• Open Access

Grafting Single Molecule Magnets on Gold Nanoparticles



The chemical synthesis and characterization of the first hybrid material composed by gold nanoparticles and single molecule magnets (SMMs) are described. Gold nanoparticles are functionalized via ligand exchange using a tetrairon(III) SMM containing two 1,2-dithiolane end groups. The grafting is evidenced by the shift of the plasmon resonance peak recorded with a UV–vis spectrometer, by the suppression of nuclear magnetic resonance signals, by X-ray photoemission spectroscopy peaks, and by transmission electron microscopy images. The latter evidence the formation of aggregates of nanoparticles as a consequence of the cross-linking ability of Fe4 through the two 1,2-dithiolane rings located on opposite sides of the metal core. The presence of intact Fe4 molecules is directly proven by synchrotron-based X-ray absorption spectroscopy and X-ray magnetic circular dichroism spectroscopy, while a detailed magnetic characterization, obtained using electron paramagnetic resonance and alternating-current susceptibility, confirms the persistence of SMM behavior in this new hybrid nanostructure.

1 Introduction

With their giant and directionally-bistable magnetic moment, single molecule magnets (SMMs) are intensely investigated as core components of new molecule-based spintronic devices, in which molecular spins are read-out and manipulated using electric currents.[1, 2] In this direction, a first important step consists in stably interfacing these magnetic molecules with conducting substrates, taking advantage of the perfectly defined and chemically modifiable structure of SMMs.[3] Ultraflat metal substrates,[4] metal nanogaps,[5] graphene[6] and carbon nanotubes[7] allowed the electrical addressing of SMMs using STM methods[4] or three-terminal transport measurements.[5-7] The magnetic moment dynamics of SMMs is, however, extremely sensitive to the environment[8] and, since now, only few types of systems have proven the persistence of slow magnetic relaxation when grafted to metallic surfaces or carbon nanotubes.[9] By inserting a bis(phthalocyaninato)terbium(III) complex in a gold nanojunction, Vincent et al. found fingerprints of the tunneling dynamics of the electronic magnetic moment coupled to the nuclear spin.[5a] Moreover, some of us have demonstrated that tetrairon(III) (Fe4) propeller-shaped SMMs with a S = 5 ground spin state retain their magnetic hysteresis once chemically grafted to a gold surface.[9a],[9b] In spite of the low anisotropy barriers (Δ/kB ≈ 15 K) and subkelvin blocking temperatures, this class of SMMs exhibits the chemical and structural robustness required even for vapor-phase processing,[10] a very rare property for a polynuclear complex.[11]

We have now explored the chemical grafting of Fe4 complexes on a third type of metallic substrate, namely gold nanoparticles (Au NPs). Owing to the versatility of the surface chemistry of gold, Au NPs have been functionalized with a wide variety of molecules, ranging from simple capping agents like thiols or amines, to more complex moieties,[12] including biomolecules and drugs.[13] The rich assortment of functionalities which can be introduced together with the peculiar intrinsic properties of Au NPs have opened exciting perspectives for their application in a wide range of strategic technological areas, such as sensing,[14] nanomedicine,[13-15] and catalysis.[16] Nevertheless, only an isolated and preliminary essay with the very fragile Mn12 SMM is documented,[17] and the combination of the optical, transport, and magnetic properties of Au NPs with the quantum magnetization dynamics of SMMs is, to the best of our knowledge, unprecedented. As NPs have enhanced reactivity and are covered by surfactants, it is not straightforward to extend to NPs the deposition protocol developed for flat surfaces. On the other side a hybrid nanostructure constituted by SMMs and NPs would represent a convenient model system to investigate the interaction between SMMs and a conducting substrate. As an important advantage over flat supports, NPs can accommodate much larger quantities of SMMs by virtue of the large surface area of the material. As a consequence, a much wider spectrum of experimental techniques can be applied.

Here we show that Au NPs can be functionalized with Fe4 molecules which retain their SMM behavior in the new hybrid material. The grafting of SMMs promotes extensive aggregation of the NPs, in much the same way as with α,ω-dithiols,[18] suggesting that Fe4 SMMs can bridge Au NPs through their two sulphur-terminated ligands (see Figure 1).

