This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 159.226.37.139

This content was downloaded on 11/12/2015 at 01:54

Please note that terms and conditions apply.

Chemical potential fluctuations in topological insulator (Bi0.5Sb0.5)2Te3-films visualized by

photocurrent spectroscopy

View the table of contents for this issue, or go to the journal homepage for more

2015 2D Mater. 2 024012

(http://iopscience.iop.org/2053-1583/2/2/024012)

Home Search Collections Journals About Contact us My IOPscience

2DMater. 2 (2015) 024012 doi:10.1088/2053-1583/2/2/024012

PAPER

Chemical potential fluctuations in topological insulator(Bi0.5Sb0.5)2Te3-films visualized by photocurrent spectroscopy

ChristophKastl1,2, Paul Seifert1,2, XiaoyueHe3, KehuiWu3, Yongqing Li3 andAlexanderHolleitner1,2

1 Walter Schottky Institut andPhysik-Department, TechnischeUniversitätMünchen, AmCoulombwall 4a, D-85748Garching, Germany2 Nanosystems InitiativeMunich (NIM), Schellingstr. 4, D-80799München, Germany3 BeijingNational Laboratory forCondensedMatter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, People’s

Republic of China

E-mail: holleitner@wsi.tum.de

Keywords: topological insulator, surface state, nanoscale optoelectronics, photo-thermoelectric current, atomic force microscopelithography, nanofabrication, dynamic plowing lithography

Supplementarymaterial for this article is available online

AbstractWe investigate the photocurrent properties of the topological insulator (Bi0.5Sb0.5)2Te3 on SrTiO3-substrates.We find reproducible, submicron photocurrent patterns generated by long-range chemicalpotentialfluctuations, occurring predominantly at the topological insulator/substrate interface.Wefabricate nano-plowed constrictions which comprise single potential fluctuations. Hereby, we canquantify themagnitude of the disorder potential to be in themeV range. The results further suggest adominating photo-thermoelectric current generated in the surface states in such nanoscaleconstrictions.

In recent years, a class of solid state materials, calledthree-dimensional topological insulators, has emerged[1–3]. In the bulk, a topological insulator behaves likean ordinary insulator with a band gap. At the surfaces,topologically non-trivial, gapless states exist showingremarkable properties such as a helical Dirac disper-sion and the suppression of backscattering of spin-polarized charge carriers [4, 5]. In principle, thin filmsof topological insulators have two surfaces; the bottomone facing the underneath substrate and the top one.Predominantly, the top surface states have beenexperimentally characterized. For instance, due to apenetration depth of ∼3 nm, an angle resolved photo-emission spectroscopy (ARPES) is mainly limited tothe topological insulator–vacuum interface [4, 6–8].This top surface is also approached by high-resolutionscanning tunneling microscopes [5, 9–11], and it isalways exposed to the environmental conditions of theexperiments. However, the prototypical topologicalinsulator compounds bismuth telluride, bismuthselenide and their respective alloys are not chemicallyinert under ambient conditions [12–14]. This can havean adverse impact on the topological states, because aninterfacial inversion layer can form with topologicallytrivial electronic properties [15]. In contrast, the

bottom surface states are buried, encapsulated andtherefore protected against degradation, which is anessential prerequisite for future electronic devicesbased on topological insulators.

We demonstrate that a photocurrent spectroscopyallows addressing electronic states at the topologicalinsulator/substrate interface, comprising both surfaceandbulk states.Weverify this scheme in circuits of thin(Bi0.5Sb0.5)2Te3-films on SrTiO3-substrates. Such sub-strates with a large dielectric constant are recently dis-cussed to allow a reduction of potential fluctuationswithin two-dimensionalmaterials [16].We apply a so-calleddynamicplowing lithographybasedonanatomicforce microscope (AFM) to the (Bi0.5Sb0.5)2Te3-films.Therewith, we thin down the films by a few nan-ometers and in turn, record a two-dimensionalphotocurrent map. The maps exhibit reproduciblephotocurrent patterns with positive and negativeamplitudes andwith a lateral extension of up tomicro-meters. We observe that the patterns are not changed,when the film-height is reduced by only a few nan-ometers, and that they are even stable after severalcooling cycles, exposures to atmospheric conditions,and/or lithography steps. We interpret these findingsin a way that the photocurrent patterns arise from

