Simple Algorithm
Quantum Computing · Rydberg State · 30 min

SU(2), SO(3) and the Double Cover

Why a 360° rotation flips the sign of a spinor.

Rotate an electron by 360∘360^\circ and its state vector comes back with a minus sign. The sign is invisible on the Bloch sphere, yet neutron interferometers measured it in 1975. It is the trace of a two-to-one map from the group of qubit gates onto the group of rotations.

What you’ll learn

  • State and prove that Φ:SU(2)→SO(3)\Phi:SU(2)\to SO(3), Φ(U)jk=12Tr⁡(σjUσkU†)\Phi(U)_{jk}=\tfrac12\Tr(\sigma_jU\sigma_kU^\dagger), is a surjective homomorphism with kernel {±I}\{\pm\id\}.
  • Explain why π1(SO(3))≅Z2\pi_1(SO(3))\cong\Z_2: a 2π2\pi loop of rotations is not contractible, a 4π4\pi loop is.
  • Compute the phase kicked back by a controlled Rn^(2π)R_{\hat n}(2\pi) and connect it to the neutron-interferometry experiments.
  • Decide which spin-jj representations descend to SO(3)SO(3), and state what Bargmann's theorem says about projective representations.

Two groups and the map between them

Identify R3\R^3 with the traceless Hermitian 2×22\times2 matrices via r⃗↦r⃗⋅σ⃗\vec r\mapsto\vec r\cdot\vec\sigma. Lengths are encoded in determinants: det⁡(r⃗⋅σ⃗)=−∣r⃗∣2\det(\vec r\cdot\vec\sigma)=-\abs{\vec r}^2. Conjugation AdU:r⃗⋅σ⃗↦U(r⃗⋅σ⃗)U†\mathrm{Ad}_U:\vec r\cdot\vec\sigma\mapsto U(\vec r\cdot\vec\sigma)U^\dagger preserves Hermiticity, trace and determinant, so it acts on R3\R^3 as a linear isometry.

DerivationProof
  1. Φ(UV)=Φ(U) Φ(V)\Phi(UV)=\Phi(U)\,\Phi(V)

    Homomorphism. Φ(U)\Phi(U) is the matrix of AdU\mathrm{Ad}_U in the basis (σ1,σ2,σ3)(\sigma_1,\sigma_2,\sigma_3), and AdUV=AdUAdV\mathrm{Ad}_{UV}=\mathrm{Ad}_U\mathrm{Ad}_V.

  2. Φ(SU(2))⊆SO(3)\Phi\big(SU(2)\big)\subseteq SO(3)

    Image. AdU\mathrm{Ad}_U preserves −∣r⃗∣2=det⁡(r⃗⋅σ⃗)-\abs{\vec r}^2=\det(\vec r\cdot\vec\sigma), so Φ(U)∈O(3)\Phi(U)\in O(3). SU(2)≅S3SU(2)\cong S^3 is connected, det⁡Φ(U)∈{±1}\det\Phi(U)\in\{\pm1\} depends continuously on UU, and det⁡Φ(I)=1\det\Phi(\id)=1.

  3. Φ(e−iθn^⋅σ⃗/2)=Rn^(θ)\Phi\big(e^{-i\theta\hat n\cdot\vec\sigma/2}\big)=\mathcal R_{\hat n}(\theta)

    Surjective. Rodrigues' formula (Excited State lesson) identifies Φ(Rn^(θ))\Phi(R_{\hat n}(\theta)), and by Euler's rotation theorem every element of SO(3)SO(3) is a rotation about some axis.

  4. ker⁡Φ={±I}  ⟹  SO(3)≅SU(2)/{±I}\ker\Phi=\{\pm\id\}\;\Longrightarrow\;SO(3)\cong SU(2)/\{\pm\id\}

    Kernel. If UσkU†=σkU\sigma_kU^\dagger=\sigma_k for all kk, then UU commutes with I,X,Y,Z\id,X,Y,Z, which span all 2×22\times2 matrices, so U=λIU=\lambda\id. Then det⁡U=λ2=1\det U=\lambda^2=1 forces λ=±1\lambda=\pm1. □\square

Check your understanding

Which rotation of the Bloch sphere does U=iXU=iX represent?

  1. A half turn (π\pi) about x^\hat x
  2. A quarter turn (π/2\pi/2) about x^\hat x
  3. None: iX∉SU(2)iX\notin SU(2) because det⁡(iX)=−1\det(iX)=-1
  4. A reflection through the yy–zz plane
Show the answer

A. A half turn (π\pi) about x^\hat x

Rx(π)=−iXR_x(\pi)=-iX, so iX=−Rx(π)iX=-R_x(\pi). The sign lies in ker⁡Φ\ker\Phi, so Φ(iX)=Rx^(π)\Phi(iX)=\mathcal R_{\hat x}(\pi): the same half turn as the XX gate.

