Temperature-Induced Coexistence of a Conducting Bilayer and theBulk-Terminated Surface of the Topological Insulator Bi2Te3Paula M. Coelho,† Guilherme A. S. Ribeiro,† Angelo Malachias,† Vinicius L. Pimentel,‡ Wendell S. Silva,†

Diogo D. Reis,† Mario S. C. Mazzoni,† and Rogerio Magalhaes-Paniago*,†

†Departamento de Física, Universidade Federal de Minas Gerais, Belo Horizonte, MG, CEP 30123-970, Brazil‡Laboratorio Nacional de Nanotecnologia, CP 6192, Campinas, SP, 13083-970, Brazil

ABSTRACT: Topological insulators such as Bi2Se3 and Bi2Te3 have extremely promising transport properties, due to theirunique electronic behavior: they are insulators in the bulk and conducting at the surface. Recently, the coexistence of two types ofsurface conducting channels has been observed for Bi2Se3, one being Dirac electrons from the topological state and the otherelectrons from a conventional two-dimensional gas. As an explanation for this effect, a possible structural modification of thesurface of these materials has been hypothesized. Using scanning tunneling microscopy we have directly observed the coexistenceof a conducting bilayer and the bare surface of bulk-terminated Bi2Te3. X-ray crystal truncation rod scattering was used to directlyshow the stabilization of this epitaxial bilayer which is primarily composed of bismuth. Using this information, we have performeddensity functional theory calculations to determine the electronic properties of the possible surface terminations. They can beused to understand recent angular resolved photoemission data which have revealed this dual surface electronic behavior.

KEYWORDS: Topological insulators, 2D crystals, surface structure, electronic properties, scanning tunneling microscopy,X-ray diffraction

Topological insulators (TI) have revealed extremelyinteresting electronic properties. While the bulk has an

insulating behavior, Dirac fermions localized at the surface areresponsible for the charge conduction, which is protectedagainst back scattering by symmetry properties.1−3 A subtlemechanism based on the effects of the spin orbit couplingunderlies this phenomenology, which is, therefore, observed incompounds formed by heavy elements, such as Bi2Te3 andBi2Se3. Novel phenomena, such as the possible observation ofMajorana fermions,4 arises in connection with the interaction ofTI’s surface with other materials. The coexistence of Diracfermions with other states has, therefore, become a topic ofincreasing interest. In this context, it is worth mentioning thedescription of a two-dimensional electron gas on the surface ofBi2Se3 due to band-bending effects observed by angular-resolved photoemission electron spectroscopy (ARPES).5 Also,the appearance of an interfacial Bi bilayer has been reported forboth Bi2Te3 and Bi2Se3 TI’s.6,7 For the former, an epitaxialgrowth of Bi on top of the Bi2Te3 surface was carried out, whilein the latter the surface termination by bismuth was observed

by low energy ion scattering (LEIS) after a cleaving procedureat room temperature. In this case, it remained unclear whichmechanism was responsible for the bilayer formation; that is, ifit was selenium evaporation or Bi diffusion from the bulk. Theformation of a bismuth bilayer was hypothesized and shown tobe stable by ab initio calculations.7

In this work, we present the first experimental evidence that atemperature-controlled process leads to a self-assembled Bibilayer on top of the TI Bi2Te3. Our data strongly support theevaporation hypothesis in the question raised by ref 7. Incontrast with most experiments, where the surface of Bi2Te3 isexposed by cleaving, in our case the surface was prepared by ionsputtering and subsequent annealing at high temperatures (573K). Scanning tunneling microscopy (STM), X-ray crystaltruncation scattering (CTR), and calculations based on thedensity functional theory (DFT) formalism were employed to

