Fractionalized Topological Insulators in 3D

Eduardo FradkinDepartment of Physics and

Institute for Condensed Matter TheoryUniversity of Illinois, Urbana, Illinois, USA

• AtMa Chan, Shinsei Ryu, Taylor Hughes and Eduardo Fradkin, Phys. Rev. B 87, 085132 (2013)• AtMa Chan, T. Kvorning, Shinsei Ryu and Eduardo Fradkin, Phys. Rev. B 93, 155122 (2016)• Peng Ye, Taylor Hughes, Joseph Maciejko, and Eduardo Fradkin, ArXiv: 1603.02696

Seminar at Instituto Balseiro, San Carlos de Bariloche, June 30, 2016

1

Motivation

• To provide an effective field theory of topological phases

• Hydrodynamic low energy theory for the charge degrees of freedom as in the FQHE

• Distinction between topologically protected response and topological phase: when is a topological phase?

• Candidates: in 2D Chern-Simons; in 3D: axion+BF theory

• How are topological insulators related to topological phases?

• 2D vs 3D; Z vs Z2

• Fractionalization in 3D

2

What is a topological phase?

• Fractional Quantum Hall fluids of 2DEG in high magnetic fields

• Spin liquid phases of frustrated quantum antiferromegnets

• Deconfined phases of discrete gauge theories

• Topological phases do not break any symmetries

• Natural representation in terms of gauge theories

• Gapped bulk excitations and gapless edge states (not always)

• Bulk excitations: vortices of the incompressible fluid and carry fractional charge (charged fluid) and fractional statistics

• On a closed 2D manifold with g handles: ground state degeneracy kg where k is an integer that depends on the topological phase

• Effective field theory: topological QFT, e.g Chern-Simons, BF theory

3

Hydrodynamic Theory of the FQH States (Fröhlich and Zee, Wen; ~ 1990)

• Example: Laughlin FQH states @ filling fraction ν=1/k

• Gapped states with a conserved charge current

4

• Gapped fluid: effective action is local and gauge invariant

• Broken time reversal invariance and parity (magnetic field)

• Effective action: Chern-Simons gauge theory, a topological QFT of deconfined excitations with fractional charge and statistics

@µjµ = 0 =) jµ =12⇡

✏µ⌫�@⌫A�

Aµ ! Aµ + @µ� =) jµ ! jµ

S

e↵

[jµ] =k

4⇡

Zd

3

x✏

µ⌫�Aµ@⌫A� �e

2⇡

Zd

3

xA

ext

µ ✏

µ⌫�@⌫A�

Discrete Gauge Theories

• Discrete gauge theories (e.g. ℤ2 gauge theory) have confining and deconfined (“free charge”) phases.

• The topological sectors of deconfined phases are described by the algebra of Wilson and ’t Hooft loops on non-contractible cycles on the torus

• d=2 spacial torus they have a 4-fold degenerate vacuum

• Fractionalized excitations: e (ℤ2 charge) and m (ℤ2 flux), and 𝜓 (fermion).

• Equivalent to Kitaev’s “Toric Code” states

• Described by a BF theory (“double” Chern-Simons)

• Q: Can we have fractionalized phase in a confining theory?

5

S. Shenker and E. Fradkin (1979)

Topological Insulators

• The electronic states of solids are Bloch states |un(k)⟩ which form energy bands labeled by a band index n and (quasi) momentum k, defined on a Brillouin zone (BZ)

• The BZ is a torus (2-torus in 2D, 3-torus in 3D)

• Insulator: integer number of bands are filled

• Topological insulator: the bands have a non-trivial topological structure

• In some cases the bands have an integer-valued topological invariant: the Chern number

• In other cases there is a ℤ2 topological invariant

• Classification of topological insulators and topological superconductors (A. Kitaev; S. Ryu, A. Schnyder, and A. Ludwig; ~2006)

6

7

• Prototype: IQH on 2D lattices in magnetic fields (Thouless, Kohmoto, Nightingale, den Nijs) (TKNN)

• Bloch states |un(k)⟩ are multivalued functions on the BZ (torus) and must be defined on patches

• Bloch states are charcterized by the Chern number Cn of the band (M. Berry; B. Simon)

Hall conductance is a topological invariant (TKNN, 1983)

�(n)xy

=e2

2⇡~C(n)

A(n)j (k) = i

Dun(k)

���rkj

���un(k)E

Z

BZd2k F (n) =

I

�dk · A(n)(k)

Cn =12⇡

I

�dk · A(k)

• Berry connection

• Berry curvature

• Chern number

Topology of electronic states

Dirac Fermions and Band Crossings

• Band crossings occur in a number of 3D materials such as Bi2Se3.