RECEIVED

26 February 2015

REVISED

17April 2015

ACCEPTED FOR PUBLICATION

24April 2015

PUBLISHED

27May 2015

© 2015 IOPPublishing Ltd

long-range potential fluctuations at the interfacebetween the topological insulator and the substrate.The observed photocurrent amplitudes are maximumwhen the Fermi-level is tuned to the charge neutralitypoint, which is consistent with a reduced screening ofthe potential fluctuations at a low electron density.Additionally, we laterally pattern the (Bi0.5Sb0.5)2Te3-films into narrow constrictions by the use of the nano-plowing lithography. Such circuits effectively com-prise single potential fluctuations. Correspondingphotocurrentmaps allow us to quantify themagnitudeof the effective disorder potential to be in the order ofmeV. The results suggest that a photo-thermoelectriccurrent within the surface states dominates the photo-current in such narrow constrictions. We further dis-cuss a photovoltaic current within the bulk states andthe impact of lateral boundaries on the electrostaticpotential landscape.

Our photocurrent studies on the narrow constric-tions pave the way towards nanoscale optoelectroniccircuits made out of topological insulators, compris-ing quantum point contacts [17–19] and mesoscopicchannels [20]. Such nanoscale circuits are particularlypromising, since it was recently demonstrated that thephotocurrent amplitude scales inversely with the lat-eral width of the circuits fabricated from topologicalinsulators [21, 22]. Our experiments further reveal thepossibility to selectively probe the transport propertiesof electronic states buried at the interface between atopological insulator and an insulating substrate. Theexperiments disclose the impact of potential fluctua-tions on the transport properties of topological

insulators. This finding may have implications foroptimizing the growth of thin topological films as wellas for the interpretation of linear magneto-resistancedata in terms of electrostatic disorder [23, 24].

Thin films of (Bi0.5Sb0.5)2Te3 with a thickness of15 nm (∼15 quintuple layers) are grown by molecularbeam epitaxy on a SrTiO3(111)-substrate (supple-mentary note 1) [23, 25, 26]. Ternary topological insu-lators such as (Bi1−xSbx)2Te3 or Bi2Te2Se showsubstantially reduced bulk conductivity [6, 25, 27]compared to simple binary compounds as for exampleBi2Se3, Sb2Te3, and Bi2Te3. The latter exhibit an inher-ent bulk doping [28, 29]. The combination of(Bi0.5Sb0.5)2Te3 as a ternary, bulk insulating topologi-cal insulator and SrTiO3 as an ultrahigh-k dielectricsubstrate allows for a full range control of the chemicalpotential from p-doping to n-doping via electricalback-gating at low temperatures [25]. The films arepatterned into 50 μmwide Hall bar devices using stan-dard optical lithography and plasma etching. Ohmiccontacts to the films as well as a gate electrode at thebackside of the substrates are formed by metallizationof chromium and gold. Photocurrent measurementsare carried out at temperatures from 4.2 K up to roomtemperature using a confocal scanning laser micro-scope with an excitation wavelength λ = 806 nm anda diffraction limited spatial resolution of ∼900 nm.The optical excitation occurs at a normal angle of inci-dence and a linear polarization. The current is mea-sured between electrically unbiased contacts, namedsource and drain, using a current–voltage converter(figure 1(a)). By scanning the laser spot across the

Figure 1. Local photocurrents in topological insulators. (a)Microstructured thinfilm of (Bi0.5Sb0.5)2Te3 on SrTiO3-substrate.Photocurrent ismeasured between unbiased source and drain contacts. The charge carrier density is adjusted via the global backgatevoltageVgate. (b)High resolution photocurrentmap of the regionmarked by the rectangle in (a). Red (blue) denotes a current fromsource (drain) to drain (source). Dashed lines indicate sample boundary. Scale bar is 10 μm. Experimental parameters arePlaser = 70 μWandTbath = 77 K.

