Tunable band gap of Bi³⁺-doped anatase TiO₂ for enhanced photocatalytic removal of acetaminophen under UV-visible light irradiation

Pharmaceutical pollutants tend to be discharged into the aquatic environment from several sources that include: manufacturing facilities (Pérez et al. 2017), disposal of various consumer products of a chemical nature (Douziech et al. 2018), and hospital waste (Verlicchi et al. 2010; Zhang et al. 2010). The presence of these pollutants has negatively impacted both humans and aquatic species (Halpern et al. 2008; Kostich et al. 2014), raising concerns about the effect of the increased level of such pharmaceutical residues on human health and the environment (Jones et al. 2005). Acetaminophen, also known as paracetamol, is the most widely used pharmaceutical in the treatment of fever and headache, and consequently is the most common pharmaceutical residue detected in wastewater (Petrie et al. 2015). The main concern associated with the excessive use of acetaminophen is how easily it accumulates in the aquatic environment due to its characteristics that include high solubility and hydrophilicity (Granberg & Rasmuson 1999). The treatment of such pharmaceutical residue using common chemical treatment processes, e.g. chemical oxidation, results in the production of secondary pollutants resulting from the chemical reaction (Wu et al. 2012). Hence, other more environmentally friendly treatment methods need to be explored.

One such method for the environmental remediation of pollutants is the process of photocatalysis, which involves the use of a semiconductor photocatalyst to carry out redox reactions on its surface, speeding up the rate of chemical reactions (Wu et al. 1999). When light with photons equal to or greater than the band gap of the semiconductor is absorbed, photo-generated electrons and holes migrate to the surface of the semiconductor where redox reactions take place, by which an effective photocatalyst is one in which the charge couple redox potential is within the photocatalyst's band gap range (Ibhadon & Fitzpatrick 2013). For the conduction band, the bottom energy level regulates the reducing potential of the photoelectrons, whereas the valence band top energy level regulates the oxidizing potential of the photo-generated gaps (Ola & Maroto-Valer 2015). Hence, a suitable semiconductor photocatalyst should have appropriate band gap energy. Other characteristics of an ideal photocatalyst include ease of production (Li et al. 2016), stability (Meng et al. 2011; Huang et al. 2013), cost effectiveness (Zhao & Liu 2008), safety for humans and the environment (Lee et al. 2010), easy activation using solar light (Asahi et al. 2001), as well as the capability to effectively catalyze reactions (Qu & Duan 2013). This process of photocatalysis has been widely implemented in pollution control for both water and air, wherein it has the capacity to remove waste and degrade toxic substances into non-toxic forms (Chong et al. 2010; Luengas et al. 2015).

Different photocatalysts have been reported in the literature, including: GaAs, PbS, CdS, ZnO and TiO₂ (Zhang et al. 2009; Wang et al. 2013; Etacheri et al. 2015). Among these photocatalysts, GaAs, PbS, and CdS suffer from instability in aqueous media and are toxic (Gupta & Tripathi 2012). As for ZnO, it exhibits lack of stability when dissolved in water where it produces Zn(OH)₂ on the surface of the particles, inactivating the catalyst with time (Daneshvar et al. 2007). On the other hand, the characteristics of TiO₂ make it an ideal photocatalyst, as it is cheap (Macwan et al. 2011), non-toxic (Choi 2006) and photo-stable in solution (Christy et al. 2009). Other characteristics of TiO₂ that are essential for a photocatalyst include high surface area and high porosity, which result in an improved reaction rate because of the improved level of interaction of the reactants with the active sites (Yu et al. 2003). As a result, the use of TiO₂ as a photocatalyst has been widely explored in the literature as displaying promise in environmental remediation applications (Pelaez et al. 2012).

