Magnetostrictive iron gallium thin films grown onto antiferromagnetic manganese nitride: Structure and magnetism

The study of magnetic bilayer systems and the inherent interface effects that occur continue to advance our ability of tailoring magnetism. The strong coupling between strain and magnetism in ferromagnetic FeGa alloys, due to their well-known magnetostrictive properties,^1–3 expands the possibilities for such tailoring especially when coupled with various substrates. Studies of FeGa thin films grown onto non-magnetic MgO, GaAs, and ZnSe/GaAs focused on their magnetic properties as a function of composition and thermal treatment.^4–11 For FeGa grown on semiconducting ZnSe/GaAs(001), the observed tetragonal distortions from the ideal cubic bulk lattice, attributed to a specific arrangement of Ga atoms within the structure, were found to be responsible for interesting magnetic properties, such as perpendicular magnetic anisotropy (PMA).^7,9,10 Continuing to investigate FeGa films onto various substrates can lead to a better control of the magnetic properties and to an overall understanding of the changes that occur compared to their bulk counterpart.

In this letter, we report the growth of FeGa thin films onto antiferromagnetic (aFM) manganese nitride [MnN(001) and Mn₃N₂(001)] substrates. For comparison, we also investigate similar composition films grown onto non-magnetic MgO(001). Since the surface morphology could also impact the film properties, we add a distinct perspective on these films by studying their surfaces with scanning tunneling microscopy (STM). We also combine ex-situ structural, compositional, and magnetic characterizations, revealing that a combination of strain and interface effects lead to an effective tailoring of magnetic anisotropies and to the appearance of PMA.

The $Fe 100−x$Ga_x films are grown by molecular beam epitaxy (MBE); MgO(001) is also used as a starting substrate for the manganese nitride films (see Yang et al. for growth details¹²). The Fe to Ga flux ratios are set such that samples with compositions x around 15 at.% Ga are obtained. The sample temperature T_S is varied from 360 °C to 420 °C between different samples. An additional FeGa/MnN(001) sample with x = 23.6 ± 2.4 is also prepared (at 320 °C) for further comparisons. The growth is monitored in real-time using reflection high-energy electron diffraction (RHEED). Following preparation, the samples are investigated with STM. Ex-situ magnetic force microscopy (MFM) imaging is performed to probe the magnetic domain structure of our films. The growth orientation and film composition are characterized by X-ray diffraction (XRD) and Rutherford backscattering spectrometry (RBS), respectively. The magnetic properties are determined using vibrating sample magnetometry (VSM) measurements at room temperature (RT).

From the RHEED patterns shown in Fig. 1, we infer that all FeGa films grow with a 45° in-plane rotation with respect to their corresponding substrate; combined with XRD results (from Fig. 2) we derive the following epitaxial relationships: $(001)FeGa||(001)MgO,MnN,Mn3N2, [110]FeGa||[100]MgO,MnN,Mn3N2$⁠. The growth temperature, film thickness t, and Ga content of each sample are indicated in the table from Fig. 1. RHEED patterns recorded during growth suggest that the in-plane lattice constant (LC) for all films remains unchanged after about 20 nm; thus all films (given that t > 100 nm) can be considered bulk-like. The RHEED patterns for samples grown on MgO (823 and 826) are streaky, consistent with high-quality epitaxial growth. Faint 2× streaks are observed for sample 823, indicating the formation of a 2 × 2 surface reconstruction; this is not the case for sample 826, most likely due to the increased growth temperature. By looking at the samples grown onto aFM substrates (825 and 828), the spottier patterns are an indication of rougher surfaces; in addition, no fractional streaks could be observed. The differences between the four FeGa films are also consistent with STM results (discussed below).

Shown in Fig. 2 are XRD measurements taken with a Cu Kα source for the FeGa films with x $≈$ 15. We find that all films are single-crystalline and grow with a (001) orientation on all substrates. From the peak positions in XRD, we calculate the out-of-plane LCs for the FeGa films; these values are then compared (see top inset from Fig. 2) with in-plane values extracted from RHEED spacing measurements^13,14 and with in-plane values obtained from separate (101) and (112) XRD reflections.¹⁵ Samples 823 and 826 seem to have close to cubic symmetry, although the RHEED and in-plane XRD values are in slight disagreement. For samples 825 and 828, grown onto aFM substrates, the out-of-plane XRD values are larger than the RHEED values by about 2.1%, indicating pronounced tetragonal distortions compared to samples grown on MgO; also note that the RHEED and in-plane XRD values are in excellent agreement, proving that the distortions happen throughout the bulk and are not simply due to a surface effect. A similar tetragonal distortion (⁠$≈$ 2.4%) is calculated for the most Ga-rich (x $≈$ 23.6) sample, 824 (t = 48 ± 4 nm) grown onto MnN (001), finding the out-of-plane and in-plane LCs to be 2.93 ± 0.01 and 2.86 ± 0.02 Å, respectively. Pronounced tetragonal distortions were also observed by Eddrief et al. for 36-nm thick $Fe 100−x$ Ga_x (0 $≤x≤$ 29.4) films grown on ZnSe at 180 °C.⁷ Interestingly, while the out-of-plane LC changed considerably and continued to increase with Ga content, the in-plane LC remained unchanged and equal to that of pure Fe films. This unusual distortion was attributed to Ga-pairing along the [001] growth direction, in agreement with theoretical predictions.¹⁶ Although our films are grown onto manganese nitride substrates and for a limited number of concentrations, a similar trend is observed for samples 824, 825, and 828, with the in-plane LC staying about the same.

