Particles as Poincaré Representations

Contents
  1. Symmetries act unitarily: Wigner's theorem
  2. The unitary generators of the Poincaré group
  3. The universal cover $\SL(2,\C)$
  4. The invariants: mass and spin
  5. Orbits, little groups, and induced representations
  6. Massive particles: spin
  7. Massless particles: helicity
  8. The particle dictionary

What is a particle? Every earlier part of this treatise uses the word operationally — a silver atom in a Stern–Gerlach magnet (Experiment: Stern–Gerlach), a photoelectron, a positron track — but relativistic quantum mechanics is where the word acquires a definition. The answer, found by Wigner in 1939 [Wigner:1939], is that a particle species is an irreducible unitary representation of the Poincaré group: relativity supplies the symmetry group of spacetime (Section 39.3), quantum mechanics demands that symmetries act unitarily on rays (Postulate 77.1), and irreducibility is the statement that the system has no smaller relativistically invariant part — that it is elementary. The two labels that classify these representations, mass and spin, are exactly the two numbers experiment attaches to every particle ever observed. This chapter carries out Wigner's programme in full; the standard modern accounts are [Weinberg:1995] [Sakurai:2017].

Throughout, the metric signature is \((+,-,-,-)\) with \(\eta = \operatorname{diag}(1,-1,-1,-1)\) as in Minkowski Space and Its Symmetries, Greek indices run over \(0,1,2,3\) and Latin ones over \(1,2,3\), the totally antisymmetric symbol is fixed by \(\epsilon_{0123} = +1\), and \(\hbar\) and \(c\) are kept explicit (Section 77.3).

Remark 94.1 (Level of rigor).

The generators used below are unbounded operators, defined on dense domains by Stone's theorem for the one-parameter unitary subgroups they generate; commutators are understood on the common invariant dense domain of smooth vectors of the representation. This chapter manipulates them algebraically, as the physics literature does [Weinberg:1995]; the functional-analytic scaffolding (domains, essential self-adjointness, integrability of the algebra to the group) is classical representation theory and is not restated here. Two standing assumptions are recorded once: every representation in this chapter is strongly continuous — \((\Lambda,a) \mapsto U(\Lambda,a)\ket{\psi}\) is continuous for each state, which is what Stone's theorem consumes — and Schur's lemma, cited from the finite-dimensional Theorem 5.153, is used in its unitary infinite-dimensional (spectral-projection) form [Weinberg:1995].

Symmetries act unitarily: Wigner's theorem

Quantum states are unit rays, not vectors: \(\ket{\psi}\) and \(\ee^{\ii\alpha}\ket{\psi}\) describe the same physical state (Postulate 77.1), and every measurable prediction is built from transition probabilities \(\abs{\braket{\varphi}{\psi}}^{2}\) (Postulate 77.4). A symmetry operation must therefore be, in the first instance, a map of rays that preserves those probabilities.

Definition 94.2 (Symmetry transformation).

Let \(\mathcal{H}\) be the Hilbert space of a quantum system and write \([\psi]\) for the unit ray of a normalized \(\ket{\psi}\). A symmetry transformation is a bijection \(T\) of the set of unit rays such that

\begin{equation}\tag{94.1} \abs{\braket{\varphi'}{\psi'}}^{2} = \abs{\braket{\varphi}{\psi}}^{2} \qquad\text{whenever } [\psi'] = T[\psi],\ [\varphi'] = T[\varphi]\ep \end{equation}
Theorem 94.3 (Wigner).

Let \(\dim\mathcal{H} \ge 2\). Every symmetry transformation \(T\) is implemented by an operator \(U\) on \(\mathcal{H}\) — \(T[\psi] = [U\psi]\) for every unit ray — which is either unitary and linear, or antiunitary and antilinear. \(U\) is unique up to an overall phase, and the two cases are mutually exclusive.

Derives Theorem 94.3.

The theorem is due to Wigner [Wigner:1931]; the complete proof given in Wigner's Theorem on Quantum Symmetries follows Bargmann [Bargmann:1964].

Corollary 94.4 (Connected symmetry groups act unitarily).

Let \(G\) be a connected Lie group acting by symmetry transformations \(T_g\), and let \(U(g)\) implement \(T_g\) as in Theorem 94.3. Then every \(U(g)\) is unitary.

Proof.

Ray composition gives \(U(g)U(h) = \omega(g,h)\,U(gh)\) for a phase \(\omega(g,h)\), since both sides implement \(T_{gh}\) and Theorem 94.3 fixes the implementation up to a phase. In a connected Lie group every element is a finite product of exponentials, \(g = \exp X_1 \cdots \exp X_n\): the image of the exponential map contains a neighbourhood of the identity, and the subgroup generated by any neighbourhood of the identity is open, hence closed — its complement is a union of its own open cosets — hence all of the connected group. Each factor is a square: \(\exp X = \left(\exp \tfrac{X}{2}\right)^{2}\). Hence \(U(g)\) equals, up to a phase, a product of operators \(U(h_i)^{2}\). The square of a unitary operator is unitary, and so is the square of an antiunitary one (two antilinear maps compose to a linear map, and isometry is preserved); phases do not affect (anti)unitarity. A product of unitaries is unitary.

The phases \(\omega(g,h)\) cannot in general be removed by rephasing the \(U(g)\): one speaks of a projective unitary representation. For the Poincaré group the obstruction was computed by Bargmann [Bargmann:1954]:

Theorem 94.5 (Bargmann).

Every continuous projective unitary representation of the proper orthochronous Poincaré group arises, after a choice of phases, from an ordinary strongly continuous unitary representation of its universal covering group \(\R^{4} \rtimes \SL(2,\C)\).

Derivation pending.

Bargmann's theorem for the Poincaré group: the proof computes the second cohomology of the Lie algebra \(\mathfrak{iso}(3,1)\) and shows every local phase cocycle is a coboundary, then lifts to the simply connected cover; see [Bargmann:1954]. The cohomological machinery is not yet developed in this treatise.

Remark 94.6 (The cover is physical).

Theorem 94.5 is why the classification below runs over representations of the cover — in which a rotation by \(2\pi\) need not be the identity, only its square is — and this is not a mathematical nicety but an observed fact: a neutron interferometer splits a beam, rotates the spin of one arm through \(2\pi\) with a magnetic field, and finds the recombined intensity reversed, returning to itself only after \(4\pi\), exactly as the sign \(U(2\pi) = -\identity\) of a spin-\(\tfrac12\) representation demands [Werner:1975]. By contrast, for the Galilei group of Part III — Classical Mechanics the obstruction does not vanish: its central charge is the mass itself [Bargmann:1954], which is why nonrelativistic quantum mechanics superselects mass while relativistic quantum mechanics treats it as a spectrum label.

The unitary generators of the Poincaré group

Proposition 94.7 (Composition law).

The transformations \(x \mapsto \Lambda x + a\) of Minkowski space, with \(\Lambda\) in the proper orthochronous Lorentz group \(L^{\uparrow}_{+}\) (Definition 39.5; Proposition 39.6 makes it a group) and \(a \in \R^4\), form a group with

\begin{equation}\tag{94.2} (\Lambda', a')\,(\Lambda, a) = (\Lambda'\Lambda,\ a' + \Lambda' a)\ep \end{equation}
Proof.

\(x \mapsto \Lambda x + a \mapsto \Lambda'(\Lambda x + a) + a' = (\Lambda'\Lambda)x + (\Lambda' a + a')\); the identity is \((\identity,0)\) and \((\Lambda,a)^{-1} = (\Lambda^{-1}, -\Lambda^{-1}a)\).

