Pulsed-laser epitaxy of topological insulator Bi₂Te₃ thin films

Studies on three dimensional topological insulators (TIs) such as Bi₂Te₃, Bi₂Se₃, and Sb₂Te₃ over the past decade have been intense due to their unique physical properties and potential applications for spintronics and quantum computing.^1–6 The unusual electronic properties are rooted in the effects of the strong spin-orbit coupling (SOC) that results from the heavy elements Bi, Sb, and Te. In these materials, SOC is sufficiently strong to invert the bulk conduction and valence bands and thus change the topological class. The primary physical implication of this manifests on the surfaces, which are required by time-reversal symmetry to be metallic with linear Dirac dispersion and locking of the direction of the spin to the direction of momentum.^1,2 As such, the combination of (i) the ability of techniques like pulsed laser deposition (PLD) and molecular beam epitaxy (MBE) to create atomically thin materials and heterostructures and (ii) the unusual electronic properties of topological insulators have both been key to the discovery of the quantum anomalous Hall effect,³ axion insulators,⁴ and chiral Majorana fermions.⁵ Furthermore, such a combination also enables us to gain precise control over quantum phenomena, for example, room temperature magnetization switching in a TI/ferromagnet heterostructure.⁶ 

Till date, the main technique used to produce these atomic-scale topological insulators has been MBE,^7,8 where individual elements are co-supplied using thermal evaporation. The growth by MBE is simple for these systems due to the volatility of the anion (Se or Te), which can be supplied in excess while maintaining a perfect stoichiometry. This volatility, however, makes the growth by PLD challenging due to the highly energetic growth processes and, as such, less effort has been paid to obtain TI films by PLD.^9–14 In contrast to MBE, PLD uses a single stoichiometric target that is ablated using a high power excimer laser, which produces a plasma plume with energy reaching a few to hundreds of eV.^15,16 This process transfers a small amount of material to the growing film per pulse (typically ≪ monolayer per pulse). PLD is advantageous for complex materials, in particular, multicomponent oxides, due to the ease of changing materials and combining multiple target materials for the creation of heterostructures.^17,18 However, the high energies used often create complications, such as non-stoichiometric transfer from the target to the films as well as impact damage; these challenges are particularly amplified when the materials used are volatile, as with topological insulators^9–14 and materials like iron based superconductors Ba(Fe_1-xCo_x)As₂.¹⁹ These effects, however, can be controlled by carefully tuning the growth conditions, in particular, the substrate temperature, background pressure, laser fluence, and laser spot geometry.¹⁶ Thus, the key hypothesis in this work is can thermodynamic and kinetic growth parameters be balanced to enable the growth of high-quality TIs using PLD where one or multiple elements are volatile? This approach will both answer fundamental questions regarding the non-equilibrium nature of the growth mechanisms inherent to PLD and open new routes to develop novel hybrid materials beyond topological insulators by enabling the creation of heterostructures with other functional materials with TIs—examples include the FeSe/SrTiO₃ interfacial superconductor,²⁰ the multitude of novel 2-dimensional layered materials,²¹ and hybrid complex oxide/chalcogenide heterostructures.²² 

In this work, we demonstrate that it is possible to grow high-crystalline quality Bi₂Te₃ on Al₂O₃ (0001) substrates using PLD by mapping and finding optimum growth conditions. Films were characterized using X-ray diffraction (XRD) for lattice parameter and rocking-curve, Rutherford backscattering spectroscopy (RBS) for film stoichiometry, atomic force microscopy for surface morphology, and low temperature resistivity and Hall effect to probe defect concentration. It is found that successfully growing Bi₂Te₃ by PLD requires fine tuning of the growth temperature (T_G) and background argon gas pressure (P_Ar). For P_Ar < 200 mTorr, the films are non-stoichiometric and highly disordered. Above 200 mTorr, the rocking curve width decreases to a minimum value and the Te-to-Bi ratio reaches the ideal stoichiometric value of 3:2. Furthermore, an ideal temperature range is found between 300 and 350 °C; below 300 °C, the films are found to be highly disordered, while above 350 °C, the films decompose.

The Bi₂Te₃ films were grown on (0001) Al₂O₃ substrates by PLD using a Bi₂Te₃ ceramic target. Before the growth, the Al₂O₃ substrates were thermally annealed at 1200 °C for 6 h to achieve atomically flat surfaces with step-terrace features. The laser fluence was tuned to about 2 J/cm², and the repetition rate was 5 Hz. Argon gas was used to backfill the chamber during growth and was varied between 10 and 350 mTorr. Prior to the growth, the substrates were heated in a background pressure of 10⁻⁶ mTorr to the growth temperature, then the growth pressure of Ar (10 mTorr–350 mTorr) was maintained for 20 min to ensure equilibration of the temperature. The target was preablated for 5 min to ensure that the surface was not contaminated with atmospheric gasses. After the growth, the films were cooled in the same atmosphere to room temperature.

