Tunnel Magnetoresistance in Self-Assemblies of Exchange-Coupled Core/Shell Nanoparticles

We report the precise control of tunneling magnetoresistance (TMR) in devices of self-assembled core/shell Fe 3 O 4 /Co 1 − x Zn x Fe 2 O 4 nanoparticles (0 ≤ x ≤ 1). Adjusting the magnetic anisotropy through the content of Co 2+ in the shell, provides an accurate tool to control the switching ﬁeld between the bistable states of the TMR. In this way, diﬀerent combinations of soft/hard and hard/soft core/shell conﬁgurations can be envisaged for optimizing devices with the required mag-netotransport response


I. INTRODUCTION
The possibility to manipulate the electrical resistive state of magnetic/non-magnetic multilayers by an external magnetic field (giant magnetoresistance, GMR) was demonstrated already 30 years ago. [1,2] The strong coupling between the electron spin and charge degrees of freedom and the development of the tools for their manipulation, triggered the growth of a new field called spintronics. [3,4] The fabrication of magnetic tunnel junctions (MTJ) constitutes one of the most important advances in this field since then. [5,6] A MTJ is composed of two layers of ferromagnetic conductors separated by an insulating tunneling barrier, typically of ≈1 nm. The different density of states at the Fermi level, N(E F ), of the spin up/down subbands of the ferromagnetic metals imply a spin-dependent tunneling probability. Therefore, the electrical resistance of the device switches between high/low resistance states as the magnetic field changes the relative orientation of the magnetizations of the two magnetic layers (tunneling magnetoresistance, TMR). The MTJ devices present high versatility and a great degree of functionalization, allowing to combine electrodes and barriers of different nature, where large tunneling magnetoresistance, up to hundreds of percents at room temperature, was obtained [7,8]. However their fabrication is a challenge, involves advanced thin film deposition techniques and complex microfabrication procedures.
Tunneling magnetoresistance has also been studied in simpler nanostructures as granular or disordered single films [9,10], where the grain boundaries act as tunnel junction barriers.
However, the characteristic of the barrier cannot be controlled in these nanostructures and lower TMR values are obtained.
On the other hand the spectacular advances of the chemical synthetic methods produced over the last few years, offer an affordable route for the synthesis of complex nanostructures, with a precise control of their chemical composition, shape and size. [11] These can be assembled in crystal-like structures over large areas, in which the organic capping layer or a non-magnetic shell protecting the particles acts as a tunneling barrier that controls the electronic transport. [12][13][14][15][16][17] Spin-dependent electrical transport and large magnetoresistance was also observed in devices formed by assembling of conducting magnetic nanoparticles (MNPs) [14,18,19] or binary nanoparticles superlattice [20][21][22].
However, an important challenge that must be addressed in this field is the design of strategies to tune the switching field of the TMR devices, which is entirely determined by the anisotropy of the magnetic material. [9,10,[23][24][25][26] Therefore, a good handling over the coercivity of the magnetic nanoparticles would allow the control over the TMR of the assemblies, in a similar approach as that used in multilayers. [27] In this regard, an exciting possibility is the fabrication of devices based on self-assemblies of exchange coupled core/shell MNPs with tailored magnetic properties. [28] The coercive field in these systems can be finely modified through the interface magnetic coupling [29][30][31][32][33][34], the core size and shell thickness, [35][36][37] or the magnetic anisotropy of the components. [23,[38][39][40] Devices of this type should provide a way to manipulate at will the characteristic switching field of TMR by controlling the magnetic coupling across the core/shell interface.
In this way, core-shell nanoparticles combine the properties of multilayered based tunnel junctions and granular or disordered thin films, offering very high versatility with a simple fabrication process.
Here we report the precise control of the TMR in self-assemblies of half metallic ferrimag- . Progressive replacement of Co 2+ by Zn 2+ in the shell reduces the magnetic anisotropy and shifts the maximum of the TMR of the self-assembled device in a perfect correlation with the magnetic response. These results demonstrate the feasibility of tunning the TMR switching field in self-assembled devices formed by magnetic core/shell nanoparticles.

