ProofComplete

Proof of Schur's Lemma

Schur's lemma is the cornerstone of representation theory, controlling the structure of intertwining operators between irreducible representations. It underlies the orthogonality relations, the Wigner-Eckart theorem, and selection rules throughout physics.


Statement

Theorem8.3Schur's Lemma

Let ΁1:G→GL(V1)\rho_1: G \to \mathrm{GL}(V_1) and ΁2:G→GL(V2)\rho_2: G \to \mathrm{GL}(V_2) be irreducible representations of a group GG over C\mathbb{C}.

(Part I) If T:V1→V2T: V_1 \to V_2 is a linear map satisfying T΁1(g)=΁2(g)TT\rho_1(g) = \rho_2(g)T for all g∈Gg \in G (an intertwining operator), then either T=0T = 0 or TT is an isomorphism.

(Part II) If V1=V2=VV_1 = V_2 = V and ΁1=΁2=΁\rho_1 = \rho_2 = \rho, then any intertwining operator T:V→VT: V \to V satisfying T΁(g)=΁(g)TT\rho(g) = \rho(g)T for all g∈Gg \in G is a scalar multiple of the identity: T=ÎģIT = \lambda I for some Îģ∈C\lambda \in \mathbb{C}.


Proof

Proof

Part I.

Consider the kernel ker⁥T={v∈V1:Tv=0}\ker T = \{v \in V_1 : Tv = 0\}. If v∈ker⁥Tv \in \ker T, then T(΁1(g)v)=΁2(g)(Tv)=΁2(g)⋅0=0T(\rho_1(g)v) = \rho_2(g)(Tv) = \rho_2(g)\cdot 0 = 0, so ΁1(g)v∈ker⁥T\rho_1(g)v \in \ker T. Thus ker⁥T\ker T is an invariant subspace of V1V_1 under ΁1\rho_1.

Since ΁1\rho_1 is irreducible, the only invariant subspaces of V1V_1 are {0}\{0\} and V1V_1. Therefore either ker⁥T=V1\ker T = V_1 (i.e., T=0T = 0) or ker⁥T={0}\ker T = \{0\} (i.e., TT is injective).

Consider the image im T={Tv:v∈V1}⊂V2\mathrm{im}\,T = \{Tv : v \in V_1\} \subset V_2. If w=Tv∈im Tw = Tv \in \mathrm{im}\,T, then ΁2(g)w=΁2(g)Tv=T΁1(g)v∈im T\rho_2(g)w = \rho_2(g)Tv = T\rho_1(g)v \in \mathrm{im}\,T. Thus im T\mathrm{im}\,T is an invariant subspace of V2V_2 under ΁2\rho_2.

Since ΁2\rho_2 is irreducible: either im T={0}\mathrm{im}\,T = \{0\} (i.e., T=0T = 0) or im T=V2\mathrm{im}\,T = V_2 (i.e., TT is surjective).

Combining: if T≠0T \neq 0, then TT is both injective and surjective, hence an isomorphism V1≅V2V_1 \cong V_2.

Part II.

Since we work over C\mathbb{C} and VV is finite-dimensional, TT has at least one eigenvalue Îģ∈C\lambda \in \mathbb{C} (by the fundamental theorem of algebra applied to the characteristic polynomial det⁥(T−ÎģI)=0\det(T - \lambda I) = 0).

Consider S=T−ÎģIS = T - \lambda I. For any g∈Gg \in G: S΁(g)=(T−ÎģI)΁(g)=T΁(g)−Îģ΁(g)=΁(g)T−Îģ΁(g)=΁(g)(T−ÎģI)=΁(g)SS\rho(g) = (T - \lambda I)\rho(g) = T\rho(g) - \lambda\rho(g) = \rho(g)T - \lambda\rho(g) = \rho(g)(T-\lambda I) = \rho(g)S

So SS intertwines ΁\rho with itself. By Part I, either S=0S = 0 or SS is an isomorphism. But Îģ\lambda is an eigenvalue of TT, so ker⁥S=ker⁥(T−ÎģI)≠{0}\ker S = \ker(T-\lambda I) \neq \{0\}, which means SS is not an isomorphism. Therefore S=0S = 0, i.e., T=ÎģIT = \lambda I. □\square

■

ExampleConsequences of Schur's Lemma

Orthogonality of characters. For compact groups, Schur's lemma combined with integration over the group yields: âˆĢG·Îģ(g)â€žĪ‡Îŧ(g) dÎŧ(g)=δÎģÎŧ\int_G \overline{\chi_\lambda(g)}\chi_\mu(g)\,d\mu(g) = \delta_{\lambda\mu}. This makes the character table an effective tool for decomposing representations. Commutant algebra. For a representation ΁=⨁ini΁i\rho = \bigoplus_i n_i\rho_i, the algebra of all operators commuting with ΁(g)\rho(g) for all gg is ⨁iMni(C)\bigoplus_i M_{n_i}(\mathbb{C}) (block diagonal matrices of size nin_i on each isotypic component). If ΁\rho is irreducible, the commutant is just C⋅I\mathbb{C}\cdot I.

RemarkSchur's Lemma over the Reals and for Infinite Dimensions

Over R\mathbb{R}, Part II generalizes: TT belongs to an associative division algebra over R\mathbb{R}, which by Frobenius' theorem is R\mathbb{R}, C\mathbb{C}, or H\mathbb{H} (the quaternions). This trichotomy classifies real irreducible representations into real type (commutant R\mathbb{R}), complex type (commutant C\mathbb{C}), and quaternionic type (commutant H\mathbb{H}). Time-reversal symmetry in physics selects the type: Kramers degeneracy for half-integer spin arises from the quaternionic structure. For infinite-dimensional representations on Hilbert spaces, Part I holds for bounded intertwining operators between topologically irreducible representations, but Part II requires the additional assumption that TT is bounded.