Microstructure and transport properties of epitaxial topological insulator Bi₂Se₃ thin films grown on MgO (100), Cr₂O₃ (0001), and Al₂O₃ (0001) templates

Topological insulators (TIs) have received considerable attention due to their novel properties arising from strong spin-orbit coupling and massless Dirac-cone-like surface states protected by time reversal symmetry (TRS).^1,2 The special kind of surface states is expected to exhibit various unique quantum phenomenon, which may bring revolutionary developments in low power electronics and topological quantum computation. While such unique systems offer nontrival surface states that can be utilized to perform dissipationless spin transport, it is equally important to break TRS of TIs to create a variety of exotic topological effects including the half-integer quantum Hall effect,³ the topological magnetoelectric effect,⁴ and the magnetic monopole.⁵ One major drawback for the implementation of TIs into real electronics devices, such as field effect transistors, is their linear energy spectra, which normally allows incident electrons to pass through a potential barrier via Klein tunneling without reflection, leading to an appreciable off-current and thus a poor on/off current ratio.^6,7 Therefore, opening a surface energy gap and generating of massive surface carriers by breaking TRS is the key for both fundamental physics research and new materials displaying exotic phenomena aimed at technological applications.

Two routes for breaking TRS have been developed by introducing ferromagnetic ordering within TIs. The first involves doping the TI host with specific elements, for example, by which ferromagnetism has been observed in Mn-doped Bi₂Te₃ single crystals^8,9 and Mn- and Cr-doped Bi₂Se₃ thin films.^10,11 However, it is difficult to separate the surface and the bulk phases using this approach. Furthermore, doping of magnetic elements inevitably introduces crystal defects, magnetic scattering centers, and impurity states in the insulating gap, which are detrimental to mobility and the transport of surface states in TIs.^12–14 In contrast, the second route is based on introducing ferromagnetism to the TI surface by proximity to a ferromagnetic insulator (FMI). Suitable FMIs have the potential to achieve strong and uniform exchange coupling via contact with TIs without significant spin-dependent random scattering of helical carriers on magnetic atoms.^15–17 To date, the prototype three dimensional (3D) TI material, Bi₂Se₃, has drawn much attention for the study of proximity effect with FMIs. Several FMI materials have been investigated to interface with Bi₂Se₃, such as EuS, GdN, and Y₃Fe₅O₁₂ (YIG). The reference for low ferromagnetic transition temperature (T_c) of these materials, however, minimizes device applications, for example, the T_c for Bi₂Se₃/EuS, Bi₂Se₃/GdN, and Bi₂Se₃/YIG are ∼15.7 K, ∼13 K, and 130 K, respectively.^18–21 

Here, we employ MgO and Cr₂O₃ controllable defect-induced room temperature FMI materials, as substrates for Bi₂Se₃ growth and demonstrate proximity-induced ferromagnetism in Bi₂Se₃. The MgO and Cr₂O₃ are selected as the FMI templates not only based on processing perspectives related to epitaxial thin film growth but also for device applications. Using domain matching epitaxy,²² the misfit strain is relieved within a couple of monolayers as a result of the large lattice misfit; the lattice misfit for MgO/Bi₂Se₃ is ∼16% and for Cr₂O₃/Bi₂Se₃ is ∼39%. As a result, the critical thickness of pseudomorphic growth is less than one monolayer and the misfit strain can be engineered and confined near the interface, with the rest of the film grown free of lattice misfit strain. From the practical point of view, both MgO and Cr₂O₃ are considered potential candidates for tunnel barriers in magnetic tunnel junctions, where the contact resistance is spin dependent and becomes comparable to the spin independent resistance of the normal metal.^23–25 In addition to being crucial components for nonvolatile magnetoresistive random-access memory and low-noise magnetic sensors based on the tunneling magnetoresistance (MR) effect,²⁶ MgO and Cr₂O₃ show room-temperature ferromagnetism (RTFM), associated with a strain-dependent defect-mediated mechanism. Li et al.²⁷ showed that undoped MgO thin films grown by pulsed laser deposition (PLD) have the RTFM signature and that the ferromagnetism exhibits strong correlation between magnesium vacancies (V_Mg) and the crystallinity. They found that reduced crystallinity increases the ferromagnetic spin-order of MgO thin films due to increased V_Mg. As for Cr₂O₃, though bulk Cr₂O₃ is an antiferromagnetic with T_N = 307 K, He et al.²⁸ showed the existence of a roughness-insensitive ferromagnetic state at the Cr₂O₃ (0001) surface, and confirmed its spin-polarized property using experimental and theoretical approaches. Later, Punugupati et al.²⁹ demonstrated that epitaxial Cr₂O₃ thin films exhibit ferromagnetic-like hysteresis loops with high saturation and finite coercivity up to 400 K due to oxygen related defects whose concentration is controlled by the strain present in the films.

