Crossed complex of a filtered space

Intuition Beginner

A crossed complex is the algebraic record of a space built in stages, read one floor at a time. Start from a filtered space: a growing chain $X_0\subseteq X_1\subseteq X_2\subseteq \cdots$ of subspaces filling out the whole space. The crossed complex turns each floor of that staircase into an algebraic object, and records how each floor sits inside the next by a boundary rule.

In degree one you record the loops: the ways of moving along the bottom two floors, organised as a groupoid over the set of base points. In every higher degree $n$ you record the $n$-dimensional fillings — disks, balls, and their higher analogues — whose rim sits one floor below. These higher records are commutative groups, modules acted on by the loops in degree one. The boundary rule sends each $n$-dimensional filling to its rim, one degree down.

Why bother? The fundamental group sees only one-dimensional loops. A space has higher-dimensional content the loops cannot reach. The crossed complex keeps the loops in degree one, the two-dimensional fillings in degree two with their Whitehead crossed-module structure, and all higher fillings above, with the boundary maps linking them. It is the smallest gadget that remembers all of this at once.

Visual Beginner

A vertical tower of boxes, one per degree. The bottom box, labelled degree one, holds a groupoid: dots for the base points and arrows for loops between them. Above it, boxes labelled degree two, three, and up each hold a module, drawn as a flat commutative group. Downward arrows labelled "boundary" run from each box to the one below, sending a higher filling to its rim. Sideways arrows show the degree-one loops acting on every box above.

The two structures to keep in view are the downward boundary maps and the sideways action of the degree-one loops. Together they make the tower a crossed complex rather than a loose stack of unrelated groups.

Worked example Beginner

Build the crossed complex of the circle from its smallest filtration: one dot, one arc.

Step 1. The filtration is $X_0=\{p\}$, a single point, and $X_1=S^1$, the loop, with every higher floor equal to $S^1$.

Step 2. Degree one records loops in the bottom two floors. The loops of the circle based at $p$ form the group of whole numbers: a loop is counted by how many times it winds around. So degree one is the whole numbers, organised as a one-object groupoid over the single base point $p$.

Step 3. Degree two records two-dimensional fillings of $S^1$ whose rim sits in $X_1$. There are no two-cells in this build, and the circle has no two-dimensional filling content, so degree two is the group with one element. Every higher degree is also the one-element group.

Step 4. The boundary rule from degree two to degree one has nothing to send, since degree two is a single element. The whole crossed complex is: the whole numbers in degree one, and the one-element group in every higher degree.

What this tells us: the circle's crossed complex is concentrated in degree one, where it reproduces the fundamental group. The higher floors are empty because the circle has no higher filling content. The crossed complex faithfully records that the only interesting information lives in dimension one.

Check your understanding Beginner

Formal definition Intermediate+

A crossed complex $C$ (over a set $C_0$ of objects) is a sequence of groupoids and modules $$ \cdots \xrightarrow{;\delta_{n+1};} C_n \xrightarrow{;\delta_n;} C_{n-1} \xrightarrow{;\delta_{n-1};} \cdots \xrightarrow{;\delta_3;} C_2 \xrightarrow{;\delta_2;} C_1 \rightrightarrows C_0 $$ subject to the following data and axioms ^{[Brown-Higgins-Sivera §7.1]}.

Degree one. $C_1$ is a groupoid with object set $C_0$; its source and target maps are the two arrows $C_1\rightrightarrows C_0$. Write $C_1(x,y)$ for the morphisms $x\to y$ and $C_1(x)=C_1(x,x)$ for the vertex group at $x\in C_0$.

Higher degrees. For each $n\ge 2$, $C_n$ is a totally disconnected groupoid over $C_0$: a family of groups $C_n(x)$, one per object $x\in C_0$, abelian for $n\ge 3$. Each $C_n$ carries a (right) action of the groupoid $C_1$: an element $a\in C_n(x)$ and an arrow $p\in C_1(x,y)$ produce $a^p\in C_n(y)$, with $(a^p{)}^q=a^{pq}$ and $a^{1_x}=a$. For $n=2$ the groups $C_2(x)$ may be nonabelian; for $n\ge 3$ they are abelian, so the action makes $C_n$ a module over the groupoid $C_1$.

