2.1 Structural Properties

MBE was employed to grow ≈15 nm thick FGT films on epitaxial graphene on SiC (0001) templates synthesized by surface graphitization.^[22] Rutherford backscattering spectrometry (RBS) analyses (see Figure S1 in the Supporting Information) revealed the films to exhibit an average composition close to x = 0, namely Fe_4.8GeTe₂. This is similar to what has been reported for bulk single crystals synthesized by chemical vapor transport.^{[7, 23]} In situ growth monitoring by reflection high-energy electron diffraction (RHEED) showed a pattern composed of isolated streaks which indicates growth of a film with a smooth surface (Figure 1a, bottom panel). The position of the streaks with respect to the pattern measured for epitaxial graphene on SiC(0001) before growth (for SiC(0001), the in-plane (IP) lattice parameter is 3.08 Å)^[24] attested the evolution to a larger in-plane lattice constant around 4.06 Å, as expected for FGT formation.^[9]

To study the in-plane structure in more detail, synchrotron-based grazing incidence diffraction (GID) was employed. Figure 1b shows the in-plane intensity distribution plotted in hexagonal coordinates according to the symmetry of the SiC(0001) substrate. This map was obtained for the 15 nm FGT film covered by a 5 nm thick Te capping layer, which was used to reduce FGT surface oxidation upon prolonged air exposure. For convenience, we use the four-component vector notation for hexagonal symmetry: (H K. L) = (H K −[H + K] L). The only substrate contribution within the investigated area in reciprocal space appears as a single (2$$\bar{1}$$.0)-type reflection. The other most intense and prominent diffraction features originate from the FGT film. There are two different sets of reflections, namely FGT(11.0) and (22.0), and FGT(H0.0) with [H = 1,…,3]. They both refer to orientations with their c-axes along the surface normal. The three most intense diffraction features at (11.0), (22.0), and (30.0) reveal an FGT in-plane lattice parameter of 4.049(13) Å. This is, in general, agreement with the value obtained by RHEED, as well as literature values for Fe_5−xGeTe₂ (0 < x ≤ 0.4) bulk crystals^{[7, 8, 10]} and thin films.^{[9, 13]}

The intensity modulation along the FGT(11.0) ring exhibits a maximum in the azimuth which is aligned along the SiC[2$$\bar{1}$$.0] in-plane substrate direction. This indicates a preferential in-plane orientation, i.e., the (11.0) net planes of FGT and SiC are parallel, as we previously observed for Fe₃GeTe₂ films grown on graphene.^[25] Nevertheless, the FGT arc maxima are broader than those of Fe₃GeTe₂. For the arc crossing the (11.0) reflection, a full width at half maximum of 8.9° (Figure 1b, inset) is found, whereas for Fe₃GeTe₂ this value is about 4.7°.^[25] Even though this relatively large value is not uncommon for epitaxially grown vdW materials,^[26-28] it reveals that the in-plane mosaicity in the FGT film is clearly larger than that measured for the Fe₃GeTe₂ one (grown with virtually identical conditions, except the Fe flux). This difference might be associated with the higher energies required for the stabilization of layered Fe–Ge–Te phases exhibiting a Fe composition higher than 3.^[17] The existence of in-plane mosaicity can also be inferred from the rather diffuse streaks of the RHEED pattern (Figure 1a, bottom panel). Similar RHEED patterns were reported by Ribeiro et al.^[9] for Fe₅GeTe₂ films grown by MBE on Al₂O₃(0001) substrates. Interestingly, they could observe an evolution from diffuse to sharp RHEED streaks by performing postgrowth annealing of the films at a temperature of 550 °C under Te flux. This suggests that postgrowth thermal treatment as well as higher growth temperatures might improve in-plane order for FGT films grown on graphene. This will be the topic of further investigations. There are two more ring-like contributions detectable in the map and in the attached line profile taken at K = 0, which cannot be related to FGT. These are most probably due to a smaller fraction of tetragonal FeTe agreeing very well with FeTe(200) and (220) in-plane lattice spacing. Interestingly, the c-axis of tetragonal FeTe points along the direction perpendicular to the surface as well. Finally, a sequence of low-index polyrings with different lattice spacings is also detected, which is originating from the topmost Te capping layer.

Complementary to GID we have performed a laboratory-based omega/2theta measurement at an X-ray wavelength of 1.54056 Å, which probes exclusively out-of-plane features of the layer (see Figure 1c). A set of FeTe reflections confirms the presence of tetragonal FeTe with its out-of-plane orientation along SiC [00.1] as it could already be anticipated from the GID in-plane data. A further set of eight subsequent FGT(00.L) reflections with L = 3, 6, …, 24 proves (in conjunction with the presented in-plane data) the formation of FGT with a Fe₅GeTe₂ structure,^[8] since they fit well to the position and intensity ratios of the numerically derived powder pattern (black bars). Based on the positions of the contributions at L = 3, 6, 18, and 21, we have deduced an out-of-plane lattice parameter of 29.16(9) Å, in agreement with previous reports for bulk crystals and thin films.^{[7-10, 13]} The formation of [00.1]-oriented Fe₃GeTe₂ can be excluded as it would result into a clearly different diffraction pattern.^{[16, 29]} Note that these measurements were performed for uncapped FGT films, which explain the absence of contributions related to pure Te.

