Vector Operators
The Clebsch-Gordan decomposition tells us how a tensor product of irreducibles breaks
apart. The Wigner-Eckart theorem is the payoff. It says that once the irreducible content
is known, the way one set of operators relates to another is forced to be a single number.
The objects on which this rigidity acts are vector operators. Each is a
triple of operators that transform under rotations the way the three coordinates of a
vector do. We begin by making that phrase precise.
How Rotations Act on Operators
Fix a representation \(\Pi\) of \(SO(3)\) on a finite-dimensional vector space \(V\), and
let \(\operatorname{End}(V)\) be the space of linear operators \(C : V \to V\). A rotation
\(R\) already acts on the vectors of \(V\) through \(\Pi(R)\). It acts on an operator
\(C\) by transporting the operator along with the space,
\[
R \cdot C = \Pi(R)\,C\,\Pi(R)^{-1}.
\]
The reading is concrete. To apply the rotated operator to a vector, we first undo the
rotation with \(\Pi(R)^{-1}\), apply \(C\) in the original frame, and then rotate the
result back with \(\Pi(R)\).
This is the same conjugation pattern by which a change of basis acts on a
matrix. A direct check using \(\Pi(R_1 R_2) = \Pi(R_1)\Pi(R_2)\) shows it is itself a
representation of \(SO(3)\), now on the space \(\operatorname{End}(V)\):
\[
R_1 \cdot (R_2 \cdot C) = \Pi(R_1)\Pi(R_2)\,C\,\Pi(R_2)^{-1}\Pi(R_1)^{-1}
= (R_1 R_2) \cdot C.
\]
The map \(R \mapsto (C \mapsto R \cdot C)\) is smooth, its entries being polynomials in
those of \(\Pi(R)\) and \(\Pi(R)^{-1}\), so the conjugation action is a representation
in the sense used throughout.
The Defining Equivariance
Alongside this action on operators sits the most familiar representation of \(SO(3)\) of
all: the rotation of ordinary vectors in \(\mathbb{R}^3\), where \(R\) sends
\(\mathbf{v}\) to \(R\mathbf{v}\). A vector operator is a triple
\(\mathbf{C} = (C_1, C_2, C_3)\) of operators on \(V\) whose three components rotate among
themselves in lockstep with the coordinate directions. To state this cleanly, assemble the
triple into a single operator depending linearly on a direction \(\mathbf{v} \in
\mathbb{R}^3\),
\[
\mathbf{v} \cdot \mathbf{C} = \sum_{j=1}^{3} v_j\, C_j.
\]
Definition: Vector Operator
Let \(V\) be a complex vector space, of any dimension, on which \(SO(3)\) acts by a
group homomorphism \(\Pi\) into the invertible linear operators on \(V\). When
\(V\) is nonzero and finite-dimensional and \(\Pi\) is smooth, this is a
representation
of \(SO(3)\). A triple \(\mathbf{C} = (C_1, C_2, C_3)\) of operators on \(V\) is a
vector operator if
\[
(R\mathbf{v}) \cdot \mathbf{C} = \Pi(R)\,(\mathbf{v} \cdot \mathbf{C})\,\Pi(R)^{-1}
\quad \text{for all } R \in SO(3) \text{ and all } \mathbf{v} \in \mathbb{R}^3.
\]
Unwinding the definition shows exactly what it demands. The right-hand side is the action
on operators applied to \(\mathbf{v} \cdot \mathbf{C}\). The left-hand side first rotates
the direction \(\mathbf{v}\) and then forms the combination. The equation says that rotating
the direction and rotating the operator are the same thing. When \(V\) is nonzero and
finite-dimensional and \(\Pi\) is smooth, in the language of the
intertwining maps
of representations, the linear map \(\mathbf{v} \mapsto \mathbf{v} \cdot \mathbf{C}\) from
\(\mathbb{R}^3\) into \(\operatorname{End}(V)\) intertwines the rotation action on
\(\mathbb{R}^3\) with the conjugation action on \(\operatorname{End}(V)\). A vector operator
is then precisely such an intertwiner. This single observation is what lets the representation
theory take over.
