Materials with layered van der Waals crystal structures are exciting research topics in condensed matter physics and materials science due to outstanding physical properties associated with their strong two dimensional nature. Prominent examples include bismuth tritelluride and triselenide topological insulators (TIs), which are characterized by a bulk bandgap and pairwise counter-propagating spin-polarized electronic surface states. Angle-resolved photoemission spectroscopy (ARPES) of ex-situ grown thin film samples has been limited by the lack of suitable surface preparation techniques. We demonstrate the shortcomings of previously successful conventional surface preparation techniques when applied to ternary TI systems which are susceptible to severe oxidation. We show that in-situ cleaving is a simple and effective technique for preparation of clean surfaces on ex-situ grown thin films for high quality ARPES measurements. The method presented here is universally applicable to other layered van der Waals systems as well.
Materials with layered van der Waals (vdW) structures are an area of intense research in condensed matter physics and materials science due to remarkable physical and electronic properties associated with their two dimensional (2D) nature.1,2 The significance of these materials, characterized by weak interlayer vdW bonding, was particularly highlighted in 2004 by the mechanical exfoliation of graphite, which led to the experimental confirmation of the existence of isolated 2D layers.3 The benefits of these materials, however, were in fact realized several years earlier in the context of epitaxial heterostructure growth of highly lattice mismatched crystals through the technique of vdW epitaxy.4,5 Recently, vdW epitaxy has been suggested as a key technique for tailor-made 2D heterostructure crystals that would build on the success of graphene, but compensate for its weaknesses.6
Topological insulators (TIs) are a recently discovered class of materials,7–10 which exhibit robust time-reversal symmetry protected topological surface states that reside inside of an insulating bulk bandgap.11,12 The unique electronic properties associated with these surface states make TIs intriguing candidates for exploring advanced quantum phenomena and device applications.13,14 The prototypical three dimensional TIs, bismuth tritellurides (Bi2Te3) and triselenides (Bi2Se3), are layered 2D vdW crystals, in which ∼1-nm-thick quintuple layers (e.g., Te-Bi-Te-Bi-Te layers) are separated by vdW gaps.
The experimental method of choice for studying the bandstructures and electronic properties of 2D materials is angle-resolved photoemission spectroscopy (ARPES).15 ARPES is a surface sensitive technique that probes the distribution of electrons in solids and is particularly important for studying TI systems with complicated electronic properties due to contributions from the intermixing of excessive bulk carriers and surface Dirac fermions.15 Atomically flat, clean surfaces are required for ARPES measurements which make bulk crystals with natural cleavage planes ideal systems for bandstructure measurements.16 Measurements on in-situ cleaved single crystal Bi2Te3 and Bi2Se3 have been very successful.8,17 In contrast, measurements on thin films, which are the backbone of device structures, are more challenging.18 Maintaining clean surfaces on thin films are best achieved by transferring films directly into the ARPES measurement chamber after growth, without breaking vacuum. Such a scenario requires the ultra-high vacuum (UHV) growth system to be attached to the ARPES measurement system. This is a rare constellation as, e.g., molecular beam epitaxy (MBE) systems are used for a specific materials class only and are generally not feasible for high-throughput, synchrotron-based ARPES end stations. If in-situ growth capabilities are not available and air exposure is therefore unavoidable, surface preparation techniques are required to restore the thin film sample surface in the UHV chamber of the ARPES system.
In this letter, we compare MBE-grown TI thin film samples with ambient environment exposure in terms of their bandstructure and surface morphology after conventional surface preparation steps. We find that Ar+ ion sputtering and annealing, which is one of the most common surface preparation techniques applied to ex-situ grown binary TIs, leads to damaged surfaces and is not sufficient to clean heavily oxidized films. We also find that protective capping layers deposited on TI thin films following growth are only applicable within limits due to film/cap material compatibility, the low desorption temperature of Bi2Te3 (and Bi2Se3), and the changes in the TI surface stoichiometry. We show that in-situ cleaving is the most reliable and promising method for the preparation of 2D layered thin film samples, especially, when they are doped with air-sensitive species such as lanthanides or 3d transition metals.
