Topological aspects of the superconductor

Haldane model
D. Oliver, Tatiana G. Rappoport
Instituto de Física, Centro de Ciências Matemáticas e da Natureza, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brasil

ABSTRACT
Topological phase transitions are associated with changing of topological numbers. In the Haldane model, the
electronic bands have a nontrivial topology characterized by the impossibility of defining uniquely the phase of
the wave function over the entire Brillouin zone. In this work, we consider a chiral p-wave superconductor
modelled by Haldane Hamiltonian with an on-site attractive interaction. By applying the BCS theory to this
model, we analyze its characteristics and the phase transitions as a function of the different parameters of the
Haldane model. In addition, we present some aspects of the topological nature of the appearing gap.

HALDANE MODEL – TOPOLOGICAL APPROACH

Starting with a honeycomb lattice, the Haldane model consists basically consists, if we consider the tight-
binding theory, by joining the usual hopping between nearest-neighbors (NN) with a possibility of hopping
between next-nearest-neighbors (NNN). Although, this second hopping will be complex. This consideration was
used before in Haldane’s original paper in order to understand the break of time symmetry by magnetic fields in
the quantum Hall effect, making the Haldane model the first topological insulator beyond QHE.
Typically, we also consider the NN hopping t1 to be equal between all of the atoms of the lattice. Analogically, FIGURE 5: Logarithm of the norm of the wave FIGURE 6: Logarithm of the norm of the wave FIGURE 7: Phase of the wave function for the
we do the same consideration for the NNN hopping. Once it is complex, this is valid for its module t2 and its function for the up band for a=1 nm, t1=1 eV, t2=0.2 function for the down band for a=1 nm, t1=1 eV, down band for a=1 nm, t1=1 eV, t2=0.2 eV, ø =
eV, ø = π/2 and M=0 eV. t2=0.2 eV, ø = π/2 and M=1.2 eV. π/2 and M=1.2 eV.
phase ø. In addition to that, we add a mass term M in order to break the inversion symmetry. By making all of
these considerations, we can write the Hamiltonian as: By taking a look at Figures 5 and 7 and applying the same method, we get Cup=1 by joining the three
singularities, the counter-clock wise phase vortices and its contributions to the first Brillouin zone. Thus, for a
(1) value of M greater than the topological phase transition, we observe six singularities at the logarithm of the
norm of the wave function of the upper band. But, by comparing those with the phase, we can separate the in
Since the Haldane model consists in a two-band model, its Hamiltonian is a two-by-two Hermitian matrix. So, two groups of three: those where the phase gradient goes clock wise and the ones that its gradients goes
we can separate it into the identity and Pauli matrices, I and σi: counter-clock wise, which provide a null contribution to the first BZ, what gives Cup=1.

(2) SUPERCONDUCTOR HALDANE MODEL

Here, by considering the complex phase equal to π/2, the contribution that multiplies the identity matrix
Superconductivity can be induced by proximity effects in small samples of graphene. In addition to that,
vanishes. In this case, the eigenvectors of this system are undefined all over the Brillouin zone because of
anomalous Andreev reflection and transport can be demonstrated analytically in graphene junctions with
some points where its two componentes goes to zero, what makes the phase of the wave function undefined.
superconductors. These and another superconductivity effects with graphene picks our interest in apply BCS
To avoid this, we can rewrite the eigenvectors by multiplying it by a phase. Although, if the initial points that
theory in the Haldane model, which is also honeycomb based. Here, the electron-electron interaction presents
presented problem goes right, this transformation gives us another set of points in the k-space that presentes
two superconductivity parameters: the s wave, which is associated with the on-site interaction energy and the
the same issue. So, the wave function cannot be defined univocally all over the BZ; we have to cut the BZ into
p wave, connected to the NN electron-electron interaction. Considering just the s-wave pairing and taking it
two areas to describe the Bloch waves. For example, we consider the wave function for the down band:
into account in the previous Hamiltonian gives us:

(3) (6)

