Tuning the magnetodynamic properties of all-perpendicular spin valves using He⁺ irradiation

Pseudo-spin-valves (PSVs) are multilayer stacks consisting of two ferromagnetic (FM) layers separated by a nonmagnetic (NM) metallic spacer. They have received a great deal of attention due to their applications in spintronics, which include magnetic read heads, sensors, and magnetoresistive random access memory (MRAM).^1–3 More recently, spin-transfer torque (STT) MRAM^4,5 and spin-torque nano-oscillators (STNOs)^6,7 have been investigated; these utilize the STT effect, which describes how a current of spin-polarized charges can exert a torque on the magnetization of a ferromagnetic layer, driving that layer into precession or switching.^8–10 

In early studies, STNOs were generally fabricated from spin valves with easy-plane magnetic anisotropy materials, such as NiFe and Co; interesting dynamics of propagating spin wave and localized bullet solitons were reported.^11,12 A novel nano-scopic and magnetodynamical object—the magnetic droplet soliton—was then unveiled in orthogonal STNOs with perpendicular magnetic anisotropy (PMA) free layer and easy-plane fixed layer.^13–24 Following this trend, stabilized droplets were found in all-PMA spin valve STNOs,²⁵ thus enabling further insights into droplet dynamics, such as its size and nucleation boundary. Obviously, tailored PMA not only plays a crucial role in stabilizing magnetic droplets, it is increasingly important for future generations of STT-MRAM.^3,26 The engineering of PMA thus attracts much interest. For example, irradiation with ions such as Ar⁺, He⁺, or Ga⁺ has proven to be capable of modifying the magnetic properties of multilayers, the PMA of which is sensitive to surface or interface structures.^27–31 

In this paper, we tailor the magnetodynamic properties in an all-PMA [Co/Ni]/Cu/[Co/Pd] PSV using He⁺ irradiation. We achieve a range of magnetic configurations, from all-perpendicular (↑↑), through orthogonal (→↑), to all in-plane (⇉) by precisely controlling the He⁺ fluences. Hysteresis loop (HL), magnetoresistance (MR), and broadband ferromagnetic resonance (FMR) measurements are performed. We find that the coercivity H_c, GMR, and effective magnetization M_eff of both the Co/Ni and the Co/Pd layers can be controlled consistently. At the same time, the damping and inhomogeneous broadening of the Co/Ni layer show rather dramatic improvements, while those of Co/Pd seem unaffected. These observations are likely to prove essential to spintronic applications.

Thin-film deposition was performed using a commercial AJA ATC Orion 8 sputtering system. A full stack of seed layer Ta (5)/Cu (15)/Ta (5)/Pd (3), all-perpendicular spin valve [Co (0.5)/Pd (1.0)]×5/Co (0.5)/Cu (6)/[Co (0.3)/Ni (0.9)]×4/Co (0.3), and capping layer Cu (3)/Pd (3) was deposited on thermally oxidized Si substrate (numbers in parentheses are thicknesses in nanometers). The sample was then cleaved into nine pieces of about 7×7 mm for He⁺ irradiation. Using He⁺ ions at energies around 15 keV, recoils are limited to 1-2 atomic distances and all the ions stop deep into the substrate. As a result, He⁺ irradiation is a soft process that allows a precise control of interface intermixing maintaining both the crystalline texture and grain size of the pristine films. The fluence (F) of He⁺ ions was varied from 2 to 50×10¹⁴ He⁺/cm². Hysteresis loop measurements, where the applied field is either along (in-plane, IP) or normal (out-of-plane, OOP) to the film plane, were conducted using an alternating gradient magnetometer (AGM). To measure GMR, a commercial Picoprobe with G- and S-pads was applied to the extended films. The dc current was then injected into the film with a Keithley 6221 current source, and dc voltage was picked up by a Keithley 2182 nanovoltmeter. The broadband ferromagnetic resonance measurement was carried out using a NanOsc Instruments PhaseFMR-40 with a coplanar waveguide (CPW). The FMR frequency f was between 3 and 40 GHz. An OOP magnetic field μ₀H was swept from 0 to 1.3 T at a given frequency. All of the measurements were performed at room temperature.

Figure 1(a) shows the OOP hysteresis loops for different fluences. The nonirradiated film (F = 0) shows clear two-step switching; the higher (μ₀H_c = 93 mT) and lower (μ₀H_c = 67 mT) coercivities correspond to Co/Pd and Co/Ni layers, respectively, indicating strong PMA. As He⁺ fluences increase, H_c undergoes continuous reduction and the squareness values remain approximate up to F = 10×10¹⁴ He⁺/cm² for both Co/Ni and Co/Pd, suggesting that their PMA values were modified by He⁺ irradiation. Further quantitative study on PMA will be carried out by FMR later. As for higher fluences (F = 30×10¹⁴ He⁺/cm²), Figure 1(b) shows that both OOP and IP HLs show a continuous change, followed by a sharp drop in magnetization, indicating that one FM layer remains OOP magnetized while the other layer favors easy-plane—that is, an orthogonal spin-valve. Figure 1(c) shows that the saturation field for IP is much lower than that of OOP, which means that the remanent magnetic states of both Co/Ni and Co/Pd turn to in-plane magnetized for the highest fluence (F = 50×10¹⁴ He⁺/cm²).

