Influence of bridged monomer on porosity and sorption properties of mesoporous silicas functionalized with diethylenetriamine groups

The names of the samples along with physico–chemical characterization are presented in Table 1 and the scheme of the synthesis—in Fig. 1. The molar ratio 14:2:2 (sample R2–R4) or 18:2 (sample R1) was chosen to keep constant number of silicon atoms in each reaction mixture (note that one mmol of bis-silylated bridged monomer, i.e., BTESE, BTESB, BTMSD contains 2 mmol of Si atoms). As follows from our previous investigations the addition of 10% of TMPET results in a significant decrease of porosity from the one hand but very well DICL sorption uptakes from the other (Barczak et al. 2018) thus it was decided to keep that ratio here. It was suggested in the literature that the addition of bis-silylated bridged monomer can oppose destructive action of amino silane (Barczak et al. 2009b). Indeed looking at the adsorption isotherms presented in Fig. 2a it can be seen that samples synthesized with the addition of BTESE (sample R2), BTESB (sample R3) or BTMSD (sample R4) have more developed porosity than the sample R1. In the case of the sample R3 the shape of isotherm with H1 type of the hysteresis loop is typical for highly ordered SBA-15 mesoporous structure. The remaining samples exhibit H2/H3 type of hysteresis loop indicating rather disordered mesoporous structure. Pore size distributions presented in Fig. 2b have two maxima: one located in the range 3.3–4.0 nm and the other located in the range 8.5–11.7 nm. The second maximum is due to the primary mesopores created due to the removal of the template (Pluronic P123). In Table 1 the values of porous structure parameters are shown. Addition of the bis-silylated bridged monomer increase the average size of primary mesopore which is particularly visible for the sample R3 (11.7 nm). Sample R2 has the highest value of specific surface area (S_BET) and micropore volume (V_micro) among all the samples studied. Interestingly total pore volumes (V_total) of the samples R2–R4 are two times higher than that of the sample R1. All the observed changes testify that addition of bis-silylated bridged monomer remarkably influences the porous structure of the final samples.

As can be seen from the low-angle XRD patterns (Fig. 3) all the materials exhibit an ordering, albeit with different degree and type. In the case of the sample R3 there are three well resolved peaks in the range of 2θ = 0.8–1.8°, one sharp reflection centered at 2θ = 0.8° indexed as (100), and two minor but distinct reflections at 2θ = 1.5° and 2θ = 1.7°, indexed as (110) and (200), respectively. This pattern, related to the hexagonal p6mm symmetry, is characteristic of well-developed SBA-15 mesostructure (Barczak et al. 2010a, 2009b). In the case of the sample R1 there are two reflections related to the cubic Ia3d symmetry: a well resolved one at 0.9° and a smaller hump at 1.1° indexed as (211) and (220) respectively (Barczak et al. 2009b; Liu et al. 2002). Decreased peak intensity of the (220) reflection can be related to the presence of organic moieties (coming from the co-condensed TMPET monomer) and inside the pores which could false diffract (Visuvamithiran et al. 2013). In the case of the samples R2 and R4 it is difficult to clearly assign a specific space group because there is only one reflection centered at 2θ = 1.8°. This also means that the ordering of the samples R2 and R4 is deteriorated, most probably due to presence of specific co-monomers in the initial mixture adversely affecting the formation of the mesophase during the synthesis.

Interestingly the addition of BTESB bridged monomer preserve the p6mm symmetry of the final R3 sample, in contrast to remaining bridged monomers. Another interesting observation is that the specific combination of monomers can induce a change of the symmetry from hexagonal to cubic, despite the fact that the synthesis conditions were set to obtain samples with hexagonal symmetry. We reported previously that the presence of vinyltriethoxysilane co-monomer can result in cubic symmetry (Barczak et al. 2009b); from this research it follows that also TMPET behaves similarly. Previously it was observed for 3-mercaptopropyltriethoxysilane (Garcia-Bennet et al. 2004). Most probably this is a more common feature of the hybrid silica structure synthesized by co-condensation of pure silica monomer (like TEOS) and organic monomer with pendant group (like TMPET) and is governed by the amount of co-monomer and most probably other synthesis parameter (e.g., processing temperature). The detailed explanation of this phenomenon exceeds the scope of this paper.

