Carbon Nitride-Perovskite Composites: Evaluation and Optimization of Photocatalytic Hydrogen Evolution in Saccharides Aqueous Solution

: The application of hybrid photocatalysts made of carbon nitride and lead-free perovskites, namely DMASnBr 3 / g-C 3 N 4 and PEA 2 SnBr 4 / g-C 3 N 4 , for the H 2 evolution from saccharides aqueous solution is described. The novel composites were tested and compared in terms of hydrogen evolution rate (HER) under simulated solar light, using Pt as a reference co-catalyst, and glucose as a representative sacriﬁcial biomass. The conditions were optimized to maximize H 2 generation by a design of experiments involving catalyst amount, glucose concentration and Pt loading. For both materials, such parameters a ﬀ ected signiﬁcantly H 2 photogeneration, with the best performance observed using 0.5 g L − 1 catalyst, 0.2 M glucose and 0.5 wt% Pt. Under optimized conditions, DMASnBr 3 / g-C 3 N 4 showed a 5-fold higher HER compared to PEA 2 SnBr 4 / g-C 3 N 4 , i.e., 925 µ moles g − 1 h − 1 and 190 µ moles g − 1 h − 1 , respectively (RSD ≤ 11%, n = 4). The former composite, which a ﬀ ords an HER 15-fold higher in aqueous glucose than in neat water, provided H 2 also with no metal co-catalyst (around 140 µ moles g − 1 h − 1 ), and it was reusable for at least three photoreactions. Encouraging results were also collected by explorative tests on raw starch solution (around 150 µ moles g − 1 h − 1 ).


Introduction
The search for new photocatalytic systems working under solar light for hydrogen production is increasingly triggering the interest of the scientific community. In the framework of novel photocatalysts, graphitic carbon nitride (g-C 3 N 4 ) has emerged in the last decade as one of the most promising material to run H 2 photoproduction from water under visible light, due to cost-effective and easy synthesis, chemical stability, narrow band-gap and band potentials suitable to perform relevant redox reactions in aqueous solution [1,2]. Its intrinsic limitations, namely relatively high charge carriers recombination, low surface area and restricted light harvesting in the visible region, have been in part compensated by several strategies. For instance, metal and non-metal doping, structural and morphological modifications, dye-sensitization, and combination with co-catalysts of different nature (e.g., carbon nanotubes, carbon dots, bimetallic deposition) have proved to be rewarding in terms of enhanced photocatalytic activity towards H 2 generation from water also in the presence of sacrificial agents [2,3]. Among these, mainly fine chemicals have been used, as recently reviewed by Nasir et al. [2], while just a few studies were undertaken in aqueous biomass solutions or directly in wastewaters [4][5][6].

Results and Discussion
Based on the preliminary findings obtained with carbon nitride-perovskite composites in TEOA solution [12,13], the two best performing materials, 33% DMASnBr 3 /g-C 3 N 4 and 5% PEA 2 SnBr 4 /g-C 3 N 4 , are investigated here for a systematic study of H 2 photoproduction from glucose aqueous solution. Glucose was selected as a biomass-derived sacrificial agent because in the wastewaters from food industry sugars are present at considerable amounts [4,5], and Pt was used as the reference metal co-catalyst because of its excellent properties for water reduction due to the large work function, resulting in a strong Schottky barrier effect [15,16].

Comparison between DMASnBr 3 /g-C 3 N 4 and PEA 2 SnBr 4 /g-C 3 N 4 after Chemometric Optimization
With the aim of investigating the behavior of the two catalysts and maximizing H 2 evolution, a chemometric approach was chosen to easily individuate the effects of the key operational parameters of heterogeneous photocatalysis, namely the relative amounts of catalyst, glucose and metal [17][18][19][20]. Accordingly, a full 2 3 experimental design was setup considering glucose concentration (x 1 ), catalyst amount (x 2 ) and Pt loading (x 3 ). As reported in Table 1, the experimental domain was comprised of two levels (−1 and +1) of each variable.