2

2DMater. 2 (2015) 024012 CKastl et al

sample, the photocurrent I x y( , )photo is detected foreach position on the sample. Figure 1(b) shows such atwo-dimensional photocurrent map for the regionmarked by the dashed rectangle in figure 1(a). We findreproducible submicron photocurrent patterns withpositive (red) and negative (blue) amplitudes. A posi-tive (negative) amplitude corresponds to a currentflowing from source to drain (drain to source). Thephotocurrent patterns stand in contrast to photo-currents routinely observed in planar circuits based onlow-dimensional materials such as graphene, MoS2 orsemiconducting nanowires. In such nano-circuits, thephotoresponse is typically dominated by photovoltaicand photo-thermoelectric currents generated at inter-faces which are for example metal interfaces at ohmiccontacts [30–32], step edges of ultrathin layeredmate-rials [33, 34] as well as heterojunctions and built-infields induced by electrostatic or chemical doping [35–37]. Here, rather than being localized near metal con-tacts, the photoresponse comprises complex spatialpatterns spanning across the entireHall bar.

We excite the (Bi0.5Sb0.5)2Te3 with a photonenergy of 1.54 eV (λ = 806 nm ), which creates elec-tron–hole pairs above the band-gap of Egap∼ 0.2 eV.Time-resolved ARPES on topological insulators showsthat high energy photogenerated charge carriers relaxwithin fs to a heated Fermi–Dirac distribution via car-rier–carrier and optical phonon scattering. Afterwardsthe non-equilibrium Fermi–Dirac distribution equili-brates with the lattice on a ps-time scale [38, 39].Then, the local photocurrent j x y( , )photo can arise

either due to a photovoltaic effect or due to a photo-thermoelectric effect. In the former case, photo-generated electrons and holes are spatially separatedresulting in a current density j .PV In turn, the localcurrent density generates an electric field

ρ= −E jPV PV where ρ is the local resistivity [37]. In thelatter case, the laser induced heat gradient T pro-duces a thermoelectric field = −E S TPTE with theSeebeck coefficient S [33]. As will be discussed in thisletter, the photocurrent profile provides informationon the local potential landscape of the (Bi0.5Sb0.5)2Te3-films. Circular and linear photogalvanic effects arenegligible for the present experiments, because thephoto-excitation occurs at an energy above the bandgap and at a normal angle of incidence [32, 39, 40].Wepoint out that the photocurrent amplitudes depend onthe device geometry. For the region denoted as ‘I’ and‘II’ in figure 1(b), themean amplitude, i.e. the strengthof the photoresponse, is approximately constant.However, as the channel widens, the current ampli-tude decreases. This becomes evident at the transitionbetween the regions ‘II’ and ‘I’. The geometry-depen-dent reduction of the amplitude is discussed in detailin [21] for Bi2Se3-based circuits. It can be understoodas a decreased density of current stream lines that con-nect the local photocurrent to the contacts [22]. In thefollowing, we elaborate that chemical potential

fluctuations at the interface between the(Bi0.5Sb0.5)2Te3-films and the SrTiO3-substrates giverise to the observed photocurrent patterns.

We use the tip of an AFM and employ a so-calleddynamic plowing lithography to thin down the(Bi0.5Sb0.5)2Te3-films with nm-precision (figure 2(a)and supplementary note 2) [41, 42]. In tapping mode,the external drive amplitude of the AFM-tip isincreased and the set-point of the feedback loop isdecreased such that the interaction between the tipand the films is sufficient to mechanically removesample material from the top surface of the(Bi0.5Sb0.5)2Te3. Although initially proposed for struc-turing soft materials such as polymers, the AFM-litho-graphy can equally be applied to semiconductors,metals, and topological insulators by using speciallycoated tips [41, 43]. The (Bi0.5Sb0.5)2Te3-films have apristine thickness of z= 15 nm (∼15 quintuple layers).They are successively thinned down by Δz= 5 nm andΔz= 9 nm for a lateral extension of ∼10 μm (lowerpanels of figures 2(b)–(d)). The main graphs offigures 2(c) and (d) show the corresponding photo-current maps taken after the lithography steps. Sur-prisingly, the photocurrent patterns of the filmthinned by Δz= 5 nm (figure 2(c)) are identical to theones of the pristine film (figure 2(b)). Consequently,the data suggest that the chemical potential fluctua-tions are independent of the top surface morphologyand that the photocurrent mainly stems from the elec-tronic states located at the bottom interface betweenthe substrate and the topological insulator film. Wefurther note that in-between the photocurrent mea-surements shown in figure 2, the films are exposed toambient conditions and temperature cycles. Hereby,the sequence of experiments depicted in figures 2(b)and (c) demonstrates that the optoelectronic responsenear the bottom interface is independent of the envir-onmental conditions.

