III. RESULTS AND DISCUSSION

Figure 1 shows a schematic of the matrix-free superparamagnetic nanocomposite. Fe/Fe_xO_y core–shell nanoparticles are first formed via solvothermal synthesis using TOP and OLA as surfactants, with 1-ODE as the high boiling point solvent. Solvothermal synthesis was chosen as it is known to give significantly better control over the synthesis of iron nanoparticles when compared to other techniques, e.g., reduction of iron salts.^{Reference Huber8} We have previously shown that magnetic nanoparticles with a uniformly spherical shape and an incredibly tight size distribution could be formed reproducibly and predictably in an extended LaMer synthesis.^{Reference Vreeland, Watt, Schober, Hance, Austin, Price, Fellows, Monson, Hudak, Maldonado-Camargo, Bohorquez, Rinaldi and Huber29} Here, continuous addition of the reaction precursor leads to a steady state growth regime which eliminates Ostwald ripening, a known source of polydispersity. This method was then extended to grow strongly magnetic zero-valent iron nanoparticles, in what has been named the reversible agglomeration mechanism.^{Reference Bleier, Watt, Simocko, Lavin and Huber27} In this approach, magnetic nanoparticles nucleate and grow until a critical susceptibility is reached. At this point, magnetic dipole interactions overcome dispersive forces leading to a magnetic agglomeration event. This is followed by the precipitation of the agglomerated nanoparticles and a phase change in the system, giving an example of true thermodynamic reaction control in nanoparticle synthesis. As the onset of agglomeration is directly a result of an increase in particle susceptibility, nanoparticle growth is arrested within a very small size window, yielding a strongly magnetic nanoparticle ensemble with a tight size distribution. With the continued addition of precursor, a second nucleation event can then occur, leading to subsequent growth and a second agglomeration/precipitation event. By continuously adding precursor, multiple reversible agglomeration steps can be performed indefinitely, meaning the mechanism is unique in its scalability. This was an important aspect when choosing this synthetic approach in the current study as for effective application of a useful nanocomposite, greater than typical lab-scale quantities need to be produced. The nanoparticle product can then be resuspended in most common organic solvents through the addition of sufficient energy, e.g., heat or sonication, confirming the reversibility of the mechanism. In the synthesis of iron nanoparticles, solvents and reagents are typically dried and degassed to remove oxygen and avoid oxidation. This is reasonable for research scale quantities, however when dealing with the larger quantities required for scale-up, this can become impractical. Therefore, the following nanoparticle products were all formed using the solvents and reagents as received directly from the supplier.

FIG. 1.

Schematic showing the as-synthesized matrix-free superparamagnetic nanocomposite. Fe/Fe_xO_y core–shell magnetic nanoparticles are synthesized using TOP and OLA as surfactants. A ligand exchange procedure is then performed in situ, attaching an alkyl chain diamine to the nanoparticle surface. The amine expressed superparamagnetic nanoparticles are then cross-linked with a triepoxide, n,n-diglycidyl-4-glycidyloxyaniline (red). By cross-linking directly to a covalently bound stabilizer species, we substantially reduce the amount of organic fraction in the nanocomposite when compared to polymer matrices and increase the packing fraction of the magnetic component. The interparticle distance can then be tuned easily by varying the alkyl chain length of the diamine.

The results of scale-up synthesis of superparamagnetic nanoparticles are shown in Fig. 2. Figure 2(a) shows 1 g of Fe/Fe_xO_y core–shell nanoparticles synthesized using the reversible agglomeration mechanism, separated from solution by the application of an external magnetic field. Ensemble SAXS size analysis is given in Fig. 2(b), showing the raw data overlaid with the model fit and corresponding residuals. The model was applied assuming noninteracting nanoparticles, with a spherical shape and Gaussian size distribution as observed in TEM [Fig. 2(c)]. The model fits well with no significant deviations and with low residuals, indicating that the nanoparticles are well suspended in solution. SAXS gave an average size of 15.2 ± 1.2 nm for the nanoparticle ensemble. The resulting Gaussian curve of the size distribution is given in Fig. S1(a). TEM analysis showed the nanoparticles to be uniformly spherical, with an iron oxide shell 2.5 nm thick which forms after the nanoparticle product is exposed to air post-synthesis (inset). The iron oxide shell could be indexed to the (220) and (311) planes of the cubic spinel phase, a characteristic of either magnetite (Fe₃O₄) or maghemite (γ-Fe₂O₃) (see Supplementary Material Fig. S2). Unfortunately, TEM is not able to distinguish between the two phases.