Variable
Level Codes

−1 +1
Glucose concentration (M), x 1 0.025 0.2 Catalyst amount (g L −1 ), x 2 0.5 2 Pt loading (wt%), x 3 0. 5 3 HERs obtained under the different conditions (mean values of duplicate tests), according to the experimental matrix, are reported in Table 2. The values were used as the experimental response (y), and they were modeled (Matlab ® software) relative to each variable (x i ), in line with the following equation: Figure 1 shows the plot of the coefficients (b i ) of the model, while the response surfaces are gathered in Figure 2.  The values were used as the experimental response (y), and they were modeled (Matlab ® software) relative to each variable (xi), in line with the following equation: y = b0 + b1x1 + b2x2 + b3x3 + b12x1x2 + b13x1x3 + b23x2x3, (1) Figure 1 shows the plot of the coefficients (bi) of the model, while the response surfaces are gathered in Figure 2. The models elaborated on the experimental results are: for DMASnBr3/g-C3N4 and PEA2SnBr4/g-C3N4, respectively.  Table 1); the number close to each curve represents the HER.
As apparent from the significance of the coefficients (Figure 1), the picture is similar for both composites, as the three variables affect significantly the HER (p < 0.001), although glucose concentration to a minor extent for PEA2SnBr4/g-C3N4 (p < 0.01); in particular, the photoreaction is favored by high glucose concentration and low levels of catalyst in the suspension and metal photodeposited on the catalyst surface. In addition, some interactions are statistically relevant, in  Table 1); the number close to each curve represents the HER.
The models elaborated on the experimental results are: for DMASnBr 3 /g-C 3 N 4 and PEA 2 SnBr 4 /g-C 3 N 4 , respectively.

