3.2. Characterization
Spectroscopic characterization of the materials was essentially carried out by ATR-FTIR spectroscopy and the 13C(1H) CP-MAS Solid State NMR technique. These techniques allowed us to unambiguously identify the signals expected for the various functional groups formed in the reticulation reactions, and the typical fingerprints of the moieties from the constituting synthons (cyclodextrins, calixarenes, linkers). Then, the amount of silver present was determined by ICP techniques.
In general, the FT-IR spectra (as an exemplificative example, the spectrum of Cat3 is depicted in
Figure 3) show the -OH stretching band centered at ca. 3370 cm
−1, due to the presence of the βCD synthon, whose typical fingerprint band system can also be easily spotted out in the 1300–1000 cm
−1 region. The fingerprint area of the spectrum does not allow the unambiguous identification of the signals relating to the calixarene constituent, as they are superimposed on other signals. Similarly, possible signals from the graphene component can hardly be detected. Conversely, the most interesting section of the spectrum is the region 1750–1500 cm
−1, where the Amide-I-like and Amide-II-like signals of the urethane and urea groups can be found. The presence of at least two signals in this region undoubtedly accounts for the actual occurrence of the cross-linked nanosponge network. These signals can possibly be subjected to deconvolution analysis in order to detect the contributions of the different groups possibly present. The relevant attributions are summarized in
Table 1. Indeed, the recognition of the three different groups provides positive evidence of the incorporation of the A or K chain extender in the relevant materials.
The
13C(
1H) CP-MAS solid-state NMR spectra enable unambiguously identifying the actual presence of the possible calixarene component in materials NSGO3, Cat3, and Cat4, thanks to the presence of the relevant signals in the aliphatic and aromatic carbon regions. Complete signal attributions are summarized in
Table 2 (the spectrum of Cat3 is depicted in
Figure 4 for exemplificative purposes). In detail, along with the expected signals for the βCD subunits, spectra of the latter materials also clearly feature the typical intense band in the aliphatic C region (
t-butyl and methylene C atoms) and a system of four signals in the aromatic C region. In addition, the spectra of all materials feature the signals relevant to the linker and chain extender units in the aliphatic C spectral region, together with a signal cluster centered around 160 ppm in the urethane/urea carbonyl region (which, again, accounts for the actual formation of the polymer network). Notably, the graphene carbonaceous component does not provide signals of detectable intensity. This finding can be partly explained by the fact that it is indeed only a minor component of the composite. Moreover, it must also be taken into account that the CP-MAS technique relies on the enhancement of the C signals due to the cross-polarization with the
1H nuclei present in the sample. Thus, the
13C nuclei of the graphene component, which are not bound to any H atom, can hardly benefit from this effect.
More convincing evidence of the actual presence of the graphene component in the materials can be obtained by Raman spectroscopy. Raman spectra of the materials pGO, RGO, and RGOAg are depicted in
Figure 5, whereas those of the photocatalyst materials Cat1–Cat6 are in
Figure 6. In general, the spectra feature the well-known systems of bands in the ranges 1000–1800 cm
−1 (given by the superimposition of the D, D″, D*, D′, and G bands) and 2400–3400 cm
−1 region (with the 2D and S3 bands). It is well known that the amount of structural defects and the extent of the graphene-like domains, as well as the number of graphene layers, can be related by the intensity ratios between the D and G bands (
ID/
IG), the 2D and G bands (
I2D/
IG), and the 2D and S3 bands (
I2D/
IS3) [
48,
49,
50,
51,
52,
53]. These ratios can be obtained after the proper deconvolution analyses. The reduction of pGO to RGO causes only small variations in the shape of the spectrum, and the deconvolution analysis of the signals shows a slight variation of the above ratios. Considering RGOAg material, it is possible to note that the introduction of silver nanoparticles causes modifications in the spectrum. The band system in the 1000–1800 cm
−1 region shows a shift in the position of the maximum and the appearance of several shoulders. The deconvolution analysis indicates in particular a shift to higher wavelength numbers of the D band (at 1371 cm
−1 approx.) and the occurrence of three new bands (1256, 1361, 1437 cm
−1). At the same time, the 2400–3400 cm
−1 region also shows a shift of the 2D and S3 bands and a substantial loss of resolution. These observations indicate the existence of strong interactions in the composite material between the metallic and carbon components, which undergo significant structural modifications.
