advanced30 min studyLesson 408 in Physics

Berry connection and gauge dependence

Compute a geometric connection with an explicit sign convention and separate local gauge freedom from loop holonomy.

Before you begin

On this page

The question for today

What you’ll understand

Calculate the connection and curvature of a normalized two-state eigenvector family.
Track the connection under a parameter-dependent eigenvector phase.
State the gap and patch assumptions behind geometric phases and Chern integrals.

An eigenvector can change smoothly as control parameters vary, yet its phase at each parameter value remains arbitrary. The Berry connection records how neighboring choices compare. Its value depends on that phase convention, while suitable closed-loop phase factors and the associated curvature are gauge invariant. This distinction is the quantum version of describing a geometric object with coordinates that can change. A connection is useful precisely because its convention-dependent pieces combine into quantities an experiment can compare.

A local phase choice creates a connection

Let $|u(\boldsymbol\lambda)\rangle$ be a normalized, nondegenerate eigenvector chosen smoothly on a patch of parameter space. Define the connection with the convention $\mathcal A_\mu=i\langle u|\partial_\mu u\rangle$ . Differentiating normalization shows the inner product is purely imaginary, so the connection is real. If a parameter has units, the corresponding connection has inverse units, ensuring its line integral is dimensionless. The eigenvector family is meaningful only where the selected eigenvalue remains isolated from the rest of the spectrum.

Changing the local representative to $e^{i\chi(\boldsymbol\lambda)}|u\rangle$ changes the connection by minus the derivative of the phase. The sign follows directly by differentiating the exponential and multiplying by the leading $i$ . Some sources define the connection with the opposite sign and adjust their phase formulas accordingly. Use one complete convention throughout. A gauge change relabels the same family of physical rays; changing the relative components of a state generally changes the ray and is not merely this gauge freedom.

\mathcal A_\mu=i\langle u|\partial_\mu u\rangle,\qquad |u^{\prime}\rangle=e^{i\chi}|u\rangle,\qquad \mathcal A^{\prime}_\mu=\mathcal A_\mu-\partial_\mu\chi

The dimensionless chi is a smooth local phase. This sign convention gives the adiabatic geometric factor exp(i gamma) with gamma equal to the connection line integral in a periodic gauge.↔ Scroll wide equations horizontally.

Curvature survives a change of representative

The antisymmetric derivative of the connection is its curvature. Mixed derivatives of a smooth gauge phase cancel, making this quantity gauge invariant. In a simply covered smooth patch, Stokes' theorem equates a connection integral around a boundary to the curvature integral through its oriented surface. If no single smooth eigenvector covers the surface, patch transition phases must be included. Applying a one-patch formula across a singular gauge choice can lose an essential phase contribution.

For a normalized spinor pointing along a unit direction, take the positive upper component real. Its connection vanishes in the polar direction and equals minus the squared lower-component magnitude in the azimuthal direction. The curvature is nonzero even though a different gauge can make the connection vanish along one particular path segment. Thus a locally removable phase accumulation along a segment does not imply that every loop has trivial geometric phase or that the whole family admits one global smooth eigenvector.

|u\rangle=\begin{pmatrix}\cos(\theta/2)\\e^{i\phi}\sin(\theta/2)\end{pmatrix},\quad \mathcal A_\theta=0,\quad \mathcal A_\phi=-\frac{1-\cos\theta}{2},\quad \mathcal F_{\theta\phi}=\partial_\theta\mathcal A_\phi=-\frac{\sin\theta}{2}

Theta and phi are dimensionless angles in radians. This gauge is smooth near the north pole but singular as a representative at the south pole; the physical ray remains regular there.↔ Scroll wide equations horizontally.

Holonomy and topology require different qualifications

For a closed parameter loop with a consistent endpoint identification, the observable geometric object is the phase factor, not a particular real number chosen to represent its angle. A single-valued gauge phase factor can change that angle by an integer multiple of a full turn. The spinor family above gives minus half the oriented solid angle. Changing the enclosed solid angle changes the phase continuously, so an arbitrary Berry phase is geometric and need not be quantized or protected against path deformations.

