Transgression form

A differential form whose exterior derivative is the difference of two characteristic forms coming from different connections
Transgression form

Let π:PM\pi:P\to M be a with Lie algebra g\mathfrak g, and let ω0,ω1\omega_0,\omega_1 be two on PP with corresponding Ω0,Ω1Ω2(P;g)\Omega_0,\Omega_1\in\Omega^2(P;\mathfrak g).

Fix an Ad\operatorname{Ad}-invariant symmetric multilinear map

p:g××gn factorsR, p:\underbrace{\mathfrak g\times\cdots\times\mathfrak g}_{n\ \text{factors}}\longrightarrow \mathbb R,

i.e. an invariant polynomial datum as used to build a . Define the affine path of connections

ωt:=ω0+t(ω1ω0),t[0,1], \omega_t := \omega_0 + t(\omega_1-\omega_0),\qquad t\in[0,1],

and set η:=ω1ω0Ω1(P;g)\eta:=\omega_1-\omega_0\in\Omega^1(P;\mathfrak g). (By , η\eta is horizontal and Ad\operatorname{Ad}-equivariant.) Let Ωt\Omega_t be the curvature of ωt\omega_t.

Definition (transgression form)

The transgression form associated to pp and the pair (ω0,ω1)(\omega_0,\omega_1) is the (2n1)(2n-1)-form Tp(ω0,ω1)T_p(\omega_0,\omega_1) on MM uniquely characterized by the requirement that its pullback to PP is

πTp(ω0,ω1)  =  n01p ⁣(ηΩtn1)dt, \pi^*T_p(\omega_0,\omega_1) \;=\; n\int_0^1 p\!\big(\eta\wedge \Omega_t^{\,n-1}\big)\,dt,

where Ωtn1\Omega_t^{\,n-1} denotes the wedge product of (n1)(n-1) copies of Ωt\Omega_t and p(ηΩtn1)p(\eta\wedge \Omega_t^{n-1}) means the R\mathbb R-valued form obtained by feeding the g\mathfrak g-valued factors into pp.

Because the integrand is basic (horizontal and GG-invariant), Tp(ω0,ω1)T_p(\omega_0,\omega_1) is well-defined on the base. It is designed so that the difference of the corresponding Chern–Weil forms is exact:

dTp(ω0,ω1)=p(Ω1)p(Ω0)on M, d\,T_p(\omega_0,\omega_1)=p(\Omega_1)-p(\Omega_0) \quad\text{on }M,

which is the content of the . Specializing ω0\omega_0 to a reference connection produces the usual .

Examples

  1. Degree 1 (linear invariant polynomial).
    For n=1n=1 and an Ad\operatorname{Ad}-invariant linear functional p:gRp:\mathfrak g\to\mathbb R, the formula reduces to

    Tp(ω0,ω1)=p(ω1ω0),dTp=p(Ω1)p(Ω0). T_p(\omega_0,\omega_1)=p(\omega_1-\omega_0), \qquad dT_p = p(\Omega_1)-p(\Omega_0).
  2. Degree 2 on a trivial bundle (the usual Chern–Simons 3-form).
    On a trivial bundle and in a global gauge, a connection is represented by a g\mathfrak g-valued 1-form AA (see ). For p(X,Y)=tr(XY)p(X,Y)=\operatorname{tr}(XY) (degree 22), taking ω0=0\omega_0=0 gives the standard 3-form

    CS(A)=tr ⁣(AdA+23AAA), \operatorname{CS}(A)=\operatorname{tr}\!\Big(A\wedge dA+\frac{2}{3}A\wedge A\wedge A\Big),

    with dCS(A)=tr(FF)d\,\operatorname{CS}(A)=\operatorname{tr}(F\wedge F), where F=dA+AAF=dA+A\wedge A is the .

  3. Abelian case.
    If GG is abelian (e.g. U(1)U(1)), then Ad\operatorname{Ad} is trivial and AA=0A\wedge A=0. For a degree 1 invariant polynomial, the transgression between two U(1)U(1)-connections A0,A1A_0,A_1 is simply T(A0,A1)=A1A0T(A_0,A_1)=A_1-A_0, and dT=dA1dA0dT = dA_1-dA_0.