Ultra-low carrier concentration and surface dominant transport in Sb-doped Bi2Se3 topological insulator nanoribbons
UUltra-low carrier concentration and surface dominant transport in Sb-doped Bi Se topological insulator nanoribbons Seung Sae Hong, Judy J. Cha, Desheng Kong, and Yi Cui Department of Applied Physics and Department of Materials Science and Engineering,Stanford University, Stanford, California 94305, USA (Dated: October 1, 2018)A topological insulator is a new state of matter, possessing gapless spin-locking surface statesacross the bulk band gap which has created new opportunities from novel electronics to energyconversion. However, the large concentration of bulk residual carriers has been a major challengefor revealing the property of the topological surface state via electron transport measurement. Herewe report surface state dominated transport in Sb-doped Bi Se nanoribbons with very low bulkelectron concentrations. In the nanoribbons with sub-10nm thickness protected by a ZnO layer,we demonstrate complete control of their top and bottom surfaces near the Dirac point, achievingthe lowest carrier concentration of 2 × cm − reported in three-dimensional (3D) topological in-sulators. The Sb-doped Bi Se nanostructures provide an attractive materials platform to studyfundamental physics in topological insulators, as well as future applications. The exotic electronic properties of the surface state,due to its spin-momentum locked Dirac cone in the elec-tronic band structure, define a topological insulator asa new class of quantum matter[1–5]. Moreover, it ispredicted to offer exciting physics in condensed mattersystems, such as elusive quasi-particles, spin transport,and fault tolerant quantum information processing[2–8].Bismuth selenide (Bi Se ) and its relative compoundsare one of the most promising candidates to realize theideal three-dimensional topological insulator due to theirlarge bulk band gap and simple surface band structure[9].There have been significant advances to probe the sur-face state in these materials by various methods, such asangle-resolved photoemission spectroscopy (ARPES)[10–12] and scanning tunneling microscopy (STM)[13–15].Transport measurements in bulk crystals have demon-strated the existence of these surface states as well[16–18].Nanoscale topological insulator devices offer a uniqueopportunity to study the surface state as the surfaceto volume ratio is increased to manifest the surfaceeffect[19, 20]. Moreover, on mesoscopic length scale, onecan study transport reflecting the fundamental nature ofcarriers, as shown in previous cases like graphene[21, 22].However, in the case of topological insulators, materialimperfection issues in the bulk blur surface state signa-tures and limit further in-depth transport studies. One ofthe critical obstacles is the dominant bulk electrons out-numbering surface state electrons[23–26]. Moreover, thematerial is very sensitive to environmental contaminationas sample degradation by environmental exposure hasbeen observed in a few studies[16, 27]. Here, we reportSb doping for Bi Se nanoribbons to suppress the bulkconductivity and achieve surface state dominant trans-port in nanodevices. Transport experiments confirm thesignificant suppression of the bulk contribution withoutdegradation of electron mobility. In addition, we deposit FIG. 1. a , A schematic of vapor-liquid-solid growth of Sb-doped Bi Se nanoribbons. By shifting the dopant sourcelocation along the temperature gradient in the tube furnace,the relative vapor pressure of two sources and the incorpo-rated dopant level are controlled. b , A scanning electron mi-croscopy (SEM) image of as-grown nanoribbons. c , A highresolution TEM image of a nanoribbon and its diffraction pat-tern (inset). d , A low magnification TEM image of a nanorib-bon (top) and its elemental line profile by EDX (bottom),indicating the homogeneous Sb dopant distribution. Insetimage shows the entire morphology of the nanoribbon (17 µ mlong). an oxide layer