Berry Phase

The idea: phase that remembers the path

Imagine walking on the surface of the Earth carrying a spear that you keep pointing in a fixed compass-like direction — never twisting it in your own hand, only letting the ground steer it. Start at the equator pointing east, walk north to the pole, turn and walk back down a different meridian, then return along the equator to your start. You will find the spear now points in a different direction, rotated by an angle equal to the area you enclosed divided by the radius squared — the solid angle of the loop. The spear "remembers" the geometry of where it has been even though you never deliberately rotated it. This is holonomy, and it is exactly the classical shadow of the Berry phase.

In quantum mechanics, the "direction" of the spear is replaced by the phase of a wavefunction. Slowly change the knobs of a Hamiltonian — a magnetic field direction, a crystal momentum, a molecular configuration — around a closed loop and return everything to where it began. The system, if it started in an eigenstate, comes back to the same eigenstate, but its phase has been rotated by an amount fixed by the geometry of the loop. That rotation is the Berry phase, sometimes called the geometric phase.

The adiabatic setup

Let the Hamiltonian depend on a set of slowly-varying parameters R(t) — for instance the three components of an applied magnetic field. At each instant it has eigenstates:

H(R) |n(R)⟩ = E_n(R) |n(R)⟩

The adiabatic theorem says that if R(t) changes slowly compared to ħ/ΔE (where ΔE is the gap to the nearest other level), a system prepared in state |n⟩ stays in the instantaneous state |n(R(t))⟩ — it never jumps to another level. Naively you would write the state as just |n(R(t))⟩ times the familiar dynamical phase:

θ_n(t) = −(1/ħ) ∫₀ᵗ E_n(R(t')) dt'

Berry's 1984 insight was that this is incomplete. The honest state carries an additional phase γ_n(t):

|ψ(t)⟩ = e^{iγ_n(t)} · e^{iθ_n(t)} · |n(R(t))⟩

Plug this into the Schrödinger equation and the geometric phase is forced upon you:

γ_n(t) = i ∫ ⟨n(R)| ∇_R |n(R)⟩ · dR

Berry connection and Berry curvature

The integrand defines the Berry connection, a vector field in parameter space:

A_n(R) = i ⟨n(R)| ∇_R |n(R)⟩

The connection is gauge dependent: you are free to multiply each eigenstate by an arbitrary phase |n⟩ → e^{iξ(R)}|n⟩, which shifts A → A − ∇ξ. For an open path this freedom means the accumulated γ can be transformed away — there is no invariant content. But around a closed loop C, the single-valuedness of the wavefunction forces ξ to return to itself (mod 2π), and the line integral becomes gauge invariant:

γ_n = ∮_C A_n(R) · dR     (gauge invariant, mod 2π)

By Stokes' theorem this loop integral equals the flux of the Berry curvature through any surface S bounded by C:

F_n = ∇_R × A_n
γ_n = ∬_S F_n · dS

The Berry curvature plays the role of a magnetic field in parameter space; the connection is its vector potential. Degeneracies — points where E_n meets another level — act as magnetic monopoles of this fictitious field, and the Berry phase counts how much of their flux the loop encircles.

The spin-½ in a magnetic field — γ = −½Ω

The cleanest example. A spin-½ in a magnetic field has H = −μ B·σ/2; the parameter space is the set of field directions, a unit sphere. As B traces a closed loop, the Berry curvature is exactly that of a unit monopole at the origin B = 0 (the spin-up/spin-down degeneracy). The flux of a unit monopole through a cap is the solid angle, so:

γ_± = ∓ ½ Ω(C)

where Ω(C) is the solid angle the field direction sweeps out as seen from the center of the sphere, and the sign depends on whether the spin is aligned (+) or anti-aligned (−). The factor of ½ is the spin quantum number — for a spin-S particle the monopole has strength S and γ = −SΩ. This single formula is the canonical Berry phase, the one the visualization above traces out: parallel-transport the state vector around a loop on the sphere, and it returns rotated by the enclosed solid angle.

Loop on the parameter sphere	Solid angle Ω	Berry phase γ_+ = −½Ω
Tiny loop near the north pole	≈ 0	≈ 0
Equator (great circle)	2π (half the sphere)	−π
Cone at fixed polar angle ϑ	2π(1 − cos ϑ)	−π(1 − cos ϑ)
Whole sphere (Ω = 4π)	4π	−2π ≡ 0 (mod 2π)

Geometric phase vs. dynamical phase

The two phases the state accumulates could not be more different in character. The dynamical phase is the "boring" one that bookkeeps energy and time; the geometric phase is the one that encodes the trip itself.

Property	Dynamical phase θ_n	Geometric (Berry) phase γ_n
Formula	−(1/ħ)∫E_n dt	∮ A_n·dR
Depends on elapsed time?	Yes — grows with duration	No — reparametrization invariant
Depends on energy?	Yes	No
Depends on path shape?	Only through E along it	Yes — only on the geometry
Survives the slow (adiabatic) limit?	Diverges (∝ time)	Stays finite and fixed
Gauge invariant on a closed loop?	Yes	Yes (mod 2π)
Classical analog	Action accumulated	Holonomy / Hannay angle

