The quantum spin Hall effect and topological insulators Xiao-Liang Qi, and Shou-Cheng Zhang
Citation: Physics Today 63, 1, 33 (2010); doi: 10.1063/1.3293411 View online: https://doi.org/10.1063/1.3293411 View Table of Contents: http://physicstoday.scitation.org/toc/pto/63/1 Published by the American Institute of Physics
Articles you may be interested in Foundational theories in topological physics garner Nobel Prize Physics Today 69, 14 (2016); 10.1063/PT.3.3381
Two-dimensional van der Waals materials Physics Today 69, 38 (2016); 10.1063/PT.3.3297
Topological phases and quasiparticle braiding Physics Today 65, 38 (2012); 10.1063/PT.3.1641
Surprises from the spin Hall effect Physics Today 70, 38 (2017); 10.1063/PT.3.3625
Topological quantum computation Physics Today 59, 32 (2006); 10.1063/1.2337825
Fifty years of Anderson localization Physics Today 62, 24 (2009); 10.1063/1.3206091 The quantum spin Hall effect and feature topological insulators Xiao-Liang Qi and Shou-Cheng Zhang
In topological insulators, spin–orbit coupling and time-reversal symmetry combine to form a novel state of matter predicted to have exotic physical properties.
Xiao-Liang Qi is a research associate at the Stanford Institute for Materials and Energy Science and Shou-Cheng Zhang is a professor of physics at Stanford University in Stanford, California.
In the quantum world, atoms and their electrons can crystals.6–8 QSH systems are insulating in the bulk—they have form many different states of matter, such as crystalline solids, an energy gap separating the valence and conduction bands— magnets, and superconductors. Those different states can but on the boundary they have gapless edge or surface states be classified by the symmetries they spontaneously break— that are topologically protected and immune to impurities or translational, rotational, and gauge symmetries, respectively, geometric perturbations.9–12 Inside such a topological insula- for the examples above. Before 1980 all states of matter in tor, Maxwell’s laws of electromagnetism are dramatically al- condensed-matter systems could be classified by the principle tered by an additional topological term with a precisely quan- of broken symmetry. The quantum Hall (QH) state, discovered tized coefficient,12 which gives rise to remarkable physical in 1980,1 provided the first example of a quantum state that has effects. Whereas the QSH state shares many similarities with no spontaneously broken symmetry. Its behavior depends only the QH state, it differs in important ways. In particular, QH on its topology and not on its specific geometry; it was topo- states require an external magnetic field, which breaks time- logically distinct from all previously known states of matter. reversal (TR) symmetry; QSH states, in contrast, are TR invari- Recently, a new class of topological states has emerged, ant and do not require an applied field. called quantum spin Hall (QSH) states or topological insula- tors (see PHYSICS TODAY, January 2008, page 19). Topologically From quantum Hall to quantum spin Hall distinct from all other known states of matter, including QH In a one-dimensional world, there are two basic motions: for- states, QSH states have been theoretically predicted and ex- ward and backward. Random scattering can cause them to perimentally observed in mercury telluride quantum wells,2,3 mix, which leads to resistance. Just as we have learned from 4,5 in bismuth antimony alloys, and in Bi2Se3 and Bi2Te3 bulk basic traffic control, it would be much better if we could spa- tially separate the counterflow directions into two separate lanes, so that random Spinless 1D chain Spinful 1D chain collisions could be easily avoided. That simple traffic control mechanism turns out to be the essence of the QH effect.1 The QH effect occurs when a strong 2 = 1 + 1 4 = 2 + 2 magnetic field is applied to a 2D gas of electrons in a semiconductor. At low tem- perature and high magnetic field, elec- trons travel only along the edge of the semiconductor, and the two counterflows Quantum Hall Quantum spin Hall of electrons are spatially separated into different “lanes” located at the sample’s top and bottom edges. Compared with a 1D system with electrons propagating in both directions, the top edge of a QH bar Figure 1. Spatial separation is at the heart of both the quantum Hall (QH) and contains only half the degrees