A protected annealing process for the production of high quality colloidal oxide nanoparticles with optimized physical properties

Abstract

Colloidal synthesis of inorganic materials is usually achieved in solvents under milder conditions of temperature than those usually used for the preparation of bulk compounds. As commonly shown on oxide nanoparticles with sizes of more than typically 20 nm, this leads to particles exhibiting an altered crystallinity with the presence of more or less extended defects that may impact the physical properties of the particles. Considering that post-annealing treatments on powders of nanoparticles inevitably lead to their sintering (i.e. growth and irreversible aggregation), the influence of these defects has rarely been discussed. For a few years, we have been investigating a post-synthesis process of “protected annealing”. The basic principle relies on thermal treatment of preformed particles that have been previously dispersed into a sol–gel silica matrix. After annealing at temperatures up to 1000 °C, the dissolution of the host matrix allows recovery of a suspension of particles with the same size as the pristine particles and an almost perfect crystallinity. Investigation of the physical properties of these particles permitted us to show their optimized properties, thus discriminating the influence of altered crystallinity compared to small size or surface effects and high surface area. Results are presented in the case of oxide phosphors (YVO₄:Eu, YAG:Ce), photocatalytic TiO₂ and magnetic γ-Fe₂O₃ particles. Finally, an extension of the process is shown in the case of a reactive protected annealing, when the thermal treatment is associated with a change of chemical composition of the particles, thus allowing the investigation of systems that are difficult to obtain through conventional colloid chemistry. This latter strategy is presented in the cases of Zn₂SiO₄:Mn nanophosphors, Co-doped γ-Fe₂O₃ and N-doped TiO₂.

---

Introduction

Nanoparticles research represents a very active field of investigations motivated by specific properties associated with their small size. This indeed leads to a high surface to volume ratio, leading to processing issues, for example, of transparent coatings and nano-objects that may be manipulated and coupled to other systems such as biological species. It is also well-known that inorganic materials considered as nano-objects may present new physical properties due to their small sizes which are close to the characteristic lengths of some physical properties. Well-known examples are quantum confinement in semiconductors, plasmonic properties of noble metal particles, coulomb blockade in metals, superparamagnetism in magnetic materials, etc. These original properties are interesting from a fundamental point of view of the solid state physics of nanomaterials, and also because they could be used in many innovative applications.

Chemical synthesis of well-controlled nanoparticles often puts a limit on investigations. Although many works report on the elaboration of nanoparticles as agglomerated nanopowders, colloidal suspensions of individualized particles allow further processing of materials with a controlled microstructure or investigation of the properties of single nano-objects. Top-down approaches such as grinding or dispersion of nanopowders may be considered for some applications but they do not allow a good control of size and dispersion state unless a low yield size selection step is achieved. Thus, colloid chemistry still appears as the most powerful approach and this field of chemistry has made huge progress in the last twenty years. Main improvements were obtained through increasing the control of nucleation/growth steps while preserving particle dispersion. Starting from preliminary works on precipitation in water from salts, the extension toward the use of organometallic precursors in high boiling point coordinating solvents has paved the way toward an always increasing number of new systems.^1,2

Concerning their size, nanoparticles are usually considered to be less than 100 nm. It is worth noticing that one could separate nanoparticles into two classes. The first one corresponds to the 1 to about 10 to 20 nm range, for which the influence of size and surface on physical properties is significant. The second class corresponds to the 20 nm to 100 nm range, for which properties are not so drastically affected by size and surface effects and, in this case one can expect to benefit from both small size and bulk material properties. Nevertheless, this is true only if a good crystallinity of particles is obtained through the elaboration process.

This question is usually irrelevant to very small particles which are obtained with a good crystallinity through nucleation/growth from individual seeds. In the case of particles with a larger size, the growth rate is much higher and may proceed from the aggregation of several primary particles leading to porosity or defects in the structure of particles. Moreover, these particles may undergo a transition due to a competition between the surface and volume energy so that two different crystalline structures may coexist. Fig. 1 presents typical TEM images of different oxide nanoparticles of the second class made by conventional colloid chemistry, either through direct precipitation from metal salts (YVO₄,³ Fe₂O₃ (ref. 6 and 7)), thermohydrolysis (ZrO₂,⁸ TiO₂ (ref. 5 and 9)) or solvolysis (ref. 4 and 10). The inhomogeneous contrast in most of these particles is clear evidence of their poor overall crystallinity. The improvement of crystallinity may be obtained by playing on the experimental parameters of elaboration, mainly the temperature similarly to what is known for common solid state reactions for bulk materials synthesis. This has led to processes such as hydrothermal treatments or solvothermal routes that indeed improve the crystallinity of particles from dissolution/reprecipitation processes which can lead to the crystal reconstruction. It remains that in many cases, the elaboration of particles in the 20 to 80 nm size range is still a challenge as post-treatments in solution are not completely efficient. Poor crystallinity thus results in the alteration of physical properties that does not allow us to take full advantage of the intrinsic properties of the bulk materials. Moreover, it prevents us from answering the question of knowing exactly for which size a particle exhibits properties that are not affected by its surface.

Fig. 1 TEM images of various oxide nanoparticles obtained from aqueous precipitation. (a) YVO₄:Eu (ref. 3), (b) YAG:Ce (ref. 4), (c) TiO₂ (ref. 5), and (d) Fe₂O₃ (ref. 6).

In this context, further improvement of crystallinity may only be obtained through increasing the temperature of a post-annealing treatment to reach temperatures that correspond to those used for bulk material synthesis. These temperatures, which are commonly of more than 700 °C, are not compatible with solvents, and such treatments achieved on powders of nanoparticles lead to their sintering (irreversible aggregation and grain growth).

