Degradation of carbamazepine by high-voltage direct current gas–liquid plasma with the addition of H2O2 and Fe2+

Carbamazepine (CBZ) is a typical psychotropic pharmaceutical which is one of the most commonly detected persistent pharmaceuticals in the environment. The degradation of CBZ in the aqueous solution was studied by a direct current (DC) gas–liquid phase discharge plasma combined with different catalysts (H2O2 or Fe2+) in this study. The concentrations of reactive species (H2O2, O3, and NO3−) and •OH radical yield in the liquid were measured during the discharge process. The various parameters that affect the degradation of CBZ, such as discharge powers, initial concentrations, initial pH values, and addition of catalysts, were investigated. The energy efficiency was 25.2 mg·kW−1·h−1 at 35.7 W, and the discharge power at 35.7 W was selected to achieve the optimal balance on the degradation effect and energy efficiency. Both acidic and alkaline solution conditions were conducive to promoting the degradation of CBZ. Both H2O2 and Fe2+ at low concentration (10–100 mg/L of Fe2+, 0.05–2.0 mmol/L of H2O2) were observed contributing to the improvement of the CBZ degradation rate, while the promotional effect of CBZ degradation was weakened even inhibition would occur at high concentrations (100–200 mg/L of Fe2+, 2.0–5.0 mmol/L of H2O2). The degradation rate of CBZ was up to 99.1%, and the total organic carbon (TOC) removal efficiency of CBZ was up to 67.1% in the plasma/Fe2+ (100 mg/L) system at 48 min, which suggested that high degradation rate and mineralization efficiency on CBZ could be achieved by employing Fe2+ as a catalyst. Based on the intermediate products identified by Ultra Performance Liquid Chromatography Tandem Mass Spectrometry (UPLC–MS), the possible degradation pathways were proposed. Finally, the growth inhibition assay with Escherichia coli (E. coli) showed that the toxicity of plasma/Fe2+-treated CBZ solution decreased and a relatively low solution toxicity could be achieved. Thus, the plasma/catalyst could be an effective technology for the degradation of pharmaceuticals in aqueous solutions.


