Two-dimensional growth of conductive ultra-thin Sn films on insulating substrate with an Fe buffer layer

Topological materials have attracted emerging interest as a new class of condensed matters, which host exotic transport phenomena originating from a non-trivial electronic structure.^1,2 Thin film synthesis of topological materials is a fundamental subject for inducing their intriguing functionalities such as quantum spin Hall effect (QSHE) in a two-dimensional (2D) topological insulator (TI) quantum well³ and quantum anomalous Hall effect (QAHE) in a magnetically doped three-dimensional (3D) TI film⁴ and heterostructures.⁵ Among a variety of TI materials, the α-Sn (gray tin) and two-dimensional stanene are intriguing candidates in a thin film form.⁶ The α-Sn is known as a zero-gap semiconductor so that transformation to 3D-TI is theoretically proposed to occur owing to the opening of a non-trivial gap induced by the quantum confinement effect⁷ or the epitaxial strain,^6,8,9 similar to HgTe.¹⁰ In addition, a large spin–orbit coupling of α-Sn causes highly effective spin-charge conversion (Rashba–Edelstein effect), which was detected at the heterointerface with a ferromagnet.¹¹ Besides, the ultra-thin limit of α-Sn (a few layer stanene) is also a platform to investigate the 2D superconductivity through substrate engineering.¹² Considering these intriguing quantum phenomena in α-Sn, heterostructures of Sn with a well-controlled thickness will provide a unique arena for experimental investigations of the emergent topological phenomena.

Despite the theoretical predications^6,13,14 and spectroscopic measurements on the electronic structure^7–9,12 of the α-Sn film grown on a wide variety of substrates, only a few have reported the electrical transport properties of the α-Sn film with a thickness as thin as a few nanometers.^12,15,16 This may be strongly relevant to the issue that the thin-film growth of Sn on an insulating substrate leads to the inhomogeneous and 3D island-based growth.¹⁷ A phase transition temperature from a low-temperature phase α-Sn (diamond structure) to a high-temperature phase β-Sn (tetragonal) occurring at around 13 °C (Ref. 18) and formation of the intermixing alloys¹⁹ probably account for the poor wettability on the insulating substrate. Thus, the low-temperature growth below 0 °C has been usually implemented to the growth of α-Sn thin films. An alternative way to stabilize the atomically thin Sn film is by employing conductive substrates such as Bi₂Te₃,²⁰ Ag,²¹ Cu,²² and Au,²³ which allow epitaxial growth of 2D stanene. Herein, the growth of ultra-thin Sn films on metal substrates at a milder temperature condition motivated us to investigate the effectiveness of the metallic buffer layer on the insulating substrates. In this study, we selected Fe as a buffer layer owing to the epitaxial growth of Fe (110) on Al₂O₃ (0001).²⁴ By the insertion of the Fe buffer layer on the insulating Al₂O₃ substrate, we obtained the 2D-layered ultra-thin Sn films at room temperature using molecular-beam epitaxy (MBE). The structural characterizations and transport measurements in the Fe-buffered Sn thin films demonstrate that the Fe buffer layer supports the 2D layer growth below 1.0 nm (corresponding to three atomic layers) by suppression of the cohesion of Sn films.

We grew the ultra-thin Sn films with the insertion of the Fe buffer layer on the Al₂O₃ (0001) substrate by MBE at room temperature. Fe and Sn beam equivalent pressures were kept at 3.2 × 10⁻⁶ Pa and 1.0 × 10⁻⁶ Pa, respectively, to obtain the growth rate of the Fe buffer and Sn films of about 3.0 Å/min and 0.5 Å/min. Lattice constants were determined by the x-ray diffraction (XRD) pattern. Surface morphology and flatness were characterized using in situ reflection high energy electron diffraction (RHEED) and ex situ atomic force microscopy (AFM). Longitudinal and Hall resistances of the samples were measured as functions of temperature and magnetic field in the physical property measurement system (PPMS, Quantum Design, Inc.) with a standard five-terminal configuration.