This is the Poincaré group of Section 39.3, restricted to its identity component. By Corollary 94.4 a relativistic quantum system carries unitary operators \(U(\Lambda, a)\), projectively composing by Equation (94.2); by Theorem 94.5 we may and do take a true representation of the cover. Stone's theorem then provides Hermitian generators: for translations and for Lorentz transformations respectively,

\begin{equation}\tag{94.3} U(\identity, a) = \exp\left(-\frac{\ii}{\hbar}\,a^{\mu}\hat{P}_{\mu}\right)\ec \qquad U(\Lambda(\omega), 0) = \exp\left(-\frac{\ii}{2\hbar}\,\omega_{\mu\nu}\hat{J}^{\mu\nu}\right)\ec \end{equation}

with \(\omega_{\mu\nu} = -\omega_{\nu\mu}\) the parameters of Section 39.2.7 and \(\hat{J}^{\mu\nu} = -\hat{J}^{\nu\mu}\). The factor \(\hbar\) makes \(\hat{P}_{\mu}\) a momentum and \(\hat{J}^{\mu\nu}\) an angular momentum; the physical dictionary is

\begin{equation}\tag{94.4} \hat{H} = c\,\hat{P}^{0}\ec\qquad \hat{J}_{i} = \tfrac{1}{2}\,\epsilon_{ijk}\hat{J}^{jk}\ec\qquad \hat{K}_{i} = \hat{J}^{i0}\ec \end{equation}

energy, angular momentum and boost generator (cf. the nonrelativistic operators of Section 77.5).

Proposition 94.8 (The Poincaré algebra of the generators).

On the common dense domain of Remark 94.1,

\begin{align} \comm{\hat{J}^{\mu\nu}}{\hat{J}^{\lambda\rho}} &= \ii\hbar\left( \eta^{\mu\lambda}\hat{J}^{\rho\nu} + \eta^{\nu\rho}\hat{J}^{\lambda\mu} + \eta^{\mu\rho}\hat{J}^{\nu\lambda} + \eta^{\nu\lambda}\hat{J}^{\mu\rho} \right)\ec\tag{94.5}\\ \comm{\hat{J}^{\mu\nu}}{\hat{P}^{\lambda}} &= \ii\hbar\left( \eta^{\nu\lambda}\hat{P}^{\mu} - \eta^{\mu\lambda}\hat{P}^{\nu} \right)\ec\tag{94.6}\\ \comm{\hat{P}^{\mu}}{\hat{P}^{\nu}} &= 0\ep\tag{94.7} \end{align}
Proof.

Let \(\set{X_a}\) be a basis of the abstract Lie algebra with real structure constants, \(\comm{X_a}{X_b} = f_{ab}{}^{c}X_c\), and let \(\dd U\) denote the induced representation of the algebra, so that \(U(\exp tX_a) = \exp\left(-\tfrac{\ii}{\hbar}t\,T_a\right)\) defines the Hermitian generator \(T_a = \ii\hbar\,\dd U(X_a)\). Since \(\dd U\) preserves brackets,

\begin{equation*} \comm{T_a}{T_b} = (\ii\hbar)^{2}\comm{\dd U(X_a)}{\dd U(X_b)} = (\ii\hbar)^{2} f_{ab}{}^{c}\,\dd U(X_c) = \ii\hbar\, f_{ab}{}^{c}\, T_c\ep \end{equation*}

The structure constants of \(\mathfrak{iso}(3,1)\) were derived in Section 14.3.2 from the realization \(J_{AB} = x_A \pp_B - x_B \pp_A\), \(P_C = \pp_C\) (Equation (14.86)): the brackets are Equations (14.90) and (14.97) together with \(\comm{P_A}{P_B} = 0\), restated for the physical signature in Equations (39.18), (39.19) and (39.20) — granting the identification of that realization with the Lie algebra of \(L^{\uparrow}_{+}\) in the parameterization Equation (94.3), a construction recorded as pending at Equation (14.86) and in Section 39.3.2. Multiplying each by \(\ii\hbar\) and raising indices with \(\eta\) gives Equations (94.5), (94.6) and (94.7).

Corollary 94.9 (Rotations, boosts, translations).

Specializing indices in Equations (94.5) and (94.6) via Equation (94.4),

\begin{equation}\tag{94.8} \begin{aligned} \comm{\hat{J}_i}{\hat{J}_j} &= \ii\hbar\,\epsilon_{ijk}\hat{J}_k\ec & \comm{\hat{J}_i}{\hat{K}_j} &= \ii\hbar\,\epsilon_{ijk}\hat{K}_k\ec & \comm{\hat{K}_i}{\hat{K}_j} &= -\ii\hbar\,\epsilon_{ijk}\hat{J}_k\ec\\ \comm{\hat{J}_i}{\hat{P}_j} &= \ii\hbar\,\epsilon_{ijk}\hat{P}_k\ec & \comm{\hat{K}_i}{\hat{P}_j} &= \frac{\ii\hbar}{c}\,\delta_{ij}\hat{H}\ec & \comm{\hat{K}_i}{\hat{H}} &= \ii\hbar c\,\hat{P}_i\ec \end{aligned} \end{equation}

and \(\hat{H}\) commutes with \(\hat{P}_i\) and \(\hat{J}_i\). The relative sign between the first and third brackets — rotations close on rotations, boosts close on rotations with the opposite sign — is the non-compactness of the Lorentz group (Equation (39.3)) seen at the algebra level.

Proof.

Direct substitution. As a sample, \(\comm{\hat{K}_1}{\hat{P}_1} = \comm{\hat{J}^{10}}{\hat{P}^{1}} = \ii\hbar\left(\eta^{01}\hat{P}^{1} - \eta^{11}\hat{P}^{0}\right) = \ii\hbar\,\hat{P}^{0} = \tfrac{\ii\hbar}{c}\hat{H}\), and \(\comm{\hat{J}_1}{\hat{J}_2} = \comm{\hat{J}^{23}}{\hat{J}^{31}} = \ii\hbar\,\eta^{33}\hat{J}^{12}\cdot(-1)^{0} = \ii\hbar\,\hat{J}^{12} = \ii\hbar\,\hat{J}_3\), all other \(\eta\) factors vanishing because the index pairs share no common value. The remaining entries are the same computation with permuted indices.

The universal cover $\SL(2,\C)$

Theorem 94.5 directs us to the covering group. It is concretely available: to each four-vector associate the Hermitian \(2\times2\) matrix

\begin{equation}\tag{94.9} X = x^{\mu}\sigma_{\mu} = \begin{pmatrix} x^{0}+x^{3} & x^{1}-\ii x^{2}\\ x^{1}+\ii x^{2} & x^{0}-x^{3}\end{pmatrix}\ec \qquad \sigma_{\mu} = (\identity, \sigma_{1},\sigma_{2},\sigma_{3})\ec \end{equation}

with \(\sigma_i\) the Pauli matrices (Experiment: Stern–Gerlach). The map is a linear bijection onto Hermitian matrices, and

\begin{equation}\tag{94.10} \det X = (x^{0})^{2} - \vect{x}^{2} = x_{\mu}x^{\mu}\ep \end{equation}
Proposition 94.10 ($\SL(2,\C)$ covers the proper orthochronous Lorentz group).

For \(A \in \SL(2,\C)\) define \(\Lambda(A)\) by \(\left(\Lambda(A)x\right)^{\mu}\sigma_{\mu} = A\,(x^{\mu}\sigma_{\mu})\,A^{\dagger}\). Then \(A \mapsto \Lambda(A)\) is a group homomorphism onto \(L^{\uparrow}_{+}\) with kernel \(\set{\identity,-\identity}\): the map is two-to-one, and \(\SL(2,\C)\), being simply connected, is the universal cover.

Proof.