Boundaries. For each $n\ge 2$ there is a $C_1$-equivariant morphism ${\delta }_n:C_n\to C_{n-1}$ over $C_0$ (so ${\delta }_n$ maps $C_n(x)$ to $C_{n-1}(x)$). For $n=2$, ${\delta }_2:C_2\to C_1$ lands in the vertex groups: ${\delta }_2(C_2(x))\subseteq C_1(x)$.

The axioms are:

• CC1 (${\partial }^2=0$). ${\delta }_{n-1}{\delta }_n=0$ for all $n\ge 3$, and ${\delta }_2$ followed by source-and-target into $C_0$ is constant: ${\delta }_{2}c\in C_1(x)$ is a loop for $c\in C_2(x)$.
• CC2 (equivariance). Each ${\delta }_n$ is $C_1$-equivariant: ${\delta }_n(a^p)=({\delta }_{n}a{)}^p$ for $a\in C_n(x)$, $p\in C_1(x,y)$, where the action on $C_1$ in degree $1$ is by conjugation, ${\delta }_2(c{)}^p=p^{-1}({\delta }_{2}c)p$.
• CC3 (crossed module at the bottom). ${\delta }_2:C_2\to C_1$, with the conjugation action of $C_1$ on itself and the given action on $C_2$, is a crossed module of groupoids: for $c,c^{\prime }\in C_2(x)$, the Peiffer identity $c^{\prime }{}^{{\delta }_{2}c}=(c{)}^{-1}{c}^{\prime }c$ holds.
• CC4 (module condition above). For $n\ge 3$, ${\delta }_2{C}_2(x)$ acts by the identity action on $C_n(x)$, so the $C_1$-action on $C_n$ descends to an action of the fundamental groupoid ${\pi }_{1}C=\mathrm{coker}{\delta }_2$; thus each $C_n$ ($n\ge 3$) is a ${\pi }_{1}C$-module.

Here ${\pi }_{1}C=C_1/\mathrm{im}{\delta }_2$ is the fundamental groupoid of the crossed complex, and $H_n(C)=\ker {\delta }_n/\mathrm{im}{\delta }_{n+1}$ its homology modules. A morphism $f:C\to D$ is a family $f_n:C_n\to D_n$ over a map $f_0:C_0\to D_0$, commuting with all ${\delta }_n$ and equivariant over $f_1$. Crossed complexes and their morphisms form the category $\mathsf{Crs}$.

The fundamental crossed complex of a filtered space $X_{\ast }$ is the crossed complex $$ C(X_*) = \Pi X_*, \qquad C(X_*)0 = X_0, \quad C(X)1 = \pi_1(X_1, X_0), \quad C(X)n = \pi_n(X_n, X{n-1}, ) \ (n \ge 2), $$ where ${\pi }_1(X_1,X_0)$ is the fundamental groupoid of $X_1$ on the base-point set $X_0$, and the $C(X_)n$for$n \ge 2$aretherelativehomotopygroupsoftheconsecutivefiltrationpairs,withonegroupperbasepoint[ref:TOD{O}_{R}EFBrown-Higgins-Sivera\S 7.1].Theboundary$\delta_n$istherelative-homotopyboundary$\partial \colon \pi_n(X_n, X{n-1}) \to \pi_{n-1}(X_{n-1}, X_{n-2})$composedwiththemapinducedbyinclusion(for$n \ge 3$),andfor$n = 2$itistheWhiteheadboundary$\partial \colon \pi_2(X_2, X_1) \to \pi_1(X_1)$.The$C_1$-action is the standard action of the fundamental groupoid on relative homotopy groups.

Counterexamples to common slips

• A crossed complex is not a chain complex of ${\pi }_1$-modules. Degree one is a groupoid, not a module, and $C_2$ need not be abelian; CC3 imposes a nonabelian crossed-module law there, not a linear one. Forgetting the nonabelian bottom collapses the invariant onto homology and discards exactly the data Whitehead's crossed module carries.
• ${\delta }_2$ does not land in a normal subgroup automatically; CC2 (equivariance) is what forces $\mathrm{im}{\delta }_2$ to be a normal totally-disconnected subgroupoid of $C_1$, and CC3 (Peiffer) is what forces the action of $C_2$ on itself through ${\delta }_2$ to be conjugation.
• The action of $C_1$ on $C_n$ for $n\ge 3$ descending to ${\pi }_{1}C$ is a theorem (CC4), not a definition: one must check $\mathrm{im}{\delta }_2$ acts identically before speaking of a ${\pi }_{1}C$-module structure.