The contents of nitrogen, carbon and hydrogen are given in Table 1. Different nitrogen contents indicate successful co-condensation, albeit with different functionalization efficiencies (FE, calculated as a ratio of the measured and theoretical content of nitrogen). While the addition of BTESE does not change significantly the FE the same is not true for the BTESB and BTMSD—samples R3 and R4 have significantly lower nitrogen contents and, consequently, lower FE values. The common feature of those two above-mentioned monomers is that they possess the benzene rings thus some additional interactions between the rings and amine groups can be responsible for the resulting lower functionalization efficiency. Thus TMPET molecules can be specifically oriented (in a statistical sense) during the synthesis which may result in their limited hydrolysis and co-condensation.

To further investigate the chemistry of the studied samples FTIR spectra were collected (Fig. 4). The most intensive band with the maximum at ~ 1075 cm⁻¹ which has a shoulder at 1150–1250 cm⁻¹ can be attributed to stretching modes of the siloxane framework. The bands located at ~ 950 cm⁻¹, ~ 800 cm⁻¹ and ~ 450 cm⁻¹ can be assigned to Si–O vibrations in the Si–OH moieties, symmetric stretching vibrations of Si–O–Si, and bending vibrations of Si–O–Si, respectively (Sanaeishoar et al. 2015).

The presence of the –СН₂– link in the alkyl chain of TMPET and ethylene bridge of BTESE is confirmed by a group of absorption bands in the region of ~ 2800–3000 cm⁻¹, attributed to the stretching modes of CH₂ groups, and possibly CH₃ groups (fragments of residual ethoxy groups as well us unremoved traces of template). The higher intensity of those signals for the samples R3 and R4 may testify to considerable amount of unremoved Pluronic P123. Two absorption bands of stretching vibrations of amine groups located in the region of 3300–3500 cm⁻¹ could not be detected due to the presence of a wide band of the physically adsorbed water extending from ~ 3700 cm⁻¹ to ~ 3000 cm⁻¹. In the case of the samples R3 and R4 the presence of benzene rings should be manifested by not very intense bands in the range of 3000–3100, ca. 1600 and ca. 1500 cm⁻¹ although due to low concentrations of the other co-monomers co-condensed with TEOS there are no clear signals on the FTIR spectra.

Thermal stability of the synthesized samples was evaluated by means of thermal gravimetry (Fig. 5). Sample R1 reveals a total mass loss of approximately 20%, due to the three distinctive but often overlapping processes occurring during thermal treatment: the evaporation of physically adsorbed water (< 150 °C, ~ 4% mass loss) decomposition of organic fragments (150–450 °C, ~ 14% mass loss) and dehydroxylation of surface silanols (> 450 °C, ~ 2% mass loss). Two distinct DTG maxima observed at 75 °C and 290 °C correspond to the first two processes. Interestingly, the location of those maxima vary from sample to sample. The first maximum is centered at 70–75 °C for the samples R1, R3 and R4 but it is shifted to 85 °C for the sample R2. This means that water is bound stronger in the pores of R2 sample than in the case of the rest of the samples. This may be due to the highest content of micropores (cf. Table 2) where a significant part of water can be stored. The second maximum on DTG curves is centered at ~ 285 °C for the samples R1 and R2 and at ~ 360 °C for the samples R3 and R4. This maximum arises from the decomposition of pendant propyldiethylenetriamine groups as well as the organic bridges. In the case of R3 and R4 samples, this maximum is shifted towards higher temperatures which means that the benzene bridges make the silica samples more resistant to thermal decomposition. In the case of the samples R3 and particularly R4 the third maximum arises at 600 °C due to the remarkable dehydroxylation of the surface silanols. The results of TG and DTG analyses clearly show that co-condensation of the monomers was successfully achieved.

The SEM and TEM data presented in Figs. 6 and 7, respectively show that the morphology is typical for SBA-15, i.e., it is composed of the “sausage-like” hexagonal motifs often curved and connected to others along long edges. The TEM microphotographs reveal that those motifs are formed by the ordered mesoporous network of long uniform hexagonal mesopores with the diameter of several nm, which is consistent with the pore sizes obtained from the nitrogen sorption data (cf. Table 1).

To reveal possible application of the fabricated xerogels for removal of pharmaceuticals as well as to correlate their physico-chemical properties with sorption efficiencies, adsorption of diclofenac sodium salt (DICL) was studied. Diclofenac poses a significant risk to the environment and to human health thus it is considered as one of the priority pollutants in the European Water Framework Directive. DICL is a commonly prescribed in large quantities, and—as a result—it has being detected in wastewaters along with its metabolites and products of its photo-transformational changes. Chronic exposure to DICL has adverse effects for aquatic animals and humans (Barczak et al. 2018).