of freedom. the quantum spin Hall (QSH) effects. (a) A spinless one-dimensional system has That unique spatial separation is illus- both a forward and a backward mover. Those two basic degrees of freedom are trated in figure 1a by the symbolic equa- spatially separated in a QH bar, as illustrated by the symbolic equation tion “2 = 1 [forward mover] + 1 [backward “2 = 1 + 1.” The upper edge contains only a forward mover and the lower edge mover]” and is the key reason why the has only a backward mover. The states are robust: They will go around an impu- QH effect is topologically robust. When rity without scattering. (b) A spinful 1D system has four basic channels, which an edge-state electron encounters an im- are spatially separated in a QSH bar: The upper edge contains a forward mover purity, it simply takes a detour and still with up spin and a backward mover with down spin, and conversely for the keeps going in the same direction lower edge. That separation is illustrated by the symbolic equation “4 = 2 + 2.” (figure 1), as there is no way for it to turn
© 2009 American Institute of Physics, S-0031-9228-1001-020-3 January 2010 Physics Today 33 ab Figure 2. (a) On a lens with antireflection coating, light waves reflected by the top (blue line) and the bottom (red line) sur- faces interfere destructively, which leads to suppressed reflection. (b) A quantum spin Hall edge state can be scattered in two di- rections by a nonmagnetic impurity. Going clockwise along the blue curve, the spin ro- tates by π; counterclockwise along the red curve, by −π. A quantum mechanical phase factor of −1 associated with that difference of 2π leads to destructive interference of the two paths—the backscattering of elec- trons is suppressed in a way similar to that of photons off the antireflection coating.
back. Such a dissipationless transport mechanism could be magnetic moment, the TR symmetry is broken and the two extremely useful for semiconductor devices. Unfortunately, reflected waves no longer interfere destructively. In that the requirement of a large magnetic field severely limits the sense the robustness of the QSH edge state is protected by application potential of the QH effect. the TR symmetry. Can we get rid of the magnetic field and still separate the The physical picture above applies only to the case of sin- traffic lanes for the electrons? In a real 1D system, forward- gle pairs of QSH edge states. If there are two forward movers and backward-moving channels for both spin-up and spin- and two backward movers in the system—as, for example, down electrons give rise to four channels, as shown in the unseparated 1D system shown in figure 1b—then an elec- figure 1b. The traffic lanes for the electrons can be split in a tron can be scattered from a forward- to a backward-moving TR-invariant fashion, without any magnetic field, as illus- channel without reversing its spin and without the perfect trated in the figure by the symbolic equation “4 = 2 + 2.” We destructive interference, and thus there is dissipation. Con- can leave the spin-up forward mover and the spin-down sequently, for the QSH state to be robust, the edge states must backward mover on the top edge and move the other two consist of an odd number of forward movers and an odd channels to the bottom edge. A system with such edge states number of backward movers. That even–odd effect, charac- Z is said to be in a QSH state, because it has a net transport of terized by a so-called 2 topological quantum number, is at spin forward along the top edge and backward along the bot- the heart of the QSH state9,13 and is why a QSH insulator is tom edge, just like the separated transport of charge in the also synonymously referred to as a topological insulator. QH state. Charles Kane and Eugene Mele from the University of Pennsylvania,9 and Andrei Bernevig and one of us Two-dimensional topological insulators (Zhang)10 from Stanford University, independently proposed Looking at figure 1b, we see that the QSH effect requires the in 2005 and 2006 that such a separation, and thus the QSH counterpropagation of opposite spin states. Such a coupling state, can in principle be realized in certain theoretical models between the spin and the orbital motion is a relativistic effect with spin–orbit coupling. (The fractional QSH state was also most pronounced in heavy elements. Although