Controlling In-Plane Magnetic Anisotropy of Co Films on MgO Substrates using Glancing Angle Deposition

The ability to control the in‐plane magnetic anisotropy of a thin film is important for magnetic device applications. One way of accomplishing this task is by glancing angle deposition (GLAD). Herein, thin Co layers have been deposited using GLAD magnetron sputtering on MgO(001) and MgO(110) substrates. For Co films on MgO(001), the in‐plane anisotropy direction can be directly controlled via the deposition angle. In contrast, for Co on MgO(110), the anisotropy due to the deposition angle competes with the anisotropy induced by the substrate, while the growth parameters determine which contribution dominates. On the other hand, while on MgO(001) the deposition angle as well as the film thickness affect the strength of the Co in‐plane anisotropy, no influence of these parameters on the magnetic properties is found for films on MgO(110).

correlate the microstructure with the magnetic anisotropy. We found that whereas on the (001) surface the deposition conditions fully control the magnetic properties of a Co film, the (110)-surface-induced anisotropies dominate over the deposition-induced effects.

The GLAD Process
The samples were grown by GLAD in a home-built, eight-source magnetron sputter deposition system (top view of the chamber is shown in Figure 1a, and the sample geometry is shown in Figure 1b). Five of the sources are mounted in the horizontal plane, while Pt and Co are 17°below the horizontal plane, having their focus in the center of the chamber.
The deposition geometry is illustrated in Figure 1b. The substrate is mounted in the center of the chamber and it has two axes of rotation, φ S around the (vertical) z-axis and θ S around the surface normal (of the substrate) N S (for example, θ S ¼ 90°, rotating the substrate axis from S 2 to S 1 ). The angles relevant for the film growth, α and β, are shown in Figure 1b and are defined as follows: α i is the angle between the substrate normal N S and the direction to the ith deposition source (position vector M i ), i.e., ð90°À α i Þ is the incident deposition angle onto the sample (the GLAD angle); p i is the projection of M i onto the sample plane, and β is the angle between the coordinate axis of the substrate S 1 and the projected vector p i . In other words, β is the in-plane direction of deposition, i.e., the imprinting angle. Both α and β can be chosen via the azimuthal φ S and polar θ S substrate rotations. A more detailed discussion of the dependence of thin film magnetron sputtering on the deposition geometry can be found in Chapter 6 in ref. [29].
The samples investigated here were mounted such that β ¼ 0°i s parallel to the 001 h idirection of MgO(110) and parallel to 100 h i of MgO(001), respectively. In the right panel of Figure 1, a top-down view of a MgO(110) substrate is shown to clarify the angle β. In the illustrated example, the imprinting angle of β ¼ 13°is achieved by a sample rotation of θ S ¼ 30°. It also shows the definition of the angle φ H , i.e., the direction of the applied (in-plane) magnetic field H for the characterization experiments conducted on the samples. Here, the example of φ H ¼ 60°is shown. By definition, for φ H ¼ 0°and β ¼ 0°, the imprinting direction and the direction of the applied field coincide. The geometry of GLAD-induced columnar growth, along with all relevant angles and orientations, is illustrated in Figure 2.

Sample Growth
Pt/Co/Pt thin film trilayers were deposited on MgO(001) and MgO(110) substrates using DC magnetron sputtering. The films were deposited in an ultrahigh vacuum chamber, with a base pressure < 4 Â 10 À9 mbar, using 99.999% pure Ar as a sputtering gas at a pressure of % 2 Â 10 À3 mbar. Prior to growth, the substrates were baked in UHV at 650°C for 3 h. The substrates were left to cool to room temperature before a Pt layer was deposited with a deposition rate of 0.19 Å s À1 (and a thickness on the order of 10 Å; see respective sample description, e.g., in Figure 3 for precise value). This layer was deposited at "normal incidence" for cosputtering conditions (i.e., with α Pt ¼ 17°) and with continuous rotation around the substrate normal, which resulted in epitaxial growth of Pt. Subsequently, a Co layer was deposited under GLAD conditions with varying α Co ¼ 80°, 60°, 67°, 50°; β Co ¼ 0°, 30°, 60°; and thicknesses from 20 to 500 Å. An effective deposition rate of 0.16 À 0.27 Å s À1 was used depending on α, corresponding to a deposition rate of 0.44 Å s À1 for deposition under normal incidence (α ¼ 0°). Finally, the sample was capped with a nominally 10 Å-thick Pt layer (at α Pt ¼ 17°a nd using continuous sample rotation) to protect against oxidation. the x-y source plane around the sample manipulator axis (z axis). b) Orientation of the substrate in the reference frame of the chamber with φ S being the rotation around the z-axis and θ S around the surface normal (of the substrate) N S . The deposition sources are located at M i , i.e., at an angle α i away from the surface normal N S . Consequently, ð90°À αÞ is the GLAD angle and β the in-plane deposition angle. c) Geometry for a MgO(110) substrate with the angles defined relative to its crystallographic axes.