• The conduction and valence bands can become (nearly) degenerate at a finite number of points of the BZ, e.g. k=0

• Sometimes the degeneracy is protected by symmetry, e.g. time reversal

• The spin-orbit interaction is large in many materials

• Close to the degeneracy point the electronic states are represented as 4-spinors: the two band indices and the two angular momentum projections

• Locally in momentum space this is a Dirac fermion and the effective Hamiltonian is the Dirac Hamiltonian

8

3D Z2 Topological Insulator (Qi, Hughes, Zhang; Fu, Mele, Kane)

• A good example of a ℤ2 TI is Bi2Se3

• A simple lattice model (using the 4 x 4 Dirac matrices) (“Wilson fermions’’)

9

• 0< M < 2: negative energy states (near k=0) have negative parity, positive energy states have positive parity• Parity of the states at the time-reversal-invariant points (π,π,π), (π,π,0), and (π,0,0) are reversed relative to (0,0,0)• The product of the parities is a topological invariant since it does not change under smooth changes of the Hamiltonian

H =

Z

BZ

d3p

(2⇡)

3 †

(p)

⇣sin(p) · ↵+ M(p)�

⌘(p) (p)

M(p) = M + cos px

+ cos py

+ cos pz

� 3

Topologically Protected Surface Weyl Fermions

• The surface of a 3D Z2 TI behaves as the locus of a topological domain wall

• Parity Anomaly through a Callan-Harvey effect

• Early theory by Boyanovsky, Dagotto and EF (1986)

• The 3D ℤ2 TI has two-component massless Dirac (Weyl) fermions

• The fermions have opposite chirality (helicity) on the two opposite surfaces

• The helicity of the surface Weyl states has been measured in polarized angle-resolved photoemission experiments (ARPES) (Hasan, Shen...) and the Dirac spectrum with STM (Kapitulnik, Yazdani...)

10

ARPES Experiments in Bi2Se3 (Hsieh et al)

11

Strategy

• We will use functional bosonization to derive the effective field theory of topological insulators

• It is equivalent to bosonization in D=1+1 as an operator identity for free gapless fermions

• For D>1+1 we will get an effective hydrodynamic field theory for gapped fermions

• Extension to interacting fermions and fractionalization

12

Z[Aex

] =

RD

⇥¯ ,

⇤exp

�iSF [

¯ , , Aex

]

�

hjµ1(x1

)jµ2(x2

) · · · i =1i

�

�A

ex

µ1(x

1

)1i

�

�A

ex

µ2(x

2

)· · · lnZ[Aex]

Use gauge invariance of the fermion path integral: shift Aex to Aex + a, where a is a gauge transformation:

13

Z[Aex + a] = Z[Aex].fµ⌫ [a] = 0

Z[Aex] =ZD[a]

pure

Z[Aex + a]

Z[A

ex

] =

ZD[a, b]Z[a]⇥ exp

⇣� i

2

Zd

Dx bµ⌫···✏

µ⌫···↵�(f↵� [a]� f↵� [A

ex

])

⌘

hjµ1(x1)jµ2(x2) · · · i = h✏µ1⌫1�1···@⌫1b�1···(x1)✏µ2⌫2�2···

@⌫2b�2···(x2) · · · i

j

µ(x) ⌘ ✏

µ⌫�⇢···@⌫b�⇢···(x) , @µj

µ = 0

• This procedure (Schaposnik and EF, 1994) is meaningful only if the effective action is local (see also, Burgess and Quevedo, 1994)

• This works in 1+1 dimensions for gapless relativistic fermions

• For D>1+1 it works only if there is a finite energy gap

• This leads to a hydrodynamic description

• For systems with a Fermi surface one obtains the Landau theory of the Fermi liquid

14

Example: Polyacetylene

• Fermions in d=1 with a spontaneously broken translation symmetry: broken chiral symmetry (Class AIII)

15

✓ R

L

◆! ei�3✓

✓ R

L

◆⇢(x)! ⇢(x + a)) ✓ = kF a

Topological invariant ⌫ =✓(+1)� ✓(�1)

2⇡

Charge conjugation (particle-hole) ✓ = ⌫⇡ mod 2⇡

Z[A

ex

] =

ZD[a, b] exp

i

Zd

DxL!