A concrete instance fixes the idea. Let \(V\) be the space of polynomial functions on
\(\mathbb{R}^3\), on which \(SO(3)\) acts by rotating the argument,
\((\Pi(R) f)(\mathbf{x}) = f(R^{-1}\mathbf{x})\), and let \(X_j\) be multiplication by
the \(j\)-th coordinate, \((X_j f)(\mathbf{x}) = x_j\, f(\mathbf{x})\). Both operations
carry polynomials to polynomials. The triple \(\mathbf{X} = (X_1, X_2, X_3)\), the
position operators of physics, is a vector operator. Rotating the coordinate that one
multiplies by is the same as conjugating the multiplication operator by the rotation.
Since \((R\mathbf{v}) \cdot \mathbf{x} = \mathbf{v} \cdot (R^{-1}\mathbf{x})\) for
\(R \in SO(3)\),
\[
\begin{align*}
\bigl(((R\mathbf{v}) \cdot \mathbf{X})\, f\bigr)(\mathbf{x})
&= ((R\mathbf{v}) \cdot \mathbf{x})\, f(\mathbf{x}) \\\\
&= (\mathbf{v} \cdot R^{-1}\mathbf{x})\, f(\mathbf{x}), \\\\
\bigl(\Pi(R)\,(\mathbf{v} \cdot \mathbf{X})\,\Pi(R)^{-1} f\bigr)(\mathbf{x})
&= \bigl((\mathbf{v} \cdot \mathbf{X})\,\Pi(R^{-1}) f\bigr)(R^{-1}\mathbf{x}) \\\\
&= (\mathbf{v} \cdot R^{-1}\mathbf{x})\, f(\mathbf{x}),
\end{align*}
\]
which is the defining identity above.
The Infinitesimal Form
Differentiating the defining identity at the identity rotation converts it into a bracket
condition on the Lie algebra \(\mathfrak{so}(3)\), often the more practical form. Let
\(V\) be finite-dimensional, let \(\pi\) be the
Lie algebra representation associated to \(\Pi\),
and take the
basis \(E_1, E_2, E_3\) of \(\mathfrak{so}(3)\)
with \([E_j, E_k] = \sum_{l} \varepsilon_{jkl} E_l\), where \(E_j\) is the hat map of
\(\mathbf{e}_j\), so that \(E_j \mathbf{e}_k = \mathbf{e}_j \times \mathbf{e}_k =
\sum_{l} \varepsilon_{jkl}\, \mathbf{e}_l\). The symbol \(\varepsilon_{jkl}\) is \(+1\)
on a cyclic permutation of \((1,2,3)\), \(-1\) on a non-cyclic permutation, and \(0\)
when an index repeats. Fix \(j\) and \(k\), and put \(R = e^{tE_j}\) and
\(\mathbf{v} = \mathbf{e}_k\) in the defining identity. The left side is
\((e^{tE_j}\mathbf{e}_k) \cdot \mathbf{C}\), with derivative
\((E_j \mathbf{e}_k) \cdot \mathbf{C} = \sum_{l} \varepsilon_{jkl}\, C_l\) at \(t = 0\).
The right side is \(\Pi(e^{tE_j})\, C_k\, \Pi(e^{tE_j})^{-1} = \Pi(e^{tE_j})\, C_k\, \Pi(e^{-tE_j})\),
whose two factors have derivatives \(\pi(E_j)\) and \(-\pi(E_j)\) at \(t = 0\), so
the product rule gives the derivative \(\pi(E_j)\, C_k - C_k\, \pi(E_j)\). Hence a vector operator satisfies
\[
[\pi(E_j), C_k] = \sum_{l} \varepsilon_{jkl}\, C_l.
\]
Conversely, because \(SO(3)\) is connected, this bracket condition implies the
defining identity. We state the converse without proof and do not use it.
The components \(C_k\) rotate under the infinitesimal generators exactly as the basis
vectors of \(\mathbb{R}^3\) rotate under the cross product. The same structure constants
\(\varepsilon_{jkl}\) appear on both sides, which is the algebraic shadow of the
correspondence between the cross product and the bracket on \(\mathfrak{so}(3)\).
Endomorphisms as a Representation
The proof of the Wigner-Eckart theorem turns the geometric statement about vector operators
into an algebraic one about a representation, and then lets Schur's lemma finish the job.
The bridge is an identification of the operator space \(\operatorname{End}(V)\) with a
tensor product, plus one special feature of the rotation group. Its irreducibles are their
own duals. We establish both here.
