Fermi Level Manipulation through Native Doping in the Topological

Jun 6, 2018 - Fermi Level Manipulation through Native Doping in the Topological ... by controlling the Se vacancy concentration using post-growth anne...
0 downloads 0 Views 2MB Size
www.acsnano.org

Fermi Level Manipulation through Native Doping in the Topological Insulator Bi2Se3

Lee A. Walsh,†,‡ Avery J. Green,§ Rafik Addou,† Westly Nolting,§ Christopher R. Cormier,† Adam T. Barton,† Tyler R. Mowll,§ Ruoyu Yue,† Ning Lu,† Jiyoung Kim,† Moon J. Kim,† Vincent P. LaBella,§ Carl A. Ventrice, Jr.,§ Stephen McDonnell,∥ William G. Vandenberghe,† Robert M. Wallace,† Alain Diebold,§ and Christopher L. Hinkle*,† Downloaded via UNIV OF TOLEDO on June 30, 2018 at 10:45:54 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.



Department of Materials Science and Engineering, University of Texas at Dallas, Richardson, Texas 75080, United States Tyndall National Institute, University College Cork, Lee Maltings Complex, Cork T12R5CP, Ireland § Colleges of Nanoscale Science and Engineering, SUNY Polytechnic Institute, Albany, New York 12203, United States ∥ Department of Materials Science and Engineering, University of Virginia, Charlottesville, Virginia 22904, United States ‡

S Supporting Information *

ABSTRACT: The topologically protected surface states of three-dimensional (3D) topological insulators have the potential to be transformative for high-performance logic and memory devices by exploiting their specific properties such as spin-polarized current transport and defect tolerance due to suppressed backscattering. However, topological insulator based devices have been underwhelming to date primarily due to the presence of parasitic issues. An important example is the challenge of suppressing bulk conduction in Bi2Se3 and achieving Fermi levels (EF) that reside in between the bulk valence and conduction bands so that the topologically protected surface states dominate the transport. The overwhelming majority of the Bi2Se3 studies in the literature report strongly n-type materials with EF in the bulk conduction band due to the presence of a high concentration of selenium vacancies. In contrast, here we report the growth of near-intrinsic Bi2Se3 with a minimal Se vacancy concentration providing a Fermi level near midgap with no extrinsic counter-doping required. We also demonstrate the crucial ability to tune EF from below midgap into the upper half of the gap near the conduction band edge by controlling the Se vacancy concentration using post-growth anneals. Additionally, we demonstrate the ability to maintain this Fermi level control following the careful, low-temperature removal of a protective Se cap, which allows samples to be transported in air for device fabrication. Thus, we provide detailed guidance for EF control that will finally enable researchers to fabricate high-performance devices that take advantage of transport through the topologically protected surface states of Bi2Se3. KEYWORDS: topological insulator, Fermi level, doping, molecular beam epitaxy, bismuth selenide

T

Bi2Se3 is one of the most widely studied 3D TIs, with a range of reports on its behavior and epitaxial growth.5−8 One of the primary challenges with using Bi2Se3 is the widely reported, severe n-type doping as determined by angle-resolved photoemission spectroscopy (ARPES),6,9−11 where the measured Fermi level (EF) resides in the bulk Bi2Se3 conduction band (as opposed to within the bulk band gap), prohibiting the exploitation of the spin-momentum-locked, topologically protected surface states. This Fermi level position is

opological insulators (TIs) are a class of material that, due to the “protection” by time reversal symmetry of their surface (3D TIs) or edge (2D TIs) states, have the potential to enable incredibly energy efficient and defecttolerant logic and memory devices.1,2 These surface/edge states exhibit a number of interesting properties including spinmomentum locking, where the spin of an electron is perpendicularly locked to its transport direction, resulting in spin-polarized transport on the surface/edge and the significant suppression of backscattering by impurities and defects. These properties make TIs attractive for a variety of nanoelectronic applications, including spin-transfer torque nonvolatile memory3 and defect-tolerant, ballistic field-effect transistors.4 © 2018 American Chemical Society

