Wigner Rotation

A massive spin-1 particle has three states. The $W$ boson has three, the $Z$ has three, the $\rho$ meson has three: $m = -1, 0, +1$ , exactly the $2s+1$ you were promised. The photon is spin-1. The photon has two. The usual answer is that gauge invariance removes the third one. That describes the bookkeeping; it does not explain it, and it leaves you believing the longitudinal photon is there, lurking, merely unobservable. In 1939 Eugene Wigner showed this is exactly backwards. His question was: what is a particle? His answer: an irreducible representation of the Poincaré group, whose internal states are labelled by the subgroup of Lorentz transformations that leave its four-momentum alone — the little group. For a massive particle that subgroup is $SO(3)$ , rotations in the rest frame, with representations of dimension $2s+1$ . A massless particle has no rest frame, and its little group is not $SO(3)$ at all. It is $ISO(2)$ , the symmetry group of a Euclidean plane — and $ISO(2)$ does not have the representations $SO(3)$ has.

The third photon state was never gauged away. It was never there.

Key takeaways

Signature $(+,-,-,-)$ , $p\cdot p = +m^2c^2$ , timelike positive. Reference momenta: massive $k = (mc,\,0,0,0)$ ; massless $k = (E/c)(1,\,0,0,1)$ .
The owned object — the LITTLE GROUP. For a chosen standard boost $L(p)$ taking $k$ to $p$ , and any Lorentz $\Lambda$ , define $\;\boxed{W(\Lambda,p) = L^{-1}(\Lambda p)\ \Lambda\ L(p)}$ . Then $W k = k$ in one line: $W$ is in the subgroup that fixes the reference momentum. That subgroup is the little group, and it is what a particle's spin states actually transform under: $\;U(\Lambda)|p,\sigma\rangle = \sum_{\sigma'} D^{(j)}_{\sigma'\sigma}\big(W(\Lambda,p)\big)\,|\Lambda p,\sigma'\rangle$ .
MASSIVE $\Rightarrow$ $SO(3)$ $\Rightarrow$ $2s+1$ states. The generators annihilating $k = (mc,0,0,0)$ are exactly the three rotations. Computed, not asserted.
MASSLESS $\Rightarrow$ $ISO(2)$ , and $ISO(2) \neq SO(3)$ . The generators annihilating $k = (E/c)(1,0,0,1)$ are $J_3$ and two commuting "translations" $T_1 = K_1 - J_2$ , $T_2 = K_2 + J_1$ , with $[T_1,T_2] = 0$ , $[J_3,T_1] = T_2$ , $[J_3,T_2] = -T_1$ . That abelian ideal is the whole story: $\mathfrak{so}(3)$ is simple and has no such ideal, so the two algebras are not isomorphic, and no amount of taking limits turns one into the other.
⚠ THE ONE SOFT STEP, AND IT IS AN EMPIRICAL INPUT, NOT A THEOREM. The two translation generators are non-compact; their unitary representations either carry a continuous internal label (Wigner's "continuous-spin" particles) or act trivially. We impose that they act trivially because no continuous-spin particle has ever been seen. That is physics, not mathematics, and this page says so.
The consequence. With $T_1, T_2$ trivial only $J_3$ survives, so the state carries a single label — the helicity $h = \pm s$ , the two values exchanged by parity. A massless particle of any spin has exactly TWO states, not $2s+1$ . Photon: 2, not 3. Graviton: 2, not 5.
The bridge to canon. For a massive particle and pure boosts, $W(\Lambda,p)$ is the Thomas–Wigner rotation of `aq-sr-thomas-wigner-rotation` — confirmed here as $\Omega = 12.680383^\circ$ at $\beta_1 = \beta_2 = 0.6$ , $\theta = 90^\circ$ , from an entirely different construction. Its formula is recalled, not re-derived.
Key idea: a particle is not a little ball with properties attached. It is a representation, and what it can be is dictated by the geometry of the momentum it carries. Give it a rest frame and you get rotations and $2s+1$ states. Take the rest frame away and the geometry changes character — the sphere becomes a plane — and two of the photon's three states go with it.

The core insight

"How many polarisation states does a particle have?" is not a question about the particle. It is a question about the stabiliser of its four-momentum — the subgroup of Lorentz transformations that leave $p^\mu$ untouched and are therefore free to shuffle the internal labels. Massive and massless momenta have stabilisers of different shape: one compact ( $SO(3)$ , the rotations of a sphere), the other not ( $ISO(2)$ , the symmetries of a plane). Everything downstream — $2s+1$ versus $2$ , why light has two polarisations, why the graviton has two and not five — is a corollary of that one geometric fact.

