Fe2O3 and Gd2O3 nanoparticles loaded in mesoporous silica: insights into influence of NPs concentration and silica dimensionality

Fine Fe2O3 and Gd2O3 magnetic nanoparticles (NPs) with sizes 7 nm and 10 nm embedded into mesoporous silica have been prepared using a wet-impregnation method. A comparative study of the reactant concentration along with the hosting matrix symmetry on mesostructuring and the magnetic properties of the nanocomposites have been investigated. Reactants with four different concentrations of Fe3+ and Gd3+ ions and silica matrices with two different kinds of symmetry (hexagonal and cubic) have been utilized for the study. The structural characterization of the samples has been carried out by the N2 adsorption/desorption method, high-energy X-ray diffraction (HE-XRD), TG/DTA, and high resolution transmission electron microscopy (HRTEM). The magnetic properties of the nanocomposites have been examined by means of SQUID magnetometry. It has been found that a range of different magnetic states (diamagnetic, paramagnetic, ferromagnetic, superparamagnetic) can be induced by the feasible tailoring of the particle concentration, the porous matrix symmetry and the composition. Furthermore, the existence of a “critical concentration limit” for embedding the particles within the body of the matrix has been confirmed. Exceeding the limit results in the expulsion of nanoparticles on the outer surface of the mesoporous matrix. Revelation of the relationships between particle concentration, matrix symmetry and magnetic properties of the particular composite reported in this study may facilitate the design and construction of advanced intelligent nanodevices.


Introduction
Novel magnetic composite materials based on mesoporous silica have found applications in the elds of catalysis, adsorption, chromatography, chemical sensors and biomedicine. [1][2][3][4][5][6] Particularly in biomedicine, nanocomposites consisting of mesoporous silica loaded with magnetic NPs and specic drugs appear very promising for diagnostic and therapeutic applications. [7][8][9][10] The employment of magnetic resonance imaging, hyperthermia treatment or hi-tec stimuli-responsive targeted drug delivery has improved dramatically with the introduction of these kinds of systems. Undoubtedly, one of the most valuable benets of these materials is their versatility. It stems from their specic inner structure and design. Crucial hosting matrix characteristics like its symmetry and pore size, volume and area are easily tunable during the fabrication process. This enables the further introduction of other structures (e.g. nanoparticles, drug molecules) of the desired size into the matrix body. On the other hand, durability, high thermal stability and low toxicity are the qualities that amorphous silica preserves over a broad range of conditions. Several preparation methods including the one-pot synthesis, wet-impregnation or in situ methods have been reported for mesoporous silica based magnetic composite materials. Zhao et al. 11 have developed an in situ synthesis of magnetic mesoporous silica via a sol-gel process with subsequent precipitation and oxidation. An iron precursor (NH 4 ) Fe(SO 4 ) 2 $6H 2 O has been utilized in order to obtain Fe 3 O 4 nanocomposites. The nal materials exhibited a pore volume density of 0.64-0.96 cm 3 g À1 , a high saturation magnetization value 1.11-5.77 emu g À1 and a high adsorption capacity (up to 212 mg g À1 for lysozyme) depending on the amounts of the reactants. A magnetic mesoporous silica composite has been fabricated by a sol-gel method in a nitrogen atmosphere where iron containing molecules are dissolved to form the sol into which the mesoporous matrix is subsequently merged. 