Photomagnetism in 5 nm nanocrystals of the alkali cation free-CoFe Prussian blue analogue embedded in a silica matrix

5 nm nanocrystals of CoFe Prussian blue analogue totally exempt from any alkali cation were prepared. Their photomagnetic properties were compared to those of the corresponding powder made of 150 nm particles as well as to those of 5 nm nanocrystals of CoFe PBAs embedded in comparable silica matrices, made of CoIIFeIII pairs and prepared under various conditions. The photomagnetic investigation of the nanoparticles exempt of any alkali cation clearly shows that they are transformed by light and the comparison of their photomagnetic properties to those of the powder made of particles of bigger size with the same chemical composition suggests that the species involved in the switching properties are surface species. Furthermore, the comparison of the magnetic properties of nanoparticles prepared under various conditions also suggests that the aggregation state of the nanoparticles in the porous channels of the silica matrix, by modulating inter-particle interactions, plays a predominant role in the magnetic properties of the nanoparticles assemblies.


Introduction
Since their physical properties can be tuned by the application of various external stimuli, switchable molecular materials are promising candidates for the preparation of multifunctional devices [1][2][3][4][5][6][7]. Among them, photoswitchable compounds are of particular interest because the associated photo-induced electronic transitions, and therefore the addressing time, are extremely short. Spin-crossover [1] and charge transfer [8] compounds are among the most studied photoswitchable inorganic systems. These compounds are usually obtained in powder or in single-crystal form, so that their integration into real devices necessitates their preparation as nano-objects as well as the knowledge of the effect of size reduction on their properties. Nevertheless, still few works evidence photoswitchable properties in nanoparticles [9][10][11][12][13][14][15][16], and the effect of size reduction on these properties remains an open question. Several CoFe Prussian blue analogue (PBA) nanoparticles were prepared [14][15][16][17][18][19][20] and a photomagnetic effect was evidenced in some of them [14][15][16]. In this family of compounds, the photomagnetic effect has been widely studied in samples prepared by precipitation in aqueous solution generally made of particles having size ranging from several tens to several hundred nanometers. In such compounds, it has been shown that the photomagnetic effect is due to the photo-induced Co III (Low Spin (LS)) Fe II  Co II (High Spin (HS)) Fe III electron transfer in diamagnetic Co III (LS)Fe II pairs [21][22][23][24]. It has also been shown that these diamagnetic Co III (LS)Fe II pairs form during the synthesis when alkali cations are inserted within interstitial sites of the well-known face centered cubic (fcc) structure of PBAs and the photomagnetic effect depends on the amount and on the nature of the inserted alkali cations [25,26].
In order to investigate the size reduction effect on the photomagnetic properties of CoFe PBAs, we developed an original strategy based on the use of the nanoporosity of ordered mesoporous silica monoliths as nanoreactor, which allows a precise control of both the size and chemical composition of the particles [27,28]. The growth of the nanoparticles being limited by the silica walls, the diameter of the particles is fixed by the diameter of the pores. In this study, the diameter of the pores and therefore the diameter of the PBA nanoparticles are fixed at 5 ± 0.5 nm. The silica walls are 4 nm thick, so that the distance between two cylindrical pore centers of the silica matrix is 9 nm [27]. Furthermore, as the chemical composition of the reaction medium in the pores can be precisely adjusted, the control of the chemical composition of the nanoparticles should be possible. We started the investigation of size reduction effect on the photo-induced switching phenomenon by studying one of the most well-documented CoFe PBA photomagnetic compound, with the chemical formula Rb2Co4[Fe(CN)6]3.3•nH2O [25]. Indeed, this compound in powder form combines an important photomagnetic effect and a reproducible synthesis. It is mainly made of Co III (LS)Fe II pairs in its ground state. In the metastable photo-excited state, it mainly contains Co II (HS) Fe III pairs like the alkali cation free CoFe PBA (chemical formula Co 4 [Fe(CN) 6 ] 2.7 •nH 2 O), which is very often taken as a model compound for the photo-excited state, and what we also did for the study of the size reduction effect. Thus, by studying 5 nm nanocrystals of the CoFe PBA containing two rubidium cations per unit cell and of the alkali cation free-CoFe PBA (their synthesis is described in reference 29), we recently evidenced a peculiar core-shell structure of their ground states [29]. The contributions of the well-defined core and shell components were extracted from linear combinations of X-ray absorption spectra. The two nanocomposites are named RbNanoCoFe and KNanoCoFe thereafter. The 'alcali cation free' compound is named KNanoCoFe here because it was prepared from the K 3 [Fe(CN) 6 ] metallic salt (it is named NanoCoFe in reference 29). If the amount of potassium cation in the final product is so low that potassium can be detected neither by elemental analysis nor by EDX, the compound (and especially the surface of the compound) could be not completely exempt of it. The core-shell structure that was proposed from the X-ray absorption spectroscopy investigation at the transition metal K-edges is schematized in Figure 1.  [29].