\(AXA^{\dagger}\) is Hermitian when \(X\) is, so \(\Lambda(A)\) is a well-defined real-linear map of \(\R^{4}\); it preserves \(x_{\mu}x^{\mu}\) by Equation (94.10) because \(\det(AXA^{\dagger}) = \abs{\det A}^{2}\det X = \det X\), hence \(\Lambda(A)\) is a Lorentz transformation (Definition 39.1). Homomorphism: \(\Lambda(AB)x \leftrightarrow AB\,X\,B^{\dagger}A^{\dagger} \leftrightarrow \Lambda(A)\Lambda(B)x\). Kernel: if \(AXA^{\dagger} = X\) for all Hermitian \(X\), then \(X = \identity\) gives \(AA^{\dagger} = \identity\), so \(A\) is unitary and commutes with every \(\sigma_i\); a \(2\times2\) matrix commuting with all Pauli matrices is a multiple \(\lambda\identity\) (write it in the basis \(\set{\identity,\sigma_i}\) and use \(\sigma_i\sigma_j = \delta_{ij}\identity + \ii\epsilon_{ijk}\sigma_k\)), and \(\det A = \lambda^{2} = 1\) forces \(\lambda = \pm1\). Connectedness and image: polar decomposition writes every \(A \in \SL(2,\C)\) uniquely as \(A = V\ee^{M}\) with \(V \in \SU(2)\) and \(M\) traceless Hermitian; \(\SU(2)\) is the three-sphere \(\abs{a}^{2}+\abs{b}^{2}=1\) of rows \((a,b),(-\bar b,\bar a)\) and the traceless Hermitian \(M\) form \(\R^{3}\), so \(\SL(2,\C)\) is connected and simply connected (\(S^{3}\times\R^{3}\); the polar factors depend continuously on \(A\), the positive square root and logarithm being continuous on positive matrices, and \(\pi_1(S^{3}) = 0\)). Its continuous image \(\Lambda(\SL(2,\C))\) is then a connected subgroup of \(L\) containing \(\identity\), hence lands in \(L^{\uparrow}_{+}\) (Proposition 39.6: both defining conditions are clopen). That it is all of \(L^{\uparrow}_{+}\) follows from Lemmas 94.11, 94.12 and 94.13: every rotation and every boost is attained, and every \(\Lambda \in L^{\uparrow}_{+}\) is a boost times a rotation. Finally, a surjective continuous homomorphism of Lie groups with discrete kernel is a covering map — a standard fact used here without restatement. In particular \(L^{\uparrow}_{+}\), as the continuous image of the connected \(\SL(2,\C)\), is itself connected: it is the identity component of \(L\), as announced in Section 39.2.4.

Lemma 94.11 (Unitary elements give rotations).

For a unit vector \(\vect{n}\) and \(V = \exp\left(-\ii\tfrac{\theta}{2} \vect{n}\cdot\vect{\sigma}\right) \in \SU(2)\), the map \(\Lambda(V)\) fixes \(x^{0}\) and rotates \(\vect{x}\) by the angle \(\theta\) about \(\vect{n}\). Every rotation arises this way, and \(V(\theta = 2\pi) = -\identity\).

Proof.

Since \((\vect{n}\cdot\vect{\sigma})^{2} = \identity\) (from \(\sigma_i\sigma_j = \delta_{ij}\identity + \ii\epsilon_{ijk}\sigma_k\) and \(\abs{\vect{n}} = 1\)), the exponential series (Equation (7.40)) splits into \(V = \cos\tfrac{\theta}{2}\,\identity - \ii\sin\tfrac{\theta}{2}\,\vect{n}\cdot\vect{\sigma}\). A direct computation with the same product identity gives, for \(X = x^{0}\identity + \vect{x}\cdot\vect{\sigma}\),

\begin{equation*} V X V^{\dagger} = x^{0}\identity + \left[ \cos\theta\,\vect{x} + \sin\theta\,(\vect{n}\times\vect{x}) + (1-\cos\theta)(\vect{n}\cdot\vect{x})\,\vect{n} \right]\cdot\vect{\sigma}\ec \end{equation*}

which is Rodrigues' rotation formula about \(\vect{n}\) by \(\theta\). Every rotation has an axis and an angle: for \(R \in \SO(3)\), \(\det(R - \identity) = \det\left((R-\identity)\transpose\right) = \det(R^{-1} - \identity) = \det(R^{-1})\det(\identity - R) = -\det(R - \identity)\), so \(\det(R - \identity) = 0\) and \(R\) fixes a unit vector \(\vect{n}\); on the invariant plane orthogonal to \(\vect{n}\), \(R\) is an orientation-preserving linear isometry, i.e. a rotation by some \(\theta\). Hence all of \(\SO(3)\) is attained; at \(\theta = 2\pi\), \(V = \cos\pi\,\identity = -\identity\).

Lemma 94.12 (Hermitian elements give boosts).

For \(H = \exp\left(\tfrac{\eta}{2}\vect{m}\cdot\vect{\sigma}\right)\) with \(\vect{m}\) a unit vector and \(\eta \in \R\), the map \(\Lambda(H)\) is the pure boost of rapidity \(\eta\) along \(\vect{m}\): for \(\vect{m} = \vect{e}_3\),

\begin{equation}\tag{94.11} x^{0} \mapsto x^{0}\cosh\eta + x^{3}\sinh\eta\ec\quad x^{3} \mapsto x^{0}\sinh\eta + x^{3}\cosh\eta\ec\quad x^{1,2} \mapsto x^{1,2}\ep \end{equation}

The dictionary to the velocity parameterization of Lorentz Transformations is \(\beta = \tanh\eta\), \(\gamma = \cosh\eta\); since \(\eta\) ranges over all of \(\R\) and \(\vect{m}\) over all directions, every boost is attained.

Proof.

For \(\vect{m} = \vect{e}_3\) we have \(H = \operatorname{diag}\left(\ee^{\eta/2}, \ee^{-\eta/2}\right)\), and in Equation (94.9) the conjugation \(HXH^{\dagger} = HXH\) multiplies the diagonal entries \(x^{0} \pm x^{3}\) of \(X\) by \(\ee^{\pm\eta}\) while leaving the off-diagonal entries \(x^{1} \mp \ii x^{2}\) fixed (\(\ee^{\eta/2}\ee^{-\eta/2} = 1\)): \(x^{0}+x^{3} \mapsto \ee^{\eta}(x^{0}+x^{3})\), \(x^{0}-x^{3} \mapsto \ee^{-\eta}(x^{0}-x^{3})\), which is Equation (94.11). A general axis \(\vect{m}\) is \(R\,\vect{e}_3\) for a rotation \(R\); conjugating \(H\) by the corresponding \(V\) of Lemma 94.11 rotates the boost axis, since \(\Lambda\) is a homomorphism.

Lemma 94.13 (Boost–rotation decomposition).

Every \(\Lambda \in L^{\uparrow}_{+}\) factors as \(\Lambda = B R\) with \(B\) a pure boost and \(R\) a rotation.

Proof.

Let \(u = \Lambda e_0\), where \(e_0 = (1,0,0,0)\). Then \(u_{\mu}u^{\mu} = 1\) and \(u^{0} = \Lambda^{0}{}_{0} \ge 1\) (Definition 39.5), so \(u = (\cosh\eta,\ \sinh\eta\,\vect{m})\) for some rapidity \(\eta \ge 0\) and unit vector \(\vect{m}\). Let \(B\) be the boost of Lemma 94.12 along \(\vect{m}\) with rapidity \(\eta\); then \(B e_0 = u\), so \(R = B^{-1}\Lambda\) satisfies \(R e_0 = e_0\). A Lorentz transformation fixing \(e_0\) preserves the spatial slice \(x^{0} = 0\) and the Euclidean norm on it (both statements from \(\Lambda\eta\Lambda\transpose = \eta\), Equation (39.2): the first column of \(R\) is \(e_0\), hence so is the first row), so \(R \in \Ogrp(3)\); and \(\det R = \det\Lambda = 1\) places \(R \in \SO(3)\).

