LETTER pubs.acs.org/NanoLett
Room-Temperature Tunnel Magnetoresistance in Self-Assembled Chemically Synthesized Metallic Iron Nanoparticles Julien Dugay, Reasmey P. Tan,* Anca Meffre, Thomas Blon, Lise-Marie Lacroix, Julian Carrey,* Pier F. Fazzini, Sebastien Lachaize, Bruno Chaudret, and Marc Respaud Universite de Toulouse, INSA, UPS, LPCNO, 135, avenue de Rangueil, F-31077 Toulouse, France CNRS; LPCNO, F-31077 Toulouse, France
bS Supporting Information ABSTRACT: We report on room temperature magnetoresistance in networks of chemically synthesized metallic Fe nanoparticles surrounded by two types of organic barriers. Electrical properties, featuring Coulomb blockade, and magnetotransport measurements show that this magnetoresistance arises from spin-dependent tunnelling, so the organic ligands stabilizing the nanoparticles are efficient spin-conservative tunnel barrier. These results demonstrate the feasibility of an all-chemistry approach for room temperature spintronics. KEYWORDS: Spintronics, nanoparticles, magnetic materials and devices, organicinorganic nanostructures, magnetoresistance, Coulomb blockade
T
he simplest spintronic device based on chemically synthesized magnetic nanoparticles (MNPs) consists of a self-assembly of MNPs deposited between two microscopic contacts. In these systems, the stabilizing organic ligands surrounding the MNPs are expected to act as a tunnel barrier, so that tunnel magnetoresistance (TMR) is expected.1 An important issue is whether the organic ligands surrounding the MNPs could be an efficient tunnel barrier up to room temperature. So far, the large majority of studies have focused on Fe3O4 NPs due to the relative ease of their synthesis and handling.25 However, the transport in such systems is dominated by either intraparticles or surface spindisorder effects.2,3,5 Even when TMR has convincingly been observed, some doubts remain on whether the tunnel barrier is related to the organic ligands surrounding the MNPs or to grain boundaries.4 The only way to discard these doubts is to study assemblies of pure metallic nonoxidized MNPs made of Co, Fe, Ni, or their alloys and stabilized by an organic shell. So far, the best results have been reported on nanogranular materials elaborated by a physical process (coevaporation) and composed of metallic ferromagnetic grains dispersed in a matrix. New hybrid systems where the particles were embedded in organic semiconductors such as fullerene C60,6 rubrene,7 or Alq38 exhibited MR values in the range of ∼10102 % at low temperature. However, in the best cases MR always decreased down to 0.1% at room temperature.68 This strong decay of MR as a function of T is probably due to spin flip process enhanced by transferred charges at localized states within the organic barrier. In other words, the conduction in the high temperature regime of these types of barriers seems to be dominated by other mechanisms than direct tunneling. Such a phenomenon has been recently addressed in r 2011 American Chemical Society
details in the case of magnetic tunnel junctions with Alq3 organic barriers.9 Studies on systems of metallic MNPs entirely synthesized by chemistry are rather scarce. Since the first observation of TMR on Co NPs by Black et al. in 2000,1 the only other metallic systems studied (large CoFe NPs superlattices) have evidenced exotic MR properties but no clear spin-dependent tunnelling.1013 In all these experiments, MR disappeared above 20 K. Note that Black et al. thermally annealed the deposited Co nanocrystals at 400 C under reducing atmosphere to get measurable resistance.1 This procedure, which results in the shrinkage of the assembly, is likely to drastically modify the surrounding organic ligands and damage the tunnel barrier. Here, we report the observation of room-temperature TMR in assemblies of fully metallic iron MNPs synthesized by organometallic chemistry. Two systems of iron NPs stabilized by different organic ligand mixtures are studied. These NPs have been obtained by adapting the general procedures defined by Dumestre et al.14 and Lacroix et al.15 In both cases, the decomposition of {Fe[N(SiMe3)2]2}2 is performed under H2 pressure in the presence of amine and acid surfactants to ensure the stabilization of metallic NPs. Details of the synthesis procedures are given in Supporting Information. The collected NPs were washed three times with toluene in order to purify them from the large excess of ligands and any remaining Fe molecular species. When dried, it yielded black materials containing 70% of iron as determined by Received: June 24, 2011 Accepted: November 11, 2011 Published: November 18, 2011 5128
dx.doi.org/10.1021/nl203284v | Nano Lett. 2011, 11, 5128–5134
Nano Letters
LETTER
Figure 1. (a) TEM image of the cubic Fe nanoparticles (sample I). The scale bar is 200 nm. Insert: size distribution of the NPs fitted by a log-normal distribution. (b) TEM image of the spherical Fe NPs (sample II). The scale bar is 100 nm. Insert: size distribution of the NPs fitted by a log-normal distribution.