Numerical examples

System	Loop / setup	Geometric phase
Spin-½ in field, cone at ϑ = 60°	Ω = 2π(1 − cos60°) = π	γ = −π/2
Spin-1 (S = 1), equatorial loop	Ω = 2π	γ = −SΩ = −2π ≡ 0
Photon in coiled fiber (Tomita–Chiao)	Helix solid angle Ω	Polarization rotates by Ω
Graphene, electron circling a Dirac point	Winding once around the cone	π Berry phase
Foucault pendulum (classical Hannay)	Earth loop at latitude λ	2π(1 − sin λ) precession
Aharonov–Bohm (related geometric phase)	Loop around flux Φ	γ = qΦ/ħ

JavaScript — solid angle and the spin-½ Berry phase

// Solid angle of a cone at polar angle theta (radians), measured from the pole.
function coneSolidAngle(theta) {
  return 2 * Math.PI * (1 - Math.cos(theta));   // steradians
}

// Berry phase of a spin-1/2 whose field traces that cone.
function berryPhaseSpinHalf(theta) {
  return -0.5 * coneSolidAngle(theta);          // gamma_+ = -Omega/2
}

const theta = Math.PI / 3;                       // 60 degrees
console.log(`Solid angle: ${coneSolidAngle(theta).toFixed(4)} sr`);   // 3.1416 (= pi)
console.log(`Berry phase: ${berryPhaseSpinHalf(theta).toFixed(4)} rad`); // -1.5708 (= -pi/2)

// Equatorial great circle: half the sphere.
console.log(`Equator gamma = ${berryPhaseSpinHalf(Math.PI/2).toFixed(4)}`); // -pi

// Solid angle of an arbitrary spherical polygon via the spherical-excess
// (Gauss-Bonnet) sum of interior angles minus (n-2)*pi.
function polygonSolidAngle(interiorAngles) {
  const n = interiorAngles.length;
  const sum = interiorAngles.reduce((a, b) => a + b, 0);
  return sum - (n - 2) * Math.PI;                // spherical excess = enclosed area = Omega
}

// A spherical triangle with three right angles (one octant of the sphere).
console.log(`Octant Omega = ${polygonSolidAngle([Math.PI/2, Math.PI/2, Math.PI/2]).toFixed(4)}`);
// pi/2 steradians, i.e. 1/8 of 4*pi. Berry phase for spin-1/2: -pi/4.

// Discrete Berry phase from a chain of states (Bargmann / Pancharatnam product).
// Each state is a 2-component complex spinor; phase = -arg of the loop product.
function discreteBerryPhase(states) {  // states: [[re,im,re,im], ...], closed loop
  const dot = (a, b) => {  //  for 2-spinors stored as [re0,im0,re1,im1]
    let re = 0, im = 0;
    for (let k = 0; k < 2; k++) {
      const ar = a[2*k], ai = -a[2*k+1];        // conjugate of a
      const br = b[2*k], bi = b[2*k+1];
      re += ar*br - ai*bi;
      im += ar*bi + ai*br;
    }
    return [re, im];
  };
  let pr = 1, pi = 0;
  for (let i = 0; i < states.length; i++) {
    const [re, im] = dot(states[i], states[(i+1) % states.length]);
    const nr = pr*re - pi*im, ni = pr*im + pi*re;
    pr = nr; pi = ni;
  }
  return -Math.atan2(pi, pr);                    // Berry phase (gauge invariant)
}

Where the Berry phase shows up

• Topological insulators & quantum Hall effect. The integer Hall conductance is the Chern number — the Berry curvature of the filled bands integrated over the Brillouin zone, divided by 2π. Graphene's π Berry phase gives its half-integer quantum Hall plateaus.
• Anomalous Hall and spin Hall effects. Berry curvature in momentum space acts like a magnetic field, deflecting electrons sideways even without an external B.
• Modern theory of polarization. The electric polarization of a crystal is a Berry phase of the valence Bloch states (King-Smith–Vanderbilt), and adiabatic charge pumping is quantized by it (Thouless pump).
• Molecular physics. Near a conical intersection of potential surfaces, the electronic wavefunction picks up a π geometric phase — the molecular Aharonov–Bohm effect — reshaping vibrational spectra and reaction rates.
• Optics. Pancharatnam's phase for polarized light; the Tomita–Chiao coiled-fiber experiment; Pancharatnam–Berry metasurface lenses that steer light by spatially patterning geometric phase.
• NMR, neutron, and atomic interferometry. Direct measurements of geometric phase by cycling spins through closed loops in field space.
• Quantum computation. Holonomic / geometric quantum gates use Berry phases, which are robust against certain timing and amplitude errors because they depend only on geometry.

Common mistakes

• Thinking it depends on speed. The Berry phase is reparametrization invariant — traverse the loop twice as slowly and it is unchanged. Only the dynamical phase cares about timing.
• Trying to gauge it away. On an open path the phase is gauge dependent and meaningless on its own. Only the closed-loop integral (or an interference of two paths) is physical.
• Confusing it with the Aharonov–Bohm phase. They are cousins — both are geometric/topological — but the Aharonov–Bohm phase comes from a real electromagnetic vector potential in real space, while the generic Berry connection lives in abstract parameter space.
• Forgetting the adiabatic requirement (in the original form). Berry's derivation assumes the loop is slow versus ħ/gap. If a degeneracy is crossed or the gap closes, the simple eigenstate-tracking picture fails; use the Aharonov–Anandan generalization for fast cyclic evolution.
• Dropping the factor of ½ for spin. For spin-½ the phase is −Ω/2, not −Ω; only a unit monopole (spin replaced by charge 1) gives the bare solid angle.
• Assuming a single phase is observable. An overall phase is invisible. You must compare two interfering amplitudes — that is why every measurement uses interferometry or a phase difference.