Key theorem with proof Intermediate+

Theorem. Let $X_$beafilteredspace.Thegradeddata$C(X_*)n = \pi_n(X_n, X{n-1}, )$($n \ge 2$),$C(X_)1 = \pi_1(X_1, X_0)$,withtherelative-homotopyboundaries$\delta_n$andthefundamental-groupoidaction,satisfythecrossed-complexboundaryidentity:$\delta{n-1},\delta_n = 0$for$n \ge 3$,and$\delta_2$isacrossedmoduleofgroupoids.Inparticular$C(X_*)$ is a crossed complex.*

The construction follows Brown-Higgins-Sivera ^{[Brown-Higgins-Sivera §7.1]} and Whitehead's homotopy systems ^{[Whitehead 1949]}.

Proof. Fix a base point $v\in X_0$ and abbreviate $C_n={\pi }_n(X_n,X_{n-1},v)$. The boundary ${\delta }_n:C_n\to C_{n-1}$ is the composite $$ \delta_n \colon \pi_n(X_n, X_{n-1}, v) \xrightarrow{\ \partial\ } \pi_{n-1}(X_{n-1}, v) \xrightarrow{\ j_*\ } \pi_{n-1}(X_{n-1}, X_{n-2}, v), $$ where $\partial$ is the boundary map in the long exact sequence of the pair $(X_n,X_{n-1})$ and $j_{\ast }=j_{\#}$ is induced by the inclusion of pairs $j:(X_{n-1},v)\hookrightarrow (X_{n-1},X_{n-2})$ (regarding the absolute group as the relative group of the pair with second term a point, then including). For $n=2$, ${\delta }_2=\partial :{\pi }_2(X_2,X_1,v)\to {\pi }_1(X_1,v)$ is the Whitehead boundary directly.

Step 1: ${\delta }_2{\delta }_3=0$. Compute the composite ${\delta }_2{\delta }_3:C_3\to C_1$. By definition ${\delta }_3=j_{\ast }{\partial }^{(3)}$ where ${\partial }^{(3)}:{\pi }_3(X_3,X_2,v)\to {\pi }_2(X_2,v)$, and $j_{\ast }:{\pi }_2(X_2,v)\to {\pi }_2(X_2,X_1,v)$ is induced by inclusion. Then ${\delta }_2={\partial }^{(2)}:{\pi }_2(X_2,X_1,v)\to {\pi }_1(X_1,v)$. Consider the segment of the long exact sequence of the pair $(X_2,X_1)$: $$ \pi_2(X_2, v) \xrightarrow{\ j_*\ } \pi_2(X_2, X_1, v) \xrightarrow{\ \partial^{(2)}\ } \pi_1(X_1, v). $$ Exactness at ${\pi }_2(X_2,X_1,v)$ gives $\mathrm{im}{j}_{\ast }=\ker {\partial }^{(2)}$. Now ${\delta }_2{\delta }_3={\partial }^{(2)}\circ j_{\ast }\circ {\partial }^{(3)}$, and the inner composite ${\partial }^{(2)}\circ j_{\ast }$ vanishes because $j_{\ast }$ lands in $\ker {\partial }^{(2)}$ by exactness. Hence ${\delta }_2{\delta }_3=0$. The identical argument with the pair $(X_n,X_{n-1})$ in place of $(X_2,X_1)$ gives ${\delta }_{n-1}{\delta }_n={\partial }^{(n-1)}\circ j_{\ast }\circ {\partial }^{(n)}=0$ for every $n\ge 3$, since $j_{\ast }:{\pi }_{n-1}(X_{n-1},v)\to {\pi }_{n-1}(X_{n-1},X_{n-2},v)$ lands in $\ker {\partial }^{(n-1)}$ by exactness of the pair $(X_{n-1},X_{n-2})$. This is CC1.

