G.
Muscas
*ab,
G.
Concas
c,
S.
Laureti
a,
A. M.
Testa
a,
R.
Mathieu
d,
J. A.
De Toro
e,
C.
Cannas
f,
A.
Musinu
f,
M. A.
Novak
g,
C.
Sangregorio
h,
Su Seong
Lee
i and
D.
Peddis
*a
aIstituto di Struttura della Materia – CNR, 00016 Monterotondo Scalo (RM), Italy. E-mail: davide.peddis@cnr.it
bDepartment of Physics and Astronomy, Uppsala University, Box 516, SE-751 20 Uppsala, Sweden. E-mail: giuseppe.muscas@physics.uu.se
cDipartimento di Fisica, Università di Cagliari, S. P. Monserrato-Sestu km 0.700, 09042 Monserrato, Italy
dDepartment of Engineering Sciences, Uppsala University, Box 534, SE-751 21 Uppsala, Sweden
eInstituto Regional de Investigación Científica Aplicada (IRICA), Departamento de Fısica Aplicada, Universidad de Castilla-La Mancha – 13071 Ciudad Real, Spain
fDipartimento di Scienze Chimiche e Geologiche, INSTM, Università di Cagliari, S. P. Monserrato-Sestu km 0.700, 09042 Monserrato, Italy
gInstituto de Fisica, Universidade Federal do Rio de Janeiro, Rio de Janeiro RJ 21945-970, Brazil
hIstituto di Chimica dei Composti Organo-Metallici, Consiglio Nazionale delle Ricerche and INSTM via Madonna del Piano 10, I-50019 Sesto Fiorentino, Italy
iInstitute of Bioengineering and Nanotechnology, 31 Biopolis Way, The Nanos, Singapore 138669, Singapore
First published on 8th November 2018
This paper aims to analyze the competition of single particle anisotropy and interparticle interactions in nanoparticle ensembles using a random anisotropy model. The model is first applied to ideal systems of non-interacting and strongly dipolar interacting ensembles of maghemite nanoparticles. The investigation is then extended to more complex systems of pure cobalt ferrite CoFe2O4 (CFO) and mixed cobalt–nickel ferrite (Co,Ni)Fe2O4 (CNFO) nanoparticles. Both samples were synthetized by the polyol process and exhibit the same particle size (DTEM ≈ 5 nm), but with different interparticle interaction strengths and single particle anisotropy. The implementation of the random anisotropy model allows investigation of the influence of single particle anisotropy and interparticle interactions, and sheds light on their complex interplay as well as on their individual contribution. This analysis is of fundamental importance in order to understand the physics of these systems and to develop technological applications based on concentrated magnetic nanoparticles, where single and collective behaviors coexist.
E = KaVpsin2ϑ − MSVpμ0Hcos(φ − ϑ) | (1) |
The magnetic properties of ferromagnetic materials are strongly connected to their crystalline structures, which determine their magneto-crystalline anisotropy and their symmetry axis. For typical (non-textured) nanoparticle ensembles and polycrystalline continuous media, the easy axes of the constituent grains/particles are randomly oriented. In polycrystalline materials, the exchange interactions force the spins to align over a correlation length (Lcorr), which can be larger than the grain size (D). Therefore, the effective anisotropy for the correlated volume will be the average of the grains within such a volume, resulting in a reduced magnitude with respect to the anisotropy of the individual grain.30 Within this context, the Random Anisotropy Model (RAM) has been employed to describe systems where the exchange interactions dominate, extending the correlation volume beyond the structural correlation length, i.e., the grain size D. The RAM has been extensively used for describing amorphous materials31 since the 1970s, and more recently it has been also used for nano-crystalline metal alloys.32 In addition, a modified RAM has been proposed to explain the properties of Cox(SiO2)1−x nanogranular films,33,34 suggesting that, with opportune modifications, this phenomenological model can be generally applied regardless of the nature of the interactions, exchange or dipolar, between particles.30 Indeed, the same model was able to describe the field dependence of the blocking temperature of 2D arrays of iron oxide nanoparticles deposited as monolayers.35
In the present paper, we propose the application of the RAM to analyze the magnetic properties of strongly interacting nanoparticle ensembles, extending the model for the first time to the pure dipolar interaction case. The effective anisotropy energy of such systems was deeply investigated, identifying quantitatively the individual contribution of the single particle anisotropy and interparticle interaction energies. Specific nanoparticle systems were selected to test the RAM. In the first stage, the influence of interparticle interactions in ensembles of maghemite nanoparticles was analyzed by comparing the magnetic properties of the nanoparticles isolated by a thick silica coating with a packed ensemble of the same particles. Then, the investigation was extended to more complex ensembles of (Co,Ni)Fe2O4 nanoparticles, with the same size but with different interaction strength and single particle anisotropy.