TiO₂ has three main phases: anatase, rutile and brookite. The anatase phase is the most active form, used in photocatalysis applications (Zhang et al. 2000). However, a disadvantage of this photocatalyst is its lack of visible light absorption due to its large band gap energy of 3.2 eV, thus limiting its application under visible light irradiation (Yin et al. 2009; Landmann et al. 2012). To overcome this drawback, different methods have been reported in the literature to extend its spectral response to visible light, including metal (Fiszka Borzyszkowska et al. 2016) and non-metal (Burda et al. 2003) doping, coupling of TiO₂ with other semiconductor materials (Liu et al. 2007a), fabrication of nanocomposite photocatalysts with different morphologies (Liu et al. 2007b; Wang et al. 2007; Zhong et al. 2014; Chen et al. 2017), and dye sensitization (Park et al. 2000). Among these methods, doping has been found to be an effective way to extend the light absorption to the visible region by introducing additional energy levels between the valence band and conduction band of the photocatalyst acting as a trap for the charge carriers, facilitating their separation from the bands. This in turn allows more charge carriers to successfully diffuse to the surface of the photocatalyst (Kim et al. 2005; Yang et al. 2010).

Bismuth based composite oxide semiconductors such as bismuth titanate (Feng Yao et al. 2003) show enhanced photocatalytic activity under visible light irradiation, which can be attributed to their narrow band gaps compared to TiO₂. In such a case, band gap reduction is attributed to the hybridized valence band of Bi⁺³ 6S lying above the O 2p and the Ti⁴⁺ 3d level just below the Bi³⁺ 6p in the conduction band. As a result, band gap energy reduction is assumed to be due to the excitation of a 6S electron of Bi³⁺ into the 3d of Ti⁴⁺. This behavior has been observed in other bismuth-based materials such as bismuth vanadate (Huang et al. 2014), bismuth molybdate (Shimodaira et al. 2006) and bismuth tungstate (Zhang & Zhu 2012). In these semiconductors, the hybrid valence bands formed between O 2p and Bi³⁺ 6S levels and conduction band between V; 3d, Mo; 4d and W; 5d orbitals respectively and the Bi³⁺ 6p (Walsh et al. 2009). Accordingly, the absorption extends to a longer wavelength due to the excitation of a 6S electron of Bi³⁺ into the d orbital representing the metal ion. Thus, incorporating Bi³⁺ ion into the TiO₂ band structure is anticipated to have a similar effect, shifting light absorption to the longer wavelengths, i.e. visible region. This in turn induces high photocatalytic activity under visible light irradiation of the bismuth doped photocatalysts.

In this work, we present a facile propylene oxide (PO) assisted sol-gel preparation of TiO₂ and bismuth-doped TiO₂. The performance of the as-prepared photocatalysts was evaluated by the removal of acetaminophen from an aqueous solution over a period of 4 hours under UV and visible light irradiation.

Titanium (IV) n-butoxide (Ti(OⁿBu)₄, 97%), Bi(NO₃)₃.5H₂O, 2-propanol (i-PrOH, 99.7%), hydrochloric acid (HCl, 37%), acetylacetone (acac, 99.3%) and propylene oxide (PO, 99%) were purchased from Sigma Aldrich and used as received. All chemicals were used without further purification. Solutions were prepared using doubly distilled water passed through a Milli-Q apparatus.

The TiO₂ photocatalyst was prepared through mixing Ti(OⁿBu)₄ solution and the hydrolysis solution that were prepared as indicated below.

Ti(OⁿBu)₄ solution: The TiO₂ photocatalyst was synthesized using Ti(OⁿBu)₄, 2-propanol (i-PrOH) used as solvent, acetylacetone (acac) as the complexing agent and PO as the gelation promoter using the sol-gel method at room temperature. In a typical synthesis, Ti(OⁿBu)₄ was dissolved in i-PrOH in a i-PrOH:Ti(OⁿBu)₄ v/v ratio of 5:1 where acac was added in a molar ratio Ti(OⁿBu)₄: acac of 1:2. The prepared solution was kept under constant stirring for 2 hours followed by the addition of PO where the PO: Ti(OⁿBu)₄ molar ratio was 10:1.