From our previous works it follows that the adsorption of DICL is very fast and usually 2–4 h are needed to reach equilibrium and the pH 6 is optimal (Barczak 2018; Barczak et al. 2018). Thus the sorption experiments were run for 16 h at pH 6 (unbuffered solution). The observed static sorption capacities (SSC) are given in Table 2. As it can be seen, samples R1 and R2 exhibit highest SSC values of 841.8 and 808.7 mg g⁻¹, respectively, while the samples R3 and R4 have lower SSC values of 607.6 and 394.5 mg g⁻¹, respectively. The higher SSC values for R1 and R2 are obviously related to higher number of protonated amine groups ready to electrostatically interact with DICL anions. In the case of samples R3 and R4, the lower number of amine groups results in lower SSC values. However, despite the similar nitrogen content (R3: 1.87% vs. R4: 1.96%) there is a huge difference in SSC values. Both samples have similar porous structure so the reason of this behaviour must by related more to the surface chemistry than porosity. Looking at the values of potential ζ collected in Table 1, it can be seen that all the samples have positive values of ζ which means that the surface of all the sorbents can electrostatically attract DICL anions (pKa value of DICF which is 4.15 (Barczak et al. 2018) thus at pH 6 the soluble ionized form of DICF is almost exclusively present in the solution). Among all the samples, R4 has the lowest value of ζ (24.6) which means that the electrostatic interactions are the most suppressed for this adsorption system. The highest mass loss > 450 °C observed on the TG curve of this sample (Fig. 5a) due to the dehydroxylation testifies to the abundance of silanol group present on the surface (reflected in lower ζ value). Thus the resulting charge density is not only controlled by the amine groups but also by the silanol groups, whose are mostly deprotonated at pH 6. Thus, even for surface-functionalized mesoporous silica, the effective charge density is not only controlled by the incorporated functionalities but also by the number and form (protonated vs. deprotonated) of the silanol groups (Rosenholm et al. 2007).

The adsorption isotherms are presented in Fig. 8 along with their fitting using three common adsorption models: Langmuir, Freundlich and Sips (also called Langmuir–Freundlich). The proper formulas can be found in one of our previous works (Barczak 2018; Barczak et al. 2018). Fitting parameters are given in Table 2. The correlation coefficients show that Sips and Langmuir models fit the results better than the Freundlich model. However, correlation coefficients are noticeably better for Sips than for Langmuir model. Interestingly the K_L and K_LF constants, (also called affinity constants) are significantly higher for R4 sample when compared with the rest of the samples. It should be mentioned that the fitting results should be interpreted cautiously since Sips equation has three parameter so it can fit per se the isotherms better that two-parameter models of Langmuir and Freundlich.

The presence of adsorbed DICL is unquestionably confirmed by FTIR spectroscopy—spectra of loaded sorbents are presented in Fig. 3 (for convenience, the spectrum of DICL sodium salt is also given at each spectrum). Comparing the spectra of the samples before and after adsorption it can be seen that there are additional signals characteristic of DICL, namely: 3322 cm⁻¹ (amine stretching), 1560 cm⁻¹ (aromatic symmetric CC stretching), 1506 and 1453 cm⁻¹ (asymmetric ring stretching and symmetric deformation of methylene groups, respectively), 1576 and 1381 cm⁻¹ (stretching modes of carboxylate group), 1302 and 1279 cm⁻¹ (C–N stretching).

In this work regeneration was achieved by immersing the loaded sorbents in 0.9% NaCl at 40 °C. Full desorption was achieved after 48 h which was confirmed by FTIR spectra of the samples after desorption exhibiting lack of signals characteristic of DICL. Desorption curves aligned as releasing profiles are shown in Fig. 9. Although all the profiles look similar there is a very important difference between them: amount of desorbed DICL after only one minute of contact with NaCl solution vary significantly from sample to sample: for R1 = 30% of DICL is immediately released, for R2: ~ 20%, while for R3 and R4 only 7% and 2%, respectively. Obviously, the immersion of sorbents in 0.9% NaCl solution (ionic strength: 154 mmol) suppresses the electrostatic interactions between the silica surface and deprotonated DICL anions which results in immediate release of some portion of the drug. This means that the significant amount of DICL is physically bounded to the surface via electrostatic interactions. It was already suggested and confirmed by both experimental findings and theoretical calculations that the first layer of adsorbed DICL is strongly bonded by hydrogen bonds formation between protonated amine groups and carboxylate anions but the next layers of the adsorbed DICL are due to the electrostatic Coulombian interactions (Barczak et al. 2018). Those findings are in agreement with the current state-of-the-art where above-mentioned mechanisms along with hydrophobic interactions are usually considered as three possible mechanisms responsible for the diclofenac adsorption.

