ebook img

Quantum Spin Hall Effect in Inverted Type II Semiconductors PDF

0.28 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Quantum Spin Hall Effect in Inverted Type II Semiconductors

Quantum Spin Hall Effect in Inverted Type II Semiconductors Chaoxing Liu1,2, Taylor L. Hughes2, Xiao-Liang Qi2, Kang Wang3 and Shou-Cheng Zhang2 1 Center for Advanced Study, Tsinghua University,Beijing, 100084, China 2 Department of Physics, McCullough Building, Stanford University, Stanford, CA 94305-4045 and 3 Department of Electrical Engineering, UCLA, Los Angeles, CA 90095-1594 Thequantumspin Hall(QSH)stateisatopologically non-trivialstateofquantummatterwhich preservestime-reversalsymmetry;ithasanenergygapinthebulk,buttopologically robustgapless states at the edge. Recently, this novel effect has been predicted and observed in HgTe quantum 8 0 wells[1, 2]. In this work we predict a similar effect arising in Type-IIsemiconductor quantumwells 0 made from InAs/GaSb/AlSb. Because of a rare band alignment the quantum well band structure 2 exhibits an “inverted” phase similar to CdTe/HgTe quantum wells, which is a QSH state when the Fermi level lies inside the gap. Due to the asymmetric structure of this quantum well, the n effects of inversion symmetry breaking and inter-layer charge transfer are essential. By standard a self-consistentcalculations,weshowthattheQSHstatepersistswhenthesecorrectionsareincluded, J andaquantumphasetransitionbetweenthenormalinsulatorandtheQSHphasecanbeelectrically 8 tunedby thegate voltage. 1 PACSnumbers: ] l l a Recently, a striking prediction of a quantum spin Hall cal edge states. h - (QSH) insulator phase in HgTe/CdTe quantum wells[1] The quantum well structures in which we are inter- s wasconfirmedintransportexperiments[2]. The QSHin- e ested are asymmetric with AlSb/InAs/GaSb/AlSb lay- m sulatorphaseisatopologicallynon-trivialstateofmatter ers grown as shown in Fig. 1. This is an unusual quan- reminiscentoftheintegerquantumHalleffect,butwhere . tum well system due to the alignment of the conduc- t time-reversalsymmetryispreservedinsteadofbeingbro- a tion and valence band edges of InAs and GaSb. The kenbythelargemagneticfield. Thestateischaracterized m valence band edge of GaSb is 0.15 eV higher than the by a bulk charge-excitation gap and topologically pro- conduction band edge of the InAs layer. The AlSb lay- - d tected helical edge states, where states of opposite spin ers serve as confining outer barriers. The “conduction” n counter-propagate on each edge[3, 4, 5].Unfortunately, subbands are localized in the InAs layer while the “va- o high-qualityHgTe/CdTequantumwellsareveryspecial, lence” subbands are localized in the GaSb layer as il- c andonlyafewacademicresearchgroupshavetheprecise lustrated in Fig. 1 (a). In this work we will focus on [ materialcontrolneededto carryoutsuchdelicateexper- the regime where the lowest electron and hole subbands 1 iments. We are therefore lead to search for other, more E1,H1,whicharederivedfromthes-likeconductionand v conventional, materials that exhibit the QSH effect. p-like heavy-hole bands respectively, are nearly degener- 1 ate, and all other subbands are well-separatedin energy. 3 In this work we introduce a new material with the 8 QSH phase, the InAs/GaSb/AlSb Type-II semiconduc- Whenthequantumwellthicknessisincreasedtheenergy 2 torquantumwellintheinvertedregime[6,7,8,9,10,11]. of the E1 (H1) band edge is decreasing (increasing). At . some critical thickness a level crossing occurs between 1 We will show that this quantum well exhibits a subband 0 inversiontransition as a function of layerthickness, sim- E1 and H1, after which the band edge of E1 sinks below 8 ilar to the HgTe/CdTe system, and can be character- that of H1, putting the system into the inverted regime 0 ized by an effective four-band model near the transition. of Type-II quantum wells. Since the H1 band disperses v: This model is similar to the model for HgTe/CdTe[1], downwards and the E1 band disperses upwards, the in- i but contains terms describing the strong bulk inversion versionofthebandsequenceleadstoacrossingofthetwo X bands,seeFig. 