Where this lives today

Every quantum field theory is built on this classification. A field's physical polarisation count is not a modelling choice — it is fixed by the little group of the momentum, and a field's extra components are gauge redundancy precisely because the little group has no labels for them. Gauge symmetry is unavoidable for a massless spin-1 field because a four-component vector field must somehow describe a two-state particle.

It is also why the graviton has two polarisations, and LIGO sees exactly two. A massive spin-2 particle would have $2s+1 = 5$ states; the graviton is massless, its little group is $ISO(2)$ , and it has two — the $+$ and $\times$ modes. That the observed gravitational-wave strain has exactly two independent polarisations is not a happy accident of Einstein's equations. It follows from the momentum being null, and would hold for any massless spin-2 field in any theory.

The physical picture

Fix a mass. The four-momenta available to a particle then lie on a fixed surface in momentum space — the mass shell $p\cdot p = m^2c^2$ , a hyperboloid — and every Lorentz transformation slides $p$ around on it without ever leaving it. A particle's states therefore come in a family: one copy of "the internal states" for each point $p$ of the shell. But which internal states? To even ask the question you must first agree what "spin up" means at each $p$ , and there is no canonical answer. So you make a choice: pick a reference momentum $k$ , pick a standard boost $L(p)$ carrying $k$ to $p$ , and define the states at $p$ to be the states at $k$ , dragged over by $L(p)$ . Now apply a Lorentz transformation $\Lambda$ . It takes the particle from $p$ to $\Lambda p$ — but the state you land on is not the one the standard convention assigns to $\Lambda p$ , because $\Lambda L(p)$ and $L(\Lambda p)$ are two different routes from $k$ to $\Lambda p$ . Undo the standard one and see what is left:

W(\Lambda,p) \;=\; L^{-1}(\Lambda p)\;\Lambda\;L(p) .

This takes $k$ to $p$ , then to $\Lambda p$ , then back to $k$ . It is a round trip. It ends where it started, so it leaves $k$ alone — and that is the entire construction. The transformation $W$ does not move the momentum at all. All it can do is stir the internal labels, and how it stirs them is the whole content of "spin". The subgroup of Lorentz transformations that leave a given four-momentum invariant is the little group of that momentum. Wigner's insight is that the little group, and not the Lorentz group, is what a particle's spin states represent. And here is why the massless case differs, in one sentence. For a massive particle you can boost to rest, $k = (mc,0,0,0)$ , and the transformations fixing $k$ are obviously the rotations. A massless particle has no rest frame. Its reference momentum must be null, $k = (E/c)(1,0,0,1)$ ; and the transformations leaving a null vector alone are not the rotations of a sphere. They are the symmetries of a plane — and a plane has translations.

Figure

THE PUNCHLINE — 3 states versus 2, and the reason is the shape of the little group. LEFT: a massive spin-1 particle. Boost to its rest frame; the transformations that fix its four-momentum k = (mc, 0, 0, 0) are the ROTATIONS, SO(3), the symmetry of a sphere. Its representations have dimension 2s+1, so spin 1 gives THREE states: m = −1, 0, +1. RIGHT: a massless spin-1 particle. There is no rest frame; the reference momentum k = (E/c)(1,0,0,1) is NULL, and the transformations fixing it are ISO(2) — the symmetries of a PLANE: one rotation about the momentum direction, plus TWO commuting translations. ISO(2) is not SO(3) and never becomes it: so(3) is simple, iso(2) has an abelian ideal. Requiring the two translations to act trivially — an EMPIRICAL input, since no continuous-spin particle has ever been seen — leaves only the rotation about p̂, whose eigenvalue is the HELICITY. Parity flips it. TWO states, ±s, for any spin. The photon's missing third state was not gauged away. IT WAS NEVER THERE

The core equations

The construction, and the one line that makes it work:

\boxed{\;W(\Lambda, p) \;=\; L^{-1}(\Lambda p)\;\Lambda\;L(p)\;} \qquad\Longrightarrow\qquad W(\Lambda,p)\,k \;=\; L^{-1}(\Lambda p)\,\Lambda\,p \;=\; L^{-1}(\Lambda p)\,(\Lambda p) \;=\; k \qquad (1)

where: $L(p)$ is the chosen standard Lorentz transformation with $L(p)k = p$ . Physical reading: $W$ fixes the reference momentum, and it took one line. Whatever else $W$ does, it cannot move $p$ ; it can only act on the internal labels. $W$ depends on the choice of $L$ , but the physics does not — a different choice conjugates every $W$ by a fixed element, relabelling the states without changing the count. How the states actually transform — Wigner, Ann. Math. 40, 149 (1939):

