Let M be compact riemannian manifold without boundary, and G be a compact Lie group. A principle G-bundle over M, denoted as P(M,G) is a manifold P with a free right action P×G∋(p,g)↦pg∈P of G such that M=P/G, and P is locally trivial, i.e., for every point x∈M, there is a neighbourhood U such that the primage π−1(U) of the canonical projection is isomorphic to U×G in the sense that it preserver the fiber and G-equivariant, more precisely, there is a diffeomorphism Φ:π−1(U)→U×G such that Φ(p)=(π(p),ϕ(p)) and ϕ:π−1(U)→G satisfying ϕ(pg)=(ϕ(p))g for all p∈π−1(U) and g∈G.
Now for any other manifold F on which G acts on the left G×F∋(g,f)↦gf∈F, the associated fiber bundle P×ρF is the quotient space P×F/∼, where [p,f]∼[pg,g−1f], for all g∈G. With the projection ˜π:[p,g]↦π(p) this is a principle bundle over M with typical fiber F. The local trivialization is induced by the one Φ:π−1(U)→U×G as
˜Φ:˜π−1(U)→Uα×F[p,f]↦(π(p),ϕ(p)f).
We will mainly encounter the associated bundle Ad(P):=P×cG and ad(P):=gP:=P×adg, where c is the conjugate action defined by c:G∋g↦cg∈Aut(G), cg(h)=ghg−1 for h∈G, and ad=dId(Ad), and Adg=de(cg), e,Id are the unit elements in G and g respectively.
The are three descriptions of gauge group (the group of gauge transformation) G(P) of a principle G-bundle P(M,G) in the literature [^1], [^2]. Firstly, a gauge transformation can be viewed as G-bundle automorphisms, i.e., S:P→P is a diffeomorphism such that are preserve the fiber, π∘S=S and that are equivariant, S(pg)=(S(p))g for all p∈P, g∈G. It can be verified that S has an inverse S−1∈AutM(P).
Secondly, a gauge transformation can be viewed as a smooth map u:P→G which is equivariant, i.e.
u(pg)=g−1u(p)g,∀p∈P,g∈G.
In fact, given S∈AutM(P), it correspond (bijective) to u by the relation S(p)=pu(p), for any p∈P. It is clearly that it preserve the fiber and the equivariant follows from the action is free.
Lastly, the gauge group is isomorphic to the group of sections of Ad(P). In fact, given u∈G(P), the corresponding section ˉu:M→Ad(P) is given by ˉu(π(p))=[p,u(p)], for all p∈P.
There are also three different definitions of connections on a principle G-bundle P=P(M,G), firstly as a horizontal distribution” H⊂TP such that
- for every p∈P,
TpP=Hp⊕Vp,
where Vp=ker(π∗)⊂TpP is the vertical subspace. What’s more π∗|p:Hp→Tπ(p)M is a linear isomorphism; - for every p∈P and for every g∈G,
(Rg)∗p(Hp)=Hpg,
i.e., H is G-invariant under the right action.
Secondly, a connection on P(M,G) can be viewed as an equivariant g-valued 1-form with fixed values in the vertical direction, i.e. A∈Ω1P(g), satisfies
Apg(vg)=g−1Ap(v)g,∀v∈TpP,g∈GAp(pξ)=ξ,∀p∈P,ξ∈g.
It related with the horizontal distribution H as Hp=kerAp.
Lastly, in terms of an associated bundle gP, we can view a connection as a linear map
∇:Ω0M(gP)→Ω1M(gP),
satisfying the Leibnitz rule.
The set of smooth connections is denoted by A(P).
[1]: Atiyah, M. F., & Bott, R. (1983). The yang-mills equations over riemann surfaces. Philosophical Transactions of the Royal Society of London. Series A, Mathematical and Physical Sciences, 523-615.
[2]: Wehrheim K. Uhlenbeck compactness[M]. 2004.
发表回复