The hydrolysis solution: The hydrolysis solution was made by mixing distilled water and i-PrOH (i-PrOH:H₂O, 1:20) where the H₂O:Ti(OⁿBu)₄ molar ratio was 10:1. The hydrolysis reaction was initiated through the addition of Ti(OⁿBu)₄ solution to the hydrolysis solution drop wise under vigorous stirring. The pH of the final solution was adjusted to pH = 2 through the addition of 37% HCl. This reaction mixture was stirred for an additional period of 3 hours and then aged for a few days for the formation of the gel. After drying, powder was air-calcinated at 400 °C for 3 hours.

For the preparation of the Bi³⁺-doped TiO₂ photocatalysts, the same procedure as that for the preparation of TiO₂ photocatalyst was implemented, where molar amounts of Bi were added. Samples of 1%, 3%, 5%, and 10% of Bi-doped TiO₂ were prepared, respectively, where Bi was first dissolved in a i-PrOH:H₂O v/v ratio of 1:5 followed by the addition of a few drops of 37% HCl. The Bi³⁺ solution was then added to the Ti(OⁿBu)₄ solution prior to the PO addition. Finally, hydrolysis solution was added dropwise under vigorous stirring. All as-prepared Bi-doped photocatalysts were air-calcinated at 400 °C for 3 hours.

According to the above equation, through the construction of a plot of (αhν)² versus hν (eV), the band gap energy E_g can be obtained by extrapolating a line through the steep linear part of the curve to the hν axis as shown in Figure 1, whereas the band gap values are tabulated in Table 1.

X-ray diffraction characterization was performed using a Shimadzu-6100 powder XRD diffractometer with Cu-Kα radiation. The working voltage and current of the X-ray tube were 40 kV and 40 mA respectively, where λ = 1.542 Å. The diffraction data were collected in the 2θ angle range of 20–80 degrees at a rate of 2 degrees/min. Figure 2 displays the XRD results for all the photocatalysts used in this study.

Scanning electron microscope images and energy-dispersive X-ray spectroscopy were obtained using JEOL JSM–6010LA. After film preparation on ITO-coated glass, samples were rinsed and allowed to dry. Images and data were then collected at accelerating voltage of 20 kV as presented in Figures 3 and 4 and Table 1.

Surface area and porosity were characterized using N₂ adsorption at 77 K using a Quantachrome Autosorb-1 volumetric gas sorption instrument. Before measurements, samples were degassed at 150 °C for 1 hour. Brunauer-Emmett-Teller (BET) theory was used to calculate surface area, and pore size distributions were determined by the Barett-Joyner-Halenda (BJH) model based on the desorption branch of the N₂ isotherms. Specific surface area and porosity data analysis are presented in Table 1.

X-ray photoelectron spectroscopy measurements of all as-prepared photocatalysts were performed using the Kratos Axis Ultra DLD spectrometer (Kratos Analytical Ltd, Manchester, UK) with Al K_α1 X-ray source. The energy of an X-ray photon of 1.486 keV with pass energy of 160 eV was used for the survey spectrum and 20 eV for narrow scans. All spectra were collected at a 54° take-off angle and analyzed area of (700 × 300 μm) using the combination of electrostatic and magnetic lens (hybrid mode). The C 1 s peak from adventitious hydrocarbon at 284.8 eV was used as an energy reference to correct for charging. Surface charging effects were minimized using a charge balance operating at 3.6 V and 1.8 V maintained as filament bias. During XPS analysis, sample charging was neutralized using an electron flood gun. All spectra were recorded under ultra-high vacuum conditions below 5 × 10⁻¹⁰ mbar. All data processing was carried out using the Casa XPS software (Casa Software Ltd) package. For calculation of the elemental composition of the prepared photocatalysts, the determined areas under the peaks were normalized using the Scofield sensitivity factors. For data evaluation, the background of the Auger photoelectron peaks was subtracted by applying a Shirley-type of background and the data were curve-resolved using an 80% Gaussian/20% Lorentzian sum.