Received: July 4, 2013Revised: August 9, 2013Published: August 16, 2013

Letter

pubs.acs.org/NanoLett

© 2013 American Chemical Society 4517 dx.doi.org/10.1021/nl402450b | Nano Lett. 2013, 13, 4517−4521

characterize our samples. Based on our results, we propose asystematic process to produce heterojunctions of Bi-bilayer-Bi2Te3/Bi2Te3 with potential applications to device fabrication.Our Bi2Te3 sample was grown by the Bridgman method, in

which metallic high purity (99.99%) Bi and Te were mixedstoichiometrically, evacuated to 10−6 mbar in a quartz tube, andheated at 600 °C for approximately 4 days. The sample wassubsequently cooled down to room temperature at a constantrate of 30 °C/h. After cooling, a single crystal was visuallydetached from the sample and exhibited the usual graphite-likeaspect, in which it was possible to remove a few layers byexfoliation. The surface of Bi2Te3(0001) was prepared through10 cycles of 15 min of sputtering of argon+ ions at 1.25 KeVand annealing at 300 °C for 30 min under ultrahigh vacuumconditions (better than 2 × 10−10 mbar). The sample was thencooled down before the STM images were taken. Aftersputtering and annealing the STM measurements were

performed with an Omicron UHV/STM microscope at roomtemperature with an electrochemically etched and UHV-annealed tungsten tip. Scanning voltage and current valueswere −1.5 V and 100 picoAmps, respectively. The imagespresented in Figure 1 are typical of the whole sample aschecked for at least three different regions.The atomic structure of Bi2Te3 is composed of quintuple

atomic layers (with the sequence Te−Bi−Te−Bi−Te) ofapproximately 1 nm of height with van der Waals interactionskeeping them connected one with another. Similar to graphene,cleavage of these materials is possible and usually leads to theformation of a few quintuple layers.8,9 Indeed, the firstobservation from Figure 1a is the presence of extended terracesof typically 20 nm in length which in principle could beassociated to the formation of complete quintuple layers. A lineprofile over all terraces (Figure 1b) confirms this idea, showingwell-defined plateaus whose heights are multiples of 1 nm. The

Figure 1. Scanning tunneling microscopy of Bi2Te3. (a) Low-resolution STM images of the Bi2Te3 surface terraces. Two types of terraces can beidentified: bulk terminated (green) and bilayer terminated (blue). (b) Height profile along terraces (following the white line of part a) showing thedifferences in height of 1 nm (green) and 0.4 nm (blue). (c) High-resolution STM images showing the hexagonal structure of the bare surface.

Figure 2. Structural configuration of a bilayer formed on top of bulk terminated Bi2Te3. (a) Top view of bulk-terminated Bi2Te3. (b) Top view of abilayer on top of Bi2Te3. (c) Side view of a bilayer on top of Bi2Te3. The hexagonal planar structure of the bilayer6 is nearly the same as Bi2Te3.Green spheres: Bulk tellurium atoms. Red spheres: Bulk bismuth atoms. Orange spheres: first monolayer atoms. Yellow spheres: second monolayeratoms.

Nano Letters Letter

dx.doi.org/10.1021/nl402450b | Nano Lett. 2013, 13, 4517−45214518

high-resolution STM image shown in Figure 1c suggests thatsuch terraces have the usual hexagonal Te termination expectedfor quintuple layers, without any changes in chemicalcomposition. Indeed, the impurities which may be seen inFigure 1c on such terraces are quite similar to the clover type ofdefects described by Urazhdin et al.,10 which corresponds toexcess bismuth on top of a surface Te layer, rather than theimpurity resonances due to vacancies.11

However, a second characteristic terrace height is clearlyidentified in Figure 1b (indicated by blue lines), this timeextending 0.45 nm above the previous terraces, which wouldamount to approximately two additional layers. Considering thealternate sequence of bismuth and tellurium and the possibilityof desorption of tellurium, two terminations should beconsidered: one with Te and Bi alternate layers and anotherwith a Bi bilayer. This possibility is illustrated in Figure 2, whichpresents a structural model of the bulk-terminated surface (a)and of a hexagonal bilayer on Bi2Te3 (top view b and side viewc). To compare with other experimental techniques, the overallbilayer coverage was estimated employing the scanning probeimage processor (SPIP) image analysis software.12 By defining aheight threshold of 0.45 nm for the contour of the terraces(over a measured scan area of 120 nm × 60 nm), we calculatedthe Bi2Te3 coverage (bulk-terminated) to be 63.7%.To evaluate the epitaxy and composition of the surface,

