Chern Numbers

A Chern number is an integer obtained by integrating the curvature of a complex vector bundle over a closed even-dimensional parameter space. For a $U(1)$ line bundle over a closed oriented surface $M$, the first Chern number is

$$C = \frac{1}{2\pi} \int_M F,$$

where $F$ is the real curvature two-form in the convention that parallel transport has phase factor $e^{i\int A}$.

This page explains the two-dimensional $U(1)$ case most often used in Berry phase and quantum Hall physics. Higher Chern classes and characteristic-class theory belong to more advanced topology pages.

For the physics-side local field whose integral appears here, see Berry Curvature.

Why Chern Numbers Matter

Chern numbers are the curvature-integral version of a topological invariant. They appear when local geometric data cannot be patched into one globally trivial description.

In quantum mechanics, the most common examples are:

• Berry curvature of an eigenstate line bundle over a closed parameter surface;
• Berry curvature of a Bloch band over a two-dimensional Brillouin-zone torus;
• the integer in the Thouless–Kohmoto–Nightingale–den Nijs description of the integer quantum Hall effect;
• the flux of an effective monopole through a parameter sphere around a degeneracy.

The key point is not merely that an integral happens to be an integer. The integer is stable under smooth deformations that do not close the relevant spectral gap or change the bundle topology.

Required Data

For the basic $U(1)$ Chern number, one needs:

• a closed oriented two-dimensional manifold $M$;
• a complex line bundle $L\to M$;
• a $U(1)$ connection with local one-forms $A_\alpha$;
• a globally defined curvature two-form $F$.

The local connection forms depend on gauge choices:

$$A_\beta = A_\alpha-d\chi_{\beta\alpha}$$

on overlaps. The curvature satisfies

$$F=dA_\alpha=dA_\beta$$

on each patch and is globally defined.

The first Chern number is

$$C_1(L) = \frac{1}{2\pi} \int_M F.$$

The factor $2\pi$ matches the phase convention for $U(1)$ holonomy. Some mathematics texts use anti-Hermitian connection forms and write equivalent formulas with a factor of $i/2\pi$. The invariant content is the same after translating conventions.

Why the Integral Is Quantized

Quantization comes from patching. Suppose a closed oriented surface $M$ is covered by two patches $U_N$ and $U_S$ whose overlap deformation-retracts to a circle. On the overlap, the local frames are related by

$$e_S=e^{i\chi}e_N.$$

With the convention used here,

$$A_S=A_N-d\chi.$$

Since $F=dA_N=dA_S$, split the integral into the two patches and use Stokes theorem. With compatible orientations, the boundary terms combine into

$$\int_M F = \oint_{\text{overlap}} d\chi.$$

The transition function $e^{i\chi}$ is single-valued on the overlap circle. Therefore $\chi$ can change by only an integer multiple of $2\pi$ after one circuit:

$$\oint d\chi = 2\pi n, \qquad n\in\mathbb Z.$$

Thus

$$\frac{1}{2\pi}\int_M F = n.$$

The exact sign of $n$ depends on the orientation convention for the surface and overlap circle, but the conclusion is invariant: the curvature integral is an integer.

Nonzero Chern Number Means No Global Gauge

If a smooth global connection one-form $A$ existed on all of a closed surface $M$, then

$$F=dA$$

globally. Stokes theorem on a manifold without boundary would give

$$\int_M F = \int_M dA = \int_{\partial M}A = 0.$$

Therefore a nonzero Chern number implies that no single smooth global gauge covers the entire surface. One must use patches, transition functions, or a formulation in terms of globally defined projectors.

This is the same warning that appears in Berry-phase calculations: a singular-looking local gauge may be a patch artifact, while the obstruction to removing all such patching is topological.