Topology: why 4π is trivial but 2π is not

SU(2)≅S3SU(2)\cong S^3 is simply connected, and Φ\Phi is a two-sheeted covering map. So SO(3)≅S3/{±1}=RP3SO(3)\cong S^3/\{\pm1\}=\R P^3. Concretely, SO(3)SO(3) is the solid ball of rotation vectors θn^\theta\hat n with θ≤π\theta\le\pi, with antipodal boundary points identified, because a half turn about n^\hat n equals a half turn about −n^-\hat n.

DerivationProof by path lifting
  1. γ~(t)=e−iπtZ,γ~(0)=I,γ~(1)=−I\tilde\gamma(t)=e^{-i\pi tZ},\qquad\tilde\gamma(0)=\id,\quad\tilde\gamma(1)=-\id

    Lift γ\gamma to SU(2)SU(2) starting at I\id. Lifts through a covering map are unique, so this is the lift, and it is not closed.

  2. A homotopy of loops based at I∈SO(3)\id\in SO(3) lifts to a homotopy of paths in SU(2)SU(2) with fixed endpoints. The endpoint −I-\id cannot move continuously within the discrete fiber {±I}\{\pm\id\}, so γ\gamma is not null-homotopic.

  3. π1(SO(3))≅ker⁡Φ≅Z2\pi_1\big(SO(3)\big)\cong\ker\Phi\cong\Z_2

    γ\gamma traversed twice lifts to e−2πitZe^{-2\pi itZ}, t∈[0,1]t\in[0,1], a closed loop in S3S^3, which is contractible since π1(S3)=0\pi_1(S^3)=0. In general, for a universal covering, π1\pi_1 of the base is isomorphic to the fiber group. □\square

Predict before you look

Through what total angle (degrees) must you rotate a spin-½ about a fixed axis before its state vector returns to itself exactly, sign included?

Show the answer

720 °

Rn^(θ)=cos⁡θ2 I−isin⁡θ2 n^⋅σ⃗R_{\hat n}(\theta)=\cos\frac\theta2\,\id-i\sin\frac\theta2\,\hat n\cdot\vec\sigma has period 4π4\pi in θ\theta: R(2π)=−IR(2\pi)=-\id and R(4π)=+IR(4\pi)=+\id.

Measuring the sign

On a single qubit, −I-\id is a global phase. Aharonov and Susskind (1967) pointed out that it becomes observable relative to a reference: split a spin's wavefunction, rotate one branch by 2π2\pi, and recombine. In 1975 two neutron-interferometry groups did exactly that, precessing the neutron spin in a magnetic field in one arm (Rauch et al.; Werner et al.). The interference intensity oscillated with period 4π4\pi in the precession angle, not 2π2\pi.

In circuit language the reference is a control qubit. A controlled rotation is ∣0⟩ ⁣⟨0∣⊗I+∣1⟩ ⁣⟨1∣⊗Rn^(θ)\dyad0\otimes\id+\dyad1\otimes R_{\hat n}(\theta), so at θ=2π\theta=2\pi:

C-Rn^(2π)=∣0⟩ ⁣⟨0∣⊗I−∣1⟩ ⁣⟨1∣⊗I=Z⊗IC\text{-}R_{\hat n}(2\pi)=\dyad0\otimes\id-\dyad1\otimes\id=Z\otimes\id
A controlled 2π2\pi rotation is a ZZ gate on the control. The spinor sign is kicked back as a relative phase (phase kickback), whatever the axis and whatever the target state.

Representations: spinors versus vectors

Every irreducible unitary representation of SU(2)SU(2) is a spin-jj representation D(j)D^{(j)} of dimension 2j+12j+1, j∈{0,12,1,32,… }j\in\{0,\tfrac12,1,\tfrac32,\dots\}, with D(j)(e−iθn^⋅σ⃗/2)=e−iθ n^⋅J⃗D^{(j)}(e^{-i\theta\hat n\cdot\vec\sigma/2})=e^{-i\theta\,\hat n\cdot\vec J}. Evaluate at −I=Rz^(2π)-\id=R_{\hat z}(2\pi) in the JzJ_z eigenbasis: each eigenvalue mm contributes e−2πime^{-2\pi im}, and all mm in the multiplet differ from jj by integers.

D(j)(−I)=(−1)2j I2j+1D^{(j)}(-\id)=(-1)^{2j}\,\id_{2j+1}
Integer jj: −I↦I-\id\mapsto\id, so D(j)D^{(j)} factors through SO(3)SO(3) (vectors, tensors, orbital angular momentum). Half-integer jj: −I↦−I-\id\mapsto-\id, a genuine representation of SU(2)SU(2) but only a projective one of SO(3)SO(3). These are the spinors.

