← quantum · contents

§ Phase iv · Spin · Chapter 03

The Bloch sphere

A 23-year-old in Zürich, sitting under the most feared theorist in Europe, drew a small diagram in 1928 that turned the abstract algebra of spin into something you could literally point at. Two angles, one ball, every pure qubit state mapped to a single arrow. The picture outlasted both the student and his professor, and now sits at the center of every quantum-computing textbook in print.

In the autumn of 1927, a twenty-two-year-old student named Felix Bloch walked across the courtyard of the Eidgenössische Technische Hochschule in Zürich and presented himself at the office of Wolfgang Pauli. Pauli was thirty, already famous, already feared. He had published the exclusion principle two years earlier, had spent the last twelve months hammering Heisenberg’s matrix mechanics into a workable spin formalism, and had a reputation for shouting “completely wrong” at speakers before they finished their first slide. Bloch had been raised on Bach, mathematics, and the orderly Swiss-German habits of the Zürich middle class. He was the wrong personality to be Pauli’s student. He became one anyway.

Pauli’s spin formalism, which Bloch would learn from the master’s own lecture notes, was an algebraic triumph and a pedagogical disaster. Three two-by-two matrices, an exclusion rule, and a set of commutation relations were enough to predict the Stern-Gerlach experiment, the anomalous Zeeman effect, and the fine structure of hydrogen. But the matrices did not look like anything. A student could memorize σ_x, σ_y, σ_z and still have no mental image of what a spin state was, what it meant to rotate one, or why two of them anti-commuted. The young Bloch, who would later develop a draftsman’s habit of drawing every theorem before he believed it, found this intolerable. The world had pictures of orbits, pictures of standing waves, pictures of probability clouds, pictures of pendulums. It had no picture of a spin.

So he drew one. The argument is so short it can be reproduced on a napkin. A pure spin-1/2 state is a complex two-component vector. Two complex numbers, four real numbers. Normalize the state and the four collapses to three. Throw away the global phase, which no measurement can ever see, and the three collapses to two. Two real numbers parameterize a sphere. Therefore every pure spin state, every qubit, every two-level quantum system in nature, sits at exactly one point on a unit ball. The diagram was not in his thesis (his thesis was about Bloch waves in crystal lattices, and would found solid-state physics a year later), but it floated through Zürich seminars from then on, and the sphere has carried his name ever since.

To make the picture work, you need a parameterization. The natural one comes straight from spherical coordinates. Write the spin-up eigenstate of σ_z as |↑⟩ and the spin-down eigenstate as |↓⟩. Any pure qubit state can be written as a superposition of these two basis vectors with complex coefficients. Use the freedom to rotate the global phase to make the coefficient of |↑⟩ a real, non-negative number, and call its arccosine θ/2. Whatever phase is left over on the |↓⟩ coefficient, call φ. You get

|ψ⟩ = cos(θ/2) |↑⟩ + e^(iφ) sin(θ/2) |↓⟩

with θ running from 0 to π and φ running from 0 to 2π. These are exactly the polar and azimuthal angles of a point on the unit sphere. The north pole, θ = 0, is the pure spin-up state. The south pole, θ = π, is the pure spin-down state. Every other point on the surface is a genuine superposition, with the latitude controlling the mix of up and down, and the longitude controlling the relative phase.

The factor of two in the half-angle θ/2 is the one feature that catches everyone the first time. Pauli’s algebra rotates spinors by half the angle that a vector in real space turns through. A 360-degree rotation of your physical apparatus brings the laboratory back to where it started, but it flips the sign of the spin wavefunction; you need a 720-degree rotation to recover the original state. On the Bloch sphere this looks tidy: a 720-degree turn of the laboratory walks the state vector once around the ball, and a 360-degree turn walks it halfway. That sign-flip is the famous double-cover of SO(3) by SU(2), and it is the reason a spin-1/2 particle behaves nothing like an arrow you draw on a piece of paper.

Once you fix the parameterization, the Pauli matrices become almost embarrassingly geometric. The eigenstates of σ_z are |↑⟩ and |↓⟩, which sit at the north and south poles. The eigenstates of σ_x are (|↑⟩ + |↓⟩)/√2 and (|↑⟩ - |↓⟩)/√2, which sit on the equator at +x and -x. The eigenstates of σ_y are (|↑⟩ + i|↓⟩)/√2 and (|↑⟩ - i|↓⟩)/√2, which sit on the equator at +y and -y. Six special points on a ball, one pair per Pauli matrix, each pair antipodal. The whole equator is the set of equal-weight superpositions, differing only in the relative phase φ. Walk around the equator and you walk through every possible relative phase a qubit can carry.

