Influence of solution electrical conductivity and ionic composition on the performance of a gas–liquid pulsed spark discharge reactor for water treatment

In plasma-based water treatment (PWT), an electrical discharge splits water and background gas molecules or atoms into a range of reactive species that then react with organic and inorganic contaminants causing their degradation. The ability of the electrical discharge plasma to form both oxidative and reductive species in situ without chemical additives makes PWT an attractive, and for some applications, a superior water treatment technology.^1–10 As an example, Stratton et al.⁵ showed that free and solved electrons as well as argon species are responsible for the removal of toxic poly- and perfluoroalkyl substances (PFAS), rendering PWT the most effective technology for the removal of these compounds available today.

When PWT is applied for the treatment of groundwater, wastewater, membrane concentrate, or ion exchange brine, the physicochemical properties of those solutions, in particular, solution electrical conductivity, impact the plasma reactor performance.^11–16 Solution conductivity has been shown to control the plasma electron density, the area of the plasma, and the intensity of the plasma-generated UV light,^8,10,17–21 although these relationships appear to be heavily influenced by the gas composition, pulse shape, and electrode geometry. Plasma chemistry and thus the reactive species production have also been investigated in relation to solution electrical conductivity. For direct-in-liquid discharges, spectroscopic studies have shown that the emission intensities of OH, H, and O increase with conductivity from 5 to 150 μS/cm and then decrease with further increase in conductivity up to 1000 μS/cm due to increased quenching of radicals at higher plasma temperatures and higher ion concentrations in the plasma.^{14,16,19,22,23} For gas discharges contacting water, Thagard et al.¹⁷ demonstrated that both an increase in the solution conductivity from 50 to 2000 μS/cm and an increase in the solution pH from 3.5 to 10.5 lower the generation rate of hydrogen peroxide (H₂O₂). The conductivity effect was not attributed to the UV light-induced decomposition of hydrogen peroxide, as assumed in the literature,¹⁰ but to the changes in the plasma chemistry that accompany changes in the solution conductivity and the pH effect on the bulk liquid reactions between plasma-generated radicals and hydroxide ions.

Ions present in the solution can influence the plasma and bulk liquid chemistry both directly and indirectly. Directly, halogen ions such as chloride, bromide, and iodide ions react with OH radicals to form various oxygenated species such as chlorate or bromate ions.^11,24–29 The intermediate species of these chain reactions, including hypochlorous acid (HOCl), Cl^•, and $Cl2∙−$⁠, have been hypothesized to be involved in the degradation of many organic contaminants.^27,28 Indirect influences involve a reaction between a metal ion such as ferrous ion and plasma-generated H₂O₂ to form OH radicals and have been shown to improve degradation rates of phenol, 4-chlorophenol, and ibuprofen,^30–32 among other compounds.

Our recent work in which we plasma-treated short- and long-chain PFAS in ion exchange (IX) regenerant still bottom solutions with conductivities ranging from 52 to 80 mS/cm (mainly due to chloride ions) showed that the chloride ion concentration controls the treatment effectiveness by influencing the interfacial physical and chemical processes.³³ This study builds on our prior work by investigating the impact of solution conductivity, up to 45 mS/cm, and its ionic composition on the removal of perfluorooctanoic acid (PFOA) and Rhodamine B dye in a gas discharge reactor with a point–point electrode geometry. PFOA was chosen as a model compound due to its environmental importance and its tendency to accumulate at gas–liquid interfaces. However, because PFOA cannot be oxidized, Rhodamine B was used to investigate the impact of the treatment on oxidative reactions. The work compares the degradation rates of the two compounds at a range of solution electrical conductivities of different ionic compositions and attempts to explain the observed results by examining the discharge characteristics, solution surface tension, and the production rates of short- and long-lived reactive species using a point–point discharge propagating over the surface of water in argon gas. The results of this study are expected to inform future plasma reactor design for treating high electrically conductive solutions.