11 The structural changes in the regular host matrix body in response to the incorporation of iron oxide nanoparticles (into the vacant mesopores) is evidenced by the decrease in the diffraction intensity observed in the XRD spectra. Another community of authors reported on the wet impregnation method 12,13 for the preparation of mesoporous silica containing magnetic nanoparticles. Here, metal nitrate solution is mixed with a porous matrix and allowed to dry. Aerwards, the product is calcinated in air. In the case of iron nitrate utilization and calcination in an oxygen atmosphere, Fe 2 O 3 NPs are formed. Jin et al. 14 synthetized a magnetic composite via the pH-adjusting method. They added iron salt solution into the matrix reaction mixture and the adjusted pH value to 7 changing Fe(NO 3 ) 3 into Fe(OH) 3 . The above described methods advantage are control of the magnetic nanoparticle size, use as a nanoreactor, however, the nanoparticles ll the pores and reduce the adsorption capacity of other molecules limiting the size of molecules which can be loaded into the pores to the pores diameter. An important advantage of all the methods mentioned above is the ability to control the size of the nanoparticles. The regular pores of the matrix serve as nanoreactors which constrain the dimensions of the structures embedded inside. Since the pore volume is strictly limited, its occupation by the nanoparticle signicantly reduces its adsorption capacity, and the shape and size of other molecules. 15 Silica nanoparticles doped with gadolinium oxide exhibit promising application potential in biomedicine. Apart from low cytotoxicity of the gadolinium content responsible for the enhanced paramagnetic effect in proton paramagnetic resonance, another quality can be attributed to the system; aer specic modication (3-aminopropyltrimethoxysilane), DNA molecules are allowed to bind to the particles' surface electrostatically. The combination of these properties favors the utilization of the system as an imaging agent or for targeted drug delivery. 3 Similar materials, however, containing NPs on the basis of iron, are adept for hyperthermia treatment. 17 Wang et al. 16 reported the fabrication of silica NPs with large pores, where the co-precipitation method was employed for Fe 3 O 4 NPs introduction.
A host of studies devoted to particular magnetic composites containing nanoparticles of iron and gadolinium oxides has been reported. However, work dealing with their systematic comparison based on an analysis of the hosting matrix symmetry along with the particles' composition and concentration has not yet been performed.
Hence, our objective was to design and examine a variety of systems from the structural and magnetic point of view. The wealth of experimental data was further analyzed with the aim to nd general rather than specic features and trends within the similar systems of the series. For the purpose of the study, a total set of 16 samples were prepared. Namely, nanoparticles of Gd 2 O 3 and Fe 2 O 3 were introduced in four concentrations into the silica matrices of the two different symmetries.

Preparation of blank mesoporous matrices
The SBA-15 mesoporous matrix with a hexagonal symmetry (P6mm) was prepared following the procedure described by Zhao et al. 18 The synthesis was performed in a molar ratio: 1TEOS : 5.9HCl : 193H 2 O : 0.017 P-123. 30 g distilled water was mixed with 120 g 2 M HCl in polypropylene beaker and stirred at 400 rpm at 35 C. 4 g P-123 was added into the reaction mixture. Aer Pluronic dissolution, 8 g TEOS was added into the beaker. The reaction mixture was stirred continuously for 24 hours (400 rpm, 35 C). Further, the mixture was aged in an oven for 24 hours at 80 C. Later the mixture was washed with distilled water. The acquired white powder was kept aging, the product was ltered under vacuum and dried in the air at room temperature. Finally, the dried powder was calcinated in an air atmosphere at 500 C for 7 hours.
The mesoporous matrix SBA-16 with a cubic symmetry (Im 3m) was prepared by the procedure described by Kim et al. 19 The molar ratio of reactants was: 1TEOS : 0.4HCl : 144H 2 -O : 0.0016 P-123 : 0.0037 F-127. 20 g of distilled water was mixed with 92.7 g HCl in a polypropylene beaker, subsequently, 0.38 g P-123 and 1.9 g F-127 were added and stirred at 400 rpm and 35 C. Aer the dissolution of both Pluronics, 8.5 g TEOS was added dropwise. This mixture was stirred at constant conditions (400 rpm, 35 C) for 15 minutes. Further, the mixture was kept ageing in an oven at 100 C for 24 hours. The nal product was ltered, several times and washed with ethanol and allowed to dry at ambient temperature. Aer drying, white powder was calcinated at 500 C in air atmosphere for 7 hours.