on top of the nanoribbons as a protectivelayer, which enables ultralow carrier concentration con-trollable near the Dirac point and helps realize surfacedominant transport in the topological insulator nanos-tructures.To suppress the bulk conductivity in topological in-sulator nanoribbons, we synthesize Bi Se nanoribbonswith Sb doping. Sb is known as an effective com-pensation dopant to reduce bulk electron density to10 cm − and does not destroy the topological surface a r X i v : . [ c ond - m a t . m e s - h a ll ] S e p TABLE I. Transport parameters of nanoribbon samples ofdifferent Sb concentrations (T=2K)Sample Sb R sheet R Hall n Hall number doping (Ω) (Ω) (10 cm − )B1 0.0% 43 -7.4 84.5B2 <
2% 177 -14.7 42.5S1 3.0% 422 -39.9 15.7S2 5.5% 975 -51 12.3S3 7.0% 1350 -63.1 9.9 state[16, 28]. Bi Se nanoribbons are synthesized viavapor-liquid-solid growth mechanism using gold parti-cles as catalysts[19, 29], and Sb vapor is introducedfrom a Sb Se source material placed at the lower tem-perature zone (Fig. 1a). As-grown ribbons are 100-300nm thick, their widths vary from 200nm to severalmicrometers, and their lengths are up to tens of microm-eters (Fig. 1b). Sb-doped Bi Se nanoribbons are ina single crystalline rhombohedral phase, which is thesame as undoped Bi Se nanoribbons. The distribu-tion of Sb dopants appears to be spatially uniform inthe nanoribbon, as confirmed by energy-dispersive X-rayspectroscopy (EDX) (Fig. 1c-1d).The Sb doping concentration in Bi Se nanoribbonscan be tuned systematically by controlling the tempera-ture of the Sb Se source material. For different Sb dop-ing concentrations, we fabricated nanodevices with Hallbar electrodes to characterize the basic carrier types anddensities. The entire set of samples (Table I) shows n-type carrier dominant transport as their Hall resistancesare negative values. By introducing more Sb dopantsinto the ribbon, its sheet resistance increases more thanan order of magnitude and the dramatic change in Hallresistance also indicates a much lower carrier density,implying the bulk electron contribution is reduced sig-nificantly (Table I). At a high Sb doping concentration(5-7%), the carrier density reaches 10 cm − . Consid-ering the carrier density of the surface states from bothtop and bottom surfaces near the bulk conduction bandedge is ∼ cm − [31], such low carrier density in theSb-doped nanoribbons suggests that the transport is nolonger dominated by large amount of bulk electrons.Additional electronic transport studies confirm thatthe bulk electron contribution is reduced significantly bythe Sb doping. In Figure 2a, temperature dependent re-sistances from low Sb concentration samples (B1, B2) fol-low typical metallic behavior. In contrast, for the sampleof high Sb concentration (S2), the resistance starts to in-crease and saturates at low temperature. An increasein resistance upon reducing temperature is likely dueto freezing out of the bulk carriers. Moreover, electro-static gating experiments in field effect transistor devicesmanifest drastic difference between low Sb concentrationsamples and high Sb concentration samples (Fig. 2b). FIG. 2. a , The resistance profile versus temperature of sam-ple B1, B2, and S2. The most metallic sample (B1) showsa monotonic decrease by cooling. The sample with a smallamount of Sb doping (B2) follows a metallic temperature de-pendence except the small upturn at T < b , Gating response of the resistance from samplesof different doping concentrations (SiO g = -60V ∼ -70V. c , High field magnetoresistance versusinverse magnetic field (1/B). Background curve (either linearor parabolic) is subtracted from original magnetoresistancecurve. SdH oscillations from an undoped sample (B F ∼ ∼ × cm − andSdH oscillation from sample B2 (B F ∼ × cm − . No oscillatory featureis observed from sample S1, S2, and S3 within the field limit(8T). The curves are displayed with an offset for visual clar-ity. d , Magnetoresistance near zero magnetic field, showingclear feature of weak