In our previous paper we have shown that the main driving adsorption force between protonated amine groups and diclofenac anion is the hydrogen bond formation (this applies to DICL anions layers having direct access to the protonated amine groups, the remaining anions interact via electrostatic attraction). Thus it was interesting to investigate whether the organic bridges affect the stabilization of the final complex. To verify this the DFT calculations for the R2 and R3 sorbents were performed in a fashion described in our previous papers (Barczak et al. 2018), i.e., we estimated energetic effect associated with the adsorption of DICL^⊖ on the protonated amine groups. For simplicity, we used here the aminopropyl group endcapped with hydrogen atom, i.e., CH₃–(CH₂)₂–NH₂ as in ref. (Barczak et al. 2018) instead of using CH₃–(CH₂)₂–NH–(CH₂)₂–NH–(CH₂)₂–NH₂ group. This time the protonated …N^⊕H₃⋯H₂O group was allowed to interact with (MeO)₃Si–(CH₂)₂–Si(OMe)₃ and (MeO)₃Si–Ar–Si(OMe)₃ systems (note that further simplification consisted in replacement of the ethoxy group with methoxy group was introduced in order to further reduce the system size). It was also assumed that the …N^⊕H₃⋯H₂O group is a member of a neighboring silica chain (or other structure present in the bulk phase). The structures are shown in Fig. 10. Note that additional hydrogen bond(s) are formed between …N^⊕H₃⋯H₂O and methoxy groups. The calculations in which the DICL^⊖ replaces water molecule clearly show that there is hardly a difference between the stabilization energies, being slightly lower than − 80 kcal mol⁻¹, when (MeO)₃Si–(CH₂)₂–Si(OMe)₃ and (MeO)₃Si–Ar–Si(OMe)₃ systems are considered. In addition, the presence of bridges lowers down the stabilization by ca. 20 kcal mol⁻¹, as in the case considered in our previous work the energetic effect is − 103.2 kcal mol⁻¹ (cf. Fig. 8 of the reference (Barczak et al. 2018)). In passing we note, that there is yet another possibility of binding DICL^⊖, i.e., via the second oxygen atom of the COO^⊖ group. This does not change the conclusions as far as energetics is concerned.

Another important conclusion coming from the desorption studies is that the addition of bridged monomer markedly affects the releasing rate of DICL, particularly in the time interval 0–300 min. In the case of the samples R3 and R4 the releasing profile is close to linear in that range without the so-called burst effect (i.e., initial abrupt releasing of the matrix) which is a very desirable feature of any controlled releasing system (CDS). In the literature there is a lot of reports describing applications of silica as CDSs due to their tailored porosity and surface chemistry (Krajnović et al. 2018; Rosenholm and Lindén 2008; Song et al. 2005; Vallet-Regí et al. 2007). Thus the synthesis approach presented here based on co-condensation of three monomers can be applied for more precise tuning of the releasing profiles of the pharmaceuticals from the silica-based CDSs.

Since reusability of any sorbent is one of the crucial parameters bringing it closer to the practical applications, the regenerated samples were used in two next adsorption/desorption cycles to check whether the SSC values will be significantly reduced. The obtained SSC values after three cycles were slightly lower when compared with initial SSC values for fresh samples (83%, 87%, 90% and 88%, for the samples R1, R2, R3, R4, respectively). Thus all the samples retain over 80% of their initial adsorption capacity, thus they are suitable for multiple usage. The decrease of SSC values is most probably related to the gradual degradation of porosity and surface chemistry of the samples. This decrease is evident in the case of the sample R1, which testifies that the addition of the bridged co-monomer can positively affect adsorption efficiency of the resulting materials, most probably due to the lower degradation rate of the organically modified silicas (Croissant et al. 2017).

It is interesting to compare the sorption data obtained here with other reports describing adsorption of diclofenac. First it should be noted that, as already shown by us, the uptake amount of other drugs (such as ibuprofen, naproxen or penicillin G) are close to that of DICL due to the similar adsorption mechanism (Barczak 2019). Thus results obtained for DICL can be extended to a wider range of pharmaceuticals. The materials obtained here have higher capacitances than majority of the sorbents reported in the literature (cf. Table 3). It is also worthwhile to mention that such high uptakes observed here are due to two complementary mechanisms: H-bonding and electrostatic interactions. The former is responsible for stronger chemical adsorption and the latter for weaker multilayered physisorption.