The Key Role of Carbon Materials in the Biological and Photocatalytic Reduction of Nitrates for the Sustainable Management of Wastewaters

Abstract

This work explores the influence of material properties and experimental conditions on both biological and photocatalytic nitrate reduction processes. For the biological route, results demonstrate that carbon supports, specifically carbon gels, with open porosity, slight acidity, and high purity enhance E. coli adhesion and promote the formation of highly active bacterial colonies. However, carbon supports of bacteria, produced from waste biomass, emerge as a sustainable and cost-effective alternative, improving scalability and environmental value. The complete conversion of nitrates to nitrites, followed by full nitrite reduction, is achieved under optimized conditions. Photocatalytic nitrate reduction under solar radiation is also proposed as a promising and ecofriendly upgrade method to conventional wastewater treatment. Graphene oxide (GO) was used to enhance the photocatalytic activity of TiO₂ nanoparticles for the degradation of nitrates. The efficiency of nitrate reduction is found to be highly sensitive to solution pH and the physicochemical nature of the photocatalyst surface, which governs nitrate interactions through electrostatic forces. TiO₂–GO composites achieved up to 80% nitrate removal within 1 h and complete removal of 50 mg/L nitrate within 15 min under optimized conditions. The screening of hole scavengers revealed that formic acid, in combination with the TiO₂–GO composite, delivered exceptional performance, achieving complete nitrate reduction in just 15 min under batch conditions at an acidic pH.

1. Introduction

Nitrate (NO₃⁻) contamination of water resources is a growing environmental and public health concern due to its high solubility, persistence, and well-known role in eutrophication and health risks [1]. In the United States, the Environmental Protection Agency (EPA) sets the maximum allowable nitrate concentration at 10 mg/L. European legislation is comparatively less stringent; Directive (EU) 2020/2184 sets the limit at 50 mg/L, which aligns with the World Health Organization’s guideline [2]. Nevertheless, even this higher threshold is frequently exceeded in regions classified as nitrate-vulnerable zones, where groundwater contamination often forces residents to rely on bottled water instead of local supplies.

Sludge carried by irrigation water is difficult to collect, and it is typically discharged without treatment, except in certain hydroponic greenhouse systems. In contrast, wastewater from domestic, industrial, and often livestock sources is generally collected through sewerage systems. Domestic wastewater and farm slurry from livestock tend to exhibit similar nitrate (NO₃⁻) concentrations upon entering wastewater treatment plants (WWTPs) [3]. Treatment is usually carried out through biological processes designed to degrade both organic matter and nitrates. However, numerous studies have shown that the efficiency of biological denitrification is strongly dependent on the carbon-to-nitrogen (C/N) ratio. Optimal performance is achieved at C/N ratios close to five, whereas efficiency significantly decreases as the nitrogen concentration increases, with substantial limitations observed at a C/N ratio of 2.7 [4].

To improve the denitrification efficiency, external organic carbon sources can be added, but this increases both operational costs and the volume of residual waste. Furthermore, the conventional activated sludge process (ASP) is known to be energy inefficient [4,5]. To address these challenges, microbial communities can be immobilized on various types of support materials to form active biofilms. The effectiveness of these biofilms depends on the composition and morphology of the microbial communities, which, in turn, are subordinate to the properties of the support materials used [6]. These biofilms are often integrated into alternative reactor designs, such as rotating drum biological contactors (RDBCs), to enhance treatment performance [7].

Alternative denitrification methods involve physical, chemical, and catalytic approaches [8]. Physical processes include techniques such as reverse osmosis, ion exchange, electrodialysis, and electroreduction. Each of these methods has specific advantages and limitations that must be evaluated individually. Common challenges associated with physical treatments include high energy consumption, membrane fouling, and the need for frequent maintenance.

Different catalytic approaches have also been explored for the reduction of nitrates in aqueous solutions. Cai et al. [9] investigated the electrochemical reduction of nitrate to nitrogen (N₂) using a two-electrode system with Pd–Cu supported on mesoporous carbon as the cathode catalyst. This method enables the effective reduction in highly concentrated nitrate solutions (up to 650 mg L⁻¹), avoiding the limitations associated with biological treatments.

Photocatalysis has emerged as a sustainable solution, especially under solar light, but its efficiency depends strongly on catalyst design and reaction conditions. The photodegradation of nitrates was first reported by Kudo et al. in 1987 using Pd–TiO₂ as a catalyst [10]. In general terms, nitrate reduction can be described as consecutive reactions where nitrates are reduced to nitrites, which can be finally converted to nitrogen or ammonium. In order to improve the rate of photoreduction reactions, hole scavengers are often added to the reactant mixture to limit the formation of oxidant species. The nitrate reduction is sensitive to pH, hence an effective pH regulation is necessary during the reaction, since basic conditions significantly promote the formation of ammonium ions [11,12]. In acid, the media reaction develops following the progressive reduction:

$${\mathrm{N}\mathrm{O}}_3^{-}\left(\mathrm{a}\mathrm{q}\right)+{2\mathrm{e}}^{-}+{2\mathrm{H}}^{+}\left(\mathrm{a}\mathrm{q}\right)\to {\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)$$

(1)

$$2{\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{6\mathrm{e}}^{-}+{8\mathrm{H}}^{+}\left(\mathrm{a}\mathrm{q}\right)\to {\mathrm{N}}_2\left(\mathrm{g}\right)+4{\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)$$

(2)

$${\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{6\mathrm{e}}^{-}+{8\mathrm{H}}^{+}\left(\mathrm{a}\mathrm{q}\right)\to {\mathrm{N}\mathrm{H}}_4^{+}\left(\mathrm{a}\mathrm{q}\right)+2{\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)$$

(3)

While in basic media,

$${\mathrm{N}\mathrm{O}}_3^{-}\left(\mathrm{a}\mathrm{q}\right)+{2\mathrm{e}}^{-}+{\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\to {\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{2\mathrm{O}\mathrm{H}}^{-}\left(\mathrm{a}\mathrm{q}\right)$$

