Simple Algorithm
Quantum Computing · Rydberg State · 28 min

Spin-½ and the Pauli Algebra

Stern–Gerlach, su(2) commutators and anticommutators.

In 1922 a beam of silver atoms passing through an inhomogeneous magnet split into two spots instead of a smear. That two-valuedness is spin-½, and its whole algebra fits on one line, σjσk=δjkI+iϵjklσl\sigma_j\sigma_k=\delta_{jk}\id+i\epsilon_{jkl}\sigma_l. From that line follow rotations, uncertainty relations, and the fact that every spin-½ state points somewhere.

What you’ll learn

  • Derive the Pauli product rule, with its commutator (su(2)\mathfrak{su}(2)) and anticommutator (Clifford) halves, and the identity (a⃗⋅σ⃗)(b⃗⋅σ⃗)=(a⃗⋅b⃗)I+i(a⃗×b⃗)⋅σ⃗(\vec a\cdot\vec\sigma)(\vec b\cdot\vec\sigma)=(\vec a\cdot\vec b)\id+i(\vec a\times\vec b)\cdot\vec\sigma.
  • Prove that every pure spin-½ state is spin-up along exactly one axis, and derive sequential Stern–Gerlach statistics.
  • Prove that the 2-dimensional irreducible representation of su(2)\mathfrak{su}(2) is unique up to unitary equivalence, via ladder operators.
  • State the spin uncertainty relations and the Larmor precession of a magnetic moment.

Stern–Gerlach: two spots, and filters in sequence

A magnetic moment μ⃗\vec\mu in an inhomogeneous field feels F⃗=∇(μ⃗⋅B⃗)\vec F=\nabla(\vec\mu\cdot\vec B), so a gradient along zz pushes with Fz≈μz ∂zBzF_z\approx\mu_z\,\partial_zB_z. A silver atom's moment comes from its single unpaired 5s5s electron. Classically, randomly oriented moments would smear across the screen; Stern and Gerlach found two separate deposits. Each component of the electron's spin, Sn^=n^⋅S⃗S_{\hat n}=\hat n\cdot\vec S, has exactly two possible measurement outcomes, ±ℏ2\pm\tfrac\hbar2.

Consequently a Stern–Gerlach filter along n^\hat n followed by an analyzer along m^\hat m gives P(+m^∣+n^)=∣⟨+m^∣+n^⟩∣2=12(1+m^⋅n^)=cos⁡2Θ2P(+\hat m\mid+\hat n)=\abs{\braket{+\hat m|+\hat n}}^2=\tfrac12(1+\hat m\cdot\hat n)=\cos^2\tfrac\Theta2, where Θ\Theta is the angle between the axes. Each filter prepares a new state and erases what earlier filters selected.

Predict before you look

Spins leave a +z+z filter, pass a second filter that keeps only spin-up along an axis 60∘60^\circ from zz, then hit a zz analyzer. What percentage of the spins leaving the first filter end up −z-z? (Without the middle filter the answer would be 0.)

Show the answer

19 %

P=cos⁡230∘⋅sin⁡230∘=34⋅14=316≈18.75%P=\cos^2 30^\circ\cdot\sin^2 30^\circ=\tfrac34\cdot\tfrac14=\tfrac3{16}\approx18.75\%. Inserting a filter increases the −z-z count, because the middle measurement erases the +z+z preparation. A middle filter at 90∘90^\circ maximizes the effect: 12⋅12=25%\tfrac12\cdot\tfrac12=25\%.

The Pauli algebra

DerivationProof
  1. XY=(0110)(0−ii0)=(i00−i)=iZXY=\begin{pmatrix}0&1\\1&0\end{pmatrix}\begin{pmatrix}0&-i\\i&0\end{pmatrix}=\begin{pmatrix}i&0\\0&-i\end{pmatrix}=iZ

    Direct multiplication. In the same way YZ=iXYZ=iX and ZX=iYZX=iY, while reversing the order flips the sign: YX=−iZYX=-iZ.

  2. σj2=I,σjσk=−σkσj(j≠k)\sigma_j^2=\id,\qquad\sigma_j\sigma_k=-\sigma_k\sigma_j\quad(j\ne k)

    Each Pauli is Hermitian and unitary, so it squares to I\id. Distinct Paulis anticommute. Together these give the product rule.

  3. (a⃗⋅σ⃗)(b⃗⋅σ⃗)=∑j,kajbk σjσk=∑jajbj I+i∑j,k,lϵjkl ajbk σl(\vec a\cdot\vec\sigma)(\vec b\cdot\vec\sigma)=\sum_{j,k}a_jb_k\,\sigma_j\sigma_k=\sum_ja_jb_j\,\id+i\sum_{j,k,l}\epsilon_{jkl}\,a_jb_k\,\sigma_l

    Expand and insert the product rule.

