Enhanced positive and negative exchange bias in FeF₂/Ni with dusted interfaces

The exchange bias (EB) phenomenon has widely been studied in the last few decades because of its relevance in magnetic sensing, storage media and fundamental knowledge of physical phenomena.^1–5 The exchange interaction between antiferromagnetic (AFM) and ferromagnetic (FM) spins produces a shift of the hysteresis loop and the pinning of the FM layer. This effect is used in sensors and spin-valves showing giant or tunnel magnetoresistance.^6–8 Moreover, EB plays an important role in novel spin–orbit torque functionalities, as current-induced switching of nano-oscillators or spin logic devices.^9–12 

The exchange coupling is a short-range interaction; therefore, this effect is mainly produced at the AFM/FM interface. Many theoretical models and experimental evidence have corroborated this origin.^13–17 Pinned spins at the AFM interface determine the magnitude of the exchange bias field (H_EB).^18–21 However, in certain systems, the bulk AFM structure may determine the interfacial configuration and, therefore, define the EB properties of the system. The role of the AFM bulk was confirmed in experimental thin film measurements and theoretical calculations.^22–25 

Despite the short-range nature of the exchange interaction, EB effects were also observed in AFM and FM films separated by non-magnetic (NM) spacers. The EB field monotonously decreases with the thickness of the NM spacer and vanishes beyond a few nanometers.^26–30 The spacer layer prevents the interlayer exchange coupling in these systems. Recently, a model based on dipolar interactions between induced AFM domains and FM moments was proposed to account for the experimental results.³¹ The model also explains the twofold sign of EB experimentally observed in AFM/NM/FM trilayers, where the hysteresis loop exhibits both positive and negative shifts tuned by magnitude of the cooling field.³¹ 

An interesting intermediate case is found between a magnetically exchange coupling and a dipolar coupling across an NM spacer, i.e., AFM/NM/FM heterostructures with an ultra-thin spacer layer, formed by a discontinuous NM material. This configuration is referred to as a diluted or “dusted” interface. Atomic-scale numerical models simulated diluted interfaces with either magnetic or nonmagnetic defects, taking into account intermixing of AFM and FM sites or the insertion of NM impurities at the interface.^32–36 These studies predict a notable effect on the magnitude of the EB field, which can be maximized for a certain degree of interfacial spin disorder. Only a few experimental structures were used to investigate this effect. For example, a large enhancement of the EB field was found in [Pt/Co]_n/Pt/AFM (AFM = FeMn, IrMn) by the insertion of an ultra-thin discontinuous Pt spacer.^37,38 However, no enhancement was observed for Al, Ru, and Cu spacers. The perpendicular EB enhancement in this multilayer structure was attributed to the induced transition from in-plane to perpendicular anisotropy by the Co-Pt spin–orbit coupling and not to a fundamental effect on pinning centers at the interface. Other studies revealed that Pt insertion acts as a barrier against Ir and Mn diffusion into the Co layer, improving the perpendicular EB effect.^39,40 A Pt spacer also yielded an increase in EB in IrMn/Co bilayers with in-plane anisotropy only. However, the insertion of other metallic impurities as Cu, Ta, and Au produced a reduction of the EB field for any spacer thickness.⁴¹ On the contrary, Vinai et al. found an EB enhancement in a similar system with Cu dusting (IrMn/Cu/Co).⁴² Mn dusting also increased the EB amplitude of IrMn/CoFe bilayers. Although the Mn insertion could modify the chemical composition of the IrMn layer at the interface, the unidirectional anisotropy enhancement was much larger in dusted interfaces than that of bilayers with a higher content of Mn in the AFM layer.^43,44 Other elements such as Ru, Pd, Ta, Gd, and Tb yielded an opposite effect, a reduction of the EB field by inserting an ultra-thin spacer layer. Therefore, the effect of dusting the interfaces on the EB phenomenon is still an intriguing issue that can be exploited to improve the properties of EB-based devices.

FeF₂/Ni bilayers with compensated AFM interfaces exhibit positive and negative EB, which is determined by the magnitude of the cooling field. Here, we show that this feature is also replicated with nonmagnetic Pd or Cu dusted interfaces. We observe that not only the Pd or Cu dusting enhances the EB magnitude but also the effect is symmetric for the two signs of EB. Moreover, dusting the interfaces reduces the field cooling threshold from negative to positive EB, which demonstrates that NM atoms at the interface affect the whole domain configuration of the AFM. The twofold sign of EB is a distinctive feature of antiferromagnetically coupled AFM/FM systems.⁴⁵ Large cooling fields align net AFM moments with the FM magnetization during the cooling process. Thus, during the magnetization curve, the AFM-FM coupling is frustrated above the saturation field; therefore, the FM reversal occurs at positive fields in the decreasing branch of the hysteresis loop.