L = �b✏µ⌫@µ(a⌫ �Aex

⌫ ) +✓

2⇡✏µ⌫@µa⌫ + · · ·

D=2+1 Chern Insulator

• Free fermions with broken time reversal invariance: integer quantum Hall states and the quantum anomalous Hall state

• The states are characterized by a topological invariant, the Chern number Ch

• In this case we obtain

16

L = �bµ✏µ⌫�@⌫(a� �Aex

� ) +Ch

4⇡✏µ⌫�aµ@⌫a�.

•The first term is the BF Lagrangian

•The hydrodynamic field bμ couples to flux tubes

•The statistical gauge field aμ couples to quasiparticle worldlines

Quantized Hall conductance �xy

= Che2

h

3D Topological Insulator

• Example: massive relativistic fermions with a conserved U(1) charge.

• Topological invariant: the winding number

17

L = �bµ⌫✏µ⌫�⇢@�(a⇢ �Aex

⇢ ) +✓

8⇡2

✏µ⌫�⇢@µa⌫@�a⇢ �1

4⇡2g2

@µa⌫@µa⌫ + · · ·

with ✓ = ⌫⇡ mod 2⇡

If time-reversal (particle-hole) is imposed, the topological class is Z2

The effective action for the external gauge field has an axion term (Qi, Hughes, Zhang, 2009)

The field bμν is an antisymmetric (Kalb-Ramond) tensor field

Physical Picture

The effective field theory we found (without the topological term) has the same form as the topological field theory of a d=2 gapped superconductor (Hansson, Oganesyan and Sondhi)

18

L = �2k

4⇡bµ✏µ⌫�@⌫(a� �Aex

� ) + · · ·

•Whereas in our case k=1, in the SC k=2.

• for k>1 this is a fractionalized state

•In the case of the SC fractionalization follows from pairing: the Bogoliubov quasiparticles have spin but no charge

•The k=2 BF theory also describes the topological limit of the deconfined phase of the Z2 gauge theory

Topological Insulator in 2+1 Dimensions

19

L = �2k

4⇡bµ✏µ⌫�@⌫(a� �Aex

� ) +Ch

4⇡✏µ⌫�aµ@⌫a�

=Kij

4⇡↵i

µ✏µ⌫�@⌫↵j� +

k

2⇡bµ✏µ⌫�@⌫Aex

�

K =✓

0 �k�k Ch

◆(↵1

µ, ↵2µ) = (bµ, aµ)

• In general the effective field theory has the K-matrix form (as in the fractional quantum Hall fluids) (Wen and Zee) with a degeneracy k2 on the torus

• A theory with k>1 is a fractionalized topological Chern insulator

•To derive this state once has to resort to flux attachment or parton constructions

The effective action for the external gauge field is

20

S

e↵

[Aex] =1

8⇡

2

Zd

4

x ✓✏

µ⌫⇢�@µA

ex

⌫ @⇢Aex

�

If the bulk mass gap is allowed to have a chiral twist, one finds a current

jµ = ✏µ⌫�⇢@⌫b�⇢ =1

4⇡2

✏µ⌫�⇢@⌫

�✓@�Aex

⇢

�

If the chiral angle has a jump from 0 to π at the surface one obtains a surface

Hall conductivity with 1/2 of the quantized value

�xy

=14⇡

=e2

2h

Electromagnetic Response

Fractionalized ℤ2 Topological Insulator• Fractionalized U(1) TI in 2D: FQH state at zero field

• It is found in models of 2D TIs with “flat’’ bands (Sun et al, Bernevig et al, Neupert et al)

• Not much is known on a 3D fractionalized ℤ2 TI

• One can obtain an effective field theory using a parton construction (assuming that the partons are deconfined)