Step 2: ${\delta }_2$ lands in a crossed module. The pair $(X_2,X_1)$ presents the Whitehead crossed module ${\Pi }_2(X_2,X_1,v)=(\partial :{\pi }_2(X_2,X_1,v)\to {\pi }_1(X_1,v))$ of 03.12.53. That unit proved that $\partial$, with the action of ${\pi }_1(X_1,v)$ on ${\pi }_2(X_2,X_1,v)$ by basepoint transport, satisfies the crossed-module axioms CM1 (equivariance) and CM2 (the Peiffer identity). Since $C_2={\pi }_2(X_2,X_1,v)$, $C_1={\pi }_1(X_1,v)$, and ${\delta }_2=\partial$ here are exactly that data, ${\delta }_2$ is a crossed module. So CC3 holds at the bottom, and CC2 is the CM1 equivariance carried verbatim from the pair.

Step 3: equivariance and the module condition above. The fundamental groupoid ${\pi }_1(X_1,v)$ acts on each ${\pi }_n(X_n,X_{n-1},v)$ by transporting the base point along loops, and each ${\delta }_n$ is equivariant because the boundary and inclusion-induced maps are natural for basepoint transport (CC2). For $n\ge 3$ the relative groups ${\pi }_n(X_n,X_{n-1},v)$ are abelian by the Eckmann-Hilton interchange on two free cube coordinates 03.12.52, and $\mathrm{im}{\delta }_2=\mathrm{im}{\partial }^{(2)}$ acts identically on them: an element of $\mathrm{im}{\delta }_2$ is the boundary of a 2-cell class, which acts on ${\pi }_n(X_n,X_{n-1})$ ($n\ge 3$) through the inclusion $X_1\hookrightarrow X_{n-1}$, and a loop bounding a disk in $X_2\subseteq X_{n-1}$ acts identically because the disk furnishes a homotopy of the transport to the constant transport. Thus the action descends to ${\pi }_{1}C={\pi }_1(X_1)/\mathrm{im}{\delta }_2={\pi }_1(X_2)$ and each $C_n$ ($n\ge 3$) is a ${\pi }_{1}C$-module (CC4). Assembling CC1-CC4, $C(X_{\ast })$ is a crossed complex. $\square$

Bridge. Step 1 is the foundational reason a crossed complex is a complex at all: the boundary square vanishes not by a calculation in the groups but because each ${\delta }_n$ routes through an absolute homotopy group sitting in the kernel of the next boundary, by exactness of the pair. This is exactly the mechanism that makes a filtered space produce a chain-like object, and it generalises the elementary fact that the connecting map of a long exact sequence composed with the next inclusion vanishes. Putting these together, the crossed complex $C(X_{\ast })$ builds toward the higher homotopy van Kampen theorem, where $C=\Pi$ is shown to preserve the colimits of connected filtered spaces, and the crossed-module bottom proved in Step 2 appears again in 03.12.53 as the value of ${\Pi }_2$ on the pair $(X_2,X_1)$. The bridge is the recognition that the nonabelian degree-two data and the abelian higher data are bound into a single object by one boundary identity.

Exercises Intermediate+

Advanced results Master

The crossed complex $C(X_{\ast })=\Pi X_{\ast }$ is the value of a functor $\Pi :\mathsf{FTop}\to \mathsf{Crs}$ from filtered spaces to crossed complexes, and on the connected filtered spaces of 03.12.54 it is the carrier of the central theorem of the subject. The boundary identity proved at Intermediate tier makes $C(X_{\ast })$ a complex; the crossed-module bottom and the groupoid action make it nonabelian and basepoint-sensitive in exactly the degrees where abelian homology loses information. Three structural facts organise the rest of the theory.

