Journal of Magnetism and Magnetic Materials 262 (2003) 56–59
Magnetotransport in core-shell Fe–Fe oxide nanostructures L. Savinia,*, E. Bonettia, L. Del Biancoa, L. Pasquinia, L. Signorinia, M. Coissonb, V. Selvagginic a
Department of Physics, University of Bologna and INFM, Via C. Berti Pichat 6/2, Bologna I-40127, Italy b DISPEA, Politecnico di Torino and INFM, C.so Duca degli Abruzzi 24, Torino I-10129, Italy c Department of Physics, Politecnico di Torino and INFM, C.so Duca degli Abruzzi 24, Torino I-10129, Italy
Abstract Magnetic and magnetotransport measurements were performed on gas-phase synthesized Fe nanoparticles subjected to surface oxidation and cold consolidation. Two samples were investigated with a-Fe volume fraction of 0.15 and 0.60. The sample with smaller metallic fraction is below the percolation threshold for metallic conduction and the conduction mechanism is dominated by thermally activated processes across the oxide. In this case, by lowering the temperature, an increase of the negative magnetoresistance is observed up to 5% at 50 K in a magnetic field of 70 kOe. The magnetoresistance dependence on the sample magnetization, temperature and sample composition is discussed considering the magnetic correlations present in these nanostrucuterd systems. r 2003 Elsevier Science B.V. All rights reserved. PACS: 75.47.m; 61.46.+w Keywords: Magnetoresistance; Nanoparticles; Iron; Iron oxide
Nanostructured magnetic systems are generally characterized by the coexistence of two or more phases on the length scale of few nanometers [1]. Interesting examples are the core-shell Fe nanoparticles constituted by two different magnetic phases: the a-Fe cores and the surrounding oxide layer [2,3]. In fact, it has been observed that the oxide shell strongly affects the magnetic behavior through size and interface effects. In this paper, we report the magnetic, electrical and magnetotransport properties of the iron–iron oxide nanocomposite obtained by cold-consolidation of passivated iron nanoparticles. Fe–Fe oxide *Corresponding author. Tel.: +39-051-209-5298; fax: +39051-209-5153. E-mail address:
[email protected] (L. Savini).
core-shell nanoparticles were prepared by inert gas condensation and consequent oxygen passivation. The nanoparticle powders were pressed in situ under an uniaxial pressure of 1.5 GPa to obtain pellets. The sample microstructure was investigated by X-ray diffraction (XRD) using a Philips PW1710 diffractometer equipped with Cu Ka radiation and graphite monochromator in the diffracted beam. Magnetization measurements were performed between –70 and +70 kOe in the temperature interval 5–300 K by means of an extraction magnetometer. The magneto-resistance was measured in the parallel configuration in the same temperature interval by the standard fourprobe technique. Two samples were obtained at different evaporation temperature to vary the particle mean
0304-8853/03/$ - see front matter r 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0304-8853(03)00018-0
L. Savini et al. / Journal of Magnetism and Magnetic Materials 262 (2003) 56–59 200
Fe 110
T= 5 K T= 283 K
D8
Magnetization (emu/g)
Fe 220
Fe 211
440 Fe 200
150 333, 511
400
311
220
Intensity (a.u.)
57
D18
100 50
D8
0 -50 -100
D18 40
60 80 2θ (degrees)
100
Fig. 1. XRD patterns of samples D8 and D18. The peak positions of the a-Fe and of the spinel oxide phases are indicated.
size. The XRD patterns of the two samples are reported in Fig. 1. The angular peak positions of the oxide phase are consistent with either magnetite (Fe3O4) or maghemite (g-Fe2O3). The peak broadening prevents from distinguishing between these two structures. The Rietveld analysis [4] of the XRD data showed a volume-weighted crystallite size of the oxide phase of B2 nm for both samples. The small oxide crystallite size indicates that a high fraction of oxide atoms are subjected to topological disorder and the oxide may be considered a disordered spinel. The volumeweighted crystallite size of the iron phase DFe is 8 and 18 nm for the two samples that will be indicated respectively as D8 and D18 in the following. Furthermore the a-Fe volume fraction, corresponding to the particle cores, is 0.15 and 0.60 in samples D8 and D18, respectively, in agreement with the values obtainable for spherical particles with a 2 nm thick oxide layer and with direct TEM observations [5]. The magnetization curves of both samples are depicted in Fig. 2. No saturation is attained up to 770 kOe, the nonsaturating trend being more evident in sample D8. The extrapolated values of the saturation magnetization Ms slightly increase with reducing temperature. At T=50 K, Ms is 68 and 175 emu/g in samples D8 and D18, respectively. These values are smaller than what are expected for the corresponding mixtures of a-Fe and spinel oxide. These results are consistent with
-150 -200
-60
-40
-20 0 20 Magnetic Field ( kOe )
40
60
Fig. 2. Magnetization vs. applied magnetic field of D8 and D18 samples at T=5 and 283 K.
