SU(2) and the Spin Double Cover

The double cover is live engineering, not a curiosity. Every single-qubit gate in a quantum computer is literally an element of $\mathrm{SU}(2)$, and the fact that a $2\pi$ pulse returns $-\mathbb{1}$ is exploited in composite-pulse and geometric (Berry-phase) gates. Neutron and atom interferometers keep testing the spinor sign flip to ever-higher precision, and the same $-1$ underlies the spin-statistics connection that forbids two electrons sharing a state. In condensed matter the classification of topological insulators turns on whether the symmetry group is $\mathrm{SO}(3)$ or its cover, and anyonic statistics generalize spinor two-valuedness to braid groups. At the largest scale the electroweak sector is a gauged $\mathrm{SU}(2)_L$ on left-handed fermion doublets — the abstract cover Cartan wrote down in 1913 becomes a dynamical force, simulated in lattice gauge theory using exactly the $S^3$ parametrization of equation (2).

Picture $\mathrm{SU}(2)$ as the 3-sphere $S^3$ and $\mathrm{SO}(3)$ as what you get by gluing every point $U$ to its antipode $-U$. That antipodal gluing turns $S^3$ into $\mathbb{RP}^3$ and is exactly the covering map $\Phi$: two points sit above every rotation. Now trace the matrices $U(\theta)$ for turning about a fixed axis. As $\theta$ runs $0\to2\pi$ the physical rotation returns to start, but $U$ has travelled from $+\mathbb{1}$ to $-\mathbb{1}$ — half-way around the sphere, an open path — and only at $\theta=4\pi$ does $U$ close its loop. The first figure makes the two-to-one wrapping and the $2\pi$-versus-$4\pi$ distinction one picture.

Nothing near the identity distinguishes the groups: their generators obey the same commutators, so every classical experiment sees only $\mathrm{SO}(3)$. The difference is a bookkeeping of large rotations — which loops are contractible — and it shows up only in the phase of a quantum amplitude. The cleanest way to watch it is the trace of the fundamental spin-$\tfrac12$ representation, $\chi_{1/2}(\theta)=2\cos(\theta/2)$: because the half-angle halves the frequency, it does not repeat with period $2\pi$ but swings to $-2$ at a full turn and back to $+2$ only after two. The second figure plots that curve — the sign flip made a measurable number.

The operational definition is algebraic: $\mathrm{SU}(2)$ is the unitary $2\times2$ matrices of determinant one. Unitarity alone gives the four-parameter $\mathrm{U}(2)$; pinning the determinant removes a phase and leaves the three real parameters of a rotation.

where: $U$ — a $2\times2$ complex matrix; $U^\dagger$ — the conjugate transpose; $\det U$ — the determinant, fixed to $1$ (unimodular). Reading it: three real parameters survive, matching the three angles of a rotation. Solving those two constraints in coordinates shows every element is fixed by one pair of complex numbers on a unit sphere — where the manifold $S^3$ appears explicitly.

where: $a,b$ — complex parameters with $|a|^2+|b|^2=1$, the unit-sphere condition, so $\mathrm{SU}(2)\cong S^3\subset\mathbb{R}^4$. Reading it: $a=x_0+ix_3$, $b=x_2+ix_1$ makes $(x_0,x_1,x_2,x_3)$ a point of the 3-sphere. Exponentiating the generators gives the closed form for a finite rotation — and the tell-tale half-angle. The fact $(\hat n\cdot\vec\sigma)^2=\mathbb{1}$ collapses the series to a sine and a cosine.

where: $\theta$ — rotation angle; $\hat n$ — unit rotation axis; $\vec\sigma=(\sigma_1,\sigma_2,\sigma_3)$ — Pauli matrices. Reading it: the argument carries $\theta/2$, not $\theta$ — the whole reason $2\pi$ lands on $-\mathbb{1}$. The generators $\tfrac12\sigma_i$ close into the same structure constants as angular momentum, so infinitesimally the groups are identical; the anticommutator is the extra $2\times2$ fact that makes (3) collapse.

where: $\sigma_i$ — Pauli matrices, $i=1,2,3$; $\epsilon_{ijk}$ — Levi-Civita symbol; $[\,,\,]$, $\{\,,\,\}$ — commutator and anticommutator. Reading it: the left relation is $\mathfrak{su}(2)\cong\mathfrak{so}(3)$; the right, special to $2\times2$, drives the half-angle. The homomorphism to $\mathrm{SO}(3)$ is conjugation: identify $\vec x\in\mathbb{R}^3$ with $X=\vec x\cdot\vec\sigma$, let $U$ act by $X\mapsto UXU^\dagger$, and trace against $\sigma_i$ to read off the rotation matrix.

