Protective Effect of Polyoxometalates in {Mo132}/Maghemite Binary Superlattices Under Annealing

The binary assembly DDA-{Mo132}/OA-γ-Fe2O3 (DDA = didodecyldimethylammonium, {Mo132} = [Mo132O372(CH3COO)30(H2O)72]42−, OA = oleic acid) constitutes one of the two examples in the literature of binary superlattices made of a mixing of nanocrystals and oxo-clusters. In a precedent work, we reported in details the preparation of such magnetic binary systems and studied the effect of the nature of the polyoxometalates (POMs) on the magnetic properties. In the present paper, we study the stability of this model binary assembly under heating at various temperatures. Indeed, especially if magnetic and/or transport properties are targeted, an annealing can be essential to change the phase of the nanocrystals in a more magnetic one and/or to desorb the organic capping of the nano-objects that can constitute an obstacle to the electronic communication between the nano-objects. We gave evidence that the maghemite organization in the binary assembly is maintained until 370°C under vacuum thanks to the presence of the POMs. This latter evolve in the phase MoO3, but still permits to avoid the aggregation of the nanocrystals as well as preserve their periodical arrangement. On the contrary, an assembly made of pure γ-Fe2O3 nanocrystals displays a clear aggregation of the nano-objects from 370°C, as attested by transmission and scanning electronic microscopies and confirmed by magnetic measurements. The stability of the magnetic nanocrystals in such POMs/nanocrystals assemblies opens the way to (i) the elaboration of new binary assemblies from POMs and numerous kinds of nanocrystals with a good control on the magnetic properties and to (ii) the investigation of new physical properties as exchange coupling, or magneto-transport in such systems.

Beyond obtaining a binary system with specific properties, one issue can be crucial, specifically in the magnetism and transport fields: the annealing of the binary assembly. Indeed, an annealing can be necessary to change the phase of the nanocrystals and form hard magnetic phases (Zeng et al., 2002;Lisiecki et al., 2007;Parker et al., 2007;Demortière and Petit, 2011;Dong et al., 2011), or to desorb the organic capping around the nanoobjects that can be an obstacle to transport measurements (Chen et al., 2013). As the surrounding ligand has a key role in the formation of the binary superlattices (Boles and Talapin, 2015;Wei et al., 2015;Huang et al., 2016;Travesset, 2017), the ligand desorption has to be done after the formation of the binary assembly. However, due to the needed high temperatures, the annealing procedure can induce coalescence phenomena leading to an increase in the size of the NCs, a modification of the shape and a change in the organization (Klemmer et al., 2003;Alloyeau et al., 2010;Demortière and Petit, 2011). In this context, one of the main challenges is to be able to anneal the nanostructured film while preventing from the aggregation of the nano-objects and keeping the periodicity in the binary system, both for addressing individual magnetic objects and for controlling the charge transport (Dong et al., 2011;Altantzis et al., 2016).
Some years ago, we reported one of the two examples in the literature consisting in the binary assembly of two kinds of nanoobjects with drastic different natures : nanocrystals and oxoclusters. We described the formation of DDA-{Mo 132 }/OA-γ-Fe 2 O 3 (DDA = didodecyldimethylammonium; OA = oleic acid) binary superlattices made of the assembly of polyoxometalates (POMs) and maghemite nanocrystals (MNCs) (Breitwieser et al., 2016). By a judicious choice of the ligand, we showed a finetuning of the magnetic properties of the MNCs film. In this article, we further characterize this DDA-{Mo 132 }/OA-γ-Fe 2 O 3 model system and study its stability after annealing at high temperature by trying to investigate three crucial issues : (i) the preservation of the periodical arrangement; (ii) the assessment of the highest temperature of annealing before collapsing the system; (iii) the intrinsic stability of the nano-objects constitutive of the binary assembly.
Synthesis of the Nano-Objects γ-Fe 2 O 3 nanocrystals were made by slightly modifying the synthesis reported by Park et al. (2004) as described in our precedent article (Breitwieser et al., 2016). In particular, 1octadecene was used (instead of 1-hexadecene) in order to obtain ∼7.7 nm iron oxide nanocrystals. Briefly, the iron-oleate complex was first prepared then dissolved in a mixture of 1octadecene and oleic acid at room temperature. This mixture was heated to the boiling point of 1-octadecene (b.p. 317 • C) under vigorous stirring and was refluxed for 30 min and then cooled to room temperature giving rise to a black gel containing the nanocrystals. The nanocrystals were separated and washed four times with a large excess of ethanol. After four washings 40 µL of oleic acid (OA) are added in a volume ratio of 1:75. The nanocrystals were finally dispersed in chloroform to give an oily ferrofluid.
The DDA-Mo 132 was prepared by counter-cation exchange of the NH 4 -Mo 132 precursor as described before (Breitwieser et al., 2016) by stirring the NH 4 -Mo 132 precursor in an excess of DDA in a water-chloroform biphasic solution.

