Magnetic Memory from Site Isolated Dy(III) on Silica Materials.

Achieving magnetic remanence at single isolated metal sites dispersed at the surface of a solid matrix has been envisioned as a key step toward information storage and processing in the smallest unit of matter. Here, we show that isolated Dy(III) sites distributed at the surface of silica nanoparticles, prepared with a simple and scalable two-step process, show magnetic remanence and display a hysteresis loop open at liquid 4He temperature, in contrast to the molecular precursor which does not display any magnetic memory. This singular behavior is achieved through the controlled grafting of a tailored Dy(III) siloxide complex on partially dehydroxylated silica nanoparticles followed by thermal annealing. This approach allows control of the density and the structure of isolated, "bare" Dy(III) sites bound to the silica surface. During the process, all organic fragments are removed, leaving the surface as the sole ligand, promoting magnetic remanence.

The pressure to increase information storage densities has grown tremendously during recent decades, which necessitates downsizing storage elements. Ultimately, the size of such an element is limited to a single atom. In that case, the magnetic moment originating from the atom’s quantum mechanical spin could be used to encode bits of information. This goal, however, requires sufficient temporal stability of the magnetic moment, which is manifested in the appearance of magnetic remanence, i.e., the capability of a magnetic system to retain its magnetic moment after removal of the external, magnetizing field.

In this context, the discovery of single-molecule magnets (SMMs) has been a major breakthrough to envision data storage at the nanometer scale. SMMs are unique molecules in which the magnetic moment resists reorientation[1] in the absence of a magnetic field. The guideline of most studies on SMMs has been the quest for the most stable nanomagnet, i.e., the most coercive at the highest temperature.[2] However, most SMMs are produced and studied in the condensed crystalline phase in which the single magnetic molecules cannot be addressed one-by-one. To solve this problem, several strategies have been investigated in recent years aiming at the isolation of atomic scale entities presenting magnetic remanence: heterogenization (surface immobilization) of SMMs[3] by grafting,[4−8] surface functionalization with organic ligands,[7,9] physisorption on a support,[10−12] supramolecular interactions,[13] or encapsulation[14,15] (Figure A,B), for example. However, these strategies present significant challenges. In particular, it is difficult to preserve the magnetic remanence behavior upon heterogenization as the SMM and the surface are produced separately and then assembled together. In other words, the properties of magnetic entities at the surface differ from the properties in the bulk or of individual molecular entities. Most of the SMMs are based on transition metals or f-elements and result from the combination of two key parameters: a large spin and a significant magnetic anisotropy. This anisotropy is a property of the metal center used, and its magnitude can be significantly modulated by its coordination sphere. Hence, heterogenization, which often changes the coordination sphere of the metal center, can significantly alter the required magnetic properties.

Figure 1

Immobilization strategies of single-molecule and single-atom magnets. (A) Immobilization of single-molecule magnets by grafting (a), through surface decoration with organic ligands (b), or by physisorption on the surface (c). (B) Immobilization on carbon nanotubes by supramolecular interaction (a) or encapsulation (b) (part of the nanotube wall is not represented for clarity). (C) Isolated atoms on flat metallic surface protected by an insulating decoupling layer. (D) Our strategy for immobilization of ions on metal oxide surfaces, based on a grafting step (a) and a thermolytic step (b).

As illustrated recently, single-atom deposition directly onto the surface is the ultimate size limit.[15] The magnetic remanence for single atoms of Ho on MgO/Ag(100) persists up to 30 K (Figure C).[16] This strategy showed unprecedented magnetic properties, but has limited thermal stability (the atoms being immobile only up to 50 K). Heterogenization on atomically flat surfaces has generated significant interest, as STM techniques enable manipulation of individual molecules or atoms. However, this approach requires complex techniques for the characterization of the materials. The development of strategies to rapidly screen the magnetic properties of the materials using conventional magnetometric techniques is thus highly desirable, but these techniques often fail when applied to flat surface analogues because of the small amount of magnetic material, which is only present at the surface. In that respect, we have demonstrated that the surface chemistry of a silica layer on passivated silicon wafers is analogous to that of high surface area silica nanoparticles and that identical surface species were found upon immobilization of molecules on both supports.[17,18]

On this basis and in view of the exceptional magnetic properties of lanthanide ions,[19−21] we targeted an approach to prepare a class of magnetic nanomaterials consisting of isolated lanthanide ions at the surface of nonmagnetic nanoparticles, which could then be applied to flat surfaces. Of all the challenges to be met to elaborate such a material, the thermal stability and the absence of organic ligands are of utmost importance. In particular, one would like to conceive a material that can be cycled between blocking temperatures and room temperature without any alteration of its magnetic properties. The approach would benefit from a simple, large-scale synthetic process. Toward this goal, we have developed a strategy for producing fully inorganic isolated lanthanide ions on a silica matrix.[22−24] Our approach combines the site isolation of silica partially dehydroxylated at 700 °C (so-called SiO2–700, presenting an average of ca. 0.8 surface silanol per nm2),[25] the unique magnetic properties of Dy(III) ions[26]—a Kramers ion with a large moment and often exhibiting a significant magnetic anisotropy—and the thermolytic molecular precursor approach, which allows clean removal of organic ligands and generation of isolated metal ions in an anisotropic environment provided by the surface.[27−31]

