Growth of high-quality Bi₂Se₃ topological insulators using (Bi_1-xIn_x)₂Se₃ buffer layers Free

Topological insulators (including Bi₂Se₃, Bi₂Te₃, Sb₂Te₃, and their alloys) show great promise for electronic and optoelectronic applications. This is due to their unusual band structure: a bulk bandgap crossed by linearly dispersing surface states.^1–3 These surface states contain two-dimensional Dirac electrons, which exhibit low mass and spin-momentum locking, leading to a large electron velocity and a significant reduction in scattering. These electrons have the potential to be used in a variety of applications, from spintronics to quantum computing.^4,5 For most applications, thin films of these materials grown on a wafer scale are required to fabricate potential devices. Molecular beam epitaxy (MBE) can be used to grow wafer-scale films of all the aforementioned topological insulators (TIs).⁶ Bi₂Se₃ (as well as the other TIs mentioned) has a layered tetradymite structure with unit cell of ordering Se-Bi-Se-Bi-Se, corresponding to one quintuple layer (1QL ∼ 1 nm). In the a-b plane within the QL, the atoms are covalently bonded, but in the c-direction, the QLs are only weakly bound to one another through van der Waals bonds. This enables growth of TIs by van der Waals epitaxy on a variety of lattice-mismatched substrates. Unfortunately, the majority of these films show significant electron doping regardless of substrate choice, indicating a Fermi energy pinned in the conduction band, rather than within the band gap. This Fermi energy position results in a large density of trivial electrons, obscuring the topological effects. Some alloys (notably Bi_2-xSb_xTe_3-ySe_y) do exhibit small bulk doping densities,^7,8 but the bandgap of these materials is smaller than the simpler material, Bi₂Se₃. In addition, the different TI materials have different intrinsic optical and electronic properties, so restricting research to one specific material is not optimal for device applications. It is therefore desirable to be able to grow all TI materials with Fermi energies within the bulk bandgap.

There are three possible mechanisms for introducing bulk carriers: (1) defects within the bulk of the film (vacancies, antisite defects, and interstitials); (2) band bending at the top or bottom surface due to oxidation or other surface states; and (3) defects at the film/substrate interface. To test for bulk defects, one can investigate how the sheet and bulk carrier densities scale with film thickness. If these defects are a major contributing factor to the overall carrier density, we expect the three-dimensional bulk density to remain relatively constant as film thickness increases while the two-dimensional sheet density should increase. In our films, however, we have found that the two-dimensional sheet density remains relatively constant as the film thickness increases, indicating that defects in the bulk of the film are a relatively minor contribution to the overall carrier density.⁹ In addition, we have investigated the growth and electrical properties of Bi₂Se₃, Bi₂Te₃, and Bi₂(Se_1-xTe_x)₃ (BST) alloys. In bulk crystals, Bi₂Se₃ tends to be n-type, while Bi₂Te₃ tends to be p-type. The bulk BST alloy, therefore, shows significantly reduced trivial carrier conduction.^10–15 However, in our thin films, all alloy concentrations exhibited n-type carriers, with the largest carrier density in the pure Bi₂Te₃ films.¹⁶ This again implies that bulk defects are not the primary cause of excess doping in our TI films. Finally, cross-sectional transmission electron microscopy images of the TI film/substrate interface (for TI films grown on sapphire) show that the first few QLs are highly defective.^17–21 This is supported by reflection high energy electron diffraction (RHEED) images, which are usually of poor quality for the first few QLs. We therefore conclude that defects at the film/substrate interface may be playing a large role in the electrical properties of our TI films.

Recently, there have been some efforts to grow In₂Se₃ by molecular beam epitaxy.^22,23 In₂Se₃ is a trivial bulk insulator with the same crystal structure and similar lattice constant to Bi₂Se₃. This makes In₂Se₃ a good candidate for use as a buffer layer between the TI and the substrate and for use in heterostructure devices.^24–26 However, the lattice match between In₂Se₃ and Bi₂Se₃ is not perfect (∼3.4% lattice mismatch). In addition, indium is known to diffuse into Bi₂Se₃ at the Bi₂Se₃/In₂Se₃ interface.²⁷ We have therefore chosen to investigate the use of the (Bi_1-xIn_x)₂Se₃ (BIS) alloy as a buffer layer. For indium concentrations above ∼30%, BIS is a trivial band insulator with a gap size dependent on indium concentration.²⁸ Using large bismuth concentrations results in a BIS buffer layer with a lattice constant closer to the desired Bi₂Se₃ film while reducing indium diffusion into the overgrown TI layer. Despite the many advantages of the BIS alloy, literature describing the optimal growth procedure is scarce.

