Invariant Inner Products
The existence of invariant inner products is crucial for proving complete reducibility and understanding unitary representations.
Let be a representation of over or . An inner product on is -invariant if: for all and .
Equivalently, each is a unitary (or orthogonal) transformation with respect to the inner product.
Let be a finite-dimensional representation of a finite group over or . Then there exists a -invariant inner product on .
Proof: Start with any inner product on . Define:
This is positive definite (since is) and -invariant by construction.
Complete Reducibility: With a -invariant inner product, if is a subrepresentation, then the orthogonal complement is also a subrepresentation. Thus decomposes.
Unitary Representations: A representation with a -invariant inner product is called unitary (over ) or orthogonal (over ). All finite-dimensional representations of finite groups and compact groups are unitarizable.
Matrix Form: If we choose an orthonormal basis with respect to a -invariant inner product, then each is represented by a unitary matrix: .
A representation over is unitary if there exists an inner product on such that: for all , where denotes the adjoint with respect to the inner product.
For compact groups (including finite groups), every finite-dimensional representation is equivalent to a unitary representation. This is proved using Haar measure to average an arbitrary inner product. The unitary perspective provides powerful tools: orthogonality of matrix coefficients, Peter-Weyl theorem, and connections to harmonic analysis.
For non-compact groups, not all representations are unitarizable. Understanding which representations admit invariant inner products is a deep question in harmonic analysis and the theory of automorphic forms.
The existence of invariant inner products dramatically simplifies the structure theory. With a fixed invariant inner product, we can use orthogonal projections, compute direct sum decompositions explicitly, and apply tools from functional analysis and spectral theory.