

\quad \cdots\to\pi_n(A)\xrightarrow{i_*}\pi_n(X)\xrightarrow{j_*}\pi_n(X,A)\xrightarrow{\partial}\pi_{n-1}(A)\to\cdots

Derivation hint: $\partial$ restricts a relative map to the face $I^{n-1}\subset A$ ; exactness is the standard kernel/image bookkeeping of mapping-cube boundaries.

DefinitionThe Hurewicz Homomorphism

Fix a generator (orientation class) $[S^n]\in H_n(S^n)\cong\mathbb{Z}$ . For $[f]\in\pi_n(X,x_0)$ define

6

\quad h\big([f]\big)=f_*\big([S^n]\big)\in H_n(X;\mathbb{Z})

Derivation hint: a based map $f:S^n\to X$ induces $f_*:H_n(S^n)\to H_n(X)$ ; push the fundamental class forward. Well-definedness uses homotopy invariance of $H_*$ . The map $h$ is a homomorphism, natural in $X$ , and lands in (singular or reduced) homology.

TheoremHurewicz Theorem

Let $X$ be path-connected.

(Absolute, $n\geq 2$ ). If $X$ is $(n-1)$ -connected (i.e. $\pi_i(X)=0$ for $0\leq i\leq n-1$ ), then $\tilde H_i(X)=0$ for $i<n$ and the Hurewicz map

7

\quad h:\pi_n(X)\xrightarrow{\ \cong\ }H_n(X;\mathbb{Z})

Derivation hint: build $X$ from cells of dimension $\geq n$ up to homotopy; the bottom nontrivial cells generate both groups identically. is an isomorphism, and $H_{n+1}(X)$ is a quotient of $\pi_{n+1}(X)$ .

(Case $n=1$ ). $h:\pi_1(X)\to H_1(X)$ is the abelianization map, with kernel the commutator subgroup:

8

\quad H_1(X;\mathbb{Z})\cong\pi_1(X)^{\mathrm{ab}}=\pi_1(X)/[\pi_1(X),\pi_1(X)]

Derivation hint: $H_1$ is abelian, so $h$ factors through the abelianization; surjectivity and injectivity-on-the-quotient is Poincaré's original observation.

(Relative version). If $(X,A)$ is $(n-1)$ -connected ( $n\geq 2$ ), $A$ simply connected and nonempty, then $H_i(X,A)=0$ for $i<n$ and $h:\pi_n(X,A)\to H_n(X,A)$ is an isomorphism.

ProofSketch via CW approximation

1. Replace $X$ by a CW complex (CW approximation preserves $\pi_*$ and $H_*$ ). Since $X$ is $(n-1)$ -connected, by cellular approximation we may take $X$ with a single $0$ -cell and no cells in dimensions $1,\dots,n-1$ . 2. Then the $n$ -skeleton $X^{(n)}$ is a wedge of $n$ -spheres $\bigvee_\alpha S^n_\alpha$ . For a wedge of $n$ -spheres, both $\pi_n$ and $H_n$ are free abelian on the spheres, and $h$ matches generators, hence is an isomorphism on $X^{(n)}$ . 3. Attaching $(n+1)$ -cells via maps $\varphi_\beta:S^n\to X^{(n)}$ modifies both groups by the same relations: in $\pi_n$ we kill the classes $[\varphi_\beta]$ (using $\pi_n(X^{(n)})\to\pi_n(X^{(n+1)})$ surjective, an issue requiring $n\geq2$ so $\pi_n$ is abelian and the action is trivial), in $H_n$ we kill the boundaries $\partial e_\beta = (\varphi_\beta)_*[S^n]$ . 4. Higher cells (dimension $\geq n+2$ ) affect neither $\pi_n$ nor $H_n$ . The commuting squares of $h$ on skeleta force the isomorphism [7]. 5. The cokernel statement for $H_{n+1}$ comes from comparing the cellular chain boundary with the relative Hurewicz map on $(X,X^{(n)})$ . $\blacksquare$

∎

Figure

The Hurewicz map: wrapping spheres become homology cycles

WH

Witold Hurewicz

1904–1956

Polish-American mathematician, student of Hahn and Menger in Vienna, later at MIT. In a celebrated series of four notes in the Proceedings of the Amsterdam Academy (1935–1936), Hurewicz introduced the higher homotopy groups as group objects, proved their abelianness, defined the homomorphism now bearing his name, and established the theorem connecting homotopy to homology — together with the notion of $n$ -connectedness and aspherical spaces (now Eilenberg–MacLane spaces $K(\pi,1)$ ). He also discovered the exact sequence of a fibration. Hurewicz died in a fall from a Mayan pyramid at Uxmal while attending a topology conference in Mexico.

