Exchange coupling of a perpendicular ferromagnet to a half-metallic compensated ferrimagnet via a thin hafnium interlayer

TL;DR

This study demonstrates that a thin hafnium layer can serve as both a diffusion barrier and an exchange coupling medium between a perpendicular ferromagnet and a compensated ferrimagnet, with the coupling sign changing at the ferrimagnet's compensation temperature.

Contribution

It introduces hafnium as an effective interlayer for exchange coupling in magnetic heterostructures involving zero-moment half metals, overcoming limitations of traditional materials like ruthenium.

Findings

Hafnium acts as a diffusion barrier at 0.7 nm thickness.

Exchange coupling energy reaches 0.11 mJ/m^2 with thin Hf layers.

Coupling sign reverses at the ferrimagnet's compensation temperature.

Abstract

A thin Hafnium film is shown to act both as an effective diffusion barrier for manganese at a thickness of 0.7 nm, and as an effective exchange coupling layer in a sandwich structure with perpendicular magnetic anisotropy. The magnetic layers are Co$_{20}$Fe$_{60}$B$_{20}$ and the low moment ferrimagnet Mn$_2$Ru$_x$Ga (MRG). The coupling changes sign at the compensation temperature of MRG and the exchange energy reaches 0.11 mJm$^{-2}$ for the thinnest Hf interlayers. Ruthenium, the usual metal of choice for coupling ferromagnetic layers in thin film heterostructures, cannot be used with the zero-moment half metal MRG because of Ru interdiffusion. Due to its large coercivity near compensation, the MRG can act as an effective source of exchange pinning.

Full text

Magnetic thin films; perpendicular magnetic anisotropy; zero-moment half metal; interlayer exchange interaction; compensation temperature

Spin electronics has influenced the information revolution through technologies such as affordable high-density magnetic storage and high-speed nonvolatile magnetic memory, taking advantage of the giant magnetoresistance or tunneling magnetoresistance (GMR or TMR) effects to read and store information. In the magnetic thin film stacks employed as field sensors or magnetic switches, it is useful to be able to control the magnetic coupling between adjacent layers. Direct contact between an antiferromagnetic and a ferromagnetic layer can lead to exchange bias of the latter, enabling it to serve as a pinned or reference layer in any spin valve sandwich structure. Stacks may also make use of the interlayer exchange coupling Parkin, More, and Roche (1990); Parkin (1991) to form synthetic antiferromagnets (SAFs) where two ferromagnetic layers are coupled antiferromagnetically via a thin layer of nonmagnetic metal in a structure that creates no stray field. The exchange coupling can be further engineered to form structures that enable demonstration of ultrafast chiral domain wall motion Yang, Ryu, and Parkin (2015) and deterministic spin orbit torque-induced switching of a perpendicular ferromagnet without an external field Lau et al. (2016). The sign of the exchange usually oscillates with the nonmagnetic spacer thickness, and Ru has been found to be the most effective Parkin, More, and Roche (1990). A peak in the antiferromagnetic coupling is found for a ruthenium thickness of $0.9\text{\,}\mathrm{nm}$, where the exchange energy reaches $0.1\text{\,}\mathrm{mJ}\text{\,}{\mathrm{m}}^{-2}$. An early systematic study by Parkin found no coupling between 3$d$ ferromagnetic layers for the $d^{2}$ transition metals Ti, Zr or Hf Parkin (1991).

An alternative to using an exchange-biased SAF to pin the reference layer is to exchange-couple it to a low-moment uniaxial ferrimagnet that is close to compensation. This could produce the necessary exchange bias, whilst adding magnetic mass to stabilise ultra-thin CoFeB layers in perpendicular magnetic tunnel junctions (MTJs) Ikeda et al. (2010). The uniaxial anisotropy $K_{1}$ ensures that, as the magnetisation $M_{s}$ of the ferrimagnet approaches zero at the compensation temperature ($T_{\text{comp}}$), the anisotropy field $2K_{1}/\mu_{0}M_{s}$ diverges and the coercivity increases significantly Thiyagarajah et al. (2015). Thin films of the recently-discovered compensated ferrimagnetic half-metal Mn2RuxGa (MRG) Kurt et al. (2014) have been shown to exhibit high spin-polarisation and a large coercivity near compensation Borisov et al. (2016); Thiyagarajah et al. (2015), due to its perpendicular magnetic anisotropy (PMA) caused by biaxial substrate-induced strain of the cubic Heusler structure. MRG has a Heusler-type structure, with Mn on $4a$ and $4c$ sites, Ga on $4b$ sites and Ru on most of the $4d$ sites.Kurt et al. (2014); Betto et al. (2015) Some Mn-Ga antisite substitution is present in the structure Betto et al. (2015), which leads to a spin gap at the Fermi level Žic et al. (2016). Here we show that MRG can be coupled antiferromagnetically to an adjacent perpendicular CoFeB layer via a thin Hf interlayer, and the effect can be used to produce exchange pinning.