. XRR and XRD
A Rigaku Smart Lab X-ray diffractometer was used to measure X-ray reflectivity (XRR) and various types of X-ray diffraction (XRD) patterns using Cu Kα radiation. The XRR data were fitted using GenX. [30] Out-of-plane diffractograms (ω-2θ scans), as well as pole figures of selected reflections, were recorded to determine the epitaxial relationships of the Pt and Co layers and the MgO substrate.

Magnetometry
In-plane hysteresis loops were measured at room temperature on a MicroSense vibrating sample magnetometer (VSM) capable of bias fields up to AE1.8 T. The samples were mounted on a quartz rod that was rotated through various in-plane angles φ H , allowing for the angular dependence of the remanent magnetization to be determined. Hysteresis loop data points were averaged five times, except for the thinner samples with d Co ≤ 50 Å where the average was over 15 data points to improve the signalto-noise ratio.

Electron Microscopy
Cross-sectional transmission electron microscopy (TEM) samples were prepared both by focused-ion-beam (FIB) milling in a JIB-4700F Multi Beam System and by conventional methods that include mechanical thinning and Ar ion milling. Structural studies were performed by TEM imaging in a JEOL JEM-2100 microscope operated at 200 kV. Prior to the FIB preparation, the samples were sputter-coated with a %20 mm-thick Au film to protect the sample from beam damage as well as to reduce charging.