Figures 1(a) and 1(b) show XRD 2θ − θ scans for Bi₂Te₃ films (50 nm in thickness) grown at a fixed temperature of 250 °C with the Ar pressure varied from 10 to 250 mTorr [Fig. 1(a)], and at P_Ar = 150 mTorr with substrate temperature varied from 100 to 400 °C [Fig. 1(b)]. All the films were c-axis (0001) oriented, which was confirmed by the appearance of the 0003n peaks that are characteristic of the three-layer stacking sequence of the quintuple layer structure of Bi₂Te₃. This XRD result confirms that the films were single phase without any impurity phases. These data do show that the film quality varies with the growth parameters. For the growth temperature fixed at 250 °C [Fig. 1(a)] and as the growth pressure was lowered, the specific 0003n peaks vanish; at 10 mTorr, only the 0006 and 00015 peaks were resolvable. Similarly, for films grown with the Ar pressure fixed at 150 mTorr while the substrate temperature was varied from 100 to 400 °C, only 0006 and 00015 peaks were resolvable below 150 °C. This result indicates that there was significant disorder in the films, as discussed further below.

Since controlling volatility is one of the primary challenges, measuring the stoichiometry is necessary. Although not direct, the lattice constant can be used as an indirect probe of stoichiometry. Tracking the peak position of the 00015 peak in the XRD 2θ − θ curves shown in Figs. 1(a) and 1(b), the dependence of the lattice parameter on the growth parameters was extracted. These values are plotted in Fig. 1(c) as a contour plot for Ar pressure on the vertical axis versus the substrate temperature on the horizontal axis. There was a strong dependence on the pressure and only a weak dependence on the growth temperature. For low pressures, the lattice parameter was minimum at around 29.70 Å. For higher pressures, the lattice parameter increased to a value of 30.47 Å. The latter result agrees well with the bulk value, which has two implications: First, the films grown at higher pressures are nominally stoichiometric, which is confirmed by RBS, shown below. Second, this shows that the films are fully relaxed; this is consistent with the so-called van der Waals epitaxy,^23,24 where, due to the weak interlayer van der Waals bonding and the large mismatch of 8.7% between Bi₂Te₃ and Al₂O₃, the films do not maintain strain but are only coherent to the crystalline orientation of the substrate. Moreover, real time measurements made with reflection high energy electron diffraction measurements on the MBE grown films have confirmed that relaxation happens within the first few layers.²⁵ Figure 1(d) shows the same data, but now plotted as a function of the P_Ar. From this view of the data, the different pressure regimes can be seen clearly. For the low pressure range, the lattice parameter depended linearly on P_Ar and then saturated for higher pressures to the bulk lattice parameter. The smaller-than-bulk lattice parameter for low Ar pressure grown films is indicative of a lower film quality and, as detailed below, is correlated with the Bi:Te ratio deviating from the ideal bulk value.

The XRD data further enable quantifying the crystal quality by comparing the full width at half maximum (FWHM) value of the rocking curves, which are found to exhibit both temperature and pressure dependences. Figure 2(a) shows the FWHM values as a function of Ar pressure for fixed temperatures of 250, 325, and 350 °C, while Fig. 2(b) as a function of temperature for Ar pressure fixed at 150 and 250 mTorr. A minimum of about FWHM = 0.17° [see Fig. 2(c)] occurs at P_Ar between 200 and 300 mTorr and at T_G between 300 and 350 °C. This value is more than an order of magnitude lower compared to the largest value of FWHM for samples grown at 250 °C, which is greater than 2.0°. The dependence of the rocking curve data on the growth parameters can be more completely seen as a contour plot as functions of P_Ar and T_G, as shown in Fig. 2(d). The optimized growth region is the area with the minimum FWHM, indicated by blue, where the FWHM is less than 0.2°. This optimized condition occurs approximately between 200 ≤ P_Ar < 350 mTorr and 300 ≤ T_G < 350 °C. Furthermore, within this optimized condition, the surface morphology as characterized by atomic force microscopy (AFM) shows a flat surface [see Figs. 2(e) and 2(f)]. The root-mean-square (RMS) roughness derived from the AFM image was ∼4 nm, which is similar to that of 3 nm for the MBE grown films.²³ 

The stoichiometry was directly measured for a subset of films using RBS, which is the only direct probe of stoichiometry with accuracy to a level of 1-2%. The films used for RBS measurements were grown at T_G = 250 °C with P_Ar = 10, 50, and 250 mTorr, as shown in Fig. 3. For larger P_Ar, the elemental concentrations were found to be close to the bulk values; the corresponding Te/Bi ratios are determined to be 0.73, 1.21, and 1.43 for the films grown at 10, 50, and 250 mTorr, respectively. Therefore, within experimental errors, the films grown under optimized conditions were stoichiometric, and for reduced Ar pressures, the films lost Te.