(2) |
(3) |
L corr defines a correlation volume (VN), where the effective anisotropy constant Keff results from the averaging over the number of particles (N) within such a volume,38 as described by the sketch in Fig. 1, as reported in ref. 30:
(4) |
The number of correlated particles N can be defined as the ratio between the correlation volume and the volume of a single particle, taking into account the volume fraction x that the particles occupy in the ensemble:
(5) |
Therefore, we can define the volume of the cluster VN as the effective volume of the magnetic material interacting within the correlation length Lcorr:
(6) |
As the interparticle interactions increase, the correlation length expands and the anisotropy is mediated within a larger number of particles, thus reducing its effective magnitude. On the other hand, the action of an external magnetic field reduces the correlation length32 and prevents the “cluster” formation. It is worth mentioning that the expressions (4) and (6) tend to the anisotropy and volume of an individual particle when the interactions are very weak and Lcorr → D.
For a non-interacting system, the dependence of the superparamagnetic blocking temperature from the field μ0H is described by the law:34,35
(7) |
(8) |
(9) |
Finally, eqn (7) can be re-written by substituting Ka, Vp and HK with the effective values for the clusters Keff, VN and HNK, respectively:
(10) |
Note that another expression for T*(H) has been recently considered.39,40 The experimental values of T*(H) can be extracted from the temperature dependence of the difference between field-cooled and zero-field-cooled curves (MFC–MZFC). For non-interacting particles, an estimate of the distribution of the anisotropy energy barrier can be obtained from:
(11) |
Because of interparticle interactions, eqn (11) gives only a rough indication of the distribution of anisotropy energies. As an example, the (MFC–MZFC) curve and its negative derivative are reported in ESI† in Fig. S1 for sample S17. Within the Néel model, the blocking temperature TB can be considered as the temperature for which the relaxation time is equal to the measuring time of the experimental technique. In a real system of nanoparticles, where a finite size distribution always exists, TB is often defined as the temperature at which 50% of the sample is in the superparamagnetic state.41,42 Since the blocking temperature of a single particle is proportional to its anisotropy barrier, an estimate of TB, i.e., T*, can be obtained as the temperature at which the integral of d(MZFC–MFC)/dT reaches 50% of its maximum value, i.e., the temperature at which (MFC–MZFC) is reduced to half of its value at the lowest temperature.
The experimental T*(H) values can be fitted according to eqn (10). By combining eqn (3)–(6), and (8), one needs as input parameters the experimental value of MS, the particle size D and the volume fraction x. The only free parameters of the fit are the effective interaction strength Aeff, the parameter C and the intrinsic anisotropy constant of the magnetic material Ka. Finally, by using such values in eqn (3)–(5), and (6), further information about the cluster and the individual particles can be obtained.
At small cooling fields, for the non-interacting sample S17, TB ≈ T*, which corresponds to a blocking temperature ≈22 K. On the other hand, sample RCP8 is a strongly interacting sample with T* ≈ 77 K. Due to such strong interparticle interactions, responsible of a collective behavior, TB ≈ T* is not more valid, thus it is not possible to get the information about the blocking temperature and the anisotropy energy of individual particles of this sample. Further analysis was carried out for both S17 and RCP8 samples, measuring (MFC–MZFC) with different cooling fields. The values of T* vs. the applied field were fitted according to eqn (10), as shown in Fig. 2 with data summarized in Table 1. For the isolated (i.e., non-interacting) particles in S17, De Toro et al. measured a value of Ka ≈ 30 × 103 J m−3 by means of AC susceptibility data.25 The Ka value extracted from the RAM fitting is 34(5) × 103 J m−3, in good agreement with the AC results. Furthermore, the same procedure applied to RCP8 allowed the estimation of the intrinsic anisotropy constant of the single non-interacting particles as 41(5) × 103 J m−3, in agreement, within the experimental error, with the previous values (Table 1). This confirms that the RAM analysis allows determination of the correct value of the anisotropy constant even in a strongly interacting system, where the blocking temperature of the individual NPs is hidden by the collective behavior. Note that demagnetization effects influence the MZFC and MFC curves for RCP8.45 Anyway, we have verified that this does not affect the T* determination beyond the experimental error. Furthermore, the single particle TB can be calculated from (10) for the case of μ0H = 0 T, leading for sample RCP8 to an intrinsic single particle value of TB = 31(4) K, compatible with the experimental data of 22(4) K of S17 for isolated-like particles.