Figure 1(a) shows XRD patterns of 2-nm-thick Fe directly on Al₂O₃ (denoted as 2-Fe, bottom panel) and 0.5-nm-thick Sn on 2-Fe (denoted as 0.5-Sn/2-Fe, top panel) grown on Al₂O₃ (0001), which were obtained by Cu Kα excitation. The insets schematically show the cross sectional sample geometry. We observed a Fe (110) diffraction peak with the clear Laue thickness fringes, indicating a (110) oriented Fe film with a flat surface morphology. As can be seen, the diffraction peak of Fe (110) remains at the same position regardless the presence of top 0.5-Sn films. Judging from this result, the intermixing between Sn and Fe hardly occurs at room temperature during the deposition. As a reference, we prepared thick Sn films directly on Al₂O₃ (0001). Only the diffraction peak of β-Sn (200) appeared in the film thickness range from 2 nm to 6 nm with a rough surface morphology of about 4 nm (see Fig. S1 in the supplementary material). This verifies that the β-Sn phase is more preferable than the α-Sn phase on Al₂O₃ at room-temperature growth. In addition, 3D island growth becomes pronounced when a thick Sn film of about 6 nm is grown, showing a similar behavior with the previous study on the SrTiO₃ substrate.¹⁷ For the thinner Sn film with a thickness less than 2 nm, it is hard to determine the preferable phase due to the weak XRD peak intensity (Fig. S1). Nonetheless, it is apparent that the fabrication of monolayer α-Sn or 2D stanene on Al₂O₃ would be difficult experimentally by room-temperature growth, despite the theoretical proposal that a strong bonding between Sn and the top-most Al atom stabilizes monolayer α-Sn or 2D stanene on Al₂O₃.¹⁴ In contrast to the 3D island growth of a few-nm-thick Sn film on Al₂O₃ (0001), 2 nm and 4 nm Fe were grown with a very flat surface [Fig. 2(j) and Fig. S5(j)] on the same substrate. On the flat surface of Fe (110), we expect both the suppression of the 3D island growth of β-Sn (200) and the stabilization of the 2D growth of α-Sn (111) driven by the similar in-plane lattice alignment as discussed in supplementary material, Fig. S4.

To determine the electronic states of the ultra-thin Sn films, we applied the surface sensitive core-level photoemission spectroscopy for the 2-Fe and 0.5-Sn/2-Fe films using the He-IIα light source with photon energy hν = 40.81 eV, as shown in Fig. 1(b). First, appearance of the Sn peaks only in the spectrum of the 0.5-Sn/2-Fe bilayer sample (blue solid curve), by comparing with that of the 2-Fe single layer sample (red solid curve), explicitly indicates the coverage of Sn on the Fe (110) buffer. We performed fitting for the 0.5-Sn/2-Fe data with the standard least-squares-fitting method. The peak positions and a Gaussian width for the Sn 4d components were used as free parameters, and a spin–orbit splitting energy of 1.08 eV was fixed in the fitting procedure.²⁵ The fitting with 0.5-Sn/2-Fe reveals that there are two binding states for the Sn atoms. In particular, two peaks depicted by black broken curves correspond to Sn 4d 3/2 and 4d 5/2 spin–orbit satellites at kinetic energies of 11.38 eV and 12.46 eV, corresponding to the binding energies of 25.01 eV and 23.93 eV, respectively, which is consistent with the previous investigation of 4d core levels in bulk Sn²⁶ and in the 1/3-atomic-layer-thick Sn film with a honeycomb lattice grown on Ge (111) (Refs. 27 and 28) and Si (111) (Ref. 29). Such an agreement in the photoemission spectrum implies that the Sn film formed on the Fe buffer layer is likely to be an α-phase. Besides, an additional broader component at a higher binding energy of ∼1.1 eV from the above peaks appears as shown with a green shaded area. We assigned it as a chemical shift that originates from different chemical environments of Sn. We exclude the possibility of oxidization in our in situ photoemission spectroscopy measurements because of the absence of the O 2s peak at around 22–23 eV (Ref. 30). In addition, the chemical shifts expected in SnO₂ and SnO are around 2.0 eV and 1.3 eV, respectively,³¹ both of which are larger than the shift observed in this study (∼1.1 eV). Although the thickness of Sn is roughly 1.5 atomic layers, the interaction at the interface of Sn and Fe might induce the chemical shift by bond formation or charge transfer.³² 