🧠 Intuition

Think of $H_n(X)$ as a census: it counts $n$ -dimensional cycles, but a sphere wrapped twice around another sphere with opposite-cancelling regions can register as a trivial census entry while still being a nontrivial map. Homotopy refuses to let cancellation happen — it tracks the map itself, with all its folding and degree-counting fine structure. Below the first hole there is nothing to fold, so the census and the map-count agree exactly: that is Hurewicz. The Hopf map $S^3\to S^2$ is the first place they diverge — $\pi_3(S^2)=\mathbb Z$ but $H_3(S^2)=0$ — because the linking of preimage circles is invisible to homology.

ExampleHomotopy and homology of

S^n

in the critical degree

Given: the $n$ -sphere $S^n$ , $n\geq 2$ . It is $(n-1)$ -connected: $\pi_i(S^n)=0$ for $i<n$ (cellular approximation — any map $S^i\to S^n$ with $i<n$ is homotopic to a non-surjective, hence nullhomotopic, map).

Steps: 1. Apply Hurewicz [7]: $h:\pi_n(S^n)\xrightarrow{\cong}H_n(S^n)$ . 2. Compute the right side: $H_n(S^n;\mathbb Z)\cong\mathbb Z$ , generated by the fundamental class $[S^n]$ . 3. Therefore $\pi_n(S^n)\cong\mathbb Z$ , generated by the identity map, with $h(\mathrm{id})=[S^n]$ . 4. A self-map $f:S^n\to S^n$ then satisfies $h([f])=\deg(f)\cdot[S^n]$ , so $[f]$ is detected by degree.

Answer: $\boxed{\pi_n(S^n)\cong\mathbb Z,\ \text{generated by }\mathrm{id},\ \text{class}=\deg}$ (dimensionless invariant).

Example

H_1

of a surface from its fundamental group

Given: the closed orientable genus- $g$ surface $\Sigma_g$ , with

\pi_1(\Sigma_g)=\big\langle a_1,b_1,\dots,a_g,b_g \,\big|\, \textstyle\prod_{i=1}^g[a_i,b_i]=1\big\rangle.

Steps: 1. $\Sigma_g$ is not simply connected, so use the $n=1$ Hurewicz [8]: $H_1=\pi_1^{\mathrm{ab}}$ . 2. Abelianize: all commutators $[a_i,b_i]$ become trivial, so the single relation becomes $0=0$ and disappears. 3. The abelianized group is free abelian on the $2g$ generators $\bar a_i,\bar b_i$ .

Answer: $\boxed{H_1(\Sigma_g;\mathbb Z)\cong\mathbb Z^{2g}}$ (rank $=2g$ , the first Betti number).

⚠️ Common mistake

$\pi_n$ is NOT functorial in unbased spaces without care, and is NOT homotopy-invariant of the underlying set alone. The group depends on the basepoint only up to isomorphism for path-connected $X$ , but the isomorphism induced by a path is generally not canonical — it is mediated by the $\pi_1$ -action on $\pi_n$ . For non-simply-connected spaces this action can be nontrivial, so " $\pi_n(X)$ " without specifying how basepoint-change acts is incomplete data. Likewise, do not assume $\pi_n$ commutes with arbitrary colimits the way $H_n$ does: there is no excision and no Mayer–Vietoris for homotopy groups in general.

CounterexampleHomology cannot replace homotopy:

S^2

and the Hopf fibration

Consider $S^2$ . We have $H_n(S^2)=0$ for all $n\geq 3$ , so homology declares $S^2$ "uninteresting" above degree 2. But the Hopf map $\eta:S^3\to S^2$ gives

9

\quad \pi_3(S^2)\cong\mathbb Z,\qquad H_3(S^2)=0.