Structural Properties
Out-of-plane XRD ( Figure 3) for films on MgO(001) shows, apart from the substrate peak, weak hcp Co(0002) and either fcc Co or CoPt (111) reflections at 2θ ¼ 44.4°and 41.2°, respectively. The lack of other reflections for films on MgO(001) shows that there is no dominant crystalline order along the out-of-plane direction. This behavior is independent of the Co film thickness (from 20 to 500 Å). In contrast, samples on MgO(110) also show reflections that can be attributed to an epitaxial Pt(220) layer (2θ ¼ 67.5°), as well as three Co reflections. First, at 2θ ¼ 44.4°, there is the hexagonal Co(0002) peak, then at 41.2°possibly the (111) peak of either fcc Co or CoPt, and lastly at 83.8°the hexagonal Co(1013) peak. We used XRR to determine the thicknesses of the grown layers. We confirmed that the layer thicknesses match the nominal values of d buffer&cap Pt ¼ 13 Å and the stated d Co within the uncertainties, taking the interfacial roughness into account. The roughnesses are on average σ buffer Pt ¼ 5 Å, σ Co ¼ 5 to 35 Å, and σ cap Pt ¼ 13 Å, were σ Co increases with d Co and deposition angle α as evidenced by the roughness increase ( Figure 3, inset). The large value of σ cap Pt is due to the roughness of the underlying Co layer. The inset to Figure 3) shows representative data for three-layer stacks with increasing deposition angle, exhibiting a decrease in contrast (i.e., an increase in roughness).
The pole figure of the Co(0002) reflection on MgO(001) (Figure 4a) shows that there is intensity predominantly in the out-of-plane direction, i.e., a weak peak is visible at the origin in accordance with the out-of-plane XRD results in Figure 3. In contrast, for films on MgO(110) (Figure 4b), two pairs of relatively localized off-center spots with two-fold symmetry are found, inheriting the two-fold symmetry from the underlying MgO(110) substrate. Two equal intensity spots are located at ϕ ¼ 0°=180°with χ % 45°, and two stretched spots with different intensity at ϕ ¼ 90°=270°with χ between 25°and 40°.   (110). [25] In our case, the expected Co(1100) reflection was not observed. It is thus clear that while the MgO(110) substrate templates directional order onto the Co films, the deposition geometry is also affecting the film growth, however, without allowing for a direct control of the crystallographic orientation. Three samples, two films on MgO(110) (Figure 5a,b) and one on MgO(001) (Figure 5c), were investigated in detail using TEM. Imaging was done from two in-plane directions, along the 001 and the 110 zone axes of the MgO(110) substrates and along 100 h i and 010 h i of the MgO(001) substrate, respectively, corresponding to φ ¼ 0°and 90°for both types of substrates. The TEM images of the film on MgO(001) shown in Figure 5c are from a sample in which imprinting of the magnetic easy axis was successful. In contrast, Figure 5b,d are from a film on MgO(110) where the easy axis did not follow the imprinting direction. Interestingly, for certain deposition geometries, also films on MgO(110) can show successful imprinting, accompanied by a similar polycrystalline appearance (Figure 5a) as in the case of films on MgO(001). In all samples, analyzing the TEM images shown in Figure 5, we indeed found tilted columnar structures with a tilt angle that closely follows the tangent rule tan α ¼ 2 tan γ, where γ is the tilt angle of the columns. [19,31] The geometry of the inclined columns, along with all relevant angles and dimensions, is shown as side and top views in Figure 2. Figure 5a,c exhibit γ % 36°and 38°, which compares well with the expected 41°from the tangent rule for α ¼ 60°. For the film shown in Figure 5b,d, in contrast, the tilt is γ % 45°, which is further away from γ ¼ 70°expected for α ¼ 80°. Note that at these high angles, deviations from the rule can be expected. Along φ ¼ 0°(the 001 h i zone axes), the columns are more difficult to see since they are viewed from a less steep angle and therefore appear wider. The projected column widths w 0°a nd w 90°, visible along the two orthogonal cuts of the samples, were used to estimate the in-plane angle β (see Figure 2). Assuming a constant column width along the deposition direction, and crossing the ridges perpendicular to it, they are related through w 90°= w 0°¼ tan β, where w φ are the projected widths of the columns along the respective φ (or cut) directions (see Figure 2). For the three investigated samples, we found β to be 35ð7Þ° (Figure 5a), 43ð11Þ° (Figure 5b), and 29ð9Þ° (Figure 5c), respectively, which are very close to the respective nominal imprinting angle β. The averages (standard deviations) for β are given, as determined by the analysis of the widths in the respective TEM image.
The most noticeable feature when comparing the TEM images of Co films on different MgO substrate systems is that the films on MgO(110) (Figure 6a) show a higher degree of crystallinity. Co films on MgO(001) (Figure 6b), in contrast, are characterized by polycrystallinity and a high degree of disorder. In Figure 6a,b, we show the corresponding diffractograms. For Co on MgO(110), clear features of hcp Co with its hexagonal motif can be seen, whereas, for Co on MgO(001), the diffractogram is disordered. It shows no clear symmetry from which the dominance of either www.advancedsciencenews.com www.pss-a.com the hcp or fcc phase could be concluded. The multitude of spots instead points toward a mixed fcc-hcp phase. Indeed, when looking at the selected area electron diffraction images in Figure 6c taken of the sample, both Co phases are present. Note that for the structurally more ordered sample, magnetic imprinting failed, and the magnetic anisotropy rather followed the direction dictated by the MgO(110) substrate.

