Improved Performance of Sulfur-Driven Autotrophic Denitrification Process by Regulating Sulfur-Based Electron Donors

Abstract

Sulfur-driven autotrophic denitrification (SADN) has demonstrated efficacy in nitrate (NO₃⁻) removal from the aquatic environment. However, the insolubility of elemental sulfur (S⁰) (maximum 5 μg/L at 25 °C) limited the NO₃⁻ removal rate. In this study, we investigated the performance of a laboratory-scale S⁰-packed bed reactor (S⁰-PBR) under various volumetric NO₃⁻ loading rates. By filling with smaller S⁰ particles (0.5–1 mm) and introducing chemical sulfide (30–50 mg S²⁻-S/L), a high NO₃⁻ removal rate (1.44 kg NO₃⁻-N/(m³·d)) was achieved, which was substantially higher than previously reported values in SADN systems. The analysis of the average specific NO₃⁻ removal rates and the half-order kinetic constants jointly confirmed that the denitrification performance was significantly enhanced by decreasing the S⁰ particle sizes from 10–12 mm to 1–2 mm. The smaller S⁰ particles with a larger specific surface area improved the mass-transfer efficiency. Dosing chemical S²⁻ (20 mg S²⁻-S/L) to trigger the abiotic polysulfuration process increased the specific NO₃⁻ removal rate from 0.366 to 0.557 g NO₃⁻-N/g VSS/h and decreased the portion of removed NO₃⁻-N in the form of nitrous oxide (N₂O-N) from 1.6% to 0.7% compared to the S²⁻-free group.

1. Introduction

The widespread use of nitrogenous fertilizers in agriculture and inadequate wastewater treatment has significantly increased nitrate (NO₃⁻) pollution in aquatic environments [1]. Zhang et al. (2021c) [2] measured NO₃⁻ data from 71 major rivers in 30 provinces in China, revealing that the NO₃⁻ concentration in approximately 8% of rivers exceeded the World Health Organization limit of 10 mg NO₃⁻-N/L [3]. The increasing NO₃⁻ loading to coastal zones has induced a severe algae boom, leading to the formation of a “dead zone” [4]. NO₃⁻ can be converted into nitrite (NO₂⁻) or nitrosoamines in the esophagus, which easily aroused methemoglobinemia, blue-baby syndrome, carcinoma, and mutation, thereby posing a severe threat to human life and health [5,6]. Traditional physical/chemical methods (e.g., reverse osmosis, ion exchange, and electrodialysis) for NO₃⁻ removal from wastewater have drawbacks such as high operational cost, low selectivity, and the generation of secondary brine wastes [7].

Alternatively, the biological denitrification process was considered an effective approach for removing NO₃⁻. During this process, NO₃⁻ was sequentially reduced to NO₂⁻, nitric oxide (NO), nitrous oxide (N₂O), and di-nitrogen (N₂) [8]. Organic matters were the most commonly used electron donors to perform heterotrophic denitrification (HD), while hydrogen (H₂), elemental sulfur (S⁰), and iron compounds were utilized by autotrophic denitrifiers [7]. In practice, both insufficient and excessive supplements of organic matter in the HD process would result in poor performance of NO₃⁻ removal and organic residue in the effluent, respectively [9]. Organic supplementation increased the cost and caused biofouling due to the high production of biomass sludge [10].

The autotrophic denitrification process can potentially replace HD because of negligible residual organics in an effluent, given that inorganic matters are utilized as electron donors [10]. Autotrophic denitrifiers exhibited lower biomass production due to the lower biomass yields of 0.4–0.57 g VSS/g NO₃⁻-N [11] than 0.8–1.2 g VSS/g NO₃⁻-N for heterotrophic denitrifiers [12]. The sophisticated hydrogen-delivering systems involved high operating and maintenance costs, which hindered the application of hydrogen-driven autotrophic denitrification [13]. Recently, sulfur-driven autotrophic denitrification (SADN) with the stoichiometry shown as follows [14] (Equation (1)) has gained increasing attention because S⁰ was non-toxic, readily available, and chemically stable under normal conditions and could be used “on demand” without overdosing concerns [15]. The yields (Y) of SADN were relatively low, 0.24 g COD/COD [16], resulting in substantial sludge reduction. SADN was more economical than HD, with estimated costs of $0.45/per kg·N removed versus $1.05/per kg·N removed [17]. In addition, N₂O, as an intermediate product during the biological denitrification process, is a potent greenhouse gas with approximately 300 times the global warming potential of carbon dioxide (CO₂) [18]. Less N₂O is produced in the SADN process [19].