The structure of the nanoparticles can be described as follow. They are constituted of a 3 nm spherical core approximately made of 20 unit cells of the well-known face-centered cubic structure of the PBAs, which roughly correspond to 70 CoFe pairs. Whereas these CoFe core pairs are in the Co III (LS)Fe II state in RbNanoCoFe, they are in the Co II (HS)Fe III state in KNanoCoFe. Whatever the state of the CoFe core pairs, the core is surrounded by a similar shell, the thickness of which corresponds to 2 to 3 metallic layers. This shell, mainly made of Co II (HS) and Fe III species, exhibits a structure different from that of the core [29]. Work is in progress to better characterize it.
RbNanoCoFe exhibits an important photomagnetic effect. The nanocrystals were therefore also studied by X-ray absorption spectroscopy in their photo-excited state. This study shows that after irradiation in the photo-excited state, the spectra of the nanocrystals can no longer be decomposed into two well-defined contributions corresponding to a core and a shell as in the ground state. The nanocrystals rather exhibit one single phase very close to the well-known fcc structure of PBAs mainly made of Co II (HS)Fe III pairs [29]. Thus, the effect of light on the core of the nanocrystals almost corresponds to the well-documented photo-induced Co III (LS)Fe II  Co II (HS)Fe III electron transfer, already evidenced in bigger size particles. Unexpectedly, the shell of the nanocrystals also undergoes a significant photo-induced structural change. This photo-induced transformation can be either the consequence of the photo-induced Co III (LS)Fe II  Co II (HS)Fe III electron transfer taking place in the core and accompanied by a lengthening of the lattice parameter and therefore of the volume of the core or another photomagnetic effect that has not been described yet. In order to better understand the behaviour of the shell of the nanocrystals under irradiation, one has to study the effect of light on similar shell species in a compound totally exempt from photomagnetic Co III (LS)Fe II pairs. As these photomagnetic Co III (LS)Fe II pairs form in the presence of alkali cations, one solution lies in the preparation of similar nanocrystals totally exempt of alkali cations. In this work, we report the preparation of nanocrystals totally exempt from any alkali cation and which therefore contain the minimum possible amount of photomagnetic Co III (LS)Fe II pairs, if not any of them. In this way, the core of the nanoparticles is expected to be mainly composed of non-photomagnetic Co II (HS)Fe III pairs surrounded by a shell similar to that of KNanoCoFe and RbNanoCoFe. The nanocrystals are called HNanoCoFe thereafter. They were characterized by X-ray diffraction and infrared spectroscopy and their magnetic properties were compared to those of KNanoCoFe. The photomagnetic properties of HNanoCoFe were then investigated and compared to those of the corresponding CoFe PBA powder made of 150 nm particles (HCoFe). At last, the magnetic properties of HNanoCoFe in the photo-excited state are compared to those of 5 nm nanocrystals of CoFe PBAs embedded in comparable silica matrices, made of Co II Fe III pairs and prepared under various conditions.