In the following, we carried out MR measurements on all samples. The MR data are plotted in Fig. 2(a). It is clear that the switching fields H_s show similar decreasing trends as the HL data in Fig. 1. We compared the H_c and H_s from both the HL and MR measurements, shown in Fig. 2(b). As expected, H_c and H_s are in good agreement with each other. It should be noted that, for F = 30 and 50×10¹⁴ He⁺/cm², the overlapping signals of both Co/Ni and Co/Pd prevented the separation of H_c and H_s; they were thus not extracted. The MR curve in Fig. 2(a) for 30 × 10¹⁴ He⁺/cm² shows a clear MR jump at μ₀H = 40 mT, which is consistent with the H_c of Co/Pd at the same fluence as the red arrow in Fig. 1(b). Both HL and MR curves are saturated at μ₀H = 140 mT, which results from the saturation of the IP magnetized Co/Ni layer. In Fig. 2(c), we calculated the GMR values defined by GMR = [R(AP)-R(P)]/R(P).^32,33 These show a linear decreasing trend from 1.14% to 0.4% with fluence values, which can be understood as the He⁺ irradiation intermixing the interfaces of [Co/Ni]/Cu and Cu/[Co/Pd].^34,35 The GMR maintains a value of 0.4% at the highest fluence.

We now turned our attention to the magnetodynamical properties of He⁺-irradiated PSVs by conducting OOP FMR measurements. Figure 3(a) shows the typical signal for different fluences at a frequency f = 25 GHz. All spectra are fitted with a sum of symmetrical and antisymmetrical Lorentzian derivates,^36,37 as shown by the solid lines in Fig. 3(a). For fluences from 0 to 10 × 10¹⁴ He⁺/cm², only one resonance peak appears. One more peak then appears for F = 20 and 30 × 10¹⁴ He⁺/cm²; again, one peak occurs for F = 50 × 10¹⁴ He⁺/cm². By fitting all the data at different frequencies, we plotted the resonance field H_res as a function of f in Fig. 3(b, c). We already know that the PMA of our Co/Ni is weaker than Co/Pd;^38,39 we thus identify the single peak for the lower fluences with Co/Ni, and conclude that the Co/Pd resonance peaks are beyond the measured frequency range because of its stronger PMA. However, as the fluences increase, the PMA of Co/Pd decreases, and the Co/Pd peaks can be observed for 20 and 30 × 10¹⁴ He⁺/cm². For 50 × 10¹⁴ He⁺/cm², the single peak could correspond to Co/Ni (see below) and Co/Pd might be too weak to detect. We then fit the H_res with the OOP Kittel equation,⁴⁰ 

where μ₀ is the permeability of free space and M_eff is the effective magnetization. γ/2π is the gyromagnetic ratio, obtained by fitting with Eq. 1; it shows a weak dependence on fluence and stays within 30.5±0.4 and 31.6±0.2 GHz/T for Co/Ni and Co/Pd, respectively. For F = 50 × 10¹⁴ He⁺/cm², we attribute the single peak to Co/Ni. This is based on i) γ/2π is 30.1 GHz/T, i.e. much closer to that of Co/Ni, and ii) in Fig. 2(a), the OOP MR for 50 × 10¹⁴ He⁺/cm² shows a maximum at μ₀H = 0.09 T suggesting that Co/Pd and Co/Ni have different M_eff, i.e. their FMR responses should not overlap.

The extracted μ₀M_eff values are plotted in Fig. 3(d). As the fluence increases from zero to its highest value, μ₀M_eff of Co/Ni increases from -0.60 to 0.31 T. The negative value of μ₀M_eff at low fluences implies that the PMA is sufficient to overcome the demagnetizing energy, and hence the easy axis is normal to the film plane; so is that of Co/Pd (All-PMA). Interestingly, we observed that the M_eff of Co/Ni changes sign to positive at 30 × 10¹⁴ He⁺/cm², which indicates that Co/Ni is IP, while that of Co/Pd remains negative (OOP), in agreement with the results of HL in Fig. 1(b). At sufficiently high fluences (50 × 10¹⁴ He⁺/cm²), the PMA is reduced to the point where the demagnetization field dominates and the easy axes of both Co/Ni and Co/Pd lie in the film plane, as confirmed by the positive μ₀M_eff value of Co/Ni in Fig. 3(c) and the only clear switching in Fig. 1(c). The saturation magnetization M_s is calculated from AGM data and exhibits no clear dependence on fluence for either Co/Ni or Co/Pd, with μ₀M_s = 1.0 ±0.1 T and 1.2 ±0.1 T, respectively. We thus claim that the anisotropy field H_k shows a decreasing trend with fluence, since M_eff = M_s - H_k. We have hence demonstrated that irradiation with He⁺ allows us to tune the magnetic structures from all-PMA, through orthogonal, to all-easy-plane.