Bi2Te3 and Gd-doped Bi2Te3 thin films (∼50–200 nm) grown by MBE on c-plane sapphire (α-Al2O3) substrates were chosen for this study. When compared to the simple binary system, rare earth doping adds the complication of severe surface oxidation; a situation commonly found in doped materials.19 Incoherent epitaxial growth of high-quality Bi2Te3 thin films can be achieved on c-plane sapphire despite a lattice mismatch of 9.1%, making Bi2Te3 a typical example of a vdW epitaxy system.20 Details about the MBE growth of these materials can be found in Refs. 19–21. The elemental compositions of the thin films used for this study were determined from Rutherford Backscattering Spectrometry (RBS) using 2.3 MeV Helium ions and Particle Induced X-ray Emission (PIXE) using 1 MeV Hydrogen ions.
Atomic force microscopy (AFM) (Park XE-70) and scanning electron microscopy (SEM) (FEI Magellan 400 XHR) were used to investigate the topography of as-grown (and thus air-exposed) and surface-treated films. ARPES measurements were performed at the high energy resolution spectrometer end station at BL 10.0.1 at the Advanced Light Source at Lawrence Berkeley National Laboratory and on a lab-based system at Stanford University.
Surface oxidation and degradation from environmental contamination of air-transferred TI thin films has created challenges for examining surface electronic properties. Previous ARPES studies aimed at addressing surface contamination issues on ex-situ grown binary chalcogenides have been successful in demonstrating adsorbate removal using sputter-and-anneal cleaning processes, but suffer from superficially broad spectral features in ARPES measurements due to ion-sputtering induced defects and enhanced surface roughness.16,18 While these results show promise, the inherent destructive nature of this technique may not be compatible with ternary or quaternary TI alloy systems which are more prone to severe oxidation and non-stoichiometric surfaces resulting from different sputtering yields.
The surface morphology of Bi2Te3 as well as Gd-doped Bi2Te3 thin films is characterized by well aligned triangular domains which reflect the three-fold symmetry of the system19–21 [see Fig. 1(a)]. Figure 1(b) shows a typical example of a triangular domain found on the surface of a Gd-doped Bi2Te3 thin film. This image reveals the spiral-like terrace-step morphology, similar to the domains observed on Bi2Te3, which consists of atomically flat terraces, separated by steps ∼1 nm in height, corresponding to the height of 1 quintuple layer in Bi2Te3. However, when compared to equivalent thickness binary Bi2Te3 thin films, the introduction Gd into the host Bi2Te3 crystal increases the disorder in the system and results in the growth of smaller domains, with less faceted terraces and more disjointed boundaries.
Since previous experience with surface-sensitive measurements on Gd:Bi2Te3 thin films have indicated severe surface oxidation,19 five cleaning cycles of Ar ion-sputtering (1 kV, 10−7 Torr) and annealing at °C were used to obtain unoxidized surfaces. Core level photoemission spectra and low-energy electron diffraction (LEED) were used to monitor the cleaning process between cycles. Figure 1(d) shows the ARPES spectra intensity plot in energy-momentum space, along the Γ-K direction, after surface cleaning (see inset for LEED pattern). The featureless spectra in Fig. 1(d), indicative of significant charging, shows that the cleaning process has not been sufficient to remove the oxidized surface layers on the Gd:Bi2Te3 thin film. An AFM image of the surface morphology of the Gd:Bi2Te3 sample after surface cleaning is shown in Fig. 1(c). This image reveals that after surface cleaning, the triangular domains are less well defined and show evidence of increased surface roughness. The root mean square (RMS) roughness of the surface treated sample in Fig. 1(c) (∼1.23 nm) has nearly doubled when compared to the RMS roughness obtained from Fig. 1(b), which has not had any surface cleaning steps (∼0.66 nm). These results suggest that for systems which are more prone to severe oxidation alternatives to conventional surface preparation techniques, which have previously been successful on binary TI systems, are needed. These results are particularly relevant to other ternary or quaternary TI materials which have been the recent focus of extensive growth efforts while searching for quantum anomalous Hall systems by doping or alloying TI materials with lanthanides or 3d transition metal elements.