Corollary 94.14 ($\SU(2)$ covers $\SO(3)$; half-integer spin is projective).

Restricting Proposition 94.10 to the rotation subgroup: \(\SU(2) \to \SO(3)\) is a two-to-one covering with kernel \(\pm\identity\). A representation of \(\SU(2)\) in which \(-\identity \mapsto -\identity\) — half-integer spin, as classified in Lemma 94.28 below — is therefore a projective representation of the rotation group itself, with precisely the \(4\pi\) periodicity observed in [Werner:1975] (Remark 94.6).

The invariants: mass and spin

Which combinations of the generators are the same in every inertial frame? By Schur's lemma (Theorem 5.153) an operator commuting with the whole representation acts as a number on an irreducible one: such Casimir operators are the invariant labels we are looking for.

Definition 94.15 (Pauli–Lubanski vector).
\begin{equation}\tag{94.12} \hat{W}_{\mu} = \tfrac{1}{2}\,\epsilon_{\mu\nu\rho\sigma}\, \hat{J}^{\nu\rho}\hat{P}^{\sigma}\ec \qquad \epsilon_{0123} = +1\ep \end{equation}
Lemma 94.16 (Invariance identity for $\epsilon$).

For every \(\omega \in \mathfrak{so}(3,1)\), i.e.\ \(\omega_{\mu\nu} = -\omega_{\nu\mu}\),

\begin{equation}\tag{94.13} \omega^{\alpha}{}_{\mu}\epsilon_{\alpha\nu\rho\sigma} + \omega^{\alpha}{}_{\nu}\epsilon_{\mu\alpha\rho\sigma} + \omega^{\alpha}{}_{\rho}\epsilon_{\mu\nu\alpha\sigma} + \omega^{\alpha}{}_{\sigma}\epsilon_{\mu\nu\rho\alpha} = 0\ep \end{equation}
Proof.

The left side is totally antisymmetric in \(\mu\nu\rho\sigma\): swapping any two of them exchanges two terms and flips the sign of the remaining two through the antisymmetry of \(\epsilon\). A totally antisymmetric four-index array in four dimensions is a multiple of \(\epsilon_{\mu\nu\rho\sigma}\); contracting the left side with \(\epsilon^{\mu\nu\rho\sigma}\) shows the multiple is proportional to \(\omega^{\alpha}{}_{\alpha} = \eta^{\alpha\beta}\omega_{\beta\alpha} = 0\).

Lemma 94.17 (Properties of $\hat{W}$).
  1. \(\comm{\hat{W}_{\mu}}{\hat{P}_{\lambda}} = 0\);

  2. \(\hat{W}_{\mu}\hat{P}^{\mu} = \hat{P}^{\mu}\hat{W}_{\mu} = 0\);

  3. \(\hat{W}\) is a four-vector: \(\comm{\hat{J}_{\mu\nu}}{\hat{W}_{\lambda}} = \ii\hbar\left(\eta_{\nu\lambda}\hat{W}_{\mu} - \eta_{\mu\lambda}\hat{W}_{\nu}\right)\).

Proof.

(i) By Equations (94.6) and (94.7), \(\comm{\hat{J}^{\nu\rho}\hat{P}^{\sigma}}{\hat{P}_{\lambda}} = \comm{\hat{J}^{\nu\rho}}{\hat{P}_{\lambda}}\hat{P}^{\sigma} = \ii\hbar\left(\delta^{\rho}_{\lambda}\hat{P}^{\nu} - \delta^{\nu}_{\lambda}\hat{P}^{\rho}\right)\hat{P}^{\sigma}\); contracted with \(\epsilon_{\mu\nu\rho\sigma}\) each term dies because \(\hat{P}^{\nu}\hat{P}^{\sigma}\) (respectively \(\hat{P}^{\rho}\hat{P}^{\sigma}\)) is symmetric in indices that \(\epsilon\) antisymmetrizes.

(ii) \(\hat{W}_{\mu}\hat{P}^{\mu} = \tfrac12 \epsilon_{\mu\nu\rho\sigma}\hat{J}^{\nu\rho}\hat{P}^{\sigma} \hat{P}^{\mu} = 0\) for the same symmetry reason, using Equation (94.7); the reversed order follows from (i).

(iii) Insert Equations (94.5) and (94.6) into \(\comm{\hat{J}_{\mu\nu}}{\tfrac12\epsilon_{\lambda\alpha\beta\gamma} \hat{J}^{\alpha\beta}\hat{P}^{\gamma}}\) and use the Leibniz rule for commutators; the resulting four \(\epsilon\eta\)-contractions combine, by Lemma 94.16 applied with \(\omega\) the antisymmetrized pair \((\mu\nu)\), into exactly the two terms \(\ii\hbar(\eta_{\nu\lambda}\hat{W}_{\mu} - \eta_{\mu\lambda}\hat{W}_{\nu})\).

Theorem 94.18 (The Casimir invariants of the Poincaré group).

\(\hat{P}^{2} = \hat{P}_{\mu}\hat{P}^{\mu}\) and \(\hat{W}^{2} = \hat{W}_{\mu}\hat{W}^{\mu}\) commute with every generator of Proposition 94.8. On an irreducible representation each is a real number.

Proof.

\(\comm{\hat{P}^{2}}{\hat{P}_{\lambda}} = 0\) by Equation (94.7), and \(\comm{\hat{P}^{2}}{\hat{J}_{\mu\nu}} = 0\) because Equation (94.6) makes \(\hat{P}\) a four-vector and the square of a four-vector is a scalar: \(\comm{\hat{J}_{\mu\nu}}{\hat{P}_{\lambda}\hat{P}^{\lambda}} = \ii\hbar\left(\eta_{\nu\lambda}\hat{P}_{\mu} - \eta_{\mu\lambda}\hat{P}_{\nu}\right)\hat{P}^{\lambda} + \ii\hbar\,\hat{P}_{\lambda}\left(\delta^{\lambda}_{\nu}\hat{P}_{\mu} - \delta^{\lambda}_{\mu}\hat{P}_{\nu}\right) = \ii\hbar\left(\hat{P}_{\mu}\hat{P}_{\nu} - \hat{P}_{\nu}\hat{P}_{\mu}\right) + \ii\hbar\left(\hat{P}_{\nu}\hat{P}_{\mu} - \hat{P}_{\mu}\hat{P}_{\nu}\right) = 0\) identically. The same computation with Lemma 94.17(iii) kills \(\comm{\hat{J}_{\mu\nu}}{\hat{W}^{2}}\), and \(\comm{\hat{W}^{2}}{\hat{P}_{\lambda}} = 0\) is Lemma 94.17(i). Schur's lemma (Theorem 5.153, with the qualification of Remark 94.1) then makes each a scalar on an irreducible representation; Hermiticity makes the scalar real.

Definition 94.19 (Mass).

On an irreducible representation write

\begin{equation}\tag{94.14} \hat{P}^{2} = m^{2}c^{2}\,\identity\ep \end{equation}

For the representations selected by observation (Remark 94.22 below) \(m^{2} \ge 0\); the number \(m \ge 0\) is the mass of the particle. The value of \(\hat{W}^{2}\) will emerge as the spin in Theorem 94.29 and Proposition 94.32.