Binary Superlattices
Binary superlattices were obtained as detailed before (Breitwieser et al., 2016). Briefly, colloidal binary solutions were prepared from equal volumetric mixtures of alkylammonium-{POMs} and oleic acid-capped maghemite MNCs dispersed in CHCl 3 . Supported coassemblies were then obtained using a dip method by drying the mixture directly on horizontally immersed highly ordered pyrolytic graphite (HOPG) or amorphous carbon TEM grids at controlled temperature and in a CHCl 3 saturated atmosphere. The volume of binary solution was deposited according to the substrates size. The TEM grids were immersed in 20 to 30 µL giving ∼1-10 MNCs layers. HOPG substrates of 10 × 5 mm 2 were immersed in 200 to 800 µL giving thickness from dozen to hundred MNCs layers. It is worth mentioning that samples made from the binary mixture were accompanied with reference samples of pure maghemite nanocrystals grown in identical conditions for better comparison.

Thermal Treatment
The HOPG or the copper grid modified by the binary film was introduced in a furnace (NEYTECH Qex) connected to a vacuum pump. The targeted temperature was reached with a ramp of 10 • C/min then the sample was left at this temperature during 1 h under vacuum. The powder of POMs was ground then placed in a crucible to be subjected to exactly the same treatment.

Samples Characterization
A 100∼kV transmission electron microscope (TEM, JEOL JEM-1011) was used to characterize both maghemite and binary assemblies.
Conventional TEM (Bright field and high resolution), highangle annular dark field imaging in scanning transmission electron microscope mode (STEM-HAADF) and X-ray energydispersive spectroscopy elemental mapping (STEM-XEDS) experiments were performed on a JEOL-2100F microscope installed at IMPMC (Paris, France), operating at 200 kV, equipped with a field emission gun, a JEOL detector with an ultrathin window allowing detection of light elements and a scanning TEM (STEM) device, which allows Z-contrast imaging in HAADF mode.
Imaging of the reference and binary samples deposited on HOPG were performed with a SU-70 Hitachi FEG-SEM.
Hysteresis loops at 3 K and thermal dependence of the γ-Fe 2 O 3 magnetization (3-300K) were measured with a commercial SQUID magnetometer (Cryogenic S600) with applied fields up to 4 T.
To characterize the POMs powder, thermogravimetric analysis (TGA) was performed under N 2 between 20 and 600 • C with a SDT-Q600 device. IR spectra were recorded from a KBr pellet on a Jasco FT/IR 4100 spectrometer. The in-situ X-ray powder diffraction (XRPD) patterns were recorded at the X-ray diffraction platform of the IMPMC using an X'Pert Pro MPD Panalytical diffractometer equipped with a Co Kα radiation source (λKα 1 = 1.78897 Å, λKα 2 = 1.79285 Å), an X'Celerator detector. The sample was gently ground and placed on a Si zero background holder. Then the experiment was performed in Bragg Brentano geometry at 293 K.