Experimental methods
As prepared and calcinated mesoporous matrices were characterized by infrared spectra using Nicolet 6700 FT-IR instrument at ambient temperature with the following settings: transmission mode, a range of 300-4000 cm À1 (32 scans), KBr technique.
Blank as well as modied calcinated samples were examined by the nitrogen adsorption/desorption volumetric method at 77 K utilizing Nova 1200e Quantachrome analyzer. Samples were degassed during 8 hours at 420 K before measurement. This method provided information on the samples' surface area (BET method), external surface area (t-plot method), pore diameter and pore volume (DFT method).
X-ray powder diffraction was used for phase/composition analysis of nanoparticles embedded inside the porous matrices. Diffraction measurements were carried out by synchrotron radiation of energy 60 kV and wavelength l ¼ 0.0207 nm at PETRA III accelerator at DESY, Hamburg. Capton capillaries were lled with powder samples and the scattering intensity was measured as a function of the scattering vector, q, being dened as q ¼ (4p/l)sin q, where 2q is the scattering angle. The obtained diffraction patterns were processed via FIT2D soware employing CeO 2 as a calibration standard. Particles' size was determined by Scherer formula.
Prepared systems were also examined by TEM (Transmission electron microscopy) with a JEOL 2100 operating at 200 kV in STEM mode. Powder samples were dispersed by means of ultrasonication in methanol and added to carbon coated copper grids of mesh size 200.
Magnetic measurements were performed by MPMS 5XL SQUID based magnetometer from Quantum Design. Static dcmagnetizations were recorded in the temperature range 2-300 K in zero-eld-cooling (ZFC) and eld-cooling (FC) protocols. Magnetization vs. applied eld (magnitude) loops were measured at 2 K and 300 K up to 50 000 Oe. In order to determine coercive eld, each sample was thermally demagnetized by heating to room temperature followed by cooling the sample in the absence of an applied eld down to the measuring temperature 2 K or 300 K.

Results and discussion
The general idea of particles' introduction into the hollow matrices is illustrated in Scheme 1. The combination of two different metal ions (Gd 3+ or Fe 3+ ), a pair of matrices with 2Dhexagonal or 3D-cubic symmetry and four different nanoparticle concentrations resulted in the preparation of the series of 16 nanocomposite samples.
FT-IR spectra for the synthesized (red) and calcinated (blue) SBA-15 matrix are shown in Fig. 1. In both IR spectra, (for calcinated as well as synthesized) the O-H bond valence vibration at $3400 cm À1 from silanol Si-O-H bonds can be recognised. The presence of the organic surfactant is conrmed by C-H bond valence vibrations between 3000-2800 cm À1 and deformation vibrations between 1450-1350 cm À1 . These vibrations are absent in the calcinated sample spectra indicating surfactant removal from the pores. Characteristic signals of Si-O-Si bonds are present at 1094 cm À1 and 964 cm À1 for valence vibrations and at 800 cm À1 and 460 cm À1 for the deformation vibration. The occurrence of peaks at 1635 cm À1 and between 2360-2340 cm À1 is attributed to water and carbon dioxide, respectively. These molecules were absorbed from air and their presence is evidence of empty pores in the calcinated matrix.
