Berry Connection

The Berry connection is the local object that records how the phase of an instantaneous eigenstate changes as parameters move. It is not itself gauge invariant, but its line integrals around closed loops and its curvature contain physical information.

For a smoothly parameter-dependent Hamiltonian,

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

the Berry connection in the convention used here is

$$\mathbf A_n(R) = i\langle n(R)|\nabla_R n(R)\rangle.$$

In differential-form notation,

$$A_n = i\langle n(R)|d n(R)\rangle.$$

The compact formula is easy to write and easy to misuse. The connection depends on a phase convention for $\lvert n(R)\rangle$, so the right question is not “what is the Berry connection at a point?” but “how does this local connection transform, and what gauge-invariant quantities does it help compute?”

Setup: A Local Eigenvector Gauge

Let $R=(R^1,\ldots,R^d)$ denote external parameters. Assume that on the region under discussion $E_n(R)$ is nondegenerate and separated from nearby levels. At each $R$, the physical instantaneous state is the ray spanned by $\lvert n(R)\rangle$, not the particular vector representative.

A smooth choice

$$R \longmapsto \lvert n(R)\rangle$$

is a local gauge. Another equally valid choice is

$$\lvert n'(R)\rangle = e^{i\chi(R)} \lvert n(R)\rangle.$$

The Berry connection is built from the chosen representative, so it is local gauge data. This is the same kind of distinction that appears between a vector potential and the magnetic field, but now the “space” is parameter space and the gauge freedom is the quantum phase of an eigenvector.

Coordinate and One-Form Forms

In coordinates, write

$$A_i^{(n)}(R) = i\langle n(R)|\partial_i n(R)\rangle, \qquad \partial_i=\frac{\partial}{\partial R^i}.$$

Then the one-form is

$$A_n = \sum_i A_i^{(n)}\,dR^i.$$

Along a parameter path $R(t)$, the connection pulls back to

$$A_t(t) = \sum_i A_i^{(n)}(R(t))\dot R^i(t) = i\left\langle n(R(t)) \middle| \frac{d}{dt}n(R(t)) \right\rangle.$$

For a closed loop $C$, the Berry phase is

$$\gamma_n[C] = \oint_C A_n = \oint_C \mathbf A_n(R)\cdot dR.$$

The corresponding phase factor is $e^{i\gamma_n[C]}$. This is the object that survives a change of local gauge.

Why the Connection Is Real

The eigenvector is normalized:

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

Differentiating gives

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

Since

$$\langle \partial_i n|n\rangle = \langle n|\partial_i n\rangle^*,$$

the quantity $\langle n|\partial_i n\rangle$ is purely imaginary. Multiplying by $i$ makes

$$A_i^{(n)} = i\langle n|\partial_i n\rangle$$

real. The Berry connection is therefore a real-valued phase connection in this convention.

Gauge Transformation

Under

$$\lvert n(R)\rangle \mapsto \lvert n'(R)\rangle = e^{i\chi(R)} \lvert n(R)\rangle,$$

one finds

$$\begin{aligned} A_n' &= i\langle n'|d n'\rangle \\ &= i\left( i\,d\chi+\langle n|dn\rangle \right) \\ &= A_n-d\chi. \end{aligned}$$

In vector notation,

$$\mathbf A_n' = \mathbf A_n-\nabla_R\chi.$$

Thus $\mathbf A_n$ is not an observable field. It is a connection: its transformation law tells different phase conventions how to represent the same underlying geometry.

For a closed loop,

$$\gamma_n[C] \mapsto \gamma_n[C]-\oint_C d\chi.$$

If $\chi$ is single-valued on the loop, the change is an integer multiple of $2\pi$, so $e^{i\gamma_n[C]}$ is unchanged.

Parallel-Transport Gauge

The Berry connection measures the phase rotation of the chosen eigenvector along a path. A parallel-transport gauge along $R(t)$ sets

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

On an open interval this can always be done locally by choosing a phase $\chi(t)$ satisfying

$$\frac{d\chi}{dt} = A_t(t).$$

Around a closed loop, however, the gauge that keeps $A_t=0$ along the trip may return to the initial ray with a different vector phase. That mismatch is the holonomy, and in an adiabatic quantum evolution it is the Berry phase after the dynamical phase has been removed.

This is the cleanest way to read the connection: it is the rule that says which phase choice counts as “not rotating” as the eigenspace moves.

Relation to Berry Phase

The Berry phase page focuses on the physical cyclic adiabatic evolution:

$$\alpha_n = -\frac{1}{\hbar} \int_0^T E_n(t)\,dt + \gamma_n[C].$$

The connection page isolates the geometric term:

$$\gamma_n[C] = \oint_C A_n.$$

The dynamical phase depends on elapsed time and energy. The connection integral depends on how the eigenspace twists over parameter space. This is why two protocols with the same timing but different parameter loops can acquire different geometric phases, and why two protocols with the same loop but different speeds can share the same geometric phase in the adiabatic limit.