all materials predicted,10 though it has yet to be experimentally observed.) have spin–orbit coupling, only a few of them turn out to be Although a QSH edge consists of both backward and for- topological insulators. In 2006 Bernevig, Taylor Hughes, and ward movers, backscattering by nonmagnetic impurities is Zhang proposed a general mechanism for finding topological forbidden. To understand that effect, we start with an anal- insulators2 and predicted in particular that mercury telluride ogy from daily life. Most eyeglasses and camera lenses have quantum wells—nanoscopic layers sandwiched between a so-called antireflection coating. As shown in figure 2a, re- other materials—are topological insulators beyond a critical d flected light from the top and the bottom surfaces interfere thickness c. The general mechanism is band inversion, in with each other destructively, leading to zero net reflection which the usual ordering of the conduction band and valence and thus perfect transmission. However, such an effect is not band is inverted by spin–orbit coupling.2,4 robust, as it depends on the matching between the optical In most common semiconductors, the conduction band wavelength and the thickness of the coating. is formed from electrons in s orbitals and the valence band is Just like the reflection of a photon by a surface, an elec- formed from electrons in p orbitals. In certain heavy elements tron can be reflected by an impurity, and different reflection such as Hg and Te, however, the spin–orbit coupling is so paths also interfere with each other. As shown in figure 2b, large that the p-orbital band is pushed above the s- orbital an electron in a QSH edge state can take either a clockwise band—that is, the bands are inverted. Mercury telluride or a counterclockwise turn around the impurity, and during quantum wells can be prepared by sandwiching the material that turn the spin rotates by an angle of π or −π to the oppo- between cadmium telluride, which has a similar lattice con- site direction. Consequently, the two paths, related by TR stant but much weaker spin–orbit coupling. Therefore, in- symmetry, differ by a full π −(−π)=2π rotation of the elec- creasing the thickness d of the HgTe layer increases the tron spin. A profound and yet deeply mysterious principle of strength of the spin–orbit coupling for the entire quantum 1 quantum mechanics states that the wavefunction of a spin- ⁄2 well. For a thin quantum well, as shown in the left column of particle obtains a negative sign upon a full 2π rotation. Thus figure 3a, the CdTe has the dominant effect and the bands the two backscattering paths always interfere destructively, have a normal ordering: The s-like conduction subband E1 is which leads to perfect transmission. If the impurity carries a located above the p-like valence subband H1. In a thick quan-
34 January 2010 Physics Today www.physicstoday.org tum well, as shown in the right column, the opposite ordering occurs due to in- a NORMAL INVERTED d creased thickness of the HgTe layer. d < 6.5 nm d > 6.5 nm d The critical thickness c for band inver- sion is predicted to be around 6.5 nm. The QSH state in HgTe can be de- CdTe CdTe CdTe HgTe CdTe scribed by a simple model for the E1 HgTe and H1 subbands2 (see the box on page E1 H1 36). Explicit solution of that model d d gives one pair of edge states for > c in H1 E1 the inverted regime and no edge states d d in the < c, as shown in figure 3b. The pair of edge states carry opposite spins and disperse all the way from valence b 0.05 0.05 band to conduction band. The crossing of the dispersion curves is required by TR symmetry and cannot be re- moved—it is one of the topological sig- 00 natures of a QSH insulator.
ENERGY (eV) Less than one year after the theo- ENERGY (eV) retical prediction, a team at the Univer- −0.05 −0.05 −0.02−0.01000.010.02 −0.02 −0.01 0.01 0.02 sity of Würzburg led by Laurens −1 −1 Molenkamp observed the QSH effect in WAVENUMBER (Å ) WAVENUMBER (Å ) HgTe quantum wells grown by molec- c 20 3
) ular-beam epitaxy. The edge states ) 7
10 Ω provide a direct way to experimentally Ω 16 h distinguish the QSH insulator from the 6 10 12 e2 trivial insulator. The two edge states of 2 5 the QSH insulator act as two conduct- 10 8 ing 1D channels, which each contribute 4 e2 h 10 4 one quantum of conductance, / . That RESISTANCE (