For that reason, our work over the last few years has been to develop a so-called protected annealing process, which consists of achieving a thermal treatment at very high temperatures on particles that are dispersed in an inorganic matrix, inert at the considered temperatures but that can be further dissolved to recover the annealed particles as colloidal suspensions.¹¹

We mainly focused our work on oxide materials with either luminescence or magnetic properties, to investigate the influence of defects on these properties. An extension called the ‘‘reactive protected annealing process’’ involves chemical modification of the particles through solid state reactions that may be involved in addition to the annealing process. We showed that this approach also allows the elaboration of nanosized systems that cannot be synthesized through conventional colloid chemistry mainly because reaction between precursors is ineffective for the precipitation of the expected phase either in terms of chemical composition, crystallinity or crystalline phase, size control or redox state of the constituents.

Basic principle of the protected annealing process

The basic principle of the process is schematically described in Fig. 2. Pristine particles are obtained through conventional colloid chemistry. In the case of oxides, it is generally achieved through a precipitation reaction starting from precursor salts dissolved in water, but particles resulting from glycothermal reaction and further dispersed in ethanol are also considered. Particle dispersion may be achieved using their electrostatic charge at the appropriate pH, but it may also be improved through the addition of a surface complexing agent, usually a polymer such as polyacrylic acid or polyethylene glycol.

Fig. 2 (a) Schematic description of the protected annealing process. The initial point (i) is an aqueous or alcoholic solution of nanoparticles, which are then dispersed in a surfactant templated silica matrix. After gelation, this composite is calcinated at 450 °C in air to obtain a mesoporous silica matrix embedding the nanoparticles (ii). A high temperature treatment is applied, leading to annealed nanoparticles dispersed in the silica matrix (iii). The last step of the process is to remove the silica matrix by a fluorhydric acid treatment, leading to an aqueous solution containing the annealed nanoparticles (iv). (b) SEM image of a typical mesoporous silica matrix containing TiO₂ nanoparticles (adapted from ref. 12).

The first step is the elaboration of a composite material consisting of a host matrix in which the nanoparticles are homogeneously dispersed. This is achieved by simple mixing of the colloidal suspension with a solution of the precursors of the matrix which will lead to the composite system after aging and drying. The choice for the host matrix has to meet mainly three requirements. First, the matrix should be processable from precursors in solution in order to allow the appropriate mixing and dispersion of the particles. Second, it must be inorganic and inert to undergo high temperature treatments without degradation or reaction with the particles. Finally, it must be dissolvable under mild conditions in order to preserve the annealed particles.

In our case, we focused our attention on silica that was found to satisfy all the above listed requirements, but other experiments were also performed using zinc oxide. Silica appears as a good candidate first because its sol–gel chemistry is well-known. It can be achieved under conditions that are compatible with the stability of colloidal particles (low ionic strength, in water or water–ethanol mixtures). Surface complexation by hydrosiloxane groups may also provide some stabilization of the particles. The sol-to-gel transition allows the entrapment of the particles that thus remain well dispersed during solvent removal. Secondly, silica can be dissolved through reaction with KOH or with HF. The kinetics of dissolution of the annealed sol–gel silica was found to be greatly improved when considering the surfactant-templated mesoporous silica. This approach only requires the adjunction of a copolymer that will act as a template for the pore formation, but may also improve the particle dispersion through adsorption on their surface. The porosity may also limit secondary solid state reaction between the nanoparticles and the silica phase through reducing their common interface.

The thermal treatment is achieved in two steps. The first one is typically at 450 °C in order to remove the organic template. The second step corresponds to the annealing treatment of the particles. The maximum possible temperature is determined either by reaction between silica and the particles or by the temperature at which crystallization of silica into cristobalite is observed. In the latter case, the main limitation is due to the very difficult dissolution of this phase. For the systems we investigated, maximum annealing temperatures were typically of 1000 °C, mostly related to silica crystallization. In some cases, the annealing treatments may be achieved under controlled atmosphere, for example H₂–N₂ mixture, in order to control the oxidation state of some elements present in the particles. This is the case, for example, of luminescent YAG:Ce nanoparticles in order to preserve the active 3+ state of the cerium ions.⁴

Structural evolution through annealing

After calcination and matrix dissolution, the particles are recovered as a colloidal suspension, usually in water with controlled pH or by adding surface modifiers such as polyacrylic acid (PAA) to ensure good dispersion. IR spectroscopy and chemical analysis allow us to check that appropriate conditions (dissolution time, HF concentration) and further washing lead to a complete elimination of all silica species. Dynamic light scattering analysis confirms that the particle size and dispersion state are globally maintained, with only small aggregates that can be easily removed by centrifugation. This is the first evidence of the ability of the silica matrix to prevent particle sintering during the thermal treatment.

Size and morphology. Fig. 3 shows typical images of YVO₄:Eu and TiO₂ particles before and after thermal treatment at 1000 °C. Regarding the homogeneity of the contrast in the volume of the particles, a clear improvement is observed, which proves the almost perfect reconstruction of the pristine particles into monocrystalline nanoparticles. This is also confirmed by the facetted shape of the YVO₄ particles, a clear indication of their monocrystalline nature.

Fig. 3 TEM images of pristine YVO₄:Eu and TiO₂ particles respectively (a and c) and corresponding images of annealed particles (b and d).

TEM observations of YVO₄ and YAG:Ce samples after their annealing revealed a surprising microstructure with the presence of cavities (Fig. 4). In some cases, the cavities exhibit a shape and an orientation that correspond to some specific crystallographic planes. In the case of YAG:Ce particles, cavities are observed in almost all the particles with a very spherical shape.

Fig. 4 TEM images of annealed particles exhibiting cavities resulting from pore coalescence: (a) YVO₄:Eu nanoparticles and (b) annealed YAG:Ce nanoparticles.

The origin of these cavities is explained as resulting from the coalescence of pores that were present within the pristine particles. Such a porosity may explain the poor homogeneity in their TEM images. It can be characterized by specific area measurements and BET analysis on a powder of pristine particles. For example, YVO₄ particles exhibit a specific surface of 206 m² g⁻¹.¹³ This corresponds to the value expected when considering that pristine particles are formed through the agglomeration of about 7 nm sized primary particles, thus revealing a possible mechanism for the particle formation. A similar microstructure seems to be commonly reported on oxide particles of sizes above 10 nm, such as YAG,⁴ ZrO₂,⁸ and TiO₂,⁹ suggesting a similar mechanism of particle formation. It is interesting to note that some differences are observed concerning the efficiency of the annealing process to remove all the porosity.

