Ab Initio Guided Low Temperature Synthesis Strategy for Smooth Face–Centred Cubic FeMn Thin Films

Abstract

The sputter deposition of FeMn thin films with thicknesses in the range of hundred nanometres and beyond requires relatively high growth temperatures for the formation of the face-centred cubic (fcc) phase, which results in high thin film roughness. A low temperature synthesis strategy, based on local epitaxial growth of a 100 nm thick fcc FeMn film as well as a Cu nucleation layer on an α-Al₂O₃ substrate at 160 °C, enables roughness values (R_a) as low as ~0.6 nm, which is in the same order of magnitude as the pristine substrate (~0.1 nm). The synthesis strategy is guided by ab initio calculations, indicating very strong interfacial bonding of the Cu nucleation layer to an α-Al₂O₃ substrate (work of separation 5.48 J/m²)—which can be understood based on the high Cu coordination at the interface—and between fcc FeMn and Cu (3.45 J/m²). Accompanied by small lattice misfits between these structures, the strong interfacial bonding is proposed to enable the local epitaxial growth of a smooth fcc FeMn thin film. Based on the here introduced synthesis strategy, the implementation of fcc FeMn based thin film model systems for materials with interface dominated properties such as FeMn steels containing κ-carbide precipitates or secondary phases appears meaningful.

1. Introduction

The material combination FeMn in its face-centred cubic (fcc) phase plays an important role in structural materials such as high manganese steels, which exhibit excellent mechanical properties [1,2,3,4]. FeMn thin films can serve as model systems to gain fundamental understanding required for the design of those steels: Single phase model systems—deposited with the highly efficient combinatorial thin film composition-spread method and combined with ab initio calculations—were shown to be a powerful tool to rationalize composition-structure-property relationships [5]; Multiphase model systems with multilayers separated by flat (2D) interfaces are expected to render fundamental and systematic insights required for the design of materials with interface dominated properties, for example, FeMn steels containing κ-carbide precipitates [6,7]. Furthermore, FeMn thin films offer antiferromagnetic properties and are used in magnetic thin film devices, for example, combined with at least one ferromagnetic layer in a multilayer thin film exhibiting the so-called exchange bias (EB) effect [8,9,10,11,12,13].

The multi-layered FeMn based thin film model systems as well as magnetic devices require sharp interfaces for a meaningful determination and understanding of interface dominated properties like coherency effects [14] or EB [15]. The sputter deposition of FeMn based model systems on rigid oxide substrates requires relatively high deposition temperatures of up to 450 °C for the fcc phase formation, which is reported to be accompanied by high thin film roughness of up to 170 nm for film thicknesses in the order of 2 µm [16,17,18,19]. Due to the high surface roughness, the implementation of fcc FeMn thin films in layered thin film model systems did not yet appear meaningful. On the other hand, (ultra) thin fcc FeMn layers in magnetic devices with high quality interfaces are accomplished by stabilizing the fcc lattice by epitaxial thin film growth on metallic nucleation layers at low deposition temperatures [20,21,22,23,24] but layer thicknesses are restricted to several nm [22]. The template effect necessitates a low lattice mismatch between FeMn and the nucleation layer and thus a similar arrangement of the atoms at the interface [20,25]. The lattice parameter of fcc Cu (a_Cu = 3.615 Å [26]) is very similar to the lattice parameter of fcc FeMn (a_FeMn ≈ 3.59–3.74 Å [27]), whereby the latter is determined by the Fe to Mn ratio, giving rise to a low lattice mismatch <4% depending on the Mn content in a cube-on-cube oriented Cu/FeMn interface arrangement. If Cu were used as (bulk) substrate material, its mechanical properties (plastic deformation) would preclude analysis techniques such as nanoindentation. However, nanoindentation is possible if a thin Cu nucleation layer on a rigid (oxide) substrate material is employed. Schmidt et al. used MgO (a_MgO = 4.210 Å [28]) substrates for the molecular beam epitaxy (MBE) synthesis of the EB FeMn/Co. thin-films due to its cubic structure and lattice mismatch of about 16% for cube-on-cube orientation on FeMn [15]. Sapphire (α-Al₂O₃) substrates (a_α-Al2O3 = 4.759 Å [29]) possess a hexagonal structure but exhibit a lattice mismatch of about 8% in (0001) stacking with Cu(111). Cu was shown to grow epitaxially on both oxide substrates: Cu(100) on MgO(100) [30] as well as Cu(111) on MgO(111) [30] or α-Al₂O₃ (0001) [31]. Moreover, Sigumonrong et al. have suggested that a small lattice misfit accompanied by strong interfacial bonding—characterized by a large work of separation—enables local epitaxial thin film growth and this notion was verified experimentally by the local epitaxial growth of 1.5 µm thick sputter deposited V₂AlC thin films on α-Al₂O₃($11\overline{2}0$) [32]. Further examples for local epitaxial growth are Ti_1−xAl_xN [33] and Ti_1−xAl_xN/Ti_1−yAl_yN multilayer [34] thin films on ferritic steel substrates or α-(Cr,Al)₂O₃ thin films on α-Al₂O₃ substrates [35].