The fundamental crossed complex is a free crossed resolution on skeleta. For the skeletal filtration of a connected CW complex $X$, $C(X_{\ast }{)}_n={\pi }_n(X^{(n)},X^{(n-1)})$ is the free ${\pi }_{1}X$-module (for $n\ge 3$) on the $n$-cells of $X$, and $C(X_{\ast }{)}_2$ is the free crossed ${\pi }_1(X^{(1)})$-module on the 2-cells, with ${\delta }_n$ the cellular boundary read in the universal cover. Thus $C(X_{\ast })$ is a free crossed resolution of ${\pi }_{1}X$ when $X$ is aspherical, and in general its homology is $H_n(\tilde{X})$ as a $\mathbb{Z}[{\pi }_{1}X]$-module. This is Whitehead's homotopy system ^{[Whitehead 1949]} recast functorially.

The higher homotopy van Kampen theorem. The functor $\Pi$ carries certain colimits of connected filtered spaces to colimits of crossed complexes: if $X_{\ast }$ is the union of connected filtered subspaces glued along connected filtered intersections satisfying the connectivity condition $\varphi$, then $\Pi X_{\ast }$ is the corresponding colimit of crossed complexes ^{[Brown-Higgins 1981]}. This is the theorem that computes relative ${\pi }_2$ in cases classical methods cannot, and it is the reason crossed complexes — rather than chain complexes — are the target: a colimit of nonabelian crossed modules captures the gluing of two-dimensional homotopy that an abelianised invariant destroys.

Tensor product and monoidal closed structure. The category $\mathsf{Crs}$ carries a tensor product $\otimes$ and an internal hom, making it monoidal closed ^{[Brown-Higgins-Sivera §7.1]}. For crossed complexes $C,D$ the tensor product $C\otimes D$ is the crossed complex generated by elements $c\otimes d$ in total degree $|c|+|d|$, subject to bilinearity, the action rules, and boundary formulas modelled on the Leibniz rule $$ \delta(c \otimes d) = (\delta c) \otimes d \pm c \otimes (\delta d), $$ with the low-degree terms governed by the crossed-module and groupoid structure (in degree one the tensor of two groupoids involves their free product over the object sets, and the cross-terms $c\otimes d$ with $|c|=1$ encode the action).

The right adjoint is an internal hom $\mathrm{CRS}(D,E)$: there is a natural bijection $$ \mathsf{Crs}(C \otimes D, E) \cong \mathsf{Crs}(C, \mathrm{CRS}(D, E)), $$ and the elements of $\mathrm{CRS}(D,E)$ in degree zero are the crossed-complex morphisms $D\to E$, while the higher degrees are the homotopies and higher homotopies between them. The homotopy-classification significance is direct: homotopies of crossed-complex morphisms are exactly the degree-one elements of the internal hom, so $\mathrm{CRS}(D,E)$ packages the entire homotopy theory of maps out of $D$ into a single crossed complex. The tensor product satisfies an exponential law for filtered spaces, $\Pi (X_{\ast }\otimes Y_{\ast })\cong \Pi X_{\ast }\otimes \Pi Y_{\ast }$ on suitably nice (e.g. CW) filtrations, which is the engine behind the equivalence of the crossed-complex and cubical $\omega$-groupoid models and behind the classifying-space functor $B$ that realises a crossed complex as a space.

Synthesis. The crossed complex is the foundational reason the higher homotopy van Kampen theorem can be both stated and computed: it is the single object that holds the nonabelian degree-two crossed module of 03.12.53 and the abelian higher modules of 03.12.52 together under one boundary, and the central insight is that the boundary square vanishes by exactness while the bottom stays nonabelian. This is exactly the combination that lets $\Pi$ preserve colimits — gluing filtered spaces along subspaces becomes a colimit of crossed complexes — which generalises the groupoid Seifert-van Kampen theorem from dimension one to all dimensions at once. The tensor product is dual to the internal hom in the monoidal closed structure, and putting these together the homotopy theory of crossed-complex maps is internalised into $\mathrm{CRS}(D,E)$, so that homotopies become elements rather than external data; the bridge to the cubical model is the exponential law $\Pi (X_{\ast }\otimes Y_{\ast })\cong \Pi X_{\ast }\otimes \Pi Y_{\ast }$, which builds toward the equivalence of 03.12.57 and appears again in the classifying-space construction. The crossed complex is therefore not a convenient bookkeeping device but the precise algebraic shadow a filtered space casts: a nonabelian chain complex whose vanishing boundary square, crossed-module bottom, and groupoid action are each forced by the homotopy theory they record.