(4)

$$2{\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{6\mathrm{e}}^{-}+{4\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\to {\mathrm{N}}_2\left(\mathrm{g}\right)+{8\mathrm{O}\mathrm{H}}^{-}\left(\mathrm{a}\mathrm{q}\right)$$

(5)

$${\mathrm{N}\mathrm{O}}_2^{-}\left(\mathrm{a}\mathrm{q}\right)+{6\mathrm{e}}^{-}+{6\mathrm{H}}_2\mathrm{O}\left(\mathrm{l}\right)\to {\mathrm{N}\mathrm{H}}_4^{+}\left(\mathrm{a}\mathrm{q}\right)+{8\mathrm{O}\mathrm{H}}^{-}\left(\mathrm{a}\mathrm{q}\right)$$

(6)

Photocatalytic reduction also offers additional advantages, notably the absence of reagent consumption [8,13], thus avoiding the careful maintenance of the experimental conditions in Fe-based reductions. The preparation method of the photocatalyst (TiO₂) strongly determines its performance [14]. Several parameters should be controlled. The anatase phase is more active than the rutile phase in nitrate degradation, achieving high conversions with 100% selectivity to N₂ [15]. The synthesis of nano-sized TiO₂ and doping with Pd have also been proposed as effective tools to enhance the degradation of nitrate species under UV (365 nm) irradiation [12,13]. From a sustainability perspective, the reaction becomes significantly more attractive when solar light is used as the energy source, eliminating the need for external energy input. In this context, composites of TiO₂ with carbon-based materials, such as graphene oxide (GO), offer several benefits over noble metal dopants [16]. These include reduced material costs, lower corrosion susceptibility, and improved charge separation due to the donor–acceptor properties and light sensitization effects of the carbon phase.

Recent advances in nitrate removal have increasingly focused on both photocatalytic and biological strategies. Several studies have explored photocatalytic nitrate reduction using modified TiO₂ systems. For instance, a Z-scheme g-C₃N₄/Pd–Cu/rGO/TiO₂ hybrid catalyst achieved nearly 58% NO₃⁻ removal under visible light, demonstrating enhanced electron transfer compared to monometallic catalysts [17]. Similarly, Co–TiO₂/GO composites developed for electrochemical nitrate reduction delivered NH₃ yields of 7.4 mg h⁻¹ cm⁻² and high Faradaic efficiency (56.5%) while maintaining excellent stability [18]. Another study combined TiO₂@Fe₃O₄ with chitosan, offering a hybrid approach integrating adsorption and photocatalysis for efficient nitrate removal [19]. Moreover, Ag–TiO₂ systems coupled with formic acid as a hole scavenger achieved selective N₂ production under visible light, providing valuable insights into the mechanistic pathways of photocatalytic nitrate reduction [20].

Regarding biological treatment, recent work has examined biofilm and electro-biofilm systems for nitrate removal under challenging conditions. A moving bed biofilm reactor (MBBR) using activated carbon carriers and acetate as a carbon source achieved efficient nitrate reduction in synthetic groundwater [21]. Likewise, a microbial electrolysis cell integrated with a constructed wetland and an Fe₃O₄/GAC anode demonstrated improved performance even at low C/N ratios, highlighting the potential of electro-biotic hybrid systems [22].

The novelty of the present work lies in its dual-strategy approach for sustainable nitrate removal, combining (i) low-cost, activated carbons (ACs) from biomass to enhance biofilm-based biological denitrification and (ii) TiO₂–GO nanocomposites with enhanced photocatalytic reactivity under solar light, addressing energy efficiency and enabling complete nitrate conversion to N₂. By optimizing operational parameters such as pH and the use of scavengers, this work demonstrates how biological and photocatalytic processes can achieve complete nitrate conversion to nitrogen gas, offering a cost-effective, scalable, and environmentally friendly solution for nitrate pollution.

2. Results and Discussion

2.1. Biological Treatment

2.1.1. Materials Characterization

Three different carbon materials (carbon xerogel, CA; bio-ACs, A; and commercial ACs, N) were selected to develop E. coli biofilms for nitrate decomposition. Samples were selected by trying to cover a wide range of properties in order to obtain correlations between the performance of biofilms and textural and chemical properties of substrates.

Almond shells were selected as the precursor for one of the ACs used as bacterial support. Almond shells are an abundant and inexpensive biomass waste, particularly in Mediterranean countries. They typically consist of 38.48% cellulose, 28.82% hemicellulose, and 29.54% lignin, and after carbonization and activation, they present a low content of inorganic matter.

In contrast, carbon xerogels (CA sample), prepared from pure reactants (RF), are consequently more expensive materials, but they are free of inorganic components, and their porosity can be designed to favor macro-, meso-, or microporosity by fitting different parameters of synthesis during polymerization, drying, and carbonization processes [23].

The carbonization processes of both precursors were simulated using a thermobalance, in which a small amount of each precursor was heated under a N₂ flow at a rate of 10 °C/min up to 800 °C (slow pyrolysis). Figure 1 shows the TG-DTG profiles obtained for the carbonization of the lignocellulosic material (almond shells, denoted as AS) and the raw RF polymer (xerogel). The decomposition of commercial pure microcrystalline cellulose (denoted as MCC) was also included as a reference.

As observed, MCC decomposes in a single step, with only one peak appearing in the DTG profile at around 360 °C. The TG profile of AS initially shows a weight loss due to moisture desorption, followed by the decomposition of hemicellulose. The TG curve begins to decrease at lower temperatures, but with a moderate slope compared to MCC, indicating interactions between the different biomass components. The maximum decomposition rate occurs at approximately 375 °C (DTG profile), confirming that cellulose is the main component. Cellulose decomposition is complete at around 400 °C. After this point, the slope of the TG profile decreased again, associated with the greater stability of the lignin phase. The average yield of biochar obtained at 800 °C from AS in various batch furnace experiments is around 18%. In this case, the sample from almond shells tested as bacteria support was thereafter activated, undergoing an additional weight loss of 28%.

