Synthesis of triethoxysilylated cyclen derivatives, grafting on magnetic mesoporous silica nanoparticles and application to metal ion adsorption

The synthesis through click chemistry of triethoxysilylated cyclen derivative-based ligands is described. Different methods were used such as the copper catalyzed Huisgen's reaction, or thiol–ene reaction for the functionalization of the cyclen scaffold with azidopropyltriethoxysilane or mercaptopropyltriethoxysilane, respectively. These ligands were then grafted on magnetic mesoporous silica nanoparticles (MMSN) for extraction and separation of Ni(ii) and Co(ii) metal ions from model solutions. The bare and ligand-modified MMSN materials revealed high adsorption capacity (1.0–2.13 mmol g−1) and quick adsorption kinetics, achieving over 80% of the total capacity in 1–2 hours.


Introduction
The anchoring of cyclen derivatives on silica-based materials has attracted attention over the last two decades. The main application concerns magnetic resonance imaging [1][2][3][4][5][6][7][8][9] or radiolabelling 10,11 with metal-DOTA-functionalized silica-based nanoparticles for theranostic applications, but not only. Other applications relate to ESR studies with Cu(II), 12 or Cu(II) catalysis, 13 Pd(II) recovery, 14 or photoresponsive materials with Eu(III) or Tb(III). 15,16 Therefore applications of silica-based materials functionalized with cyclens are very diverse, promising, and present a high interest. Two ways are used for the functionalization of silica materials with cyclen derivatives. The rst one consists of anchoring a reactive function such as a chloro or amino group on silica using silanes such as chloro or aminopropyltriethoxysilane and then to react this function with an activated acid of DOTA derivative for aminopropyltriethoxysilane or with an amino of cyclen derivative for chloropropyltriethoxysilane. The drawback of this strategy is that the coupling reaction is usually not quantitative and the remaining reactive function on the surface of the material is not completely functionalized. The second possibility consists in preparing triethoxysilylated cyclen derivatives and then coupling the derivatives on the surface of the materials. The preparation involves the coupling reaction of silanes (iodopropyltriethoxysilane, aminopropyltriethoxysilane) with cyclen or activated DOTA derivatives.
Magnetic mesoporous silica nanoparticles (MMSN) have gained much attention the last decade for bio-application, sensing or catalysis and the eld has been comprehensively reviewed. [17][18][19][20][21][22] In the course of our studies on functionalized MMSN for metal complexation, 23 we were interested in the preparation of triethoxysilylated cyclen-based ligands, through click Cu-catalyzed azide-alkyne Huisgen's reaction 24 with azidopropyltriethoxysilane or click thiol-ene reaction 25 with mercaptopropyltriethoxysilane, which have not been investigated for this purpose. The cyclen derivatives were then graed on the surface of MMSN and the materials were used for the extraction and separation of Ni(II) and Co(II) ions.
FTIR spectra were recorded in the 4000-400 cm À1 range using 32 scans at a nominal resolution of 4 cm À1 with a Perkin Elmer 100 FT spectrophotometer equipped with an ATR unit. TEM images were recorded with JEOL 1200 EXII microscope and an JEOL 2200 FS (JEOL Europe SAS, Croissy Sur Seine, France). For the purpose of TEM analysis, the sample particles were dispersed in ethanol and then dropped onto copper grids covered with porous carbon lms. TGA analyses were performed with a thermal analyser STA 409 Lux® (Netzsch) in the range 25-800 C by a heating speed of 5 C min À1 .

