Influence of the Bismuth Precursor on Phase Purity and Photocatalytic Activity of Hydrothermally Synthesised BiVO₄ for Dye and Antibiotic Degradation

2.1 Materials

Basic bismuth nitrate Bi₅O(OH)₉(NO₃)₄ (“bismutum subnitricum”, ≥98%, Merck 101878) and bismuth(III) nitrate pentahydrate Bi(NO₃)₃·5H₂O (reagent grade, crystals, ≥98%, Sigma-Aldrich 248592) were used as bismuth sources for samples BVO–1 and BVO–2, respectively. Ammonium metavanadate NH₄VO₃ (ACS reagent, ≥99.0%, Sigma-Aldrich), methylene blue, methyl orange, tetracycline hydrochloride, and ciprofloxacin (analytical grade) were used as received. Deionised water was used throughout.

2.2 Hydrothermal Synthesis and Calcination

The synthesis followed a recently reported hydrothermal protocol [14] with the bismuth source as the sole variable. In a typical preparation, 2.9 g of the bismuth precursor was dissolved in 50 mL of deionised water under continuous stirring. Separately, 0.70 g of NH₄VO₃ was dissolved in 30 mL of deionised water and heated to 60 °C until fully dissolved. The vanadium solution was added dropwise to the bismuth solution and the mixture was stirred for 30 min. The resulting yellow suspension was transferred into a Teflon-lined stainless-steel autoclave, sealed, and held at 150 °C for 24 h. After natural cooling to room temperature, the solid was recovered by vacuum filtration, rinsed thoroughly with deionised water, and dried in an oven at 80 °C for 2 h. The dried powder was subsequently calcined in static air at 600 °C for 2 h (heating rate 5 °C/min). The two products are designated BVO–1 (Bi precursor Bi₅O(OH)₉(NO₃)₄) and BVO–2 (Bi precursor Bi(NO₃)₃·5H₂O). Because the two bismuth sources differ in molar mass, the fixed precursor mass of 2.9 g corresponds to different Bi/V molar ratios. The 0.70 g of NH₄VO₃ provides 5.98 mmol of vanadium in both preparations, whereas the bismuth charged is 5.98 mmol for Bi(NO₃)₃·5H₂O (M=485.07 g/mol) and 9.92 mmol for Bi₅O(OH)₉(NO₃)₄ (M=1461.98 g/mol). The Bi/V molar ratio is therefore 1.00 for BVO–2 and 1.66 for BVO–1, so the BVO–1 preparation carries a bismuth excess of approximately 66%.

2.3 Characterisation

The crystalline structure and phase composition were analysed by X-ray diffraction (XRD) on a Rigaku SmartLab SE Basic diffractometer with Cu Kα radiation (λ = 1.5406 Å), scanned over 5–80° 2θ for BVO–1 and 10–80° 2θ for BVO–2, at a step of 0.01°; each pattern was refined over its full measured range, and the two are compared from 10° 2θ, the window common to both measurements. Rietveld refinement was carried out in BGMN v4.2.22 through the Profex 5 interface [15,16] against the ICDD structure models of clinobisvanite ms-BiVO₄ (PDF 00–014–0688), Bi₅O₇NO₃ (PDF 00–051–0525), and β-Bi₂O₃ (PDF 00–027–0050, Harwig 1977). The morphology was examined by field-emission scanning electron microscopy (FESEM) on a JEOL JSM-IT700HR operated at 10 kV, and the average particle size was determined by image analysis (ImageJ; n≥120 particles per sample). Energy-dispersive X-ray spectroscopy (EDS) maps and standardless ZAF quantification were acquired on the same instrument at 10 kV (live time 30 s) with oxygen quantified directly; the limitations of these conditions are set out in Section 3. X-ray photoelectron spectroscopy (XPS, Kratos Axis Supra, monochromatic Al Kα) was used to determine the chemical states of Bi, V, and O in BVO–2; binding energies were referenced to C 1s at 284.8 eV, and the spectra were fitted with pseudo-Voigt line shapes on a Shirley background. Diffuse reflectance spectra (Shimadzu UV-2600i Plus, integrating sphere, 190–800 nm) were converted to the Kubelka–Munk function F(R), and the optical band gap was obtained from the Tauc plot of (F(R)hν)² against photon energy for a direct allowed transition.