II. EXPERIMENTAL PROCEDURE
Fe 3 O 4 /Co 1−x Zn x Fe 2 O 4 core/shell nanoparticles were fabricated by seed mediated high temperature decomposition of metal-acetylacetonates in benzyl ether assisted by oleic acid and oleylamine, based on the method described in Refs. [39,41,42] The self-assembly of the core/shell MNPS was done at the liquid-air interface following the procedure reported in Refs. [20][21][22]. In the assembly process schematized in the upper panel of Figure 1, a drop of 10 µL of solution with 5 mg/mL of nanoparticles in hexane is drop-casted onto the surface of triethylene glycol in a Teflon container, which was then covered by a glass slide. In order to transfer the assemblies to a substrate, the teflon vessel of 1.5x1.5x1.0 cm 3 was designed with a 30 • inclined base-plane where the substrate was located previously and was completed cover by the triethylene glycol. A self-assembled structure is formed after complete evaporation of hexane (between 10-15 min). After that the triethylene glycol was removed very slowly using a syringe in order to gently deposit the assembled film on the substrate. All the samples received a thermal treatment in a vacuum atmosphere (∼10 −3 Torr) in order to reduce the organic coating of the particles and to promote a closer contact between them. The decomposition temperature of the organic nanoparticle coating was determined from thermogravimetric analysis. The self-organized nanoparticles were heated from room temperature up to 400 • C at heating rate of 15 • C/min, kept at 400 • C by 30 min and then cooled to room temperature at 15 • C/min. to perform the structural characterization of the self-assemblies of core/shell nanoparticles, they were transferred from the triethylene glycol surface to commercial silicon nitride TEM grids followed by thermal annealing. Atomic force microscopy (AFM) measurements were done in a Veeco (now Bruker) Dimension 3100 SPM in tapping mode using a standard tip.
The 2 µm scans were done using a scan frequency of 1Hz and after waiting 30 minutes for thermal stabilization and noise reduction. No modification of the surface was observed after the measurements.
The magnetic properties were studied using a commercial superconducting quantum interference device magnetometer (SQUID, MPMS Quantum Design). To perform the measurements the self-assembled nanoparticles were transferred them from the triethylene glycol surface to glass substrate (4 mm × 6 mm) followed by the thermal annealing. The magnetoresistive devices were fabricated by thermal evaporation of the Au/Cr electrodes on glass substrates. The Au/Cr patterns of 7 µm channel length and 6 mm channel width, were fabricated by photolithography as shown in the middle panel of Figure 1. Then, the selfassembled of core/shell nanoparticles floating on the triethylene glycol surface were transferred to the prepatterned glass substrates, and the obtained films were thermally annealed.
The magnetotransport measurements were performed using a Keithley 4200 source-measure unit in a two probe configuration, with a maximum applied field of ±12 kOe. fitted with a Gaussian function, the mean particle sizes ( D ) were calculated, resulting 7.7 nm and 9.6 nm, for core and core/shell systems, respectively. From high resolution TEM (HRTEM) image it is noticed that the core is monocrystalline and the shell growth epitaxial over the core for most of the nanoparticles. Moreover different crystalline orientations for core and shell can be observed for most of the nanoparticles as noticed in Figure 2

III. RESULTS AND DISCUSSION
where the (044) and (222) crystalline planes of spinel phase are signaled for core and shell, respectively. The core/shell structure is confirmed by dark field, as shown in Figure 2 Table I. From these data, and assuming a Fe 3 O 4 core of 7.7 nm of diameter, we calculated the shell stoichiometry which shows a systematic evolution consistent with the nominal concentration.  The self assemblies of Fe 3 O 4 -core/Co 1−x Zn x Fe 2 O 4 -shell nanoparticles were obtained by the liquid-air interface process [20][21][22] as explained in the Experimental Procedure Section.
In order to perform the different measurements, the self-assembly and the subsequent thermal treatment was reproduced using a commercial silicon nitride support grid for TEM characterization, and a glass substrate patterned with two Au electrodes separated by ∼7 µm (as shown in the SEM image of Figure 1) for the magnetotransport studies. The topog-raphy of the annealed assemblies was analyzed by atomic force microscopy. Images acquired at different regions of the films reveal a large homogeneity with uniform and smooth surface, as observed in the bottom panel of Figure 1. From the AFM height profile cross section at the film boundary, an average film thickness of 20 nm was measured, which corresponds to two layers of nanoparticles.
Homogeneity and narrow size distribution are essential conditions to reach large area of self-organization; for core/shell nanoparticles, as observed from Figure 3 and Figure S1 Figure   S1(f) it is also noticed that the self-organized nanoparticles are separated by a gap of ∼ 1 nm.