where: $R(U)$ — the rotation induced by $U$; $\tfrac12\,\mathrm{tr}(\,\cdot\,)$ — the inner product on traceless Hermitian matrices. Reading it: $U=\mathbb{1}$ gives $R_{ij}=\tfrac12\mathrm{tr}(\sigma_i\sigma_j)=\delta_{ij}$, the identity, as it must. Because $U$ and $-U$ conjugate identically, the map is two-to-one with kernel the two-element centre; quotienting by it gives $\mathrm{SO}(3)$.

where: $\ker\Phi$ — elements mapping to the identity rotation; $\mathbb{Z}_2$ — the cyclic group of order two. Reading it: $\mathrm{SU}(2)$ double-covers because the fibre over each rotation has two points. The global difference is a single homotopy invariant: the 3-sphere has no non-contractible loops, projective 3-space has exactly one non-trivial class.

where: $\pi_1$ — the fundamental group of loops up to homotopy; $S^3,\mathbb{RP}^3$ — the 3-sphere and real projective 3-space. Reading it: this $\mathbb{Z}_2$ is the topological root of the spinor sign, and it equals the kernel above. The irreducibles, built in aq-mp-group-representations, are indexed by a spin label with odd or even dimension as $j$ is integer or half-integer.

where: $D^{(j)}$ — the spin-$j$ irreducible; $j$ — the spin label; half-integer values are genuine on $\mathrm{SU}(2)$ only. Reading it: $j=\tfrac12$ is the fundamental (spinor) representation, dimension $2$. The trace of $D^{(j)}$ depends only on the angle and sums to a compact ratio — the character that diagonalizes the theory in aq-mp-characters-orthogonality-theorems.

where: $\chi_j$ — the character of $D^{(j)}$ at angle $\theta$; $m$ — the weight (eigenvalue of $J_3$), $m=-j,\dots,j$. Reading it: $\chi_j(0)=2j+1$ recovers the dimension, and $\chi_{1/2}(\theta)=2\cos(\theta/2)$ is the curve plotted above.

How the cover behaves in the key regimes and on its special elements:

1 · Why the 2π sign flip is measurable (confidence builder)

A neutron beam is split into two coherent paths. In one arm the spin is rotated by an angle $\phi$ about a fixed axis with a magnetic field; the other arm is left alone. If the $-1$ at $2\pi$ were an unobservable global phase, the recombined intensity would repeat with period $2\pi$. What period does it actually show?

Step 1. Take the incident spin along the field axis, an eigenstate; rotating it by $\phi$ multiplies its amplitude by the diagonal entry of $U(\phi)$, namely $e^{-i\phi/2}$. The unrotated arm contributes $1$, so recombining gives $A(\phi)=\tfrac12\big(1+e^{-i\phi/2}\big)$.

Step 2. The intensity is $I(\phi)=|A(\phi)|^2=\tfrac14\big(2+2\cos\tfrac{\phi}{2}\big)=\cos^2\!\tfrac{\phi}{4}$. The relative phase between the arms is $\phi/2$ — again the half-angle — so interference oscillates at half the rotation rate.

Step 3. At $\phi=0$, $I=1$ (bright); at $\phi=2\pi$, $I=\cos^2\tfrac{\pi}{2}=0$ (a dark fringe — full destructive interference); at $\phi=4\pi$, $I=\cos^2\pi=1$ (bright again): period $4\pi$, not $2\pi$.

Step 4. The $-1$ is thus a relative phase, directly visible as the beam darkening at a single $2\pi$ rotation of one arm. This is exactly the Rauch–Werner result of 1975: minimum at $2\pi$, maximum at $4\pi$, confirming the double cover in the laboratory.

2 · Computing the rotation R(U) explicitly

Take $U=\exp\!\big(-\tfrac{i}{2}\cdot\tfrac{\pi}{2}\,\sigma_3\big)$, a $90^\circ$ "rotation" about $\hat z$. Find $R(U)$ from equation (5), verify it is a proper rotation, and check that $-U$ gives the same $R$.

Step 1. The half-angle law gives $U=\cos\tfrac{\pi}{4}\,\mathbb{1}-i\sin\tfrac{\pi}{4}\,\sigma_3=\mathrm{diag}\big(e^{-i\pi/4},\,e^{+i\pi/4}\big)$, so $U\sigma_1 U^\dagger=\big(\begin{smallmatrix}0&e^{-i\pi/2}\\ e^{+i\pi/2}&0\end{smallmatrix}\big)$.