Indeed, annealing of TiO₂ particles at 1000 °C leads to a complete densification of the particles. In the case of YVO₄ (at the same temperature), densification is either complete or coalescence of the porosity produces the formation of one or two cavities as shown above. In the case of YAG, we observe a partial coalescence of the pores into several cavities. This difference between the investigated systems may be understood as a consequence of their refractory nature and their ability to sinter.

Crystalline phases. More detailed investigations of the structural evolution occurring through annealing were achieved using X-ray diffraction. We first focus on the evolution of the crystalline phase as a function of the annealing temperature. It is indeed well-known that crystal structures exhibited by nanoparticles may differ from the corresponding bulk materials, either due to the kinetics formation of a metastable phase during the precipitation process, or more commonly because of the high surface to volume ratio in nanoparticles that may lead to the thermodynamical stabilization of the polymorph having the lowest surface energy. The latter effect has been reported in many examples such as TiO₂, ZrO₂ or BaTiO₃ for which the anatase,¹⁴ tetragonal,¹⁵ or cubic¹⁶ crystalline phases have been respectively observed in the case of nanoparticles although they are less stable in the bulk compounds. In the case of Fe₂O₃ nanoparticles, the γ, ε, β, and α-polymorphs can be successively stabilized upon crystallite growth.¹⁷ The γ to α-Fe₂O₃ transition was explained as resulting either from a nucleation and growth process starting at or close to the surface,¹⁸ or a martensitic phase-like transformation with an activation energy that decreases when the particle size increases.¹⁹ Surface chemistry, especially related to the presence of surface hydroxyl groups, was also shown to be important to account for phase stability through a modification of surface energy.²⁰

Using a protected annealing process, a similar behaviour was observed in TiO₂ and Fe₂O₃ particles for which the anatase and maghemite structures were respectively preserved even for annealing temperatures up to 900 °C. As shown in Fig. 5, maghemite (γ-Fe₂O₃) nanoparticles dispersed in a silica matrix retain their spinel structure up to 900 °C, while a transition into the stable hematite (α-Fe₂O₃) form is observed at 450 °C for powder of particles. This result, similar to what has been previously reported by other authors,^21,22 proves the efficiency of the confinement in the silica matrix to prevent the growth of particles through coalescence and sintering. It can also be noted that the absence of transition is a clear indication of the good dispersion of the particles instead of the formation of aggregates that would be able to grow upon heating, thus leading to a phase transition.

Fig. 5 X-ray diffraction profiles (λ = 1.5418 Å) of 7 nm Fe₂O₃ nanoparticles either annealed as a powder or dispersed in a silica matrix (annealing temperatures: 90 °C, 290 °C, 540 °C and 840 °C). The peaks related to a Si calibrant are indicated by filled circles. Inset: an enlarged view of synchrotron X-ray data (λ = 0.7211 Å), showing the P4₁32 extra reflections (asterisks) related to Fe vacancies ordering. Adapted from ref. 6.

Evolution of the coherence length. The structural coherence length, as determined from Scherrer analysis of the diffraction line width,²³ gives an indication of the evolution of the size and the crystallinity of the particles upon annealing. The evolution in the case of unprotected γ-Fe₂O₃ particle nanopowders (Fig. 6a) confirms that the γ-Fe₂O₃ → α-Fe₂O₃ phase transition is directly related to the growth of the particles.

Fig. 6 (a) Coherence length value, L_c, as a function of the annealing temperature for crude Fe₂O₃ particles (open and filled circle), for Fe₂O₃ particles embedded within silica (filled triangles) and for YVO₄:Eu nanoparticles also embedded into silica (filled squares). (b) Williamson–Hall plot for YVO₄:Eu nanoparticles before (open circles) and after (filled circles) application of the annealing process at 1000 °C.

In the case of particles dispersed in a silica matrix, one would have expected first an increase of the coherence length up to the point at which particles are completely annealed and monocrystalline, and then a saturation up to a temperature at which there is either a parasitic reaction with silica, or a significant diffusion within the matrix leading to some kind of Ostwald ripening. Such a behaviour is actually not observed in the case of YVO₄ (and similarly YAG), in which the coherence length values increase continuously starting from about 16 nm at 400 °C and reaching 47 nm at 1100 °C (Fig. 6a). At this point, this length almost corresponds to the geometrical size of the particle as determined by TEM suggesting that the saturation might have taken place above the maximum temperature of the annealing process. In the case of protected γ-Fe₂O₃ particles, the coherence length remains almost constant up to 1000 °C. This is explained by the small size of the initial particles (7 nm) that are already monocrystalline after synthesis.

Defects annealing and strain relaxation. Coherence length values only give a partial view of the particle crystallinity. Deeper insights into the crystalline order may be obtained by considering the (hkl)-dependence of the X-ray diffraction line widths. The size of the diffraction spots in reciprocal space is actually defined by the sum of two terms, one is related to the finite size L_c of the diffraction domains and the second one to the strain effects associated with local fluctuations of d-spacings within the sample. Local variations of d-spacings usually account for non-uniform distortions due to external stresses, crystalline defects or local heterogeneities of the chemical composition.

Assuming a Cauchy line shape for the Bragg reflections, the full width at half maximum H of each peak can be written as:

where θ represents the diffraction angle and λ the X-ray wavelength. The width of the distribution in d-spacings, Δd/d is then represented by the variable ε and can be determined experimentally by a linear regression of a scattered plot of βcos θ/λ versus 2sin θ/λ.²⁴Fig. 6b presents the evolution of the Williamson–Hall plot in the case of YVO₄:Eu particles before and after application of the annealing process at 1000 °C. We note that points obtained from each diffraction peak are far less dispersed in the case of the annealed particles with an overall linear regression coefficient which has decreased. This shows that the annealing treatment resulted in a strong reduction of the lattice strains and a final coherence length of about 50 nm, which is in good agreement with the mean particle size measured by TEM.

