Wigner's classification of the unitary irreducible representations of the Poincaré group

Intuition Beginner

A particle moving through spacetime carries no forces of its own; what it does carry is a way of responding when you change your point of view. Shift the origin of time or space, boost to a moving frame, or rotate your laboratory, and the particle's quantum state has to transform consistently. The collection of all such changes of viewpoint that preserve the structure of special relativity is the Poincaré group. Wigner's discovery was that an elementary particle simply is one irreducible way of representing this group on a space of quantum states. There is nothing else to a free elementary particle beyond how it answers the question "what happens to me when you change frames?"

Why does this matter? Before Wigner, "mass" and "spin" were properties you measured and then tried to fit into equations by hand. Wigner turned this around. He showed that the only labels an elementary particle can carry are exactly the two numbers that survive every change of relativistic frame: a mass and a spin. Mass measures the length of the energy-momentum vector. Spin measures the leftover rotation a particle feels once you have used up the freedom to boost it to a standard frame. Everything else about how the particle transforms is fixed once these two numbers are chosen.

The everyday picture is a spinning ball you carry on a train. The train's motion changes where the ball is and how fast it moves, but the rate at which the ball spins about its own centre is something all observers agree on. Mass is the analogue of the ball's rest energy; spin is the analogue of that frame-independent rotation. Wigner's theorem says these two are the whole story for a free particle. This classification builds toward all of relativistic quantum theory.

Visual Beginner

A schematic of energy-momentum space, with the energy axis pointing up and the three momentum directions compressed into a horizontal plane. Three surfaces appear. The upper sheet of a hyperboloid is the set of energy-momentum vectors of fixed mass $m$: a boost slides a point along it without leaving it. A cone touching the origin holds the energy-momentum vectors of a massless particle, like light. A single dot at the origin marks the vacuum, with no energy and no momentum, which looks the same to everyone. Arrows along each surface show that a Lorentz boost moves a particle's energy-momentum vector but keeps it on its own surface.

The picture captures the central idea: each surface is one orbit of the Lorentz group acting on energy-momentum space, and each orbit, together with one extra rotational label, names one kind of elementary particle.

Worked example Beginner

Take a single massive particle of mass $m$ and check that its energy-momentum vector stays on one fixed surface no matter how you boost it, while a massless one stays on a different surface. Use units where the speed of light is $1$.

Step 1. Energy-momentum is a four-component object $p=(E,p_x,p_y,p_z)$. Relativity assigns it the combination $E^2-p_x^2-p_y^2-p_z^2$, and this combination equals $m^2$ for a particle of mass $m$. So a mass-$m$ particle lives on the surface $E^2-p_x^2-p_y^2-p_z^2=m^2$.

Step 2. Pick a concrete case: $m=5$, and a particle momentarily at rest, so $p_x=p_y=p_z=0$ and $E=5$. Check: $5^2-0-0-0=25=m^2$. The point sits on the surface.

Step 3. Now boost the particle to momentum $p_x=12$, with the other two momentum components still zero. Relativity forces the energy to grow so the combination is preserved: $E^2-12^2=25$, giving $E^2=169$, so $E=13$. Check: $13^2-12^2=169-144=25=m^2$. The boosted point still sits on the same surface.

Step 4. Repeat for a massless particle, $m=0$. Its surface is $E^2-p_x^2-p_y^2-p_z^2=0$, the light cone. A photon with $p_x=7$ and the other momenta zero has $E^2=49$, so $E=7$. Boost it to $p_x=20$: now $E=20$. In every case $E=|p|$, and the point stays on the cone.

Step 5. Compare. The massive particle can be brought to rest ($p=0$, $E=m$); the massless one can never be brought to rest, since $E=0$ would force $p=0$ and erase the particle. This single fact is why massive and massless particles get classified by different leftover-rotation groups.

What this tells us: a particle's mass is the fixed length of its energy-momentum vector, and that length is the same for all observers. Boosting moves a particle along its surface but never off it. The massive surface has a special "rest" point; the massless surface does not. These two geometric facts are exactly what split Wigner's classification into a massive case and a massless case.

Check your understanding Beginner

Formal definition Intermediate+

Minkowski spacetime is ${\mathbb{R}}^{1,3}$ with the metric $\eta =\mathrm{diag}(+1,-1,-1,-1)$, giving the indefinite form $⟨x,x⟩=(x^0{)}^2-(x^1{)}^2-(x^2{)}^2-(x^3{)}^2$. The Lorentz group $\mathrm{O}(1,3)$ is the group of linear maps preserving $\eta$. Its identity component, the proper orthochronous Lorentz group ${\mathrm{SO}}^{+}(1,3)$, consists of the Lorentz transformations with determinant $+1$ that preserve the direction of time. The Poincaré group is the semidirect product $$ \mathcal{P} = \mathbb{R}^{1,3} \rtimes \mathrm{SO}^+(1,3), $$ with elements $(a,\Lambda )$ acting on $x\in {\mathbb{R}}^{1,3}$ by $(a,\Lambda )\cdot x=\Lambda x+a$ and composition $(a_1,{\Lambda }_1)(a_2,{\Lambda }_2)=(a_1+{\Lambda }_1{a}_2,{\Lambda }_1{\Lambda }_2)$. The spacetime translations ${\mathbb{R}}^{1,3}$ form an abelian normal subgroup; the Lorentz transformations act on them by the defining linear action.