Preparation of MMSN
Synthesis of magnetic iron oxide nanoparticles (MIONs). The MIONs were synthesized by thermal decomposition according to a previous reported protocol. 26 Briey, 0.18 g of hydrated iron oxide(II) were mixed with 5 g of docosane and 3.2 g of oleic acid. The mixture was rst stirred under vacuum for 30 min and then placed under Ar ow for another 30 min. Then the reaction occurred at 340 C for 1.5 h. Aer completion of the reaction the MIONs were washed and recovered by three successive centrifugations at 20 krpm for 10 min and redispersion by ultrasounds for few minutes: (1) by adding 15 mL of pentane and 30 mL of a mixture of ether : ethanol (2 : 1, v/v) to the as-synthesized product; (2) by adding 2 mL of pentane and 30 mL of a mixture of ether : ethanol (2 : 1, v/v) to the collected black product; (3) by adding 30 mL of a mixture of ether-: ethanol (1 : 1, v/v) to the collected black product. Finally, the MIONs were stored in 15 mL of chloroform and stabilized by adding 200 mL of oleylamine. Synthesis of MMSN. 27 2 g of CTAB were rst dissolved in 250 mL of distilled water for 1 h at 70 C under stirring. The surfactant solution was then cooled down to 40 C and 5 mL from the previous synthesized MIONs dispersed in chloroform were added. The emulsion was rst put under vigorous stirring without heating for 30 min and then the temperature was risen to 70 C and kept for additional 40 min under stirring. Hence, the phase transfer of MIONs succeeded, 300 mL of distilled water slightly basied by 300 mL of 2 M NaOH was added to the solution under stirring. When the solution reached 60 C, 7.5 mL of TEOS followed by 12.5 mL of EtOAc were quickly added to the mixture. Then the condensation reaction was conducted for 3 h at 70 C under stirring. The nal MMSN were recovered by centrifugation, concentrated to 100 mL and washed two times: twice with water and twice with EtOH at 95%. Aer each centrifugation step the MMSN were redispersed in the appropriate washing solvent under sonication for few minutes. To extract the CTAB the MMSN were soaked twice in a solution of ammonium nitrate (6 g L À1 in EtOH 90%) and put under stirring at 70 C for 2 h. Aer each extraction, the washing steps by centrifugation described above in water and ethanol were performed. Finally, the MMSN were redispersed in 25 mL of EtOH 96%. IR n (ATR) (cm À1 ): 2979.6, 1067.8, 963.6, 797.2, 451.3. TGA (air, 5 C min À1 , 20-800 C) loss mass 14.0%. Zeta potential: z ¼ À32 mV, pH ¼ 7.24. DLS: 183 nm.

Preparation of MMSN-Ln
The MMSN-Ln-Boc/ t Bu were dissolved in TFA/CH 2 Cl 2 (5 : 1) solution, the mixture was stirred at room temperature for 60 minutes. Then the NPs were collected by centrifugation (20 000 rpm at 25 C for 10 min). The material MMSN-L1 was washed successively with EtOH (2 Â 30 mL), saturated NaHCO 3 solution (2 Â 30 mL), distilled water (2 Â 30 mL), acetone ( For isotherm experiments stock solutions of Ni 2+ and Co 2+ (50 mM) were prepared using nitrate salts and the nal concentrations were adjusted by dilution with Milli-Q water (varying between 0.5-25 mM). Samples (10 mg) of magnetic nanoparticles (MNPs) were mixed with 10 mL of metal cation solution in plastic tubes of 50 mL and put on a shaker for 24 hours. Aer each experiment, MNPs were centrifuged (7000g) for 10 min and an aliquot (1 mL) was separated to determine the metal concentration in the remaining solution. The samples were rst diluted 5-10 times (depending on the initial metal concentration) and titrated aerwards with EDTA using xylenol orange as an indicator. For each sample, the titrations were repeated 3 times, and the average concentration was calculated.
The uptake of metal cation by the magnetic nanoparticles was calculated according to the equation: where C 0 is the initial metal concentration, which was also measured by titration; C e is the equilibrium metal concentration; V is the metal solution volume which was kept constant; m is the NPs weight.
