High temperature magnetic stabilization of cobalt nanoparticles by an antiferromagnetic proximity effect

Thermal activation tends to destroy the magnetic stability of small magnetic nanoparticles, with crucial implications in ultra-high density recording among other applications. Here we demonstrate that low blocking temperature ferromagnetic (FM) Co nanoparticles (TB<70 K) become magnetically stable above 400 K when embedded in a high N\'eel temperature antiferromagnetic (AFM) NiO matrix. The origin of this remarkable TB enhancement is due to a magnetic proximity effect between a thin CoO shell (with low N\'eel temperature, TN; and high anisotropy, KAFM) surrounding the Co nanoparticles and the NiO matrix (with high TN but low KAFM). This proximity effect yields an effective AFM with an apparent TN beyond that of bulk CoO, and an enhanced anisotropy compared to NiO. In turn, the Co core FM moment is stabilized against thermal fluctuations via core-shell exchange-bias coupling, leading to the observed TB increase. Mean-field calculations provide a semi-quantitative understanding of this magnetic- proximity stabilization mechanism.

Unfortunately, most high-K materials require high-temperature annealing processes to obtain the desired phase, which could hamper their implementation in certain structures. Thus, FM-AFM exchange coupling alternatives may be an appealing option. In fact, it has been demonstrated [4] that ferromagnetic-antiferromagnetic (FM-AFM) interfacial exchangecoupling is an effective method, later patented by Seagate [12], to increase the effective K of FM nanoparticles. However, a T B enhancement beyond RT using this approach has been rarely reported [22][23][24][25][26] (where often broad particle size distribution can partly account for the "apparent" T B increase [22][23][24][25]). The reason for this scarcity is that high Néel temperature (T N ) AFMs tend to have a low anisotropy constant (e.g., NiO), and vice versa (e.g., CoO), while substantial values of both properties are required for high-temperature stabilization.
This limitation could, in principle, be overcome by exploiting proximity effects, i.e., the interfacial synergetic hybridization of the properties of two AFM materials having complementary properties (here, high T N and high K). Although this phenomenon is best known in superconductivity [30], proximity effects in bi-or multi-layered magnetic systems (i.e., magnetic proximity effects) have also been studied [31]. In contrast, and despite their strong technological presence, proximity effects involving nanoparticles have been hardly explored [32,33].
In this Letter we demonstrate a proximity effect between two AFMs (a CoO shell and a NiO matrix) on FM particles (Co) and the resulting thermal stabilization of the NPs well above RT (with an ~10-fold enhancement of T B to exceed 400 K), and propose a mean-field model to gain insight into the nature of such AFM proximity effect.
The low-temperature hysteresis loops of the Co/CoO-NiO samples (S-series) measured after field cooling are shown in Figure 1(b). The loops show rather large coercivities (µ 0 H C~0 .4 T) and loop shifts (i.e, H E , exchange bias) µ 0 H E~0 .4 T. In contrast, the loops exhibit a rather small vertical shift (less than 1% of M S ). In the reference samples, where NiO is replaced by Nb (R-series), H C is considerably smaller (µ 0 H C~1 0 mT) and no loop shifts are observed [ Figure 1 Remarkably, the T=300 K hysteresis loops shown in Figure 1(c) evidence that the samples are not superparamagnetic (i.e., with remanence, M R , and H C >0). Not only is µ 0 H C~6 mT, but H E is surprisingly large (e.g., µ 0 H E =14 mT for S50He). In contrast, the reference samples have vanishing M R and H C at T = 300 K, revealing their superparamagnetic state ( Figure S1 [34]). The present results demonstrate that Co nanoparticles of a few nm can be made magnetically stable above RT, up to at least T=400 K [41]. The origin of the enhanced magnetic stability must reside in some coupling existing between the Co nanoparticles and the high-T N AFM matrix, NiO (T N =520 K), since using CoO alone as matrix limits the T B enhancement to 290 K [T N (CoO)] [4]. However, NiO is known to have a low anisotropy [42], leading to small H E and low T B [H E ] (often below RT) [25,27,[43][44][45][46]. This highlights that using a high-T N material is not sufficient in itself to reach high-temperature stability.