The TG-DTG profiles obtained during the carbonization of RF polymers are evidently very different than those discussed for biomass pyrolysis. The decomposition of the chemical structure of the polymers is associated with two main processes, the breakage of C-O bonds and C-C at around 350 and 550 °C, respectively [23,24], carbonization being complete at around 700 °C. Desorption at a low temperature (below 200 °C) is mainly associated with the removal of solvents. At a glance, the first conclusion observed is that the proportion of the residue increases. The carbonization of RF polymers yields around 50% wt. of char at 800 °C, regarding 18% obtained from biomass residues. This fact can partially compensate for the cost of production of the different materials.

The physicochemical properties of the samples tested are summarized in

Table 1. Textural properties and pH_PZC of the materials (CA—carbon aerogel; A—AC from almond shells; N—commercial AC Norit).

Support	Ash (wt%)	pHPZC	SBET (m2 g−1)	VT (cm3 g−1)	W0 (cm3 g−1)	L0 (nm)	Vmeso (cm3 g−1)
CA	null	6.3	594	0.69	0.21	2.5	0.48
A	0.3	10.6	913	0.47	0.32	1.4	0.15
N	4.8	11.0	1233	0.60	0.56	1.5	0.04

S_BET = BET surface area, V_T = total pore volume, W₀ = micropore volume, L₀ = mean micropore width, and V_meso = V_T-W₀ mesopore volume, obtained by difference according to the Gurvich rule.

. As expected, the CA sample is a carbon material with negligible ash content. Among the ACs, the A sample (prepared almond shells) shows a relatively low ash content compared to the commercial N sample. The CA sample (carbon xerogel), produced at a lower temperature and without activation, exhibits a lower pH_PZC, indicating weakly acidic behavior. In contrast, both ACs display a markedly basic character.

From a textural point of view, the most notable differences are also observed between CA and the ACs, as clearly shown by the shapes of their N₂ adsorption isotherms (Figure 2). Both AC samples exhibit Type I isotherms, typical of predominantly microporous materials. In contrast, the CA sample displays a Type IV isotherm with a pronounced hysteresis loop, indicating a mesoporous structure. Consequently, the ACs exhibit well-developed microporosity (W₀) and higher surface area values compared to CA (Table 1). However, the total pore volume (measured by the volume of liquid nitrogen adsorbed at P/P₀ = 0.9, Figure 2) and the mesopore volume are both higher in the carbon aerogel sample.

2.1.2. Performance of Adhered Biofilms

The performance of E. coli biofilms in nitrate removal, when supported on the various carbon materials, is presented in Figure 3. Control experiments conducted in the absence of microorganisms showed that nitrate adsorption or degradation by the carbon supports alone was negligible. Regardless of the carbon material used, the nitrate concentration remained stable. Therefore, the decrease in nitrate concentration observed after immobilizing the biofilms on the supports can be attributed solely to the bacterial activity involved in the nitrate reduction process.

It is well-known [25] that the reduction of NO₃⁻ by bacteria occurs following a four-step process: microorganisms reduce nitrates NO₃⁻ progressively to nitrites NO₂⁻, nitric oxide NO, nitrous oxide N₂O, and finally to nitrogen gas N₂. In this study, a gas-phase analysis was not performed; instead, only the evolution of species in a solution (nitrate and nitrite) was monitored.

Figure 3a shows the kinetics of NO₃⁻ reduction by E. coli biofilms supported on pure carbon aerogels (CA) compared with those supported on activated carbons (N and A). The subsequent degradation of nitrite intermediates by the different biofilms is presented in Figure 3b. It is noteworthy that both nitrates and nitrites were completely reduced and eliminated from the aqueous solution by the E. coli biofilms, regardless of the nature of the carbon support. However, the biofilm developed on CA exhibited a higher activity than those formed on the ACs, resulting in a faster reduction in both nitrate and the subsequent nitrite. The concentration of NO₂⁻ increased as NO₃⁻ decreased, indicating a sequential reduction pathway. Because of the imposed anaerobic conditions, bacteria exhibit a higher affinity for NO₃⁻ than for NO₂⁻, such that the concentration of NO₂⁻ does not begin to decrease until NO₃⁻ is almost completely depleted. Once NO₃⁻ is consumed, bacteria proceed to reduce NO₂⁻ completely. Initially, the concentration of NO₂⁻ is zero, but it increases as the reaction progresses. This behavior suggests a mechanism involving the adsorption of NO₃⁻ onto the biofilm surface, followed by the desorption of NO₂⁻ while nitrate remains readily available in the medium. This confirms the role of NO₂⁻ as a reaction intermediate. It also implies that the adsorption of NO₃⁻ is faster and energetically more favorable than its immediate reduction. If reduction occurred directly on the adsorbed NO₃⁻ without desorption, no accumulation of NO₂⁻ in the solution would be observed.

Interestingly, the curves in Figure 3b are nearly symmetrical, indicating that the rates of NO₂⁻ formation and subsequent reduction are relatively similar. The lower affinity of the biofilms for NO₂⁻ also accentuates the differences in biofilm activity: broader curves correspond to slower nitrite formation from nitrate and a slower subsequent reduction to N₂.

The morphology of the bacteria colonies in supported biofilms was analyzed by SEM (Figure 4). On the CA sample, bacteria tend to cluster together, while in the case of the activated carbons (ACs), they appear more dispersed. In both the CA and A samples, bacteria are connected to each other and to the carbon surface through excreted fibers, which are less frequently observed in the N sample. These filaments are biosynthesized polysaccharides and may play a role in electron transfer processes [26], potentially influencing the redox activity of the bacteria.

Since the incubation of E. coli on the various supports was conducted in buffered solutions at pH = 7, the AC surfaces would be positively charged. This condition is generally favorable for interaction with E. coli, a Gram-negative bacterium [27]. Nevertheless, more structured colonies were observed in CA.

