Topology of Incompressible Fermionic

Quantum Fluids
R. Shankar
The Institute of Mathematical Sciences, Chennai

IMSc@50
One day (14 March 2012) Meeting on

Topology and Differential geometry in quantum physics

Outline

The Quantum Hall Effect and Topological order

Topological Insulators

Kitaev-Hubbard Model

Outline

The Quantum Hall Effect and Topological order

Topological Insulators

Kitaev-Hubbard Model

IQHE

FQHE

H =Chern Invariant

H =No. of edge channels

Non-interacting fermions on a lattice

Z
H=

ddk
C (k ) (h(k ) F ) C(k )
(2)d

, = 1, , NB
h(k ) = U(k ) (k ) U (k )
The ground state depends only on U(k ). Many hamiltonians
will share the ground state with H. In particular if the Fermi
level lies in a gap,
Z
Q =

ddk
C (k )q(k )C(k )
(2)d

q(k ) = U(k ) 0 U (k ) 0
0

= sgn((k ) F ) 0

d = 2, NB = 2



1 (k )
0
h(k ) = U(k )
U (k )
0
2 (k )


1 0
q(k ) = U(k )
U (k )
0 1
(k ) ~
= n
Thus q(k ) defines a map from the 2-torus to the 2-sphere. This
map breaks up into topological sectors characterised by the
integer valued Chern invariant.
TKNN work showed that a physical property shared by all
ground states in a topological sector is,
H =

e2

( = Chern Invariant)

d = 2, NB bands, NF occupied

U(NB )
q(k ) = U(k )U (k )
U(NF ) U(NB NF )
Z



i
=
d 2 k ij tr q i q, j q
8

The Haldane model

Non-zero Chern number without a magnetic field

h(k ) = x px (k ) + y py (k ) + M(k )
px (k ) = t(1 + cos k1 + cos k2 )
py (k ) = t(sin k1 sin k2 )
M(k ) = M + (sin k1 + sin k2 + sin(k1 k2 ))
I

M >> , = 0, M << , = 1. Topological

transition

from = 0 to = 1 phase at M = 3 23 .

Magnetic field not necessary for non-zero Chern number.

Other time reversal symmetry breaking terms can also
induce it.

Degeneracy on a torus

Magnetic tranlations

Ta Tb Ta1 Tb1 |iN = ei2

I

If = qp , then lowest unitary irreducible representation is

q-dimensional.

If system is translationally invariant, then all eigenstates

are atleast q-fold degenerate

Chern invariant for interacting systems

Gapless edge states for interacting systems

Topological order

Topological order

Systems with topological order are characterised by

quasi-particles with fractional quantum numbers and
statistics.

Their universal properties ( quasi-particle quantum

numbers and statistics) are described by topological field
theories.

Quantum number frationisation occurs in strongly

correlated fermion systems (Hubbard models): Spinons
and Holons of RVB theory. Can be understood in the
framework of emergent gauge fields.
G. Baskaran and P.W. Anderson, Phys. Rev. B 37, 580 (1988)
T. Senthil and Fisher, Phys. Rev. B 62, 7850 (2000)

Outline

The Quantum Hall Effect and Topological order

Topological Insulators

Kitaev-Hubbard Model

Topological Insulators: Theory

Topological Insulators: Real materials

Time reversal in Quantum mechanics

2 = eiH(t2 t1 ) 1
If there is an operator T such that,
T 1 = eiH(t2 t1 ) T 2
Then the system is time reversal invariant.
Time reversal for Bloch hamiltonians,
T u(k ) = ( y u (k ))
if
y h (k ) y = h(k )
Then system is time reversal invariant

Band Pairs
T un (k ) = un (k )
T un (k ) = un (k )
n (k ) = n (k )
Fijn (k ) = Fijn (k )

The 2-d Z2 invariant

Consider a system with 2 occupied time reversed bands:
I

The total Berry flux carried by them will always be zero.

If the hamiltonian is smoothly perturbed, maintaining time

reversal invariance, and they touch (h(k ) becomes
degenerate), it will always happen at pairs of points
(k , k ).

Time reversal invariance ensures that the flux exchanged

by bands at these two points is always equal.

The change of flux in each band is always even and hence

the Chern index of each band modulo 2 is invariant under
smooth time reversal symmetric perturbations.

in general,

=
is invariant.

NB
1X

n=1

|n | modulo 2

The 2-d Z2 invariant

Consequences at the edge

Even number of edge pairs (per edge) for = 0

Odd number of edge pairs (per edge) for = 1
No backscattering for a single pair due to time reversal
symmetry.

The 3-d Z2 invariants

Parameterise the 3-d torus by ki , i = x, y , z.
I

There are many time reversal invariant planes. eg ki = 0, .

The 2-d Z2 invariants of these planes are all topological

invariants.

Of these, there are 4 independent invariants which can be

chosen to be,
x kx = , y ky = , z kz =
0 kz =0 kz =

If 0 = 1, then strong topological insulator".

Consequences at the surface

Even number of surface Dirac cones for = 0

Odd number of surface Dirac cones for = 1

Gaplessness for odd number of Dirac points protected by

time reversal symmetry.