\boxed{\;U(\Lambda)\,|p,\sigma\rangle \;=\; \sum_{\sigma'} D^{(j)}_{\sigma'\sigma}\big(W(\Lambda,p)\big)\;|\Lambda p,\ \sigma'\rangle\;} \qquad (2)

where: $D^{(j)}$ represents the little group, not the Lorentz group. Physical reading: the momentum transforms the obvious way, $p \to \Lambda p$ ; the spin label transforms under $W$ . This is the definition of spin in relativistic quantum theory, and "how many values can $\sigma$ take?" becomes "what are the unitary irreps of the little group?" — a question with different answers for $m > 0$ and $m = 0$ . The massive little algebra, from $k = (mc,0,0,0)$ :

\{X : Xk = 0\} = \mathrm{span}\{J_1, J_2, J_3\}, \qquad [J_i, J_j] = \varepsilon_{ijk}J_k \qquad\Longrightarrow\qquad \boxed{\;\mathfrak{so}(3) \;\Rightarrow\; 2s+1 \ \text{states}\;} \qquad (3)

where: $J_i$ , $K_i$ are the real $4\times4$ generators. Physical reading: the boosts all move $k$ (they give it momentum); the rotations do not. So the little algebra is exactly $\mathfrak{so}(3)$ — you are simply rotating the particle in its rest frame — and the unitary irreps of $SO(3)$ have dimension $2s+1$ . That is where $2s+1$ comes from, and it comes from having a rest frame. The massless little algebra, from $k = (E/c)(1,0,0,1)$ — and it is a different animal:

T_1 = K_1 - J_2, \qquad T_2 = K_2 + J_1, \qquad J_3 \qquad\text{with}\qquad \boxed{\;[T_1,T_2] = 0, \quad [J_3,T_1] = T_2, \quad [J_3,T_2] = -T_1\;} \qquad (4)

where: $T_1, T_2, J_3$ are the only generators annihilating the null $k$ — a three-dimensional algebra, same dimension as $\mathfrak{so}(3)$ , and not isomorphic to it. Physical reading: this is $\mathfrak{iso}(2)$ , the algebra of the Euclidean plane — two commuting translations and one rotation. The tell is $[T_1,T_2] = 0$ : an abelian ideal. $\mathfrak{so}(3)$ is simple and has no abelian ideal at all, so no change of basis, no limit and no clever redefinition can identify the two. The photon is not a massive particle with a state deleted; it is a representation of a different group.

Deriving the little group

Derivation AThe massive case, and the bridge back to canon — the Wigner rotation IS the Thomas-Wigner rotation

The little group is defined by an invariance condition, so the honest way to find it is to solve that condition, not assert the answer. Write the six real generators of the Lorentz algebra as explicit $4\times4$ matrices, act with each on the reference momentum, and compute the null space — the generators that annihilate it. That gives the little algebra as pure linear algebra, with no group theory quoted. For a massive $k$ it recovers $\mathfrak{so}(3)$ ; and then, to check that the abstract $W(\Lambda,p)$ is the concrete object canon already owns, we compute it for two boosts and compare against a number `aq-sr-thomas-wigner-rotation` has already published.

Step 1 — the generators, and the one sign that matters. With $(E_{ab})_{cd} = \delta_{ac}\delta_{bd}$ , take $K_i = E_{0i} + E_{i0}$ (boosts) and $J_z = E_{21} - E_{12}$ (rotations, pinned by their physical action: $\exp(\theta J_z)$ carries $\hat{\mathbf x}$ toward $+\hat{\mathbf y}$ ). These are real, they satisfy $X^{\mathsf T}\eta + \eta X = 0$ , and they generate the group. Why: the rotations close, $[J_i,J_j] = +\varepsilon_{ijk}J_k$ ; the boosts do not, $[K_i,K_j] = -\varepsilon_{ijk}J_k$ — the minus is real, not a convention artefact, once $J$ is pinned by its action, and `aq-sr-thomas-wigner-rotation` owns that commutator. It is used here, not re-derived.

Step 2 — annihilate $k = (mc,0,0,0)$ . Act with each of the six generators on $k$ and solve $Xk = 0$ . The boosts fail immediately: $K_i k = mc\,\hat{\mathbf e}_i \neq 0$ — a boost gives a particle at rest a momentum, which is what a boost is for. The rotations all annihilate it, because a rotation cannot move a vector that points along the time axis. The null space is exactly three-dimensional:

\{X : Xk = 0\} = \mathrm{span}\{J_1, J_2, J_3\} = \mathfrak{so}(3).

Why: this is the entire content of "a massive particle has a rest frame". The little group is the rotation group of that frame, its unitary irreps are the familiar spin- $s$ representations, and their dimension is $2s+1$ . The $2s+1$ of every undergraduate quantum-mechanics course is a statement about the stabiliser of a timelike vector.

Step 3 — check that $W$ is really in it, and then bridge. From (1), $W(\Lambda,p)k = k$ for any $\Lambda$ and any $p$ on the shell; building $W$ numerically confirms it to machine precision, and $W$ comes out block-diagonal, $W = \mathrm{diag}(1,R_3)$ with $R_3$ orthogonal, $\det = +1$ — a pure spatial rotation, not one time component touched. Why: that is the only thing a Lorentz transformation is permitted to do to a particle's internal labels.