Structural disorder (strains and defects) can also be assessed through a more precise analysis of diffraction data, for example by using synchrotron radiation. Analysis performed on data from YAG nanoparticles showed a lattice parameter of 12.0785 (1) Å in the pristine particles that decreased down to 12.01410 (3) Å for annealed particles, much closer to the bulk value (12.01300 (1) Å). This effect is consistent with the observations of Hosokawa and colleagues on unprotected nanopowders annealed at 1000 °C.²⁵ Similar to results on other systems,²⁶ the authors suggest that the presence of hydroxyl groups in the porosity of the pristine particles causes the observed lattice expansion.

However, only large densities of extended defects such as stacking faults or dislocations induce a significant line broadening or a change in the lattice parameter value, while a low density of uncorrelated point defects will rather impact the repartition of the scattered intensity in the reciprocal space. For instance, for γ-Fe₂O₃ particles below 15 nm synthesized by room-temperature co-precipitation of ferric and ferrous salts in water, the main source of disorder is a randomized distribution of iron vacancies within the octahedral sites of the spinel structure (Fd3̄m space group). Partial vacancy ordering occurs by subsequent annealing above 300 °C (Fig. 5), as evidenced by the appearance of additional peaks corresponding to the P4₁32 space group.⁶

Influence of crystallinity and annealing on physical properties

The initial idea of this work was to investigate the influence of bulk defects on the physical properties of nanoparticles, and provide a way for optimization in cases where it would have been strongly limiting. This would also allow more precise investigation of the size for which surface properties begin to alter the physical properties significantly as compared to those of the bulk compounds. The answer to this question can indeed only be provided in the case of well-crystallized particles in order to discriminate between bulk and surface defects. Finally, it must be noted that since elaboration processes were shown to lead to porous particles in some cases, the difference between surface atoms and bulk defects is not always so clear. Concerning the physical properties that may be affected by bulk defects, surface and size, and since we are concerned mainly with oxides, we focused our attention on rare-earth doped oxides (YVO₄ and YAG:Ce) and magnetic iron oxide nanoparticles.

Protected annealing process applied to oxide nanophosphors. Luminescent nanoparticles are the subject of intense investigations motivated by innovative applications requiring low size and optical transparency, such as biological labels or light emitting displays or devices. Apart from quantum dots, which have been largely explored and are still being investigated, rare earth doped oxides represent a new class of materials that are of potential interest. In this case, the idea is to take advantage of specific emission properties related to rare-earth doped oxides: high photostability, thin emission lines, energy transfer processes, relatively long emission lifetimes, and very specific properties such as up-conversion and delayed emission. It must also be noted that these compounds are insulators, and that emission occurs via localized excited states although eventually some energy transfer processes are involved. When compared to quantum dots, no confinement effects are expected, and localized excitation states conduce to anticipate a lower trapping efficiency on surface states.

Many optimized compositions have been developed, usually for industrial purposes. Our research activities on rare-earth doped nanoparticles are mainly concentrated on two systems: YVO₄:Eu and YAG:Ce. The reason for this choice was first given by the performance of the bulk materials, but also because of the possible synthesis of nanoparticles through conventional colloid chemistry routes.^3,10

The emission properties of Eu³⁺-doped YVO₄ pristine nanoparticles have been investigated in detail.²⁸ The spectroscopic features were nearly similar to those of the bulk materials with a maximum excitation wavelength at 280 nm (vanadate absorption) and the main emission at 617 nm from the ⁵D₀–⁷F₂ transition of Eu³⁺. A maximum quantum yield of 10 to 20% was found corresponding to a maximum Eu doping concentration of 20%. Although it is not so low when compared to other luminescent particles dispersed in water without surface modification, the difference in the bulk properties (68% of quantum yield for 5% Eu doping) was initially attributed to alteration of energy transfers and quenching of the emission through surface defects and hydroxyl groups.²⁸

Nevertheless, this result was not satisfactory since the large size of the particles (30 to 50 nm) led us to expect a rather limited contribution of the surface. The protected annealing process was achieved and properties of the resulting annealed particles were compared to those of pristine particles and bulk reference materials.

Fig. 7a displays the evolution of emission quantum yield of the different samples (pristine particles, annealed particles and bulk references) as a function of their europium content. These plots give a good indication of the emission yield and of the efficiency of the energy transfers leading to emission from the europium. It appears clearly that, contrary to pristine particles, annealed samples display an overall evolution that is now very similar to the one of the bulk references, with an optimum doping at 5%, attesting that energy transfers have been largely improved by the annealing treatment due to a better crystallinity and a large decrease of porosity.

Fig. 7 (a) Evolution of the luminescence quantum yield of Y_1−xEu_xVO₄ as-made nanoparticles (●), annealed nanoparticles (□) and bulk material (□), as a function of the Eu doping concentration x (adapted from ref. 13). (b) Wide-field fluorescence microscopy image of annealed Y_0.78Yb_0.2Er_0.02VO₄ nanoparticles deposited onto a glass slide. The nanoparticle emission is centered at 550 nm under a 980 nm excitation (8 kW cm⁻²). This image was taken with an acquisition time of 500 ms.²⁷

Nevertheless, the optimum quantum yield of the annealed particles (40%) remained lower than for the bulk reference (68%) which seemed surprising considering the low porosity and the specific surface area of the particles, their excellent crystallinity and similar energy transfer processes attested by the same shape of the curves shown in Fig. 7a. The reason was found when considering a dielectric effect intrinsic to the small size of the particles.¹³ Indeed, light emission measurements were achieved on annealed nanoparticles dispersed in water, whose refractive index (n = 1.33) is much lower than that of the bulk YVO₄ (n ∼ 2). In a rough approximation,^29,30 the radiative lifetime τ_rad of a transition from an ion embedded in a media of refractive index n can be written as:

A decrease of the effective refractive index results in an increase of radiative lifetime and thus a decreased quantum yield when assuming a constant non-radiative decay time. To verify this hypothesis, YVO₄ nanoparticles were embedded into two media (SiO₂ and TiO₂) with two different refractive indexes (nSiO₂ = 1.5 and nTiO₂ = 2) and annealed for 12 hours at 600 °C. The ⁵D₀ level lifetime of Eu³⁺ was then measured for these two samples. Experimentally, we found τSiO₂/τTiO₂ = 2.6 which corresponds exactly to the value (2.6) calculated from eqn (2).

