Parallel Transport

Parallel transport is the local rule behind Berry holonomy. It says how to choose the phase of an instantaneous eigenvector as parameters move, so that the representative is not being artificially rotated by a local phase convention.

For a normalized nondegenerate eigenstate along a path $R(t)$,

$$H(R(t))\lvert n(t)\rangle = E_n(t)\lvert n(t)\rangle,$$

the Berry connection component along the path is

$$A_t(t) = i\langle n(t)|\dot n(t)\rangle.$$

A parallel-transport gauge along the path is a phase convention satisfying

$$A_t(t)=0.$$

Equivalently,

$$\langle n(t)|\dot n(t)\rangle=0.$$

This does not mean the eigenvector is constant. It means its change is orthogonal to itself, so the change is not a pure local phase rotation.

Why the Condition Makes Sense

Normalization implies

$$\langle n(t)|n(t)\rangle=1.$$

Differentiating gives

$$\langle \dot n|n\rangle + \langle n|\dot n\rangle = 0.$$

Because $\langle \dot n|n\rangle=\langle n|\dot n\rangle^*$, the quantity $\langle n|\dot n\rangle$ is purely imaginary. It is therefore exactly the component of $\lvert\dot n\rangle$ that changes the phase of $\lvert n\rangle$.

The parallel-transport condition removes that component:

$$\langle n|\dot n\rangle=0.$$

The remaining change of the eigenvector is the change of the ray or line itself in Hilbert space, not a chosen phase spin.

How to Reach the Parallel Gauge

Start from any smooth choice $\lvert n(t)\rangle$. Rephase it:

$$\lvert n'(t)\rangle = e^{i\chi(t)} \lvert n(t)\rangle.$$

The connection component transforms as

$$A_t'(t) = A_t(t)-\dot\chi(t).$$

Therefore choosing

$$\chi(t) = \int_0^t A_\tau(\tau)\,d\tau$$

sets $A_t'=0$ along the open path. Locally along a one-dimensional path, this is always possible for a smooth eigenvector.

The word “locally” matters. Around a closed loop, this phase choice may fail to return to the original vector phase. That failure is the geometric phase.

Endpoint Mismatch Around a Loop

Suppose the ray returns to itself after a closed parameter loop:

$$R(T)=R(0), \qquad \mathbb C\lvert n(T)\rangle = \mathbb C\lvert n(0)\rangle.$$

Choose a parallel-transport gauge along the loop. Then

$$A_t^{\mathrm{pt}}=0$$

at every point on the path, but the endpoint can satisfy

$$\lvert n(T)\rangle_{\mathrm{pt}} = e^{i\gamma_n[C]} \lvert n(0)\rangle_{\mathrm{pt}}.$$

The local phase rotation was set to zero all along the path. The remaining endpoint mismatch is therefore global. It is the Berry holonomy:

$$e^{i\gamma_n[C]} = \exp \left( i\oint_C A_n \right).$$

This is the cleanest way to understand why a geometric phase can survive even when the state has been carried with “no local phase change.”

Relation to the Adiabatic Theorem

The adiabatic theorem and parallel transport answer different questions.

The adiabatic theorem says, under gap and slowness assumptions, the state stays close to the instantaneous eigenspace:

$$\lvert\psi(t)\rangle \approx e^{i\alpha_n(t)} \lvert n(t)\rangle.$$

Parallel transport says how to choose the phase of $\lvert n(t)\rangle$ along the path:

$$i\langle n(t)|\dot n(t)\rangle=0.$$

The theorem controls leakage to other eigenspaces. The transport gauge controls the phase convention inside the followed eigenline. Berry phase appears when both ideas are put together: the state follows a line, and the line has nontrivial closed-loop transport.

Relation to the Schrödinger Phase

The physical Schrödinger state is not automatically in a parallel-transport gauge. It accumulates the dynamical phase

$$-\frac{1}{\hbar} \int_0^t E_n(t')\,dt'$$

as well as the geometric phase. If one writes the adiabatic state using a parallel-transport representative, then the local geometric term is hidden in the endpoint mismatch:

$$\lvert\psi(t)\rangle \approx \exp \left( -\frac{i}{\hbar} \int_0^t E_n(t')\,dt' \right) \lvert n(t)\rangle_{\mathrm{pt}}.$$

For a closed loop,

$$\lvert n(T)\rangle_{\mathrm{pt}} = e^{i\gamma_n[C]} \lvert n(0)\rangle_{\mathrm{pt}},$$

so the final state contains both the dynamical phase and the Berry phase.

Open Paths

For an open path, one can impose $A_t=0$, but the resulting endpoint phase is not by itself gauge invariant. The initial and final fibers sit over different parameter values, and a gauge transformation can change their phase conventions separately.

To extract an observable open-path geometric phase, one must specify endpoint comparison data, an interferometric reference, or a convention such as Pancharatnam’s phase. Closed loops are the first clean case because the initial and final base point are the same.

Sphere Analogy

On a sphere, a tangent vector can be parallel transported around a loop and return rotated. Locally it was kept “as parallel as possible,” but globally the surface curvature leaves a memory of the path.

Berry parallel transport works the same way structurally:

$$\text{local no-turn rule} \quad \longrightarrow \quad \text{global mismatch}.$$

The transported object is not a tangent arrow on an ordinary sphere unless the physical model makes that identification. In the spin-$1/2$ magnetic-field example, the relevant sphere is the sphere of field directions, and the transported object is the eigenstate phase line over that sphere.