1(b). Historically,theinvertedregimeof asymmetry (BIA) and structural inversion asymmetry r InAs/GaSb/AlSbquantumwellswasdescribedasasemi- a (SIA). In fact, due to the unique band alignment of metal without a gap[6]. However, Ref. [7] first pointed InAs/GaSb/AlSb,theelectronsubbandandtheholesub- bandarelocalizedindifferent quantumwelllayers. Addi- out that due to the mixing between E1 and H1, a small gap(E inFig1(b))isgenerallyopened,leadingtobulk tionally, the band alignment forces one to consider self- g insulating behavior. This hybridization gap was later consistent corrections[12, 13, 14] which we will discuss demonstrated in experiments[8, 9]. Therefore, just like below. Our results show that the asymmetric quantum in the HgTe/CdTe quantum wells, the inverted regime well,withstrongbuilt-inelectricfield,canbeelectrically of InAs/GaSb quantum wells should be a topologically tuned through the phase transition using front and back non-trivial QSH phase protected by the bulk gap. gates. Whilethisisofsignificantfundamentalinterest,it alsoallowsonetoconstructaquantumspinHallfieldef- Thisseeminglysimpleconclusioniscomplicatedbythe fect transistor (FET) that exhibits an insulating “OFF” unique features of type II quantum wells: the electron- statewithnoleakagecurrent,andanearlydissipationless subbandandhole-subbandareseparatedintwodifferent “ON” state with non-zero conductance via the topologi- layers. There are several separate consequences of this 2 (a) parts H=H0+HBIA+HSIA. (1) Front Barrier GaSb Barrier Back gate AlSb 0.7eV AlSb gate V 1.6eV E 1.6eV V In the basis {|E1+i,|H1+i,|E1−i,|H1−i}, and keeping f 1 H b terms only up to quadratic powers of k, we have InAs 1 0.36eV M(k) Ak+ 0 0 (b) H0 =ǫ(k)I4×4+ A0k− −M0(k) M0(k) −A0k− (2)  0 0 −Ak+ −M(k) E H1 Eg wMh2ekr2eaIn4×d4ǫ(isk)th=eC40×+4Cid2ken2.titTyhmisaitsrisxim, Mply(kt)he=HMam0i+l- tonian used by BHZ. The zinc-blende structure has two E different atoms in each unit cell, which breaks the bulk 1 inversion symmetry and leads to additional terms in the k bulk Hamiltonian[16]. When projected onto the lowest subbands the BIA terms are FIG.1: (a)Bandgapandbandoffsetdiagramforasymmetric 0 0 ∆ek+ −∆0 AislcSobn/nIencAtesd/GtoaSabfrqounatngtautmewwheillles.thTehreiglhetftbaArlrSiebrbisacrroinenrelcatyeedr H = 0 0 ∆0 ∆hk− . (3) toabackgate. TheE1 subbandislocalized intheInAslayer BIA  ∆ek− ∆0 0 0  and H1 is localized in the GaSb layer. Outer AlSb barriers  −∆0 ∆hk+ 0 0  provide an overall confining potential for electron and hole Finally the SIA term reads states. (b) Schematic band structure diagram. The dashed lineshowsthecrossingoftheE1andH1statesintheinverted regime,andduetothehybridizationbetweenE1 andH1,the 0 0 iξek− 0 gap Eg appears. HSIA =−iξ0e∗k+ 00 00 00. (4)  0 0 0 0 Here we recognize the SIA term as the electron k-linear fact. First, the hybridization between E1 and H1 is re- Rashba term; the heavy-hole k-cubic Rashba term is ne- duced, but this is just a quantitative correction. Second, sincethereisnoinversionsymmetryinthequantumwell glected. The parameters ∆h,∆e,∆0,ξe depend on the quantum well geometry. growthdirection,SIAtermsmaybelargeenoughtocom- Now we address, from pure band structure considera- pete with the reduced hybridization. In addition, BIA tions, whether or not a QSH phase exists in this model. may also play a role for this system. Therefore, both WithoutH andH the Hamiltonianis blockdiag- SIA and BIA must be included properly to make a cor- BIA SIA onal and each block is exactly a massive Dirac Hamilto- rectprediction,whileinHgTe/CdTequantumwellsthese nianin(2+1)d. Byitself,eachblockbreakstime-reversal two types of terms were ignored because BIA terms are symmetry,butthetwo2×2blocksaretime-reversalpart- small when compared with the hybridization, and the ners so that the combined system remains