The pore volume, diameter and surface area were established by the nitrogen adsorption/desorption method at 77 K.   For the cases of higher precursor concentrations (0.5 M and 4 M), the adsorbed nitrogen volume signicantly decreases when comparing samples containing Gd 3+ to the samples containing Fe 3+ . This indicates that aer calcination, the pores show a higher propensity for lling by gadolinium oxide than by iron oxide. The data from the adsorption measurements were utilized for calculation of the specic adsorption parameters (specic surface area S BET , pore diameter d DFT and pore volume V DFT ) for all 16 studied samples. The complete summary of the values established for the nanocomposites containing Fe 3+ and Gd 3+ are shown in Tables 1 and 2, respectively. With the aim of determining the composition and structural phases of nanoparticles embedded inside the matrices, high energy X-ray diffraction (HE-XRD) experiments using synchrotron radiation were performed. The diffraction patterns of the nanocomposites are demonstrated in Fig. 3. For all samples modied with iron precursor, (Fig. 3(a) and (c)) exhibit one broad peak at 2q At higher concentrations of Fe 3+ ions (0.5 M and 4 M), the presence of a crystalline phase was detected as shown in Fig. 4(Ic, Id, IIc and IId). We assume that due to the relatively high precursor concentration in the silica matrix pores, the volume capacity of the pores was not sufficient for connement of the progressive growth of the Fe 2 O 3 NPs. As a consequence, the excess precursor runs over the pores and the particles also formed one external silica surface. This could rationalize the detection of the hematite phase with high crystallinity. Diffraction patterns of the samples with a higher concentration of iron precursor which (Y ¼ 0.5 M; 4 M) show clear evidence of the crystalline phase, and were further used for the estimation of the average particle size via Scherrer formula. 20 The hematite nanoparticles had average sizes of D ¼ 16.8 nm and D ¼ 18 nm for the Fe@SBA-15 and Fe@SBA-16 samples with elevated precursor concentration, respectively, and were found to be signicantly higher than the pore sizes of both silica matrices. This supports the assumption of particle formation out the pores as discussed above.
Intensity vs. scattering angle of the samples containing Gd 3+ nanoparticles is demonstrated on Fig. 3(b) and (d). All the samples exhibit a broad peak in the vicinity of 2q ¼ 3.09 , while the samples with a higher concentration of gadolinium nanoparticles (0.5 M a 4 M) are characteristic of the additional broad peak occurring at 2q ¼ 6.2 . All of these features are assigned to mesoporous matrix with the amorphous nature. In spite of the progressive abundance of nanoparticles in the matrices, only weak peaks corresponding to Gd 2 O 3 (space group Ia 3 (no. 206), JCPDS no. 43-1014) phase are observed in the diffraction patterns. We assume that almost all diffraction peaks assigned to Gd 2 O 3 are hidden under the broad peaks of the silica matrix. Diffusive patterns typical of all samples containing Gd 3+ NPs, Fig. 4III(a-d) and IV(a-d) point to differences between the   nanocomposites modied by iron and gadolinium, even though they are prepared with the same concentration and silica matrix symmetry. This suggests that the mechanism of impregnation and growth of the nanoparticles in the mesopores is rather different in the case of Fe 3+ and Gd 3+ ions. TEM images taken in the STEM mode are shown in Fig. 5 and  6. The process of progressive pore lling with increasing concentration of Fe 3+ or Gd 3+ precursors is represented by pronounced darkness of the corresponding area. In the case of Fe@SBA-15 nanocomposite with the highest ion concentration, spherical Fe 2 O 3 NPs conned by cylindrical pores of hexagonal arrangement are clearly recognized. On the other hand, a similar contrast between the nanoparticles and silica matrix in the Gd 2 O 3 nanocomposites was not observed. A detailed TEM study of all the nanocomposite samples show that a small portion of the larger particles are present on the matrices' external surface at the highest concentration of Fe 2 O 3 nanoparticles in the 2D (SBA-15) and 3D (SBA-16) matrices as seen in Fig. 6. This nding is in accordance with the results of the XRD analysis and both support the assumption of the existence of a "critical concentration". Apparently, with increasing concentration of Fe 3+ precursor, the critical limit can be exceeded when the particles are not only formed inside of pores, but are also   expelled out from the porous system. We suppose that the strong XRD signal documented for Fe@SBA-15 and Fe@SBA-16 with NPs concentrations Y ¼ 0.5 M and Y ¼ 4 M, see Fig. 3(a), (c) and 4(Ic, Id, IIc and IId) comes from the crystalline phase located on the external surface of the matrices. A similar process of NPs expulsion on the external surface of the nanocomposites containing Gd 2 O 3 NPs was not observed.