A convenient method to protect the as-grown surface of MBE films against oxidation and decay is to deposit a film-compatible protective capping layer that can be removed by thermal desorption after the sample has been transferred through air. For compound materials, such as GaAs and Bi2Te3 (Bi2Se3), the most common choice is a low-temperature deposited amorphous chalcogen layer. While As-capping of III-V materials has been very successful,22 the effectiveness of Te and Se capping layers to protect and preserve the as-grown surface of air-sensitive Bi2Te3 and Bi2Se3 thin films is less well understood due to challenges associated with the thermal decapping process.
To address this issue, AFM and RBS measurements were used to investigate the decapping process of Te-capped binary Bi2Te3 thin films. Te-capping layers were chosen for this study to avoid a Se-Te exchange at the Bi2Te3 surface.23 Figure 2(a) shows an AFM image of a Bi2Te3 sample covered by an amorphous ∼35-nm-thick Te cap, deposited in-situ at room temperature following thin film growth. Decapping was attempted by annealing Te-capped samples in an UHV environment ( Torr) using temperatures ranging from 125 °C to 250 °C for up to 1 h. Figures 1(b) and 1(c) show AFM images of a Te-capped sample after annealing for 1 h at 250 °C, which reveal that the characteristic Bi2Te3 surface morphology has not been recovered after de-capping attempts and that the Te-capping layer appears to have developed a more well defined microcrystallite structure. RMS roughness values obtained from the as-grown Te-capped sample in Fig. 2(a) and the sample after attempted cap removal in Fig. 2(b) were ∼1.83 nm and ∼2.39 nm, respectively.
RBS measurements, performed to quantify changes in the Te cap and the underlying film, show visible differences in the RBS spectra before (black line) and after (red line) annealing [see Fig. 2(d)]. The knee in the Te signal, indicated by the arrow, shows the position of the Te in the cap layer and indicates a redistribution of Te after annealing. In addition, the Bi signal was found to be higher and the Te signal was wider following the anneal, which implies that Te has surprisingly left the Bi2Te3 film and moved into the Te cap layer. Fitting the obtained data to a structural model reveal the following film compositions and film and cap thicknesses. Before annealing, the cap thickness was Å, which increases to Å after the 1 h anneal. In contrast, the film thickness was reduced from Å to Å, while the atomic percentage of Bi in the film was increased from at. % to at. %, and the atomic percentage of Te in the film was reduced from at. % to at. %, which indicates that the previously stoichiometric Bi2Te3 film has decomposed into Bi2.2Te2.8 after the attempted decapping process. Increasing the anneal temperature would likely enable complete cap removal, but would lead to irreversible damage to the underlying film. These results suggest that Te-capping layers are not an effective method of preserving the as-grown surface of TI thin films.
While in-situ cleaving is a standard technique used to create clean surfaces on bulk TI single crystals for ARPES studies,8,17 as well as for preparing substrates for the subsequent thin film growth,24–26 in-situ cleaving of thin films is uncommon. However, for TI thin films and vdW bonded systems in general, this technique should also be applicable. The weak vdW bonding between quintuple layers along the c-axis in vdW systems naturally lends itself to easy cleavage along the basal plane. To demonstrate the efficacy of in-situ cleaving of ex-situ grown TI thin films for ARPES measurements, a ∼110-nm-thick Gd:Bi2Te3 thin film was prepared for in-situ cleaving by first gluing the substrate side of the sample to the ARPES sample holder with conductive silver epoxy. A small ceramic post (∼2 mm in diameter) was then attached with epoxy to the film side of the sample. Once in the measurement system, under UHV conditions, a portion of the film was mechanically separated by striking the post with a wobble stick, leaving the freshly cleaved (001) sample surface exposed for ARPES characterization [see Fig. 3(a)].
Figures 3(b)–3(e) show small and large scale images of the surface morphology of the Gd:Bi2Te3 film before and after in-situ cleaving. These images indicate that the surface of the sample has clearly been transformed as a result of cleaving since the once present characteristic triangular domains are no longer visible on the sample surface. Instead, the dominant surface feature on the cleaved sample is micrometer long segments with a torn appearance that resembles nanosheets on exfoliated graphite.27 Cross-sectional SEM (not shown) on the air-exposed and cleaved sample indicates that sections with thicknesses of ∼10–25 nm were removed from the film surface during cleaving.