To investigate dusting effects in compensated AFM/FM interfaces exhibiting positive and negative EB, FeF₂ (70 nm)/NM (t_NM)/Ni (20, 40 nm)/Al (2 nm) wedge-shaped trilayers were fabricated by electron beam evaporation onto MgF₂ (110) single crystals. FeF₂ deposited at a base pressure of 5 × 10⁻⁷Torr and 300 °C grows epitaxially on MgF₂ with a surface orientation (110), which corresponds to a magnetically compensated plane of the FeF₂ crystallographic structure. Subsequent metals were deposited at 150 °C and protected with an Al capping layer. Figure 1 illustrates the preparation of wedge-shaped films. The shutter is located at the home position for the evaporation of FeF₂, Ni, and Al, but it is positioned at x = 0 to start with the NM material wedge. The distance between home and x = 0 positions is about 2 mm. Thus, in this area, there is a clean interface between FeF₂ and Ni without NM material. This is the reference area to compare the effect of NM atoms on EB. The NM layers extend from x = 0, and its thickness increases with x. The shutter speed was synchronized with the deposition rate to set the thickness gradient of the NM wedge. Deposition rates were 0.1 nm/s for FeF₂ and Ni and 0.03 nm/s for NM materials. A typical thickness gradient for the NM layer was 0.3 nm/mm.

Samples were cooled from room temperature at low (≤200 Oe) and high (2 kOe) cooling fields (H_FC). The EB field was obtained from hysteresis loops measured by the magneto-optical Kerr effect below the FeF₂ Néel temperature (78 K). The laser spot was around 100 μm in diameter. Consecutive hysteresis loops were recorded at steps of 100 or 200 μm. Thus, both the reference area and the dusted area of the sample were probed from the same measurement sequence, which allows for a straightforward comparison of EB in dusted and clean FeF₂/Ni exchange-coupled interfaces. Figure 2 shows examples of magneto-optical measurements at 5 K for positively and negatively shifted hysteresis loops. The EB sign was tuned by H_FC. The large difference in the EB magnitude is due to the Ni thickness; 40 nm in the Pd wedge and 20 nm in the Cu film.

Figure 3 shows the dependence of H_EB with the NM-spacer thickness for NM = Pd and Cu. The reference signal for H_EB is based on the initial areas where H_EB was kept constant. The deviation of the EB field from this value indicates the starting point of the NM wedge, indicated as t_NM = 0 in Fig. 3. This point was checked for each sample with the shutter shadow marked on the substrate. Both positions agree within experimental error (±1 mm). Thus, t_NM < 0 indicates in Fig. 3 the area of direct FeF₂/Ni contact, while t_NM > 0 corresponds to the NM wedge.

Figure 3(a) shows the H_EB dependence for Pd dusting at 5 K and two different cooling fields. H_FC = 100 Oe produces a hysteresis loop shifted in the negative direction (blue squares), whereas H_FC = 2 kOe produces a positive EB (red circles). The fine blue line corresponds to the absolute value of the negative EB for H_FC = 100 Oe. The overlap with the positive EB dependence demonstrates that the two EB signs mirror each other, not only at a clean FeF₂/Ni reference demonstrated earlier^46–48 but also in the dusted region (FeF₂/Pd/Ni interface) where the enhancement was found. The maximum up to 24% enhancement of H_EB, above the magnitude of clean interfaces, occurs at an estimated 0.1 nm Pd thickness. Since for this thickness, it is not possible to form a continuous spacer layer, a direct exchange interaction between FM and AFM spins must still be present. The enhancement of the exchange anisotropy can be attributed to isolated Pd atoms trapped at the interface. As the density of Pd atoms at the interface increases, the exchange coupling diminishes and becomes comparable to the interaction in clean interfaces at t_Pd = 0.3 nm. Above this spacer thickness, H_EB decreases monotonously. The EB improvement keeps with temperature, and at T = 65 K, this increase is also 24% for t_Pd ≈ 0.1 nm.

