Since the prediction by Slonczewski and Berger that a spin-polarized current can exert enough torque on a magnetic layer to significantly affect its magnetization [Slonczewski 1996, Berger 1996], the spin-transfer torque (STT) effect has been intensely investigated for potential applications in spintronic devices, such as STT-magnetoresistance random access memory (STT-MRAM) [Zhu 2008], spin torque oscillators (STO) [Katine 2008], and domain-wall memory [Parkin 2008]. While such devices are typically divided into in-plane and perpendicularly magnetized free or fixed layer geometries, we recently proposed a so-called tilted polarizer STO, where the fixed layer magnetization is tilted out of the film plane in order to simultaneously achieve zero-field operation and high output power [Zhou 2008, Zhou 2009a]. The tilt angle introduces an additional degree of freedom, which leads to a surprisingly rich phase diagram of spin torque switching and precession [Zhou 2009b].
A tilted spin polarizer, with both in-plane and out-of-plane spin polarization components, can be experimentally achieved using materials with strong tilted magnetocrystalline anisotropy. We have previously reported on using (111)-oriented L10 FePt and FePtCu with tilted magnetocrystalline anisotropy to fabricate pseudo spin valves (PSVs) for spin torque devices [Zha 2009a, 2009b, 2009c]. However, L10 FePt and FePtCu have a number of drawbacks such as a relatively low spin polarization [Seki 2008], undesirably high damping factor [Seki 2006], and prohibitive cost due to their high Pt content. Very recently, D022-ordered Mn3−xGa (x = 0−1) was theoretically predicted to be a nearly half-metallic ferrimagnet with 88% spin polarization at the Fermi surface, and was consequently proposed to have great potential for STT devices [Balke 2007, Winterlik 2008, Wu 2009]. The large magnetocrystalline anisotropy (Keff = 1.2 × 107 erg/cm3), the low magnetization (Ms ≤ 250 emu/cm2), and the expected high degree of spin polarization make Mn3−xGa ideal as a tilted polarizer [Winterlik 2008, Wu 2009], provided the appropriate crystalline orientation of the D022 phase can be realized.
In this letter, we report the successful fabrication of (112)-textured D022 Mn2.3−2.4Ga (MnGa hereafter) thin films with a tilted magnetization and PSVs based on these films. This is a crucial first step in the eventual realization of the aforementioned STT devices. In single D022 MnGa films, we observe a small magnetoresistance (MR) with a parabolic field dependence consistent with ordinary MR typically observed in all metals. In MnGa/Cu/CoFe PSVs, a small negative giant magnetoresistance (GMR) is observed between the MnGa and CoFe layers. In order to obtain a sizable positive GMR effect, an ultrathin CoFe layer is inserted at the MnGa/Cu interface. To the best of our knowledge, we are the first to report on the fabrication and subsequent demonstration of MR in D022 MnGa-based PSVs.
All film stacks were deposited at room temperature on thermally oxidized Si substrates using a magnetron sputtering system (AJA ATC Orion-8) with a base pressure better than 5 × 10−8 torr. Deposition of a 6 nm Ta underlayer was followed by MnGa deposition from a Mn60Ga40 alloy target. This bilayer was subsequently annealed in situ at 400 °C for 35 min to form the D022 (112)-textured MnGa phase. For the PSVs, a 5 nm Cu spacer and a 5 nm Co50Fe50 (CoFe) layer were deposited after cooling down to room temperature. Finally, a 3 nm Ta capping layer was deposited on both single MnGa films and PSVs for oxidation protection. Three different MnGa thicknesses of 15, 25, and 50 nm, respectively, were employed to fabricate PVSs. The final Mn70Ga30 film composition was determined using energy dispersive X-ray spectrometry (EDX). Note that, the achieved composition is located in the 66–74 at.% Mn range, where the D022 phase is expected to appear [Niida 1996].