deciding between the two main possibilities (Te−Bi or Bi−Bitermination) we have performed X-ray crystal truncation rod(CTR) measurements and first-principles calculations. CTR isthe technique of choice to determine the presence oftermination layers and is also sensitive to strain and differenttypes of stacking of epilayers.13 The Bi2Te3 sample was taken tothe bending magnet X-ray beamline XRD2 of the BrazilianSynchrotron Radiation Laboratory (LNLS). The CTR surfacemeasurements were performed using a Huber four-circlediffractometer. An orientation matrix for Bi2Te3 was builtprior to the measurements using eight different Bragg peaks.Using this matrix, the 00L and 01L crystal truncation rod lineswere followed.In Figures 3a and b we show 00L and 01L scans performed

to determine the surface termination. The 00L line correspondsto a specular reflectivity scan, where one is sensitive only to thevertical position of the atoms within the layers, with noinformation about the horizontal position of bismuth andtellurium atoms. This is the case since the horizontalcomponent of the momentum transfer Kf−Ki is equal to zero.The 01L scans, however, are sensitive both to the vertical andhorizontal atomic positions of the atoms of bulk Bi2Te3 as wellas any epilayer which is epitaxial with respect to the underlyingsurface.For comparison, the X-ray truncation rod scattering intensity

was calculated using the structure factor for the bulk (a semi-infinite crystal) Fb and the bilayer-terminated surface FBTS:

13

∝ | | + − | + |I q F q F q F q( ) frac ( ) (1 frac) ( ) ( )b2

b BTS2

where Fb(q ) = ∑n=2n=∞f ne

−iq ·rne−μn and FBTS(q ) = f Bi + f1e−iq ·r1e−β.

β is the attenuation factor due to the absorption of X-rays byone atomic monolayer ,and f n is the X-ray atomic scatteringfactors of individual atoms (either bismuth or tellurium). q =Hb1 + Kb2 + Lb3 is the X-ray scattering vector, H, K, and L arethe reciprocal lattice coordinates, and b 1, b2, and b3 are thehexagonal Bi2Te3 reciprocal lattice vectors.14 The parameterfrac stands for the fraction of the surface which consist ofBi2Te3 coverage (bulk-terminated) and f1 is equal to f Bi or f Te

depending on the termination model. The vectors rn representthe atomic positions of the atoms obtained from ref 14.Considering the structure factor of Bi2Te3, the truncation

rods are terminated in Bragg peaks for every (0 0 3n) reflectionfor the 00L truncation rod and (0 1 3n+2) for the 01L. InFigure 3c and d, we show least-squares fits of the experimentaldata considering several structural models for the surfacetermination: bulk terminated Bi2Te3, bismuth bilayer (Bi2)terminated, 1/2 bulk terminated, and 1/2 Bi2 (or 1/2 TeBi)bilayer terminated.These fits were performed minimizing a modified chi-square

parameter χ2 = ∑i=1,N[ln(IH,K,Li) − ln(Iexp)]2/N (where N is the

number of data points), for both truncation rods concom-itantly. Minimization of χ2 by solely varying the Bi2Te3 coverageyielded an optimal value for frac = (57 ± 5)%, in agreementwith STM results.The fit of both scans reveals that the bilayer is completely

epitaxial with respect to the bulk structure. It also shows that allpossible differences in lattice parameter of the bilayer structureto the bulk Bi2Te3 are accommodated by strain, which shouldbe of the order of 3.4%, if one considers the difference in latticeparameter between Bi2Te3 and Bi2 in the (111) orientation.7 Asshown in Figure 3c and d an analysis among the differentstructural models rules out pure bulk and bilayer terminations,showing that the STM results are representative of the whole

Figure 3. X-ray crystal truncation rod scattering for the Bi2Te3 surface.Geometry of CTR scans (a) reflectivity (00L) and (b) off-axis (01L)CTR scans. Comparison of calculated CTR line profiles for severalsurface termination models and experimental data for (c) 00L and (d)01L scans.