Operators as a Tensor Product
Let \(V\) be a finite-dimensional representation of \(SO(3)\) and \(V^*\) its
dual representation.
There is a natural way to build an operator on \(V\) out of a covector and a vector. Given
\(\varphi \in V^*\) and \(v \in V\), the rule
\[
w \longmapsto \varphi(w)\, v
\]
is a linear operator on \(V\). It reads off the \(\varphi\)-component of its input and
returns that multiple of \(v\).
This assignment \((\varphi, v) \mapsto \varphi(\,\cdot\,)v\)
is bilinear, so by the
universal property of the tensor product
it factors through a unique linear map
\[
\Psi : V^* \otimes V \longrightarrow \operatorname{End}(V),
\quad \Psi(\varphi \otimes v)(w) = \varphi(w)\, v.
\]
The universal property is stated for real spaces, but its proof uses only the vector
space operations and so applies verbatim over \(\mathbb{C}\).
Proposition (Endomorphisms as a Tensor Product)
Let \(V\) be a finite-dimensional representation of \(SO(3)\). The map \(\Psi\) above is
an isomorphism of representations
\[
\operatorname{End}(V) \cong V^* \otimes V,
\]
where \(SO(3)\) acts on \(\operatorname{End}(V)\) by conjugation,
\(R \cdot C = \Pi(R)\,C\,\Pi(R)^{-1}\), and on \(V^* \otimes V\) by the
tensor product representation
\(\Pi^* \otimes \Pi\).
Proof:
First, \(\Psi\) is a vector-space isomorphism. Pick a basis \(v_1, \dots, v_n\) of \(V\)
with dual basis \(\varphi_1, \dots, \varphi_n\) of \(V^*\), so that
\(\varphi_i(v_k) = \delta_{ik}\). Then \(\Psi(\varphi_i \otimes v_j)\) is the operator
sending \(v_k\) to \(\delta_{ik} v_j\). This is the elementary matrix \(E_{ji}\), with
a \(1\) in row \(j\) and column \(i\). As \((i, j)\) ranges over all pairs, these are
exactly the \(n^2\) elementary matrices, a basis of \(\operatorname{End}(V)\). So
\(\Psi\) carries a basis of \(V^* \otimes V\) to a basis of \(\operatorname{End}(V)\)
and is an isomorphism of vector spaces.
Next, \(\Psi\) intertwines the two actions. On the tensor product, a rotation acts by
\((\Pi^* \otimes \Pi)(R) = \Pi^*(R) \otimes \Pi(R)\) with \(\Pi^*(R) =
[\Pi(R^{-1})]^\top\). Applying \(\Psi\) to a rotated elementary tensor and evaluating
on \(w \in V\), we obtain
\[
\begin{align*}
\Psi\big((\Pi^*(R)\varphi) \otimes (\Pi(R) v)\big)(w)
&= (\Pi^*(R)\varphi)(w)\, \Pi(R) v \\\\
&= \varphi\big(\Pi(R)^{-1} w\big)\, \Pi(R) v \\\\
&= \Pi(R)\,\big[\varphi(\Pi(R)^{-1} w)\, v\big] \\\\
&= \Pi(R)\,\Psi(\varphi \otimes v)\big(\Pi(R)^{-1} w\big) \\\\
&= \big(\Pi(R)\,\Psi(\varphi \otimes v)\,\Pi(R)^{-1}\big)(w).
\end{align*}
\]
The second line uses the definition of the dual action, namely that
\((\Pi^*(R)\varphi)(w) = \varphi(\Pi(R)^{-1} w)\). The third uses linearity of
\(\Pi(R)\) to move the scalar \(\varphi(\Pi(R)^{-1} w)\) inside.
The result is the conjugation action on \(\operatorname{End}(V)\)
applied to \(\Psi(\varphi \otimes v)\). Thus \(\Psi\) intertwines, and being an
invertible intertwiner, it is an isomorphism of representations.
The Irreducibles Are Self-Dual
The second ingredient is special to the rotation group. We first record which
representations of \(SO(3)\) are irreducible, in the labelling of the previous pages.
There \(V_m\) denotes the irreducible representation of \(\mathfrak{sl}(2,\mathbb{C})\)
of dimension \(m + 1\), \(\sigma_m\) the
representation of \(\mathfrak{so}(3)\)
obtained from it through the isomorphism \(\mathfrak{su}(2) \cong \mathfrak{so}(3)\),
and \(\Sigma_m\), for even \(m\), the representation of \(SO(3)\) supplied by the
lifting criterion.