A Chern number is a stronger global construction. On a closed oriented two-dimensional parameter manifold, an isolated nondegenerate band defines a complex line bundle, and the curvature integral divided by a full turn is an integer. Smoothness, a maintained spectral isolation, and global patch compatibility are essential assumptions. A gap closing can invalidate the bundle and allow the integer to change. The curvature of this parameter-space family should not be confused with an ordinary magnetic field living in real laboratory space.

e^{i\gamma(C)}=\exp\!\left(i\oint_C\mathcal A_\mu\,d\lambda^\mu\right),\qquad C_1=\frac{1}{2\pi}\int_{\mathcal M}\mathcal F\in\mathbb Z

The loop expression uses a periodic gauge or the appropriate endpoint/patch factors. The Chern integral assumes a smooth isolated eigenline over a closed oriented surface M; F is the curvature two-form.↔ Scroll wide equations horizontally.

Work it through

Two gauges on the same latitude

For the displayed spinor, traverse $\theta=\pi/3$ once with $\phi$ increasing from zero to $2\pi$ . Compute the geometric phase in the given gauge and in the gauge $|u^{\prime}\rangle=e^{i\phi}|u\rangle$ . Compare their phase factors.

At this latitude $\cos\theta=1/2$ , so $\mathcal A_\phi=-1/4$ . Integrating over one positive azimuthal turn gives $\gamma=-\pi/2$ . Both endpoint vectors agree in this periodic gauge.
The new phase has derivative one, giving $\mathcal A^{\prime}_\phi=-1/4-1=-5/4$ . The new loop integral is $\gamma^{\prime}=-5\pi/2$ . The multiplying phase also agrees at the endpoints, so it is an admissible loop gauge.
The two answers differ by $-2\pi$ , whose exponential is one. Both geometric factors equal $-i$ . Comparing connection values alone would mistakenly assign different physics to the same eigenvector rays.

The result

The connection and phase angle change with gauge, while the closed-loop geometric factor remains −i.

The living notebook

Explore the model

Interactive lab

Interactive lab · quantum

Latitude geometry is visible; gauge phase is not

Reset and set both controls to $(60^\circ,0^\circ)$ , then Play. Predict constant Bloch z=0.5 and basis probabilities 0.75 and 0.25 as the azimuth precesses. This traces the spinor family's latitude. The lab contains no common-phase control, connection readout, spectral gap, or Berry-phase accumulator, so the loop phase must be calculated separately.

Bloch vector / P(0)P(1)Phase latitude

Probability |0⟩: 71.1%
Probability |1⟩: 28.9%
Bloch (x, y, z): 0.78, 0.45, 0.42

Mixing angle θ65°Relative phase φ30°

t = 0 s

A normalized two-level pure state, not an interacting field simulation. Playback adds one radian per second to the relative phase; basis probabilities stay fixed. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forA ray path supplies geometric data, while a gauge representative and a phase calculation supply its holonomy.

Pause & reason

Check your understanding

Interactive checks need JavaScript. You can still read the explanation.

Read the checkpoint explanation

Differentiating the gauge phase gives the minus d chi term. Its exterior derivative vanishes for a smooth phase, and compatible loop endpoint factors preserve the holonomy.

Your turn

Try it yourself

In a dimensionless planar parameter patch take $\mathcal A_x=-by/2$ and $\mathcal A_y=bx/2$ , with $b=1/4$ . Traverse the rectangle $0\leq x\leq2$ , $0\leq y\leq1$ counterclockwise. Compute the loop phase and show what the gauge $\chi=xy$ does to it.

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Compute the curvature as the x derivative of Ay minus the y derivative of Ax. A gauge gradient has zero closed-loop integral when the phase is smooth and single valued.

Compare with the solution

The curvature is $b/2-(-b/2)=1/4$ . Multiplying by the oriented area two gives $\gamma=1/2$ radians. Directly, the right and top edges contribute $1/4$ each and the other two contribute zero. The new connection is $(-by/2-y,bx/2-x)$ . The extra line integral is minus the net change of xy around the loop, which is zero, so the phase remains one half radian.

Keep the idea

Take it with you

A Berry connection depends on eigenvector phase, with a sign fixed by its definition.
Curvature and compatible closed-loop phase factors are gauge invariant.
Quantized Chern integrals require a globally isolated eigenline over a closed oriented surface.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.