2.4 Photocatalytic Activity Tests

Photocatalytic experiments were performed for four pollutants spanning two chemical classes, the dyes methylene blue (MB, cationic; initial concentration 5 ppm; monitored at 663 nm) and methyl orange (MO, anionic; 5 ppm; 464 nm), and the antibiotics tetracycline hydrochloride (TC; 10 ppm; 357 nm) and ciprofloxacin (CIP; 10 ppm; 271 nm). In each run, 50 mg of catalyst was dispersed in 100 mL of pollutant solution (catalyst loading 0.5 g/L). The suspension was stirred in the dark for 60 min to establish adsorption-desorption equilibrium, with aliquots withdrawn at t=−60 min, −30 min, and 0 min. The illumination phase was initiated at t=0 min using four 10 W Philips TUV germicidal lamps emitting at 253.7 nm (UV-C) [17]. This source was selected so that both samples are excited above the absorption edge of every phase present, which allows the two catalysts to be compared under identical illumination. No measurements were performed under visible light, so the activities reported here describe behaviour under UV-C only. Aliquots (3 mL) were collected at 30 min intervals up to 120 min of illumination, filtered, and analysed by UV-Vis spectrophotometry (Shimadzu UV-1900) at the pollutant-specific monitoring wavelength. The conversions reported below therefore describe the loss of the monitored absorbance band of the parent pollutant. The apparent rate constant k was extracted from the Langmuir–Hinshelwood pseudo-first-order kinetic model (Equation (1)):

3.1 Phase Composition and Morphology

Two BiVO₄ samples were prepared under identical hydrothermal and calcination conditions from the two bismuth precursors and are designated BVO–1 (Bi₅O(OH)₉(NO₃)₄) and BVO–2 (Bi(NO₃)₃·5H₂O). Their morphology, phase composition, surface chemical states and optical absorption were examined first, prior to the evaluation of their photocatalytic performance.

The FESEM micrographs in Figure 1(a,b) show that the morphology differs between the two products. BVO–1 (Figure 1a), prepared from Bi₅O(OH)₉(NO₃)₄, appears as a heterogeneous aggregate of small particles (≈100–400 nm) in which at least three distinguishable morphologies coexist: faceted blocky crystallites 200–400 nm in size, small spherical features 30–50 nm, and elongated rod- or plate-like fragments. How these habits relate to the phases identified below is considered after the diffraction results. BVO–2 (Figure 1b), prepared from Bi(NO₃)₃·5H₂O, is, by contrast, uniform: it consists of faceted polyhedral crystallites ≈1–2 μm in size with little aggregation. ImageJ analysis of the 20 000× frames yields a mean equivalent diameter of 200 nm for BVO–1 and 1.2 μm for BVO–2, about six times smaller for the multi-phase product, a size difference that is relevant to the interpretation of the photocatalytic data discussed below.

Elemental analysis by EDS detected only bismuth, vanadium and oxygen in both samples. The elemental maps acquired for BVO–1 (Figure 1c) show bismuth, vanadium and oxygen distributed uniformly across the imaged field, with no regions of segregation resolvable at this magnification, and the atomic percentages obtained for the two samples are compared in Figure 1d. These percentages are semi-quantitative and are reported for the elements present and their distribution rather than for composition. At a landing voltage of 10 kV the overvoltage of the vanadium K line is only 1.83, bismuth is quantified on its M lines, and the vanadium Lα and oxygen Kα lines are separated by 14 eV, well below the resolution of the detector in that range. The composition of the two samples is therefore taken from the charged stoichiometry and from the diffraction analysis that follows.

The phase composition of the two samples was examined by X-ray diffraction, and the patterns are compared in Figure 2a. The XRD pattern of BVO–2 exhibits sharp, well-defined reflections that can be indexed in full to the monoclinic-scheelite phase of BiVO₄ (clinobisvanite, PDF 00–014–0688), with no additional reflections, indicating the formation of phase-pure BiVO₄. In contrast, the pattern of BVO–1 contains, in addition to the clinobisvanite reflections, extra intensity near 2θ≈28° that cannot be accommodated by ms-BiVO₄ alone [18], indicating the coexistence of additional crystalline phases.

