Geometric (Berry) Phase
Phase from the shape of a path, not its duration.
Carry a spin slowly around a closed loop and it comes back with an extra phase that depends only on the shape of the loop: minus half the solid angle it enclosed. Berry's phase is the simplest example of a holonomy, and it reappears in topological band theory and in geometric quantum gates.
What you’ll learn
- Derive from the adiabatic theorem and prove it is gauge invariant modulo .
- Compute for a spin-½ following a field, and the Chern number of that eigenstate bundle.
- Generalize to non-adiabatic cyclic evolutions (Aharonov–Anandan) and to discrete paths (the Bargmann invariant).
- Split a total phase into dynamical and geometric parts, and use that split to design a purely geometric gate.
Where the extra phase comes from
Let a Hamiltonian depend on parameters , say a magnetic-field vector, and let be a nondegenerate eigenstate with energy , separated by a gap from the rest of the spectrum. Move slowly around a closed loop in total time . The adiabatic theorem says the state stays in the instantaneous eigenspace (with error for smooth paths), so the only freedom left is a phase.
Adiabatic ansatz: the dynamical phase, plus an unknown extra phase .
Insert into the Schrödinger equation; the terms cancel. The components of outside are the non-adiabatic transitions neglected here.
Project onto . Since , is purely imaginary, so is real.
Integrate over one cycle. Time has dropped out: depends on the path, not on the speed. is the Berry connection.
Spin-½: minus half the solid angle
Take with on the unit sphere: a spin in a field of fixed strength whose direction you steer. The ground state is the spin aligned with the field, , whose Bloch vector is .
Differentiate the Bloch parametrization.
Only the component survives. This gauge is smooth everywhere except the south pole, where is multivalued.
Integrate around the circle of latitude , traversed with increasing.
is the solid angle of the polar cap. Letting the field magnitude vary too, , the curvature is : a monopole of strength at the degeneracy . By Stokes the formula holds for every loop.
A spin adiabatically follows a field that sweeps once around the equator. Ignoring the dynamical phase, what factor does the aligned state pick up?
- It depends on the sweep speed
Show the answer
C.
The equator encloses a hemisphere, , so and , whichever cap you use.
Without adiabaticity: Aharonov–Anandan and Pancharatnam
Adiabaticity is not essential. Aharonov and Anandan (1987) noticed that any cyclic evolution, , traces a closed curve in projective space, and the part of not accounted for by energy is geometric:
Precess a state with polar angle once around .
The total phase is the double-cover sign, the same for every state.
The dynamical phase, using .
Geometric phase = total phase minus dynamical phase, matching the solid angle of the cap the Bloch vector circles. The of the double cover splits into a geometric part and a dynamical part.
Carry a qubit along geodesics , the octant triangle . What is the magnitude of its geometric phase, in degrees?
Show the answer
45 °
The triangle covers one octant, , so . With amplitudes: , whose argument is .
Geometric gates, and the frontier
A phase fixed by the path is insensitive to how fast the path is traversed. The octant loop makes this concrete. , then , then carries , each step a quarter great circle about an axis perpendicular to the Bloch vector, so throughout and no dynamical phase accrues. The net result is : a purely geometric phase. Make the rotations conditional on a control qubit and that phase is kicked back where you can see it.
Key takeaways
- Adiabatic transport around a loop adds , with : independent of speed, gauge invariant modulo , equal to the flux of the Berry curvature.
- For spin-½, : the curvature is a monopole of strength at the degeneracy, and the Chern number is an integer.
- Aharonov–Anandan: for any cyclic evolution, total phase = dynamical phase + geometric phase, and for a qubit the geometric part is of the Bloch-vector loop.
- Discretely, for the geodesic triangle; the octant loop gives a phase.
- Geometric phases underlie topological band invariants and geometric or holonomic gates, whose robustness depends on the noise.
Check yourself
In the interactive lesson these are scored with your confidence, and what you miss comes back for review.
You traverse the same loop twice as slowly. What happens to the dynamical phase and to the Berry phase?
- Both double
- Dynamical phase unchanged; Berry phase halves
- Dynamical phase doubles; Berry phase unchanged
- Both are unchanged
Show the answer
C. Dynamical phase doubles; Berry phase unchanged
grows with the duration, while is a line integral over the path, independent of its parametrization.
Under with smooth and single valued, how does the Berry phase of a closed loop change?
- By an integer multiple of
- By , an arbitrary real number
- It is multiplied by
- It flips sign
Show the answer
A. By an integer multiple of
, so . Since is single valued, can only wind by around the loop, so is invariant.
A spin-½ adiabatically follows a field around a loop enclosing solid angle . What are the Berry phases of the aligned and the anti-aligned states?
- and
- for both
- and
- and
Show the answer
D. and
with gives . The two eigenstates see opposite curvature: monopoles of strength .
A qubit undergoes a cyclic evolution with total phase and dynamical phase , while its Bloch vector encloses solid angle . Which relation holds modulo ?
- , always
Show the answer
B.
Aharonov–Anandan: total = dynamical + geometric, with geometric part for a qubit, and no adiabaticity needed. In the precession example , and .
Further reading
- M. V. Berry, Quantal phase factors accompanying adiabatic changes, Proceedings of the Royal Society of London A (1984).
- B. Simon, Holonomy, the quantum adiabatic theorem, and Berry's phase, Physical Review Letters (1983).
- Y. Aharonov & J. Anandan, Phase change during a cyclic quantum evolution, Physical Review Letters (1987).
- J. Samuel & R. Bhandari, General setting for Berry's phase, Physical Review Letters (1988).
- P. Zanardi & M. Rasetti, Holonomic quantum computation, Physics Letters A (1999).