Introduction
Organic pollutants are among the most common sources of pollution and have attracted significant attention in recent years due to their wide range of potential influences . Pharmaceuticals and personal care products (PPCPs) such as triclosan, carbamazepine, diclofenac (Leong et al. 2021;Yu et al. 2013) and emerging contaminants (ECs) such as X-ray contrast media and flame retardant (Koch et al. 2019;Wang et al. 2019b) are not easily detected in the environment, yet it can stay in the environmental media for a long time and has a potential impact on the ecological environment as well as human health (Amarasiri et al. 2019;Gonsioroski et al. 2020). With the increase of drug consumption, hazardous drug residues have been detected in water, soil, aquatic organisms, and other environmental Responsible Editor: Ricardo A. Torres-Palma media at different levels. What is more, PPCPs have been monitored in the inlet and outlet water of sewage plants, surface water, and groundwater in some countries and regions, and the concentration is up to ug/L (Watkinson et al. 2007). However, it has been proven that even low concentrations of hormone drugs have a significant impact on human health (Santos et al. 2010). Therefore, the potential hazards of PPCPs should not be ignored.
As a typical representative of PPCPs, carbamazepine (CBZ , Table 1) is consumed about 1014 tons per year worldwide due to its good therapeutic effects (Mohapatra et al. 2014). Edward Archer et al. (Archer et al. 2017) found that the removal rate of CBZ in traditional sewage treatment plants was less than 15%. Residual CBZ had been found in surface water, groundwater, and even drinking water, and its long-term accumulation may cause certain harm to aquatic systems, such as damaging the livers of aquatic organisms (Jarvis et al. 2014) and changing the physiological function of vertebrates (Suwalsky et al. 2006). There is evidence that the concentration of CBZ in the wastewater discharged from sewage treatment plants can cause harmful effects (Magureanu et al. 2015). Therefore, to ensure the safety of effluents discharged from sewage treatment plants, it is urgent to explore new technologies to deal with such pollutants.
The application of advanced oxidation processes (AOPs) to reduce harmful effects of emerging contaminants had been proved to be an effective way (Kim et al. 2020). 2021 The main characteristic of AOPs is that they are effective means to generate highly reactive species (RS) for chemical oxidation of macromolecular organic contaminants into harmless mineralized products. The most commonly used AOPs include photocatalysis (Baaloudj et al. 2021;Rabahi et al. 2019), ozonation (Ostman et al. 2019), fenton oxidation (Giri and Golder 2014), ultraviolet (UV) (Kamagate et al. 2018;Zeghioud et al. 2018), etc. However, oxidative process such as ozone is laborious to keep in reserve and UV treatment is powerless to achieve a high level of mineralization (Santos et al. 2015). Therefore, there is still room for the development of more reliable technologies for removing contaminants.
Recently, plasma processes for the treatment of pollutants such as volatile organic compounds (VOCs) (Dobslaw et al. 2018) and wastewater (Liu et al. 2021) have aroused extensive attention of scholars due to their efficient and clean technologies (Qu et al. 2021;Xu et al. 2022a, b). Plasma was generated by high-voltage discharge at the liquid or gas-liquid interface. High-energy electrons generated by high-voltage discharge collided with surrounding gas molecules and reactive species such as short-lived free radicals (•OH, •H, •O, etc.) and long-lived molecules (H 2 O 2 , O 3 , etc.) formed through direct and indirect electron collision process and chemical reaction (Malik et al. 2016). Xin et al. (Xin et al. 2020) applied gas-liquid plasma to remove bromoamine acid in dye wastewater and efficient removal was achieved by the system within 15-min treatment. Feng et al. (Feng et al. 2016) used a wire cylinder DBD reactor to remove atrazine in aqueous solution, and 93.7% of degradation rate and 12.7% of TOC removal rate were achieved after 18 min of treatment. Since the sole plasma exhibited restricted penetration depth into the liquid, short-lived reactive species generated by plasma, such as •OH (considered to be one of the most important reactive species to degrade liquid-phase harmful organic compounds), unable to diffuse to the deep parts of the solution, resulting in low mineralization level and limited efficiency to a certain extent. The H 2 O 2 generated by the plasma has a long lifetime and can reach a relatively far position in solution by diffusion, the addition of Fe 2+ can fully utilize H 2 O 2 to trigger the catalytic degradation of fenton reaction to generate •OH, thus achieving improved organic pollutants degradation rates. Specifically, homogeneous catalysts possess remarkable properties as the lower investment cost, no phase-transfer problems, high activity, and selectivity. Sang et al. (Sang et al. 2019a) studied the degradation of aniline by dielectric barrier discharge (DBD) combined with Fe 2+ and found that the removal efficiency increased obviously when Fe 2+ was added. Fahmy et al. (Fahmy et al. 2018) employed Fe 2+ as a catalyst in discharge plasma treatment and 95.05% of decolorization efficiency was obtained after 20 min time of treatment. Feng et al. (Feng et al. 2015) discovered that the removal rate of TOC and the degradation efficiency of 3,4-DCA increased dramatically with adding Fe 2+ . In previous study, we have explored the effect of H 2 O 2 and Fe 2+ on the degradation of antibiotics by DBD in an alternating current (AC) power supply condition. Different discharge forms will affect the composition and concentration of reactive species in the liquid induced by plasma, and these reactive species are an important factor in the degradation of pollutants in water. Compared with the dielectric barrier discharge excited by the AC high-voltage power supply, the structure of the DC high-voltage discharge plasma is simpler, and the price of power supply device is relatively low. However, compared with the dielectric barrier discharge plasma, there are fewer studies on DC high-voltage discharge plasma for the degradation of pollutants in water, especially the effect of the liquid phase reactive species and the addition of catalysts on the degradation of organics in the water caused by this plasma. Therefore, this research is a continuation of our previous research and hopes to provide an option for actual wastewater treatment in the future. In this study, CBZ was selected as the target organic contaminate to examine the degradation effected by DC gas-liquid phase plasma with a catalyst (Fe 2+ or H 2 O 2 ) system. Both of these catalysts should promote the removal rate of organic substances. The hydroxyl radical yield (•OH) and three representative long-lived reactive species (H 2 O 2 , O 3 , and NO 3 − ) that may promote the oxidization of CBZ have been quantified. Moreover, the influence of various parameters like initial solution concentration, initial solution pH value, and the catalyst concentration on the performance of the reactor was investigated to obtain the optimal condition on CBZ degradation for a higher degradation rate. Furthermore, the primary degradation intermediates of CBZ and potential degradation pathways were identified by analyzing UPLC-MS results. The possible toxicity of intermediate products was also evaluated through growth inhibition assays using the E. coli bacterium as the testing organism.

Experimental apparatus
The schematic diagram of our experimental device to generate the gas-liquid phase discharge is illustrated in Fig. 1. Briefly, the experimental setup primarily consists of a direct current (DC) high-voltage power supply, plasma device, flowmeter, and air pump. The plasma device consists of a hollow stainless steel needle surrounded by a quartz glass tube with an inner diameter of 2.6 mm placed inside a quartz container. The needle (8 cm length and 2 mm inner diameter) is the anode electrode which is connected to the gas inlet and driven by a positive high-voltage DC power supply. The ground electrode is the CBZ aqueous solution. Air was used as working gas in this experiment at a flow rate of 1.5 L/min. Discharge occurs in the gas channel of the quartz tube under the CBZ liquid interface, and 40 mL of CBZ solution (unless otherwise specified, 20 mg/L was used) was treated for each test.
The applied voltage and discharge current of the plasma were measured by a high-voltage probe (Tektronix P6015A) and a current probe (Tektronix P6021) via a digital oscilloscope (Tektronix MSO 5104). The average discharge power (P) in the system was calculated by the following Eq. (1) (Xiao et al. 2014): where P represents the average discharge power, U(t) and I(t) are the discharge voltage and current at time t, respectively, and T is the discharge period. The optical emission spectra from the gas-liquid plasma was recorded by a spectrometer (Maya2000 Pro, Ocean Optics, USA) with an fiber probe. The fiber probe was mounted on the exit of the glass tube nozzle and about 20 mm apart from the plasma source.