The Mn2Ru0.7Ga used in this study has $T_{\text{comp}}=$240\text{,}\mathrm{K} so data can be conveniently collected at temperatures above and below compensation. Stacks of MgO//MRG(35)/Hf($t$)/CFB(1)/MgO(1.4)/Ta(1) [thicknesses in $\mathrm{nm}$], where $t=$ $0.30.40.50.60.70.80.91.11.20.3\text{\,}\mathrm{nm}$, were grown on $10\text{\times}10\text{\,}\mathrm{mm}$ single-crystal MgO (100) substrates in Shamrock magnetron sputter deposition tool. Here CFB denotes Co*20*Fe*60*B*20* (nominal target composition). The MRG was DC co-sputtered from stoichiometric Mn2Ga and Ru targets, at a substrate temperature of $380\text{\,}\mathrm{\SIUnitSymbolCelsius}$.Kurt *et al.* ([2014](#bib.bib7)); Thiyagarajah *et al.* ([2015](#bib.bib6)) After cooling to room temperature, Hf and CFB were DC sputtered and, finally, MgO and Ta were RF sputtered. Stacks were post-annealed under a vacuum of $\sim${10}^{-6}\text{\,}\mathrm{mbar} in an applied out-of-plane magnetic field of $800\text{\,}\mathrm{mT}$ at $300\text{\,}\mathrm{\SIUnitSymbolCelsius}$ for $30\text{\,}\mathrm{m}\mathrm{i}\mathrm{n}\mathrm{u}\mathrm{t}\mathrm{e}\mathrm{s}$ to enhance the PMA of CFB. In-plane magnetised reference stacks Yuasa and Djayaprawira (2007) of composition Ta(5)/Ru(10)/Ta(5)/Ni81Fe19(5)/Ir22Mn78(10)/Co90Fe10(2.4)/X($t$)/CFB(3)/MgO(1.4)/Ta(5), where X = Hf or Ru and $t$ = 0.5, 0.7, 0.9 and $1.0\text{\,}\mathrm{nm}$, were deposited on thermally oxidised $20\text{\times}20\text{\,}\mathrm{mm}$ Si/SiO2 coupons at room temperature. Reference stacks were post-annealed in an in-plane applied field at $350\text{\,}\mathrm{\SIUnitSymbolCelsius}$ for $60\text{\,}\mathrm{m}\mathrm{i}\mathrm{n}\mathrm{u}\mathrm{t}\mathrm{e}\mathrm{s}$ to set the exchange bias at IrMn/CoFe interface. The stacks are illustrated in Figure 1 a) and b).

Anomalous Hall effect (AHE) was measured using the 4-point van der Pauw method, with indium contacts, an applied current of $1\text{\,}\mathrm{mA}$ and an applied magnetic field of up to $1\text{\,}\mathrm{T}$. Additional Hall measurements up to $14\text{\,}\mathrm{T}$ were carried out in a Quantum Design PPMS system. Room temperature magnetisation measurements with magnetic field applied in-plane or perpendicular to the films were carried out using a Quantum Design SQUID with a maximum field of $5\text{\,}\mathrm{T}$.

If we are to use MRG, and more generally the Mn-containing Heusler alloys, in an exchange coupling system, we have to consider three factors when deciding what metal interlayer to use it with: Does it prevent interdiffusion between the MRG and the adjacent layers?Hayakawa et al. (2006) Does it maintain perpendicular anisotropy of the overlayer?Liu, Cai, and Sun (2012); Pai et al. (2014); Oh et al. (2014) Does it provide sufficient coupling between the MRG and an adjacent ferromagnetic layer?Radu et al. (2012) An ideal material will do all three, but the selection is limited. Initially, a number of candidates were chosen, namely W, Zr, Ru, Mo, Hf, Ta and TiN. Some of them were effective at blocking diffusion between layers 111The diffusion profile has been recorded by Time-of-Flight Secondary Ion Mass Spectroscopy, in preparation., which is necessary to preserve spin-polarisation and device integrity (Hayakawa et al., 2006). Only Mo, Hf and Ta also maintained the PMA of the CFB overlayer, and of these Hf was the only one to demonstrate strong effective exchange coupling between MRG and CFB. An order of magnitude smaller effect was found with the thinnest Ta ($0.4\text{\,}\mathrm{nm}$). Ruthenium, the usual material of choice for exchange coupling ferromagnetic films, is unsuitable due to the lack of PMA in Ru/CFB/MgO heterostructures Worledge et al. (2011) and the incorporation of Ru into the MRG layer, which alters the interfacial magnetic properties.