For kinetic tests, the magnetic nanoadsorbents (40 mg) were mixed with 40 mL of metal cation solutions (with 10 mM initial metal concentration) and the uptake was measured aer set interval of times. For that purpose, an aliquot of 1 mL was separated, diluted 10 times and the remaining metal concentration was evaluated by titration with EDTA and xylenol orange.
For selectivity test, mixed metal solution was prepared with nal concentration of 5 mM + 5 mM Ni and Co ions. The overall concentration was measured by titration with EDTA and xylenol orange as an indicator. For adsorption experiment, 20 mg magnetic nanoparticles were mixed with 20 mL metal solution in 50 mL falcon tubes, and the tubes were put on a shaker for 24 hours. Aer 24 hours, the nanoparticles were separated from metal solution by centrifugation (7000g) and the remaining metal content was measured by titration ( Table 2). The magnetic nanoparticles were dried under nitrogen atmosphere and the adsorbed metal distribution was determined by EDS analysis.
For the study of stability and reuse, 30 mg of each nanoparticle samples were mixed in 30 mL metal solution with 10 mM metal ion concentration for 24 hours. Aerwards, the solutions with nanoparticles were centrifuged (7000 rpm) and the metal content was estimated by titrations with EDTA and xylenol orange as an indicator. The desorption was carried out by adding 30 mL nitric acid (1 mM) and put on a shaker for 24 hours. Nanoparticles were centrifuged for 15 min (7000 rpm) and the amount of desorbed metal ions was estimated by titrations with EDTA. Aerwards, the nanoparticle samples were washed with MilliQ water and second round of adsorption experiment was carried out, with the same metal ion concentration and volume. Adsorption and desorption experiments were repeated 3 times. under anhydrous conditions, 30 using CuBr(PPh 3 ) 3 as catalyst in dry THF at 60 C (Scheme 1).
Ligand L2-t Bu was obtained through a thiol-alkene click reaction between the suitable alkene and 3-mercaptopropyltriethoxysilane. The synthesis of L2-t Bu is summarized in Scheme 2. The commercial cyclen 2 was treated with three equivalents of tert-butyl 2-bromoacetate in the presence of sodium acetate to afford compound 3, 31 which was reacted with allyl bromide in acetonitrile at room temperature in the presence of potassium carbonate to give 4 in an overall 60% yield from 2. The alkene 4 and (3-mercaptopropyl)triethoxysilane were irradiated at 365 nm using 2,2-dimethoxy-1,2diphenylethanone (DMPA) as initiator, and the desired ligand L2-t Bu was obtained in quantitative yield.
Through these two synthesis routes, two new cyclenderivatives that are, to the best of our knowledge, not described in the literature, were successfully synthesized. Nevertheless, these syntheses, although successful, have reaction times lasting several days. Thus, a modication of the synthesis procedure was explored to synthesize L1-Boc by click chemistry under microwaves, which allows to obtain the molecule in less than 1 h versus 48 hours through standard protocol. Briey, precursor 1, the silylated azide and 6 mol% of catalyst were dissolved in dry THF and put in the microwaves at 100 C, 200 Wmax, for 3 cycles of 10 min. The obtaining of the right molecule was assessed through HNMR analysis. Since the microwave approach was successful to synthesize L1-Boc the protocol was adapted for the synthesis of the derivative named L3-t Bu, which could not be obtained through standard approaches. Thereby precursor 5 (see ESI † for the synthesis of 5), 32 the silylated azide and 12 mol% of catalyst were dissolved in dry THF and stirred under microwaves at 100 C (200 Wmax) for 4 cycles of 10 min (Scheme 3). This new cyclen derivative was then identied by 1 H NMR, 13 C NMR, 29 Si NMR and HRMS.  During the experiments, to set up the best protocols for microwave assisted synthesis of L1-Boc and L3-t Bu two key parameters were identied: the catalyst amount and the number of 10 min cycles. Indeed, it was observed that with less catalyst the reaction did not occur fully and a mixture of starting and expected molecules was obtained. In addition, increasing the reaction time (6 cycles of 10 min for example) tended to damage the molecules: the 1 H NMR signal at 2.63 ppm (4H) for L1-Boc and at 2.59 ppm (4H) for L3-t Bu representative of the cyclen were missing.
