Where the Hopf fibration really shows up in light

A fully polarized beam is a normalized Jones vector, a point on S³. Dropping the global phase gives a point on the Poincaré sphere S². That map is the Hopf map, and each fibre is the set of Jones vectors that differ only by global phase. The experiment below tests the one place this phase becomes measurable: the Pancharatnam (geometric) phase picked up when a beam passes through a cycle of three polarization states.

Poincaré sphere (S²) with the three polarization states and geodesic triangle.

Fibres of S³ over the same three states, stereographically projected to R³. Each fibre is a circle; any two distinct fibres link once.

View
yaw pitch
Precision of the loop phase
photons N

Model: ideal shot-noise limit, fringe read at quadrature (or scanned), both interferometer ports counted, loop arm transmits amplitude V because the polarizer chain removes the rest. No path-length jitter, detector noise or wave-plate error. For sensing a displacement of one vertex, ordinary two-port polarimetry gives F = 1 per photon, while the loop-phase channel reaches at most 0.25 (only at the singular, orthogonal-state limit) in this model.

Holonomy: horizontal lift of the loop in S³ (dashed curve, right panel)

Each leg is lifted so the phase never changes along it. The lift starts on fibre A and ends on the same fibre, short of closing by the thick arc. That arc is the geometric phase, −Ω/2.

Noise budget against the Fisher bound (N from the slider above)
fast jitter σ (rad)run-to-run drift σ (rad)vertex error σ (°, Poincaré sphere)dark counts per port, 10^xinterferometer contrast

A polarizer angle error α moves a linear-state vertex by 2α. Fast jitter lowers contrast by exp(−σ²/2). Drift between the loop run and the A→A reference run adds directly to the phase error.

Fit your own fringe data
Sensing one vertex, and the three-photon tritter
vertexdirection on sphere (°)

Tritter: three photons with these polarizations enter a balanced three-port Fourier splitter. Output probabilities depend only on the overlaps and the triad phase, so A, B, C set that phase directly.

Derivations
With Bloch vectors a,b,c and ρ=(1+n·σ)/2: Δ3 = ⟨A|B⟩⟨B|C⟩⟨C|A⟩ = Tr(ρaρbρc) = (W + iT)/4, W = 1+a·b+b·c+c·a, T = a·(b×c). So arg Δ3 = atan2(T,W) = Ω/2 and V² = |Δ3|² = (1+a·b)(1+b·c)(1+c·a)/8. Successive projections A→B→C→A give the conjugate, so the measured shift is −Ω/2. Gradient: ∇_b φ = [W(c×a) − T(a+c)]/(W²+T²), cyclic for a and c. ∇_b V² = [a(1+b·c)+c(1+a·b)](1+c·a)/8. Fringe: loop arm amplitude s=√u·V, reference q=√(1−u), ports P± = (q²+s² ± 2qs·cos(φ+ψ))/2 (rate model). F = Σ(∂P)²/P = 4q²s²sin²x·(q²+s²)/((q²+s²)² − 4q²s²cos²x), largest at x=π/2: F = 4q²s²/(q²+s²). Maximizing over u gives u = 1/(1+V) and F = (2V/(1+V))². A balanced splitter gives 2V²/(1+V²). Scanning the piezo uniformly gives the average F = 2·min(q², s²). Cramér–Rao: σ(φ) ≥ 1/√(NF), σ(Ω) = 2σ(φ). Vertex sensing: F_θ = F·(∂φ/∂θ)² for the phase channel. Two-port polarimetry with an optimal probe gives 1. Tritter: p(o) = (1/∏nₖ!) Σ_{σ,τ∈S3} ∏ⱼ U_{oⱼσ(j)} U*_{oⱼτ(j)} ⟨ψ_τ(j)|ψ_σ(j)⟩, U = Fourier matrix/√3.

Bench protocol to run this for real

  1. Laser (HeNe or 633 nm diode) into a Mach-Zehnder interferometer, with equal arms and a piezo mirror to scan path length.
  2. In one arm, place a polarizer to prepare state A, then step through states B and C with a quarter-wave plate + half-wave plate pair, and return to A with a final polarizer. Leave the other arm as a reference that is polarization-matched to A.
  3. Record the fringe pattern with the loop closed. Compare against a run where the arm goes straight from A to A. Wave plates are unitary rotations and add their own dynamical phase, so either use polarizers at A, B and C (accepting the intensity loss, which is the V in the readout) or calibrate the wave-plate phase out with the A-to-A run.
  4. Prediction: the fringes shift by an amount equal to minus half the solid angle of the geodesic triangle ABC on the Poincaré sphere, with the sign set by traversal direction. Reverse the loop and the shift should flip sign.
  5. Control: vary only the dynamical path length with the piezo. This moves fringes but should not change the loop-induced offset.

Status: the geometric phase itself is established physics (Pancharatnam 1956, Berry 1984, with polarization experiments by Bhandari and Samuel and later groups). This page does not test any claim that c, hydrogen or Zipf's law follow from the Hopf map; it only shows the verified link between the fibre structure and an observable.