The direct evidence of progressive lling of the silica pores depending on the increasing concentration of Fe 3+ and Gd 3+ precursors is demonstrated by the TEM-EDS measurements, see Fig. 7. The very scarce but explicit presence of Fe 3+ and Gd 3+ ions was documented unambiguously even in samples with the lowest concentration of metal precursors, Fig. 7(a) and (d). Further, the density of points along with the (colour) contrast were found to be gradually enhanced for the sample series with the increasing concentration of the metal precursors, revealing the progressive incorporation of the NPs into the matrix pores.
A deeper investigation of the differences between Fe 3+ and Gd 3+ samples along with the elucidation of the particle's growth mechanism was carried out by means of thermal analysis experiments, see Fig. 8. At rst, we studied thermal decomposition of four mesoporous samples SBA-15 and SBA-16 impregnated by Fe(NO 3 ) 3 $9H 2 O and Gd(NO 3 ) 3 $6H 2 O, Fig. 8(a).
While the DTA/TG curves of Gd@SBA-15 (dark blue line) and Gd@SBA-16 (light blue line) are almost identical and both point to sample decomposition in three subsequent steps, the decomposition of the samples Fe@SBA-15 (red line) and Fe@SBA-16 (orange line) proceed with only one sharp observed peak. Moreover, differences in the decomposition process of samples containing Fe 2 O 3 NPs loaded in the SiO 2 matrix with different symmetries of SBA-15 (Fe@SBA-15) and SBA-16 (Fe@SBA-16) were conrmed. This is due to differences in the symmetry and pore sizes of SBA-15 (hexagonal) and SBA-16 (cubic symmetry).
Thermal decomposition of the pure nitrate salts, Fe(NO 3 ) 3 -$9H 2 O and Gd(NO 3 ) 3 $6H 2 O, which serve as metal precursors was also examined, see Fig. 8(b). While Fe(NO 3 ) 3 $9H 2 O (red line) decomposes in one sharp step in the temperature range 130-180 C, the decomposition of Gd(NO 3 ) 3 $6H 2 O (blue line) is slower and it takes place in three steps in the temperature range 100-530 C.
Taking into account the results of the DTA/TG analysis, the XRD and TEM observations can be explained. When the concentration of the Gd or Fe salts is low during decomposition of the nitrates in the pores of the silica matrix, the released molecules of water or nitrogen oxide can freely diffuse from the pores and during decomposition, and the formation of the corresponding oxides Fe 2 O 3 or Gd 2 O 3 takes place. When the concentration of the salts increases, the nanoparticles with the higher concentration start to ll the pores and the release of gases during the decomposition of the salts from the pore system is more difficult.
On the contrary, in the case of Gd(NO 3 ) 3 $6H 2 O the decomposition is slow and the gases are allowed to diffuse slowly out of the pores. However, in case of the Fe(NO 3 ) 3 $9H 2 O which decomposes in one fast step, the partial pressure inside of the pores is so high that the nanoparticle plugs which block the pores are pushed out onto the external surface of the silica.  The magnetic properties of all the samples were also scrutinized in order to recognise and highlight differences between them. Susceptibility and magnetization dependences on the temperature and applied eld were recorded and compared for this purpose. Fig. 9 shows the temperature dependence of the magnetic dc-susceptibility of the samples obtained in the ZFC and FC protocols.