Figure 1.

a) Structure of Fe4-thioctic single molecule magnet with the black arrows indicating the spins’ arrangement in the ground state (iron atoms are drawn as large light gray spheres, sulphurs as small light gray spheres, oxygen atoms as spheres, and carbons as dark gray backbone; hydrogen atoms and tert-butyl groups of the β-diketonate ligands are omitted for clarity); b) Schematic representation of the used ligand exchange procedure.

2 Results and Discussion

A single batch of several hundred milligrams of hexadecylamine (HDA)-capped Au NPs, HDA-NPs hereafter, was prepared modifying a previously reported synthetic procedure (see Experimental Section) and used for the following exchange reactions after accurate purification. HDA-NPs were chosen as starting material because, on a gold substrate, the replacement of an N-terminated ligand with a S-terminated one is a favored process, due to the strong affinity of Au for S. Moreover, HDA was individuated as the optimal ligand for the exchange reaction due to its compatibility with the SMM, despite other amines are known to provide a better grade of Au NPs monodispersity.[19] To best exploit the sensitivity of the plasmonic properties to the surface coverage of Au NPs,[20] the experimental conditions were tuned to obtain HDA-NPs with average size of about 5 nm, as shown by the TEM image reported in Figure 2a (see the Supporting Information for methods and statistics). For the functionalization of Au NPs we used a tetrairon(III) complex with a propeller-like structure, [Fe4(thioctic)2(dpm)6] (Fe4-thioctic), where Hdpm is dipivaloylmethane and H3 thioctic is a tripodal ligand obtained by esterification of (±)-α-lipoic acid (“thioctic acid”) with pentaerythritol (see Figure 1a). These cyclic disulfides derived from thioctic acid are known to interact strongly with metal surfaces[21] and NPs.[22] The two 1,2-dithiolane rings located on opposite sides of the metal core indeed promote an efficient grafting of Fe4-thioctic on ultraflat gold substrates.[23]

Figure 2.

TEM images of HDA-NPs (a), and of the same Au NPs after functionalization with ADM (b) or with Fe4-thioctic (c). Image d is a magnification of the part of figure c contained in the white rectangle.

The hybrids were then assembled by exchanging HDA ligands with Fe4-thioctic in CH2Cl2 solution (Figure 1b). Based on simple geometrical considerations reported in the Supporting Information, an ideal total coverage of the NPs surface requires 12.0% by weight of Fe4-thioctic. To ensure a complete ligand exchange we added a 2.5-fold excess of SMMs. Upon this functionalization, the red color, typical of Au NPs of this size, turned to violet-blue and the particles started to flocculate, suggesting the formation of an extended network of Fe4-thioctic-capped NPs (Fe4-NPs). This allowed the purification of the sample through repeated dispersion in CH2Cl2 by sonication and successive spontaneous precipitation over several hours. In contrast, if the sample was treated with ethanol and successively redispersed in dichloromethane, the red color of the pristine HDA-NPs was restored (see Supporting Information). As ethanol is a well-known disrupting agent for Fe4 core, as shown by X-ray absorption experiments (see below), this observation supports the hypothesis that aggregation of Fe4-NPs was caused by Fe4-thioctic molecules acting as bridges between NPs. Fe4-NPs are seen to form extended three-dimensional networks (Figure 2c) where individual NPs can be visualized only at the borders (Figure 2d). Distances between adjacent NPs are shorter than 2.5 nm, thus compatible with the presence of Fe4-thioctic molecules as linker agents.

To prove that the observed behavior is actually induced by the Fe4-thioctic linker, we repeated the same ligand exchange procedure using the methyladamantyl ester of thioctic acid (ADM), which cannot simultaneously bind to two different Au NPs having only one sulphur-terminated ligand (see the Supporting Information for the synthetic procedure). No significant color change was detected in this case and no aggregation was observed in TEM images (Figure 2b).