1.1S⁰ + NO₃⁻ + 0.76H₂O + 0.4CO₂ + 0.08NH₄⁺ → 0.08C₅H₇O₂N + 1.1SO₄²⁻ + 0.5N₂ + 1.28H⁺

(1)

The orthorhombic α-S₈⁰, as the only steady form of S⁰ under ambient conditions, was hardly soluble in water (5 μg/L, 25 °C) due to the high bond strength between S-S bonds in S₈⁰-rings and its large molecular size [20]. Owing to this problem, the bioavailability of S⁰ is too poor for sulfur-respiring bacteria, such as S⁰-oxidizing bacteria (S⁰OB) and S₈⁰-reducing bacteria (S⁰RB). Preliminary microbial hydrolysis of S⁰ was required as S⁰ was only taken up by sulfur-respiring bacteria after its solubilization [16,21,22]. The low solubility resulted in lower kinetics than the conventional HD or sulfate (SO₄²⁻) reduction process, which could be seen as the main bottleneck preventing the S⁰-driven bioprocess from realistic applications. Some studies demonstrated a positive relationship between the denitrification rate and factors affecting the surface area of S⁰, including S⁰ concentration [10], particle morphology [23], and size [24,25]. Additionally, the bioavailability of biogenic sulfur (S_bio⁰) particles is superior to chemical sulfur (S_chem⁰) due to its micro-crystallinity structure and higher specific area [26,27].

The nucleophilic attack between sulfide (S²⁻) and S⁰ under neutral or alkaline conditions results in the cleavage of S₈⁰ rings and the formation of polysulfide (S_n²⁻) as detailed in Equation (2) [20]. This chemical reaction has received much attention due to the higher solubility and bioavailability of S_n²⁻ [20,28,29]. Therefore, the polysulfide-involved SADN (PiSADN) process (Equation (3)) might be an effective method for realizing high-rate NO₃⁻ removal. However, the main challenge is how to continuously generate S_n²⁻ in situ. Although promoting the sulfidogenic bacteria activity for S⁰/SO₄²⁻ reduction to trigger the polysulfuration process was an option [30], organic supplementation might lead to a failure of the SADN system because the faster growth rate of heterotrophic NO₃⁻-reducing bacteria than autotrophic NO₃⁻-reducing bacteria [31]. Interestingly, a recent study by Qiu et al. (2022) [32] proposed a novel PiSADN process for S⁰-packed bed reactor (S⁰-PBR), and the polysulfuration was induced by an autotrophic biological sulfur disproportionation (SD) process (Equation (4)). It was difficult to continuously obtain the precursor biogenic S²⁻ through the SD process because the reaction was thermodynamically unfavored under standard conditions [33]. The novel PiSADN process was only adaptable for low-strength wastewater treatment and required sufficient alkalinity supplementation and complex internal recirculation devices.

$${\mathrm{HS}}^{-}+\frac{n-1}{8}{{\mathrm{S}}_8}^0 \leftrightarrow {{\mathrm{S}}_{\mathrm{n}}}^{2-}+{\mathrm{H}}^{+}$$

(2)

S_n²⁻ + 6NO₃⁻ + 2H₂O → S_n-5²⁻ + 5SO₄²⁻ + 3N₂ + 4H⁺

(3)

4S⁰ + 4H₂O → SO₄²⁻ + 3HS⁻ + 5H⁺

(4)