Analysis methods
A high-performance liquid chromatography (Agilent 1260 Infinity II, USA), consisted of a C18 column (250 × 4.6 mm i.d., 5 μm; Agilent, USA) maintained at 30.0 °C, was used to measure CBZ concentration at 286.0 nm by a UV detector. Mobile phase A was acetonitrile, while mobile phase B was Milli-Q water, and they were at the ratio of 60/40 (v/v) with a flow rate of 1.0 ml/ min. The sample volume of the degradation solution was 20 μl. The data were analyzed by using the software Agilent ChemStation (Agilent, USA). The degradation rate of CBZ for each sample was calculated according to Eq. (2) (Markovic et al. 2015): where η (%) indicates the degradation rate of CBZ, C 0 (mg/L) is the initial concentration of CBZ, and C t (mg/L) represents the residual concentration of CBZ after different plasma discharge treatment time (mg/L).
The energy efficiency Y(mg·kW −1 ·h −1 ), defined by the amount of CBZ degraded per unit energy delivery in the reactor, is calculated by Eq. (3) (Wang et al. 2016): where Y (mg·kW −1 ·h −1 ) represents the energy efficiency, C 0 (mg/L) refers to the initial concentration of CBZ, V (L) was the reactor volume, η (%) indicates the degradation rate of CBZ, P (kW) is the average discharge power, and t (h) is the plasma discharge time.
To investigate the mineralization variations of the plasma-treated CBZ solution, the concentration of the total organic carbon (TOC) in the initial CBZ solution and subsequent samples was acquired via a TOC/TN b analyzer (Mutil N/C 3100, Analytik Jena, Germany). The TOC removal rate (%) of CBZ was calculated according to Eq. (4): where ω (%) represents the TOC removal rate of CBZ, TOC 0 is the concentration of TOC in the initial CBZ solution, TOC t is the concentration of TOC in the CBZ solution after discharge.
To elucidate variations in toxicity of the plasmatreated of CBZ, the antimicrobial test of the samples against E. coli was carried out, according to Wen et al. (Wen et al. 2015) method with minor modifications. The specific operation is as follows: 0.5 mL of CBZ solution with different treatment times (6-48 min) in the plasma/ Fe 2+ (100 mg/L) system, untreated CBZ solution 0.5 mL untreated CBZ solution (pH = 6.4, 20 mg/L), and blank control were added to a series of tubes containing 3 mL of sterilized LB medium, and then 50 µL E. coli cultured overnight was added to each of tubes. LB medium was used as negative control in this test and all tubes were incubated in an incubator at 37 ℃ at 200 rpm. The growth of E. coli was analyzed by measuring the absorbance at λ = 600 nm and the toxicity of samples was estimated through the inhibition index (I ind ) according to the following Eq. (5) (Silambarasan and Vangnai 2016): where A 0 and A are the absorbance values at 600 nm for the inoculated solutions in the absence (control samples) and presence of CBZ or its degraded metabolites, respectively.

Measurement of reactive species in deionized water
Terephthalic acid (TA) was used as a chemical probe to measure the •OH radical yield, which reacted with •OH to form 2-hydroxyterephthalic acid (HTA), and the •OH radical yield was measured indirectly using the fluorescent properties of hydroxyterephthalic acid (HTA) (Mason et al. 1994;Tampieri et al. 2021). The concentration of H 2 O 2 was determined by hydrogen peroxide (H 2 O 2 ) content assay kit (AKAO009C, Beijing Boxbio Science & Technology Co., Ltd, China). A spectrophotometer (photolab 7100, WTW, Germany) was used to detect concentrations of O 3 and NO 3 − in the gas-liquid plasma-treated solution. The relevant test kits were 00,607 and 09,713, and measurement methods were operated according to the instruction manual (Shen et al. 2019;Zhang et al. 2015).

Mass Spectrometry (MS)
To better understand the degradation mechanism of CBZ, the major degradation intermediates of CBZ were investigated qualitatively by UPLC-MS (ACQUITY UPLC LCT premier XE, Waters, USA) and the possible degradation pathways of CBZ were analyzed. The analytical parameters of MS were as follows: fragmentation at 125.0 V, capillary voltage of 4.0 kV, desolation gas (nitrogen, 99.99% purity) flow rate of 10.0 L/min, nebulizer pressure of 40.0 psi, temperature of 300℃, scan range of m/z = 100-1000, and argon (99.99% purity) as the collision gas.

Electrical parameters of the gas-liquid phase plasma
The electrical properties of plasma were closely related to the degradation effect of pollutants; therefore, the variations of current and voltage during plasma discharge and average discharge power were detected and calculated, respectively.
The typical voltage and current waveform of the gas-liquid air plasma are presented in Fig. 2. As shown, the peak value of discharge voltage and current were approximately 6.0 kV and 0.3A, respectively. In this work, the variation range of the discharge powers was from 20.9 to 56.4 W.