  4. =(a⃗⋅b⃗) I+i(a⃗×b⃗)⋅σ⃗=(\vec a\cdot\vec b)\,\id+i(\vec a\times\vec b)\cdot\vec\sigma

    Recognize (a⃗×b⃗)l=ϵljkajbk(\vec a\times\vec b)_l=\epsilon_{ljk}a_jb_k. With a⃗=b⃗=n^\vec a=\vec b=\hat n this gives (n^⋅σ⃗)2=I(\hat n\cdot\vec\sigma)^2=\id, the source of e−iθn^⋅σ⃗/2=cos⁡θ2I−isin⁡θ2 n^⋅σ⃗e^{-i\theta\hat n\cdot\vec\sigma/2}=\cos\tfrac\theta2\id-i\sin\tfrac\theta2\,\hat n\cdot\vec\sigma. □\square

With Sj=ℏ2σjS_j=\tfrac\hbar2\sigma_j, the commutator half becomes the angular-momentum algebra [Sj,Sk]=iℏ ϵjklSl[S_j,S_k]=i\hbar\,\epsilon_{jkl}S_l, valid for every spin. The anticommutator half, {Sj,Sk}=ℏ22δjkI\acomm{S_j}{S_k}=\tfrac{\hbar^2}{2}\delta_{jk}\id, is special to spin-½: it says Sj2=ℏ24IS_j^2=\tfrac{\hbar^2}4\id, hence S⃗2=34ℏ2I=s(s+1)ℏ2I\vec S^2=\tfrac34\hbar^2\id=s(s+1)\hbar^2\id with s=12s=\tfrac12. Together with Tr⁡(σjσk)=2δjk\Tr(\sigma_j\sigma_k)=2\delta_{jk}, the Paulis and I\id form an orthogonal basis of all 2×22\times2 matrices, so every qubit observable is a0I+a⃗⋅σ⃗a_0\id+\vec a\cdot\vec\sigma.

Check your understanding

Which identity holds for spin-½ but fails for spin 1 (units ℏ=1\hbar=1)?

  1. [Sx,Sy]=iSz[S_x,S_y]=iS_z
  2. S⃗2=s(s+1)I\vec S^2=s(s+1)\id
  3. Sx2=14IS_x^2=\tfrac14\id
  4. S±=Sx±iSyS_\pm=S_x\pm iS_y shift SzS_z eigenvalues by ±1\pm1
Show the answer

C. Sx2=14IS_x^2=\tfrac14\id

Commutators, the Casimir S⃗2=s(s+1)I\vec S^2=s(s+1)\id and the ladder action hold in every representation. Sx2=14IS_x^2=\tfrac14\id comes from the Pauli anticommutator, special to s=12s=\tfrac12. For spin 1, SxS_x has eigenvalues −1,0,1-1,0,1, so Sx2S_x^2 has eigenvalues 1,0,11,0,1 and is not a multiple of I\id.

Uniqueness: every two-level angular momentum is a spin-½

DerivationProof with ladder operators
  1. J±=J1±iJ2,[J3,J±]=±J±,[J+,J−]=2J3J_\pm=J_1\pm iJ_2,\qquad[J_3,J_\pm]=\pm J_\pm,\qquad[J_+,J_-]=2J_3

    J±J_\pm raise or lower J3J_3 eigenvalues by 1.

  2. J⃗2=J−J++J32+J3  ⟹  J⃗2=mmax⁡(mmax⁡+1) I\vec J^2=J_-J_++J_3^2+J_3\;\Longrightarrow\;\vec J^2=m_{\max}(m_{\max}+1)\,\id

    J⃗2\vec J^2 commutes with every JjJ_j, so by Schur's lemma it is a scalar on an irreducible space. Evaluate it on an eigenvector of the largest J3J_3 eigenvalue mmax⁡m_{\max}, where J+J_+ gives 00.

  3. ∥J−∣m⟩∥2=⟨m|J+J−|m⟩=mmax⁡(mmax⁡+1)−m(m−1)\norm{J_-\ket m}^2=\mel{m}{J_+J_-}{m}=m_{\max}(m_{\max}+1)-m(m-1)

    Use J+J−=J⃗2−J32+J3J_+J_-=\vec J^2-J_3^2+J_3. Lowering from mmax⁡m_{\max} must stop in finite dimension, and the norm vanishes only at m=−mmax⁡m=-m_{\max}. The chain mmax⁡,…,−mmax⁡m_{\max},\dots,-m_{\max} spans an invariant subspace, hence everything: the dimension is 2mmax⁡+12m_{\max}+1, so s=mmax⁡∈12Ns=m_{\max}\in\tfrac12\N.