FIG. 2.

(a) 1 g of Fe/Fe_xO_y core–shell nanoparticles synthesized using OLA and TOP as surfactants, attracted to an applied external magnetic field. (b) Ensemble SAXS analysis performed on an aliquot of nanoparticles measured an average diameter of 15.2 ± 1.2 nm. (c) Corresponding TEM image. (Inset) High-resolution TEM image showing partial surface oxidation has occurred during the sample preparation step.

The magnetic properties of the 15.2 ± 1.2 nm Fe/Fe_xO_y core–shell nanoparticles were then investigated using vibrating sample magnetometry (VSM), with the results given in Fig. 3. Saturation magnetization, M _sat, of the nanoparticles was measured to be 96 Am²/kg at 50 K [Fig. 3(a)]. This is larger than the value expected for magnetite (92 Am²/kg) or maghemite (76 Am²/kg), indicating the presence of a zero-valent iron core. We then fit the M _sat curve using the Langevin equation, with the fit shown in Fig. 3(a). The fit gave a magnetic diameter of 10.2 nm, which is consistent with what is observed for the zero-valent iron core observed in TEM.^{Reference Unni, Uhl, Savliwala, Savitzky, Dhavalikar, Garraud, Arnold, Kourkoutis, Andrew and Rinaldi30} HRTEM (Fig. S1) did not observe lattice spacings of the iron core characteristic of the bcc crystal structure of α-Fe, indicating that the nanoparticles possess either an amorphous core or a distorted lattice.^{Reference Monson, Ma, Stevens, Lavin, Leger, Klimov and Huber31} Both of which are known to reduce M _sat when compared to α-Fe. An additional factor reducing the M _sat may be the strong iron–phosphine interaction introducing a magnetically dead layer at the surface.^{Reference Huber8} ZFC–FC experiments were then performed to identify the blocking temperature, T _B, and the corresponding onset of superparamagnetism [Fig. 3(b)]. The measurement time chosen was 100 s, as is typical for DC magnetometry.^{Reference Concas, Congiu, Muscas and Peddis32} The peak is relatively broad and centered with a T _B of 233 K, which is significantly reduced when compared to zero-valent iron nanoparticles of a similar size, due to the presence of the oxide shell.^{Reference Watt, Bleier, Austin, Ivanov and Huber33} Here, oxidation is beneficial as it ensures that the nanoparticles are superparamagnetic at room temperature. The broadness of the peak indicates that there may be some interparticle interactions occurring in the nanoparticle ensemble, which is a reasonable assumption as the product is measured as an agglomerated magnetic pellet, as is produced from the reversible agglomeration mechanism.^{Reference Bleier, Watt, Simocko, Lavin and Huber27} At higher frequencies, hysteresis may be observed, and understanding their performance under such switching environments will be essential for their eventual application.^{Reference Yun, Kim, Paik, Meng, Jo, Kikkawa, Kagan, Allen and Murray34}

FIG. 3.

Magnetic characterization of the 15.2 ± 1.2 nm Fe/Fe_xO_y core–shell nanoparticles performed using a VSM. (a) Magnetic moment versus field plot shows that the nanoparticles have a saturation magnetization, M _sat, of 96 Am²/kg at 50 K. The M _sat curve was fit using the Langevin equation, indicating that the nanoparticles had a magnetic diameter of 10.2 nm. (b) Zero field, cooled-field cooled (ZFC–FC) plot showing a blocking temperature of 233 K, which indicates that the nanoparticles are superparamagnetic at room temperature.

The reversible agglomeration mechanism is characterized by the nucleation and growth of nanoparticles followed by agglomeration and renucleation steps, a cycle that can be repeated indefinitely to reach large scale synthesis.^{Reference Bleier, Watt, Simocko, Lavin and Huber27} Therefore, to calculate the number of reversible agglomeration cycles needed to synthesize 1 g of Fe/Fe_xO_y nanoparticles, we first determine the critical nuclei radius at the reaction temperature of 523 K. The critical nuclei radius is defined by^{Reference Park, Joo, Kwon, Jang and Hyeon35}