Sodium chloride (NaCl, reagent grade ≥99.0%), potassium chloride (KCl, ≥99.0%), hydrochloric acid (HCl, 12N), calcium chloride (CaCl₂, ≥99.0%), barium chloride (BaCl₂, ≥99.0%), sodium fluoride (NaF, ≥99.0%), sodium nitrate (NaNO₃, ≥99.0%), potassium nitrate (KNO₃, ≥99.0%), magnesium nitrate [Mg(NO₃)₂, ≥99.0%], calcium nitrate [Ca(NO₃)₂, ≥99.0%], water (HPLC grade), methanol (HPLC grade), and perfluorooctanoic acid (PFOA, ≥96%) were purchased from Fisher Scientific and were used as received. A labeled internal standard for PFOA was purchased from Wellington Laboratories (Guelph, ON). Lithium chloride (LiCl, ≥99.0%), magnesium chloride hexahydrate [MgCl₂(H₂O)₆, ≥99.0%], dimethyl sulfoxide (DMSO, ≥99.7%), and titanium (IV) oxysulfate in sulfuric acid (27%–31%) were purchased from Sigma-Aldrich and were also used as received.

Experiments were performed in a cubic reactor with a characteristic length of 14 cm, constructed of clear acrylic with a lid fitted to allow liquid recirculation, gas inlet, and electrode integration (Fig. 1). The electrodes consisted of two tungsten rods (d = 1/8 in.) sharpened to a point and positioned in the headspace gas 0.2 cm above the liquid surface, unless stated otherwise, and 2.2 cm apart from each other thereby forming a point–point electrode configuration with a constant plasma–liquid contact area. For some experiments, both electrodes were raised to 2.5 cm above the liquid surface to prevent the plasma from contacting the liquid. The headspace was continuously purged with argon at 4.3 l/min. All experiments were performed in a semi-batch mode with a liquid recirculation rate of 2.0 l/min and a liquid depth of 3.0 cm. In order to maintain a constant liquid temperature of 10 °C, a heat exchanger was integrated into the liquid recirculation loop unless otherwise stated.

The treated solution consisted of 300 ml of de-ionized water containing 21 μM of Rhodamine B, 21 μM of caffeine, or 21 μM of PFOA with NaCl, LiCl, KCl, HCl, MgCl₂, CaCl₂, BaCl₂, NaF, NaNO₃, KNO₃, Mg(NO₃)₂, or Ca(NO₃)₂ to adjust the solution conductivity between 0.3 and 45 mS/cm. For experiments performed at a fixed conductivity of 45 mS/cm at different pH values, HCl and NaCl were added to adjust the pH. Hydroxyl radical scavenging experiments were performed by adding 5 vol. % DMSO to solutions containing 21 μM Rhodamine B and various salts at conductivities between 1.2 and 45 mS/cm.

Figure 2 shows a schematic of the reactor used to determine the influence of plasma-generated light on hydrogen peroxide generation and Rhodamine B removal. One 3.5 ml quartz cuvette, which is optically transparent for wavelengths >170 nm, and one 3.5 ml plastic cuvette, transparent for wavelengths >400 nm, were placed 3 cm away from the plasma location. Experiments were performed with de-ionized water, 1.5 mM hydrogen peroxide solution at ∼0 mS/cm, or 2.1 μM Rhodamine B-NaCl solutions with conductivities of ∼0, 1.2, and 45 mS/cm in each cuvette.

The resistances of NaCl solutions were measured by applying a potential of 28 V with an Eventek KPS305D power supply terminated with 0.25 mm diameter platinum wires (Thermo Fisher Scientific, Standard Grade). The platinum wires were submerged 0.5 cm into the solution 2.2 cm apart. The voltage and current were measured using a Tektronix MDO 3032 oscilloscope with a Tektronix TPP0500B voltage probe and a Tektronix TCP0030A current probe. The resistance was calculated using Ohm's law.