Derivation hint: the long exact sequence of the fibration $S^1\to S^3\to S^2$ gives $\pi_3(S^3)\cong\pi_3(S^2)$ since $\pi_3(S^1)=\pi_2(S^1)=0$ . This is exactly the first degree above the Hurewicz range, where the theorem's conclusion fails: $\pi_3\neq H_3$ . Homology is blind to the Hopf invariant (the linking number of fibers), which $\pi_3(S^2)$ records perfectly.

CounterexampleHigher homotopy of a wedge:

\pi_3(S^2\vee S^2)

One might guess homotopy groups of wedges add like homology. They do not. While $H_n(X\vee Y)=H_n(X)\oplus H_n(Y)$ in reduced homology, homotopy groups of wedges contain extra Whitehead product terms. For instance $\pi_3(S^2\vee S^2)$ is larger than $\pi_3(S^2)\oplus\pi_3(S^2)=\mathbb Z^2$ : it contains an extra $\mathbb Z$ generated by the Whitehead product $[\iota_1,\iota_2]$ of the two inclusions. Homotopy is genuinely nonlinear.

connection

The Hurewicz theorem is the engine behind the Whitehead Theorem: a map of simply connected CW complexes inducing isomorphisms on all homology is a homotopy equivalence — proved by feeding Hurewicz into the relative homotopy of the mapping cylinder. It connects to Eilenberg–MacLane Spaces, where $K(\pi,n)$ has a single nonzero homotopy group and Hurewicz pins down its bottom homology. The relative version interlocks with the Long Exact Sequence of a Fibration and underlies Postnikov Towers. The failure of Hurewicz above the connectivity range is precisely what Stable Homotopy Theory and the Serre Spectral Sequence are designed to manage, and rationally it is corrected by Rational Homotopy Theory (where $\pi_*\otimes\mathbb Q$ and $H_*(\;;\mathbb Q)$ are governed by minimal models).

🔬 Where this appears

⏳ History

- 1895–1904 — Henri Poincaré introduces the fundamental group and (in Analysis Situs) the relation

H_1\cong\pi_1^{\mathrm{ab}}

, the prototype of the

n=1

Hurewicz statement.

- 1932 — Eduard Čech proposes higher homotopy groups at the Zürich ICM, but the result is suppressed: Alexandroff and Hopf object that the groups are abelian, deeming them "uninteresting" (a misjudgment).

- 1935–1936 — Witold Hurewicz publishes his four Amsterdam notes, founding the theory: definitions, abelianness,

n

-connectedness, aspherical spaces, and the Hurewicz isomorphism theorem.

- 1941 — Hurewicz and Steenrod develop the homotopy exact sequence of fibrations, supplying the relative Hurewicz tools.

- 1944–1945 — Eilenberg and MacLane construct

K(\pi,n)

spaces, making the Hurewicz bottom-degree principle systematic.

- 1949 — J.H.C. Whitehead proves the Whitehead theorem and develops CW complexes, the natural setting for the modern Hurewicz proof.

- 1951 — Jean-Pierre Serre's thesis introduces spectral sequences for fibrations, computing many

\pi_n(S^k)

and showing how far beyond Hurewicz one can go (Serre finiteness).

- 1953 — Hurewicz's mod-

\mathcal C

generalization (Serre classes) refines the theorem:

\pi_n\to H_n

is an iso modulo a Serre class.

- 1958 — Adams resolves the Hopf invariant one problem, demonstrating the depth of phenomena Hurewicz only opened the door to.

open-problem

The Hurewicz theorem gives complete information only in the first nonvanishing degree; beyond it lies the central hard problem of the subject. The homotopy groups of spheres $\pi_{n+k}(S^n)$ (the stable stems for $k$ fixed, $n$ large) are known only in a finite range and their complete structure — encoded by the stable homotopy category, chromatic filtration, and the Adams and Adams–Novikov spectral sequences — remains open in general. Even the unstable groups resist: there is no algorithm producing $\pi_n(X)$ for arbitrary finite simply connected $X$ that is efficient, though they are in principle computable (Brown's theorem). The mod- $\mathcal C$ Hurewicz theorem and rational homotopy theory tame the rational picture completely, but the torsion — where all the genuine difficulty lives — is governed by deep number-theoretic structures (the image of $J$ , $K$ -theory, the Kervaire invariant, resolved only in 2016 by Hill–Hopkins–Ravenel except in dimension 126).

Higher Homotopy Groups and the Hurewicz Theorem

Formal Statement

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also