a noncollinear arrangement of the atomic spins at the oxide interface and with the higher oxide/a-Fe ratio of sample D8. Recent investigations on this kind of systems show that ferromagnetic single domain Fe cores are surrounded by magnetically disordered oxide shells, which display a spin-glasslike behavior at low temperature as a result of topological disorder and frustration of magnetic interactions [3,5]. At high temperature, the moments of the oxide phase become able to fluctuate and tend to be polarized by the Fe cores. The magnetization change below 10 kOe is primarily ascribed to the alignment of the M vectors of ferromagnetic cores, while the high-field approach to saturation involves the ordering of spin-glasslike shells. The electrical resistance and its temperature dependence were observed to be strongly affected by the composition. The resistance of the sample D8 increases with reducing temperature according to the empirical law R=R0exp(A/T)p [6], with A=30 K and p=0.7, while the resistance of D18 increases by only a factor 2 in the same temperature interval. This temperature trend indicates that the conduction mechanism across the oxide is a thermally activated process, and further it is the dominant conduction channel in sample D8 where the a-Fe content is below the percolation threshold for electrical conductivity. In sample D18 the oxide conduction channel, in parallel with
L. Savini et al. / Journal of Magnetism and Magnetic Materials 262 (2003) 56–59
58
0
0 -1
MR × 100
4
-4
D18 T = 50 K T = 282 K
-2 -3
0 0
0
100
10
200 T (K)
300
20
30
-4
D18 T = 50 K T = 282 K
40 H (kOe)
50
60
70
Fig. 3. Magnetic field dependence of magnetoresistance at T=50 and 282 K for D8 and D18 samples. Inset: temperature dependence of MRMAX.
a metallic channel due to direct contact between percolating cores [5], is progressively activated by increasing the temperature and the resistance decreases moderately. A negative magnetoresistance (MR) was observed in both samples. Owing to the increase of the electrical resistance of sample D8, MR data could be actually obtained only in the interval 50 K–RT. The MR curves of the two samples at T=50 and 282 K are shown in Fig. 3, where the MR is defined as [R(H)R(0)]/R(0). No evident trend towards saturation is observed. The maximum absolute value of MR (MRMAX) is reported as a function of temperature in the inset of Fig. 3; MRMAX exhibits opposite trend in the two samples. When the measured MR is plotted as a function of the reduced sample magnetization m=M/Ms, a boxlike shape is obtained in all cases, as shown in the inset of Fig. 4 for sample D8 at T=50 K. MR vs. m is expected to exhibit a parabolic shape for granular systems containing uncorrelated magnetic scatterers (such as superparamagnetic particles [7] or paramagnetic clusters [8]). We recall that in sample D8 the electrical transport occurs via thermally activated processes across the oxide shells. The transmission probability of an electron across a nonmetallic layer is affected by the magnetic states of the regions separated by the layer itself, or constituting the layer [9]. The
0 -2
2
D8
-6
D8
6
MRMAX
MR × 100
-2
-4 -6
-5 0.00
-1.0 -0.5 0.0 0.5 1.0
m 0.25
0.50
0.75
1.00
m2
Fig. 4. Magnetoeresistance plotted as a function of the reduced magnetization m2 at T=50 and 282 K for D8 and D18 samples. Inset: MR vs. m at T= 50 K for the D8 sample.