To quantify the phase composition, the XRD data were analysed by Rietveld refinement; the refined patterns are shown in Figure 2b,c, and the corresponding lattice parameters, phase fractions, and reliability factors are summarised in Table 1. For BVO–2, the single-phase refinement (Figure 2c) shows a good agreement between the observed (Y_obs) and calculated (Y_cal) profiles, converging to R_wp=12.44 lattice parameters a=5.1999 Å, b=5.0987 Å, c=11.7078 Å, and γ=90.375^°, indicating phase-pure clinobisvanite within the detection limit of the method. For BVO–1, a two-phase model (ms-BiVO₄ + Bi₅O₇NO₃) reduces R_wp to ≈18 (≈28-σ) residual peak at 2θ=28.4°. Candidate third phases, namely pucherite (o-BiVO₄), α-Bi₂O₃, and Bi₆O₈(OH)₄(NO₃)₆, were each tested and refined to zero weight fraction. The inclusion of β-Bi₂O₃ (Harwig 1977) accounts for the residual peak and lowers R_wp to 14.15% and χ² to 4.75 (Figure 2b). The refined composition of BVO–1 is therefore approximately 25% ms-BiVO₄, 59% Bi₅O₇NO₃, and 16–19% β-Bi₂O₃ (Table 1). A parallel refinement with the β-Bi₂O₃ lattice fixed to literature values reproduces a comparable weight fraction (19.2 %, R_wp=15.49 supporting the phase assignment independently of any lattice flexibility. The spread between these two refinement strategies, from 15.9% to 19.2%, is an order of magnitude larger than the estimated standard deviations returned by the refinement. Those deviations describe the statistical precision of the fit alone and exclude the systematic contribution of the choice of structural model, so they are treated here as lower bounds on the uncertainty and the phase fractions are quoted accordingly. The phase assignment itself is unaffected, since both strategies require β-Bi₂O₃ to account for the residual intensity at 2θ=28.4°. Overall, the refinement results indicate that Bi(NO₃)₃·5H₂O yields single-phase clinobisvanite within the detection limit of the method, whereas Bi₅O(OH)₉(NO₃)₄ yields a three-phase product in which BiVO₄ accounts for only about a quarter of the mass.

The combined morphological and crystallographic evidence is consistent with a two-stage decomposition pathway for BVO–1. In the hydrothermal stage, Bi₅O(OH)₉(NO₃)₄ only partially hydrolyses, dehydrating to the oxynitrate framework Bi₅O₇NO₃ according to Bi₅O(OH)₉(NO₃)₄ → Bi₅O₇NO₃ + 2 H₂O + 3 HNO₃; in this scheme only about one-fifth of the bismuth reacts with VO₃⁻, which would account for the observed ≈25 wt% ms-BiVO₄. During subsequent calcination at 600 °C, a portion of the residual Bi₅O₇NO₃ undergoes further thermal decomposition. The chosen temperature lies within the metastability window of β-Bi₂O₃ (≈330–650 °C), which favours formation of the β polymorph over the thermodynamically stable α-Bi₂O₃ and is consistent with the Rietveld-refined α:0 wt%, β: 16 to 19 wt% partition. In contrast, Bi(NO₃)₃·5H₂O is expected to release Bi³⁺ directly under hydrothermal conditions, which would enable complete reaction with VO₃⁻, consistent with the phase-pure ms-BiVO₄ obtained. The three coexisting morphologies observed in Figure 1a may be associated with the three quantitatively refined phases: blocky crystallites with ms-BiVO₄, small spherical particles with β-Bi₂O₃, and elongated plates or rods with layered Bi₅O₇NO₃. This association rests on particle size and habit rather than on a compositional measurement of individual particles.