  4. J+=∣↑⟩ ⁣⟨↓∣,J1=12(J++J−)=12X,J2=12i(J+−J−)=12YJ_+=\ketbra{\uparrow}{\downarrow},\qquad J_1=\tfrac12(J_++J_-)=\tfrac12X,\qquad J_2=\tfrac1{2i}(J_+-J_-)=\tfrac12Y

    In dimension 2, s=12s=\tfrac12 and J3=12diag⁡(1,−1)J_3=\tfrac12\diag(1,-1) in the basis ∣↑⟩,∣↓⟩\ket\uparrow,\ket\downarrow. Here ∥J+∣↓⟩∥2=s(s+1)−m(m+1)=34+14=1\norm{J_+\ket\downarrow}^2=s(s+1)-m(m+1)=\tfrac34+\tfrac14=1, so a phase choice for ∣↓⟩\ket\downarrow makes J+∣↓⟩=∣↑⟩J_+\ket\downarrow=\ket\uparrow exactly. The basis change is the unitary VV. □\square

Physically, then, any two-level system whose three observables obey angular-momentum commutators is a spin-½, whatever it is made of: photon polarization (the Stokes parameters), a two-level atom (the pseudo-spin of the optical Bloch equations), or a transmon restricted to its lowest pair of levels. This is why the qubit has a single, universal Bloch sphere.

Uncertainty, precession and spin qubits

Applied to [σx,σy]=2iσz[\sigma_x,\sigma_y]=2i\sigma_z, Robertson's inequality ΔA ΔB≥12∣⟨[A,B]⟩∣\Delta A\,\Delta B\ge\tfrac12\abs{\expval{[A,B]}} gives Δσx Δσy≥∣⟨σz⟩∣\Delta\sigma_x\,\Delta\sigma_y\ge\abs{\expval{\sigma_z}}. Spin-½ also has a sharper, state-independent relation. From σj2=I\sigma_j^2=\id:

(Δσx)2+(Δσy)2+(Δσz)2=∑j(1−rj2)=3−∣r⃗∣2  ≥  2(\Delta\sigma_x)^2+(\Delta\sigma_y)^2+(\Delta\sigma_z)^2=\sum_j\big(1-r_j^2\big)=3-\abs{\vec r}^2\;\ge\;2
Equality holds exactly for pure states. No state is sharp in all three components; a pure state is sharp along n^=r⃗\hat n=\vec r and maximally uncertain in both perpendicular directions.

The electron's magnetic moment is μ⃗=−gμBS⃗/ℏ=−gμB2σ⃗\vec\mu=-g\mu_B\vec S/\hbar=-\tfrac{g\mu_B}2\vec\sigma with g≈2.0023g\approx2.0023, so H=−μ⃗⋅B⃗=gμB2 B⃗⋅σ⃗H=-\vec\mu\cdot\vec B=\tfrac{g\mu_B}2\,\vec B\cdot\vec\sigma. Comparing with ℏ2Ω⃗⋅σ⃗\tfrac\hbar2\vec\Omega\cdot\vec\sigma gives Ω⃗=gμBB⃗/ℏ\vec\Omega=g\mu_B\vec B/\hbar: the Bloch vector undergoes Larmor precession, r⃗˙=Ω⃗×r⃗\dot{\vec r}=\vec\Omega\times\vec r, at ωL/2π=gμBB/h≈28\omega_L/2\pi=g\mu_BB/h\approx28 GHz per tesla. Driving at ωL\omega_L with a transverse field is electron spin resonance, the subject of the Rabi lesson.

Key takeaways

  • Stern–Gerlach shows each spin component has two values. Filters along axes at angle Θ\Theta pass a fraction cos⁡2(Θ/2)\cos^2(\Theta/2), and each filter erases earlier selections.
  • σjσk=δjkI+iϵjklσl\sigma_j\sigma_k=\delta_{jk}\id+i\epsilon_{jkl}\sigma_l packs the commutators (su(2)\mathfrak{su}(2)) and anticommutators (Clifford algebra) into one rule, giving (a⃗⋅σ⃗)(b⃗⋅σ⃗)=(a⃗⋅b⃗)I+i(a⃗×b⃗)⋅σ⃗(\vec a\cdot\vec\sigma)(\vec b\cdot\vec\sigma)=(\vec a\cdot\vec b)\id+i(\vec a\times\vec b)\cdot\vec\sigma.
  • The 2-dimensional irrep of su(2)\mathfrak{su}(2) is unique up to unitary equivalence: any two-level system with angular-momentum commutators is a spin-½ in disguise.
  • For spin-½ every pure state is spin-up along one axis and ∑j(Δσj)2=3−∣r⃗∣2≥2\sum_j(\Delta\sigma_j)^2=3-\abs{\vec r}^2\ge2; higher spins have neither property. A magnetic moment precesses at ωL=gμBB/ℏ\omega_L=g\mu_BB/\hbar.