Figure 2. (a) Sketch of dielectrophoresis process used to trap the NPs between gold electrodes. (b) Cross section micrograph of sample I (Fe nanocubes). A thick film (100 nm) of alumina is sputtered to prevent oxidation. Note that the shape of the Fe nanocubes has been damaged as a result of the FIB preparation. The scale bar is 100 nm. (c,d) Cross section of sample II at two magnifications. The scale bars are 1 μm and 100 nm for (c) and (d), respectively. (e) Energy dispersive X-ray spectroscopy analysis of sample II. The scale bar is 200 nm. The top Pt layer has been added after the transport measurements and was required for FIB thinning of the samples (see Supporting Information).
chemical microanalysis. All syntheses and manipulations were performed under either an inert Ar atmosphere or a reductive dihydrogen atmosphere, using a glovebox and Fisher-Porter bottles, respectively. These precautions allowed us to prevent any oxidation of Fe and preserve the metallic surface character. Among the different systems, we studied cubic NPs surrounded by hexadecylamine [CH3(CH2)15NH2]/hexadecylammonium [CH3(CH2)15NH3+Cl] (sample I) and spherical NPs surrounded by a mixture of hexadecylamine/palmitic acid [CH3(CH2)14COOH] (sample II). The metallic character of the Fe NPs was confirmed by M€ossbauer spectroscopy, and no trace of
either remaining paramagnetic precursors or Fe oxides has been evidenced in the sensitivity limit of the technique (see Supporting Information). As evidenced by TEM, the resulting cubic (spherical) NPs display a mean size a = 8.9 (11.1) nm and a lognormal standard deviation σ = 0.08 (0.09) (see Figure 1). In both systems of NPs, the mean interparticle distance is in the range of 2 nm (see Figure 1 and Supporting Information). We used dielectrophoresis16 to trap the NPs between fingerlike electrodes displaying a 5 μm gap (see Figure 2). To prevent the oxidation of the NPs, this deposit was made in a glovebox under an inert atmosphere. Then, the devices were capped with 5129
dx.doi.org/10.1021/nl203284v |Nano Lett. 2011, 11, 5128–5134
Nano Letters
LETTER
Figure 3. (a) Magnetization of sample I. Data from the as-grown powder and from a sample where the NPs were trapped by dielectrophoresis between electrodes are shown. (b) Magnetization measurements performed on a powder of sample II.
Figure 4. Currentvoltage [I(V)] characteristics measured as a function of temperature for (a) sample I (c) sample II. The circles represent a fit using eq 1 with ζ = 3.5 (2.0) and VT = 1.5 (4.4) V for sample I (II). (b,d) Corresponding differential resistivity obtained from I(V) curves for (b) sample I (d) sample II.