Received: May 6, 2018 Accepted: June 6, 2018 Published: June 6, 2018 6310

DOI: 10.1021/acsnano.8b03414 ACS Nano 2018, 12, 6310−6318

Article

Cite This: ACS Nano 2018, 12, 6310−6318

Article

ACS Nano consistently attributed to the presence of a large concentration of Se vacancies that form during the growth process, causing ntype doping.12,13 There have also been reports that clearly show an excess of bulk conduction in Bi2Se3 using electrical measurements,3,14 which seems to confirm the ARPESmeasured EF position and the attribution of Se vacancy formation during the growth process. However, a closer look, with the hindsight knowledge to be reported later in this paper, reveals that those bulk conduction measurements were performed on Bi2Se3 that utilized a thermal treatment postgrowth. This thermal treatment was performed to desorb an intentionally deposited Se capping layer, used to protect the sample surface during transfer between the growth system and device fabrication.3,14 The Se cap is then desorbed at temperatures between 240 and 270 °C for 1 h,3,14−16 which as we will show, is high enough to induce additional Se vacancies in the Bi2Se3 once the amorphous Se cap has been removed. Importantly though, some of the papers that observe EF in the conduction band (from ARPES) simultaneously observe Fermi levels within the band gap in Shubnikov−de Haas measurements.11,17 An excellent recent study by Frantzeskakis et al. demonstrated that illumination-induced band bending contributes significantly to the observed Fermi level position in ARPES measurements.18 Performing a study of the EF position as a function of illumination time, it was shown that EF shifts in excess of 150 meV (toward the conduction band) were observed after a 1 h long illumination. These shifts were in addition to the more commonly observed doping due to surface decoration related “aging” effects19,20 and were attributed to the photodissociation and ionization of the adsorbed species. These results indicate that it is possible that other groups have unknowingly grown near-intrinsically doped Bi2Se3, and the use of ARPES and/or excessive Se decap temperatures have masked this possibility. Here we demonstrate the growth of near-intrinsic Bi2Se3 films by molecular beam epitaxy (MBE) and the crucial ability to control and manipulate the Se vacancy concentration, and thus the EF position, by post-growth annealing. This EF control is vitally important for fabricating devices that utilize the topologically protected surface states. Additionally, through a detailed study, we identify the optimum temperature that successfully removes the commonly used Se protective capping layer (used for transporting Bi 2Se3 in air) while also maintaining the midgap EF. These findings will finally enable researchers to develop and fabricate high-performance devices that take advantage of transport through the topologically protected surface states of Bi2Se3.

Figure 1. Schematic diagrams of the quintuple layer structure of (a) stoichiometric Bi2Se3 and (b) Bi2Se3 with Se vacancies. (c) Averaged STS spectrum of the as-grown Bi2Se3 sample, with the Fermi level 14.0 mV below the Dirac point, indicating a nearintrinsic film with no doping. (d) Averaged STS spectrum of the Bi2Se3 sample following a 30 min anneal at 220 °C, with the Fermi level 70.0 mV above the Dirac point, indicating n-type doping due to the formation of Se vacancies caused by the anneal. (e) Histogram showing the spread of EF positions from STS spectra taken at different spots on the as-grown sample showing very little variation across the surface of the undoped film. (f) XPS VB spectra confirming the EF is slightly below midgap for the MBE as-grown Bi2Se3. Also shown for comparison is a typical VB measurement for purchased chemical vapor transport (CVT)-grown material, which has EF at the bulk conduction band edge. (g) Scatter plot showing the Fermi level shift as determined using shifts in the XPS spectra and the STS Dirac point position as a function of anneal temperature, indicating increased n-type doping with increasing anneal temperature. The XPS shifts are the average shifts of the Bi 4f and VB spectra. The error bars shown for the XPS measurements are due to the error in determining the peak position for the Bi 4f and the band offset for the VB. Each STS data point is the average of 30 measurements, and the error bars show the statistical spread.