As the thermogravimetric analysis indicates that approximately 7% of residual mass remains in the systems after the thermal treatment at 400 • C in vacuum atmosphere, and infrared spectroscopy measurements do not detect organic molecules (see Figure S3 in Supplemental Material [43]), we conclude that the gap between the nanoparticles is formed by amorphous carbon. The nanoparticles size distributions measured from the TEM micrographs, are shown in the middle panels of Figure 3. From the fitting of the histograms with a Gaussian function, the mean nanoparticles size D was calculated and summarized in Table II, which vary between 9.2-9.9 nm for all the systems. We also notice that the nanoparticle size and also the superstructure of the self-assembly is preserved at higher annealing temperature, however at 600 • C the nanoparticles start to coalesce (see Figure S2 Supplemental Material [43]). From the HRTEM images (shown in Figure 2 for x =0 and in Figure 3 for x =0.75 and x =1) it is observed that the core/shell microstructure is preserved after the annealing at 400 • C, where different interplanar distances and crystallographic planes orientation for the inner and outer part of the particle can be measured. As mentioned before, this morphology is confirmed by dark field images, as shown in Figure 3 for x=0.00, 0.25 and 0.50. Although the core/shell microstructure is preserved, we can not discard some degree of interdiffusion at the interface as reported for similar nanoparticles systems [44,45]; however, as we are going to discuss later, the magnetoresistance measurements confirms the half-metallic nature of the Fe 3 O 4 core.  where M , f and K are the magnetization, film thickness (or volume fraction), and magnetic anisotropy of the core (c) and shell (sh), respectively. [28,51,52] Instead, if the size of the soft magnetic phase is larger than δ crit exchange spring behavior is found, the magnetization reversal is nonuniform and lower coercivities are obtained. [48,49,51] For CoFe 2 O 4 the values reported for δ w span in the range of 13-20 nm, [28,53]   reports the parameters that characterize the magnetic properties of dispersed nanoparticles before the annealing process. Both systems, the dispersed nanoparticles and the annealed assemblies, present qualitative and quantitative similar behavior, with an enhancement of the effective magnetic anisotropy when the concentration of Co in the shell increases, which points out that the magnetic behavior is governed by the hard/soft rigid coupling magnetization inversion process, as analyzed previously. However, the annealed assemblies present an approximately 20% larger T B , H C and K ef f , probably due to an increase of the dipolar interaction and the improvement of the crystallinity in the annealed assemblies.
The magnetic measurements demonstrate that the effective magnetic anisotropy of the system can be controlled by adjusting the shell composition, without appreciably modifying the morphology and the overall magnetic saturation, as observed from the inset of Figure   4(b). Given that the anisotropy of the system is to a great extent responsible of the switching field of the TMR, devices made of self-assembled core/shell nanoparticles provide an ideal system for studying spin-dependent transport between magnetic nanoparticles. Although both materials at the core/shell structure are strongly exchange coupled and behave as a unique magnetic entity with an average magnetic anisotropy, the conductivity of each phase is different. While the Fe 3 O 4 core is half-metallic, the Co 1−x Zn x Fe 2 O 4 shell is a semiconductor. Therefore, in order to observe TMR properties it is crucial to have Fe 3 O 4 phase to provide the spin polarized transport, whereas the role of the shell is to modulate the switching field by tuning the magnetic anisotropy, while acting as a tunnel barrier.
The electronic transport in the annealed self-assembled devices was studied from the current-voltage (I − V ) measurements and from the temperature dependence of the resistivity ρ, as reported in Figure 5. From this figure it can be affirmed that the electron conduction in the devices is given by two independent mechanisms: thermally activated hopping, which is revealed from the temperature dependence of ρ; and the tunneling conduction manifested by the non-Ohmic behavior in the I-V curve with the characteristic V 3 dependence at low temperature, as discussed next. From the Simmons's model, which considers inelastic tunneling across an insulating barrier, the I − V curves can be quantitatively linked to the physical parameters of the system, i.e. the tunnel barrier height (h) and width (w), the effective contact area, etc. [57]. This model also considers linear dependence approximation of the barrier potential profile with V and w. For voltage smaller than the potential barrier, the Simmons's model can be approximated with the well known polynomial relationship: [58] where G 0 is the equilibrium conductance and F ∝ ξ 2 , where ξ = w/h is the shape factor of the barrier. [59] From the plot shown in the inset of Figure 5 measured by SEM and the film thickness obtained by AFM microscopy, we have plotted ρ(T ) in Figure 5(b), which suggests that a thermally activated transport mechanism is also involved in the conduction of the devices. The temperature dependence follows the relation ln(ρ) ∝ T −1/α , where α = 0.4(1) was found as the best fitted parameter for all the systems.