Orbits, little groups, and induced representations

Since the \(\hat{P}^{\mu}\) commute (Equation (94.7)), diagonalize them: a basis \(\ket{p,\sigma}\) with \(\hat{P}^{\mu}\ket{p,\sigma} = p^{\mu}\ket{p,\sigma}\), where \(\sigma\) enumerates whatever degeneracy remains. Translations then act by \(U(\identity,a)\ket{p,\sigma} = \ee^{-\ii\,a\cdot p/\hbar} \ket{p,\sigma}\), and by Equation (94.6) a Lorentz transformation carries the eigenvalue \(p\) to \(\Lambda p\): \(\hat{P}^{\mu}\,U(\Lambda)\ket{p,\sigma} = (\Lambda p)^{\mu}\,U(\Lambda)\ket{p,\sigma}\) (conjugate \(\hat{P}\) through \(U(\Lambda)\) with Equations (94.3) and (94.6)). An irreducible representation therefore lives on a single orbit of momenta.

Lemma 94.20 (Orthochronous transformations preserve the sign of $p^{0}$ on causal vectors).

If \(p_{\mu}p^{\mu} \ge 0\), \(p \neq 0\), and \(\Lambda \in L^{\uparrow}_{+}\), then \(\operatorname{sgn}\left((\Lambda p)^{0}\right) = \operatorname{sgn}\left(p^{0}\right)\).

Proof.

\((\Lambda p)^{0} = \Lambda^{0}{}_{0}p^{0} + \Lambda^{0}{}_{i}p^{i}\). By the Cauchy–Schwarz inequality of Section 6.3 and the row identity \(\left(\Lambda^{0}{}_{0}\right)^{2} = 1 + \sum_i\left(\Lambda^{0}{}_{i}\right)^{2}\) (the \(\mu=\nu=0\) component of Equation (39.2) applied to \(\Lambda^{-1} = \eta\Lambda\transpose\eta\)),

\begin{equation*} \textstyle\abs{\Lambda^{0}{}_{i}p^{i}} \le \sqrt{\sum_i\left(\Lambda^{0}{}_{i}\right)^{2}}\; \abs{\vect{p}} < \Lambda^{0}{}_{0}\,\abs{p^{0}} \end{equation*}

whenever \(\abs{\vect{p}} \le \abs{p^{0}}\) with \(p \ne 0\) (causality) — strictly, since either the causal inequality is strict or \(\sum_i (\Lambda^{0}{}_{i})^{2} < (\Lambda^{0}{}_{0})^{2}\). Hence the first term dominates and carries the sign of \(p^{0}\) (\(\Lambda^{0}{}_{0} \ge 1\), Definition 39.5).

Proposition 94.21 (Orbit classification).

Under \(L^{\uparrow}_{+}\) the nonzero momenta fall into the disjoint orbits

orbitinvariantsstandard momentum $k$
massive, positive energy$p^{2} = m^{2}c^{2} > 0$, $p^{0} > 0$$(mc, 0, 0, 0)$
massive, negative energy$p^{2} = m^{2}c^{2} > 0$, $p^{0} < 0$$(-mc, 0, 0, 0)$
massless, positive energy$p^{2} = 0$, $p^{0} > 0$$(\kappa, 0, 0, \kappa)$
massless, negative energy$p^{2} = 0$, $p^{0} < 0$$(-\kappa, 0, 0, \kappa)$
tachyonic$p^{2} = -\mu^{2}c^{2} < 0$$(0, 0, 0, \mu c)$

together with the fixed point \(p = 0\); here \(\kappa > 0\) is any fixed reference momentum (different \(\kappa\) label the same orbit).

Proof.

\(p^{2}\) is invariant by definition of \(L\), and \(\operatorname{sgn}(p^{0})\) is invariant on the causal orbits by Lemma 94.20, so the listed classes are unions of orbits. Each class is a single orbit because its standard momentum reaches every member: a rotation (Lemma 94.11) aligns \(\vect{p}\) with \(\vect{e}_3\), and the boost Equation (94.11) with the appropriate rapidity then produces any admissible \(p^{0}\): \(\cosh\eta = p^{0}/(mc)\) in the massive case, \(\ee^{\eta} = p^{0}/\kappa\) in the massless case, \(\sinh\eta = p^{0}/(\mu c)\) in the tachyonic case, the spatial part then matching automatically — which also shows both signs of \(p^{0}\) lie on one tachyonic orbit, \(\sinh\) being unbounded in both directions; the negative-energy rows follow by applying the same transformations to their own standard momenta. The composition of these transformations is the standard boost \(L(p)\), fixed once and for all on each orbit:

\begin{equation}\tag{94.15} L(p) = R(\hat{\vect{p}})\,B_{3}(\eta_p)\ec\qquad L(p)\,k = p\ec \end{equation}

with \(R(\hat{\vect{p}})\) a fixed choice of rotation carrying \(\vect{e}_3\) to the unit vector \(\hat{\vect{p}}\) and \(B_{3}\) the boost Equation (94.11); the conventions are \(R(\vect{e}_3) = \identity\) — so that \(L(k) = \identity\) — and a fixed measurable choice of \(R(\hat{\vect{p}})\) elsewhere (no continuous global choice exists); any such choice serves.

Remark 94.22 (Observation selects two orbit classes).

Nature realizes the massive and massless positive-energy orbits, and only those. Negative-energy representations would make every system unstable against decay into ever-lower energies — the resolution of this instability is the reinterpretation of the negative-frequency modes in quantum field theory, anticipated by the positron (Experiment: The Positron) — and no tachyonic particle (\(p^{2} < 0\)) has ever been observed. This is an empirical input to the classification, not a mathematical one, and this treatise records it as such (cf. the front-matter scope rules).

Definition 94.23 (Little group).

The little group (stabilizer) of the standard momentum \(k\) is

\begin{equation}\tag{94.16} G_{k} = \set{\Lambda \in L^{\uparrow}_{+} \mid \Lambda k = k}\ep \end{equation}
Proposition 94.24 (Massive little group).

For \(k = (mc,0,0,0)\), \(G_{k} = \SO(3)\), with cover \(\SU(2)\).

Proof.

\(\Lambda k = k\) says \(\Lambda e_0 = e_0\); the proof of Lemma 94.13 showed exactly that such a \(\Lambda\) is a rotation, and conversely rotations fix \(e_0\). The cover statement is Corollary 94.14.

Proposition 94.25 (Massless little group).

For \(k = (\kappa,0,0,\kappa)\) the Lie algebra of \(G_{k}\) is spanned by

\begin{equation}\tag{94.17} \hat{J}^{3} = \hat{J}^{12}\ec\qquad \hat{A} = \hat{J}^{01} + \hat{J}^{13}\ec\qquad \hat{B} = \hat{J}^{02} + \hat{J}^{23}\ec \end{equation}

with brackets

\begin{equation}\tag{94.18} \comm{\hat{J}^{3}}{\hat{A}} = \ii\hbar\,\hat{B}\ec\qquad \comm{\hat{J}^{3}}{\hat{B}} = -\ii\hbar\,\hat{A}\ec\qquad \comm{\hat{A}}{\hat{B}} = 0\ep \end{equation}

This is the algebra \(\mathfrak{iso}(2)\) of the Euclidean group of the plane — a rotation (\(\hat{J}^{3}\)) and two commuting translations (\(\hat{A},\hat{B}\)) — cf. the semidirect structure Equation (14.99) in two Euclidean dimensions.

Proof.

