Degradation of Oxytetracycline in Saturated Porous Media by In Situ Chemical Oxidation Using Oxygen-Doped Graphitic Carbon Nitride and Peroxymonosulfate: Laboratory-Scale Column Experiments

Abstract

Oxytetracycline (OTC) is frequently detected in groundwater and soil, posing substantial risks to the subsurface environment via persistence, phytotoxicity, changing bacterial communities, and antibiotic resistance. In situ chemical oxidation (ISCO) is one of the best alternatives for removing OTC from groundwater. However, its feasibility has rarely been investigated using columns for which optimal conditions can be obtained for practical applications. Thus, a system consisting of oxygen-doped graphitic carbon nitride (OgCN) and peroxymonosulfate (PMS) (OgCN/PMS) was tested for OTC removal using continuous-flow experiments with columns packed with sand and glass beads (GBs). The sand column exhibited better adsorption and degradation of OTC than the GB column in pulse injection experiments, regardless of whether OgCN was packed. Additional experiments were performed using a column saturated with the OTC solution and another filled with deionized water to simulate ISCO, using GB as the medium, to evaluate the net OTC removal by catalytic oxidation, excluding adsorption. Performance improved with increased OgCN packing, PMS dosage, retention time, and pH. Anions slightly affected the performance due to scavenging and propagation of radicals. These findings indicate the high potential of OgCN/PMS for ISCO and the usefulness of column experiments in field applications.

1. Introduction

Pharmaceuticals, particularly antibiotics, are used worldwide to prevent infections in humans and animals [1]. Among the broad-spectrum antibiotics, tetracyclines (TCs) are favored for medical and veterinary applications because of their high efficiency and low cost [1,2,3,4]. Many antibiotics are released into aquatic systems through various pathways, such as excretion, treatment of aquatic animals, and runoff of livestock waste [1,5,6,7,8,9]. This leads to the accumulation of antibiotics in different environmental media, such as soil, surface water, groundwater, and sediment, over time. For example, the oxytetracycline (OTC) content is as high as 100 ng/L, 50,000 ng/g, and 183,500 ng/g in groundwater, soil, and manure, respectively, which are frequently applied to soil [10]. In addition, the maximum DT50 of OTC has been reported to be 63 days, which is higher than most antibiotics [1]. The diverse existence of antibiotics in the natural environment significantly affects bacterial communities and ecosystems while promoting antibiotic resistance [11]. In particular, TCs are phytotoxic, inhibit plant growth, and decrease the content of chlorophyll and carotenoids. Oxytetracycline (OTC) is bioconcentrated in aquatic plants consumed by humans [12]. Thus, removing OTC from soil and groundwater is essential for protecting the subsurface environment, considering the persistence and risks of OTC in the soil environment.

Considering the risk posed by OTC to human health and the ecosystem, in situ chemical oxidation (ISCO) is regarded as the best option for the removal of OTC from soil and groundwater, because OTC is not readily biodegradable [13]. ISCO is based on the activation of oxidants, involving injecting catalysts and powerful oxidants directly into polluted zone of soil or groundwater to break down organic contaminants on-site [14]. The oxidants include hydrogen peroxide (H₂O₂), peroxydisulfate (S₂O₈^2–, PDS), peroxymonosulfate (HSO₅⁻, PMS), and permanganate (MnO₄⁻), leading to the generation of reactive species, such as hydroxyl radical (^●OH), sulfate radical (SO₄^●−), and singlet oxygen (¹O₂) [15,16]. On the other hand, the catalysts include homogeneous transition metal ions, such as Fe²⁺, as well as heterogeneous metallic particles such as Fe⁰ and Fe₃O₄ [17,18]. However, they have disadvantages of high cost and leaching of metal. Moreover, the leached metals can be transported to clean subsurface zones, accumulate, and block the pores in soil [19].

In this regard, heterogeneous carbon nanomaterials are more advantageous than homogeneous or heterogeneous metallic activators for organic contaminant degradation [20]. They include graphene oxide, carbon nanotubes, and graphitic carbon nitride (gCN), which have excellent thermal and chemical stability, availability, and well-developed pores. However, using metallic materials is not recommended because of metal leaching, accumulation of non-reactive species from the leached metals, and pore blockage [21]. Therefore, the combination of the carbon nanomaterials and oxidant(s) is one of the best options for ISCO without the secondary contaminants.