A Spellman SL30P1200 high voltage power supply was used with a custom-built circuit, shown in Fig. 3, to generate pulsed plasma. The system was operated using a 2 nF load capacitor that was charged to −30 kV through a 3 MΩ charge resistor and discharged at 40 Hz. The voltage and current in the plasma reactor were measured using a North Star PVM-1 high voltage probe and a Tektronix TCP0150 current probe connected to a Tektronix MDO 3032 oscilloscope. Example current and voltage waveforms are shown in Fig. 4. The oscillation in the voltage and current is the result of the underdamped nature of spark discharges. Both energy per pulse and the inception voltage were independent of the salt type used to adjust the solution conductivity. Similarly, the pulse duration was independent of the conductivity.

The concentration of PFOA was determined using a UPLC-MS-MS (Thermo Scientific, Vanquisher-TSQ ALTIS) triple quadrupole system in the negative ionization mode equipped with a Phenomenex Luna Omega column (2.1 mm × 100 mm, 1.6 μm). Samples were diluted with 75% methanol and 25% water, then sonicated and centrifuged prior to injection (10 μl). For quality assurance (QA)/quality control (QC), all samples were spiked with 2 ng of labeled internal standard. Eight-point calibration in the range of 9 and 2000 ng/l was used for the quantification using C-13 isotopic dilution or internal standard methods. Detection limits were approximately 9 ng/l. A detailed description of the analytical methods, QA procedures, and individual detection limits can be found in our previous work.³³ The concentration of Rhodamine B was determined spectrophotometrically (Shimadzu UV-1800) by measuring its absorbance at 554 nm. Caffeine concentration was measured using high performance liquid chromatography (HPLC, Thermo Scientific Accela 3000, Texas, USA) with a C18 HPLC column (4.6 mm × 150 mm, 4 μm, Phenomenex) and a detection wavelength of 273 nm. The mobile phase consisted of acetonitrile and water with a linear solvent gradient from 5% acetonitrile to 25% acetonitrile at a flow rate of 1 ml/min. Chlorate ion concentration was measured using an ion chromatograph (Dionex Integrion HPIC) equipped with a Dionex IonPac AS16 RFIC (4 × 250 mm²) column and a conductivity detector. A 35 mM sodium hydroxide solution was used as the eluent with a constant flow of 1 ml/min. H₂O₂ concentration was determined spectrophotometrically by measuring the absorbance of the complex formed by the reaction of H₂O₂ and titanium sulfate at 410 nm.³⁴ Surface tension measurements were carried out using the pendant drop method with a DataPhysics 15plus tensiometer. The solution was dispensed from an electronically controlled syringe through a needle with a diameter of 1.65 mm to form a pendant drop. The surface tension was then calculated using the Young–Laplace equation using DataPhysics SCA 20 software. Three individual drops were measured for each sample. Conductivity and pH were measured using a ThermoScientific Orion 4 Star meter. Optical emission spectroscopy (OES) measurements were performed by placing an optical fiber connected to an Avaspec Multichannel Fiber Optic Emission Spectrometer (Avantes, AvaSpec-ULS2048) inside a quartz tube 3 cm above the discharge location. Spectra were recorded from 190 to 1100 nm with a 50 ms integration time.

For the point–point reactor electrode configuration used (Fig. 1), the discharge is forced to propagate between the two electrodes over the liquid surface, which has multiple advantages when examining the influence of solution conductivity on plasma characteristics. The fixed distance between the electrodes in the gas phase allows for a constant plasma–liquid contact area. Normally, changing solution conductivity results in changes in the plasma–liquid contact area due to shorter propagation of plasma streamers, which influences many of the physicochemical properties of the discharge.^12,35–38 Additionally, because the discharge occurs in the gas phase, there is no influence of the solution conductivity on the inception voltage of the breakdown. However, this was only true for the interelectrode distance of up to 2.2 cm, which is the value used in this study. At distances greater than 2.2 cm, the conductive channel between the two electrodes failed to be established. In that case, the liquid served as a conductive path between the high voltage and the ground electrode resulting in two corona-like discharges occurring simultaneously from the two electrodes to the liquid surface. By fixing the electrode gap to 2.2 cm, the discharges at all conductivities explored in this work formed spark discharges over the liquid surface, which acted as the primary conductive path from high voltage to ground, allowing for the analyses of process characteristics without having to consider changes in the plasma volume or plasma area.

