nLab Haldane model


For the Haldane phase see instead at Heisenberg model.



In solid state physics, the Haldane model is a model for effectively 2-dimensional crystalline Chern insulators (ie. topological phases of matter “without symmetry protection”) obtained by starting with a simple model for a graphene-like 2d semi-metal and then adding a mass term and interactions (similar to spin-orbit coupling) which break the time-reversal symmetry and the spatial inversion symmetry.


Consider the honeycomb crystal lattice and consider the following three lattice sites a i 2a_i \in \mathbb{R}^2:

(from DTC)

The graphene-like semi-metal model

The Hilbert space of quantum states of the Haldane model is the direct sum of copies of 2\mathbb{C}^2 (regarded as the defining/fundamental representation of SU(2)) indexed by the sites in a Brillouin zone of this hexagonal lattice.

The un-deformed Bloch-Hamiltonian at momentum/wave vector kk in the Brillouin torus 𝕋^ 2\widehat{\mathbb{T}}^2 is

(1)H 0(k)t 1i=13(cos(ka i)σ xσ ysin(ka i)), H_0(k) \;\coloneqq\; t_1 \cdot \underoverset{i = 1}{3}{\sum} \; \big( cos(k \cdot a_i) \sigma_x - \sigma_y sin(k \cdot a_i) \big) \,,


  1. t 1t_1 \in \mathbb{R} is some real parameter,

  2. the sum ranges over the three unit cell sites shown above,

  3. ka ik \cdot a_i denotes the canonical inner product (evaluation pairing) of the wave vector with a position space lattice vector,

  4. σ x,σ y,σ z\sigma_x, \sigma_y, \sigma_z are the Pauli matrices acting as linear operators of the on-site copies of the SU(2)-representation 2\mathbb{C}^2.

This undeformed Hamiltonian (1) is a simple but good model for the electronic band structure of a graphene-like 2-dimensional semi-metal.

Deformation by the Haldane mass term

Consider now also the following next-to-mearest neighbour site vectors b i 2b_i \in \mathbb{R}^2 in the above honeycomb lattice:

(from DTC)

The full Bloch-Hamiltonian of the Haldane model is the deformation of H 0(k)H_0(k) (1) by a mass term of the following form:

(2)H(k)H 0(k)+(M+2t 2j=13sin(kb i))σ zmassterm, H(k) \;\coloneqq\; H_0(k) \;+\; \underset{ \mathrm{mass}\;\mathrm{term} }{ \underbrace{ \left( M \;+\; 2 t_2 \underoverset{j = 1}{3}{\sum} sin(k \cdot b_i) \right) \cdot \sigma_z } } \,,


  1. t 2t_2 \in \mathbb{R} is a second real parameter

  2. the sum is now over the above three next-to-nearest neighbour site vectors b ib_i.

Notice that (only) for t 2=0t_2 = 0 this reduces to the deformation by a constant mass term (which is often understood as the default meaning of “mass term”):

(3)H(k) t 2=0=t 1i=13(cos(ka i)σ xσ ysin(ka i))+Mσ zconstantmassterm, H(k)_{t_2 = 0} \;=\; t_1 \cdot \underoverset{i = 1}{3}{\sum} \; \big( cos(k \cdot a_i) \sigma_x - \sigma_y sin(k \cdot a_i) \big) \;+\; \underset{ \mathclap{ \mathrm{constant}\;\mathrm{mass}\;\mathrm{term} } }{ \underbrace{ M \sigma_z } } \,,


Phase diagram

First of all, the Haldane model

(graphics from SS 22)

For M0M \neq 0 and as the parameter t 2t_2 (2) increases, the Haldane model passes, consecutively, through

  1. at 0t 2/M<1330 \leq t_2/M \lt \frac{1}{3 \sqrt{3}} – in particular where the mass term is constant (3) at t 2=0t_2 = 0 –, a topologically trivial insulator-phase with Berry curvature concentrated (see below) around the would-be Dirac points of the graphene-like semi-metal phase which has been gapped out by the constant mass term;

  2. at t 2/M=1330.19t_2/M = \frac{1}{3 \sqrt{3}} \sim 0.19 a non-trivial topological semi-metal-phase with a single band node;

  3. at 133<t 2/M\frac{1}{3\sqrt{3}} \lt t_2/M a non-trivial topological insulator-phase.