The nanocomposites containing iron oxide ( Fig. 9a and b) exhibit the hallmarks of superparamagnetic systems: (i) the presence of a maximum in the ZFC curves at the blocking temperature T B (T B $45 K), (ii) the merging of the ZFC/FC curves above the blocking temperature. On the other hand, the samples containing Gd 3+ NPs manifest paramagnetic behaviour that is typical of almost all Gd 3+ salts, 21 bulk and Gd 2 O 3 particles 22 at room temperature. These samples also show enhanced susceptibility with increased nanoparticles' concentration ( Fig. 9c and d). Intriguingly, this dependence was not broken in the case of samples containing Fe 3+ NPs. As it is seen in Fig. 9a and b, nanocomposites with Fe 2 O 3 NPs loaded in both 2D (Fe@SBA-15, see Fig. 9a) and 3D (Fe@SBA-16, see Fig. 9b) matrices are characteristic of lower susceptibility values when compared the highest (4 M, dark red) to the lowest (0.5 M, red) concentration. This most likely points to the existence of a "critical" Fe 2 O 3 NPs concentration at which the internal porous system is optimally lled by the particles. Above this "critical" concentration, a portion of NPs is ejected out of the internal surface. Since the growth mechanism of NPs is controlled by the size of the internal porous system, the Fe 2 O 3 particles located out of pores are allowed to exhibit larger sizes (about 80 nm, see Fig. 6). However, the critical size for superparamagnetic particles of Fe 2 O 3 is 35 nm, 23 NPs exceeding this limit are not in a superparamagnetic state. As a consequence, the susceptibility is reduced in samples with the highest concentration of Fe 2 O 3 NPs (Fe@SBA-15 4 M and Fe@SBA-16 4 M), Fig. 9a and b.
Magnetization curves (M(H)) measured at temperatures 2 K and 300 K, Fig. 10, conrm the results from temperature dependence of the susceptibility. In all four samples with the lowest concentration of Fe 3+ and Gd 3+ NPs Y ¼ 0.01 M (namely: Fe@SBA-15 0.01 M, Fe@SBA-16 0.01 M, Gd@SBA-15 0.01 M, Gd@SBA-16 0.01 M) the diamagnetic state was observed at room temperature (300 K) and paramagnetic behaviour appears with decreasing temperature to 2 K due to the thermally activated processes.
In the nanocomposites containing Fe 3+ NPs with higher concentrations, the samples Fe@SBA-15, Fe@SBA-16, 0.5 M and 4 M (see Fig. 10c and d), typical superparamagnetic behaviour was conrmed as well as from the M(H) loops. At a temperature of 300 K (T > T B $ 45 K), Fig. 10c-red line, 10d-red line, the magnetic moments of Fe 2 O 3 NPs can freely uctuate in the external magnetic eld leading to a lack of superparamagnetism and coercivity. Below T B (T < T B $ 45 K), the  magnetic moments are blocked in the external magnetic eld direction and coercivity caused by ferromagnetic interaction appears, Fig. 10c-blue line and 10d-blue line.
On the other hand, the samples containing Gd 3+ NPs Gd@SBA-15 and Gd@SBA-16, Fig. 10g and h, with the same nanoparticle concentration as the samples containing Fe 3+ NPs show like-paramagnetic behaviour in the temperature range 10-300 K, which was conrmed by the measured M(H) curves (Fig. 10) and the ZFC/FC curves (Fig. 9). The eld dependence of the magnetization measured at 2 K in samples with higher concentration of Gd 3+ NPs also displays a weak "wasp waist" 24 remanencefree hysteresis, Fig. 10g blue line, 10hblue line. This non-typical behaviour indicates that the anisotropy eld in the studied Gd 3+ NPs is larger than 50 000 Oe and may be the result not only on the reduction of the dimensions from the bulk to the nanoscale and a dramatic increase in surface area but also by the associated crystalline lattice expansion. This expansion may result in longer Gd-Gd bonds and weaker ferromagnetic coupling.

Conclusion
Nanocomposite materials containing Fe 2 O 3 and Gd 2 O 3 nanoparticles with the same concentrations were prepared by the nanocasting method. Structural analysis conrms that the nanocasting provides a simple procedure for which the silica matrix serves as a nanoreactor for the growth of the nanoparticles. Temperature and eld dependencies of the magnetization of all samples were compared. The composite containing Fe 2 O 3 nanoparticles show superparamagnetic behaviour with a blocking temperature around 45 K. Otherwise, paramagnetic properties were observed for the sample with Gd 2 O 3 (above 10 K). Additionally, due to free pores, the silica matrix could serve as a medium for the encapsulation of drugs to create magnetically vectored drug delivery systems.

Conflicts of interest
There are no conicts to declare.