Epitaxial growth and magnetic properties of kagome metal FeSn/elemental ferromagnet heterostructures

Topological semimetals have generated significant interest over the last decade. An emerging family of topological materials are binary kagome metals T_mX_n (T = Fe, Co, Mn; X = Sn, Ge; m:n = 1:1, 3:2, 3:1), which host both topological band features and flatbands near the Fermi level, giving rise to strong anisotropic anomalous and spin Hall effects.^1–10 Of these, FeSn is considered a prototypical kagome metal, which crystallizes in the hexagonal P6/mmm structure with lattice parameters a = 5.297 Å and c = 4.448 Å.^11,12 The structure follows an ABAB stacking sequence where the basal A plane comprises a network of Fe atoms that makes up a kagome pattern with a Sn atom occupying the center of the resulting hexagon, yielding a planar composition of Fe₃Sn as shown in Fig. 1(a). Within each kagome layer, Fe spins couple ferromagnetically. The B plane is made up of monolayer stanene, Sn₂, resulting in an overall composition of FeSn. The neighboring Fe₃Sn planes are coupled antiferromagnetically, giving rise to A-type antiferromagnetism with a bulk Néel temperature of 365 K.^13,14 Angle-resolved photoemission spectroscopy studies on bulk FeSn crystals have revealed the presence of Dirac fermions in surface and bulk states as well as the existence of flatbands near the Fermi level.^15,16 Recently, flatbands were also discovered in FeSn thin films originating from their surface kagome layers.¹⁷ These authors predicted that the surface flatband in FeSn may generate novel spin–orbit torques (SOT) when coupled with ferromagnets (FMs) due to a strong contribution from the Berry curvature arising from the surface state.

To draw on these exotic SOTs and to use them effectively in devices, it is essential to synthesize high-quality crystalline thin films and interface them with other magnetic materials. Molecular beam epitaxy (MBE) has been employed previously to grow single-phase FeSn films directly on insulating perovskite oxide substrates, although realizing continuous films is a challenge. For instance, FeSn films grown on (111)-oriented perovskite LaAlO₃ and SrTiO₃ substrates typically crystallize as discrete 3D islands with a lateral feature size of ∼20–200 nm,^18,19 while continuous films grown on SrTiO₃ (111) have also been reported.²⁰ Epitaxial growth of FeSn has been achieved with a combination of Pt and Ru as nonmagnetic metal buffer layers using magnetron sputtering.²¹ While these buffer layers provide an effective template for epitaxy, Pt is known to exhibit strong spin–orbit coupling, which may complicate efforts to isolate and quantify SOT signatures generated by FeSn and similar topological materials.²² In this work, we show that continuous and epitaxial FeSn films can be grown on Co and Fe underlayers using MBE. A cartoon schematic of the stack is shown in Fig. 1(b). The ferromagnetic (FM) metals act as a buffer layer to facilitate uniform FeSn films on Al₂O₃ substrates, as revealed by electron microscopy, reflection high-energy electron diffraction, and x-ray scattering. Exchange bias is observed in the Fe/FeSn and Co/FeSn bilayer junctions, confirming the presence of exchange interactions between the elemental ferromagnets and the antiferromagnetic FeSn films.

Heterostructures of Fe/FeSn and Co/FeSn were grown via MBE (Omicron modified LAB-10 system, base pressure ∼5 × 10⁻¹⁰ Torr) on (0001)-oriented Al₂O₃ (MTI Corp.) through sublimation of Fe (99.95%, slug, Alfa Aesar) and Co (99.95%, shots, Alfa Aesar) and evaporation of Sn (99.9999%, shots, Alfa Aesar) from Knudsen cells. The corresponding cells were heated to ∼1175 and ∼1400 °C for Fe and Co buffer layer deposition, respectively. During FeSn deposition, the Fe and Sn cells were maintained at ∼1125 and ∼1040 °C, respectively, with codeposition utilized to obtain FeSn. Film compositions were calibrated using a quartz crystal microbalance and Rutherford backscattering spectrometry (RBS). Prior to deposition, the substrates were sonicated in an acetone bath for 15 min, loaded into the chamber, and heated to a temperature of ∼425 °C. The FM layer was deposited on Al₂O₃ (0001) substrates at ∼400 °C for Fe and ∼450 °C for Co, and then the sample was cooled immediately to room temperature. Next, the Fe and Sn cell temperatures were adjusted to obtain an ∼1:1 stoichiometry, and FeSn layer deposition was carried out at ∼450 °C. The sample was cooled following deposition without any postgrowth annealing. The film quality was monitored in situ through reflection high-energy electron diffraction (RHEED) with an operating voltage of 14.5 kV.