The surface structure was characterized by RHEED [Figs. 2(a)–2(d)] and AFM with line profiles [Figs. 2(e)–2(h)] for 0.5-Sn, 2-Fe on Al₂O₃ (0001), and Sn/Fe bilayers with Sn thicknesses of 0.5 nm and 1.5 nm. The RHEED pattern was taken along the azimuth of Al₂O₃[1–100]. First, the 0.5-Sn and 2-Fe directly grown on Al₂O₃ were compared as a reference for direct growth. The RHEED patterns are completely different; the 0.5-Sn [Fig. 2(a)] shows a polycrystalline ring pattern, while the 2-Fe exhibits a spotty pattern with streaky lines [Fig. 2(b)]. In fact, the root mean square (rms) of the surface roughness for the 0.5-Sn film directly grown on Al₂O₃ [Fig. 2(e)] is comparable to the estimated thickness of the Sn supplement, indicating the 3D island growth mode. By contrast, the 2-Fe grown on Al₂O₃ is atomically flat with an rms as small as 0.13 nm and a very small surface thickness fluctuation, as presented in Fig. 2(f), which allows it to be used as a buffer layer. After deposition of Sn on the 2-Fe buffer, the spot with the streaky pattern becomes weaker for 0.5-Sn [Fig. 2(c)] and completely vanishes for 1.5-Sn [Fig. 2(d)]. This degradation was observed in the surface morphology measured by AFM; the rms of the 0.5-Sn surface remains at a comparable value to that of Fe, while with further increase in d_Sn to 1.5 nm, the rms of the 1.5-Sn/2-Fe becomes as large as 1.3 nm, which is comparable to the estimated thickness of Sn on the Fe buffer.

With these structural characterizations, we discuss the growth mode of Sn on Al₂O₃ and the Fe buffer layer, as schematically illustrated in Figs. 2(i)–2(l). The 0.5-nm-thick Sn directly grown on Al₂O₃ shows the 3D growth mode without in-plane percolation [Fig. 2(i)]. It should be noted that the two-terminal resistance of this film was beyond the measurement limit of the digital multimeter. This result is not surprising because of the poor wettability of Sn on the oxide substrate such as involving SrTiO₃ (Ref. 17). For the 0.5-Sn on the 2-nm-thick Fe buffer layer, the 2D flat surface of the 2-Fe [Fig. 2(j)] is kept as exemplified by both the spot with streaky RHEED pattern [Fig. 2(c)] and the AFM image [Fig. 2(g)]. Such a contrast of 0.5-Sn/Al₂O₃ and 0.5-Sn/2-Fe clearly indicates that the Fe buffer layer enables the subsequent 2D growth of Sn films [Fig. 2(k)] below the critical thickness d_c^Sn (∼1.0 nm, the definition of d_c^Sn will be explained later in Fig. 3). The drastic change in the surface morphology with increasing d_Sn in Fig. 2(h) implies that the growth mode changes from 2D [Fig. 2(k)] to 3D [Fig. 2(l)].

The comparative experiments using the 4-nm-thick Fe buffer (d_Sn-Sn/4-Fe) have revealed the same trend; the 2D growth mode assisted by the insertion of Fe persists for d_Sn ≤ 1.0 nm (see Fig. S5 in the supplementary material). Thus, it can be inferred that the Fe buffer layer provides a good wettability originating from a better affinity of Sn with Fe rather than that with oxide substrates. At the initial growth stage on Fe, we speculate that Sn atoms are adsorbed onto the surface and bonded with Fe as exemplified by the chemical shift in Fig. 1(b), leading to 2D-layered growth mode. Then, the next Sn atoms are adsorbed and bonded onto the first Sn layer, inducing 2D growth of the Sn thin films. When the thickness exceeds d_c^Sn, growth mode dramatically changes to island-based growth. Note that the 2.0-nm-thick Sn layer grown directly on the Al₂O₃ substrate forms the β-phase [Fig. S1(a)], which is absent on the Fe buffer layer (Fig. S2). Such a contrast in the XRD patterns demonstrates that the main role of the Fe buffer layer is to suppress the formation of the β-phase.