II. Results and Discussion
The nanocrystals HNanoCoFe were prepared within the pores of an ordered mesoporous silica monolith exhibiting a 2D hexagonal organization by adapting a procedure described elsewhere [27]. Briefly, the first PBA precursor (Co(II) cation), added during the synthesis of the silica monolith, is homogeneously distributed as a low-condensed species within the cylindrical and parallel pores after thermal decomposition of the surfactant. The second PBA precursor ([Fe(CN) 6 ] 3-entity) is then introduced in the pores via a simple impregnation step. Thus, the reaction medium can be precisely adjusted. The concentration of Co 2+ ions is determined by the amount of Co salt added during the synthesis of the monolith and the concentrations of all other species correspond to those of the impregnation solution. In order to prepare nanocrystals completely free of alkali cation, all K + cations of the potassium ferricyanide solution were replaced by protons through two salt metathesis reactions described in the experimental part. Ferricyanide is first precipitated as a silver salt, in which no trace of potassium could be detected by EDX. [Fe(CN) 6 ] 3complexes are redissolved in water by addition of hydrochloric acid, giving an acidic ferricyanide aqueous solution exempt of potassium cations. The ferricyanide concentration of the solution is precisely determined by UV-Vis absorption spectroscopy. Impregnation of the silica monolith by this solution leads to the formation of PBA in the pores of the monolith through a nucleation-growth process in confined space resulting in spherical 5 nm monocrystalline CoFe PBA nanoparticles (TEM images of such nanoparticles have already been presented several times [30], representative images are shown in Figure S1). The X-ray powder diffractogram ( Figure S1) of HNanoCoFe is very close to that of KNanoCoFe, showing that the nanocrystals in both compounds exhibit a very similar crystallographic structure. In a previous work, we investigated by infrared spectroscopy such nanocomposites [29] and we could identify four main cyanide bands, which were assigned as follow: a band at 2166 cm -1 assigned to Co II (HS)-NC-Fe III core species in a core mainly made of Co II (HS)-NC-Fe III pairs, a band at 2130 cm -1 assigned to Co III (LS)-NC-Fe II core species in a core mainly made of Co III (HS)-NC-Fe II pairs, a band at 2190 cm -1 assigned to Co II (HS)-NC-Fe III shell species, and a band at 2119 cm -1 assigned to Fe III -CN terminal species or a minority of Co III (LS)-NC-Fe II pairs in a core mainly made of Co II (HS)NCFe III core species. The infrared spectra of HNanoCoFe and of KNanoCoFe over the energy range corresponding to the stretching vibration of cyanide are shown in Figure 2. Both spectra exhibit the same bands, showing that the nanocrystals contained in both compounds are very close, except that HNanoCoFe is completely exempt of alkali cation. The spectra of both compounds, HNanoCoFe and of KNanoCoFe, exhibit a main band at 2165 cm -1 . This band is assigned to cyanide in the Co II (HS)-NC-Fe III linkages forming the well-known fcc structure of CoFe PBAs [31][32][33][34][35][36][37] and contained in the cores of the nanocrystals. The same position of this band in both compounds shows that the core of the nanoparticles are made of exactly the same Co II (HS)-NC-Fe III linkages. Both spectra exhibit a second band of weaker intensity situated at 2125 cm -1 and at 2119 cm -1 on the spectra of HNanoCoFe and KNanoCoFe respectively. Given the energy position and the intensity of this band, it can be ascribed either to very small amounts of Co III (LS)-NC-Fe II linkages in the well-known fcc structure of CoFe PBAs, that is to say in the core of the nanocrystals, or to terminal Fe III -CN species belonging to the very last layer at the surface of the nanocrystals. The presence of terminal Fe III -CN species is expected because i) all the Fe(CN) 6 entities, are linked by the six N sides of the cyanides to Co cations except those belonging to the last layer at the surface, where Fe III -CN terminal groups can be found, and ii) the nanocrystals were formed in the presence of an excess of [Fe(CN) 6 ] 3-entities. Furthermore, the slight but significant shift of this band from one compound to the other supports the assignment to terminal Fe III -CN species. Indeed, if this band around 2120 cm -1 was due to the presence of a small amount of Co III (LS)-NC-Fe II linkages in the core of the nanocrystals mainly made of Co II (HS)-NC-Fe III linkages, the shift of this band from one compound to the other could be explained by different network strains resulting from different Co III (LS)-NC-Fe II /Co II (HS)-NC-Fe III linkages ratio in both compounds. In that case, this presence of Co III (LS)-NC-Fe II linkages would also induce a shift of the band at 2165 cm -1 because of the different network strains on the two types of linkages. The fact that the band assigned to the Co II (HS)-NC-Fe III linkages is exactly at the same position in both compounds is rather in favor of the presence of Fe III -CN terminal species. In that case, the shift of the band assigned to