Although all biofilms were capable of completely removing the initial nitrate load, the differences in performance are likely related to the distinct characteristics of the supports that influence biofilm architecture. When correlating the physicochemical properties of the supports with colony morphology and denitrification performance, it becomes evident that the best results—specifically, the fastest NO₃⁻ reduction kinetics—are achieved with biofilms supported on the mesoporous, pure carbon CA sample, in comparison with those supported on ACs. The porous structure and surface chemistry of the support materials influence bacterial interaction and biofilm formation. The open, coral-like mesoporous structure of the CA sample provides surface irregularities and roughness that facilitate bacterial anchoring. Given that nitrate adsorption onto the supports is negligible and that micropores are inaccessible to bacteria due to their size, the highly microporous surfaces of the ACs seem to have a limited influence on biofilm performance. As shown in previous works [28], some oxygenated functionalities (-OH, -COOH, -CHO, or -C=O) can form hydrogen bonds with bacterial cell walls, promoting the attachment to the carrier surface. The lower carbonization temperature used for the CA sample likely preserves a higher concentration of such oxygenated surface groups (OSG). Comparing the performance of ACs, both present a marked basic character, but sample A presents a certain mesopore volume and lower ash content than sample N. This led to an intermediate character between CA and N samples, thus showing poorly structured colonies like on N, but bacteria are attached to the carbon surface by filaments like on CA. In a previous work [29], we demonstrated that doping carbon gels with different metal oxides (SiO₂, TiO₂, and Al₂O₃) progressively limits the activity of E. coli biofilms as the acidic character of the oxide increases. This could help explain the lower performance observed for sample N.

The best biofilm performance was observed when E. coli was supported on the CA material, which can be attributed to its neutral or slightly acidic surface character, mesoporous structure, and absence of impurities. Furthermore, the presence of oxygen-containing surface functionalities can further enhance the formation of active bacterial colonies. Additional advantages of the use of carbon gels are related to the proper sol–gel process, which allows for the incorporation of heteroatoms or metallic phases into the carbon matrix, enabling the precise tuning of surface hydrophobicity and electrochemical behavior.

Moreover, to reduce reactant consumption, carbon gels can be synthesized as thin films deposited on various supports. Once the optimal combination of pore structure and surface chemistry is established, different types of monolithic ceramic structures (e.g., honeycombs, foams) can be coated using spin or dip-coating techniques with an appropriate mixture of reactants [30,31]. Polymerization and carbonization are then carried out on the impregnated monoliths. This process yields stable, thin films, while the open channels of the monoliths facilitate wastewater diffusion without significant mass transfer limitations, even under atmospheric pressure or in the presence of fouling. Using biomass as a precursor is also a promising and cost-effective approach to produce materials with excellent properties for biofilm support. Based on the results discussed, biochar can serve as an effective alternative to ACs, eliminating the need for the activation step and thereby saving time, energy, and chemicals. The potential of biochar as a bacterial support has been recently reviewed [32]. The authors concluded that biochar could contribute to the improved performance of biofilm through various mechanisms, including electron transfer, microbial immobilization or alterations in microbial structure at the genetic level. Many different modifications of biochar, namely doping coconut biochar with humid acid, improve the nitrate removal of bioreactors around 33% [33].

A key advantage of carbon materials lies in their versatility with a wide variety of available precursors, along with multiple carbonization, activation, and functionalization strategies, providing a broad design space for tailoring properties. This flexibility often results in materials that are more cost-effective than other commonly used supports or adsorbents, including inorganic oxides, zeolites, and metal–organic frameworks (MOFs).

2.2. Photocatalytic Reduction of Nitrates

2.2.1. Catalyst Characterization

Table 2. BET surface area (S_BET), total pore volume (V_T), pH at the point of zero charge (pH_PZC), crystallite size, and band-gap energy (E_g) of materials of photocatalysts.

Catalyst	SBET (m2 g−1)	VT (cm3 g−1)	pHPZC	Crystalline Phase (%)	Crystallite Size (nm)	Eg (eV)
P25	52		6.5	85 anatase	22	3.20
TiO2	118	0.11	3.5	100 anatase	8	3.12
TiO2–GO	120	0.17	3.0	100 anatase	4	2.95

summarizes the surface area (S_BET), crystalline phases, point of zero charge (pH_PZC), and band-gap energies (E_g) for the photocatalysts tested.

The TiO₂ nanoparticles synthesized exhibited an acidic surface character with a pH_PZC around 3.5, which is attributed to the nature of the precursor employed during synthesis. Similarly, the TiO₂–GO composite showed a slightly lower pH_PZC value, likely influenced by the incorporation of GO. In contrast, the commercial P25 material showed an almost neutral surface with a pH_PZC close to 6.5. The functionalities of materials were analyzed by FTIR (Figure 5a), which confirms the purity of the phases as well as the interplay of both phases in the composite. In both TiO₂ and the TiO₂–GO composite, the typical band between 800–950 cm⁻¹ corresponding to the Ti-O-Ti bond is observed. The most characteristic absorption bands of GO correspond to the vibration bands of the C-OH groups located around 3000–3500 cm⁻¹, carbonyl groups (C=O) at 1720 cm⁻¹. At 1250 cm⁻¹ and at 1050 cm⁻¹, the observed characteristic bands correspond to the stretching vibration mode of C-OH and ether C-O-C groups, respectively. It is noteworthy that the intensity of the peak associated with the hydroxyl and ether groups decreases notably in TiO₂–GO, suggesting possible bonds of TiO₂ with the surface groups of GO at these positions.

XRD patterns (Figure 5b) of both TiO₂ and TiO₂–GO showed only the peak distribution associated with the anatase phase, although the presence of GO in the composite seems to increase the crystallinity of the oxide, resulting in narrower peaks. No GO peaks were detected in the diffractogram of the TiO₂–GO composite. XRD patterns of P25 confirm the well-known presence of both anatase–rutile crystallographic phases in this sample.

Together, structural and chemical properties and textural properties are also involved in the adsorptive performance of the samples (Table 2). In terms of surface area, TiO₂ and TiO₂–GO showed significantly higher S_BET values (118 and 120 m² g⁻¹, respectively), consisting of anatase-phase particles with average crystallite sizes of 8 nm for TiO₂ and 4 nm for TiO₂–GO. On the other hand, P25 showed a lower surface area of 52 m² g⁻¹ and a mixture of anatase and rutile phases (85 and 15% of anatase and rutile, respectively). The porosity analysis suggests that P25 primarily contains macropores or interparticle voids, as indicated by its Type II nitrogen adsorption isotherm. In contrast, the TiO₂–GO composite exhibits mesoporous or even microporous characteristics, owing to the structural influence of GO, as previously reported [34].