Native Superlattices
In this study we considered 7.7 ± 0.50 nm Maghemite (γ-Fe 2 O 3 ) nanocrystals synthesized by thermal decomposition of iron chloride at high temperature by following a method (Ngo et al., 2013) similar to that reported by Park et al. (2004) (see Experimental section). Their low size dispersion (∼9%) favors their self-assembly in a hexagonal network characterized by a center-to-center inter-particle distance (d cc ) equal to 11.0 ± 0.5 nm ( Figure 1A and Supplementary Figure 1). Such large distances show that the maghemite film is not compact. This is explained by the excess of oleic acid added at the end of the procedure and necessary to obtain the ordered film (see Experimental section  (Breitwieser et al., 2016). The suitable C 60 -like hollow spherical POM shape (with icosahedral symmetry (Müller et al., 1998), its large size (2.9 nm) and the use of hydrophobic long alkyl chains as counter-cations that confer them core-shell structure and solubility properties similar to those of MNCs favor the co-cristallization (Breitwieser et al., 2016). Supported binary assemblies were then obtained upon solvent evaporation either on immersed amorphous carbon coated TEM grid or HOPG. The TEM images exhibit typical [111]-projected hexagonal lattices of nanocrystal bilayers showing large domains (>500 nm 2 ) in which the presence of DDA-{Mo 132 } can be observed in the octahedral interstices of the hexagonal MNCs lattice ( Figure 1B). The center-to-center interparticle distances between the MNCs in the binary superlattices are equal to 11.0 ± 0.5 nm, similar to that of the pure maghemite organization. This confirms that the POM is small enough to insert in the octahedral interstices without changing the MNCs inter-particle distance.
The magnetic properties of the films (pure Maghemite and DDA-{Mo 132 }/OA-γ-Fe 2 O 3 binary assemblies) have been characterized using a superconducting quantum interference device (SQUID) magnetometer. The superparamagnetic behavior of the MNCs was probed by performing fieldcooling/zero-field cooling (FC/ZFC) measurements of magnetization vs. temperature (3-300 K) in an in-plane field of 100 Oe. The ferromagnetic-superparamagnetic (i.e., blockedunblocked) transition can be determined by the temperature dependence of the magnetization that exhibits a maximum in ZFC curves. It is important to mention that the corresponding temperature (called blocking temperature, T B ) depends on both the MNCs volume (magnetic anisotropy) and on their dipolar interactions. For isolated MNCs (no interaction), the blocking temperature is directly related to the anisotropy magnetic energy (K eff V) of the material, where K eff is the anisotropic constant (that depends on the surface, the crystallinity and the shape) and V the volume of the MNCs. When we consider assemblies of magnetic NCs, they are no longer isolated and the presence of dipole-dipole interactions induces an interaction field (Dormann et al., 1998). The magnetic energy of a nanoparticle interacting with its neighbors, corresponding to the barrier energy (E b ), will then be written:  (Mørup et al., 2010). It appears that the dipolar interaction energy is directional and will depend on the geometrical arrangement of the MNCs in the assembly (Kechrakos and Trohidou, 1998;Mørup et al., 2010). The isolated 7.7 ± 0.70 nm Maghemite (γ-Fe 2 O 3 ) nanocrystals are characterized by a blocking temperature (T B ) equal to 55 K while it is 57 K in the film (Figure 2A, black squares). This latter value, very close to that of isolated MNCs, confirms the formation of a non-compact film characterized by very low interaction between MNCs due to the presence of oleic acid excess. The FEG-SEM images of the binary superlattice show long-range ordered domains of DDA-{Mo 132 }/OA-γ-Fe 2 O 3 MNCs on HOPG ( Figure 3A). As the presence of the POMs do not induce any change in the inter-particle distances in the MNCs network, the binary superlattice exhibits the same magnetic properties as the maghemite film ones with a T B equal to 57 K (Figure 2A, red circles). The hysteresis loops (magnetization vs. applied magnetic field) show the same ferromagnetic behavior at 3 K with a coercitivity H c equal to 230 Oe for both samples ( Figure 2B). As discussed in the introduction, one important challenge in this domain concerns the annealing of the film that can bring important improvements in magnetic and/or transport properties. The main difficulty of this treatment relies on the possibility to keep the periodicity and avoid the aggregation of the NCs in the binary superlattice. To further characterize the DDA-{Mo 132 }/OA-γ-Fe 2 O 3 MNCs binary system, we studied its stability under annealing at various temperatures.