Several systems have demonstrated the high potential of ISCO in batch experiments, including biochar and PDS for acetaminophen removal [22], N–S-co-doped mesoporous graphite-like carbon nanosheets and PDS for TC removal [23], and activated carbon and PDS for phenol removal [24]. However, it is difficult to find reports on continuous flow experiments in columns packed with porous media, although the feasibility of practical applications in subsurface environments can be verified. ISCO is generally conducted by the injection of catalysts, such as carbon nanomaterials, into the subsurface contaminated zone, followed by the continuous flow of liquid-phase oxidants [14]. ISCO shares common advantages and limitations with batch processes, such as high efficiency and oxidant consumption by organic matter [25]. However, different from batch processes, the migration of contaminated flume from the contaminated zone to clean zone must be considered via the experiments with continuous oxidant flow [14] for proper evaluation of ISCO alternatives.

In this regard, the performance of a system of oxygen-doped gCN (OgCN) and PMS (OgCN/PMS) was evaluated via continuous experiments using a porous-medium column in which OgCN was injected. Previous studies have demonstrated the excellent performance of OgCN/PMS in OTC degradation and transport through porous media [26,27,28]. Moreover, gCNs are non-toxic [26]; therefore, they may be left at the point of injection after the completion of ISCO operations.

2. Materials and Methods

2.1. Chemicals and Materials

OTC hydrochloride (>95%) and potassium PMS (OXONE^®, 2KHSO₅·KH₂SO₄·K₂SO₄ > 99%) were purchased from Merck KGaA (Darmstadt, Germany). Urea (NH₂CONH₂), oxalic acid dihydrate ((COOH)₂.2H₂O), sodium chloride (NaCl), potassium chloride (KCl), sodium hydroxide (NaOH), sodium bicarbonate (NaHCO₃), sodium sulfate (Na₂SO₄), sulfuric acid (H₂SO₄), and sodium nitrate (NaNO₃) were purchased from Samchun Chemical Co., Ltd. (Seoul, Republic of Korea). Deionized water (OTC/DIW) was prepared using an AquaPuri 5 series (Youngin Chromass Co., Ltd., Anyang-si, Republic of Korea).

OgCN was synthesized by heating a mixture of 20 g urea and 8 g oxalic acid at 550 °C for 4 h [29].

2.2. Column Experiments

Glass columns (ID 25 mm × L 145 mm) equipped with two end plates and stainless steel screens were used for the experiments. They were packed with either 112 g of sand or 103.4 g of glass beads (GBs). The pore volumes and porosities of the packed columns were measured by standard gravimetric methods [30].

The hydrodynamic dispersion coefficients (D_L, cm² min⁻¹) of the sand and GB columns were obtained from the breakthrough curves (BTCs) of a tracer (KCl). The columns were flushed with 10 PVs of DIW, 2.23 PVs of 10 mM KCl solution, and 10 PVs of DIW. The flow rate was 2.0 mL min^−1, and the KCl concentrations in the effluents were measured using a conductivity probe (Vernier, Beaverton, OR, USA). BTCs were analyzed using Hydrus-1D version 4.17 software (Equation (1)) [31]:

$${\displaystyle \frac{\partial \left(\theta C\right)}{\partial t}}=D_L{\displaystyle \frac{{\partial }^2\left(\theta C\right)}{\partial z^2}}-{\displaystyle \frac{\partial \left(vC\right)}{\partial z}}-\rho {\displaystyle \frac{\partial S}{\partial t}}$$

(1)

where θ is the volumetric water content (cm³ cm⁻³), C is the concentration of OTC in the liquid phase (mg L⁻¹), z is depth (cm), ν is Darcy’s velocity (cm min⁻¹), $\rho$ is the dry density of sand (or glass beads) (mg cm⁻³), and S is the OTC concentration in solid phase (mg g⁻¹), which was neglected when an inert solute, i.e., KCl, was used.