The photocatalytic performance is directly related to the interaction of the samples with the radiation, determined spectroscopically. Figure 6 shows the diffuse reflectance UV-Vis spectra of the materials, and the Figure 6 inset shows the transformed Kubelka–Munk function as a function of the energy of light for the determination of the band-gap of the materials. All photocatalysts displayed narrower band-gap energies than P25, with values of 3.12 eV for TiO₂, 2.95 eV for TiO₂–GO, and 3.2 eV for P25. The increase in absorption at the wavelength around 380 nm and mainly in the visible range is directly related to the interaction of TiO₂ and GO, the inherent light-absorbing ability of carbon materials, as well as to the electronic transitions between the carbons and phases of TiO₂, resulting in a lower band-gap (E_g = 2.95 eV, Table 2). The reduced band-gap in the TiO₂–GO material can be ascribed to the inherent light absorption capacity of carbon materials, as well as to the electronic transitions between carbon and TiO₂ phases [34,35].

2.2.2. Photocatalytic Performance of the Samples

Among the different composites developed in our laboratories for environmental applications [36,37,38,39], we selected TiO₂–GO. This choice was based on the significant band-gap narrowing observed (Table 2) with the incorporation of only a small amount of dopant (4 wt% GO) in this composite. These results are in good agreement to those exposed by Lin et al. [18]. Figure 7 shows the photocatalytic reduction profile of nitrate ion over the bare TiO₂, TiO₂–GO, and P25 catalyst and the experiment without a catalyst (photolysis) at pH 2.5. The NO₃⁻ reduction by photolysis is around 8%, and the adsorption experiments (equilibrium) show values around 3–5% of the initial NO₃⁻ concentration under dark conditions. The high photocatalytic activity of P25 regarding the bulk TiO₂ sample arises from the synergistic interaction between its anatase and rutile phases, as previously analyzed.

The incorporation of GO into TiO₂ also improved the photocatalytic characteristics of the TiO₂–GO composite, as clearly reflected in its superior performance, achieving even better results than the commercial P25 sample and enabling up to 80% nitrate removal within 1 h. The observed differences may be attributed to the combined effects of the TiO₂–GO lower band-gap energy, slightly reduced pH_PZC, and known photoluminescence quenching compared to bare TiO₂ [34,40,41]. Additionally, the presence of graphene derivatives appears to facilitate more efficient photogenerated electron transfers and suppresses charge carrier recombination under solar radiation, possibly involving different photocatalytic mechanisms [42].

The catalytic role of carbon materials in nitrate reduction has been widely investigated due to their tunable surface chemistry and morphology. Prior studies have shown that both textural properties and surface doping significantly influence catalytic performance and selectivity toward either N₂ or NH₄⁺ formation. For instance, N-doped carbon supports have been reported to enhance Pd–Cu or Pd–Sn catalyst activity while reducing undesired NH₄⁺ formation, suggesting that electron donor sites introduced by N-functionalities play a critical role in reaction pathways [43,44]. On the other hand, biochars prepared from various biomasses such as straw, rapeseed, rice husk, softwood, and sewage sludge have shown activity under UV irradiation, enabling the reduction of nitrates to NO₂ and NO [45]. Generally, the materials exhibit a low surface area (~30 m² g⁻¹) and a strongly basic character (pH_PZC ≈ 9.5), though their composition varies significantly with the biomass precursor.

Mohamed et al. [46] prepared mesoporous TiO₂ samples using different templates. These samples exhibited significant activity in photocatalytic NO₃⁻ reduction under UV irradiation, achieving conversion rates of 45–60% in contrast to only 7% obtained with the benchmark P25. This improvement was associated with the role of mesoporosity, the formation of heterojunctions between the TiO₂ phases, the particle size and the crystallite size, the latter having a decisive influence on the activity. Moreover, it is also presumed that such high activity is due to the presence of organic moieties acting as hole scavengers, thus improving the catalytic performance. The formation of heterojunctions on TiO₂, active in the NO₃⁻ reduction, has also been systematically studied by doping with noble metals as Pd or other transition metals as Fe, Cu, and Ag [47]. Results are influenced by factors such as metal leaching and the crystal size of doping agents, showing also, in some cases, a certain period of induction prior to the onset of nitrate removal. Shaban et al. [16] prepared C/TiO₂ photocatalysts by a sol–gel procedure using glycine as a carbon source. The E_g values decreased from 3.2 eV for undoped TiO₂ to 1.8 eV after doping with carbon, facilitating the NO₃⁻ reduction also under sunlight radiation. However, the reaction rate was lower compared with experiments carried out under UV radiation. These catalysts are highly active, reaching the total photocatalytic reduction of nitrates following pseudo first-order reaction kinetics. Nitrites are observed as intermediate products with negligible formation of ammonia.

2.2.3. Influence of pH

As was commented in the Introduction section, the reduction of NO₃⁻ strongly depends on the pH (Equations (1)–(6)). Nevertheless, the pH of solutions determines not only the reaction pathway through the ability of the reactant (H⁺) but also the charge of either of the species in solutions and the catalysts surface. The catalysts surface becomes positively charged at pH values below the pH_PZC of catalysts and negatively charged in more basic conditions, thus also conditioning the electrochemical interactions with the NO₃^– ions.

Figure 8 shows that the pH and surface nature influence the photocatalytic performance of the samples. The increasing activity with the increasing acidity of the solution is clear for both samples. Considering P25 (pH_PZC = 6.5, Table 2), at pH 2.5, about 60% of the nitrates are degraded in 1h, while at pH 10, only about 20% are degraded in the same period (Figure 8a). This effect is due to the fact that at an acidic pH the TiO₂ surface is positively charged, forming TiOH₂⁺ species and facilitating the adsorption of nitrate ions (NO₃⁻) to the active sites of the catalyst. On the contrary, at higher pH values, the surface of TiO₂ is negatively charged (TiO⁻), and as a consequence, the adsorption of nitrate ions is disadvantaged due to electrostatic repulsions. The mechanism of adsorption and degradation of NO₃⁻ on the TiO₂ surfaces was analyzed by FTIR in situ experiments [13]. The NO₃⁻ is adsorbed on the Ti sites with the formation of (M-O)₂=NO structures, which are converted to monodentate nitrite (M-O-N-O) as an intermediate reaction.