Magnetic Properties
Magnetization curves were recorded at different in-plane angles φ H . As an example, the hysteresis loops for a 100 Å-thick Co film on MgO(001) with α ¼ 60°and β ¼ 30°are shown in Figure 7a for various φ H . We found the magnetization within the growth series (see Section 2.2) to always have its easy axis in-plane, which may be unexpected given the only weakly out-of-plane oriented hcp structure of the Co layer (as determined by XRD), however, not surprising considering the demagnetization factor of a thin film. For thin (≤30 Å) Co films grown at room temperature on MgO(001), a four-fold magnetic anisotropy can be expected, with the easy axis directions being along Co 110 h i. [32] The remanent magnetization, M r , was extracted from the loops at each in-plane angle φ H and plotted as a function of φ H in Figure 7b. The normalized remanence M r =M s ðφ H Þ was fitted with a model for twofold anisotropy, M r =M s ¼ Aj sinðφ H þ φ HA Þj þ B, where A, B, and φ HA are fitting parameters. The fit to the data points shown in Figure 7b reveals the easy axis of magnetization at φ EA ¼ 121°, which is at 90°from the hard axis, i.e., φ EA ¼ φ HA þ 90°. Figure 8 shows anisotropy plots from three 50-Å-thick Co films deposited under the same conditions on MgO(001), only varying in β. For each film, the magnetic easy axis is always at 90°from the imprinted axis at β, thereby demonstrating perfect control of the magnetic anisotropy using GLAD. In total, we deposited 14 samples on MgO(001) with varying d Co and α, all showing full anisotropy control via β (with the easy axis at 90°þ β).
In contrast, for films on MgO(110), the situation is more complicated. For example, when depositing 50 Å-thick Co films (α ¼ 60°, β ¼ 30°) on MgO(110) under nominally identical conditions, we obtained in only a few cases almost perfect anisotropy control with φ EA ¼ 138°, while most of the samples showed φ EA close to 180°. The latter has the easy axis essential almost www.advancedsciencenews.com www.pss-a.com perfectly aligned with the 001 h i direction of the substrate (180°), i.e., the substrate orientation governs the magnetic anisotropy of the film irrespective of the GLAD angle. The situation was the same for other combinations of d Co , α, and β, and only occasionally anisotropy control was achieved, i.e., the is no correlation between anisotropy control and the deposition parameters.
In principle, for thin films, one can expect a thickness dependence of the in-plane uniaxial anisotropy energy K U , usually increasing with decreasing film thickness d. [16,22] A measure of the anisotropy energy is given by the difference in remanent magnetization ΔðM R =M S Þ between the easy and hard axes. Plots of ΔðM R =M S Þ, which are proportional to the anisotropy energy, are shown as a function of d Co and the deposition angle α in Figure 9a,b, respectively. For films on MgO(001), the observed thickness trend agrees with the expectation that thinner films have higher anisotropy energy than thicker ones due to their shape anisotropy. In the XRD spectra of the thicker films (>300 Å), signs of increased crystallinity of the Co layer were found (see Figure 3), which could be the cause for the decrease of M R =M S as well. Note that the error is larger for thinner samples since there is less material to measure, and the relative error in fitted thickness used to calculate the magnetic moments is larger. In terms of deposition angle, a complex dependence of K U on α has been reported, with the energy maximum found for an intermediate angle. For d Co ¼ 50 Å, the smallest energy is found for α ¼ 50°, and it is continuously increasing for 60°a nd 70°until there is a decrease at very high angles (80°).
In contrast, films grown on MgO(110) show no clear trend. The anisotropy energy of most films on MgO(110) is several times larger than those on MgO(001), demonstrating that the anisotropy, in this case, is caused by a much stronger effect, competing with the imprinted anisotropy from GLAD. This effect clearly originates from the substrates, as the easy axis always points along the 001 h i direction of MgO(110). Note that in rare cases, we found the anisotropy of films on MgO(110) to be comparable to those on Mg(001); however, we did not The red curve is a fit to the data using a two-fold anisotropy model, with the red arrow pointing along the easy axis. The imprinted easy axis is expected to be at 90°from the in-plane deposition axis (green arrow). In this case, β ¼ 30°and the resulting φ EA ¼ 121°agrees well with the expected value. www.advancedsciencenews.com www.pss-a.com study their behavior in more detail due to the inconsistent dependencies.