Step 2. Read off column one via $R_{i1}=\tfrac12\mathrm{tr}(\sigma_i\,U\sigma_1U^\dagger)$: contracting with $\sigma_1$ gives $\cos\tfrac{\pi}{2}=0$, with $\sigma_2$ gives $\sin\tfrac{\pi}{2}=1$, with $\sigma_3$ gives $0$. So column one is $(0,1,0)^{\mathsf T}$; repeating for the rest yields

Step 3. Check it is a proper rotation: the columns are orthonormal so $R^{\mathsf T}R=\mathbb{1}$, and $\det R=0\cdot(0\cdot1-0)-(-1)\cdot(1\cdot1-0)+0=+1$. $R(U)$ is the $90^\circ$ rotation about $\hat z$, and $\det R=+1$ confirms $R\in\mathrm{SO}(3)$.

Step 4. Now $-U=\mathrm{diag}\big(e^{+i3\pi/4},e^{-i3\pi/4}\big)$. Because $R$ depends on $U$ only through the pair $(U,\,U^\dagger)$ and every entry carries one $U$ and one $U^\dagger$, the overall sign cancels: $(-U)\sigma_j(-U)^\dagger=U\sigma_jU^\dagger$. Hence $R(-U)=R(U)$ — the two antipodal matrices $\pm U$ produce the identical rotation, the two-to-one structure made explicit.

3 · Adding two spin-½ particles by characters

Combine two spin-$\tfrac12$ particles. Use the character to decompose $D^{(1/2)}\otimes D^{(1/2)}$, confirm the dimension count, and evaluate the identity at $\theta=\pi/2$.

Step 1. Characters multiply on a tensor product: $\chi_{1/2}(\theta)^2=\big(2\cos\tfrac{\theta}{2}\big)^2=2+2\cos\theta$.

Step 2. Since $\chi_1(\theta)=1+2\cos\theta$ and $\chi_0=1$, their sum $\chi_1+\chi_0=2+2\cos\theta$ matches Step 1. Therefore $D^{(1/2)}\otimes D^{(1/2)}=D^{(1)}\oplus D^{(0)}$ — the spin-$1$ triplet plus the spin-$0$ singlet, the Clebsch–Gordan series of aq-mp-tensor-products-clebsch-gordan.

Step 3. Dimension check: $2\times2=4$ states, and $(2\cdot1+1)+(2\cdot0+1)=3+1=4$. The counts agree: 4 = 3 + 1.

Step 4. Spot-check at $\theta=\pi/2$: left side $\chi_{1/2}^2=(\sqrt2)^2=2$; right side $(1+2\cos\tfrac\pi2)+1=2$. Both equal $2$, confirming $D^{(1/2)}\otimes D^{(1/2)}=D^{(1)}\oplus D^{(0)}$.

The spin double cover is where three strands meet: the classification of Lie-group representations, the $2\times2$ matrices Pauli wrote for the electron, and the geometry of spinors. The person who first saw the spinor representations — years before any physicist needed them — was a French geometer working on pure representation theory.

Texts. Stone and Goldbart, Mathematics for Physics, ch. on Lie groups and $\mathrm{SU}(2)$ · Arfken, Weber and Harris, Mathematical Methods for Physicists, 7th ed., group-theory chapter · Hassani, Mathematical Physics, chapters on Lie groups and their representations · Nakahara, Geometry, Topology and Physics, 2nd ed., §4 and §13 (covering groups and spin) · Sakurai and Napolitano, Modern Quantum Mechanics, ch. 3 (rotations and $\mathrm{SU}(2)$) · Hall, Lie Groups, Lie Algebras, and Representations, 2nd ed. The key numerical anchors are: (1) $R(U)_{ij}=\tfrac12\mathrm{tr}(\sigma_i U\sigma_j U^\dagger)$ reproduces the $\hat z$-rotation matrix of Worked Example 2 with $\det R=+1$ and $R(-U)=R(U)$; (2) $U(2\pi)=-\mathbb{1}$, $U(4\pi)=+\mathbb{1}$ from equation (3); (3) $\dim D^{(j)}=2j+1$, $\chi_{1/2}(\theta)=2\cos(\theta/2)$, and $\chi_{1/2}^2=\chi_1+\chi_0$; (4) $\pi_1(\mathrm{SO}(3))=\mathbb{Z}_2$, $\pi_1(\mathrm{SU}(2))=0$.

Prerequisites. The Rotation Group SO(3) · Lie Groups · Lie Algebras and the Exponential Map · Group Homomorphisms and Quotient Groups · Group Representations. Leads to. Reducible and Irreducible Representations · Tensor Products and Clebsch–Gordan Decomposition · Characters and Orthogonality Theorems.

Ask the tutor: Ask me to walk through the belt trick, to compute $R(U)$ for any axis and angle, or to show why projective representations of $\mathrm{SO}(3)$ force nature onto $\mathrm{SU}(2)$.