Table I shows the energy per pulse (EPP), calculated by integrating the product of the instantaneous voltage and current measurements [Eq. (1)], delivered to the reactor as a function of solution conductivity. The EPP measurements were performed six times and the error was determined as shown in Table I,

The observed increase in the delivered energy with conductivity has important implications for the plasma reactor performance as it has been shown to be accompanied by the changes in the plasma temperature, electron density, production rates of reactive species, and the intensity of plasma-generated shockwaves and (V)UV light emissions.^{23,38–41,42} The same phenomena may occur for the system presented in this study, but it is important to note that in the majority of the previously investigated and published gas–liquid plasma systems, the plasma and the water were placed in a series forcing the total discharge current to flow across the plasma–liquid boundary. Additionally, many of these studies used nanosecond duration pulses. In this work, the plasma and the solution are arranged in parallel allowing a portion of the (microsecond) discharge current to remain entirely in the plasma and not to cross the boundary. This will influence the magnitude and the means by which solution conductivity impacts various plasma phenomena as well as energy distribution in the plasma reactor between the two phases.

To estimate the fraction of the EPP converted into heat, the temperature of the solution was measured as a function of the solution conductivity. For those experiments, the reactor was sealed to prevent any gas from entering or leaving the system and the liquid recirculation system was removed to minimize heat losses and maximize temperature changes. The reactor was operated for 10 min during which changes in the liquid temperature were measured. Assuming a constant heat capacity C_p for water at 20 °C, the energy transferred into the solution as heat (Q) during the 10-min interval was calculated using

The energy converted to heat per pulse (EPP_h) was calculated by dividing Q by the total number of pulses in 10 min.

As solution conductivity increased, the measured EPP in the reactor also increased (Table I). It is unclear what causes this increase in EPP; however, this energy does not appear to be used to increase the chemical reactivity of the plasma but is simply wasted as heat (EPP_h).

The implication of these findings on plasma and bulk liquid chemistry was assessed by measuring the concentration of H₂O₂ produced by the plasma and qualitatively examining the spectral intensities of different plasma-generated reactive species. The rate of H₂O₂ production was measured in the liquid as a function of solution conductivity for four salts (electrolytes). As shown in Fig. 5, the production of H₂O₂ decreases with conductivity, regardless of the salt type. Although conductivity appears to be an important factor influencing the production rate of hydrogen peroxide, conductivity-controlled secondary processes can affect hydrogen peroxide production for discharges in gases and liquids. These secondary processes include a decrease in the plasma–liquid contact area, which decreases the available recombination volume for OH radicals (not applicable to this work as the reactor was designed to maintain a constant plasma–liquid interfacial area), electrode material sputtering into the reactor, direct chemical reactions with inorganic ions, OH, H, solvated electrons, UV light, plasma temperature, and plasma electron density.^{12,23,39,40,43} Many of those cited studies reported a decrease in the production rate of H₂O₂ with increasing solution conductivity similar to what was seen here.

In this work, electrode erosion is expected to be insignificant due to the low rate of electrode erosion for a gas discharge, especially for the lanthanated tungsten used, which is a high melting point electrode material. The possibility of direct chemical reactions with anions influencing the H₂O₂ chemistry was examined and determined to be unimportant by comparing the hydrogen peroxide production rates in the presence of chloride ions (which react with OH radicals at near-diffusion controlled rates²⁵), fluoride ions (which do not react with OH radicals), and nitrate ions (which react with solvated electrons also at near-diffusion controlled rates⁴⁴). The minor role of cations in the production of hydrogen peroxide was confirmed by comparing H₂O₂ production rates in sodium chloride and magnesium chloride solutions, where they were found to be equal. Based on these findings, the reduced H₂O₂ production at higher conductivities was likely due to either a direct decomposition of hydrogen peroxide by plasma-generated UV light or a change in the concentration/availability of OH radicals, the main H₂O₂ precursors, produced by the plasma.