On passing to the spectra of Cat1–Cat6 materials, the concomitant presence of the overwhelming NS component makes the analysis much more complicated, because of the superimposition of the signals relevant to the organic matrix to those of the carbonaceous component. The spectra appear poorly resolved even in the 1000–1800 cm−1 region and the signals are in any case displaced compared to the characteristic position of the pristine pGO. There is also the complete loss of resolution of the 2400–3400 cm−1 area, where the graphene band system is no longer clearly distinguishable because intense signals are even visible around 2900 cm−1 attributable to the aliphatic C-H stretching. Therefore, deconvolution analysis of the spectral signals becomes pointless. Nevertheless, spectra undoubtedly provide convincing evidence of the actual presence of the graphene component embedded within the NS network, and of their interaction as well.
The photocatalysts were characterized by ICP analysis, in order to determine their actual Ag loadings. Results are summarized in
Table 3. Notably, RGOAg shows a loading close to the theoretical one of 50%, in agreement with what is reported in the literature. For materials Cat1–Cat6, silver loadings between 5% and 9% were found, which are lower than the theoretical 10% expected. It is noteworthy that the materials that best retain the Ag component are Cat5 and Cat6, i.e., those decorated with the amine chain extender (consistent with the ability of amine ligands to effectively coordinate the Ag
+ ion). The ex-ante loading procedure leads to significant metal loss, which can be justified by possible metal leaching during the work-up procedures. The ex-post loading procedure appears poorly effective in the case of Cat2, probably due to a lack of strong binding sites. Conversely, the ex-post loading of Cat4 works much better, probably because of the favorable interaction between the soft Lewis acid Ag
+ cation with the calixarene component cavities.
Finally, morphological characterization was performed using SEM techniques. Some representative micrographs (namely for RGOAg, Cat3, and Cat4) are shown in
Figure 7 for exemplificative scope (the complete micrographs are collected in
Supplementary Materials, Figure S1).
Pristine RGOAg shows a dense covering of tiny silver particles. The materials obtained with the ex-ante silver loading strategy (after having been crushed in a mortar and sieved at 150 μ) appear as compact masses with a quite clean surface, over which few irregular Ag particles are present (roughly several tenths or even few μ in size, as it can be spotted out by inspection of SEM images), but for those rare regions where the enclosed RGOAg component is occasionally exposed. This finding suggests that a small part of the silver particles formerly present in the RGOAg component could have been mechanically detached from the carbonaceous support during the polymer network synthesis. Conversely, the presence of tiny surface silver particles (most of which appear smaller than 0.3 μ, from inspection of SEM images) is much more apparent for the materials prepared with the ex-post loading strategy, supporting the hypothesis that, in the latter cases, Ag particles are not forced to be bound to the carbonaceous support embedded into the NS network. As a final remark, it is interesting to notice the materials containing the calixarene chain extender (Cat3 and Cat4) appear as relatively smooth masses, whereas the other NSs appear to have a much more wrinkled surface. This can be tentatively explained considering that the use of the calixarene moiety, bearing four nucleophilic –OH groups, enables obtaining a more reticulated network.
3.3. Preliminary Adsorption Tests
Before studying the photocatalytic properties of Cat1–Cat6 materials, a preliminary assessment of the intrinsic adsorption abilities of these NS–graphene composite matrices was needed, for understanding the effects from the presence of the chain extender component. For this purpose, four model composite materials, NSGO1–NSGO4, each loaded with 10% pGO (
w/
w), were prepared. In detail, NSGO1 features the diisocyanate H crosslinker only; in NSGO2, diol D as chain extender is present (as in Cat1 and Cat2); in NSGO3, the calixarene K is present (as in Cat3 and Cat4); finally, NSGO4 contains the triamine A (as in Cat5 and Cat6). The adsorption tests were carried out at pH 4.4 and 6.7 (in analogy with some previous works) in order to verify the possible effect of the pH medium, which could be particularly interesting in the case of the potentially pH-sensitive material NSGO4. The dyes 1–6 only were considered, because they are far more viable to quantify by UV–Vis spectrophotometry. The obtained results are summarized in
Table 4.
Collected data generally indicate a good affinity of the various materials towards the substrates, with few exceptions. The material that presented the best average adsorption ability was the non-functionalized NSGO1 (69%), followed by NSGO2 (65%) containing the aliphatic D chain extender. The presence of the calixarene component in NSGO3 has a rather negative effect, as the average percentage drops down to 41%. The amino chain extender of NSGO4 also has a negative effect, but in this case, it is important to take into account the pH value too. In fact, due to the presence of the tertiary amine groups, which are partly protonated even under neutral pH conditions, the nanosponge backbone is positively charged. This enhances the electrostatic effects affecting the adsorption properties of the material. Consistently, it is observed that the average adsorptions rise from 36% at pH 4.4 to 56% at pH 6.7. Furthermore, the significant decrease in affinity for NSGO3, particularly in the case of the dyes 1, 4, 5 and 6 (namely, Naphthol blue-black, Methyl orange, Bromochresol green and Rhodamine B), is not trivial, because the presence of K, by increasing the hydrophobic character of the material, should favor effective van der Waals interactions with the large conjugated structure of the organic guests. However, it must be taken into account that these dyes have anionic sulfonate or carboxylate groups, involving the presence of a localized charge. The charged groups interact effectively with the aqueous solvent, overall causing a decrease in affinity for the highly hydrophobic NS backbone. Indeed, it should also be kept in mind that an increase in the hydrophobic character makes the material less permeable to the aqueous solvent medium. Conversely, for those dyes such as 2 and 3 (Malachite green and Toluidine blue), carrying a positive charge largely delocalized over their conjugate structure, the increase in the hydrophobic character of the material has a limited outcome on their possible adsorption.
