Reducing a GL(n)-structure to O(n) using a bundle metric

A fiberwise inner product reduces the structure group of a frame bundle from GL(n) to O(n).
Reducing a GL(n)-structure to O(n) using a bundle metric

Let EME\to M be a smooth real rank-nn vector bundle over a smooth manifold MM. Its frame bundle Fr(E)M\mathrm{Fr}(E)\to M is a principal GL(n,R)\mathrm{GL}(n,\mathbb R)-bundle.

A reduction of structure group from GL(n,R)\mathrm{GL}(n,\mathbb R) to a subgroup HGL(n,R)H\subset \mathrm{GL}(n,\mathbb R) is, by definition, a principal HH-subbundle QFr(E)Q\subset \mathrm{Fr}(E) such that

Q×HRnE Q\times_H \mathbb R^n \cong E

as associated bundles.

Example (bundle metric gives an O(n)-reduction)

A bundle metric on EE is a smooth choice of inner product ,x\langle\cdot,\cdot\rangle_x on each fiber ExE_x.

Given such a metric, define

O(E):={uFr(E):u:RnEx is an isometry for the standard inner product}. O(E):=\{u\in \mathrm{Fr}(E): u:\mathbb R^n\to E_x \text{ is an isometry for the standard inner product}\}.

Then:

  • O(E)MO(E)\to M is a principal O(n)\mathrm{O}(n)-bundle, where O(n)\mathrm{O}(n) is a acting by postcomposition on frames.
  • The inclusion O(E)Fr(E)O(E)\subset \mathrm{Fr}(E) is a reduction of the principal Fr(E)M\mathrm{Fr}(E)\to M from GL(n)\mathrm{GL}(n) to O(n)\mathrm{O}(n).
  • Conversely, any principal O(n)\mathrm{O}(n)-subbundle QFr(E)Q\subset \mathrm{Fr}(E) determines a unique bundle metric by declaring frames in QQ to be orthonormal.

Specializing to E=TME=TM recovers the construction of the .

Examples

  1. Riemannian manifolds.
    For E=TME=TM, choosing a Riemannian metric on MM produces the reduction Fr(TM)O(TM)\mathrm{Fr}(TM)\supset O(TM).

  2. Trivial bundle with standard metric.
    For the trivial bundle M×RnM\times\mathbb R^n, the standard Euclidean inner product yields O(E)M×O(n)O(E)\cong M\times \mathrm{O}(n) as a reduction of M×GL(n)M\times \mathrm{GL}(n).

  3. Changing the metric changes the reduction.
    Two different bundle metrics on the same underlying bundle generally produce different O(n)\mathrm{O}(n)-subbundles of Fr(E)\mathrm{Fr}(E), even though both reductions have the same associated bundle EE.