X-ray diffraction (XRD) and reflectivity (XRR) were measured using a Rigaku SmartLab diffractometer and analyzed using the software program GenX 3.6 to probe crystallinity, thickness, and interface/surface roughness. RBS was carried out at the Materials Research Laboratory, University of Illinois Urbana-Champaign and the Laboratory for Surface Modification at Rutgers University and analyzed using SIMNRA software package. Scanning electron microscopy (SEM) was performed using a Zeiss Supra 50VP field-emission SEM operated at 14 kV. High-angle annular dark-field (HAADF)-scanning transmission electron microscopy (STEM) with energy-dispersive x-ray spectroscopy (EDS) was performed using an aberration-corrected JEOL NEOARM operating at 200 kV. The images and spectra were recorded using a 2 cm camera length and 27 mrad convergence angle. Transport measurements and magnetometry were carried out using a physical properties measurement system (Quantum Design) with a vibrating sample magnetometry attachment. The diamagnetic signal from the substrate was not subtracted as the magnetization data were dominated by the Fe/Co ferromagnetic signal at the measurement fields used.

Previous attempts to grow kagome metals directly on oxide substrates have often resulted in island growth and noncontinuous films.^18,19,23 While Pt and Ru buffer layers have been employed to mitigate the propensity for island growth,²¹ we use Fe and Co as buffer layers, as both materials are also ferromagnetic. These 3d transition metals have been grown epitaxially on oxide substrates including MgO and Al₂O₃.^24–28 For instance, Shiratsuchi and coauthors showed that under certain conditions, epitaxial Fe (110) can grow as quasi-2D continuous films on Al₂O₃ (0001) substrates.²⁵ Co (0001) films on Al₂O₃ (0001) have also been grown epitaxially, despite a large mismatch of ∼9% between them.²⁷ Prior to the deposition of our bilayer films, the FeSn stoichiometry was calibrated by altering the MBE fluxes and measuring the film composition of monolithic FeSn films grown directly on MgO (111) substrates using RBS, yielding general compositions of Fe_1+xSn_1−x (x = 0–0.07). The calibration films became more iron-rich in a linear fashion with increasing Fe flux as expected. The flux rates from these calibration growths were then used in the synthesis of the bilayer FM/FeSn structures. Interestingly, the bilayer films with different stoichiometries had a noticeably different surface morphology. Figure S1 in the supplementary material shows electron micrographs from Fe/FeSn films with nominal compositions of Fe_0.99Sn_1.01, Fe_1.03Sn_0.97, and Fe_1.07Sn_0.93 respectively, based on the calibration data obtained from the monolithic FeSn films on MgO. Fe_0.99Sn_1.01 films showed a large volume of granular particles that were 200–500 nm in size on the surface. Upon increasing the Fe content in the film, these particles decrease in quantity and at a composition of Fe_1.07Sn_0.93, the particles are largely absent from the surface. Based on these observations, and previous reports of Sn dewetting from films,^29,30 we hypothesize that these particles comprise Sn (or a phase that is predominately Sn) that is not incorporated into the films. Informed by these calibration depositions, all the following measurements were carried out on FeSn samples with an estimated composition of Fe_1.05±0.02Sn_0.95±0.02.

Figure 2 shows RHEED images obtained after the deposition of each layer juxtaposed with a schematic of their hypothesized surface atomic configuration to show the epitaxial structure of the bilayer films. Oxygen atoms on an O-terminated Al₂O₃ (0001) surface form an irregular hexagonal pattern where four sides have length s₁ = 2.87 Å and two sides have length s₂ = 2.53 Å. Fe atoms in the (110) plane also form an irregular hexagonal pattern with four sides of length s₁ = 2.48 Å and two sides of length s₂ = 2.87 Å, and in the Co (0001) plane, the atoms form a regular hexagon with sides of length s = 2.51 Å. Similarly, the atoms in the FeSn (0001) plane form a regular hexagon with sides of length s = 2.65 Å. These repeating hexagons stack on top of each other, leading to epitaxial growth of Fe/Co and FeSn on Al₂O₃. Images of RHEED patterns obtained after growth of each layer show clear streaks, indicating that the surfaces are smooth. Additionally, the same patterns from the FeSn layer were observed with every 60° in-plane rotation, suggesting a sixfold rotational symmetry as expected from the hexagonal lattice.