amorphous alumina layers (100 or 200 nm) using a RF sputtering setup coupled to the glovebox, which avoids any subsequent oxidation. Figure 2 displays representative transmission electronic microscopy (TEM) cross sections of selected regions thinned by focused ion beam (FIB). We emphasize that they have been obtained after the samples have been measured. The TEM pictures evidence thick layers of disordered assemblies of Fe NPs separated by the organic ligands. Note that the original cubic shape and the crystalline structure of the NPs have been damaged during the FIB thinning process due to the high energy of Ga ions. The energy dispersive X-ray analysis (EDX) confirms the nature of each layer and the absence of alumina diffusion inside the NPs layer. Magnetic properties measured on a powder of NPs give a saturation magnetization MS = 230 ( 23 Am2 kg1 (220 Am2 kg1) and a small coercive field μ0HC = 23 mT (4 mT) at T = 2 K (300 K) for sample I, and MS = 240 ( 24 Am2 kg1 (229 Am2 kg1) and μ0HC = 36 mT (18 mT) at T = 2 K (300 K) for sample II (see Figure 3). The MNPs are not fully saturated at large magnetic field and a linear increase of the magnetization (0.5%/T)
is evidenced above 2 T. Figure 3a displays the magnetization measurements with the magnetic field applied in the plane of a typical device elaborated with Fe nanocubes. In comparison with the powder sample, the assembly of NPs displays a higher remanence and coercivity at 2 K (μ0HC = 44 mT) and a superparamagnetic behavior at 300 K. The increase of the coercivity is a well-known consequence of magnetic interaction reduction in such a deposit compared to a compact powder.17 No exchange bias features are observed in the hysteresis loop measured at 2 K after a field cooling performed at 5 T from 300 K. This confirms the absence of any significant oxidation of the Fe NPs when elaborating the devices. Therefore, the small slope of the magnetization at large magnetic field is attributed to the presence of spin canting at the surface of the NPs. Magnetotransport measurements were performed on 32 devices. Among them, 21 displayed measurable resistances (R) ranging from 5 to 6 MΩ at room temperature, while 13 over the 21 devices exhibited MR properties. The 11 devices that displayed no detectable current have probably a resistance higher than our range of measurement or do not display any percolation paths. 5130
dx.doi.org/10.1021/nl203284v |Nano Lett. 2011, 11, 5128–5134
Nano Letters
LETTER
Figure 5. (a) Magnetoresistance of Fe nanocubes at T = 4.4 K. Insert: low field region. The black line and the red line indicate the sweep up and back of the applied magnetic field. (b) Temperature dependence of R(H) characteristics for V = 10 V. (c) Magnetoresistance of Fe nanocubes at room temperature with the magnetic field applied in and out of the plane of the device (black and red line respectively). (d) Temperature dependence of the tunnel magnetoresistance of Fe nanocubes (solid squares and circles), coercive field deduced from the magnetoresistance curves (open diamonds) and from magnetization measurements (open down triangles).
The 13 magnetoresistive devices qualitatively displayed a similar behavior. In this Letter, results from two unique devices, one for each type of organic barrier mixtures, will be shown. Complementary specific studies presented in the Supporting Information originate from other devices. Electrical measurements are presented in Figure 4. For both samples, the resistivity increases strongly at low temperature and low bias voltage, showing a suppression of current typical of Coulomb blockade in NPs assemblies (see Figure 4b,d). This is also depicted in the currentvoltage characteristics [I(V)] as a function of temperature shown in Figure 4a,c; the I(V) curves change from an ohmic behavior at room temperature to a highly nonlinear one at low temperature with a progressive opening of the gap. Numerical studies predict that I(V) characteristics follow at T = 0 K ðV VT Þ ζ ð1Þ I VT where VT is the threshold voltage above which conduction occurs.18 This power-law is related to the progressive opening of conduction channels, and ζ is indicative of the dimensionality of the conduction paths across the assembly. ζ = 1 and ζ = 5/3 have been calculated and further confirmed experimentally for one- and two-dimensional (1D and 2D) disordered assemblies of NPs respectively.19 In the case of 3D assemblies, only experimental estimations have been reported with values larger than for the 2D case.13 Here, we found ζ = 3.5 ( 0.2 (2.0 ( 0.2) and VT = 1.5 and 4.4 V for sample I and II, respectively. These values of ζ are in the same range than the ones reported in multilayers of NPs.1 The basic MR behaviors are the same for both ligands mixtures. Figures 5 and 6 display the magnetic field dependence of the resistivity R(H) for sample I and II, respectively. The external
magnetic field is applied in the sample plane. The MR curves evidence two contributions. At high magnetic field, the first one is characterized by a linear decrease of R. The associated MR is weakly dependent on temperature (see Supporting Information). More interestingly, the second contribution occurs at low applied magnetic field
