The Quantum Spin Hall Effect and Topological Band Theory

E E

k=a

k=b

k=a

k=b

The Quantum Spin Hall Effect and Topological Band Theory

I. II. Introduction
- Topological band theory

Two Dimensions : Quantum Spin Hall Insulator

- Time reversal symmetry & Edge States - Experiment: Transport in HgCdTe quantum wells

III. Three Dimensions : Topological Insulator

- Topological Insulator & Surface States - Experiment: Photoemission on BixSb1-x and Bi2Se3

IV. Superconducting proximity effect

- Majorana fermion bound states - A platform for topological quantum computing? Thanks to Gene Mele, Liang Fu, Jeffrey Teo, Zahid Hasan + group (expt)

The Insulating State

Characterized by energy gap: absence of low energy electronic excitations Covalent Insulator
e.g. intrinsic semiconductor

Atomic Insulator
e.g. solid Ar

The vacuum

electron

4s

Egap ~ 10 eV
3p

Dirac Vacuum Egap = 2 mec2 ~ 106 eV

positron ~ hole

Egap ~ 1 eV Silicon

The Integer Quantum Hall State

2D Cyclotron Motion, Landau Levels
E

Egap = hc

Energy gap, but NOT an insulator

Quantized Hall conductivity :

J y = xy Ex

Jy B Ex

e2 xy = n h
Integer accurate to 10-9

Haldane Model
IQHE in a crystal with zero net magnetic field
(Haldane PRL 1988)

Graphene in a periodic magnetic field B(r)

+ + + + + + + + + + + B(r) = 0 Zero gap, Dirac point B(r) 0 Energy gap

e xy = h
k

Band structure indistinguishable from an ordinary insulator

Egap

Topological Band Theory

The distinction between a conventional insulator and the quantum Hall state is a topological property of the manifold of occupied states

r | (k ) : Brillouin zone (a torus) a Hilbert space

1 n= d 2k k u (k ) k u (k ) = Integer 2 i BZ

Classified by the Chern (or TKNN) topological invariant (Thouless et al, 1982)

Insulator IQHE state

: n=0 : xy = n e2/h

The TKNN invariant can only change at a quantum phase transition where the energy gap goes to zero

Analogy: Genus of a surface : g = # holes g=0

g=1

Edge States
Gapless states must exist at the interface between topologically distinct phases y

IQHE state n=1

Vacuum n=0

n=1
x

n=0

Edge states ~ skipping orbits

Smooth Interpolation

2 Band Model : Dirac Eq.

E
M<0

0
M>0

H = iv( x x + y y ) + M ( x) z
M(x) =M0: Gap Eg = 2M0
E ( k x , k y ) = v 2 | k |2 + M 0 2

Eg

x
Gapless Chiral Fermions

E
Eg

M(x) ~ M0 sgn(x) : Domain Wall bound state

0 (k y , x) e

ik y y M 0 | x|/ v

1 i

E0 (k y ) = vk y
ky

Quantum Spin Hall Effect in Graphene

Kane and Mele PRL 2005

The intrinsic spin orbit interaction leads to a small (~10mK-1K) energy gap
Simplest model: |Haldane|2 (conserves Sz)

H H = 0

0 H Haldane = H 0

* H Haldane 0

J E

Bulk energy gap, but gapless edge states

Spin Filtered edge states Edge band structure

vacuum

QSH Insulator

/a

Edge states form a unique 1D electronic conductor

HALF an ordinary 1D electron gas Protected by Time Reversal Symmetry (conservation of Sz is NOT essential) Elastic Backscattering is forbidden. No 1D Anderson localization

Topological Insulator : A New B=0 Phase

2D Time reversal invariant band structures have a Z2 topological invariant, = 0,1

=0 : Conventional Insulator
E

=1 : Topological Insulator

Edge States
Kramers degenerate at

time reversal invariant momenta k* = k* + G

k*=0

k*=/a

k*=0

k*=/a

is a property of bulk bandstructure.

Easiest to compute if there is extra symmetry:

1. Spin rotation symmetry : Sz conserved Quantum spin Hall effect Independent spin Chern integers n = n 2.

= n, mod 2

Inversion (P) Symmetry : determined by Parity of occupied 2D Bloch states at time reversal invariant points.in bulk Brillouin zone

Quantum Spin Hall Effect in HgTe quantum wells

Bernevig, Hughes and Zhang, Science 06

HgxCd1-xTe HgxCd1-xTe

HgTe

d < 6.3 nm : Normal band order E

d > 6.3 nm : Inverted band order E

6 ~ s 8 ~ p

8 ~ p 6 ~ s

k
Egap~10 meV

Conventional Insulator

Quantum spin Hall Insulator with topological edge states

Observation of QSH insulator in HgTe quantum Wells

Konig, Wiedmann, Brune, Roth, Buhmann, Molenkamp, Qi, Zhang Science 2007

I
0

Landauer Conductance

d< 6.3 nm normal band order conventional insulator

e2 I =2 V h

d> 6.3nm inverted band order QSH insulator

Conductance 2e2/h independent of W for short samples (L<Lin)

