Characterization and evaluation of the photocatalytic activity of oxides based on TiO₂ synthesized by hydrolysis controlled by the use of water/acetone mixtures

Experimental

All reagents used (titanium tetraisopropoxide, 97%; isopropanol, 99.5%; ponceau 4R (P4R), 75%; reactive red 120 (RR120)—purity not informed by the supplier; Methanol, 99.8%; hexahydrated hexachloroplatinic acid, 37.5%; hydrochloric acid, 37% and sodium hydroxide, 98%) were of analytical grade, provided by Sigma–Aldrich. Acetone 99.5%, was provided by Synth. All solutions were prepared with ultrapure water obtained from an Elix 5 Milli-Q® water purification system.

Preparation of photocatalysts

The oxides were obtained by the sol-gel method, involving the solubilization of titanium tetraisopropoxide in isopropanol at 3 °C under ultrasonic stirring for 20 min, followed by its hydrolysis by the addition of water/acetone mixture by drip and precipitation under ultrasonic stirring.

The water/acetone mixtures were prepared with deionized water and different proportions of acetone (0%, 25%, 50% and 75% v/v). The resulting amorphous solids were washed with distilled water, centrifuged and sintered using a conventional oven at 400 °C for 5 h.

The standard photocatalyst, synthesized in aqueous medium, was called W1. The other oxides, synthesized by hydrolysis using different water/acetone mixtures (25%, 50%, 75% v/v of acetone), were named W1-25, W1-50, W1-75, respectively.

Characterization of the photocatalysts

The photocatalysts were characterized by different methodologies. By X-ray diffraction (XRD) using a XDR600 (Shimadzu, Kyoto, Japan) powder diffractometer operating at 40 kV and 120 mA, using Cu Kα radiation. The diffractograms were scanned in the range between 10 and 80° under a rate of 0.5° min⁻¹. Finally, they were refined by the method of Rietveld using the software “FullProf ” (Roisnel & Rodriguez-Carvajal, 2001). As criteria of mounting, the S factors were maintained between 1.22 and 1.31 (Table S1, Supplemental Information).

The Raman spectra were acquired at room temperature using a Bruker RFS 100/S spectrometer coupled to a 1,064 nm laser operating at 100 mW. Each Raman spectrum, with spectral resolution of 4 cm⁻¹, is the result of the accumulation of 128 scans.

The diffuse reflectance spectra were obtained using a double beam UV-1650 (Shimadzu, Kyoto, Japan) spectrophotometer, estimating the band energy by Kubelka–Munk treatment (Patterson, Shelden & Stockton, 1977). In these measures, barium sulfate was used as reference.

The N₂ adsorption–dessorption isotherms were obtained using an ASAP 2020 (Micrometrics) analyzer. The adsorption data were analyzed by the method proposed by Brunauer, Emmett and Teller (BET) for the surface area and the method of Barrett–Joyner–Halenda (BJH) for pore volume.

Transmission electron microscopy (TEM) images were obtained using a JEM-2100 (Jeol, Tokyo, Japan) microscope. In the preparation of the samples, suspensions containing the powders dispersed in acetone were used with the aid of a cutting-edge ultrasound. These suspensions were deposited on copper grids and air dried. From the images, obtained with the aid of the image editing software “ImageJ”, it was possible to calculate the particle size randomly selecting approximately 100 particles per image.

Photocatalytic assays

Four liters of an aqueous solution containing 100 mg L⁻¹ of the photocatalyst were used in each photodegradation assay, in combination with a concentration equivalent to 12.0 ppm of dissolved organic carbon of the dye—corresponding to 31.3 mg L⁻¹ of P4R or 43.5 mg L⁻¹ of RR120—used as oxidizable substrates. Detailed experimental assembly for the photodegradation assays was described in a previous study (Oliveira et al., 2012).

A commercial 400 W high-pressure mercury lamp (HPLN) without the protective bulb was employed as radiation source. Under this condition, its estimated photonic flux in the UVA was of 3.3 × 10⁻⁶ Einstein/s (Machado et al., 2008), with an irradiance inside the reactor of 100 W/m². During discoloration and dye mineralization monitoring, aliquots were collected every 20 min, in a total reaction interval of 140 min. The dyes discoloration was evaluated, without pH correction, by measuring the absorbance of the solutions, collected at different reaction times. Monitoring was done in the maximum absorbance wavelength in the visible of each dye—507 nm for P4R and 512 nm for RR120—using a UV-1201 (Shimadzu, Kyoto, Japan) spectrophotometer.

Mineralization was monitored from dissolved organic carbon (DOC) measurements, using a TOC-VCPH/CPN (Shimadzu, Kyoto, Japan) analyzer, aiming to identify the most efficient photocatalyst. These experiments were restricted only to the monitoring of P4R photodegradation.

The most efficient photocatalyst was also submitted to photodegradation tests using Remazol Red (RR120), comparing its performance with that presented by the commercial catalyst Evonik Degussa TiO₂.

These assays were conducted at least in triplicate and separately for each dye.

The reuse of the most efficient photocatalyst was evaluated using P4R as oxidizable substrate. For this, after each reaction the photocatalyst was separated from the supernatant by decanting, washed with distilled water, centrifuged and dried at 70 °C for 24 h, and then reused under the same described conditions using a new load of the same dye. Each test was performed in quadruplicate in order to compensate for the losses that occurred during the washing of the catalyst, in order to ensure a constant mass in each cycle.

Subsequently, hydrogen production assays were done using the most effective synthesized photocatalyst, as well as the commercial catalyst Evonik Degussa TiO₂ and the W1 oxide. In these experiments, the concentration of catalyst was similar to that used in the assays of dye degradation, being this oxide loaded by photoreduction with 0.5% m/m of Pt, furnished by a solution of hexachloroplatinic acid. So, the Pt-loaded photocatalyst was then suspended in 750 ml of a water/methanol mixture containing 20% v/v of methanol, this last being used as sacrificial reagent. These assays occurred under continuous stirring. The pH of the reaction medium was adjusted in 6.2 using solutions 0.1 mol L⁻¹ of HCl or NaOH. Finally, the potential of reuse of the photocatalyst used in such assays was evaluated in at least three photocatalytic cycles. In the reuse assays, only the pH adjustment of the reaction medium was performed at the beginning of each new cycle. The first cycle was equivalent to the first hydrogen production test, carried out for 5 h. Thus, the total reaction time was 15 h.

For all photocatalytic assays the results are the averages of at least three individual experiments.

For operator protection and better use of radiation produced by the lamp, the reactor was positioned in a box internally covered with aluminum film, Fig. 1.

The reactor, built in borosilicate glass, has a cooling jacket connected to a thermostat bath on its outside which keeps the temperature of the reaction medium stabilized at 20 °C throughout the reaction. Before each experiment, the reactor was purged with N₂ for 20 min to eliminate dissolved gases, especially oxygen. The same HPLN lamp reported above was used as radiation source. For analysis of the gases produced during the reaction, aliquots of one mL of these gases were collected at intervals of 30 min of reaction, in a total period of 5 h. These samples were analyzed at 230 °C in a Shimadzu GC-17A gas-phase chromatograph equipped with thermal conductivity detector (TCD) and a Carboxen™ 1010 Plot capillary column. Argon, with flow of 40 mL min⁻¹, was employed as carrier gas.

Characterizations

By analyzing the X-ray diffraction (XRD) data, Fig. 2, it is possible to infer that all oxides have well-defined diffraction peaks suggesting crystallinity for these materials most likely due to the heat treatment used in the synthesis process. In addition, according to reports found in the literature and the crystallographic files JCPDS 21-1272 (anatase) and 29-1360 (brookite), all oxides exhibit major peaks characteristic of the anatase phase, and secondary peaks related to the brookite phase (Di Paola, Bellardita & Palmisano, 2013; Neto et al., 2017; Patrocinio et al., 2015; Hu, Tsai & Huang, 2003).

The mean size and mean deformation of crystallite were calculated from the data obtained from Rietveld refinement, as presented in Table 1. The diffractograms, accompanied by the respective calculated diffraction profiles, experimentally obtained profiles, and residual curves and Bragg diffractions adjusted by Rietveld method can be seen in Fig. S1, in the Supplemental Information.

Rietveld refinement data demonstrates that the percentage of brookite phase increases from 13% to 22% with the addition of acetone as co-solvent in the hydrolysis of titanium tetraisopropoxide. Despite this, the increase in the proportion of acetone from 25% to 75% did not result in an equivalent increase in the percentage of brookite phase, suggesting that the use of acetone only interfered in hydrolysis, affecting the organization of critical nuclei in the oligomeric network of titanium, in order to preorder the crystallization of the mentioned phase. On the other hand, the average crystallite size of the anatase phase was about 30% lower for W1-75, compared to the other oxides, including the W1, prepared without addition of acetone during its synthesis. This suggests that the excess of acetone should promote a significant reduction in the average crystallite size of the anatase phase, favoring the increase in the average crystallite size of brookite. Thus, the mean deformation of the crystallite follows the same trend, that is, if the secondary phase becomes larger it will present larger deformations when compared with the primary phase.

As in the X-ray diffractograms, the Raman spectra also evidence the mixed composition of two crystalline phases, Fig. 3. In all oxides, five main bands attributed to the anatase phase are observed respectively at 145 cm⁻¹ (E_g), 198 cm⁻¹ (E_g), 399 cm⁻¹ (B_1g), 519 cm⁻¹ (B_1g) and 640 cm⁻¹ (E_g) (Sahoo, Arora & Sridharan, 2009). Between 200 and 500 cm⁻¹ four bands of lower intensity are observed: at 247 cm⁻¹ (A_1g), 323 cm⁻¹ (B_1g), 368 cm⁻¹ (B_2g) and 456 cm⁻¹ (B_2g), attributed to the phase brookite. In addition to these bands, this phase features a band of greater intensity around 150 cm⁻¹ which may be superimposed with the band identified at 145 cm⁻¹, attributed to anatase, thus influencing the width of the E_g Raman mode (Di Paola, Bellardita & Palmisano, 2013; Iliev, Hadjiev & Litvinchuk, 2013).

From the expanded spectrum, Fig. 4, it was possible to observe a small enlargement of the band centered at 145 cm⁻¹ in the Raman spectrum of the photocatalyst W1-75. This effect should be related to the smaller particle size since the lifetime of the vibrational mode tends to be shorter as particle size decreases, which ends up resulting in band enlargement (Liu et al., 2012; Zhu et al., 2012).

From the diffuse reflectance spectra, Fig. 5—Insert, expressed in terms of the Kubelka–Munk’s function, Fig. 5, and by applying the direct method, it was possible to estimate the band gap energies (Eg) of the photocatalysts (Liu & Li, 2012). The estimated Eg were as follows: 3.23 eV for W1, 3.24 eV for W1-25, 3.22 eV for W1-50, and 3.23 eV for W1-75, indicating that the estimated band gap energies have not undergone major changes, which agree with the data reported in the literature for pure TiO₂ (Martin-Somer et al., 2020; Neto et al., 2017; Resende et al., 2017). Most likely, this stems from the synthesis conditions adopted in this work, where none dopant material was added. It is known that the E_g displacement to lower energies occurs preferably in synthesis that promote the doping of oxides with metal cations (Santos et al., 2015b), non-metallic anions (Liu et al., 2014), co-doping (Kuvarega, Krause & Mamba, 2015), and self-doping (Chen et al., 2011).

As for N₂ adsorption and desorption of these oxides, Fig. 6, the analysis of the adsorption–desorption isotherms suggests that they are type IV (IUPAC, 1985), characteristic of mesoporous materials with an average pore diameter between 2 and 50 nm, Table 2. Hysteresis profiles are very close to those of type H2, associated with more complex mesoporous structures, in which the distribution of pore sizes and their shape are not well defined (Guan-Sajonz et al., 1997). It is also evident that the photocatalysts W1-50 and W1-75, synthesized by hydrolysis using the highest percentages of acetone, present slightly more steeper isotherms compared to the oxides W1 and W1-25, also exhibiting greater heterogeneity in pore distribution compared to these same oxides.

Table 2 presents the morphological parameters related to the synthesized oxides. In general, oxides obtained from hydrolysis using water/acetone mixtures did not undergo significant morphological changes, since for W1 the oxide porosity is practically the same presented by W1-50 and W1-75. On the other hand, the surface area of these two oxides is between 10% and 12% larger than that of W1. This may favor the adsorption of organic matter on their surfaces, which can consequently favor the photocatalytic efficiency. In addition, it was observed an inverse correlation between the surface area and the average particle size, except for the W1-25 that presented wide variation on its particle size.

The TEM images, Fig. 7, suggest a dense aspect to the particles, which have irregular spherical shape and a strong tendency to aggregation, giving rise to clusters of TiO₂. This behavior should be related to the high level of hydrolysis provided by the synthesis method (Jiang, Herricks & Xia, 2003). However, agglomeration appears to have been minimized by addition of acetone as co-solvent in the hydrolysis, evidencing that its use decreased the hydrolysis rate of the precursor. This, consequently, should favor particle dispersion. On the other hand, the particle sizes estimated from these images do not suggest a role of acetone on this property, as can be seen by the values estimated for the particle size: (14 ± 1) nm, (17 ± 3) nm, (10 ± 1) nm and (13 ± 1) nm, respectively for W1, W1-25, W1-50 and W1-75. The histograms can be viewed in Fig. S2, in the Supplemental Information.

Photocatalytic activity: degradation/mineralization of organic compounds

Table 3 presents the photocatalytic performance of the synthesized oxides and of the commercial oxide TiO₂-P25, in the degradation of the two azo dyes used in this study as oxidizable substrates. For comparative purposes, the dyes were also submitted to direct photolysis, in order to evidence the role of the photocatalysts in the photodegradation.

The discoloration (k_dis) and mineralization (k_min) rate constants were estimated from the application of the kinetic model of Languimir–Hinschelwood (Hoffmann et al., 1995; Machado et al., 2012), considering that the kinetic regimen in these photocatalytic processes follows a pseudo-first order kinetics (Machado et al., 2003, 2012; Santos et al., 2015b; França et al., 2016). Graphs containing the kinetic data corresponding to these values are presented in the Suplemental Information (Figs. S3–S5).

The expected low efficiency both in degradation and discoloration via direct photolysis, compared to the results achieved by the photocatalysts can be related to the energy of the incident photons, provided by the radiation source (Machado et al., 2008), and to the very low rate of formation of radical species, produced by homolytic scission of labile bonds present in these dyes (Kumar et al., 1999).

In the experiments involving the participation of the photocatalysts, the degradation occurred more efficiently due the participation of reactive oxygen species, among them the hydroxyl radicals (HO∙) and superoxide radical-ions (O_∙−²), generated mainly by water decomposition. Such species, due their low selectivity (Machado et al., 2012), together with secondary radical species produced during the photocatalytic process, tend to promote the oxidation of organic substrates present in the reactional medium (Oancea & Meltzer, 2014; Santos et al., 2015a). The dissolved oxygen, present in the aqueous medium, as example, when reduced by the semiconductor, contributes with the formation of O_∙−² and perhydroxyl radicals, which, although less oxidizing than HO∙ (Machado et al., 2012), are very important in promoting the degradation of organic substrates.

The values of the apparent rates of discoloration and mineralization, observed in the reactions mediated by the photocatalysts evaluated in the present study, Fig. S3—Supplemental Information, suggest that these reactions occur in two stages, following kinetics of apparent pseudo-first order. Initially, the reaction occurs at a rate lower than in the second stage, when the apparent rate constant, in some cases, is three times higher. The higher rate constant in the second stage should be a consequence of the more favored adsorption of the fragments of organic matter formed in the first stage of the process, combined with the good availability of oxygen and water, important for the formation of radicals responsible for the oxidation of organic matter (França et al., 2016).

The mineralization of P4R mediated by the oxides W1-25, W1-50 and W1-75 increased respectively 5.7%, 18.6% and 24.3% more than the result obtained using W1, when 70% mineralization was achieved. It is noteworthy that the hydrolysis process which gave rise to this oxide, occurred exclusively in the presence of water. It should be noted that the mineralization achieved using TiO₂ P25 as photocatalyst was only 8% higher than that obtained when W1-75 was employed.

Although the mineralization and discoloration of P4R conducted using W1-75 presented the best performance among the synthesized oxides, the result observed was only 4.8% higher than that achieved using W1-50. Considering the proportion of acetone used in the synthesis of W1-75 and its limited photocatalytic performance, W1-50 was considered as the most effective catalyst for mineralizing P4R, being therefore preferably applied in the following stages of the present study. Since W1-25 presented intermediate performance to that observed for W1 and W1-50 catalysts, evaluating its efficiency, regarding the degradation of RR120, was therefore considered unnecessary.

The good photocatalytic activity presented by these oxides, in particular the W1-50, can be attributed mainly to the mixed composition of the phases and high crystallinity obtained after heat treatment, confirmed by the XRD and Raman spectra. The presence of an additional phase tends to introduce defects that tend to favor the photocatalytic activity of a photocatalyst (Kandiel et al., 2010). Brookite, for having conduction band approximately 0.14 eV more negative than anatase, ends up favoring the interfacial electron transfer by imposing an energy barrier for the return of the excited electrons to the valence band of anatase, which tends to favor the coexistence of charge carriers (Kandiel et al., 2010; Di Paola, Bellardita & Palmisano, 2013; Patrocinio et al., 2015).

Reuse assays were performed using the recycled W1-50 in the photocatalytic degradation of the dye P4R. The recycled W1-50 was separated by decantation after the first photocatalytic test. It was then washed with distilled water, centrifuged and dried at 70 °C for 24 h. After this procedure, the recycled oxide was used to promote the degradation of P4R present in a new solution. The discoloration level remained at 100% while the mineralization performance decreased about 30%. This loss of performance should be related to photocatalyst poisoning caused by species adsorbed on the catalyst at the end of each photocatalytic cycle, compromising the availability of active sites (Nakhjavani et al., 2015). It is important to consider that the recycled catalyst was not submitted to any prior purification procedure aiming the removal of contaminants incorporated by adsorption the previous cycles. The discoloration and mineralization profiles, as well as the kinetics of discoloration and mineralization in this reuse assay, are available in the Supplemental Information, Fig. S4.

Table 3 also presents the performance of the oxides W1, W1-50 and TiO₂ P25 in the mineralization and discoloration of the dye RR120. In this case, although RR120 has a more complex chemical structure than P4R, presenting two azo groups and two triazine groups, the performance achieved by W1-50 was comparable to that presented when using TiO₂ P25 differing only by the kinetic constants of mineralization (k_min). The residual total organic carbon (TOC) observed after degradation of both P4R and RR120 (Figs. S3 and S5; Supplemental Information), should be related to the presence of short-chain carboxylic acids, recalcitrant to photocatalytic degradation (França et al., 2016). Studies have shown that the triazine groups present in the chemical structure of RR120, when photocatalytically oxidized, give rise to cyanuric acid, very resistant to degradation (Chun & Yizhong, 1999; Wang, 2000; Camarillo & Ricón, 2011).

Photocatalytic hydrogen production

The profiles of hydrogen production as function of the reaction time, Fig. 8, show a superior performance of W1 and W1-50 compared to TiO₂-P25.

The process mediated by W1-50 produced approximately 56 mmols of gaseous H₂, while in the same period TiO₂ P25 produced 43% less. On the other hand, W1 produced approximately 3% less hydrogen than W1-50. In addition, it is explicit that the production of H₂ using the oxides presented in this study increased until the end of the assay, suggesting that the photocatalytic process was still in its propagation stage. H₂ production using TiO₂ P25 presented a different profile, suggesting typical accommodation of processes in stages near termination. It is known that TiO₂ P25 is the result of the crystalline composition between anatase and rutile. The advantage of the photocatalysts presented in this work should be in the combination of anatase and brookite that tends to increase the photocatalytic efficiency of the semiconductor. Liu et al. (2014), for example, demonstrated that the recombination of the photoinduced charge carriers is minimized when the semiconductor oxide have a structure based on this kind of phase composition. This behavior occurs due to the most negative cathode potential of the conduction band of the brookite phase, more negative than the proton reduction potential and the cathode potential of the conduction band of anatase, thus favoring its conversion to H₂ (Kandiel et al., 2010; Patrocinio et al., 2015; Tay et al., 2013). Besides, this phenomenon facilitates the interfacial transfer of electrons while an energy barrier is established, which hinders their return, thereby prolonging the coexistence of the charge carriers. With this, both the oxidative (metanol oxidation) and the reductive process (H₂ production) end up being favored.

In terms of specific hydrogen production rate (SHPR), the production mediated by W1-50 (138.5 mmol h⁻¹ g⁻¹) was 60% higher than the achieved using TiO₂ P25 (86.4 mmol h⁻¹ g⁻¹). Even the SHPR of W1 (126.5 mmol h⁻¹ g⁻¹) was higher than that of the commercial photocatalyst. It is observed, therefore, that the variant of the sol–gel synthesis proposed in this study resulted in pure photocatalysts, such as the W1-50, which present SHPR much higher than that of TiO₂ P25, as well as of photocatalysts recently reported in the literature. Selcuk, Boroglu & Boz (2012), in a study involving a catalyst resulting from TiO₂ codopage with platinum and nitrogen, reported, under the best operating conditions, a SHPR of 13 µmol h⁻¹ g⁻¹, a value significantly lower than the achieved using W1-50. This study involved the use of a 400 W mercury lamp as a source of radiation and a solution containing 10% methanol. In another study, Lin & Shih (2016) using a TiO₂ doped with copper and nitrogen, obtained a SHPR equal to 27.4 mmol h⁻¹ g⁻¹, a value approximately 5 times lower than the achieved using the W1-50 in the present study. These authors also used a 400 W mercury lamp as a source of radiation. In this case, the catalyst was suspended in a solution containing 20% methanol.

In addition, the reuse of the W1-50 was evaluated for the collection of information related to its photostability. These tests consisted in evaluating the reproducibility of the catalytic action of this oxide by performing three consecutive photocatalytic cycles of 5 h each using the same initial conditions applied to the system, with the exception of the pH of the medium, adjusted at the beginning of each additional cycle. These results are presented in Fig. 9.

Although there is an increase in H₂ production in the other cycles, compared to the first cycle, during the photocatalytic cycles W1-50 presented a similar profile of H₂ production in the three cycles, Fig. 9. In the second cycle, the SHPR increased by about 10% (154.7 mmol h⁻¹ g⁻¹) compared to the first cycle, whereas in the third cycle this increase was of 9%. The good photostability, reproducibility and significant yield in H₂ production during these experiments may be related to the absence of contaminants in the catalyst in the different cycles. Certain oxides based on TiO₂, obtained from associations, anchoring and doping with other substances, show losses in the capability of H₂ production as the photocatalytic cycles succeed. The reason for this has been pointed out as being due the photodesorption of compounds associated or anchored or by photoreduction of metals on TiO₂ surface, thus contaminating the reaction sites (Zhang et al., 2013; Yuan et al., 2015).