An algebra element \(\tfrac12\omega_{\mu\nu}\hat{J}^{\mu\nu}\) stabilizes \(k\) iff \(\omega^{\mu}{}_{\nu}k^{\nu} = 0\), i.e.\ \(\omega^{\mu}{}_{0} + \omega^{\mu}{}_{3} = 0\) for every \(\mu\). Writing this out with \(\eta\): for \(\mu = 0\) it gives \(\omega_{03} = 0\); for \(\mu = 1\), \(\omega_{01} = \omega_{13}\); for \(\mu = 2\), \(\omega_{02} = \omega_{23}\); for \(\mu = 3\) it repeats \(\omega_{03} = 0\). Three free parameters remain, multiplying exactly the three combinations Equation (94.17). The brackets follow from Equation (94.5) by direct substitution; as a sample, in \(\comm{\hat{J}^{12}}{\hat{J}^{01}}\) the only surviving term of Equation (94.5) is \(\eta^{\mu\rho}\hat{J}^{\nu\lambda}\) with \((\mu\nu\lambda\rho) = (1,2,0,1)\), giving \(\ii\hbar\,\eta^{11}\hat{J}^{20} = \ii\hbar\,\hat{J}^{02}\), and \(\comm{\hat{J}^{12}}{\hat{J}^{13}} = \ii\hbar\,\eta^{11}\hat{J}^{32} = \ii\hbar\,\hat{J}^{23}\), whose sum is \(\ii\hbar\,\hat{B}\); the other two brackets are the same bookkeeping, with the two disjoint-index commutators \(\comm{\hat{J}^{01}}{\hat{J}^{23}} = \comm{\hat{J}^{13}}{\hat{J}^{02}} = 0\) making \(\comm{\hat{A}}{\hat{B}}\) collapse to \(\comm{\hat{J}^{01}}{\hat{J}^{02}} + \comm{\hat{J}^{13}}{\hat{J}^{23}} = \ii\hbar\left(-\hat{J}^{12} + \hat{J}^{12}\right) = 0\).

Lemma 94.26 (Lorentz-invariant measure on a mass shell).

On the orbit \(p^{2} = m^{2}c^{2} \ge 0\), \(p^{0} > 0\), the measure

\begin{equation}\tag{94.19} \dd\mu(p) = \frac{\dd^{3}p}{2p^{0}}\ec\qquad p^{0} = \sqrt{\vect{p}^{2} + m^{2}c^{2}}\ec \end{equation}

satisfies \(\int f(\Lambda^{-1}p)\,\dd\mu(p) = \int f(p)\,\dd\mu(p)\) for every \(\Lambda \in L^{\uparrow}_{+}\) and integrable \(f\).

Proof.

Write the integral over the orbit as a four-dimensional one,

\begin{equation*} \int f(p)\,\frac{\dd^{3}p}{2p^{0}} = \int \dd^{4}p\; \delta\!\left(p_{\mu}p^{\mu} - m^{2}c^{2}\right) \theta\!\left(p^{0}\right) f(p)\ec \end{equation*}

which follows from the composition rule for the Dirac delta (Proposition 7.113): at fixed \(\vect{p}\) the argument \(\left(p^{0}\right)^{2} - \vect{p}^{2} - m^{2}c^{2}\) has the single positive root Equation (94.19) with derivative \(2p^{0}\). In the four-dimensional form each factor is invariant under \(p \mapsto \Lambda p\): the delta because \(p^{2}\) is, the step function by Lemma 94.20, and \(\dd^{4}p\) because \(\abs{\det\Lambda} = 1\).

Theorem 94.27 (Wigner's induced representations).

Fix a causal positive-energy orbit of Proposition 94.21 — the ones Lemma 94.26 equips with an invariant measure, and the only ones used in the sequel — with standard momentum \(k\), and a unitary irreducible representation \(D_{\sigma'\sigma}\) of (the cover of) the little group \(G_{k}\) on a space with basis \(\set{\sigma}\). For half-integer representations everything is read in the covering group: \(L(p)\) then denotes the fixed lift of the standard boost Equation (94.15) assembled from the \(\SL(2,\C)\) elements of Lemmas 94.11 and 94.12, so that \(W(\Lambda,p)\) below lies in the cover of \(G_{k}\) and \(D(W)\) is unambiguous; for integer spin the formulas descend to \(L^{\uparrow}_{+}\). Define one-particle states by \(\ket{k,\sigma}\) carrying \(D\), and

\begin{equation}\tag{94.20} \ket{p,\sigma} = U\!\left(L(p)\right)\ket{k,\sigma}\ec \end{equation}

with \(L(p)\) the standard boost Equation (94.15) and the normalization \(\braket{p',\sigma'}{p,\sigma} = 2p^{0}\,\delta^{3}(\vect{p}'-\vect{p})\,\delta_{\sigma'\sigma}\). Then

\begin{align} U(\identity,a)\,\ket{p,\sigma} &= \ee^{-\ii\,p\cdot a/\hbar}\,\ket{p,\sigma}\ec \tag{94.21}\\ U(\Lambda,0)\,\ket{p,\sigma} &= \sum_{\sigma'} D_{\sigma'\sigma}\!\left(W(\Lambda,p)\right)\, \ket{\Lambda p,\sigma'}\ec \qquad W(\Lambda,p) = L(\Lambda p)^{-1}\,\Lambda\,L(p)\ec \tag{94.22} \end{align}

defines a unitary representation of the covering group; the element \(W(\Lambda,p)\) — the Wigner rotation — indeed lies in \(G_{k}\).

Proof.

\(W(\Lambda,p)k = L(\Lambda p)^{-1}\Lambda L(p)\,k = L(\Lambda p)^{-1}(\Lambda p) = k\), so \(W(\Lambda,p) \in G_{k}\). Composition: applying Equation (94.22) twice,

\begin{equation*} W(\Lambda',\Lambda p)\,W(\Lambda,p) = L(\Lambda'\Lambda p)^{-1}\Lambda' L(\Lambda p)\, L(\Lambda p)^{-1}\Lambda L(p) = W(\Lambda'\Lambda, p)\ec \end{equation*}

so the \(D\) factors compose correctly and \(U(\Lambda')U(\Lambda) = U(\Lambda'\Lambda)\); translations compose by inspection of Equation (94.21), and the mixed products follow from Equation (94.2) applied to \(\Lambda a\). Unitarity: for wave packets \(\ket{\psi} = \sum_{\sigma}\int\dd\mu(p)\, \psi_{\sigma}(p)\ket{p,\sigma}\) the chosen normalization gives \(\norm{\psi}^{2} = \sum_{\sigma}\int\dd\mu(p) \abs{\psi_{\sigma}(p)}^{2}\), and by Equation (94.22) a Lorentz transformation replaces \(\psi_{\sigma}(p)\) by \(\sum_{\sigma'}D_{\sigma\sigma'}(W)\,\psi_{\sigma'}(\Lambda^{-1}p)\): the sum over \(\sigma\) is preserved because \(D\) is unitary, and the integral because the measure is invariant (Lemma 94.26).

Derivation pending.

Completeness of the classification: that the representation defined by Equations (94.21) and (94.22) is irreducible precisely when \(D\) is, and that every irreducible unitary representation of the cover with \(\hat{P}\)-spectrum on the given orbit is equivalent to an induced one, is Mackey's imprimitivity theorem, specialized by Wigner to this group [Wigner:1939]. The systematic imprimitivity machinery is not yet developed in this treatise.

Massive particles: spin

By Proposition 94.24 and Theorem 94.27, classifying massive particles means classifying the unitary irreducible representations of \(\SU(2)\) — equivalently, of its algebra \(\comm{\hat{J}_i}{\hat{J}_j} = \ii\hbar\,\epsilon_{ijk}\hat{J}_k\) (Equation (94.8)).

Lemma 94.28 (Unitary irreducible representations of $\mathfrak{su}(2)$).

