Positive and negative exchange bias in CoNi/Gd/CoNi trilayers and CoNi/Gd bilayers

Abstract

Temperature-dependent magnetic properties of CoNi/Gd/CoNi trilayers and CoNi/Gd bilayers were investigated experimentally between room temperature and 90 K. The films were deposited by DC magnetron sputtering at room temperature. Strong positive and negative exchange bias were observed due to the diffusion of Co and Ni into the Gd layer. The larger of the Gd or CoNi layer magnetization determines the exchange bias direction due to the antiferromagnetic exchange coupling between Gd and Co, and between Gd and Ni. Cooling magnetic field strength did not change the exchange bias field strength. However, the exchange bias field became zero when hysteresis loop measurements were done in large magnetic field intervals. This may be the sign of the existence of interface domain wall for exchange bias at lower field interval and its annihilation in the larger field interval.

Introduction

Exchange bias (EB) refers to the shift of hysteresis loop from the zero field axis. Meiklejohn and Bean discovered it in surface oxidized Co particles in 1965 [1]. Mostly ferromagnetic/antiferromagnetic (FM/AF) bilayers have been investigated due to their technological importance and possibly due to their convenient form in order to study EB. EB is believed to be due to the unidirectional anisotropy developed by the exchange interaction at the interface of ferromagnetic (FM) and antiferromagnetic (AFM) layers when the FM/AFM bilayer is cooled through the Néel temperature of the AFM layer under an external magnetic field. Lately, there has been considerable theoretical and experimental effort to understand the phenomena due to its technological importance (see reviews [2], [3], [4], [5]). It has been observed in several systems (see Ref. [2] for details). In general, the hyteresis loop shifts opposite to the external cooling field in most systems (negative EB). However, recently, the shift of the hyteresis loop from negative field to positive field, thus shift in the external cooling field direction (positive EB), has been observed in Fe/FeF₂ and Fe/MnF₂ when the bilayers were cooled under an extremely high external magnetic field [6], [7], [8]. Several models have been put forward to explain the formation of positive and negative EB [9], [10], [11], [12]. Under small external fields (but large enough to saturate the FM layers), the interface coupling determines the interface spin configuration of the AFM layers. The hyteresis loop shifts to negative field for both ferromagnetic and antiferromagnetic interface coupling. However, If there is antiferromagnetic interface coupling between the FM and AFM layers, the Zeeman energy of the AFM spins at the interface in the high external field is comparable to the FM/AFM interface exchange energy. Therefore, the spins of the AFM layer at the interface align toward the external field and the hyteresis loop shifts in the positive field direction. This positive EB model has been supported experimentally [13], [14], [15]. Nice demonstration of positive and negative EB can be seen in exchange biased Co films and Co/Pt multilayers [13].

Positive and negative EB have been also observed in ferromagnetic/ferrimagnetic transition metal rare-earth alloy bilayers and multilayers [16], [17], [18], [19], [20]. Both Co/CoGd₂ multilayers [16] and TbFe/NiFeMo bilayers [17] cross over from negative to positive EB or vice versa, when they go through compensation point by either changing the temperature or the composition of ferrimagnetic alloys. The magnetization of a ferrimagnet is zero at the compensation point. Above the compensation point, transition metals dominate in the magnetization of the ferrimagnet and ferromagnetically couples to a ferromagnetic layer, hence negative EB occurs. Below compensation point, rare-earth metals contribute more to the magnetization of the ferrimagnetic layer and couple to the ferromagnetic layer antiferromagnetically and positive EB results in this case. It should be noted here that positive EB without unidirectional anisotropy could result in rare-earth-dominated ferrimagnetic/ferromagnetic bilayers with antiferromagnetic coupling at the interface of the bilayer if there is a difference between their magnetizations [18].

Multilayers of transition metals and rare-earth metals can couple antiferromagnetically. Therefore, antiferromagnetically coupled RE/TM multilayers form an artificial ferrimagnet. Camley and co-worker theoretically predicted different magnetic phases in Fe/Gd multilayers depending on the temperature and the external field [21]. At high fields, both Gd and Fe aligns in the field direction. At moderate fields, the spin of Fe and Gd make different angles with respect to the external field (twisted or spin flop phase). In small fields, the layer with the larger moment will remain in the field direction (Gd or Fe aligned state). Several investigators confirmed this prediction experimentally (for example [22]). The theoretical calculation was based on the minimization of the interface exchange energy and the Zeeman energy of the multilayer. The interfaces were considered sharp and no magnetic anisotropy was taken into account. However, ferrimagnetic transition metal rare-earth alloy may occur naturally at the interfaces of the multilayer due to solid state reaction between transition metal layers and rare-earth metal layers. This type of alloying is quite common and requires a negative heat of mixing and fast diffusion of one component into the other [23]. Such alloying has been observed in Fe/Tb [24], and in Co/Gd multilayers [25]. It was observed that Co diffuses into Gd strongly when the Co layer was deposited on top of the Gd layer than when the Gd layer was deposited on top of the Co layer. The effects of such preferential or unidirectional diffusion on magnetic properties of Fe/Tb multilayers [24] and CoNi/Gd bilayers [26] can be found in the literature. Experimentally, it was observed that CoNi(top)/Gd bilayers have higher coercivity than Gd(top)/CoNi bilayers [26].