RESULTS AND DISCUSSION STS Study of Air-Exposed Bi2Se3 Samples. The structures of stoichiometric Bi2Se3 and, alternatively, Bi2Se3 with Se vacancies (Bi2Se3‑η) are demonstrated schematically in Figure 1a and b, respectively. The initial Fermi level position of the as-grown Bi2Se3 film is determined from ex situ scanning tunneling spectroscopy (STS) and X-ray photoelectron spectroscopy (XPS) measurements. The STS spectrum for an asgrown Bi2Se3 sample is shown in Figure 1c. STS dI/dV is proportional to the local surface density-of-states (DOS), and the minimum of dI/dV reveals the position of the Dirac point of the Bi2Se3 surface states. The spectrum is averaged from at least 10 separate spectra acquired across a 10 × 10 nm2 area of the sample. The tunneling current levels of the as-grown Bi2Se3

sample were the lowest of all samples, likely due to surface contamination or the growth of a thin surface oxide layer (Figure S1), as this sample was exposed to ambient conditions during transfer to the ex situ STS instrument (we will discuss these effects in more detail later). In Figure 1c, the Dirac point of the material is centered at +14.0 mV, indicating that the EF (0 mV in the STS spectra) is 14.0 meV below the Dirac point, in between the bulk conduction and valence bands. This indicates that the film has a very low density of Se vacancies, similar to that expected for stoichiometric Bi2Se3. In contrast, 6311

DOI: 10.1021/acsnano.8b03414 ACS Nano 2018, 12, 6310−6318

Article

ACS Nano

Figure 2. XPS spectra of (a) the Bi 5d core level before and after decap showing the absence of any detectable Bi in the capped sample due to attenuation caused by the Se cap. After decap the Bi 5d doublet is present, indicating complete removal of the Se layer. The Se 3d core level shows the peak shift after decap, indicating the transition from Se−Se bonding in the as-capped sample to Bi−Se bonding after decap. (b) O 1s and C 1s core levels before and after decap showing no detectable oxygen in either sample and a small C peak in the as-capped sample, which disappears after decap. (c) TPD measurements of the Se cap desorption as a function of anneal temperature, showing the expected proportions of 78Se:80Se isotopes (1:2.1).28

Figure 1d shows the STS spectra for the same sample after a 30 min ultra-high-vacuum (UHV) anneal at 220 °C. The observed EF shift toward the conduction band is caused by an increase in the density of Se vacancies formed during the anneal (as shown schematically in Figure 1b), which act as electron donor n-type dopants and are the primary defects in Bi2Se3 (evaporation energy of Se is 95.5 kJ/mol, while that of Bi is 179.0 kJ/mol).21 In this paper, the Fermi level is hereafter referenced as EC − EF, i.e., the Fermi level position with respect to the bulk conduction band minimum (EC). A histogram of the measured Fermi level positions for asgrown Bi2Se3 films is shown in Figure 1e. STS spectral maps with resolutions that ranged between 5 and 25 spectra/μm were acquired to provide statistically significant numbers of spectra, from which the difference between the ED and EF could be measured and analyzed. The centroid of this histogram is at 18.0 ± 11.4 mV; that is, the Fermi level is 18.0 ± 11.4 meV below the Dirac point, taken from 94 spectra. The spread of EF positions indicates only very minor variations in the as-grown doping levels across the surface of the film. XPS valence band (VB) spectra are shown in Figure 1f and provide a second, independent, technique to confirm the EF position for as-grown Bi2Se3 films. For the MBE-grown material, the bulk VB is located at a binding energy of 130.0 ± 30.0 meV, i.e., 130 meV below the Fermi level (which is at 0 eV binding energy in XPS), indicating that EF is 202 meV below the conduction band edge for the 332 meV22 band gap Bi2Se3. The EF position determined from the combination of XPS and STS results confirms the near-intrinsic nature of the as-grown films. We can determine the Dirac point position within the band gap using the combined XPS and STS results on the same as-grown sample, finding it to be 184 meV below the conduction band minimum, below midgap as expected.22,23