The influence of plasma-generated light on the H₂O₂ concentration was examined by placing a 1.5 mM H₂O₂ solution in a quartz cuvette, which transmits light above 170 nm, 3.0 cm away from the discharge location (Fig. 2). As a control, a de-ionized water sample was placed in the second cuvette at the same distance. In those cuvettes, H₂O₂ was neither being formed (cuvette with the de-ionized water) nor destroyed (cuvette with pre-prepared H₂O₂) after 60 min of treatment (Fig. 6).

Figure 7 shows OES spectra recorded for aqueous NaCl solutions at 0.6, 12, and 45 mS/cm. For clarity, individual spectra are shown in the supplementary material (Figs. S1–S3). Table II lists the observed transitions of the key identified species.

Inspection of the raw spectroscopic data for all solution conductivities revealed the well-known presence of atomic oxygen and atomic hydrogen lines along with numerous emissions from argon atoms. The intensities of O, H, and Ar lines were influenced by solution conductivity but due to the challenges associated with spectroscopic analyses of atmospheric pressure plasmas^46,47 their observed changes were not utilized to infer the properties of the plasma. As a result, the effect of the conductivity on the plasma and electron temperature and electron density was not determined in this work. Gasanova² observed no changes in the plasma temperature for in-liquid discharges between 0.5 and 2 mS/cm and measured an increase of 3.5 times in electron density when increasing solution conductivity from 0.5 to 10 mS/cm. Wang et al.²¹ also reported a minimal change in the plasma temperature for a gas discharge contacting a thin liquid film up to 36 mS/cm but reported a nearly constant electron density for solution conductivities up to 28 mS/cm that was accompanied by its significant increase from 28 to 36 mS/cm. Unfortunately, neither Gasanova² nor Wang et al.²¹ were able to correlate the observed changes in the electron density to the generation of H₂O₂ due to concurrent changes in the plasma–liquid contact area with conductivity.

Spectroscopic analysis revealed that the emission from OH at 309 nm was extremely weak, and its intensity did not appear to change with the conductivity. As a result, the observed decrease in the H₂O₂ production rate cannot be explained by examining the raw spectroscopic data even though H₂O₂ is likely formed in the gas phase in the outer sheath layer of the plasma channel by the recombination of OH radicals. Shih et al.⁴⁰ used chemical actinometry to calibrate their spectroscopic data and reported a decrease in the OH intensity up to a conductivity of 1 mS/cm for an in-liquid discharge which correlated well with the measured decrease in the H₂O₂ generation. Wang et al.²¹ similarly observed a decrease in H₂O₂ production up to a conductivity of 36 mS/cm and used simulations to predict a similar decrease in OH concentration with increasing solution conductivity.

Due to this convoluted relationship between the properties of the plasma and the resulting bulk liquid chemistry, we decided to explore the degradation of caffeine, a compound that reacts predominantly with OH radicals and is not susceptible to photolysis.^48,49 Caffeine is a non-surfactant and according to the triple mechanism degradation model,⁹ participates in so-called sub-surface reactions (i.e., reactions that occur beneath the plasma–liquid interface with dissolved species).

The degradation of caffeine (represented as a pseudo-first order rate constant) decreases with increasing conductivity similarly to what was observed for H₂O₂ (Fig. 8). Both hydrogen peroxide generation and caffeine degradation results suggest that the hydroxyl radical production rate or availability decreases with increasing solution conductivity. The reduced OH radical production may be the result of the lower energy deposited into the plasma at high conductivity due to a larger current flowing through the water as a consequence of the lower solution resistance (Table I).