This then prompts a question regarding the observed contraction of the lattice for Te-deficient films: In a simplistic view, non-stoichiometry can be accommodated by several different point defect-types in the Bi₂Te₃ structure, e.g., antisites (i.e., Bi occupying Te-site), interstitial Bi, and Te-vacancies.²⁶ Based on the larger ionic radius of Bi relative to Te, Bi antisite defects would cause an expansion of the lattice parameter. Similarly, any interstitial accumulation of excess Bi, most likely in between the quintuple layers, would also cause an out-of-plane expansion. Regarding Te-vacancies, there are two unique lattice positions of the Te atoms, and based on a simple ionic picture of the bonding, it can be argued that these may cause either an expansion or contraction. The Te atom positioned inside the quintuple is fully coordinated by Bi ions, and a vacancy at this position would cause the Bi³⁺ ions to repel each other, thus expanding the lattice. The Te that is positioned on the outside of the quintuple layer is directly bonded to Bi within the quintuple layer and to the Te in the adjected quintuple layer via van der Waals bonds; a vacancy at this position could cause a contraction of the lattice because the Bi³⁺ ion would attract the Te^2- in the adjacent layer; furthermore, since the Te atoms at the outer position is more weakly bonded, then thermodynamically, the formation of vacancies on these sites should be more probable. This would be a scenario that is consistent with the experimental observations of a reduction in the lattice parameter and a Te-deficiency. However, this considers only the ionic nature of the bonding and ignores the more complex covalent nature and even possible formation of defect complexes, and a full answer would necessitate the investigation of defect structures using, e.g., scanning tunneling microscopy.

The effect of growth pressure on the stoichiometry of Bi₂Te₃ was further evaluated by transport measurements. The transport properties of Bi₂Te₃ films grown at different pressures are shown in Fig. 4(a). Here, the growth temperature of these films was fixed at 325 °C. The films grown at lower pressures showed higher values of conductivity indicating more charged defects. To determine the carrier density of the optimum film, we measured the Hall effect of the sample grown at 250 mTorr. The inset of Fig. 4(b) shows the Hall resistivity data at 5 K, which were found to be linear over the entire range of magnetic fields up to 14 T. The carriers were found to be holes as indicated by the positive slope of the Hall resistivity curve. As shown in Fig. 4(b) on the left axes, the 2D carrier density was found to decrease slightly with decreasing temperature from ∼12 × 10¹³ cm⁻² at room temperature to ∼9 × 10¹³ cm⁻² at 5 K, and the corresponding 3D carrier density of ∼2.5 × 10¹⁹ cm⁻³ at room temperature and ∼1.9 × 10¹⁹ cm⁻³ at 5 K, which is slightly larger as compared to previous reports,^27–31 and above the threshold for the transport being dominated by the topological surface states.^32,33 Finally, as shown in Fig. 4(b), on the right axis, the mobility was found to increase from ∼100 cm²/Vs at room temperature to 186 cm²/Vs at 5 K.

The change in FWHM, lattice parameter, and stoichiometry with growth parameters (T_G, P_Ar) together indicates that temperature mainly affects crystallization, while Ar pressure plays a predominant role in determining stoichiometry through kinetic processes in the PLD plume. For low temperatures, there is low crystal quality regardless of pressure, which points to insufficient surface kinetics for the film to crystallize. This trend occurs for T_G ≤ 300 °C. For T_G > 350 °C, the FWHM again increases, which marks the upper limit for the growth of Bi₂Te₃; this is typical near the boundary of thermodynamic instability where a significant over-pressure of Te would be necessary to maintain the film. Previously, a study reported that Bi₂Se₃ can be grown under the optimized condition at 300 °C and 75 mTorr.¹⁴ Compared to our growth condition (325 °C, 250 mTorr) for Bi₂Te₃, they share nearly identical optimized growth temperature, but require quite different background pressures of Ar. Considering this observation in light of the different atomic masses and volatility of Te versus Se suggests that this is an effect of plume kinetics. Scattering with the background gas can reduce the kinetic energy of the elements, thus suppressing impact damage, but at high pressures may induce non-stoichiometric growth through different diffusion rates for different atomic species. This process is directly dictated by the mass, as lighter elements will diffuse more readily. Hence, the background pressure should be lower for Se compounds than Te compounds, as observed here.

In summary, this work shows that it is possible to grow chalcogenide epitaxial thin films with volatile elements using PLD, which has long been considered challenging. As such, key questions remain. In particular, (1) how to improve the quality beyond what is reported here? And (2) what other materials can be successfully grown by PLD? Question (1) is fundamentally related to the subject of achieving surface dominated transport in topological insulators because the Bi₂Te₃ family of materials is prone to charge defects that push the Fermi level into the bulk conduction or valence band, thereby obscuring the intrinsic effects of the topological surface states. Regarding question (2), many of the most exciting 2-dimensional layered materials that exhibit intriguing quantum phenomena are transition metal chalcogenides where the transition metals are refractory (Nb, Mo, Ta, W, etc.). Therefore, they cannot be easily grown by MBE as these refectory metals are hard to be thermally evaporated. PLD, in contrast, can ablate these materials and, as demonstrated in this work, can be a great choice if volatile anions can be controlled.

This work was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), Materials Sciences and Engineering Division (dc transport, RBS, and related data analysis) and by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. DOE (synthesis and structural characterization). Q.L. was supported by the DOE, BES, Computational Materials Sciences Program for his experimental assistance. L.N. was supported by the University of Tennessee Governor’s Chair program for recording the RBS spectra.