Considering that the radiative lifetime of the emission from the nanoparticles is increased by a lower effective refractive index than in the bulk compound, this explains their lower quantum yield if non-radiative processes are not affected by such dielectric effects. This appears as an intrinsic limitation of the optical properties of the nanoparticles when compared to bulk materials, annealed particles thus being the best that can be obtained in terms of emission properties.

A straightforward application of the protected annealing process concerns the investigation of up-conversion properties of Yb–Er doped samples. This process is based on the excitation of two Yb ions at about 975 nm that transfer their energy to a neighboring Er ion which further emits photons in the visible range (mainly around 520–550 nm and 660 nm). This process is well-known to be very sensitive to non-radiative relaxations through phonons from the matrix, and is all the more sensitive to defects and surface groups such as hydroxyls with high energy phonons. Unsurprisingly, the first experiments have shown that no up-conversion could be detected on pristine YVO₄:Yb:Er particles. The same experiments made on annealed particles showed a drastic enhancement of the emission yield, to a point that detection at the single particle level could be achieved as shown in Fig. 7b.²⁷ This confirms the high efficiency of the annealing in reducing non-radiative processes. Moreover, the estimation of the quantum yield provided a value of about 1.3% for an excitation intensity of 8 kW cm⁻², which is not so low in comparison with those reported on halide nanoparticles whose phonon energy is much more favorable for the UC process. Our interpretation is that the excellent quality of the annealed particles compensates the less favorable phonon energy range assuming that the reported halide particles may not be so well crystallized.

Protected annealing process applied to iron oxide nanoparticles. Size reduction in magnetic materials gives rise to new phenomena which are not observed in the corresponding bulk phases, due to a strong interplay between intrinsic properties, finite size and surface effects, shape anisotropy or interparticle interactions.³¹ Superparamagnetic behavior and enhanced magnetic anisotropy, that were found to have a strong impact on magnetic recording and bio-technologies,^32,33 are typical illustrations of such effects. Saturation magnetization and coercivity values are also well known to show a marked domain size dependence, the former effect being ascribed to spin canting at the surface of the particles.^34–36 Because of this strong correlation between size and magnetic behavior, a low sized dispersion is often required to get optimized properties. Various chemical routes have thus been proposed to produce ultra-fine magnetic particles with uniform sizes,³⁷ including template-assisted routes using reverse micelles, hydrothermal and sol–gel processes, hot-injection or seed-assisted growth methods. Nevertheless, different magnetic properties are often reported for the same mean particle size depending on the synthesis protocol that is used.^38–40 Recent results have shown that these differences could be explained by the fact that geometric and magnetic size distributions do not systematically coincide.^41,42 Even for optimized chemical routes that yield uniform particle sizes in the single domain regime, the size of the magnetic domains was found to be substantially smaller than the geometric size. For iron oxide nanoparticles, which are extensively studied for biomedical applications, the presence of antiphase boundaries, twins or dislocations is assumed to disrupt magnetic couplings and to have a detrimental impact on magnetic properties.^41,42 The investigation of magnetic nanomaterials is thus quite complex, as one should take into account intrinsic phenomena but also extrinsic parameters based on geometric size distribution, non-uniform chemical composition and lack of crystallinity.

We have recently investigated the influence of post-synthesis heat treatments carried out up to 900 °C on 7 and 14 nm sized γ-Fe₂O₃ nanoparticles prepared by room temperature coprecipitation in water.⁶ Thermal annealings up to 900 °C were shown to afford a control of the crystalline state of the particles, preventing both grain growth and transformation into hematite (see Fig. 5 and 6a). Heat treatments above 290 °C led to a partial ordering of the Fe vacancies, evidenced by synchrotron X-ray powder diffraction and FT-IR spectroscopy measurements. An increased magnetic anisotropy was also reported, presumably representative of a surface contribution. This latter effect could be ascribed to a modification of the surface state upon heating, involving the removal of surface hydroxyl groups. For the 7 nm particles, a slight decrease in saturation magnetization was observed after annealing. Analysis of the thermal dependence of the saturation magnetization indicated that this result was related to an increased number of weakly coupled and misaligned spins at the surface of the particles.⁶ All experimental findings were actually consistent with a propagation of the shell of non-collinear spins toward the center of the particles. Comparison with Monte Carlo simulations suggested that this expansion was mainly driven by the increased surface anisotropy.⁴³ In contrast, for the 14 nm particles, an increased saturation magnetization was reported after thermal treatment, confirming that the previous evolution was mostly ascribed to surface effects. For large particles, the volume contribution dominates and the saturation magnetization value increases due to Fe vacancy ordering, as reported by other authors.⁴⁴

Extension to a reactive protected annealing process

As previously shown, the protected annealing process makes possible the thermal treatment of particles preserving their size and agglomeration state. The permissible temperatures are compatible with those commonly used for solid state reactions to form ceramics. Initially developed to improve the crystallinity of preformed particles, the extension of the process to the elaboration of new compositions that cannot be reached by colloid chemistry is straightforward. In this case, the idea is to target compositions in which one component can be prepared by colloid chemistry in order to control the final size of the product.