A similar H_EB dependence was observed for FeF₂/Cu/Ni interfaces, as illustrated in Fig. 3(b). Low cooling fields show an EB enhancement of 21% by dusting with a Cu layer of 0.1 nm. The improvement was also noted for the interval 0 < t_Cu < 0.3 nm, as in Pd samples.

It should be noted that a cooling field H_FC = 2 kOe was not sufficient to produce positive EB in clean FeF₂/Ni interfaces. The dusted interface favors the transition to positively shifted hysteresis loops (red circles), and H_EB becomes positive for t_Cu = 0.3 nm. This finding implies that dusting Cu atoms at the interface influence the bulk AFM domain structure. It was demonstrated that positive EB requires the Zeeman energy of AFM domains to overcome the antiparallel interaction between AFM and FM spins during the cooling process.^45,49 Thus, the presence of positive EB for t_Cu = 0.3 nm and not for t_Cu = 0 nm, for the same cooling field, implies a reconfiguration of bulk AFM domains.

Figure 4 illustrates qualitatively how NM atoms can contribute to the EB enhancement in compensated interfaces and the reconfiguration of AFM domains. The magenta line marks the interface between AFM spins (blue arrows) and the FM layer (black arrows). The topography of the interface is a source of uncompensated spins (red arrows), which can also be generated in the bulk by crystallographic defects (white holes). The imbalance in the two signs of AFM spins leads to the formation of bulk domains during the cooling process, separated by a narrow domain wall (DW) (brown line). In FeF₂/Ni interfaces, the net domain magnetic moment aligns opposite to the FM magnetization, because of the antiparallel coupling between AFM and FM spins–origin of positive EB in this system.^45,50 Unbalanced AFM-FM couplings at the interface highlighted by ellipses provide the pinning for the EB, Fig. 4(a). The introduction of NM atoms at the interface (green dots) modifies the density of pinning centers, as shown in Fig. 4(b). They can lead to favorable AFM-FM interactions—as in domain 1—or frustrated couplings that increase the domain energy—as in domain 2. In the latter case, it might be more advantageous to freeze this area in the opposite direction, which moves the DW as shown in Fig. 4(c). This configuration enhances the magnitude of the EB field by increasing the number of pinning sites. In addition, NM atoms can also displace AFM atoms during the deposition, increasing the roughness of the interface. This is a source of uncompensated spins that increase the net magnetic moment of the domain and, consequently, reduce the cooling field required for positive EB, as observed in Fig. 3(b). It is worth noting that impurity distributions in real samples can be different and much more complex due to intermixing and alloying effects, broader grain boundaries, or cluster formations. Thus, different NM materials can lead to different trends. We also prepared dusted interfaces of Ag, Ti, V, and SiO₂ in FeF₂/Ni bilayers but no EB enhancement was observed for these elements. Thus, Fig. 4 connects two main concepts to our experimental observations: the density of uncompensated pinned moments that determine the magnitude of the EB and the net magnetic moment of AFM domains that sets the EB sign with the cooling field.

In summary, our findings demonstrate that ultrathin spacer layers can generate a twofold effect in compensated AFM-FM interfaces: (i) enhancement of the EB by the generation of new pinning centers in the vicinity of NM atoms and (ii) threshold reduction of the cooling field necessary for positive EB by rearrangement of AFM domains with higher net magnetic moments. Moreover, the EB enhancement is symmetric in interfaces that exhibit both positive and negative EB. The intrinsic improvement of the interfacial exchange coupling energy due to new pinning centers can be used to increase and tune the sign of the EB field without disrupting the FM/AFM coupling, even at a compensated AFM surface.

This is a highly collaborative research study. The experiments were conceived jointly, the data were extensively debated, and the paper was written by multiple iterations between all the coauthors. Sample preparation and measurements at UCSD were supported by the Office of Basic Energy Science, U.S. Department of Energy, BES-DMS funded by the Department of Energy's Office of Basic Energy Science, DMR under Grant No. DE FG02 87ER-45332. This work received funding from No. AEI FIS2016-76058, UE FEDER “Una manera de hacer Europa;” the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie Grant Agreement Nos. 734801 and AEI-PID2019-104604RB, and the Basque Country Grant Nos. IT1162-19 and PIBA 2018-11. M.K. and F.T. acknowledge support from Financiamiento Basal para Centros Científicos y Tecnológicos de Excelencia (No. CEDENNA FB180001). The UPV/EHU authors are grateful for the technical and human support provided by the Laser Facility and Magnetic Measurements units of SGIker UPV/EHU.