In this research, we present structural, magnetic, and magnetotransport characterization of TI/FMI heterostructures. Two hybrid structures are studied based on proximity-induced ferromagnetism in a TI: Bi₂Se₃/Cr₂O₃/c-sapphire and Bi₂Se₃/MgO/c-sapphire. The key advantage here is that the TRS breaking occurs mostly at the TI/FMI interface, rather than affecting the majority of bulk states. Furthermore, the approach avoids the creation of secondary phases, clusters or defect agglomerations, which result from the doping 3d transition metal impurities into TIs.

All samples presented here were grown using PLD (λ = 248 nm and τ = 25 ns). Three heterostructures were studied: Bi₂Se₃/Cr₂O₃/c-sapphire, Bi₂Se₃/MgO/c-sapphire, and Bi₂Se₃/c-sapphire, which are referred hereafter as Cr₂O₃-bilayer, MgO-bilayer, and control sample, respectively. This enables us to compare the structural, magnetic, and transport properties of Bi₂Se₃ thin films grown on a non-magnetic substrate to those grown on FMI substrates. The Cr₂O₃ film was deposited from a Cr target that was held at 650 °C under an oxygen partial pressure of 5 × 10⁻² Torr with laser energy density ∼3.2 J/cm². The MgO thin film was deposited at the same temperature under an oxygen partial pressure of 5 × 10⁻⁵ Torr along with∼2.8 J/cm² laser energy density. The detailed Bi₂Se₃ growth conditions have been reported previously,³⁰ using alternating laser pulses; here, a 1:1 ratio of Bi₂Se₈: Se pulses was used at 150 °C under 3 × 10⁻¹ Torr Ar pressure, to reduce the formation of Se vacancies. All Bi₂Se₃ films have the same thickness, ∼15 nm, well above the 6 nm threshold, below which the wavefunctions of the top and bottom surfaces overlap substantially, resulting in a hybridization gap near the Dirac point where the transport characteristics are similar to those brought by strong magnetic interactions.^31–33 The thicknesses of the Cr₂O₃ and MgO films were around 100 nm.

To characterize the crystal structure, x-ray diffraction (XRD) was performed using a Rigaku X-ray diffractometer with CuKa1 radiation (λ = 1.54056 A °). Phi-scans were performed by Philips X-Pert Pro X-ray diffractometer equipped with a high resolution goniometer to investigate the crystallographic relationship between substrate and grown layer. Microstructural and growth characteristics of the heterostructures were investigated using a JEOL-2010 field emission transmission electron microscope (TEM), operating at 200 kV with a point-to-point resolution of 1.8 A, and, with high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) using a FEI probe-corrected Titan G2 60–300 operated at 200 kV. Magnetotransport characterization was performed on 10 mm × 10 mm thin films using a standard four-probe method in the van der Pauw geometry in magnetic fields up to 5 T in an Ever Cool Quantum Design Physical Property Measurement System (PPMS) system with a base temperature as low as 2 K. A custom-built gold-coated pogo pin setup was used to make contacts and ensures the same distance between the contacts on the films. Angular-dependence measurements were performed with a horizontal rotator in the PPMS. The magnetic properties were measured by applying magnetic field parallel and perpendicular to the Bi₂Se₃ (003 n) surface using a Super Conducting Quantum Interference Device.

