Horizontal distribution

A smooth choice of horizontal tangent subspaces complementing the vertical spaces in a fiber bundle.
Horizontal distribution

Let π:EM\pi:E\to M be a surjective submersion. For each eEe\in E, write VeE=ker(dπe)TeEV_eE=\ker(d\pi_e)\subset T_eE.

Definition. A horizontal distribution is a smooth assignment

eHeETeE e \longmapsto H_eE \subset T_eE

of a constant-rank subspace such that for every eEe\in E,

TeE=HeEVeE. T_eE = H_eE \oplus V_eE.

“Smooth” means that locally there exist smooth on EE whose values span HeEH_eE at each point.

A horizontal distribution is the pointwise version of a ; the two viewpoints are equivalent. The distribution is called integrable precisely when it is involutive, in the sense of .

Examples

  1. Constant horizontals on a product. On M×FM\times F, taking H(x,f)E=TxM{0}H_{(x,f)}E=T_xM\oplus\{0\} defines a horizontal distribution.
  2. Transverse foliation. If EE is foliated by submanifolds that project locally diffeomorphically onto MM (for example, graphs of local sections), then their tangent spaces define a horizontal distribution.
  3. From a connection. Any Ehresmann connection (or principal connection) determines a horizontal distribution by declaring HeEH_eE to be the horizontal subspace at ee.