of 11
As apparent from the significance of the coefficients (Figure 1), the picture is similar for both composites, as the three variables affect significantly the HER (p < 0.001), although glucose concentration to a minor extent for PEA 2 SnBr 4 /g-C 3 N 4 (p < 0.01); in particular, the photoreaction is favored by high glucose concentration and low levels of catalyst in the suspension and metal photodeposited on the catalyst surface. In addition, some interactions are statistically relevant, in particular x 1 -x 2 and x 2 -x 3 for DMASnBr 3 /g-C 3 N 4 . In line with the response surfaces (see Figure 2), HER increases by keeping the catalyst amount at the low level but with the highest concentration of sacrificial agent, whereas both catalyst and co-catalyst at the lowest level enhance H 2 evolution. For PEA 2 SnBr 4 /g-C 3 N 4 , x 1 -x 2 and x 1 -x 3 are significative (p < 0.01 and p < 0.05, respectively); accordingly, the response surfaces show that also in this case, HER is maximized with low Pt loading and catalyst amount, and the high level of sacrificial biomass is useful, especially when working with 0.5 wt% Pt.
These findings can be explained considering that at low sacrificial substrate concentration, the mass transfer of glucose from water to the catalyst surface is the bottleneck of photocatalytic reaction, whilst in concentrated solutions, interfacial reactions govern the process, due to the saturation of glucose on the catalyst [17]. Furthermore, excessive Pt loading reduces the catalytic surface available for light absorption; additionally, a high concentration of suspended powder causes scattering of the incident radiation, thus decreasing the overall process [15,18].
Based on these outcomes, the selected conditions were 0.2 M glucose, 0.5 g L −1 catalyst, and 0.5 wt% Pt. These are advantageous, involving the use of small amounts of both catalyst and metal co-catalyst, and suggesting that large sample dilution could be avoided when using food industry wastewaters, where the total sugar content is up to tens of grams per liter [5].
For the best performing composite, DMASnBr 3 /g-C 3 N 4 , the optimal conditions were further confirmed by validation of the model, working at the test point (x 1 = 0; x 2 = 0; x 3 = 0). Being the mean HER (407 ± 69 µmoles g −1 h −1 , p = 0.05, n = 4) not significantly different from the value predicted by Equation (1), viz. 390 µmoles g −1 h −1 , it is demonstrated that the model is effective and it can be applied to the whole experimental domain.
The chemometric study provided key information. First, it enabled pointing out the most convenient conditions for H 2 evolution, which resulted in a three-fold increase in HER relative to the first pilot work [13]; at the same time, the results collected clearly highlight the superior photocatalytic activity of DMASnBr 3 /g-C 3 N 4 .
Given the very small difference between the MHPs' band gaps-2.85 and 2.74 eV for DMASnBr 3 and PEA 2 SnBr 4 , respectively [12,13]-and the higher surface area of PEA 2 SnBr 4 /g-C 3 N 4 (5.5 m 2 g −1 vs. 3.4 m 2 g −1 of DMASnBr 3 /g-C 3 N 4 ), it could be speculated that the better performance of the first catalyst derives from the different band alignments of the MHPs' valence band (VB) and conduction band (CB) edges with those of carbon nitride (gathered in Figure 3). This results in a different electron transfer and charge carriers stabilization mechanism.
The synergistic effect between the two constituents of the DMASnBr 3 -based photocatalyst, which works as a Z-scheme heterojuction, has been explained in terms of favorable band alignment and prolonged charge carrier lifetimes [13]. In the case of PEA 2 SnBr 4 , this is also true, but it is known that 2D perovskites have higher excitonic binding energies that may slightly reduce the charge carrier dynamics, thus affording a lower photocatalytic efficiency [21].
Another reason rationally stands in the better dispersibility of the DMA composite in aqueous phase due to its lower hydrophobicity. Because of the low surface area of the composites, good mechanical stirring is important; on the other hand, the fast sedimentation of the powders can be an advantage to recover them from water after photocatalysis.
Given the very small difference between the MHPs' band gaps-2.85 and 2.74 eV for DMASnBr3 and PEA2SnBr4, respectively [12,13]-and the higher surface area of PEA2SnBr4/g-C3N4 (5.5 m 2 g −1 vs. 3.4 m 2 g −1 of DMASnBr3/g-C3N4), it could be speculated that the better performance of the first catalyst derives from the different band alignments of the MHPs' valence band (VB) and conduction band (CB) edges with those of carbon nitride (gathered in Figure 3). This results in a different electron transfer and charge carriers stabilization mechanism. The synergistic effect between the two constituents of the DMASnBr3-based photocatalyst, which works as a Z-scheme heterojuction, has been explained in terms of favorable band alignment and prolonged charge carrier lifetimes [13]. In the case of PEA2SnBr4, this is also true, but it is known

Further Investigation of DMASnBr 3 /g-C 3 N 4
The photocatalytic behavior of the composite containing the DMASnBr 3 perovskite was further studied. Table 3 summarizes the HERs collected by key tests aimed at assessing the role of the metal co-catalyst, of the glucose biomass and to evaluate the contribute of "direct water splitting", i.e., H 2 evolved from water in the absence of the sacrificial donor [22]. Table 3. Key experiments for evaluating the photocatalytic system based on 33% DMASnBr 3 /g-C 3 N 4 (simulated solar light). As is apparent, the model biomass has a major role in sustaining H 2 evolution, which is 15-fold higher relative to the HER observed in neat water (the contribution of "direct water splitting" is <7%). In addition, the metal co-catalyst largely rules the photoreaction to give gas-phase H 2 , both in the presence of the sacrificial agent and in pure water. Notice that appreciable HER was gained also with no metal, underlining the potentiality of such photoactive material. This evidence corroborates a photoreaction mechanism typical of the photocatalytic systems for H 2 production from water [1,3,23], wherein: (1) The catalyst absorbs the radiation generating the characteristic charge separation (holes and electrons); (2) The sacrificial organic material serves as scavenger of the oxidizing species and, by undergoing gradual oxidation (photoreforming), supplies electrons; (3) The metal works as an electron collector and is the active site for hydrogen ions reduction while hindering charge carriers' recombination and backward reactions [3].
Regarding the DMASnBr 3 /g-C 3 N 4 composite, valence band holes-not hydroxyl radicals in the solution-are generated in g-C 3 N 4 and these trigger biomass oxidation to boost H 2 formation from water, while avoiding a complete aqueous phase reforming of the organic substrate until mineralization [1], which instead occurs in titanium dioxide photocatalysis, especially with Pd co-catalyst [16,22,24]. The negligible H 2 evolution (<0.008 µmoles h −1 ) observed by irradiation of neat water and 0.2 M glucose, as the control tests (Table 3), further substantiated the proposed mechanism. The synergistic effect between carbon nitride and perovskite described above and assessed in foregoing work [13] was here confirmed in glucose solution under the best conditions. Indeed, HER was 74 and 40 µmoles g −1 h −1 for g-C 3 N 4 and DMASnBr 3 individually tested, respectively, against 925 µmoles g −1 h −1 of the hybrid catalyst (see Table 3).
The values of apparent quantum yield (AQY), calculated as the percent ratio H 2 moles/incident photons moles [5] and turn over number (TON), calculated as the ratio H 2 moles/Pt moles [16], for DMASnBr 3 /g-C 3 N 4 in the optimized conditions are reported in Table 4. Table 4. Apparent quantum yield (AQY) and turn over number (TON) values for DMASnBr 3 /g-C 3 N 4 (0.5 g L −1 catalyst, 0.5 wt% Pt). These results are very interesting and highlight the efficiency of the composite catalyst that, despite the very low surface area (<4 m 2 g −1 , 18-times lower than that of the commercial nanometric P25 TiO 2 [25]), promoted HER equal to about one third of the latter (2906 µmoles g −1 h −1 ), under the same conditions. To point out the role of surface area and to provide comparable data, we tested a lower amount of P25 TiO 2 (0.028 g L −1 ) to have approximately the same catalytic surface area of 0.5 g L −1 of the composite. HER was 377 µmoles g −1 h −1 , 2.5-times lower than that afforded by employing the composite.

Sample
Additional trials were devoted to investigate the stability of the catalyst upon subsequent irradiations. As is shown in Figure 4, no loss of efficiency was noticed using the recycled powder in a new sample solution, and only a 25% decrease in H 2 evolution in a third treatment.  These results are very interesting and highlight the efficiency of the composite catalyst that, despite the very low surface area (<4 m 2 g −1 , 18-times lower than that of the commercial nanometric P25 TiO2 [25]), promoted HER equal to about one third of the latter (2906 µmoles g −1 h −1 ), under the same conditions. To point out the role of surface area and to provide comparable data, we tested a lower amount of P25 TiO2 (0.028 g L −1 ) to have approximately the same catalytic surface area of 0.5 g L −1 of the composite. HER was 377 µmoles g −1 h −1 , 2.5-times lower than that afforded by employing the composite.
Additional trials were devoted to investigate the stability of the catalyst upon subsequent irradiations. As is shown in Figure 4, no loss of efficiency was noticed using the recycled powder in a new sample solution, and only a 25% decrease in H2 evolution in a third treatment. The X-ray diffraction (XRD) profiles overlaid in Figure 5 attest the preservation of the pristine structure in the composite employed after the three sequential photoreactions (18 h total irradiation time) in aqueous glucose. The X-ray diffraction (XRD) profiles overlaid in Figure 5 attest the preservation of the pristine structure in the composite employed after the three sequential photoreactions (18 h total irradiation time) in aqueous glucose. Although in the XRD pattern of the as-synthesized catalyst the fingerprints of the perovskite are not visible, a broad signal centered on the principal peak of the carbon nitride that stabilizes an amorphous structure of the perovskite is present, as previously discussed [13]. Superimposable patterns were observed on both fresh and recycled materials (see Figure 5b); moreover, no peaks related to metallic Sn were found after catalysis (instead observed working in TEOA solution, data not shown). These findings support both the integrity of the hybrid material and the maintenance of the perovskite in the photocatalyst. Thus, this can be utilized for at least three consecutive irradiations, as a proof of photo-chemical stability that strengthens the applicability of DMASnBr3/g-C3N4 for a potential practical use.
Such new outcomes prompted us to test the composite in the presence of a polysaccharide sacrificial substance, and starch was chosen as the representative bio-polymer because of its low cost, large availability and attractiveness as bio-oxygenated for H2 photogeneration [5,24].
Hydrogen gas was effectively produced from starch solution (see Table 4), at a concentration of 4.5 g L −1 (the maximum solubility [5]), although at lower amount compared to glucose. This is justified considering that that the rate of hydrogen evolution from water decreases with increasing molecular weights and structure complexity of carbohydrates [15][16][17]24]. Additionally, since the VB holes are the oxidizing species of the present catalytic system, an intimate contact between the sacrificial substrate and the catalyst surface is certainly required. This could be more favored with monosaccharides than polysaccharides with a branched skeleton such as starch, which realistically has slower mass transfer kinetics from solution to the composite surface sites [16,24].
Such new findings highlight that DMASnBr3/g-C3N4 is able to carry forward H2 photoproduction exploiting the sacrificial role of a raw polysaccharide such as soluble starch, which thus can be employed directly without any pre-treatment, e.g., enzymatic hydrolysis or microwave-assisted hydroxylation [26,27].

Materials and Methods
Preparation and characterization of the two new composites 5% PEA2SnBr4/g-C3N4 and 33% DMASnBr3/g-C3N4 was reported by these authors in very recent works [12,13].
Although in the XRD pattern of the as-synthesized catalyst the fingerprints of the perovskite are not visible, a broad signal centered on the principal peak of the carbon nitride that stabilizes an amorphous structure of the perovskite is present, as previously discussed [13]. Superimposable patterns were observed on both fresh and recycled materials (see Figure 5b); moreover, no peaks related to metallic Sn were found after catalysis (instead observed working in TEOA solution, data not shown). These findings support both the integrity of the hybrid material and the maintenance of the perovskite in the photocatalyst. Thus, this can be utilized for at least three consecutive irradiations, as a proof of photo-chemical stability that strengthens the applicability of DMASnBr 3 /g-C 3 N 4 for a potential practical use.
Such new outcomes prompted us to test the composite in the presence of a polysaccharide sacrificial substance, and starch was chosen as the representative bio-polymer because of its low cost, large availability and attractiveness as bio-oxygenated for H 2 photogeneration [5,24].
Hydrogen gas was effectively produced from starch solution (see Table 4), at a concentration of 4.5 g L −1 (the maximum solubility [5]), although at lower amount compared to glucose. This is justified considering that that the rate of hydrogen evolution from water decreases with increasing molecular weights and structure complexity of carbohydrates [15][16][17]24]. Additionally, since the VB holes are the oxidizing species of the present catalytic system, an intimate contact between the sacrificial substrate and the catalyst surface is certainly required. This could be more favored with monosaccharides than polysaccharides with a branched skeleton such as starch, which realistically has slower mass transfer kinetics from solution to the composite surface sites [16,24].
Such new findings highlight that DMASnBr 3 /g-C 3 N 4 is able to carry forward H 2 photoproduction exploiting the sacrificial role of a raw polysaccharide such as soluble starch, which thus can be employed directly without any pre-treatment, e.g., enzymatic hydrolysis or microwave-assisted hydroxylation [26,27].

Materials and Methods
Preparation and characterization of the two new composites 5% PEA 2 SnBr 4 /g-C 3 N 4 and 33% DMASnBr 3 /g-C 3 N 4 was reported by these authors in very recent works [12,13].
Surface area measurements were carried out by the Brunauer, Emmett and Teller (B.E.T.) single point method using a Flowsorb II 2300 (Micromeritics, Norcross, GA, USA) apparatus. The sample was accurately weighed and degassed at 80 • C for 1.5 h, under a continuous stream of a N 2 -He (30:70) mixture, and then it was put in liquid nitrogen for gas adsorption. Evonik Aeroxide ® P25 TiO 2 (10-50 nm, 60.8 m 2 g −1 [25]) was supplied by Evonik Industries AG (Hanau, Germany). H 2 evolution experiments were conducted in distilled water containing 0.025-0.2 M glucose (99.9%, Carlo Erba Reagents) or 4.5 g L −1 soluble starch (ACS reagent, Carlo Erba Reagents), irradiated in Pyrex glass containers [28]. After addition of the catalyst (0.028, 0.5, 2 g L −1 ), the suspension was deoxygenated by Ar bubbling (20 min) and irradiated for 6 h, under magnetic stirring. When using Pt as the co-catalyst, chloroplatinic acid (H 2 PtCl 6 , 38% Pt basis) from Sigma-Aldrich (Milan, Italy) was employed as the metal source. Since Pt is in situ photodeposited on the catalyst surface, after Ar bubbling, a small volume from an H 2 PtCl 6 aqueous solution (0.15, 3 or 15 g L −1 ) was added using a 10-100 µL micropipette to the catalyst suspension, directly in the photoreactor. This was closed with sleeve stopper septa and was irradiated, as described in the following, achieving simultaneous Pt deposition and H 2 production [23,[28][29][30]. Irradiation was done under simulated solar light using a Solar Box 1500e (CO.FO.ME.GRA S.r.l., Milan, Italy) set at a power factor of 500 W m -2 , and equipped with UV outdoor filter made of IR-treated soda lime glass. The apparent photon flux, measured as previously described [5], was 1.53 × 10 -7 photons moles s −1 . The headspace H 2 was quantified by gas chromatography coupled with thermal conductivity detection (GC-TCD) [28]. The results obtained relative to H 2 evolution are shown in the paper as micromoles of gas per gram of catalyst per hour (µmoles g −1 h −1 , HER). Reproducibility was appraised on the RSD% from four independent photoproduction runs. For catalyst recycling tests, the catalyst was recovered by filtration (0.45 µm nylon membrane), washed with plenty of distilled water, left drying at room temperature for several weeks in the dark, and finally re-used under the optimized conditions. Room temperature Cu-radiation XRD spectra were acquired by a Bruker D8 diffractometer (Billerica, MA, USA).

Conclusions
Lab-scale tests proved the superior photocatalytic activity of DMASnBr 3 /g-C 3 N 4 against PEA 2 SnBr 4 /g-C 3 N 4 for H 2 evolution from glucose solution, under simulated solar light, using Pt as a co-catalyst. The systematic optimization of the experimental conditions by a DoE provided HER higher than 900 µmoles g −1 h −1 using the former composite. The results herein collected highlight the importance to work under selected conditions to maximize the H 2 yield while reducing the use of catalyst and co-catalyst. The comparison with P25 TiO 2 strengthened the catalytic efficiency of DMASnBr 3 /g-C 3 N 4 , while avoiding the handling of nano-sized powders. Notably, a rewarding H 2 formation (around 140 µmoles g −1 h −1 ) was gained, also avoiding any metal deposition. DMASnBr 3 /g-C 3 N 4 showed good photochemical stability to be utilized for two consecutive runs without any loss of efficiency and with a decrease of 25% in the third cycle, as verified by H 2 evolution measurement and XRD analysis. The system turned out to work also in raw starch solution as sustainable, model sacrificial biomass, providing ca. 150 µmoles g −1 h −1 . These outcomes corroborate the potential application of this new class of photocatalysts for clean energy retrieval under sustainable conditions and deserve further investigation to move the photoproduction experiments on sugar-rich wastewaters under natural solar light.