When the film thickness is reduced to 6 nm (inbetween the triangles in figure 2(d)), the photocurrentresponse changes. At the vertical step edges (blue andred triangle in figure 2(d)), the photoresponse isenhanced by approximately one order of magnitude.Additionally, the sign of the current is reversed at theopposite step edges. The findings are consistent with arecent study, where a photo-thermoelectric effect wasdemonstrated at step edges of 4–12 quintuple layers(4–12 nm) thin Bi2Te3- and Sb2Te3-platelets [44]. Inthis interpretation, the photo-thermoelectric currentresults from the thickness dependence of the electro-nic band structure due to confinement effects. In addi-tion, defect states and imperfections in the nano-plowed top surface may further influence the bandstructure. We expect a substantial quantum mechan-ical coupling due to hybridization of the top and bot-tom surface states only when the thickness of the(Bi0.5Sb0.5)2Te3-films is less than 3–4 nm. We note,however, that for thicknesses below 6 nm, we find thatthe dynamic plowing lithography rather destroys the

3

2DMater. 2 (2015) 024012 CKastl et al

films, i.e. the films rupture into parts (data notshown).

The data shown in figure 2 are measured at zerogate voltage. Figures 3(a)–(c) depict photocurrentmaps as the Fermi-level is tuned from the p-type intothe n-type regime via the backgate voltage. The laserexcitation occurs in the center of the pristine Hall barwhere the geometry of the sample does not constrainthe photocurrent. The exact form of the photocurrentpatterns depends on the applied backgate voltage,which suggests that they are caused by electrostaticpotential fluctuations within the (Bi0.5Sb0.5)2Te3-films. At Vmax≈ 10 V (figure 3(b)), the mean ampli-tude of the photoresponse is maximum. This is evenbetter visible in the top graph of figure 3(d), which

depicts the mean photocurrent amplitude Iphoto as a

function of the gate voltage. Each data point is calcu-lated by averaging the absolute value of the current

Iphoto for an entire two-dimensional photocurrent

map. In particular, the photoresponse is enhancednear the charge neutrality point, where the overallconductance exhibits a minimum (Vmin infigure 3(d)). In our interpretation, the enhancementof the photocurrent near the charge neutrality pointcan have two origins. On the one hand, the maximum

of Iphoto in figure 3(d) can be related to the poor

screening near the charge neutrality point of the bot-tom surface states, since a reduction of the screeningleads to larger fluctuations in the effective electrostaticpotential [45–47]. On the other hand, both the See-beck coefficient S as well as the local resistivity ρ showan enhancement near the charge neutrality pointwhich can effectively lead to an overall increased pho-toresponse of the two-dimensional (Bi0.5Sb0.5)2Te3-film [37, 48].

In a next step, we focus on single potential fluctua-tions to uncover the dominating photocurrentmechanism within the bottom surface states. Inparticular, we create a constriction parallel to thesample boundary by completely removing the(Bi0.5Sb0.5)2Te3-film along a thin straight line(figure 4(a) and supplementary note 2). In this way, acircuit is formed consisting of pristine two-dimen-sional source and drain regions as well as the narrowconstriction. With a lateral width d= (1.8 ± 0.3) μm,the constriction comprises a sequence of single poten-tial fluctuations (figure 4(b)). In figure 4(c), the pho-tocurrent amplitude versus the gate voltage is plottedfor each excitation position along the constriction. Incontrast to the data presented in figure 3, all photo-current puddles change sign at a gate voltage Vzero.

Intriguingly, Vzero is not constant along the

Figure 2.Dynamic plowing lithography. (a) The tip of an atomic forcemicroscope (AFM) is used formechanicalmilling of a(Bi0.5Sb0.5)2Te3-film. (b) Photocurrentmap of pristinematerial with thickness 15 nm.Dashed lines indicate sample boundary.(c) Photocurrentmap after thefilm is thinned by 5 nm.Characteristic photocurrent patterns remain unchanged. (d) Photocurrentmap after thefilm is thinned by 9 nm in total. At vertical step edges, the photoresponse is enhanced (triangles). Colorscale is clippedfor contrast. Scale bars are 5 μm. Lower panels depict schematic side views of thefilmwith an indicated thickness after eachAFM-step.Experimental parameters areVgate = 0 V,Plaser = 65 μW, andTbath = 77 K.

