Efficiency of Ozone Quenching Agents at Different Temperature, pH, and Hydrodynamic Conditions

ABSTRACT

Quenching of excess aqueous ozone (O₃) residual is needed to avoid off-gassing and oxidative damage to downstream components. Eight quenching agents (QAs), calcium thiosulfate (Ca-Thio), sodium bisulfite (Na-Bi), hydrogen peroxide (H₂O₂), sodium thiosulfate (Na-Thio), sodium metabisulfite (Na-Metabi), potassium metabisulfite (K-Metabi), sodium sulfite (Na-S), and ascorbic acid (AscAcid) were tested at varying pHs (7.5, 7.9, and 8.5) and temperatures (4°C, 15°C, and 23°C). Temperature had significant effects on initial quenching (≤30 s), with less occurring at 4°C than 23°C; but with extended contact time this trend reversed. Quenching efficiency was not substantially affected by pH. Ca-Thio and Na-Thio were inefficient but quick, had lower mass requirements, and minimal handling concerns. Na-Metabi, Na-Bi, and Na-S were quick and efficient, but K-Metabi was the only QA that quenched (26–31%) less O₃ than stoichiometry at all conditions. These sulfite-based QAs had more handling concerns and greater mass requirements. H₂O₂ was slower than other QAs but 2–9× quicker than predicted by rate coefficients and efficient when given longer contact time (≤5 min). H₂O₂ had some handling concerns but the lowest mass requirements (assuming 5 min contact time) and creates desirable AOP conditions. AscAcid consistently quenched (22–113%) more O₃ than the stoichiometric ratio but had higher mass requirements, except in colder temperatures. Hydrodynamics, QA-diffuser design, and supply chain were also important considerations.

Graphical abstract

KEYWORDS:

• Ascorbic Acid
• Quenching Agent
• Scavenger
• Hydrogen Peroxide
• Bisulfite
• Kinetics
• Ozone
• Redox
• Thiosulfate

Introduction

Ozonation is an effective oxidation process utilized by drinking and wastewater treatment plants for contaminant degradation and pathogen inactivation. Water quality goals are achieved by creating a dissolved ozone (O₃) residual in water, which then passes through a baffled O₃ contactor to achieve a target O₃ exposure (CT). Under conditions where excess dissolved O₃ (DO₃) residual is present in the contactor effluent, an O₃ quenching agent (QA) is applied to prevent downstream off-gassing (e.g., protect worker safety) and to safeguard equipment (e.g., rubber o-rings, pumps) from oxidative damage (Bromley and Wert 2002; Cataldo 2019; Langlais, Reckhow and Brink 2019; Schulz 2014). The QA is typically applied in at least one chamber prior to the ozonated water exiting the contactor to allow sufficient contact time for hydraulic mixing and complete DO₃ residual quenching. In most treatment facilities, a chemical feed system for quenching DO₃ residual is considered a standard part of the O₃ system design. Drinking water treatment plants targeting disinfection often require quenching of excess DO₃ residual during certain water quality conditions (e.g., low temperature, low dissolved organic carbon (DOC) concentration, low pH). Wastewater treatment plants often have greater DOC concentrations and temperature that accelerate O₃ decomposition and are less likely to require QA addition.

The selection of an O₃ QA is determined by several factors related to system design and operation. The chemical reaction kinetics of the QA are a primary consideration with respect to stoichiometry (i.e., low mass/molar ratio), rapid reaction time (i.e., O₃ reaction rate constant), minimal secondary reactions (i.e., downstream oxidant demand, disinfection product formation), and need for additional advanced oxidation process (AOP) conditions. Chemical feed system design considerations may include QA stability (e.g., non-reactive, nonvolatile, low freezing point), material compatibility (e.g., storage tank, feed pumps, delivery lines), and application point (e.g., contactor location, diffuser design, mixing efficiency). Material handling and employee safety considerations (e.g., nontoxic, nonflammable, non-corrosive) are also factored into QA selection (K. L. Rakness 2005; Richey et al. 2000). Practical considerations may include product purity (i.e., NSF 60 certification), product availability (i.e., delivery frequency, supplier location), and cost (Richey et al. 2000).