The quantum-mechanical relevance forces a covering. By Wigner's theorem, symmetries act on a Hilbert space only up to phase, so the relevant group is the universal cover, with $\mathrm{SL}(2,\mathbb{C})$ replacing ${\mathrm{SO}}^{+}(1,3)$ via the two-to-one homomorphism $\mathrm{SL}(2,\mathbb{C})\to {\mathrm{SO}}^{+}(1,3)$. The covering Poincaré group is $\tilde{\mathcal{P}}={\mathbb{R}}^{1,3}⋊\mathrm{SL}(2,\mathbb{C})$. Genuine and projective representations of $\mathcal{P}$ correspond to ordinary representations of $\tilde{\mathcal{P}}$.

The Lie algebra $\mathfrak{p}$ has generators $P^{\mu }$ (translations, $\mu =0,1,2,3$) and $M^{\mu \nu }=-M^{\nu \mu }$ (Lorentz), with relations $$ [P^\mu, P^\nu] = 0, \quad [M^{\mu\nu}, P^\rho] = i(\eta^{\nu\rho} P^\mu - \eta^{\mu\rho} P^\nu), $$ $$ [M^{\mu\nu}, M^{\rho\sigma}] = i(\eta^{\nu\rho} M^{\mu\sigma} - \eta^{\mu\rho} M^{\nu\sigma} - \eta^{\nu\sigma} M^{\mu\rho} + \eta^{\mu\sigma} M^{\nu\rho}). $$ The translations form an abelian ideal; the $M^{\mu \nu }$ span $\mathfrak{so}(1,3)$ acting on the $P^{\mu }$ as on a vector.

The centre of the enveloping algebra is generated by two Casimir invariants. The first is $P^2={\eta }_{\mu \nu }{P}^{\mu }{P}^{\nu }=(P^0{)}^2-{\mathbf{P}}^2$, whose value on an irreducible representation is the scalar $m^2$, the mass-squared. The second is built from the Pauli–Lubanski vector $$ W^\mu = -\tfrac{1}{2},\varepsilon^{\mu\nu\rho\sigma}, P_\nu M_{\rho\sigma}, $$ with ${\epsilon }^{0123}=+1$ the totally antisymmetric symbol. The invariant $W^2={\eta }_{\mu \nu }{W}^{\mu }{W}^{\nu }$ commutes with all of $\mathfrak{p}$; on a massive irreducible representation it takes the value $-m^{2}s(s+1)$, where $s$ is the spin. These two scalars $(m^2,s)$ are the labels Wigner's classification produces.

A unitary representation of $\tilde{\mathcal{P}}$ is a strongly continuous homomorphism $U:\tilde{\mathcal{P}}\to \mathcal{U}(\mathcal{H})$ into the unitary group of a Hilbert space. It is irreducible when the only closed $U$-invariant subspaces are $\{0\}$ and $\mathcal{H}$. Two are unitarily equivalent when an isometry intertwines them. Wigner's classification is the complete list, up to equivalence, of the irreducible such representations.

Counterexamples to common slips

• The Poincaré group is not a direct product. The boost-and-translate composition $(a_1,{\Lambda }_1)(a_2,{\Lambda }_2)=(a_1+{\Lambda }_1{a}_2,{\Lambda }_1{\Lambda }_2)$ rotates the second translation by ${\Lambda }_1$ before adding it. Treating translations and Lorentz transformations as commuting loses the entire orbit structure on which the classification rests.
• The little group is not the full stabiliser inside $\mathrm{O}(1,3)$ in general; one works in the identity component (or its cover). For a massive standard momentum $\bar{p}=(m,0,0,0)$ the stabiliser in ${\mathrm{SO}}^{+}(1,3)$ is $\mathrm{SO}(3)$, and in the cover it is $\mathrm{SU}(2)$, which is what makes half-integer spin admissible.
• A boost is not a rotation. The massless little group $\mathrm{ISO}(2)={\mathbb{R}}^2⋊\mathrm{SO}(2)$ contains two non-compact "translation" generators in addition to the compact rotation. Demanding that these non-compact generators act by the identity is an extra physical input; representations where they act by a non-identity action carry a continuous "infinite-spin" label and are not realised by known particles.
• The label $s$ is the spin, but the dimension of the massive representation as an $\mathrm{SU}(2)$-module is $2s+1$, realised on the internal index, not on the Hilbert space, which is infinite-dimensional because momentum ranges over the whole mass shell.

