I think this will be an easy question for numerous people.

Let $\pi:P\rightarrow M$ be a principal bundle and $\rho:G\rightarrow GL(V)$ a representation.

The space of $k$ forms on $M$ with values in $P\times_G V$ (denote as $\Omega^k(M;P\times_G V)$ can be identified with with the space of horizontal, right invariant $k$-forms on $M$ (denote as $\Omega^k_G(P;V))$.

Ie, there is an isomorphism:

$\Omega^k_G(P;V)\cong \Omega^k(M;P\times_G V)$.

I am reading through some lecture notes which say

Let $\overline{\zeta}\in \Omega^k_G(P;V)$. Define $\zeta_{\alpha}=s_{\alpha}^*\overline{\zeta}\in \Omega^k(U_{\alpha};V)$. ($s_{\alpha}$ is the local section $s_{\alpha}:U_{\alpha}\rightarrow P$).

It then asks to show that $\{\zeta_{\alpha}\}$ define a form in $\Omega^k(M;P\times_G V)$ by showing that the 'gluing' equation is satisfied

$\zeta_{\alpha}=\rho(g_{\alpha\beta})\circ \zeta_{\beta}$.

Here $g_{\alpha\beta}$ is the transition functions related to the local trivialisations which satisfy $s_{\beta}(m)=s_{\alpha}(m)g_{\alpha\beta}(m)$.

I have managed to show that the required equation holds. My question is - why do the constructed $\zeta_{\alpha}\in \Omega^k(U_{\alpha};V)$ define forms in $\Omega^k(M;P\times_G V)$? I understand that $P\times_G V$ has the structure of a fibre bundle with typical fibre $V$, but I am not sure why the gluing equation is important. I am guessing it has something to do with the fact that because the equation holds, one is able to extend the local definition go a global one. I'm not sure. If someone can help that would be great.

  • 1
    $\begingroup$ In general, a vector bundle can be specified by the data of a vector space $V$, a cover $U_\alpha$ of your manifold, and functions $h_{\alpha\beta} : U_\alpha\cap U_\beta \to GL(V)$ satisfying the cocycle condition. Under this definition of a vector bundle, a section is a collection of functions $\zeta_\alpha : U_\alpha \to V$ such that $\zeta_\alpha(x) = h_{\alpha\beta}(x) \zeta_\beta(x)$. The associated vector bundle has the same transition functions as the principal bundle, but now acting via the rep, so the transition functions are $\rho(g_{\alpha\beta})$. $\endgroup$ – Eric O. Korman Mar 16 '15 at 6:19
  • 1
    $\begingroup$ So on $U_\alpha$ you just have a trivial vector bundle with fiber $V$ but then the transition functions tell you how to glue (i.e. identify) these trivial vector bundles together on overlaps. To get a global section you then need the sections on each piece to be compatible via the transition functions. $\endgroup$ – Eric O. Korman Mar 16 '15 at 6:22

This is a situation where generalizing makes things clearer. Suppose that $X$ is a manifold and $G$ is a lie group acting freely and properly discontinuously on $X$. It follows that the quotient $X/G$ exists in the category of manifolds and we have a smooth map $ \rho : X \to X/G$. Assume that $ \pi : E \to X $ is a $G$-equivariant vector bundle. That is, $E$ is a vector bundle with a $G$ action such that $ \pi $ is $G$-equivariant and $ g : E_x \to E_{gx}$ is a linear isomorphism. Note that if $s$ is a section of $E$, then $g s g^{-1}$ is a section of $E$. Therefore $G$ acts on the sections of $E$.

The quotient $E/G$ is a vector bundle on $X/G$. In the quotient $E/G$, $E_x$ and $E_{gx}$ get identified via multiplication by $g$. Moreover, if $ U \subseteq X/G$ is open, then $$ H^0(E/G,U) = H^0(E,\pi^{-1}(U))^G $$ If this is not clear, try and draw some cartoons in the case when $G$ is a finite group. This is a special case of the following more general fact: suppose that $E$ and $F$ are $G$-equivariant vector bundles on $X$. Then $$ {\rm Hom} \, (E/G,F/G) = {\rm Hom}(E,F)^G $$

Now consider the following exact sequence of $G$-equivariant vector bundles on $X$: $$ 0 \to V \to TX \xrightarrow{\rho'} \rho^* T(X/g) \to 0 $$ We call $V$ the vertical subbundle: It consists of vectors which are tangent to the orbits. The isomorphism $ \rho' : TX / V \to \rho^* T(X/G)$ induces an isomorphism $(TX/V)/G \cong T(X/G)$. If $E$ is a $G$-equivariant vector bundle on $X$ then $$ {\rm Hom} \, (T(X/G),E/G) = {\rm Hom} \, ((TX/V)/G,E/G) = {\rm Hom} \, (TX/V,E)^G $$ This says that $E/G$-valued 1-forms on $X/G$ are the same as $E$-valued 1-forms $\eta$ on $X$ such that

  • $\eta(v) = 0 $ for tangent vectors $v \in V$
  • $L_g^* \eta = g \cdot \eta $

extending this to $p$-forms is easy: the first condition becomes $\eta(v_1,\dots,v_p) = 0 $ whenever one of the $v_i$ is tangent to a orbit and the second does not change.

Now let us apply this framework to your problem. Assume that $M$ is a manifold, $G$ is a Lie group, $ \pi : P \to G$ is a principal $G$-bundle and $V$ is a $G$-representation. Then $ P / G = M $. The vector bundle $P \times V$ is $G$-equivariant with $G$-action given by $$ (p,v) \cdot g = (pg,g^{-1}v) $$ The quotient bundle $P \times V / G = P \times_G V $ is a vector bundle on $M$. A $P \times_G V$-valued $p$-form on $M$ is the same as a $V$-valued $p$-form $\eta$ on $P$ such that

  • $\eta(v_1,\dots,v_p) = 0 $ when one of the $v_i$ is tangent to the fiber
  • $R_g^* \eta = g^{-1} \cdot \eta $

The formulas look a little different because $G$ acts on $P$ from the right instead of the left. This is exactly this isomorphism $\Omega^p_G(P,V) \cong \Omega^p( M, P\times_G V) $ which you were asking about.


Same thing as your other question, here is a note I took: https://www.evernote.com/shard/s318/sh/e4354637-e8a0-4a89-9043-4507303f7006/af3513ed77440494665094c5f4270cc3

  • $\begingroup$ Thanks for the notes... yes they are good lecture notes but the author's notes have lots of different notation and there are a lot of things I think he takes for granted which I don't understand. Still. they're very good $\endgroup$ – beedge89 Apr 1 '15 at 10:51
  • $\begingroup$ I think this is the price to pay to be more explicit, diff. geo. for gauge theories works with objects that can have several "lives" in different structures $\endgroup$ – ubugnu Apr 1 '15 at 11:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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