CW Complexes and Cell Attachments

In topology, we seek flexible ways to decompose and rebuild complicated spaces from simple, standardized pieces. CW complexes provide the ultimate framework for this, allowing us to construct a vast array of spaces—from surfaces to infinite-dimensional manifolds—by systematically gluing together disks of increasing dimension. This method isn't just elegant; it turns the abstract algebra of homology into a powerful, computable tool, making it indispensable for understanding the shape of spaces in modern mathematics and theoretical physics.

Defining the Building Process: Attaching Cells

A CW complex is built inductively, like a topological skyscraper, starting with a discrete set of points (0-cells). The construction proceeds dimension by dimension. For each dimension $n$, we add $n$-dimensional disks, but we don't attach them arbitrarily. Each disk is attached via a specific gluing map defined on its boundary.

Formally, we begin with a 0-skeleton, $X^0$, which is just a collection of points. To build the 1-skeleton, $X^1$, we take 1-disks (intervals) and attach their two boundary points (0-spheres) to points in $X^0$. This creates a graph. The process continues: to form the $n$-skeleton, $X^n$, we take a collection of $n$-disks, $D^n$. The boundary of each $n$-disk is an $(n-1)$-sphere, $S^{n-1}$. We attach these disks to the existing $(n-1)$-skeleton, $X^{n-1}$, via continuous attaching maps ${\varphi }_{\alpha }:S^{n-1}\to X^{n-1}$. The complex $X$ is the union of all skeletons, equipped with the weak topology: a set is open in $X$ if and only if its intersection with each closed cell (the image of a disk) is open.

This "closure-finite, weak topology" structure is the source of the name "CW." The closure-finite (C) condition means the closure of each cell intersects only finitely many other cells. The weak (W) topology ensures the space is coherently assembled from its cells. A key feature is the characteristic map ${\Phi }_{\alpha }:D^n\to X$, which extends the attaching map ${\varphi }_{\alpha }$ to the entire disk and maps the interior homeomorphically onto its designated open cell.

Cellular Homology: A Computational Power Tool

The true utility of the CW structure shines in homology theory. Cellular homology is a homology theory specifically designed for CW complexes that is often dramatically easier to compute than singular homology. It works directly with the cell structure.

The cellular chain complex is defined using the skeletons. The $n$-th chain group, $C_n^{CW}(X)$, is the free abelian group generated by the $n$-cells of $X$. The boundary homomorphism $d_n:C_n^{CW}\to C_{n-1}^{CW}$ encodes how each $n$-cell is attached to the $(n-1)$-skeleton. The degree of this map is given by the degree of the attaching map of the $n$-cell onto each $(n-1)$-cell. Crucially, the homology groups of this chain complex, $H_n^{CW}(X)$, are isomorphic to the singular homology groups $H_n(X)$.

This isomorphism transforms homology from an abstract, hard-to-compute invariant into a combinatorial procedure. For a finite CW complex, you have finitely generated chain groups in each dimension, and the boundary maps can often be calculated by examining the cell attachments. This makes computing the Betti numbers and torsion coefficients of a space a manageable exercise in linear algebra.

Canonical Examples: Spheres and Projective Spaces

The abstract definition comes to life through essential examples. The $n$-sphere $S^n$ has a simple CW structure with just two cells: one 0-cell (a base point) and one $n$-cell attached via the constant map from $S^{n-1}$ to that point. This immediately yields its homology: $C_0\cong \mathbb{Z}$, $C_n\cong \mathbb{Z}$, and all other chain groups are zero. All boundary maps are zero, so $H_0(S^n)\cong H_n(S^n)\cong \mathbb{Z}$.

More instructive are the projective spaces. Real projective space ${\mathbb{RP}}^n$ can be constructed by attaching one cell in each dimension from 0 to $n$. The attaching map for the $k$-cell is the canonical double cover $S^{k-1}\to {\mathbb{RP}}^{k-1}$. This results in a cellular chain complex where the boundary map $d_k:C_k\to C_{k-1}$ is multiplication by 0 if $k$ is odd, and by 2 if $k$ is even. Computing the homology from this gives the well-known result: for ${\mathbb{RP}}^n$, $H_k$ is $\mathbb{Z}$ for $k=0$ and (if $n$ is odd) for $k=n$; it is ${\mathbb{Z}}_2$ for odd $k<n$; and it is 0 otherwise.