Thus, the goal of our paper is to develop a low temperature synthesis strategy based on local epitaxial growth on a Cu nucleation layer to enable the growth of smooth fcc FeMn films with surface roughness values in the order of magnitude of the pristine substrate. This development is guided by ab initio calculations identifying a layer system comprising a rigid substrate material, Cu nucleation layer and fcc FeMn thin film that is characterized by strong interfacial bonding across both interfaces. The work of separation for each interface in the layer systems MgO(001)/Cu(001)/Fe_0.6Mn_0.4(001) and α-Al₂O₃(0001)/Cu(111)/Fe_0.6Mn_0.4(111) is determined as a measure for the interfacial strength. The layer system α-Al₂O₃(0001)/Cu(111) will be demonstrated to exhibit the highest interfacial strength across the substrate/nucleation layer interface. Furthermore, the low temperature growth of a combinatorially deposited 100 nm thick FeMn layer onto the Cu nucleation layer is employed. This thin film exhibits a continuous Fe-Mn concentration gradient parallel to the substrate surface and thus allows to determine the Mn concentration range for the formation of fcc FeMn on the Cu nucleation layer.

2. Materials and Methods

To determine the work of separation between different layers of the systems MgO/Cu/FeMn and α-Al₂O₃/Cu/FeMn, density functional theory [36] was used. Vienna ab initio simulation package (VASP 5.4.1, Computational Materials Physics, University of Vienna, Vienna, Austria) and projector augmented wave potentials [37,38,39] were employed. The augmented wave potential parametrization was performed within the generalized-gradient approximation by Perdew, Burke and Ernzerhof [40]. The Blöchl approach [41] was applied for the total energy. The integration in the reciprocal space was carried out on 9 × 9 × 1 k-point Monkhorst-Pack mesh [42]. Fe_0.6Mn_0.4 was randomized using special quasirandom structures (SQS) [43], whereby locally self-consistent Green’s function (LSGF) package [44,45] was employed. In particular, the Warren-Cowley short-range order parameter [46] within seven coordination shells was applied to treat randomness in these SQS configurations. The cells obtained from the LSGF code were used as input in the VASP code. Fe_0.6Mn_0.4 was spin polarized (antiferromagnetic ordering). Full structural optimization was made for each configuration employing the convergence criterion for the total energy of 0.01 meV and a 400 eV cut-off. Two interface configurations were studied: MgO(001)/Cu(001)/Fe_0.6Mn_0.4(001) and α-Al₂O₃(0001)/Cu(111)/Fe_0.6Mn_0.4(111). The interfaces were characterized by the work of separation [32,47], calculated from the total energy change per unit area upon separation of a slab from the remaining layered structure, for example, Fe_0.6Mn_0.4(001) separation from MgO(001)/Cu(001). MgO(001)/Cu(001)/Fe_0.6Mn_0.4(001) was constructed by 6 layers of MgO(001), 6 layers of Cu(001) and 6 layers of Fe_0.6Mn_0.4(001), accounting for 48 atoms. The layer thickness convergence was performed by considering 6 additional layers of Cu(001). The work of separation differed by only 0.02 J/m² between 6 and 12 layer Cu configurations so that 6 layers were henceforth considered. The stacking sequence for MgO(001)/Cu(001) was adopted from literature [48]. α-Al₂O₃(0001)/Cu(111)/Fe_0.6Mn_0.4(111) was span by 7 layers of O terminated α-Al₂O₃(0001), 6 layers of Cu(111) and 6 layers of Fe_0.6Mn_0.4(111). Cu(111) and Fe_0.6Mn_0.4(111) were constructed by A-B-C-A-B-C(111) stacking of the corresponding cubic lattice. The stacking of α-Al₂O₃(0001)/Cu(111) was implemented based on a previous work by Hashibon et al. [49]. The number of layers implemented for the oxide substrates is assumed to be sufficient based on ab initio studies on the surface energy of the binary oxides by Lazar et al. [50] and Sigumonrong et al. [51]. All interfaces were constrained to the corresponding 0 K lattice parameters of the substrates (MgO and α-Al₂O₃) obtained herein.