$${r_{\rm{c}}} = - {{2\gamma } \over {{G_{\rm{V}}}}}\quad ,$$

where γ is the surface free energy of iron, which at 523 K is equal to 2.3 J/m².^{Reference Schonecker, Li, Johansson, Kwon and Vitos36,Reference Grochola, Russo, Yarovsky and Snook37} G _v is the Gibbs energy per volume of iron, which in the molten state of Fe(0) nuclei at high temperatures is equal to −1.57 × 10⁹ J/m³.^{Reference Park, Joo, Kwon, Jang and Hyeon35} As the nanoparticles are being formed under an inert atmosphere and the evolution of CO gas by the decomposition of Fe(CO)₅ creates a reducing environment, the nuclei can be assumed to be truly zero-valent iron. Therefore, the critical nuclei radius for Fe(0) at 523 K is 2.93 nm. If we assume that the first drop of precursor nucleates and all iron is consumed by surviving nuclei, then we can calculate how much additional precursor is needed to grow the nanoparticles to their maximum size, before agglomeration occurs. Each individual drop has a volume of ∼10 μL.^{Reference Tripp, Good, Motta, Kass and Murphy38} Therefore, as we are introducing a 3:1 ODE:Fe(CO)₅ precursor mixture, each drop contains 2.5 μL of Fe(CO)₅, which equates to 1.06 mg of Fe(0). Taking a nuclei size of 2.93 nm and a density of 7874 kg/m³, we can calculate that each drop contains 1.28 × 10¹⁵ nuclei. Then, for a 1 g reaction, the maximum size before agglomeration is 15.2 nm. Each maximum-sized nanoparticle is therefore calculated to have a mass of 1.45 × 10⁻²⁰ kg. Then, if we consider that each of the 1.28 × 10¹⁵ nuclei has grown to this maximum size, we can calculate that one reversible agglomeration cycle has consumed 1.85 × 10⁻⁵ kg, or 0.0185 g, of Fe(0). Taking the above assumptions and knowing that the final mass of Fe(0) nanoparticles formed is 1 g, we can calculate that a minimum of 54 cycles of reversible magnetic agglomeration have occurred.

Figure 4 shows the results from scaling-up the reversible agglomeration mechanism by an order of magnitude, to 10 g. At this large scale, a peristaltic pump was needed to deliver sufficient quantities of Fe(CO)₅ precursor solution to the reaction flask [Fig. 4(a)]. Likewise, the size of the reaction flask dictates that magnetic stirring is not sufficient, and an overhead stirring setup was required. This introduces challenges with controlling an air-free environment; however, this was achieved by maintaining a positive pressure of flowing nitrogen throughout the reaction. Specific Tygon tubing was chosen with low gas permeability and high chemical resistance, to protect against premature decomposition of the Fe(CO)₅. Figure 4(b) shows 10 g of Fe/Fe_xO_y core–shell nanoparticles synthesized using the reversible agglomeration mechanism attracted to an external magnetic field. SAXS analysis gave an average diameter of 13.7 ± 2.5 nm for the nanoparticle ensemble [Fig. 4(c)]. Again, the raw data were model fit assuming a spherical shape and Gaussian size distribution [Fig. S1(b)]. TEM analysis showed a slight loss of size and shape control, which can be attributed to thermal and concentration gradients associated with the larger reaction vessel required to carry out the synthesis (Fig. S3). It is also important to note here that the slight deviation from a spherical shape will lead to an increase in the calculated size distribution from the SAXS analysis.^{Reference Li, Senesi and Lee39} We can then apply the same calculations as above to determine the number of reversible agglomeration cycles needed to synthesize 10 g of 13.7 nm nanoparticles. We calculate that 1.36 × 10⁻⁵ kg of Fe(0) is required for each cycle, therefore the 10 g scale-up reaction is the result of a minimum of 735 cycles of reversible magnetic agglomeration. The number of cycles needed has increased by more than an order of magnitude due to the reduction in overall nanoparticle size. Despite the slight loss of size and shape control, the magnetic properties of the ensemble are expected to be uniform as arresting of nanoparticle growth by magnetic agglomeration is defined by particle susceptibility, not shape.^{Reference Bleier, Watt, Simocko, Lavin and Huber27}

FIG. 4.

(a) Experimental setup to increase mass output to 10 g; a peristaltic pump is required to deliver such large volumes of iron precursor. (b) 10 g of Fe/Fe_xO_y nanoparticles attracted to an external magnetic field. (c) SAXS analysis gave an average size of 13.7 nm ± 2.5 nm.