In the case of TiO₂–GO (Figure 8b), a similar effect is observed, and the highest nitrate degradation is observed at acidic pH values. A comparison between P25 and TiO₂–GO across the tested pH range highlights the significant impact of surface chemistry modifications induced by the presence of GO.

Under acidic conditions (pH < 4), the TiO₂–GO composite exhibits notably enhanced photocatalytic activity compared to P25. This improvement may be attributed to the increased formation of positively charged surface species, likely due to basic functional groups on the GO or electronic interactions between the GO and TiO₂ phases. However, under basic conditions (pH > 7.5), the opposite effect is observed. The TiO₂–GO composite becomes nearly inactive at pH 10, possibly due to the deprotonation of acidic functional groups on the GO surface or alterations in electron transfer dynamics between phases. The abundance of oxygen-containing surface groups (OSGs) on the GO may contribute to the amphoteric behavior of the TiO₂–GO composite and influence charge transfer processes through intrinsic donor–acceptor interactions. As a result, the influence of pH on the TiO₂–GO composite is more pronounced than on P25. Hence, the superior photocatalytic activity of the TiO₂–GO composites relative to P25 can be attributed to strong interfacial interactions between TiO₂ and GO, facilitated by the oxygenated functional groups present on GO (such as epoxy and hydroxyl groups) [40]. These interactions promote optimal assembly and interfacial coupling during synthesis, resulting in GO platelets effectively embedded within the TiO₂ matrix, which enhance the overall photocatalytic performance of the TiO₂–GO composite under solar radiation [34,48].

2.2.4. Effect of Hole Scavengers

As demonstrated in the previous section, nitrates adsorbed on the TiO₂ surface can be photoreduced directly by accepting electrons from the conduction band. However, hole scavengers are commonly used to enhance photocatalytic reduction reactions [12,49], and the efficiency of the process is therefore related to the nature of the scavenger. In this study, the influence of ethanol, oxalic acid, and formic acid (0.08 M) on the performance of P25, bare TiO₂ and TiO₂–GO for the nitrate reduction at pH 2.5 is shown in Figure 9a and Figure 9b, respectively. Product analysis confirmed that NH₄⁺ was not detected under any conditions, regardless of reaction time, photocatalyst, or scavenger. As an example, Figure 9d shows the concentration of nitrate, nitrite, and ammonium over the TiO₂–GO catalyst using formic acid (0.08 M). Small amounts of NO₂⁻ as an intermediate were observed simultaneously with the decrease of nitrate concentration. After 60 min of reaction, no nitrogen-containing species remained in the solution, indicating that nitrate could be completely reduced to N₂.

In the case of P25, the nitrate conversion after 1 h of irradiation was approximately 60% in the absence of scavengers (Figure 9a). As expected, the addition of scavengers enhanced nitrate conversion, following the trend ethanol < oxalic acid < formic acid. This effect can be due to the enhanced separation of electrons and holes and/or the generation of reductive carboxyl anion radicals (CO₂•⁻) during the decomposition of the hole scavenger [49]. These radicals have a strong reductive potential (E°(CO₂/CO₂•⁻) = –1.8 V for converting nitrate to nitrite (E°(NO₃⁻/NO₂⁻) = 0.94 V), to N₂ (E°(NO₃⁻/N₂) = 1.25 V), and for nitrite to N₂ (E°(NO₂⁻/N₂) = 1.45 V) [47,50].

The effect of the initial concentration of formic acid (HCOOH, 0.04, 0.08, and 0.012 M) was also evaluated on the nitrate reduction (Figure 9c). It was found that the reduction rate of nitrate enhanced with the increase in formic acid concentration from 0.04 to 0.08 M, although an increase to 0.12 M leads to a decrease in the photoreduction, indicating overdosing. A similar trend has been observed in other photocatalytic systems [51]. Shi et al. develop Z-scheme ZnSe/BiVO₄ that has high selective photocatalysts of NO₃⁻ into N₂. They used formic acid, ethanol, and acetic acid as scavengers. The best performance was obtained with formic acid, which was attributed to the simultaneous apportioning of H⁺ and CO₂•⁻ radicals [51]. These findings highlight the critical role of acid concentration: insufficient H⁺ availability at low formic acid concentrations slows the reaction and favors NO₂⁻ formation, whereas excessive concentrations promote NH₄⁺ formation. Therefore, identifying the optimal formic acid concentration is essential for maximizing the nitrate reduction efficiency.

When anatase-based photocatalysts are used, the performance of the TiO₂ and TiO₂–GO composite is enhanced by the presence of formic acid, although the increase in activity is not as significant as in the case of P25 (Figure 9b). In the absence of a scavenger, the conversion of nitrates on P25 and TiO₂–GO was 60% and 80%, respectively, after 1 h of reaction (Figure 9a and Figure 9b, respectively). This difference is progressively attenuated with an increasing concentration of formic acid in the solution (Figure 9c) to the extent that the performance of both materials is quite similar at 0.08 M. However, the performance of TiO₂–GO is lower than that of P25 at a higher concentration (0.12 M) of formic acid.

Furthermore, it was observed that the highest performance of the TiO₂–GO composite was not achieved with formic acid, but with oxalic acid. Complete nitrate conversion was achieved in just 15 min when TiO₂–GO was used with oxalic acid as the scavenger. These are the best experimental conditions observed throughout this series of experiments, which again demonstrate the strong influence of interactions between NO₃⁻, the scavenger, and the catalyst surface on the final conversion and selectivity achieved. Considering the stronger effect of pH on the performance of P25 (Figure 8a), the superior performance of oxalic acid compared to formic acid can be attributed to the greater acid strength of oxalic acid (pKa = 1.2) compared to formic acid (pKa = 3.8).