The observed removal for PFOA increases with the salt content, regardless of the salt type (Fig. 9). PFOA has a strong tendency to partition to gas–liquid interfaces and has been shown to participate in above-surface reactions,^{3,5,9,50–54} which is the reason why the presence of different anions in the bulk solution exhibit no effect on its removal. In addition, PFOA is non-oxidizable but can be degraded thermally at temperatures as low as 400 °C and possibly in reactions with argon ions and hot electrons.^55,56 Regardless of the mechanism of PFOA degradation, the most significant phenomenon contributing to the increase in its removal rate with increasing conductivity relates to an increase in the interfacial concentration of PFOA with conductivity. As shown in Fig. 10, the surface tension of PFOA solutions, an indirect representation of interfacial compound concentration, decreases with an increase in the salt content, indicating that the Gibbs surface excess concentration of PFOA at the gas–liquid interface increases with increasing conductivity.⁵¹ This so-called salting out effect occurs when an electrolyte interacts with a non-electrolyte (in this case PFOA), causing it to become less soluble at high salt concentrations.⁵¹ Because PFOA is degraded at the plasma–liquid interface, its removal rate will be directly proportional to its surface concentration (Fig. 11).

Rhodamine B is another surface-active compound whose degradation by electrical discharge plasmas has been studied extensively. However, unlike PFOA, Rhodamine B reacts with a wide range of reactive oxidative and chlorinated species.^5,6,26,28 Figure 12 shows the observed pseudo-first order rate constant for the removal of Rhodamine B as a function of solution conductivity and salt type. Non-chlorine salts are plotted in Fig. 12(a) and chlorine salts in Fig. 12(b). Unlike PFOA, the removal of Rhodamine B is strongly dependent on the type of the salt present. The salting out effect responsible for the increased removal of PFOA with conductivity played no role in the removal of Rhodamine B, as the surface tension of the rhodamine B solution did not change with conductivity, regardless of the salt type (Figure S4 in the Supplementary Material). The results shown in Fig. 12 make it apparent that non-chlorine salts behave differently than chlorine salts. Different mechanisms these salts elicit on the dye degradation are discussed next.

Non-chlorine salts. For non-chlorine salts, there is no impact of solution conductivity or ion composition on the removal of Rhodamine B [Fig. 12(a)]. In addition, based on the caffeine degradation data, which showed that OH concentration is decreasing with increasing solution conductivity, OH cannot be the dominant oxidative species responsible for Rhodamine B removal. However, unlike caffeine, Rhodamine B can be degraded by UV light.⁵⁷ To isolate the Rhodamine B solution from reactive species while exposing it to plasma-generated light, the high voltage and the ground electrode were raised to 2.5 cm above the liquid surface to prevent the plasma from contacting the liquid surface and still allow the arc to be generated between the two electrodes. Experiments were performed with this modified electrode configuration at 1.2 and 45 S/cm bulk liquid solution conductivities with both sodium chloride and sodium fluoride electrolytes (Fig. 13). For most experiments (except for the 45 mS/cm NaCl, which will be discussed later), removal was the same for both electrode configurations indicating that the plasma-generated light is the primary agent responsible for the degradation of Rhodamine B. If OH radicals were assisting the degradation of Rhodamine B, there should also be a decrease in the removal rate at higher solution conductivities, which suggests that OH does not participate in the degradation of Rhodamine B. The lack of influence of OH radicals on the degradation of the dye could be explained by the transport limitations of these reactive species, which are absent for photolysis.

The similarities among the magnitudes of the Rhodamine B rection rate constants at different solution conductivities [Fig. 12(a)] imply that the intensity of the plasma-generated light responsible for the dye degradation may be independent of solution conductivity. Others have reported increased UV emission intensity with increasing solution conductivity. For example, Gasanova² observed that the intensity of the light generated by in-liquid discharges increased as solution conductivity was increased from 0.5 to 2 mS/cm and the exposure time decreased from 1400 ns at 0.5 mS/cm to 500 ns at 2 mS/cm due to shorter pulse durations at higher conductivities. Lukes et al.⁵⁸ also observed increases in UV light emissions with increasing solution conductivity for liquid discharges in solutions between 0.1 and 0.5 mS/cm which was attributed to changes in the pulse duration and the mean pulse power. In this work, the pulse width did not change significantly with increasing solution conductivity, which may explain why light emissions appear to be independent of solution conductivity. The experiments with PFOA identical to those whose results are shown in Fig. 13 (W/O contact cases) resulted in no PFOA degradation, which is consistent with the recalcitrant structure of this compound, which has been shown to be unresponsive to any direct UV or chlorine treatments.^53,59