Optical emission spectroscopy (OES)
The types of reactive species formed in the gas-phase or at the gas-liquid interface can be identified according to their characteristic peaks on the optical emission spectroscopy (OES) of the discharge. The typical optical spectrum of the gas-liquid phase discharge with air over the range of 200 to 900 nm is presented in Fig. 3. The discharge generates a significant UV radiation which belongs to transitions of the OH band from 305.0 to 320.0 nm (Rashid et al. 2020). In particular, the distinct peaks in the range of 315.0-400.0 nm correspond to diverse nitrogen species (Aggelopoulos et al. 2020). In addition to those lines above, the transitions of H α and H β are found at 656.3 nm and 486.1 nm, respectively. Furthermore, the distinct peaks were also detected at 777.3 and 844.6 nm could be attributed to the formation of excited atomic oxygen (Rezaei et al. 2014). Notably, the atomic oxygen come from oxygen molecules and water molecules excited by high energy particles, while the atomic hydrogen was generated by high energy particles exciting water molecules. Identifying excited particles formed in the air phase can explain the source of liquid products more clearly since the liquid phase product will be triggered by the gas phase product. In the air gas-liquid discharge, the reaction between charged particles and water molecules is to generate more reactive species (H 2 O 2 , NO 3 − , O 3 and •OH).