Three Dimensional Topological Insulators

In 3D there are 4 Z2 invariants: (0 ; 123) characterizing the bulk. These determine how surface states connect.
Surface Brillouin Zone
Fu, Kane & Mele PRL 07 Moore & Balents PRB 07 Roy, cond-mat 06

4 1

3 2

OR
k=a k=b k=a k=b

2D Dirac Point

0 = 1 : Strong Topological Insulator

Fermi surface encloses odd number of Dirac points Topological Metal
Can only exist at a surface Robust to disorder (antilocalization)

EF

0 = 0 : Weak Topological Insulator

Fermi surface encloses even number of Dirac points Normal Metal
less robust.

Bi1-xSbx
Pure Bismuth
semimetal

Alloy : .09<x<.18
semiconductor Egap ~ 30 meV

Pure Antimony
semimetal

Ls La
E

EF EF

La Ls Egap

EF

La Ls T L

L
0
8 i =1

Inversion symmetry (1)

= 2 n (i )
n

Predict Bi1-xSbx is a strong topological insulator: (1 ; 111).

Experiments on Bi1-x Sbx

Map E(kx,ky) for (111) surface states below EF using Angle Resolved Photoemission Spectroscopy
D. Hsieh, D. Qian, L. Wray, Y. Xia, Y. S. Hor, R. J. Cava and M. Z. Hasan, Nature (08)

5 surface state bands cross EF between and M Proves that Bi1-x Sbx is a Strong Topological Insulator

Experiments on Bi2 Se3

Y. Xia, L. Wray, D. Qian, D. Hsieh, A. Pal, H. Lin, A. Bansil, D. Grauer, Y.S. Hor, R.J. Cava, M.Z. Hasan, arXiv:0812.2078

Bi2Se3 is a strong topological insulator with a simple surface Fermi surface.

EF

Similar to graphene, except only a

single Dirac point

Superconducting Proximity Effect

Fu, Kane PRL 08

s wave superconductor Topological insulator

Surface states acquire superconducting gap due to Cooper pair tunneling

BCS Superconductor :
ck c k ei

-k

(s-wave, singlet pairing)

Superconducting surface states

k -k k
Dirac point

cc

k k

surface e

(s-wave, singlet pairing)

Half an ordinary superconductor Highly nontrivial ground state

Majorana Fermion at a vortex

Ordinary Superconductor :
Andreev bound states in vortex core: E 0 E , -E , Bogoliubov Quasi Particle-Hole redundancy :

= h / 2e

= 2

=0

E ,

= E ,

Surface Superconductor :
Topological zero mode in core of h/2e vortex: E Majorana fermion : 0 0 Half a state 0 E=0 Two separated vortices define one zero energy fermion state (occupied or empty)

Topological Quantum Computation

Kitaev 2003

2 separated Majoranas = 2 degenerate states (1 qubit) 2N separated Majoranas = N qubits Quantum information immune to local decoherence Adiabatic braiding of vortices performs unitary operations on N qubits
Manipulate Majoranas by controlling phases on superconducting junctions
S - TI - S line junction Network of line junctions
phase 2/3 0 2/3

S TI S

A wire for Majoranas

Extra quasiparticle in each junction affects current

Conclusion
A new electronic phase of matter has been predicted and observed - 2D : Quantum spin Hall insulator in HgCdTe QWs - 3D : Strong topological insulator in Bi1-xSbx and Bi2Se3 Experimental Challenges - Spin dependent Transport Measurements - Transport and magneto-transport expts on Bi1-xSbx and Bi2Se3 - Superconducting proximity effect : - Characterize S-TI-S junctions - Create the Majorana bound states - Detect the Majorana bound states

Theoretical Challenges - Effects of disorder on surface states and critical phenomena - Effects of electron-electron interactions - Other Materials?

The challenges : Find suitable topological insulator (Bi1-x Sbx ? Eg ~ 30 meV) Find suitable superconductor which makes good interface ( Nb ? ) Optimize proximity induced gap and discrete Andreev bound states Control the superconducting phases with Josephson junctions Measure current difference when Majoranas are fused

Evidence for good contact between BiSb and Nb : minimal Shottky barrier Observed super current may be dominated by bulk electrons

A first step:
Control phases on a single tri-junction with currents I1, I2

2
I2

1
0
I1

Zero energy bound state predicted at tri-junction

Detect zero energy state by tunneling Predict a zero bias tunneling anomaly when a bound state is present.