Different studies have used oxalic acid as a scavenger for NO₃⁻ reduction. For example, Li et al. [52] analyzed the photoreduction of nitrates by oxalic acid at pH 2.4. They concluded that the reaction was accelerated by the scavenger and that there was an optimal concentration. Oxalic acid is adsorbed onto the TiO₂ surface, forming oxalate complexes. However, in excess, oxalic acid occupies all surface sites, which hinders the adsorption of nitrates and thus decreases conversion. The reaction involves the formation of HNO₃•^– and C₂O₄•^– radicals, which decompose to CO₂•^–, and the main reaction product is ammonia. Oxalic acid acts as a sacrificial electron donor.

Nevertheless, our results are more consistent with those reported by Sowmya et al. [53]. They found that the reaction on TiO₂–Ag depends on the oxalic acid concentration, which is oxidized following a pseudo-first-order kinetic. NH₄⁺ was not detected, and NO₂⁻ was only observed at short reaction times (Figure 9d). Clearly, doping with either Ag or GO significantly affects the nature of the interactions between the reactants and catalysts. This results in a decreased BG value and improved charge separation, changing not only the conversion values, but also the reaction pathway and product distribution compared to results obtained with bare TiO₂.

Figure 10 illustrates the photocatalytic reduction of nitrate on TiO₂–GO under solar radiation. When the TiO₂–GO photocatalyst is exposed to photons with energy hν ≥ E_g, electrons (e⁻) in the valence band (VB) are excited to the conduction band (CB), leaving behind positively charged holes (h⁺) in the VB and generating electron–hole pairs. The conduction band electrons can act as reducing agents, converting NO₃⁻ to lower oxidation state species, such as NO₂⁻, and finally to N₂ through multi-electron transfer steps. Furthermore, photogenerated electrons in TiO₂ can be transferred to GO, promoting charge separation under solar radiation. Hole scavengers (formic acid) can consume the photogenerated holes, reducing electron–hole recombination and enhancing the availability of electrons for nitrate reduction. This fact modifies selectivity toward N₂ formation instead of undesired by-products like NH₄⁺, as previously observed (Figure 9d).

Table 3. Comparison of recent studies reported in the literature on nitrate reduction.

Approach	Catalyst/System	Conditions	Nitrate Removal	Selectivity	Ref.
Photocatalysis	Ag–TiO2/Formic acid	Visible light, pH 3	High N2 selectivity, fast kinetics	N2 (no NH4+)	[20]
Photocatalysis	g-C3N4/Pd–Cu/rGO/TiO2 hybrid	Visible light	58%	N2	[17]
Photocatalysis	TiO2@Fe3O4–Chitosan composite	UV light, adsorption + photocatalysis	70–80%/1 h	N2	[19]
Photocatalysis	Carbon/TiO2 nanoparticles	pH 3 and 0.04 M of formic acid	100%/60 min	N2	[16]
Photocatalysis	Z-scheme ZnSe/BiVO4	Hg lamp, formic acid	89%/50 min	91% N2	[51]
Photocatalysis	TiO2 (P25)	pH 2.5, solar light, no scavenger	60%/1 h	N2 (no NH4+)	This work
Photocatalysis	TiO2–GO	pH 2.5, solar light, and oxalic acid (0.08 M)	100%/15 min	N2 (no NH4+)	This work
Biological	Moving bed biofilm reactor + AC filter	Synthetic groundwater, acetate C-source	High efficiency	N2	[21]
Biological	Wetland + microbial electrolysis (Fe3O4/GAC anode)	Electrobiotic integration	88.9%	NO2− or NH4+ in the effluent	[22]
Biological	E. coli biofilm on carbon xerogel (CA)	pH 7, anaerobic, and ethanol as C-source	100% NO3− + NO2− removal	N2	This work
Biological	E. coli biofilm on activated carbon (A, N)	pH 7, anaerobic	pH 7, anaerobic	N2	This work

summarizes recent studies on various catalysts, experimental conditions, and performance for photocatalytic and biological nitrate reduction, enabling a direct comparison of catalysts, operating parameters, and key findings between the literature and the present work.

3. Materials and Methods

3.1. Synthesis of Carbon Materials and TiO₂–GO Composites

Three carbon materials with different origins and properties were evaluated as support for Escherichia Coli (E. coli) bacteria. The first material, a carbon xerogel (CA), was synthesized from resorcinol–formaldehyde (R–F) mixtures via sol–gel polymerization. The resulting gel was dried in an oven at atmospheric pressure (100 °C, overnight) and subsequently carbonized at a relatively low temperature (500 °C). The synthesis procedure has been described in detail in previous work [23]. The second material, denoted as A sample, was derived from almond shells. These were initially carbonized at 800 °C, followed by activation under a CO₂ atmosphere for 4 h to enhance porosity. The third carbon support, referred to as N sample, served as a commercial reference. It is an acid-washed, steam-activated carbon with a high pore volume, commercially available under the name RX3-Extra from Norit (Cabot Corporation, Boston, MA, USA).

TiO₂ and TiO₂–GO catalysts were prepared and used in the photocatalytic reduction of NO₃⁻. GO was obtained from an aqueous dispersion of graphite oxide (prepared by the modified Hummers’ method) with subsequent dispersion in an ultrasonic bath [34,54]. The TiO₂–GO compound was synthesized with this GO dispersion by the sol–gel method with an optimum GO content of about 4 wt% and 200 °C (in N₂ atmosphere), as previously described [34]. Briefly, ammonium hexafluorotitanate (IV), (NH₄)₂TiF₆ (0.1 mol/L), and boric acid, H₃BO₃ (0.3 mol/L), were added to the GO dispersion (1 g L⁻¹) and heated at 60 °C for 2 h under stirring. The GO loading was fitted at ca. 4 wt%, where the optimal assembly and interfacial coupling of the TiO₂ nanoparticles with the GO sheets was reached [34]. Bare TiO₂ was also prepared by the same method without adding any carbon material. The commercial Degussa P25 (Evonik, Essen, Germany) was also studied for comparative purposes.