As stated above, soluble S_n²⁻ remarkably enhances the bioavailability of S⁰ and thus facilitates the NO₃⁻ removal performance in the SADN system. However, there are many challenges in the generation pathways for precursor biogenic S²⁻. Dosing organic matter in the SADN system poses a risk to the stability of the microbial community. The SD process is endergonic and is easily inhibited by the presence of high NO₃⁻ loading. These concerns have hindered the development of high-rate in situ PiSADN applications. As such, we investigated the feasibility of establishing an in situ PiSADN system by adding chemical S²⁻ directly for the treatment of high-loading wastewater. Moreover, although the literature found that S⁰ particle size was a key factor affecting the denitrification rate, the underlying kinetic mechanisms were not fully understood. Therefore, a laboratory-scale S⁰-PBR was continuously operated for 163 days under different volumetric loading rates of NO₃⁻. A laboratory-scale S⁰-PBR was continuously operated for 163 days under different volumetric loading rates of NO₃⁻. The specific aims of this study were to (a) demonstrate the feasibility of achieving a high NO₃⁻ removal rate in the long-term S⁰-PBR by introducing smaller S⁰ particles and chemical S²⁻; (b) investigate the effect of different S⁰ particle sizes and S²⁻ initial concentrations on NO₃⁻ removal, NO₂⁻ accumulation, and N₂O production; and (c) analyze the underlying mechanisms of optimized NO₃⁻ removal in the bioreactor. This work might facilitate a better understanding of how to achieve an efficient SADN process in the S⁰-PBR.

2. Materials and Methods

2.1. Bioreactor Setup and Operation

A laboratory-scale plexiglass S⁰-PBR (dimension: 8 cm diameter × 40 cm height) was operated under anaerobic conditions with an effective volume of 1.55 L. The outlet was set at 36 cm from the base. The S⁰-PBR was covered with aluminum foil to prevent the growth of phototrophic bacteria during the entire operation period. The sample port with a 0.8 cm diameter was set at a height of 30 cm, while the bottom of the S⁰-PBR was provided with an outlet port (2 cm diameter). Two peristaltic pumps (KCM-B146, Kamoer, Shanghai, China) were used in the S⁰-PBR operation, i.e., one for feeding and the other for suction.

The S⁰-PBR was filled with S_chem⁰ (0.5–1 mm) and activated carbon (0.5–1 mm) particles with a volume ratio of around 2/3 and 1/3. The inoculation sludge was obtained from the aeration tank of a municipal wastewater treatment plant (Tangxun Lake wastewater treatment plant, Wuhan, China), and the total inoculum mass was approximately 5.4 g. The bioreactor was operated continuously in an up-flow mode at 25 ± 2 °C in a temperature-controlled room.

The S⁰-PBR was fed with synthetic wastewater, as per Qiu et al. (2020) [30]. A step-wise increase in influent volumetric loading, 0.06 kg NO₃⁻-N/(m³·d) to 1.92 kg NO₃⁻-N/(m³·d), was achieved in Stage I (days 1–127) by increasing the NO₃⁻ concentration of 20 to 400 mg NO₃⁻-N and decreasing the hydraulic retention time (HRT). The influent flow rate was increased from 3.23 mL/min to 5.17 mL/min by adjusting the feeding pump, and accordingly, the HRT decreased from 8 h to 5 h. In Stage II (days 128–151), the chemical S²⁻ solution was provided into the S⁰-PBR while maintaining the same influent NO₃⁻ loading rate as the latter Stage I (days 100–127). In Stage III (days 156–163), the working conditions of the S⁰-PBR were identical to the latter Stage I while ceasing the supplement of chemical S²⁻. Sufficient NaHCO₃ was added to the synthetic wastewater, acting as an alkalinity source and inorganic carbon for S⁰OB growth. The details of the three operational conditions are presented in

Table 1. Operational conditions of the S⁰-PBR.

Stages	Stage I	Stage II	Stage III
NO3−-N (mg/L)	20–400	400	400
HRT (h)	8–5	5	5
Loading (kg NO3−-N/(m3·d))	0.06–1.92	1.92	1.92
S2− (mg S/L)	-	30–50	-

.

2.2. Batch Experiments

To investigate the appropriate S⁰ particle size, Test I was performed in 100 mL serum bottles placed in a chamber (20 °C, 200 rpm), including four groups with different S⁰ sizes, i.e., 10–12 mm, 7–9 mm, 3–5 mm, and 1–2 mm, respectively. All bottles were sealed with butyl rubber stoppers and purged with N₂ to obtain anaerobic conditions. The sludge was taken from the S⁰-PBR, and the concentration in each bottle was 0.445 g MLVSS/L. A total of 50 mg NO₃⁻-N/L and 1 g S⁰ particles with the above-mentioned different sizes were added. The purpose of providing excessive S⁰ was to avoid the impact of S⁰ limitation on the denitrification process. In addition, 2 g/L NaHCO₃ was provided to balance the pH and support the bacterial growth. The trace elements were the same as the S⁰-PBR feed solution. The batch test was conducted in duplicate for 34 h, during which samples were taken at 0 h, 3 h, 6 h, 9.5 h, 12 h, 21.5 h, and 34 h to measure NO₃⁻ and NO₂⁻.

