Production of Fe nanoparticles from γ-Fe2O3 by high-pressure hydrogen reduction

In this work, the reduction of iron oxide γ-Fe2O3 nanoparticles by hydrogen at high pressures is studied. Increasing the hydrogen pressure enables reduction of γ-Fe2O3 to α-Fe at significantly lower temperatures. At low pressures, a temperature of 390 °C is necessary whereas at 530 bar complete reduction can be realized at temperatures as low as 210 °C. This leads to significant improvement in the final particle morphology, maintaining high surface-to-volume ratio of the nanoparticles with an average size of 47 ± 5 nm which is close to that of the precursor γ-Fe2O3. Neck formation, coalescence and growth during reduction can be significantly suppressed. Investigations of magnetic properties show that saturation magnetization of the reduced α-Fe nanoparticles decreases with particle size from 209 A m2 kg−1 at 390 °C reduction temperature to 204 A m2 kg−1 at 210 °C. Coercivity for the fine iron particles reaches 0.076 T which exceeds the theoretical anisotropy field. This is attributed to nano-scale surface effects.


Introduction
Iron nanoparticles have important applications due to their unique chemical and magnetic properties. 1 For example, it has been shown that compared to micrometer-sized particles, the catalytic activity of nanostructured iron for the Fischer-Tropsch process is much higher due to the larger surface area. 2 Similar catalytic activity to platinum-based materials for chemical hydrogen storage applications has been reported for amorphous Fe nanoparticles. 3 It has also been demonstrated by Dirba et al. that ne Fe nanoparticles are crucial for formation of the a 00 -Fe 16 N 2 phase using low-temperature nitriding with ammonia. 4 Furthermore, since iron has the highest saturation magnetization among all ferromagnetic elements, yet is a cheap, abundant and environmentally friendly material, numerous applications of Fe nanoparticles in magnetism are promising. Exchange-coupled nanocomposite magnets for remanence and maximum energy product enhancement, consisting of nanoscale a-Fe as the magnetically so phase and SmCo 5 , Nd 2 Fe 14 B or Sm 2 Fe 17 N 3 as the hard phases, have been investigated. [5][6][7][8] Furthermore, the potential in biomedical applications, such as magnetic resonance imaging 9 or hyperthermia mediated drug release for cancer therapy 10 has also been discussed if a proper surface coating (e.g. Au) is provided. 11 Synthesis of nanoparticles with controlled particle size, morphology and surface modications can be achieved by solution chemistry techniques. [12][13][14] This approach is well suited for biomedical applications where minimizing reactivity and agglomeration is important. However, for magnetic nanocomposites and catalysts, the presence of an organic shell around the nanoparticles is not favorable, because it competes with the exchange length (<5 nm for typical hard magnetic materials; e.g. 1.9 nm in for Nd 2 Fe 14 B 15 ) in the former and hinders active surface sites for the latter. Therefore, in this work we demonstrate how to produce ne a-Fe nanoparticles using a gas-solid high-pressure hydrogen reduction process.
In literature, reduction of iron oxides to a-Fe with hydrogen is typically conducted at temperatures of at least 390 C-500 C. [16][17][18] This choice of temperature is supported by hygrometry measurements, where the maximum rate of H 2 O production was from 389 C to 522 C 19,20 depending on the heating rate. Heat treatment at such elevated temperatures leads to particle coalescence and growth of their effective size, followed by a detrimental reduction of their surface area due to temperature-enhanced diffusion which increases exponentially with temperature. A sintering prevention layer, such as Al 2 O 3 , 21 hydroxyapatite 22 or carbon 23 can be helpful, which, however, covers the produced nanoparticles and, therefore is also not suitable for catalyst or exchange-coupled nanocomposite applications due to the additional shell.

Nanoscale Advances
PAPER Assuming a single-step reaction, hydrogen reduction of Fe 2 O 3 can be written as  (1) is endergonic and endothermic, and does not occur spontaneously at ambient conditions. Taking the expression for K r into account, a shi of the chemical equilibria to the right (towards products) and complete reduction of the iron oxide to Fe at lower temperatures is achievable at higher hydrogen partial pressure p H 2 . Therefore, in this work, we attempt to achieve complete reduction of Fe 2 O 3 at signicantly lower temperatures by using high hydrogen pressures. This approach should result in improved morphology of the produced nanoparticles since the detrimental coarsening during the reduction step is suppressed.