Magnetic properties were characterized using a physical property measurement system (PPMS) equipped with a vibrating sample magnetometer (VSM), and an alternating-gradient magnetometer (AGM) with a maximum field of 14 kOe. In addition to standard major hysteresis loop analysis, we employed a first-order reversal curve (FORC) technique [Davies 2004, Dumas 2007]. First, a family of FORC curves was measured. Each curve started at a successively more negative reversal field HR, and was then measured with an increasing applied field H parallel to the film plane. Then, a mixed second-order derivative of the magnetization M(H, HR) was used to generate a FORC distribution ρ ≡ − ∂2 M(H, HR)/2 ∂ H ∂ HR, which was plotted against (H, HR) coordinates on a contour map. Crystallographic structures were investigated by X-ray diffraction (XRD) with Cu Kα radiation in a symmetric scan geometry. Current-in-plane (CIP) electron transport properties were determined by a standard four-point tester with the current orthogonal to the magnetic field (transverse configuration).
Fig. 1. (a) XRD patterns of 15, 25, and 50 nm single MnGa films, respectively; (b) in-plane and out-of-plane VSM hysteresis loops and (c) current-in-plane magnetoresistance curve of a single 15 nm MnGa film; (d) in-plane and out-of-plane VSM hysteresis loops of a single 50 nm MnGa film.
View All | Next
RESULTS AND DISCUSSION
Fig. 1 shows structural and magnetic properties of single MnGa films. Following [Winterlik 2008], we identify the peaks at 41.36° in Fig. 1(a) as the (112) diffraction peak of the D022 phase of Mn3−xGa. Importantly, with increasing film thickness, we observe enhanced (112) texture as the relative diffraction intensity becomes stronger. Hysteresis loops of the 15 nm MnGa film in Fig. 1(b) exhibit in-plane coercivity (Hc∂) of 6.8 kOe and out-of-plane one (Hc⊥) of 1.5 kOe, as well as a 120 emu/cm3 saturation magnetization (MS). A 50 nm MnGa [see Fig. 1(d)] reveals Hc∂ = 11.1 kOe, Hc⊥ = 9.3 kOe, and MS = 160 emu/cm3. These values are consistent with the ferrimagnetic structure of D022 MnGa film previously reported [Wu 2009]. The squareness ratios of the out-of-plane and the in-plane loops are less than 1, indicating that the easy magnetization axis lies in neither the film plane nor along the normal direction, as expected for highly textured (112) D022 MnGa films. Also, it likely reflects the domain formation in the D022 MnGa layer resulting from the varying local tilted anisotropy directions of different D022 (112) MnGa grains. The improved magnetic properties of the 50 nm MnGa film are consistent with its enhanced structural properties [see Fig. 1(a)]. This indicates that the chemical ordering of the D022 phase of MnGa increases (inducing a higher magnetocrystalline anisotropy) with thickness under the same annealing condition. The CIP-MR is found to be 0.12% at ±14 kOe for the 15 nm single MnGa film [see Fig. 1(c)], which we ascribe to ordinary magnetoresistance of a metal. A relatively large resistance of about 70 Ω implies a moderate crystallinity, consistent with the XRD results [see Fig. 1(a)].
Fig. 2. (left) Current-in-plane magnetoresistance curves and (right) hysteresis loops of Ta (6 nm)/MnGa (15 nm)/CoFe (y nm)/Cu (5 nm)/CoFe (5 nm)/Ta (3 nm), (a and b) y = 0; (c and d) y = 0.5; (e and f) y = 1.0; and (g and h) y = 1.5. The arrows indicate the onset of the MnGa reversal.