Step 4 — the bridge. Take $L(p)$ to be the standard pure boost, put the particle at $p_1 = B(\mathbf v_1)k$ with $\mathbf v_1 = 0.6c\,\hat{\mathbf x}$ , and act with $\Lambda = B(\mathbf v_2)$ , $\mathbf v_2 = 0.6c\,\hat{\mathbf y}$ . Then

W = L^{-1}(\Lambda p_1)\,B(\mathbf v_2)\,B(\mathbf v_1) = \mathrm{diag}(1, R_3), \qquad \Omega = \arctan\frac{(R_3)_{yx}}{(R_3)_{xx}} = \mathbf{12.680383^\circ} .

Why: that is canon 1902's number, to every digit it prints. `aq-sr-thomas-wigner-rotation` obtains $12.680383^\circ$ from the half-rapidity formula $\tan(\Omega/2) = \tfrac19$ ; this page obtains it from $L^{-1}(\Lambda p)\Lambda L(p)$ , a construction that never mentions polar decomposition. The Wigner rotation of the little group and the Thomas–Wigner rotation of two boosts are the same object. Canon derives the angle; this page identifies what it is — an element of the little group of a massive particle, which is exactly why it can rotate a spin without changing a momentum. *Verified in verify_aq-sr-wigner-rotation.py: the six generators checked to satisfy $X^{\mathsf T}\eta + \eta X = 0$ ; $[J_i,J_j] = +\varepsilon_{ijk}J_k$ and $[K_i,K_j] = -\varepsilon_{ijk}J_k$ confirmed entrywise; the null space of $X \mapsto Xk$ computed by SVD for $k = (1,0,0,0)$ and found to be exactly $\mathrm{span}\{J_1,J_2,J_3\}$ ; $W(\Lambda,p)k = k$ and $W = \mathrm{diag}(1,R_3)$ with $R_3^{\mathsf T}R_3 = \mathbb{1}$ , $\det R_3 = +1$ ; and $\Omega = 12.680383^\circ$ , agreeing with 1902's $2\arctan\tfrac19$ .*

Go deeper · Derivation B · 3 minThe massless case — ISO(2), not SO(3), and the one step that is physics rather than mathematics

Derivation BThe massless case — ISO(2), not SO(3), and the one step that is physics rather than mathematics

Everything above hinged on a rest frame. A massless particle has none: $p\cdot p = 0$ is invariant, and $(mc,\mathbf 0)$ has $p\cdot p = m^2c^2 > 0$ , so no boost brings a null vector to rest. The reference momentum must itself be null and the little group must be recomputed from scratch. The answer is not a deformation of $SO(3)$ — it is a group of a different kind, and the state count changes discontinuously as a result. This card computes the algebra, proves it is not $\mathfrak{so}(3)$ , and then flags, in the open, the one step that appeals to experiment rather than deduction.

Step 1 — solve $Xk = 0$ for a null $k$ . Take $k = (1,0,0,1)$ (units $E/c = 1$ : energy $1$ , momentum along $\hat{\mathbf z}$ ). Acting with the six generators and computing the null space gives a three-dimensional algebra again, but spanned by different things:

T_1 = K_1 - J_2, \qquad T_2 = K_2 + J_1, \qquad J_3 .

Why: $J_3$ survives because a rotation about the momentum direction leaves a vector along $\hat{\mathbf z}$ alone. The other two are the surprise — mixtures of a boost and a rotation. Neither $K_1$ nor $J_2$ annihilates $k$ alone; only the combination does. Boost along $\hat{\mathbf x}$ and the null vector tips; rotate about $\hat{\mathbf y}$ by just the right amount and it tips back.

Step 2 — compute the commutators, and read the shape.

[T_1, T_2] = 0, \qquad [J_3, T_1] = T_2, \qquad [J_3, T_2] = -T_1 .

Why: that is $\mathfrak{iso}(2)$ , the algebra of the Euclidean group of the plane — two commuting translations and one rotation that mixes them, exactly as translations and rotations behave on a flat sheet of paper. The little group of a massless particle is the symmetry group of a plane.

Step 3 — prove it is not $SO(3)$ in disguise. Both algebras are three-dimensional, so dimension alone proves nothing. But $\mathfrak{iso}(2)$ contains $\mathrm{span}\{T_1,T_2\}$ : a two-dimensional abelian ideal ( $[T_1,T_2] = 0$ , and $[J_3, T_i]$ stays inside the span). And $\mathfrak{so}(3)$ is simple — it has no non-trivial ideal at all, abelian or otherwise, since $[J_1,J_2] = J_3$ escapes any two-dimensional subspace you try. ∎ Why: this is the whole article, in one line of algebra. An isomorphism preserves ideals; $\mathfrak{iso}(2)$ has one and $\mathfrak{so}(3)$ does not, so they are not isomorphic. No massless limit of a massive particle can smoothly become a massless particle's little group. The state count is allowed to jump because the group jumps.