RESISTANCE (k perfect transmission is possible be- 103 0 cause of the principle of antireflection −1.0−0.50.00.51.01.52.0 −1.0 −0.5 0.0 0.5 1.0 1.5 2.0 explained earlier. In contrast, a trivial GATE VOLTAGE (V) GATE VOLTAGE (V) insulator phase is “really” insulating, with vanishing conductance. Such a Figure 3. Mercury telluride quantum wells are two-dimensional topological sharp conductance difference between insulators. (a) The behavior of a mercury telluride–cadmium telluride quantum thin and thick quantum wells was ob- well depends on the thickness d of the HgTe layer. Here the blue curve shows the served experimentally, as shown in potential-energy well experienced by electrons in the conduction band; the red figure 3c. curve is the barrier for holes in the valence band. Electrons and holes are trapped From two to three dimensions laterally by those potentials but are free in the other two dimensions. For quan- tum wells thinner than a critical thickness d ≃ 6.5 nm, the energy of the lowest- From figure 3b we see that the 2D topo- c logical insulator has a pair of 1D edge energy conduction subband, labeled E1, is higher than that of the highest- energy valence band, labeled H1. But for d > d , those electron and hole bands states crossing at momentum k =0. c Near the crossing point, the dispersion are inverted. (b) The energy spectra of the quantum wells. The thin quantum well of the states is linear. That’s exactly the has an insulating energy gap, but inside the gap in the thick quantum well are dispersion one gets in quantum field edge states, shown by red and blue lines. (c) Experimentally measured resistance theory from the Dirac equation for a of thin and thick quantum wells, plotted against the voltage applied to a gate massless relativistic fermion in 1D, and electrode to change the chemical potential. The thin quantum well has a nearly thus that equation can be used to de- infinite resistance within the gap, whereas the thick quantum well has a quan- scribe the QSH edge state. Such a pic- tized resistance plateau at R = h/2e2, due to the perfectly conducting edge states. ture can be simply generalized to a 3D Moreover, the resistance plateau is the same for samples with different widths, topological insulator, for which the sur- from 0.5 μm (red) to 1.0 μm (blue), proof that only the edges are conducting. face state consists of a single 2D mass- less Dirac fermion and the dispersion forms a so-called Dirac cone, as illus- trated in figure 4. Similar to the 2D case, the crossing point— Fang’s group at the Chinese Academy of Sciences, the two of the tip of the cone—is located at a TR-invariant point, such us predicted that Bi2Te3 , Bi2Se3 , and Sb2Te3 , all with the lay- as at k = 0, and the degeneracy is protected by TR symmetry. ered structure in figure 4a, are 3D topological insulators, 4 6 Liang Fu and Kane predicted that the alloy Bi1−xSbx whereas a related material, Sb2Se3 , is not. x would be a 3D topological insulator in a special range of , As in HgTe, the nontrivial topology of the Bi2Te3 family and with angle-resolved photoemission spectroscopy is due to band inversion between two orbitals with opposite (ARPES) Zahid Hasan and coworkers at Princeton University parity, driven by the strong spin–orbit coupling of Bi and Te. observed the topological surface states in that system.5 How- Due to such similarity, that family of 3D topological insula- ever, the surface states and the underlying mechanism turn tors can be described by a 3D version of the HgTe model (see out to be extremely complex. In collaboration with Zhong the box). First-principle calculations show that the materials www.physicstoday.org January 2010 Physics Today 35 have a single Dirac cone on the surface. The spin of the sur- momentum. By mapping all of momentum space, the ARPES face state lies in the surface plane and is always perpendicu- experiments show convincingly that the surface states of lar to the momentum, as shown in figure 4b. Bi2Te3 and Bi2Se3 consist of a single Dirac cone. Such a state is