As mentioned above, insoluble S⁰ would be converted into soluble S_n²⁻ in the presence of S²⁻ at alkaline conditions. Based on this point, Test II was launched to investigate whether S_n²⁻ could promote the SADN process. Two sets of experiments with the presence of 0 and 20 mg S/L chemical S²⁻ were performed in different serum bottles containing 1 g S⁰ particles (1–2 mm) and 0.473 g/L MLVSS. Controls lacking S²⁻ to monitor the conventional SADN process with S⁰ as the only electron donor. This test lasted for 32 h, during which samples were collected at 0 h, 6 h, 9.5 h, 22 h, 27.5 h, and 32 h to measure NO₃⁻, NO₂⁻, N₂O, and S²⁻ concentrations. Other operational conditions were the same as those mentioned above.

2.3. Chemical Analysis

The NO₃⁻, NO₂⁻, S²⁻ and SO₄²⁻ in the water samples were measured after filtering using disposable Millipore filter units (pore size: 0.22 μm). NO₃⁻ and NO₂⁻ were measured using an ultraviolet–visible spectrophotometer (UV5500PC, Shanghai Metash Instruments Co., Ltd., Shanghai, China). Dissolved N₂O in water samples was analyzed using a gas chromatograph (7890 plus GC, Lunan Ruihong Chemical Instrument, Tengzhou, Shandong, China) fitted with an HP-Plot Molesieve column (30 m × 0.53 mm × 25 μm) and an electron capture detector (ECD). SO₄²⁻ was quantified using an ion chromatograph (883 Basic IC plus, Metrohm, Switzerland) with a conductivity detector. Total dissolved sulfide (H₂S, HS⁻, and S²⁻) was determined using the methylene blue method [34]. The concentration of S_n²⁻ was indicated by the dissolved zero-valent sulfur atoms in polysulfide ions, which was measured by the above UV at a wavelength of 285 nm after filtration [35,36]. pH and temperature were measured with portable meters (Multi-Parameter Meter, HQ40D, Hach, Loveland, CO, USA). MLSS and MLVSS in the sludge used in batch tests were determined according to APHA (2005) [34].

3. Results and Discussion

3.1. Optimization of the S⁰-PBR Performance

To enhance the S⁰-PBR performance, a long-term continuous-flow experiment focused on reducing the size of S_chem⁰ particles and facilitating the formation of S_n²⁻. Three operational conditions were applied in the bioreactor. In Stage I, the S_chem⁰ particles (0.5–1 mm) were used as the main filler. The volumetric loading rate of the reactor was step-wise increased to investigate the feasibility of enhancing NO₃⁻ removal capability by reducing S_chem⁰ particle size in S⁰-PBR. In Stage II, on the basis of optimum S_chem⁰ particle size, the chemical precursor S²⁻ (30–50 mg S²⁻-S/L) was added to form S_n²⁻ to accelerate the SADN process. In Stage III, the external S²⁻ addition was completely eliminated so that the polysulfuration process was inhibited, highlighting the positive effect of S_n²⁻ as an electron donor on the SADN process.

In Stage I (day 1–127), the influent NO₃⁻ concentration increased from 20 to 400 mg NO₃⁻-N/L, and HRT decreased from 8 h to 5 h, resulting in a step-wise increase in volumetric loading rate from 0.06 kg NO₃⁻-N/(m³·d) to 1.92 kg NO₃⁻-N/(m³·d). Of note, even with an influent NO₃⁻ loading as low as 0.06 kg NO₃⁻-N/(m³·d) in the early phase of Stage I (day 1–40), the average NO₃⁻ removal efficiency was only 89.3% (Figure 1a,b). The main reason might be attributed to the low abundance of S⁰OB in the inoculation sludge and its slow growth rate, which led to a start-up period of S⁰-PBR as long as 28 days [37]. Yang et al. (2016a) [38] mentioned that the volumetric denitrification loading rate was less than 0.1 kg NO₃⁻-N/(m³·d) when the MLVSS concentration remained below 0.3 g/L in the anoxic filter. After the adaption period, NO₃⁻ was occasionally detected in the S⁰-PBR effluent during days 40–99 (Figure 1a). The accumulation of functional biomass could explain this result due to the relatively strong biomass retention capacity of packed-bed reactors [39]. The S⁰OB could become the dominant microbial community in the S⁰-PBR since the absence of organic supplementation.