N doped TiO₂. Anatase particles are remarkable for their photocatalytic properties that find many applications in the field of depollution or self-cleaning materials. Nevertheless, applications are limited to irradiation wavelengths whose energy is higher than the band gap of the semiconductor, i.e. in the UV range. In the case of a solar irradiation, only the small UV part of the spectrum is used. A large amount of work has thus been done to develop efficient photocatalysts working under visible light, it consists in either looking for new active compounds with a lower gap or through doping of TiO₂ to induce some absorption in the visible range. In the latter case, nitrogen doping appeared as a promising approach according to the first results reported by Asahi.⁴⁵ Efficient nitrogen insertion in the TiO₂ matrix was obtained through a high temperature treatment under ammonia at temperatures that are typically above 700 °C. This has been achieved on powders, leading to doped materials, but a similar treatment performed on nanoparticles causes an irreversible aggregation that prevents further processing of thin films for transparent self-cleaning windows. Such treatments achieved directly on thin films are not possible because the required temperatures are not compatible with the glass substrate. We thus applied successfully the protected annealing process by i) dispersing TiO₂ nanoparticles within a porous silica matrix, ii) performing the high temperature reaction of nitridation, and iii) recovering the particles that can be dispersed as thin films on glass substrates.^46,47 Pictures of the powders and of the resulting colloidal suspensions are shown in Fig. 8a, displaying the successful insertion of nitrogen in TiO₂ particles which is evidenced by an enhanced coloration. Detailed investigations proved that this insertion was concomitant to some Ti⁴⁺ reduction which is detrimental to photocatalysis. Nevertheless, these Ti³⁺ centers could be almost completely removed by appropriate post-thermal treatment in air, and thus a substantial activity under 390 nm could be achieved.

Fig. 8 Examples of application of a reactive protected annealing process. (a) Pictures of N-doped TiO₂ colloids obtained from TiO₂ particles dispersed in a porous silica matrix and annealed under NH₃ at 700 °C, then in air at different temperature to eliminate residual Ti³⁺.⁴⁶ (b) Magnetization versus magnetic field loops measured at 10 K for Co doped γ-Fe₂O₃ nanoparticles dispersed in silica, after annealing at temperatures ranging between 100 and 600 °C.⁵⁷ (c) X-ray diffraction of Zn₂SiO₄:Mn particles obtained through reaction of SiO₂ nanoparticles in a Mn-doped ZnO matrix, evidencing a size induced phase transition and the corresponding change of luminescence from green to yellow.⁴⁸

Co doped γ-Fe₂O₃. As already mentioned, nanoparticles of iron oxides have been extensively studied for applications in biomedicine, as contrast agents for MRI or new vectors for drug delivery, hyperthermia cancer treatment or tissue engineering.^33,49 To reach an optimum efficiency, each application requires particles with tailored properties related to chemical composition and saturation magnetization, magnetic anisotropy, but also size and shape distributions. For instance, recent theoretical studies have revealed that the specific loss power, which measures the heating produced by the relaxation of magnetic particles placed in a rf magnetic field, presents a sharp maximum as a function of the particle size.⁵⁰ Designing nanoparticles with a narrow size distribution is thus mandatory for hyperthermia therapeutics and optimized protocols have been developed such as seed-assisted growth and heating-up processes^51,52 or alternatively post-synthesis size selection steps.^53,54 In this context, the modification of the magnetic characteristics such as saturation magnetization or magnetic anisotropy through doping, in order to improve the relaxivity or the specific loss power of iron oxide colloids, should preserve a low size dispersion. It is admitted that insertion of cobalt within the MFe₂O₄ spinel structure (M: transition metal) results in an increased magnetic anisotropy.^55,56 For instance, the anisotropy constant of cobalt ferrite CoFe₂O₄ is about one order of magnitude larger than those of maghemite and magnetite. The previous approach of a reactive protected annealing process has, in this way, been extended to the high-temperature cobalt doping of γ-Fe₂O₃ particles.⁵⁷ The doping strategy involves adsorption of Co²⁺ ions on preformed magnetite particles, and it is followed by further oxidation of maghemite and diffusion of the dopants in the volume of the particles upon heating. A large (up to 3 times) and progressive increase of the anisotropy constant could be achieved upon Co insertion even at very low doping levels (Co–Fe atomic ratio of 0.9%). Increased magnetic anisotropy was evidenced by an increase of the blocking temperature T_B and this trend was confirmed by the M(H) curves measured at 10 K, i.e. in the blocked regime, which also showed a significant increase in the coercive field value, H_c upon heating (Fig. 8b). Note that a direct comparison between H_c values is relevant only because the particle size remains unchanged after annealing due to the confinement in the silica matrix.

Manganese doped zinc silicate phosphors. Manganese doped zinc silicate (ZSM) is a phosphor having a special interest for light emitting displays, especially those involving VUV excitation. Several works have been devoted to the elaboration of colloidal suspensions of nanoparticles. Crystalline particles with luminescence properties were reported using hydrothermal⁵⁸ or solvothermal⁵⁹ routes, leading to more or less controlled dispersions. In our case, we investigated the elaboration of colloidal suspensions of nanoparticles obtained by using a reactive protected annealing process.⁴⁸ In addition to the investigation of this original process, the objective was to recover dispersible particles with an excellent crystallinity as expected after high temperature treatment, thus exhibiting improved emission properties compared to those of the particles obtained through usual colloid chemistry techniques at lower temperatures. The process was achieved considering the dispersion of silica particles into a ZnO matrix containing Mn²⁺ ions. Thermal treatment at high temperatures (700 °C to 900 °C) of the obtained composite powder ensured the Zn₂SiO₄:Mn phase formation, and the further dissolution of the ZnO matrix led to the recovery of colloidal suspensions of nanoparticles. Characterization of the obtained particles evidenced the role of the ZnO matrix in limiting the aggregation and growth of the Zn₂SiO₄:Mn particles, validating the concept of this protected calcination strategy. As shown in Fig. 8c, the size and the crystal structure of the particles were found to depend on the calcination temperature: yellow emitting β-Zn₂SiO₄:Mn particles with a diameter of 40–70 nm were obtained at about 800 °C, while green emitting α-Zn₂SiO₄:Mn particles with a diameter of 80–100 nm were obtained at about 900 °C. The phase transition induced by the size change is discussed in terms of surface/volume energy of the two phases. The Mn insertion was characterized, indicating that a maximum of about 2% could be obtained before the formation of a parasitic ZnMn₂O₄ phase. The Mn²⁺ insertion was shown to stabilize the β-phase, which is explained by a better fit of the Mn ions inside the octahedral sites that are present only in this structure. Finally, the emission yield of the particles was measured to be 16% when compared to a bulk reference. This value, quite high for oxide nanoparticles in solution, is expected to be limited not only by residual defects but also by dielectric effects as previously displayed for highly crystalline YVO₄:Eu particles, considering the high emission lifetime of the emitting level (several tens of milliseconds).