Sample | K a (J m−3) | L corr (nm) | T B (K) | M S (A m−1) |
---|---|---|---|---|
S17 | 34(5) × 103 | 8.0(5) | 22(4) | 225(1) × 103 |
RCP8 | 41(5) × 103 | 18.2(1) | 31(4) | 225(1) × 103 |
Two ensembles of ferrite nanoparticles with the same average size of ∼4.5 nm (Fig. S3, ESI†) were synthetized by the polyol process. A pure CoFe2O4 sample (CFO) has been compared with the one in which about 50% of cobalt was substituted with nickel (CNFO, (CoNi)Fe2O4). Both samples represent concentrated systems with volume-filling factors of 52% and 39% for CFO and CNFO, respectively. Their particles are in close proximity, but not in contact since they are coated by polyol (see Section 2.2 and Fig. S4 in ESI†) preventing possible super-exchange interactions.46 A detailed description of the samples is provided in the ESI† and in ref. 23.
The M vs. H curves of the two samples recorded at 5 K are reported in Fig. 3a. By reducing the amount of cobalt, the magnetic anisotropy decreases, as reflected by a decrease in the coercive field μ0HC and the saturation field μ0Hsat (Table 2). The latter parameter is defined as the minimum field that reverses the moment of the particles with the highest anisotropy energy. It was evaluated as the field at which the difference between the two branches of the M(H) curves was reduced down to 1% of their maximum value.23 The larger anisotropy energy of sample CFO can be ascribed to its higher Co content. Cobalt ions have a marked magneto-crystalline anisotropy far above that of nickel and iron, indeed its orbital magnetic moment is not quenched by the crystal field, with a particular strong spin–orbit coupling for Co2+ ions in octahedral sites.47–49 The difference in saturation magnetization among CFO and CNFO can be explained by the different magnetic moments of Co and Ni and an unusual cationic distribution, as discussed in ref. 23.
Sample | Experimental formula | 〈DTEM〉 (nm) | M S (A m2 kg−1) | μ 0 H C (T) | μ 0 H sat (T) |
---|---|---|---|---|---|
CFO | Co0.97Fe2.00O4.00 | 4.5(1) | 130(10) | 0.88(7) | 3.1(8) |
CNFO | Co0.40Ni0.58Fe2.00O4.00 | 4.6(1) | 77(4) | 0.50(7) | 2.4(3) |
The thermal dependence of the magnetization measured by ZFC, FC and (MFC–MZFC) analysis gives interesting information about the blocking process (Fig. 3b–e and Table 3). In this case, the susceptibility is lower than that of sample RCP8, as well as the packing fraction. This enforces the assumption that the demagnetization effects have limited influence on these samples. Although the M(H) curves indicate a different magnetic anisotropy, the two samples show equal values of Tirr and Tmax, within the experimental error. This scenario is substantially confirmed by the (MFC–MZFC) magnetization curves measured after a field cooling of 2.5 mT (see Fig. 3d and e). For both samples, (MFC–MZFC) decreases with increasing temperature, as expected for an ensemble of magnetic monodomain particles. The T* values follow the same behavior of Tmax and Tirr, suggesting similar blocking temperatures for both samples. On the other hand, the values of T*(H) show different field dependences. A further confirmation of the unexpected situation, i.e., different magnetic anisotropy and the same blocking temperature, is provided by Mössbauer spectroscopy carried out at 300 K (Fig. 4). Generally speaking, the total Mössbauer spectrum of an ensemble of magnetic mono-domain particles is typically represented by the superposition of magnetic (six lines) and quadrupole (two lines) patterns. These two components are due to particles with superparamagnetic relaxation time that is, respectively, longer or shorter compared to the timescale of Mössbauer spectroscopy (tw = 5 × 10−9 s). In the samples under investigation, the fraction of particles being superparamagnetic at room temperature is similar, actually slightly higher for CNFO with respect to CFO (≈82 vs. 75%, respectively). This result confirms the landscape drawn by M vs. T measurements in low field, indicating that despite having different magnetic anisotropies, the superparamagnetic relaxation times of the two samples are very similar, as observed by the different time windows of two experimental methods.