The photocatalytic evaluation of Bi³⁺-doped TiO₂ as-prepared photocatalysts in the removal of acetaminophen in aqueous solution was performed using a Luzchem Photoreactor (Mod. LZC-1, Luzchem Research Inc., ON, CAN) equipped with two UVA lamps centered at ∼350 nm, two UVB lamps centered at ∼300 nm with a peak of 313 nm, two UVC (254 nm) germicidal lamps, and seven cool white fluorescent tubes. Temperature was kept constant at 25 °C for all the reactions. In a typical photocatalytic run, powdered photocatalyst in the amount of 0.1 gL⁻¹ was suspended in aqueous solution of 10⁻⁴ M acetaminophen in a 100-mL quartz beaker. The suspension was magnetically stirred in the dark over 30 min to achieve a complete adsorption/desorption equilibrium and then it was irradiated with a UV-Vis light source. The light intensity was measured using a digital light meter (TES-1330A) and it was 11,460 lux (44 W/cm²). Aliquots of the reaction medium were periodically sampled and filtered using a polytetrafluoroethylene (PTFE) membrane filter (Millipore, 0.22 μm) prior to analysis. The acetaminophen concentration was measured employing a spectrophotometric method by using a double beam UV-Vis SPECORD 210 PLUS spectrophotometer and following the decrease in the absorbance of acetaminophen at 250 nm over time definite intervals.

Tauc plots of the UV-Vis DRS spectra of the pure TiO₂ and bismuth-doped TiO₂ as-prepared photocatalysts are presented in Figure 1. Pure TiO₂ exhibited a direct allowed band gap of 3.08 eV. On the other hand, all bismuth doped as-prepared photocatalysts showed quite similar light absorption shifts towards the visible region of the spectrum, with the lowest band gap of 2.8 eV for the 10% Bi³⁺-doped TiO₂ photocatalyst, i.e. a red shift of approximately 40 nm from that of pure TiO₂. This red shift observed for all doped as-prepared photocatalysts can be attributed to the presence of bismuth ion within the TiO₂ lattice and lack of presence as a secondary phase as indicated solely by the XRD spectra presented in Figure 2.

Figure 2 displays the XRD patterns of the as-prepared pure and bismuth-doped TiO₂ photocatalysts with varying amounts of Bi³⁺ dopant. From the figure, it can be observed that the peaks for the doped TiO₂ samples coincide with those of the pure TiO₂ in which all the patterns displayed indicate the presence of the pure anatase TiO₂ phase with no peaks corresponding to another phase or newly formed lattice structure. Furthermore, the fact that there was no observed shift in the diffraction angles of the XRD peaks indicates that doped Bi³⁺ ions did not cause any changes in the phase composition to the TiO₂ lattice structure at the studied doping level. The fact that Bi³⁺ species were incorporated into the TiO₂ lattice without changing its phase does not necessarily rule out the possibility that bismuth may present as a separate phase onto the surface of TiO₂, as in the case of Bi₂O₃ as claimed by Wang et al. (2008). Similar results have been reported by Yang et al. in the work performed on TiO₂ photocatalysts prepared through coupling with Bi₂O₃ and doping with Si (Yang et al. 2014). However, in this study, the XRD peaks of the as-prepared photocatalysts corresponding to Bi₂O₃ were not found even when samples were calcinated at 700 °C. Therefore, the absence of this compound is not conclusive, since it is believed that presence of minor amounts of Bi₂O₃ cannot be detected by XRD (Ji et al. 2009).

SEM images of pure and Bi³⁺-doped as-prepared photocatalysts are depicted in Figure 3; the pure and bismuth-doped TiO₂ photocatalyst particles seem to be irregular in shape. These particles displayed a rock-like morphology, where the sizes of the particles of the doped samples seemed to be larger than the un-doped sample under the same magnification. This can be attributed to the presence of particle agglomeration. Similar agglomeration results were reported by Aware & Jadhav (2016) in the work performed on the doping of TiO₂ nanoparticles with zinc using the same method used in the current study, the sol-gel method.