Conclusions and perspectives

Physical properties of nanoparticles usually differ significantly from those of the bulk counterparts. It is commonly accepted that it results from both effects intrinsic to small size (quantum confinement, superparamagnetism, dielectric confinement, etc.) and more extrinsic effects associated to surface states, which are largely enhanced in this size range. The influence of defects in the volume of the particles, resulting from an altered crystallinity, has rarely been discussed. Their influence could be non-negligible considering the low temperature of synthesis required by colloid chemistry and the growth mechanism of particles, especially when considering sizes of several tens of nanometers. The main difficulty in the investigation of such effects comes from the possibility to produce particles with optimized crystallinity while preserving their size, surface and dispersion state. In this context, we have developed for a few years a so-called “protected annealing process”. It consists of (i) dispersing preformed colloidal nanoparticles into a refractory matrix, (ii) achieving thermal treatments of annealing up to high temperatures (typically 1000 °C) and (iii) dissolving the matrix to recover the annealed particles as a colloidal suspension. Applications of this process were investigated in our group on various systems, especially oxides such as luminescent rare-earth doped oxides (YVO₄:Eu, YAG:Ce), photocatalytic TiO₂ and magnetic γ-Fe₂O₃. This allowed us to provide new insights into the physical properties of these systems and to discriminate between size, surface and altered crystallinity effects. As a further development of this strategy, we also studied the application of the process for the elaboration of systems obtained following a solid state reaction at high temperature, commonly used in ceramic synthesis, but still preserving the size and dispersion state of the particles. We show that, in the case of Zn₂SiO₄:Mn nanophosphors, Co doped γ-Fe₂O₃ and N doped TiO₂, such a “reactive protected annealing process” gives access to new systems by increasing the temperatures of synthesis when compared to those used in conventional colloid chemistry. This strategy is a way to join colloid chemistry and conventional solid state chemistry and should lead to numerous investigations for the elaboration of complex systems with excellent crystallinity and thus optimized physical properties.

The authors gratefully acknowledge the contribution of Caroline Bertail, Geneviève Mialon, Morgan Gohin, Amélie Revaux, Charlotte Vichery and Lucie Devys to this work.