Discussion
By employing the GLAD process, [19][20][21] shadowing will lead to tilted columnar growth. The resulting structural anisotropy of the tilted columns, and possibly also of their noncircular cross-section, [22] will further affect their magnetic anisotropy. However, the overall magnetic properties of the films not only stem from the columnar magnetic properties but are also determined by the shape and surface anisotropies of the film (i.e., depending on the film thickness) and substrate interactions. For example, for Fe films on amorphous substrates, high-incidence angles result in needle-like columns and low-incidence angles in an extended wall-like structure (perpendicular to the plane of incidence), with the in-plane magnetic easy axis parallel and perpendicular to the incidence plane, respectively. [16] Note that the columnar structures in thicker films have a defined crystallographic orientation, which is dependent on the growth rate. [16] For ultrathin Fe layers on sapphire, in contrast, where the films are polycrystalline, full control of the in-plane magnetic anisotropy has been achieved via the deposition angle. [18] The behavior of Co films can be expected to be different from that of Fe films, resulting from differences in magnetocrystalline anisotropy. However, note that while polycrystalline Co has a larger magnetocrystalline anisotropy than Fe, [18] the change of crystal structure in Co columns from hcp to fcc can lead to a lower anisotropy. [22] For example, for high-incidence angle deposited Co films on amorphous substrates, a transition in the magnetic anisotropy was observed at a critical thickness of 300 Å, below which the easy axis is in-plane and above which the easy axis is tilted away from the film plane. [22] The tilted easy axis of the thicker films, which is confined to the incidence plane, has been tied to the microstructure of the fcc Co columns. In contrast, the in-plane anisotropy of the thinner films was explained by the fact that instead of pronounced columns, they are dominated by elongated nuclei. Given the low magnetic anisotropy of randomly oriented fcc Co grains, the shape anisotropy of the elliptic nuclei is believed to be the reason for their in-plane easy axis perpendicular to the incidence plane. [22] The presence of elongated nuclei in the early stages of film growth was determined by atomic force microscopy (AFM) imaging.
In our Co films on MgO(001), in which imprinting was achieved with the easy axis always perpendicular to the in-plane deposition angle, we find no evidence of uniaxially elongated Co nuclei from AFM imaging, which could explain this observation via the shape anisotropy tied to the microstructure. Further, while the in-plane anisotropy increases with Co film thickness (Figure 9), no transition of the magnetic anisotropy from inplane toward the out-of-plane direction is observed, which would have been the result of pronounced columnar growth pulling the easy axis out of the plane. As can be seen for % 300 Å-thick Co films on MgO(001) in Figure 5c and 6b,c, the columns are not very pronounced compared to the TEM image shown in ref. [22].
Our Co films on MgO(001) show a two-fold in-plane anisotropy (Figure 8), strictly tied to be perpendicular to the in-plane deposition angle. This means that the MgO(001) substrate with its four-fold symmetry has no direct effect on the structural properties of the Co films, as it would be the case for physical vapordeposited films under normal incidence. [32] In contrast, Co films on MgO(110) show a strong influence of the substrate on the magnetic properties, which makes this substrate orientation unsuitable for GLAD deposition. It should be noted that successful imprinting has been achieved for extreme angles (α ¼ 80°); however, given the observed differences in magnetic properties of nominally almost identically grown films, it appears that other effects, e.g., the polishing and surface quality of the commercial substrates, have a stronger influence on the film parameters than any of the GLAD parameters.

Conclusions
Here, we have demonstrated the control of the in-plane magnetic anisotropy of Co thin films on MgO(001) using GLAD. By carrying out VSM measurements with varying azimuthal angles, we found that the direction of the magnetic easy axis is always perpendicular to the in-plane deposition angle β. The strength of the anisotropy energy, as estimated by the difference in remanent magnetization between the easy and hard axes, is determined by the film thickness and the GLAD angle. In contrast, on MgO(110), the crystalline anisotropy dominates over the imprinting effects, yielding a magnetic anisotropy that is dictated by the substrate. Therefore, the easy axis is along the substrate's 001 h i direction, independent of the film thickness, the GLAD, or the deposition angle. While GLAD results in a tilted columnar microstructure on both MgO(001) and (110) substrates, the anisotropy induced by the substrate, if present, is stronger than the GLAD-induced anisotropy tied to the microstructure. The control of the magnetic anisotropy of magnetic layers on crystalline substrates is of great importance for magnetic device fabrication, and industry-compatible GLADbased magnetron sputtering is a promising approach. www.advancedsciencenews.com www.pss-a.com