Known to be excellent thermoelectric materials, Bi2Te3 impossible to construct in a purely 2D system. For example, a 2D graphene sheet has four Dirac cones (see the article by and Bi2Se3 have been investigated independently, particularly at Princeton, where Hasan’s group observed in ARPES exper- Andrey Geim and Allan MacDonald, PHYSICS TODAY, August iments the single Dirac-cone surface state of Bi Se samples 2007, page 35). The 2D HgTe quantum well at the crossover 2 3 d d prepared by Robert Cava and coworkers.7 Furthermore, the point = c has two Dirac cones. In a sense, the cones are spa- group’s spin-resolved measurements showed that the elec- tially separated, with one placed on the top surface and the tron spin indeed lies in the plane of the surface and is always other on the bottom surface, similar to spatial decomposition perpendicular to the momentum, in agreement with theory. illustrated in figure 1 for the 1D surfaces of a 2D system. (Par- ticle physicists have been using similar ideas—treating a 3D However, the experiments also observed bulk carriers co- lattice as the surface of a 4D lattice—in numerical simulation existing with the topological surface states. A pure topological- of fermions so as to avoid getting an unwanted doubling of insulator phase without bulk carriers was first observed in neutrinos.) As we discuss below, a single Dirac cone on the Bi Te by Yulin Chen and Zhi-xun Shen’s group at Stanford 2 3 surface directly leads to novel topological properties.11,12 in ARPES experiments on material prepared by Ian Fisher and colleagues.8 As shown in figure 4c, the observed surface Topological classification of insulators states indeed disperse linearly, crossing at the point with zero Mathematicians group geometric objects into broad topolog- ical classes. Objects with different shapes, such as a donut Models of topological insulators and a coffee cup, can be smoothly deformed into each other and can therefore be grouped into the same topological class. The essence of the quantum spin Hall effect in real materials can Mathematicians also developed the concept of a topological be captured in explicit models that are particularly simple to invariant that uniquely defines the topological class. Topo- solve. The two-dimensional topological insulator mercury tel- logical materials in general, and topological insulators in par- luride can be described by an effective Hamiltonian that is essen- ticular, can be defined by physically measurable topological tially a Taylor expansion in the wave vector k of the interactions invariants in topological field theories. between the lowest conduction band and the highest valence We can first divide insulators into two broad classes, ac- band:2 cording to the presence or absence of TR symmetry. The QH state is a topological insulator state that breaks TR symmetry. David Thouless and coworkers showed that the physically Mk() Ak(+xy ik ) 0 0 Ak ik −M k measured integer QH conductance is given by a topological H kk1(−xy ) () 0 0 invariant called the first Chern number (see the article by () =()+ Mk Ak ik , 0 0 () −(xy − ) Joseph Avron, Daniel Osadchy, and Ruedi Seiler, PHYSICS Ak ik Mk 0 0 − (+xy ) − () TODAY, August 2003, page 38). For a generally interacting sys- tem, the topological properties of the QH state can be de- k C Dk2 Mk M−Bk2 ( ) = + , ( ) = , scribed by an effective topological field theory based on the Chern–Simons theory.14 Although Duncan Haldane con- where the upper 2 × 2 block describes spin-up electrons in the s- structed a model of the QH effect without the external mag- like E1 conduction and the p-like H1 valence bands, and the netic field, that state still breaks TR symmetry. lower block describes the spin-down electrons in those bands. For a long time it was widely believed that both TR sym- The term ϵ(k)1 is an unimportant bending of all the bands (1 is M metry breaking and two-dimensionality are necessary for an the identity matrix). The energy gap between the bands is 2 , insulator to be topological, but in 2001 the first model of a B A and , typically negative, describes the curvature of the bands; TR-invariant topological insulator was introduced.15 That M B incorporates interband coupling to lowest order. For / < 0, the model was originally defined in 4D, but TR-invariant topo- eigenstates of the model describe a trivial insulator. But for thick logical insulators in 3D and 2D can be obtained through a M quantum wells, the bands are inverted, becomes negative, simple dimension-reduction procedure.12 Shuichi Murakami, and the solution yields the edge states of a quantum spin Hall Naoto Nagaosa, and Zhang, and in parallel MacDonald and insulator. Another model9 for the 2D topological insulator