terminal Fe III -CN groups can be explained by different interactions between the CN terminal group and K + and H + cations in the case of KNanoCoFe and only H + cations in the case of HNanoCoFe (due to different synthesis conditions). At last, both spectra exhibit a shoulder of significant intensity and approximately located at 2190 cm -1 in KNanoCoFe and at 2197 cm -1 in HNanoCoFe. We proposed in reference 29 to assign this shoulder, present with variable intensity on most of the spectra of CoFe PBAs, to Co II (HS)-NC-Fe III surface species. These species located in the shell of the nanoparticles exhibit a structure different from the Co II (HS)-NC-Fe III linkages contained in the cores. As in the case of the previous band, its slight frequency shift from one compound to the other supports the assignment to surface species. The magnetic properties of HNanoCoFe were first studied before irradiation. The temperature dependence of the zero-field-cooled (ZFC) and field-cooled (FC) magnetization curves of HNanoCoFe are shown in Figure 4. They are very close to those already reported for the nanocomposite directly prepared from the K 3 [Fe(CN) 6 ] salt (KNanoCoFe) [28]. The ZFC magnetization curves of HNanoCoFe and KNanoCoFe exhibit a maximum at very close temperatures (5.4 K for HNanoCoFe against 4.6 K for KNanoCoFe), which is in agreement with very close chemical compositions, very close structures and very close particles sizes for both compounds. In order to further compare both compounds, Alternating-Current (AC) magnetic susceptibility measurements were performed. AC susceptibility measurements were already performed for KNanoCoFe; the whole study reported in reference 28 showed that the nanocrystals contained in KNanoCoFe mainly behave as superparamagnets but their magnetic behaviour slightly deviates from pure superparamagnetism. The temperature dependence of the magnetization relaxation time  was well reproduced with an Arrhenius law (equation 1) where  0 is the attempt time and its value varies over the 10 -8 to 10 -12 s range for pure superparamagnets, k B is the Boltzmann constant and E a is the average energy barrier for the spin-flip. In the case of KNanoCoFe, the value obtained for  0 shows deviation from pure superparamagnetic behaviour, which was explained by three possible reasons: i) weak interparticle interactions, ii) surface effects or iii) presence of Fe(CN) 6 vacancies responsible for some kind of frustration and cluster-glass-like behaviour [28]. As the origin of the deviation from pure superparamagnetism is not known, other expressions likely to reproduce the data will not be considered here. The parameters extracted from the AC susceptibility measurements for HNanoCoFe ( Figure S2a) and the fit of the Arrhenius equation to the temperature dependence of the magnetization relaxation time  are reported in Table 1, where they are compared to those for KNanoCoFe. The  0 values are rather comparable for HNanoCoFe and KNanoCoFe. The amount of alkali cation being extremely small in KNanoCoFe (the amount of alkali cations is below the detection threshold for elemental analyses) [38] and zero in HNanoCoFe, the chemical composition of the coordination polymer is therefore close to Co 4 [Fe(CN) 6 ] 8/3  4/3 (where  stands for [Fe(CN) 6 ] vacancies) in both compounds, so that the amount of Fe(CN) 6 vacancies can be considered as being the same in both compounds. Therefore, the slightly higher  0 value in HNanoCoFe can arise from slightly weaker interparticle interactions in HNanoCoFe due to different aggregation states of the nanoparticles in the pores or from different surface behaviours. Nevertheless, the behaviour of HNanoCoFe is not far from that of KNanoCoFe and similar conclusions can be drawn, in agreement with comparable structure, chemical composition and particle size for both compounds.  The magnetization of HNanoCoFe was recorded as a function of time under irradiation in the visible range at 10 K (Figure 3). Under irradiation, the magnetization of HNanoCoFe increases slightly but significantly and the magnetization after irradiation once the laser is switched off is clearly higher than before irradiation. Therefore, HNanoCoFe exhibits a weak but unquestionable photomagnetic effect. It is worth noticing here that KNanoCoFe also exhibits a photomagnetic effect at least comparable to that of HNanoCoFe ( Figure S3). As the sample was irradiated at 10 K, temperature above the blocking temperature of the nanocrystals, the increase in magnetization cannot be assigned to a reorientation of the magnetic moments of the nanocrystals in the external magnetic field only. The temperature dependence of the magnetization was measured before and after irradiation to check the reversibility of the phenomenon ( Figure S4). The two curves being very close to one another, the temperature at which the initial magnetization is fully recovered was determined as the temperature for which the difference of the magnetization after and before irradiation goes to zero (115 K, Figure S4). Unexpectedly, HNanoCoFe exhibits an unquestionable and reversible photomagnetic effect. In order to better