Tangent and Cotangent Bundles - Main Theorem
Frobenius' theorem characterizes when a distribution of tangent spaces integrates to give a foliation. It provides the fundamental criterion for solving systems of partial differential equations geometrically.
Let be a -dimensional distribution on (a smooth assignment of -dimensional subspaces ). The following are equivalent:
- is involutive: if are vector fields tangent to , then is also tangent to
- is integrable: every point has a neighborhood with a chart such that the last coordinates are constant on each integral manifold
- Through each point passes a unique maximal connected integral manifold
This theorem bridges algebra (the Lie bracket condition) and geometry (existence of integral submanifolds). It's essential for understanding when local symmetries extend to global geometric structures.
The involutivity condition for all is algebraic and checkable, while integrability is a geometric property. Frobenius' theorem shows these are equivalent, providing a practical test for existence of foliations.
On with coordinates , consider the distribution spanned by
Then , which is not in the span of . Thus this distribution is not involutive and cannot be integrated.
A distribution is involutive if and only if for every covector field annihilating (meaning for all ), we have
This is equivalent to the exterior derivative also annihilating when restricted appropriately.
A foliation of dimension on is a decomposition of into disjoint connected submanifolds (called leaves) of dimension , such that locally the foliation looks like the decomposition of into slices.
The 2-torus admits a 1-dimensional foliation by curves for each slope . When is rational, the leaves are closed circles. When is irrational, each leaf is dense in .
Every finite-dimensional Lie algebra arises as the Lie algebra of vector fields generating a local Lie group action on some manifold.
This deep result shows that abstract Lie algebras have geometric realizations, connecting representation theory with differential geometry.