Pontryagin class via Chern–Weil theory

Characteristic cohomology classes of a real vector bundle defined from curvature, using the complexification in Chern–Weil theory.
Pontryagin class via Chern–Weil theory

Let MM be a and let π:EM\pi:E\to M be a real vector bundle of rank rr. Choose a bundle metric on EE and a compatible \nabla (so the structure group reduces to O(r)O(r)). Let FΩ2(M;so(E))F_\nabla\in\Omega^2(M;\mathfrak{so}(E)) be its .

Definition (Pontryagin forms and Pontryagin classes)

Let EC:=ERCE^{\mathbb C}:=E\otimes_{\mathbb R}\mathbb C be the complexification, and let C\nabla^{\mathbb C} be the induced complex connection. Define the Pontryagin forms by

pk()  :=  (1)kc2k(C)Ω4k(M), p_k(\nabla)\;:=\;(-1)^k\,c_{2k}(\nabla^{\mathbb C})\in \Omega^{4k}(M),

where c2k(C)c_{2k}(\nabla^{\mathbb C}) is the (2k)(2k)th Chern form of the complex bundle ECE^{\mathbb C} (defined as in ).

Then:

  1. Each pk()p_k(\nabla) is closed: dpk()=0d\,p_k(\nabla)=0, where dd is the .
  2. The de Rham class [pk()]HdR4k(M)[p_k(\nabla)]\in H^{4k}_{\mathrm{dR}}(M) is independent of the choice of compatible connection.
  3. The kkth Pontryagin class pk(E)H4k(M;Z)p_k(E)\in H^{4k}(M;\mathbb Z) is the unique integral class whose real image equals [pk()][p_k(\nabla)].

Equivalently, pk(E)p_k(E) is the Chern–Weil class associated to the structure group O(r)O(r) (or SO(r)SO(r) in the oriented case) by applying an AdAd-invariant polynomial on so(r)\mathfrak{so}(r) corresponding to the kkth elementary symmetric polynomial in the squares of the formal curvature eigenvalues.

Naturality holds: for any f:NMf:N\to M,

pk(fE)=fpk(E). p_k(f^*E)=f^*p_k(E).

Examples

  1. Trivial bundle / flat connection. If EM×RrE\cong M\times\mathbb R^r with the flat connection, then F=0F_\nabla=0, hence pk()=0p_k(\nabla)=0 for all k1k\ge 1, and thus pk(E)=0p_k(E)=0.

  2. Underlying real bundle of a complex line bundle. Let LML\to M be a complex line bundle with c1(L)=xH2(M;Z)c_1(L)=x\in H^2(M;\mathbb Z). For the underlying real rank-2 bundle LRL_{\mathbb R} one has

    p1(LR)=x2H4(M;Z), p_1(L_{\mathbb R})=x^2\in H^4(M;\mathbb Z),

    because p1=(1)c2((LR)C)p_1=(-1)c_2\big((L_{\mathbb R})^{\mathbb C}\big) and (LR)CLL(L_{\mathbb R})^{\mathbb C}\cong L\oplus \overline{L}.

  3. Dimensional vanishing. If dimM<4k\dim M < 4k, then every 4k4k-form vanishes and hence pk(E)=0p_k(E)=0 in de Rham cohomology (and therefore in rational cohomology) for degree reasons.