U(1) Bundles and Quantum Phase

A $U(1)$ bundle is the geometric structure behind a smoothly varying phase choice. In quantum mechanics, it is the natural language for global phase freedom, local gauge choices, Berry phase, and Aharonov–Bohm holonomy.

The group $U(1)$ is the circle group of unit complex numbers:

$$U(1) = \{e^{i\theta}:\theta\in\mathbb R\}.$$

Multiplying a normalized state vector by an element of $U(1)$ changes its representative but not its ray:

$$\lvert\psi\rangle \mapsto e^{i\theta}\lvert\psi\rangle.$$

Changing a phase convention never changes a physical prediction. What can be observable is the phase produced by transport around a loop, compared with a reference path; that relative phase is independent of the convention used to calculate it.

Helpful background. Fiber Bundles, First Look explains local frames, while Rays and Global Phase explains why normalized vectors have phase freedom.

Global phase versus local gauge

For one isolated state vector, the transformation

$$\lvert\psi\rangle\mapsto e^{i\theta}\lvert\psi\rangle$$

does not change probabilities or expectation values. This is the ray principle explained in Rays and Global Phase.

For a family of states depending on a parameter $R$, the phase may vary:

$$\lvert\psi(R)\rangle \mapsto e^{i\chi(R)}\lvert\psi(R)\rangle.$$

This is a gauge choice over the parameter space. The local representative changes, but the ray at each point is the same. Derivatives and path integrals can change under this local phase choice unless they are assembled into gauge-invariant quantities.

Complex Line Bundles

A complex line bundle assigns a one-dimensional complex vector space to each point of a base space $M$:

$$L_p\simeq\mathbb C, \qquad p\in M.$$

Choosing a nonzero local vector $e(p)$ in each fiber is a local frame. A section can be written as

$$s(p)=c(p)e(p).$$

Changing the local frame by a phase

$$e(p)\mapsto e^{i\chi(p)}e(p)$$

is a $U(1)$ gauge transformation. The component $c(p)$ changes in the opposite way so that the section $s(p)$ itself is unchanged.

This is the bundle version of a familiar quantum fact: the representative vector can change by phase while the underlying ray or line is fixed.

Principal U(1) Bundle Picture

There are two closely related pictures:

• a complex line bundle, whose fibers are copies of $\mathbb C$;
• a principal $U(1)$ bundle, whose fibers are phase choices.

For a normalized state representative, the allowed phases form a circle. If a ray is fixed, all normalized representatives of that ray differ by $U(1)$:

$$\lvert\psi\rangle, \quad e^{i\theta}\lvert\psi\rangle.$$

Thus the phase fiber over a ray is a copy of $U(1)$. The associated complex line bundle lets complex amplitudes live over the same base.

This page uses whichever picture is more transparent in context. For most Berry phase calculations, the complex line bundle and its $U(1)$ gauge freedom are enough.

Gauge Transformations on Overlaps

Suppose two patches $U_\alpha$ and $U_\beta$ overlap. A local frame on $U_\alpha$ and a local frame on $U_\beta$ can be related by

$$e_\beta(p) = g_{\beta\alpha}(p)e_\alpha(p),$$

where

$$g_{\beta\alpha}(p)\in U(1).$$

Thus

$$g_{\beta\alpha}(p)=e^{i\chi_{\beta\alpha}(p)}.$$

The functions $g_{\beta\alpha}$ are transition functions. They tell how local phase conventions are glued together. To test whether the line bundle is trivial, change each local frame by a smooth phase $h_\alpha:U_\alpha\to U(1)$. The new transition functions are

$$e_\alpha'=h_\alpha e_\alpha, \qquad g_{\beta\alpha}'=h_\beta g_{\beta\alpha}h_\alpha^{-1}.$$

If compatible choices of the $h_\alpha$ make every $g_{\beta\alpha}'=1$, the new local frames agree on overlaps and form one nowhere-zero global frame. This is global triviality. Multiplying every frame by the same globally defined phase would leave the $U(1)$ transition functions unchanged; it is the freedom to choose frames separately on the patches that tests whether they can be glued into a global one.

U(1) Connections

A $U(1)$ connection supplies a rule for comparing phases at nearby base points. With the convention used in the nearby geometry pages, the line-bundle connection one-form is written

$$\Gamma=-iA,$$

where $A$ is a real one-form in a chosen local frame.