The changes in the growth mode of the Sn thin film can also be characterized by electrical properties of d_Sn-Sn/Fe bilayers. Figure 3 displays the total sample thickness (d_tot) dependence of sheet conductance (1/R_xx) at 300 K. The sheet conductance of pure Fe (solid green circles) linearly increases as its thickness increases, indicating a uniform distribution of electric current along the thickness. By the linear fitting (the green broken line), we extract the resistivity of the Fe film (ρ_Fe) to be about 2.6 × 10⁻⁵ Ω cm, which is comparable to that in the previous study.³³ It is worth noting that the crossing point of the linear fitting line and the bottom axis corresponds to the existence of a low conductive dead layer within a finite value of thickness (d₀^Fe) roughly at 1.5 nm. When Sn is deposited on the 2-Fe buffer, the 1/R_xx of the d_Sn-Sn/2-Fe (red squares) linearly increases with d_Sn up to the saturation value. Here, the critical Sn thickness (d_c^Sn) is defined from the intersection point of two fitting lines (one part is the proportional increase and the other is the horizontal saturation plateau). The saturation behavior of 1/R_xx above the d_c^Sn indicates that the excess Sn layer does not contribute to the electrical conduction, which agrees well with the grain-like rough surface structure discussed in Fig. 2 [as shown by 1.5-Sn/2-Fe in Figs. 2(h) and 2(l)]. Similarly, d_tot dependence of the 1/R_xx of Sn on the 4-Fe buffer layer (d_Sn-Sn/4-Fe, blue squares) shows consistent results (see also Fig. S5). The Fe-buffer-layer-thickness (d_Fe) dependent d_c^Sn is presented in the inset of Fig. 3, giving an average d_c^Sn value of about 1.0 nm. These observations clearly evidence that the ultra-thin Sn films below d_c^Sn ∼1.0 nm are under the regime of 2D growth mode and host a good electrical conduction.

We consider the model that the Sn and Fe layers contribute to conductance in parallel, and total sheet conductance (1/R_xx) is a sum of conductances of Sn and Fe individual layers.^34,35 The relationship between 1/R_xx and the film total thickness d_tot (d_tot = d_Sn + d_Fe) for Sn/Fe bilayers can be expressed as follows:

where ρ_Fe and ρ_Sn are the resistivities of Fe and Sn layers, respectively, and d_Fe and d_Sn are the thicknesses of Fe and Sn layers, respectively. Equation (1) holds only when d_Sn ≤ d_c^Sn, where the linear increase in 1/R_xx of Sn/Fe bilayers constitutes the evidence of 2D growth of the uniform ultra-thin Sn film. The slope of such a linear relationship represents the average conductivity of ultra-thin Sn films, (ρ_Sn)⁻¹. The extracted value of ρ_Sn with the order of 10⁻⁵ Ω cm is about one order smaller than the value reported for α-Sn in previous works.^16,36,37 In contrast, β-Sn is a superconductor with a low resistivity at the magnitude order of 10⁻⁶ Ω cm at room temperature.³⁸ Judging from this behavior, both 0.5-nm-thick Sn on 2-Fe and 4-Fe are close to the α-Sn phase rather than to β-Sn.

Figures 4(a) and 4(b) show temperature dependent sheet resistance (R_xx) for the d_Sn-Sn/2-Fe and the d_Sn-Sn/4-Fe bilayers, respectively. With increasing d_Sn from 0.0 nm to 3.0 nm, a dramatic reduction in R_xx was observed in the whole temperature range of 2–300 K owing to the increase in Sn conduction. For 1.0-, 1.5-, and 3.0-Sn on the 2-Fe buffer, these three samples at d_Sn ≥ d_c^Sn show similar R–T curves due to the small contribution of the additional 3D Sn layer. The resistivity of Sn films (ρ_Sn) can be extracted by Eq. (1) at each temperature (see the details in supplementary material, Fig. S6). Figure 4(c) shows temperature dependence of resistivity ρ_Sn exhibiting weak metallic behavior, which also proves the suppression of β-Sn formation.