We first prepared and characterized a Dy(III) molecular precursor bearing thermally labile (tBuO)3SiO– ligands, Dy[OSi(OtBu)3]3[κ2-HOSi(OtBu)3] (1).[24] Grafting of 1 (1.05 equiv/silanol or ca. 0.26 mmol g–1) on SiO2–700 led to the formation of [(≡Si–O)Dy[(OSi(OtBu)3)2(κ2-HOSi(OtBu)3)], 1/SiO (Figure A). Treatment of 1/SiO at 400 °C under high vacuum (10–5 mbar) afforded the fully inorganic material Dy@SiO as a white solid. The entire process can be carried out on gram scale. The absence of organic ligands in Dy@SiO was confirmed by FTIR spectroscopy, which shows the disappearance of ν(C–H) bands and the reappearance of isolated silanol bands upon thermal treatment (Figure B), similarly to what was observed for the corresponding Cr(II), Cr(III), and W(VI) analogues[29−31] used as single-site catalysts. The nature of the Dy sites in 1/SiO and Dy@SiO was determined by X-ray absorption near edge structure (XANES) spectroscopy at the Dy LIII-edge. All Dy species (1, 1/SiO and Dy@SiO) have similar edge energies (Figure C), indicating that during grafting of 1 and thermal decomposition of 1/SiO the Dy oxidation state remains +III, with no evidence of metallic Dy.[32] The extended X-ray absorption fine structure (EXAFS) analysis at the Dy LIII-edge confirmed that, upon grafting 1 onto SiO2, the coordination sphere is largely unchanged, with one (tBuO)3SiO– ligand being replaced by a surface silanol moiety while retaining a chelating κ2 neutral silanol ligand. Thermal treatment resulted in larger changes in the Dy coordination sphere. The best fit to the EXAFS spectrum of Dy@SiO is obtained by adding two neutral oxygen ligands at 2.42 Å to the three short oxygen neighbors at 2.20 Å. Those two additional oxygen atoms are presumably due to siloxane (Si–O–Si) oxygens interacting with the metal center. Taking into account such additional ligands is necessary for a good fit of the EXAFS spectra of Dy@SiO. In contrast, the EXAFS results show no Dy–Dy scattering paths, consistent with site isolation and with the coordination of the Dy(III) sites in Dy@SiO solely to the surface (Figure F and Table S3). Overall, EXAFS shows that Dy in Dy@SiO is bound to only 5 oxygen atoms and is therefore coordinatively unsaturated. The oxygen ligands arise from the surface and should therefore impose a very asymmetric environment. In fact, 1,10-phenanthroline, a bidendate ligand, coordinates to Dy as evidenced by a shift of the aromatic ring (C=C and C=N) in IR (Figure S17),[33] thus showing that Dy was not incorporated in the bulk of the material similarly to what was observed for Cr and Yb.[23,30]

Figure 2

(A) Grafting of 1 on SiO2–700 to form 1/SiO followed by a thermal annealing under high vacuum yielded Dy@SiO (color code: carbon (gray), silicon (orange), oxygen (red), and dysprosium (blue)). (B) FTIR transmission spectra of SiO2–700, 1/SiO, and Dy@SiO. Grafting of 1 (1.05 equiv/silanol or ca. 0.26 mmol g–1) on SiO2–700 leads to the disappearance of most isolated silanols while a broad band appears at 3650 cm–1, consistent with the presence of unreacted silanols interacting with organic ligands. (C) Dy LIII-edge XANES of 1, 1/SiO, and Dy@SiO. (D) k3 weighted EXAFS fits in k-space (top) and R-space (bottom) of 1. The fit of 1 includes five O atoms in the first coordination sphere: three at 2.11 Å, and one each at 2.39 and 2.53 Å. (E) k3 weighted EXAFS fits in k-space (top) and R-space (bottom) of 1/SiO. Aside from a shortening of the Dy–O paths of the κ2 neutral silanol at 2.26 and 2.40 Å compared to the molecular structure, the structure of 1/SiO is similar to 1 (Figures S4 and S5, Table S3). (F) k3 weighted EXAFS fits in k-space (top) and R-space (bottom) of Dy@SiO. The best fit was obtained by including five O ligands, three with short (2.20 Å) and two with longer (2.42 Å) Dy–O distances (Figures S4 and S6, Table S3).