Previous | View All | Next
The magnetotransport and magnetic properties of PSVs of MnGa (15 nm)/CoFe(0, 0.5, 1.0, 1.5 nm)/Cu (5 nm)/CoFe(5 nm) are shown in Fig. 2. The unique shape of the MR curve in Fig. 2(a), for MnGa/Cu/CoFe, is due to a combination of effects. The most prominent contribution to the MR is a negative GMR between the CoFe and MnGa layers. Near positive saturation the two layers are parallel and a relatively high-resistance state is observed. As the applied field is reduced (dark-filled squares) a sharp drop in resistance near zero field is observed as the CoFe (5 nm) free layer switches and becomes antiparallel to the MnGa layer. With a further decrease in applied field, the resistance remains nearly constant until roughly –5 kOe where the MnGa layer begins to switch. Finally, a high-resistance state, corresponding to parallel alignment of the CoFe and MnGa, is once again achieved at negative saturation. This situation is similar to Fe/Cu/GdCo spin valves [Yang 2006] and FeCoGd/AlO/FeCo tunnel junctions [Bai 2008], where a negative GMR is also observed. In addition to the dominant negative GMR contribution, a small peak in the resistance near zero field is also observed, and is most likely due to either anisotropic MR in the transverse measurement geometry or a small positive GMR component. Finally, a small parabolic background, most obvious at high fields, is observed due to the ordinary MR of the MnGa layer, similar to Fig. 1(c).
CoFe usually gives positive bulk and interface spin asymmetry coefficients with Cu [Li 2002]. Considering, ab initio calculations [Winterlik 2008] of the electronic structure for D022 Mn3Ga, it is found that the minority (majority) density of states exhibits a maximum (minimum) at the Fermi energy and the bulk spin asymmetry coefficient is positive [Tsymbal 1996]. However, based on the negative GMR found in this PSV [see Fig. 2(a)], we could not rule out the possibility of a negative interface spin asymmetry coefficient at the MnGa/Cu interface. To obtain a positive interface spin asymmetry coefficient, we employ a thin CoFe insertion layer at the MnGa/Cu interface. We, therefore, expect to not only obtain conventional positive GMR, as anticipated for a PSV, but also enhancement of the MR [Vouille 1999].
As shown in Fig. 2(c), when an ultrathin 0.5 nm CoFe layer is inserted between the spacer and fixed layers, we find a positive MR of 0.07%. This suggests that GMR from spin-dependent interface scattering at the CoFe/Cu/CoFe interfaces is greater than net negative MR from MnGa alone. The MR is found to increase further as the thickness of the CoFe insertion layer is increased as shown in Fig. 2(e) and (g). The shape of the MR loops indicates a two-step switching for spin valves, which progressively weakens as the CoFe insertion layer becomes thicker, virtually disappearing for 1.5 nm layers. On the other hand, the easy magnetization axis is probably gradually pushed into the film plane by the thicker CoFe.
Fig. 3. (a and c) CIP-MR curves and (b and d) in-plane AGM hysteresis loops of Ta (6 nm)/MnGa (25 nm)/CoFe (1.5 nm)/Cu (5 nm)/CoFe (5 nm)/Ta (3 nm) and Ta (6 nm)/MnGa (50 nm)/CoFe (1.5 nm)/Cu (5 nm)/CoFe (5 nm)/Ta (3 nm), respectively.
Previous | View All | Next
The hysteresis loops in Fig. 2(b,d,f,h) show a two-stage switching which leads to the observed GMR effect. The large vertical separation between the upper and lower plateaus in the hysteresis loops is due to the much higher moment (four times as high) of the CoFe free layer than the fixed layer. With increasing thickness of the CoFe insertion layer, the switching plateau becomes gradually smaller and the coercivity of the MnGa fixed layer also gradually decreases, indicating the switching of the soft CoFe insertion layer assists the hard MnGa switching due to exchange interactions.
To further improve the CIP-MR we explore the effect of the MnGa thickness on magnetotransport. As shown in Fig. 3(a) and (c), MR increases to 3.88% when tMnGa = 25 nm. However, when increasing that thickness of MnGa to 50 nm, we find the MR decreases to 2.95%. This decrease is most likely due to current shunting through the relatively thick MnGa layer [Zha 2009d]. Fig. 3(b) and (d) exhibit the in-plane magnetic properties for these two-spin valves with tMnGa = 25 and 50 nm. The separate switching between the free CoFe layer and the fixed-MnGa/CoFe bilayer corresponds to GMR.