The surface morphology was probed by atomic force microscopy (AFM). Figure 1d shows a typical AFM height image for an uncapped FGT film. Its coverage appears mainly continuous over the micrometer-sized graphene/SiC terraces, with the exception of some holes appearing close to surface steps. These substrate regions are known to exhibit few-layer thick graphene ribbons and patches, whereas the terraces are mostly covered in monolayer—and, to a lesser extent, bilayer—thick graphene.^{[28, 30]} A similar effect was observed in Fe₃GeTe₂ grown on graphene/SiC(0001),^[25] which we associated with the lower surface reactivity of the thicker graphene areas relative to that of the terraces,^[22] leading to a lower FGT growth rate near the steps. Moreover, it is possible to observe nanoislands with tetragonal shapes scattered over the surface (Figure 1d, inset). Based on the GID and X-ray diffraction (XRD) findings, we suggest that they are composed of tetragonal FeTe,^[31] forming due to phase segregation during growth or during subsequent sample cooling. In spite of these unwanted surface effects—which are expected to be mitigated by further refinement of both the FGT and graphene growth protocols—it is reasonable to conclude that the FGT film offers a good surface homogeneity over large areas. The root-mean-square (RMS) roughness for different surface regions (free of FeTe nanoislands) varies between 0.8 and 1.1 nm, which is close to the thickness of one monolayer of Fe₅GeTe₂ (a single layer corresponds to 1/3 of the unit cell with c = 29.16(9) Å). It is important to note that the AFM investigations discussed here were conducted on noncapped FGT films. Although oxidation in FGT^[32] is not as strong as in 2D magnetic halides,^[33] surface oxidation taking place within a few hours after air exposure might increase the surface roughness.

The structure of the FGT/graphene heterostructure was also investigated by scanning transmission electron microscopy (STEM). Figure 2a depicts a high-angle annular dark-field (HAADF)-STEM image of the FGT film grown on graphene/SiC(0001). It is possible to observe the layered structure of the film; in that single layers are clearly separated by vdW gaps, in agreement with previous TEM characterization of FGT bulk crystals and thin films.^{[7, 9]} Most of the film is formed of well-oriented layers with the expected FGT single-layer thickness (≈1 nm, see Figure 2b). This value agrees with the c lattice constant obtained from the XRD data. Stacking and rotational faults are also evident. Strikingly, STEM images reveal that, although the overall vdW-layered structure is preserved, single layers with a thickness larger than 1 nm are also present. Figure 2c, which shows an enlarged area from an interface region in Figure 2a, reveals some layers as thick as 2 nm. To the best of our knowledge, the experimental realization of Fe–Ge–Te-layered phases in which the thickness of a single layer is larger than that of Fe₅GeTe₂ has not been reported. In a simple picture, such a thickness increment is anticipated if one considers that “filling” the 2D backbone structure of Fe₅GeTe₂ with more Fe atoms will result in the formation of additional slabs of Fe, following the same trend as experimentally observed when transitioning from Fe₃GeTe₂ to Fe₅GeTe₂ (for which the single-layer thickness evolves from ≈0.8 to ≈1.0 nm).^{[7, 8, 16, 34]} In fact, Liu et al.^[35] used first-principles calculations to study different Fe–Ge–Te-layered phases including Fe₆GeTe₂ and Fe₇GeTe₂. They predicted them to exhibit a thickness for single layers of around 1.1 and 1.2 nm, respectively. Our XRD data (Figure 1c) do not directly confirm the formation of such phases, which shows that they are not dominant in the film. However, the shift of the (00.9) reflection toward a lower angle (in comparison to the value calculated for Fe₅GeTe₂), as well as the existence of an additional, unidentified reflection closed to (00.18) (superimposed to SiC (00.6)), could be related to the existence of layered phases with larger c lattice constants.