Proposition (Irreducible Representations of \(SO(3)\))
Every irreducible finite-dimensional complex representation \(V\) of \(SO(3)\) is
isomorphic to \(\Sigma_m\) for exactly one even integer \(m \geq 0\), namely
\(m = \dim V - 1\), and each \(\Sigma_m\) with \(m\) even is irreducible. In
particular two irreducible representations of \(SO(3)\) of the same dimension are
isomorphic, and the dimensions that occur are exactly the odd numbers
\(2\ell + 1\), \(\ell = 0, 1, 2, \dots\).
Proof:
Let \((\Pi, V)\) be an irreducible representation of \(SO(3)\) and \(\sigma\) the
associated representation
of \(\mathfrak{so}(3)\), so that \(\Pi(e^{X}) = e^{\sigma(X)}\). The group
\(SO(3)\) is
connected,
so \(\sigma\) is irreducible by the
group-algebra correspondence
(Point 1). The irreducible representations of \(\mathfrak{so}(3)\) are, up to
isomorphism, the \(\sigma_m\), one in each dimension \(m + 1\), so there is an
isomorphism \(T : V \to V_m\) with \(T\,\sigma(X) = \sigma_m(X)\,T\) for all
\(X \in \mathfrak{so}(3)\), and \(m = \dim V - 1\). Then
\(R \mapsto T\,\Pi(R)\,T^{-1}\) is a representation of \(SO(3)\) on \(V_m\)
satisfying \(T\,\Pi(e^{X})\,T^{-1} = T e^{\sigma(X)} T^{-1} = e^{\sigma_m(X)}\), by
property (e) of the matrix exponential
in any basis, so its associated representation is \(\sigma_m\). The
lifting criterion
excludes odd \(m\). Conversely, for even \(m\) the lifting criterion provides
\(\Sigma_m\) with associated representation \(\sigma_m\), which is irreducible, so
\(\Sigma_m\) is irreducible by the correspondence (Point 1), and
\(\dim \Sigma_m = m + 1\) is odd. Finally, if two irreducible representations of
\(SO(3)\) have the same dimension \(m + 1\), their associated representations are
both isomorphic to \(\sigma_m\), hence to each other, and the correspondence
(Point 2) makes the representations themselves isomorphic.
Intertwiners pass from the group to the Lie algebra. Let \(T : V \to W\) be an
intertwining map of two finite-dimensional representations \(\Pi\) and \(\Sigma\) of
\(SO(3)\). Differentiating \(T\,\Pi(e^{tX}) = \Sigma(e^{tX})\,T\) at \(t = 0\)
gives \(T\,\pi(X) = \sigma(X)\,T\) for the associated representations, and since
\(T\) is complex-linear the same identity holds for the
complex-linear extensions
of \(\pi\) and \(\sigma\) to \(\mathfrak{sl}(2,\mathbb{C})\), transported through
\(\mathfrak{su}(2) \cong \mathfrak{so}(3)\). The hypothesis of that extension,
\(\mathfrak{su}(2) \cap i\,\mathfrak{su}(2) = \{0\}\), holds because a matrix
that is both skew-Hermitian and Hermitian is zero. An isomorphism of
\(SO(3)\)-representations is therefore also an isomorphism of the corresponding
\(\mathfrak{sl}(2,\mathbb{C})\)-representations, and by the proposition an irreducible
representation of \(SO(3)\) of dimension \(m + 1\) is, as a representation of
\(\mathfrak{sl}(2,\mathbb{C})\), isomorphic to \(V_m\). This is the form in which the
Clebsch-Gordan decomposition of the previous page will be applied.