Notes and references

1. C. B. Murray, D. J. Norris and M. G. Bawendi, J. Am. Chem. Soc., 1993, 115, 8706.
2. J. Park, K. An, Y. Hwang, J.-G. Park, H.-J. Noh, J.-Y. Kim, J.-H. Park, N.-M. Hwang and T. Hyeon, Nat. Mater., 2004, 3, 891.
3. A. Huignard, T. Gacoin and J.-P. Boilot, Chem. Mater., 2000, 12, 1090.
4. A. Revaux, G. Dantelle, N. George, R. Seshadri, T. Gacoin and J.-P. Boilot, Nanoscale, 2011, 3, 2015.
5. Commercial TiO₂ suspensions CristalACTiV™ S5-300A from CristalGlobal.
6. C. Vichery, I. Maurin, P. Bonville, J.-P. Boilot and T. Gacoin, J. Phys. Chem. C, 2012, 116, 16311.
7. R. Massart, IEEE Trans. Magn., 1981, 17, 1247.
8. D. Carriere, M. Moreau, P. Barboux and J.-P. Boilot, Langmuir, 2004, 20, 3449.
9. E. Puzenat and P. Pichat, J. Photochem. Photobiol., A, 2003, 160, 127.
10. R. Kasuya, T. Isobe and H. Kuma, J. Alloys Compd., 2006, 820, 408.
11. G. Mialon, M. Gohin, T. Gacoin and J.-P. Boilot, ACS Nano, 2008, 2, 2505.
12. E. Allain, S. Besson, C. Durand, M. Moreau, T. Gacoin and J.-P. Boilot, Adv. Funct. Mater., 2007, 17, 549.
13. G. Mialon, S. Turkcan, A. Alexandrou, T. Gacoin and J.-P. Boilot, J. Phys. Chem. C, 2009, 113, 18699.
14. A. Navrotsky, J. Chem. Thermodyn., 2007, 39, 2.
15. R. C. Garvie, J. Phys. Chem., 1965, 69, 1238.
16. G. Caboche, F. Chaput, J.-P. Boilot and J.-C. Niepce, Silic. Ind., 1993, 58, 103.
17. S. Sakurai, A. Namai, K. Hashimoto and S.-I. Ohkoshi, J. Am. Chem. Soc., 2009, 131, 18299.
18. T. Belin, N. Millot, N. Bovet and M. Gailhanou, J. Solid State Chem., 2007, 180, 2377.
19. G. Schimanke and M. Martin, Solid State Ionics, 2000, 136–137, 1235.
20. A. Navrotsky, L. Mazeina and J. Majzlan, Science, 2008, 319, 1635.
21. C. Chanéac, E. Tronc and J.-P. Jolivet, Nanostruct. Mater., 1995, 6, 715.
22. C. Chanéac, E. Tronc and J.-P. Jolivet, J. Mater. Chem., 1996, 6, 1905.
23. P. Scherrer, Gött. Nachr., 1918, 2, 98.
24. G. K. Williamson and W. H. Hall, Acta Metall., 1953, 4, 22.
25. S. Hosokawa, Y. Tanaka, S. Iwamoto and M. Inoue, J. Alloys Compd., 2008, 451, 309.
26. I. F. Guilliat and N. H. Brett, J. Mater. Sci., 1970, 5, 615.
27. G. Mialon, S. Turckan, G. Dantelle, D. Collins, M. Hadjipanayi, R. Taylor, T. Gacoin, A. Alexandrou and J.-P. Boilot, J. Phys. Chem. C, 2010, 114, 22449.
28. A. Huignard, V. Buissette, A.-C. Franville, T. Gacoin and J.-P. Boilot, J. Phys. Chem. B, 2003, 107, 6754.
29. R. S. Meltzer, S. P. Feofilov, B. Tissue and H. B. Yuan, Phys. Rev. B: Condens. Matter, 1999, 60, R14012.
30. V. LeBihan, A. Pillonnet, D. Amans, G. Ledoux, O. Marty and C. Dujardin, Phys. Rev. B: Condens. Matter Mater. Phys., 2008, 78, 113405.
31. X. Batlle and A. Labarta, J. Phys. D: Appl. Phys., 2002, 35, R15.
32. S. Laurent, D. Forge, M. Port, A. Roch, C. Robic, L. Vander Elst and R. N. Muller, Chem. Rev., 2008, 108, 2064.
33. Q. A. Pankhurst, N. K. T. Thanh, S. K. Jones and J. Dobson, J. Phys. D: Appl. Phys., 2009, 42, 224001.
34. E. Tronc, A. Ezzir, R. Cherkaoui, C. Chanéac, M. Noguès, H. Kachkachi, D. Fiorani, A.-M. Testa, J.-M. Grenèche and J.-P. Jolivet, J. Magn. Magn. Mater., 2000, 221, 63.
35. S. Morup, J. Magn. Magn. Mater., 2003, 266, 110.
36. A. Millan, A. Urtizberea, N. J. O. Silva, F. Palacio, V. S. Amaral, E. Snoeck and V. Serin, J. Magn. Magn. Mater., 2007, 312, L5.
37. A. G. Roca, R. Costo, A. F. Rebolledo, S. Veintemillas-Verdaguer, P. Tartaj, T. Gonzalez-Carreño, M. P. Morales and S. J. Serna, J. Phys. D: Appl. Phys., 2009, 42, 224002.
38. M. P. Morales, S. Veintemillas-Verdaguer, M. I. Montero and C. J. Serna, Chem. Mater., 1999, 11, 3058.
39. P. Dutta, A. Manivannan, M. S. Seehra, N. Shah and G. P. Huffman, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 70, 174428.
40. X. Batlle, N. Pérez, P. Guardia, O. Iglesias, A. Labarta, F. Bartolomé, L. M. García, J. Bartolomé, A. G. Roca, M. P. Morales and C. J. Serna, J. Appl. Phys., 2011, 109, 07B524.
41. B. Luigjes, S. M. C. Woudenberg, R. De Groot, J. D. Meeldijk, H. M. Torres Galvis, K. P. de Jong, A. P. Philipse and B. Erné, J. Phys. Chem. C, 2011, 115, 14598.
42. M. Levy, A. Quarta, A. Espinosa, A. Figuerola, C. Wilhelm, M. Garcia-Hernandez, A. Genovese, A. Falqui, D. Alloyeau, R. Buonsanti, P. Cozzoli, M. Garcia, F. Gazeau and T. Pellegrino, Chem. Mater., 2011, 23, 4170.
43. O. Iglesias and A. Labarta, J. Magn. Magn. Mater., 2005, 290–291, 738.
44. M. P. Morales, C. J. Serna, F. Bodker and S. Morup, J. Phys.: Condens. Matter, 1997, 9, 5461.
45. R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki and Y. Taga, Science, 2001, 293, 269.
46. M. Gohin, I. Maurin, T. Gacoin and J.-P. Boilot, J. Mater. Chem., 2010, 20, 8070.
47. M. Gohin, E. Allain, N. Chemin, I. Maurin, T. Gacoin and J.-P. Boilot, J. Photochem. Photobiol., A, 2010, 216, 142.
48. C. Bertail, S. Maron, V. Buissette, T. Le Mercier, T. Gacoin and J.-P. Boilot, Chem. Mater., 2011, 23, 2961–2967.
49. Q. A. Pankhurst, J. Connely, S. K. Jones and J. Dobson, J. Phys. D: Appl. Phys., 2003, 36, R167.
50. R. Hergdt, S. Dutz and M. Röder, J. Phys.: Condens. Matter, 2008, 20, 385214.
51. S. Sun and H. Zeng, J. Am. Chem. Soc., 2002, 124, 8204.
52. S. G. Kwon, Y. Piao, J. Park, S. Angappane, Y. Jo, N. M. Hwang, J. G. Park and T. Hyeon, J. Am. Chem. Soc., 2007, 129, 12571.
53. E. Blums and A. Yu. Chukhrov, J. Magn. Magn. Mater., 1993, 122, 110.
54. R. Massart, E. Dubois, V. Cabuil and E. Hasmonay, J. Magn. Magn. Mater., 1995, 149, 1.
55. J. C. Slonczewski, Phys. Rev., 1958, 110, 1341.
56. G. Salazar-Alvarez, J. Sort, A. Uheida, M. Muhammed, S. Suriñach, M. D. Baró and J. Nogués, J. Mater. Chem., 2007, 17, 322.
57. C. Vichery, I. Maurin, J.-P. Boilot and T. Gacoin, J. Appl. Phys., 2012, 111, 07B541.
58. T. S. Ahmadi, M. Haase and H. Weller, Mater. Res. Bull., 2000, 35, 1869.
59. J. S. An, J. H. Noh, I. S. Cho, H. S. Roh, J. Y. Kim, S. H. Han and K. S. Hong, J. Phys. Chem. C, 2010, 114, 10330.

---