Even though at the present stage we cannot unambiguously identify their composition and structure, the existence of single layers with thickness larger than that of Fe₅GeTe₂ is strong evidence for the formation of new Fe–Ge–Te-layered phases with Fe composition exceeding 5, which are possibly metastable.^[17] This illustrates the great potential of vdW epitaxy via MBE for the large-scale realization of metastable phases of 2D magnets, similar to what have been observed for 2D semiconducting dichalcogenides.^{[36, 37]} Further investigations are needed in order to identify the reasons for the formation of such phases and to controllably realize them. Finally, note that the average RBS composition obtained for our films (Fe_4.8GeTe₂) does not pose a contradiction to the existence of layers with higher Fe concentration. It is known that Fe_5−xGeTe₂ single layers are usually Fe-deficient with 0 < x ≤ 0.4^{[8, 10, 38, 39]} (x can be even close to 1 if the Fe₄GeTe₂ structure stabilized by Seo et al.^[17] is considered, which interestingly also exhibits a single-layer thickness of around 1 nm). Fe deficiency taking place in the “normal” FGT layers could, in principle, compensate for the higher amount of Fe present at the less abundant, thicker layers.

Another very important feature observed in the STEM images is the sharp nature of the FGT/graphene interface. Figure 2d depicts the same area as in Figure 2a, rendered using a custom color map to aid the visualization of the graphene layers, since the HAADF-STEM images provide an atomic number (Z) contrast. The continuous graphene layers are observed (dark green color) directly interfacing the crystalline FGT film (in orange). A more detailed view of the FGT/graphene interface, obtained for another sample region, is provided in Figure 2e–h. In agreement with the previous reports,^[22] they confirm the existence of a bilayer thick graphene (with an interlayer distance of 0.34 nm), and the carbon-rich buffer layer very close to SiC.

2.2 Magnetic and Electric Properties

The magnetic and electric properties were investigated via magnetotransport measurements, superconducting quantum interference device (SQUID) magnetometry, X-ray absorption spectroscopy (XAS), and X-ray magnetic circular dichroism (XMCD). These measurements were performed on pieces cut from the same 1 cm² sample consisting of an ≈15 nm FGT film (capped with 5 nm Te) grown on graphene/SiC(0001).

Figure 3a displays the Hall resistivity ρ_xy measured via magnetotransport during upward and downward sweeps of an external magnetic field H, where the arrows indicate the field sweeping directions. The linear background is caused by the ordinary Hall effect (OHE) with the negative slope observed for the whole temperature range suggesting electron-like transport behavior. However, the measured OHE results from parallel conduction through the FGT and graphene transport channels exhibiting hole- and electron-like transport characteristics, respectively.^[25] Consequently, the sign as well as the absolute value of the OHE slope depend on the relative conductivities of the FGT and graphene films. The extracted effective carrier density of 1.0 × 10¹⁴ cm⁻² and the sheet resistance of 176 Ω sq⁻¹ (at room temperature) in our FGT/graphene heterostructure are close to the values obtained for the previously studied Fe₃GeTe₂/graphene heterostructure (7 × 10¹³ cm⁻² and 150 Ω sq⁻¹, respectively).^[25] The observed hysteresis loop superimposed onto the OHE at each temperature originates from the anomalous Hall effect (AHE), its signal strength being proportional to the out-of-plane magnetization. The nearly square-shaped loops and the finite coercivity observed for temperatures up to 360 K provide clear evidence for ferromagnetism well above room temperature with a robust PMA. Similar results obtained for an ≈10 nm FGT film are shown in Figure S2 (Supporting Information). For the determination of the Curie temperature, we adopt the commonly used method based on the Arrott plot;^{[17, 40]} ρ_A² is plotted against µ₀H/ρ_A, where ρ_A is the AHE part of the Hall resistivity after subtracting the linear background of the OHE contribution (see also Figure S3 in the Supporting Information). From the linear fit in the high-field ranges of the Arrott plots in Figure 3b, the intercepts at the ρ_A² axis are extracted for each temperature. As shown in Figure 3c, these intercepts exhibit a linear temperature dependence in accordance with Arrott's theory.^[40] Finally, a Curie temperature of T_C ≈ 390 K is extracted from the extrapolation to zero intercept by a linear fit (red line in Figure 3c; see also Figure S2 in the Supporting Information). The same T_C is obtained from the temperature decays of the remanent AHE resistivity and the coercive field (Figure S3, Supporting Information). This value is significantly higher than those previously reported for single crystals (or flakes) as well as large-area films of Fe₃GeTe₂ (T_C ≈ 220 K),^{[25, 29, 41, 42]} Fe₄GeTe₂ (T_C ≈ 270 K),^[17] and Fe_5−xGeTe₂ (0 < x ≤ 0.4, T_C ≈ 280–320 K).^{[7, 9, 21, 39, 43]} Only the transition temperature recently reported for MBE-grown Fe_5−xGeTe₂/Bi₂Te₃ heterostructures (T_C as high as 570 K) surpasses this value.^[13]

