Carlos
Moya
*ab,
Mariona
Escoda-Torroella
ab,
Javier
Rodríguez-Álvarez
ab,
Adriana I.
Figueroa
ab,
Íker
García
a,
Inés Batalla
Ferrer-Vidal
a,
A.
Gallo-Cordova
c,
M.
Puerto Morales
c,
Lucía
Aballe
d,
Arantxa
Fraile Rodríguez
ab,
Amílcar
Labarta
ab and
Xavier
Batlle
*ab
aDepartament de Física de la Matèria Condensada, Universitat de Barcelona, Martí i Franquès 1, 08028 Barcelona, Spain. E-mail: carlosmoyaalvarez@ub.edu; xavierbatlle@ub.edu
bInstitut de Nanociència i Nanotecnologia (IN2UB), Universitat de Barcelona, 08028 Barcelona, Spain
cDepartment of Nanoscience and Nanotechnology, Instituto de Ciencia de Materiales de Madrid (ICMM-CSIC), Sor Juana Inés de la Cruz 3, 28049 Madrid, Spain
dALBA Synchrotron Light Facility, CELLS, 08290 Barcelona, Spain
First published on 3rd January 2024
Iron oxide nanoflowers (IONF) are densely packed multi-core aggregates known for their high saturation magnetization and initial susceptibility, as well as low remanence and coercive field. This study reports on how the local magnetic texture originating at the crystalline correlations among the cores determines the special magnetic properties of individual IONF over a wide size range from 40 to 400 nm. Regardless of this significant size variation in the aggregates, all samples exhibit a consistent crystalline correlation that extends well beyond the IONF cores. Furthermore, a nearly zero remnant magnetization, together with the presence of a persistently blocked state, and almost temperature-independent field-cooled magnetization, support the existence of a 3D magnetic texture throughout the IONF. This is confirmed by magnetic transmission X-ray microscopy images of tens of individual IONF, showing, in all cases, a nearly demagnetized state caused by the vorticity of the magnetic texture. Micromagnetic simulations agree well with these experimental findings, showing that the interplay between the inter-core direct exchange coupling and the demagnetizing field is responsible for the highly vortex-like spin configuration that stabilizes at low magnetic fields and appears to have partial topological protection. Overall, this comprehensive study provides valuable insights into the impact of crystalline texture on the magnetic properties of IONF over a wide size range, offering a deeper understanding of their potential applications in fields such as biomedicine and water remediation.
The apparent SPM behaviour makes these IONF ideal magnetic platforms for both biomedicine and water remediation purposes since the net magnetization of the IONF can be controlled at will by the application of a weak external magnetic field so that the particle agglomeration is effectively reduced.15–21
Crucial to optimizing IONF designed for the above applications is probing the 3D magnetic texture associated with the actual demagnetized state, an issue that is becoming more cumbersome than expected. The reason is that there are several aspects affecting the demagnetized state of the IONF such as the distribution of sizes and magnetic anisotropy, and the crystallographic texture of the inner cores within the IONF. The interaction between cores can lead to the formation of magnetic superstructures within a single IONF, potentially leading to supramagnetic correlations over several groups of IONF. Hence, a thorough understanding of both the internal crystallographic and magnetic structures of single IONF is essential.
While conventional structural and magnetic measurements are the backbone for a precise understanding of the nanoscopic properties of the IONF, the study of the specific features within the individual IONF is essential to achieve ultimate control over the functionality of the IONF arrangements. This calls for a correlative study combining standard techniques with advanced complementary techniques sensitive to local magnetic texture, such as synchrotron-based Transmission X-Ray Microscopy (TXM) combined with X-ray magnetic circular dichroism (XMCD),22–24 together with numerical calculations that reproduce the experimental results.
