Skip to content

Chern Numbers

A Chern number is a quantized flux of Berry curvature through a closed two-dimensional parameter space. In the simplest nondegenerate U(1)U(1) case,

Cn=12π∫MFn,C_n = \frac{1}{2\pi} \int_M F_n,

where MM is a closed oriented surface and FnF_n is the Berry-curvature two-form of an isolated eigenstate line bundle over MM.

The formula is short, but every word matters. A Chern number is not just any Berry-curvature integral. The surface must be closed, the eigenstate or band must be defined over the whole surface, and the relevant spectral gap must stay open. Under those assumptions the result is an integer that cannot change under smooth deformations.

The mathematical canonical page is Chern Numbers. This page explains the physics meaning, assumptions, and examples.

For a nondegenerate eigenstate family

H(R)∣n(R)⟩=En(R)∣n(R)⟩,H(R)\lvert n(R)\rangle = E_n(R)\lvert n(R)\rangle,

choose a local phase convention and define the Berry connection

An=i⟨n∣dn⟩.A_n = i\langle n|dn\rangle.

The Berry curvature is

Fn=dAn.F_n=dA_n.

Locally, the Berry phase around a small loop is the curvature flux through a small spanning surface. Globally, the Chern number integrates this curvature over a closed surface:

Cn=12π∫MFn.C_n = \frac{1}{2\pi} \int_M F_n.

The factor 1/(2π)1/(2\pi) appears because quantum phases are defined modulo 2π2\pi. In this convention, the holonomy phase factor is ei∮Ane^{i\oint A_n}.

The standard U(1)U(1) Chern number requires:

  • a closed oriented two-dimensional parameter space MM;
  • an isolated nondegenerate eigenstate, or isolated Bloch band, over all of MM;
  • a Berry curvature FnF_n defined globally, even if the Berry connection is only local;
  • a fixed orientation and normalization convention.

In band theory, MM is often the two-dimensional Brillouin-zone torus. In a two-level spin example, MM can be a sphere surrounding a degeneracy in parameter space.

If the state ceases to be isolated at some point of MM, the Chern number for that state is not defined. This is why gap closings are central to topological phase transitions.

The integer comes from patching local phase choices. A single smooth eigenvector gauge may not exist on all of MM. On overlapping patches, two local eigenvectors differ by a phase:

∣nβ⟩=eiχβα∣nα⟩.\lvert n_\beta\rangle = e^{i\chi_{\beta\alpha}} \lvert n_\alpha\rangle.

The Berry connections differ by

Aβ=Aα−dχβα.A_\beta = A_\alpha-d\chi_{\beta\alpha}.

When the curvature integral is split across patches, Stokes’ theorem turns the answer into the winding of the transition phase around overlaps. Single-valued transition functions can wind only an integer number of times around U(1)U(1), so

12π∫MFn∈Z.\frac{1}{2\pi} \int_M F_n \in \mathbb Z.

The toolkit page gives the detailed two-patch derivation. The physics lesson is that the integer is a global obstruction to choosing one smooth phase convention everywhere.

If a smooth global Berry connection AnA_n existed on a closed surface, then

Fn=dAnF_n=dA_n

globally. Stokes’ theorem would give

∫MFn=∫MdAn=∫∂MAn=0,\int_M F_n = \int_M dA_n = \int_{\partial M}A_n = 0,

because ∂M=∅\partial M=\varnothing. Therefore a nonzero Chern number means that no single smooth global eigenvector gauge covers the entire surface.

This is not a failure of quantum mechanics. It is the point: the state family is globally twisted even though it looks ordinary on small patches.

For the spin-1/21/2 Berry phase convention used in this volume,

F+=−12sin⁡θ dθ∧dϕ.F_+ = -\frac12 \sin\theta\,d\theta\wedge d\phi.

Over the full sphere,

∫S2F+=−2π.\int_{S^2}F_+ = -2\pi.

Thus

C+=12π∫S2F+=−1.C_+ = \frac{1}{2\pi} \int_{S^2}F_+ = -1.

The degeneracy at zero field acts as a Berry-curvature monopole enclosed by the sphere. The opposite eigenstate has the opposite Chern number in the matching convention. The sign depends on the Hamiltonian, eigenstate, and orientation choices; the integer quantization does not.

For a two-dimensional isolated Bloch band, the parameter space is the Brillouin zone:

BZ≃T2.\mathrm{BZ} \simeq T^2.

In a common notation, the band Chern number is

Cn=12π∫BZΩn(k) d2k,C_n = \frac{1}{2\pi} \int_{\mathrm{BZ}} \Omega_n(\mathbf k)\,d^2k,

where Ωn\Omega_n is the Berry-curvature component compatible with the chosen orientation. Equivalently,

Cn=12π∫BZFn.C_n = \frac{1}{2\pi} \int_{\mathrm{BZ}}F_n.

This number is stable under smooth changes of the Hamiltonian that keep the band isolated over the whole Brillouin zone.

In a noninteracting two-dimensional band insulator, filled isolated bands can contribute to the Hall conductivity through their Chern numbers:

σxy=e2h∑occupied nCn,\sigma_{xy} = \frac{e^2}{h} \sum_{\text{occupied }n}C_n,

up to orientation and electron-charge sign conventions.

This formula is the integer quantum Hall bridge, not a complete derivation. It assumes a filled-band setting, a gap at the Fermi level, and the usual linear-response framework. Chern Numbers in Band Theory develops occupied projectors, direct and indirect gaps, the Kubo reduction, symmetry constraints, and numerical computation. Integer Quantum Hall Effect adds Landau filling, disorder localization, chiral edges, flux insertion, plateau transitions, and metrology. The role of this page is to explain the integer carried by Berry curvature.