The three-dimensional case will matter most. The rotation of ordinary vectors,
extended to \(\mathbb{C}^3\) by letting \(R\) act on complex vectors by the same
matrix, is irreducible. Indeed, let \(W \subseteq \mathbb{C}^3\) be a nonzero
invariant subspace. It is invariant under each \(e^{tE_j}\), hence under
\(E_j\), the derivative at \(t = 0\), because \(W\) is closed. The generator \(E_3\)
has the eigenvectors \(\mathbf{e}_1 + i\mathbf{e}_2\), \(\mathbf{e}_3\), and
\(\mathbf{e}_1 - i\mathbf{e}_2\) with the distinct eigenvalues \(-i\), \(0\), and
\(i\). Write a nonzero \(w \in W\) as \(w = w_{+} + w_0 + w_{-}\) along these
eigenvectors. Then \((E_3^2 + I)\,w = w_0\), \(E_3(E_3 - iI)\,w = -2\,w_{+}\), and
\(E_3(E_3 + iI)\,w = -2\,w_{-}\) all lie in \(W\), so \(W\) contains one of the three
eigenvectors. If \(\mathbf{e}_3 \in W\), then \(E_2 \mathbf{e}_3 = \mathbf{e}_1\) and
\(E_1 \mathbf{e}_3 = -\mathbf{e}_2\) lie in \(W\), so \(W = \mathbb{C}^3\). If
\(\mathbf{e}_1 \pm i\mathbf{e}_2 \in W\), then
\(E_1(\mathbf{e}_1 \pm i\mathbf{e}_2) = \pm i\,\mathbf{e}_3 \in W\), and we are back
in the first case. By the proposition, \(\mathbb{C}^3\) is therefore isomorphic to
\(\Sigma_2\), and as a representation of \(\mathfrak{sl}(2,\mathbb{C})\) to \(V_2\).
Proposition (Self-Duality of the Irreducibles)
Every finite-dimensional irreducible representation \(V\) of \(SO(3)\) is isomorphic to
its dual: \(V \cong V^*\).
Proof:
The dual \(V^*\) has the same dimension as \(V\), and passing to the
dual representation
preserves irreducibility,
an invariant subspace of \(V^*\) being the annihilator of an invariant subspace of
\(V\), of complementary dimension. Thus \(V\) and \(V^*\) are irreducible
representations of \(SO(3)\) of the same dimension, hence isomorphic by the
classification of irreducibles
above.
Combining the two propositions gives the form in which the operator space enters the proof.
For an irreducible \(V\), self-duality lets us replace \(V^*\) by \(V\). An isomorphism
\(T : V^* \to V\) of representations gives the isomorphism
\(T \otimes I : V^* \otimes V \to V \otimes V\), which intertwines the two tensor
product representations. Hence
\[
\operatorname{End}(V) \cong V^* \otimes V \cong V \otimes V.
\]
The right-hand side is a tensor product of two irreducibles, and its irreducible content
is exactly what the
Clebsch-Gordan decomposition
computes. This is the lever. A question about operators has become a question about how the
standard three-dimensional representation sits inside \(V \otimes V\), which the previous
page already answered.
The Wigner-Eckart Theorem
Everything is now in place. On an irreducible representation, the operator space contains
at most one slot into which a vector operator can map, and Schur's lemma forces any two vector
operators to fill that slot in proportion. The result is a rigidity statement of unusual
economy. A single complex number relates one vector operator to another.
Theorem (Wigner-Eckart, Irreducible Case)
Let \((\Pi, V)\) be an irreducible finite-dimensional complex representation of
\(SO(3)\), and let \(\mathbf{A}\) and \(\mathbf{B}\) be two vector operators on
\(V\), with
\(\mathbf{A}\) nonzero. Then there is a constant \(c \in \mathbb{C}\) such that
\[
\mathbf{B} = c\, \mathbf{A}.
\]
Before the proof, it is worth saying what the statement buys. A vector operator on an
\(n\)-dimensional space is, on its face, three \(n \times n\) matrices, a great many
numbers. The theorem says that once one nonzero vector operator \(\mathbf{A}\) is
known on an irreducible \(V\), every other vector operator on \(V\) is a scalar multiple of
it. All the apparent freedom collapses to one number.