Key theorem with proof Intermediate+

Theorem (Wigner's classification; Wigner 1939, Mackey machine form). Let $\tilde{\mathcal{P}}={\mathbb{R}}^{1,3}⋊\mathrm{SL}(2,\mathbb{C})$. Suppose the action of $\mathrm{SL}(2,\mathbb{C})$ on the dual $\widehat{{\mathbb{R}}^{1,3}}\cong {\mathbb{R}}^{1,3}$ of momentum characters is regular. Then every irreducible unitary representation of $\tilde{\mathcal{P}}$ is unitarily equivalent to one induced from a pair $(\mathcal{O},\sigma )$, where $\mathcal{O}$ is an orbit of $\mathrm{SL}(2,\mathbb{C})$ on momentum space and $\sigma$ is an irreducible unitary representation of the little group $L_{\bar{p}}$ stabilising a chosen standard point $\bar{p}\in \mathcal{O}$. Inequivalent pairs give inequivalent representations. The orbits and their little groups are:

1. $m^2>0$, $p^0>0$ (forward mass hyperboloid): $L_{\bar{p}}=\mathrm{SU}(2)$, $\sigma$ the spin-$s$ irrep, $s\in \{0,\frac{1}{2},1,\dots \}$;
2. $m^2=0$, $p^0>0$ (forward light cone minus origin): $L_{\bar{p}}=\widetilde{\mathrm{ISO}(2)}$; finite-dimensional $\sigma$ are characters of the $\mathrm{SO}(2)$ factor, labelled by helicity $h\in \frac{1}{2}\mathbb{Z}$;
3. $m^2<0$ (tachyonic single-sheet hyperboloid): $L_{\bar{p}}=\mathrm{SU}(1,1)\cong \mathrm{SL}(2,\mathbb{R})$;
4. $p=0$ (origin): $L_{\bar{p}}=\mathrm{SL}(2,\mathbb{C})$ itself; the zero-momentum representations, including the vacuum.

Proof. Write $G=A⋊H$ with $A={\mathbb{R}}^{1,3}$ abelian and $H=\mathrm{SL}(2,\mathbb{C})$. The argument is the Mackey machine in four steps: diagonalise $A$, decompose by $H$-orbits, identify each fibre as induced from the little group, and verify irreducibility and exhaustiveness.

Step 1: spectral decomposition over the dual of $A$. The restriction $U{|}_A$ is a unitary representation of the abelian group $A={\mathbb{R}}^{1,3}$. By the spectral theorem (the SNAG theorem for locally compact abelian groups), $U{|}_A$ is diagonalised by a projection-valued measure $E(\cdot )$ on the Pontryagin dual $\hat{A}\cong {\mathbb{R}}^{1,3}$, the momentum space, with $U(a,I)={\int }_{\hat{A}}{e}^{i⟨p,a⟩}dE(p)$. The generators $P^{\mu }$ are the spectral coordinates: $P^{\mu }=\int p^{\mu }dE(p)$. Thus $\mathcal{H}\cong {\int }_{\hat{A}}^{\oplus }{\mathcal{H}}_{p}d\mu (p)$ for a measure $\mu$ and measurable field of Hilbert spaces.

Step 2: the Lorentz action permutes the spectrum. The semidirect relation $U(0,\Lambda )U(a,I)U(0,\Lambda {)}^{-1}=U(\Lambda a,I)$ forces the projection-valued measure to be covariant: $U(0,\Lambda )E(S)U(0,\Lambda {)}^{-1}=E(\Lambda S)$ for Borel $S\subseteq \hat{A}$. Hence the support of $\mu$ is $H$-invariant. When $U$ is irreducible, the system $(U{|}_A,U{|}_H)$ is a transitive system of imprimitivity, so $\mu$ is supported on a single $H$-orbit $\mathcal{O}\subseteq \hat{A}$, and $\mu$ is the unique (up to equivalence) $H$-quasi-invariant measure class on $\mathcal{O}$. The orbits of $\mathrm{SL}(2,\mathbb{C})$ acting through ${\mathrm{SO}}^{+}(1,3)$ on $\hat{A}$ are exactly the level sets of $P^2=(p^0{)}^2-{\mathbf{p}}^2$ together with the sign of $p^0$, enumerated as the four families above.

Step 3: identification with an induced representation. Fix a standard momentum $\bar{p}\in \mathcal{O}$ and let $L=L_{\bar{p}}=\{h\in H:h\cdot \bar{p}=\bar{p}\}$ be the little group. Choose a Borel section $q:\mathcal{O}\to H$ with $q(p)\cdot \bar{p}=p$ (a "boost to $p$"). Mackey's imprimitivity theorem states that a representation of $G$ carrying a system of imprimitivity based on the homogeneous space $\mathcal{O}=H/L$ is unitarily equivalent to one induced from a representation of the isotropy group $A⋊L$. The character ${\chi }_{\bar{p}}(a)=e^{i⟨\bar{p},a⟩}$ of $A$ is $L$-invariant, so $A⋊L$ carries the representation ${\chi }_{\bar{p}}\otimes \sigma$ for any unitary irrep $\sigma$ of $L$. The induced representation $U^{(\mathcal{O},\sigma )}={\mathrm{Ind}}_{A⋊L}^G({\chi }_{\bar{p}}\otimes \sigma )$ acts on $L^2(\mathcal{O},\mu ;V_{\sigma })$ by $$ \bigl(U^{(\mathcal{O},\sigma)}(a,\Lambda),\psi\bigr)(p) = e^{i\langle p, a\rangle},\sqrt{\tfrac{d\mu(\Lambda^{-1}p)}{d\mu(p)}};\sigma!\bigl(W(\Lambda, p)\bigr),\psi(\Lambda^{-1}p), $$ where $W(\Lambda ,p)=q(p{)}^{-1}\Lambda q({\Lambda }^{-1}p)\in L$ is the Wigner rotation, the residual little-group element produced by boosting to $p$, applying $\Lambda$, and un-boosting. The square root is the Radon–Nikodym cocycle making the representation unitary.