To better understand the interaction between the CoFe insertion layer and the MnGa fixed layer, the FORC technique is employed. The family of FORC curves is shown in Fig. 4(a), where black dots represent the starting point for each FORC. As the reversal field is decreased, a sharp drop in magnetization is found near HR = −150 Oe, which corresponds to the CoFe free layer switching. A family of FORC curves highlighting the CoFe free layer reversal at small fields is shown in Fig. 4(a) (inset). The CoFe free layer switching is a highly irreversible process which manifests itself as a very narrow peak with a large intensity centered at (H, HR) ∼ (−50 and 150 Oe), which then dominates the FORC distribution. In order to highlight the switching of the MnGa (50 nm)/CoFe (1.5 nm) bilayer, the FORC distribution is plotted only for HR < −2 kOe in Fig. 4(b). Interestingly, the FORC diagram is nearly featureless for −6.5 kOe < HR < −2 kOe indicating reversible switching processes. This region is also highlighted with a bracket in Fig. 4(a), and is associated with the highly reversible switching of the CoFe insertion layer. The onset of irreversible switching occurs for reversal fields HR < −6.5 kOe, and is indicated with a horizontal dashed line in Fig. 4(b), and the red-dashed FORC in Fig. 4(a), respectively. For HR < −6.5 kOe, we begin to see peaks in the FORC distribution that corresponds to irreversible switching of the MnGa layer as negative saturation is approached. We can interpret the reversal of the MnGa (50 nm)/CoFe (1.5 nm) bilayer as being that of a classic bilayer exchange-spring magnet [Davies 2005, Fullerton 1998, Nagahama 1998], where the MnGa and CoFe can be identified as the hard and soft components, respectively. This exchange spring interaction explains the lack of a clear two-step behavior in the MR data in samples with thick CoFe insertion layers. Essentially, after the CoFe free layer switches, the CoFe insertion layer begins to reversibly switch leading to a gradual decrease in the MR as saturation is approached.
Fig. 4. (Color online). (a) Complete family of FORC curves of the MnGa (50 nm)/CoFe (1.5 nm)/Cu (5 nm)/CoFe (5 nm) spin valve. The inset highlights the reversal of the CoFe free layer at small fields. (b) Corresponding FORC diagram highlighting the reversal of the MnGa(50 nm)/CoFe(1.5 nm) bilayer. The red dashed lines in (a) and (b) indicate the onset of irreversible switching for the MnGa/CoFe bilayer.
Previous | View All
In summary, PSVs using (112)-textured D022 MnGa as a fixed layer have been demonstrated and a MR up to 3.88% has been achieved. A negative to positive transition in the MR is realized by insertion of a thin CoFe layer at the MnGa/Cu interface. Reversal of the MnGa/CoFe bilayer has been analyzed and shows exchange-spring like behavior which explains the lack of a two-step reversal typically observed in the MR response of spin valves. Some degree of material optimization is still needed. Unfortunately, the relatively large negative CIP-MR of the D022 MnGa layer always cancels some of the positive CIP-GMR in our PSVs. However, the current-perpendicular-to-plane (CPP) configuration might eliminate the negative contribution to the MR, and hence hopefully achieve even higher MR values.
The authors are grateful to S. Lidin for giving them access to the PPMS. C. L. Zha would like to thank Dr. N. A. N. Thi for fruitful discussions. Support from The Swedish Foundation for strategic Research (SSF), The Swedish Research Council (VR), the Göran Gustafsson Foundation, and the Knut and Alice Wallenberg Foundation is gratefully acknowledged. J. Nogués would like to thank the Wenner Gren Center Foundation, the Catalan DGR (2009SGR1292), and the Spanish MICINN (MAT2007-66309-C02) projects for partial financial support. Johan Åkerman is a Royal Swedish Academy of Sciences Research Fellow supported by a grant from the Knut and Alice Wallenberg Foundation.