Proof
Proof:
Recall that, since \(V\) is a finite-dimensional representation, a vector
operator is an
intertwining map
\[
\phi_{\mathbf{A}} : \mathbb{R}^3 \longrightarrow \operatorname{End}(V),
\quad \phi_{\mathbf{A}}(\mathbf{v}) = \mathbf{v} \cdot \mathbf{A},
\]
from the rotation action on \(\mathbb{R}^3\) to the conjugation action on
\(\operatorname{End}(V)\), and likewise \(\phi_{\mathbf{B}}\) for \(\mathbf{B}\). Both
extend complex-linearly to intertwiners out of the complexification \(\mathbb{C}^3\),
which carries the three-dimensional irreducible representation \(\Sigma_2\) of
\(SO(3)\), as shown in the previous section. By the same section,
\(\phi_{\mathbf{A}}\) and \(\phi_{\mathbf{B}}\) are then intertwiners of the
associated representations of \(\mathfrak{sl}(2,\mathbb{C})\), and it is there that
we compare them. As representations of \(\mathfrak{sl}(2,\mathbb{C})\),
\(\mathbb{C}^3 \cong V_2\) and \(V \cong V_m\) with \(m + 1 = \dim V\), and the
representation associated to \(\Pi \otimes \Pi\) is \(\pi \otimes \pi\), the
outer-product formula
evaluated at \((X, X)\). Tensoring an isomorphism \(V \to V_m\) with itself, and
using \(\operatorname{End}(V) \cong V^* \otimes V \cong V \otimes V\) from the previous
section, the last step by the
self-duality of the irreducible \(V\),
we obtain \(\operatorname{End}(V) \cong V_m \otimes V_m\) as representations of
\(\mathfrak{sl}(2,\mathbb{C})\).
The
Clebsch-Gordan decomposition
writes \(V_m \otimes V_m\) as \(V_{2m} \oplus V_{2m-2} \oplus \cdots \oplus V_0\), a
direct sum of irreducible invariant subspaces \(N_1, \dots, N_r\) in which \(V_2\)
occurs at most once. It occurs exactly once when \(m \geq 1\), and not at all
when \(m = 0\), that is, when \(V\) is one-dimensional. This single occurrence is the
multiplicity-freeness established on the previous page, read off the decomposition.
Because each \(N_k\) is invariant, the projection
\(p_k : \operatorname{End}(V) \to N_k\) along the other summands is an intertwiner,
so \(p_k \circ \phi_{\mathbf{A}} : \mathbb{C}^3 \to N_k\) is an intertwiner between
irreducibles. By the
first part of Schur's lemma
it is zero unless \(N_k \cong V_2\), and the same holds for \(\phi_{\mathbf{B}}\).
Suppose first that \(V\) is one-dimensional. Then no \(N_k\) is isomorphic to
\(V_2\), so \(p_k \circ \phi_{\mathbf{A}} = 0\) for every \(k\), hence
\(\phi_{\mathbf{A}} = 0\), contradicting that \(\mathbf{A}\) is nonzero. Hence
\(m \geq 1\), and exactly one summand, call it \(U\), is isomorphic to \(V_2\).
The same projections show that \(\phi_{\mathbf{A}}\) and \(\phi_{\mathbf{B}}\) land in
\(U\), since every \(p_k\) with \(N_k \neq U\) annihilates both. Because
\(\mathbf{A}\) is nonzero, \(\phi_{\mathbf{A}}\) is a nonzero intertwiner
\(\mathbb{C}^3 \to U\) between isomorphic irreducibles. If \(\mathbf{B} = 0\) the
conclusion holds with \(c = 0\), so assume \(\phi_{\mathbf{B}}\) is nonzero as well.
Applying
the part of Schur's lemma comparing two intertwiners
to the pair \(\phi_{\mathbf{B}}, \phi_{\mathbf{A}} : \mathbb{C}^3 \to U\), we obtain a
constant \(c \in \mathbb{C}\) with \(\phi_{\mathbf{B}} = c\, \phi_{\mathbf{A}}\).
Evaluating on \(\mathbf{v} \in \mathbb{R}^3\) gives \(\mathbf{v} \cdot \mathbf{B} =
c\,(\mathbf{v} \cdot \mathbf{A})\) for every direction, which is precisely
\(\mathbf{B} = c\, \mathbf{A}\).
One Number for the Whole Operator
The force of the theorem is that the proportionality constant \(c\) is the only datum
not fixed by symmetry. Two vector operators on an irreducible representation, however
different the quantities they are built from, must be proportional. Once one of them
is nonzero, the other is a scalar multiple of it. The geometry of rotation has used up
every other degree of freedom. The next section extends this from a single
irreducible to a larger space, where the same mechanism reappears as a rule about the
matrix entries of operators taken between two of its irreducible pieces.
