A surface state of topological insulators (TIs) is of great importance as it provides a spin-momentum-locked Dirac dispersion which can be exploited for the spintronic applications in forms of p-n junction1, topological-insulator-superconductor heterostructure2, and even spin-dependent thermometric device3. Also, it was recently demonstrated that the spin injection from the topological insulator enables the magnetization switching in the adjacent magnetic layer through the spin-orbit torque4,5. However, those application exploiting peculiar topological surface states are often hindered by contributions from the bulk free carriers which mask or overwhelm the charge transport of surface carriers. Even for the sample having little bulk carriers, it is not straightforward to design the thin-film devices including a proper choice of the adjacent capping or substrate layer, for example, because of an additional charge accumulation or depletion at the TI surfaces which sometimes occurs in unexpected and uncontrollable ways6,7.

Several surface sensitive techniques have been employed to investigate TI surface states. In particular, its evolution as a function of time after the sample cleavage has been revealed by angle-resolved photoemission spectroscopy8,9, scanning tunneling spectroscopy10, Hall measurement11, and so on. Terahertz (THz) emission spectroscopy also has proven to be another important surface-sensitive technique12. THz light can be generated near the sample surface upon the illumination of laser pulses, and Luo et al. demonstrated that THz emission amplitude observed for doped Bi2Se3 depends strongly on the charge carrier density which they attributed to the contribution of Dirac fermions together with bulk carriers13. Also, Zhu et al. examined the time-dependent changes of the THz emission responses from Bi2Se3 after the sample cleavage, and demonstrated that the THz emission amplitude is changed sensitively upon the time-dependent evolution of the surface space-charge region14.

In this work, we investigate the evolution of the surface electronic state of the Bi1.5Sb0.5Te1.7Se1.3 (BSTS), which is one of the most bulk insulating topological insulators15, by exploiting THz emission technique. We cleaved the sample in various environmental conditions at room temperature, and found strong variations in the amplitude and phase of emitted THz wave as a function of time after the sample cleavage. In particular, when the sample is cleaved in an ambient air, THz emission responses exhibit non-monotonic time-dependent variations with a clear minimum at a few hours after the cleavage which we attribute to the signature of the possible flat band bending. Interestingly, the surface state evolution of the BSTS sample exhibits a remarkable controllability by adjusting the atmospheric condition; we could accelerate, decelerate, and even stop the evolution of the surface electronic states. We discuss the maturation process of the surface electronic state in the ambient air in terms of the Se vacancy formation, chemical adsorption and reaction of O2 and H2O molecules.

Results and Discussion

THz emission mechanism

We investigate the surface electronic state by examining the amplitude and phase of THz wave emitted from the sample upon the radiation of near-infrared laser pulse. A schematic of an optical configuration near the sample is shown in Fig. 1(a). Among several THz generation mechanisms, we consider the optical rectification and the laser-induced surge current as a primary source for BSTS16. Particularly in the latter case, the emitted THz wave can provide important information about the charge dynamics as well as about the surface band bending. Laser-induced surge current arises in terms of two representative mechanisms, i.e., photo-Dember effect and the band-bending-induced charge acceleration, as depicted in Fig. 1(b,c), respectively. In the case of photo-Dember effect, a transient dipole is induced by a diffusion of photo-induced carriers, and hence the polarity of THz wave is determined by a mobility difference between electron and hole17. In the case of band-bending effect, on the other hand, the transient dipole is induced by the acceleration of electron and hole along the opposite directions by the built-in electric field in the space charge region, and hence the polarity of THz wave depends on the details of the band bending, i.e., whether it is upward or downward as depicted in Fig. 1(c)18. Therefore, by investigating the amplitude and phase of emitted THz wave, we can obtain information about the electrodynamics (particularly about the charge carrier mobility) or about the band bending in the space charge region17,18.