This work reports such a study of different-sized IONF synthesized through polyol routes, ranging from 40 to 400 nm. Despite the significant variation in the size of the aggregate, these samples exhibit a consistent crystalline correlation that extends beyond the individual cores yielding high values of both the saturation magnetization and the initial susceptibility, but with almost zero remnant magnetization and coercivity. Magnetic TXM imaging of tens of individual IONF reveals a predominantly demagnetized state, with a seemingly slight correlation between the agglomerate size and the particular magnetic texture. These experimental findings are supported by micromagnetic simulations, where the role of the interplay between the demagnetizing field and the magnetic interactions between neighbouring cores in stabilizing 3D vortex-like spin configurations is unambiguously shown. These results highlight that the manipulation of the crystalline texture of IONF opens new opportunities for designing highly efficient drug delivery carriers with enhanced properties, as well as materials with excellent sorption capacity for environmental remediation.
High-resolution TEM (HRTEM) images depicted in Fig. 2(a), (d) and (g) and in Fig. S1–S3 in the ESI† reveal common trends in the three samples, regardless of IONF size. For example, IONF are composed of multicore aggregates that resemble petal-like nanostructures and interplanar distances consistent with monophasic γ-Fe2O3. Moreover, samples exhibit the same family of planes through large areas of aggregates, which suggests the existence of a crystal correlation between the cores. For example, this is apparent in sample IONF40, where the crystal plane (210) is homogenously distributed over the whole IONF (see Fig. 2(b)). Regarding the crystalline texture of the samples, one would expect Small Angle Electron Diffraction (SAED) patterns for a single IONF to look like those of polycrystalline particles, showing concentric ring-like patterns due to the small size of the cores. However, Fast Fourier transform (FFT) and SAED patterns shown in Fig. 2(c), (f) and (i), corresponding to IONF40, IONF100, and IONF400, respectively, contain a set of relatively ordered spots extending along small arcs of the fully polycrystalline rings. The average angle subtended by the arcs of the spots in the three samples is within 7–14°. This fact indicates that the IONF are neither entirely polycrystalline nor single crystal but show a significant crystalline texture among the cores within each IONF. As expected, the lowest value of the subtended angle is found in sample IONF40, which is made of fewer cores than the other two particle types. Interestingly, although the sample IONF400 contains many more cores than the other two, the average value of the subtended angle is still small, indicating the prevalence of a strong preferential orientation of the cores within the IONF, even in such large core aggregates. This trend can also be visualized by indexing the spots of a SAED pattern of a single IONF400 to the planes oriented in the [111] zone axis of a simulated diffraction pattern of γ-Fe2O3 (see Fig. S4 and Table S1 in the ESI†).
The crystalline correlation among the cores was further investigated by X-ray diffraction (XRD) (see Fig. 3). At first glance, XRD data show the typical peaks associated with the γ-Fe2O3 phase, ruling out the presence of other parasitic iron oxides phases, such as wüstite (FeO) or hematite (α-Fe2O3).28,29 It is worth mentioning that although these patterns could also be indexed to the Fe3O4 phase, the actual fraction of this phase in the samples, if present, should be very small due to the oxidative acid treatment during the synthesis process (see details in the Methods section). The values of the crystal size determined by the Scherrer method for the three samples are within a small range between 13 and 15 nm.30 Note that these values are more than twice the TEM core sizes, indicating a crystalline order that extends well beyond individual cores. Table 1 summarizes the structural parameters of the three samples obtained from the overall structural characterization techniques.
