Category Theory, Pushout and Pullback

Pushout

The pushout is a special case of the colimit, which was described in the previous section. Start with a morphism f from Z into X, and another morphism g from Z into Y. The pushout, if it exists, is an object P, and morphisms from X into P and from Y into P, such that the entire square commutes. In other words, you can go from Z to X to P, or from Z to Y to P, and the result is the same. Furthermore, P is an initial object having this property. If Q also builds a commutative square, a morphism from P to Q makes the entire diagram, {Z,X,Y,P,Q}, commute.

The simplest example comes from sets and arbitrary functions on sets. Let Z be a subset of both X and Y, so that the morphisms into X and Y are inclusion. Let P be X union Y, with inclusion from X / Y into P as the pushout morphisms. If Q is some other extension of X and Y, map c ∈ X or Y into P, giving d, and into Q, giving e. Then map d to e. This is the unique morphism from P into Q that makes everything commute, hence P is initial, and is the unique (up to isomorphism) pushout of X and Y.

If X and Y are abelian groups, or modules, the pushout P is the direct product of X and Y, mod the relations f(c)-g(c), for each c in Z. Both X and Y embed in the direct product, and map into the quotient module. These are the pushout morphisms. Verify that P is an initial object.

Finally, X and Y could be topological spaces, whence P becomes the adjunction space. It is basically the disjoint union of X and Y, glued together at Z.

Pullback

The pullback is just like the pushout, but the arrows are reversed. Let X and Y map into Z, and the pullback P maps into X and Y, such that the square commutes, and P is terminal. In other words, P is the limit of our 3 object diagram. These are not as common as pushouts, though they are sometimes used in connection with fiber bundles.