Epitaxial Mn₂Au thin films for antiferromagnetic spintronics

The usual concepts of spintronics rely on the magnetization of ferromagnets (FMs) for data storage and manipulation. An alternative concept is based on the use of antiferromagnets (AFMs), e.g. by employing the direction of the staggered magnetization for data storage. An indirect read-out of this direction is possible by coupling the AFM to a FM creating an exchange spring, but also direct approaches based on magnetotransport effects exist: for instance an anisotropic magnetoresistance (AMR) effect was observed in the AFMs Sr₂IrO₄ [1] and FeRh [2]. With Cu point contacts on Sr₂IrO₄ even large magnetoresistance effects up to 28% could be obtained [3]. A prominent example is the large tunneling anisotropic magnetoresistance effect (TAMR $\simeq$ 160%) between the AFM IrMn on one side and non-magnetic Pt on the other side of an MgO barrier [4]. For such devices, ordered metallic AFM compounds combining high Néel temperature and large spin–orbit coupling effects are required [5]. However, only few compounds with such properties exist. Examples are CuMnAs [6] and Mn₂Au [7, 8]. One benefit of antiferromagnets for spintronics is the potentially ultrafast spin-dynamics, which in AFMs is typically 2–3 orders of magnitude faster than in FMs [9]. Another advantage of AFMs is their insensitivity in external fields due to their zero net magnetization. However, this property also poses a major challenge in antiferromagnet-based spintronics: New approaches for the controlled manipulation of the AFM spin-orientation have to be identified and investigated. For ordered AFMs with strong spin–orbit interaction it was recently suggested to use short current pulses for the manipulation of the spin-axis direction. Considering the tetragonal AFM compound Mn₂Au, based on Néel-order spin–orbit torques fields current densities of 10⁸–10⁹ $\text{Ac}{{\text{m}}^{-2}}$ were estimated to reorient the spin-axis within the easy ab-plane [10]. Such high current densities are of the same order of magnitude as required for spin–orbit torque switching of e.g. Ta/CoFeB structures and can only be applied as short pulses ($\simeq$10–100 ns) to narrow conductor channels ($\simeq$1–10 nm) using thin films ($\simeq$1–10 nm) [11, 12]. Band structure calculations [13] predict strong anisotropies (up to 50%) of the density of states of Mn₂Au for the different alignments of the staggered moment, i.e. either parallel to the easy (1 1 0) or to the hard (0 0 1) axis of the tetragonal compound. This anisotropy significantly exceeds the corresponding anisotropy of IrMn (about 10%), thus a larger TAMR is expected for Mn₂Au.

Experimentally it was confirmed that Mn₂Au indeed is a simple antiferromagnet with the magnetic moments on the Mn sites aligned in the ab-plane and a Néel temperature well above 1000 K [8]. It was also demonstrated that (1 0 1)-oriented Mn₂Au thin films can be grown by MBE, showing exchange bias effects with Fe layers [7]. However, for the predicted current switching of the AFM domains, the current has to be applied within the ab-plane requiring (0 0 1)-oriented thin films, whose growth and characterization is described here.

Within the preparation and analysis cluster described in [14] epitaxial Mn₂Au thin films in (0 0 1)-orientation were prepared by rf-sputtering from a single stoichiometric target on Al₂O₃(1 $\bar{1}$ 0 2) substrates with Ta(0 0 1) buffer layer. The growth rates amounted to $\simeq$0.1 nm s⁻¹, typically a buffer layer thickness of 18 nm was used for growing Mn₂Au thin films of typically 62 nm thickness.

Figure 1 shows the results of a x-ray diffraction (XRD) $\Theta/2\Theta$-scan of a Al₂O₃(1 $\bar{1}$ 0 2)/Ta(1 0 0)/Mn₂Au(0 0 1) sample indicating negligible impurity phases and a small portion of misaligned Mn₂Au(1 0 1). The inset shows a rocking curve (ω-scan) of the Mn₂Au (0 0 2) peak indicating a mosaicity of ${{0.5}^{{}^\circ}}$. The Mn₂Au phase formation was observed for substrate temperatures ${{T}_{\text{S}}}$ during deposition between 500 °C and 650 °C. However, the most narrow XRD rocking curve was obtained with ${{T}_{\text{S}}}=600$ °C. Figure 2 shows an XRD φ-scan of off-specular (1 0 1) and equivalent peaks of Mn₂Au(0 0 1) and (1 1 0) and equivalent peaks of Ta(1 0 0) demonstrating the in-plane order and epitaxial relation of the Mn₂Au thin film and buffer layer. The scans show that the ab-plane of Mn₂Au(0 0 1) has the same orientation as the ab-plane of the Ta buffer layer, as expected due to the small lattice mismatch of only 0.6%. The surface order of the Mn₂Au(0 0 1) thin films is demonstrated by the electron diffraction image (RHEED) shown in figure 3. The arrangement of the sharp RHEED spots on semi-circles is characteristic for scattering from a 2-dimensional, i.e. smooth surface with structural in-plane order.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Temperature dependent measurements of the resistivity R(T) of a lithographically patterned Al₂O₃(1 $\bar{1}$ 0 2)/Ta(1 0 0), 18 nm/Mn₂Au(0 0 1), 62 nm/AlO_x, 2 nm thin film revealed metallic behavior of the Mn₂Au layer with a residual resistivity ratio of RRR $\simeq$ 3 (figure 4). The room temperature resistivity amounts to ${{\rho}_{300\text{K}}}\simeq 22$ $\mu \Omega$cm, which is similar to the specific resistivity of the Ta buffer layer and a typical value for a good metal. The small specific resistivity of the Mn₂Au thin films is beneficial considering the required current densities and associated heating effects for the predicted current induced switching of AFM domains [10].