OTC removal in the sand and GB columns was compared with pulse injection of OTC. The adsorption of OTC on the columns was investigated by flushing with 10 PVs of DIW, 24.5 PVs of 5 mg L⁻¹ OTC solution, and then with 10 PVs of DIW (Figure 1A). OTC removal by PMS was evaluated by repeating the experiments, replacing DIW with a 0.25 mM PMS solution (Figure 1B). This was repeated using sand and GB columns packed with 750 mg of OgCN (Figure 1C). The BTCs of OTC were analyzed using Hydrus-1D (Equation (2)). We employed a two-site model (Equations (2)–(6)), where OTC was adsorbed onto one type of site (S₁), whereas the pseudo-first-order degradation of OTC occurred on another type of site (S₂) [32,33]:

$$S=S_1+S_2$$

(2)

$$S_1=f{K}_{d}C$$

(3)

$$S_2=\left(1-f\right)K_{d}C$$

(4)

$${\displaystyle \frac{\partial S_1}{\partial t}}=f{K}_d{\displaystyle \frac{\partial C}{\partial t}}$$

(5)

$${\displaystyle \frac{\partial S_2}{\partial t}}=\omega \left(\left(1-f\right)K_{d}C-S_2\right)$$

(6)

where f is the fraction of all sites, K_d is the partition coefficient (L kg⁻¹) calculated by the concentration of OTC on solid divided by that in liquid under equilibrium (C_S/C_e), and ω is the pseudo-first-order reaction rate coefficient (min⁻¹). The K_d was determined to be 1.767 and 0.074 L kg⁻¹ for sand and GB, respectively, by separate experiments.

Simulated ISCO experiments were performed using a series of two (2) columns packed with GB or sand [34]. The pore of the first column (column 1) was completely filled with a 5 mg L⁻¹ OTC solution to emulate the subsurface zone saturated with groundwater that is polluted with OTC. The pore of the second column (column 2) was filled with DIW, representing downflow zone saturated with clean groundwater. The effluent of column 1 was introduced to column 2 and was flown through column 2, as DIW or PMS solution then the effluent.

Subsequently, continuous experiments were conducted using these three systems. Firstly, column 1 was not packed with OgCN where DIW was introduced, denoted “OTC/DIW” (Figure 1D). Secondly, PMS aqueous solution was introduced to the columns, denoted as “OTC/PMS” (Figure 1E). Thirdly, the first GB or sand column was packed with 750 mg OgCN, where PMS solution was injected, denoted “OTC/OgCN/PMS” (Figure 1F).

The flow rate of the DIW and PMS solutions was 2 mL min⁻¹ for all experiments.

2.3. Analytical Methods

OTC concentration was measured with a high-performance liquid chromatography (HPLC) system (YL9100 Plus, Yongin, Republic of Korea) equipped with a C18 column (Eclipse Plus, Agilent, CA, USA), because of the complex structures of OTC, having many chromophores (Table S1) [35]. The mobile phase consisted of acetonitrile, 0.01 M oxalic acid, and methanol (20:70:10, v/v/v). The flow rate of the mobile phase, temperature, injection volume, and detection wavelength was 1.0 mL/min, 25 μL, 30 °C, and 360 nm, respectively [22]. PMS concentration was measured by a colorimetric method using KI using UV-vis absorption at λ = 395 nm [36].

3. Results and Discussion

3.1. Dispersivity of Sand and GB Columns

The BTCs of KCl of sand and GB columns, where OgCN was pre-packed, are shown in Figure 2. KCl started to be detected approximately at 0.9 pore volume (PV) and reached its initial concentration at around 1.35 PV for both columns. The C/C₀ was around 0.5 at a PV of 1.0 in both columns, indicating that the tracer (KCl) migration was not hindered significantly. Hydrus-1D was a good fit for the BTCs of KCl, and the dispersion in both columns was similar, considering the values of D_L (

Table 1. Parameters of tracer BTCs.

	DL (cm2 min−1)	r2
GB	0.1545	0.9908
Sands	0.1374	0.9897

).

3.2. OTC Removal in Sand and GB Columns with Pulse Injection of OTC

The BTCs of OTC are displayed in Figure 3, and the ω and K, calculated by Hydrus-1D, are given in

Table 2. Parameters of Hydrus-1D for the BTCs of OTC in slug input of OTC to GB and sand column.

		Kd (L kg−1)	ω (min−1)	r2
GB	OTC/DIW	0.074	0.169	0.934
	OTC/PMS	0.074	0.231	0.883
	OTC/OgCN/PMS	0.074	0.460	0.902
Sand	OTC/DIW	1.767	0.219	0.946
	OTC/PMS	1.767	0.484	0.934
	OTC/OgCN/PMS	1.767	0.744	0.841

. In OTC/DIW, sand and GB columns showed similar and low ω. The BTC of the sand column was more asymmetric with a longer tailing than that of the GB column in OTC/DIW (Figure 3A,B), implying that more sorption was involved in the transport of OTC in the sand column than in the GB column, as suggested by K_d. However, the OTC adsorption by sand and GB was 0.077 ± 0.004 and 0.003 ± 0.004 mg/g, respectively, suggesting low contribution of adsorption both in sand and GB columns.