Experimental
Commercial maghemite g-Fe 2 O 3 nanoparticles from Alfa Aesar with an average particle size of 20-40 nm according to the manufacturer's datasheet, were used as a precursor. Reduction experiments were done in a hydrogen (99.999% purity, Linde) atmosphere in a custom-made autoclave system equipped with high pressure/high temperature vessel (series 4740, Parr Instrument Company) and an external vertical tube heater assembly (Model 4921, Parr Instrument Company). The working volume of the vessel was 75 cm 3 . 0.050 g of iron oxide nanoparticles were lled in a steel crucible and mounted in the pressure vessel. Aer twice evacuating and purging with Ar (99.999% purity, Air Liquide), the reactor was heated to target temperature under continuous pumping and then pressurized with hydrogen. The reduced samples were handled in an Ar lled glovebox (p(O 2 ) < 0.1 ppm, MBraun) in order to avoid oxidation. Magnetic properties of the nanoparticles were measured using vibrating sample magnetometer (VSM) (7400 Series, Lake Shore). Microstructural investigations were carried out using a scanning electron microscope (SEM, Philips XL30 FEG, 15 kV acceleration voltage). For scanning transmission electron microscopy (STEM) a JEOL JEM 2100F was used. The microscope was operated at 200 kV. The sample was prepared on a lacey carbon Cu grid in a glovebox and transferred to the TEM in a Gatan 648 double tilt vacuum transfer holder under nitrogen atmosphere to prevent oxidation. In addition to highresolution (HR) TEM imaging and energy dispersive X-ray spectroscopy (EDS) in scanning transmission electron microscopy (STEM), nanobeam diffraction was performed to acquire electron diffraction data from different regions of the particle. The full width half maximum of the electron beam was 2 nm. Crystal structure and phase purity of the nanoparticles were investigated by X-ray diffraction (XRD) using a powder diffractometer (STOE, Stadi P) with Mo K a1 -radiation (l ¼ 0.70930 A) in Debye-Scherrer geometry with powders sealed in quartz capillaries to avoid oxidation. X-ray photoelectron spectroscopy (XPS) measurements were performed on compacted nanoparticle pellets using a PHI Versaprobe 5000 spectrometer with monochromatic Al K a radiation.