Step 4: irreducibility and exhaustiveness. The Mackey machine theorem (proved in the Advanced section) states that for a regular semidirect product, $U^{(\mathcal{O},\sigma )}$ is irreducible if and only if $\sigma$ is irreducible, and $U^{(\mathcal{O},\sigma )}\cong U^{({\mathcal{O}}^{\prime },{\sigma }^{\prime })}$ if and only if $\mathcal{O}={\mathcal{O}}^{\prime }$ and $\sigma \cong {\sigma }^{\prime }$ (after conjugating standard points). Step 2 shows every irreducible $U$ has its spectrum on one orbit, and Step 3 realises it as an induced representation; the imprimitivity theorem guarantees nothing else exists. The four orbit families with their little groups complete the list. The compact massive little group $\mathrm{SU}(2)$ has the spin-$s$ irreps; the non-compact massless little group, once its translation generators are required to act by the identity, reduces to $\mathrm{SO}(2)$ with helicity characters. $\square$

Bridge. Wigner's classification builds toward all of relativistic quantum field theory, and the foundational reason it works is exactly the semidirect-product structure of the Poincaré group: an abelian translation factor whose dual is momentum space, acted on by a Lorentz factor whose orbits are the mass shells. This is exactly the non-relativistic free-particle story of 09.03.03, where $E(3)={\mathbb{R}}^3⋊\mathrm{SO}(3)$ had spherical orbits and an $\mathrm{SO}(2)$ little group; the relativistic case generalises the orbits to mass hyperboloids and light cones and the little group to $\mathrm{SU}(2)$ or $\mathrm{ISO}(2)$. The central insight is that the two Casimir invariants $P^2$ and $W^2$ are exactly the orbit label and the little-group label, so mass and spin are the complete set of relativistic quantum numbers. This pattern — diagonalise the abelian factor, decompose by orbits, induce from the stabiliser — is the bridge connecting the Mackey machine to the physicist's particle table, and it appears again in the covariant field realisations of Bargmann and Wigner. Putting these together, "particle = irrep" is not a slogan but a theorem: the imprimitivity theorem identifies every irreducible unitary representation of the Poincaré group with a single mass shell carrying a single little-group representation.

Exercises Intermediate+

Lean formalization Intermediate+

Mathlib has $L^2$ spaces, the Fourier transform on ${\mathbb{R}}^n$, and Stone's theorem, but no Poincaré-group representation theory. The intended formalisation reads schematically:

import Mathlib.Analysis.Fourier.FourierTransform
import Mathlib.MeasureTheory.Function.LpSpace
import Mathlib.GroupTheory.SemidirectProduct

/-- The proper orthochronous Lorentz group SO^+(1,3) (placeholder). -/
structure LorentzGroup where
  toMatrix : Matrix (Fin 4) (Fin 4) ℝ
  preservesEta : sorry
  detOne : sorry
  orthochronous : sorry

/-- The covering Poincaré group ℝ^{1,3} ⋊ SL(2,ℂ). -/
abbrev PoincareCover := SemidirectProduct (Fin 4 → ℝ) (SpecialLinearGroup (Fin 2) ℂ) sorry

/-- A massive Wigner representation: orbit = mass-m hyperboloid,
    little group SU(2), spin label s. -/
noncomputable def wignerMassive (m : ℝ) (hm : 0 < m) (s : ℕ) :
    PoincareCover → (massShell m →₂[μ] su2Module s) →L[ℂ] (massShell m →₂[μ] su2Module s) :=
  sorry  -- induced representation with Wigner-rotation cocycle

/-- Wigner's classification: every irreducible unitary rep is induced
    from (orbit, little-group irrep). -/
theorem wigner_classification :
    ∀ (U : UnitaryRep PoincareCover), Irreducible U →
      ∃ (O : LorentzOrbit) (σ : LittleGroupIrrep O), U ≃ᵤ inducedRep O σ :=
  sorry  -- Mackey imprimitivity theorem

The proof gap is large. Mathlib needs the indefinite Minkowski form, ${\mathrm{SO}}^{+}(1,3)$ and its $\mathrm{SL}(2,\mathbb{C})$ cover as Lie groups, the Pauli–Lubanski vector and the two Casimirs, the orbit decomposition of momentum space into mass shells, the little groups, the SNAG theorem for ${\mathbb{R}}^{1,3}$, and the Mackey imprimitivity theorem with its quasi-invariant orbit measures and Radon–Nikodym cocycle. Each is formalisable from the analytic core but unpackaged; the imprimitivity theorem in particular has no Mathlib analogue.