This value is close to the dependence found in the Efros's variable range hopping model ln(ρ) ∝ T −1/2 , [60] and it is consistent with the behavior measured in other nanoparticles arrays. [14,17,21,24,61]. Moreover, ρ(T ) in the core/shell assemblies is, at least, two orders of magnitude smaller than the reported for pellets of Co  [62][63][64]). This result shows that the tunnel conduction strongly increases the conductivity of the core/shell assemblies compared to the values measured in the semiconductor shell material. Notice that the larger the tunneling current, the lower the resistivity of the devices. Therefore, the material that has the lower tunneling barrier will be the one that is most influenced by this contribution. Consistently, from Figure 5 we have measured that the assemblies of nanoparticles with a ZnFe 2 O 4 shell present the lowest energy barrier, and a systematic increase with decreasing x is observed.
The main findings of this work are summarized in Figure 6 and 7. Magnetoresistance curves measured at different temperatures for the five devices studied in this work are shown in Figure 6 and Figure S7 of Supplemental Material [43]. At a given temperature, the switching field of the magnetoresistance curve monotonously decreases when the Zn concentration of the shell increases. Moreover, as shown in Figure 7, each sample shows a smooth decrease of the switching field with temperature, consistently with the temperature evolution of the coercive field. Also, a saturation behavior of TMR at high magnetic field is observed when x decreases, in good correspondence with the M(H) curves, see Figure 4.
For spin-polarized intergrain tunneling the TMR is related to the macroscopic magnetization. [9,10,25] For a ferromagnetic-insulator granular system, the electron tunneling across the insulating barrier was calculated including an additional exchange energy arising from the interaction between the tunneling electron spin and the non-parallel magnetic moment of the neighboring grains. [25,65]. Assuming that the exchange energy can be expressed in terms of the spin correlation function of two ferromagnetic neighboring grain, the magnetoresistance can be expressed as: where J accounts for the magnetic correlations when the electron tunnels through the in- to the calculations for the intergrain tunneling transport of manganese perovskites [66].
The TMR amplitude, which is in the 0.3 − 1.5% range depending on the composition, is similar to other reported values for self assembled nanoparticles, [16,21,24] however, it is much smaller than the calculated from the Julliere's model: T M R = 2P 2 (1−P 2 ) on the basis of the spin polarization values of magnetite. This reduction may be due to the fact that the tunneling probability decreases exponentially with the barrier width. According to Refs. [67,68] for TMR multilayers, the optimal barrier thickness is in the 1 − 1.5 nm range; however, the tunnel current between Fe 3 O 4 cores in the self-assembled structure must pass through the insulator barrier of ∼ 3 nm width, which is composed by the cobalt ferrite shell and the amorphous carbon nanoparticles coating. On the other hand, although it is known that the TMR diminishes with increasing bias voltage, [53] the high resistance of the magnetic nanoparticles devices determines the experimental parameters, and high voltage bias ∼ 100 V had to be applied to perform the transport measurements at low temperature. These factors make evident the importance of optimizing the different stages of the fabrication process in order to increase the conductivity to produce large amplitude and low switching field TMR devices based on core/shell magnetic nanoparticles. However, irrespective of the absolute value of TMR, this study demonstrates that the switching field of TMR can be tailored at will in self-assemblies of exchange coupled core-shell nanoparticles, synthesized by an affordable chemical route.

IV. CONCLUSION
In summary, we have fabricated self-assemblies of core/shell nanoparticles with controlled TMR. Particularly, we have shown that the magnetic properties can be finely tuned by changing the shell composition which provides a tool to adjust the TMR switching field.
We shows whereas that the Fe 3 O 4 core provides the spin polarized transport, the shell acts as tunnel barrier in the self assembly and also modulates the switching field by tuning the magnetic anisotropy through the interface exchange coupling. This approach shows the feasibility to use assemblies of exchange coupled magnetic nanoparticles in TMR devices, where different combinations of soft/hard and hard/soft core/shell configurations can be envisaged. In this way, combinations of materials can be carefully designed to move across the rigid coupling to exchange bias regime, to design devices with tailored magnetotransport response, which gives a promising base for the design of core/shell nanoparticles based devices for fundamental studies or for spintronic applications.