Fiber Bundles, First Look

A fiber bundle is a space that looks locally like a product of a base space and a typical fiber, even if it is globally twisted or patched together in a nontrivial way.

The basic picture is:

$$\pi:E\to M,$$

where $M$ is the base space, $E$ is the total space, and the fiber over a point $p\in M$ is

$$E_p=\pi^{-1}(p).$$

In quantum mechanics, fiber bundles organize families of vector spaces, phase choices, eigenspaces, tangent spaces, and gauge frames over parameter spaces.

Helpful background. Manifolds, First Look introduces coordinate patches; Vector Spaces and Dual Spaces introduces local bases.

Families of spaces over a parameter manifold

For a Hamiltonian $H(R)$, each parameter value $R$ has an eigenspace. A bundle organizes that family of spaces, including how local bases fit together. A connection adds a rule for comparing vectors in neighboring fibers; the bundle and the comparison rule are distinct pieces of mathematics.

The Data of a Bundle

A fiber bundle consists of:

• a total space $E$;
• a base space $M$;
• a projection map $\pi:E\to M$;
• a typical fiber $F$;
• local identifications of $\pi^{-1}(U)$ with $U\times F$ on sufficiently small patches $U\subset M$.

The projection map tells which base point an element of $E$ lies over. If $e\in E$, then

$$\pi(e)=p$$

means that $e$ belongs to the fiber over $p$.

The fiber over $p$ is the set

$$E_p = \{e\in E:\pi(e)=p\}.$$

For vector bundles, each fiber $E_p$ is a vector space. For principal bundles, each fiber looks like a group. This first-look page mostly uses vector and phase examples.

Local Product, Global Twisting

The defining feature is local product structure. For each small enough patch $U\subset M$, there is a local trivialization

$$\phi_U:\pi^{-1}(U)\to U\times F$$

that respects the projection to the base:

$$\pi^{-1}(U) \quad \text{looks like} \quad U\times F.$$

This does not mean the whole bundle is globally a product $M\times F$. Different patches may be glued by nontrivial transition functions on overlaps.

On an overlap $U_\alpha\cap U_\beta$, two local trivializations are related by a transition function

$$g_{\alpha\beta}:U_\alpha\cap U_\beta\to G,$$

where $G$ is the structure group acting on the fiber. For a complex line bundle, $G$ is often $U(1)$. For a rank-$k$ complex vector bundle, $G$ may be $U(k)$ after choosing Hermitian inner products on the fibers.

Trivial Bundle

The simplest bundle is the product bundle

$$E=M\times F, \qquad \pi(p,f)=p.$$

The fiber over $p$ is

$$E_p=\{p\}\times F.$$

This is globally untwisted. A single global product coordinate describes every fiber at once.

Many first calculations silently assume a trivial bundle. That is often harmless locally, but it can miss global phase, winding, and topological effects.

Sections

A section is a choice of one element in each fiber. Formally, a section is a map

$$s:M\to E$$

such that

$$\pi\circ s=\operatorname{id}_M.$$

In words: $s(p)$ lies in the fiber over $p$.

For a product bundle $E=M\times F$, a section is the same as a function

$$f:M\to F,$$

because

$$s(p)=(p,f(p)).$$

Every vector bundle has a global zero section, $s(p)=0_p$. The subtle question is whether one can choose a nowhere-zero section, or a whole basis of sections forming a global frame. For a line bundle, a nowhere-zero global section would trivialize the bundle. General fiber bundles need not have a global section at all.

Frames and Gauge Choices

For a rank-$k$ vector bundle, a local frame on a patch $U$ is a choice of basis

$$e_1(p),\ldots,e_k(p)$$

for each fiber $E_p$ over $p\in U$.

In a local frame, a section can be written

$$s(p)=s^a(p)e_a(p).$$

A different local frame changes the component functions $s^a(p)$. This is the bundle-theoretic source of gauge transformations: the geometric section is the same, but its local coordinate representative changes.

For a complex line bundle, a local frame is a nowhere-zero local vector $e(p)$. A section is

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

Changing the frame by a phase

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

changes the component by the inverse phase so that $s(p)$ itself is unchanged.

Tangent and Cotangent Bundles

The tangent bundle of a manifold $M$ is

$$TM = \bigcup_{p\in M}T_pM.$$

Its projection sends a tangent vector to the point where it lives:

$$\pi(v_p)=p.$$

A vector field is a section of $TM$:

$$X(p)\in T_pM.$$

Similarly, the cotangent bundle is

$$T^*M = \bigcup_{p\in M}T_p^*M.$$

A one-form is a section of $T^*M$:

$$\alpha(p)\in T_p^*M.$$

Higher differential forms are sections of exterior-power bundles such as $\Lambda^kT^*M$.

These examples explain why tangent spaces at different points should not be casually identified. They are fibers in a bundle, not one single vector space repeated by default.

Quantum Eigenstate Bundles

Let a Hamiltonian depend smoothly on parameters:

$$H(R).$$

Suppose a nondegenerate eigenvalue $E_n(R)$ remains separated from the rest of the spectrum on a parameter region $M$. At each $R\in M$, the eigenspace is a complex line

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

The collection of these lines forms a complex line bundle over $M$:

$$L = \bigcup_{R\in M}L_R.$$

A local normalized eigenvector choice is a local section of the associated unit vector representatives. It is not unique:

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

This phase freedom is not a nuisance added after the fact. It is the local-frame freedom of a complex line bundle.