Advanced results Master

Theorem (Mackey machine for regular semidirect products; Mackey 1949–58). Let $G=A⋊H$ with $A$ locally compact abelian, $H$ acting continuously, and the dual action of $H$ on $\hat{A}$ regular (orbits locally closed, a Borel cross-section exists). For an orbit $\mathcal{O}=H\cdot \chi \subseteq \hat{A}$ with little group $L=H_{\chi }$ and an irreducible unitary representation $\sigma$ of $L$, the induced representation $U^{(\mathcal{O},\sigma )}={\mathrm{Ind}}_{A⋊L}^G(\chi \otimes \sigma )$ is irreducible. Every irreducible unitary representation of $G$ is equivalent to some $U^{(\mathcal{O},\sigma )}$, and $U^{(\mathcal{O},\sigma )}\cong U^{({\mathcal{O}}^{\prime },{\sigma }^{\prime })}$ if and only if $\mathcal{O}={\mathcal{O}}^{\prime }$ and $\sigma \cong {\sigma }^{\prime }$.

This is the engine. Wigner's 1939 result is its application to $G=\mathcal{P}$, predating the general theorem; Mackey's imprimitivity theorem (1949–53) and his theory of induced representations of locally compact groups (1952) supplied the proof that the construction is both irreducible and exhaustive. Folland Ch. 6 §6.6 gives a complete modern account.

Theorem (massive orbits and spin; Wigner 1939). For $m>0$ and $p^0>0$, the orbit is the forward hyperboloid ${\mathcal{O}}_m^{+}=\{p:P^2=m^2,p^0>0\}\cong \mathrm{SL}(2,\mathbb{C})/\mathrm{SU}(2)$, diffeomorphic to ${\mathbb{R}}^3$ via $\mathbf{p}$. The little group is $\mathrm{SU}(2)$, whose unitary irreps are the spin-$s$ representations of dimension $2s+1$ for $s\in \{0,\frac{1}{2},1,\dots \}$. The irreducible representation $U^{(m,s)}$ acts on $L^2({\mathcal{O}}_m^{+},d{\Omega }_m;{\mathbb{C}}^{2s+1})$ with the Lorentz-invariant measure $d{\Omega }_m=d^3\mathbf{p}/(2p^0)$, via the Wigner rotation. The Casimirs take the values $P^2=m^2$ and $W^2=-m^{2}s(s+1)$.

The Lorentz-invariant measure $d^3\mathbf{p}/(2p^0)$ is the unique (up to scale) ${\mathrm{SO}}^{+}(1,3)$-invariant measure on the mass shell; its invariance is what makes the Radon–Nikodym cocycle in the induced representation a pure phase, so the representation is genuinely unitary without the explicit square-root factor.

Theorem (massless orbits and helicity; Wigner 1939). For $m=0$, $p^0>0$, the orbit is the forward light cone ${\mathcal{O}}_0^{+}=\{p:P^2=0,p^0>0\}\setminus \{0\}$. The little group of $\bar{p}=(1,0,0,1)$ is $\widetilde{\mathrm{ISO}(2)}={\mathbb{R}}^2⋊\widetilde{\mathrm{SO}(2)}$. The finite-dimensional unitary irreps are those in which the ${\mathbb{R}}^2$ ("null rotation") generators act by the identity; they are the characters of $\widetilde{\mathrm{SO}(2)}$, labelled by helicity $h\in \frac{1}{2}\mathbb{Z}$. The representation $U^{(0,h)}$ acts on $L^2$ of the light cone with the invariant measure $d^3\mathbf{p}/(2|\mathbf{p}|)$; the Casimirs are $P^2=0$, $W^2=0$, with $W^{\mu }=h{P}^{\mu }$ proportional to the momentum.

For a massless particle $W^{\mu }$ and $P^{\mu }$ are both null and parallel, so $W^2=0$ regardless of helicity; the helicity is recovered as the proportionality constant $W^{\mu }=h{P}^{\mu }$, an invariant because both sides transform as four-vectors. Photons ($h=\pm 1$) and gravitons ($h=\pm 2$) are massless irreps; parity, when imposed, pairs $\pm h$.

Theorem (the remaining orbits). Two further orbit types complete the list. The tachyonic orbits $P^2=-{\kappa }^2<0$ form a single-sheeted hyperboloid $\cong \mathrm{SL}(2,\mathbb{C})/\mathrm{SU}(1,1)$ with little group $\mathrm{SU}(1,1)\cong \mathrm{SL}(2,\mathbb{R})$; their unitary irreps are infinite-dimensional, and no causal local field is built from them. The orbit $p=0$ has little group all of $\mathrm{SL}(2,\mathbb{C})$; the identity little-group representation is the unique Poincaré-invariant state, the vacuum.

Theorem (Bargmann–Wigner realisation; Bargmann–Wigner 1948). Each massive spin-$s$ irrep $U^{(m,s)}$ is realised on the space of totally symmetric rank-$2s$ spinor fields ${\psi }_{{\alpha }_1\cdots {\alpha }_{2s}}(x)$ satisfying a Dirac-type equation in each index. The resulting covariant wave equations (Klein–Gordon for $s=0$, Dirac for $s=\frac{1}{2}$, Proca for $s=1$) are the field-theoretic incarnations of the abstract induced representation.