Figure 3.Photocurrents at the charge neutrality point. (a) Photocurrentmap forVgate =−30 V, (b) +10 V, and (c) +24 V. Scale bars

are 2 μm. Insets sketch the position of the Fermi-level in the bottom surface state. (d) Absolute value of photocurrent Iphoto

averaged over two-dimensionalmaps versus gate voltage (top), and corresponding sheet conductanceG (bottom). Letters a, b, and cindicatemeasurements of (a), (b), (c).Vmin highlights theminimumof the conductanceG. Experimental parameters arePlaser = 70 μWandTbath = 4.2 K.

4

2DMater. 2 (2015) 024012 CKastl et al

constriction (figure 4(d)). Therefore, we interpretVzero(x) to mimic the variation of the local chemicalpotential [30], since the Fermi-level can be tuned by

the gate voltage as −( )E V V~ ,F gate Dirac1/2

with

VDirac the gate voltage at which the Fermi-level alignswith the Dirac point of the bottom surface state [45].Within a plate capacitor model for the gate capaci-tance, the maximum peak-to-peak voltage ΔVzero of∼9 V in figure 4(d) translates to a value on the order of6 meV for the chemical potential fluctuations in the(Bi0.5Sb0.5)2Te3-film.

The macroscopic Hall bar geometry does not pro-vide the possibility to directly measure the constric-tion’s conductance by a four-terminal configuration.Therefore, we combine different two-terminal config-urations of the Hall bar to indirectly extract the indivi-dual dark conductance of the source contact and theone of the AFM-fabricated constriction as a functionof Vgate (see figure 4(e)). Notably, the minimum con-ductance of the constriction is shifted to a more nega-tive gate voltage by ΔV ~ 90 Vgate compared to thepristine source contact. The shift suggests that at theconstriction, the energetic position of the electronicstates is altered. As a consequence, also the Dirac pointof the surface state will be shifted downwards inenergy. It is well known that plasma etching inducesdefects at unprotected surfaces [49, 50]. For the(Bi0.5Sb0.5)2Te3-films, such defects could occur at thevertical facets of the lateral edges. As was recently pro-posed [51], the defects may result in a band bending.This explains the observed voltage shift and thereforeenergy shift of the electronic states at the edges of the(Bi0.5Sb0.5)2Te3-films. We point out that we observe asimilar shift when the constrictions are fabricated atthe center of the Hall-bar by the dynamic plowinglithography (supplementary note 3). For such

constrictions, all edges are fabricated by the AFMlithography. Hereby, we conclude that the occurrenceof defect states and their influence on the optoelec-tronic properties of a topological insulator are genericto nanofabricated circuits.

We model the optoelectronic response of the bot-tom surface states by a photothermoelectric current.Starting with the extracted dark conductance G of theconstriction (figure 4(e)), we calculate the gate-voltagedependence of the Seebeck coefficient according to theMott formula [48, 52]

π= − ⋅ ⋅Sk T

e G

G

E3

1 d

d, (1)B

F

2 2⎛⎝⎜

⎞⎠⎟

assuming aDirac systemwith = × −v 3.5 10 m sDirac5 1

[53]. The result is presented in figure 4(f). In ourinterpretation, there exists one dominating potentialfluctuation for each excitation position. Therefore, S(Vgate) can be shifted by ΔVzero when the laser isscanned along the constriction. Then, a laser inducedheat gradient generates a thermovoltage due to thedifference in thermopower along the constriction.The thermovoltage is given by the following expres-sion [53]

Δ Δ

Δ Δ

= + − ⋅

= ⋅( ) ( )V S V V S V T

S T , (2)

th gate zero gate⎡⎣ ⎤⎦

with a typical value of ΔVzero∼ 9 V (figure 4(d)). Wefurther determine a local maximum temperatureincrease of Δ =T 3 K for the used laser power fromfinite element simulations. In turn, we can compute agate voltage dependence of the thermovoltage Vth

according to equation (2) (line in figure 4(g)). Thecomputed dependence qualitatively captures both theexperimentally observed ambipolar sign change of thephotocurrent at Vgate∼ 0 V as well as a higher