Annealing at 370 • C
The previous samples (Maghemite and DDA-{Mo 132 }/OA-γ-Fe 2 O 3 binary assemblies) are both annealed at 370 • C for 1 h under vacuum (see experimental section). The effect of annealing for the reference sample (maghemite film) and for the binary superlattice results in clear differences on the film morphology and on their corresponding magnetic properties (Figure 3).
SEM-FEG images of the annealed binary superlattice give evidence that the integrity of the organization is maintained despite a drastic decrease of the d cc inter-particle distance from 11.0 to 8.9 ± 0.50 nm after annealing ( Figure 3B). Conversely, the annealed OA-γ-Fe 2 O 3 (reference sample) exhibits a high coalescence of the MNCs in surface, while the organization is preserved in the bulk ( Figure 3C). Indeed, compact underlying layers of organized MNCs are observed, characterized by an inter-particle distance d CC equal to 9.6 ± 0.50 nm. This coalescence between MNCs has a strong influence on their magnetic properties. Indeed, a drastic increase of both the T B ( T B = 159 K) from 57 to 216 K (Figure 3D, blue down triangles) and the coercive field ( Hc = 307 Oe) from 230 to 537 Oe (Figure 3E, blue down triangles) are observed. The ZFC curve is relatively large with a shoulder at 75 K. Such magnetic behaviors can be explained by the presence of coalesced MNCs leading to different sizes. The resulting widening of the size distribution causes the increase of the distribution of barrier energy E b according to two aspects. On the one hand, the bigger MNCs (coalesced) are characterized by higher magnetic anisotropy (K eff V) compared to that of 7.7 nm MNCs. Hence, the distribution of size will induce an increase of the distribution of the magnetic anisotropy energy of the MNCs. On the other hand, the size distribution associated to the decrease of the interparticle distance and the inhomogeneity of the film will contribute to an increase in the dipolar magnetic interactions and an inhomogeneous interaction field. The shoulder in the ZFC curve (100 Oe) at 75 K illustrates these effects. It corresponds to the magnetic response of the underlying film of non-coalesced 7.7 nm MNCs observed on the SEM-FEG images ( Figure 3C). The shift from 57 to 75 K is related to the stronger dipolar magnetic interactions between the 7.7 nm MNCs due to the decrease of their inter-particle distance.
For the binary superlattices, as no coalescence occurs, the size distribution is not modified. In this case, magnetic properties intermediate between the native and annealed maghemite films are observed: an increase of T B ( T B = 35 K) from 57 to 92 K ( Figure 3D, green up triangles) and of the coercive field ( Hc = 144 Oe) from 230 to 374 Oe ( Figure 3E). This magnetic behavior is directly related to an increase of magnetic dipolar interactions related to the decrease in the inter-NPs distance after annealing.
If we compare the hysteresis loops for both annealed maghemite and binary superlattices, we can note some differences in the approach to the saturation that is slower for the annealed maghemite film compared to the binary superlattice while its coercivity is higher ( Figure 3E). Monte-Carlo simulations by Kechrakos et al. give evidence that coalescence of NCs induces an increase of the coercivity due to a strong random dipolar fields generated in the sample. In addition, they show that the coalescence induces an increase of the remanence (Kechrakos and Trohidou, 2003). This is consistent with the strong dipolar magnetic interactions in the annealed inhomogeneous maghemite film compared to the binary superlattice. The presence of POMs in the binary superlattice constitutes a physical barrier between the MNCs preventing their coalescence during annealing while keeping their organization.