The Mn2Ru0.7Ga composition used here has a room-temperature magnetisation of $40\text{\,}\mathrm{kA}\text{\,}{\mathrm{m}}^{-1}$ and coercivity $\mu_{0}H_{c}=$0.7\text{,}\mathrm{T}. For $t=$0.3\text{\,}\mathrm{nm}, we found neither PMA nor exchange coupling of CFB for any of the seven interlayer materials.

The effect of the exchange coupling is nicely illustrated for a Hf interlayer with $t=$0.4\text{,}\mathrm{nm} in Figure [2](#S0.F2), where the CFB layer is perpendicular and minor hysteresis loops can be readily measured using the AHE. The narrow hysteresis loop of the CFB with a coercivity of $16\text{\,}\mathrm{mT}$ is exchange shifted by $\mu_{0}H_{ex}$ = $175\text{\,}\mathrm{mT}$ in the direction opposite to the saturation field. These measurements have been repeated for greater Hf thickness, and $\mu_{0}H_{ex}$ decreases monotonically with $t$, falling to zero at $t\approx$1.0\text{\,}\mathrm{nm} (Figure 3). Note that the exchange shift of the loop in Figure 2 is towards negative fields for ferromagnetic effective interlayer coupling. This is the same as the normal exchange bias effect caused by direct contact between antiferromagnetic and ferromagnetic layers.

Complete hysteresis loop for $t=$0.4\text{,}\mathrm{nm} is shown in the inset Figure [3](#S0.F3), and the values of $\mu_{0}H_{ex}$ measured by AHE and SQUID magnetometry are compared in Figure [3](#S0.F3). In agreement with the AHE data, the greatest exchange shift $\mu_{0}H_{ex}=$174\text{\,}\mathrm{mT} is found for the thinnest interlayer, with $t=$0.4\text{,}\mathrm{nm}, and the shift has disappeared at $t=$0.9\text{\,}\mathrm{nm}. The exchange energy $J_{ex}$ is estimated as $\mu_{0}H_{ex}M_{\text{CFB}}t_{\text{CFB}}$.(Meiklejohn and Bean, 1956; Meiklejohn, 1962) Taking $M_{\text{CFB}}=$0.6\text{,}\mathrm{MA}\text{,}{\mathrm{m}}^{-1}$$, the maximum value of $J_{ex}$ is $0.11\text{\,}\mathrm{mJ}\text{\,}{\mathrm{m}}^{-2}$. This is comparable to the antiferromagnetic exchange coupling for CoFe/CFB layers separated by a Ru interlayer of thickness $t_{\text{Ru}}$ $\approx$ $0.9\text{\,}\mathrm{nm}$Jung, Lim, and Lee (2012). It has to be noted that this thickness corresponds to the second AFM peak of Ru which is easily attainable, while reports on the first AFM peak showing stronger coupling ($2.2\text{\,}\mathrm{mJ}\text{\,}{\mathrm{m}}^{-2}$ - $5\text{\,}\mathrm{mJ}\text{\,}{\mathrm{m}}^{-2}$) with $t_{\text{Ru}}$ = $0.4\text{\,}\mathrm{nm}$-$0.5\text{\,}\mathrm{nm}$ are scarceParkin, More, and Roche (1990); Lau et al. (2013); Yakushiji et al. (2015).