Figure 4. Long-range electrostatic potential fluctuations and thermoelectric photocurrent generation. (a) AFM lithography of a1.8 μmwide constriction. (b) Photocurrentmap of the nano-plowed constriction atVgate =−40 Vwith sequence of single potentialfluctuations. Shaded areas indicate lithography and boundary. (c) Photocurrent versusVgate along the constriction. The photocurrentchanges sign for all positions around a varyingVzero close to−2 V. Scale bars are 2 μm. (d)Variations ofVzero along the length of theconstriction extracted from (c). (e) The dark conductance through the constriction (black line) is shifted inVgate compared to the darkconductance of the source contact (gray line). (f) Seebeck coefficient S calculated from equation (1) for the constriction.(g) ThermovoltageVth calculated from equation (2) andmeasured photocurrent as function of gate voltage. The thermovoltage andthe photocurrent change polarity atVgate∼−2 V. Experimental parameters arePlaser = 70 μWandTbath = 4.2 K.

5

2DMater. 2 (2015) 024012 CKastl et al

photocurrent amplitude for negative Vgate (opencircles in figure 4(g)). Hereby, the model suggests thata photo-thermoelectric current within the surfacestates dominates the photocurrent in the constriction.Along this line, the values of Vth∼ 1.2 μV andIphoto∼ 200 pA in figure 4(g) yield a resistanceR∼ 6 kΩ, which is consistent with the local resistanceat the laser spot. For Vgate < 50 V (figure 4(g)), ourmodel clearly starts to deviate from the experimentaldata. In this regime, the bulk valence band statesincreasingly contribute to the photocurrent, which areneglected in the calculation. For each laser positionalong the constriction, the sign of ΔVzero can switch(figure 4(d)). In turn, Vth switches polarity accordingto equation (2). This explains the polarity change ofthe photocurrent for certain positions along theconstriction (figures 4(b) and (c)). We point out thatthe observed sign change excludes a photovoltaic effectto dominate the photoresponse of the bottom surfacestates. This can be seen by the following simpleargument. Let us assume that at a negative gate voltage,the chemical potential fluctuations give rise to adoping profile of p-p+-p along the constriction. Whenthe global back-gate is tuned to positive voltages, itshifts the doping profile finally to a profile of n−-n-n−.For all gate voltages, however, the direction of the localbuilt-in fields and accordingly the polarity of a localphotovoltaic current would remain unchanged [37].With this argument, the photovoltaic effect can onlyinfluence the variation of Vzero at each laser positionalong the constriction (figure 4(d)), but it does notexplain the ambipolar dependence versus Vgate infigure 4(c).

In our interpretation, the laser spot defines anextrinsic length scale of the optoelectronic dynamicsin the (Bi0.5Sb0.5)2Te3-films. However, the occurrenceof micrometer sized photocurrent patterns suggestslong-range potential fluctuations within the bottomsurface state on a comparable length scale. As arguedabove, we interpret them to stem from the interactionwith the SrTiO3-substrate and defects at the edges ofthe (Bi0.5Sb0.5)2Te3-films. Moreover, the laser spotprobes a certain number of current stream lines thatconnect to the macroscopically distant contacts[21, 22]. Then, the AFM-defined constrictions confineall current stream lines on a lateral width which com-pares well with the laser spot diameter. In this sense,the constrictions are quasi-one-dimensional, andthere is always one dominating potential fluctuationand therefore, current direction per laser position. Inturn, we can measure a zero-crossing of the photo-current amplitude versus gate voltage (figure 4(c)). Inthe center of the (Bi0.5Sb0.5)2Te3-films, the excitationspot generates a photocurrent isotropically, i.e. alsointo directions which are perpendicular to the currentstream lines. Hereby, additional current stream linespossibly outside of the laser spot are involved, whichthen connect equally to source and drain. Thisexplains that the corresponding photocurrent patternsrather do not change their polarity as a function of gate

voltage (figures 3(a)–(c)). Consistently, this two-dimensional photocurrent response is maximum atthe charge neutrality point (figure 3(d)). For the lateraledges of the (Bi0.5Sb0.5)2Te3-films, the data suggestthat the involved states are energetically lowered(figure 4(e)). We point out that the cooling cycles canslightly shift the voltage Vmin of the minimum con-ductance (as defined in figure 3(d)), which can be con-sidered approximately as the gate voltage for thecharge neutrality point. However, the distinctionbetween optoelectronically quasi-one-dimensionaland quasi-two-dimensional responses is independentof the temperature cycling. We note that the photo-current patterns reproducibly persist up to room tem-perature. Therefore, we conclude that the bathtemperature only plays a negligible role to describe theoptoelectronic phenomena. However, due to anincreasing dielectric constant of the SrTiO3-substrateat low temperatures, the backgating only works suffi-ciently well below∼77 K (figures 3 and 4).