Full proof set Master

Proposition (the crossed-complex boundary identity, full form). For a filtered space $X_$andtheboundaries$\delta_n = j_* \circ \partial^{(n)}$definedabove,$\delta_{n-1}\delta_n = 0$forall$n \ge 3$,andtheseboundariesare$C_1$-equivariant.Consequently$C(X_*)$ satisfies CC1 and CC2.*

Proof. Write ${\partial }^{(n)}:{\pi }_n(X_n,X_{n-1},v)\to {\pi }_{n-1}(X_{n-1},v)$ for the connecting map of the pair $(X_n,X_{n-1})$ and $j_{\ast }:{\pi }_{n-1}(X_{n-1},v)\to {\pi }_{n-1}(X_{n-1},X_{n-2},v)$ for the inclusion-induced map, so ${\delta }_n=j_{\ast }{\partial }^{(n)}$. The composite is $$ \delta_{n-1}\delta_n = \big(j_* \partial^{(n-1)}\big)\big(j_* \partial^{(n)}\big) = j_* ,\big(\partial^{(n-1)} j_*\big), \partial^{(n)}. $$ The inner factor ${\partial }^{(n-1)}{j}_{\ast }$ is the composite $$ \pi_{n-1}(X_{n-1}, v) \xrightarrow{\ j_*\ } \pi_{n-1}(X_{n-1}, X_{n-2}, v) \xrightarrow{\ \partial^{(n-1)}\ } \pi_{n-2}(X_{n-2}, v). $$ In the long exact sequence of the pair $(X_{n-1},X_{n-2})$ the segment $$ \pi_{n-1}(X_{n-1}, v) \xrightarrow{\ j_*\ } \pi_{n-1}(X_{n-1}, X_{n-2}, v) \xrightarrow{\ \partial^{(n-1)}\ } \pi_{n-2}(X_{n-2}, v) $$ is exact at the middle term, and moreover ${\partial }^{(n-1)}\circ j_{\ast }=0$ because $j_{\ast }$ is the map from the absolute group into the relative group and the connecting map kills its image — this is the standard composite $\partial \circ j=0$ in the homotopy long exact sequence (consecutive maps compose to zero). Hence ${\partial }^{(n-1)}{j}_{\ast }=0$, so ${\delta }_{n-1}{\delta }_n=j_{\ast }\cdot 0\cdot {\partial }^{(n)}=0$. For $n=3$ the same computation with ${\delta }_2={\partial }^{(2)}$ (no $j_{\ast }$ at the bottom, since ${\delta }_2$ targets the absolute ${\pi }_1(X_1)$) gives ${\delta }_2{\delta }_3={\partial }^{(2)}{j}_{\ast }{\partial }^{(3)}=0$, using ${\partial }^{(2)}{j}_{\ast }=0$ from the pair $(X_2,X_1)$.

For equivariance, the fundamental groupoid ${\pi }_1(X_1,X_0)$ acts on every relative group ${\pi }_n(X_n,X_{n-1})$ by transporting the base point along a path in $X_1\subseteq X_{n-1}$; the connecting map ${\partial }^{(n)}$ and the inclusion-induced $j_{\ast }$ are both natural transformations of functors of the basepoint, so they commute with transport. Concretely, for $a\in C_n(x)$ and $p\in C_1(x,y)$, ${\delta }_n(a^p)=(j_{\ast }{\partial }^{(n)}a{)}^p=({\delta }_{n}a{)}^p$, with the action on $C_1$ being conjugation in degree one. Thus CC1 and CC2 hold. $\square$

Proposition ($\Pi$ is a functor and preserves the bottom crossed module). The assignment $X_ \mapsto C(X_*)$extendstoafunctor$\Pi \colon \mathsf{FTop} \to \mathsf{Crs}$,andforevery$X_*$thetruncation$\tau_{\le 2} \Pi X_* = \big(\delta_2 \colon C_2 \to C_1\big)$istheWhiteheadcrossedmodule$\Pi_2(X_2, X_1, )$ofthepair$(X_2, X_1)$.