(4) (5) After solving the eigen-problem, we must defined the real gap ΔR as the difference between the minimum
energy of the upper band and the maximum of the down band all over the first BZ. Here, once we have four
As we can see, the wave function of the down band is connected by a gauge transformation. The same thing bands, we must set the gap as the difference between the minimum energy of the two up bands and the
happens for the wave function of the up band. It is important to say that these singularities are not physical, maximum of the two down bands. We have to search the minimum energy between the two up bands, because
once, if we measure any observable using any one of the two expressions for the wave function, we observe no they cross each other, as we can see in Figure 8.
singularity at all. But, for the wave function and the Berry connection, which are not physical observables,
there are always some singularities for the Hamiltonian presented here. The position of these singularities in
the reciprocal space depends on the gauge choice, in other words, the phase presented in Eq. (5). In principle,
its positions also have no physical meaning at all. But, in our problem these singularities appears in the Dirac
cones K and K’. So to calculate the total Berry curvature (2π times the Chern number), we will write small
circles around the K (or K’) points and set these as Region 1. In this, we use the expression for the wave
function that does not present singularities in these. For the rest of the BZ, we set as Region 2 and use the
expression that has presented singularities inside the circles, once it will be well defined here.
Although, this way to get the Chern number, passing through the Berry parameters can be very difficult. So, in
order to get observe the topological aspects of the system, we will focus on analyzing the wave function. Once
all its two components goes to zero, we can look to the logarithm of the norm of each one of the wave
functions (one for each band) and search for points where it diverges.
FIGURE 8: Slice of the band structure at ky = 0 t for a=1 nm, t1=1 eV,
When we work with a two-component vector we can define a global phase, that will not have any contribution t2=0.2 eV, ø = π/2 and Δ = 0.5 eV.
to the norm or any other physical observable, in order to make one of the components real. In this situation,
the remaining phase in the other component will be the called as the phase of the wave function. If we are in a Now, we try to see how the gap behaves. But, instead of changing M, once we are considering a massless
point in the k-space that all components vanish, we can say that this phase is not defined univocally, which will superconductor model, we variate the superconductor parameter Δ and search for the closing/opening
be seen by the appearing of a vortex (or an anti-vortex) in the plot of this phase in the k-space. This will mechanism of the gap.
provide us the Chern number for each band.
TOPOLOGICAL PHASE TRANSITION
To analyze the topological behavior, we look to the gap between the the two bands of our problem. More
precisely, we look to its closing/opening mechanism because of the addition of the mass and its topological
feature, that will be defined by the evaluation of the Chern number. Just remembering, since the Haldane
model is a two-band one, the real gap ΔR consists to be the difference between the minimum energy of the
upper band and the maximum energy of the lower band all over the Brillouin zone.

FIGURE 9: Real gap for a=1 nm, t1=1 eV and t2=0.5 eV and ø = π/2;

FIGURE 1: Real gap for a=1 nm, t1=1 eV and t2=0.5 eV and ø = π/2;
In Figure 2, we see the closing\opening mechanism of the gap. According to the Chern number, this represents
the changing between a topological gap and an ordinary one. In addition, it can be demonstrated that it occurs
exactly at the Dirac point for M=3√3t2.
To extend what is already done to a 4-band model. To begin, we must take a look to the phase of the wave
function previously defined. After finding the four eigenvectors of the superconductor Hamiltonian, we find
that two components of each one of the eigenvectors have its respective phases defined all over the Brillouin
zone. So, they will not present the phase vortex that we have looked before. So, basically, we must look to
the norm of the wave function and to its two remaining phases and, by following the feature of the vortex on
the k-space, calculate the Chern number for each band. By doing this, we get C=-2 for the down bands ψ1- and
ψ3- and C=2 for the up bands ψ2+ and ψ4+ for Δ less than 3√3t2, analogous to the massive case. After the first
closing/opening of the gap, we get C=-2 just for the down band ψ3- and C=2 for the up band ψ4+, once ψ1- and
ψ2- provide C=0. Thus, after the second closing/opening of the gap, all bands present C=0. So, we can
conclude that we can get a topological phase transition just for some bands, since they change its Chern
numbers two at a time.

CONCLUSIONS
From our analysis, we can evaluate the Chern number for each band of our problem just by searching for
singularities in the norm of its wave function. These singularities will result in a vortex in the k-space of the
wave function phase and, just by looking to each vortex contribution into the Brillouin zone, we can find the
topological index C.

Analyzing the superconductor Haldane model itself, we conclude that a topological phase transition can
happen independently for each band of the problem, since we observe the closing/opening of the gap twice as
FIGURE 2: Logarithm of the norm of the wave FIGURE 3: Logarithm of the norm of the wave FIGURE 4: Phase of the wave function for the we increase the superconductor parameter and observe changes in Chern numbers in both of them, but for
function for the down band for a=1 nm, t1=1 eV, function for the down band for a=1 nm, t1=1 eV, down band for a=1 nm, t1=1 eV, t2=0.2 eV, ø = distinct bands.
t2=0.2 eV, ø = π/2 and M=0 eV. t2=0.2 eV, ø = π/2 and M=1.2 eV. π/2 and M=1.2 eV.
REFERENCES
In Figure 2 we can see, in red, the singularities at the norm of the wave function corresponding to this band in
the case of a massless system. By comparing this picture with the correspondent phase (Figure 4), we can [1] F. D. M. Haldane. Phys. Rev. Lett. 61, 002015 (1988).
[2] Shuichi Murakami and Naoto Nagaosa. Phys. Rev. Lett. 90, 057002 (2003).
observe that, at these points in, we get a phase vortex spinning clock wise, assuming that it consists on the [4] Bruno Uchoa and A. H. Castro Neto. Phys Rev. Lett. 98, 146801 (2007).
[5] Yi-Cai Zhang, Zhihao Xu and Shizhong Zhang. arXiv:1511.03833 (2015).
phase going from –π to π. If we consider only the first Brillouin zone, which is a hexagon in this case), we say
that each rotation contribute with one third of a complete rotation. Summing all over the three contributions
from the three singularities to the first Brillouin zone, we have one complete rotation clock wise in the first BZ. ACKNOWLEDGEMENTS
This corresponds to a Chern number Cdown=-1 for the down band.
By taking a look to the Figure 3, where we take M=1.2 eV, we can observe that the singularities disappear. Even
thought the phase on Figure 4 remains the, we only consider the phase vortex for the points where the
logarithm of the norm of the wave function diverges. So, we have Cdown=0, and conclude that we have a
topological phase transition for the down band.

E-mail : diego.oliver@if.ufrj.br