Step 4 — and now the step that is NOT mathematics. $T_1$ and $T_2$ are non-compact generators. In a unitary representation their eigenvalues can be any real numbers, and the general irrep of $ISO(2)$ is labelled by a continuous parameter $\xi \geq 0$ — the "radius" in the translation plane. If $\xi > 0$ the particle carries a continuous internal degree of freedom: Wigner's continuous-spin particles, with an infinite tower of helicities. We set $\xi = 0$ — that is, we require $T_1$ and $T_2$ to act trivially — because no such particle has ever been observed. Why: say this out loud. It is an empirical input, not a theorem. Nothing in the Poincaré group forbids continuous-spin representations; they are perfectly good unitary irreps, and they were in Wigner's 1939 classification. We discard them because the world does not contain them, and honesty about that is worth more than a tidy derivation.

Step 5 — count. With $T_1, T_2$ trivial, the only generator left acting on the internal space is $J_3$ — rotation about the momentum direction — and its eigenvalue is a single number, the helicity $h$ . So the internal space is one-dimensional: one state, one helicity. Parity reverses the momentum but not the angular momentum, so it maps $h \to -h$ , and a parity-invariant theory must contain both. Hence exactly two states, $h = \pm s$ , for any spin. Why: the photon has $h = \pm1$ and nothing else. No $h = 0$ . Massive spin-1 has three states because $SO(3)$ has a three-dimensional irrep; the photon has two because $ISO(2)$ , with the translations killed, has only one-dimensional ones. The count is spin-independent — the graviton has $2$ , not $5$ . And where parity is not a symmetry there is only one state: the Weyl neutrino of the pre-1998 Standard Model. *Verified in verify_aq-sr-wigner-rotation.py: the null space of $X \mapsto Xk$ for $k = (1,0,0,1)$ computed by SVD, found three-dimensional, and shown to be spanned by $\{K_1 - J_2,\ K_2 + J_1,\ J_3\}$ ; $T_ik = 0$ asserted entrywise; $[T_1,T_2] = 0$ , $[J_3,T_1] = T_2$ , $[J_3,T_2] = -T_1$ confirmed; the simplicity of $\mathfrak{so}(3)$ checked by exhausting its two-dimensional subspaces and showing none is an ideal; and the state counts $2s+1$ versus $2$ tabulated.*

Common trapThe one soft step, in the open — we do not derive the photon's two states from mathematics alone

This page proves, by explicit computation, that the little group of a massless momentum is $ISO(2)$ and not $SO(3)$ . That is a theorem, and the verification script checks every commutator.

What is NOT a theorem is the step from $ISO(2)$ to "two states". $ISO(2)$ has perfectly respectable unitary irreps in which the translation generators act non-trivially, labelled by a continuous $\xi > 0$ : massless particles carrying a continuous internal degree of freedom, an infinite ladder of helicities. Wigner classified them in the same 1939 paper. They are not mathematically sick. They are simply not observed, and that is the only reason we throw them away. So the honest chain is: the little group is $ISO(2)$ (proved); if the translations act trivially (imposed, on experimental grounds) the states are labelled by helicity alone (proved); and with parity there are exactly two (proved). Anyone who tells you the photon's two polarisations follow from group theory alone has quietly skipped the middle step.

Scope, likewise. What is derived above is the Lie algebra of the little group, by linear algebra. The full unitary representation theory — induced representations, the measure on the mass shell, the Casimirs $P^2$ and $W^2$ — is stated, not derived. It is Wigner's paper, and it is fifty-six pages long.

Common trapThis article was authored from scratch, because its base was somebody else's physics

The base for this page carries eight equation cards. Seven belong to other articles — its polar decomposition, boost-composition law, exact half-rapidity angle and non-relativistic limit are all `aq-sr-thomas-wigner-rotation`'s owned results, and its precession rate is `aq-sr-thomas-precession`'s. All three of its derivations likewise. Shipping it faithfully would have shipped a duplicate of live canon.

Only one card — the little group, equation (2) — is genuinely this article's, and the base does nothing with it beyond printing it. So the article was rebuilt around it: a base substitution, recorded here rather than hidden.