Given the limited electron-scavenging capability from the solid S⁰ interface by S⁰OB [40,41], the NO₃⁻ reduction rate was significantly higher than the NO₂⁻ reduction rate, leading to NO₂⁻ accumulation in the SADN system. However, NO₂⁻ was undetectable during days 1–99 (Figure 1a), even during days 95–99, corresponding with a high volumetric loading rate of 1.44 kg NO₃⁻-N/(m³·d) and high average NO₃⁻ removal efficiency of approximately 100% (Figure 1a,b). The efficient denitrification performance of the S⁰-PBR could be attributed to the use of smaller S⁰ particles (0.5–1 mm) with larger specific surface areas in the S⁰-PBR than those in the literature (2–16 mm) [10,32]. The higher surface areas of S⁰ improved the mass-transfer efficiency during S⁰OB utilization of S⁰ [10]. Consequently, the dissolution process of S_chem⁰, considered the main rate-limiting step, was largely promoted [10,19,42].

However, during days 100–127, the average effluent NO₃⁻ concentration increased to 78 mg NO₃⁻-N/L, and the NO₃⁻ removal efficiency continuously declined to 62.7% on day 127 (Figure 1a,b). In addition, overloading of the S⁰-PBR was evidenced by the detection of NO₂⁻ and high-level average N₂O concentration of 32 mg N₂O-N/L in the effluent (Figure 1a,c). These could be seen as reliable markers of the reactor overloading in the SADN process [10,43,44]. Therefore, the maximum NO₃⁻ removal loading rate of the S⁰-PBR was considered as 1.44 kg NO₃⁻-N/(m³·d) during days 95–99, which was 1.88 times higher than the result obtained by Koenig et al. (2001) [24] who used bigger size of S⁰ particles, 2.8–5.6 mm.

Of note, the practical effluent SO₄²⁻ concentration during days 1–40 and 43–60 averaged 447 mg SO₄²⁻/L and 542 mg SO₄²⁻/L (Figure 1d), respectively, which were significantly higher than the theoretical values that were 155 mg SO₄²⁻/L and 364 mg SO₄²⁻/L. It is well known that the sulfur-based autotrophic disproportionation process occurs only after NO₃⁻ is depleted in the SADN system [45,46,47], especially near the effluent side of S⁰-PBR [32]. Hijnen et al. (1992) [48] pointed out that the volumetric loading rate of S⁰-PBR should be kept above the minimum limitation, 0.22 kg NO₃⁻-N/(m³·d), to prevent the head loss caused by the SD process. It was reasonable to infer that the occurrence of the SD process at these two early periods of Stage I, due to the relatively low volumetric loading rates, 0.06–0.15 kg NO₃⁻-N/(m³·d), and the nearly complete NO₃⁻ removal in the S⁰-PBR (Figure 1a,b). According to Equation (4), SO₄²⁻ in excess amount of theoretical production in these two periods was likely to come from the SD process. In addition, the ratios of SO₄²⁻ production to NO₃⁻ removal were becoming closer to the theoretical value as the volumetric loading rate increased (Figure 1d). As NO₃⁻ loading rates were in the range of 0.76–1.92 kg NO₃⁻-N/(m³·d) during days 80–127 (Figure 1b), this ratio was almost equivalent to the theoretical value of 7.54 mg SO₄²⁻/mg NO₃⁻-N, suggesting the inhibition of high NO₃⁻ loading rate on SD process. A previous study also reported that the SD process was completely inhibited when influent NO₃⁻ loading exceeded 0.72 kg NO₃⁻-N/(m³·d) and the concentration of the sulfur-heterologous electron acceptors (e.g., NO₃⁻, NO₂⁻, and dissolved oxygen) increased to 1.1 mg/L [49].