The lattice parameters and epitaxial orientation of the films were obtained from x-ray diffraction (XRD). Figures 3(a) and 3(b) show 2θ-ω XRD scans from three Fe/FeSn (F1, F2, and F3) and Co/FeSn (C1, C2, and C3) bilayers of varying FeSn thicknesses. The scattering vector was fixed perpendicular to the film surface, and excluding the substrate peaks, only 000l, hh0, and 000l reflections were observed from FeSn, Fe, and Co, respectively. The presence of significant Laue oscillations about the Fe, Co, and FeSn Bragg peaks points to sharp interfaces between each layer. The thickness fringes arising from the Co and Fe layers are convoluted with the FeSn 0002 Bragg peak positions, especially in films with thinner FeSn layers. To extract quantitative information from the diffraction data, we simulated the 2θ-ω patterns from these heterostructures using GenX.³¹ Through comparisons between the measured and simulated data, the thickness and c axis parameters of each layer are determined. The parameters obtained for Fe (d₁₁₀ = 2.03 Å, corresponding to a = 2.87 Å) and Co (c = 4.07 Å) are within 0.1% and 0.25% of bulk values.³² The average c axis parameter for the FeSn layer in Fe/FeSn films is 4.44 ± 0.01 Å and in Co/FeSn films is 4.40 ± 0.03 Å. For comparison, the c axis length for bulk FeSn is 4.448 Å.¹¹ The origin of these deviations from bulk c axis values is likely due to the small degree of off-stoichiometry within the films. Noteworthily, in Co/FeSn films, the obtained lattice parameters a_FeSn and c_FeSn of sub-5 nm films are different than those of 40 nm films: a_4 nm = 5.402 Å and a_40 nm = 5.429 Å, whereas c_4 nm = 4.37 Å and c_40 nm = 4.42 Å. We attribute the decreased a axis and c axis parameter in ultrathin FeSn films on Co to chemical intermixing between FeSn and Co layers leading to Co incorporation in FeSn near the interface. As Meier and coauthors have shown,¹³ in Fe_1−xCo_xSn, both a axis and c axis parameters decrease with increasing Co concentration. In contrast, we observe minimal change in lattice parameters for 8 and 40 nm FeSn films on Fe.

To determine if the films are epitaxial, in-plane diffraction measurements were performed. ϕ scans along a fixed χ angle, shown in Figs. 3(e) and 3(f), reveal six distinct in-plane $2 1 ¯ 1 ¯0$ Bragg peaks from the FeSn layer in Fe/FeSn and Co/FeSn, respectively. The peaks are evenly spaced 60° apart from each other and are spaced ∼30° (±5°) apart from the six $2 1 ¯ 1 ¯0$ Bragg peaks of Al₂O₃. As shown schematically in Fig. S2 in the supplementary material, this 30° rotation provides better crystallographic alignment between Al₂O₃ (0001) and FeSn (0001). A small deviation, ∼5°, is observed between the ideal 30° rotation, the origin of which is unknown but likely arises from the presence of the Fe or Co buffer layer. This confirms the epitaxial nature of the FeSn layers with respect to the substrate, and by correlation, likely the ferromagnetic layers Fe and Co as well. The epitaxial relationship is as follows: Al₂O₃ [0001] || FeSn [0001] and Al₂O₃ [ $2 1 ¯ 1 ¯0$] || FeSn [ $10 1 ¯0$]. Due to the small thicknesses of the ferromagnetic layers, no in-plane Bragg peaks were observed from either Co or Fe. The full width at half-maximum (FWHM) of the FeSn $2 1 ¯ 1 ¯0$ peaks from the film grown on Co is 1.9°, which is smaller than that of the film on Fe at 6.9°, indicating a better degree of in-plane crystallinity for the film grown on the Co buffer. This is likely due to the 5% mismatch between Co and FeSn hexagons compared to 7% between Fe and FeSn hexagons, as well as the better symmetry matching between FeSn and the Co (0001) plane compared to the Fe (110) plane. However, this is not reflected in the rocking curve measurements from the 0002 FeSn peaks as both Fe/FeSn and Co/FeSn films have comparable FWHM values of 0.09° and 0.07°, respectively, for FeSn films ∼10 nm in thicknesses (Fig. S3 in the supplementary material). For comparison, the Al₂O₃ 0006 rocking curve had a FWHM of 0.03° under the same measurement conditions, which points to both films having some degree of mosaicity. As the FeSn film thickness is increased, the rocking curve FWHM values also increase with 40-nm thick films exhibiting FWHM on the order of 0.2°. Figure S3 in the supplementary material shows a compilation of the FWHM of films plotted against their thicknesses and a clear trend is observed, indicating that the film crystalline quality declines slightly at larger thicknesses. While the films are epitaxial, they do not appear to be coherently strained to the Al₂O₃ substrate as the RHEED patterns show clear differences in k_x spacings between the specular and scattered streaks of each layer within the heterostructure (Fig. S4 in the supplementary material).