Sample | D IONF (nm) | σ RSD (%) | D core (nm) | σ RSD (%) | D XRD (nm) |
---|---|---|---|---|---|
IONF40 | 38 ± 10 | 26 | 7 ± 1 | 31 | 13 ± 4 |
IONF100 | 122 ± 14 | 11 | 5 ± 1 | 20 | 15 ± 3 |
IONF200 | 186 ± 28 | 15 | 20 ± 4 | 20 | 20 ± 3 |
IONF400 | 379 ± 15 | 36 | 4 ± 1 | 23 | 13 ± 3 |
Sample | T = 298 K | T = 5 K | ||||
---|---|---|---|---|---|---|
M S (emugγ-Fe2O3−1) | M r (emu gγ-Fe2O3−1) | H C (Oe) | M S (emu gγ-Fe2O3−1) | M r (emu gγ-Fe2O3−1) | H C (Oe) | |
IONF40 | 78 ± 4 | 1.4 ± 0.3 | 4 ± 1 | 87 ± 5 | 14.1 ± 0.5 | 95 ± 3 |
IONF100 | 70 ± 3 | 0.7 ± 0.2 | 4 ± 1 | 88 ± 4 | 13.4 ± 0.4 | 104 ± 3 |
IONF200 | 77 ± 5 | 4.6 ± 0.5 | 31 ± 1 | 84 ± 1 | 21.5 ± 1.5 | 227 ± 2 |
IONF400 | 72 ± 4 | 0.7 ± 0.2 | 2 ± 1 | 88 ± 12 | 10.9 ± 0.5 | 106 ± 3 |
M ZFC/MFC curves as a function of temperature were recorded under H = 50 Oe (see Fig. S7 in the ESI†). They did not show any hint of a blocking temperature, and the onset of irreversibility between the FC and ZFC branches was just at the temperature at which the cooling process started (RT). Additionally, the FC curves exhibited very weak temperature dependence, if any, and no Curie-like increase in magnetization was detected at low temperatures. These results suggest an arrangement of the magnetic moments with significant magnetic correlations throughout the IONF but yielding an almost fully demagnetized state at zero field, similarly to that previously found for nanocrystalline Fe3O4 nanoparticles.33
Fig. 6 Illustrative examples of magnetic TXM images recorded at polar angle θ = 0° that are representative of the three main categories of moment textures found within the 53 analyzed IONF, labelled as A (a), B (b), and C (c). These particles correspond to those highlighted in Fig. 5(a), and (b). Note that magnetic moments are pointing inwards and outwards the plane for θ angles below and above 90°, respectively, and that the length and direction of the arrow indicate the magnitude and direction of the normalized magnetization vector. |
By adding up all the local magnetization vectors within each individual IONF, the net normalized magnetization value, mz, in a scale of −1 to 1 has been obtained. As an example, the IONF in Fig. 6(a) shows a net magnetization of −0.0056, that is, very close to zero corresponding to a demagnetized state. By repeating this process for each of the 53 IONF analyzed, we find that mz as a function of θ for this subset is +0.02 (−7.5°), −0.0057 (0°), +0.034 (5°), and +0.019 (7.5°). Within our statistics, one can conclude that a nearly demagnetized state is thus generally found in IONF, in good agreement with the very low remanence found by macroscopic magnetic characterization of the samples (see Fig. 4, and Fig. S1–S8 in the ESI†), and the micromagnetic simulations that will be shown further below. In addition, as displayed in Fig. S10 in the ESI,† there is no apparent correlation between the IONF size (estimated in pixel units from the TXM images) and mz, which is less than 0.12 in all the cases studied. Note that a precise estimation of the particle size cannot be determined from the TXM images since the spatial resolution is about two orders of magnitude worse than that of the HRTEM images shown in Fig. 2 (0.1 nm spatial resolution from HRTEM vs. 7.7 nm pixel size in TXM).
The analysis performed on the orientation of the local magnetic moments seem to indicate the existence of different types of magnetic textures within individual IONF. Within our subset of 53 particles, three types of spin arrangements within individual IONF have been identified. About 50% of the IONF show a vortex-like configuration with a central tube of magnetization pointing out of the image plane (see Fig. 6(a)), hereafter denoted as type A. About 20% of them, labelled as type B, are compatible with two relatively large black and white regions of somewhat homogeneous magnetization (Fig. 6(b)), resembling magnetic domains. The remaining 30% of IONF fall into an intermediate category, labelled as type C, of black, grey, and white regions of comparable sizes (Fig. 6(c)). Within our error bars, a slight tendency towards more homogeneous textures (such as type B) is observed in IONF with smaller sizes. In contrast, a tendency for more complex twisted moment textures with higher vorticity (such as type A) tends to be found in larger IONF which could be related to increasingly complex correlations between the magnetic moments of the cores associated with their progressively larger number in the aggregate.