Reference spectra of a purchased chemical vapor transport grown sample24 are also shown for comparison, with a VB offset of 300.0 ± 40.0 meV, showing that EF lies near the bulk conduction band edge for the purchased Bi2Se3 and indicating a highly n-doped film due to a large density of Se vacancies.23 Further structural and chemical characterization that confirms the high-quality crystalline nature of the MBE-grown Bi2Se3 can be found in the Supporting Information. Next, we systematically measured the MBE-grown Bi2Se3 sample following a succession of increasing vacuum anneal temperatures to quantify the change in the Fermi level position as a function of the anneal. Both XPS and STS measurements were performed in situ to monitor EF in conjunction with changes in the material chemistry after each successive anneal. Each anneal is performed for 30 min at the specified temperature, after which the sample is cooled to room temperature before the XPS and STS measurements are performed. No changes in the film stoichiometry are detected by XPS, indicating that the Se loss (∼1012/cm3 as we will show later) is below the level of XPS detection (∼0.1%). The chemical analysis can be found in the Supporting Information. Figure 1g shows the EF positions measured by both XPS and STS plotted relative to the conduction (EC) and valence band edges (EV). The EF positions for the sample as-loaded are found to be close to ED for both XPS (EC − EF = 202.0 ± 40.0 meV) and STS (EC − EF = 202.0 ± 11.4 meV) measurements, in agreement with the results in Figure 1c. Following each successively higher anneal temperature, a consistent shift in EF toward EC is observed. After annealing at 220 °C, the EF is positioned at 127.0 ± 40.0 meV and 114.0 ± 40.0 meV below the conduction band edge from the XPS and STS measurements, respectively. These are total EF shifts toward the conduction band of 107.0 meV (XPS) and 120.0 meV (STS) 6312

DOI: 10.1021/acsnano.8b03414 ACS Nano 2018, 12, 6310−6318

Article

ACS Nano

Figure 3. (a) Averaged STS spectrum of a sample with the Se cap removed through annealing at 170 °C. EF is at the Dirac point (EC − EF = 184.0 meV). (b) Scatter plot showing the shift in the XPS spectra and the STS Fermi level position as a function of anneal temperature for a decapped Bi2Se3 sample, indicating n-type doping of the material following higher anneal temperatures. The XPS shifts are the average of the shifts measured in the Bi 4f and VB spectra. The error bars shown for the XPS measurements are due to the error in determining the Bi 4f peak position and VB offset. Each STS data point is the average of 10 measurements, and the error bars show the range measured. (c) Bi 4f 7/2 core level from XPS showing a shift to higher BE as a function of anneal temperature, indicating an EF shift. (d) Te 3d core level spectra showing no detectable Te in the sample, ruling out Te counter-doping.

capping layer) and no detectable carbon after removal of the Se cap. The broad feature observed near the C 1s region is a Se Auger feature. The absence of C and O after decap shows that the Se cap successfully protects the Bi2Se3 surface. To understand the correct temperature to perform the decapping process, it is first necessary to determine the temperature at which the cap desorbs, which was investigated using temperature-programmed desorption (TPD) measurements.29 There are two significant Se desorption events that occur below 400 °C for Se-capped Bi2Se3. The first of these desorption events results from the sublimation of the Se capping layer, as is shown in Figure 2c, which is a measurement of the partial pressures of the 80Se and 78Se isotopes as a function of annealing temperature. A sliding thermocouple contact was used to measure the temperature of the sample during the TPD measurements so that samples could be mounted on the sample holder in UHV via the load lock. Because of the difference in the thermal contact resistance from sample to sample, there was some variance in the measured temperature of the Se peak. By plotting the previously published vapor pressure of Se vs temperature,30 it was determined that the onset of desorption of Se should occur at approximately 105 °C. Therefore, to calibrate the temperature curve of each sample, a linear offset was used to adjust the onset temperature of the first desorption peak to be 105 °C. The second peak occurs at ∼300 °C and is due to the decomposition of the Bi2Se3 film, as shown in Figure S5. Anneals that approach or exceed this temperature were found to damage and desorb the entire Bi2Se3 film. In conjunction with the TPD measurements, a range of decapping temperatures were studied using XPS to understand the temperature at which a clean Bi2Se3 surface was recovered. An anneal temperature of 170 °C for 30 min was found to