When PMS was introduced, the OTC in the effluents decreased for both columns (Figure 3C,D), which was attributed to the activation of PMS, leading to OTC degradation. The self-activation of PMS to ¹O₂ was responsible for OTC removal in the GB column [37,38]. On the other hand, it was proposed that O₂^●− can be generated via one electron transfer from PMS to the Si of a high oxidation state in sand, probably O = Si(δ⁺) [39]. O₂^●− is an important precursor of ¹O₂, which plays a major role in organic pollutant degradation in various catalytic systems [15,23,24,27,40].

OTC removal was greatly improved in both columns when OgCN was packed (Figure 3E,F), as confirmed by the raised values of ω (Table 1) via PMS activation by OgCN [27]. This also verifies the excellent potential of OgCN for the removal of organic pollutants by ISCO in subsurface environments.

3.3. OTC Removal in Simulated ISCO

The results of the simulated ISCO experiments using the GB columns are presented in Figure 4A,C. In OTC/DIW, OTC was flushed out of GB column 1 and transferred to GB column 2, resulting in an increase or decrease in the OTC concentration in GB column 2. The time courses of the OTC in OTC/PMS and OTC/OgCN/PMS were similar to those in OTC/DIW. However, the OTC concentration decreased as PMS was introduced and was further reduced in OTC/OgCN/PMS. The OTC in the effluents of column 1 at 1.44 PV was 2.44, 1.45, and 0.38 mg/L for OTC/DIW, OTC/PMS, and OTC/OgCN/PMS, respectively. In contrast, the maximum OTC concentration in the effluents of column 2 in OTC/DIW decreased as PMS was injected and OgCN was additionally packed. The overall removal of OTC in OTC/PMS and OTC/OgCN/PMS was 35.33% and 66.49%, respectively, mainly attributed to column 1. These results suggest the oxidative removal of OTC via the self-activation of PMS and elevated activation of PMS by OgCN. It has been reported that OgCN reduces PMS to produce ^●OH and SO₄^●− by electron transfer from electron-rich O atoms to PMS and that the electron-poor C atoms adjacent to the O atoms accept electrons from PMS to form SO₅^●− [24,26,41].

The activation of PMS in both the OTC/PMS and OTC/OgCN/PMS systems was confirmed by PMS consumption (Figure 4B,D). The PMS showed BTCs in the effluents as the PMS solution passed through the columns. The PMS concentration in OTC/OgCN/PMS was substantially lower than that in OTC/PMS in both columns 1 and 2 during the entire operation period, indicating that OgCN promoted the decomposition of PMS. It was also shown that PMS predominantly decomposed in column 1, either by self-activation without OgCN (OTC/PMS) or catalysis by OgCN (OTC/OgCN/PMS) (Figure 4D).

It is certain, by the results in Figure 4, that the OTC/OgCN/PMS system is efficient in OTC removal. However, the possibility of byproduct formation must be considered. This was investigated by dissolved organic carbon (DOC) removal and the changes in UV-vis spectra, as provided in Figure S1A and B, respectively. The removal of DOC was significantly lower than that of OTC, suggesting the formation of OTC degradation byproducts. The spectrum of OTC showed two absorption peaks at 260–280 nm, which is assigned to the aromatic ring A and dimethylamino group, and at 340–380 nm, which is attributed to the B, C, and D rings and chromophores [42]. The spectra of column 1 showed that the peaks decreased gradually. In particular, the peak at 340–380 nm decreased faster than that at 260–280 nm, indicating that rings of B, C, and D were cleaved preferentially to ring A, probably because of less energy. This is consistent with the results of our previous work [22], which demonstrated that OTC undergoes cleavage of aromatic rings, followed by hydroxylation/demethylation at C4, deamination, and secondary alcohol oxidation in the presence of OgCN and PMS to form various organic compounds.