anti-localization from samples of highSb concentration (S1, S2). Each curve is normalized by zerofield resistance and displayed with an offset for visual clarity. Low Sb-doped samples (B1, B2) show weak gating de-pendence by a bottom gate, which is reasonable as theribbon thicknesses ( > ∼ ∼ g ), which suggests that the oscillation originates frombulk electrons. Without Sb doping, the bulk SdH oscil-lation periodicity in an inverse magnetic field is small. Itindicates the cross section of the Fermi surface is large,consistent with high bulk carrier concentration. As Sbconcentration increases, the oscillation period gets largerand disappears at the highest doping level. That is, thebulk electron pocket crossing the Fermi level eventuallybecomes too small so that the oscillation period is toolarge to be detected in our magnetic field range (8T).The small cross section of the bulk Fermi surface meansthe contribution of bulk electrons is greatly reduced. Inthe low magnetic field, weak anti-localization (WAL), thequantum correction which emerges strongly in the spin-orbit coupled surface state[31], is absent in the MR traceof the undoped sample (Fig. 2d, B1) as the large bulkelectron contribution masks surface state transport. Onthe contrary, higher Sb concentration samples (Fig. 2d,S1, S2) manifest a sharp cusp near zero magnetic field,which is a characteristic feature of WAL. Owing to thereduced bulk electron contribution by Sb doping, all thetransport measurements including temperature depen-dence, field effect gating, and magnetotransport consis-tently indicate significant contribution from surface statetransport.Even though the 5-7% Sb doping in the aforementionedexperiments can effectively reduce the carrier concen-tration to ∼ cm − and lower the Fermi level closeto the conduction band edge, nanoribbons are still ex-posed to extrinsic contaminations during device fabrica-tion, which have the issue of the additional increase ofthe bulk carrier concentration, as reported by previousstudies[16, 27]. Therefore, we hypothesize that the Sb-doped Bi Se should have much lower carrier concentra-tion if protected from the ambient environment. Here wedeposit an insulating Zinc Oxide (ZnO) layer on top ofthe nanoribbons to protect samples from external con-tamination. The sputtered ZnO layer covers the entiresurface of the nanoribbons, which would prevent degra-dation and extrinsic doping that might happen duringthe standard fabrication process.Here we fabricate a bottom-gate device using Sb-dopedand ZnO protected Bi Se nanoribbon with a large thick-ness ( ∼ ∼ FIG. 3. a , Device schematic (top) and optical image (bot-tom) of a thick sample ( ∼ µ m. b , Gatevoltage dependence of longitudinal sheet resistance (R sheet )and Hall resistance (R xy ) measured at low field (B < g , im-plying that the Fermi level of the bottom surface is near theDirac point at zero gating voltage. The anomalous kink in thelongitudinal resistance curve is not well understood, and it isnot reproducible in other samples. c , Temperature-dependentresistance curve at different V g (negative). The curve changessignificantly as more holes are added, reflecting the conven-tional metallic temperature dependence of induced carriers.A band diagram (inset) shows band bending at the bottomsurface induced by gating. d , Temperature dependent resis-tance curve at different gate voltages (positive). In a banddiagram (inset), the Fermi level does not cross any bulk bandby positive gating, explaining the qualitatively similar tem-perature curves over the wide range of gating voltage. tom surface in this particular device is close to the Diracpoint. In Figure 3b, its resistance decreases by either di-rection of gating voltage, and Hall resistance increases asmore electronic states on the bottom surface (p-type car-riers for negative V g and n-type carriers for positive V g )are populated. At zero V g , resistance increases with de-creasing temperature, and saturates at low temperature,similar to the case of Sb-doped