understand this photomagnetic effect, the magnetic properties of HNanoCoFe in the photo-excited state (HNanoCoFe*) were compared firstly to those of HNanoCoFe in the ground state. Secondly, the photomagnetic effect exhibited by HNanoCoFe was compared to that of the corresponding powder with the same structure and chemical composition and made of 150 nm particles. At last, the magnetic properties of HNanoCoFe in the photo-excited state (HNanoCoFe*) were compared to those of 5 nm nanocrystals of CoFe PBAs embedded in comparable silica matrices, made of Co II Fe III pairs and prepared in various conditions. The ZFC/FC magnetization curves of HNanoCoFe and HNanoCoFe* are shown in Figure 4. The ZFC/FC magnetization curves of HNanoCoFe* are different from those of HNanoCoFe, which confirms the weak but significant photomagnetic effect exhibited by HNanoCoFe. The FC magnetization value at 2 K and the ZFC magnetization maximum value at T max increase by 10% and 20% respectively after irradiation. The temperature of the ZFC magnetization maximum is called T max thereafter. Within the error bar, the T max values determined from the temperature dependence of dM ZFC /dT ( Figure S5, Table  1) are the same before and after irradiation for HNanoCoFe and HNanoCoFe*. Irradiation cannot change the aggregation state of the nanoparticles. Therefore, the increase in magnetization reflects an increase of the magnetic moment of each nanocrystal after irradiation, which can have several origins. First, a photo-induced oxidation or reduction of Co or Fe ions by an agent external to the nanocrystals could happen, but it is unlikely since the photomagnetic effect is reversible. Second, even if there is no alkali cation inserted in the nanocrystals, the photo-induced Co III (LS)Fe II  Co II (HS)Fe III electron transfer in very few Co III (LS)Fe II diamagnetic pairs cannot be excluded. Indeed, very few Co III (LS)Fe II pairs located in the core of the nanocrystals could be formed due to the random distribution of Fe(CN) 6 vacancies. Nevertheless, no IR band clearly assigned to such Co III (LS)Fe II diamagnetic pairs were detected in the infrared spectrum of HNanoCoFe (as discussed above) whereas the bands of cyanide species linked to Fe II ions are significantly more intense than those of cyanide species linked to Fe III ones. Third, a photoinduced transformation of the structure of the nanocrystals without electron transfer like the photoinduced structure transformation of the shell of RbNanoCoFe detected by X-ray absorption spectroscopy [29] is also possible. In that case, the phenomenon would essentially affect the surface of the nanocrystals. These two last phenomena differ by the location of the involved species: the photo-induced Co III (LS)Fe II  Co II (HS)Fe III electron transfer would take place in the core of the nanoparticles, having the well-known fcc structure of PBA (in the ground state, the surface species are mainly made of Co II (HS) and Fe III species with a peculiar structure), while the photo-transformation of the peculiar surface species would take place in the shell of the nanocrystals. In order to identify which of these two phenomena occurs, one has to vary the surface area-to-volume ratio. The higher the surface area-to-volume ratio, the stronger the surface effects are expected to be. CoFe PBA particles completely exempt from alkali cation and having bigger size than the nanocrystals embedded in the silica matrix were also synthesized (see experimental part) and studied. The sample is named HCoFe thereafter.
On the SEM images of HCoFe ( Figure S6), one can see cuboidal particles, the average size of which can be estimated around 150 nm. The infrared spectrum of HCoFe ( Figure S7) shows the same cyanide bands as the spectrum of HNanoCoFe except that i) the relative intensities of the bands are different and ii) they are slightly shifted due to confinement or size reduction effect. Such a shift is always observed when the IR spectra of nanoparticles embedded in the silica matrix are compared to those of the powder made of particles of bigger size with same chemical composition [27]. Both compounds are therefore made of the same species. The main band at 2160 cm -1 is assigned to Co II (HS)Fe III core species. The relative intensities of the bands, which were assigned to surface species on the spectrum of HNanoCoFe (Co II (HS)Fe III shell species at 2197 cm -1 and Fe III -CN terminal species at 2125 cm -1 ) are weaker in the powder made of the 150 nm particles, which confirm the assignment of these bands to surface species. The main difference between HNanoCoFe and HCoFe is therefore the surface area-to-volume ratio. The particles of HCoFe were diluted with a silica powder in order to reach the same Co/Si ratio as in HNanoCoFe and the ZFC/FC magnetization curves of this sample were recorded before and after irradiation in the same conditions that were used for HNanoCoFe ( Figure S8). The ZFC/FC magnetization curves of HCoFe in the silica powder are nearly the same before and after irradiation. This suggests that the amount of photo-switchable species is much smaller in HCoFe than in HNanoCoFe.