OTC removal was significantly enhanced when sand columns were used (Figure 5) compared to the GB columns. Substantial OTC removal was observed in column 1 of OTC/DIW, which was attributable to adsorption, as verified by K_d. OTC removal was enhanced in OTC/PMS and increased in OTC/OgCN/PMS via PMS activation by Si and OgCN, as shown in Table 2 and Figure 3. The OTC concentration in OTC/DIW, OTC/PMS, and OTC/OgCN/PMS was 1.71, 1.28, and 0.63 mg/L, respectively, at PVs of 1.81–1.87. The OTC concentrations in column 2 followed the order: OTC/DIW > OTC/PMS > OTC/OgCN/PMS. However, they were less than 0.55 mg/L, suggesting a significant OTC removal in column 2, probably via adsorption. Oxidation by residual reactive species in column 1 entering column 2 can be considered. However, it may not be plausible considering the short life of ^●OH (20 ns), SO₄^●− (30–40 μs) and ¹O₂ (3.7 ± 0.4 μs) [43,44].

It should be noted that the total OTC removal in OTC/PMS (78.64%) was similar to that in OTC/OgCN/PMS (88.96%); however, OTC was predominantly removed in columns 2 and 1 in OTC/PMS and OTC/OgCN/PMS, respectively. This suggests that the sand in column 2 was mainly responsible for the OTC removal in OTC/PMS, whereas it was predominantly contributed by the catalytic degradation by OgCN and PMS in column 1 in OTC/OgCN/PMS. The role of PMS activation was verified by the significantly lower PMS concentration and higher PMS decomposition in OTC/OgCN/PMS than in OTC/PMS, both in column 1 and in column 2 than in column 1, both in OTC/PMS and OTC/OgCN/PMS (Figure 5B,D).

The effects of the operating conditions in the simulated ISCO, such as the packed OgCN amount, PMS dose, empty bed retention time (EBCT), pH, and the presence of anions, were investigated using GB columns. The effects of adsorption were excluded to better evaluate the catalytic activities of OgCN and PMS. Therefore, inert materials such as GB were suitable, whereas the OTC removal in the sand columns significantly depended on adsorption.

3.4. Effects of OgCN and PMS Dose

Figure 6 shows the effects of OgCN and PMS doses on OTC degradation in the OTC/OgCN/PMS system using GB columns. The OTC degradation was promoted by increasing the concentration of OgCN (Figure 6A,C). OTC concentration was 4.29, 3.16, and 1.93 mg/L when OgCN was 0.2, 0.5, and 0.75 g, respectively, at 0.90–0.93 PVs in column 1. In column 2, the PV of the maximum OTC concentration decreases as OgCN increases. The OTC removal in column 1, where OgCN was packed, was promoted with increasing OgCN, while that in column 2 was not correlated with OgCN but showed a relatively similar removal of 25.06 ± 2.86%. This indicates that OTC removal in column 2 was not affected by the performance of column 1.

Increasing the PMS dose from 0.1 to 0.25 mM decreased and increased the OTC removal in columns 2 and 1, respectively. However, it was not significantly enhanced as the PMS dose was further increased in either column (Figure 6B,D). The similar OTC removal in column 1 at 0.25−0.75 mM PMS was attributable to the PMS scavenging and self-quenching of ^●OH and SO₄^●− under excessive PMS dose [45,46].

3.5. Effects of EBCT

The time courses of the OTC at different empty bed contact times (EBCT) are shown in Figure 7A,B. In the effluents of column 1 for both OTC/PMS and OTC/OgCN/PMS, the initial OTC concentrations within 1 PV decreased, whereas those over 1 PV increased as the EBCT increased. This suggests that the dispersion, retardation, and removal of OTC are enhanced by longer retention times. This was more significant when the EBCT increased from 36 min to 71 min than when it increased from 18 min to 36 min. It is also shown that the maximum OTC concentrations in column 2 in OTC/PMS were reached at 2.94, 3.27, and 4.07 PVs, while those in OTC/OgCN/PMS were at 2.71, 2.96, and 3.75 PVs, for the EBCT of 18, 36, and 71 min, respectively.