Fig. 1(a) displays a XRD θ-2θ pattern of a Cr₂O₃-bilayer sample. Both the Cr₂O₃ and Bi₂Se₃ layers exhibit c-axis preferred orientation crystallographic peaks, suggesting films are highly textured. To confirm the epitaxial growth of the films and establish the epitaxial relationships, XRD in-plane Φ-scans were used. Fig. 1(b) shows the Φ-scans obtained by exciting (104) lattice planes of Cr₂O₃ and sapphire and (015) planes of Bi₂Se₃. The three fold symmetry illustrates that the Cr₂O₃ is rhombohedral and grows epitaxially on c-sapphire without any in-plane rotation. The presence of Bi₂Se₃ six peaks indicates that the film is epitaxial as a result of the existence of two domains. A schematic showing the alignment of two different domain variants of Bi₂Se₃ on a Cr₂O₃ template is seen in Fig. 1(c). In one case, the Bi₂Se₃ basal plane is aligned with that of Cr₂O₃ and in the other there is a 60°/180° rotation. The epitaxial relationships are written as (0001)Bi₂Se₃||(0001)Cr₂O₃||(0001)Al₂O₃ and [2-1-10]Bi₂Se₃||[2-1-10]Cr₂O₃|| [2-1-10]Al₂O₃ (or) [2-1-10]Bi₂Se₃|| [11-20]Cr₂O₃|| [11-20]Al₂O₃.

The out-of-plane XRD θ-2θ pattern of the MgO-bilayer sample is shown in Fig. 2(a). The MgO is completely orientated in the (111) plane, and Bi₂Se₃ exists only the (003 n) reflections, suggesting a highly textured single-phase heterostructure. The in-plane orientation relationship between Bi₂Se₃, MgO, and c-sapphire substrate was established by Φ-scan; the scan results are shown in Fig. 2(b). It can be seen that the Φ-scan for (200) planes of MgO results in six sharp peaks, confirming the epitaxial growth. The presence of these reflections is attributed to the threefold symmetry of MgO film growth which is epitaxial with two domain types that have 60° in-plane rotation with respect to each other about the [111] growth direction. The occurrence of two in-plane orientations is explained by stacking sequences in the [111] direction, which make up a face centered cubic (fcc) structure. Any fcc on a hexagonal (0001) or rhombohedral (001) systems, when grown in the [111] direction, has two possible orientations. An fcc structure is described as having an ABCABC… type stacking in the closed-packed direction. During the nucleation on the closed packed plane (cpp) of Al₂O₃, the first monolayer plane sets in one orientation, which is the same everywhere if the substrate surface has no steps. Fixing this first monolayer cpp as a reference and denoting it as A, the next cpp monolayer grows as a B or a C layer. Upon further growth in the [111] direction, the fcc can stack as either ABCABC… or ACBACB…. These two possibilities are random and manifest themselves in two different fcc with a 60° in-plane rotation with respect to each other. Here, all constituent layers within the heterostructure were found to be epitaxial with the relationships: (0001)Bi₂Se₃||(111)MgO||(0001)Al₂O₃ and [10-10]Bi₂Se₃||[-110]MgO||[10-10]Al₂O₃; a schematic showing the relative orientations between each layer is presented in Fig. 2(c).