Figure 1
figure 1

(a) Schematic diagram showing THz emission from the sample surface upon the illumination of near-infrared laser pulses. Incident laser and outgoing THz light are p-polarized and propagate along 45 degrees away from the surface normal direction. The blue hemi-cylinder means the cleaved layer of the BSTS crystal. (b,c) Cartoons describing two THz generation mechanisms, i.e., photo-Dember effect and band bending-induced charge acceleration, respectively. In both cases, thick arrow indicates a transient electric dipole moment formed by displacement of photo-induced electrons and holes, and its direction determines the polarity of the emitted THz wave.

Evolution of emitted THz waves after the sample cleavage

Figure 2(a) shows time(t)-domain electric-field profiles of THz waves ETHz(t) emitted from BSTS obtained after the cleavage of the sample in an ambient air (Case I) with an oxygen content of 20.5% and a relative humidity of about 33%. Each measurement of ETHz(t) took about two minutes, and it has been monitored as a function of time after the cleavage (tcl) up to about 14 hours. Just after the cleavage, i.e., tcl ~ 0.04 hours (red one in the bottom), the wavefront of the THz wave shows a negative sign of the electric field in a temporal range between 4.5 and 7.8 ps. As time goes on after the cleavage, i.e., with an increase of tcl, this feature becomes weaker, and almost smeared out at tcl = 0.46 hours (thick black line in the bottom side). Then ETHz(t) develops to have an another distinct waveform (blue line) at tcl = 1.17 hours which looks different from that of the initial THz wave. After that, the waveform is restored to the initial one, and then has a deep valley at the wavefront in the saturation limit, e.g., at tcl = 13.7 hours. It should be noted that ETHz(t) displays an isotropic azimuth dependence at tcl > 13.7 hours (as demonstrated in Supplementary information), and hence we exclude the optical rectification from the candidates of possible THz generation mechanisms. Instead, we consider the surge current as a dominant THz emission mechanism at tcl > 13.7 hours as well as tcl ~ 0 hours when ETHz(t) has a similar waveform. Therefore, the time-dependent evolution of the THz waveform can provide us with valuable information of the photocarrier dynamics near the sample surface.

Figure 2
figure 2

(a) Time(t)-domain electric-field profiles of emitted THz waves ETHz(t) from Bi1.5Sb0.5Te1.7Se1.3 displayed in the chronological order after its cleavage in air (Case I). Five ETHz(t) are highlighted, and they are obtained at tcl = 0.04, 0.46, 1.17, 1.88, and 2.88 hours after the cleavage of the sample in an ambient air. (b) Integrated ETHz(t) between t = 4.5 ps and 7.8 ps given as a function of tcl. (c) Integrated THz wave spectrum THz(ω) between ω/2π = 0.04 THz and 0.79 THz plotted as a function of tcl. Note that both values in (b,c) are normalized by the values at tcl = 9 hours. Inset shows three representative THz(ω) obtained from the Fourier transform of ETHz(t).

We examine an evolution of the emitted THz wave in more detail by integrating the THz electric field ETHz(t). When the THz wave is generated by the surge current mechanism, the THz electric field should be proportional to the time-derivative of the surge current J(t), i.e., 19. Therefore, the temporal integration of ETHz(t) can provide information about the transient photocurrent along the surface normal direction19,20. To examine the initial charge dynamics, in particular, we integrate the wavefront side, i.e., ETHz(t) in the temporal range between 4.5 ps and 7.8 ps. This temporal weight TW of ETHz(t) is normalized with the value at tcl = 9 hours, and plotted as a function of tcl in Fig. 2(b). TW has a negative value just after the cleavage which corresponds to the dip structure in ETHz(t). As time goes on, TW approaches a zero level, and crosses over it to have a small but positive value around 1.17 hours. Then it returns back to have a negative value, and is finally saturated from about 6 hours after the cleavage.

We can find a similar behavior of the THz emission response from the spectral weight SW analysis. Electric field amplitude of the THz wave in a frequency domain, i.e., THz(ω) is obtained by the Fourier transform of ETHz(t), and the inset of Fig. 2(c) displays THz(ω) at tcl = 0.04, 0.46 and 1.17 hours. We integrate these spectra from 0.38 THz to 0.79 THz, and display the SW values as a function of tcl in Fig. 2(c). The time(tcl)-dependent SW shows a similar behavior with TW; as time goes on, SW decreases first, and it increases until it saturates at about tcl = 6 hours. It is interesting to note that it has a local maximum around tcl = 1.17 hours between two minima. Actually, these minimum SWs are obtained when TW crosses the zero level before and after having the positive maximum at tcl = 1.17 hours. These clearly demonstrate that the THz waveform experiences a symmetric time-dependent changes with respect to tcl = 1.17 hours, and there exist two competing contributions to the observed THz emission signal as a function of tcl.