The finite-dimensional unitary irreducible representations of the algebra \(\comm{\hat{J}_i}{\hat{J}_j} = \ii\hbar\,\epsilon_{ijk} \hat{J}_k\) are labelled by \(s \in \set{0, \tfrac12, 1, \tfrac32, \ldots}\). The representation labelled \(s\) has dimension \(2s+1\), carries \(\hat{\vect{J}}^{2} = \hbar^{2}s(s+1)\,\identity\), and admits a basis \(\ket{s,\sigma}\), \(\sigma = -s, -s+1, \ldots, s\), with \(\hat{J}_3\ket{s,\sigma} = \hbar\sigma\ket{s,\sigma}\).

Proof.

Let \(\hat{J}_{\pm} = \hat{J}_1 \pm \ii\hat{J}_2\), so that \(\hat{J}_{\pm}^{\dagger} = \hat{J}_{\mp}\) and, from the algebra,

\begin{equation}\tag{94.23} \comm{\hat{J}_3}{\hat{J}_{\pm}} = \pm\hbar\,\hat{J}_{\pm}\ec\qquad \comm{\hat{J}_{+}}{\hat{J}_{-}} = 2\hbar\,\hat{J}_3\ec\qquad \hat{\vect{J}}^{2} = \hat{J}_{\mp}\hat{J}_{\pm} + \hat{J}_3^{2} \pm \hbar\hat{J}_3\ep \end{equation}

\(\hat{\vect{J}}^{2}\) commutes with each \(\hat{J}_i\) (same computation as in Theorem 94.18, three-dimensional case), so on an irreducible representation \(\hat{\vect{J}}^{2} = \lambda\hbar^{2} \identity\) with \(\lambda \ge 0\) (it is a sum of squares of Hermitian operators). Diagonalize the Hermitian \(\hat{J}_3\): eigenvalues \(\hbar\sigma\). The first bracket in Equation (94.23) shifts \(\sigma\) by \(\pm1\) under \(\hat{J}_{\pm}\), while

\begin{equation}\tag{94.24} \norm{\hat{J}_{\pm}\ket{\sigma}}^{2} = \bra{\sigma}\hat{J}_{\mp}\hat{J}_{\pm}\ket{\sigma} = \hbar^{2}\left[\lambda - \sigma(\sigma\pm1)\right] \ \ge 0\ep \end{equation}

In a finite-dimensional space the ladder must terminate in both directions: there are extremal eigenvalues \(\sigma_{\pm}\) with \(\hat{J}_{+}\ket{\sigma_{+}} = 0\) and \(\hat{J}_{-}\ket{\sigma_{-}} = 0\), so Equation (94.24) forces \(\lambda = \sigma_{+}(\sigma_{+}+1) = \sigma_{-}(\sigma_{-}-1)\), whose only solution with \(\sigma_{+} \ge \sigma_{-}\) is \(\sigma_{-} = -\sigma_{+}\). Writing \(s = \sigma_{+}\): the ladder descends from \(\sigma_{+}\) to \(\sigma_{-}\) in unit steps, so \(2s \in \Z_{\ge0}\) and \(\lambda = s(s+1)\). The states \(\set{\ket{s,\sigma}}_{\sigma=-s}^{s}\) span an invariant subspace (the three generators are linear combinations of \(\hat{J}_3\) and \(\hat{J}_{\pm}\), which act within it), hence by irreducibility the whole space: the dimension is \(2s+1\). Conversely each such \(s\) is realized — Equation (94.24) defines the matrix elements, and the brackets Equation (94.23) are verified directly — so the list is exhaustive. (Infinite-dimensional unitary irreducible representations of a compact group do not occur: every unitary representation of \(\SU(2)\) decomposes into finite-dimensional ones — the Peter–Weyl theorem [Weinberg:1995]; the treatise's own account of \(\SU(2)\), Section 14.2.2, is still pending.)

Theorem 94.29 (Massive particles).

The irreducible positive-energy representations with \(m > 0\) are classified by the pair \((m, s)\) with \(2s \in \Z_{\ge0}\): mass and spin. The representation \((m,s)\) has \(2s+1\) polarization states for each momentum, and

\begin{equation}\tag{94.25} \hat{W}^{2} = -m^{2}c^{2}\,\hbar^{2}\,s(s+1)\,\identity\ep \end{equation}
Proof.

The classification is Theorem 94.27, Proposition 94.24 and Lemma 94.28, modulo the completeness statement recorded as pending after Theorem 94.27 (irreducibility and exhaustiveness of the induced family). For Equation (94.25) evaluate \(\hat{W}^{2}\) — a Casimir, so any convenient states will do (Theorem 94.18) — on the standard-momentum states \(\ket{k,\sigma}\), \(k = (mc,\vect{0})\). There \(\hat{P}^{\sigma}\) may be replaced by its eigenvalue, in both factors of \(\hat{W}^{2}\): \(\hat{W}\) preserves the momentum eigenspace by Lemma 94.17(i). In Equation (94.12) only \(\hat{P}^{0} = mc\) survives,

\begin{equation*} \hat{W}_{0}\ket{k,\sigma} = \tfrac12\epsilon_{0\nu\rho 0}\,\hat{J}^{\nu\rho}mc \ket{k,\sigma} = 0\ec \qquad \hat{W}_{i}\ket{k,\sigma} = \tfrac{mc}{2}\,\epsilon_{i\nu\rho0}\,\hat{J}^{\nu\rho} \ket{k,\sigma} = -mc\,\hat{J}_i\ket{k,\sigma}\ec \end{equation*}

the last step because \(\epsilon_{ijk0} = -\epsilon_{0ijk}\) (one cyclic shift of four indices is an odd permutation) and \(\hat{J}_i = \tfrac12\epsilon_{ijk}\hat{J}^{jk}\) Equation (94.4). Hence on these states \(\hat{W}^{2} = \hat{W}_{0}\hat{W}^{0} + \hat{W}_{i}\hat{W}^{i} = -\sum_i\left(mc\,\hat{J}_i\right)^{2} = -m^{2}c^{2}\,\hat{\vect{J}}^{2}\), and Lemma 94.28 supplies the eigenvalue \(\hbar^{2}s(s+1)\): the Pauli–Lubanski square measures the rest-frame angular momentum — the spin.

Phenomenon 94.30 (Quantization of intrinsic angular momentum).

The intrinsic angular momentum of every particle species is quantized: measured along any axis it takes exactly \(2s+1\) equally spaced values \(\hbar\sigma\), \(\sigma = -s,\ldots,s\), with \(2s\) a nonnegative integer characteristic of the species. For the electron and for silver atoms, \(s = \tfrac12\): a magnetic-gradient analyzer splits an unpolarized beam into exactly two components (Section 80.2), never a classical continuum; light transfers angular momentum to matter in units set by \(\hbar\) per photon [Beth:1936].

Derivation. Theorem 94.29: a massive particle species is an irreducible representation \((m,s)\), its polarization space is the \((2s+1)\)-dimensional representation of Lemma 94.28, and any axis may be taken as the \(3\)-axis by Proposition 94.24, so a measurement of the spin projection resolves the \(2s+1\) eigenvalues \(\hbar\sigma\) of \(\hat{J}_3\) (Postulate 77.3). The massless counterpart — helicity \(\pm\hbar\) for the photon — is Proposition 94.32 below. The experimental record is analyzed in Experiment: Stern–Gerlach and [Beth:1936].

Massless particles: helicity

Proposition 94.31 (Massless representations).