EDS analysis was used to investigate the quantitative chemical composition for the constituent elements of pure and bismuth-doped TiO₂ as-prepared photocatalysts. Figure 4(a)–4(e) show the binding energy (BE) (keV) in the x-axis versus intensity of X-rays (counts) in the y-axis of the EDS analysis spectrum of as-prepared Bi³⁺-doped TiO₂ photocatalysts. Based on the EDS spectra, the existence of all corresponding elements, i.e. the peaks of Bi, Ti and O, respectively, present in the as-prepared photocatalysts were confirmed. A carbon peak is also present in the EDS spectra, which is due to the use of carbon tape during analysis. Moreover, the increased carbon content in the EDS spectra may also be attributed to possible contamination due to the hydrocarbon used in the preparation method. Hence, it is tempting to say that the presence of carbon does not dramatically affect as-prepared photocatalysts' performance, since it is present in all photocatalysts in almost similar amounts. No additional impurity peaks were detected, which confirms phase purity of the as-prepared photocatalysts, as further supported by XRD and XPS analysis.

The N₂ adsorption-desorption isotherms for the pure and Bi³⁺-doped TiO₂ as-prepared photocatalysts are presented in Figure 5, with the surface area, pore size and volume of the pores as presented in Table 1. All as-prepared photocatalysts were characterized by type IV isotherms, which indicates the presence of mesoporous material. The surface areas for the 1%, 3%, 5% and 10% Bi-doped photocatalysts were reported at 33.78, 38.27, 42.86 and 60 m²/g respectively, compared to the pure TiO₂ photocatalyst's surface area of 125.3 m²/g. According to Table 1, the change caused by Bi³⁺ doping is clear in the 10 mol % Bi-doped photocatalyst exhibiting the highest surface area (60 m²/g). Moreover, as Bi³⁺ dopant increased, the pore size became larger and the surface area was reduced to a large extent compared with pure TiO₂.

The elemental composition and valence state of the as-prepared photocatalysts were analyzed by XPS study. The XPS results are presented in Figure 6 and Table 2. The elements Ti, Bi and O in the as-prepared photocatalysts were confirmed in the wide spectrum. It also shows carbon; the high C 1 s content is due to the hydrocarbon contamination introduced from the laboratory environment as shown from EDS analysis. Thus, the observed C 1 s peak was used as an energy reference for determining the peak positions of core level spectra.

For more information about the valence state of titanium and bismuth ions – Ti 2p and Bi 4f – present in the as-prepared photocatalysts, high-resolution spectra were analyzed with respect to Bi³⁺ doping level as shown in Figures 7 and 8. Table 3 displays the fitted peak position corresponding to the BE values of Ti 2p spectra using the Lorentzian-Gaussian curve fitting method. For the pristine TiO₂, Ti 2p peak is distinguishable as two peaks at 462.32 eV for 2p_3/2 and 456.59 eV for 2p_1/2, which correspond to the characteristics peak of Ti⁴⁺ ion (Bapna et al. 2011). After doping bismuth into the TiO₂, the deconvolution of these peaks shifted towards lower binding energies of about 0.17 eV and peak width increased, indicating a small contribution of Ti³⁺ ion present in the sample. The shifts of these peaks are consistent with what was previously reported by Liu et al. (2012). These results may indicate an oxygen deficiency in TiO₂ lattice (Hamdy et al. 2012). It is conceivable that bismuth cations are substituted for Ti ion in the Ti_1-xBi_xO₂ matrix, because of the decreased Ti 2p peak area in comparison to the pure counterpart.

It is seen from Figure 8 that the Bi 4f peak is split into two peaks located at ∼159.06 eV and ∼164.41 eV corresponding to Bi 4f_5/2 and Bi 4f_7/2 peaks, as fitted by the Gaussian-Lorentzian product function after Shirley-type background subtraction. Both are mainly assigned to Bi-O bonds. Spin orbital splitting energy of the Bi 4f doublet is equal to ∼5.35 eV, which is a slight increase of about 0.15 eV with increase in the bismuth doping content x in the sample. Furthermore, each of the Bi 4f_5/2 and Bi 4f_7/2 peaks could be fitted by two sub-peaks. The sub-peaks located at 159.19 and 158.24 eV are assigned to Bi 4f_7/2-O, whereas 166.5 and 164.62 eV are of Bi 4f_5/2-O bonds for Bi³⁺ (Mekki et al. 1996). Table 4 displays the fitted Bi 4f XPS peak of the samples.