In Stage II, 30 to 50 mg S/L chemical S²⁻ was added into the S⁰-PBR in sequence while keeping the volumetric loading rate constant at about 1.92 kg NO₃⁻-N/(m³·d), same as that during days 100–127 in Stage I (Figure 1b). As a result of the addition of 30 mg S²⁻-S/L, the downward trend of NO₃⁻ removal efficiency was terminated and replaced by an upward trend during days 128–142, showing that the average NO₃⁻ removal efficiency was increased to 85.3% from 81.3% (Figure 1b). In addition, upon the overloading of influent NO₃⁻, the average N₂O concentration of 2 mg N₂O-N/L in effluent samples was far lower than that during days 100–127 (32 mg N₂O-N/L) without chemical S²⁻ addition, decreasing N₂O accumulation by 93.8%. The result differs from previous points that the S²⁻ could precipitate with soluble copper cofactors in the N₂O reductase, leading to a rise in N₂O production [50]. However, it was also observed in a previous study by Yang et al. (2016) [51] that the bio-poisoning chemical S²⁻ could be instantly oxidized into S_n²⁻ by membrane-bound sulfide-quinine reductase presented in almost S⁰OB, and S_n²⁻ acting as a bioavailable electron donor could contribute to N₂O reduction.

However, the average NO₂⁻ concentration increased from 6 mg NO₂⁻-N/L to 15 mg NO₂⁻-N/L during days 145–151 in Stage II, when the dosage increased to 50 mg S²⁻-S/L. The bioavailability of insoluble S_chem⁰ in this S⁰-PBR could be greatly improved by adding a higher S²⁻ concentration to promote the chemical polysulfuration process. The formation of S_n²⁻ and the higher competitive capacity of the nitrate reductase for electrons than nitrite reductase explained well the severe NO₂⁻ accumulation [52,53]. Moreover, it has been reported that high-level chemical S²⁻ exerted an inhibitory effect on nitrite reductase activity and ceased the NO₂⁻ reduction process [54,55,56]. As a result of the severe NO₂⁻ accumulation, the denitrification microorganisms in the S⁰-PBR could be further restrained [41,57,58] and caused an undesirable NO₃⁻ removal performance, exhibiting the average NO₃⁻ removal efficiency declined to 80.8% from 85.2% during days 128–142 (Figure 1d).

In Stage III (days 156–163), the operational conditions were identical to those during days 100–127 in Stage I, with an overloading volumetric loading rate of 1.92 kg NO₃⁻-N/(m³·d) and no external chemical S²⁻ addition. The average NO₃⁻ removal efficiency decreased further to 42.1% (Figure 1b). Simultaneously, the NO₂⁻ accumulation was aggravated to 19 mg NO₂⁻-N/L (Figure 1a), suggesting the deterioration of SADN performance in the S⁰-PBR with high influent NO₃⁻ loading applied. The high volumetric loading rate completely prevented the SD process in Stage III, as evidenced by the similar practical SO₄²⁻ production (1267 mg SO₄²⁻/L) and theoretical value (1327 mg SO₄²⁻/L) (Figure 1d). Therefore, the deterioration could be attributed to the lack of precursor, such as biogenic/chemical S²⁻, to induce a chemical polysulfuration reaction.

3.2. The Short-Term Effects of Varying S⁰ Particle Sizes and Chemical S²⁻ Addition on the SADN Process

Batch experiments were categorized into four groups based on the diameters of S⁰ particles, i.e., 1–2 mm, 3–5 mm, 7–9 mm, and 10–12 mm, for evaluating the effect of varying particle sizes on the SADN process.

NO₃⁻ removal fastened as the S⁰ particle size decreased (Figure 2a). NO₃⁻ was almost completely removed at 12 h when the S⁰ size was smaller than 5 mm (Figure 2a). Comparatively, the residual NO₃⁻ concentrations were approximately 16 mg NO₃⁻-N/L and 12 mg NO₃⁻-N/L at 12 h in the groups with the S⁰ particle sizes of 10–12 and 7–9 mm, respectively. Meanwhile, as a result of a lower NO₂⁻ reduction rate and faster NO₃⁻ reduction rate, the build-up of NO₂⁻ was gradually formed in all groups (Figure 2b), which was consistent with points that the capability of electron-scavenging for nitrite reductase was weaker than nitrate reductase [59,60].