resistivity depends on the term cosyij ; where yij is the angle of tilt between magnetization vectors outside the nonmetallic layer (if nonmagnetic) or within the layer itself (if magnetic). In the present case, the oxide shells play the role of the non metallic layer. There may be a nonzero angle of tilt either between the magnetization vectors of two particle cores, or between those of the two shells in contact. The boxlike shape of the MR vs. m curves can be ascribed to the occurrence of two magnetization processes: the first one (named M1) is characterized by a large change of M without a significant MR; the second one is responsible for most of the observed MR. The M1 process occurs at low applied magnetic field and, on the basis of microstructure and magnetization results, it can be mainly related to the core magnetization, while the second one originates with the alignment of M in the oxide. In a first approximation, the total magnetization is pictured as the sum of these two processes, occurring in different magnetic field ranges. The M1 and M2 processes have been disentangled exploiting the strong change in the slope of the m vs. H curves: a parameter m1max was defined by the relation dH/dm(m1max)=1.5 kOe. Fig. 4 displays the MR vs. m2 curves, where m2 is defined as m2 = (mm1max)/(1m1max) and is therefore representative of the shell magnetization
L. Savini et al. / Journal of Magnetism and Magnetic Materials 262 (2003) 56–59
alone. With this representation, the curves for sample D8 become similar to the flat-top parabolas observed in granular bimetallic materials containing magnetically correlated scatterers [10]. Further, the curves tend to be more parabolic on lowering the temperature. The above observations support the following picture for sample D8: since the conduction is dominated by thermally activated processes at the oxide interface between two metallic cores, the initial alignment of the cores magnetization induces the M1 process but almost negligible MR. Higher fields tend to align the M vectors in the oxide interface as well, thereby increasing the transmission probability and giving rise to MR. The nearly parabolic shape observed at low temperature can be explained by the occurrence of a spin-glass-like arrangement of the shell magnetization, induced by competing magnetic
interactions [5]: in fact, in this case the cosyij term can be factorized as hcosyi i cosyj Em22 ; where yi is the angle between the i-shell vector and the field axis. Conversely, at high temperatures, the increasing influence of the core magnetization brings about statistical correlations between yi and yj leading to a flat-top parabola [10]. Following the same arguments, the increase of MRMAX with lowering T in sample D8 (inset of Fig. 3) can be attributed to an increasingly disordered initial pattern of the M vectors in the shells. At low temperatures, after completion of the M1 process, the shell magnetization is still highly disordered (owing to the emerging spin-glass-like character); when the temperature is raised, it is increasingly influenced by the aligned core magnetization, so
59
that the initial pattern is less random, and MRMAX is reduced. The different MRMAX vs. T behavior observed for sample D18 can be understood considering that a metallic conduction channel is also active because of the direct metallic contact between the cores. However, we have seen that the MR effect is dominated by high-field magnetization processes in the oxide phase, where the electrical transport is thermally activated: therefore the temperature rising is expected to enhance the MR by virtue of the increasing importance of the nonmetallic conduction channel. Further investigations are needed to confirm this last hypothesis. The present research work has been supported by INFM–PRA ELTMAG.
References [1] A. Hernando, J. Phys.: Condens. Matter 11 (1999) 9455. [2] S. Gangopadhyay, G.C. Hadjipanayis, S.I. Shah, V. Papaefthymiou, A. Kostikas, J. Appl. Phys. 70 (1991) 5888. [3] L. Del Bianco, A. Hernando, M. Multigner, C. Prados, J.C. Sanchez-Lopez, A. Fernandez, C.F. Conde, A. Conde, J. Appl. Phys. 84 (1998) 2189. [4] L. Lutterotti, P. Scardi, J. Appl. Crystallogr. 23 (1990) 246. [5] L. Del Bianco, D. Fiorani, A. Testa, E. Bonetti, L. Savini, S. Signoretti, Phys. Rev. B 66 (2002) 174418. [6] P. Sheng, Phys. Rev. B 21 (1980) 2180. [7] C.L. Chien, J.Q. Xiao, J.S. Jiang, J. Appl. Phys. 73 (1993) 5309. [8] P. Allia, M. Coisson, V. Selvaggini, P. Tiberto, F. Vinai, Phys. Rev. B 63 (2001) 180404R. [9] J. Inoue, S. Maekawa, Phys. Rev. B 53 (1996) R11927. [10] P. Allia, M. Knobel, P. Tiberto, F. Vinai, Phys. Rev. B 52 (1995) 15398.