Fig. 7 System set up formed building a spheroid of 160 nm in diameter with cubic cores of 16 nm in edge length made up of cells of 4 × 4 × 4 nm. |
Both the local arrangement of the moments throughout the IONF and the z-component of the net magnetization of the IONF were studied when a magnetic field was applied along the z-axis following discrete time steps and various convergence criteria to decide when to initiate the next step (see details in the Methods section). In addition, simulations were carried out varying the exchange constant Ai between neighbouring cells in different cores to find out the effect of the inter-core exchange constant on the magnetic configuration of the IONF. As a general trend, the hysteresis loops became more square-like and had higher remnant magnetization and coercivity as Ai increased. Interestingly, the hysteresis loop (see Fig. 8(a)) best resembling that of the real systems was found for Ai = 0.1Aw (Aw being the exchange constant between cells belonging to the same core) thus, this value was selected as a set parameter in the following calculations.
One should mention that the same distribution of anisotropies was used in all the simulations shown in this work, and hence, the results were dependent on this realization of the system. Thus, for instance, the detail in the non-reversible part in the hysteresis loop (small jumps and other anomalies) in Fig. 8(a) was related to this configuration of anisotropies. Nevertheless, we checked that other anisotropy distributions gave similar overall behaviour, and only non-significant small differences were found among the results of those simulations. In fact, the system was large enough as to get a kind of self-averaging of the anisotropy disorder.
Regarding the moment arrangements as a function of the applied magnetic field, four stages of the moment configuration following a hysteresis loop are shown in the snapshots in Fig. S11 in the ESI,† corresponding to an equatorial plane perpendicular to the z-axis. At low magnetic field including the remnant states, moment arrangements with high vorticity are found (see Fig. S11(a) and (d) in the ESI† and Fig. 8(b) for remnant states after saturation in opposite directions of the magnetic field). This is because of the combined effects of the demagnetizing field acting in the whole IONF and the weak magnetic interaction between neighbouring cores that causes a certain magnetic correlation among them extending throughout the IONF. This type of complex structure is what gives the system its low remanence and coercivity values, and it vanishes as the system tends to saturation along the field direction. It is also worth stressing that the moments around the central axis of the IONF (along the z-axis) have a significant out-of-plane component (see Fig. 8(c)) of the magnetization, even at remanence. In fact, these central moments form a kind of axial magnetization tube, conferring a certain polarity to the moment vortex that coincides with the direction of the last applied magnetic field (see Fig. 8(c)). In addition, the direction of rotation of the vortex is reversed in the two snapshots at −1.269 kOe in Fig. S11(b) and (c) in the ESI,† since they were simulated while the magnetic field was decreasing and increasing, respectively, before and after saturation along the negative direction of the field axis. It must be pointed out that the moment configurations at remanence (shown in Fig. 8(b) and S8(b) in the ESI†) are in qualitative agreement with the experimental 2D magnetic TXM maps of single IONF labelled as categories A and C (see Fig. 6(a) and (c)). Furthermore, both the vorticity of the moment configuration and the central tube of out-of-plane magnetization are common features found in the OOMMF simulations at remanence and the magnetic TXM data of IONF falling into the categories A and C (Fig. 6(a) and (c)).
Further insight into the vorticity of the magnetization can be gained by studying as a function of the magnetic field. Since is a discrete field, a discrete approximation must be taken when calculating the vorticity. Thus, Fig. 8(a) shows the average z-component of the vorticity as the particle follows the hysteresis loop. Notice that abrupt jumps in the magnetization are highly correlated with corresponding changes in the vorticity, as these jumps are the result of reconfigurations of the moment arrangement. In addition, the chirality of the vortex is reversed when the system is saturated and then de-saturated as shown by the snapshots in Fig. S8 in the ESI.† When the applied field is reduced from saturation, the vortex starts to form in the chirality that lowers the energy according to the anisotropy distribution present in the system, which is in opposite directions whether the applied field is increasing or decreasing. At low magnetic fields, vorticity decreases since anisotropies have a more dominant effect in the system, and thus, does not align with a perfect vortex.