    This is the non-trivial Chern insulator-phase of the Haldane model.

Berry curvature

A curious property of the Haldane model, possibly not shared by all 2d Chern insulators, is that its Berry curvature is strongly localized around the (would-be) nodal Dirac points, hence that the Berry connection is essentially a flat connection on the complement of a small neighbourhood retract of the (would-be) nodal points.

(eg. Atteia 16, Sec. 2-3.4, p. 4, DTC here)

(from Atteia 16, Fig. 2.7)

Concretely, the above graphics shows the Berry curvature in the Haldane model for constant mass term, hence in the case that the interaction paramater t 2t_2 (2) vanishes:

(from DTC)

Since blue and red coloring denotes Berry curvature of opposite sign, the figure makes it plausibly manifest that the integrated Berry curvature – and thus the first Chern number of the valence bundlevanishes for M0M \neq 0, t 2=0t_2 = 0. This is the statement that the Chern insulator-phase of the Haldane model at t 2=0t_2 = 0 is topologically trivial.

In constract, as t 2/M>133t_2/M \gt \frac{1}{3 \sqrt{3}}, the Berry curvature is still concentrated around the would-be Dirac points, but now it has the same sign everywhere:

(from DTC)

This makes it clear that the integral of the Berry curvature over the Brillouin torus – hence the first Chern number of the valence bundle – is positive in this phase. This is the statement that the Chern insulator-phase of the Haldane model at t 2/M>133t_2/M \gt \frac{1}{3\sqrt{3}} is non-trivial.

K-Theory classification?

The established K-theory classification of topological phases of matter asserts that (twisted equivariant) topological K-theory of gapped valence bundles over a full Brillouin torus classifies crystalline topological insulator-phases. For the non-symmetry protected phases relavant for the Haldane model, this is the KU-theory-classification of Chern insulators.

There has not been much systematic discussion of how to generalize this statement to a classification of topological semi-metal-phases in some flavor of topological K-theory. But the above phase behavior of of the Haldane model under deformation by constant mass terms (i.e. at t 2=0t_2 = 0) supports the conjecture that topological semi-metal phases are classified by the (twisted, equivariant) flat differential K-theory of the complement T^ 2{k 1,,k N}\widehat{T}^2 \setminus \{k_1, \cdots, k_N\} of the NN nodal points k:=k 1,k 2,,k N\vec k := k_1, k_2, \cdots, k_N inside the Brillouin torus.

Namely, under this assumption the relevant K-theory charges form the following double complex of exact sequences:

(graphics from SS 22)

(To see that the boundary of this diagram is as shown, use the stable homotopy type of the punctured torus as shown there.)

Here the squiggly dashed arrow indicates the mathematical lifting problem which accurately encodes the above statement that topological charges associated to accidental/spurious band nodes (those which actually make for a trivial semi-metal phase, in that there exists a constant mass term that gaps out the node crossings), give rise to local Berry curvatures whose global topological charge cancels out and thus implies that also the globally gapped phase is topologically trivial:

(graphics from SS 22)

This accurately reflects what is seen above in the phase diagram of the Haldane mode, for constant mass terms MM (which are those that correspond to passage to K 1\mathrm{K}^{-1}, by Karoubi K-theory) but in the absence of interactions (t 2=0t_2 = 0).



The original article:

see also


Much of the above material follows:

Some of the above material is taken from:


Realization in experiment:

  • Gregor Jotzu, Michael Messer, Rémi Desbuquois, Martin Lebrat, Thomas Uehlinger, Daniel Greif, Tilman Esslinger:

    Experimental realization of the topological Haldane model with ultracold fermions, Nature 515 (2014) 237–240 (doi:10.1038/nature13915)

Last revised on June 14, 2022 at 03:06:13. See the history of this page for a list of all contributions to it.