with a colleagues at the University of Texas in Austin, developed the honeycomb lattice can also be simply solved to gain explicit theory of the intrinsic spin Hall effect in doped semiconduc- understanding. tors and identified spin–orbit coupling as the crucial ingre- The 3D topological insulator in the Bi2Te3 family can be dient; later, Murakami, Nagaosa, and Zhang extended the described by a similar model:6 theory to TR-invariant insulators. Kane and Mele first intro- duced the topological band theory of TR-invariant QSH in- M k Ak ik Ak () 2(+xy ) 0 1 z sulators in 2D and showed that they fall into two distinct Ak ik −M k Ak Z 2(−xy ) () 1 z 0 topological classes, generally referred to as the 2 classifica- H ()kk1 =()+ , 9 Akz M k Ak ik tion. That beautiful topological band theory was soon gen- 0 1 () −(−2 xy ) eralized to three dimensions.11 The two of us and our col- Akz 0 −Ak(+xy ik ) −M()k 1 2 leagues have developed a unifying topological field theory CDk,MM−Bkk2 Dk2 k2 −Bk2 that defines the general concept of a topological insulator in ( ) = +1 z +21⊥ ( ) = z 2 ⊥ , terms of a physically measurable topological field theory.12
in the basis of the Bi and Te bonding and antibonding pz orbitals We now have two precise definitions of TR-invariant B B topological insulators, one in terms of noninteracting topo- with both spins. The curvature parameters 1 and 2 have the M B logical band theory11 and one in terms of topological field the- same sign. As in the 2D model, the solution for / 1 <0 M B ory.12 If we approximate an insulator with noninteracting describes a trivial insulator, but for / 1 > 0, the bands are in- verted and the system is a topological insulator. electrons filling a certain number of bands, the topological band theory can evaluate an explicit topological invariant
36 January 2010 Physics Today www.physicstoday.org a b E ky
Figure 4. In three-dimensional topologi- k cal insulators, the linearly dispersing edge K x Γ states of figure 3b become surface states K described by a so-called Dirac cone. (a) The crystal structure of the 3D topolog-
ical insulator Bi2Te3 consists of stacked quasi-2D layers of Te-Bi-Te-Bi-Te. The ar- rows indicate the lattice basis vectors. The surface state is predicted to consist of a single Dirac cone.6 (b) Angle-resolved photoemission spectroscopy maps the c energy states in momentum space. Spin- dependent ARPES of the related com-
pound Bi2Se3 reveals that the spins (red) of 0 the surface states lie in the surface plane and are perpendicular to the momentum.7 SSB (c) This ARPES plot of energy versus
wavenumber in Bi2Te3 shows the linearly dispersing surface-state band (SSB) above 0.2 Bi BVB the bulk valence band (BVB). The dashed white line indicates the Fermi level. The Te1 BINDING ENERGY (eV) blue lines meet at the tip of the Dirac Te2 K Γ K cone.8 0.4 −0.20 0.2 WAVENUMBERk (Å−1 )
Z α e2 ħc θ that can give only binary values of 0 or 1: a 2 classification where = / ≈ 1/137 is the fine-structure constant, is a pa- that defines trivial and nontrivial insulators. For materials rameter, and ϵμνρτ is the fully asymmetric 4D Levi-Civita ten- with inversion symmetry, a powerful algorithm developed sor. Unlike the Maxwell action, Sθ is a topological term—it by Fu and Kane4 can be easily integrated into electronic struc- depends only on the topology of the underlying space, not ture calculations to numerically evaluate the topological on the geometry. Written using the field tensor, the term is band invariant. However, since all insulators in nature are independent of the metric. necessarily interacting, it is important to have a general def- Since the E field is invariant under TR, whereas the B inition of topological insulators that is valid for interacting field changes sign, Sθ naively breaks TR symmetry. For a pe- systems and is experimentally measurable. Both problems riodic system, however, there are two values of θ, namely 12 were solved with the topological field theory, which can be θ = 0 or θ = π, that preserve the TR symmetry.12 One can easily generally defined for all insulators, with or without inter - understand that conclusion by an analogy. If we have a 1D actions. In the noninteracting case, both definitions agree. ring with a magnetic flux inside, a general value of the flux Surprisingly, the topological field theory can be explained Φ would break the TR