3.2. Characterization Techniques

Textural characterization of samples was carried out by physical adsorption of N₂ at −196 °C using Quantachrome Autosorb-1 equipment (Boston Beach, FL, USA). From the isotherm data, parameters such as the BET surface area (S_BET) and the volume (W₀) and the surface (S_micro) and width (L₀) of micropores were estimated by applying Brunauer–Emmett–Teller, Dubinin–Radushkevich, and Stoeckli equations, respectively [55,56,57]. The total pore volume (V_T) was considered as the volume of N₂ adsorbed at P/P₀ = 0.95. The point zero of charge (pH_PZC) of the powder materials was determined following a pH drift test. Solutions with varying initial pHs were prepared using HCl or NaOH and 50 mL of NaCl as the electrolyte containing 0.15 g of the material. The PZC value of the material was determined by intercepting the obtained final-pH vs. initial-pH curve with the straight line final-pH = initial-pH [58]. Thermogravimetric (TG) analysis of the composites was obtained using a SHIMADZU TGA–50H thermobalance (Tokyo, Japan) by heating the sample in air flow up to 950 °C, with a heating rate of 20 °C min⁻¹. The crystalline properties of photocatalysts were determined by X-Ray diffraction (XRD) (PANalytical X’Pert MPD, Malvern, UK) and the analysis of the band-gap by spectroscopic DRUV-Vis experiments (UV-Vis JASCO V-560, JASCO, Tsukuba, Japan). The spectra have been collected in diffuse reflectance mode and subsequently transformed into equivalent Kubelka–Munk absorption units. Scanning electron microscopy (SEM) images, were recorded with a LEO model GEMINI-1430-VP (Carl Zeiss, Oberkochen, Germany) to analyze the morphology of the samples, as well as the architecture of the bacteria forming the biofilm.

3.3. Nitrate Removal

3.3.1. Biological Experiments

Bacteria used in the experiments was Escherichia Coli, ATCC^® 25922^TM strain (Thermo Fisher Scientific, Waltham, MA, USA), which was first incubated at 37 °C using a buffered media at pH 7 with tryptic soy broth (TSB) before being immobilized on the different supports. Bacteria were supported on the different solids adding 1 mL of this suspension to 0.4 g of support suspended in 20 mL of TSB, and the mixtures were shaken at 37 °C for 3 days. Afterward, the colonized supports were filtered and washed repeatedly with sterilized distilled water.

The denitrification process of water was studied by bacteria immobilized on the different supports (0.2 g), which were added to 100 mL of a solution containing 10 mg L⁻¹ of nitrate from NaNO₃ and 1.3 mL of ethanol. The suspension was buffered at pH 7 with an appropriate phosphate solution. The flasks used as batch bioreactors were flushed with argon atmosphere to obtain anaerobic conditions and then placed in a thermostatic rotary shaker at 25 °C. Periodically, the concentration of nitrate was measured directly in the bioreactors with a selective electrode supplied by Mettler, and simultaneously, a small volume of solution (1 mL) was withdrawn for the determination of the nitrite concentration. This parameter was determined using a Hitachi model U2000 spectrophotometer (Hitachi, Tokyo, Japan) at 543 nm after coupling diazotized sulphanilamide with N-1-naphthyl-ethylendiamine [59].

3.3.2. Photocatalytic Reduction of Nitrate

The photocatalytic reduction of nitrate was performed under simulated solar radiation at room temperature (average 25 °C) using a SOLAR BOX 1500e (CO.FE.MEGRA, Milano, Italy), provided with a 1500 W Xenon lamp (500 W m^–2 of irradiance power). The photochemical reactor consists of a glass reactor loaded with 100 mL of solution containing the nitrate solution (50 mg L⁻¹). The catalyst concentration was fixed as 0.5 g L⁻¹. The suspension was magnetically stirred under nitrogen stream to displace possible dissolved oxygen, allowing the adsorption equilibrium on the catalysts (dark phase) to be established. Aliquots were collected at specific time intervals (5, 15, 30, and 60 min) and filtered with 0.45 μm cellulose acetate filters before determining the nitrate concentrations and the nitrite ions by ion chromatography (LC-10A, Shimadzu, Tokyo, Japan). The content of NH₄⁺ was analyzed according to the standard colorimetric method (Nessler’s reagent, Merck, Germany).

The experiments to determine the effect of pH and the nature of the hole scavenger (0.08 M) in photocatalysis were conducted using the same methodology previously described, with the pH adjusted prior to the dark phase to the values (2.5, 4.0, 7.5, and 10.0) using diluted HCl for acidic conditions and diluted NaOH for basic conditions, measured with a pH meter (CRISON micropH 2002, Barcelona, Spain). The photocatalytic reduction of nitrate was studied in the presence of formic acid (HCOOH), oxalic acid (C₂H₂O₄), or ethanol (EtOH) as hole scavengers, respectively. Formic acid was selected to study the effect of scavenger concentration at 0.04, 0.08, and 0.12 M.

4. Conclusions

This study demonstrates that optimizing both biofilm-based biological denitrification and solar-driven photocatalysis can achieve efficient and sustainable nitrate removal from water. Using E. coli biofilms supported on biomass-derived carbon materials, we found that carbon xerogels (CA) exhibited the best performance due to their mesoporous structure, low ash content, slightly acidic surface character, and abundance of oxygenated groups, all of which favored bacterial adhesion and activity. In contrast, activated carbon, particularly the commercial sample N, was less effective, likely due to their microporosity and less favorable surface chemistry.

For photocatalysis, the TiO₂–GO composite outperformed commercial P25 under acidic conditions, achieving complete nitrate removal at an initial concentration of 50 mg/L within 15 min when oxalic acid was used as a scavenger. The synergistic effects between TiO₂ and GO enhanced the charge separation, light absorption, and active surface species formation. By combining biological and photocatalytic strategies, this work allows for cost-effective and scalable nitrate removal. Future studies should investigate reactor design optimization, long-term stability testing, and pilot-scale demonstrations to advance this dual-approach system toward practical deployment in real-world water treatment applications.