• It requires having at least two species of a TI and finding a state in which the species are “paired’’

• In this case the difference of the two hydrodynamic fields becomes massive (Higgs mechanism)

• In the low-energy limit the different hydrodynamic fields become identified with each other

21

22

Projection: aIµ ⌘ aµ ) L =

✓

8⇡2k✏µ⌫�⇢@µAex

⌫ @�Aex

⇢

In the simplest case we get a theory at a higher level k

•This system has a fractional surface Hall conductance σxy=e2/2kh

•Fractional Callan-Harvey effect

•On opposite surfaces the Hall conductance has opposite signs

•surface gapped phases with broken time reversal invariance, and quantum phase transitions

Effective Field Theory with k>1

L =� ✏µ⌫�⇢bµ⌫

kX

I=1

@�aI⇢ � e ✏µ⌫�⇢@⌫b�⇢A

ex

µ

+✓

8⇡2

kX

I=1

✏µ⌫�⇢@µaI⌫@�aI

⇢ �1

4g2

kX

I=1

(f Iµ⌫)2

23

•Fractional charge and fractional statistics on the surface are endpoints of Wilson arcs of the bulk

•Fractionalized bulk extended excitations (Wilson surfaces coupled to the BF antisymmetric tensor fields) are Wilson loops of the surface states

Wilson arc

domain wall

bulk

domain wall

bulk

Wilson loop

Wilson surface

Gapped Surface Phases

Quantum Phase Transitions and Modular Symmetry

• Gapped topological phases with broken time reversal invariance described by Chern-Simons gauge theory

• The bulk theory has an extended duality (modular SL(2,ℤ)) symmetry of the coupling z=(θ/2π)+i4π/g2 (Witten 1995) which induces a modular symmetry on the boundary degrees of freedom (S. Kivelson and EF, 1996)

• Quantum critical points are fixed points under subgroups of SL(2,ℤ) with finite universal values of σxx and σxy

• Similarity of the quantum Hall plateau transitions but not driven by disorder

24

3D Fractionalized Phases with Confinement

• Most theories of fractionalized phases in 3D (and in 2D) use parton constructions (e.g. in the FQH fluids, in spin liquids of frustrated magnets, and even in HTSC).

• The fermion (the electron) is regarded as a composite object, e.g. 𝜓≡𝜓1𝜓2⋯𝜓2n+1, and one imagines the that spectrum is made of these fermionic partons.

• The Hilbert space is now much too big and needs to be projected down, and this leads to the confinement of the partons (as in hadrons)

• The projection takes place because the partons are coupled to gauge fields (SU(2n+1) in this example) which typically are in a confined phase.

• One way out is that the gauge theory may be in a deconfined phase, e.g. by condensing groups (pairs) of partons in some representation (higher than the fundamental) or that the gauge theory has a Chern-Simons term (e.g. in 2+1 dimensions).

• We will now see that in 3+1 dimensions there are other options.

25

3D Fractionalized Phases with Confinement

• We will focus on the compact abelian subgroup U(1)2n, the maximal torus of SU(2n+1)(`t Hooft, ~1981), with 𝜓i ↦ 𝜓i exp(i(𝜃i-𝜃i+1)), and the associated 2n compact abelian gauge fields are {a𝜇i}

• Monopoles of U(1)2n have proliferated and form a magnetic condensate that confines electrically charged excitations (“dual Meissner effect”).

• However this process may be incomplete and leave behind some discrete symmetry which may be deconfined and have non-trivial topology (Seiberg, Kapustin, ~2014) (inspired by SPT phases in CMP).

• We constructed phases of this type with condensates that involve both monopoles and charges (they are not dyonic).

• We find time-reversal invariant phases with fractional values of 𝜃

• This construction generalizes the concept of “flux attachment” to 3D

26

Conclusions and Outlook

• Topological Insulators offer a new arena in which to find new topological phases of matter

• Unlike the FQH fluids of the 2DEG in magnetic fields the fractionalized TI phases require non-trivial interactions

• They have been realized in relatively simple models but, so far, they have not been realized in experiment

• 2D fractionalized TI’s are extensions of the FQH fluids and have wave functions with interesting structure

• The 3D fractionalized TI have novel features

• What is the structure of their wave functions?

• Classification?

27