Let π : P → M \pi:P\to M π : P → M be a principal G-bundle
with Lie algebra g \mathfrak{g} g . The adjoint bundle is the vector bundle
a d ( P ) ≔ P × A d g → M ,
\mathrm{ad}(P)\coloneqq P\times_{\mathrm{Ad}} \mathfrak{g}\;\to\;M,
ad ( P ) : = P × Ad g → M , associated to the adjoint action of G G G on the Lie algebra
g \mathfrak{g} g .
A principal connection on P P P induces a connection ∇ \nabla ∇ on a d ( P ) \mathrm{ad}(P) ad ( P ) (equivalently, it induces the operator on tensorial forms described by the exterior covariant derivative
). The covariant exterior derivative
d ∇ : Ω k ( M ; a d ( P ) ) → Ω k + 1 ( M ; a d ( P ) )
d_\nabla:\Omega^k(M;\mathrm{ad}(P))\to \Omega^{k+1}(M;\mathrm{ad}(P))
d ∇ : Ω k ( M ; ad ( P )) → Ω k + 1 ( M ; ad ( P )) is the unique graded derivation extending ∇ \nabla ∇ and satisfying the usual Koszul formula: for α ∈ Ω k ( M ; a d ( P ) ) \alpha\in \Omega^k(M;\mathrm{ad}(P)) α ∈ Ω k ( M ; ad ( P )) and vector fields X 0 , … , X k X_0,\dots,X_k X 0 , … , X k ,
( d ∇ α ) ( X 0 , … , X k ) = ∑ i = 0 k ( − 1 ) i ∇ X i ( α ( X 0 , … , X i ^ , … , X k ) ) + ∑ 0 ≤ i < j ≤ k ( − 1 ) i + j α ( [ X i , X j ] , X 0 , … , X i ^ , … , X j ^ , … , X k ) ,
\begin{aligned}
(d_\nabla \alpha)(X_0,\dots,X_k)
&= \sum_{i=0}^k (-1)^i\,\nabla_{X_i}\bigl(\alpha(X_0,\dots,\widehat{X_i},\dots,X_k)\bigr) \\
&\quad + \sum_{0\le i<j\le k} (-1)^{i+j}\,\alpha([X_i,X_j],X_0,\dots,\widehat{X_i},\dots,\widehat{X_j},\dots,X_k),
\end{aligned}
( d ∇ α ) ( X 0 , … , X k ) = i = 0 ∑ k ( − 1 ) i ∇ X i ( α ( X 0 , … , X i , … , X k ) ) + 0 ≤ i < j ≤ k ∑ ( − 1 ) i + j α ([ X i , X j ] , X 0 , … , X i , … , X j , … , X k ) , where the bracket is the Lie bracket
of vector fields on M M M .
Locally, if A ∈ Ω 1 ( U ; g ) A\in \Omega^1(U;\mathfrak{g}) A ∈ Ω 1 ( U ; g ) is a local connection form and we identify a d ( P ) ∣ U ≅ U × g \mathrm{ad}(P)|_U\cong U\times \mathfrak{g} ad ( P ) ∣ U ≅ U × g , then for α ∈ Ω k ( U ; g ) \alpha\in \Omega^k(U;\mathfrak{g}) α ∈ Ω k ( U ; g ) ,
d ∇ α = d α + [ A ∧ α ] ,
d_\nabla \alpha = d\alpha + [A\wedge \alpha],
d ∇ α = d α + [ A ∧ α ] , with d d d the exterior derivative
.
Examples Sections of a d ( P ) \mathrm{ad}(P) ad ( P ) (degree 0). For ϕ ∈ Ω 0 ( M ; a d ( P ) ) \phi\in \Omega^0(M;\mathrm{ad}(P)) ϕ ∈ Ω 0 ( M ; ad ( P )) , locally ϕ : U → g \phi:U\to\mathfrak{g} ϕ : U → g , one has
d ∇ ϕ = d ϕ + [ A , ϕ ] ,
d_\nabla \phi = d\phi + [A,\phi],
d ∇ ϕ = d ϕ + [ A , ϕ ] , which is the usual covariant derivative in the adjoint representation.
Bianchi identity on the base. Let F ∈ Ω 2 ( U ; g ) F\in \Omega^2(U;\mathfrak{g}) F ∈ Ω 2 ( U ; g ) be the local curvature 2-form. Then
d ∇ F = 0
d_\nabla F = 0
d ∇ F = 0 is the Bianchi identity written as a covariant closure condition.
Abelian structure group. If G G G is abelian, then [ A ∧ α ] = 0 [A\wedge \alpha]=0 [ A ∧ α ] = 0 for all α \alpha α , so d ∇ d_\nabla d ∇ reduces to the ordinary exterior derivative on g \mathfrak{g} g -valued forms.