One card deserves a note rather than a dismissal. Its Ungar formula

\cos\Omega = \frac{(1+\gamma_1+\gamma_2+\gamma_{12})^2}{(1+\gamma_1)(1+\gamma_2)(1+\gamma_{12})} - 1

is correct, and the card does define $\gamma_{12}$ — as "the Lorentz factor of the relative velocity, $\gamma_{12} = \gamma_1\gamma_2(1 - \boldsymbol\beta_1\cdot\boldsymbol\beta_2)$ ." That definition is wrong. The formula is exact only for the composed-velocity factor, $\gamma_{12} = \gamma_1\gamma_2(1 + \boldsymbol\beta_1\cdot\boldsymbol\beta_2)$ — note the plus. At $\boldsymbol\beta_1 = (0.8,0,0)$ , $\boldsymbol\beta_2 = (0.3,0.5,0)$ the true angle is $14.520^\circ$ and the card's own definition returns $17.105^\circ$ : wrong by $17.8\%$ . The defect hides in plain sight because it is invisible for perpendicular boosts — $\boldsymbol\beta_1\cdot\boldsymbol\beta_2 = 0$ makes the two readings coincide, and perpendicular is the case everybody tests. A right equation with a wrong definition is a wrong result.

Limiting cases

Case	Little group	States	Meaning
$m > 0$ , any $s$	$SO(3)$	$\mathbf{2s+1}$	there is a rest frame; the stabiliser of a timelike vector is the rotations
$m = 0$ , any $s$	$ISO(2)$	$\mathbf{2}$ ( $h = \pm s$ )	no rest frame; the stabiliser of a null vector is the plane's symmetries
$T_1, T_2$ act non-trivially	$ISO(2)$ , $\xi > 0$	$\infty$	continuous spin — allowed by the group, never observed
two pure boosts, $m > 0$	$W = R(\Omega)$	—	the Thomas–Wigner rotation. $\Omega = 12.680383^\circ$ at $\beta = 0.6$ , $\theta = 90^\circ$

Worked examples

1 · Count the photon's states — and find out what the W boson ate

A massive spin-1 particle has three polarisation states; the photon has two. Where did the third go, and where does the $W$ boson's third come from?

Step 1 — identify. The $Z$ boson is spin-1 with $m_Z = 91.19\ \mathrm{GeV}/c^2$ . Its reference momentum is timelike, its little group is $SO(3)$ , and the spin-1 irrep has dimension $2s+1 = \mathbf{3}$ : transverse-left, transverse-right, longitudinal.

Step 2 — apply. The photon is spin-1 with $m = 0$ . Its reference momentum is null, its little group is $ISO(2)$ , and with the translations acting trivially the states are labelled by helicity alone: $h = +1$ and $h = -1$ . Two. There is no $h = 0$ — not because it is forbidden by a constraint, but because the little group has no such label to offer.

Step 3 — the boundary number, and it is a counting argument. Now the electroweak bookkeeping. Before symmetry breaking: four massless gauge bosons ( $W^1, W^2, W^3, B$ ), $2$ states each $= \mathbf{8}$ ; plus a complex Higgs doublet, $\mathbf{4}$ real scalars. Total $\mathbf{12}$ . After: three massive bosons ( $W^\pm, Z$ ), $3$ each $= 9$ ; one massless photon, $2$ ; one physical Higgs, $1$ . Total $9 + 2 + 1 = \mathbf{12}$ . ✓

Step 4 — what it means. The $W$ did not conjure a third state out of nowhere. It ate one. Each boson that becomes massive absorbs one degree of freedom from the Higgs doublet — a Goldstone boson — and that absorbed state becomes its longitudinal polarisation. The count is forced: becoming massive means changing your little group from $ISO(2)$ to $SO(3)$ , and $SO(3)$ demands one more state than $ISO(2)$ supplies. The Higgs mechanism is, from here, a bookkeeping device for a change of little group — and the Higgs had to exist because you cannot give a gauge boson a mass without finding it a third state somewhere.

2 · The Wigner rotation of a massive particle IS the Thomas–Wigner rotation

Boost a massive particle to $0.6c$ along $\hat{\mathbf x}$ , then apply a further boost of $0.6c$ along $\hat{\mathbf y}$ . Compute $W(\Lambda,p)$ and see what it is.

Step 1 — identify. $k = (1,0,0,0)$ (units $mc = 1$ ). Take $L(p)$ = the standard pure boost. Then $p_1 = B(\mathbf v_1)k = (1.25,\ 0.75,\ 0,\ 0)$ , and $\Lambda = B(\mathbf v_2)$ gives $\Lambda p_1 = (1.5625,\ 0.75,\ 0.9375,\ 0)$ .

Step 2 — apply (1). $W = L^{-1}(\Lambda p_1)\,\Lambda\,L(p_1)$ , built as an explicit $4\times4$ product. The result is block-diagonal:

W = \begin{pmatrix} 1 & 0 & 0 & 0\\ 0 & 0.97560976 & -0.21951220 & 0\\ 0 & 0.21951220 & 0.97560976 & 0\\ 0&0&0&1 \end{pmatrix}, \qquad Wk = k\ \checkmark

Not one time component is touched. $W$ is a pure spatial rotation about $\hat{\mathbf z}$ — an element of $SO(3)$ , exactly as (3) says it must be.