Sample | T max (K) | T irr (K) | T* (K) | L 2NP | L corr (μ0H = 0 T) | L corr (μ0H = 2.5 mT) | μ 0 H limit |
---|---|---|---|---|---|---|---|
CFO | 229(4) | 255(4) | 174(2) | 10.1(1) nm | 8.4(1) nm | 8.3(1) nm | — |
CNFO | 231(4) | 255(4) | 170(2) | 10.8(1) nm | 15.1(1) nm | 14.8(1) nm | 0.087(1) T |
In order to investigate the interaction regimes among the particles, the remanent magnetization was measured following the IRM (Isothermal Remanent Magnetization) and DCD (Direct Current Demagnetization) protocols (see Section 3 of ESI† for details). For non-interacting single-domain particles with uniaxial anisotropy and magnetization reversal by coherent rotation, the two remanence curves are related via the Wohlfarth equation50mDCD(H) = 1 − 2mIRM(H), where mDCD(H) and mIRM(H) represent the reduced terms MDCD(H)/MDCD(5T) and MIRM(H)/MIRM(5T), with MDCD(5T) and MIRM(5T) being the remanence values for the DCD and IRM curves for a reversal field of 5 T, respectively. Kelly et al.51 rewrote the Wohlfarth equation to explicitly reveal the deviations from a non-interacting case:51
ΔM = mDCD(H) − 1 + 2mIRM(H) | (12) |
The negative ΔM values are usually taken as indicative of the prevalence of demagnetizing (e.g., dipole–dipole) interactions; positive values are attributed to the interactions promoting the magnetized state (e.g., direct exchange interactions). For both samples, the ΔM plots (Fig. 5) show a strong negative deviation, indicating the prevalence of magnetic dipolar interactions among the nanoparticles.52 The average reversal field, extracted from the field position of the negative peak, is higher for CFO due to the larger cobalt content (μ0Hrev ≈ 0.85 vs. 0.56 T, for CFO and CNFO, respectively). Furthermore, the amplitude of the peak (IΔM), proportional to the interaction magnitude,13 shows again a larger value for CFO (−0.34 vs. −0.17) (see Section 3 of ESI† for description), as expected considering the dipolar interaction energy to be roughly Edip = (μ0μ2)/(4πd3),53 where μ is the magnetic moment of the single particle and d is the distance between the particles’ center (considered as the point dipole). Due to similar average particle sizes and center-to-center interparticle distances (d) in the two samples, the higher saturation magnetization of CFO's particles leads to stronger dipolar interactions with respect to CNFO, being Edip/kB ≈ 44 vs. 12 K, respectively. This result is at odds with a very similar ZFC peak (Tmax) and T*, observed for the two samples.
Fig. 5 The ΔM-plots of samples CFO (black circles) and CNFO (red triangles) are reported. The solid lines are a guide to the eye. |
The M vs. H curves evidence quite different anisotropy energies for the two samples, but similar blocking energies emerge from the (MFC–MZFC) curves measured at low field (2.5 mT), and Mössbauer spectra measured with no applied magnetic field. Such results suggest the presence of a collective behavior induced by interactions, which creates a condition of higher effective anisotropy, in particular for CNFO, but which manifests only under low applied field; on the other hand, the high field applied in M(H) curves suppresses the coupling and lets the single particle anisotropy dominate.