Results and discussion
The initial g-Fe 2 O 3 is a typical brown rust-like powder. The average particle size is expected to be around 20-40 nm  according to manufacturer's data. SEM and TEM images of the powder are presented in Fig. 1. Particles show a broad size distribution from ne 20 nm nanoparticles to several large particles in the range of hundreds of nanometers. The average particle size estimated from the SEM images is 52 AE 31 nm (152 particles counted). The average crystallite size obtained from XRD peak broadening is 55 AE 5 nm. The TEM image shows that the ne nanoparticles mostly have well-dened hexagonal shapes. Fig. 2 shows XRD patterns of the initial g-Fe 2 O 3 and the same nanoparticles reduced at a hydrogen pressure of 100 bar for 3 hours at different temperatures in the range from 230 C to 390 C. At 390 C, in agreement to previous reports, reduction can be achieved even at low pressures (not shown here). 100 bar hydrogen pressure is sufficient for complete reduction at temperatures down to 270 C within 3 h. Further lowering of the reduction temperature to 230 C leads to partial reduction and remaining g-Fe 2 O 3 reections are evident in the XRD data.
In a next step, the reduction of g-Fe 2 O 3 nanoparticles at temperatures below 230 C was investigated at hydrogen pressure above 100 bar. The corresponding XRD results are presented in Fig. 3. At 230 C and 200 bar hydrogen pressure, a small signal from the highest intensity iron oxide diffraction peak can be identied at 16.1 .
Therefore, the pressure was increased further to 300 bar at which the g-Fe 2 O 3 nanoparticles are fully reduced; only reections from a-Fe are visible in the corresponding diffractogram. At lower temperatures of 220 C and 210 C, a hydrogen pressure of 400 bar and 530 bar is necessary to complete the reduction in 3 hours. Thus, by increasing the hydrogen pressure to 530 bar, it is indeed possible to reduce the temperature necessary for complete reduction of g-Fe 2 O 3 nanoparticles to a-Fe from 390 C to 210 C. The pressure of 530 bar was the limit for the autoclave, a further decrease in the reduction temperature for higher pressures can be expected. On the other hand, it is possible that at a certain pressure the benets from the lower temperature will be outweigh by a reduction of the diffusion barrier 25 and gas-pressure sintering effects. 26 In order to investigate the coalescence and growth of the nanoparticles at the proposed high-pressure, low-temperature  Nanoscale Advances hydrogen reduction, free-standing nanoparticles were examined using SEM. The reduction at 390 C leads to severe bonding of the nanoparticles into sponge-like agglomerates with sizes in several tens of micrometers (Fig. 4a). A coalescence of the nanoparticles into large grains with much smaller surface area is observed in the high-resolution SEM image in Fig. 4b. Bridging between neighboring nanoparticles and their coalescence occur also at 310 C (Fig. 4c) and at 270 C (Fig. 4d), although, with much smaller increase in the size. However, aer reduction at 230 C (Fig. 4e) and especially at 210 C (Fig. 4f), neck formation and coalescence are minimized, and individual nanoparticles can be distinguished. Thus, the coalescence of the nanoparticles during the reduction process has been suppressed. Clustering still takes place due to magnetostatic interactions, therefore monodisperse particles could not be achieved. The average size of the obtained phase-pure a-Fe nanoparticles is comparable with that of the g-Fe 2 O 3 precursor. A particle size analysis was conducted by the line interception method in which an average size was calculated for the grains along 4 lines (horizontal, vertical and both diagonals) 27 in the respective SEM images. The obtained sizes of the nanoparticles are summarized in Table 1, including a comparison with the crystallite size, extracted from the XRD peak broadening. Usage of lower reduction temperatures is clearly benecial and results in ne a-Fe nanoparticles with a size smaller than 50 nm. Clearly, lowering of the reduction step temperature results in much ner a-Fe particles. Error in the SEM data is caused mainly by a limited number and inaccurate size measurements of the analyzed nanoparticles in the SEM images, especially for large concrescent particles at high reduction temperatures. The extracted value of the crystallite size is additionally inuenced by the broadening of the XRD reections due to microstrain. Fig. 5 summarizes the results of hydrogen reduction experiments. The data points represent temperaturepressure pairs where complete reduction of Fe 2 O 3 to a-Fe was achieved according to the XRD data. It can be concluded that reduction of iron oxide nanoparticles at elevated hydrogen pressures and consequently lower temperatures can indeed be used for production of high-quality iron nanoparticles without the detrimental particle coarsening.
Due to the broadening of the XRD reections in nanoparticles and a possibility that some fraction of the material is amorphous, small amounts of residual iron oxide could remain   unnoticed in the XRD data. For this reason, XPS measurements of the initial g-Fe 2 O 3 powder and the reduced a-Fe nanoparticles were conducted to analyze electronic structure of Fe (Fig. 6). Prior to the XPS measurements, the samples were sputtered for 10 minutes using Ar + ions at 2 kV to remove the adventitious carbon contamination and oxidized layers on the sample surface due to contacts with atmospheric oxygen. The Fe2p spectra for the initial g-Fe 2 O 3 nanoparticles show emissions at 710.8 and 724.5 eV (a spin-orbit splitting of 13.7 eV) for the Fe2p 1/2 and Fe2p 3/2 , respectively. The observed shake-up satellite peaks are shied from the parent peaks by 8.5 eV. Within the measurement accuracy, the spectra correspond to those reported in literature for g-Fe 2 O 3 . 28,29 The Fe2p spectra of the obtained a-Fe nanoparticles reveal a textbooklike 30 Fe2p 1/2 and Fe2p 3/2 emissions at 706.8 and 719.9 eV, respectively, with 13.1 eV spin-orbit splitting and absence of shake-up satellites. The XPS results indicate complete reduction of the g-Fe 2 O 3 powder at 210 C and 530 bar hydrogen pressure. It has to be noted that the measured initial survey spectra show oxygen and carbon on the surface which can be removed aer in situ Ar ion sputtering. This indicates surface oxidation of the a-Fe nanoparticles.