Possible flat band-bending at the BSTS surface

It is important to point out that the existence of two different contributions to the THz signal can be clearly resolved by closely examining THz waveforms in the intermediate states. Just after the cleavage and also at the saturation stage, ETHz(t) has a well-defined single-cycle waveform. On the other hand, at tcl = 1.17 hours when TW and SW have a global and local maximum, respectively, we suppose that ETHz(t) has another distinct form with a one-and-half cycle oscillation. Interestingly, the THz waveform in between, i.e., at 0.04 < tcl < 1.17 hours, is not well characterized by any of these two waveforms. Instead, it looks that both contributions coexist. For example, ETHz(t) at tcl = 0.46 and 1.88 hours has a dip at the wavefront and additional double peaks following it which resemble the characteristic features of ETHz(t) at tcl = 0.04 and tcl = 1.17 hours, respectively. Furthermore, whereas THz(ω) at tcl = 0.04 and tcl = 1.17 hours have a single maximum (inset of Fig. 2(c)), THz(ω) at tcl = 0.46 hours has a dip around 0.3 THz between two maxima which is a typical spectral feature appearing when there are two consecutive pulses with a finite time interval. Actually, experimental results of ETHz(t) can be reproduced quite well by a linear combination of aforementioned distinct waveforms as . The fitting curves based on this equation are plotted as solid lines on top of the experimental data in Fig. 3(a). Obtained coefficients S1 and S2 are displayed as a function of tcl in Fig. 3(b). These values take completely opposite time-dependent changes at tcl < 3 hours, and then only S1 keeps increasing until it becomes saturated at about tcl = 6 hours. These results demonstrate that there exist two different THz emission mechanisms, and they contribute to the THz response differently upon the evolution of the surface state as a function of time.

Figure 3
figure 3

(a,b) Fitting analysis of ETHz(t) with a linear combination of two waves at tcl = 0.04 hours and 1.17 hours. Fitting curves of ETHz(t) are shown with symbols in (a). The weighting factors S1 and S2 for each contribution are plotted in (b). (c,d) Cartoons describing interaction between Bi1.5Sb0.5Te1.7Se1.3 and atmospheric molecules, i.e., H2O and O2. Just after the sample cleavage (c), O2 and H2O molecules are adsorbed at the surface, and they act as electron acceptors. Then the chemical reaction occurs between the sample and the molecules (d), and it leads to the n-type doping in the sample surface.

As a dominant THz emission mechanism in the saturation stage, we already excluded the optical rectification, and now consider the charge acceleration due to the downward band bending because of following reasons. When BSTS is exposed to the air, it is well accepted that the final surface state has the downward band bending9,11. Since the downward band bending is largely attributed to the extrinsic contributions, such as a selenium vacancy, chemical adsorption and reaction7,8,9,21, it is expected to have a formation of a larger number of charged scattering centers which will reduce the contribution of photo-Dember effect. Therefore, the increasing tendency of the THz emission amplitude before its saturation should be attributed to the larger contribution of the downward band bending. This leads to the assignment of the THz emission mechanism for the contribution S1 to be the charge acceleration due to the downward band bending. Therefore, the fact that the contribution of S1 becomes almost disappeared at about tcl = 1.17 hours possibly indicates that the flat band bending can be achieved in the intermediate state during the maturation of the surface state in the atmospheric condition.