Measurement of RS in deionized water induced by plasma and plasma/catalyst
The plasma-induced aqueous RS has attracted much attention due to its significant role in transforming and degrading organic compounds. As a significant indicator of the performance of gas-liquid plasma, the generation of RS in aqueous solution is frequently monitored (Vanraes et al. 2015). The plasma discharge could produce various ions and free radicals and further lead to the generation of RS in liquid phase by physical and chemical reactions, which would interact with the organic pollutants synergistically (Ognier et al. 2015;Rozas et al. 2017 , HO 2 •, etc.) ones. The simplified formation mechanisms of these species during the discharge are as follows (6)-(20) (Dojcinovic et al. 2011;Reddy and Subrahmanyam 2012): The •OH yield and the aqueous concentrations of H 2 O 2 , O 3 , and NO 3 − generated by plasma treatment and catalyst at discharge power of 35.7 W are presented in Fig. 4, respectively. As shown in Fig. 4a, the concentrations of the •OH radicals reached 4.40 uM, 5.69 uM and 0.66 uM under the condition of sole discharge, H 2 O 2 (0.1 mM), Fe 2+ (100 mg/L) after 4 min of treatment, respectively. •OH was formed in the discharge process mainly due to high-energy electrons' bombardment on H 2 O (Eq. (9) and the conversion from H 2 O 2 and O 3 (Eq. (13) and Eq. (17)), thus leading to a gradually enhancing •OH concentration for the sole plasma and plasma/H 2 O 2 (0.1 mM) system from 0 to 4 min. There were no accumulation effects of •OH radicals in the liquid due to its short lifetime ( about 10 −9 s); therefore, the •OH production rate was calculated indirectly by using the amounts of HTA formation per unit time. Consequently, the production rates of •OH in the liquid were estimated to be of the order of 1.83 × 10 −8 M s −1 and 2.37 × 10 −8 M s −1 for the sole plasma and plasma/ H 2 O 2 (0.1 mM) system under the present experimental condition. The values were higher than the production rates of the order of 10 −9 M s −1 under the condition of pulsed discharges on the surface of liquid (Kanazawa et al. 2011) and 1.67 × 10 −8 M s −1 for the direct discharge in water at an applied voltage of 45 kV and input power delivered to the water of 64 W (Sahni and Locke 2006). Shen et al. (Shen et al. 2019) have estimated •OH production which were on the order of 1.2 × 10 −8 M s −1 and 2.2 × 10 −8 M s −1 for the air and oxygen gas-liquid plasma treatment, respectively. As a matter of fact, power supply, applied voltage, plasma device, liquid conductivity, and plasma instability would have a significant impact on the •OH generation (Yue et al. 2022). After comprehensive comparison, the production rate of •OH in this work was similar to that in these researches. The measured •OH concentrations decreased slightly after 4 min of plasma discharge treatment. When the discharge time reaches 4 min, there may not be sufficient TA to react with the •OH radical to transform into HTA at a TA initial concentration of 0.2 mM. In addition, the decomposition of HTA due to the reaction with • OH radicals and other species must be considered, thereby leading to a decreasing •OH yield after 4 min (data not shown). It is noteworthy that the measured •OH production rate in the plasma /Fe 2+ (100 mg/L) system was always kept in a low level of the order of 2.75 × 10 −9 M s −1 , and the reason may be that the reaction probability between Fe 2+ and •OH was higher than that of TA and •OH due to the significant difference in concentration between Fe 2+ and TA. Due to the reaction between •OH and Fe 2+ , •OH might be scavenged by Fe 2+ (Eq. (31)) (Abou Dalle et al. 2017;Lodha and Chaudhari 2007), thus leading to a low •OH measured value in plasma/ Fe 2+ (100 mg/L) system. Figure 4b depicts the variations of H 2 O 2 under different treatment conditions. The aqueous H 2 O 2 concentration increased markedly from 0 to 2.37 mM, 0.1 to 2.85 mM, 0 to1.45 mM for the sole plasma, plasma/H 2 O 2 (0.1 mM), plasma/Fe 2+ (100 mg/L) systems during 0-12 min, and later changed to 2.64 mM, 3.76 mM, and 1.58 mM at 48 min of treatment. The H 2 O 2 concentration measured in the sole plasma system was higher than that reported in these researches, nearly 60 ppm (1.76 mM) at 18 kV after 25 min of plasma treatment (Manoj Kumar Reddy et al. 2013), and 15 mg/L (0.44 mM) in underwater parallel-multi-tube air discharge plasma jet system (Rashid et al. 2020). The H 2 O 2 concentration more than 120 mg/L (3.52 mM) reported in Sang et al. (Sang et al. 2019b) was higher than that achieved in the present paper. However, the value was obtained using a relatively small aqueous solution (10 mL), while it is 40 mL in this work. In fact, due to the differences in experimental conditions and environments, plasma devices, and power supplies, the concentration of reactive species would have been quite different. The H 2 O 2 concentration was produced by plasma mainly due to the high-energy electrons impact O 2 to produce oxygen atoms and later react with water to form H 2 O 2 (Eq. (6) and Eq. (11)) (Hayashi et al. 2016). In addition, the combination of •OH was also an aspect of enhancement of H 2 O 2 concentration (Eq. (10)). As shown in Fig. 4b, when the treatment time exceeded 12 min, the H 2 O 2 concentration remains substantially unchanged in the sole plasma system. Significantly, the H 2 O 2 concentration of the plasma/H 2 O 2 (0.1 mM) system was much higher than that in the other two systems during the whole discharge period. This phenomenon is probably due to the additional H 2 O 2 that may accelerate and enhance the formation of various reactive species, which was mentioned by Liu et al. (Liu et al. 2016). The production of these reactive species may promote the increase of H 2 O 2 to a certain extent. However, the specific reasons still require further study. Furthermore, for the plasma/Fe 2+ (100 mg/L) system, as H 2 O 2 had been continuously generated by plasma and the relatively low reaction efficiency for the Fenton reaction (Eq. (30)) due to the nearly neutral solution pH in the early term (Feng et al. 2006), hence the aqueous H 2 O 2 concentration increased steadily. It is worth mentioning that as with declining solution pH, the H 2 O 2 generated by plasma may not be adequate to meet the demand of the rapidly enhancing reaction efficiency of the Fenton reaction (Xu et al. 2020), thus resulted in a higher consumption of the accumulated H 2 O 2 . Consequently, the H 2 O 2 concentration declined from 1.91 mM (24 min) to 1.58 mM (48 min). Figure 4c exhibits the change of aqueous O 3 concentration under different treatment conditions. The O 3 concentration, respectively, increased from 0 to 0.59 mg/L, 0.052 to 1.30 mg/L, 0 to 0.21 mg/L after 48 min of treatment for the sole plasma, plasma/H 2 O 2 (0.1 mM), plasma/Fe 2+ (100 mg/L) system. The concentration of O 3 by plasma/ H 2 O 2 (0.1 mM) treatment in water was the highest compared with the other two groups. The probable cause was the dissolved O 2 reacted with ·O generated from the additional H 2 O 2 decomposition to form more O 3 (Eq. (8) and Eq. (22)) (Reddy and Subrahmanyam 2012). For the plasma/Fe 2+ (100 mg/L) system, although O 3 could be continuously generated by plasma, it is worthy to mention that Fe 2+ would consume a portion of plasma-generated O 3 to produce Fe 3+ (Eq. (23)) (Huang et al. 2019;Malik et al. 2019). The consumption could mainly account for the lower aqueous concentration of O 3 compared with that from the sole plasma system.
As illustrated in Fig. 4d, the aqueous NO 3 − concentration increased steadily from 0 to 105 mg/L, 0 to 151.3 mg/L, 0 to 135.6 mg/L for the sole plasma, plasma/H 2 O 2 (0.1 mM), plasma/Fe 2+ (100 mg/L) system after 48 min of discharge treatment. The aqueous NO 3 − was mainly generated by plasma discharge, and it kept increasing during the whole treatment period for all systems. Specifically, in the plasma/ H 2 O 2 (0.1 mM) system, the added H 2 O 2 would react with NO and NO 2 generated by plasma to promote the formation of NO 3 − (Eqs. (24) and (25)) thus leading to the highest aqueous NO 3 − concentration in the late period. Notably, in the plasma/Fe 2+ (100 mg/L) system, the aqueous NO 3 − concentration was higher compared with the sole plasma system. When air was bubbled, nitrogen molecules around the discharge electrode would be excited and decomposed into active nitrogen species. Due to the Fenton reaction (Eq. (30)) produced more •OH, active nitrogen species could react with •OH to generate NO 3 − according to Eqs.(26-29) (Bian et al. 2009), thus promoted the measured NO 3 − concentration. In present work, the measured NO 3 − concentrations were lower than the detected NH + 4 −N and NO − 3 −N of 316.53 mg/L in the residual atrazine solution (Zhu et al. 2014) and higher than that of 1.96 mM (66.6 mg/L) with bubbling air and 2.34 mM (79.5 mg/L) with bubbling pure nitrogen under the condition of pulsed high-voltage discharge for 30 min (Bian et al. 2009). It must be emphasized that relatively high concentration of nitrate was formed during plasma treatment in air atmosphere, thus leading to a relatively low pH value, even though a certain contribution to the pH variation during the plasma treatment may come from the N-containing heterocyclic structure of CBZ. However, the degraded solution with low pH cannot be directly discharged into the environment without secondary treatment. How to reduce the secondary problems produced in the process of CBZ degradation still requires further studies, using oxygen enriched air (Bian et al. 2009), adjusting energy density (Dojcinovic et al. 2011), shortening the processing time, and adopting a lower gas flow may play a role.