In Figure 4 a) and b) full AHE loops are recorded above ($300\text{\,}\mathrm{K}$) and below ($200\text{\,}\mathrm{K}$) $T_{\text{comp}}$. The applied field sequence was $+\rightarrow-\rightarrow+$. We analyse the behaviour in terms of the direction of the CFB magnetisation and the MRG $4c$ sublattice one, which is the dominant contribution to the MRG electronic structure at the Fermi level Betto et al. (2015). At $300\text{\,}\mathrm{K}$ ($T>T_{\text{comp}}$), the $4c$ sublattice is the minority sublattice and hence is antiparallel to the net magnetisation. CFB switches after passing zero applied field (Figure 4a), the coupling is ferromagnetic with the net MRG moment and antiferromagnetic with the $4c$ moment. At $200\text{\,}\mathrm{K}$ (Figure 4b), the $4c$ sublattice is the majority sublattice, parallel to the net MRG magnetisation. CFB now switches before reaching zero applied field, corresponding to antiferromagnetic coupling with the net moment and with the $4c$ sublattice. The sign of the effective coupling relative to the net magnetisation changes when crossing $T_{\text{comp}}$. Also note that the MRG contribution to the AHE changes sign when crossing $T_{\text{comp}}$ in accordance with our previous results.Thiyagarajah et al. (2015)

In order to highlight the dependence of the exchange coupling on the sublattice moment, as opposed to the net moment, we measured minor loops in the range $\pm$0.5\text{,}\mathrm{T}$$, both above and below $T_{\text{comp}}$ (Figure 4 c and d) after saturating in a field of $+14\text{\,}\mathrm{T}$. Two measurements were made in each case. First after saturation at the measurement temperature, and subsequently after saturation on the other side of $T_{\text{comp}}$ and heating (Figure 4c) or cooling (Figure 4d) back to the measurement temperature in zero field. When crossing $T_{\text{comp}}$ in zero field the net MRG magnetization, $M$, changes from positive to negative while the direction of the $4c$ magnetization, $M_{4c}$, is preserved and therefore all four possible combinations between $M$ and $M_{4c}$ are accessible. We find that the coupling is always antiferromagnetic with respect to $4c$, but the alignemnt is ferromagnetic with respect to the net moment above $T_{\text{comp}}$ and antiferromagnetic below.

Finally, we show in Figure 5 the SQUID loops for the in-plane reference stack of Figure 1b) where a CoFe layer is separated by a hafnium or ruthenium interlayer from the upper CFB layer. The data show clear evidence of antiferromagnetic exchange coupling for Ru($0.9\text{\,}\mathrm{nm}$), and strong ferromagnetic coupling when the spacer is Hf($0.7\text{\,}\mathrm{nm}$), in contrast with early reportsParkin (1991). Comparing the exchange coupling energy obtained with IrMn and Ru in our reference stacks, the exchange shifts are $\mu_{0}H_{ex}^{\text{IrMn}}~{}\approx~{}$22\text{,}\mathrm{mT} and $\mu_{0}H_{ex}^{\text{Ru}}~{}\approx~{}$33\text{\,}\mathrm{mT}, hence the exchange coupling energies are $J_{ex}^{\text{IrMn}}~{}\approx~{}$0.04\text{,}\mathrm{mJ}\text{,}{\mathrm{m}}^{-2} and $J_{ex}^{\text{Ru}}~{}\approx~{}$0.15\text{\,}\mathrm{mJ}\text{\,}{\mathrm{m}}^{-2}.

The strength of the coupling via a $0.7\text{\,}\mathrm{nm}$ Hf interlayer is only about $20\text{\,}\mathrm{\char 37\relax}$ of that for a similar thickness of Ru, but for a $0.4\text{\,}\mathrm{nm}$ interlayer it is the same as that of the Ru ($t_{\text{Ru}}$ = $0.9\text{\,}\mathrm{nm}$), and it shows no oscillations of sign, neither in the MRG-based stacks nor in the reference stacks. This suggests a coupling mechanism that is different to the accepted one for Ru interlayers.Bruno (1995)