Berry Curvature Formula

For a nondegenerate eigenstate line bundle over a closed oriented surface $M$, the Berry connection is locally

$$A_n = i\langle n\rvert dn\rangle,$$

and the Berry curvature is

$$F_n=dA_n.$$

The corresponding Chern number is

$$C_n = \frac{1}{2\pi} \int_M F_n.$$

If the eigenstate is described by the projector

$$P_n=\lvert n\rangle\langle n\rvert,$$

then in local coordinates

$$F_{ij}^{(n)} = i\,\operatorname{Tr} \left( P_n \left[ \partial_iP_n, \partial_jP_n \right] \right).$$

This projector formula is often safer than an eigenvector formula when no global smooth phase convention exists.

Spin-1/2 Sphere Example

For the local spinor convention used in Berry Connection as a Mathematical Object,

$$F_+ = -\frac12\sin\theta\,d\theta\wedge d\phi.$$

Integrate over the unit sphere with the standard orientation:

$$\begin{aligned} \int_{S^2}F_+ &= -\frac12 \int_0^\pi\int_0^{2\pi} \sin\theta\,d\theta\,d\phi\\ &= -2\pi. \end{aligned}$$

Therefore

$$C_+ = \frac{1}{2\pi}\int_{S^2}F_+ = -1.$$

The opposite eigenline has the opposite Chern number in the corresponding convention. This example is often described as a unit Berry monopole at the degeneracy enclosed by the parameter sphere. The sign depends on which eigenstate, Hamiltonian sign, and surface orientation are chosen.

Constant Curvature on a Torus

Let $T^2$ have coordinates $(k_x,k_y)$ with periods $2\pi$. If

$$F = \frac{N}{2\pi}\,dk_x\wedge dk_y, \qquad N\in\mathbb Z,$$

then

$$\int_{T^2}F = \frac{N}{2\pi}(2\pi)(2\pi) = 2\pi N.$$

Thus

$$C=\frac{1}{2\pi}\int_{T^2}F=N.$$

This model captures the normalization used for Bloch bands over a Brillouin-zone torus. A real band curvature need not be constant, but its total integral is still an integer when the isolated band defines a line bundle over the closed Brillouin zone.

Quantum Hall Connection

In the integer quantum Hall effect for noninteracting electrons in a periodic two-dimensional system, a filled isolated Bloch band contributes an integer Chern number over the Brillouin-zone torus. In common conventions, the Hall conductivity is proportional to the sum of occupied-band Chern numbers:

$$\sigma_{xy} = \frac{e^2}{h} \sum_{\text{occupied }n} C_n,$$

up to sign conventions for electron charge, Berry-curvature orientation, and coordinate orientation.

This page does not derive linear response or the full quantum Hall effect. It supplies the mathematical invariant that the physics pages use: a curvature integral over a closed parameter space that cannot change under smooth gap-preserving deformations. Integer Quantum Hall Effect is the canonical material treatment of plateaus, edges, localization, and metrology.

Stability and Gap Closing

Chern numbers are stable because they are integer-valued and vary continuously under smooth deformations of the connection and bundle data. A continuous path of gapped Hamiltonians cannot change $C_n$ for an isolated band unless the assumptions fail.

The usual failure mode is a gap closing. If the band touches another band, the eigenline may cease to be well-defined at that parameter value. After such a singular event, the Chern number assigned to a band can change.

This is why topological phase transitions are often associated with spectral gap closings. The Chern number is not a local curvature value; it is a global invariant of an isolated family.

Relation to Winding

Winding counts how a loop wraps around a one-dimensional target such as $U(1)$ or around a missing point in the plane. A Chern number is a higher-dimensional cousin: it counts the obstruction encoded by curvature over a closed surface.

The two are connected by patching. In the two-patch argument above, the Chern number becomes the winding of the transition function

$$e^{i\chi}:S^1\to U(1)$$

around the overlap circle. Thus the curvature-integral integer and the transition-function winding are two views of the same topology.