The Bargmann–Wigner construction is the dictionary between Wigner's abstract Hilbert-space representation and the physicist's covariant fields. The covariant field carries a finite-dimensional, non-unitary representation of $\mathrm{SL}(2,\mathbb{C})$; the wave equation projects onto the unitary subspace, which is exactly $U^{(m,s)}$. This is why relativistic wave equations exist and why they take the specific forms they do.

Synthesis. Wigner's classification is the foundational reason that "mass" and "spin" are the only labels a free elementary particle carries, and the central insight is that these two labels are exactly the two Casimir invariants $P^2$ and $W^2$ of the Poincaré algebra. Putting these together, the Mackey machine diagonalises the abelian translation factor, the spectrum lands on a single Lorentz orbit, and that orbit's little group supplies the second label; this is exactly the orbit-and-stabiliser structure that identifies the irreducible representation with a mass shell carrying a little-group irrep. The massive case generalises the non-relativistic $E(3)$ story of 09.03.03 with $\mathrm{SU}(2)$ replacing the spherical-orbit little group, and the massless case is dual to it in the precise sense that the compact $\mathrm{SO}(3)$ degenerates to the non-compact $\mathrm{ISO}(2)$ as the mass shell flattens onto the light cone. The bridge between the abstract representation and the covariant fields is the Bargmann–Wigner construction, which builds toward every relativistic wave equation and shows that the Klein–Gordon, Dirac, and Proca equations are projections onto the unitary little-group content. The whole arc — semidirect product, abelian dual, orbit decomposition, induced representation, Casimir labels — is the same machine that classifies the irreducibles of any regular semidirect product, specialised to the symmetry group of spacetime, and it is the rigorous content of the slogan that a particle is an irreducible unitary representation of the Poincaré group.

Full proof set Master

Proposition (uniqueness of the orbit measure on the massive shell). The measure $d{\Omega }_m=d^3\mathbf{p}/(2p^0)$ with $p^0=\sqrt{{\mathbf{p}}^2+m^2}$ on the forward mass hyperboloid ${\mathcal{O}}_m^{+}$ is, up to positive scale, the unique ${\mathrm{SO}}^{+}(1,3)$-invariant Borel measure.

Proof. The hyperboloid ${\mathcal{O}}_m^{+}=\mathrm{SL}(2,\mathbb{C})/\mathrm{SU}(2)$ is a homogeneous space of the Lorentz group with compact stabiliser. A homogeneous space $H/L$ carries a (unique up to scale) $H$-invariant measure precisely when the modular functions satisfy ${\Delta }_H{|}_L={\Delta }_L$; here $H=\mathrm{SL}(2,\mathbb{C})$ and $L=\mathrm{SU}(2)$ are both unimodular, so an invariant measure exists and is unique up to scale. It remains to identify it with $d^3\mathbf{p}/(2p^0)$. Parametrise ${\mathcal{O}}_m^{+}$ by $\mathbf{p}\in {\mathbb{R}}^3$ through $p=(\sqrt{{\mathbf{p}}^2+m^2},\mathbf{p})$. The four-dimensional Lorentz-invariant distribution $\delta (P^2-m^2)\theta (p^0)d^{4}p$ restricts, after integrating out $p^0$ against the delta, to $$ \int \delta!\bigl((p^0)^2 - \mathbf{p}^2 - m^2\bigr)\theta(p^0), dp^0, d^3\mathbf{p} = \int \frac{d^3\mathbf{p}}{2\sqrt{\mathbf{p}^2 + m^2}} = \int \frac{d^3\mathbf{p}}{2 p^0}, $$ using $\delta (f(x))=\sum \delta (x-x_i)/|f^{\prime }(x_i)|$ with $f(p^0)=(p^0{)}^2-{\mathbf{p}}^2-m^2$, whose positive root has $|f^{\prime }|=2p^0$. The left side is manifestly Lorentz-invariant ($d^{4}p$ and $\delta (P^2-m^2)\theta (p^0)$ are each invariant under ${\mathrm{SO}}^{+}(1,3)$), so $d^3\mathbf{p}/(2p^0)$ is the invariant measure. By the uniqueness just established it is the only one up to scale. $\square$

Proposition (irreducibility of the massive representation). For $m>0$ and spin $s$, the induced representation $U^{(m,s)}$ on $L^2({\mathcal{O}}_m^{+},d{\Omega }_m;{\mathbb{C}}^{2s+1})$ is irreducible.

Proof. By the Mackey machine theorem it suffices that $\sigma$, the spin-$s$ representation of the little group $\mathrm{SU}(2)$, is irreducible, which it is. For completeness, the direct argument: let $T$ be a bounded operator commuting with every $U^{(m,s)}(a,\Lambda )$. Commuting with the translations $U^{(m,s)}(a,I)$, which act by multiplication by $e^{i⟨p,a⟩}$ varying over the orbit, forces $T$ to be a decomposable operator, $T={\int }_{{\mathcal{O}}_m^{+}}^{\oplus }T(p)d{\Omega }_m(p)$ with each $T(p)\in \mathrm{End}({\mathbb{C}}^{2s+1})$ — this is the double-commutant statement that the only operators commuting with all multiplications by orbit functions are themselves fibrewise multiplications. Commuting with the Lorentz part then forces $T(\Lambda p)=\sigma (W(\Lambda ,p))T(p)\sigma (W(\Lambda ,p){)}^{-1}$, so the field $p\mapsto T(p)$ is determined by its value $T(\bar{p})$ at the standard point, transported by the transitive Lorentz action. Finally $T(\bar{p})$ commutes with $\sigma (L)=\sigma (\mathrm{SU}(2))$ for every little-group element, and since $\sigma$ is irreducible, Schur's lemma gives $T(\bar{p})=cI$. Hence $T=cI$, and $U^{(m,s)}$ is irreducible. $\square$