Parallel transport along a path $C$ gives the phase factor

$$\exp \left( i\int_C A \right).$$

Under a frame change

$$e\mapsto e^{i\chi}e,$$

the connection one-form transforms as

$$A\mapsto A-d\chi.$$

The curvature is

$$F=dA.$$

It is gauge invariant:

$$d(A-d\chi) = dA.$$

Berry Line Bundle

For a nondegenerate eigenstate of a parameter-dependent Hamiltonian,

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

the physical eigenspace at each $R$ is the line

$$L_R = \{\lambda\lvert n(R)\rangle:\lambda\in\mathbb C\}.$$

The collection of lines $L_R$ over parameter space is the Berry line bundle. A choice of normalized eigenvector $\lvert n(R)\rangle$ is a local gauge choice.

The Berry connection in that gauge is

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

Under

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

it transforms as

$$A_n\mapsto A_n-d\chi.$$

The Berry curvature is

$$F_n=dA_n.$$

In a smooth eigenvector frame that is periodic around the loop, the geometric phase factor is the $U(1)$ holonomy

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

If the frame does not return to the same vector, include the endpoint overlap as explained in Berry Connection as a Mathematical Object. Berry Phase explains the adiabatic conditions under which this holonomy appears in physical evolution.

Aharonov–Bohm Phase

The Aharonov–Bohm effect also has a $U(1)$ bundle interpretation. A charged particle’s wavefunction is locally described by complex phases, and the electromagnetic vector potential acts as a $U(1)$ connection.

In the common minimal-coupling convention $p\mapsto p-qA$, transport around a closed loop gives a phase factor

$$\exp \left( \frac{iq}{\hbar} \oint_C A_i\,dx^i \right).$$

If the accessible region excludes a flux tube, the magnetic field may vanish along the particle paths while the holonomy around the excluded region is nontrivial. This is a global feature of the $U(1)$ connection over the punctured configuration space.

The unit convention and loop-level interpretation are summarized in Holonomy. This page emphasizes the phase-bundle structure.

Trivial and Nontrivial Phase Bundles

On a contractible patch of a smooth parameter manifold, a $U(1)$ bundle admits one smooth phase convention. Locally, it looks like

$$U\times U(1).$$

Globally, the bundle may require several patches. Nontrivial transition functions can prevent a single smooth phase convention from covering the entire base space.

This is why gauge singularities should be treated carefully. A singular-looking local phase convention may indicate a bad patch, not a singular physical state. Conversely, the impossibility of removing all patching data can encode real topology.

A global frame need not be parallel

Bundle triviality does not say that a connection can be gauged to zero. Consider the product line bundle $S^1\times\mathbb C$ with a fixed global unit frame and connection $A=\alpha\,d\theta$, where $\alpha\in\mathbb R$ is dimensionless. Here $d\theta$ is a smooth one-form on the circle even though no single real-valued angle $\theta$ covers it. Every complex line bundle over a circle is in fact trivial; see Hatcher, Example 1.12. Nevertheless, the connection has holonomy $e^{2\pi i\alpha}$, as calculated in Holonomy.

To set $A'=A-d\chi=0$ requires $d\chi=\alpha\,d\theta$, so locally $\chi=\alpha\theta+\text{constant}$. The frame change must be a single-valued function on the circle:

$$e^{i\chi(\theta+2\pi)}=e^{i\chi(\theta)} \quad\Longleftrightarrow\quad e^{2\pi i\alpha}=1 \quad\Longleftrightarrow\quad \alpha\in\mathbb Z.$$

The phase $e^{i\chi}$ must be single-valued; a chosen real lift $\chi$ need not be periodic. More generally, a gauge transformation with integer winding $m$ shifts $\alpha$ to $\alpha-m$. Thus $\alpha$ modulo integers labels the inequivalent connections in this family. Varying it continuously changes the holonomy while leaving the product bundle unchanged. Nontrivial holonomy here obstructs a global parallel frame, not a global frame.