In Figure 3 we present the normalized plasmon resonance peaks for HDA-NPs, ADM-NPs and Fe4-NPs in CH2Cl2 solution (the complete electronic spectra showing also ligand absorptions are available in Supporting Information). HDA-NPs and ADM-NPs display sharp peaks at 522 nm and 530 nm, respectively; this relatively small red-shift is due to the different nature of the ligand donor atoms, as predicted by hard-soft theory.[24] By contrast, Fe4-NPs manifest a very broad peak with maximum around 580 nm. It is well-known that aggregation of NPs induces a low-energy shift of the plasmonic peak, which becomes more pronounced with increasing the dimension of the aggregate.[25] A remarkable broadening of the plasmon resonance peak was concomitantly observed upon addition of a cross-linking agent, like DNA, to a solution of gold NPs.[25b] The optical properties of Fe4-NPs are therefore indicative of the formation of NPs aggregates, in accordance with TEM image reported in Figure 2c.

Figure 3.

Electronic spectra (normalized at the plasmon resonance peak) of the three types of investigated NPs in CH2Cl2 solution.

The electronic spectra of both Fe4-NPs and pure Fe4-thioctic feature a common strong absorption peak around 260 nm (see Supporting Information), which suggests that the tetrairon(III) core remains intact in the hybrid material.

To exclude that this signal may originate from free Fe4 complexes in solution, reflecting an instability of the hybrids or their incomplete purification, we used proton nuclear magnetic resonance (1H-NMR). According to several works,[26] when organic molecules are grafted to NPs the NMR peaks of their protons in closest proximity to the metallic surface undergo a dramatic broadening and may even disappear completely from the spectra. Such an effect is evident if one compares the 1H-NMR spectrum of ADM and of ADM-NPs in CD2Cl2: all the signals produced by the 1,2-dithiolane ring protons (δ = 1.87–1.95 ppm and δ = 2.50–3.62 ppm) become undetectable when the ligand is bound to the metal NP (see Supporting Information). The 1H-NMR spectrum of Fe4-thioctic in CD2Cl2 is complicated by the paramagnetic nature of the metal ions. It exhibits an intense, broad peak at 10.7 ppm originating from the tBu protons of dipivaloylmethanide ligands,[27] along with very weak and broad signals. A CD2Cl2 suspension of Fe4-NPs features no detectable proton NMR signals, confirming the formation of NPs aggregates that respond as paramagnetic solids (see Supporting Information) and indicating that free SMMs are not present in detectable amounts.

To further characterize the Fe4-NPs hybrid material, we performed X-ray Photoelectron Spectroscopy (XPS) measurements on Au4f, O1s, S2p, Fe2p and N1s regions (See the Supporting Information for experimental spectra and set-up). While Au in the NPs produces an intense peak, the photoemission signals from the other elements are very weak, in agreement with the literature,[28] and almost at the limit of instrumental sensitivity. This is not surprising considering the small percentage of Fe4-thioctic expected in the samples. While not allowing a quantitative analysis, XPS data confirm qualitatively the presence of species containing S and Fe. Notice that an exceedingly weak peak was also detected in the N1s region, associated to the contribution of residual HDA ligands.

X-ray absorption spectroscopy (XAS) and X-ray magnetic circular dichroism (XMCD) have been widely employed to investigate molecules, including Fe4 complexes.[3],[9a],[9b],[11, 23, 29] The characterization was performed at the Fe L2,3 edges using a low photon flux in order to avoid radiation damage.[30] XAS spectra were measured at a temperature of 2.2 K and under a 30 kOe applied magnetic field parallel to the X-ray propagation vector. The same set-up was used to record the field dependence of XMCD signal and extract an X-ray detected magnetization curve normalized to the isotropic contribution.

XAS spectra are presented in Figure 4a for two different circular polarizations of the X-ray beam (σ and σ+). The line shape is typical for Fe4-based SMMs, which feature iron(III) ions in a distorted octahedral coordination environment. However, additional and conclusive proof of the intactness of the Fe4 core is provided by the XMCD spectrum. This is defined as the difference in absorption of X-rays with opposite circular polarization (σ – σ+), normalized by the maximum absorption signal in the isotropic spectrum ½ (σ + σ+). The maximum XMCD amplitude at 709.1 eV (ca. 40%) and the vanishing signal at 707.8 eV are considered fingerprints of the ferrimagnetic spin structure schematized in Figure 1a, where the central and peripheral iron ions provide opposing magnetic contributions.[31] Our synchrotron investigation thus indicates that Au-NPs are capped by intact SMMs. By contrast, when a dichloromethane solution of Fe4-NPs is treated with ethanol, and the resulting precipitate redissolved in dichloromethane to recover the initial plasmonic peak, a drop-cast sample of this solution shows a much weaker XMCD signal (see the Supporting Information). The field dependence of the XMCD signal at 709.1 eV was also measured at the same temperature and the results are reported in Figure 4b.