Key takeaways

  • SU(2)≅S3SU(2)\cong S^3 acts on Bloch vectors by conjugation, giving a surjective homomorphism Φ:SU(2)→SO(3)\Phi:SU(2)\to SO(3) with kernel {±I}\{\pm\id\}.
  • SO(3)≅RP3SO(3)\cong\R P^3 has π1=Z2\pi_1=\Z_2: the 2π2\pi rotation loop is not contractible, the 4π4\pi loop is.
  • The spinor sign Rn^(2π)=−IR_{\hat n}(2\pi)=-\id is observable relative to a reference. A controlled 2π2\pi rotation is ZZ on the control, and neutron interferometry saw the 4π4\pi period in 1975.
  • D(j)(−I)=(−1)2jD^{(j)}(-\id)=(-1)^{2j}: integer spins represent SO(3)SO(3), half-integer spins only projectively, as Bargmann's theorem allows.

Check yourself

In the interactive lesson these are scored with your confidence, and what you miss comes back for review.

Question 1

What is ker⁡Φ\ker\Phi for Φ:SU(2)→SO(3)\Phi:SU(2)\to SO(3)?

  1. {I}\{\id\}
  2. {±I}\{\pm\id\}
  3. {eiαI:α∈R}\{e^{i\alpha}\id:\alpha\in\R\}
  4. {±I,±iI}\{\pm\id,\pm i\id\}
Show the answer

B. {±I}\{\pm\id\}

Only scalars commute with all Paulis, and det⁡(λI)=λ2=1\det(\lambda\id)=\lambda^2=1 forces λ=±1\lambda=\pm1.

Question 2

A control qubit in ∣+⟩\ket+ drives a controlled Rx^(2π)R_{\hat x}(2\pi) on a target in an arbitrary state ∣ψ⟩\ket\psi. What is the output?

  1. ∣+⟩⊗(−∣ψ⟩)\ket+\otimes(-\ket\psi)
  2. An entangled state
  3. ∣+⟩⊗∣ψ⟩\ket+\otimes\ket\psi
  4. ∣−⟩⊗∣ψ⟩\ket-\otimes\ket\psi
Show the answer

D. ∣−⟩⊗∣ψ⟩\ket-\otimes\ket\psi

Controlled-Rn^(2π)=Z⊗IR_{\hat n}(2\pi)=Z\otimes\id for every axis, so ∣+⟩∣ψ⟩↦∣−⟩∣ψ⟩\ket+\ket\psi\mapsto\ket-\ket\psi.

Question 3

Which spin-jj representations of SU(2)SU(2) are genuine representations of SO(3)SO(3)?

  1. Only j=0j=0
  2. Half-integer jj
  3. Integer jj
  4. All jj, since su(2)≅so(3)\mathfrak{su}(2)\cong\mathfrak{so}(3)
Show the answer

C. Integer jj

D(j)(−I)=(−1)2jID^{(j)}(-\id)=(-1)^{2j}\id is trivial exactly for integer jj, and only then does D(j)D^{(j)} factor through SU(2)/{±I}SU(2)/\{\pm\id\}.

Question 4

In SO(3)SO(3), which loop of rotations about a fixed axis can be continuously shrunk to the constant loop?

  1. The 2π2\pi loop
  2. The 4π4\pi loop
  3. Neither: no rotation loop is contractible
  4. Both: SO(3)SO(3) is simply connected
Show the answer

B. The 4π4\pi loop

π1(SO(3))=Z2\pi_1(SO(3))=\Z_2 is generated by the 2π2\pi loop. Its square, the 4π4\pi loop, lifts to a closed loop in SU(2)≅S3SU(2)\cong S^3, which contracts.

Further reading

  • Y. Aharonov & L. Susskind, Observability of the sign change of spinors under 2π rotations, Physical Review (1967).
  • H. Rauch, A. Zeilinger, G. Badurek, A. Wilfing, W. Bauspiess & U. Bonse, Verification of coherent spinor rotation of fermions, Physics Letters A (1975).
  • S. A. Werner, R. Colella, A. W. Overhauser & C. F. Eagen, Observation of the phase shift of a neutron due to precession in a magnetic field, Physical Review Letters (1975).
  • N. J. Ross & P. Selinger, Optimal ancilla-free Clifford+T approximation of z-rotations, Quantum Information & Computation (2016).
  • B. C. Hall, Lie Groups, Lie Algebras, and Representations: An Elementary Introduction, 2nd ed. (Springer, 2015).