As the surface area to volume ratio of the particles is much smaller in HCoFe than in HNanoCoFe, this suggests that the photo-switchable species are located at the surface of the particles. An X-ray absorption spectroscopy investigation at the Co and Fe K-edges of HNanoCoFe and HCoFe is planned in order to clearly identify the photo-switchable species and to study further this peculiar switching phenomenon exhibited by the nanoparticles. At last, the magnetic properties of HNanoCoFe* (photo-excited state) were compared to those of 5 nm nanocrystals of CoFe PBAs embedded in comparable silica matrices, made of Co II Fe III pairs and prepared in various conditions: HNanoCoFe (ground state; excess of H + and [Fe(CN) 6 ] 3-), KNanoCoFe (ground state; excess of H + , [Fe(CN) 6 ] 3-and K + ) [28] and RbNanoCoFe* (photo-excited state, excess of H + , [Fe(CN) 6 ] 3-and K + ; large excess of Rb + ) [29]. The magnetic properties of RbNanoCoFe in the ground and the photo-excited states were briefly presented in reference 29. These properties are recalled and further discussed here. The ZFC/FC magnetization curves are shown in Figure S9a. Before irradiation, the ZFC/FC magnetization is low because the core of the nanocrystals is mainly made of Co III (LS)Fe II diamagnetic pairs. The increase in magnetization upon cooling and the irreversibility of the magnetization between the ZFC and FC conditions below 12 K, can be explained by the presence of the peculiar surface contribution mainly made of Co II (HS) and Fe III ions [29]. Contrary to the few Co II (HS) ions (0.7 per unit cell) dispersed in the diamagnetic core, the magnetic moments of the majority Co II (HS) and Fe III surface species can interact and result in the weak but significant collective magnetic behaviour, reflected by the irreversibility between the FC and ZFC magnetization curves. After irradiation, the ZFC magnetization curve of RbNanoCoFe* exhibits a maximum at 8.3 K. The irreversibility between the ZFC and FC magnetization curves of RbNanoCoFe* up to 20 K suggests the occurrence of inter-particle interactions, which are supported by AC susceptibility measurements for RbNanoCoFe* reported in reference 29. Indeed, the AC magnetic data were well fitted by using a Vogel-Fulcher model, which is well suited to take into account inter-particle interactions. All the magnetic data are gathered in Table 1. The magnetic data for HNanoCoFe*, HNanoCoFe and KNanoCoFe are also gathered in Table 1. The AC magnetic measurements for HNanoCoFe* are shown in Figure S2b. It is striking that the magnetic properties of HNanoCoFe* are much closer to those of HNanoCoFe and KNanoCoFe than to those of RbNanoCoFe*. The main differences between the magnetic properties of the three first compounds and RbNanoCoFe* can be explained in terms of different inter-particle interactions. It is worth recalling here that the particles having the same size as the diameter of the cylindrical pores can aggregate forming chains inside the channels [40], which are separated from each other by 4 nm walls. First, the T max value for RbNanoCoFe* (8.3 K) is significantly higher than the T max value for HNanoCoFe*, HNanoCoFe and KNanoCoFe (5.4 K, 5.4 K and 4.6 K, respectively) and it is usually assumed that T max shifts to higher temperature with increasing inter-particle interactions [41][42][43][44]. Second, the peak of the ZFC magnetization curve, broader for RbNanoCoFe* than for HNanoCoFe ( Figure S10), indicates a broader distribution of energy barriers and relaxation times in RbNanoCoFe*, assignable to a broader distribution of interparticle interactions strength [45] (for all compounds the distribution of the particle size is very narrow as it is determined by the diameter of the pores, 5 nm here). Third, the ZFC and FC magnetization curves of HNanoCoFe*, HNanoCoFe and KNanoCoFe do not show the same important irreversibility displayed by those of RbNanoCoFe* up to 11 K above T max , which is another sign of the broader distribution of energy barriers in RbNanoCoFe* due to a broader distribution of inter-particle interactions. Such strengthened inter-particle interactions in RbNanoCoFe* have to be related to the aggregation state of the nanocrystals (the nanocrystals can be either well dispersed within the pores or aggregated along the cylindrical pores forming nanocrystals chains). The presence of an excess of Rb + cations in the reaction medium during the formation of the nanocrystals probably favors their aggregation. Given the strong effect of aggregation states on the magnetic properties, the pure effect of light on the magnetic properties of such nanocomposites can only be evidenced by comparing the magnetic properties of a same sample before and after irradiation. In the case of HNanoCoFe this effect is weak but significant and is yet to be fully understood.