Phase, Curvature, and Topology

For a $U(1)$ connection, the local curvature is the two-form

$$F=dA.$$

If $C=\partial\Sigma$ and a smooth gauge exists on $\Sigma$, then

$$\oint_C A = \int_\Sigma F.$$

But global topology can matter even when $F=0$ on the accessible region. A flat connection on a circle or punctured plane can have nontrivial holonomy because the loop cannot be contracted without leaving the space.

Thus $U(1)$ geometry has two complementary diagnostics:

• curvature measures local phase twisting;
• holonomy measures closed-loop phase transport.

Homotopy and Winding explains the loop-level version of this statement. Chern Numbers explains how curvature flux over closed surfaces becomes quantized, and Topological Invariants explains the deformation-stability language common to both.

Common Mistakes

• Concluding that all phase structure is unphysical because one global phase is unobservable.
• Treating a local gauge choice as if it were a globally defined state.
• Forgetting that $A$ is gauge dependent while $F=dA$ is gauge invariant.
• Applying Stokes theorem across a surface where the gauge or state is not defined.
• Confusing the Berry line bundle over parameter space with the particle’s physical-space wavefunction.
• Mixing the Berry convention $A_n\mapsto A_n-d\chi$ with an electromagnetic convention that uses different signs.
• Assuming a $U(1)$ bundle must be trivial because every small patch looks like a product.

Exercises

1. Show that $U(1)$ is a group under multiplication.

Solution

Closure follows because

$$e^{i\theta}e^{i\phi} = e^{i(\theta+\phi)} \in U(1).$$

The identity is $1=e^{i0}$. The inverse of $e^{i\theta}$ is

$$e^{-i\theta}.$$

Associativity is inherited from complex multiplication.

2. For the Berry connection $A_n=i\langle n\rvert dn\rangle$, show that $\lvert n\rangle\mapsto e^{i\chi}\lvert n\rangle$ gives $A_n\mapsto A_n-d\chi$.

Solution

Compute

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

3. Show that $F=dA$ is invariant under $A\mapsto A-d\chi$.

Solution

The transformed curvature is

$$F' = d(A-d\chi) = dA-d^2\chi.$$

Since $d^2=0$,

$$F'=dA=F.$$

4. On the product line bundle over a circle, let $A=\alpha\,d\theta$ in a global unit frame. Compute the $U(1)$ holonomy for one loop with increasing $\theta$. For $\alpha=\tfrac12$ and $\alpha=1$, decide whether a smooth single-valued gauge transformation can set $A$ to zero. Does either connection require a nontrivial bundle?

Solution

The holonomy phase factor is

$$\exp \left( i\oint A \right) = \exp \left( i\int_0^{2\pi}\alpha\,d\theta \right) = e^{i2\pi\alpha}.$$

For $\alpha=\tfrac12$, the holonomy is $-1$. The candidate $e^{i\theta/2}$ changes sign after one turn and is not a gauge transformation on the circle, so the connection cannot be gauged to zero. For $\alpha=1$, the holonomy is $1$ and $e^{i\theta}$ is single-valued; with $\chi=\theta$ locally, $A'=d\theta-d\theta=0$. Both connections were defined on the same trivial product bundle. Their different holonomies concern the connection, not the existence of a global frame.

5. Why does a global phase not matter for one state, while a Berry phase can matter around a loop?

Solution

A single global phase multiplies all amplitudes of one state by the same factor, so it cancels in probabilities and expectation values. A Berry phase compares phase choices transported around a path in parameter space. For a closed loop, the endpoint ray is the same but the transported representative may differ by a gauge-invariant phase factor. To observe it, compare the transported state with a coherent reference, for example by recombining two interferometer arms. A phase multiplying a single isolated state still cancels from its probabilities.

6. Why can a singular-looking gauge potential be a patch artifact?

Solution

A local gauge potential is written in a chosen frame or patch. If that frame fails somewhere, the potential can look singular even when the underlying bundle and curvature are regular there. One must check whether another patch removes the singularity and whether transition functions on overlaps carry the real global information.

References

• M. V. Berry, “Quantal phase factors accompanying adiabatic changes”, Proceedings of the Royal Society A 392, 45-57, 1984.
• A. Hatcher, Vector Bundles & K-Theory, version 2.2, 2017, Example 1.12, p. 24.
• M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.
• B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase”, Physical Review Letters 51, 2167-2170, 1983.