The geometry has a second payoff that is harder to appreciate at first. The Pauli matrices do not just label the axes of the sphere; they generate the rotations of it. Apply the unitary exp(-iασ_z/2) to a state and the arrow on the sphere rotates by angle α about the z axis. Apply exp(-iασ_x/2) and it rotates about x. Any single-qubit gate any quantum computer will ever execute is a rotation of this ball about some axis through its center. The Hadamard gate is a 180-degree rotation about an axis halfway between x and z; the phase-flip Z gate is a 180-degree rotation about z; the bit-flip X gate is a 180-degree rotation about x. Quantum-circuit pedagogy spends weeks of lecture time on this, but the entire content is one diagram: a sphere with arrows on it.

The most consequential rotation of all is the one Larmor first described in 1897 for a classical magnetic moment in a magnetic field. The quantum version is the same picture exactly. Put a spin-1/2 particle in a uniform magnetic field along z. Its Hamiltonian is H = -ωσ_z/2, where ω is the Larmor frequency proportional to the field strength and the gyromagnetic ratio of the particle. The time evolution operator is exp(-iHt/ℏ), which for this Hamiltonian is exactly a rotation about z at angular frequency ω. A state initially tilted at polar angle θ from the z axis precesses around z at constant latitude, completing one circuit every period 2π/ω. The Bloch arrow traces a horizontal circle on the sphere, exactly the way a spinning top precesses under gravity.

|ψ(0)⟩ = cos(θ/2) |↑⟩ + e^(iφ₀) sin(θ/2) |↓⟩

U(t) = exp(-i H t / ℏ) = exp(+i ω σ_z t / 2)

U(t) = diag( exp(+i ω t / 2), exp(-i ω t / 2) )

|ψ(t)⟩ = exp(+i ω t / 2) cos(θ/2) |↑⟩ + exp(-i ω t / 2) e^(iφ₀) sin(θ/2) |↓⟩

|ψ(t)⟩ = cos(θ/2) |↑⟩ + e^(i(φ₀ - ω t)) sin(θ/2) |↓⟩

The picture has, in the century since it was drawn, propagated through every branch of physics that touches a two-level system. Atomic physicists draw it for the ground and excited states of a driven atom. NMR spectroscopists draw it for the nuclear spins in their samples, with the Larmor precession around the static field being the principal motion they have to control. Quantum optics draws it for the polarization of a photon, where |↑⟩ and |↓⟩ become horizontal and vertical polarization and the equator carries the diagonal and circular states. Quantum computing draws it for the qubit, where every single-gate operation is a rotation and every measurement is a projection onto a chosen axis. The same ball with the same six labelled points appears, with minor relabelling, in five different chapters of five different graduate textbooks.

The fact that all of these systems share a picture is not an accident. Any pure quantum state of a two-level system, no matter what physical degree of freedom it represents, is exactly the same mathematical object, a unit complex two-vector modulo a global phase. The Bloch sphere is the space of such objects. When two completely different experiments (an electron spin in a magnetic field, the polarization of a single photon, a superconducting flux qubit in a dilution refrigerator) yield interference patterns that look identical, the deep reason is that they all live on the same sphere and are being rotated about the same axes by their respective Hamiltonians. The picture is universal because the underlying mathematics is.

Bloch went on, of course, to do more than draw spheres. He left Europe in 1933 after the Nazi race laws cost him his position in Leipzig, spent stints with Bohr in Copenhagen and Fermi in Rome, and eventually settled at Stanford in 1934. During the war he worked on radar and on isotope separation. In 1945, returning to physics, he wondered whether the radio-frequency techniques of radar could detect the magnetic moments of atomic nuclei directly. They could. The 1946 paper in which he reported nuclear magnetic resonance was, in effect, the Bloch sphere made experimental: drive the spins with a tuned coil, watch them precess at the Larmor frequency, listen to the signal as the population on the sphere dephases. He shared the 1952 Nobel Prize for the result, and the same idea now drives every MRI scanner ever sold. The teaching diagram he had drawn in Zürich in 1928, used for years as a way to make spin look like something a student could point at, turned out also to be the operating manual for a five-million-dollar imaging instrument.