bulk crystal[16]. The in-crease of resistance at high temperature (T > g (Fig. 3c), there is a noticeablechange in the temperature curve - resistance starts todrop down by cooling. As we have observed this metallictemperature dependence when the Fermi level is cross-ing the bulk band, we conclude that it becomes a mixedstate with bulk carriers as increasing p-type carriers. Onthe other hand, the temperature curve does not changequalitatively by inducing more n-type carriers (positiveV g , Fig. 3d).The asymmetric trends in the temperature dependenttransport reflect the characteristic band structure nearthe Dirac point in Bi Se topological insulators. FromARPES studies[10, 30], the Dirac point of surface statesis just above the bulk valance band edge, while it is rel-atively far apart from the bulk conduction band edge( ∼ g range (+80V), posi-tive gating only creates more electrons from the surfaceband but not from the bulk band, which will not changethe overall shape of the temperature curve. In contrast,bulk electronic states are easily populated from the va-lence band by negative gating. Now, the ribbon is in themixed state, evolving from surface dominant transportto bulk dominant transport by adding more bulk holecarriers by gating, showing the dramatic change in thetemperature-dependent resistance curves. This temper-ature dependence study clearly shows that we can inde-pendently tune the Fermi level of one surface near theDirac point. This is an interesting system to control-lably build either a topological insulator junction of dualtypes of carriers or a single surface junction to studynovel proximity effects in the future[7, 8].In addition to the manipulation of one surface in thethick ribbons, we eventually want to control the Fermilevel of the both top and bottom surfaces to achieve ul-tralow carrier concentration near the Dirac point. Toshift the Fermi level of the entire ribbon together byelectrostatic gating, the nanoribbons need to be fur-ther thinned down. We etched a thick, Sb-doped ribbon( ∼ × cm − at zero V g (Fig. 4b and 4d). Since thisconcentration value is more than three times lower thanthat ( ∼ cm − ) of the surface states from the bothtop and bottom surfaces of which Fermi level is near thebulk conduction band edge, we believe that the Fermilevel of the both top and bottom surfaces is completelywithin the bulk band gap. In other words, the Fermilevel crosses only the surface Dirac cone above the Diracpoint.In this ZnO-protected nanoribbon device already with FIG. 4. a , Device schematic (top) and optical image (bottom)of thin ribbon sample ( ∼ µ m. b ,Gate voltage dependence of longitudinal resistance (R xx ) andHall resistance (R xy ) measured at low magnetic field (B < g > -50V); increases as induced p-type carriers fromthe bottom surface form a mixed state with decreasing n-typecarriers (-65V < V g < -50V); and decreases again as the entiresample becomes a hole conductor (V g < -65V). c , Conduc-tance ( σ ) versus V g curve is linear, except near the minimumconductance. d , Electron density (n Hall ) plot as a functionof V g . It depends on V g linearly in the wide range of volt-age (-90V to +80V). e , Semi-log scale plot of carrier densitynear its charge neutrality point (Dirac point). Band diagramsof top and bottom surfaces (inset) for samples with pure p-type conduction (left), mixed conduction (middle), and puren-type conduction (right). very low carrier concentration, its small thickness of 6nmmakes it possible to shift the Fermi level of the entire rib-bon by electrostatic gating across the Dirac point. WhenV g is changed from positive to negative bias, its longitu-dinal resistance (R xx ) first increases and reaches a largepeak value ( ∼ g of -50V, and then decreaseswhen further lowering the gating voltages below V g of-65V (black curve, Fig. 4b). The Hall resistance (redcurve, Fig. 4b) changes significantly too, as its magni-tude increases by more than 10 times and switches its signalso at V g of -60V. These results clearly