The average specific NO₃⁻ removal rates within the first 12 h in groups with S⁰ sizes of 10–12 mm, 7–9 mm, 3–5 mm, and 1–2 mm applied were 0.672 g NO₃⁻-N/g VSS/h, 0.678 g NO₃⁻-N/g VSS/h, 0.850 g NO₃⁻-N/g VSS/h, and 0.910 g NO₃⁻-N/g VSS/h, respectively (Figure 2c). Additionally, it has been reported that a half-order reaction model could be used to explain the kinetics of the SADN process [24]. The half-order kinetic constants in groups with S⁰ sizes of 10–12, 7–9, 3–5, and 1–2 mm applied were calculated to be 0.382 mg-N^1/2/L^1/2/h, 0.435 mg-N^1/2/L^1/2/h, 0.545 mg-N^1/2/L^1/2/h, and 0.565 mg-N^1/2/L^1/2/h (Figure 2c), suggesting that the reaction rate constant increased with the specific surface area of S⁰ [25]. The smaller S⁰ size with a higher specific area not only provided a larger area for biofilm growth but, more importantly, reduced the mass-transfer resistance of insoluble S⁰ [10,29].

Generally, it was assumed that the saturation constants K_s was as low as 0.22 mg S/L in the SADN process [24,47], indicating that the affinities between S⁰ and the enzymes related in S⁰ oxidiation, such as SDO/SOR/Hdr were strong. Given that S⁰ was only taken up by S⁰OB after its solubilization and diffusion [16,21], the mass-transfer resistance of insoluble S⁰ became the main rate-limiting factor in the SADN system. The specific surface area of insoluble sulfur was the key parameter affecting the population of hydrolysis bacteria attached to its surface and the dissolution kinetics [61]. A kinetic model focusing on S⁰ hydrolysis as a prior and rate-limiting step was proposed, where both activities of hydrolytic biomass and autotrophic denitrifying bacteria in the SADN process were considered [19,62]. The model demonstrated that the specific surface area of S⁰ was the dominant factor affecting the denitrification rate.

To investigate how S²⁻ or S_n²⁻ promoted NO₃⁻ removal, batch Test II with an initial dosage of 20 mg S²⁻-S/L was conducted. During 0–9.5 h, the specific NO₃⁻ removal rates and NO₃⁻ consumption slope k in the S²⁻-added group were 0.557 g NO₃⁻-N/g VSS/h and 0.0465 (Figure 3a,b), respectively, significantly higher than the S²⁻-free group of 0.366 g NO₃⁻-N/g VSS/h and 0.0364. It could be due to the fact that the lower Gibbs energy was required when S²⁻ with the relatively high solubility served as the additional electron donor, compared with the conventional SADN process [53]. However, the quietly close NO₃⁻ removal rates (3.0 mg NO₃⁻-N/(L·d) versus 3.6 mg NO₃⁻-N/(L·d)) were found in the two groups with S⁰ and chemical S²⁻ as a single electron source by Qi et al. (2023) [63]. This result indicated that the acceleration of NO₃⁻ removal in this study should be mainly attributed to the S_n²⁻ formation instead of chemical S²⁻ participation. In the presence of chemicals S²⁻ and S⁰, the abiotic polysulfuration process was triggered (Equation (2)). As a result of the product of soluble S_n²⁻ with higher bioavailability, the NO₃⁻ removal rate was remarkably enhanced in the S²⁻-added group. This result was consistent with previous studies [29,30,32] and the improvement in NO₃⁻ removal efficiency in Stage II (day128–142) in the S⁰-PBR.