Based on the information given above, the magnetic structure of our CoNi(top)/Gd/CoNi(bottom) trilayers and CoNi(top)/Gd bilayers can be presented schematically as shown in Fig. 1. A thin Gd layer may be amorphised totally by Co and Ni diffusion into the Gd layer. The formed alloy couples strongly to the CoNi layer and it can be magnetically hard (Fig 1a). If another CoNi layer is added at the bottom of this bilayer it would couple ferromagnetically to the top CoNi layer (Fig. 1b). The exchange coupling at the Gd/CoNi interface may not be as strong as the coupling at the CoNi/Gd interface as the Gd layer thickness is increased. If the Gd layer were thick enough to have a larger magnetic moment than the CoNi layer, the magnetic structure of the bilayer would be as in Fig. 1c. There are two possible configurations for trilayers in this Gd thickness range. If the total Gd magnetic moment were still less than the total magnetic moment of the CoNi layers, the magnetic structure would be as in Fig. 1b. If the total Gd moment were greater than the total moment of the CoNi layers, the magnetic structure would be as in Fig. 1d. These global magnetization orientations can induce spin frustration (even domains) in the interface alloy or at the interface of the alloy (if it were considered a separate layer) depending on the temperature and the alloy composition. The diffusion gradient itself can induce a spin frustration region in the alloy for a particular temperature because there might be a part of the alloy close to compensation temperature. The role of this spin frustration on EB, which is the subject of this article, is another question that needs to be addressed. The experimental observation of strong positive and negative EB in CoNi(top)/Gd/CoNi(bottom) trilayers and CoNi(top)/Gd bilayers as a function of Gd layer thickness, bottom CoNi layer thickness and temperature will be presented. It is seen that the preferential diffusion of Co and Ni into Gd layers causes EB and the layer with larger magnetization (CoNi or Gd layer) in the bilayer or in the trilayer determines the EB direction due to the negative exchange coupling between the Gd and Co, and between the Gd and Ni.

In this experiment CoNi (Ni-20 wt%), Gd, and Ag targets were used and deposited by DC magnetron sources on Corning 1737 glass substrate to form the films. Ag films of 10.0 nm thickness were used as buffer and cap layers in all trilayers and bilayers. Depositions were done in a UHV chamber with unbaked base pressure of 10⁻⁹ Torr at room temperature (RT). Deposition pressure was 3 mTorr and ultra-high-purity Ar gas was used. Deposition rates were 0.032, 0.024, and 0.04 nm/s for Gd, CoNi, and Ag, respectively. These deposition rates and thicknesses were monitored in situ by using a quartz thickness monitor calibrated with a stylus profilometer. A Lakeshore 7300 model vibrating sample magnetometer was used for the magnetic measurements. Samples can be cooled or heated between 77 and 500 K in a liquid nitrogen cryostat.

Low temperature magnetic properties of CoNi[8 nm]/Gd[t]/CoNi[8 nm] trilayer and CoNi[8 nm]/Gd[t] bilayer were measured for t=5, 8, 10, 12, 14 and 20 nm Gd thickness. The top CoNi layer thickness was kept constant at 8 nm in all bilayers and trilayers. The CoNi/Gd/CoNi trilayers decouple completely at RT for Gd interlayer thickness greater than or equal to 6 nm [27]. For 5 nm Gd thickness no EB was observed in both trilayer and bilayer. In the bilayer (t=5 nm), the coercivity was increased from 560 Oe at RT up to 1500 Oe at 90 K. The trilayer (t=5 nm) has a similar hyteresis curve as the other trilayers and preserves the double loop state down to 200 K. As the trilayer was cooled below 200 K, ferromagnetic interlayer coupling increased between the hard CoNi/Gd bilayer and the soft CoNi layer. Although the coupling between Gd and Co or between Gd and Ni is antiferromagnetic in the ferrimagnetic interlayer alloy or at the interfaces, this can occur if Co and Ni dominate in the magnetization of the ferrimagnetic interlayer alloy or if the magnetization of both CoNi layers is larger than the Gd layer magnetization.