Detailed microstructural analysis was performed by HRTEM and HAADF-STEM on both Cr₂O₃ and MgO–bilayer samples. All the images present here were taken in the c-sapphire [11–20] zone axis. Figs. 3(a) and 3(b) show the interfaces within Cr₂O₃/c-sapphire and Bi₂Se₃/Cr₂O₃, indicating that both are clean, sharp, and reaction free. The fast Fourier transform (FFT) diffraction patterns in the insets confirm the highly epitaxial growth of the Cr₂O₃ layer. The atomically ordered structure of Bi₂Se₃ is observed in Fig. 3(b) where the horizontal arrows and the kinks in the zig-zag lines identify twin boundaries which are in agreement with the Φ-scan results. The Bi₂Se₃ layered structure, formed by Bi and Se, is stacked along the c-direction in five layer packets Se₁-Bi-Se₂-Bi-Se₁ that connect to each other by weak van der Waals bonds that are also shown in Fig. 3(b), indicating the high quality of Bi₂Se₃.

Fig. 3(c) shows the interface between MgO and c-sapphire in an MgO-bilayer. Fig. 3(d) shows more clearly the existence of a ∼2 nm spinel structure that formed at the interface when MgO was grown at 650 °C This interfacial layer, identified as MgAl₂O₄, was not observed when the deposition was carried out at 550 °C. It should be noted that the presence of this ∼2 nm interfacial layer at the MgO/c-sapphire interface does not affect the subsequent MgO growth as a rocksalt phase with the [111] axis aligned parallel to the substrate normal. The inset FFT diffraction patterns confirm the epitaxial growth of MgO. Since electrical current will only flow through the Bi₂Se₃ surface, i.e., charge transport is restricted solely to the upperlaying Bi₂Se₃ due to the highly insulating nature of MgO, the existence of ∼2 nm MgAl₂O₄ should not affect the transport properties. Fig. 3(e) shows an HRTEM image at the Bi₂Se₃ and MgO interface, showing a clean and sharp interface with no evidence of any interfacial reaction, which is crucial for a strong proximity effect in hybrid structures. Lamellar twins are found at the boundaries of the different stacking sequences parallel to the growth plane in Bi₂Se₃; these are seen in the marked region in Fig. 3(e). The microstructure results are consistent with the XRD Φ-scan with six peaks seen presented in Bi₂Se₃ bilayer.

Fig. 4 compares the resistivity versus temperature for all three sample types, directly revealing the influence of the magnetic layer on transport in Bi₂Se₃. At high temperatures, all films show metallic behavior. Although they have approximately the same temperature coefficients in resistivity, the bilayer samples have larger resistivity (or suppressed conductivity), compared to the control sample, with the highest resistivity found in the MgO-bilayer for all temperatures. The Hall effect measurements at 300 K indicate that charge carriers of Bi₂Se₃ in all samples are of n-type with a net carrier density ∼1.9 × 10¹⁹/cm³, 1.75 × 10¹⁹/cm³, and 1.84 × 10¹⁹/cm³ for the pristine, Cr₂O₃-bilayer, and MgO-bilayer, respectively. The carrier concentration does not vary significantly among the three samples. However, the mobility values of both bilayer samples are consistently smaller in magnitude than the pristine samples, having ∼426 cm²/V s for pristine and ∼334 cm²/V s and ∼108 cm²/V s cm³ for Cr₂O₃-bilayer and MgO-bilayer, respectively. One probable cause for this suppression in mobility is due to the presence of insulating magnetic layers. In addition, as the temperature decreases, the resistivity shows a minimum and then increases as the temperature decreases further, suggesting stronger electron-electron interactions in two-dimensional (2D) systems. A comparison of the low-temperature insulating behaviors of all samples is shown in the inset of Fig. 4. The insulating behavior in the bilayer samples is more pronounced than in the control sample. The onset temperature of the insulating behavior is also higher in the bilayer samples. These results suggest that the magnetic layers may be responsible for the stronger insulating behavior in the bilayer samples. The onset temperature in the MgO-bilayer sample (∼42 K) is higher than that in the Cr₂O₃-bilayer sample (∼28 K), which is explained on the basis of their microstructural differences. Fig. 5(a) shows the low magnification bright-field TEM image of a cross-section of the MgO-bilayer sample where Bi₂Se₃ contains a small misorientation between two grains, labelled A and B, respectively. The selected area electron diffraction (SAED) pattern in Fig. 5(b) further confirms that the Bi₂Se₃ spot is split into two components with coherence; the angle between these two spots is less than 2°. The observation of extra diffraction patterns with the same radius in Bi₂Se₃ is believed due to the small angle grain boundaries, referred to as sub-grain boundaries, when the Bi₂Se₃ was grown on MgO. It should be noted that the sub-grain boundaries are not distinctly sharp owing to the fact that these boundaries can run in any direction leading to an overlap of neighboring domains in the projected TEM zone direction. Nonetheless, such small-angle tilt grain boundaries are not observed in the Cr₂O₃-bilayer samples. The SAED pattern of Cr₂O₃-bilayer sample is shown in Fig. 5(c), and no evidence of the sub-grain diffraction patterns is observed. Therefore, the higher resistivity and onset temperature in the MgO-bilayer sample could be associated with grain boundary scattering, which could increase the resistivity.