Nano Letters Letter

dx.doi.org/10.1021/nl402450b | Nano Lett. 2013, 13, 4517−45214519

surface. The results strongly support 1/2 bulk−1/2 Bi2 bilayertermination.To be able to compare these terminations and ARPES

results, band structure first-principles calculations wereperformed based on the DFT formalism as implemented inthe QUANTUM-ESPRESSO15 code. Fully relativistic ultrasoftpseudopotentials16,17 were employed, which lead to a self-consistent treatment of the spin−orbit interaction. We madeuse of a plane-wave basis set defined by a cutoff energy of 35Ry, and the generalized gradient approximation was adopted forthe exchange-correlation functional within the Perdew−Burke−Ernzerhof (PBE)18 parametrization.The calculations were first performed for the standard Bi2Te3

structure terminated in Te, including five quintuple layers tocorrectly describe the dispersion relations, and the result,shown in Figure 4a, agrees very well with other first-principlescalculations.19 Next, we considered two models for theinclusion of the two additional layers on top of this structure,a Te−Bi and a Bi−Bi bilayer. The formation energies of thethree models in a Bi-rich environment are −0.05 eV, 5.38 eV,and −2.59 eV, respectively, assuming a Bi chemical potentialfrom an atomic calculation and applying the thermodynamicconstraint 2μ(Bi) + 3μ(Te) = μ(Bi2Te3) to calculate μ(Te)from the bulk Bi2Te3 chemical potential.In addition to this genuine stability, the electronic properties

of these two surface terminations are markedly distinct as it canbe seen in Figure 4b and c. The Bi2Te3:Bi2 bilayer bandstructure (Figure 4c), as already discussed in ref 6, presents awell-defined Dirac point and two dispersive bands crossing theFermi level, features that have been observed in ARPESexperiments.5 On the other side, for the Bi2Te3:BiTe system, itseems that doping effects and additional surface statescontribute to the appearance of more bands in the Fermilevel. Evidently, ARPES measurements could easily distinguishthese two models.

From all of these results, we can describe the followingscenario for the formation of the Bi bilayer on top of Bi2Te3.Once the sample is cleaved, we have probably the appearance oftwo-dimensional islands of incomplete quintuple layers, which,in principle, enables the desorption of Te atoms. However, atroom temperature, this process is not effective, and therefore,even incomplete quintuple layers should be stable at roomtemperature. Once heated, the Te desorption process increases.It seems that the thermodynamic stability of the two surfaceterminations (Bi2Te3 and Bi2Te3:Bi2) is nearly the same, sinceboth of them coexist in our samples. The procedure used heremay even enhance the formation of a pure Bismuth bilayer, dueto the preferential sputtering of tellurium atoms.In contrast, for Bi2Se3 it seems that at room temperature only

the exposed bilayer is present.8 This obviously should dependon the surface preparation. A study of the formation of a Bi2bilayer in Bi2Se3 could be easily performed, where the amountof selenium desorbed from the surface could be controlled bytemperature. This temperature-dependent behavior would alsoimpact the electronic properties of Bi2Se3 where a dualelectronic behavior was observed.5 Although these findingswere ascribed to band bending effects, this change in the surfacestructural configuration should be considered while analyzingARPES measurements.On should point out that some previous works have already

reported devices made with the Bi2Te3 family of materialswhich revealed mixed surface and bulk transport. For instance,Goyal et al.20 measured the thermal conductivity of bulk andstacked quintuple layers of Bi2Te3 and showed that they exhibitopposite behavior, possibly due to the desorption of telluriumas temperature is increased. The carrier transport in these filmswas likely of mixed volume and surface nature. Also Hossain etal.21 measured noise characteristics in Bi2Se3 which could beattributed to the mixed volume-surface transport regime. It isclear that transport and ARPES experiments have to take into