The combinatorial Fe-Mn thin films on Cu nucleation layers were deposited in a laboratory-scale sputtering system by direct current magnetron sputtering. The base pressure was below 2.2 × 10⁻⁵ Pa and the Ar deposition pressure was 0.75 Pa. A Cu target (power density 0.86 W/cm²) was facing the substrate for the deposition of the nucleation layers. A Fe target (power density 1.23 W/cm²) and a Mn target (power density 0.36 W/cm²) with inclination angles of 45° with respect to the substrate normal were used in order to achieve the desired Fe-Mn gradient. Fast acting target shutters (actuation time of 200 ms) ensured the deposition of defined Cu nucleation layers and the FeMn films directly followed by each other. Polished single-crystalline α-Al₂O₃(0001) substrates in a target-to-substrate distance of 10 cm were kept at floating potential and heated to 160 °C during deposition.

The chemical composition of the combinatorial FeMn thin film was determined by energy dispersive X-ray analysis (EDX) in a JEOL JSM-6480 scanning electron microscope (JEOL Ltd., Tokyo, Japan) equipped with an EDAX Genesis 2000 EDX system (EDAX Inc., Mahwah, NJ, USA). Spatially resolved, high-throughput X-ray diffraction (XRD) measurements were used to evaluate the local crystal structure of the combinatorial Fe-Mn thin films using a Bruker AXS D8 Discover XRD (Bruker Corporation, Billerica, MA, USA) equipped with a General Area Detector Diffraction System (GADDS). The diffractometer was operated at a current of 40 mA and a voltage of 40 kV with Cu Kα radiation at a fixed incident angle ω of 8°. ϕ-scans are performed with a frame width of 3° at 2θ = 43.39°, ω = 21.70° and χ = 20°. The primary beam was collimated with a 0.5 mm pinhole. For Bragg-Brentano XRD measurements with a step size of 0.02°, a Siemens D5000 (Siemens, Munich, Germany) diffractometer was employed and operated at a current of 40 mA and a voltage of 40 kV with Cu Kα radiation.

The surface roughness (R_a) of the FeMn thin film and a pristine α-Al₂O₃(0001) substrate was determined using a Hysitron TI-950 TriboIndenter (Hysitron Inc., Eden Prairie, MN, USA) by scanning the film surface with a cube corner diamond tip and average the R_a values from ten areas of 10 × 10 µm² or six areas of 5 × 5 µm², respectively.

Site-specific lift-outs for microstructural characterization of the interface were carried out in a FEI Helios Nanolab 660 (FEI Company, Hillsboro, OR, USA) dual-beam focused ion beam (FIB) microscope. Lamellae were extracted in growth direction of the thin films and Ga ions were used at an acceleration voltage of 30 kV. First, a 1 µm thick Pt protection layer was applied, followed by trench milling, extraction of the lamella with a manipulation needle and application of the lamella on a Cu Omniprobe grid. Final lamella thicknesses were <100 nm. The lamellae were investigated using EDX analysis in a JEOL JSM-2200FS field emission gun transmission electron microscope (TEM) (JEOL Ltd., Tokyo, Japan) in scanning TEM (STEM) mode, with a 30 mm diameter silicon drift detector from JEOL. In addition, selected area electron diffraction (SAED) investigations in the TEM were utilized to disclose the orientation dependency between the substrate and the Cu and FeMn layers. The results of the SAED were evaluated using the Java electron microscopy software (JEMS) (Version 3.8224U2012, JEMS-SAAS, Saas-Fee, Switzerland).