In summary, we use spatially resolved photocurrentmeasurements to probe the electronic states at the inter-face of (Bi0.5Sb0.5)2Te3-films and the underlyingSrTiO3-substrate. The measured photocurrent patternsare consistent with the existence of submicron chemicalpotential fluctuations near the topological insulator/substrate interface. We introduce an AFM-based nano-plowing lithography to locally thin the (Bi0.5Sb0.5)2Te3-films and to laterally define constrictions with singlepotential fluctuations. For the lateral constrictions, thephotocurrent is consistent with a dominating photo-thermoelectric effect within the surface states of thetopological insulator. We extract potential fluctuationson the order ofmeV.Our resultsmay prove essential forthe design and fabrication of nanoscale circuits fromtopological insulators.

Acknowledgments

We gratefully acknowledge financial support of theDFG-grant HO 3324/8 of the SPP 1666 on topologicalinsulators, the Center for NanoScience (CeNS) andthe Munich Quantum Center (MQC). The work inChina was supported by MOST 973 program(2012CB921703) andNSFC (91121003).

References

[1] HasanMandKaneC 2010Rev.Mod. Phys. 82 3045–67[2] Qi X L andZhang SC2011Rev.Mod. Phys. 83 1057–110[3] Brüne C, LiuC,Novik E,Hankiewicz E, BuhmannH,ChenY,

Qi X, Shen Z, Zhang S andMolenkampL 2011Phys. Rev. Lett.106 126803

[4] HsiehD et al 2009Nature 460 1101–5[5] Zhang T et al 2009Phys. Rev. Lett. 103 266803[6] KongD et al 2011Nat. Nanotechnol. 6 705–9[7] Chen Y L et al 2009 Science 325 178–81[8] BerntsenMH,GötbergO,Wojek BMandTjernbergO2013

Phys. Rev.B 88 195132[9] Zhang T, LevyN,Ha J, KukY and Stroscio J A 2013Phys. Rev.

B 87 115410

6

2DMater. 2 (2015) 024012 CKastl et al

[10] YoshimiR,TsukazakiA,KikutakeK,Checkelsky JG,TakahashiKS,KawasakiMandTokuraY2014Nat.Mater.13253

[11] BeidenkopfH, Roushan P, Seo J, Gorman L,Drozdov I,Hor Y S, Cava R J andYazdani A 2011Nat. Phys. 7 939–43

[12] BeniaHM, LinC,KernK andAst CR2011Phys. Rev. Lett. 107177602

[13] BandoH,KoizumiK,OikawaY,DaikoharaK,Kulbachinskii VA andOzakiH 2000 J. Phys.: Condens.Matter12 5607–16

[14] KongD et al 2011ACSNano 5 4698–703[15] BahramyMS et al 2012Nat. Commun. 3 1159[16] CoutoN JG, Sacépé B andMorpurgoA F 2011Phys. Rev. Lett.

107 225501[17] WharamDA, ThorntonT J,Newbury R, PepperM, AhmedH,

Frost J E F,HaskoDG, PeacockDC, RitchieDA andJones GAC1988 J. Phys. C: Solid State Phys. 21 209–14

[18] vanWees B J, Kouwenhoven LP, van derMarel D,vonHoutenH andBeenakker CW J 1988Phys. Rev. Lett. 60848–50

[19] Hof K-D, Rössler C,Manus S, Kotthaus J P,Holleitner AW,SchuhD andWegscheiderW2008Phys. Rev.B 78 115325

[20] Rössler C,HofK-D,Manus S, Ludwig S, Kotthaus J P, Simon J,Holleitner AW, SchuhD andWegscheiderW2008Appl. Phys.Lett. 93 071107

[21] Kastl C,GuanT,HeXY,WuKH, Li YQ andHolleitner AW2012Appl. Phys. Lett. 101 251110

[22] Song J CWand Levitov L S 2014Phys. Rev.B 90 075415[23] ZhangG,QinH,Chen J, HeX, Lu L, Li Y andWuK2011Adv.