To probe the surface chemical states of the phase-pure sample, XPS analysis was carried out on BVO–2 (Figure 2d–g). The survey spectrum shows the presence of Bi, V, and O, together with adventitious carbon serving as the reference for calibration (Figure 2d). The high-resolution Bi 4f spectrum reveals binding energy peaks at ≈158.7 eV (Bi 4f_7/2) and ≈164.0 eV (Bi 4f_5/2), with a spin-orbit splitting of ≈5.3 eV, characteristic of Bi³⁺ in the BiVO₄ lattice [19,20] (Figure 2e). Similarly, the V 2p spectrum shows peaks at ≈516.4 eV (V 2p_3/2) and ≈523.7 eV (V 2p_1/2), consistent with V⁵⁺ in the vanadate framework, with no evidence of lower oxidation states such as V⁴⁺ [21,22] (Figure 2f). The O 1s spectrum resolves into two components (Figure 2g): a main peak at ≈529.5 eV, assigned to lattice oxygen in Bi-O and V-O bonds (O_L), and a higher-binding-energy component at ≈531.4 eV, corresponding to chemisorbed oxygen or surface hydroxyl groups (O_C), which together carry approximately 70% and 30% of the O 1s intensity, respectively [22]. The latter component is not assigned to oxygen vacancies, since the C 1s region of the same surface contains C-O and O-C=O contributions and the present data do not separate the two possible origins at this binding energy. Overall, the XPS results indicate that bismuth and vanadium are present as Bi³⁺ and V⁵⁺, consistent with the phase-pure clinobisvanite identified above by Rietveld refinement.

The optical absorption of the two samples was compared by diffuse reflectance spectroscopy (Figure 3a). Both samples absorb strongly below about 500 nm and reflect above it, with the reflectance plateau of BVO–1 (about 55%) well below that of BVO–2 (about 85%). The Tauc plots for a direct allowed transition (Figure 3b,c) give optical band gaps of 2.55 eV for BVO–1 and 2.50 eV for BVO–2. The value for BVO–2 lies within the 2.4–2.5 eV range reported for monoclinic BiVO₄ [7,19]. The two edges are close, and the difference of 0.05 eV is comparable to the uncertainty of the extrapolation, so it is not interpreted here beyond noting that its direction is the one expected from the wider-gap Bi₅O₇NO₃ and β-Bi₂O₃ phases present in BVO–1. Both edges lie far below the 4.88 eV photon energy of the UV-C source used in the photocatalytic tests, so photon harvesting is not expected to limit either sample. Overall, the two precursors give products that differ in phase composition and primary particle size but not appreciably in optical absorption edge, and these differences are expected to influence their photocatalytic behaviour.

3.2 Photocatalytic Activity

The photocatalytic performance of BVO–1 and BVO–2 was benchmarked against four pollutants representing two dye classes (Figure 4: top row methylene blue, MB; bottom row methyl orange, MO) and two antibiotic classes (Figure 5: top row ciprofloxacin, CIP; bottom row tetracycline, TC). These pollutants were selected as model contaminants because of their widespread occurrence in textile and pharmaceutical wastewater and their persistence in aquatic environments. Each test comprised 60 min of stirring in the dark to establish adsorption-desorption equilibrium, followed by 120 min of UV-C irradiation. For each pollutant, panels a, b, e and f show the time-evolved UV-Vis spectra over BVO–1 and BVO–2, panels c and g the normalised C/C₀ profile across the dark-adsorption and illumination stages, and panels d and h the pseudo-first-order linearisation ln(C₀/C) versus t (Equation (1)). The apparent rate constants k and the corresponding linear-fit R² are shown directly on each kinetic panel.