3. Results and Discussion

3.1. Ab Initio Interface Design

For the ab initio design of a suitable layer system, Figure 1 depicts the structural models as well as the corresponding electron density distributions, which are used to explore the electronic structure analysis. For MgO(001)/Cu(001)/Fe_0.6Mn_0.4(001), the calculated work of separation value is 0.85 J/m² for the substrate/Cu interface, which is consistent with literature [48] and 3.89 J/m² for the Cu/FeMn interface. For α-Al₂O₃(0001)/Cu(111)/Fe_0.6Mn_0.4(111), the work of separation value is 5.48 J/m² for the substrate/Cu interface, which is in line with previous studies [49] and 3.45 J/m² for the Cu/FeMn interface. Hence, striking differences occur for the considered substrate/Cu interfaces. MgO(001)/Cu(001) reveals a rather low work of separation, which is still by one order of magnitude higher compared to a weak interface, for example, ideal polypropylene/Ti_0.5Al_0.5N(002) [52] but smaller than medium strong interfaces like V₂AlC/α-Al₂O₃ [32] or Pt/NbO₂ [53]. The α-Al₂O₃(0001)/Cu(111) interface exhibits a more than six times higher work of separation which is in the range of strong interfaces such as Nb/α-Al₂O₃ [54], TiO₂/α-Al₂O₃ [55], or Al/diamond [56]. On the other hand, the work of separation for the Cu/Fe_0.6Mn_0.4 interface in both layer systems is comparable and differs by 0.44 J/m². The values are in the range of other medium strong interfaces such as Ir/Ir₃Zr [57].

The distinct differences in the calculated work of separation for the substrate/Cu interface can be understood based on the Cu coordination at the interface, which is 1 between Cu and O for MgO(001)/Cu(001) but 3 for α-Al₂O₃(0001)/Cu(111) with Cu being bonded to O via (short and strong) ionic-covalent bonds. The origin of the difference in interfacial bonding can then be explained by the different electronic structures of MgO and α-Al₂O₃: The Mg-O bonds are predominantly ionic in character with significant charge transfer from Mg to O, while the Al-O bonds are characterized by (ionic) charge transfer as well as some localization (hybridization; consistent with literature on isostructural oxides [58]). For the O terminated α-Al₂O₃, the bond distance between Cu and O is 2.10 Å and the shared charge between Cu and O provides only Cu-O bonds in the interface. Compared to that, MgO is terminated by both O and Mg so that the weaker Cu-O interaction is characterized by less near neighbours and hence the lower coordination and a bond distance Cu-O of 2.25 Å.

The comparable work of separation values for the Cu/FeMn interfaces in both systems can be explained by the rather similar bonding conditions. The bond distance between Cu and Fe is 2.55 Å for Cu(001)/Fe_0.6Mn_0.4(001) compared to 2.56 Å for the Cu(111)/Fe_0.6Mn_0.4(111) interface exhibiting a 13% lower work of separation for the considered Fe_0.6Mn_0.4 lattice with its Mn content dependent lattice constant. As the nature of both the Cu as well as the Fe-Mn bonds are mainly metallic offering itinerant change with some localization between the atoms and there are no coordination modulations across the interface, hence, the bonding conditions at the Cu/FeMn interface are maintained throughout both layers.

Hence, the interfacial bonding between rigid substrate and Cu nucleation layer is theoretically shown to be extraordinary high for α-Al₂O₃(0001)/Cu(111), which may facilitate the epitaxial growth of Cu(111) on α-Al₂O₃(0001). The subsequent Cu(111)/Fe_0.6Mn_0.4(111) interface exhibits high interfacial strength as well and may thus allow for the continuation of the epitaxial growth throughout the interface between nucleation layer and FeMn thin film. Consequently, the layer system α-Al₂O₃(0001)/Cu(111)/Fe_0.6Mn_0.4(111) is selected to be appraised experimentally.