III. Conclusions
In a previous work, we evidenced the core shell structure of CoFe Prussian blue analogues and a peculiar behaviour of the surface species in the nanoparticles containing two rubidium cations per unit cell (RbNanoCoFe) under irradiation. The photomagnetic study of the nanoparticles exempt of any alkali cation (HNanoCoFe) clearly shows that these nanoparticles are significantly transformed by light, which is not the case of the powder made of bigger size particles having exactly the same chemical composition. These results raise completely new questions about size reduction effect on the photomagnetic properties of PBAs and suggest that the species involved in the switching properties could be surface species; work is in progress in order to better understand this phenomenon. Furthermore, the comparison of the magnetic properties of nanoparticles prepared in various conditions also suggests that the aggregation state of the nanoparticles in the porous channels of the silica matrix, by modulating inter-particle interactions, plays a key role in the magnetic properties of these nanoparticles assemblies.

IV. Additional Information
Supporting information is available online. Correspondence and requests for materials should be addressed to corresponding author.

V. Materials and Methods
Sample preparation: Preparation of monoliths with 2D-hexagonal organization containing cobalt(II) ions with a 0.5% Co 2+ /Si ratio: The synthesis of the monoliths has already been described elsewhere [26]. In a typical procedure, a solution was prepared by mixing 2.4 g of P123 with TMOS (26 mmol, 4 g), as a silica precursor, in a 30 mL polypropylene vial. This solution was stirred in a water bath at 50 °C until the P123 copolymer was dissolved completely and then cooled to room temperature.  6 ] 0.3 M and HNO 3 0.7 M was prepared. The calcined monoliths containing 0.5% Co 2+ /Si ions were immersed in heptane (5 mL) and impregnated with a volume of the acidic impregnating solution equal to 80 % of the porous volume. After 1 h, the heptane was removed, and the monoliths were washed five times with water and air-dried. All monoliths were ground before measurements.
HCoFe: Powder HCoFe was synthesized by addition of 400 mL of a 6.25 × 10 -4 M solution of cobalt(II) nitrate hexahydrate to 50 mL of a 10 -2 M H 3 [Fe(CN) 6 ] aqueous solution. The addition rate was regulated to last 3 h. The powders were centrifuged, washed three times with distilled water, and allowed to dry in air.
FTIR Spectroscopy: FTIR spectra were collected in the transmission mode with a PerkinElmer Spectrum 100 spectrometer.
Magnetic and photomagnetic measurements: About 10 mg of powder was spread on a piece of cardboard and some drops of nujol oil were added to fixate it. Direct Current (DC) data were recorded using a XL-7 SQUID magnetometer. Field Cooled-Zero Field Cooled (FC-ZFC) curves were recorded over the 2-15 K temperature range under a 50 Oe magnetic field. Alternative Current (AC) data were recorded using a XL7 SQUID magnetometer working with an oscillating field of 3 Oe and frequencies ranging from 0.1 to 50 Hz. For photomagnetic measurements, the sample was irradiated at 10 K in the SQUID with a laser diode (635 nm, 25 mW/cm 2 ). Samples were irradiated until reaching a photo-stationary state. The temperature dependence of the magnetization was recorded using a XL-7 SQUID magnetometer under a 5000 Oe magnetic field with a sweeping rate of 2 K.min -1 .

VI. Conflict of Interests
No conflict of interests. VII.