Relation to Curvature

The Berry curvature is obtained by differentiating the connection:

$$F_n=dA_n.$$

In local coordinates,

$$F_{ij}^{(n)} = \partial_i A_j^{(n)} - \partial_j A_i^{(n)}.$$

In three-dimensional vector notation this is often written as

$$\mathbf B_n = \nabla_R\times\mathbf A_n.$$

Unlike $A_n$, the abelian curvature $F_n$ is gauge invariant. Berry Curvature develops this local field and its applications. The mathematical derivation and projector formulas are collected in Berry Connection as a Mathematical Object and the compact band-theory formula card Berry Curvature.

When a loop $C$ bounds a surface $\Sigma$ and a smooth gauge is available on the surface, Stokes’ theorem gives

$$\oint_C A_n = \int_\Sigma F_n.$$

This statement is powerful, but it has assumptions: the eigenstate must be smooth on the surface, the surface must avoid degeneracies, and patching may be needed if no single gauge covers the region.

Example: A Real Local Gauge

Suppose a normalized eigenvector can be chosen real and smooth on a region. Then

$$\langle n|\partial_i n\rangle$$

is both real and purely imaginary, so it must vanish:

$$A_i^{(n)}=0.$$

This does not mean all geometric phase effects disappear globally. A real gauge may fail to be single-valued around a loop, or it may be impossible to choose smoothly across the entire parameter region. The Berry connection is local; global phase information can survive even when the connection is zero in a particular patch.

Example: Spin One-Half Local Gauge

For the spinor

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

one finds

$$A_+ = i\langle +;\theta,\phi\rvert d\lvert +;\theta,\phi\rangle = -\frac{1-\cos\theta}{2}\,d\phi.$$

For a loop at fixed polar angle $\theta$, with $\phi$ running from $0$ to $2\pi$,

$$\gamma_+ = \oint A_+ = -\pi(1-\cos\theta).$$

This is $-\Omega/2$, where

$$\Omega = 2\pi(1-\cos\theta)$$

is the solid angle enclosed by the loop on the parameter sphere. The sign changes with eigenstate and Hamiltonian convention, so the convention must be stated before quoting the result. The physical worked example is Berry Phase for Spin-1/2.

Common Mistakes

• Treating the Berry connection as gauge invariant.
• Forgetting that $\lvert n(R)\rangle$ is a local representative of a ray, not a unique physical vector.
• Applying the nondegenerate $U(1)$ formula through a degeneracy or level crossing.
• Using Stokes’ theorem across a surface where no smooth eigenvector gauge exists.
• Confusing the Berry connection with an electromagnetic vector potential in physical space. The analogy is useful, but the base space and gauge freedom are different.
• Dropping the dynamical phase in a physical evolution without explicitly separating it from the geometric phase.
• Quoting the spin-$1/2$ solid-angle sign without specifying the eigenstate and Hamiltonian convention.

Cross-Links

• Berry Phase
• Adiabatic Theorem Reminder
• Berry Curvature
• Holonomy
• Parallel Transport
• Berry Phase Problems
• Chern Numbers
• Non-Abelian Berry Phase Preview
• Berry Phase for Spin-1/2
• Berry Connection as a Mathematical Object
• Tangent and Cotangent Spaces
• Differential Forms
• Connections and Curvature
• Parallel Transport
• Holonomy
• U(1) Bundles and Quantum Phase
• Integration on Manifolds
• Projective Hilbert Space
• From Berry Phase to Topological Terms
• Berry Curvature Formula Card

References

• M. V. Berry, “Quantal phase factors accompanying adiabatic changes,” Proceedings of the Royal Society A 392, 45-57, 1984.
• B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase,” Physical Review Letters 51, 2167-2170, 1983.
• A. Shapere and F. Wilczek, eds., Geometric Phases in Physics, World Scientific, 1989.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
• D. J. Griffiths and D. F. Schroeter, Introduction to Quantum Mechanics, 3rd ed., Cambridge University Press, 2018.
• M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.

Exercises

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

Solution

Differentiate the normalization condition:

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

Since

$$\langle \partial_i n|n\rangle = \langle n|\partial_i n\rangle^*,$$

we get

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

Thus the quantity is purely imaginary.

2. Derive the gauge transformation $A_n'=A_n-d\chi$.

Solution

Use

$$\lvert n'\rangle = e^{i\chi}\lvert n\rangle.$$

Then

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

Therefore

$$\begin{aligned} A_n' &= i\langle n'|dn'\rangle \\ &= i \left( i\,d\chi+\langle n|dn\rangle \right) \\ &= A_n-d\chi. \end{aligned}$$

3. For $A_+=-(1-\cos\theta)d\phi/2$, compute the Berry phase around a loop of constant $\theta$.

Solution

Along the loop, $\phi$ runs from $0$ to $2\pi$, so

$$\gamma_+ = \oint A_+ = -\frac{1-\cos\theta}{2} \int_0^{2\pi}d\phi.$$

Thus

$$\gamma_+ = -\pi(1-\cos\theta).$$