OTC removal in OTC/PMS increased from 22.88% to 34.33% and 42.04% when the EBCT was increased from 18 to 71 min (Figure 7C), indicating that PMS self-activation [37] was enhanced as the retention time increased. That of OTC/OgCN/PMS also grew from 35.64% to 75.30% with increasing EBCT (Figure 7D). Notably, OTC removal in column 1 increased gradually, but that in column 2 increased from 12.55% to 26.09%, but decreased slightly to 21.99% when EBCT was 18, 36, and 71 min, respectively. It is believed that the maximum contribution of PMS self-activation was approximately 26.09% in the GB columns, considering the OTC removal in the columns where OgCN was not packed (i.e., columns 1 and 2 in OTC/PMS and column 2 in OTC/OgCN/PMS).

However, it seems reasonable that the performance would be better in field applications because the retention of the OTC and PMS solutions in the columns was much shorter than in the field, even at an EBCT of 72 min, because of the higher linear velocity. That of the flow in the columns was 5.87, 2.94, and 1.47 m/day when the RBCT was 18, 36, and 72 min, respectively, while that of groundwater in the Republic of Korea was reported to be 0.06–0.44 m/day [47].

3.6. Effects of pH

It has been reported that the performance of PMS-based advanced oxidation is significantly affected by the pH because it influences the speciation of PMS (HCO₅⁻ or SO₅⁻) [48] and the charge of OgCN and therefore, the contact between OgCN and PMS [28]. Thus, OTC removal from the columns was investigated under acidic, neutral, and alkaline conditions, that is, pH values of 3, 7, and 9, respectively.

The removal of OTC by OTC/PMS was promoted by increasing the pH from 3 to 9 (Figure 8A,C). This suggests that the self-activation of PMS is enhanced under alkaline conditions. PMS exists as its protonated form (HSO₅⁻) under the pK_a of 9.4, which is more reactive than its deprotonated form (SO₅⁻) at pH higher than the pK_a [38]. However, it was recently shown that SO₅⁻ also generates HO₂⁻ via hydrolysis (Equation (7)) [38], which forms one of the major intermediates of ¹O₂, i.e., O₂^●− (Equation (8)) [49]. In addition, HO₂⁻ generates SO₄^●− and O₂^●− via reactions with HSO₅⁻ (Equation (9)) [50]. Moreover, ¹O₂ is directly generated by HSO₅⁻ and SO₅⁻ (Equation (10)) [38], under the pH near the pK_a of PMS where both HSO₅⁻ and SO₅⁻ exist.

SO₅⁻ + OH⁻ → HO₂⁻ + SO₄²⁻ + H⁺

(7)

HO₂⁻ + H₂O → O₂^●− + H₃O⁺

(8)

HSO₅⁻ + HO₂⁻ → SO₄^●− + O₂^●− + H₂O

(9)

HSO₅⁻ + SO₅⁻ → SO₄²⁻ + ¹O₂ + H⁺

(10)

OTC removal was enhanced when OgCN was packed (OTC/OgCN/PMS), regardless of the pH (Figure 8B,D), suggesting that OgCN substantially activated PMS under the pH conditions used in this study. It is also shown that the performance in OTC/OgCN/PMS was enhanced from 66.49% to 74.33% as pH was increased. This suggests that PMS activation by OgCN is not significantly influenced by contact between them. The electrostatic attraction decreased as the pH increased because the point of zero charge of OCN is approximately 4.6 [28]. Thus, it seems reasonable that the enhancement is attributed to the alkaline and self-activation of PMS [38,49,50] (Equations (7)–(10))as well as the propagation of SO₄^●−/SO₅^●− to ^●OH [51].

3.7. Effects of Anions

The influences of anions commonly found in groundwater, such as HCO₃⁻, Cl⁻, NO₃⁻, and SO₄²⁻ [52], were investigated to examine the feasibility of field applications. The concentration of the anions was 5 mM considering the maximum values in groundwater in the Republic of Korea, i.e., 2.9–7.6 mM [52], and to keep the same ratio of them to PMS concentration. OTC removal in OTC/PMS was not significantly affected by HCO₃⁻, NO₃⁻, and SO₄²⁻ (Figure 9A,C). This demonstrates that the self-activation of PMS was not notably influenced by HCO₃⁻, NO₃⁻, and SO₄²⁻. OTC/OgCN/PMS was affected more by the anions than OTC/PMS (Figure 9B,D). OTC removal was 66.49% in the control, while it decreased to 58.63%, 52.7%, and 60.53% when HCO₃⁻, NO₃⁻, and SO₄²⁻ were introduced, respectively. The presence of Cl⁻ positively affected the performance both in OTC/PMS and OTC/OgCN/PMS.

