Let GG be a . Fix the normalized Haar measure dgdg on GG (so G1dg=1\int_G 1\,dg=1). Let (π,V)(\pi,V) and (σ,W)(\sigma,W) be finite-dimensional continuous unitary representations of GG (see ), with π,σ\pi,\sigma irreducible (see ). Choose orthonormal bases so that π(g)\pi(g) and σ(g)\sigma(g) have matrix entries πij(g)\pi_{ij}(g) and σkl(g)\sigma_{kl}(g).

Schur orthogonality asserts:

Gπij(g)σkl(g)dg  =  {1dimVδikδjl,if πσ,0,if π≄σ.\int_G \pi_{ij}(g)\,\overline{\sigma_{kl}(g)}\,dg \;=\; \begin{cases} \frac{1}{\dim V}\,\delta_{ik}\delta_{jl}, & \text{if }\pi\simeq \sigma,\\[6pt] 0, & \text{if }\pi\not\simeq \sigma. \end{cases}

This is the analytic avatar of Schur’s lemma and is one of the key inputs in the , which decomposes L2(G)L^2(G) into finite-dimensional isotypic pieces. In practice, Schur orthogonality is the tool that turns representation theory into concrete integral identities on compact groups (compare also ).

Equivalent characterizations

Equivalently, the matrix coefficients of irreducible unitary representations form an orthogonal family in L2(G)L^2(G), and within a fixed irreducible representation they are orthogonal with the explicit scale factor 1/dimV1/\dim V.