Check yourself

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

Question 1

Simplify (a⃗⋅σ⃗)(b⃗⋅σ⃗)(\vec a\cdot\vec\sigma)(\vec b\cdot\vec\sigma) for a⃗=x^\vec a=\hat x and b⃗=y^\vec b=\hat y.

  1. 00
  2. −iZ-iZ
  3. I\id
  4. iZiZ
Show the answer

D. iZiZ

(a⃗⋅b⃗)I+i(a⃗×b⃗)⋅σ⃗=0+i z^⋅σ⃗=iZ(\vec a\cdot\vec b)\id+i(\vec a\times\vec b)\cdot\vec\sigma=0+i\,\hat z\cdot\vec\sigma=iZ, i.e. XY=iZXY=iZ.

Question 2

Silver atoms leaving a +z+z Stern–Gerlach filter enter an apparatus oriented along xx. What fraction exits in the +x+x beam, and what does a subsequent zz measurement on that beam give?

  1. 12\tfrac12; then zz gives ++ with certainty
  2. 11; then zz gives ++
  3. 12\tfrac12; then zz gives ±\pm with probability 12\tfrac12 each
  4. 00, because SxS_x and SzS_z do not commute
Show the answer

C. 12\tfrac12; then zz gives ±\pm with probability 12\tfrac12 each

P(+x∣+z)=cos⁡245∘=12P(+x\mid+z)=\cos^245^\circ=\tfrac12, and the xx filter prepares ∣+x⟩\ket{+x}, for which a zz measurement is again 50/50. The earlier zz selection is erased.

Question 3

Which statement about pure states of a spin-½ is true?

  1. Every pure state is the +ℏ2+\tfrac\hbar2 eigenstate of n^⋅S⃗\hat n\cdot\vec S for exactly one unit vector n^\hat n
  2. Some pure states have ⟨S⃗⟩=0\expval{\vec S}=0
  3. Only eigenstates of SzS_z are pure
  4. Every pure state is an eigenstate of SxS_x, SyS_y and SzS_z at once
Show the answer

A. Every pure state is the +ℏ2+\tfrac\hbar2 eigenstate of n^⋅S⃗\hat n\cdot\vec S for exactly one unit vector n^\hat n

∣ψ⟩ ⁣⟨ψ∣=12(I+n^⋅σ⃗)\dyad\psi=\tfrac12(\id+\hat n\cdot\vec\sigma) with n^=⟨σ⃗⟩\hat n=\expval{\vec\sigma} a unit vector, so ∣ψ⟩\ket\psi is spin-up along n^\hat n and along no other axis.

Question 4

For a spin-½ state with Bloch vector r⃗\vec r, what is ∑j(Δσj)2\sum_j(\Delta\sigma_j)^2?

  1. 11
  2. 3−∣r⃗∣23-\abs{\vec r}^2
  3. ∣r⃗∣2\abs{\vec r}^2
  4. 33
Show the answer

B. 3−∣r⃗∣23-\abs{\vec r}^2

(Δσj)2=⟨σj2⟩−⟨σj⟩2=1−rj2(\Delta\sigma_j)^2=\expval{\sigma_j^2}-\expval{\sigma_j}^2=1-r_j^2. Summing gives 3−∣r⃗∣2≥23-\abs{\vec r}^2\ge2, with equality exactly for pure states.

Further reading

  • W. Gerlach & O. Stern, Der experimentelle Nachweis der Richtungsquantelung im Magnetfeld, Zeitschrift für Physik (1922).
  • W. Pauli, Zur Quantenmechanik des magnetischen Elektrons, Zeitschrift für Physik (1927).
  • J. J. Sakurai & J. Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 1 and 3 (Addison-Wesley, 2011).
  • B. Friedrich & D. Herschbach, Stern and Gerlach: how a bad cigar helped reorient atomic physics, Physics Today (2003).
  • J. M. Elzerman et al., Single-shot read-out of an individual electron spin in a quantum dot, Nature (2004).