Canonical: https://knowlpedia.com/fiber-bundles/horizontal-distribution/

# Horizontal distribution

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

Let \(\pi:E\to M\) be a surjective submersion. For each \(e\in E\), write \(V_eE=\ker(d\pi_e)\subset T_eE\).

**Definition.** A horizontal distribution is a smooth assignment
\[
e \longmapsto H_eE \subset T_eE
\]
of a constant-rank subspace such that for every \(e\in E\),
\[
T_eE = H_eE \oplus V_eE.
\]
“Smooth” means that locally there exist smooth [vector fields](https://knowlpedia.com/fiber-bundles/vector-field/) on \(E\) whose values span \(H_eE\) at each point.

## Equivalent viewpoints

A horizontal distribution is the pointwise version of a [horizontal subbundle](https://knowlpedia.com/fiber-bundles/horizontal-subbundle/); the two viewpoints are equivalent. The distribution is called integrable precisely when it is involutive, in the sense of [integrability of horizontals](https://knowlpedia.com/fiber-bundles/integrable-horizontal-distribution/).

## Examples

1. **Constant horizontals on a product.** On \(M\times F\), taking \(H_{(x,f)}E=T_xM\oplus\{0\}\) defines a horizontal distribution.
2. **Transverse foliation.** If \(E\) is foliated by submanifolds that project locally diffeomorphically onto \(M\) (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 \(H_eE\) to be the horizontal subspace at \(e\).