demonstrate theambipolar field effect and suggest that the entire sampleis converted from n-type to p-type. The sample conduc-tance depends on gating voltage linearly, except at theplateau of minimum conductance ( ∼ ) (Fig. 4c).The absence of zero conductance region during the am-bipolar transition is due to the gapless surface states.The origin of the conductance plateau is not definitiveyet, which could be explained as suggested in previoustheoretical studies[32–34]. The carrier density obtainedby Hall resistance is plotted in Figure 4d showing its lin-ear dependence on V g as well. The sample stays puren-type until V g = -50V, and then switches to a mixedcarrier state (-65V < V g < -50V) and p-type (V g < -65V)subsequently (Fig. 4e). Considering the order of low car-rier density ( < cm − ), now both surfaces are veryclose to the Dirac point in this range of V g . In fact, theminimum carrier density obtained from Hall resistance isabout 2 × cm − , which is 10 times lower than the con-centration at zero V g . This low carrier density value isalso comparable with that of initial single layer graphenedevices[21, 22].The Sb doping of topological insulator nanoribbonsand the protective oxide layer effectively reduce bulk car-riers originating from both intrinsic and extrinsic ma-terial defects. Consequently, we are able to carry outelectronic transport measurement to demonstrate surfacestate dominated transport as a result of fairly low bulkcarrier concentration. In the thick sample, single sur-face can be individually tuned within the wide range ofthe Fermi level. In the thin sample, both top and bot-tom surfaces are easily tunable across the Dirac pointby electrostatic gating. With minimal contribution fromresidual bulk carriers, Sb-doped Bi Se nanoribbons willoffer a great opportunity to test various topological in-sulator phenomena proposed by theory, as well as be theoptimal material for future applications. Acknowledgments
We thank K. Lai and J. R. Williams for the helpfuldiscussions, and B. Weil for the help in the manuscript preparation. Y.C. acknowledges the supports from theKeck Foundation, DARPA MESO project (N66001-11-1-4105), and King Abdullah University of Science andTechnology (KAUST) Investigator Award (No. KUS-l1-001-12). [1] J. E. Moore, Nature , 194 (2010).[2] X.-L. Qi, S.-C. Zhang, arXiv: (2010).[3] M. Z. Hasan, C. L. Kane, Rev. Mod. Phys. , 3045(2010).[4] L. Fu, C. Kane, Phys. Rev. B , 04530 (2007).[5] M. Konig et al., Science , 766 (2007).[6] X.-L. Qi, R. Li, J. Zang, S.-C. Zhang, Science , 1184(2009).[7] B. Seradjeh, J. E. Moore, M. Franz, Phys. Rev. Lett. , 066402 (2009).[8] L. Fu, C. Kane, Phys. Rev. Lett. , 096407 (2008).[9] H. Zhang et al., Nat Phys , 438 (2009).[10] Y. Xia et al., Nat Phys , 398 (2009).[11] Y. L. Chen et al., Science , 178 (2009).[12] Y. Zhang et al., Nat Phys , 584 (2010).[13] P. Roushan et al., Nature , 1106 (2009).[14] Z. Alpichshev et al., Phys. Rev. Lett. , 016401 (2010).[15] T. Hanaguri et al., Phys. Rev. B , 081305 (2010).[16] J. G. Analytis et al., Nat Phys , 960 (2010).[17] D.-X. Qu et al., Science , 821 (2010).[18] A. A. Taskin et al., Phys. Rev. Lett. , 016801 (2011).[19] H. Peng et al., Nat Mater , 225 (2009).[20] F. Xiu et al., Nat Nano , 216 (2011).[21] Y. Zhang, Y.-W. Tan, H. L. Stormer, P. Kim, Nature , 201 (2005).[22] K. S. Novoselov et al., Nature , 197 (2005).[23] J. G. Checkelsky, Y. S. Hor, R. J. Cava, N. P. Ong, Phys.Rev. Lett. , 196801 (2011).[24] H. Steinberg et al., arXiv: (2011).[25] D. Kim et al., arXiv: (2011).[26] D. Kong et al., arXiv: (2011).[27] D. Kong et al., ACS Nano , 4698 (2011).[28] Y. Zhang et al., Appl. Phys. Lett. , 194102 (2010).[29] D. Kong et. al., Nano Lett. , 329 (2010).[30] J. G. Analytis et al., Phys. Rev. B , 205407 (2010).[31] J. Chen et al., Phys. Rev. Lett. , 176602 (2010).[32] C.-X. Liu et al., Phys. Rev. B , 041307 (2010).[33] J. Linder, T. Yokoyama, A. Sudb , Phys. Rev. B ,205401 (2009).[34] D. Culcer, E. H. Hwang, T. D. Stanescu, S. Das Sarma,Phys. Rev. B82