Magnetic analyses of 1/SiO and Dy@SiO provide room temperature χMT of 13.5 and 12.4 cm3 K mol–1 respectively,[34] in agreement with the expected value of 14.2 cm3 K mol–1 for the 6H15/2 multiplet ground state of Dy(III) ions (Figures S7 and S8). The χMT values decrease on cooling with the thermal depopulation of M states to reach 11 and 10.4 cm3 K mol–1 at 2 K. The saturation magnetizations at 2 K (4.6 and 4.5 Nβ for 1/SiO and Dy@SiO, respectively) are consistent with the stabilization of the m = ±15/2 Kramers doublet state well isolated from the first excited Kramers doublet (Figures S7 and S8). Ac susceptibility measurements performed in zero external dc field for 1/SiO and Dy@SiO show that the magnetic moments of Dy(III) ions relax independently below 16 K at the nanoparticle surface, which is not the case for the molecular precursor 1. Plots of the in phase (χM′) and out of phase (χM″) ac magnetic susceptibilities in zero dc field as a function of frequency in the 2–18 K range (Figures A and S9) illustrate the slowdown of the magnetic relaxation and confirm that Dy(III) ions behave as magnets. These results indicate that the oxide surface can be viewed as the ligand that dictates the electronic properties of the Dy(III) ions. Dy@SiO relaxes more slowly than 1/SiO at all temperatures and is therefore a more persistent magnet in zero field. At 2 K the relaxation time determined using an extended Debye model (Figure S10, Tables S4 and S5) is 100 times slower in Dy@SiO than in 1/SiO. The high frequency limit (χS) corresponding to the nonrelaxing fraction of the magnetic susceptibility was fixed at zero to minimize free parameters. τ obeys the Arrhenius law above 9 K with an activation energy of 65(±3) K (Table S4, Figure S11, see Materials and Methods for details) for Dy@SiO while the activation energy (16(±1) K, Table S5 and Figure S11) is much lower in 1/SiO. At low temperature, the average relaxation time reaches a maximum at 1 s (Figure A) in Dy@SiO. Saturation of the relaxation time, which is commonly observed in Dy(III)-based SMMs,[35] prevents the storage of information at zero field. However, in the present case, a non-negligible fraction of Dy centers relaxes at time scale longer than 100 s due to the relatively large distribution of the relaxation times (Figure A, Table S4), which produces a hysteresis loop open at zero field between 2 and 5 K (Figures B and S12) in sweeping the field at 16 Oe s–1. The same hysteresis loop closes when the sweep rate is decreased down to 0.6 Oe s–1 and remains slightly opened at intermediate rate of 2.5 Oe s–1 (Figure S13a). It must be mentioned that the loops close at high field (Figure S13b). The relaxation time in Dy@SiO is almost unaffected by an external dc field (Figure S14). On the contrary, the application of a moderate field on 1/SiO slows down the relaxation time by a factor 100 (Figure S15) at 2 K, which causes the opening of a butterfly like hysteresis (Figure S16). This difference might originate from the higher number of vibration/rotation modes resulting from the presence of organic ligands in 1/SiO with respect to its fully inorganic and more rigid counterpart Dy@SiO.

Figure 3

(A) Frequency dependence of the two components (χM′ and χM″) of the ac susceptibility for Dy@SiO measured between 2 and 18 K in the absence of an external dc field. (B) Hysteresis loop for Dy@SiO measured at 2 K and at a sweep rate of 16 Oe s–1.

In contrast to traditional SMM heterogenization strategies where the magnetic properties of the molecular precursor are transferred to the surface, the magnetic memory properties of Dy@SiO appear only after grafting and thermal decomposition in our approach. This behavior most likely arises from the formation of highly constrained surface species and geometries that could not be obtained with molecular materials, thus illustrating the potential of the thermolytic precursor approach. Furthermore, thermolysis produces a non-centrosymmetric environment in Dy@SiO, with low coordinated Dy(III) exposed at the material surface as shown by EXAFS and 1,10-phenanthroline probe experiment (vide supra).

The strategy used here enabled the simple generation of fully inorganic isolated Dy(III) ions on a silica matrix, which is thermally stable up to at least 400 °C and behaves as isolated SMMs at liquid 4He temperatures. The magnetic remanence of Dy sites in Dy@SiO depends solely on the site isolation on silica and is not found in the corresponding molecular entities, likely resulting from the unusual coordination environment of surface sites and the absence of ancillary organic ligand. The synthetic approach developed here is easily amenable to large scale production and is transferable to any type of oxide surfaces including layers of ferromagnetic oxides or oxide layers on wafers. It thus opens new avenues to generate and manipulate lanthanide nanomagnets for data storage at the nanoscale and quantum information processing.