The first indication of the origin of the observed effects is the very large H E measured in the Co/CoO-NiO series at T=10 K. NiO alone cannot induce such high H E values, hence, the highly anisotropic CoO shell must be involved in the H E enhancement. However, isolated Co/CoO nanoparticles with a thin (natural oxidation) CoO shell usually exhibit very small H E [39,40]. Three main types of processes have been proposed to achieve large H E in Co/CoO systems [4,[13][14][15]47]: (i) forced oxidation of the Co particles to form thick AFM CoO shells [13][14][15], (ii) matching the crystallographic structure between the CoO shell and the matrix (which structurally stabilizes the CoO shell) [47] and (iii) coupling the CoO shell to an AFM matrix (magnetic stabilization) [4]. Our low-oxygen synthesis method allows to safely rule out the first possibility [13,15].   [31], where the overall properties of AFM 1 /AFM 2 systems are the combination of both counterparts [50][51][52]. This concept has been applied recently to other types of AFMs such as IrMn/FeMn [53] and it must take place in the Co/CoO-NiO system, where the overall T B is determined by the combined effect of the CoO shell coupled to the NiO matrix. However, to explain the high-temperature stability of the Co nanoparticles, a polarization of the Co AFM moments in the CoO shell is not sufficient; the overall anisotropy of the CoO-NiO couple, ultimately felt by the Co particles, must also remain sufficiently high. Consequently, the proximity effect between CoO and NiO has a two-fold consequence where both the Co induced magnetization and the overall anisotropy are involved [54].
For systems composed of FM nanoparticles embedded in an AFM matrix, H E is classically expressed as: where M FM is the FM magnetization, V is the volume of the ferromagnet, γ is the interfacial coupling energy per unit surface area, and the associated surface area. The evaluation of γ 0 , the 0 K coupling energy, constitutes the major difficulty in the analysis of exchange-bias systems. Here, γ 0 is only taken as an experimental parameter. Naively, the temperature dependence of γ should be proportional to the interfacial AFM staggered magnetization  Figure S2a [34]). The temperature dependence of the surface magnetization was then calculated by assuming that, for surface atoms, the number of neighbours is reduced from 12 in the bulk to 9. The temperature dependence of the surface magnetization ( Figure S2b [34]) is reminiscent of the temperature dependence of the remanent magnetization in CoO nanoparticles, which has been related to surface magnetic moments [55]. Additionally, the calculated variation of the surface magnetization reproduces correctly the temperature dependence of H E in the Co/CoO-CoO system (i.e., Co/CoO nanoparticles embedded in a CoO matrix [4]) in the whole temperature range (compare the calculated temperature dependence of the CoO surface magnetization in Figure S2 [34] to the experimental µ 0 H E (T) in Figure S3 [34]).
Obviously, expression (1) cannot explain the sizable H E measured in the Co/CoO  CoO/NiO multilayers prepared by sputtering [52]. Moreover, the total pure CoO equivalent thickness (≈1 nm corresponding to 2 pure CoO layers, each 0.268 nm thick, + 1.5 equivalent CoO layers from the 3 intermixed layers) is consistent with the oxygen-poor synthesis conditions of the nanoparticles. In the model, a given atom has 12 neighbours in total: 6 neighbours in the shell it belongs to, and 3α p (α m ) atoms in the preceding (next) shells, where the coefficients α p (α m ) are proportional to the respective surface area of each considered shell [34]. Calculations then revealed that a significant magnetization is maintained in CoO above its bulk T N (red line in Figure S2b [34]), via the proximity effect with NiO. Since the Co atoms in CoO at the interface with the core are directly exchange-coupled to the Co FM core (consequently, directly involved in exchange-bias), the existence of a significant CoO staggered magnetization above T N (CoO) directly accounts for the persistence of exchangebias in this temperature range. Note that due to the short-range nature of the interactions, the NiO-induced polarization of CoO at the CoO/Co interface becomes negligible if more than two non-intermixed CoO layers are considered.
Although this calculation demonstrates that proximity effects in Co/CoO-NiO can account for H E above T N (CoO), it does not explain the rapid decrease of H E with increasing temperature in Co/CoO-NiO compared to Co/CoO-CoO (compare Figure 3 and Figure S3 [34]). Actually, the AFM component in exchange biased systems is usually composed of nanosized grains, which are prone to superparamagnetic effects [56]. To account for the possible existence of superparamagnetic CoO grains, a reduction coefficient must be applied to H E derived from Equation (1), which does not include thermal activation effects: where  [58,59].