Common Mistakes

• Treating any integral of Berry curvature over any surface as a Chern number. The surface must be closed, oriented, and part of a well-defined bundle problem.
• Forgetting the normalization factor $1/(2\pi)$ in the $U(1)$ convention used here.
• Assuming a nonzero Chern number can be represented by one smooth global eigenvector gauge.
• Ignoring orientation and sign conventions.
• Calling a local Berry curvature peak a topological invariant. The integral is topological; the local distribution can move under deformations.
• Applying a band Chern number when the band is not isolated across the whole Brillouin zone.
• Treating the quantum Hall formula as convention-free without specifying charge and orientation conventions.

Cross-Links

• Berry Connection as a Mathematical Object
• U(1) Bundles and Quantum Phase
• Fiber Bundles, First Look
• Integration on Manifolds
• Exterior Derivative
• Holonomy
• Homotopy and Winding
• Topological Invariants
• Berry Phase
• Chern Numbers in Quantum Mechanics
• Integer Quantum Hall Effect

References

• S. S. Chern, “Characteristic classes of Hermitian manifolds,” Annals of Mathematics 47, 85-121, 1946.
• D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs, “Quantized Hall conductance in a two-dimensional periodic potential,” Physical Review Letters 49, 405-408, 1982.
• B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase,” Physical Review Letters 51, 2167-2170, 1983.
• M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.
• T. Frankel, The Geometry of Physics, 3rd ed., Cambridge University Press, 2011.
• M. Z. Hasan and C. L. Kane, “Colloquium: Topological insulators,” Reviews of Modern Physics 82, 3045-3067, 2010.
• B. C. Hall, Quantum Theory for Mathematicians, Springer, 2013.

Exercises

1. Let $F=(N/2\pi)\,dk_x\wedge dk_y$ on a torus with $0\leq k_x,k_y\lt2\pi$. Compute the Chern number.

Solution

The integral is

$$\int_{T^2}F = \frac{N}{2\pi} \int_0^{2\pi}\int_0^{2\pi} dk_x\,dk_y = 2\pi N.$$

Therefore

$$C = \frac{1}{2\pi} \int_{T^2}F = N.$$

2. Why does a nonzero Chern number forbid a smooth global gauge on a closed surface?

Solution

If a smooth global gauge existed, the curvature would be globally $F=dA$. Since the surface has no boundary, Stokes theorem would give

$$\int_M F = \int_M dA = \int_{\partial M}A = 0.$$

A nonzero value of $(1/2\pi)\int_MF$ therefore contradicts the existence of one smooth global gauge.

3. For $F_+=-\frac12\sin\theta\,d\theta\wedge d\phi$ on $S^2$, compute $C_+$ with the standard orientation.

Solution

Compute

$$\begin{aligned} \int_{S^2}F_+ &= -\frac12 \int_0^\pi\int_0^{2\pi} \sin\theta\,d\theta\,d\phi\\ &= -\frac12(2)(2\pi)\\ &= -2\pi. \end{aligned}$$

Thus

$$C_+ = \frac{1}{2\pi} \int_{S^2}F_+ = -1.$$

4. In the two-patch argument, why must $\oint d\chi$ be an integer multiple of $2\pi$?

Solution

The transition function is $e^{i\chi}$ on the overlap circle. For it to be single-valued after one circuit, the phase can change only by a multiple of $2\pi$:

$$\chi(1)-\chi(0)=2\pi n.$$

Therefore

$$\oint d\chi=2\pi n.$$

5. A two-band model is smoothly deformed while its lower band remains separated from the upper band everywhere on the Brillouin-zone torus. Can the lower-band Chern number change?

Solution

No. Under a smooth deformation that preserves the isolated-band condition over the entire torus, the Chern number varies continuously. Since it is integer-valued, it must remain constant. To change it, the band must cease to be isolated somewhere, usually through a gap closing.

6. Explain the relation between the Chern number and winding in the two-patch construction.

Solution

The curvature integral over the closed surface reduces, by Stokes theorem on two patches, to an integral of $d\chi$ around the overlap circle. The transition function $e^{i\chi}:S^1\to U(1)$ has an integer winding number. That winding is precisely the integer obtained from $(1/2\pi)\int_MF$, up to the orientation convention.