Step 3 — the boundary number. Read the angle off: $\Omega = \arctan(0.21951220/0.97560976) = \mathbf{12.680383^\circ}$ .

Step 4 — recognise it. That is exactly the number `aq-sr-thomas-wigner-rotation` publishes, by a completely different route: it polar-decomposes $B(\mathbf v_2)B(\mathbf v_1)$ and evaluates its half-rapidity formula to $\tan(\Omega/2) = \tfrac19$ exactly, giving $2\arctan\tfrac19 = 12.680383^\circ$ . Two constructions, no shared step, one number. The Thomas–Wigner rotation is the Wigner rotation of the little group. Canon owns the angle and its formula; this page supplies the identification, and with it the reason the rotation exists at all: $W$ can rotate the spin precisely because it cannot move the momentum, and it cannot move the momentum because it is, by construction, a round trip.

Physics of the portrait

Origins · 2 minEugene Wigner and how this was discovered

Eugene Wigner

1902–1995 · Budapest · Technische Hochschule Berlin · Göttingen · Princeton · the Manhattan Project

He was trained as a chemical engineer, at his father's insistence, and took his doctorate at the Technische Hochschule in Berlin in 1925 on the rates of chemical reactions. He spent his life doing something else entirely. Born in Budapest on 17 November 1902, Wigner belonged to the extraordinary Hungarian generation — von Neumann was a schoolmate — that left Europe in the 1930s and reshaped American physics. At Göttingen he was assistant to Hilbert, and there he began doing the thing he is remembered for: importing group theory into quantum mechanics, over the loud objections of physicists who called it the Gruppenpest, the group pestilence, and wanted nothing to do with it. The 1939 paper is the summit. "On Unitary Representations of the Inhomogeneous Lorentz Group" (Annals of Mathematics 40, 149–204) asks a question nobody had thought to ask so sharply: what is an elementary particle? Wigner's answer — that it is an irreducible unitary representation of the Poincaré group, labelled by mass and spin — is the definition physics still uses. In the same paper he classified the little groups, found $SO(3)$ for massive momenta and $ISO(2)$ for massless ones, and identified the continuous-spin representations that nature appears to have declined to use. He went on to build the first nuclear reactor with Fermi, and to receive the Nobel Prize in 1963 for symmetry principles in physics. He died in Princeton on 1 January 1995. The rotation on this page carries his name because he was the one who understood which group it belongs to.

History

1928

— Wigner begins applying group theory to quantum mechanics, to the vocal irritation of physicists who dub the programme the Gruppenpest.

1939

— Wigner, in Annals of Mathematics 40, 149-204, classifies the unitary irreducible representations of the inhomogeneous Lorentz group. A particle IS such a representation; the little group of its momentum is SO(3) if massive and ISO(2) if massless; the continuous-spin representations are found, and are not seen in nature.

1963

— Wigner receives the Nobel Prize in Physics for his contributions to the theory of the atomic nucleus and elementary particles, particularly through the discovery and application of fundamental symmetry principles.

1964

— Higgs, Brout and Englert show how a gauge boson acquires mass by absorbing a scalar degree of freedom; in little-group language, by acquiring the third state that ISO(2) could not supply.

Going further

Texts. Weinberg, The Quantum Theory of Fields, Vol. I, §2.5 (the standard modern treatment) · Tung, Group Theory in Physics, ch. 10 · Maggiore, A Modern Introduction to QFT, §2.5. Primary: Wigner, Annals of Mathematics 40, 149–204 (1939).

Provenance. Signature $(+,-,-,-)$ ; real $4\times4$ generators; $J_z$ pinned by its physical action ( $\exp(\theta J_z)$ carries $\hat{\mathbf x}$ toward $+\hat{\mathbf y}$ ). Every algebra statement above is recomputed in verify_aq-sr-wigner-rotation.py, which also reads this file and asserts each printed value against it. BASE SUBSTITUTION — authored from scratch: 7 of the base's 8 equation cards and all 3 of its derivations carry results owned elsewhere. Claim-scope: the Lie algebra of the little group is derived; the full unitary representation theory is stated, not derived; and the step from $ISO(2)$ to two helicities rests on an empirical input, flagged in the open.

Scope. The non-closure theorem, the half-rapidity angle, the commutator $[K_i,K_j] = -\varepsilon_{ijk}J_k$ and the value $12.68^\circ$ are owned by `aq-sr-thomas-wigner-rotation` — recalled and confirmed, not re-derived. The precession rate is `aq-sr-thomas-precession`'s. The Higgs mechanism proper belongs to a future particle-physics track; only the state count is used here.