3.2. Experimental Validation

A 10 nm layer of Cu subsequently followed by a 100 nm combinatorial FeMn layer was synthesized on an α-Al₂O₃(0001) substrate and evaluated by EDX and spatially resolved XRD to identify the Mn content range, in which the desired fcc FeMn forms. Figure 2a shows the XRD scans (2θ-scans at fixed incidence angle ω) for Mn contents between 16 and 49 at. %. For Mn contents above 39 at. % the formation of phase pure fcc FeMn is observed. Lower Mn contents appear to facilitate the formation of an additional bcc phase and are thus excluded from further analysis. Figure 2b shows the Bragg-Brentano XRD scan of the fcc region of the α-Al₂O₃/Cu/FeMn sample, which is selected for TEM analysis (TEM-EDX: ~42 at. % Mn). For comparison, the XRD scan of a sole 10 nm Cu thin film (α-Al₂O₃/Cu)—deposited and measured under the same conditions—is included. Both Bragg-Brentano scans solely exhibit the (111) peak of the fcc structure. The intensity difference is caused by the difference in thin film thickness (110 nm for α-Al₂O₃/Cu/FeMn and 10 nm for α-Al₂O₃/Cu) and thus measured volume. Hence, the fcc FeMn layer as well as the Cu layer are highly (111) oriented and exhibit a very similar d₁₁₁-value of about 2.08 Å. The presence of a (111) fibre texture was excluded by ϕ-scans (2θ = 43.39°, ω = θ°, χ = 20°) in the fcc region at ~42 at. % Mn. Figure S1 (Supplementary Material) shows the maximum intensities plotted in dependence of ϕ, exhibiting maximum intensities every 60° ϕ but minimum intensities in-between and thus refuting a fibre texture. Hence, XRD analysis indicates local epitaxial growth, which will be confirmed by high resolution techniques.

The surface roughness of the FeMn thin film is determined to be as low as R_a = 0.6 nm with a calculated standard deviation of 4% (based on ten individual measurements of 10 × 10 µm²; a representative surface area scan can be found in the Supplementary Material, Figure S2). Hence, the surface roughness is on the same order of magnitude as the one measured for the pristine α-Al₂O₃ substrate (R_a = 0.1 nm). Thus, a significant reduction of the surface roughness compared to the high temperature synthesis strategy resulting in surface roughness values of R_a = 58 nm [19] and larger [16,17] (film thickness ~2 µm) could be obtained.

The STEM high angle annular dark field (HAADF) image of the layer system α-Al₂O₃/Cu/Fe_0.58Mn_0.42 with the corresponding EDX line scan shown in Figure 3 confirms the chemical stacking of the layer system: The α-Al₂O₃ substrate is characterized by high Al K and O K intensities, while the nucleation layer shows a peak in Cu K intensity for a length of about 10 nm. The Cu layer is directly followed by the ~100 nm thick FeMn layer with correspondingly high Fe K and Mn K intensities. The uppermost region of the FeMn layer already shows significant intensities of Pt M and Ga K, which is an effect of the FIB preparation necessitating the deposition of a Pt protection layer in order to reduce Ga implantation. Spot EDX measurements of the Fe and Mn content in the thin film layer result in 58 ± 1 at. % and 42 ± 1 at. % Mn, respectively, which is close to the calculated Fe_0.6Mn_0.4 composition.

The SAED of the α-Al₂O₃/Cu/Fe_0.58Mn_0.42 layer system confirms that the α-Al₂O₃(0001) substrate is a chemically ordered structure and thus provides superstructure diffraction spots, as shown in the indexed diffraction pattern in Figure 4a. Figure 4b shows the diffraction pattern and an evaluation of the Cu/Fe_0.58Mn_0.42 layer. The Cu and the FeMn layer are not distinguishable and offer the same diffraction spots, confirming that they exhibit the same fcc structure as well as non-distinguishable lattice parameters. This was already indicated by the (macroscopic) XRD analysis discussed above, see Figure 2b. The fcc Fe_0.58Mn_0.42 is a chemically disordered solid solution, which is consistent with literature [25]. The orientation relationship between the α-Al₂O₃ substrate and the fcc Cu/Fe_0.58Mn_0.42 suggests (local) epitaxial growth and is determined as follows:

• α-Al₂O₃$\left(0006\right)$ || Cu/Fe_0.58Mn_0.42$\left(\overline{1}\overline{1}1\right)$ [or $\left(11\overline{1}\right)$] and
• α-Al₂O₃$\left(0\overline{3}30\right)$ || Cu/Fe_0.58Mn_0.42$\left(2\overline{2}0\right)$ [or $\left(\overline{2}20\right)$]

The possibilities in square brackets are due to the rotational symmetry for a 180° rotation of the diffraction pattern. Furthermore, the inverted diffraction pattern of the layer system in Figure 4c reveals that the (small, focused) diffraction spots of the α-Al₂O₃ substrate are not positioned at the exact same positions as the (broad, unfocused) spots of the fcc Cu/Fe_0.58Mn_0.42 layer, indicating on the one hand that the (local) epitaxial growth is realized but also that the slightly different lattice plane spacing of α-Al₂O₃ and fcc Cu/Fe_0.58Mn_0.42 result in strain relaxation in the probed volume. Hence, we have demonstrated that fcc Fe_0.58Mn_0.42(111) can be grown epitaxially on fcc Cu(111) which in turn grows epitaxially on α-Al₂O₃(0001).

The low surface roughness of the 100 nm thick FeMn thin film, achieved by the local epitaxial growth on the Cu nucleation layer, allows for the formation of a flat (2D) interface between the fcc FeMn if another layer is deposited on top. This is a necessary step towards realizing the deposition of a bi- or multi-layered (multiphase) thin film model systems to allow for the systematic evaluation of interface dominated properties. These model systems are relevant to explore for example FeMn steels containing precipitates or secondary phases. The single phase model systems can benefit from the low surface roughness in terms of for example, a more accurate determination of mechanical properties by nanoindentation or other roughness-sensitive analysis techniques.

4. Conclusions

In conclusion, a low temperature deposition strategy for the growth of smooth fcc FeMn thin films was developed. Ab initio calculations suggest that the layer structure α-Al₂O₃(0001)/Cu(111) exhibits higher interfacial strength than MgO(001)/Cu(001) and is thus chosen as substrate/nucleation layer combination to enable (local) epitaxial growth of fcc Cu(111) on α-Al₂O₃(0001) as well as (local) epitaxial growth of fcc FeMn(111) on fcc Cu(111). The work of separation between Cu and Fe_0.6Mn_0.4 does not differ significantly between the two configurations studied, and, as such, is not decisive regarding the selection of the substrate system. However, the α-Al₂O₃(0001)/Cu(111) interface exhibits a more than six times higher work of separation (5.48 J/m²) than the MgO(001)/Cu(001) interface (0.85 J/m²) and hence strong interfacial bonding. This distinct difference in the calculated work of separation can be understood based on the Cu coordination at the interface, which is 1 between Cu and O for MgO(001)/Cu(001) but 3 for α-Al₂O₃(0001)/Cu(111).

The low temperature deposition strategy for the growth of smooth fcc FeMn thin films was critically appraised by synthesizing a 100 nm thick (combinatorial) FeMn thin film on a 10 nm Cu nucleation layer which was in turn deposited on an α-Al₂O₃(0001) substrate (α-Al₂O₃/Cu/FeMn). XRD confirms the phase formation of fcc FeMn for Mn contents above 39 at. %. The (local) epitaxial growth of fcc Fe_0.58Mn_0.42(111) on fcc Cu(111) on α-Al₂O₃(0001) is verified by TEM/SAED with the following orientation relationship: α-Al₂O₃$\left(0006\right)$ || Cu/Fe_0.58Mn_0.42$\left(\overline{1}\overline{1}1\right)$ and α-Al₂O₃$\left(0\overline{3}30\right)$ || Cu/Fe_0.58Mn_0.42$\left(2\overline{2}0\right)$. The thin film exhibits the desired low surface roughness with R_a values as low as 0.6 nm, which is on the same order of magnitude as the α-Al₂O₃ substrate (R_a ~ 0.1 nm) and at least two orders of magnitude lower than the fcc FeMn based thin films deposited at high temperatures (R_a ≥ 58 nm). Thus, the introduced synthesis strategy opens the possibility of the successful implementation of thin film model systems for materials with interface dominated properties such as FeMn based steels containing κ-carbide precipitates.