EB was observed from t=8 nm up to t=20 nm Gd spacer thickness in trilayers, when samples were cooled from room temperature to 90 K under a 2 kOe external magnetic field. However, as-grown samples did not show EB. Once the hysteresis loop of samples were taken at RT, it was not necessary to cool them under an external magnetic field to observe EB. This suggests that the as-grown demagnetized state or multi-domain state removes EB. The degree of magnetic order, which may be represented by the remanence magnetization, is the key factor in setting the EB. The magnetic field is important for setting the magnetic order and for selecting the order direction. It may be concluded that conditions that alter the remanence state, such as grain orientations and sizes, can also effect the EB. Modifying EB field by controlling the remanence state with an external magnetic field has been reported in a ferromagnetic/antiferromagnetic system [28].

Fig. 2 shows the hysteresis loops of CoNi[8 nm]/Gd[10 nm]/CoNi[8 nm] trilayer at the given temperatures. The hysteresis loops were measured after cooling the sample to 90 K under a 2 kOe external magnetic field. Then, the sample was heated to the indicated temperature, and the hysteresis loop was measured. The trilayer couples strongly around 140 K. As temperature is decreased, the coupling gets stronger and the hysteresis loop start to shift. At 90 K, the center of the hysteresis loop shifts approximately 400 Oe opposite to the external cooling field direction (negative EB). Once again, this indicates as if the hard layer and the soft layer couple ferromagnetically. As the sample was cooled, a slope occurs in the magnetization curve starting approximately at around 1000–1200 Oe. It is to be seen that this slope is mainly due to the relative angular orientation of the hard layer (CoNi(top)/Gd bilayer) and the soft layer (bottom CoNi layer) with respect to the magnetic field. The measured magnetization of this trilayer at 1 kOe, the magnetization of the soft layer and the hard layer calculated from the hysteresis loop are shown in Fig. 3 as a function of temperature. The calculation was done up to the temperature where the double loop in the hysteresis curve almost disappeared due to strong coupling. Therefore, there is some error in the determination of each layer magnetization due to the temperature-dependent interlayer coupling. Total magnetization is decreasing as a function of the decreasing temperature. The calculated soft CoNi(bottom) layer magnetization changes slightly. The calculated hard layer magnetization decreased with decreasing temperature (Fig. 3). On the contrary, the experimentally measured magnetization of both hard layer (Fig. 4) and soft layer did not change with temperature. It is expected to see a change in the hard layer magnetization due to the temperature-dependent magnetization of ferrimagnetic alloy formed at the interface. If there is such a change, it must be within our measurement limits. Therefore, the decrease in the total magnetization of the trilayer and the sloping in the hysteresis loop are mainly due to the angular orientation of the hard layer with respect to the magnetic field which is caused by antiferromagnetic coupling at the interface. Both the soft layer and the hard layer have almost the same magnetization. Therefore, 90° coupling at around 150 K is expected, which coincides with the onset of the EB. Below 150 K, both the soft layer and the hard layer should deviate from the external field direction in order to have lower magnetization. This may be due to the increased antiferromagnetic coupling as temperature decreases. The 90° coupling was theoretically predicted [9] and experimentally observed [29].

The coercivity of the hard layer increased as temperature decreased and its hysteresis curve did not shift (Fig. 4). Since the hysteresis loop did not shear, the coupling field between the CoNi(top)/Gd must be stronger than the coercive field. This strong coupling due to diffusion may be the reason for not observing EB in this hard layer and in the CoNi[8 nm]/Gd[12 nm] bilayer.

It is believed that the EB is because of the hard ferrimagnetic alloy developed due to the diffusion at the top CoNi/Gd interface. This was supported, in addition to the XPS depth profiling, by not observing EB in samples where 2 nm Ag layer was deposited between the CoNi(top) and the Gd layer, and in the Gd(top)/CoNi bilayer sample where strong interface mixing is not expected (Fig. 5). XPS depth profile of the CoNi [8 nm]/Gd/[10 nm]/CoNi[8 nm] trilayer showed strong Co and Ni diffusion into the Gd layer, negligible oxygen (below 1 at%) and the metallic state of Gd, Co, and Ni. The depth profile results can be repeated with better depth resolution using an XPS system that can provide Zalar rotation and low energy ion bombardment. However, the current depth profiling results are significant in showing intermixing, and eliminating the role of oxygen in the EB. On average the atomic composition of the Gd layer will be CoNi_0.90Gd_0.10, which requires just a 2 nm thick CoNi layer diffusion through 10 nm Gd due to the large atomic mass difference.

As the Gd interlayer thickness was increased, the spin flop state shows up more clearly because it is energetically favorable to introduce a domain wall in the thicker Gd layer [29]. The trilayer hysteresis loops for 12 and 14 nm (Fig. 6) Gd interlayer thicknesses are similar. They show similar negative EB properties. The CoNi[8 nm]/Gd[14 nm] bilayer did show positive EB and spin flop (Fig. 7) in contrast to the negative EB in the CoNi[8 nm]/Gd[14 nm]/CoNi[8 nm] trilayer shown in Fig. 6.