Zoom In Zoom Out Reset image size

To test the antiferromagnetic nature of the Mn₂Au(0 0 1) thin films, a 5 nm thick epitaxial Fe(001) layer (with a 5 nm Al protective capping layer) was deposited in situ on top of the Mn₂Au. This bilayer should act as an exchange coupled system, if cooled in a high enough magnetic field through the blocking temperature. This method was e.g. successfully used to image antiferromagnetic CoO domains at an Fe/CoO interface [15]. Considering the expected high Néel temperature of Mn₂Au [7], we used a relatively thin film (7 nm), attempting to reduce the ordering temperature. We cooled the sample from 375 K in a field of 1 T parallel to the Fe(1 0 0), i.e. Mn₂Au(1 1 0) axis. In figure 5 we show the results of temperature dependent measurements of the magneto-optical Kerr-effect (MOKE) from 300 K down to 10 K. There are two remarkable features of these hysteresis measurements: first, there is a negligible exchange shift. The slight shift observed is not reproducible upon temperature cycling, which could indicate some training effect. Second, the coercivity of the Mn₂Au/Fe heterostructure at 300 K is much larger (15 mT) than for a bare Fe film (1 mT) and increases dramatically with decreasing temperature. The coercivity almost doubles between 300 K and 10 K with the biggest change between 300 K and 175 K. This strongly suggests the presence of an exchange anisotropy acting on the Fe film, which, however, is randomly oriented in the film plane. Obviously the field cooling protocol had little effect on the alignment of the AF domains in the field direction. This, in turn, implies that the cooling field was still applied below the blocking temperature of the Mn₂Au film. However, Wu et al previously followed a similar protocol and were able to demonstrate an exchange bias effect in Fe/Mn₂Au(1 0 1) [7]. This could be related to the different growth direction of their Mn₂Au thin films or to a different morphology, which may result in uncompensated magnetic moments at the interface. For our Mn₂Au(0 0 1) thin films magnetometry using a superconducting quantum interference device (SQUID) gave no indication for any uncompensated moment.

Zoom In Zoom Out Reset image size

Figure 6(a) shows Kerr microscopy images obtained from a Mn₂Au(60 nm)/Fe(5 nm)/Al(5 nm) sample at different field values in the film plane and at 300 K. The hysteresis shown in the center is obtained by integrating the MOKE intensity over the entire image. Close to coercitivity the domain structure is grainy indicating an Fe domain size of typically 500 nm. This is by far smaller than the domain size of a Fe film on a bare substrate, as shown in figure 6(b). Furthermore, the domains do not show a preferred orientation. Thus the domain structure visible here confirms the notion already gained from MOKE hysteresis measurements that the Fe domains are associated with an exchange coupling to the Mn₂Au layer. This coupling causes an imprint of the AFM domains onto the Fe domains, allowing for a visualization of the antiferromagnetic domain structure. It turns out that the antiferromagnetic domain structure is random in space, which explains the missing exchange bias.

Zoom In Zoom Out Reset image size

In order to exclude that the observed domain structure is dominated by properties of the Fe instead of the Mn₂Au, we repeated the investigations using Permalloy instead of Fe as a ferromagnetic layer. The obtained domain structure is identical to the Fe case, providing additional evidence for a dominating contribution of the exchange coupling with the Mn₂Au thin film to the ferromagnetic domain formation.

Epitaxial Mn₂Au thin films with (0 0 1)-orientation were grown by rf-sputtering at optimized substrate temperatures on Al₂O₃ (102) substrates with Ta(1 0 0) buffer layers. The thin films show very good metallic properties and a high degree of crystallographic ordering within the film as well as on the surface. Evidence for antiferromagnetic order forming small domains in Mn₂Au was provided via magnetic characterization of ferromagnetic thin films exchange coupled to Mn₂Au. Thus rf-sputter deposition of antiferromagnetic Mn₂Au thin films of high epitaxial quality in the required orientation for future spintronics applications was demonstrated.

This work is supported by the Graduate School of Excellence Materials Science in Mainz (MAINZ, GSC 266) and the DFG.