The 15.2 ± 1.2 nm Fe/Fe₃O₄ core–shell nanoparticles formed in the 1 g reaction were then used as the magnetic fraction for the formation of a matrix-free superparamagnetic nanocomposite. The nanoparticles are subjected to a ligand exchange procedure in situ, replacing the monoamine OLA and TOP with an alkyl-diamine ligand (1,6-diaminohexane). We fully expect the amine on the surface to be replaced by adsorption dynamics, whereas the TOP has previously been shown to be replaced by ligands with amine functionality.^{Reference Xu, Qin, Palchoudhury and Bao40} The ligand exchange procedure means the nanoparticles now express an amine functionality into the environment, which is known to be reactive toward cross-linking epoxides. The aminated nanoparticles are then reacted with a triepoxy (n,n-diglycidyl-4-glycidyloxyaniline) in chloroform. The solvent is subsequently removed to give a viscous, workable liquid that can then be cured at elevated temperatures to yield a cross-linked epoxy network characterized by covalent linkages between the magnetic nanoparticles. By crosslinking directly to a species that is covalently bound to the nanoparticle surface, the amount of organic fraction can be reduced substantially when compared to polymeric matrices. This allows for a significant increase in the magnetic fraction, optimizing the magnetic properties of the resulting nanocomposite. This approach also means that the interparticle spacing can be easily tuned, with the choice of alkyl chain length directly determining the distance between particles.

To confirm that the nanoparticles are undergoing a ligand exchange procedure at the surface, we performed ex situ experiments on the Fe/Fe_xO_y core–shell nanoparticles to mimic the conditions found during nanocomposite formation. To do this, an aliquot of purified nanoparticles were reacted with a solution containing 1,6-diaminohexane, in chloroform. The aminated nanoparticles were then washed thoroughly to remove any excess ligand not covalently bound to the surface of the particles. We then introduced a small quantity of fluorescamine, which is a spiro compound that forms highly fluorescent pyrrolinones upon reaction with primary and secondary amines.^{Reference Nakamura and Tamura41,Reference Eastwood, Fernandez, Yoon, Sheaff and Wai42} Following the reaction with fluorescamine, the nanoparticles were washed thoroughly with hexane and methanol using magnetic separation to remove any unreacted fluorescamine. The nanoparticles were then measured using a spectrofluorometer with an excitation wave length of 390 nm. An emission peak at 468 nm was observed, which is a characteristic of the as-formed pyrrolinones (Fig. S4), clearly indicating the presence of covalently bound diamine on the surface of the nanoparticles.^{Reference Gore43}

To characterize the nanocomposite, we carried out TGA along with DSC, with the results given in Fig. 5. TGA showed that our nanocomposite possessed 62 wt% of Fe/Fe_xO_y nanoparticles, significantly larger than what has previously been reported for magnetic nanoparticles in an epoxy system.^{Reference Puig, Hoppe, Fasce, Perez, Pineiro-Redondo, Banobre-Lopez, Lopez-Quintela, Rivas and Williams44} This value is also significantly larger than that has previously been prepared by covalent linkage to a nanoparticle surface; Mochalin et al. incorporated up to 14 wt% of nanodiamonds into a covalently linked epoxy-amine network in a similar fashion.^{Reference Mochalin, Neitzel, Etzold, Peterson, Palmese and Gogotsi17} In the TGA/DSC given in Fig. 5, we observe a number of different features corresponding to the various components that make up the nanocomposite. At around 210 °C, we observed an endothermic peak in the DSC which corresponds to boiling of 1,6-diaminohexane. TGA shows that this fraction accounts for 7 wt% of the nanocomposite, where the slightly higher temperature indicates that this is most likely uncrosslinked diamine that is bound to the nanoparticle surface. At 300 °C, we observe a strong exothermic peak in the DSC, which can be attributed to the high temperature fully completing the curing of the epoxy network.^{Reference Kessler45} This is followed by a large mass loss observed in the TGA between 300 and 460 °C, which can be attributed to the complete removal of the organic fraction (a further 31 wt%). From 460 to 620 °C, a number of endothermic processes can be observed in the DSC which can be assigned to the evaporation of molecular oxygen from the oxide shell of the nanoparticles. The peak shape matches well with previous observations of iron oxide reduction in alumina.^{Reference Gao, Shen, Hsia and Chen46} Here, the peak at 460 °C, the shoulder on the peak at 600 °C and the peak at 600 °C correspond to the reduction profile; α-Fe₂O₃ → Fe₃O₄ → α-Fe. The peaks are shifted to slightly lower temperatures than previously reported due to the high curvature of the 15.2 nm nanoparticles.^{Reference Gao, Shen, Hsia and Chen46}

FIG. 5.

TGA and DSC analysis of the cured nanocomposite formed using 1,6-diaminohexane.