Figure 9 and Table 5 show the O 1 s XPS peak of the Ti_1-xBi_xO₂ sample. The O 1 s XPS spectra are well fitted into two peaks situated at 529.88 eV and 531.57 eV. The BE at 529.88 eV is attributed to the contribution of crystal lattice oxygen, whereas the BE at 531.57 eV is called a shoulder peak, which arises due to chemisorbed oxygen species on the surface of samples (Xu et al. 2008). Based on the above results, XPS observations confirm that no additional/impurity phase was formed, which further supports the EDS and XRD results.

The photocatalytic activity of the as-prepared pure and Bi³⁺-doped photocatalysts was evaluated in terms of their rate of acetaminophen removal from aqueous solution at pH = 5. The results indicated a direct relation between the percentages of Bi³⁺-doped TiO₂ and the removal rate of the pharmaceutical as presented in Figures 10–13. Upon comparing the photocatalytic removal of acetaminophen, 10 mol % Bi³⁺-doped photocatalyst's performance exceeded that of the pure as-prepared TiO₂. Figure 11 presents the graph of C/C_o vs. time by which the variation of the concentration of acetaminophen is observed for the different photocatalysts under UV-Vis light with irradiation time up to 4 hours. Although the surface areas of as-prepared Bi-doped photocatalysts are significantly lower than that of pure TiO₂, their enhanced photocatalytic activities might be due to their suitable band gaps absorption that occurs in the visible light region.

The photocatalytic removal of acetaminophen follows pseudo-first-order reaction kinetics for the concentration range of the present study (Dalida et al. 2014). As a result, the variation in -ln(C_o/C) as a function of irradiation time is reported in Figure 12, by which the rate constants were calculated in Figure 13 and listed in Table 1. The observed photocatalytic activity of Bi³⁺-doped as-prepared photocatalysts follows the order: 10 > 5 > 3 > 1> pure TiO₂, where enhanced removal of acetaminophen from solution under UV-Vis light irradiation is mainly ascribed to the Bi³⁺ doping. The doping with Bi³⁺ ion into the lattice of TiO₂ reduces the band gap energy. Therefore, the excitation energy is extended to the visible region. In our studied doping range, results from photocatalytic experiments are found to be in agreement with that of UV-Visible diffuse reflectance spectra analysis. Generally, the increase in mol % of Bi³⁺ doping led to a red shift increase, thereby resulting in higher photocatalytic activity. Moreover, as evident from the enhanced photocatalytic performance of Bi³⁺-doped photocatalysts in comparison to the un-doped photocatalysts, the presence of Bi³⁺ ions in the TiO₂ lattice significantly influences photoactivity by reducing charge carrier recombination rates during the irradiation process.

In conclusion, mesoporous Bi³⁺-doped anatase TiO₂ photocatalysts have been prepared using the PO assisted sol-gel method. The characterization results displayed suggest that the Bi³⁺ ion conceivably caused substitutional changes to the crystal lattice without altering its anatase phase. All doped as-prepared photocatalysts were found to have enhanced photocatalytic activities towards the removal of acetaminophen in aqueous solution under UV-Vis solar light irradiation compared to pure TiO_2. The enhanced photocatalytic activity indicated that Bi³⁺ doping into the TiO₂ enhanced the visible light absorption, as compared to the pure TiO₂. The rate of the photocatalytic reactions was found to depend mainly on the percent doping of Bi³⁺ ion, with the highest activity achieved for the highest doping percent of 10% added within the range of this study.

This research project was financially supported by the United Arab Emirates University ‘SURE Plus’ (grant no. 31S292, Ahmed Alzamly). The authors of this paper would like to acknowledge Mr Bassam Al-Hindawi for his assistance and guidance in the performance of BET analysis as well as Ms Afra Gharib Al-Blooshi for her cooperation with obtaining the SEM images.