Figure 4.

a) XAS spectra recorded with different circular polarizations on Fe4-NPs at T = 2.2 K and H = 30 kOe (top) and XMCD spectrum obtained as the difference (σ – σ+). b) field dependence of the XMCD signal at 709.1 eV measured at T = 2.2 K. The grey solid line corresponds to the magnetization calculated with the anisotropy parameters extracted from EPR spectra. The right-hand scale was adjusted to achieve the best overlap between the XMCD and magnetization data.

With their large surface area, bulk samples of NPs can host a relevant amount of SMMs. Thus, a magnetometric characterization resulted to be feasible. First, samples of HDA-NPs and ADM-NPs were investigated in order to evaluate the magnetic contribution of the NPs. The magnetization curves at T = 2 K, shown in the Supporting Information, revealed an increase of the weak paramagnetic contribution when replacing HDA with 1,2-dithiolane-terminated ADM ligands. This trend was already observed in other S-functionalized small Au NPs and is attributed to a modification of Au-NPs electronic structure due to the introduction of 5d band holes upon formation of S-Au bonds.[32] As the paramagnetic signal is expected to change with the coverage of sulphur-terminated ligands, the contribution of Au NPs cannot be directly subtracted from the magnetic data of Fe4-NPs. However, the magnetization of Fe4-NPs at 2 K, reported in Supporting Information, is ca. one order of magnitude larger than in ADM-NPs, thus allowing a reliable investigation of the magnetic properties of Fe4 units in the hybrid nanostructure. From the linear part of the M vs H graph we calculated a susceptibility of 1·10−4 emu g−1 at 2 K. At this temperature, the molar χT product of pristine Fe4-thioctic is 13.5 emu K mol−1 and the mass percentage of SMM in our sample is therefore estimated as ca. 2.9(3)%. To further confirm the validity of this approach, by comparing the magnetization of Fe4-NPs and pristine Fe4-thioctic at 2 K and 50 kOe we obtained a rewardingly similar mass percentage of 3.5(3)%.

The SMM behavior of Fe4 systems is characterized by a well-defined frequency dependent out-of-phase signal in the ac magnetic susceptibility.[23]The in-phase (χ′) and out-of-phase (χ″) components of ac susceptibility were measured on a sample of Fe4-NPs at variable temperatures in the 100 Hz – 10 kHz frequency range and in both zero and 1 kOe static fields. A field of 1 kOe is in fact able to suppress quantum tunneling of the magnetization without a significant decrease in the susceptibility due to saturation effects. The results, reported in Supporting Information, have been analyzed using the Casimir and Du Pré formula to extract the relaxation time (τ) as well as the distribution width of τ through the empirical parameter α. The data in Figure 5 (circles) show that lnτ is a linear function of T−1 indicating that magnetic relaxation takes place via thermal activation. A linear fit was then performed to obtain the height of the anisotropy barrier (Δ) and the pre-exponential factor (τ0) in the Arrhenius law, τ = τ0exp(Δ/kBT). The values extracted from the fit are: (at H = 0 Oe) Δ/kB = 8.0(1) K and τ0 = 1.20(6)·10−6 s; (at H = 1 kOe) Δ/kB = 11.6(1) K, τ0 = 1.01(4)·10−6 s. Noticeably, the distribution width of relaxation times displayed in Figure 5 (triangles) is about twice broader than in crystalline Fe4-thioctic.[23]

Figure 5.

Temperature dependence of the relaxation time (τ) for Fe4-NPs in static fields H = 1 kOe (solid circles) and H = 0 Oe (open circles); the black lines represent the fits to the Arrhenius law. The upper part of the figure presents the α values extracted from magnetic data at H = 1 kOe (solid triangles) and at H = 0 Oe (open triangles).

In order to obtain information on the magnetic anisotropy responsible for slow magnetic relaxation, we recorded W-band (94 GHz) EPR spectra on solid Fe4-NPs at variable temperature (Figure 6). The Fe4-NPs were measured directly as a powder, while control spectra of microcrystalline Fe4-thioctic were measured blocking the powder in wax to avoid preferential orientation of the microcrystallites which may occur on application of magnetic field.