symmetry. However, for two special in terms of elementary concepts in undergraduate-level values of the flux, Φ = 0 or Φ = hc/2e, an electron’s wavefunc- electromagnetism. tion changes its phase by 0 or π when the electron circles the Inside an insulator, the electric field E and the magnetic B ring either clockwise or counterclockwise, and TR symmetry field are both well defined. In a Lagrangian-based field the- is maintained. ory, the insulator’s electromagnetic response can be described S π d3xdt ϵE2 μ B2 ϵ If we integrate out all the microscopic fermionic degrees by the effective action 0 = 1/8 ∫ ( −1/ ), with the of freedom to obtain the effective action Sθ, all nonmagnetic electric permittivity and μ the magnetic permeability, from insulators in the universe would fall into two distinct topo- which Maxwell’s equations can be derived. The integrand de- logical classes, described by effective topological field theories pends on geometry, though, so it is not topological. To see that F with θ = 0 or with θ = π. Unlike ϵ and μ, the physically meas- dependence, one can write the action in terms of μν, the 4D θ S π d3xdtF Fμν urable parameter is universally quantized, with the two pos- electromagnetic field tensor: 0 = 1/16 ∫ μν . The im- μ ν sible values defining the topologically trivial and nontrivial plied summation over the repeated indices and depends Z on the metric tensor—that is, on geometry. (Indeed, it is that insulators, respectively—the 2 classification again. dependence that leads to the gravitational lensing of light.) Such classification is valid for a periodic system. For a There is, however, another possible term in the action of real solid with a finite boundary, a topological insulator is in- the electromagnetic field: sulating only in the bulk; it has an odd number of gapless Dirac cones on the surface that describe conducting surface θα θα 3 3 μυ ρτ Sθ = ∫d xdt E · B ≡ ∫d xdt μυρτ F F states. If we uniformly cover the surface with a thin ferro- π2 π2 4 32 magnetic film, an insulating gap also opens up on the bound- ary; the TR symmetry is preserved in the bulk but broken θα 3 μ υρτ = ∫ d xdt ∂ ( AAμυρτ ∂ ), on the surface. The last identity of the equation above for 2π 4π Sθ shows that the bulk topological term is in fact a total www.physicstoday.org January 2010 Physics Today 37 derivative, expressible as a surface term, given by the expression in ab parenthesis. That surface term is the same Chern–Simons term that de- scribes the topological field theory of
the QH state. In the QH field theory, TIME the term’s coefficient specifies the value of the Hall conductance.14 Here the coefficient of θ = π translates into 1 2 SC a Hall conductance of ⁄2 e /h, half the TI conductance of the first QH plateau. TI That value is uniquely associated with the single Dirac cone on the sur- face of topological insulators. Any Figure 5. Novel behavior is predicted for topological insulators. (a) When a topo- random disorder can change a sys- logical insulator (TI, green) is coated by a thin ferromagnetic layer (gray), each tem’s Hall conductance only by an electron (red sphere) in the vicinity of the surface induces an image monopole e2 h integer multiple of / , thus the (blue sphere) right beneath it.12 When one electron winds around another (red 1 e2 h half-QH conductance of ⁄2 / can circle), it will experience the magnetic flux (arrows in the blue dome) carried by never be reduced to zero by disor- the image monopole of the other, so that the electron– monopole composite, der—the surface states are topologi- called a dyon, obeys fractional statistics. (b) When a TI is coated by an s-wave cally robust. superconductor (SC), the superconducting vortices are Majorana fermions—they Outlook are their own antiparticles. Exchanging or braiding Majorana vortices, as sketched here, leads to non-abelian statistics.17 Such behavior could form the basis for topo- The field of topological insulators is logical quantum computing. growing rapidly, and many remark- able experiments have been carried out. In nonlocal transport measure- ments in a series of HgTe devices, the Würzburg group con- memory by purely electric means, and the Majorana fermions firmed that transport current is carried by the QSH edge could be used for topological quantum computing.18 states. The topological insulators Bi2Te3 and Bi2Se3 fabricated Albert Einstein