Referring now to the STM investigations, Figs. 3(a) and 3(b) show the surface morphology of the two FeGa films deposited onto MgO. Sample 823 exhibits a pyramidal morphology with connecting pyramids that form a maze-like pattern; the inset from Fig. 3(a) [shown in derivative mode] is a zoom-in covering several pyramids. By looking at sample 826, a cloverleaf spiral morphology can be observed, with terraces showing the presence of vacancy islands and being considerably larger compared to sample 823. This type of morphology could be a result of the increased sample temperature during growth by 60 °C compared to sample 823. For samples 825 and 828 (having similar composition but grown onto MnN and Mn₃N₂, respectively), the surface morphologies [only shown for sample 828 in Fig. 3(c)] were found to be a lot rougher, consistent with RHEED investigations which show spottier patterns compared to samples grown on MgO. For the most Ga-rich sample, 824, although it was found that some regions were rougher than others, a large portion of the imaged areas had a morphology similar to that shown in Fig. 3(d). By increasing the Ga content, the surface exhibits a stepped morphology with straight and well-defined step edges. Also noticeable are a variety of surface arrangements that appear in close proximity to one another, as visible in the inset, including two distinct types of rows having different widths (emphasized by the dotted boxes) together with some featureless areas; such types of rows could occur if they have a different or varying composition. Interestingly, the smaller width rows are perfectly parallel to one another, whereas the wider rows exhibit some irregularities and distortions; we do not attribute these differences to a tip effect, since we would then expect for these distortions to occur across the entire horizontal section of the image.

To understand these various morphologies, we next address the phase diagram of FeGa.³ For $x≤$ 15, the Ga atoms take random positions within the FeGa matrix forming the disordered A2 phase. We could therefore consider that the areas shown in Figs. 3(a)–3(c) come from mostly the A2 phase. For 15 $≤x≤$ 23, the FeGa alloy is in a mixed region of A2 and more ordered D0₃. For higher x, the FeGa alloy only exhibits ordered phases, D0₃ and B2. Since sample 824 is in the right compositional range for at least a mixture of phases to occur, the observed row-structures could actually be an indication of such ordered phases. Moreover, studies on bulk FeGa alloys with x $≈$ 19 found that D0₃-phase nanostructures (⁠$≤$ 2 nm in size) were dispersed within the A2 matrix.¹⁷ Therefore, such structures should be easily observable by STM. However, it is difficult at this point to attribute the observed structures to a particular FeGa ordered phase.

The magnetic properties of the FeGa films for in-plane fields applied along [100] and [110] (and for additional out-of-plane fields applied along [001] for the samples grown onto aFM substrates) are shown in Fig. 4; all FeGa films exhibit primarily in-plane anisotropy. Starting with FeGa/MgO (x $≈$ 15), sample 823 has [100] as easy-axis, consistent with recent studies of similar composition FeGa films grown on MgO,^4,5 and the coercive field H_c is found to be $≈$ 80 Oe. From the large squareness (calculated to be $≈$ 0.95), the magnetization of sample 823 stays in the film plane for the most part. For sample 826 (not shown), we observe minute differences in the magnetic behavior compared to 823. By looking at the results for samples 825 and 828 (also with x $≈$ 15), we see that the in-plane anisotropies are completely suppressed, with loops having an almost identical behavior along both in-plane directions. It is interesting to point out that this loss of in-plane anisotropy was also observed for 72-nm thick FeGa films of similar composition grown on ZnSe.⁹ In addition to the lack of in-plane anisotropy, the shapes of the hysteresis loops are very different compared to those of samples grown on MgO. One immediate observation is that H_c increases considerably and the remnant magnetization M_r decreases, the latter being an indication that the out-of-plane component of the magnetization in these films increases. Although samples 825 and 828 show similar behavior, the difference in substrates induces slight changes in H_c and M_r values. For the most Ga-rich sample 824, we again observe a loss of in-plane anisotropy and a decrease in M_r; in addition, this sample has the highest H_c, of about 500 Oe. From this behaviour, it is evident that the magnetic properties of FeGa can be easily tuned by changing substrates.