Topology enables classifying these types of vortex structures. Each can be viewed as a point on the sphere S2. Therefore, one can map a 2D cross section of the configuration to the sphere via the stereographic projection, and classify the mapping from the sphere to the sphere according to how many times it enfolds around itself and computing the topological number W:
(1) |
Fig. 8(a) shows W following the hysteresis loop for the equatorial plane shown in Fig. 8(b). Note that there is only non-zero values of W in the non-reversible part of the hysteresis loop. Moreover, W can be viewed as the product of two numbers, which are associated with the net axial polarization of the magnetization and the winding of the moment arrangement. Thus, the abrupt jumps between negative and positive values of W in Fig. 8(a) at Hz = ±0.55 kOe are due to the reversal of the central tube of magnetization along the z-axis, while keeping the direction of the vorticity. Interestingly, a maximum value of W = 0.5 is attained for Ai = 0.1Aw, suggesting that the most stable magnetic structure from the topological point of view takes place for this small value of the inter-core exchange. Thus, although the moment configuration does not attain a complete topologically protected structure at any field, W = 0.5 is remarkably high for a magnetic structure mainly originating from demagnetizing effects and the relatively weak exchange interactions among the cores.
(2) |
The mean particle size DTEM and the standard deviation σ were computed from eqn (3) and (4), respectively, as follows.
DTEM = D0eS2/2 | (3) |
(4) |
Finally, the diameter dispersion was compared among samples by using the variation coefficient σRSD = σ/DTEM.
The crystal structure of IONF was determined by combining the analysis of HRTEM and SAED patterns obtained with a JEOL 2100 microscope. Interplanar distances (dhkl) were calculated using Gatan Microscopy Suite® software and compared to X'Pert High Score Plus patterns for bulk γ-Fe2O3 (Inorganic Crystal Structure Database, ICSD: 00-039-1346). The interplanar distances of SAED were calculated by measuring the distance between the central spot and the diffraction spots using ImageJ software in at least three different IONF for each sample. The reflections were indexed to the (hkl) planes using the above crystal structure pattern. The representation of the reciprocal lattice at the [111] zone axis was carried out by using CaRIne Crystallography software, version 3.1, and fitting the crystal lattice to the measured SAED area.38
A PANalytical X'Pert PRO MPD diffractometer with Cu Kα radiation (λ = 1.5418 Å) was used to collect XRD spectra within 10° and 70° with a step size of 0.040° of 2θ. The peak positions were compared to a reference spectrum of γ-Fe2O3 (ICSD: 00-039-1346), which was also used to determine the crystal size DXRD by Rietveld analysis of the full spectra using the FullProf Suite software.39
Fe concentration in the colloidal suspensions was determined by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) with a PerkinElmer OPTIMA 2100 DV by digesting and diluting controlled volumes of samples.
Magnetic features were evaluated in powder samples by measuring hysteresis loops M(H) recorded within ±5T at 5 and 298 K employing a Vibrating Sample Magnetometer (Oxford Instrument Model MLVSM MagLab 9 T). Magnetization values were normalized to the ICP-OES results assuming γ-Fe2O3 phase. Saturation magnetization MS was obtained by extrapolation of the high-field region of M(H) to zero field, assuming the high-field behaviour: M(H) = MS + χpH, where MS and χp correspond to the saturation magnetization and the residual susceptibility due to parasitic paramagnetic species and/or non collinear spin alignments, respectively. Coercive field Hc was computed as Hc = (|Hc+| + |Hc−|)/2, where Hc+ and Hc− were the intercepts of the hysteresis loop branches with the H-axis. Hysteresis loops at 298 K were fitted to a Langevin function (see eqn (5)).40,41
(5) |
Imaging of the M-TXM in individual IONF was carried out at MISTRAL beamline of the ALBA light source equipped with a full field TXM and a high-precision rotary stage,34 in samples deposited by drop casting onto C-coated marked TEM grids. The two components of the magnetization in the plane perpendicular to the rotation axis were studied in a tomographic series of images using polar angles θ of −7.5°, −5°, 0°, 5°, and 7.5°. To eliminate systematic positioning errors of the goniometer of the microscope, images with both polarizations were acquired sequentially at each angle.