To understand this complex behavior, we applied the modified random anisotropy model. For both samples, T* was measured from d(MFC–MZFC)/dT curves at several applied magnetic fields, and the values of T*(H) were fitted according to eqn (10) (Fig. 6a and data summarized in Table 3). In a condition of zero applied field, the correlation length calculated from (3) is larger than the single particle diameter for both samples, but only for sample CNFO this distance is long enough to enclose two full particles (L2NP), considering also the organic coating (Table 3). It is worth mentioning that the correlation length in these two samples is considerably shorter than in the RCP8 sample, as expected due to the higher particle anisotropy in CFO and CNFO. A similar situation is present even at an applied field of 2.5 mT, supporting the theory of a larger effective anisotropy visible for CNFO in M(T) measurements at low field and in Mössbauer spectra recorded in zero field. Indeed, at an applied field of 2.5 mT, the effective correlated volume of the magnetic material calculated from (6) for CFO and CNFO is ≈180 and 740 nm3, respectively. Despite the different single particle anisotropies, the average anisotropy energy within these two volumes is comparable, about 7.2(1) × 10−20 J, which translates to a similar experimental T*. In this framework, CFO shows stronger dipolar interactions, but the higher single particle anisotropy prevents the formation of large correlated clusters of particles. On the other hand, CNFO, despite showing reduced dipolar interactions with respect to CFO, is characterized by a lower NP anisotropy, which in turn allows for higher correlation lengths, at least for low enough applied fields. It is interesting to note that for CNFO, the μ0Hlimit, the applied field which reduces the correlation length below the limit of L2NP, is about 0.09 T (Fig. 6b). This explains why the effective cluster anisotropy is visible only at low field, indeed analyses of M(H) and ΔM-plot employ fields higher than μ0Hlimit, which reduce to zero the correlation length, thus leading to the dominant effect of the single particle anisotropy in both samples. It is well known that dipolar interactions affect the M(H) loop at low temperature (below blocking) by reducing the HC magnitude and leading to a concomitant lower remanence with respect to the corresponding isolated particles.26,54 Anyway, in CFO (stronger interactions), HC is greater than in CNFO due to the suppression of interparticle interaction effect caused by the much higher local single particle anisotropy. In strongly interacting particle systems, where Tmax in the ZFC curve reflects the phase transition to a collective superspin glass state, theoretical analysis55,56 and subsequent experimental works25,46 have suggested that Tmax is the result of interactions. Here we demonstrate for the first time that the individual particles’ anisotropy plays an additional opposite role, indeed higher NP energy barriers moderate the interparticle correlations and tend to reduce the glassy temperature. Thus, the overall magnetic behavior is the balance of the interplay between the single particle anisotropy and the interparticle interactions.
The second part of the work analyzed the complex interplay of single particle magnetic anisotropy and interparticle interactions in two ensembles of particles with the same average size: pure cobalt ferrite (CFO, CoFe2O4) and half-substituted cobalt–nickel ferrite (CNFO, Co0.5Ni0.5Fe2O4). The low-temperature hysteresis loops showed a strong reduction of magnetic anisotropy upon introducing nickel in place of cobalt. On the other hand, the thermally activated switching process of the magnetization analyzed by ZFC, FC and (MFC–MZFC) protocols with a low applied field of 2.5 mT, as well as by zero-field Mössbauer spectroscopy at 300 K, underlined a different picture, where almost the same average anisotropy energy emerged for the two samples. The interplay between the single particle anisotropy and the interparticle interactions is the key point to understand this intriguing behavior. Upon interpretation of the results by means of the random anisotropy model, it was shown that a strong single particle anisotropy prevailed in CFO. On the other hand, the right mix between the lower anisotropy and the quite long-range dipolar interactions produced a more correlated state in CNFO, which exhibited a blocking temperature higher than expected. This correlation was reduced by the application of an external field; indeed applied high fields allowed the single particle anisotropy to emerge. These results provide experimental evidence that the overall magnetic behavior of strongly interacting particles depends on the complex interplay between the single particle anisotropy energy and the interparticle interaction energy, with higher individual NP energy barriers playing against the collective behavior.
These conclusions are of interest not only from a fundamental point of view but also for their potentiality in technological applications. As demonstrated by the results of CFO and CNFO, it is possible to develop systems in which the thermal and the field-induced reversal of the magnetization can be tuned independently, opening new interesting perspectives in order to design materials for specific applications.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8cp03934h |
This journal is © the Owner Societies 2018 |