As a consistency test of our model, the temperature dependence of H E in the Co/CoO-CoO system was re-calculated using the same parameters as above (red line in Figure S3 [34]) . As in Co/CoO-NiO, the calculated curves (with the interfacial coupling coefficient γ 0 as an adjustable parameter) give fair account of the experimental data. The theoretical curves obtained with γ 0 =1.3×10 -3 J/m 2 , respectively accounting and neglecting thermal activation, differ only slightly close to T N ( Figure S2 [34]), indicating that in the Co/CoO-CoO system, only a minor fraction of the AFM grains become superparamagnetic as temperature is increased. This situation is related to the high K AFM characteristic of CoO [42]. Altogether, these results reflect the dual role of CoO and NiO in the magnetic stabilization of Co nanoparticles, i.e., while NiO contributes with high-T N , CoO supplies the high anisotropy.
In conclusion, we have presented the foremost example of exchange-bias particle stabilization exploiting magnetic proximity effects. Co/CoO core/shell (~5 -7

Supplemental Material Experimental Details
Deposition Conditions. Films about 350 nm thick, were grown by high-speed sequential deposition of inert gas-condensed Co nanoparticles (in a modified commercial cluster-source) and a rf-sputtered NiO matrix (from a NiO target) on to thermally oxidized Si(100) substrates using a rotating (0.3 Hz) sample holder [1,2]. Two of the samples were grown using different sputtering powers in the cluster source (50 W -S50; 80 W -S80). Note that increasing the sputtering power increases the average particle size [3] and the deposition rate from 0.2 (50 W) to 0.4 Å/s (80 W). The third sample was grown at 50 W but using twice as much carrier gas (He) -S50He-as in the other samples (5 sccm) with a view of obtaining smaller nanoparticle size [3]. Importantly, oxygen is partially released during the deposition of oxide materials (NiO in our case) by plasma techniques [4] so that the Co nanoparticles partially oxidize to form Co/CoO core/shell nanoparticles. This leads to a core-shell structure of Co/CoO nanoparticles embedded in a NiO matrix (Co/CoO-NiO). Note that in the present conditions the shell grows as CoO and not Co 3 O 4 [1,2]. Reference samples (named analogously, but starting by R) using a niobium matrix instead of NiO (while keeping the same cluster-source parameters for the nanoparticle synthesis) were also prepared. The NiO and Nb matrix targets were sputtered at 150 and 100 W, respectively, which provides deposition rates much larger than those of the Co nanoparticles.
The Co/Ni composition ratio, as derived from energy dispersive microanalysis, was lower than 5%, which implies that the concentration of Co nanoparticles is sufficiently dilute to safely neglect interparticle interactions as well as exchange-bias connectivity effects [5,6] resulting from direct contact between the nanoparticles.
Morphological characterization. Transmission electron microscopy (TEM) (FEI Tecnai F20 S/TEM operating at 200 kV) was used to estimate the particle size using grids exposed to the nanoparticle beam after the film deposition. The TEM characterization of the samples shows that the particle diameter ranges from ~5 nm (S50He) to ~7 nm (S80) with relatively narrow size distributions (see Figure S4). Note that to facilitate the analysis, the density of nanoparticles in the TEM grid is much higher than in the studied granular films. The size is somewhat approximate since the nanoparticles for TEM analysis are unprotected and hence they tend to oxidize. The Co nanoparticle size was estimated assuming complete oxidation of the particles in the TEM grid to CoO and using Co and CoO bulk density values. The molecular field on a Co atom in shell i is equal to: (S1 ) where is the number of Co or Ni atoms per unit volume, equal to 51.  Figure S2. Note that no fitting parameters are involved in this calculation.
Since the CoO shell is due to the surface oxidation of the Co nanoparticle, it is not expected to be a single crystal, but to be formed by several nanograins (see Fig. S5). TEM observations indicate that the oxide grains at the surface of Co nanoparticles are rather small, 2-3 nm in lateral size. Thus, for simplicity, the oxide grains are taken as small cylinders with their axis perpendicular to the Co nanoparticle surface and 2.5 nm in diameter. Assuming that the oxide grains are not exchange coupled to each other [7], the thermal stability will be linked to the stability of the individual grains rather than to the whole oxide shell volume at the surface of the Co nanoparticles. Consequently, the volume extracted from the fit to the model, Theoretical calculations for the Co/Co-CoO sample Figure S3. Temperature dependence of H E for the Co/CoO-CoO system: Experimental data (•) [9]; Calculated temperature dependence of H E , neglecting thermal activation ( ⋅ ) and taking thermal activation into account (). Figure S4. Transmission electron micrograph (a) and particle size distribution histogram (b), of the S50He cobalt nanoparticles. Note that these particles are partially oxidized (after exposure to ambient conditions), consequently the particles appear larger than they are.