Distinct surface-state evolution depending on environmental conditions

By monitoring time-dependent THz emission responses after the sample cleavage in the ambient air, we have demonstrated that the surface states of BSTS are strongly affected by the atmospheric conditions, and there can be a surface state condition close to the flat band bending during the maturation process. Now, we present distinct time-evolutions of the THz emission responses in different environmental conditions which reveal the effect of ambient gases on the surface electronic structure in more detail. As a Case II, we cleaved the sample in the N2 atmosphere with the oxygen content less than 0.3% and RH of about 2.0%, and kept the sample in this environment for about 7.5 hours before changing the condition to the ambient air environment as in Case I. After the cleavage in N2, there is a gradual variation of THz emission signal (Fig. 4(a,d)). Such tiny change can be further minimized by cleaving the sample in the dry air. As a Case III, we cleaved the sample and kept it for about 4.7 hours in dry air with RH at a zero level, and then changed the environmental condition to the ambient air. As shown in Fig. 4(b,e), THz emission signals exhibit essentially no time-dependent change when the sample is cleaved in the dry air.

Figure 4
figure 4

THz wave emission responses of BSTS after its cleaving in N2 atmosphere (Case II), in dry air (Case III), and in the ambient/dry air (Case IV).

In Cases II and III, the sample is exposed to the ambient air after keeping the sample in the initial condition for several hours. In Case IV, the condition is consecutively switched between humid air with a relative humidity of 23% and dry air with the humidity less than 6%. Results obtained in the initial conditions are displayed with red and blue colors for the Case II and Case III, respectively, and the results obtained in the ambient air are displayed in black. Electric-field profiles of emitted THz waves ETHz(t) are shown in (ac). Integrated ETHz(t) is plotted as a function of the elapsed time after the cleavage in (df). Solid lines in (d,e) are fitting curves using two exponential functions. The time-dependent relative humidity (RH) is displayed in (f) where a solid line is reproduced from the measured values shown with a symbol.

In both Cases of II and III, as the conditions are changed to the ambient air containing a relative humidity of about 33%, dramatic changes in ETHz(t) occur, and their time-dependences are quite similar with the result for the Case I; all of them have lambda-like shapes of TW(tcl). In particular, the distinct results for Case III obtained in dry and ambient air clearly demonstrates the important role of H2O molecules in the evolution of the surface electronic state after the sample cleavage. Although overall behaviors in the ambient air appear similar for all the Cases, it should be noted that the time scales of such variations are given distinctly. We fit time profiles of TW(tcl) with two exponential functions as , where Coffset is a constant offset level, τp and τn are relaxation times of the exponential function with positive and negative slopes, respectively, and Cp and Cn are corresponding weighting parameters. The fitting curves are shown with solid lines in each TW(tcl) plot, and the decay time (τp, τn) used for the fitting in Cases I, II, and III are given as (1.0, 1.2), (0.3, 0.5), and (1.8, 2.0), respectively, in the unit of hours. It should be noted that this process occurs the fastest for the Case II and the slowest for the Case III.

Possibility to prolong the period of flat band-bending condition

Whereas we found a clear signature of the flat band bending during the maturation process of the surface state in the ambient air, we examined whether such an intriguing condition can be maintained by further controlling environmental conditions. Figure 4(c) displays time-dependent evolutions of THz emission responses of BSTS under the consecutive switches of environmental conditions between the ambient air containing RH of 23% and the dry air with RH level less than 6%. When the sample is in the humid air condition, the THz emission responses (indicated with black color) show strong time-dependent changes, and overall behaviors appear quite similar with the results demonstrated for the Cases I–III. This includes the global maximum in the time-dependent envelop of the integrated temporal weight TW observed at tcl ~ 1.4 hours after the cleavage, and it corresponds possibly to the flat band-bending condition. When the condition is changed into the dry air, however, time-dependent change appears quite differently; as indicated with blue color, the time-dependent variations can be almost stopped before and after the flat band-bending condition. (The increase of the THz emission amplitude at the saturation stage at tcl ~ 4.2 hours is attributed to the less atmospheric absorption of THz wave in the dry air condition). This clearly demonstrates that we can prolong the period of the flat band-bending condition before reaching the irreversible saturation stage by reducing the contributions of H2O molecules in the air.