Matrix Elements and Equivariance
The irreducible case assumed the whole space \(V\) was irreducible. In applications the
interesting operators act on a large, reducible space and we care about how they connect
two irreducible pieces of it. The general theorem handles this, and the conclusion is the
one that organizes computation. Between any two fixed irreducible subspaces, the entries of
every vector operator are the same up to a single constant.
From Endomorphisms to Homomorphisms
The bridge from the previous section generalizes from operators on one space to
homomorphisms between two. If \(W_1\) and \(W_2\) are finite-dimensional representations of
\(SO(3)\), the space \(\operatorname{Hom}(W_2, W_1)\) of linear maps \(W_2 \to W_1\)
carries the conjugation-type action \(C \mapsto \Pi_1(R)\, C\, \Pi_2(R)^{-1}\), where
\(\Pi_i\) is the action on \(W_i\). The same construction as before identifies it with a
tensor product.
Proposition (Homomorphisms as a Tensor Product)
Let \(W_1\) and \(W_2\) be finite-dimensional representations of \(SO(3)\). Then there
is an isomorphism of representations
\[
\operatorname{Hom}(W_2, W_1) \cong W_2^* \otimes W_1.
\]
If in addition \(W_2\) is irreducible, hence
self-dual, this
becomes \(\operatorname{Hom}(W_2, W_1) \cong W_2 \otimes W_1\).
The proof repeats that of the endomorphism case verbatim, with the rule
\(\Psi(\varphi \otimes w_1)(w_2) = \varphi(w_2)\, w_1\) for \(\varphi \in W_2^*\) and
\(w_1 \in W_1\). Taking \(W_1 = W_2 = V\) recovers
the endomorphism isomorphism.
The General Theorem
Theorem (Wigner-Eckart)
Let \(V\) be a complex inner product space, possibly infinite-dimensional, and let
\(\Pi\) be a group homomorphism from \(SO(3)\) to the invertible linear operators on
\(V\) that preserve the inner product. Let \(W_1\) and \(W_2\) be
finite-dimensional invariant subspaces of \(V\) on which \(\Pi\) restricts to
irreducible representations, and let
\(\mathbf{A}\) and \(\mathbf{B}\) be vector operators on \(V\) for which
\(\langle w, A_j w' \rangle\) is nonzero for some \(w \in W_1\), \(w' \in W_2\), and
\(j \in \{1,2,3\}\). Then there is a constant \(c \in \mathbb{C}\) such that
\[
\langle w, B_j w' \rangle = c\, \langle w, A_j w' \rangle
\]
for all \(w \in W_1\), all \(w' \in W_2\), and all \(j = 1, 2, 3\).
Proof:
The operators \(A_j\) need not carry \(W_2\) into \(W_1\), but taking the inner product
of \(A_j w'\) against an element of \(W_1\) records only the \(W_1\)-component of
\(A_j w'\). Let \(P_1 : V \to W_1\) be the orthogonal projection onto \(W_1\). It exists
because \(W_1\) is finite-dimensional. Expanding along an orthonormal basis of
\(W_1\), obtained from any basis by the Gram-Schmidt process, produces for each
\(x \in V\) a vector \(P_1 x \in W_1\) with \(x - P_1 x\) orthogonal to \(W_1\).
Since \(W_1\) is invariant and \(\Pi\) preserves the inner product, the orthogonal
complement \(W_1^\perp\) is invariant too (for \(u \perp W_1\) and \(w \in W_1\),
\(\langle \Pi(R) u, w \rangle = \langle u, \Pi(R)^{-1} w \rangle = 0\)), so \(P_1\)
commutes with every \(\Pi(R)\). Define
\[
\phi_{\mathbf{A}} : \mathbb{R}^3 \longrightarrow \operatorname{Hom}(W_2, W_1),
\quad \phi_{\mathbf{A}}(\mathbf{v})(w') = P_1\big((\mathbf{v} \cdot \mathbf{A})\, w'\big),
\]
and likewise \(\phi_{\mathbf{B}}\). Because \(\mathbf{A}\) is a vector operator and
\(P_1\) commutes with the action, for \(w' \in W_2\)
\[
\begin{align*}
\phi_{\mathbf{A}}(R\mathbf{v})(w')
&= P_1\big(\Pi(R)\,(\mathbf{v} \cdot \mathbf{A})\,\Pi(R)^{-1} w'\big) \\\\
&= \Pi(R)\,\phi_{\mathbf{A}}(\mathbf{v})\big(\Pi(R)^{-1} w'\big),
\end{align*}
\]
so \(\phi_{\mathbf{A}}\), extended complex-linearly as before, is an
intertwining map
from \(\mathbb{C}^3\) into \(\operatorname{Hom}(W_2, W_1)\).