Figure 6.

Temperature dependence of W-band EPR spectrum of Fe4-NP and best simulation for the T = 10 K spectrum (parameters reported in the text). The upper line represents the spectrum of microcrystalline Fe4-thioctic at T = 20 K. The asterisk evidences a spurious signal of the cavity.

The spectra show the fine structure typical of a zero-field split ground state and their temperature dependence confirms the easy axis character of the investigated system.[33] Indeed, the result compares well with those obtained on a microcrystalline sample of Fe4-thioctic, suggesting a strict similarity of the anisotropy parameters in the pristine material and in the hybrid nanostructure. For a quantitative estimation of the anisotropy parameters we undertook spectral simulations[34] based on the following spin Hamiltonian:

display math

A good agreement between calculated and experimental spectra of Fe4-NPs was obtained with the following parameters: S = 5, D = −0.415(2) cm−1, E = 0.010(1) cm−1 and B40 = 1.0(1) × 10−5 cm−1 while g was fixed at 2.000(5), as expected for high-spin iron(III)-containing species. Inclusion of higher order anisotropy terms is generally necessary to correctly describe the EPR spectra of SMMs, which are intrinsically multi-spin in nature.[35] The peculiar variation of linewidth along the spectrum has been accounted for by imposing different and anisotropic linewidth parameters for different groups of resonances. The D parameter estimated from EPR spectra was employed to simulate the X-ray detected magnetization curve at low temperature. As shown in Figure 4b, the result scales very well with the XMCD data, which are unaffected by the paramagnetic contribution of the Au NPs.

Based on the above described EPR studies, the total splitting of the S = 5 manifold is |D|S2/kB = 14.9 K, to be compared to the effective barrier of about 8 K obtained by AC magnetic measurements at H = 0 Oe. This observation suggests the presence of an efficient tunnel mechanism, also confirmed by the marked increase of the barrier height when the energy degeneracy of the ±ms pairs is lifted by application of a static magnetic field. Although such an effect of the static field is quite common among SMMs, crystalline Fe4 derivatives, including Fe4-thioctic, exhibit a much weaker field effect. The pronounced tunnel efficiency in zero field could in principle be attributed to structural distortions induced by the formation of the NPs aggregates. However, a comparison with the spin Hamiltonian parameters reported for crystalline Fe4-thioctic, D = −0.430(4) cm−1 and E = 0.005(2) cm−1,[23] indicates that the zero-field splitting parameters do not change much upon grafting to the Au NPs. Differences in higher-order transverse terms cannot be ruled out at this stage, since these important sources of tunneling[35] have basically no effect on the EPR spectra of randomly-oriented samples. The observed enhancement of quantum tunneling in zero field however suggests that other mechanisms might be active, which are possibly related to the metallic nature of the substrate and require further studies. In this respect, the grafting of SMMs on NPs rather than on ultraflat surfaces seems a very convenient strategy to investigate such substrate-dependent effects, as the dynamics of the magnetization can be directly probed over a wide frequency range via traditional ac susceptometry.

3 Conclusion

In conclusion, we prepared and characterized Fe4-thioctic-capped gold NPs, a new material that combines the magnetic bistability of SMMs with the plasmonic and transport properties of gold NPs. The SMM behavior is preserved, with an increased efficiency of the tunnel mechanism that cannot be simply attributed to a distortion of cluster's geometry, caused by the intense strain inside the NPs aggregates. The metallic nature of the NPs could play a role here and it would be interesting to investigate if the excitation of localized plasmon resonance is able to affect the magnetization dynamics of the SMMs. Although the large aggregates obtained here are not suitable for transport measurements in nano-juctions, where isolated NPs are required, they might be employed in mesoscopic devices. Interestingly, our functionalization procedure can be adapted to grow the network layer-by-layer directly on the electrodes, as commonly done for di-thiol-functionalized NPs employed for gas sensing.[28d] As an appealing extension of our approach, the replacement of Au NPs with magnetic alloys may afford hybrid magnetic materials and allow the investigation of quantum effects in magneto-transport devices without requiring the manipulation of isolated SMMs.

4 Experimental Section

Synthesis: Unless otherwise stated, all reagents were purchased from Sigma Aldrich with purity at least 98%.