In this paper, we describe growth of BIS on sapphire for a variety of indium concentrations under a variety of growth conditions. We present atomic force microscopy (AFM) and x-ray diffraction (XRD) data showing the high quality of our BIS buffers. Finally, we discuss Bi₂Se₃ films grown on BIS buffers and demonstrate a reduced trivial bulk carrier density with a commensurate increase in mobility, indicating that the use of BIS as a virtual substrate for the growth of Bi₂Se₃ films results in significantly higher quality TIs.

All BIS films were grown in a dedicated Veeco GenXplor molecular beam epitaxy chamber. The films are grown on c-plane sapphire with a selenium cracking source, resulting in highly reactive species and improving the chalcogen incorporation. Bismuth and indium are both evaporated using dual-filament effusions cells; their fluxes are controlled using beam flux monitoring. Fluxes are set for a growth rate of ∼1 nm/min. Substrate temperatures are monitored using a thermocouple. Sapphire substrates are first outgassed in the loadlock before being heated to 650 °C in the MBE chamber. The substrate temperature is then lowered for film growth; exact temperatures and growth procedures will be detailed in Sec. III. The crystal quality is monitored in situ by RHEED. After growth, films are characterized using XRD, AFM, and room temperature van der Pauw four-point Hall measurements using small (<1 T) magnetic fields. Film composition is determined by growing films of pure Bi₂Se₃ and In₂Se₃ while varying the bismuth or indium flux. The thickness of these pure films is measured using x-ray reflectivity. When growing BIS films, we assume a sticking coefficient of 1 for both bismuth and indium and assume complete incorporation for both elements. We find that film thicknesses for BIS are as expected, thus validating these assumptions.

In this section, we will describe our efforts to determine the ideal growth parameters for the BIS buffer layer and for the overgrown Bi₂Se₃ material. We have broken this up into four sections: growth of BIS on a seed layer, BIS as a buffer layer, effect of substrate temperature during BIS growth, and effect BIS composition on Bi₂Se₃ properties.

Initially, we attempted to grow BIS films of a variety of concentrations directly on sapphire substrates using a codeposition method supplying bismuth, indium, and selenium simultaneously. Despite a detailed exploration of growth parameter space, in all cases, the RHEED pattern did not show the sharp lines characteristic of good film growth. An example is shown in Fig. 1(b) for a film with x = 0.5. Diffuse, spotty streaks appear which are indicative of poor-quality growth. In addition, the same RHEED pattern was observed for all substrate rotations, indicating a polycrystalline film. XRD measurements taken on these films are presented in Fig. 1(a), with indium concentration increasing from bottom to top. For pure Bi₂Se₃ films, we observe the family of peaks corresponding to the [000m] direction near 17° [0006], 27° [0009], and so on, as well as peaks near 40° from the substrate. The full width half maximum (FWHM) of the Bi₂Se₃ peaks is small, indicating good film quality. However, as the indium content increases, the film quality degrades. The FWHM becomes larger and, especially for the pure In₂Se₃ film, double peaks arise. We believe this growth difficulty is primarily due to the multiple polymorphs of In₂Se₃, as well as the many other In-Se phases.²⁹ In₂Se₃ has at least five polymorphs, usually denoted α, β, γ, δ, and κ.^30–34 The α phase is generally considered to be the only stable phase at room temperature in bulk crystals, though the β and γ phases have been observed at room temperature in thin films. Although all four polymorphs share the same $R3¯m$ structure as Bi₂Se₃, they exhibit different lattice constants. The literature on this compound is inconsistent, but most reports agree that the β phase has a slightly smaller in-plane lattice constant than the α phase, while the γ phase is much larger. We are aiming to grow only the α-phase of In₂Se₃, as it has the most similar lattice constant to Bi₂Se₃.