Effect of discharge power on CBZ degradation
The discharge power is considered a significant parameter because it is directly related to the production, transmission, and diffusion of reactive species, which plays a vital role in pollutant degradation. Figure 5a presents the degradation of CBZ under various discharge power intensities. With regard to a discharge time of 48 min, the maximum degradation efficiencies were 90.1% and 62.5% in 35.7 W and 20.9 W, respectively, while the CBZ in aqueous solution was almost completely degraded under the condition of 56.4 W. The main reason was that more reactive species was generated with the increase of discharge power, which could lead to the complete degradation of CBZ in liquid phase. Figure 5b displays the energy efficiency under various discharge power intensities during the degradation process. It can be seen that the energy yield declined over time at different discharge power. As shown in Fig. 5b, the degradation of CBZ decreased from 29.9 mg•kW − 1 h −1 (20.9 W) to 17.7 mg•kW − 1 h −1 (56.4 W) after 48 min. A possible reason for this phenomenon is that an excessive input power would make part of the energy converted into heat in the system, resulting in a waste of energy (Zhang et al. 2017). Although the energy efficiency at 20.9 W was higher than the other two discharge powers, the degradation rate of CBZ was still at a relatively low level. Furthermore, the energy efficiency at discharge power of 35.7 W is higher than that at 56.4 W during the whole discharge period. In order to achieve a relatively optimized balance of degradation efficiency and energy efficiency in the CBZ degradation process, 35.7 W was selected as the discharge power in subsequent experiments to pursue a higher CBZ removal rate and economic benefits.

Effect of initial concentration on CBZ degradation
The variations of the plasma-treated CBZ solution during the discharge period with different initial concentrations from 10.0 to 40.0 mg/L are displayed in Fig. 6. The volume of the plasma-treated solution was 40.0 ml without pH adjustment (Initial pH = 6.40). In this test, the discharge power was set at 35.7 W. It can be seen from the figure that with the increase of the initial concentration of CBZ solution, the removal efficiency of CBZ decreased at the same discharge time, and the result was similar to that of Zhu et al. (Zhu et al. 2014). A possible cause of this phenomenon is that the RS produced by discharge is at a particular level under the condition of stable discharge. At a high concentration of CBZ, CBZ molecules had lower chances to be attacked by RS (Batista et al. 2014). Meanwhile, the intermediates also consumed a part of RS and increased competition of CBZ for RS, thus leading to a drop in the efficiency of CBZ degradation under the high concentration condition (Guo et al. 2015). Significantly, the middle initial concentration of 20.0 mg/L was selected in later experiments.

Effect of initial pH value on the removal rate of CBZ
The initial pH value is a crucial factor for the removal of contaminants. Thus, different initial pH conditions (pH = 3.85, 9.05 adjusted by HCl or NaOH) and original pH (pH = 6.40, no pH adjustment) were selected to estimate the CBZ degradation. The pH variations of the plasma-treated CBZ solution during the discharge period with different initial solution pH was also investigated. Figure 7a shows the pH variations of CBZ solutions with different pH at an initial concentration of 20.0 mg/L when the volume of the plasmatreated solution and discharge power was set at 40.0 ml and 35.7 W, respectively. The solution pH declined rapidly in the first 18 min of treatment and then decreased by degrees, finally stabilizing at about pH = 2.0 for each initial pH value. Consequently, the solution pH presented a continuous downward trend for each initial pH value, this might be due to the effect of inorganic acids such as HNO 3 and HNO 2 produced during the discharge process (Dojcinovic et al. 2011).
The removal rates of CBZ at various initial pH values are shown in Fig. 7b. It can be seen that the removal efficiency of CBZ increased when the initial conditions were acidic (pH = 3.85) and alkaline (pH = 9.05), and both reached up to 98% after discharge for 48 min. The initial pH of the solution will affect the generation of RS in the discharge process (Joshi and Thagard 2013). There is a higher degradation efficiency in alkaline conditions because O 3 and H 2 O 2 were unstable in alkaline environments (Feng et al. 2006), hence they could be rapidly decomposed into •OH in aqueous solution ), more •OH was generated. As one of the reactive species, ·OH increased the degradation rate of CBZ. In the acidic solution, the concentration of dissolved ozone increased with the decrease of pH value (Sotelo et al. 1989), which played a certain role in its direct reaction with CBZ. Nevertheless, the oxidation ability of O 3 is weaker than •OH (Zhang et al. 2012), which resulted in a lower degradation of CBZ compared with an alkaline condition. By the way, an original pH without adjustment (pH = 6.40) was chosen in subsequent experiments due to the not particularly significant impact of pH.