Unlike rutheniumParkin, More, and Roche (1990), where the sign of the exchange coupling oscillates with interlayer thickness with a period of about $1.1\text{\,}\mathrm{nm}$, reminiscent of the RKKY interaction, the exchange coupling through hafnium is always antiferromagnetic in the case of MRG/Hf/CFB and ferromagnetic for the symmetric CoFe/Hf/CFB stack, and it decreases monotonically to zero with increasing interlayer thickness. The coupling is strongest for the very thin $0.4\text{\,}\mathrm{nm}$ films that are only two atomic layers thick. A possible mechanism for the coupling is hybridization of the unpolarized Hf $6s^{2}5d^{2}$ conduction electrons with the spin polarized $4c$ sub-band at the Fermi surface of MRG. This polarizes the first few layers of Hf, rather as the $5d6s$ bands of the rare-earth metals are polarized by the $4f$ core. Coupling between an atom with a nearly full $d$ band (Co) and one with a nearly empty $d$ band (Hf) is antiferromagnetic.Campbell (1969, 1972) The effective exchange coupling is ferromagnetic when $4c$ is the minority sublattice above $T_{\text{comp}}$ and antiferromagnetic when $4c$ becomes the majority sublattice below $T_{\text{comp}}$. We saw from Figure 4 that the exchange coupling changes sign at $T_{\text{comp}}$ in just the sense that is predicted. For the symmetric reference stack CoFe/Hf/CFB these two antiferromagnetic interactions leads to an overall ferromagnetic exchange as observed. Similar effect has been previously observed in $L$10-MnGa/Fe1-xCoxMa et al. (2014) as a function of $x$ which was attributed to the reversal of the interfacial spin polarisation.

Another possible mechanism is exchange coupling via pinholes in the thin interlayer which could produce the same effect, with a sign change at compensation. The $0.4\text{\,}\mathrm{nm}$ Hf layer is unlikely to be entirely defect-free but this explanation is implausible for several reasons. First it should apply to any interlayer with pinholes. Hf is the only one of the seven metals studied to exhibit the strong exchange coupling. Furthermore, no exchange was observed for $t_{\text{Hf}}=$0.3\text{,}\mathrm{nm}$$ where the case for pinholes should have been even stronger. Finally when CFB is deposited directly on MRG, no exchange is observed. We can therefore discount this explanation.

Exchange bias has previously been demonstrated by direct coupling of a compensated ferrimagnet with a ferromagnetUngureanu et al. (2010); Romer et al. (2012), but the advantages of our approach are the MRG high ordering temperature ($T_{\mathrm{C}}$ $\approx$ $550\text{\,}\mathrm{K}$)Kurt et al. (2014) and the broad compensation temperature tunability ($2\text{\,}\mathrm{K}$-$400\text{\,}\mathrm{K}$)Thiyagarajah et al. (2015). The use of a single layer of MRG or some other near-zero-moment ferrimagnet close to compensation can in principle replace the standard bottom SAF reference stack for MTJ with PMA, e.g. buffer layer/[Co/Pt]n/Ru/[Co/Pt]n/Ta or W/CFB,Cuchet et al. (2013); Yakushiji et al. (2015) provided the uniaxial anisotropy is finite so that the anisotropy field diverges when $M_{s}\rightarrow 0$ as discussed above. The difficulty with using MRG, or any Mn-based ferrimagnet in direct contact with a ferromagnet or a tunnel barrier such as MgO is manganese diffusion into the overlayer.Borisov et al. (2016); Kubota et al. (2011); Hayakawa et al. (2006) Here we have demonstrated the feasibility of a simple bcc-textured MRG/Hf/CFB reference structure that can be grown directly on MgO as an alternative. A fully bcc-textured MTJ stack with robust reference layer can therefore be realised. Furthermore, the use of MRG with high spin polarisation in conjunction with a Hf interlayer with relatively high spin transparency Pai et al. (2014) may also be advantageous for enhancing the spin-torque efficiency of the MTJ.

The research on Mn-based Heusler ferrimagnets have instigated considerable scientific interest recently due to broad range of magnetic propertiesKurt et al. (2014); Nayak et al. (2015); Betto et al. (2017), generation of sub-THz oscillationAwari et al. (2016); Mizukami et al. (2016) and successful integration in magnetic tunnel junctionBorisov et al. (2016); Jeong et al. (2016). We have previously demonstrated that the compensated half-metallic Heusler alloy MRG combines the advantageous properties of both ferro- and anti-ferromagnets without their limitations. It can be used as a source of spin-polarised currents, is immune to external fields and produces no stray field of its own. In this work we have demonstrated yet another use: as a reference layer in a spin electronic stack, either alone or as a pinning material, thereby greatly reducing the number of individual layers in a stack.

K.B. acknowledges financial support from Science Foundation Ireland within SSPP (11/SIRG/I2130). This work was partly supported by Science Foundation Ireland under grants 13/ERC/12561 and 12/RC/2278 (AMBER). Gavin D’Arcy was supported by INTEL (Irl.) Ltd. as a researcher in residence at CRANN. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No DLV-737038 (TRANSPIRE).