A pollutant-dependent reversal in activity is observed. Methylene blue, the only cationic pollutant tested, is degraded more efficiently by the three-phase BVO–1 (57.8% photodegradation after 120 min of illumination, k_MB=7.6 × 10⁻³ min⁻¹, R²=0.9797) than by phase-pure BVO–2 (37.8%, k=3.9 × 10⁻³ min⁻¹, R²=0.9954; Figure 4d) [23,24]. BVO–1 shows a larger dark adsorption of MB than BVO–2 (15.0% against 6.4%, visible as the deeper C/C₀ drop at t=0 min in Figure 4c) that is carried over into the illuminated stage. This behaviour is consistent with an electrostatic interaction between cationic MB and the residual oxynitrate and oxide phases (Bi₅O₇NO₃ and β-Bi₂O₃) that line the BVO–1 surface [25,26]. Methyl orange, which is anionic, is poorly removed by either sample (7.1% and 6.4% photodegradation, k_MO=1.6 × 10⁻³ and 0.6 × 10⁻³ min⁻¹ for BVO–1 and BVO–2, respectively; Figure 4h). The limited residual activity would likewise be consistent with electrostatic repulsion of MO⁻ from a negatively charged photocatalyst surface, and the contrast between the two dyes indicates that surface chemistry, in addition to band structure, governs the accessible substrate scope.

This behaviour is reversed for the two antibiotics (Figure 5). For ciprofloxacin, both catalysts perform similarly, with the higher rate constant for BVO–2 (45.0% against 34.7% photodegradation; k_CIP=4.8 × 10⁻³ min⁻¹ against 3.5 × 10⁻³ min⁻¹ for BVO–1; Figure 5d) [22,27]. For tetracycline, the difference is larger: BVO–2 reaches 82.1% photodegradation and k_TC=15.2 × 10⁻³ min⁻¹, roughly 1.8× faster than BVO–1 (56.2%, 8.3 × 10⁻³ min⁻¹; Figure 5h) [20,28]. The corresponding C/C₀ profile (Figure 5g) drops to below 0.2 within 120 min for BVO–2 while BVO–1 plateaus near 0.4. Moreover, the larger dark-adsorption capacity of BVO–1 for tetracycline (46.7% against 8.7%) does not translate into higher overall removal under illumination (76.7% against 83.6% total removal relative to the initial concentration), indicating that adsorption is not the rate-limiting step for antibiotic degradation; the origin of the difference is considered below.

The germicidal UV-C illumination at 253.7 nm (4.88 eV) is well above the measured absorption edges of the two samples (Figure 3) and the band gaps of all three phases present in the samples (BiVO₄ ≈2.4 eV; β-Bi₂O₃ ≈2.6 eV; Bi₅O₇NO₃ ≈3.0 eV), so all three phases can, in principle, be photoactivated under these conditions. The higher activity of BVO–2 for antibiotic degradation is therefore unlikely to originate in photon harvesting, and may instead be attributed to the fate of the photogenerated carriers. The phase-pure, micron-sized ms-BiVO₄ grains of BVO–2 are expected to present a low density of grain boundaries; their faceted habit is consistent with the {010} and {110} terminations that are known to segregate photogenerated electrons and holes onto distinct surfaces in monoclinic BiVO₄ [19,21]. BVO–1, by contrast, contains a large density of internal interfaces between dissimilar phases that may act as recombination sites rather than as effective Type-II heterojunctions, because the phase boundaries between BiVO₄, Bi₅O₇NO₃, and β-Bi₂O₃ are neither lattice-matched nor crystallographically aligned. On this interpretation, the more efficient use of the absorbed photons by BVO–2 would compensate for its larger primary particle size when the target is bulk-oxidation-limited, as in antibiotic degradation, which requires sustained •OH attack on multiple aromatic rings [29]. However, no measurement of carrier dynamics was performed in this work. Photoluminescence or electrochemical impedance spectroscopy would be required to establish this mechanism directly, and such measurements are the subject of ongoing work.

Conversely, the MB data are consistent with removal governed largely by adsorption at the catalyst surface, with the higher surface coverage of BVO–1 dominating the overall removal rate. The apparent rate constants reported here are not normalised to specific surface area. Since the primary particle size of BVO–1 is smaller than that of BVO–2 by a factor of about six, a larger specific surface area is expected for the three-phase product, and its contribution to the higher methylene blue removal cannot be separated from that of the phase composition and the associated adsorption behaviour. Overall, the two samples show complementary behaviour: BVO–1 is the more effective for the cationic dye, where adsorption contributes strongly, whereas BVO–2 is the more active for both antibiotics, so the precursor sets not only the phase composition but also the substrate scope of the product.