Proposition (inequivalence of distinct labels). $U^{(m,s)}\cong U^{(m^{\prime },s^{\prime })}$ requires $m=m^{\prime }$ and $s=s^{\prime }$.

Proof. The Casimir $P^2$ acts on $U^{(m,s)}$ as the scalar $m^2$ and on $U^{(m^{\prime },s^{\prime })}$ as $m^{\prime 2}$; a unitary intertwiner preserves the action of central elements, so $m^2=m^{\prime 2}$, and with $m,m^{\prime }>0$ this gives $m=m^{\prime }$. Given $m=m^{\prime }$, the orbits coincide, and the second Casimir $W^2=-m^{2}s(s+1)$ forces $s(s+1)=s^{\prime }(s^{\prime }+1)$, hence $s=s^{\prime }$ for non-negative half-integers. Equivalently, the restriction to the little group at a standard point recovers $\sigma$, and inequivalent $\sigma$ give inequivalent induced representations by the Mackey machine. $\square$

Proposition (helicity is a Lorentz invariant). On a massless irrep, the relation $W^{\mu }=h{P}^{\mu }$ holds with $h$ a constant, and $h$ is invariant under ${\mathrm{SO}}^{+}(1,3)$.

Proof. On the light cone $P^2=0$ and $P_{\mu }{W}^{\mu }=0$ (the identity of Exercise 3). Two null vectors that are orthogonal in the Minkowski sense are parallel: if $P^2=0$, $W^2=0$ (which holds because $W^2$ is a Casimir vanishing on this orbit, as $W^{\mu }$ is null), and $P_{\mu }{W}^{\mu }=0$, then $W^{\mu }=h{P}^{\mu }$ for some scalar $h$. Under a Lorentz transformation both $W^{\mu }$ and $P^{\mu }$ transform as four-vectors, so the proportionality constant $h$ is unchanged: $W^{\prime \mu }={\Lambda }_{\nu }^{\mu }{W}^{\nu }=h{\Lambda }_{\nu }^{\mu }{P}^{\nu }=h{P}^{\prime \mu }$. Hence $h$ is a Lorentz invariant labelling the representation. $\square$

Proposition (covariant field carries a non-unitary $\mathrm{SL}(2,\mathbb{C})$ representation). The finite-dimensional spinor index of a Bargmann–Wigner field transforms in a finite-dimensional representation of $\mathrm{SL}(2,\mathbb{C})$, and no such representation other than the identity one is unitary.

Proof. $\mathrm{SL}(2,\mathbb{C})$ is a connected non-compact simple Lie group. A finite-dimensional unitary representation of such a group has image in a compact subgroup of $\mathrm{GL}(n,\mathbb{C})$; but the only normal subgroup contained in the kernel making the quotient compact is all of $\mathrm{SL}(2,\mathbb{C})$ itself (the group is simple modulo its centre and non-compact), so any finite-dimensional unitary representation factors through the centre and is the identity on the connected component. Therefore the covariant spinor fields $(\frac{m}{2},\frac{n}{2})$ representations with $(m,n)\ne (0,0)$ are non-unitary. Unitarity is restored only on the infinite-dimensional Hilbert space of the induced representation, where the inner product integrates over the mass shell. This is the structural reason a relativistic wave equation is needed: it cuts the non-unitary finite-component field down to the unitary little-group content $U^{(m,s)}$. $\square$

Connections Master

• Quantum free particle as a representation of $E(3)$ 09.03.03. The non-relativistic free particle is the Mackey-machine classification for $E(3)={\mathbb{R}}^3⋊\mathrm{SO}(3)$, with spherical orbits in momentum space and little group $\mathrm{SO}(2)$ (or $\mathrm{SO}(3)$ at zero momentum). Wigner's classification is the relativistic upgrade: the same induced-representation construction with mass hyperboloids replacing spheres and $\mathrm{SU}(2)$ replacing the spherical little group. Reading the two side by side isolates exactly what relativity changes — the orbit geometry and the little group — while the machine is identical.

• Induced representation 07.01.07. Every Wigner representation is an induced representation ${\mathrm{Ind}}_{A⋊L}^{\mathcal{P}}(\chi \otimes \sigma )$ from the isotropy subgroup to the full Poincaré group. The induction construction, Frobenius reciprocity, and the imprimitivity characterisation of induced representations are the technical core; without induced representations the classification has no construction, only a list.