In Figs. 4(a, b), we extracted the full width at half maximum (FWHM) linewidth ΔH by fitting the FMR spectra for Co/Ni and Co/Pd. The FMR linewidth contribution normally originates from the intrinsic Gilbert linewidth ΔH_G,^41,42 inhomogeneous broadening ΔH₀,⁴³ and two-magnon scattering ΔH_TMS.^44,45 The Gilbert contribution, produced by the intrinsic spin-orbit coupling of the ferromagnetic materials, is proportional to the frequency f: ΔH_G ∝ f. Inhomogeneous broadening, independent of f, relies on sample inhomogeneity, which is probably associated with the local variation of M_eff, specifically the existence of interface roughness (see below). The two-magnon scattering, valid for defects as scattering centers in ferromagnets, is a process in which the k = 0 magnon excited by FMR is scattered into degenerate magnon states with wave vectors k ≠ 0. Since there is no contribution of ΔH_TMS when the applied field is normal to the films,^29,45 we ignore the contribution of two-magnon scattering in our OOP FMR measurements. The linewidth can then be described as

where α is the damping constant. By fitting the linewidth with Eq. 2 in Figs. 4(a, b), the extracted ΔH₀ and α are shown in Fig. 4(c). The damping of the Co/Ni layer is found to first improve substantially at low fluence (4 × 10¹⁴ He⁺/cm²), and then remain almost constant (α = 0.025 − 0.027) at higher fluences, where it is still lower than the non-irradiated value. The improvement in the damping of Co/Ni (by a factor of two) may result from the intermixing of Co and Ni layers, ultimately becoming an alloy. The literature⁴⁶ has reported that the damping of Co (α_Co = 0.005) and Ni (α_Ni = 0.028) single layers is much lower than, and respectively comparable to, our Co/Ni multilayers. The damping of the alloy of Co_1−xNi_x has been investigated in Ref. 46, where the damping shows a monotonic decrease with a decreasing in Ni concentration. For our He⁺-irradiated Co/Ni multilayers, the intermixing of the Co and Ni layers by He⁺ collision is analogous to the case of the alloying of Co and Ni. This could be one of the explanations for the reduction of damping with increasing fluence. The inhomogeneous broadening ΔH₀ of the Co/Ni layer shows a dramatic drop after irradiation, and similar behavior has been reported for irradiated Co/Ni.^29,31 The soft ion-induced intermixing may average out the role of defects at interfaces (interface roughness) by inducing a continuous Co/Ni alloy.⁴⁷ As a result, the distribution of magnetic anisotropy is reduced leading to a decrease of the inhomogeneous broadening. on the other hand, The inhomogeneous linewidth is also associated with the grain sizes and proportional to the anisotropy field, as proposed in Ref. 44, which is consistent with our experimental observations. As we indeed observed, the anisotropy field H_k and the inhomogeneous linewidth decrease as fluence increases. Regarding Co/Pd, the damping seems unaffected (α = 0.04), and the inhomogeneous broadening is slightly larger for the two measurable fluences in Figs. 4(b, c). To better understand the effect on Co/Pd, more detailed studies are needed, which is beyond the scope of this paper. The damping and inhomogeneous broadening of Co/Ni were improved simultaneously, which is critical for the free layer of STNOs, and suggests lower threshold currents and more uniform films. These parameters, however, are not as important for the fixed Co/Pd layer as for the free layer.

In conclusion, we investigated the controlled magnetic properties of [Co/Pd]/Cu/[Co/Ni] PSVs by He⁺ irradiation. By performing hysteresis loop and magnetoresistance measurements, the coercivities of Co/Ni and Co/Pd showed a continuous reduction with increasing fluence, just like the GMR. FMR results showed that the perpendicular anisotropy field H_k is progressively decreased by He⁺ irradiation. This could result from the soft ion-induced interface mixing and strain relaxation. By precisely controlling the fluence, the remanent magnetic states can be adjusted from OOP to IP for both Co/Ni and Co/Pd, which allow us to achieve magnetic structures ranging from all-perpendicular (↑↑), through orthogonal (→ ↑), to all in-plane (⇉). In addition, the damping and inhomogeneous broadening are improved simultaneously for Co/Ni, which benefits STNOs and STT-MRAM applications.

We gratefully acknowledge financial support from the China Scholarship Council (CSC), the Swedish Foundation for Strategic Research (SSF), the Swedish Research Council (VR), and the Knut and Alice Wallenberg Foundation (KAW). This work was also supported by the European Research Council (ERC) under the European Community’s Seventh Framework Programme (FP/20072013)/ERC Grant 307144 “MUSTANG.”, the French ANR project ELECSPIN number ANR-16-CE24-0018-02 and the Labex Nanosaclay.