The bilayer thickness and roughness values were further examined using XRR on the same set of films. The roughness of each layer in all films was found to be <1 nm, while the obtained scattering length densities of all layers were within 5% of their theoretical values. There is good agreement between the thicknesses calculated from the XRD simulations and XRR fits (as shown in Table S1 in the supplementary material), implying that the films are crystalline throughout their entire thickness. It is worth noting that a ∼1 nm thick FeSnO_x passive oxide layer was used in the model to account for surface oxidation, which led to better fits. Resistivity measurements from the bilayer samples exhibit metallic temperature dependence (as shown in Fig. S5 in the supplementary material), a result that is consistent with laterally continuous films.

Cross-sectional STEM and EDS analyses were conducted to probe the interface in a representative bilayer Co/FeSn film. Figure 4(a) shows a HAADF-STEM image along the $[ 10 1 ¯ 0]$ direction, revealing two distinct layers atop the insulating Al₂O₃ substrate. In this sample, the Co and FeSn layers are ∼7 and ∼35 nm, respectively, featuring sharp interfaces between Al₂O₃–Co and Co–FeSn, as illustrated in Fig. S6 in the supplementary material. The high-resolution image of the FeSn layer shown in Fig. 4(b) confirms the expected ABAB stacking of the Fe₃Sn and Sn₂ planes in FeSn. A schematic of the crystal structure along the $[ 10 1 ¯ 0]$ direction is shown in Fig. 4(c). To further understand the distribution of Al, O, Co, Fe, and Sn within the heterostructure and to visualize the intermixing at the Co–FeSn interface, EDS spectra were acquired, and corresponding elemental maps are shown in Fig. 4(e). An EDS line profile along the green line drawn in Fig. 4(e) is shown in Fig. 4(d). The Al₂O₃–Co interface is very sharp with limited intermixing and minimal roughness. However, at the Co–FeSn interface, some intermixing of Co into FeSn is present in the first 4 nm of the FeSn layer. This observation falls in line with the c axis shrinkage noticed from the XRD measurements of the ultrathin FeSn films on Co. Interestingly, an additional layer with enhanced contrast was observed in the high-resolution image of the Al₂O₃–Co interface (Fig. S6 in the supplementary material), which originates from Sn atoms, as confirmed by EDS. The Sn diffusion and accumulation at the interface does not occur by way of uniform diffusion of Sn through the Co layer but through pillar formation in the Co layer, as shown in the EDS map in Fig. S6 in the supplementary material. These narrow Sn pillars span the entire thickness of the Co layer and disperse at the interface with the Al₂O₃ substrate. Away from the pillars, the line profile and the EDS map both indicate that Sn is exclusively concentrated at the interface and is not dispersed within the Co layers. This, supported by the streaky nature of our RHEED patterns and significant Laue oscillations from the Bragg peaks, indicates that the films are smooth and continuous with well-defined interfaces.

We performed magnetometry on the Fe (7 nm)/FeSn (38 nm) and Co (8 nm)/FeSn (40 nm) samples, the results of which are displayed in Fig. 5. At 300 K, the saturation magnetizations (M_s) for the two films were found to be ∼1675 and ∼1390 kA/m, and coercive field strength (μ₀H_c) to be 9.4 and 14.4 mT, respectively. These values are consistent with previous reports in the literature for monolithic Fe and Co films,^33,34 indicating that the FeSn layers are not contributing a significant magnetization signal as expected for antiferromagnets. To elucidate the interaction between the ferromagnetic metals and antiferromagnetic FeSn layers, isothermal magnetization measurements as a function of applied field were conducted at various temperatures. The samples were field-cooled from 400 K, which is well above T_N for FeSn, under an in-plane applied field. Field scans were conducted twice at each temperature, once under positive field (2 T) cooling and again under negative field cooling (−2 T).