To determine the part of the spectrum responsible for the degradation of the Rhodamine B dye, two 3.5 ml cuvettes filled with a 1 mg/l Rhodamine B solution, one plastic, which is transparent to the visible light, and one quartz, transparent to both visible and UV light, were placed inside the reactor as shown in Fig. 2. The degradation of a 1 mg/l Rhodamine B solution after 60 min of exposure to plasma-generated light mainly occurred in the quartz cuvette (Fig. 14), indicating that only UV light plays a significant role in the degradation of the dye. This is in agreement with the literature that reports that direct photolysis of Rhodamine B's chromophore can only occur at wavelengths <290 nm.^57,60–62 The lower observed dye removal at 1.2 mS/cm compared to 45 mS/cm is likely due to the ions in the solution absorbing some of the UV light generated by the plasma.⁶³ It is important to note that the degradation of H₂O₂ also occurs at wavelengths <290 nm; however, the reason no H₂O₂ degradation was observed (Fig. 6) is likely due to the inability of the quantification method used for H₂O₂ to measure peroxide concentration changes below ∼60 μM (the lower detection limit for H₂O₂ measured as pertitanic acid).³⁴ The largest change in concentration observed for Rhodamine B was ∼12 μM (21 μM starting concentration) over an hour of exposure to plasma-generated light, and it is reasonable to expect that similar concentration changes (within the order of magnitude) would be measured for H₂O₂ decomposition (1.2 mM starting concentration). The reason that smaller changes in Rhodamine B concentrations can be measured in comparison to H₂O₂ is due to the significant difference in the molar extinction coefficients of Rhodamine B (106 000 l/cm-mol) and pertitanic acid (500 l/cm-mol). Both calculated values agree with literature reports.^64,65

Chlorine salts. The removal of Rhodamine B increases with increasing chloride ion concentration and appears to be independent of the valence number of the cation and the cation type in solution [Fig. 12(b)]. The formation of chlorine species by electrical discharges within or in contact with saline solutions has been shown to occur to varying degrees depending on the chloride ion concentration.^{11,24,26–28} Hypochlorous acid (HOCl), chlorite (⁠$ClO2−$⁠), chlorate (⁠$ClO3−$⁠), and perchlorate (⁠$ClO4−$⁠) have all been identified as products of these reactions.²⁵ The formation of chorine-based species was verified by measuring the concentration of chlorate ions as a function of solution conductivity, confirming that chloride ions, regardless of the salt type used, are converted into chlorate with higher concentrations measured at higher solution conductivity (Fig. 15).

During an electrical discharge, OH radicals have been hypothesized to be the main species responsible for the initiation of bulk liquid chlorine chemistry, as shown in reactions (3)–(11).^24,26,27 The importance of reaction (3) in the formation of chlorine species and, ultimately, the degradation of the dye were confirmed using DMSO (Fig. 16), a well-known OH radical scavenger. The observed rate of degradation of Rhodamine B (k_obs) at high solution conductivity is the result of both the degradation by UV light (k_UV−RhB) and the degradation by chlorine species (k_Cl−RhB). By scavenging the species responsible for the formation of the reactive chlorine species, the resulting k_obs was equal to that of k_UV−RhB; k_obs was found to be on average 0.025 ± 0.004 min⁻¹ for all conductivities equal to the measured k_UV−RhB rates of 0.022 ± 0.003 min⁻¹. At lower conductivities where the chloride ion concentration was not high enough to produce significant quantities of reactive chlorine species, the addition of DMSO had no impact on k_obs.