time-reversal quantum well was symmetric which minimizes SIA. Fi- invariant. As mentioned, this is the pure BHZ model nally, since the electron and hole subbands lie in two and from their argument we know that there is a topo- differentlayers,thereisanautomaticchargetransferbe- logicalphase transition signalled by the gap closing con- tweenthelayerswhichyieldsacoexistenceofp-typeand n-typecarriers. Consequently,aself-consistenttreatment dition M0 = 0, and the system is in QSH phase when ofColoumbenergyisnecessarytoaccountforthiseffect. M0/M2 <0. When HBIA and HSIA terms are included, In the following, we will discuss all of these issues and the two blocks of H0 are coupled together and the anal- ysisinBHZmodeldoesnotdirectlyapply. However,the concludethattheQSHphaseexistsinanexperimentally QSH phase is a topological phase of matter protected viable parameter range. by the band gap[3, 4, 5]. In other words, if we start The materials in these quantum wells have the zinc- from the Hamiltonian H0 in the QSH phase and turn on blende lattice structure and direct gaps near the Γ H and H adiabatically, the system will remain BIA SIA point and are thus well-described by the 8-band Kane in the QSH phase as long as the energy gap between model[15]. We will construct an effective 4-band model E1 and H1 remains finite. With realistic parameters for using the same envelope function approximation proce- an InAs/GaSb/AlSb quantum well obtained from the 8- dure as the Bernevig-Hughes-Zhang (BHZ) model[1]; al- band Kane model, the adiabatic connection between the beit a more complex one due to the SIA and BIA terms. inversion-symmetricHamiltonianH0 andthe fullHamil- The Hamiltonian naturally separates into three distinct tonian H was verified for the proper parameter regime, 3 FIG. 2: The energy dispersions of Hamiltonian (1) on a cylindrical geometry with open boundary conditions along the y-direction and periodic boundary conditions along the x-direction. (a) The dispersion for the quantum well with FIG.3: (a)-(c)Theenergydispersionscalculated from the8- GaSb layer thickness d1 = 10nm and InAs layer thickness bandKanemodelforthreewell configurations, where d1 and d2 =8.1nm, which is a normal insulator with no edge states. d2arethethicknessofGaSblayerandInAslayer,respectively. (b)Thedispersionford1=d2 =10nmquantumwellwhichis (d)Theenergygapvariationind1−d2 plane,wherebrighter aQSHinsulatorwithonepairofedgestates. Atight-binding colorsrepresentasmallergap. A,BandConthedashedblue regularization with lattice constant a = 20˚A is used in this lineindicate respectively theplace where(a), (b)and (c) are calculation. plotted. NIandQSHdenotethephasesinthecorresponding region of parameter space. whichsupports the existence ofa QSHphase in this sys- tem. Thoughthe BIA and SIA terms do not destroy the QSH state for d2 > d2c (Fig. 3 (c)). As the band in- QSHphase,theydomodifythequantumphasetransition version is only determined by the relative positions of between the QSH phase and normal insulator (NI). The E1 and H1, the quantum wells with other values of d1 transition (gap-closing) will generically occur at finite-k behave essentially the same. As the QSH phase and NI rather than at the Γ point, and a nodal region between phase are always separated by a gap closing point, we QSH and NI phases can possibly appear in the phase can determine the d1−d2 phase diagram via the energy diagram[17, 18, 19, 20]. gap. As shown in Fig. 3 (d), two gapped regimes (in A more direct way of identifying the QSH phase is red) are separated by a critical line (brightly colored) in to study the edge state spectrum. There are always an thed1,d2 plane. Thequantumwellconfigurationsshown oddnumberofKramers’spairsofedgestatesconfinedon in Figs. 3 (a), (b) and (c) are indicated by points A, B the boundary of a QSH insulator, and an even number and C, respectively. Due to the adiabatic continuity, an pairs (possibly zero) for the boundary of the NI phase. entire connected gapped region in the phase diagram is The edge state energy spectrum of the effective model in the NI (QSH) phase once one point in it is confirmed (1) can be obtained by solving this model with a simple to be in this phase. Since Fig. 3 (a) corresponds to the tight-bindingregularizationinacylindricalgeometry,the NI phase and (c) the QSH phase, we identify the right resultofwhichisshowninFig. 