Effect of H 2 O 2 on CBZ degradation
In order to investigate the effect of H 2 O 2 on CBZ degradation by plasma, H 2 O 2 was added as the additive at variant concentrations of 0.05, 0.1, 2.0, and 5.0 mM in the aqueous solution, and the results are shown in Fig. 8. Obviously, the degradation rate of CBZ could be accelerated or inhibited in the presence of H 2 O 2, depending on the applied dosage. The results indicated that the degradation value of CBZ increased with the addition of low concentrations of H 2 O 2 (0.05-0.1 mM). However, it hindered the degradation with the higher aqueous H 2 O 2 concentration (2.0-5.0 mM). Specifically, the CBZ degradation rate in the sole plasma system was 53.3% and increased to 65.4% in the plasma/ H 2 O 2 (0.1 mM) system at the discharge period of 18 min. On the contrary, when the added H 2 O 2 concentration was enhanced to 5.0 mM, the degradation rate declined from 53.3% to 37.5%. The reason why CBZ degradation could be improved when H 2 O 2 was added at the appropriate concentration might be explained by Eqs. (17) and (18)

Effect of Fe 2+ on CBZ degradation
Plasma could trigger the generation of H 2 O 2 in the liquid, so Fe 2+ was added to the treated solution to study its effect on degradation of CBZ, which involved the following reactions (Domínguez et al. 2012;Xu et al. 2022a, b): Figure 9 displays the effect of adding different concentrations of Fe 2+ in the aqueous solution combined with plasma treatment on the degradation efficiency of CBZ. It was indicated that Fe 2+ benefited the degradation at the concentration of 10-100 mg/L. However, the enhancement effect was reduced at a higher concentration from 100 to 200 mg/L. Specifically, the CBZ degradation rates in the sole plasma system 18 and 48 min were 53.2% and 90.1%, respectively, it sharply increased, respectively, to 73.3% and 99.1% with the addition of Fe 2+ in a concentration of 100 mg/L. Although the plasma process can generate •OH, it still exhibited restricted penetration depth into the liquid thus limited its performance (due to the short lifetime of •OH radical). The H 2 O 2 generated by the plasma has a long lifetime and can reach a relatively far position in solution by diffusion, the addition of Fe 2+ can trigger the catalytic degradation of heterogeneous Fenton reaction (Eq. (30)) with plasma-induced H 2 O 2 to generate •OH (Abou Dalle et al. 2017) at the position further away from the plasma discharge reaction area, which can quickly and non-selectively degrade organic pollutants, thus leading to an improved CBZ degradation rate. Furthermore, the produced Fe 3+ could be reduced to Fe 2+ by reacting with the aqueous H 2 O 2 provided by plasma discharge exposure (Eq. (31)), thereby maintaining the Fenton reaction and constantly degrading CBZ. However, the degradation efficiency declined after adding Fe 2+ into the solution to 200 mg/L compared with 100 mg/L. The lower enhancement of degradation rate at higher Fe 2+ concentration can be explained by the scavenging effect of oxidation reaction of Fe 2+ , which would result from the competition between excessive Fe 2+ and the organic substance for •OH (Eq. (32)). The •OH consumption increased, hence the enhancement effect of CBZ degradation was accordingly weakened. Through comparing the degradation rate of CBZ for plasma/Fe 2+ (100 mg/L) and plasma/H 2 O 2 (0.1 mM) system, plasma/Fe 2+ (100 mg/L) system could achieve a higher degradation rate on CBZ, thus the plasma/ Fe 2+ (100 mg/L) system was utilized in following UPLC-MS analysis.

Effect of plasma/catalyst system on TOC removal
In addition to the degradation rate, the total concentration of organic pollutants in aqueous solution can be exhibited through the removal of the TOC value, which could also reflect the mineralization efficiency of the contaminants in the solution. The mineralization rate of CBZ solution was Notably, the solution TOC was reduced by 48.0% after 48 min of the sole plasma discharge treatment, meanwhile, plasma/H 2 O 2 (0.1 mM) and plasma/Fe 2+ (100 mg/L) system resulted in a 58.7% and 64.1% TOC diminution, respectively. The mineralization rate in the plasma/Fe 2+ (100 mg/L) system was obviously higher than that in the plasma/H 2 O 2 (0.1 mM) system, which were in accordance with the tendency on the CBZ degradation rate. These results indicated that plasma/catalyst treatments could directly carbonize a portion of contaminants in the liquid phase and high mineralization rates on CBZ could be achieved. Nevertheless, certain intermediates in the processing of CBZ degradation still could not be further mineralized due to the stable amount of yielded reactive species under the constant experimental condition.

