Correlation between surface chemistry and magnetism in iron nanoparticles

To shed light on the factors governing the stability and surface properties of iron nanoparticles, a series of iron nanoparticles has been produced by hydrogenation of two different iron amido complexes: the bis[bis(trimethylsilyl)amido] Fe(ii), [Fe(N(SiMe3)2)2]2, and the bis(diphenylamido) Fe(ii), [Fe(NPh2)2]. Nanostructured materials of bcc structure, or nanoparticles displaying average sizes below 3 nm and a polytetrahedral structure, have been obtained. Depending on the synthesis conditions, the magnetization of the nanoparticles was either significantly lower than that of bulk iron, or much higher as for clusters elaborated under high vacuum conditions. Unexpectedly, hydrogenation of aromatic groups of the ligands of the [Fe(NPh2)2] precursor has been observed in some cases. Confrontation of the experimental results with DFT calculations made on polytetrahedral Fe91 model clusters bearing hydrides, amido and/or amine ligands at their surface, has shown that amido ligands can play a key role in the stabilisation of the nanoparticles in solution while the hydride surface coverage governs their surface magnetic properties. This study indicates that magnetic measurements give valuable indicators of the surface properties of iron nanoparticles in this size range, and beyond, of their potential reactivity.


List of the samples reported in this work
: Reaction conditions for each sample. Duration (48h), temperature (150°C) and hydrogen pressure (3 bar) were identical for all samples. PPO = polydimethylphenylene oxide, HMDS = hexamethyldisilazane

Potential by-products formed upon hydrogenation of the two iron precursors
Scheme S1: tentative schemes for the hydrogenation of a) [Fe[N(SiMe 3 ) 2 ] 2 ] 2 and b) [Fe(NPh 2 ) 2 ] 2 complexes showing the potential by-products formed that could play the role of ligands at the surface. The Fe atoms thus generated will nucleate and form the NPs. Hydrogen potentially dissociates (more or less easily depending on the ligands present) into hydrides as in c) ; ligands are not drawn at the surface of the NP in this generic scheme for clarity. It is noteworthy that all steps a (or b) and c can occur simultaneously and that the amines can also dissociate at the surface (activation of the NH bond) leading to surface amido ligands as described in Figures SI19. 3 Figure S1: WAXS diagrams in reciprocal space in comparison with the bcc-Fe data (PDF-01-071-4648).

Magnetic measurements
Magnetic measurements were performed with a Vibrating Sample Magnetometer (VSM, Quantum Device PPMS Evercool II). VSM studies were carried out on compact powder samples that were prepared and sealed under argon atmosphere. Magnetization measurements were normalized by the Fe mass within the measured sample, determined by inductively coupled plasma-optical emission spectrometry (ICP-OES). Magnetization values (M) are thus given in A.m 2 /kg Fe . The average magnetic moment per Fe atom ( Fe ) can be calculated as follows: where M mol is the molar mass of Fe (55.845 10 -3 kg/mol), N a = 6.022 10 23 atoms/mol is the Avogadro number, and  B = 9.27 10 -24 J.T -1 . The magnetic moment per atom can be directly compared with theoretical data. We estimate that the total error is around 5% as indicated in Table  1.
For the low temperature measurements, magnetization curves were recorded after a field cooling (FC) from 300K down to 2.5 or 5K, under a magnetic field of 5T. These hysteresis loops were compared with those recorded after a zero field cooling (ZFC) procedure (cooling from r.t. down to the low temperature in the absence of magnetic field). This procedure should unveil exchange phenomena which occur in the case of ferromagnetic metallic Fe / ferrimagnetic Fe oxide interface. Such an interface may induce an asymmetric hysteresis loop at low temperature, when measured after a FC procedure. We didn't observe such phenomenon on any of the samples studied, which allowed concluding that the NPs are metallic.
Data were analyzed following the procedures described in reference ( 1 ) using the fitting procedure of the low temperature ZFCFC, or by fitting the isothermal magnetization curves in the superparamagnetic regime. Either procedure allowed estimating the magnetic NP size and the effective magnetic anisotropy.
at 5K after FC ( Figure S8) and a coercive field of 19mT at 5K. The NPs displayed a superparamagnetic behaviour with a blocking temperature below 5K. Analysis of the zero field cooled / field cooled (ZFC/FC) magnetization recorded at 10mT evidenced a Curie Weiss behaviour, with = -3K indicative of weak dipolar interactions. The isothermal magnetization curves were analyzed with modified Langevin functions, using a NP diameter centered at 1.5nm, and an effective magnetic anisotropy K eff = 6.3 10 5 J.m -3 (See Table 1, and Figure S8). Inset showing the variation of the inverse of the magnetization versus temperature. In green, fit of the low temperature data with a simple Curie-Weiss law. Bottom left: hysteresis loops measured at 5K after FC; Bottom right: magnetization curves recorded at 10, 25, 50, 100, 200 and 300K (dots), showing a superparamagnetic behaviour, and fit of the data (solid lines).

Sample 2
The main features of the magnetic properties are shown in Figure S9 and reported in Table 1. For this nanostructured iron powder of bcc structure, the hysteresis cycle recorded at 5K after a FC displayed only a very small coercive field (13mT) and a magnetization of 205 A.m 2 .
at 5T. Note that this hysteresis loops is symmetric, which confirms the absence of Fe oxide domains. Figure S9 : Hysteresis cycle recorded from sample 2 (inset: enlargement showing the limited coercivity of the sample). Low temperature data were collected after a field cooling from 300K down to 5K under a magnetic field of 5T.