2. WefindoneKramers’s side of the diagram as the QSH regime and the left side pair of edge states with opposite spin on each edge for as the NI regime. the QSH side, and no edge states for the NI side. This OneadvantageoftheInAs/GaSb/AlSbquantumwells againconfirmstheexistenceofQSHphaseinthismodel. is that due to the large built-in electric field, the QSH- To study the InAs/GaSb/AlSb quantum well system NI phase transition can be easily tuned by external gate more systematically and quantitatively, we confirm the voltages. When we tune the gate voltage,both the band aboveanalysisbynumericallysolvingtherealistic8-band structure and the Fermi level are adjusted simultane- Kane model. In the inverted regime, there exists an in- ously. Since the QSH effect can only occur when the trinsic charge transfer between the InAs layer and GaSb Fermi level lies in the gap, we need two gates in order layer. Therefore, we need to take into account the built- to independently tune the relative position between the in electric field. The energy dispersions for different well E1 and H1 band edges and the Fermi level. In fact, thicknesses are shown in Fig. 3 (a)-(c), where we fix the such a dual-gate geometry has already been realized ex- GaSb layer thickness d1 = 10nm and vary the thickness perimentally in InAs/GaSb/AlSb quantum wells[10]. In ofInAslayerd2. Thesystemisgappedforagenericvalue thepresentwork,weperformedaself-consistentPoisson- of d2. However, at a critical thickness d2c = 9nm (Fig 3 Schrodinger type calculation[12, 13, 14] for such a dual- (b)) a crossing at finite k occurs between the subbands gate geometry shown in Fig. 1 (a). To simplify the cal- E1 and H1, which marks the phase transition point be- culation,we takethe thicknessof the AlSb barrierlayers tween the QSH and NI phases. According to the above to be much smaller than that in realistic experiments, adiabatic continuity argument, we know that the quan- which has a negligible effect in the quantum well except tum well is in a NI state for d2 < d2c (Fig. 3 (a)) and forarescalingofVf andVb. Wealsoneglectthe weakef- 4 fectsofsubbandanisotropyandintrinsicdonordefectsat Compared to the similar proposal of a gate-induced the InAs/GaSb interface. None of these simplifications phase transition in asymmetric HgTe/CdTe quantum should affect our results qualitatively. wells[21], the InAs/GaSb/AlSb quantum well is much For fixed d1 = d2 = 10 nm we explored the Vf −Vb more sensitive to the gate voltage, which makes it much phase diagramas shownin Fig. 4. There aresix distinct easier to realize such a transition experimentally. Physi- regionsinthefigure. Thedottedblacklineshowsthegap cally,this comes fromthe factthatthe electronandhole closing transitionbetween the invertedand non-inverted wavefunctionsarecenteredinseparatelayers,sothatthe regimes. In parameter regions I,II,III the system has an effect of the gate voltage on them is highly asymmetric. invertedbandstructure,butonlyregionII isintheQSH Thissimplemechanismallowsustoinvestigatethequan- phasewiththeFermileveltunedinsidethebulkgap. Re- tum phase transition from the NI to the QSH state in- gion I (III) is described by the same Hamiltonian as the situ, through the continuous variation of the gate volt- QSHphase,butwithfinitehole(electron)doping. Inthe age, rather than the discrete variation of the quantum same way, region V is the NI phase and IV, VI are the well thickness. It is also useful for developing a QSH correspondingp-dopedandn-dopednormalsemiconduc- FET. The FET is in an ‘OFF’ state when the Fermi tors. Thus,bytuningV andV tothecorrectrange,one