Populated the namespace with:
np, matplotlib, kwant, holoviews, init_notebook, interact, display_html, plt, pf, SimpleNamespace, pprint_matrix, scientific_number, pretty_fmt_complex
MoocVideo, PreprintReference, MoocDiscussion, MoocCheckboxesAssessment, MoocMultipleChoiceAssessment, MoocPeerAssessment, MoocSelfAssessment
spectrum, hamiltonian_array, h_k, pauli
HoloViewsJS, MatplotlibJS successfully loaded in this cell.
Press this button to show/hide the code used in the notebook:
Quantum Hall effect on a lattice and the Dirac Hamiltonian
Pairs of chiral edges in a 1D wire
Coupling wires and QHE without field
Dirac equation at the phase transition
Introduction: stacking wires
Looking back at the material from the past weeks, you might have the impression that the quantum Hall effect and one dimensional topological superconductors are really different topics, and not connected at all.
Xiaoliang Qi from Stanford University will now explain that this is not the case, and will also introduce this week's topic - Chern insulators.
Pairs of chiral edges in a 1D wire
Last week, we started with a phenomenological description of the quantum Hall effect. We learned that one way to get a quantum Hall state is to place electrons in an external magnetic field.
However, we still don't have a simple lattice Hamiltonian describing the quantum Hall effect - we'd like to have something like the Kitaev chain model, which was very useful to understand Majoranas in the first two weeks of the course. An added benefit of finding this tight binding model is that it would not need an external magnetic field to exhibit the unique properties of quantum Hall effect.
These models exist, and they are referred to as Chern insulators. The quantum Hall effect without an external magnetic field is also referred to as the quantum anomalous Hall effect. Duncan Haldane, from who we will hear in the next chapter, invented the first model of a Chern insulator now known as Haldane model. However, in this chapter, we will use a more natural route that fits better into the context of our course so far.
General strategy to construct a lattice model for the Quantum Hall Effect
What we will do to get a model for a Chern insulator is to follow a "domino prescription", as we did to get the Kitaev model in week 1. Our strategy will have two key aspects:
Focusing on the unique property of the quantum Hall edge that cannot exist in isolation from a bulk, and can only be present because of bulk-boundary correspondence. For the Kitaev chain, this property was the presence of unpaired Majorana modes. In the quantum Hall effect, it is the chiral edge states.
Finding a lower dimensional building block, from which we can somehow “extract“ in a clever way the exotic object we are interested in. In the Kitaev chain, these were the fermionic sites, which we could think of as a pair of Majorana modes.
So we need to find a one-dimensional system which can host a pair of chiral edge states. Since they are one-dimensional, these states will necessarily be spatially on top of each other, essentially a pair of propagating modes with opposite velocity. Once we have this building block, we can follow the domino prescription: couple the counter-propagating states in pairs, just like we did with Majorana modes. We then end up with a pair of spatially separated chiral edge states, and so a quantum Hall insulator without magnetic field.
So our plan is to turn the system on the left into the system on the right:
Voilà - we have a lattice model for the 2D quantum Hall state!
Getting a one dimensional wire with a pair of edge states
Let's focus on the first essential step of our plan: finding a one dimensional system with a pair of counter-propagating chiral states.
You might guess that the easiest way to get a pair of states moving in opposite directions is to take the usual one dimensional single electron Hamiltonian with an effective mass. Such a system has a parabolic dispersion and the low energy excitations at a finite Fermi wave-vector k=±kF move in opposite directions. The catch is that these two states have different momenta, and it is tricky (though not impossible) to deal with momentum conservation when pairing the edges. So we will avoid this approach.
The more educated guess we make a model that we have already encountered, the Dirac model H=Δkτy of the Kitaev chain at the topological phase transition.
More specifically, in the first week we wrote the Kitaev Hamiltonian in momentum space as H(k)=−(2tcosk+μ)τz+Δsinkτy. At the critical point μ=−2t, we found the Kitaev Hamiltonian to become
We see that at k≈0 we have a pair of states with wave functions the eigenvalues ±1 of τy, and with opposite and equal velocities.