Builds on. The Thomas–Wigner Rotation · Four-Momentum · Causality and the Light Cone. Leads to. Thomas Precession · The Thomas Factor One-Half · BMT.

Figure

The round trip that defines the little group. A particle's momentum p sits on the mass shell (the hyperbola). We never compare spins at different momenta directly — there is no canonical way to. Instead we agree on a STANDARD boost L(p) that carries the reference momentum k to p, and define the states at p to be the states at k, dragged over by L(p). Now hit the particle with a Lorentz transformation Λ: it moves p to Λp. But Λ L(p) and L(Λp) are two DIFFERENT routes from k to Λp. Undo the standard one, and W = L⁻¹(Λp) Λ L(p) is what is left: a round trip k → p → Λp → k. It ENDS WHERE IT STARTED, so it cannot move the momentum — it can only stir the internal labels. That is the little group, and the whole of spin lives in it. For a massive particle W is a pure spatial rotation; for the configuration β₁ = β₂ = 0.6 at right angles it is a rotation by 12.680383° — precisely canon's Thomas-Wigner angle, reached by a completely different route

What does it actually mean to say a particle IS a representation?

It means the list of things a particle can be is exhausted by the ways the symmetries of spacetime can act on it. Fix a state $|p,\sigma\rangle$ , apply every Poincaré transformation, and the states you sweep out are the particle's entire state space. If that space has no proper invariant subspace the representation is irreducible — which is the technical content of "elementary": you cannot split the particle into smaller things that transform among themselves. Wigner's classification then fixes such a representation by two numbers, the Casimirs $P^2$ (mass) and $W^2$ (spin). So "a particle is a mass and a spin" is not a description; it is a theorem about representations, and the little group is the machinery that proves it.

Where exactly does the 2s+1 of ordinary quantum mechanics come from?

From the existence of a rest frame, and from nothing else. If a particle has mass you may boost to $k = (mc,0,0,0)$ , and the Lorentz transformations that leave that momentum alone are exactly the rotations — a boost would give it momentum, and a rotation cannot move a vector pointing along the time axis. So the little group is $SO(3)$ , whose unitary irreps have dimensions $1, 2, 3, \dots = 2s+1$ . That is the whole derivation. It never once mentions the particle — only the shape of the set of transformations that fix a timelike vector. Which is why the answer is the same for an electron, a $W$ boson and a hydrogen atom, and why it fails, discontinuously, the moment the mass is zero.

Is the massless case really a limit of the massive one?

No, and that is the crux. Taking $m \to 0$ does not smoothly deliver the massless theory, because the little group itself changes: $SO(3)$ is compact and simple, $ISO(2)$ is neither, and the state count jumps discontinuously from $2s+1$ to $2$ . Something must happen to the extra states, and in a physical theory it does: as $m\to0$ the longitudinal mode of a massive vector boson decouples — its couplings vanish with $m$ — so the theory has a smooth limit even though the representation does not.

What would a continuous-spin particle actually look like?

Like nothing anyone has ever seen, which is the point. It would be massless and would carry an infinite tower of helicity states $\dots, -1, 0, +1, \dots$ all belonging to the same particle, connected by $T_1$ and $T_2$ , with a continuous parameter $\xi$ setting the scale. Any interaction would excite it into infinitely many internal states. It is not observed, and it is entirely permitted by the Poincaré group — Wigner found it in 1939 and it has been a curiosity ever since. The photon's two polarisations rest on its absence, and that is an experimental fact, not a theorem.

Why is helicity Lorentz invariant for a massless particle but not for a massive one?

Because you cannot overtake light. Helicity is the projection of spin on the direction of motion; to reverse it you must reverse the momentum relative to yourself, which means outrunning the particle. For a massive particle that is easy — boost past it, its momentum flips and its spin does not — so helicity is frame-dependent and cannot label a state invariantly, which is why massive particles carry a spin $s$ and a rest-frame projection $m$ instead. No boost outruns a massless particle, so its helicity is the same in every frame: a good invariant label. That is precisely why $J_3$ , the rotation about $\hat{\mathbf p}$ , is the only surviving generator.

Does the Wigner rotation have anything to do with the Thomas precession?

Everything — they are the same object at two scales, and this track splits them across three articles on purpose. `aq-sr-thomas-wigner-rotation` computes the finite angle between two boosts ( $12.68^\circ$ at $\beta = 0.6$ ) from non-closure. This page shows what that rotation is: an element of the little group of the momentum, which is why it can rotate a spin without touching a four-momentum, and why the same $12.680383^\circ$ falls out of $L^{-1}(\Lambda p)\Lambda L(p)$ . `aq-sr-thomas-precession` then integrates it along a worldline to get the rate. Angle, identity, rate.