The 2D magnetization maps at given polar angles were studied in 53 IONF using image processing and analysis routines based on ImageJ and homemade python codes.37 All the particles were probed simultaneously, thus, systematic errors arising from either the measurements (e.g., due to an inhomogeneous illumination across the TXM images) or the analysis procedure (mainly caused by image drift correction) were comparable for all the analysed particles. To distinguish the perimeter of each IONF in the XMCD stack, an overlay tool of ImageJ was used as a mask using XAS stacking as reference. Selection of the 53 IONF particles analysed was done after background subtraction of the XMCD image, by cutting each IONF out as a square section with the ImageJ Region of Interest (ROI) tool. The contrast level of each IONF was measured from the mean grey value inside square boxes of 2 × 2 pixels. To visualize the orientation of the magnetic moments inside a selected single IONF in a 2D map of magnetic moments a homemade python program was used to convert the grey scale of the XMCD contrast measurements to a distribution of colour arrows scaling between 0° (red, mz = +1) and 180° (blue, mz = −1).
Numerical calculations (OOMMF code 37) were performed to simulate the internal arrangement of the core moments within a single IONF.36 The real system was approximated by building a 160nm-IONF with small cubes formed by cells of 4 × 4 × 4 nm that mimicked the dimensions of a petal. The exchange constant between cells belonging to the same core was set to Aw = 7 × 10−12 J m−1, which is the value corresponding to bulk maghemite,42 while the exchange constant Ai between neighbouring cells in different cores was set to a percentage of Aw within 5% and 50% in order to account for the inter-core interactions. Therefore, considering that the saturation magnetization for maghemite is Ms = 4.8 × 105 A m−1,42 the chosen cell length of 4 nm is larger or, at most, the same order of magnitude than the magnetostatic correlation length , supporting the validity of the micromagnetic simulations. There is a significant variability in the shape of the cores in an IONF, so a uniaxial anisotropy axis accounting for the shape and surface anisotropy of each core was chosen randomly, and an average anisotropy constant was set to Ku = −5 × 103 J m−3, which is a reasonably value for γ-Fe2O3 particles of about tens of nm in size. For the cubic anisotropy, intrinsic to γ-Fe2O3, two unit-vectors were generated for each core, one having a normal distribution around the z-axis with standard deviation of 5°, and the other one perpendicular to it with a uniform distribution of orientations. This was done to account for the important crystalline texture among the cores shown by the IONF. The value of the cubic anisotropy constant was set to Kc = −1.3 × 104 J m−3.42 A white noise that simulated the effect of a temperature of T = 10 K was introduced by a specific routine to reduce the possibility of the system to get stuck in some local minima. When adding a non-zero temperature, criteria based on the convergence of the time derivative of the cell moment were discarded as the fluctuations in the magnetization were greater than any reasonable upper bound. Hence, time criteria were used for the stabilization of the magnetization after any field variation. Given that the path of a hysteresis loop follows metastable states, the results depend on the observation time and the steps taken when varying the applied field. Thus, different stopping times were used for a total number of 320 field steps. The longer the stopping time, the more detail shown in the hysteresis loop, but also the longer total simulation run time. Hence, a stopping time of 5 × 10−9 s was set as a reasonable compromise between both aspects.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3nr04608g |
This journal is © The Royal Society of Chemistry 2024 |