We also performed Hall effect measurements to evaluate the interaction between the ultra-thin Sn film and the ferromagnetic Fe buffer layer and discuss the current distribution in the Sn/Fe bilayers. Figure 5(a) shows a schematic of the Hall effect measurement configuration for the Sn/Fe bilayer and an equivalent circuit for the longitudinal current in the xz plane [Fig. 5(b)] and for the anomalous Hall current in the yz plane [Fig. 5(c)], where x is the direction of electric current and y is that of Hall voltage. Here, we consider that the shunting and short-circuit effects^39,40 play a role in the measurement of Hall resistance R_yx [see the results for the d_Sn-Sn/2-Fe in Fig. 5(d) and details in Fig. S7 in the supplementary material]. Thus, the Hall voltage (V_yx) is given by the voltage drop across R_Sn (resistance of the Sn layer), which is driven by the anomalous Hall voltage (E_Fe^A) of the ferromagnetic Fe layer via R_Fe (resistance of the Fe layer). The Hall resistance, defined as R_yx = V_yx/I_xx, can be written as follows (the detailed deduction of the formula is addressed in supplementary material, Sec. S5):

where R_yx and R_Fe^A represent the Hall resistance of the bilayer and the anomalous Hall resistance of the Fe layer, respectively. I_A is the current in the equivalent circuit in the yz plane. Because the ordinary term is relatively small and can be neglected, we only take the anomalous term into account for analysis. ρ_Fe and ρ_Sn are considered to be constant under a variation of the magnetic field as the magnetoresistance was negligibly small. Figure 5(e) shows the d_Sn dependence of the anomalous Hall resistance R_A (which is defined as the value of R_yx at B = 3 T) and a tangent of the Hall angle R_A/R_xx for the d_Sn-Sn/2-Fe and the d_Sn-Sn/4-Fe bilayers. The comparably large R_A in the Fe layer with previous studies^41,42 decreases with increasing d_Sn owing to the parasitic contribution of the nonmagnetic Sn layer. The simulations fit well with the experimental data for d_Sn below d_c^Sn, whereas they deviate from the fitting line for d_Sn > d_c^Sn (dashed line in the shaded area) because of the 2D to 3D growth mode transition. We conclude that shunting and short-circuit effects originating from the conductive Sn layer are, indeed, responsible for the reduction in the observed R_yx with increasing d_Sn. Despite the drastic reduction in R_A, a tangent of the Hall angle R_A/R_xx is kept constant upon increase in the nonmagnetic layer thickness (See R_xx data in Fig. S8). This result suggests that the longitudinal current in the whole thickness region of the bilayer contributes to the anomalous Hall effect with a large Hall angle. Stacking of the 2D Sn and ferromagnetic layers can be a platform to explore an intriguing magnetic proximity effect at the interface between conductive nonmagnetic and ferromagnetic layers.

In conclusion, we have demonstrated the 2D growth of the conductive ultra-thin Sn films at room temperature by the insertion of an Fe buffer layer on insulating Al₂O₃ (0001) substrates. We found that the Fe buffer layer suppresses the island growth of β-Sn and assists 2D growth mode of the conductive Sn layer. Investigation of the thickness dependence of the Sn films on the surface structure and the sheet resistance reveals a transition of growth mode from 2D layer-based to 3D island-based with increasing d_Sn at the critical thickness of d_c^Sn about 1.0 nm. By fitting the thickness dependent sheet conductance (1/R_xx ∼ d_Sn), the 2D grown ultra-thin Sn film possesses a resistivity ρ_Sn of about 10⁻⁵ Ω cm at 300 K, which is rather consistent with α-Sn in previous reports. Further optimization of the 2D growth of conductive ultra-thin Sn, i.e., low-temperature growth or different buffer layers, may offer a platform to stabilize α-Sn or 2D stanene films. Their topological properties combined with ferromagnetic bilayers will provide a good arena to exemplify fascinating interfacial phenomena toward spintronic device applications.

See the supplementary material for the additional information about the thin film growth procedure and characterization, XRD and AFM results of Sn films directly grown on Al₂O₃ (0001) (Fig. S1), XRD patterns of Sn grown on the Fe buffer (Fig. S2), the spectrum from photoemission spectroscopy measurement around the Fermi level for 2-Fe and 0.5-Sn/2-Fe (Fig. S3), lattice arrangement of the Sn growth on Al₂O₃ with an Fe buffer layer (Fig. S4), RHEED and AFM images of the Al₂O₃ (0001) substrates and the Sn films grown on the 4-nm-thick Fe buffer layer (Fig. S5), analysis of the thickness and temperature dependent electrical properties of Sn/Fe bilayers (Fig. S6), analysis of the Hall measurement in Sn/Fe bilayers (Fig. S7), and d_Sn dependence of the longitudinal resistance R_xx (Fig. S8).