It is known that HCO₃⁻ and NO₃⁻ scavenge SO₄^●−/^●OH to form less reactive CO₃^●− (1.59 eV) and NO₃^● (2.3–2.5 eV), respectively (Equations (11)–(16)) [53,54,55]. The performance of SO₄^●−-based systems can be decreased by SO₄²⁻ via the combination of them (Equation (17)) [56]. Also, SO₄²⁻ scavenges Cl^●/HOCl^● [57]. On the other hand, it was reported that Cl⁻ scavenges SO₄^●−/^●OH to generate Cl^●, HClO^●, or Cl₂^●, having substantial redox potentials of 2.41–2.47, 2.7, or 2.0–2.12 eV, respectively [57]. In addition, Cl₂, HOCl, and OCl⁻ are all oxidizing agents with positive redox potential, while ¹O₂ can be formed via HOCl [58].

On the other hand, the effects of anions in Figure 9 were not as dramatic as those in batch experiments, probably because of the short reaction times, i.e., EBCT. In addition, the results in Figure 9 and in other studies [35,40] suggest that batch experiments can be used to investigate the effects of anions in ISCO operations.

^●OH + HCO₃⁻ → CO₃^●− + H₂O

(11)

^●OH + CO₃⁻ → CO₃^●− + OH⁻

(12)

SO₄^●− + HCO₃⁻ → CO₃^●− + SO₄²⁻ + H⁺

(13)

SO₄^●− + CO₃⁻ → CO₃^●− + SO₄²⁻

(14)

SO₄^●− + NO₃⁻ → SO₄²⁻ + NO₃^●

(15)

^●OH + NO₃⁻ → NO₃^●−

(16)

SO₄²⁻ + SO₄^●− → S₂O₈²⁻ + e⁻

(17)

It should be noted that the effects of organic matter should be evaluated in the future because it generally exists in groundwater. The mean and 95th percentile concentration of dissolved organic matter in groundwater were reported to be 3.8 and 16.6 mg-C/L, respectively [59]. Natural organic matter (NOM), especially fulvic acid, would significantly affect the performance of ISCO using OTC/OgCN/PMS, as demonstrated by the results of batch experiments in previous studies [60,61]. The inhibition of PMS activation by NOM is attributable to the accumulation of it via the adsorption of phenolic hydroxyl and carboxyl groups and to the consumption of PMS and reactive species, such as SO₄^●−, by nucleophilic components [61]. However, the mechanisms and the degree of inhibition and/or promotion of a catalytic oxidation system must be investigated in depth separately, because they greatly differ by catalyst and oxidant [60,61,62].

4. Conclusions

In this study, the performance of oxygen-doped gCN (OgCN)/PMS was investigated via continuous experiments using columns filled with porous media to evaluate its feasibility for practical applications in ISCO.

The dispersivity of the sand and GB columns packed with 750 mg of OgCN was similar to the dispersion coefficients of 0.1374–0.1545 cm² min⁻¹. However, the results of continuous experiments with pulse injection of OTC showed that OTC removal was significant in the sand column but negligible in the GB column when both OgCN and PMS were used. However, it improved significantly when PMS was introduced and was further enhanced when OgCN was packed in both columns, whereas the performance was better in the sand column than in the GB column. This was attributed to the greater adsorption of OTC and PMS activation in the sand column than in the GB column. It was supported by a higher partition coefficient (K_d), and reaction rate constants (ω) in the sand column than in the GB column, calculated using Hydrus-1D.

The superiority of sand over GB was confirmed by simulated ISCO experiments, which showed that OTC removal and PMS decomposition were significantly better in sand than in the GB column, regardless of the presence of PMS or OgCN. The performance of OgCN/PMS was further investigated by varying the influencing factors using the GB columns to exclude the influence of adsorption. It was shown that OTC removal was improved with increasing packed OgCN amount, PMS dose, EBCT, and pH. These were attributable to the increased reactive sites, oxidant amount, contact time, and the promoted formation of O₂^●− and/or ¹O₂. OTC removal was slightly inhibited by HCO₃⁻, NO₃⁻, and SO₄²⁻, but improved by Cl⁻, probably because of scavenging and propagation of SO₄^●−/^●OH.

Overall, the results of this study confirm the high potential of the OgCN/PMS system in ISCO for pharmaceutical removal and would significantly contribute to establishing optimum conditions for field applications.