Proof. A filtered map $f:X_{\ast }\to Y_{\ast }$ satisfies $f(X_n)\subseteq Y_n$, so it restricts to maps of pairs $(X_n,X_{n-1})\to (Y_n,Y_{n-1})$ and to $(X_1,X_0)\to (Y_1,Y_0)$. Relative homotopy groups are functors on maps of pairs, the fundamental groupoid is a functor on the bottom, and the boundary ${\partial }^{(n)}$ and inclusion $j_{\ast }$ are natural, so $f$ induces $C(f{)}_n=(f{|}_{(X_n,X_{n-1})}{)}_{\ast }$ commuting with all ${\delta }_n$ and equivariant over $C(f{)}_1$. This is a morphism in $\mathsf{Crs}$. Functoriality $C(gf)=C(g)C(f)$ and $C(\mathrm{id})=\mathrm{id}$ follow from functoriality of each relative homotopy group in the map of pairs. For the truncation: in degree one and two, $C_1={\pi }_1(X_1,X_0)$ over the single basepoint $v$ reduces to ${\pi }_1(X_1,v)$, $C_2={\pi }_2(X_2,X_1,v)$, and ${\delta }_2={\partial }^{(2)}$ is precisely the boundary of the pair $(X_2,X_1)$ whose crossed-module structure was established in 03.12.53: the action of ${\pi }_1(X_1)$ is basepoint transport, CM1 is CC2 restricted to the bottom, and CM2 is CC3. Hence ${\tau }_{\le 2}\Pi X_{\ast }={\Pi }_2(X_2,X_1,v)$ as a crossed module. $\square$

Stated without proof — see Brown-Higgins-Sivera §7.1 and §9 ^{[Brown-Higgins-Sivera §7.1]}. The category $\mathsf{Crs}$ is symmetric monoidal closed under the tensor product $\otimes$ and internal hom $\mathrm{CRS}(-,-)$, with unit the crossed complex $\mathbb{Z}$ concentrated in degree zero on a point; the natural isomorphism $\mathsf{Crs}(C\otimes D,E)\cong \mathsf{Crs}(C,\mathrm{CRS}(D,E))$ holds, and on CW-filtered spaces $\Pi$ is monoidal: $\Pi (X_{\ast }\otimes Y_{\ast })\cong \Pi X_{\ast }\otimes \Pi Y_{\ast }$. The higher homotopy van Kampen theorem then states that $\Pi$ preserves the colimits of connected filtered spaces glued along connected filtered subspaces.

Connections Master

• The relative homotopy group 03.12.52 supplies the graded objects of the crossed complex: $C(X_{\ast }{)}_n={\pi }_n(X_n,X_{n-1},\ast )$ for $n\ge 2$, with the action of ${\pi }_1$ and the boundary $\partial$ both imported from that unit. The abelianness of ${\pi }_n(X_n,X_{n-1})$ for $n\ge 3$, proved there by the Eckmann-Hilton interchange, is what makes the higher degrees modules rather than nonabelian groups, and the boundary identity here is the long-exact-sequence fact $\partial \circ j=0$ specialised to consecutive filtration pairs.

• Whitehead's crossed module of a pair 03.12.53 is the bottom two stages ${\delta }_2:C_2\to C_1$ of the crossed complex: the truncation ${\tau }_{\le 2}\Pi X_{\ast }={\Pi }_2(X_2,X_1,\ast )$ is exactly that crossed module, and the Peiffer identity proved there is axiom CC3. The free crossed modules computed in that unit are the degree-two part of the free crossed resolution that the crossed complex of a CW filtration provides.

• The filtered space 03.12.54 is the geometric input: the crossed complex is the value of the functor $\Pi :\mathsf{FTop}\to \mathsf{Crs}$ on it, and the connectivity condition $\varphi$ established there is precisely what makes $\Pi$ lose no cellular information and preserve the colimits that drive the higher homotopy van Kampen theorem. The colimit computation in $\mathsf{FTop}$ proved in that unit is the geometric side that $\Pi$ must carry to colimits of crossed complexes.