The high T_C value approaching 400 K is in agreement with a recent theoretical prediction for the Fe₅GeTe₂ phase;^[17] however, it differs from other experimental reports for Fe_5−xGeTe₂. This could be related to the existence of thicker single vdW FGT layers, as observed by STEM. It is anticipated that a gradual increase of the Fe content within individual layers will result in a larger number of nearest Fe neighbors per Fe atom, enhancing the spin–pair exchange interaction and consequently the Curie temperature. In fact, recent theoretical investigations predicted PMA and T_C values of 450 and 570 K for layered Fe₆GeTe₂ and Fe₇GeTe₂, respectively.^[35] Therefore, in the case of our samples, the existence of similar phases with Fe composition higher than 5 could, in principle, lead to an effective T_C value that is superior to that of pure Fe_5−xGeTe₂. Nevertheless, in order to verify this hypothesis, future growth experiments will be conducted aiming at the stabilization, and isolation, of such phases. Finally, note that the tetragonal FeTe phase detected by GID and XRD (Figure 1b,c) and correlated to the surface nanoislands observed by AFM (Figure 1d) is not expected to contribute to the observed ferromagnetism due to its anti-ferromagnetic nature (Néel temperature of around 70 K).^[31]

The magnetic properties of the FGT film were further investigated by SQUID magnetometry. Figure 4 displays the remanent magnetization obtained for both out-of-plane (OP) and in-plane (IP) configurations as a function of temperature (a background signal caused by the SiC substrate has been subtracted; Figure S4, Supporting Information). The clearly larger remanent magnetization in the OP configuration confirms the PMA in our FGT films. Furthermore, it is clear from the decay of the OP remanent magnetization that ferromagnetic order persists well above room temperature, in agreement with the result extracted from the AHE measurements. The Fe atoms on the five inequivalent lattice sites in Fe₅GeTe₂ are expected to exhibit rather different magnetic moments.^{[7, 38]} The average magnetic moment per Fe atom obtained from the SQUID data measured at 2 K is 1.37 µ_B, considering the average Fe concentration of 4.8 as obtained by RBS measurements. This result falls into the range between 0.8 and 2.6 µ_B per Fe atom covered by reported experimental values.^[7]

XAS and XMCD measurements were employed to investigate the atomic-scale magnetic properties.^[44] Figure 5 shows the remanence of the dichroic signal at B = 0 T (normalized by its value at B = 6 T), for both grazing (20° from the sample plane) and normal incidence geometries, as a function of temperature. At low temperature, the sample has a PMA, as evidenced by the larger remanence in the perpendicular direction (red triangles). However, above T = 220 K, the measured magnetization undergoes an in-plane reorientation transition. The spontaneous magnetization persists above room temperature (Figures S6 and S7, Supporting Information).

We performed a sum-rule analysis on the XAS–XMCD spectra at T = 3 K to extract the atomic magnetic moments of Fe atoms (Figure S5, Supporting Information). We obtained an effective total magnetic moment in normal incidence (i.e., not corrected from the intra-atomic dipole operator T_z) M_eff = 1.13 ± 0.15 µ_B per Fe atom, in reasonable agreement with the value obtained by SQUID. The lower value found from this analysis might be due to the fact that not all Fe atoms probed in XAS contribute to the magnetic signal obtained by XMCD (e.g., those part of anti-ferromagnetic FeTe nanoislands do not contribute); thus, this value is an underestimation of the actual Fe magnetic moment in FGT. This analysis also allowed us to extract the orbital moments measured in normal incidence $$m_{\rm{L}}^ \bot = 0.03$$ µ_B/Fe and the difference between perpendicular and in-plane orbital moments $${{\Delta}}{m}_{\rm{L}} = m_{\rm{L}}^ \bot - m_{\rm{L}}^\parallel $$ that is proportional to the macroscopic anisotropy in itinerant magnetism materials according to Bruno's model.^[45] We found Δm_L < 0.01 µ_B, consistent with the small anisotropy observed in normal (red curve) and grazing (black curve) hysteresis loops shown in the inset of Figure 5.

A T_C above room temperature, as observed in XMCD, is in good agreement with literature values for Fe_5−xGeTe₂ with x ranging from 0 to 0.3.^{[9, 10, 21, 39]} Certain discrepancy observed between XMCD, SQUID, and magnetotransport results, in terms of remanence, anisotropy and T_C may be attributed to the fact that the different experiments probe different parts of the FGT film. While SQUID probes the whole film, the detection in XMCD is restricted to the topmost FGT layers (≈5 nm). As observed by STEM (Figure 2a), this area is formed exclusively of single layers with the interlayer spacing expected for Fe_5−xGeTe₂ with x being close to 0. Also, the grazing geometry measured in XMCD does not correspond to a purely in-plane component of the magnetization, but instead it has a contribution from the out-of-plane one. For the magnetotransport (AHE) measurements, the actual current distribution between the different FGT components and the graphene film is essential but difficult to determine.