Outline

The Quantum Hall Effect and Topological order

Topological Insulators

Kitaev-Hubbard Model

Collaborators

Syed R. Hassan
Sandeep Goyal
Archana Mishra
P.V. Sriluckshmy
(IMSc., Chennai)
David Senechal
(University of Sherbrooke)

The Kitaev Honeycomb Model

Alexei Kitaev,Anyons in an exactly solved model and beyond", Ann. Phys. (N.Y) 321, 2 (2006).

H = Jx

X
<ij>

ix jx + Jy

X
<ij>

iy jy + Jz

X
<ij>

iz jz

The Kitaev Honeycomb Model

Simple exact solution. Same degree of difficulty as 1-d

transverse field Ising model.

Solution is a spin liquid with topological order, spinons and

non-abelian anyons.

Open problem:
What is the ground state wave function in the spin basis ?
X
|GSi =
(x1 , x2 , . . . , xN )|x1 , x2 , . . . , xN i
x1 ,...,xN

A proposal for engineering the model

The Kitaev-Hubbard Model


X   t + t 0a 
X
H=
Ci
Cj + h.c + U
ni ni
2
hijia

U , Half filling:
Heff =

X
hijia

2t 0 2 a a
t 2 t 02 ~ ~
Si Sj
S S
U
U i j

(similar to) J. Chaloupka, George Jackeli and Giniyat Khaliullin, Phys. Rev. Lett. 105, 027204 (2010)

t 0 =0, Heisenberg Model

t 0 =t, Kitaev Model

The spin dependent hopping term

The t 0 term breaks time reversal symmetry.





0 a
0 a
t +t
t t
Ci
Cj + h.c T Ci
Cj + h.c
2
2
A time reversal symmetric term,




0 a
0 a
t + it
t + it
Ci
Cj + h.c T Ci
Cj + h.c
2
2
Does not give the Kitaev model at large U, half filling.

U=0



0
(k )
h(k ) =
(k )
0


0
t
t
(k ) =
f (k ) +
z + eik1 x + eik2 y
2
2
ik1
f (k ) = 1 + e
+ eik2

s
(k ) = 20 +
~ =
B

3t 0 2
t0 ~
+ |B(k
)|
4
2

1 t 0 sin k1 + cos k2 + cos k3 x


+ 1 + cos k1 t 0 sin k2 + cos k3 y


+ 1 + cos k1 + cos k2 t 0 sin k3 z

U = 0, t 0 = 0.0

U = 0, t 0 = 0.1

The bands overlap for t 0 <

6 = 0.717

U = 0, t 0 = 0.9

The bands overlap for t 0 <

6 = 0.717

U = 0: Chern Numbers

Hall conductance" at 1/4 filling

The Pancharatnam-Berry curvature for each band is given by


ij 
0
0
0
PB pp (k ) =
i pp (k ) j pp (k ) H.c.
8i
The Hall conductivity" is
Z
d 2k X
0
H =
( pp0 (k ))PB pp
2
(2)
0
pp

Angular momentum

The Orbital magnetization of Bloch electrons of the band pp0 is

Mpp0

e
2~

d 2k
( pp0 (k ))
(2)2
0

hk pp |(Hk + pp0 (k) 2)|k pp i,

Interactions

Is the physics stable against interactions ?

I

Does the gap persist ?

Do the chiral edge states persist ?

Variational Cluster Approximation

Cluster Perturbation Theory:
H = Hc + T

1
G(k , ) Gc1 () + T (k )
Variational Cluster Approximation:
Z
(h) c (h) +



d d 2 k
)Gc (k , i)
ln
det
1

T
(
k
(2)2

Minimize the grand potential with respect to the variational

parameters h.

The cluster

Variational parameters: M the total magnetization and c , the

cluster chemical potential.

The gap

The edge states

The edge states

AFQHE: Warm up problem

I

Haldane model with nearest neigbour interactions,

X
Hint = V
ni nj
ij

Mean field theory:



ni nj ij Ci Cj + h.c |ij |2

We find solutions corresponding to = 1/3 (one third of

lower band occupied).

Interpretation in terms of composite fermions: Wave

function ?

Generalising to Kitaev-Hubbard using VCPT ?

t 0 = 0 at half filling

t0 = 1
At U = 0, Dirac points persist at all t 0
At U/t >> 1,
X
~i S
~j
H = HK
S
hiji
J. Chaloupka, George Jackeli and Giniyat Khaliullin, Phys. Rev. Lett. 105, 027204 (2010)

H = HK

Siz Sjz

hiji
S. Mandal, Subhro Bhattacharjee, K. Sengupta, R. Shankar and G. Baskaran, Physical Review B 84, 155121 (2011)

The Spin Liquid remains till 0.1

From 0.1 to 1, an Antiferromagnetic phase

q = (0, ).

> 1, Neel phase q = (0, 0).

The question

Time reversal symmetry

Kitaev model (U ) is time reversal symmetric (TRS)

but microscopic theory is not. What happens at
intermediate U ?

At t 0 = 0, system is time reversal symmetric (TRS). Can

the spin liquid near t 0 = 0 be continously connected to the
spin liquid near t 0 = 1 ?

Result: Particle-Hole symmetry in this model implies that

the Mott phase is TRS. Proof to all orders in t/U, t 0 /U
perturbation theory

The Phase Diagram

Future work

Nature of the spin liquid phases at half-filling.

AFQHE states at less than quarter filling.