Chemical reactions and surface state evolution

From now on, we discuss a possible mechanism of the surface state evolution related to the band bending after the cleavage of the BSTS crystal. For topological insulators, it is usually not straightforward to determine the direction of the band bending as the surface state of topological insulators is very dispersive, and hence the Fermi level pinning to the high-density surface state is not simply applicable22,23. In this work, however, we address this issue by examining the THz emission responses in detail. As discussed in the above, the THz emission at the saturation state is attributed to the surge current driven by the downward band bending. Since ETHz(t) obtained just after the sample cleavage looks essentially the same with that at the saturation state (Fig. 2(a)), it is reasonable to consider that the downward bend banding occurs also just after the sample cleavage. Actually, this interpretation is supported by the comparison with the result for InAs. The observed THz waveform (Fig. 2(a)) shows an opposite phase to that of InAs (Supplementary information) where the THz emission is attributed to the photo-Dember effect. As the electron mobility is larger than the hole for InAs, the corresponding THz waveform is expected to have the same phase with the THz wave generated by the upward band bending (See Fig. 1(b,c)). Therefore, the opposite phase for BSTS with respect to that for InAs also suggests the downward band bending just after the sample cleavage. ETHz(t) for Case II just after the cleavage exhibits a waveform different from the other cases, but it resembles the waveform for Case I with a minimal downward band bending. Hence, this implies that the downward band bending would be much weakened in this case. Nevertheless, ETHz(t) after a few hours recovers the waveform corresponding to the downward band bending. This is attributed to the formation of the negatively charged Se vacancies VSe near the surface after the cleavage even in the N2 atmosphere21,24, and the accumulation of more electrons at the surface. This assures again our interpretation of the initial state (except for the Case II) to have the downward band bending.

When the sample is cleaved in the ambient air, the fresh surface state will be strongly affected also by an adsorption of the environmental ions, such as N2, O2, CO2, CO, and H2O, and even their chemical reactions with the sample7,8,25,26,27,28. Among these molecules, we take account of the contributions of O2 and H2O mainly, but discard the effects of other molecules by considering the inert nature of N2 and tiny contents of CO2 and CO (about 400 ppm and 0.01 ppm, respectively) in the air (See Supplementary information for the details). Since O2 has a strong electron affinity, they are adsorbed to the TI surface to attract excess electrons accumulated at the surface (Fig. 3(c))7,24. When the sample is cleaved in the dry air condition (Case III), the THz emission response shows almost no time-dependence which is attributed to the balance between the surface charge accumulation originating from the VSe and the electron depletion by O2 molecules. Such tendency to reduce the electron accumulation can be further enforced by H2O which has a large electron affinity and electric dipole moment (see Supplementary information)21. After the cleavage in the ambient air, hence the adsorption of both O2 and H2O occurs, and it tends to neutralize the surface charge state which was originally electron-accumulated. During this process, the number of charged particles near the surface is reduced, and accordingly the built-in electric field becomes smaller as indicated in the right bottom surface of Fig. 3(c). Therefore, ETHz(t) generated from the built-in field should be reduced. These behaviors could explain the history of ETHz(t) in the initial stage after the cleavage in the ambient air, i.e., tcl < 1.17 hours for Case I. With a further elapse of time, actual chemical reactions can occur between BSTS and molecules in air. For example, H2O is dissociated into H and OH in the existence of O2 gas, and accelerates the formation of VSe by forming a H2Se gas and also by leaving 2Bi(OH)3 in the sample as depicted in Fig. 3(d)8,21. This process contributes to the more charge accumulation near the surface which is accompanied by the enhancement of the band bending as indicated in the right bottom surface of Fig. 3(d). Therefore, ETHz(t) generated from the built-in field should be enhanced, and this explains the history of ETHz(t) after tcl = 1.17 hours for Case I. It is worth to emphasize that the driving mechanisms for the charge accumulation and neutralization are competing with each other in BSTS, and there can be a moment to have an intermediate state with a negligible band bending after the sample cleavage.