insisted that all fundamental laws of in nanoribbon form at Stanford and by molecular-beam epi- physics should be expressed in terms of geometry, and he ex- taxy at Tsinghua University. Scanning tunneling microscopy emplified that ancient Greek ideal by formulating the theory experiments have been carried out at Princeton, Stanford, of gravity in terms of the geometrical curvature of space and and Tsinghua universities to probe the topological surface time. Physicists are now pursuing Einstein’s dream one step states. Preliminary transport measurements indicate domi- further, exploring the fundamental laws expressed in terms nant contributions from surface states. of topological field theory. The standard model of elementary Solving Maxwell’s equations with the topological term particles contains a topological term that is identical to the Sθ included leads to predictions of novel physical properties term that defines topological insulators. Even if only a small characterized by exotic excitations. The 2D QSH insulator is number of the predicted exotic particles are observed in topo- predicted to have fractional charge at the edge and spin– logical insulators, our fundamental understanding of nature charge separation in the bulk. In introductory physics classes would be greatly enhanced. Such tabletop experiments could we learned that a point charge above a metal or an insulator become a window into the standard model16 and help reveal can be viewed as inducing an image charge below the sur- the alluring beauty and mysteries of our universe. face. A point charge above the surface of a 3D topological in- sulator is predicted to induce not only an image electric References charge but also an image magnetic monopole below the sur- 1. K. v. Klitzing, G. Dorda, M. Pepper, Phys. Rev. Lett. 45, 494 (1980). Science 314 face,12 as shown in figure 5a. Such a composite object of elec- 2. B. A. Bernevig, T. L. Hughes, S.-C. Zhang, , 1757 tric and magnetic charges, called a dyon, would obey neither (2006). 3. M. König et al., Science 318, 766 (2007). Bose nor Fermi statistics but would behave like a so-called 4. L. Fu, C. L. Kane, Phys. Rev. B 76, 045302 (2007). anyon with any possible statistics. Dislocations inside a 3D 5. D. Hsieh et al., Nature 452, 970 (2008). topological insulator contain electronic states that behave 6. H. Zhang et al., Nat. Phys. 5, 438 (2009). similarly to QSH edge states. 7. Y. Xia et al., Nat. Phys. 5, 398 (2009); D. Hsieh et al., Science 323, Axions are weakly interacting particles postulated to 919 (2009). solve some puzzles in the standard model of particle 8. Y. L. Chen et al., Science 325, 178 (2009). Phys. Rev. Lett. 95 physics16 (see the article by Karl van Bibber and Leslie Rosen- 9. C. L. Kane, E. J. Mele, , 226801 (2005); C. L. Kane, E. J. Mele, Phys. Rev. Lett. 95, 146802 (2005). berg, PHYSICS TODAY, August 2006, page 30). Those elusive 10. B. A. Bernevig, S.-C. Zhang, Phys. Rev. Lett. 96, 106802 (2006). particles are also predicted to exist inside topological mag- Phys. Rev. B 75 θ 11. J. E. Moore, L. Balents, , 121306 (2007); L. Fu, C. L. netic insulators, systems for which the parameter above be- Kane, E. J. Mele, Phys. Rev. Lett. 98, 106803 (2007); R. Roy, Phys. comes dependent on position and time. Majorana fermions Rev. B 79, 195322 (2009). are distinct from the familiar Dirac fermions: They are their 12. X.-L. Qi, T. L. Hughes, S.-C. Zhang, Phys. Rev. B 78, 195424 own antiparticles. There is still no conclusive evidence for (2008); X.-L. Qi et al., Science 323, 1184 (2009). Phys. Rev. Lett. 96 Majorana fermions in nature. But when a superconductor is 13. C. Wu, B. A. Bernevig, S.-C. Zhang, , 106401 (2006); C. Xu, J. E. Moore, Phys. Rev. B 73, 045322 (2006). close to the surface of a topological insulator, Majorana fermi- Int. J. Mod. Phys. B 6 17 14. S.-C. Zhang, , 25 (1992). ons are predicted to occur inside vortices (see figure 5b). 15. S.-C. Zhang, J. Hu, Science 294, 823 (2001). Besides teaching us about the quantum world, the exotic 16. F. Wilczek, Nature 458, 129 (2009); Nat. Phys. 5, 614 (2009). particles in topological insulators could find novel uses. For 17. L. Fu, C. L. Kane, Phys. Rev. Lett. 100, 096407 (2008). example, image monopoles could be used to write magnetic 18. C. Nayak et al., Rev. Mod. Phys. 80, 1083 (2008). ■
38 January 2010 Physics Today www.physicstoday.org