Spin-One-Half Example

For the spinor

$$\lvert +;\theta,\phi\rangle = \begin{pmatrix} \cos(\theta/2)\\ e^{i\phi}\sin(\theta/2) \end{pmatrix},$$

the Berry connection is

$$A_+ = -\frac{1-\cos\theta}{2}\,d\phi.$$

Along a path $\theta(t),\phi(t)$, the parallel-transport phase must satisfy

$$\dot\chi(t) = A_t(t) = -\frac{1-\cos\theta(t)}{2}\dot\phi(t).$$

For a closed loop, integrating this equation gives the endpoint mismatch. In the common convention for the aligned state,

$$\gamma_+[C] = -\frac{\Omega[C]}{2},$$

where $\Omega[C]$ is the oriented solid angle enclosed by the path of field directions. The worked calculation is Berry Phase for Spin-1/2.

Curvature View

Parallel transport focuses on a path. Curvature explains the infinitesimal failure of path independence. For a small loop bounding a small surface $\Sigma$,

$$\gamma_n[C] \approx \int_\Sigma F_n.$$

Thus a path-level transport rule and a local curvature field are two descriptions of the same geometry. The path description is indispensable for holonomy; the curvature description is especially useful for local calculations and topological integrals.

Common Mistakes

• Thinking $A_t=0$ means the eigenstate is physically time independent.
• Confusing removal of local geometric phase with removal of the dynamical phase.
• Assuming a parallel-transport gauge around a closed loop must return the vector to the same phase.
• Treating open-path endpoint phases as gauge invariant without endpoint comparison data.
• Forgetting that the adiabatic theorem controls transitions, while parallel transport fixes a phase convention.
• Applying the nondegenerate line-bundle condition inside a degenerate eigenspace without using the nonabelian connection.
• Taking the sphere analogy literally in systems whose parameter space is not physical space.

Cross-Links

• Berry Connection
• Holonomy
• Berry Phase
• Berry Curvature
• Adiabatic Theorem Reminder
• Berry Phase for Spin-1/2
• Born–Oppenheimer Berry Phase
• Non-Abelian Berry Phase Preview
• Projective Hilbert Space
• Parallel Transport in the Mathematical Toolkit
• Holonomy in the Mathematical Toolkit
• Connections and Curvature
• U(1) Bundles and Quantum Phase

References

• B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase,” Physical Review Letters 51, 2167-2170, 1983.
• M. V. Berry, “Quantal phase factors accompanying adiabatic changes,” Proceedings of the Royal Society A 392, 45-57, 1984.
• A. Shapere and F. Wilczek, eds., Geometric Phases in Physics, World Scientific, 1989.
• J. Anandan, “The geometric phase,” Nature 360, 307-313, 1992.
• M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.

Exercises

1. Show that $\langle n|\dot n\rangle$ is purely imaginary for a normalized state.

Solution

Differentiate $\langle n|n\rangle=1$:

$$\langle \dot n|n\rangle + \langle n|\dot n\rangle = 0.$$

Since $\langle \dot n|n\rangle=\langle n|\dot n\rangle^*$,

$$\langle n|\dot n\rangle^* = -\langle n|\dot n\rangle.$$

Therefore $\langle n|\dot n\rangle$ is purely imaginary.

2. Derive $A_t'=A_t-\dot\chi$ under $\lvert n'\rangle=e^{i\chi}\lvert n\rangle$.

Solution

Compute

$$\frac{d}{dt}\lvert n'\rangle = e^{i\chi} \left( i\dot\chi\lvert n\rangle+\lvert\dot n\rangle \right).$$

Then

$$\begin{aligned} A_t' &= i\langle n'|\dot n'\rangle\\ &= i \left( i\dot\chi+\langle n|\dot n\rangle \right)\\ &= -\dot\chi+A_t. \end{aligned}$$

3. Given $A_t(t)$ along an open path, find a phase $\chi(t)$ that sets $A_t'=0$.

Solution

The transformed connection is

$$A_t'=A_t-\dot\chi.$$

Set

$$\dot\chi=A_t.$$

One convenient choice is

$$\chi(t) = \int_0^t A_\tau(\tau)\,d\tau.$$

Then $A_t'=0$ along the path.

4. A closed loop has Berry phase $\gamma=\pi/3$. In a parallel-transport gauge, what is the relation between the final and initial eigenvector representatives?

Solution

In a parallel-transport gauge, the local connection component vanishes along the loop, but the endpoint can differ by the holonomy:

$$\lvert n(T)\rangle_{\mathrm{pt}} = e^{i\gamma} \lvert n(0)\rangle_{\mathrm{pt}}.$$

For $\gamma=\pi/3$,

$$\lvert n(T)\rangle_{\mathrm{pt}} = e^{i\pi/3} \lvert n(0)\rangle_{\mathrm{pt}}.$$

5. For the spin-$1/2$ connection $A_+=-(1-\cos\theta)d\phi/2$, what is the parallel-transport phase rate along a constant-$\theta$ loop with $\dot\phi=\omega$?

Solution

Along the path,

$$A_t = -\frac{1-\cos\theta}{2}\dot\phi = -\frac{1-\cos\theta}{2}\omega.$$

To set $A_t'=0$, choose

$$\dot\chi=A_t = -\frac{1-\cos\theta}{2}\omega.$$