In order to achieve high-quality growth of large indium content BIS films, we moved to a thin seed layer grown with sequential deposition. In this method, we deposited 5QL of Bi₂Se₃ on the sapphire substrate at 300 °C, followed by 5QL of In₂Se₃ at the same temperature.²⁶ The bilayer was then heated under a selenium overpressure to 425 °C and annealed. This annealing step resulted in the indium diffusing into the Bi₂Se₃ layers,²⁷ leading to a homogeneous 10QL 50% BIS seed layer with good structural quality [AFM shown in Fig. 3(a)]. The seed layer was then cooled to the desired growth temperature, and subsequent BIS films grown using the standard co-deposition method. Hall measurements performed on a sample of only the seed layer were completely insulating. This means sufficient numbers of indium atoms diffuse into the bottom Bi₂Se₃ layer to turn it from a topological insulator to a trivial insulator. In Figs. 1(d) and 1(e), RHEED images of a pure In₂Se₃ film grown on this seed layer are shown. The streaks are narrow and we observe different streak spacing in the $101¯0$ and $112¯0$ directions, indicating flat, single-crystal growth.³⁵ Figure 1(c) presents XRD data for five BIS films grown using this seed layer technique. All films show narrow linewidths, with the FWHM varying only from 0.18° to 0.24°. For high indium content films, no peak splitting is observed, indicating successful growth of a single In₂Se₃ polytype. We therefore determine that the subsequent deposition seed layer technique provides a good template for the growth of BIS films of all compositions.

The ultimate goal of this study is to use BIS as a buffer layer for the growth of high-quality Bi₂Se₃. In this section, we will compare the morphology and electrical properties of a Bi₂Se₃ film deposited directly on sapphire to one grown on a co-deposition BIS buffer and one grown using the BIS seed layer technique. We judge the quality of the overgrown Bi₂Se₃ films by their room temperature electrical properties: films with a higher mobility and a lower sheet density (indicating a Fermi energy closer to the Dirac point) are of better quality. It should be noted that every BIS film we have tested is insulating at room temperature. We are therefore confident that the electrical properties for Bi₂Se₃ grown on BIS come only from the Bi₂Se₃ film. In Table I, we provide the growth details for all the films we will be discussing in this paper.

We begin our discussion by considering films A, B, and C. These comprise 50 nm of Bi₂Se₃ grown directly on sapphire (A), 50 nm of Bi₂Se₃ grown on 50 nm of 50% BIS grown with co-deposition (B), or 50 nm of Bi₂Se₃ grown on 50 nm of 50% BIS grown on the seed layer (C). In all cases, after the growth of the BIS buffer, the substrate temperature was reduced to 300 °C for the growth of the Bi₂Se₃ layer. The properties of sample A are fairly standard for Bi₂Se₃ films grown on sapphire; a sheet density of ∼3 × 10¹³ cm⁻² is indicative of a Fermi energy near the bottom of the conduction band. If the Fermi energy were within the bandgap (resulting in transport only through the topological surface states), we would expect a sheet density in the 10¹² cm⁻² range. Sample B grown using co-deposition shows a significant reduction in sheet density, but also a reduction in mobility. The improvement in sheet density can be attributed to a reduction in defects in the first few Bi₂Se₃ QLs due to the presence of the buffer layer, while the reduction in mobility is due to the poor surface morphology of the co-deposition buffer. For sample C, we observe a significant increase in mobility, while the sheet density remains low, approximately half that of the Bi₂Se₃ film grown directly on sapphire. A plot of these results is shown in Fig. 2. In Fig. 2, we also show AFM images for samples A and C. Sample A shows many triangular domains and significant twinning, as expected for these films grown by van der Waals epitaxy.³⁶ However, film C shows a wider range of domain sizes but a smaller overall surface roughness. In addition, the extent of twinning as judged by the variation in orientation of the triangular domains is drastically reduced, leading to an overall improvement in mobility. We can therefore clearly conclude that growing Bi₂Se₃ on a BIS buffer with a seed layer significantly improves the film properties. In the rest of the article, we will describe how various growth parameters for the buffer layer grown on the seed layer affect the quality of the buffer and thus electrical properties and morphology of the overgrown Bi₂Se₃ films.

We first studied the effect of substrate temperature on the properties of the BIS buffer. Three samples with a buffer thickness of 50 nm and a 30% indium concentration were grown: D buffer (T_sub = 425 °C), E buffer (T_sub = 375 °C), and F buffer (T_sub = 325 °C). Atomic force microscopy images of the buffers alone are shown in Fig. 3. Figure 3(a) shows the surface morphology of the seed layer after annealing. The surface is generally smooth, with many small domains of 1–2QL in height. It should be noted that the tall dots observed on some AFM images are attributed to environmental contamination, and are not believed to be intrinsic to the film. Figures 3(b)–3(d) shows AFM images for the three buffer layers with no overgrown Bi₂Se₃. It is clear that the substrate temperature significantly affects the structural quality of the buffer. For the buffer grown at 425 °C [Fig. 3(b)], large hexagonal and triangular domains are observed, as expected. The pitlike features in this sample are caused by incomplete layer growth, most likely due to thermal roughening. For the buffer grown at 375 °C [Fig. 3(c)], we primarily observe triangular domains and spiral growth with layer step heights of 1QL, similar to what occurs during growth of pure Bi₂Se_3.³⁶ There are also fewer pitlike features in this buffer than in the one grown at 425 °C, which supports thermal roughening as their cause. Finally, Fig. 3(d) shows the buffer grown at 325 °C. The morphology for this sample is quite different that the others, comprising only hillocks with no observable crystalline domains. In addition, the single QL step heights that were observed in Figs. 3(b) and 3(c) are not present in this buffer. This was somewhat surprising, since we grow pure Bi₂Se₃ at 300 °C, and these buffers only contain 30% indium. We hypothesize that the lower growth temperature led to indium aggregation in this film, which leads to the observed hillocks. Overall, it is clear that the substrate temperature has a strong effect on buffer morphology.