compared to the as-grown sample, indicating a significant increase in n-type doping. An additional anneal at 230 °C shows no further change in EF. STS Study of Bi2Se3 Samples Utilizing a Protective Se Cap to Prevent Air Exposure. Bi2Se3 films oxidize easily upon air exposure,25 resulting in a surface oxide that, although it does not eliminate the existence of the topologically protected surface states, is an undesired interlayer for device design and fabrication. As such, the deposition of an amorphous Se cap has been used to protect the surface from oxidation during ex situ transfer to other characterization instruments or device fabrication tools, as many groups have done in a variety of materials systems.8,26,27 A Se cap ∼40 nm thick is deposited in situ in the MBE growth chamber immediately after the growth of the Bi2Se3 film. The capping method is described in detail in the Methods section, and the crystallinity and cleanliness of the Bi2Se3 surface after removal of the Se cap are further discussed in the Supporting Information. As described previously, others remove this Se cap at too high of a temperature, degrading the film and resulting in n-type doping due to the formation of Se vacancies. To maintain the film quality and Fermi level, the removal of this Se cap, while maintaining the Fermi level position close to midgap, has been investigated. XPS spectra of the Bi 5d core level (Figure 2a) before decap show the absence of any Bi-related features indicating full coverage of the cap that results in the attenuation of the Bi core level. After decap, a sharp Bi doublet is observed. The Se 3d core level spectrum shows a shift after decap consistent with the difference between Se−Se and Bi2Se3 bonding. No Bi−O or Se−O oxide features are detected in either core level. The O 1s spectra in Figure 2b show no detectable signal in the capped or decapped sample, while the C 1s spectra show a small adventitious carbon peak in the capped sample (atop the Se 6313

DOI: 10.1021/acsnano.8b03414 ACS Nano 2018, 12, 6310−6318

Article

ACS Nano result in a Se−Se free surface, indicating complete removal of the Se cap and the optimal decap temperature, whereas for an anneal at 160 °C for 30 min, a residual Se−Se related feature can still be observed in the XPS spectra (Figure S6). As such, a decap temperature of 170 °C for 30 min was used to completely remove the Se cap and obtain a clean surface, while not perturbing the underlying Bi2Se3 film, which we then used as the initial data point for the EF measurements. With a controlled capping and decapping process now in place, we repeat the study tracking EF as a function of anneal temperature on samples that were protected with the Se cap during the ex situ transfer to the XPS and STM system, eliminating any convolution due to surface oxidation. A representative STS spectrum for those samples decapped at 170 °C is shown in Figure 3a. A slightly different STS spectrum is observed when compared to the partially oxidized sample in Figure 1c, which can be attributed to the absence of surface oxidation for the decapped sample. The spectrum in Figure 3a has a similar shape to those commonly reported in the literature.26,31 Figure 3b shows the scatter plot of the EC − EF position as a function of anneal temperature. The sample decapped at 170 °C shows an EF position close to the Dirac point from XPS and STS measurements, positioning EC − EF at 224.0 ± 40.0 meV and 184.0 ± 30.0 meV, respectively. Extrinsic “dopants” such as Ca32 or Te33 have been shown to counterdope the Se vacancies to obtain Fermi levels close to ED. To ensure that unintended counterdoping is not responsible for the EF position measured in this study, the Te 3d spectra were acquired for the sample as a function of annealing temperature. The 160 °C anneal sample still has some of the Se cap present and shows no detectable Te in the film, as shown in Figure 3d. Similarly, no Te is detected after any of the anneals up to 230 °C. This confirms that the concentration of Te is below the detection limits of XPS (