Because OH radicals appear to be the primary agent responsible for the initiation of chlorine chemistry and the caffeine degradation data shown previously suggest that OH radical production decreases as solution conductivity increases, the concentration of chlorine species generated at higher solution conductivities should decrease and not increase, as shown in Fig. 15. This discrepancy can be explained by the availability of chloride at the plasma-liquid interface, which has been shown to increase with conductivity,⁶⁶ 

Of all the chorine species that may be formed in reactions (3)–(11), Cl^•, $Cl$^•−₂, and hypochlorous acid (HOCl) can all react with Rhodamine B.^26–28 Of these species, both Cl^• and $Cl$^•−₂ are highly unstable and will rapidly react with either Rhodamine B near the interface or form more stable species such as HOCl which can diffuse into the bulk liquid to react at longer timescales. To determine the role of HOCl in dye degradation, experiments were conducted at varying starting solution pH values (Fig. 17). When the pH of the solution is lowered and the solution conductivity is maintained at a constant value, the dye degradation undergoes a maximum rate, similar to the HOCl profile on the free chlorine speciation diagram (Fig. 18). Above pH ∼ 7, HOCl deprotonates to form a relatively inert ClO⁻ that does not react with Rhodamine B.⁶³ This result suggests that the primary chlorine species responsible for the removal of Rhodamine B is hypochlorous acid. To additionally verify that the observed changes originate from the chlorine species' pH dependence, and not a change in reactivity due to the protonation of Rhodamine B (pKa = 4.2),^67,68 a control experiment was performed in the presence of a mixture of sodium nitrate and sulfuric acid. Without the chlorine in solution, pH has a minimal impact on the observed removal (Fig. 17), verifying that hypochlorous acid is playing a major role in Rhodamine B dye degradation at high solution conductivity and that chlorine species can enhance contaminant removal.

This study investigated the influence of solution conductivity and ion composition on the discharge characteristics and treatment of organic contaminants using a point–point electrode configuration above water. The point–point electrode geometry allowed for the study of the influence of conductivity without the influence of a changing plasma–liquid contact area. The measurements were carried out with up to 10 electrolytes at conductivities ranging from 0.3 to 45 mS/cm. The energy delivered to the reactor increased with increased solution conductivity; however, the increase in energy delivered to the system was lost as heat. In the future, work should be done to optimize the energy delivered to form reactive species while minimizing heat loss. Data also suggested that as solution conductivity increases the hydroxyl radical generation decreases, which could be detrimental for the degradation of some contaminants.

This work demonstrated that the structure of the target organic contaminant plays a very important role on the mechanism by which it is degraded. For PFOA, increased removal at high solution conductivity is the result of an increase in surface concentration due to the salting out effect. This response was independent of the salt composition added. For Rhodamine B, salting out was not observed and only the addition of chlorine-containing salts improved observed removal above 1.2 mS/cm. Plasma-generated UV light was shown to be the primary source of Rhodamine B degradation for all non-chlorine salts regardless of solution conductivity and for chlorine salts at 1.2 mS/cm. It was determined that hypochlorous acid formed during the plasma treatment process was responsible for the improved Rhodamine B degradation at higher solution conductivities that contained chloride ions. The primary species that resulted in the formation of hypochlorous acid was determined to be hydroxyl radicals. For caffeine, a compound that cannot react with either hypochlorous acid or UV light, the hydroxyl radical generation rate directly correlated with the removal rates observed, suggesting that hydroxyl radicals are still the primary reactant responsible for some contaminant degradation. The observed differing trends for each of the studied compounds in this work make it apparent that the selection of model compounds should be done with care when optimizing reactor design, particularly, the use of dyes should be avoided due to the various mechanisms by which they can be degraded.

This study demonstrates the strong influence of solution conductivity on the treatability of various contaminants. The efficacy of plasma-based water treatment is highly dependent on the water conductivity and in some cases also strongly influenced by the ion composition of the liquid phase. Future use of plasma-based systems for water treatment should consider conductivity, contaminant structure, and ionic composition when determining the optimum conditions for treatment.

See the supplementary material for Rhodamine B surface tension measurements and optical emission spectra of discharges at different conductivities.

This project was funded by the Department of Defense's Environmental Security Technology Certification Program (ESTCP) under Contract No. ER18-B3-5015. The authors declare no competing financial interest.

The data that support the findings of this study are available from the corresponding author upon reasonable request.