Thus, aer establishing the best protocol, microwaves allowed the efficient synthesis of L1-Boc and L3-t Bu and represented a promising approach toward simple, cheaper and faster cyclen derivatives synthesis.
Core-shell nanoparticles were synthesised using $20 nm iron oxide cores obtained through thermal decomposition. A silica mesoporous shell was then grown around the core. These nanoparticles exhibited an overall diameter of 102 AE 4 nm by TEM images (Fig. 1a and b), and HRTEM showed the mesoporosity ( Fig. 1c and d). The mesoporosity was further conrmed with N 2 adsorption-desorption (BET) (Fig. 1e and f). MMSN showed type I-IV isotherm characteristic of micromesoporous materials. The specic surface area was 302 m 2 g À1 and pore diameters (DFT method) showed two populations at 1.3 nm (microporosity) and a larger pic at 2.5 nm (mesoporosity). TEM was in agreement with the average hydrodynamic diameter (183 nm) of the nanoparticles determined by DLS (Fig. S4a †). These nanoparticles were also analyzed by zeta potential, TGA and IR (Fig. S4b-S6 †).
Then ligands (L1-Boc and L2-t Bu) were used to modify the surface of MMSN by post-graing (Fig. 2). We initially investigated different graing conditions with ligand L2-t Bu, and the reaction was monitored by FTIR (gC]O 1657 cm À1 ) (see Table   S1 and Fig. S7 in ESI †). In reuxing toluene no reaction occurred. The best conditions were established as follows: in DMF at 120 C with 2 mmol L2-t Bu per g MMSN, using water and Et 3 N as catalysts. The same procedure was used with L1-Boc. The BOC and t-Bu groups were then classically cleaved using TFA to lead to MMSN-L1 and MMSN-L2, respectively. To further assess and estimate the surface functionalisation the nanoparticles were characterised by DLS (Fig. S8 †), zeta potential ( Fig. S9 †), FTIR ( Fig. S10 and S11 †), TGA (Fig. S12 †), and elemental analysis. The data are summarised in Table 1. DLS showed a slight increase of the hydrodynamic diameters from MMSN to MMSN-L1-Boc and MSN-L2-t Bu in agreement with a successful graing of the ligand on the surface of MMSN (Fig. S8 †). This also indicates that the NPs keep their colloidal stability in water even aer their surface modication. The graing was also conrmed by zeta potential measurements performed around neutral pH 7. From raw NPs to graed NPs, the zeta-potential increased due to the presence of organic coating (Fig. S9 †). Aer TFA treatment, zeta potential became highly positive with MMSN-L1 in agreement with the presence of ammoniums at neutral pH whereas a strong decrease was noticed with MMSN-L2 as carboxylate groups were deprotected. The high zeta-potential positive value for MMSN-L1 and negative one for MMSN-L2 also indicate a good colloidal stability of these nano-objects at a neutral pH. Finally, the graing content was determined by TGA analysis (Fig. S11 †) and elemental analysis (Table 1). TGA analysis showed a graing of 4.34 wt% for MMSN-L1 which corresponds to a graing of 0.12 mmol g À1 and 7.19 wt% for MMSN-L2 which corresponds to a graing of 0.15 mmol g À1 . These results were conrmed by elemental analysis on N which indicated a graing of 0.15 and 0.18 mmol g À1 for MMSN-L1 and MMSN-L2, respectively.   The prepared magnetic core-shell silica nanoparticles were studied for the adsorption of Ni 2+ and Co 2+ from aqueous solutions at room temperature.