On the massless positive-energy orbit, the representations realized by observed particles have \(\hat{A}\) and \(\hat{B}\) (Equation (94.17)) acting as zero. The little group then effectively reduces to the rotations \(\ee^{-\ii\theta \hat{J}^{3}/\hbar}\) about the momentum axis, whose irreducible unitary representations are one-dimensional,

\begin{equation}\tag{94.26} D_{\theta} = \ee^{-\ii h\theta}\ec\qquad \hat{J}^{3} = \hbar h\,\identity\ec\qquad h \in \tfrac12\Z\ec \end{equation}

and the number \(h\) — the helicity — replaces spin: a massless species carries a single polarization state per momentum.

Proof.

Suppose instead \(\hat{A}\) (say) had a nonzero eigenvalue \(a\) on some state. Since \(\comm{\hat{A}}{\hat{B}} = 0\) Equation (94.18), diagonalize both; conjugating by the rotation \(\ee^{-\ii\theta\hat{J}^{3}/\hbar}\) rotates the eigenvalue pair \((a,b)\) by the angle \(\theta\) in the plane (this is Equation (94.18) exponentiated), so the spectrum would contain the whole circle of pairs — a continuous infinity of internal states at fixed momentum, Wigner's “continuous spin” representations [Wigner:1939]. No observed massless particle has more than two polarization states, so observation forces \(a = b = 0\): \(\hat{A} = \hat{B} = 0\) on the physical representation (the same empirical selection as Remark 94.22). What remains of the little group is the one-parameter rotation subgroup; its irreducible unitary representations are one-dimensional because the group is abelian: each \(D_{\theta}\) commutes with the whole representation, so by Schur's lemma (Theorem 5.153, with the qualification of Remark 94.1) it is a scalar, every line is then invariant, and irreducibility forces dimension one (cf. the \(\SO(2)\) analysis of Section 14.2.1). Thus \(D_{\theta} = \ee^{-\ii h\theta}\) with \(\hat{J}^{3}\) acting as the number \(\hbar h\). Quantization of \(h\): in the covering group the rotation subgroup about \(\vect{e}_3\) is \(V(\theta) = \exp\left(-\ii\tfrac{\theta}{2}\sigma_3\right)\) (Lemma 94.11), which closes only at \(\theta = 4\pi\); single-valuedness of \(D\) on the cover demands \(\ee^{-4\pi\ii h} = 1\), i.e. \(2h \in \Z\). (A rotation by \(2\pi\) multiplies the state by \(\ee^{-2\pi\ii h} = (-1)^{2h}\), exactly as for massive spin, Corollary 94.14.)

Proposition 94.32 (Helicity is the projection of angular momentum on the momentum).

On a massless representation with helicity \(h\),

\begin{equation}\tag{94.27} \hat{W}_{\mu} = \hbar h\,\hat{P}_{\mu}\ec \qquad\text{equivalently}\qquad \frac{\hat{\vect{J}}\cdot\hat{\vect{P}}}{\abs{\hat{\vect{P}}}} = \hbar h\,\identity\ec \end{equation}

and \(\hat{W}^{2} = 0\).

Proof.

Evaluate the components of Equation (94.12) on the standard-momentum states, \(p = k = (\kappa,0,0,\kappa)\), replacing \(\hat{P}\) by its eigenvalue (\(k^{0} = k^{3} = \kappa\)):

\begin{align*} \hat{W}_{0} &= \tfrac12\epsilon_{0\nu\rho3}\hat{J}^{\nu\rho}\kappa = \kappa\,\hat{J}^{12} = \kappa\,\hat{J}^{3}\ec & \hat{W}_{3} &= \tfrac12\epsilon_{3\nu\rho0}\hat{J}^{\nu\rho}\kappa = -\kappa\,\hat{J}^{12} = -\kappa\,\hat{J}^{3}\ec\\ \hat{W}_{1} &= -\kappa\left(\hat{J}^{23} + \hat{J}^{02}\right) = -\kappa\,\hat{B}\ec & \hat{W}_{2} &= \kappa\left(\hat{J}^{13} + \hat{J}^{01}\right) = \kappa\,\hat{A}\ec \end{align*}

each line by direct expansion of the \(\epsilon\) sums with \(\epsilon_{0123} = +1\). On the physical representation \(\hat{A} = \hat{B} = 0\) and \(\hat{J}^{3} = \hbar h\) (Proposition 94.31), so \(\hat{W}_{\mu} = \hbar h\,(\kappa,0,0,-\kappa)_{\mu} = \hbar h\,k_{\mu}\) on these states. Both sides of \(\hat{W}_{\mu} = \hbar h \hat{P}_{\mu}\) are four-vectors built on the representation (Lemma 94.17), so equality at the standard momentum propagates to the whole orbit by conjugation with \(U(L(p))\). Contracting with itself, \(\hat{W}^{2} = \hbar^{2}h^{2} \hat{P}^{2} = 0\); taking the spatial part at momentum \(\vect{p}\) gives the stated projection formula (\(\hat{W}^{i}\) reduces to \(p^{0}\hat{J}_i\)-terms along \(\hat{\vect{p}}\)).

Remark 94.33 (The photon).

The photon is massless to spectacular experimental accuracy (\(m_{\gamma}c^{2} < 10^{-18}\,\mathrm{eV}\) [Navas:2024]) and carries helicity \(h = \pm1\): circularly polarized light exerts a measurable torque, \(\pm\hbar\) of angular momentum per photon, first detected by Beth with a torsion pendulum of birefringent plates [Beth:1936]. The two signs are exchanged by spatial reflection, so the parity-symmetric electrodynamics of Part VII — Classical Field Theory and Electrodynamics bundles \(h = +1\) and \(h = -1\) into one species, “the photon”, with two polarization states — and only two: the missing longitudinal state of a massive spin-1 particle is not a small correction but a different orbit of the Poincaré group altogether. That a state \(h = 0\) is absent is the group-theoretic seed of gauge invariance, taken up in Part XI — Quantum Field Theory and the Standard Model. The same counting will reappear for the graviton candidates of Part XIII — Frontiers and Open Problems: helicity \(\pm2\), two polarizations — the two polarizations derived for gravitational waves in The Quadrupole Formula for Gravitational Radiation.

The particle dictionary

Definition 94.34 (Particle).

A particle species is an irreducible unitary representation of the universal cover \(\R^{4}\rtimes\SL(2,\C)\) of the proper orthochronous Poincaré group, realized on the positive-energy massive or massless orbits, labelled by \((m, s)\) with \(m > 0\) and \(2s \in \Z_{\ge0}\), or by \((0, h)\) with \(2h \in \Z\).

Every column of the experimental particle tables [Navas:2024] is an instance:

species$mc^{2}$spin/helicitypolarization states
electron\(0.51099895\,\mathrm{MeV}\) [Tiesinga:2021]$s = \tfrac12$$2$ (Section 80.2)
proton\(938.27\,\mathrm{MeV}\) [Navas:2024]$s = \tfrac12$$2$ (composite, but still a $(m,\tfrac12)$ representation)
photon$0$ ($< 10^{-18}\,\mathrm{eV}$ [Navas:2024])$h = \pm1$$2$ [Beth:1936]
Higgs boson\(125.2\,\mathrm{GeV}\) [Navas:2024]$s = 0$$1$

The wave equations of the two preceding chapters are the field-theoretic realizations of these representations: the Klein–Gordon equation (The Klein–Gordon Equation) is nothing but the Casimir statement \(\hat{P}^{2} = m^{2}c^{2}\) Equation (94.14) written on wavefunctions, and the Dirac equation (The Dirac Equation) realizes the \((m,\tfrac12)\) representation on a field carrying the parity-paired \((\tfrac12,0)\oplus(0,\tfrac12)\) representation of the Lorentz group. Wigner's 1939 paper [Wigner:1939] thus stands at the head of everything Part XI builds: quantum field theory is the machinery that makes interactions compatible with this classification, and the Standard Model is a specific, experimentally determined list of which representations Nature uses.