Average exchange bias field and coercive field were calculated after ±FC, plotted against temperature from (c) Fe/FeSn and (d) Co/FeSn. At 300 K, little-to-no loop shift was observed; however, at low temperatures, a negative exchange bias was observed when the samples were field-cooled under a positive field, as shown in Figs. 5(a) and 5(b) for both Fe/FeSn and Co/FeSn, respectively. A positive exchange bias was also observed when the samples were field-cooled under a negative field as shown in Fig. S7 in the supplementary material. The shift in hysteresis loops was accompanied by an increase in H_C, further pointing to an exchange interaction between the ferromagnetic Fe and Co layers and the antiferromagnetic FeSn layer.^35, Figures 5(c) and 5(d) show the temperature dependence of exchange bias field (H_EB) and coercive fields (H_C) in both Fe- and Co-based heterostructures. H_EB was calculated by averaging the magnitude of the positive and negative exchange biases observed with ±2 T field cooling. In both films, H_EB and H_C increase with decreasing temperature. Both H_C and H_EB are larger in magnitude for Fe/FeSn than Co/FeSn. One possible explanation could be the intermixing at the Co–FeSn interface. In bulk Fe_1−xCo_xSn random alloys, the ordered antiferromagnetic moments undergo a reorientation with substitution of Co in the lattice – from planar to tilted and then to axial antiferromagnetism. At x = 1 (CoSn), the crystal becomes paramagnetic.¹² Hence, it is possible that the moments in the intermixed phase at the Co–FeSn interface are partially reoriented from planar to tilted/axial, leading to a decrease in exchange bias. The H_C and H_EB values in both Co/FeSn and Fe/FeSn are lower in comparison to the reported values for permalloy/FeSn.¹⁶ 

We have demonstrated the growth of epitaxial and continuous thin films of FeSn by utilizing elemental ferromagnetic metal buffer layers, BCC Fe (110), and HCP Co (0001) on Al₂O₃ (0001) substrates. By growing films that are slightly Sn-deficient, we show that heterostructures with smooth surfaces, well-defined interfaces, and a high degree of crystallinity can be obtained. The FeSn films are antiferromagnetic as confirmed by the presence of exchange bias in both Fe/FeSn and Co/FeSn bilayers. We predict that other common ferromagnets with FCC and BCC structures like Ni and permalloy could also be used as a buffer layer to grow FeSn and other kagome crystals, as the (111) and (110) surfaces, respectively, offer the hexagonal imprint required for epitaxy of these materials. The ability to form epitaxial ferromagnetic heterostructures with topological kagome materials opens the door to studies of spin–orbit torques and spin transport in these systems.

See the supplementary material for additional material characterization through scanning electron microscopy, x-ray diffraction, temperature-dependent resistivity, scanning transmission electron microscopy and energy dispersive x-ray spectroscopy, and dc magnetometry. We also provide a schematic of the crystallographic alignment between FeSn and Al₂O₃.

The authors at Drexel and UIUC were supported by the National Science Foundation under Grant No. ECCS-2031870, except M.-W.Y., who was supported by the National Science Foundation through the University of Illinois Urbana-Champaign Materials Research Science and Engineering Center Grant No. DMR-1720633. R.K.R. and E.A.S. were supported by the National Science Foundation under Grant No. DMR-2122102. The authors would like to thank Hussein Hijazi at Rutgers University for performing timely RBS experiments and Dr. Craig Johnson of the Materials Characterization Core (MCC) at Drexel University for the assistance in preparing samples for and performing TEM. They also acknowledge the characterization facilities at MCC of the College of Engineering at Drexel University; the diffractometer was acquired with support from the National Science Foundation under Grant No. DMR-1040166. The acquisition of the PPMS used for transport and magnetometry measurements was supported by the U.S. Army Research Office under Grant No. W911NF-11-1-0283.

The authors have no conflicts to disclose.

Prajwal M. Laxmeesha: Formal analysis (equal); Investigation (lead); Writing – original draft (lead). Tessa D. Tucker: Investigation (equal); Writing – review & editing (supporting). Rajeev Kumar Rai: Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Shuchen Li: Investigation (equal); Writing – review & editing (supporting). Myoung-Woo Yoo: Investigation (equal); Writing – review & editing (equal). Eric A. Stach: Supervision (equal); Writing – review & editing (equal). Axel Hoffmann: Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal). Steven J. May: Conceptualization (lead); Formal analysis (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available at https://doi.org/10.13011/m3-q18q-hx55.