Funct.Mater. 21 2351–5[24] Ping J, Yudhistira I, RamakrishnanN, Cho S, AdamS and

FuhrerMS 2014Phys. Rev. Lett. 113 047206[25] HeX,GuanT,WangX, Feng B, Cheng P, Chen L, Li YQ and

WuK2012Appl. Phys. Lett. 101 123111[26] Chen J et al 2010Phys. Rev. Lett. 105 176602[27] Gehring P, Gao BF, BurghardMandKernK 2012Nano Lett.

12 5137–42[28] Chen J,HeX,WuK, Ji Z, Lu L, Shi J, Smet J and Li YQ2011

Phys. Rev.B 83 241304[29] Plucinski L et al 2013 J. Appl. Phys. 113 053706[30] Park J, AhnYHandRuiz-Vargas C 2009Nano Lett. 9 1742–6[31] Prechtel L, Song L,Holleitner AW, SchuhD, Pulickel A and

WegscheiderW2012Nat. Commun. 3 646

[32] McIver JW,HsiehD, SteinbergH, Jarillo-Herrero P andGedikN. 2012Nat. Nanotechnol. 7 96–100

[33] XuX,GaborNM,Alden J S, van der ZandeAMandMcEuen P L 2010Nano Lett. 10 562–6

[34] Cheng R, ZhouD, LiH,WangC, YinA, Jiang S, Liu Y, Chen Y,Huang Y andDuanX2014Nano Lett. 14 5590–7

[35] BuscemaM, BarkelidMandZwiller V 2013Nano Lett. 13358–63

[36] ErhardN,GolamSarwar ATM,Yang F,McCombDW,Myers RC andHolleitner AW2015Nano Lett. 15 332–8

[37] LemmeM,Koppens F, Falk A, RudnerMS,Hungkun P,Levitov L S andMarcusCM2011Nano Lett. 11 4134–7

[38] Crepaldi A, Ressel B, Cilento F, ZacchignaM,Grazioli C,BergerH, BugnonP, KernK,GrioniM and Parmigiani F 2012Phys. Rev.B 86 205133

[39] Kastl C, Karnetzky C, KarlH andHolleitner AW2015Nat.Commun. 6 6617

[40] Olbrich P et al 2014Phys. Rev. Lett. 113 096601[41] Irmer B, Blick RH, Simmel F,GödelW, LorenzH and

Kotthaus J P 1998Appl. Phys. Lett. 73 2051[42] HeydeM,RademannK, Cappella B, GeussM, SturmH,

Spangenberg T andNiehusH2001Rev. Sci. Instrum. 72 136[43] Hong S S, KundhikanjanaW,Cha J J, Lai K, KongD,

Meister S, KellyMA, Shen ZX andCui Y 2010Nano Lett. 103118–22

[44] Sung JH,HeoH,Hwang I, LimM, LeeD, KangK,ChoiHC,Park JH, Jhi SH and JoMH2014Nano Lett. 14 4030–5

[45] AdamS,Hwang E and Sarma SD2012Phys. Rev.B 85 235413[46] Martin J, AkermanN,Ulbricht G, LohmannT, Smet JH,

vonKlitzing K andYacoby A 2007Nat. Phys. 4 144–8[47] Skinner B, ChenT and Shklovskii B I 2012Phys. Rev. Lett. 109

176801[48] KimD, Syers P, ButchNP, Paglione J and FuhrerMS 2014

Nano Lett. 14 1701–6[49] KawabeM,Kanzaki N,MasudaK andNamba S 1978Appl.

Opt. 17 2556–61[50] Pang SW, LincolnGA,Mcclelland RW,Degraff PD,

GeisMWandPiacentiniW J 1983 J. Vac. Sci. Technol.B 11334–7

[51] Zhang F, KaneCL andMele E J 2012Phys. Rev.B 86 081303[52] CutlerM andMottNF 1969Phys. Rev. 181 1336–40[53] Zhang J et al 2011Nat. Commun. 2 574

7

2DMater. 2 (2015) 024012 CKastl et al