Of note, NO₂⁻ accumulation occurred in both groups and was aggravated by S²⁻ addition (Figure 3c). The results were similar to the long-term performance of S⁰-PBR in Stage II and the previous study [63,64], indicating that dosing chemical S²⁻ could significantly improve the NO₃⁻ reduction process but rarely promote NO₂⁻ reduction process. The aggravated NO₂⁻ accumulation in the S²⁻-added group resulted from the imbalance rate of the NO₃⁻ and NO₂⁻ reduction process, and the imbalance could be attributed to two main reasons. Firstly, the extent of NO₂⁻ accumulation was positively correlated with the NO₃⁻ reduction rate [60]. It can also be noticed that the NO₃⁻ removal rate was faster in the S²⁻-added group (Figure 3a,b) due to the presence of S_n²⁻, which explained the severe NO₂⁻ accumulation well. Secondly, the bio-toxicity of chemical S²⁻ to nitrite reductase [54,55] and the higher competitive capacity of the nitrate reductase for electrons both hindered the NO₂⁻ reduction process [54,55,56].

Additionally, in the S²⁻-free group, only 1.6% of removed NO₃⁻-N was in the form of N₂O-N within 27.5 h. The amount of N₂O production was much lower than in the HD process, suggesting that less N₂O was produced in the SADN process [19,65]. Additionally, a further decrease in N₂O production in the S²⁻-added group was observed even in the presence of higher NO₂⁻ accumulation, only accounting for 0.7% of removed NO₃⁻-N within 27.5 h (Figure 3c). The result was consistent with the performance of the S⁰-PBR (Figure 1c) in Stage II. Similarly, a linearly proportional relationship between chemical S²⁻ concentration and N₂O emissions during autotrophic denitrification was reported in the study, including the mass ratio of S²⁻-S:NO₃⁻-N up to 5 [51]. Yang et al. (2016b) [51] also confirmed that chemical S²⁻ had no inhibitory effect on nitrous oxide reductase. Moreover, S_n²⁻ was formed in the S²⁻-added group due to the abiotic polysulfuration process. When S_n²⁻ participated in the N₂O reduction process, higher energy than S⁰-oxidation was yielded [66]. These explained the lower N₂O production in the S²⁻-added group well.

4. Conclusions

Based on the long-term performance of the S⁰-PBR, the NO₃⁻ removal loading rate could be significantly enhanced using smaller S⁰ particle fillers with a higher specific surface area. More importantly, chemical S²⁻ supplementation improved the performance of the S⁰-PBR under overloading conditions. It proved the feasibility of establishing an in situ PiSADN system by adding chemical S²⁻ directly for high-loading wastewater treatment. The conducted batch tests have clarified the kinetic dynamics between the sizes of S⁰ particles and the rate of denitrification. Furthermore, the responses of the SADN process to chemical S²⁻ were also investigated. The principle findings were summarized as follows:

♦

Utilization of smaller S⁰ particles (0.5–1 mm) within the S⁰-PBR achieved a high volumetric loading rate of 1.44 kg NO₃⁻-N/(m³·d) and a NO₃⁻ removal efficiency nearing 100%, significantly surpassing outcomes observed in S⁰-PBR employing larger S⁰ particles (2–16 mm);

♦

The supplementation of 30 mg S²⁻-S/L in the S⁰-PBR led to an increase in NO₃⁻ removal efficiency from 81.3%% to 85.3% and facilitated a 93.8% reduction in N₂O accumulation;

♦

In the batch tests with a S⁰ size of 10–12, 7–9, 3–5, and 1–2 mm applied, the average specific NO₃⁻ removal rates were 0.672 g NO₃⁻-N/g VSS/h, 0.678 g NO₃⁻-N/g VSS/h, 0.850 g NO₃⁻-N/g VSS/h, and 0.910 g NO₃⁻-N/g VSS/h, respectively, while the half-order kinetic constants were 0.382 mg-N^1/2/L^1/2/h, 0.435 mg-N^1/2/L^1/2/h, 0.545 mg-N^1/2/L^1/2/h, and 0.565 mg-N^1/2/L^1/2/h, respectively;

♦

The specific NO₃⁻ removal rates and NO₃⁻ consumption slope k in the S²⁻-added group were 0.557 g NO₃⁻-N/g VSS/h and 0.0465, respectively, significantly higher than S²⁻-free group of 0.366 g NO₃⁻-N/g VSS/h and 0.0364;

♦

The 1.6% of removed NO₃⁻-N was in the form of N₂O within 27.5 h in the S²⁻-free group, while only 0.7% of the removed NO₃⁻-N was produced as N₂O in the S²⁻-added group.