The domain structure of the films grown onto aFM substrates is shown in Figs. 5(a)–5(c), and one can see that all films exhibit stripe-like magnetic domain patterns, independent of thickness or composition. Such stripe-like domains are associated with an oscillating out-of-plane component of the magnetization in these films and consequently with the appearance of PMA.^9,18,19 As discussed above, the M-H loops for samples 824, 825, and 828 indicate such a behavior by showing an increase in the out-of-plane component of the magnetization. Returning to the case of FeGa/ZnSe,^7,9,10 in addition to having an almost identical magnetic anisotropy behavior as our films, similar stripe-like domains were also present, and they were associated with the observed tetragonal deformations due to Ga-pairing. We already established from our XRD/RHEED measurements that samples 824, 825, and 828 exhibit rather large tetragonal distortions. Whether or not a similar Ga-pairing mechanism is the cause for these distortions, it is clear that PMA is present in our films.

Fast Fourier transform (FFT) analyses on the images from Figs. 5(a)–5(c) show that samples 824 and 828 have some preferential domain alignment along [100]; that is not the case for sample 825, however, where the spherical FFT indicates more randomly oriented domains. Although all FFTs show some domain width non-uniformities, as indicated by diffuse halos with some distribution of radii, we were able to extract some average values for the domain widths λ [see Fig. 5(d)]. The most Ga-rich sample 824 exhibits the smallest λ, of 103 ± 23 nm; domain widths for 825 and 828 are found to be similar in value with 184 ± 21 nm and 190 ± 26 nm, respectively. A similar trend as we observe here, specifically a decrease in λ as x increases, was observed for both bulk and MBE-grown FeGa alloys.^9,20,21 Comparisons can be made with FeGa/ZnSe films⁹ finding that our values for films 825/828 and 824 are close (albeit slightly larger) to values of $≈$150 nm and $≈$ 80 nm, respectively (we note that the values for FeGa/ZnSe were obtained from MFM images taken at remanence). From the M-H loops, we calculate the average magnetic moment per Fe atom [Fig. 5(e)], finding a decrease in moment with increasing x (from 2.03 to 1.83 μ_B), in agreement with the expected trend and with previous reports.²² 

Due to the observed differences in morphology, hysteresis, and magnetic domain structure of the FeGa films, we must consider the influence of the substrate. The bulk structure of Mn₃N₂(001) consists of 1 Mn layer followed by 2 MnN layers, whereas MnN(001) only contains MnN layers.²³ Our RBS measurements indicate the formation of a mixed layer (of a few nm) consisting of an Fe-Mn-Ga-N alloy for samples grown onto aFM substrates. This can be understood for sample 828 based on our previous results of sub-monolayer Fe depositions onto Mn₃N₂,²⁴ which showed that Fe atoms replace Mn atoms in the Mn layers only, followed by Mn (and possibly N) release at the surface; the intermixing is somewhat surprising though for samples 824 and 825, due to the lack of Mn layers within MnN. However, from RBS, we also extract the composition of the manganese nitride layers, finding Mn:N ratios of $≈$ 1.13 for MnN, and $≈$ 1.50 for Mn₃N₂, in excellent agreement with previous reports.²³ Note that the Mn:N ratio for MnN is larger than one, implying that N vacancies are present in the structure, as also established from previous investigations.^12,23 A plausible explanation for the existence of the alloyed interface layer between FeGa and MnN would then be that an Fe and/or Ga incorporation takes place within the MnN subsurface layers due to the presence of N vacancies, with additional possible releases of Mn and/or N atoms to the surface. Although the structure of this mixed interface is not known, there is a clear correlation between its presence and: (i) the rougher surfaces observed for samples 825 and 828 and (ii) the large discrepancies found between the LCs of similar composition samples grown on MgO and manganese nitride, which could then have a direct impact on the magnetic properties through the appearance of PMA.

In summary, we have studied FeGa films deposited onto antiferromagnetic manganese nitride and insulating magnesium oxide, finding that all films are single-crystalline and grow with a 45° in-plane rotation and (001) orientation on all substrates. Using STM, we investigated how the surface morphology changes with substrate and sample temperature. Some of the STM images presented here could be a great starting point for establishing structural surface models that could perhaps then indicate how the Ga atoms are arranged in the overall bulk structure, so that direct comparisons with theoretical predictions can be made. We have also revealed that the magnetic properties of FeGa films can be easily tuned by choosing different underlying substrates. The mixed interface between manganese nitride and FeGa over-layers could give rise to the complete loss of in-plane anisotropy and also to the observed tetragonal deformations. These distortions may be a direct cause for the occurrence of perpendicular magnetic anisotropy. Studying in more detail the effect of the alloyed interface on the resulting properties of FeGa could provide further insight into the observed magnetic anisotropies. Although not entirely unavoidable, one could try to control the extent of interface alloying by using, for example, lower growth temperatures and/or slower deposition rates.

This research was supported by the National Science Foundation under Award Nos. DMR-1206636 and DMR-1507274 (VSM magnetic characterization) and in part by the Center for Emergent Materials, an NSF-funded MRSEC under Award No. DMR-1420451 (XRD characterization). The authors thank Dr. M. Kordesch for substrate back-coating. WSxM is also acknowledged for image processing.²⁵