Before painstakingly defining all these terms, let me ask my question in plain english: given a $G$-bundle, is every conjugate of a vector field by a gauge transformation an element of the Lie algebra?

For a little more context, the question arises from a calculation in Exercise 86 of Part II (Gauge Fields) of Gauge fields, knots, and gravity by Baez and Muniain. Spurred on by the comments, I have decided to spell out the terminology in full detail as follows.

Let $M$ be a smooth manifold, let $G$ be a Lie group, and let $ E\to M$ be a $G$-bundle, viz. a vector bundle whose model fiber $V$ is a (finite dimensional) vector space equipped with a representation $\rho\colon G\to \mathrm{GL}(V)$ such that all bundle transition maps belong to $\rho(G)$. Denote the endormorphism bundle of $E$ by $\mathrm{End}(E)$ and let $G(E)\subset \Gamma\bigl(\mathrm{End}(E)\bigr)$ denote the gauge group of the bundle. Elements $g\in G(E)$ are sections of $\mathrm{End}(E)$ such that for every local trivialization $\phi\colon \mathrm{End}(E)\restriction U\to U\times \mathrm{GL}(V)$ and for every $p\in U$, it holds that $\phi(g_p)\in \{p\}\times \rho(G)$. The derivative $d\rho\colon\mathfrak{g}\to \mathrm{End}(V)$ furnishes an embedding of the Lie algebra of $G$ into the set of linear transformations of $V$. Paralleling the definition of $G(E)$, we define the infinitesimal gauge group $\mathfrak{g}(E)$ to be the set of sections $\mathcal{G}\in\mathrm{End}(E)$ such that for $\phi(\mathcal{G}_p)\in \{p\}\times d\rho(\mathfrak{g})$ for every $p$ and for every local trivialization $\phi$ around $p$. Finally, on any trivialization $\phi$ of $\mathrm{End}(E)\restriction U$ let $D^{\phi}$ denote the standard flat connection on the trivialization.

Question. For all vector fields $X$ and $\rho,\phi, U$ as above is $\rho(g)^{-1}D^\phi_X\rho(g)\in \mathfrak{g}(E\restriction U)$?

  • $\begingroup$ Can you please change "$g$ be a smooth section of $\text{End}(E)$" to "$s$ be a smooth section of $\text{End}(E)$".. What does it mean to say $g(p)\in G$ for all $p\in M$? $\endgroup$ Commented Jan 13, 2019 at 7:50
  • 1
    $\begingroup$ A $G$-bundle (as defined in II.2 of the aforementioned book) consists of a bundle together with a representation of $G$ on its model fiber such that the bundle transition maps belong to (the image of) $G$. Thus it is meaningful to stipulate that a section of $\textrm{End}(E)$ take values in (the image of) $G$ (see page 214 in the aforementioned book for a discussion of this point). $\endgroup$
    – pre-kidney
    Commented Jan 13, 2019 at 8:08
  • 3
    $\begingroup$ What do you mean by $g^{-1} X g$? As you said, $g(p)$ is a linear (?) map between the fibers $E_p \to E_p$. How does such a map act on a tangent vector to $M$ at $p$? Do you assume that $E$ is a natural bundle so that you have a lift of vector fields on $M$ to vector fields on $E$? $\endgroup$ Commented Jan 13, 2019 at 11:16
  • $\begingroup$ In local coordinates around $p$ write $g=g^i_je_i\otimes e^j$ and let $Xg:= X(g^i_j)e_i\otimes e^j$. An implicit part of the question is to understand whether the condition $(g^{-1}Xg)_p\in\mathfrak{g}$ using the above definition is independent of the choice of chart. $\endgroup$
    – pre-kidney
    Commented Jan 13, 2019 at 11:48
  • $\begingroup$ I think if you add all you said in comments to question it will become easier for other to read.. you might get more attention.. $\endgroup$ Commented Jan 16, 2019 at 9:47

1 Answer 1


I suggest an alternative definition of ''$G$-bundle'', not using trivializations, that may help you sort out the confusion. It appears in many places, e.g. Kobayashi Nomizu, Fundation of Dfferential Goemetry, vol 1, chap.2. Here is a quick summary.

Let $E\to M$ be a rank $n$ vector bundle and $F\to M$ the associated frame bundle. The fiber $F_x$ over a point $x\in M$ consists of all linear isomorphisms $\varphi:\mathbb{R}^n\to E_x$. The group $\mathrm{GL}_n(\mathbb{R})$ acts freely on $F$ on the right by $\varphi\mapsto \varphi\circ g$. Let $G\subset \mathrm{GL}_n(\mathbb{R})$ be a subgroup (instead of $\rho(G)$ of your question). A reduction of the structure group of $F$ to $G$ consists of a $G$-subbundle of $F$. That is, a submanifold $F^G\subset F$ such that each $F^G\cap F_x$ consists of a single $G$-orbit. The vector bundle $E$ is said to be a $G$-bundle, or equipped with a $G$-structure, if a reduction of the structure group of its frame bundle to $G$ has been chosen. Now you can show that $E$ admits a $G$-structure if and only if there exists a cover of $M$ by open subsets, over each of which $E$ can be trivialized, i.e. is isomorphic to the trivial vector bundle with fiber $\mathbb{R}^n$, and such that the transition functions between the trivializations take value in $G$.

To make all the above rigorous you need to add smoothness conditions on maps and local triviality on all bundles. I will skip it here.

You can make a rigorous definition of a $G$-structure on a vector bundle using trivializations alone, avoiding the frame bundle, but it's a bit tedious. The definition you gave in your question is not of a $G$-bundle but the condition of existence of a $G$-structure on a vector bundle. It is not the same. $G(E)$ does not make sense unless a $G$-structure has been chosen, it is not enough to know that there is one.

Anyway, using the above definition of a $G$-bundle, it is rather easy to sort out the rest. $\mathrm{Aut}(E)$ is the bundle whose fiber at a point $x\in M$ consists of all linear automorphisms of $E_x$. $G(E)\subset\mathrm{ Aut}(E)$, for a vector bundle $E$ with a given $G$-structure $F^G\subset F$, is defined as the bundle whose fiber at a point $x\in M$ consists of all linear automorphisms $f:E_x\to E_x$ such that $\varphi\circ f\circ \varphi^{-1}\in G$ for all $\varphi\in F^G_x.$ Similarly, $\mathrm{End}(E)$ is the vector bundle whose fiber at a point $x\in M$ consists of the linear transformations $T:E_x\to E_x$. Finally, $\mathfrak{g}(E)\subset\mathrm{End}(E)$ is the vector bundle whose fiber at a point $x\in M$ are the linear transformation $T:E_x\to E_x$ such that $\varphi\circ T\circ \varphi^{-1}\in\mathfrak{g}$ for all $\varphi\in F^G_x,$ where $\mathfrak{g}\subset \mathrm{End}(\mathbb{R}^n)$ is the Lie algebra of $G$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.