Complex projective space ${\mathbb{CP}}^n$ has an even simpler CW structure because it is built from even-dimensional cells only. It has exactly one cell in each even dimension $0,2,4,...,2n$. There are no odd-dimensional cells. Consequently, all boundary maps in the cellular chain complex are zero. Its homology is therefore free abelian: $H_k({\mathbb{CP}}^n)\cong \mathbb{Z}$ for even $k$ with $0\le k\le 2n$, and $0$ otherwise. This elegant structure underscores the geometric difference between the real and complex cases.

Why CW Complexes Model Most Spaces of Interest

The power of CW complexes lies in their universality and flexibility. They are the "right" category for homotopy theory for several compelling reasons. First, any topological space is weakly homotopy equivalent to a CW complex. This means that for the purposes of studying homotopy and homology invariants, we can almost always replace a nasty space with a well-behaved CW complex without losing essential information.

Second, the category of CW complexes has excellent formal properties. Key constructions like product and quotient behave well when performed on CW complexes, often resulting in a space that can be given a natural CW structure. While the product of two CW complexes isn't automatically a CW complex in the product topology, it is in the compactly generated topology, which is suitable for homotopy theory.

Finally, the cellular approximation theorem states that any continuous map between CW complexes is homotopic to a cellular map, one that sends the $n$-skeleton of the domain into the $n$-skeleton of the codomain. This allows us to study continuous maps at the combinatorial level of cell attachments, further tightening the link between topology and algebra. For these reasons, CW complexes form the standard playground where algebraic topology is most effectively conducted.

Common Pitfalls

1. Confusing the Attaching Map with the Characteristic Map: A frequent conceptual error is to think cells are attached by the characteristic map. Remember: the attaching map $\varphi :S^{n-1}\to X^{n-1}$ is defined only on the boundary of the disk and dictates how it is glued. The characteristic map $\Phi :D^n\to X$ is defined on the entire disk and is a homeomorphism from the disk's interior to the open cell. The characteristic map is an extension of the attaching map.

1. Misunderstanding the Weak Topology: It's easy to assume a set is open if it looks open in each cell individually. The weak topology is stricter: a set is open if its intersection with the closure of each cell is open. This ensures that a cell's closure doesn't interact strangely with infinitely many other cells. Failing to account for this can lead to incorrect conclusions about continuity.

1. Overlooking the "Closure-Finite" Condition: When drawing or imagining infinite complexes, one might attach a cell whose boundary touches infinitely many lower-dimensional cells. This violates the "C" in CW. Each cell must be attached so that its closure meets only finitely many other cells. This condition is crucial for ensuring manageability in proofs and constructions, particularly those involving induction over skeletons.

1. Assuming Homology is Immediate from Cell Counts: Seeing a cell in dimension $n$ does not guarantee a $\mathbb{Z}$ summand in $H_n$. The boundary maps can cancel things out. For example, a complex with one 1-cell and one 2-cell, where the 2-cell is attached by wrapping the boundary circle twice around the 1-cell (degree 2 map), will have $H_1\cong {\mathbb{Z}}_2$, not $\mathbb{Z}$. You must compute the boundary maps in the cellular chain complex.

Summary

• CW complexes are constructed inductively by attaching disks of increasing dimension to lower-dimensional skeletons via specific attaching maps, with the resulting space given the weak topology.
• Cellular homology is a directly computable homology theory for CW complexes that is isomorphic to singular homology. The chain groups are free on the cells, and the boundary maps are determined by the degrees of the attaching maps.
• Fundamental examples include spheres (minimal cell structure) and projective spaces. Real projective space ${\mathbb{RP}}^n$ has cells in every dimension leading to torsion in homology, while complex projective space ${\mathbb{CP}}^n$ has only even-dimensional cells, resulting in simpler, torsion-free homology.
• CW complexes are the preferred objects in homotopy theory because most spaces of interest are homotopy equivalent to a CW complex, and important theorems (like cellular approximation) allow maps and homotopies to be studied combinatorially on the cell structure.