level lies inside the normal insulating gap. Then, by ad- f b can easily get the phase transition between QSH phase justing the gate voltages the FET can be flipped to the II and NI phase V. ‘ON’state by passingthroughthe transitionto the QSH phase, where the current is carried only by the dissipa- tionless edge states. This simple device can be operated with reasonable voltages as seen in Fig. 4 but would be more promising if one could enlarge the bulk insulating III gap to support room temperature operation. II ) V ( VI Inconclusion,weproposethattheQSHstatecanbere- Vf I alized in InAs/GaSb quantum wells. We presented both simple arguments based on effective model and realistic V self-consistentcalculations. Inadditionwehaveproposed IV anexperimentalsetuptoelectricallycontrolthequantum phase transition from the normal insulator to the QSH phase. This principle could be used to construct a QSH V (V) FET device with minimal dissipation. b FIG. 4: The phase diagram for different front (V ) and back We would like to thank B. F. Zhu for the useful f (Vb)gatevoltages. RegionsI,II,IIIareintheinvertedregime, discussion. This work is supported by the NSF un- in which the striped region II is the QSH phase with Fermi- der grant numbers DMR-0342832, the US Department levelin thebulkgap, andI,III arethep-dopedandn-doped of Energy, Office of Basic Energy Sciences under con- inverted system. Regions IV,V,VI are in the normal regime, tract DE-AC03-76SF00515and by the Focus Center Re- in which the striped region V is the NI phase with Fermi search Program (FCRP) Center on Functional Engi- level in the bulk gap, and IV, VI are the p-doped and n- neered Nanoarchitectonics (FENA). CXL acknowledges doped normal semiconductors. The well configuration is set the support of China Scholarship Council, the NSF of as d1 = d2 = 10nm, and the AlSb barrier thickness is taken 30nm on each side in the self-consistent calculation. V and China(GrantNo.10774086,10574076),andtheProgram f V aredefinedwithrespecttotheFermilevelinthequantum of Basic Research Development of China (Grant No. b well. 2006CB921500). [1] B. A. Bernevig, T. L. Hughes, and S.C. Zhang, Science [6] L. L. Chang and L. Esaki, Surf. Sci. 98, 70 (1980). 314, 1757 (2006). [7] M. Altarelli, Phys. Rev.B 28, 842 (1983). [2] M. Konig, S. Wiedmann, C. Brune, A. Roth, H. Buh- [8] M. J. Yang, C. H. Yang, B. R. Bennett, and B. V. mann, L. W. Molenkamp, X.-L. Qi, and S.-C. Zhang, Shanabrook, Phys. Rev.Lett. 78, 4613 (1997). Science 318, 766 (2007). [9] M. Lakrimi, S. Khym, R. J. Nicholas, D. M. Symons, [3] C. L. Kane and E. J. Mele, Phys. Rev.Lett. 95, 226801 F. M. Peeters, N. J. Mason, and P. J. Walker, Phys. (2005). Rev. Lett.79, 3034 (1997). [4] C. J. Wu, B. A. Bernevig, and S. C. Zhang, Phys. Rev. [10] L. J. Cooper, N. K. Patel, V. Drouot, E. H. Linfield, Lett.96, 106401 (2006). D. A. Ritchie, and M. Pepper, Phys. Rev. B 57, 11915 [5] C. Xu and J. Moore, Phys. Rev.B 73, 045322 (2006). (1998). 5 [11] E. Halvorsen, Y. Galperin, and K. A. Chao, Phys. Rev. [17] S. Murakami, S. Iso, Y. Avishai, M. Onoda, and N. Na- B 61, 16743 (2000). gaosa, Physical ReviewB 76, 205304 (2007). [12] I.Lapushkin,A.Zakharova,S.T. Yen,and K.A.Chao, [18] X.Dai,T.L.Hughes,X.-L.Qi,Z.Fang,andS.-C.Zhang, J. Phys: Condens. Matter 16, 4677 (2004). arXiv:0705.1516, to be published on Phys. Rev.B. [13] I.Semenikhin,A.Zakharova,K.Nilsson,andK.A.Chao, [19] M. Koenig, H. Buhmann, L. W. Molenkamp, T. L. Phys.Rev.B 76, 035335 (2007). Hughes, C.-X. Liu, X.-L. Qi, and S.-C. Zhang, [14] Y. Naveh and B. Laikhtman, Phys. Rev. Lett. 77, 900 arXiv:0801.0901 (2008). (1996). [20] T. L. Hughes, C.-X. Liu, X.-L. Qi, and S.-C. Zhang, in [15] E.O. Kane, J. Phys. Chem. Solids 1, 249 (1957). preparation. [16] R. Winkler, Spin-Orbit Coupling Effects in Two- [21] W. Yang, K. Chang, and S. C. Zhang, arXiv:0711.1900. DimensionalElectronandHoleSystems (SpringerTracts in Modern Physics, 2003).

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.