Figure 4. Comparison of electronic band diagrams for different surface terminations as calculated using DFT. (a) Bulk terminated Bi2Te3, (b) Bi2Te3with a TeBi bilayer, and (c) Bi2Te3 with a Bi2 bilayer. The green and red disks indicate bismuth and tellurium bulk atomic monolayers, respectively.Orange disks represent the topmost bismuth monolayers. The Fermi level indicated by the dashed line is set to zero.

Nano Letters Letter

dx.doi.org/10.1021/nl402450b | Nano Lett. 2013, 13, 4517−45214520

account this dual electronic behavior due to the surfacemodification.Finally, one should consider the possible applications of this

temperature-dependent effect. The process described abovemay be performed in selected regions of the Bi2Te3 crystal byapplying, for instance, microscopic tips to locally enhance thetemperature. This suggests several possibilities for thefabrication of devices, with possible preparation of very well-defined terraces in desired locations. One could, for instance,design devices where the transport is either by conventional orDirac electrons at selected regions. Terminals and gates couldbe easily implemented. Obviously, several different effectsshould be considered beforehand, namely, electronic junctionbarriers at the bismuth−Bi2Te3 interface and pinning effectsdue to the presence of bismuth impurities.In summary, we have directly observed an epitaxial bilayer

formed on the topological insulator Bi2Te3. Contrary to thesurface of this material, the bilayer is a two-dimensionalelectron gas. Combining scanning tunneling microscopy withX-ray crystal truncation rod scattering, we have shown that thisbilayer is epitaxial with respect to the substrate, and in principle,its formation could be controlled by temperature. This materialcould be used in several different devices where theconcomitant use of conventional and Dirac electrons is needed.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: rogerio@fisica.ufmg.br.Author ContributionsV.L.P., P.M.C., D.D.R., and W.S.S. performed the STMmeasurements. P.M.C., A.M., and R.M.-P. performed the X-ray experiment and analysis. G.A.S.R. and M.S.C.M. calculatedthe electronic band diagrams. The manuscript was writtenthrough contributions of all authors.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors acknowledge computational help from CEN-APAD-UFMG and experimental support from LNLS, wherethe X-ray diffraction experiments were carried out. Thescanning tunneling microscopy experiment was performed atLNNANO. This research was financed by the NationalInstitute for Nanocarbon Science (INCT-Nanocarbono) andby the funding agencies CNPq and FAPEMIG.

■ REFERENCES(1) Hsieh, D.; Qian, D.; Wray, L.; Xia, Y.; Hor, Y. S.; Cava, R. J.;Hasan, M. Z. A Topological Dirac insulator in a quantum spin Hallphase. Nature 2008, 452, 970.(2) Chen, Y. L.; Analytis, J. G.; Chu, J.-H.; Liu, Z. K.; Mo, S.-K.; Qi,X. L.; Zhang, H. J.; Lu, D. H.; Dai, X.; Fang, Z.; Zhang, S. C.; Fisher, I.R.; Hussain, Z.; Shen, Z.-X. Experimental Realization of a Three-Dimensional Topological Insulator Bi2Te3. Science 2009, 325, 178.(3) Hsieh, D.; Xia, Y.; Qian, D.; Wray, L.; Meier, F.; Dil, J. H.;Osterwalder, J.; Patthey, L.; Fedorov, A. V.; Lin, H.; Bansil, A.; Grauer,D.; Hor, Y. S.; Cava, R.; Hasan, M. Z. Observation of Time-Reversal-Protected Single-Dirac-Cone Topological-Insulator States in Bi2Te3Sb2Te3. Phys. Rev. Lett. 2009, 103, 146401.(4) Fu, L.; Kane, C. L. Superconducting Proximity Effect andMajorana Fermions at the Surface of a Topological Insulator. Phys. Rev.Lett. 2008, 100, 096407.(5) Bianchi, M.; Guan, D.; Bao, S.; Mi, J.; Iversen, B. B.; King, P. D.C.; Hofmann, P. Coexistence of the topological state a two-