In the CoNi[8 nm]/Gd[14 nm] bilayer (Fig. 7), the magnetization of the Gd layer and the coupling between both layers start to increase around 200 K as it is cooled. At around 120 K, the magnetization of the Gd layer dominates over the CoNi layer magnetization, and aligns in the external field direction to minimize the Zeman energy. As a result, the CoNi layer aligns opposite to the low external magnetic field due to the antiferromagnetic coupling between Gd and CoNi. At high external field both layers align towards the external field. However, when the field is reduced the CoNi layer reverses earlier than expected due to the strong antiferromagnetic coupling (or in other words strong coupling among Co atoms). Hence, the hyteresis loop shifts in the cooling field direction (Fig. 7). On the contrary, in the corresponding trilayer (Fig. 6), the total magnetization of the CoNi layers dominates over the Gd layer magnetization. Therefore, they align in the external field direction and the Gd layer is compressed between them by the external field (the Gd layer is exchange coupled at two interfaces). When the field is reduced, the Gd layer relaxes and CoNi layers shift opposite to the cooling field direction. Both the CoNi[8 nm]/Gd[20 nm] bilayer and the CoNi[8 nm]/Gd[20 nm]/CoNi[8 nm] trilayer (Fig. 8) show positive EB because in both cases the Gd layer magnetization dominates over the total CoNi layer magnetization. Of course, this picture does not explain the formation of the unidirectional anisotropy. It only states that the coupling direction of layers to the hard layer at the interface or the coupling between sublattices in the hard layer are determined by counter play of Zeeman energy and interface coupling energy.

Fig. 9, Fig. 10 show the EB field and coercivity, respectively, in several trilayers as a function of temperature. EB field changes strongly and inversely with temperature. These figures also show the effect of Gd and CoNi layer thickness on the coercivity and the EB field. The coercivity (Fig. 10) showed bumps below 150 K where strong coupling between layers start. Similar anomalies have been observed in FM/AFM systems, which is related to losses in the antiferromagnetic layer, around blocking temperatures or near Néel temperature for particular ratio of the FM and AFM layer thickness, and at cross over from negative EB to positive EB due to interfacial magnetic frustrations [30]. Inhomogeneities and/or temperature-dependent anisotropy development in the interface alloy can cause this peak. Clearly, all these factors depend on the diffusion process. In addition to these properties, training effect was observed [31]. More importantly, the EB was not observed if hysteresis loops were taken in larger field intervals [32], for example +4 kOe to −4 kOe in bilayers, over 4 kOe in trilayers. However, EB recovers if the hysteresis loop is measured again in a lower field interval. This may be the sign of the existence of an interface domain wall for EB at lower field interval and its annihilation in the larger field interval because of not being supported by the interface alloy. Breaking Co–Gd, Ni–Gd coupling by large field is another possibility. However, it is unlikely because the EB field did not show any dependence on the external cooling field strength up to 8 kOe. This magnitude of the external field may not be enough to align spins of sublattices towards the external field direction during cooling in order to see field-dependent EB [6], [33]. The dependence of EB on measurement field strength shows the existence of irreversible processes and supports the interpretation of EB as a minor loop effect [33].

In summary, several CoNi/Gd bilayers and CoNi/Gd/CoNi trilayers were investigated experimentally between RT and 90 K. Top CoNi layer thickness was kept at 8 nm. The Gd layer thickness was changed from 5 to 20 nm. Positive EB was observed in the bilayers of 14 and 20 nm Gd layer thickness due to the domination of the Gd layer magnetization over the CoNi layer magnetization. This causes the CoNi layer magnetization to align opposite to the external field hence positive EB. Other bilayers did not show EB probably due to strong coupling or due to their thicknesses. Only their coercivity increased as a function of cooling temperature. EB was observed in those trilayers in which the Gd interlayer thickness was equal to or thicker than 8 nm. Negative EB became positive EB for 20 nm Gd interlayer thickness in the trilayer because of the Gd layer magnetization alignment in the external field direction. Bottom CoNi layer thickness was changed from 8 to 20 nm in the trilayer with fixed 10 nm Gd interlayer thickness. EB field decreased as the bottom CoNi layer thickness was increased. EB field in all cases decreased linearly as temperature was increased. Cooling field strength did not change the EB, but it became zero if hysteresis loops were taken in larger magnetic field intervals. This may indicate the existence of interface domain wall. This study suggests us that EB investigation in structurally well-known ferromagnetic/ferrimagnetic bilayers may bring light on the role of magnetic anisotropy, sublattice coupling, and sublattice magnetization on EB.