Figure 3(f) shows a comparison of core level spectra of Gd:Bi2Te3 samples after in-situ cleaving and surface cleaning. The spectrum of the in-situ cleaved sample shows prominent peaks of Bi (5d3∕2 and 5d5∕2) and Te (5d3∕2 and 5d5∕2) and negligible intensities near the Fermi level. In contrast, the spectrum of the sample after surface cleaning shows significant intensities near the Fermi level, which come from the surface oxidation. Also, even after surface cleaning, the Bi peaks exhibit a double peak structure which is a consequence of an oxidation-induced core level shift. One may note that the core level of Gd is not present in the spectra, which is due to the relatively low photon energy and the limited energy range.
Figures 3(h) and 3(i) show ARPES spectra intensity plots taken along the Γ-K direction for different photon energies (after potassium dosing). Both the bulk conduction and valence band, as well as the “V”-shaped Dirac cone of the TI surface state, are easily resolved on the in-situ cleaved Gd:Bi2Te3 thin film, without the need for any additional surface preparation steps. The Fermi velocity along the Γ-K direction was estimated to be 4.1 × 105 m/s (2.74 eV Å), which is similar to the Fermi velocity reported for cleaved single crystal Bi2Te3.8 Stacked constant energy contours, obtained at different binding energies, are shown in Fig. 3(g). The evolution of the shape of the surface state band and the appearance of the bulk valence band (at higher binding energies) are clearly visible in the 3D representation and are consistent with previously reported cross-sections of the Dirac-like dispersion of Bi2Te3.8 Additional experiments using in-situ cleaving of ex-situ grown binary and rare earth-doped TI thin films have produced similar quality ARPES data for films with thicknesses as thin as ∼40 nm. Reverse mounting (i.e., film side attached to the sample holder, post attached to the substrate side) has also been very successful, but complicates post-cleaving surface morphology characterization.20
By comparing the bandstructures of in-situ cleaved Gd:Bi2Te3 with in-situ cleaved Bi2Te3, it is evident that Bi2Te3 yields higher quality ARPES data as the bandstructure features are more sharp. However, the degradation in quality is expected as Gd doping, which is mostly substitutional on Bi sites, still induces defects and disorder into the host Bi2Te3. Even with the defect-induced distortion, the ARPES data obtained using this technique is of sufficiently high quality to definitively conclude that the Gd:Bi2Te3 surface state remains intact, as a gap at the Dirac point is not observed. The absence of the gap is indicative that Gd does not form long range magnetic order at the measurement temperature (20 K), and time-reversal symmetry is not broken in the Gd:Bi2Te3 system.
By comparing MBE-grown TI thin film samples with ambient environment exposure in terms of their bandstructure and surface morphology after different surface preparation steps, we have demonstrated that in-situ cleaving is the most promising and reliable method for preparation of clean surfaces of ex-situ grown thin films for ARPES measurements. We found that Ar+ ion sputtering and annealing leads to damaged surfaces and is not sufficient to clean heavily oxidized films. We also found that removal processes for protective capping layers deposited on TI thin films following growth can result in unwanted and irreparable changes in the TI stoichiometry. In-situ cleaving was demonstrated to be a simple and effective method for preparation of clean TI surfaces. It is widely applicable to layered vdW thin film samples, especially to binary TIs as well as more complicated ternary or quaternary TI systems for which conventional surface preparation techniques fail.
This work was supported by a DARPA MESO Project (No. N66001-11-1-4105) and the Army Research Laboratories. S. E. Harrison was supported by the Department of Defense (DoD) through the National Defense Science & Engineering Graduate Fellowship (NDSEG) Program and the VPGE (Stanford University). We thank S. Li, H. Zhang, S.-C. Zhang, R. Chen, A. Lin, Z. Liu, and T. Sarmiento for measurement assistance and helpful discussions throughout the course of this work.