It is important to note that in TI/FMI hybrid structures, even though strong exchange coupling exists across the interface, no gap in the topological surface states is expected with in-plane magnetization. When the magnetic easy axis of the FM is out-of-plane, a gap opens in the surface states of the vicinal TI.^33,38 The magnetic properties were measured with the applied magnetic field parallel (Fig. 6) and perpendicular (Fig. 7) to the Bi₂Se₃ (003 n) surface for all bilayer sample types. The control samples are diamagnetic in both directions, consistent with previous reports.³⁹ The in-plane magnetization at 300 K and 5 K for Cr₂O₃ and MgO-bilayer samples provide the evidence of ferromagnetic signatures with clear saturation and hysteresis. Interesting, the Cr₂O₃–bilayer sample is paramagnetic with observable coercivity in the out-of-plane direction, whereas the MgO-bilayer is diamagnetic in this orientation. More detailed analysis is required for investigating the paramagnetic mechanism along the c-axis of the Cr₂O₃–bilayer sample.

The normalized MRs as a function of temperature (2 K–25 K) and magnetic field perpendicular to the films are compared in Figs. 8(a) and 8(b) for Cr₂O₃-bilayer versus control sample and MgO-bilayer versus control sample, respectively. The shape of the MR in bilayer structures are drastically different from the control samples, though they have overall positive MR, and at low magnetic field, the MR exhibits the weak antilocalization (WAL) cusp feature. This WAL could reflect the nontrival topology of the surface states and is suppressed when the temperature increases as a result of the decreasing coherence length. At higher magnetic fields (>4 T), the MR of the control sample does not saturate and follows a linear-like dependence, showing a positive, weak temperature-dependence up to B = 5 T. The linearity of the MR can be interpreted through the quantum MR model proposed for a zero gap band structure with Dirac linear dispersion.³⁴ In contrast, the bilayer samples quickly saturate at low magnetic field (∼2.5 T).

Fig. 8(c) compares the MR at T = 2 K for all samples as a function of perpendicular magnetic field. As a result of proximity with Cr₂O₃ and MgO, the localization behavior of the Bi₂Se₃ is significant altered. The enlarged part in Fig. 8(c) clearly shows that the introduction of magnetic components reduces the sharpness of bilayer samples, the cusp features at low magnetic field, indicating a weakened WAL effect, which may be associated with an increase of backscattering of Dirac fermions.