All the magnetic nanoparticles showed similar maximum adsorption capacity (varying between 1-2.13 mmol g À1 ), MMSN having the lowest capacity for both heavy metals (Table 2). For both heavy metals, adsorption isotherms were best t with Langmuir curves model (Fig. 3), having higher correlation coefficients compared to Freundlich (Fig. S13 †), indicating similar mechanisms with uniform binding sites in a monolayer adsorption.
Note that the measurements for the isotherm and the kinetics were carried out under different conditions. Final concentrations in the measurements of isotherms were considerably higher (25 mM) along with lower amount of the applied adsorbent (10 mg) to achieve saturation (Fig. 3). In case of kinetic measurements the solution was much less concentrated (10 mM), while the amount of adsorbent was, in contrast, higher in order to distinguish the rst steps in adsorption. The saturation of the adsorbent was not achieved in the kinetic measurements (Fig. 4).
The kinetic test results showed fast uptake at the rst 1-2 hours of metal interaction with magnetic nanoparticles, reaching over 80% of the total capacity. Slower uptake was found aer these 2 hours and the equilibrium was reached aer 5 hours (Fig. 4). This indicated quick and unhindered diffusion through the mesopores.
It is interesting to note that bare, not surface modied MMSN display high capacity in uptake of both Ni(II) and Co(II) cations, most probably due to their very high negative surface charge as manifested by strong zeta-potential (À32 mV). Particles modied by L2, acting presumably as a predominantly hard base in Pearson classication, display almost the same capacity as bare silica. This can be explained by some consumption and shielding of the silicate functions aer graing of the cyclen functions, which is compensated by the carboxylate ones in reactivity. On the contrary, the MMSN modied by L1, a so base in Pearson classication reveal strongly enhanced adsorption characteristics, indicating that the ligand is acting for heavy metal cations as an azo-crown functionality forming stable surface complexes with so Pearson acids such as Ni(II) and Co(II).
From the results of the selectivity tests with Ni : Co ¼ 1 : 1 mixture it is apparent that magnetic nanoparticles have lower adsorption capacity when introduced to mixed metal solution. This could be explained by possible competition for the adsorption sites between the two metal ions, making it more difficult for the ligand to bind the metals. Ni 2+ and Co 2+ cations can possess different coordination geometry when binding to  so ligands in Pearson classication. Ni 2+ cations are prone to Jahn-Teller distortion while Co 2+ usually display undistorted octahedral coordination. On the contrary, MMSN nanoparticles without any ligands revealed similar results with previous experiments (with single metals only), showing that there is no competition between Ni and Co without any ligands attached to nanoparticles.
EDS analysis gave similar results in regards with selectivity of magnetic nanoparticles (see ESI, Tables S2-S5 †). The numbers for L2 and MMSN for contents of both metals are rather similar and in general relatively low, being approximately twice higher for L1. This ratio is reecting the data on total capacity of the materials (see Fig. 3). This indicates that the enhanced surface of silica itself makes also a strong contribution to adsorption along with complexation to ligands.
Recyclability of the materials was studied, testing the uptake and release capacity in 4 consecutive adsorption-desorption cycles. The results, summarised in Fig. S14, † show unaltered functional characteristics for ligand-free reference material MMSN and functionalized MMSN-L1. In the case of MMSN-L2 it was, however, possible to see some loss of capacity.

Conclusions
In summary, click chemistry such as thiol-ene or Huisgen's reaction was very efficient for the syntheses of cyclen-based ligands possessing triethoxysilyl group. Associated with microwaves, Huisgen's reaction showed successful preparation of challenging L3-t Bu ligand. L1-Boc and L2-t Bu ligands were graed on magnetic mesoporous silica nanoparticles (MMSN) and protective groups were cleaved in order to extract and separate metal ions Ni(II) and Co(II) from model solutions. The kinetic of metal(II) uptake was very quick reaching 80% of the capacity aer 1-2 hours. The adsorption capacity of MMSN was very high, up to 2.13 mmol g À1 of Ni(II) with ligand L1 graed on MMSN.

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