dimensional electron gas on the surface of Bi2Se3. Nat. Commun.2010, 1, 128.(6) Hirahara, T.; Bihlmayer, G.; Sakamoto, Y.; Yamada, M.; Miyazaki,H.; Kimura, S.-I.; Blugel, S.; Hasegawa, S. Interfacing 2D-3DTopological Insulators: Bi(111) Bilayer on Bi2Te3. Phys. Rev. Lett.2011, 107, 166801.(7) He, X.; Zhou, W.; Wang, Z. Y.; Zhang, Y. N.; Shi, J.; Wu, R. Q.;Yarmoff, J. A. Surface Termination of Cleaved Bi2Se3 Investigated byLow Energy Ion Scattering. Phys. Rev. Lett. 2013, 110, 156101.(8) Teweldebrhan, D.; Goyal, V.; Balin, A. Exfoliation Character-ization of Bismuth Telluride Atomic Quintuples Quasi-Two-Dimen-sional Crystals. Nano Lett. 2010, 10, 1209.(9) Li, Y. Y.; Wang, G.; Zhu, X. G.; Liu, M. H.; Ye, C.; Chen, X.;Wang, Y. Y.; He, K.; Wang, L. L.; Ma, X. C.; Zhang, H. J.; Dai, X.;Fang, Z.; Xie, X. C.; Liu, Y.; Qi, X. L.; Jia, J. F.; Zhang, S. C.; Xue, Q.K. Intrinsic topological insulator Bi2Te3 thin films on Si their thicknesslimit. Adv. Mater. 2010, 22, 4002.(10) Urazhdin, S.; Bilc, D.; Tessmer, S. H.; Mahanti, S. D.; Kyratsi,T.; Kanatzidis, M. G. Phys. Rev. B 2002, 66, 161306.(11) Alpichshev, Z.; Analytis, J. G.; Chu, J.-H.; Fisher, I. R.; Chen, Y.L.; Shen, Z. X.; Fang, A.; Kapitulnik, A. STM Imaging of ElectronicWaves on the Surface of Bi2Te3: Topologically Protected SurfaceStates Hexagonal Warping Effects. Phys. Rev. Lett. 2010, 104, 016401.(12) SPIP software is released by http://www.nanoscience.com/products/spip/SPIP.html.(13) Robinson, I. K. Crystal truncation rods surface roughness. Phys.Rev. B 1986, 33, 3830.(14) Atuchin, V. V.; Gavrilova, T. A.; Kokh, K. A.; Kuratieva, N. V.;Pervukhina, N. V.; Surovtsev, N. V. Solid State Commun. 2012, 152,1119−1122.(15) Giannozzi, P.; et al. J. Phys.: Condens. Matter 2009, 21, 395502.(16) Rappe, A. M.; Rabe, K. M.; Kaxiras, E.; Joannopoulos, J. D.Optimized pseudopotentials. Phys. Rev. B 1990, 41, 1227.(17) We used the pseudopotentials Bi.rel-pbe-dn-rrkjus.UPF Te.rel-pbe-dn-rrkjus.UPF from the http://www.quantum-espresso.org dis-tribution.(18) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,3865.(19) Yazyev, O. V.; Moore, J. E.; Louie, S. G. Phys. Rev. Lett. 2010,105, 266806.(20) Goyal, V.; Teweldebrhan, D.; Balandin, A. A. Appl. Phys. Lett.2010, 97, 133117.(21) Hossain, Md. Z.; Rumyantsev, S. L.; Shahil, K. M. F.;Teweldebrhan, D.; Shur, M.; Balandin, A. A. ACS Nano 2011, 5, 2657.

Nano Letters Letter

dx.doi.org/10.1021/nl402450b | Nano Lett. 2013, 13, 4517−45214521