Degenerate Eigenspace Bundles

If an isolated eigenspace has dimension $k$, then each parameter value has a $k$-dimensional fiber:

$$E_R=\operatorname{span} \{ \lvert n,1;R\rangle,\ldots,\lvert n,k;R\rangle \}.$$

The collection of these eigenspaces is a rank-$k$ complex vector bundle, as long as the subspace varies smoothly and stays isolated from the rest of the spectrum.

A local orthonormal basis of the degenerate subspace is a local frame. On overlaps, frames can be related by $U(k)$-valued transition functions. This is the setting for nonabelian Berry connections and Wilczek-Zee holonomy.

This page does not develop the nonabelian adiabatic theorem. The important bundle point is simpler: degeneracy turns a phase line into a higher-dimensional vector fiber.

Connections Are Extra Structure

A bundle by itself tells which fiber lies over which base point. It does not tell how to compare nearby fibers.

That comparison rule is a connection. With a connection, one can define:

• covariant derivatives of sections;
• parallel transport along paths;
• curvature;
• holonomy around loops.

Without a connection, it is meaningful to say $s(p)\in E_p$ and $s(q)\in E_q$, but not to subtract them as if they lived in the same vector space.

This separation is especially important in Berry phase. The eigenstate line bundle organizes the fibers; the Berry connection gives the transport rule; holonomy gives the closed-loop phase.

Local Sections and Global Obstructions

A bundle can look trivial on every small patch while being globally nontrivial.

The Möbius band is the standard real example: locally it looks like an interval fiber over a circle, but going once around the base reverses the fiber orientation. A single local product picture cannot describe the whole band without a twist.

Complex line bundles can have analogous global structure. A local eigenvector may be smooth on one patch but fail to extend smoothly and single-valuedly over the whole parameter space. This is not merely bad notation; it can signal nontrivial topology and lead to topological invariants on later pages.

Common Mistakes

• Thinking a bundle is always globally a product $M\times F$.
• Confusing the base point $p$ with an element of the fiber over $p$.
• Treating a section as a number rather than a fiber-valued object.
• Assuming a local frame is physically or geometrically unique.
• Using a local eigenvector formula as if it were automatically global.
• Asking for parallel transport before specifying a connection.
• Treating Berry phase as mysterious after ignoring the underlying line bundle and its gauge freedom.

Exercises

1. For the product bundle $E=M\times F$, show that sections are equivalent to maps $f:M\to F$.

Solution

A section $s$ obeys

$$\pi(s(p))=p.$$

In the product bundle, every element of the fiber over $p$ has the form $(p,f)$ with $f\in F$. Therefore any section must be

$$s(p)=(p,f(p))$$

for some map $f:M\to F$. Conversely, any such map defines a section by this formula.

2. Explain why a vector field is a section of the tangent bundle.

Solution

The tangent bundle is the union of tangent spaces:

$$TM=\bigcup_{p\in M}T_pM.$$

A vector field assigns to each point $p$ a tangent vector $X(p)\in T_pM$. The projection sends that tangent vector back to its base point:

$$\pi(X(p))=p.$$

Thus $X:M\to TM$ satisfies $\pi\circ X=\operatorname{id}_M$, which is exactly the definition of a section.

3. Let two local eigenvector choices on an overlap be related by $\lvert n_\beta(R)\rangle=e^{i\chi(R)}\lvert n_\alpha(R)\rangle$. What is the transition function?

Solution

The transition function is the $U(1)$-valued function

$$g_{\beta\alpha}(R)=e^{i\chi(R)}.$$

It tells how the local frame or representative on patch $\alpha$ is converted to the one on patch $\beta$.

4. Why does a fiber bundle not automatically define parallel transport?

Solution

The bundle tells which fiber sits over each base point and how local product descriptions are patched together. It does not specify how an element of one fiber should be compared with an element of a nearby fiber. Parallel transport requires a connection, which is extra structure on the bundle.

5. What bundle is formed by an isolated $k$-fold degenerate eigenspace over a parameter manifold?

Solution

If the degenerate subspace varies smoothly and remains isolated from the rest of the spectrum, the fibers are $k$-dimensional complex vector spaces. Their union forms a rank-$k$ complex vector bundle over the parameter manifold. Local choices of orthonormal eigenbasis are local frames, related on overlaps by $U(k)$ transformations.

6. Why can a local eigenvector formula fail to define a global section?

Solution

A local formula may depend on a coordinate patch or phase convention that breaks down elsewhere. On overlaps, different formulas can be related by nontrivial transition functions. For a complex line bundle, a nowhere-zero global frame exists precisely when compatible changes of the local frames make all transition functions equal to one. If this is impossible, the obstruction belongs to the bundle’s global topology, not merely to algebraic inconvenience. Even a trivial bundle can carry a connection with nontrivial holonomy; U(1) Bundles and Quantum Phase distinguishes a global frame from a global parallel frame.

References

• T. Frankel, The Geometry of Physics, 3rd ed., Cambridge University Press, 2011.
• J. M. Lee, Introduction to Smooth Manifolds, 2nd ed., Springer, 2013.
• M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.