3.4. Photocatalytic Activity
The experimental procedures for performing the photodegradation tests are reported in the Materials and Methods section. A few comments should be reported herein. The tests rely on the simple idea of comparing the UV–Vis absorption of the pollutant solution before and after the irradiation in the presence of the photoactive material. The experimental conditions (irradiation power and time) were chosen in analogy with our previous work [
28]. The initial amounts of the different substrates were not the same (see
Section 2.7), but were suitably chosen in such a way to optimize their UV–Vis detection. It is important to stress that, in order to accomplish a correct evaluation of the catalytic activity, it is mandatory to rule out any contribution to the decrease in the solution absorbance deriving from the mere adsorption of the substrate onto the material. Hence, elution of the recovered photocatalyst with a suitable solvent (and evaluation of the substrate concentration in the eluate) is needed. Moreover, the amounts of photocatalysts in the various tests were chosen in such a way as to have roughly the same amount of silver in each test sample. In fact, all the results obtained, in terms of the mole amount of degraded organic substrate, must be normalized for the silver content in order to obtain a correct evaluation of the catalytic efficiency. Finally, the normalized results for the composite catalysts Cat1–Cat6 were compared with those for the plain photoactive species RGOAg, in order to ascertain the effect of the NS matrix on the catalytic efficiency. This effect is suitably quantified as the ratio between the normalized degradation yields for each composite catalyst and the corresponding datum for RGOAg. The obtained results are collected in
Table 5,
Table 6 and
Table 7 (data are also illustrated in in
Supplementary Materials, Figure S2).
All the photocatalysts generally show good activities, although strong differences were observed between different substrates. Considering the normalized degradation yields, dyes were degraded more efficiently than drugs, on average. In particular, dyes Naphthol blue black 1, Malachite green 2, and Toluidine blue 3 show the best degradation yields. By contrast, Rhodamine B 6 appears more resistant. On passing to drugs, fair yields can be found for Nalidixic acid 7, Tetracyclin 8, and Ciprofloxacin 12. Conversely, Diclofenac 9 and Ketoprophene 10 appear particularly resistant to photodegradation. As far as the latter two drugs are concerned, it is noteworthy that polyamine-decorated materials Cat5 and Cat6 appear to be the only photocatalysts able to promote degradation up to a fair extent, whereas even plain RGOAg has no significant effect. Because of this, it was not possible to quantitatively assess any activity enhancement due to the NS matrix for these two substrates.
Considering the effect of the NS matrix, a close inspection of data in
Table 7 reveals that, significant activity enhancement (>1.2) can be found in 33 cases out of 59 (24 out of 36 if we consider dyes only; up to 9 times, in the case of 5 with Cat1), whereas significant activity decrease (<0.8) occurs in 11 cases out of 60 (8 out of 23, on considering drugs only). As far as the average behavior of the different materials is concerned, the best performances were shown by amine-decorated materials Cat5 and Cat6 (2.3 both). Satisfactory performances were shown also by Cat1 (2.2) and Cat2 (2.1), bearing the aliphatic chain extender D. Conversely, only a fair average enhancement (1.5) was found for the calixarene-decorated material Cat3, whereas no significant average matrix effect occurred for Cat4. In perfect analogy with the preliminary adsorption tests, these data indicate that the large increase in the hydrophobic character caused by the presence of the calixarene component has a negative outcome on the overall properties of the composite material. Notably, only in this case, the different protocol used for silver loading has a significant impact on the photocatalytic performances, in agreement with the well-known reactivity–selectivity principle. Otherwise, the use of either the ex-ante or the ex-post silver loading method provides no particular advantage.