Sample 3
Magnetic measurements evidenced a superparamagnetic behaviour ( Figure 10) with a blocking temperature at 12.2K. The variation of the inverse of the magnetization against the temperature followed a straight line crossing the x axis at -4.5K. This value is indicative of limited magnetic dipolar couplings between the nanoparticles, in agreement with their good dispersion in the polymer matrix. A good fit of the ZFC/FC curve could be achieved taking into account a narrow log normal size distribution centred on a diameter of 2.2 ± 0.16 nm The value of the effective magnetic anisotropy constant was extracted from this fit : K eff = 3.8 10 5 J.m -3 . The hysteresis cycle ( Figure S10) recorded at 2.5K after a FC procedure from 300K in a magnetic field of 5T was symmetrical and showed a relatively large coercive field (103mT) for Fe nanoparticles. This hysteresis loop exactly superimposed on the one recorded after a ZFC sequence, thus confirming the metallic ferromagnetic character of the iron nanoparticles. After correction of the diamagnetic contribution from the polymer matrix, the value of the saturation magnetization measured at 5T (214 A.m 2 . ) is very close to that of bulk iron (2.22μ B = 222 A.m 2 . at 5K).  (Table 1) according to a ZFC/FC procedure evidenced a superparamagnetic behaviour ( Figure S11) with a blocking temperature at 11.3 K. The variation of the inverse of the magnetization against the temperature followed a straight line crossing the x axis at 7K indicative of limited magnetic dipolar couplings between the nanoparticles in agreement with the good dispersion of the nanoparticles in the polymer matrix as for sample 3. A good fit of the ZFC/FC curve could be achieved taking into account a narrow log normal size distribution centred on a diameter of 1.7 ± 0.15 nm, leading to a K eff value of 7.9 10 5 J.m -3 . The hysteresis cycle recorded at 2.5K was symmetrical and showed a small coercive field (37mT). Hysteresis loops recorded after ZFC and FC procedures superimposed, thus confirming the metallic character of the NPs. After correction of the diamagnetic contribution from the polymer matrix, the value of the saturation magnetization could be determined: 280 A.m 2 .

Sample 4 Magnetic measurements
(at 5T). This value is far above that of bulk iron.

Sample 5
The low temperature hysteresis loop confirmed the metallic character of the NPs. The saturation magnetization could be determined at 249 A.m 2 .kg Fe -1 (at 5T). This value is above the bulk iron.
Magnetic measurements (Table 1) according to a ZFC/FC procedure evidenced a superparamagnetic behaviour ( Figure S12) with a blocking temperature at 9.1 K. The variation of the inverse of the magnetization against the temperature followed a straight line crossing the x axis at -12 K indicative of some dipolar magnetic couplings between the nanoparticles. Dipolar interactions are here too strong to apply the ZFCFC fitting procedure. The fitting of the isothermal magnetization curves in the superparamagnetic regime allowed us to determine the average particle size 1.7 nm and the effective magnetic anisotropy K eff = 8 10 5 J.m -3 . Figure S12 : Magnetic investigation of sample 5. Top: ZFC/FC curves recorded at 2.5mT (in black).
Inset showing the variation of the inverse of the magnetization versus temperature. In green, fit of the low temperature data with a simple Curie-Weiss law. Bottom Left: hysteresis loops measured at 5K measured after FC. Bottom right: magnetization curves recorded at 25, 50, 100, 200 and 300K (dots), showing a superparamagnetic behaviour, and fit of the data (solid lines).

Sample 6
The magnetic properties were typical of metallic Fe (symmetric hysteresis loops) with a saturation magnetization of 210 A.m 2 .
, a value close to that of bulk iron as reported in Table 1 and Figure  S13. Note that this nanomaterial displayed a hysteretic behaviour at r. t. as a result of the high packing density. Figure S13 : Magnetic investigation of sample 6. Hysteresis loops measured at 5K (after FC) and 300K.

Sample 7
The magnetization measured at 5K (212 A.m 2 . ) was very near the bulk one (see Figure S14). Note that the hysteretic behaviour vanishes at r. t.. Figure S14: Magnetic investigation of sample 7. Hysteresis loops measured at 5K (after FC) and 300K.    Figure  S16). On the contrary to the bare Fe 91 case, the lower the CN, the higher the atomic magnetic moment. It can be understood in terms of interatomic distances between the central atom and its neighbours (see Figure S18). Figure S18. Histogram of the interatomic distances between the central atom and its nearest neighbours in the calculated β-Mn-type phase of iron and in the Fe 91 cluster. The shapes of coordination polyhedra are also shown. Atoms around CN=12 iron atoms exhibit the same distorted icosahedron structure. The higher magnetic moment in the bulk is probably due to a lower packing efficiency, as evidenced by the histogram (plain black line, second peak at 2.6 Å vs. plain red line, second peak at 2.5 Å). The CN=14 coordination polyhedra differ in Fe 91 and in the bulk. And the difference in the magnetic moment seems also related to the packing efficiency, which is weaker for the bulk (dashed black line) than for the cluster (dashed red line), in line with a higher magnetization in the bulk for such site.