HDA-capped gold NPs were prepared by slight modification of literature procedures.[19, 36] In a three-neck flask, HDA (25 g) was solubilized in CHCl3 (160 mL) at 40 °C under N2 flow and vigorous stirring. When the solution became colorless, a solution of HAuCl4·4H2O (0.9 g, purchased from Strem Chemicals) in CHCl3 (5 mL) was rapidly added. The mixture, heated to reflux, assumed a red color and, after 10 min, turned to orange. A solution of borane tert-butylamine complex (0.3 g) in CHCl3 (5 mL) was then quickly added. The obtained purple-red mixture was stirred under N2 atmosphere for 1 h and cooled down to room temperature. Ethanol (340 mL) was added to give a violet solution that was centrifuged at 5000 rpm for 5 min, discarding the supernatant. The dark-red precipitate was solubilized in a small volume of CH2Cl2 (15 mL) and ethanol was added (30 mL) before centrifugation. The washing procedure was repeated four times to obtain a purple-red solution containing HDA-capped gold NPs (200 mg) in CH2Cl2 (50 mL). The average diameter of the synthesized NPs was 5.1 ± 1.2 nm, as estimated from TEM images (see the Supporting Information for statistics).

To obtain ADM-capped gold NPs, a 2.5-fold excess of ADM (41 mg) was added to a solution of HDA-capped gold NPs (60 mg) in CH2Cl2 (15 mL) and the mixture was stirred for 24 h at room temperature. The adopted purification procedure was similar to that reported for HDA-capped NPs, with ethanol replaced by methanol to decrease the solvent affinity of ADM-NPs. A purple-red solution was finally obtained.

Au NPs capped by Fe4-thioctic were obtained by introducing a 2.5-fold excess of Fe4-thioctic[23] (30 mg) in a solution of HDA-capped gold NPs (100 mg) in CH2Cl2 (25 mL). The mixture was stirred for 24 h at room temperature. In this case the NPs started to flocculate, so the purification occurred via direct precipitation from the solution over several hours; the yellow supernatant was discarded, the black precipitate dispersed again in CH2Cl2 and the washing cycle repeated six times to finally afford a colorless solution over a dark precipitate. Fe4-NPs have a low solubility in CH2Cl2, so prolonged sonication was necessary to get a violet-blue solution.

Characterization: TEM images were obtained using a CM12 PHILIPS TEM, equipped with an OLYMPUS megaview G2 camera, resolution power of 0.34 nm and emitter filament tension of 100 kV. UV-vis spectra were recorded with a JASCO V-670 spectrophotometer. XAS and XMCD characterizations were performed at the DEIMOS beamline of SOLEIL synchrotron facility (France) on a sample of precipitated Fe4-NPs deposited mechanically on a Cu foil. The UHV compatible pumped 4He optical cryomagnet of the beamline was used and absorption spectra measured in total-electron yield (TEY) detection mode[37] to guarantee an optimal detection sensitivity. For the XPS maesurements setup see Supporting Information. The static and dynamic magnetic properties were measured on powders with teflon coverage in a Quantum Design MPMS SQUID and in a Quantum Design PPMS systems equipped with the alternating current measurement option. EPR spectra were recorded using a Bruker E600 continuous-wave spectrometer with a cylindrical cavity (TE011 mode) operating around 94 GHz equipped with a split-coil superconducting magnet (Oxford instruments) and a continuous flow cryostat (Oxford CF935), to achieve temperature variation.


We thank the staff of the DEIMOS beamline and of Centro di Microscopie Elettroniche “Laura Bonzi”, CNR, Sesto Fiorentino for the experimental support. Funding from the European Research Council through the Advanced Grant “MolNanoMas” (267746), from the Italian MIUR through FIRB projects “NanoPlasMag” (RBFR10OAI0), “Nanomagneti molecolari su superfici metalliche e magnetiche per applicazioni nella spintronica molecolare” (RBAP117RWN) and “Rete ItalNanoNet” is acknowledged. The financial support from Ente Cassa di Risparmio di Firenze and Fondazione Cariplo (2010–0612) is also acknowledged. Part of this work was done using equipment funded by the Agence National de la Recherche, grant ANR-07-BLANC-0275.