On passing to analyze the collected data as a function of the organic substrate, the largest average enhancement can be found for dyes 5 (5.6), 1 (3.1), and up to a lesser extent, 4 (2.4). Conversely, among drugs, significant activity enhancement can be found only for 11 (2.0; a modest average enhancement of 1.3 can be found also for 12). By contrast, no average enhancement can be found for 6, 7, and 8. These results appear quite tricky to rationalize because no correlation is apparent with the structure of the substrates, nor with their average affinity for the NS matrix as obtained from the preliminary adsorption tests. Indeed, it should be kept in mind that guest affinity for nanosponges is the outcome of a fine balance between different and mutually contrasting effects (specific and unspecific host–guest interactions, solvation effects, etc.). In the present case, the situation is complicated by the possible concomitant occurrence of further specific interaction with the graphene and the silver nanoparticle components. Therefore, any attempt to actually predict the overall outcome of all these factors would be a desperate task.
Data relevant to degradation percentages should be compared with those reported in the literature. A massive amount of experimental work has been carried out on the photodegradation of both dyes and drugs. Recent works on the subject are virtually countless, indeed, and are periodically reviewed [
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
54,
55,
56]. Plenty of papers make a claim for very satisfactory (>90%) [
10,
57,
58,
59] or even almost complete (>97%) [
60,
61,
62,
63,
64,
65] apparent removal of all the substrates considered herein. Again, it is important to stress that, unlike most reports, our data take into account the correction for the amount of undegraded substrate possibly adsorbed onto the photoactive material. We always observed that the catalyst recovered after each photodegradation test, releasing significant amounts of unreacted substrate (particularly in the case of dyes) after washing with methanol. Just to cite a typical case, dye 6 shows a very satisfactory 90% apparent removal with title RGOAg, as it can be simply evaluated from the absorbance of the aqueous solution after the irradiation experiment. However, this result reduces to a modest 18% once the unreacted dye has been washed away from the catalyst. Therefore, our results are hardly comparable with those present in previous works.
The possible recyclability of the materials was verified. In particular, their possible reuse in the degradation of dye 3, as a model substrate, was tested for five catalytic cycles. Unfortunately, the obtained results (collected in
Table 8) were not satisfactory. In fact, a significant drop in activity could be observed after the first cycle. Then, activity was roughly constant after the third cycle for Cat1–Cat3 and Cat5, whereas it decreased almost to zero for Cat2 and Cat4. These findings can be attributable to the loss of metal loading during the process. Once again, it was possible to note that the
ex-post procedure appears disadvantageous.
Finally, the photodegradation kinetics were studied in order to obtain further insights into the kinetic and mechanistic course of the process. For this purpose, RGOAg, Cat5, and Cat6 were selected (the first as the reference, the other two as the best-performing systems on average). As model substrates, dyes 3 and 4 and drug 8 were considered. The data collected, summarized in
Table 9, show some unexpected features. In fact, in some cases, the degradation followed a first-order kinetic, whereas in other cases, the data trend can be mathematically modeled using a “stretched exponential” function of the following type:
where P represents the percent of the degraded substrate,
k is the apparent kinetic constant, and the exponential “stretching” parameter
n accounts for the deviation of the kinetic course from the standard first-order trend (
n values range between 0 and 1; for
n = 1 an ordinary first-order kinetic occurs). Kinetic trends that can be described using this type of mathematical law (reported also as the Kohlrausch equation) have been occasionally reported in the literature [
66,
67,
68]. The physicochemical interpretation of this type of dependence is still currently under debate. However, the observation of this trend has been generally justified claiming the occurrence of some “inhomogeneity” in the behavior of the catalytic system. In other words, the observed behavior cannot be assimilated to that of a “simple” system in solution, but it should rather be seen as deriving from the superimposition of a “continuum” of microsystems, described by a suitable distribution function. This hypothesis rules out the idea that the photodegradation process simply occurs through the interaction between the substrate and free ROS species in solution. Rather, it suggests that a direct interaction between the substrate and the surface of the photocatalyst can play an important role. It is important to notice that this behavior was observed, in particular, with dyes 3 and 4, having a highly conjugated molecule able to interact with the graphene material, but not with the more hydrophilic tetracycline 8. This behavior, combined with the fact that dyes absorb visible radiation (produced by the light source) more effectively than drugs, could even suggest that the degradation process involves the substrate in its electronically excited state, rather than in the fundamental state. Finally, it is worth noting that, in the case of dyes, apparently there was no direct correspondence between the values of the apparent kinetic constants
k and those of the degradation percentages at 2 h reported previously in
Table 5. In the case of 3, this apparent anomaly is fully justified by the different values of the parameter
n. Moreover, in the case of 4, the reaction seems to proceed to completion only with Cat6, whereas the regression parameters indicate that the degradation seems to stop at 26% with RGOAg and 46% with Cat5. Also, for Tet with RGOAg, the regression indicates a maximum degradation percentage of 52%. The last observations do not appear easy to rationalize, and they will be subjected to detailed future mechanistic studies, which were indeed by far beyond the scope of the present work.