If several occupied bands are separated from empty bands but not from each other, individual band Chern numbers may be gauge dependent or ill-defined. The meaningful object is the total Chern number of the occupied subspace.

In nonabelian Berry-connection language, the first Chern number is

Cocc=12π∫MTr⁡Focc.C_{\mathrm{occ}} = \frac{1}{2\pi} \int_M \operatorname{Tr}F_{\mathrm{occ}}.

This is a preview of nonabelian Berry geometry. The trace is gauge invariant under changes of basis inside the occupied subspace. The detailed nonabelian construction belongs to the later preview page and to the mathematical toolkit.

Chern numbers are robust because they are integer-valued and cannot change continuously. If a Hamiltonian is deformed smoothly while the relevant state or occupied subspace remains isolated over all of MM, the Chern number remains fixed.

To change a Chern number, something singular must happen. In band problems this is usually a gap closing:

En(k,λ)=Em(k,λ)E_n(\mathbf k,\lambda) = E_m(\mathbf k,\lambda)

at some parameter value λ\lambda. At that point, the eigenbundle whose Chern number was being tracked is no longer defined. After the gap reopens, curvature flux can have been transferred between bands.

This is why Chern-number changes are tied to topological phase transitions.

The Dirac monopole shows that flux through a closed sphere can be quantized by single-valued quantum patching. Berry monopoles show the same structure in parameter space: degeneracies act as sources of Berry-curvature flux.

For a closed surface MM surrounding a degeneracy, the Chern number measures the net Berry monopole charge enclosed by MM. If no degeneracy crosses the surface during a smooth deformation, the integer cannot change.

This is the common geometry behind:

  • Dirac monopole quantization;
  • spin-1/21/2 solid-angle Berry phase;
  • Chern bands over the Brillouin-zone torus;
  • quantum Hall response in filled-band settings.
  • Integrating curvature over an open patch and calling the result a Chern number.
  • Forgetting the isolation or gap assumption for the eigenstate or band.
  • Treating the local Berry curvature distribution as topological. The integral is topological; the local distribution can move.
  • Assuming a nonzero Chern number can be represented by one smooth global eigenvector.
  • Ignoring orientation and sign conventions in CnC_n and σxy\sigma_{xy}.
  • Assigning separate Chern numbers to bands that are degenerate or mixed inside the occupied subspace.
  • Treating the Hall-conductivity formula as valid without checking the filled-band and linear-response assumptions.
  • 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. V. Berry, “Quantal phase factors accompanying adiabatic changes,” Proceedings of the Royal Society A 392, 45-57, 1984.
  • D. Xiao, M.-C. Chang, and Q. Niu, “Berry phase effects on electronic properties,” Reviews of Modern Physics 82, 1959-2007, 2010.
  • M. Nakahara, Geometry, Topology and Physics, 2nd ed., CRC Press, 2003.
  • B. A. Bernevig and T. L. Hughes, Topological Insulators and Topological Superconductors, Princeton University Press, 2013.
  1. Compute a constant-curvature Chern number.

Let MM be a torus with coordinates 0≤kx,ky<2π0\le k_x,k_y<2\pi and

F=N2πdkx∧dky,N∈Z.F = \frac{N}{2\pi} dk_x\wedge dk_y, \qquad N\in\mathbb Z.

Compute C=(1/2π)∫MFC=(1/2\pi)\int_MF.

Solution

The integral is

∫MF=N2π∫02π∫02πdkx dky=2πN.\int_MF = \frac{N}{2\pi} \int_0^{2\pi} \int_0^{2\pi} dk_x\,dk_y = 2\pi N.

Therefore

C=12π∫MF=N.C = \frac{1}{2\pi} \int_MF = N.
  1. Explain why a nonzero Chern number obstructs a global gauge.
Solution

If one smooth global connection one-form AA existed on a closed surface MM, then F=dAF=dA everywhere. Stokes’ theorem would give

∫MF=∫MdA=∫∂MA=0.\int_MF = \int_MdA = \int_{\partial M}A = 0.

Since a nonzero Chern number means ∫MF≠0\int_MF\ne0, it rules out a single smooth global gauge.

  1. Compute the spin-1/21/2 Chern number.

Use

F+=−12sin⁡θ dθ∧dϕF_+ = -\frac12 \sin\theta\,d\theta\wedge d\phi

on S2S^2 with the standard orientation.

Solution

Compute

∫S2F+=∫02π∫0π−12sin⁡θ dθ dϕ=−2π.\begin{aligned} \int_{S^2}F_+ &= \int_0^{2\pi} \int_0^\pi -\frac12 \sin\theta\,d\theta\,d\phi\\ &= -2\pi. \end{aligned}

Therefore

C+=12π∫S2F+=−1.C_+ = \frac{1}{2\pi} \int_{S^2}F_+ = -1.
  1. Why can a Chern number change when a gap closes?
Solution

The Chern number is assigned to an isolated eigenstate, isolated band, or isolated occupied subspace. If a gap closes, that object is no longer defined over the full parameter space. The assumptions behind the integer invariant fail at the gap closing. After the gap reopens, the reconstructed band or subspace can have a different Chern number because Berry-curvature flux has passed through the singular point.

  1. What is wrong with calling ∫ΣF/(2π)\int_\Sigma F/(2\pi) a Chern number for an open surface Σ\Sigma?
Solution

For an open surface, the integral can change continuously under gauge-compatible deformations and can depend on boundary data. The integer quantization argument uses a closed surface and patching around closed overlaps. On an open surface, ∫ΣF/(2π)\int_\Sigma F/(2\pi) can be a useful Berry-curvature flux, but it is not generally a topological Chern number.