Theorem: Pullback of a principal bundle is a principal bundle

The pullback construction sends principal bundles to principal bundles functorially in the base map.
Theorem: Pullback of a principal bundle is a principal bundle

Let f:NMf:N\to M be a between smooth manifolds, and let π:PM\pi:P\to M be a with structure GG.

Theorem (pullback principal bundle)

Define the pullback total space

fP:={(n,p)N×Pf(n)=π(p)}, f^*P := \{(n,p)\in N\times P \mid f(n)=\pi(p)\},

with projection πN:fPN\pi_N:f^*P\to N given by πN(n,p)=n\pi_N(n,p)=n. Equip fPf^*P with the right GG-action

(n,p)g:=(n,pg). (n,p)\cdot g := (n,p\cdot g).

Then πN:fPN\pi_N:f^*P\to N is a principal GG-bundle, called the pullback bundle of PP along ff.

Moreover, this construction is functorial: if g:LNg:L\to N is another smooth map, then (fg)P(f\circ g)^*P is canonically isomorphic to g(fP)g^*(f^*P) as principal bundles.

Examples

  1. Pullback of a trivial bundle. If PM×GP\cong M\times G, then fPN×Gf^*P\cong N\times G by the evident identification.

  2. Restriction to a submanifold. If i:ZMi:Z\hookrightarrow M is an embedding, then iPi^*P is the restriction of PP to ZZ; its total space is π1(Z)P\pi^{-1}(Z)\subset P.

  3. Pullback along a covering map. If f:NMf:N\to M is a covering and PMP\to M is nontrivial, fPNf^*P\to N may become trivial; for circle bundles this reflects the effect of the degree of ff on the corresponding cohomology class.