In the HRTEM image shown in Fig. 7a a spherical particle shape can be observed in accordance to the SEM imaging discussed above. A core-shell type oxidation layer is revealed. Since atomic resolution information is included in this image, it is possible to apply a bandpass lter for the lattice spacing of the Fe 2 O 3 (210) lattice spacing (d ¼ 2.519 A). The ltered image shown in Fig. 7b clearly shows the occurrence of this spacing in the outer or shell regions of the particle. This coincides with increased oxygen concentration in the shell, which is evident in the STEM-EDS mapping, Fig. 7c. As a spatially resolved electron diffraction (ED) method, NBD allows for the collection of kspace data with a high real space resolution. A series of NBDED patterns was acquired along indicated line in Fig. 7a and the azimuthal integrated intensities are plotted in Fig. 7e. Characteristic Bragg intensities are marked for both phases, a-Fe and g-Fe 2 O 3 , it can be observed that the core of the particle, the Fe intensities are prominent and with increasing distance from the center of the particle, the Fe 2 O 3 phase intensities increase. In the experimental geometry, only a projection can be observed, which is causing a convolution of the probed volume in both the NBD patterns and the EDS intensities. This behavior is shown in Fig. 7f, where a homogeneous distribution of an element in Fig. 7 (a) HRTEM image of a representative nanoparticle, the NBD scan direction is indicated originating at the center of the particle. (b) Bandpass/Fourier filtered image with the Fe 2 O 3 (210) distance selected, note that this lattice spacing is only observed for the shell of the particle. (c) Color-coded EDS map of the particle (red: Fe-La, green O-K). (d) Exemplary NBD electron diffraction pattern with the radius in k-space indicated. (e) Intensity profile of rotational averaged NBD patterns recorded along the scan line indicated in (a). Note that for the center of the particle, the intensities of the Fe lattice spacings are prominent and for the edge/shell of the particle the Fe 2 O 3 lattice spacings are stronger. (f) Line profiles of the characteristic line intensities plotted in (c). An ideal distribution for a homogeneous and spherical particle is also plotted.
This journal is © The Royal Society of Chemistry 2020 Nanoscale Adv., 2020, 2, 4777-4784 | 4781 Paper Nanoscale Advances a spherical particle is indicated. As discussed above, the coreshell structure of the particle becomes evident with the increased concentration of Fe in the center of the particle and with the increased O content for the shell, both deviating from the ideal distribution. These results are in agreement with the XPS survey measurements and show that surface oxidation during handling could not be fully avoided. Magnetic properties of the precursor Fe 2 O 3 and hydrogen reduced Fe nanoparticles are summarized in Fig. 8. Room temperature M(H) loops for pure samples reduced at 390 C, 310 C, 270 C and 210 C are shown in Fig. 8a with the corresponding M s and H c values as a function of the size of the reduced particles, plotted in Fig. 8b. The initial g-Fe 2 O 3 powder shows M s ¼ 70 A m 2 kg À1 which is slightly lower than the bulk value of 74 A m 2 kg À1 . 31 There is a decrease in M s with particle size for the a-Fe samples from 209 A m 2 kg À1 (390 C reduction temperature) to 204 A m 2 kg À1 (210 C reduction temperature). Similar ndings have been reported by Hsu et al., 32 where the lower saturation magnetization was explained by the semi-innite nature of ne particles. The atoms on the free surface of nanoparticles possess a lower magnetization than the bulk material due to the lack of translational symmetry and broken magnetic exchange bonds on the surface. For this reason, ner particles lead to reduced M s due to the large surface area. Another contributing factor to the reduced magnetization as a function of the reciprocal particle size is also the slight surface oxidation during handling and measurement as evident from the TEM and XPS data.
Coercivity H c scales inversely with the particle size and increases more than 10-fold from 0.0069 T (390 C reduction temperature) to 0.076 T (390 C reduction temperature). The increased coercivity for smaller particles (lower reduction temperatures) is consistent with the expectation that the nucleation eld is maximum for single-domain particles. 33 For a spherical iron particle, the single-domain diameter is about 10 nm, 15 which is smaller than the experimental particle size.
Thus, a further coercivity enhancement for even ner particles can be expected and indeed H c values of 0.106 T for the mean particle size of approximately 30 nm have been reported. 32 According to the Stoner-Wohlfarth model, 34 the maximum coercivity is dened by the anisotropy eld H a ¼ 2K 1 /m 0 M s , where K 1 is the principal anisotropy constant. For iron, K 1 z 50 kJ m À3 , m 0 M s z 2.15 T and, therefore H a z 0.06 T. However, typically the measured coercivities reach only small fraction, approximately 25% of the theoretical H a which is known as the Brown's paradox. 35,36 Interestingly, as demonstrated in the discussion above, coercivity of ne Fe particles exceeds the calculated theoretical upper limitthe anisotropy eld H a . Therefore, most likely additional surface-related phenomena take place and account for the discrepancy. For example, an increase in the anisotropy constant with the reciprocal particle diameter up to 300 kJm À3 for 2 nm a-Fe nanoparticles has been reported 37 and is attributed presumably to inuence of surface effects. Also, the observed surface oxidation may play a role due to exchange coupling between the core/shell layers.

Summary
In this work, we have demonstrated that by increasing the hydrogen pressure up to 530 bar, it is possible to lower the temperature necessary for complete reduction of g-Fe 2 O 3 nanoparticles to a-Fe from 390 C down to 210 C. This significant improvement in reduction temperature was shown to be benecial for the nal particle morphology. Coalescence and sintering of the particles accompanied by surface area loss which occurs at elevated temperatures (e.g. 390 C), can be suppressed when reduction is performed at 210 C. The saturation magnetization of the reduced a-Fe nanoparticles decreases with particle size from 209 A m 2 kg À1 at 390 C reduction temperature to 204 A m 2 kg À1 at 210 C. Coercivity shows opposite behavior and even exceeds the theoretical anisotropy eld for the ne 47 nm particles. This is attributed to surface effects. TEM investigations reveal that the Fe nanoparticles are passivated with a Fe 2 O 3 layer resulting in a core-shell structure.
These ndings are also relevant for applications such as catalysis and exchange-coupled nanocomposites, where ne iron nanoparticles with high surface area are required. The presented method can be extended to other metal-oxide systems.

Conflicts of interest
There are no conicts to declare.