To understand the origin of the weakened WAL effect in the bilayer samples, the low-field magnetoconductance data are analyzed quantitatively using the Hikami-Larkin-Nagaoka equation, and the phase coherence length (⁠$lΦ$⁠) as a function of temperatures is extracted for all samples:

where $lΦ$ is the phase coherence length, $Ψ$ is the digamma function, and $α$ is a coefficient determined by the type of localization. Theoretically, the coherence length is proportional to T^−1/2 and T^−3/4 for 2D and 3D systems, respectively.³⁵ The monotonic decrease of the coherence length with increasing temperature is observed in both bilayer and the control samples, similar to other TI systems.^36,37 The fitting gives $lΦ$ = T^−0.4289 for the control sample, $lΦ$ = T^−0.4271 for the Cr₂O₃-bilyer sample and $lΦ$ = T^−0.4144 for the MgO-bilayer sample, suggesting the dominant dephasing mechanism is e-e interactions in 2D surface states.

Yet, the Cr₂O₃-bilayer sample has the smallest $lΦ$ among all samples whereas the MgO-bilayer and control samples have similar $lΦ$⁠; $lΦ$(T) is plotted in Fig. 9. The noticeable difference in $lΦ$ between Cr₂O₃-bilayer and MgO-bilayer samples is explained by the anisotropic magnetization in Cr₂O₃ and MgO layers. The Cr₂O₃–bilayer sample is paramagnetic with observable coercivity in the out-of-plane direction, whereas the MgO-bilayer is diamagnetic in this orientation. The considerable reduction of $lΦ$ in the Cr₂O₃-bilayer is likely caused by additional inelastic scattering, such as electron-magnon scattering, to the topological surface states proximate to the magnetic layer with a perpendicular anisotropy component.^32,40

The angular dependence of the resistance was measured by rotating the samples 360° in a 7.5 T magnetic field at T = 2 K; the result are presented in Fig. 10. It is known that the WAL induced by 2D surface states is characterized by a sole dependence on the perpendicular component of the applied magnetic field (0° and 180°). A strong dependence of the resistance on angle is observed for the control sample; the data are accurately described by a cosine function. For the bilayer samples, the WAL contribution from the bottom surface states expected to be limited or may coexist with a 3D bulk Fermi surface owing to the magnetic layers. Therefore, both bilayer samples seem to suppress the surface transport channel, as well as the WAL, in the topological surface states. We attribute these transport phenomena to the increased magnetic scattering at the TI/FMI interface.

We have demonstrated the growth of new platforms for topological insulator/ferromagnetic insulator heterostructures by interfacing Bi₂Se₃ with Cr₂O₃ and MgO thin films, showing proximity-induced interfacial magnetization. The lattice misfit of Cr₂O₃/Bi₂Se₃ and MgO/Bi₂Se₃ are ∼16% and ∼39%, respectively, where the critical thickness of pseudomorphic growth is less than one monolayer. The Bi₂Se₃ bilayers are characterized by atomically sharp and reaction-free interfaces. The insulating behavior in the bilayer samples are more pronounced and the onset temperatures are higher as compared to the control sample. The MgO-bilayer sample has the largest resistivity and the highest onset temperature (∼42 K) due to the existence of small misorientations between grains in the Bi₂Se₃ thin films. This could give rise to an increase in resistivity owing to grain boundary scattering. The shortest phase coherent length (⁠$lΦ$⁠) in the Cr₂O₃–bilayer samples is attributed to the additional inelastic scattering between the Cr₂O₃ and Bi₂Se₃ interface with a perpendicular anisotropy component. Our hybrid structures demonstrate the suppression of the 2D surface channel as well as a weakened WAL in the topological surface state. However, stronger perpendicular magnetocrystalline anisotropy is required to observe the time-reversal symmetry breaking and further investigation is needed to establish the exact mechanism for ferromagnetism in these heterostructures.

The authors acknowledge the use of the Analytical Instrumentation Facility (AIF) at North Carolina State University, which is supported by the State of North Carolina and the National Science Foundation. Y. F. Lee partly supported by ECCS-1306400 would like to Dr. Xiahan Sang and Dr. Yang Liu for technical support on the HAADF-STEM, and Sandhyarani Punugupati for the stimulating scientific discussion.