At this point you might worry that the Kitaev model has superconductivity, and so the τ matrices refer to particle and hole degrees of freedom. Indeed, these ingredients should not enter in the description of the quantum Hall effect. But this is not a real issue, because we can just interpret the τ matrices as acting in the space of left- and right-movers. This flexibility in interpreting the Hamiltonian terms in different ways, depending on the context, is part of the advantage of thinking in terms of toy models.
Finally, before we go on with our plan, keep in mind that considering the phase transition point of a lower dimensional model turns out to be a fairly generic strategy to construct higher dimensional topological models.
question=("It seems that both a quantum Hall bar and a Kitaev chain can have chiral states. ""Apart from the two systems having different dimensionality"", what's the fundamental difference between the two cases?")answers=["The quantum Hall edge states go in opposite directions, while the Kitaev states go in the same direction.","The quantum Hall edge states go in the same direction, while the Kitaev states go in opposite directions.","The quantum Hall edges always cross zero energy at zero momentum while the Kitaev states don't.",("The Kitaev chiral states exist only at specific parameter values, ""while the quantum Hall edge states don't.")]explanation=("The pair of chiral states in the Kitaev model only exists at ""the phase transition point, when the chain becomes gapless. ""On the other hand, chiral edge states are a topological property of the quantum Hall state. ""They are separated by a gapped bulk which protects them, and they exist for a full range of parameter values.")MoocMultipleChoiceAssessment(question,answers,correct_answer=3,explanation=explanation)
It seems that both a quantum Hall bar and a Kitaev chain can have chiral states. Apart from the two systems having different dimensionality, what's the fundamental difference between the two cases?
The correct answer is: The Kitaev chiral states exist only at specific parameter values, while the quantum Hall edge states don't. The pair of chiral states in the Kitaev model only exists at the phase transition point, when the chain becomes gapless. On the other hand, chiral edge states are a topological property of the quantum Hall state. They are separated by a gapped bulk which protects them, and they exist for a full range of parameter values.
QHE without a magnetic field
Deriving the model Hamiltonian
Let us now couple the wires to get the quantum Hall system as promised. We take a stack of chains all extending along the horizontal x direction, like in the figure above. We stack them one next to the other along the y direction, so that we form a square lattice. If we take a large (ideally infinite) stack of chains we have a truly two-dimensional system.
Now let us make this formal by first labeling the chains by an index ny, which takes integer values. Let us also replace k→kx to denote the wave-vector along a chain. Hence a single chain has the Hamiltonian [−(2tcoskx+μ)τz+Δsinkxτy]⊗∣ny⟩⟨ny∣. The projector ∣ny⟩⟨ny∣ is needed to single out one chain from the stack.
Now all that we have to do is to couple the τy=−1 branch of one chain to the τy=+1 branch of a neighboring chain, and we will have a quantum Hall state.
A term coupling opposite movers from different chains is ∣ny⟩⟨ny+1∣⊗(τz+iτx). The first part couples neighboring chains and the matrix (τz+iτx) turns a right mover into a left mover, which is what we want. Let's call the strength of this coupling γ.
To obtain the complete Hamiltonian of the stack of chains we just need to sum over ny, and we obtain
This Hamiltonian should in principle suffice to produce a quantum Hall state.
Gap and edge states
Let's first consider a finite number of chains ny=1,…,N. If you tune each chain individually to the point μ=−2t, it's easy to check that at kx≈0 the above Hamiltonian has one right-moving edge eigenstate for ny=1 with eigenvalue ≈Δkx, and a left-moving edge eigenstate for ny=N with eigenvalue ≈−Δkx.
The next thing we need to check is that these are the only eigenstates close to zero energy. In other words, is the system gapped in the two-dimensional bulk? To see this, let's switch to an infinite stack of chains along the y direction. We then have full translational invariance, so we can go to momentum space in the y direction.
This leads us to the two-dimensional Bloch Hamiltonian
Aside from special points, this spectrum is gapped, just like we wanted. For instance it is gapped if μ<−2t−2γ. If we start from this point and increase the value of μ, the gap closes at the point μ=−2t−2γ and then reopens: