Photocatalytic degradation of caffeine in a slurry reactor with intermittent UV irradiation: optimization and response surface modelling

Pharmaceutical and personal care products (PPCPs) in the aquatic environment are of environmental concern and have received more attention among the researchers due to their persistence and non-biodegradability (Fathinia & Khataee 2015). PPCPs occur in the aquatic environment because of industrial and municipal wastewater discharges. Even though their concentration is usually at the nanogram level in wastewater, they have toxic impact on both humans and in the environment. Since PPCPs are persistent and non-biodegradable, their removal is of great concern (Barcelo & Petrovic 2008).

Caffeine (C₈H₁₀N₄O₂), a PPCP and persistent organic pollutant (POP), is a psychoactive drug consumed widely across the globe (Lovett 2005) and found in beverages like coffee, tea, and soft and energy drinks, as well as food products including chocolate and ice cream, and some medicines (Miners & Birkett 1996; Indermuhle et al. 2013; Marques et al. 2013; Ghosh et al. 2019). Caffeine has high water solubility (Ks >10,000 mg/L) (Elhalil et al. 2018), a low octanol water partition coefficient (Log K_ow = −0.07) and a long half-life (Edwards et al. 2015). Even though short-term exposure to CAF is not a threat to living things (Moore et al. 2008), its intense use can cause mutation, anxiety, tremors (Zhang et al. 2011) and cardiovascular diseases (Torres et al. 2014; Elhalil et al. 2018).

Various methods including chlorination (Gould & Richards 1984), photo-Fenton (Klamerth et al. 2010; Trovo et al. 2013), ozonation (Rosal et al. 2008; Souza & Feris 2015), UV photolysis (Shu et al. 2013) and semiconductor-based heterogeneous photocatalysis (Marques et al. 2013; Arfanis et al. 2017; Elhalil et al. 2018; Luna et al. 2018; Vaiano et al. 2018) have been investigated for CAF degradation in aqueous solutions. UV photolysis is ineffective for CAF removal (Buerge et al. 2003), other methods – for example, ozonation, photo-Fenton and chlorination – are limited by slow reaction rates (Indermuhle et al. 2013; Ghosh et al. 2019).

Heterogeneous photocatalysis is an advanced oxidation process (AOP) applied to degrade POPs and found to be effective (Zhou et al. 2017; Elhalil et al. 2018). When a semiconductor absorbs a photon energy (hν) ≥ the semi-conductor band gap (Eg), an electron (e⁻) may be promoted from the valence band (VB) to the conduction band (CB), thereby creating an electron vacancy – ‘hole’ (h⁺) (Bahnemann et al. 1997; Fujishima et al. 2008; Gaya & Abdullah 2008). The electron and hole can migrate to the catalyst surface where they participate in redox reactions with the species adsorbed on the catalyst surface. The holes can oxidize H₂O and OH⁻ to generate hydroxyl radicals (^•OH) whereas the electrons, while reacting with oxygen, can reduce molecular oxygen O₂ to O₂^•− (Mozia 2010; Laohaprapanon et al. 2015; Puga 2016). The ^•OH generated destroys the POPs, breaking them down into the relatively less toxic end products, CO₂ and H₂O (Diaz-Uribe et al. 2014). Among the semi-conductors applied as photocatalysts, such as TiO₂, ZnO, Fe₂O₃, WO₃, SnO₂, and ZrO₂ (Sudha & Sivakumar 2015; Ahmad et al. 2016; Awfa et al. 2018; Pathakoti et al. 2018), TiO₂ had attracted many researchers and been found successful due to its easy availability, low cost, non-toxicity, chemical and thermal stability, and applicability to solar energy (Bouarioua & Zerdaoui 2017; Rimoldi et al. 2017; Muangmora et al. 2020; Sujatha et al. 2020). As far as is known, no study has been conducted on CAF degradation using intermittent UV irradiation.

The objective of this study was to evaluate CAF photocatalytic degradation using intermittent irradiation from a 365 nm UV lamp (1 hour on and 10 minutes off for a trial period of 3 h 30 min) in a batch-scale reactor. Experiments were conducted in order to assess the effect of pH, TiO₂ dose, and initial CAF concentration for CAF removal. Photocatalytic degradation kinetics were also studied. Response surface methodology (RSM) using central composite design (CCD) was used to analyse the individual and interactive effects of operating variables on CAF and chemical oxygen demand (COD) removals.

Caffeine (C₈H₁₀N₄O₂, MW-194.19 g/mol) and titanium dioxide (Evonik Degussa P25, Germany TiO₂, 21 nm TEM size) were purchased from Sigma Aldrich. The specific surface of the TiO₂ photocatalyst was 50 m²/g, its band gap was 3.2 eV and it was a mixture of anatase (70%) and rutile (30%). Sodium hydroxide (99%), hydrochloric acid (36.5 to 38.0%), dichloromethane, acetonitrile and hexane, all of HPLC grade, came from Merck. COD determination was done using double-distilled water.

Figure 1 is a schematic of a laboratory-scale, batch photocatalytic reactor. The reactor, made of Plexiglas, had a working volume of 2 L. A 16 W, 365 nm, low pressure, Hg UV-A lamp was used. The lamp was in the centre of the reactor so that irradiation was uniform and was enclosed in a double-layered, quartz glass envelope. Water was recirculated through the envelope to prevent the lamp causing thermo-catalytic effects.

The photocatalytic degradation was optimized using response surface methodology (RSM) based on central composite design (CCD). The three factors, solution pH, initial CAF concentration and COD removal, were assessed for two responses, CAF degradation and COD removal. The experimental data were analysed using Design-Expert software version 7.0.0 Stat-Ease (2005).

The probable adsorption of CAF onto the TiO₂ surface was evaluated through dark control tests. A 2 L sample containing 5 mg-CAF/L and 0.5 g-TiO₂ /L, was covered with aluminium foil and kept in the dark for 24 h. The sample pH was maintained at 7.2. Analysis showed that only 9% of the CAF was adsorbed by the catalyst, perhaps because of electrostatic repulsion. This test was conducted to reduce the errors arising from non-photocatalytic adsorption, etc.

UV photolytic experiments were also conducted by irradiating a 2 L sample containing 5 mg-CAF/L for 180 min in the absence of TiO_2. The sample residual CAF concentration was determined after 180 min and CAF removal efficiency was found to be 32%.

Figures 2(a) and 2(b) show the effect of pH on CAF photocatalytic degradation. The CAF removal efficiencies obtained at pH 3.0, 5.0, 7.2, and 9.0 were 92.3, 89.1, 86.7, and 60.2%, respectively – that is, CAF photocatalytic degradation was best in acidic solution. This could arise from (i) the surface charge, and/or (ii) the agglomerated photocatalyst sizes.

The isoelectric point (pH_zc) is the pH at which a molecule has no net/zero electric charges. The pH_zc value of TiO₂ is 6.3. The TiO₂ surface carries a positive charge at low pH, while the charges on CAF and its intermediates are primarily negative and neutral. Thus, at low pH, organic molecule adsorption was facilitated and better photocatalytic degradation was promoted. The photocatalyst also exhibited more photocatalytic activity, due to the greater concentration of hydroxyl radicals. It is noted, too, that TiO₂ particles agglomerate when dispersed in water and that their agglomerated sizes vary from 0.2 to 1.2 μm (Fernandez et al. 2014). Particle agglomeration and particle-particle interaction reduce the active sites available for surface holes and electrons, and thus removal efficiency.

Pre-assessment of the optimum photocatalyst dose offers several advantages, including (i) avoiding the use of excess photocatalyst, and (ii) total adsorption of efficient photons. In this study, experiments were conducted with 5 mg-CAF/L initial concentration at pH 7.2, and aliquots were collected every 30 min and analysed for residual CAF concentration. The results are depicted in Figures 2(c) and 2(d).

Increasing the photocatalyst dose from 0.1 to 0.5 g/L raised photocatalytic degradation efficiency from 45.1 to 87.2%. Increasing it further, however, lowered removal efficiency; for example, to 57.8% at 0.9 g/L, because the excess photocatalyst dose did not facilitate light scattering but reduced light penetration into the reaction mixture (Qourzal et al. 2012; Elhalil et al. 2018). The optimum catalyst dose of 0.5 g/L was thus chosen for further studies.

The effect of initial CAF concentration on its degradation was studied by varying its initial concentration from 5 to 20 mg/L. The experiments were conducted at pH 7.2 with 0.5 g/L catalyst dose (the optimum). The results are shown in Figures 2(e) and 2(f). Hydroxyl radicals are short-lived and exist only for nanoseconds, so that they can only react at or near the location where they are formed. Increasing the CAF concentration increases the probability of collision with the oxidizing species, resulting in an increase in the degradation rate.

The k values obtained in this study coincide with the results observed by others (Chuang et al. 2011; Luna et al. 2018; Mai et al. 2018; Sacco et al. 2018; Ghosh et al. 2019; Muangmora et al. 2020). Table 2 shows comparisons of the pseudo-first-order rate constants. The half-life, t_1/2, determined was 69 min at the optimized concentration of 5 mg-CAF/L. While this coincides with the results obtained by others, Sacco et al. (2018) reported t_1/2 as 92 min and Elhalil et al. (2018) reported 87 min.

The interactions between dependent (CAF and COD removal) and independent (initial CAF concentration, catalyst dose and pH) variables were analysed using a quadratic model. For the experimental data, an RSM model with central composite design (CCD) was employed. The experiment parameters and the experimental design are shown in Tables 3 and 4.

The experimental data were thus validated, and the variables' synergistic and antagonistic effects were determined and analysed with the minimum number of experiments. The second-order polynomial equation was obtained to describe the correlation between the independent variable and CAF degradation. The ANOVA results for CAF degradation and COD removal are given in Tables 5 and 6.

ANOVA with a high coefficient of determination (R²) reveals a good agreement between the experimental and predicted values. In this study, ANOVA work yielded R² and adjusted R² values of 0.9557 and 0.9225, and 0.9369 and 0.8895, respectively, for CAF and COD removal, indicating that the quadratic model is adequate. The degree of freedom (df) indicates the number of estimated parameters used to compute the source's sum of squares. The Model F-value (test for comparing the source's mean square to the residual mean square) of 28.78 for CAF and 19.78 for COD removals obtained in this experiment implies that the model is significant. The ‘lack of fit’ is the amount by which the model's predictions miss observations. In other words, lack of fit’ is the variation of data around the fitted model. It should be insignificant for the model to fit well in the experimental data. The non-significant ‘Lack of Fit F-value’ of 163.32 and 0.19 obtained for CAF and COD removals indicates that there is a good correlation between process variables and response. The normal plot of residuals and the predicted vs actual response are shown in Figure 4. The residuals fall on a straight line, suggesting that the errors are distributed normally.

The residuals vs predicted, residuals vs run, Box-Cox plot for power transforms, and residuals vs concentration, for both CAF and COD removals, are shown in Figures 5 and 6.

The 3D surface plots are shown in Figure 7. Design-Expert software version 7.0.0 Stat-Ease (2005) with desirability approach was used to predict the maximum CAF and COD removal efficiencies (Stamatis et al. 2015). Maximum removal efficiencies were obtained at acidic pH because the surface charge on TiO₂ is then positive. Increasing the TiO₂ dose increased CAF removal efficiency up to 0.5 g-CAF/L, after which it decreased. Light screening and scattering caused the removal efficiency to decrease, while catalyst particle agglomeration also contributed substantially (Antonopoulou et al. 2012; Rani & Karthikeyan 2021). Increasing initial CAF concentrations also led to a fall in removal efficiency because the availability of hydroxyl radicals was inadequate.

The photocatalytic degradation of CAF was evaluated in a slurry photocatalytic reactor with intermittent UV irradiation. The highest CAF and COD removals were 86.7 and 66.7% respectively at optimized conditions: CAF=5 mg/L; pH=7.2 and TiO₂=0.5 g/L. CAF and COD removals followed pseudo-first-order kinetics. The experimental data were analysed with RSM modelling, using Design-Expert software. ANOVA yielded R²=0.9557 and adjusted R²=0.9225 for CAF removal, and R²=0.9369 and adjusted R²=0.8895 for COD removal, confirming good agreement between the experimental and predicted values. The RSM 3D surface plots also showed that all three operating parameters studied were significant in CAF removal. The study's results also showed that CAF molecules were not degraded completely and COD removal was less than CAF removal.

All relevant data are included in the paper or its Supplementary Information.