Characterization of the Constitutive Elements
To further characterize each nano-object in the film after annealing, transmission electronic microscopies  characterizations have been carried out for both samples. In order to preserve the microscopy copper grid, the annealing conditions were slightly modified: 310 • C for 15 min under primary vacuum. The inter-NCs zones can be properly observed in the TEM images that show coalesced NCs for the reference sample (Supplementary Figure 2) while an inter-MNCs spacing is visible in the case of the annealed binary film with a matrix coating these NCs (Figures 4A,B). TEM and HRTEM characterizations confirmed that the MNCs have kept their integrity, size and shape, even after the drastic thermal treatment. To study the fate of the POMs after annealing, we performed STEM-HAADF (Figures 4C,D1) and STEM-XEDS (Figures 4D2-D4; Supplementary Figure 3) to determine the chemical mapping. On the energy filtered images is observed a homogeneous distribution of iron ( Figure 4D2) and molybdenum ( Figure 4D3). The coexistence of the Mo and Fe signals indicates that the non-coalescence of the MNCs is well induced by the presence of a Mo oxide matrix. Note that the STEM image and the chemical mapping revealed some copper contamination due to the partial degradation of the grid after annealing ( Figure 4D4). However, the STEM technique is not accurate enough to precisely determine if the POMs have kept their integrity after annealing. To do so, a powder of DDA-Mo 132 was treated independently by the same procedure as for the binary film (370 • C, 1 h under vacuum) then characterized by infrared spectroscopy (IR) and X-ray powder diffraction (XRPD). According to the thermogravimetric analysis (Supplementary Figure 4A), an annealing at 370 • C should totally desorb the DDA counter-cations surrounding the POMs. This is confirmed by the IR spectrum after annealing (Supplementary Figure 4B) in which the two bands at 2,852 and 2,923 cm −1 corresponding to the CH bonds stretching in the DDA counter cation disappeared. However, in the IR spectrum a drastic evolution of the bands between 500 and 1,000 cm −1 characteristic of the inorganic core of the DDA-Mo 132 is also observed. This shows that the structure of the POM is no longer preserved under annealing. To determine the nature of the phase after annealing, a powder diffractogram was recorded and the MoO 3 phase could be unambiguously determined (Supplementary Figure 4C). Under annealing, the DDA-Mo 132 moves to the MoO 3 oxide but still ensures the coating of the MNCs, which permits the non-coalescence of the MNCs and the preservation of their periodical arrangement.
Limit of the Protective Effect: Annealing at 450 • C The protective effect of POMs in the film of MNCs shows a limit, as demonstrated by a similar study in which an annealing at 450 • C for 1 h under vacuum was performed. As expected, the reference sample totally coalesces at this temperature (Supplementary Figure 5A). The corresponding blocking temperature is similar to the 370 • C annealed reference (T B = 216 K) but the shoulder at 75 K is not anymore observable (Supplementary Figure 5C, compare violet and blue curves), which means that the coalescence occurs not only at the surface but also in the volume of the film. Concerning the binary assembly, the SEM-FEG images show that at 450 • C the surface layer of the binary film coalesced. However, an underlying layer of organized, non-coalesced MNCs persists even at this temperature (Supplementary Figure 5B). The corresponding magnetic properties are intermediate to the ones of the annealed binary superlattice at 370 • C and the annealed maghemite film, with a T B at 140 K. We emphasize that the ZFC curve is quite large around T B . In particular, a discrete shoulder can be observed at 100 K that could be explained by the presence of the underlayer of non-coalesced MNCs (Supplementary Figure 5C, dark green curve).

CONCLUSION
In this article, we showed that the MNCs in binary assemblies made with oxo-clusters keep their morphology and organization under annealing. Up to around 370 • C, the presence of the POMs in the octahedral sites of the MNCs lattice avoid the aggregation of the Fe 2 O 3 NCs and preserves the periodicity even if the POM structure is destroyed. At this temperature, all the organic coating is desorbed, which permits to consider conductivity measurements in such systems. Moreover, the stability of the magnetic nanocrystals in the binary assembly until this temperature opens the way to the elaboration of new binary superlattices based on other nanocrystals such as Co, or CoPt, whose magnetic properties are very dependent of cristallinity Parker et al., 2007;Demortière and Petit, 2011;Andreazza et al., 2015).

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
RB synthesized the binary assemblies, did the annealing and characterized it. AG synthesized the binary assemblies and studied the stability of POMs under annealing. TA developed the method to make stable binary superlattices based on polyoxometalates and nanocrystals. A-TN did the synthesis of the maghemite nanocrystals. BB did the XRPD characterization. NM did the STEM-XEDS characterization. AP is at the origin of the collaboration between MONARIS and IPCM. CP is at the origin of the collaboration between MONARIS and IPCM. FV and CS managed the project and supervised the involved students.

FUNDING
This work was supported by the LabEx MiChem, part of French state funds managed by the ANR within the investissements d'Avenir program under reference ANR-11-IDEX-0004-02. Recurrent funding from Sorbonne University and CNRS also supported this project.