As demonstrated for Cases I–III, the surface state evolution in the ambient air occurs with different time scales depending on the treatments done before switching the condition into the ambient air. To account for the difference in the decay time during the maturation process, we consider the effective amount of VSe at the moment when the atmospheric condition is changed into the air. For the case II, VSe keeps being formed in the N2 atmosphere, and hence there is an enough chance of the interaction between VSe and ambient molecular components which results in a fast adsorption of the molecules and accordingly the fast relaxation process. For the case III, on the other hand, the naturally formed VSe is already compensated by the adsorption of O2 in a dry air condition. Therefore, it takes a longer time for the system to be relaxed into the final state even with an additional contribution of H2O molecules. It is worth to note that the surface electronic state evolution of BSTS for all three cases investigated shows distinct behaviors when it is compared to that of Bi2Se3 which exhibits faster and monotonic variations of the THz emission14 and optical second harmonic generation responses24 after the sample cleavage. As one of possible reasons of these differences, we consider the suppression of the VSe formation for BSTS where the Se ions are trapped at the center of the quintuple layer with a strong chemical bonding with Bi ions29. This leads to a relatively slow progress of the surface state stabilization, and hence provides a chance to trace such evolutions in more detail.

In summary, we investigated a time-dependent evolution of the surface state of the Bi1.5Sb0.5Te1.7Se1.3 single crystal after the sample cleavage by examining THz light emitted from the sample surface upon the illumination of the near-infrared femtosecond laser pulses. We observed strong variations of THz emission responses as a function of time after the sample cleavage which strongly depend on environmental conditions. In particular, we observed a clear signature of the flat band bending condition of the surface electronic structure in the course of the interaction between Bi1.5Sb0.5Te1.7Se1.3 and environmental molecules. Furthermore, we could demonstrate that the evolution of the surface electronic state can be accelerated, decelerated, and even stopped by controlling the atmospheric conditions. This assures that the surface decoration can work as an efficient way to tune the Fermi level of Bi1.5Sb0.5Te1.7Se1.3 with respect to the surface Dirac state. Since such intriguing controllability could be achieved in the atmospheric condition at room temperature, it will provide a novel chance to exploit the topological surface state for the further practical researches, such as the spin photocurrent generation, with a minor contribution from the residual bulk carriers in the bulk-insulating Bi1.5Sb0.5Te1.7Se1.3.


A high-quality single crystal of Bi1.5Sb0.5Te1.7Se1.3 (BSTS) was grown using the self-flux method with stoichiometric chunks, and it has an optically flat surface along the (111) direction with a lateral size of about 2 mm in diameter. Terahertz (THz) emission experiments were done by using a conventional THz time-domain spectroscopy setup. Laser pulse has a central wavelength of 800 nm and pulse width of about 100 fs with a repetition rate of 80 MHz, and is incident onto the sample with an average power of 230 mW at the incidence angle of 45°. As the laser beam size is large with a full width at half maximum of about 1.2 mm, the emitted THz wave follows a specular reflection of the laser beam20. Hence we collect the THz beam along the angle of 45° which is guided by a pair of parabolic mirrors and detected by using a photo-conductive antenna.

The O2 content and the relative humidity (RH) are controlled by flushing the N2 gas and the dry air, and they are monitored in real time by using oxygen analyzer (Greisinger) and hygrometer (Center). The ambient air conditions for Cases I, II, and III contain O2 molecules of 20.5% and RH of 33%. The O2 content and RH in the N2 atmosphere for Case II are 0.3% and 2.0%, respectively, and the corresponding values in the dry air for Case III are 20.5% and 0.0%, respectively. Note that the maximum fluctuation of RH is about ±2.5% of each measured value. In Case IV, the condition is consecutively switched between a humid air with RH 23% and dry air with RH less than 6% whereas the O2 content is maintained at 20.5%. Upon flushing the dry air, RH drops down to 6% in less than two minutes when the first measurement is launched. The other molecules, such as CO2 and CO, are not controlled intentionally, and expected to have usual volume fractions of about 400 ppm and 0.1 ppm, respectively, in the ambient air.

Additional Information

How to cite this article: Park, S.H. et al. Possible flat band bending of the Bi1.5Sb0.5Te1.7Se1.3 crystal cleaved in an ambient air probed by terahertz emission spectroscopy. Sci. Rep. 6, 36343; doi: 10.1038/srep36343 (2016).

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.