We now consider films D, E, and F, comprising 50 nm Bi₂Se₃ grown on these buffer layers. The electrical properties of these samples are shown graphically in Fig. 4. Sample E, with the buffer grown at 375 °C shows the highest mobility and lowest sheet concentration, while sample D (425 °C buffer) is only slightly worse. However, sample F (325 °C buffer) shows a significant reduction in mobility and increase in sheet concentration. This is consistent with the buffer morphology: the buffer for sample E showed the best morphology, with large triangular domains, while the buffer for sample D was similar though showed more pits between domains. The poor quality buffer for sample F resulted in a poor quality overgrown Bi₂Se₃ film. This can also be observed in the AFM images of the overgrown Bi₂Se₃, also shown in Fig. 4. Sample E has the largest domains and highest mobility, while sample D has only slightly smaller domains and reduced mobility. Sample F, however, has much smaller domains and a correspondingly low mobility.

It is clear that buffer layers grown at higher temperatures perform better than those grown closer to the 300 °C growth temperature of Bi₂Se₃. For Bi₂Se₃ growth, we generally grow in the temperature regime of T_Se < T_sub < T_Bi, allowing the use of a selenium overpressure without excess selenium sticking on the film surface and leading to a growth rate determined purely by the bismuth flux. However, for the case of BIS growth, somewhat higher growth temperatures are clearly beneficial. We attribute this to the stability of the In₂Se₃ compound as well as the smaller surface mobility of indium compared to bismuth. In₂Se₃ melts at ∼1160 K, while Bi₂Se₃ melts at only 979 K. It is therefore not surprising that the optimal growth temperature of BIS is higher than for pure Bi₂Se₃.

Next, we studied the effect of indium concentration in the BIS buffer on the quality of the overgrown Bi₂Se₃ film. Three samples were considered, all with buffer layers grown at a substrate temperature of 425 °C: D (x = 0.3), C (x = 0.5), and G (x = 0.75). In Fig. 5, we present AFM images and electrical properties for 50 nm Bi₂Se₃ grown on these three buffer layers. From the Hall data, it is clear that the composition of the buffer does not have a strong effect on the quality of the overgrown Bi₂Se₃ film. Sample D, with the lowest indium concentration, has a slightly higher sheet density and lower mobility, but samples C and G are identical to within error bars. This is consistent with the AFM images of the overgrown films. The AFM for sample D shows many small domains, while samples C and G show reasonably large, flat domains. We therefore have the freedom to choose our buffer composition to fit our needs without drastically affecting the quality of the TI film.

In summary, we have described the growth of high-quality Bi₂Se₃ topological insulator thin films on a trivially insulating (Bi_1-xIn_x)₂Se₃ buffer layer. In order to grow a flat, single-crystal film of BIS on the sapphire substrates, we used a 10QL seed layer in which 5QL of Bi₂Se₃ was grown followed by 5QL of In₂Se₃. This bilayer was then annealed to form a BIS seed layer, after which BIS films with a variety of compositions and growth parameters were grown. BIS films grown at relatively high temperatures exhibited larger domains and overall superior morphology compared to those grown at colder temperatures. We found that Bi₂Se₃ films grown on the best quality BIS buffer layers exhibited a reduced carrier density and increased mobility at room temperature, indicating a Fermi energy closer to the Dirac point and an overall reduction in trivial carriers. Although these films likely still contain some trivial carriers, the overall density is reduced by more than a factor of two, with significant implications for electrical and spintronic device designs. These results clearly indicate that Bi₂Se₃ films are highly sensitive to the quality of the underlying surface, despite the use of van der Waals epitaxy. We suspect that further improvements in buffer quality could further improve the overgrown topological insulator film quality.

Y.W. and S.L. acknowledge funding from the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award No. DE-SC0016380. T.G. and S.L. acknowledge funding from the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award No. DE-SC0017801.