The cured nanocomposite was then characterized using VSM magnetometry, with the results given in Fig. 6. Saturation magnetization, M _sat, of the nanocomposite cured using 1,6-diaminohexane was measured to be 58 Am²/kg at 50 K. Initial mass susceptibility of the nanocomposite was observed to reduce from 1.0 × 10⁻³ m³/kg for the Fe/Fe_xO_y core–shell nanoparticles to 5.2 × 10⁻⁴ m³/kg for the cured nanocomposite. Using the known mass percentages from TGA, we can then calculate the saturation magnetization of the magnetic nanoparticles themselves in the cured composite, giving a value of 76 Am²/kg [Fig. 6(b)]. The drop in M _sat when compared to the as-synthesized nanoparticles indicates that there has been an increase in the magnetically dead layer at the surface. This is most likely due to a further increase in the thickness of the oxide shell, although other factors may be present. Iron catalysts are known to be active in epoxide ring opening transformations, which may lead to irreversible binding at the surface of the nanoparticle and the introduction of a magnetically dead layer.^{Reference Bolm, Legros, Le Paih and Zani47} While we fully expect the ligand shell to prevent access of the epoxide to the surface, the high liability of the 1,6-diaminohexane, due to the relatively short 6 carbon chain, may allow some finite residence time. To investigate whether the magnetically dead layer has increased during nanocomposite formation, we fit the M _sat curve in Fig. 6(b) with the Langevin equation. We calculated that the magnetic diameter had decreased from 10.2 to 9.5 nm.^{Reference Unni, Uhl, Savliwala, Savitzky, Dhavalikar, Garraud, Arnold, Kourkoutis, Andrew and Rinaldi30} This indicates an increase in the thickness of the oxide shell from 2.5 to 2.9 nm, which also accounts for the overall reduction in initial mass susceptibility as iron oxides possess a lower susceptibility than zero-valent iron.

FIG. 6.

VSM magnetometry of the as-cured nanocomposite, compared to the as-synthesized Fe/Fe_xO_y core–shell nanoparticles. The nanocomposite has an M _sat of 58 Am²/kg, compared to 96 Am²/kg for the nanoparticles. The initial mass susceptibility has decreased from 1.0 × 10⁻³ m³/kg to 5.2 × 10⁻⁴ m³/kg. (b) M _sat of the magnetic fraction of the nanocomposite only, giving a reduced value of 76 Am²/kg. The M _sat curve was fit with the Langevin equation which indicates that the magnetic diameter of the nanoparticles has reduced from 10.2 to 9.5 nm during the curing process.

To investigate the influence on interparticle spacing on the magnetic properties of the superparamagnetic nanocomposite, we then substituted 1,6-diaminohexane with longer alkyl chain diamines, namely 1,8-diaminooctane and 1,12-diaminododecane. Saturation magnetizations, M _sat, for the new nanocomposites were measured to be 60 Am²/kg and 62 Am²/kg for 1,8-diaminooctane and 1,12-diaminododecane, respectively. These values are similar to what was observed with 1,6-diaminohexane albeit slightly higher. The slight increase in M _sat may be due to the lower water solubility values observed as the alkyl chain length of the diamine increases, which reduces the extent of oxidation during nanocomposite preparation.

We then performed ZFC experiments to compare the onset of superparamagnetism in various nanocomposites, with the results shown in Fig. 7. The Fe/Fe_xO_y core–shell nanoparticles possessed a blocking temperature, T _B = 233 K. In comparison, the as-cured nanocomposites possessed blocking temperatures of T _B = 147 K, 142 K, and 146 K for 1,6-dimainohexane, 1,8-diaminooctane, and 1,12-diaminododecane, respectively. The reduction in T _B is consistent with an increase in the size of the oxide shell on the nanoparticles. The lack of variation in the blocking curves indicates that increasing the alkyl chain length from C6 to C12 still does not significantly change the particle interaction behavior. In fact, this indicates that an alkyl chain length as short as 6 carbons results in good separation of the magnetic fraction, as any magnetic interactions present in the system tends to drive T _B higher.^{Reference Kin, Kura and Ogawa48} The well-defined blocking temperatures located below room temperature indicate that the nanocomposite is still superparamagnetic and there is no evidence for the formation of ferromagnetic domains.

FIG. 7.

ZFC curves performed using VSM magnetometry. The Fe/Fe_xO_y nanoparticles show a broad peak with a blocking temperature, T _B, of 233 K indicating that they are superparamagnetic at room temperature. When cured into the superparamagnetic nanocomposite, we observe a reduction in T _B to 147 K, 142 K, and 146 K for nanocomposites formed using 1,6-diaminohexane, 1,8-diaminooctane, and 1,12-diaminododecane, respectively.