Proposed degradation pathways of CBZ
The primary degradation intermediates of CBZ were detected by UPLC-MS to better understand the degradation mechanism. Based on the mass spectra, ten main degradation intermediates of CBZ by gas-liquid plasma are identified and shown in Table 2. The possible degradation pathways were proposed, as shown in Fig. 11. The olefinic double bond with high frontier electron density on the central heterocyclic ring of CBZ molecule was susceptible to the attack of reactive species, leading to the formation of epoxidized product M1 (m/z 253), which was one of the most frequently detected intermediates in related researches (Li et al. 2013;Miao et al. 2005). M1 (m/z 253) could be further confirmed to be 10, 11-epoxy carbamazepine according to the further hydrolysis product of M2(m/z271) (De Laurentiis et al. 2012;Yang et al. 2017), whose further transformation product was M4(m/z 226). Meanwhile, a direct dehydrogenation reaction occurred at the sites of double •OH groups in M2, thus leading to the formation of compound M5(m/z267), which could further transform to compound M8 (m/z 224) and M7(m/z 196) due to the cleavage of aldehyde group and amide group in a certain order (Wang et al. 2019a). In addition, a facile ring contraction and amide group loss process of M1(m/z 253) led to the formation of M3(m/z 208). The same compound was identified and reported earlier in chlorine dioxide(ClO 2 ) oxidation of the CBZ molecule (Kosjek et al. 2009 Zheng et al. 2014).
Although organic substance TOC removal rate exceeded 67% in present system, complete mineralization of pollutants cannot be achieved. Recently, Leong et al. (Leong et al. 2021) found that adapted mixed cultures in batch and continuous operating systems had pretty well performance in triclosan biodegradation and mineralization aspect at a large range of initial concentrations. Moreover, a combined process of non-thermal plasma and mineral adsorber showed good performance levels for VOC removal and odor abatement (Dobslaw et al. 2018). Those studies indicated that the combination of various technologies may work better to realize the complete mineralization of refractory organics in future research.

Biological toxicity evolvement during CBZ degradation
The inhibition index (I ind ) of E. coli could reflect the microbial viability thus signifying the variations in biological toxicity during the treatment of plasma/Fe 2+ (100 mg/L) for different periods. Figure 12 demonstrates the I ind of CBZ solutions degraded for different times for E. coli. In the inhibition test, the blank tests of pH = 2 and 135 mg/L NO 3 showed very low inhibition for the growth of E. coli, and 100 mg/L Fe 2+ had almost no inhibitory effect. It needs to be emphasized that the blank test samples were only 0.5 mL in the test procedure, 0.5 mL blank test sample, and Fig. 10 The variations of TOC removal efficiency in different systems 50 µL E. coli would mix with 3 mL of sterilized LB medium, which diluted the concentration/acidity of test sample in the system and reduced the effect of 100 mg/L Fe 2+ , pH = 2 and 135 mg/L NO 3 to E.coli to a certain extent. As shown in Fig. 12, the inhibition index (I ind ) in a short period of time gradually decreased from 21.1% (0 min) to 5.8% (12 min), and then increase to 11.1% (18 min). This slight increase might have been due to the production of some small molecular-weight intermediates but with higher toxicity on E. coli at the early stages of CBZ degradation. Specifically, Table 2 Information and proposed structure of the degradation intermediates from CBZ the intermediates Acridine (M6) and Acridone (M7), which were the transformation and oxidation products of CBZ, may be more toxic than the parent compound thus strongly promoting the inhibitory effects (Donner et al. 2013). Notably, when the treatment time exceeded 24 min, the inhibition index (I ind ) declined and then stabilized at around 6.5%. According to the previous TOC removal results, more than half of the organic substances had been mineralized and those degradation intermediates with high toxicity had also been further destructed, thus leading to the reduction on the inhibition rate. Although the biological toxicity of the plasma/Fe 2+ (100 mg/L)-treated CBZ solution fluctuated in the whole treatment process, it exhibited an overall downward trend and a relatively low solution toxicity could be achieved. Therefore, it may be confirmed that plasma provides an important pathway for CBZ toxicity reduction despite some non-negligible intermediates. Nonetheless, the secondary effects such as high levels of NO 3 and low pH caused by plasma process and residual Fe 2+ are still worthy of further study and explore.

Conclusion
In summary, the degradation effects and related mechanisms on CBZ in aqueous solution by a gas-liquid plasma combined with different catalysts (H 2 O 2 and Fe 2+ ) were investigated. It is proved by experimental results that CBZ in aqueous solution can be effectively degraded by the plasma/catalyst treatment. During the discharge, RS (•OH, H 2 O 2 , O 3 , etc.) were produced to degrade aqueous organic pollutants. The parameters that affect the degradation of CBZ, such as discharge powers, initial concentrations, pH values, and addition of catalysts were all affected the CBZ degradation efficiency. Specifically, plasma combined with suitable aqueous concentrations of H 2 O 2 (0.1 mM) and Fe 2+ (100 mg/L) could lead to high removal rates for more than 97.9% and 99.1%, respectively, at 48 min with the highest energy efficiency at discharge power of 35.7 W. The TOC removal efficiency of CBZ was up to 67.1% in the plasma/Fe 2+ (100 mg/L) system at 48 min, which suggested a high mineralization efficiency. Moreover, based on various degradation intermediates identified by UPLC-MS, the proposed degradation mechanism on CBZ by plasma/Fe 2+ treatment was investigated. Cleavage of azepine ring, hydroxylation, and deamidation were considered the primary degradation pathways. The biological toxicity of the plasma/Fe 2+ (100 mg/L)-treated CBZ solution fluctuated and finally declined, and a relatively low solution toxicity could be achieved. Therefore, this work illustrated the promising potential of using plasma/catalyst systems as a highly effective method in removing pharmaceuticals in aqueous solutions.