Now \(\operatorname{Hom}(W_2, W_1) \cong W_2 \otimes W_1\) by the proposition above, and
the same passage to \(\mathfrak{sl}(2,\mathbb{C})\) as in the irreducible case
identifies \(W_2 \otimes W_1\) with \(V_a \otimes V_b\), where \(a + 1 = \dim W_2\)
and \(b + 1 = \dim W_1\). The
Clebsch-Gordan decomposition
of \(V_a \otimes V_b\) contains \(V_2\) at most once, its summands having distinct
dimensions. Projecting onto the summands as before, \(\phi_{\mathbf{A}}\) and
\(\phi_{\mathbf{B}}\) vanish on every summand not isomorphic to \(V_2\) by the
first part of Schur's lemma.
If \(V_2\) does not occur, both are zero, and all the inner products
\(\langle w, A_j w'\rangle\) would then be zero, against the hypothesis. So \(V_2\)
occurs exactly once, and both \(\phi_{\mathbf{A}}\) and \(\phi_{\mathbf{B}}\) map into
that single copy. If \(\phi_{\mathbf{B}} = 0\) the conclusion holds with \(c = 0\), so
assume it is nonzero. Since \(\phi_{\mathbf{A}}\) is nonzero as well,
comparison of two intertwiners
gives a constant \(c\) with \(\phi_{\mathbf{B}} = c\, \phi_{\mathbf{A}}\).
Finally, the orthogonal projection \(P_1\) is self-adjoint and acts as the identity on
\(W_1\), so for \(w \in W_1\),
\[
\langle w,\, (\mathbf{v} \cdot \mathbf{A})\, w' \rangle
= \langle w,\, P_1 (\mathbf{v} \cdot \mathbf{A})\, w' \rangle
= \langle w,\, \phi_{\mathbf{A}}(\mathbf{v})\, w' \rangle,
\]
and the same with \(\mathbf{A}\) replaced by \(\mathbf{B}\). Therefore
\(\langle w, (\mathbf{v} \cdot \mathbf{B}) w' \rangle = c\, \langle w, (\mathbf{v} \cdot
\mathbf{A}) w' \rangle\) for every \(\mathbf{v}\). Taking \(\mathbf{v} = \mathbf{e}_j\) yields the
claim.
Reduced Matrix Elements and Weight Sharing
The constant \(c\) is the ratio of what physics calls the reduced matrix
elements of \(\mathbf{B}\) and \(\mathbf{A}\). Read as a
statement about computation, the theorem describes the full array of numbers
\(\langle w, A_j w' \rangle\) as \(w\), \(w'\), and \(j\) vary over a pair of
irreducible subspaces. That array is determined by one scalar once the geometry-fixed
part is known. The dependence on the indices is universal, the same for every vector operator
between that pair of irreducibles. Only the overall scale carries information about
the specific operator.
This is the principle a rotation-equivariant network exploits when it couples features.
A layer mapping type-\(\ell_2\) features to type-\(\ell_1\) features through a vector
quantity has, a priori, many weights. Equivariance forces those weights to follow the
fixed index pattern and leaves at most a single free parameter for each pair of an
input type-\(\ell_2\) piece and an output type-\(\ell_1\) piece, and none unless
\(|\ell_1 - \ell_2| \leq 1 \leq \ell_1 + \ell_2\). The reduction from a full weight
matrix to at most one learnable scalar per pair of irreducible pieces is
the parameter economy usually credited with the sample efficiency of equivariant
models. Schur's lemma does the reducing, not an architectural choice.
The setting also accommodates the infinite-dimensional spaces of analysis. When \(V\) is a
space of square-integrable functions on \(\mathbb{R}^3\) under the rotation action, its
decomposition into finite-dimensional irreducible subspaces is the harmonic content of the
sphere. The matrix entries of any vector operator defined on the whole space, taken
between two such pieces, are again fixed up to a single constant factor. Pursuing that
decomposition leads to the theory of harmonic analysis on the rotation group, where the
irreducibles classified here reappear as the building blocks of every square-integrable
function.