• The 2-group / internal-category picture 03.12.56 is the bottom-degree shadow of the crossed complex: truncating to degrees one and two and applying the Brown-Spencer equivalence presents ${\tau }_{\le 2}\Pi X_{\ast }$ as a strict 2-group, locating the crossed complex inside higher-categorical homotopy theory as the strict $\omega$-groupoid model of a homotopy type.

• The cubical $\omega$-groupoid of a filtered space 03.12.57 is the equivalent cubical model: the functor $\rho X_{\ast }$ built from filtered maps of cubes is equivalent to $\Pi X_{\ast }$ as a category, and the equivalence is proved using the tensor product and exponential law $\Pi (X_{\ast }\otimes Y_{\ast })\cong \Pi X_{\ast }\otimes \Pi Y_{\ast }$ developed in the Advanced results above. The crossed complex is the globular face of the same data the cube records.

• The higher homotopy van Kampen theorem for crossed complexes 03.12.59 is the colimit theorem whose receiving category is $\mathsf{Crs}$: $\Pi$ of a union of connected filtered spaces is the colimit (pushout) of crossed complexes, and the crossed complex constructed in this unit is the object on which that theorem operates, generalising the groupoid Seifert-van Kampen theorem to all dimensions.

Historical & philosophical context Master

The crossed complex has two roots. The first is J.H.C. Whitehead's Combinatorial homotopy II (Bull. Amer. Math. Soc. 55, 1949, 453-496) ^{[Whitehead 1949]}, where the relative homotopy groups of the skeletal filtration of a CW complex, together with their boundary operators and the bottom crossed module, were assembled into what Whitehead called a homotopy system — the object now recognised as the fundamental crossed complex of a CW filtration. Whitehead used homotopy systems to study the algebraic structure of ${\pi }_2$ and the homotopy classification of 2-dimensional and 3-dimensional complexes, and his free-crossed-module description of the 2-cells is the degree-two part of the modern construction.

The second root is the higher van Kampen programme of Ronald Brown and Philip J. Higgins. Their 1981 paper Colimit theorems for relative homotopy groups (J. Pure Appl. Algebra 22, 11-41) ^{[Brown-Higgins 1981]} isolated the crossed complex of a filtered space as the carrier of a higher-dimensional van Kampen theorem and proved that the functor $\Pi$ preserves the relevant colimits, making relative second homotopy groups computable in cases beyond the reach of classical homology. The tensor product and monoidal closed structure, the equivalence with cubical $\omega$-groupoids, and the systematic role of the connectivity condition were consolidated in the 2011 monograph of Brown, Higgins, and Sivera, Nonabelian Algebraic Topology (EMS Tracts in Mathematics 15) ^{[Brown-Higgins-Sivera §7.1]}, whose §7.1 develops the crossed complex of a filtered space and §9 the tensor product. This work places the crossed complex as the basic algebraic model of a filtered homotopy type in the strict-$\omega$-groupoid lineage, parallel to and equivalent in content with the $\infty$-categorical models of Joyal and Lurie.

Bibliography Master

@book{BrownHigginsSivera2011NAT,
  author    = {Brown, Ronald and Higgins, Philip J. and Sivera, Rafael},
  title     = {Nonabelian Algebraic Topology: Filtered Spaces, Crossed Complexes, Cubical Homotopy Groupoids},
  series    = {EMS Tracts in Mathematics},
  volume    = {15},
  publisher = {European Mathematical Society},
  year      = {2011}
}

@article{BrownHiggins1981Colimit,
  author  = {Brown, Ronald and Higgins, Philip J.},
  title   = {Colimit theorems for relative homotopy groups},
  journal = {J. Pure Appl. Algebra},
  volume  = {22},
  year    = {1981},
  pages   = {11--41}
}

@article{BrownHiggins1981Cubes,
  author  = {Brown, Ronald and Higgins, Philip J.},
  title   = {On the algebra of cubes},
  journal = {J. Pure Appl. Algebra},
  volume  = {21},
  year    = {1981},
  pages   = {233--260}
}

@article{Whitehead1949CombinatorialII,
  author  = {Whitehead, J. H. C.},
  title   = {Combinatorial homotopy {II}},
  journal = {Bulletin of the American Mathematical Society},
  volume  = {55},
  year    = {1949},
  pages   = {453--496}
}