• Representations of $\mathrm{SU}(2)$ and $\mathrm{SO}(3)$ 07.07.05. The massive little group is $\mathrm{SU}(2)$, and the spin label $s$ is exactly the highest weight of the $\mathrm{SU}(2)$ irrep $\sigma$ in the induced representation. The double cover $\mathrm{SU}(2)\to \mathrm{SO}(3)$ is why half-integer spin is physically admissible: the covering $\mathrm{SL}(2,\mathbb{C})\to {\mathrm{SO}}^{+}(1,3)$ of the Poincaré group restricts on the rest-frame stabiliser to this $\mathrm{SU}(2)$ cover.

• Casimir element 07.06.10. The mass and spin labels are the eigenvalues of the two Casimir invariants $P^2$ and $W^2$ on the irreducible representation. The general principle that central elements of the enveloping algebra act as scalars on irreducibles (a Schur-lemma consequence) is what makes mass and spin well-defined invariants rather than frame-dependent quantities, and the Pauli–Lubanski construction is the specific Casimir adapted to the non-semisimple Poincaré algebra.

• Special relativity — Lorentz transformations 10.05.01. The orbits of the classification are the level sets of the Minkowski invariant $P^2=E^2-{\mathbf{p}}^2$, whose invariance under boosts is the defining content of special relativity. The mass hyperboloid, the light cone, and the tachyonic hyperboloid are exactly the causal structure of Minkowski momentum space, so Wigner's classification reads the particle spectrum directly off the geometry of special relativity.

Historical & philosophical context Master

Eugene Wigner published the classification in On unitary representations of the inhomogeneous Lorentz group, Annals of Mathematics 40 (1939) 149–204 ^{[Wigner 1939]}. The paper introduced the orbit-and-little-group method — what physicists now call the method of induced representations — years before the general theory existed, working out the mass hyperboloid, light cone, tachyonic, and zero-momentum orbits by hand and identifying the little groups $\mathrm{SO}(3)$, $\mathrm{ISO}(2)$, $\mathrm{SO}(2,1)$, and the full Lorentz group. The recognition that the rest-frame stabiliser is $\mathrm{SU}(2)$ in the covering group, hence that spin may be half-integral, supplied the representation-theoretic origin of fermions. The infinite-spin (continuous-spin) representations of the massless little group, which Wigner identified and set aside as unphysical, remain a subject of study.

The general machinery underlying Wigner's hand computation was supplied by George Mackey in a sequence of papers from 1949 to 1958, culminating in the imprimitivity theorem and the theory of induced representations of locally compact groups; the monograph Induced Representations of Groups and Quantum Mechanics (Benjamin 1968) ^{[Mackey 1968]} presents the physical reading. Valentine Bargmann and Wigner, in Group theoretical discussion of relativistic wave equations, Proc. Natl. Acad. Sci. USA 34 (1948) 211–223 ^{[Bargmann 1948]}, built the covariant spinor-field realisations that connect the abstract Hilbert-space representations to the Klein–Gordon, Dirac, and Proca equations, explaining why relativistic wave equations take their observed forms. Sternberg's Group Theory and Physics (CUP 1994) ^{[Sternberg 1994]} presents the classification as the apex of a single arc running from finite-group character theory through the compact-group Peter–Weyl theory to the non-compact Mackey machine, and Weinberg's The Quantum Theory of Fields, Vol. I (CUP 1995) makes the classification the starting axiom of relativistic quantum field theory, defining a particle species as an irreducible representation and constructing fields to realise them.

Bibliography Master

@article{Wigner1939,
  author  = {Wigner, Eugene P.},
  title   = {On unitary representations of the inhomogeneous {L}orentz group},
  journal = {Annals of Mathematics},
  volume  = {40},
  number  = {1},
  year    = {1939},
  pages   = {149--204}
}

@article{BargmannWigner1948,
  author  = {Bargmann, Valentine and Wigner, Eugene P.},
  title   = {Group theoretical discussion of relativistic wave equations},
  journal = {Proceedings of the National Academy of Sciences},
  volume  = {34},
  number  = {5},
  year    = {1948},
  pages   = {211--223}
}

@book{Mackey1968,
  author    = {Mackey, George W.},
  title     = {Induced Representations of Groups and Quantum Mechanics},
  publisher = {W. A. Benjamin},
  year      = {1968}
}

@article{Mackey1952,
  author  = {Mackey, George W.},
  title   = {Induced representations of locally compact groups I},
  journal = {Annals of Mathematics},
  volume  = {55},
  year    = {1952},
  pages   = {101--139}
}

@book{Sternberg1994,
  author    = {Sternberg, Shlomo},
  title     = {Group Theory and Physics},
  publisher = {Cambridge University Press},
  year      = {1994}
}

@book{Weinberg1995,
  author    = {Weinberg, Steven},
  title     = {The Quantum Theory of Fields, Vol. I: Foundations},
  publisher = {Cambridge University Press},
  year      = {1995}
}

@book{Folland1995,
  author    = {Folland, Gerald B.},
  title     = {A Course in Abstract Harmonic Analysis},
  publisher = {CRC Press},
  year      = {1995}
}

@book{Hall2013,
  author    = {Hall, Brian C.},
  title     = {Quantum Theory for Mathematicians},
  publisher = {Springer},
  series    = {Graduate Texts in Mathematics},
  volume    = {267},
  year      = {2013}
}