Calcium thiosulfate (Ca-Thio), sodium bisulfite (Na-Bi), and hydrogen peroxide (H₂O₂) are the most common O₃ QAs used at water treatment facilities, according to a survey of O₃ application literature and conference proceedings (Bromley and Wert 2002; Mazloum et al. 2004; Mitra et al. 2017; Neemann et al. 2001; K. L. Rakness 2005; E. C. Wert et al. 2016; Richey et al. 2000; Schulz 2014; Soucie, Sheen and Madura 2001). Little is documented on historical O₃ quenching, but based on the survey, prior to the early 2000s, H₂O₂ appeared to be most frequently used for O₃ quenching. References to the use of Ca-Thio as a QA were observed more broadly after 2000; the same year a patent application for the use of Ca-Thio was submitted to the US Patent Office by Ag Formulators Inc (Bromley and Wert 2002; Holbrook 2005; Mazloum et al. 2004; Mitra et al. 2017; Neemann et al. 2001; K. L. Rakness 2005; E. C. Wert et al. 2016; Richey et al. 2000; Schulz 2014; Soucie, Sheen and Madura 2001). Since the early 2000s, trends in quenching have largely not changed, despite documented limitations of the three common QAs (Hojjatie and Abrams 2006; Holbrook 2005). H₂O₂ has been characterized as slow, unstable (e.g., reactive, degrades), the most expensive, and requires special handling and storage (e.g., storage in stainless steel, corrosive). The additional benefit of H₂O_2, when added to O₃, is it’s an AOP which degrade taste & odor compounds and micropollutants (Bromley and Wert 2002; K. L. Rakness 2005). Na-Bi off-gases sulfur dioxide and thus requires a special scrubbing and ventilation system, as well as insulation to avoid elevated temperatures (K. L. Rakness 2005; Richey et al. 2000). Additionally, Na-Bi is subject to crystallization at temperatures below 15°C (60℉) and requires tight temperature controls in most environments (K. L. Rakness 2005; Richey et al. 2000). Ca-Thio has shown low efficiency, calcium deposits/blockages in feed lines have been noted, and there may be concerns about use restrictions and supply due to the existing patent (Bromley and Wert 2002; Holbrook 2005; K. L. Rakness 2005). Overall, Ca-Thio has less stability and handling considerations compared to the other common QAs (Bromley and Wert 2002; K. L. Rakness 2005). Both Ca-Thio and Na-Bi have also been effectively applied for dechlorination (Bedner, MacCrehan and Helz 2004; MacCrehan, Bedner and Helz 2005). There are other common dechlorination reagents that are reductants, which have yet to be assessed for O₃ quenching. Given the limitations with the three common QAs, these unevaluated reducing agents may pose an opportunity to expand the chemical alternatives for O₃ quenching.

Although this study focuses on O₃ quenching for treatment process control, most studies in the literature focus on selection of QA for sample collection. The most relevant takeaways from these studies are summarized in the supplemental information (Text S1 and Table S1). When O₃ quenching is performed at full-scale (flows on the order of mgd), quenching efficiency can affect the applied dosing requirements, dosing system design, and required infrastructure. Accounting for quenching efficiency under the range of expected water quality conditions is important to ensure the quenching system design is adequate and to prevent over- or underdosing of QA (Bromley and Wert 2002). Overdosing of a QA can result in increased dosing requirements for downstream oxidants (e.g., chlorine used as a secondary disinfectant), to overcome the oxidant demand from the excess QA (Bouland, Duguet and Montiel 2004; Mazloum et al. 2004; Richey et al. 2000). Underdosing of a QA can result in O₃ off-gassing and degradation of equipment downstream (Bromley and Wert 2002). During application, online monitoring downstream of the quenching application point is not always effective for pacing QA dosing, assessing quenching performance, and preventing over/underdosing. If monitoring is performed, online oxidation-reduction potential (ORP) is typically used, but has been evaluated with varied results (Muri 2006; Muri 2008). In some installations, DO₃ residual probes are used near the effluent chambers/channels, but they are not always sensitive enough, with typical detection limits in the range of 0.1 to 0.2 mg/L. If the O₃ process effluent stream is not well mixed, a single point measurement, whether ORP or DO₃ may not effectively reveal residual O₃ in the effluent prior to exiting the contactor.

Water quality conditions can have a direct impact on QA performance. Temperature affects chemical reaction kinetics. Several studies have shown that O₃ decay rate, which is primarily driven by reactions between O₃ and natural organic matter (NOM), increases substantially with increasing temperature (Elovitz, von Gunten and Kaiser 2000; Shin, Hidayat and Lee 2016). It would follow that QAs, with reaction rate constants similar to O₃ reactive NOM fractions (e.g., Group I compounds k_O3 >10⁶ M⁻¹s⁻¹) (Lee et al. 2013) would similarly have accelerated reactions with O₃ due to increased temperature, and slower reactions with lowered temperature. Speciation of QAs and other water matrix components (e.g., NOM) may be dependent on pH and can affect the reactions that occur and their kinetics (k_O3). For example, O₃ decay has been shown to increase with pH due to the deprotonation of NOM, which increases electrophilic O₃ decomposition (Elovitz, von Gunten and Kaiser 2000). The magnitude of pH’s effect on O₃ reaction rate constants varies widely depending on the functional groups of the reactant (e.g., QA) being protonated/deprotonated (von Sonntag and von Gunten 2015). Efficiency and rate of O₃ quenching reactions may be affected by pH due to changes in the direct reaction with O₃, as well as changes in secondary reactions (von Sonntag and von Gunten 2015). Despite the importance of quenching, there are currently no practical, comprehensive guidelines in the peer-reviewed literature for dosing O₃ QAs, and there is limited quantitative characterization of how these water quality and hydrodynamic factors may influence the efficiency and rate of quenching reactions. In highly regarded water treatment fundamental and O₃-specific textbooks, little if anything is mentioned about the need to quench residual O₃, quenching chemistry, nor guidance on how to design a quenching process at the end of ozonation (Benjamin and Lawler 2013; Crittenden et al. 2012; Langlais, Reckhow and Brink 2019; K. L. Rakness). Rakness provides the most explicit guidance on O₃ quenching, discussing the use of Na-Bi, Ca-Thio, and H₂O₂. The guidance, however, is still vague. For example, it is suggested that site-specific factors will affect quenching but generally dosing ≥10% excess of QA is recommended, and that H₂O₂ may take ≥3–5 min to react (K. L. Rakness 2005).

The lack of peer-reviewed information regarding QA efficiency has led many utilities and consultants to rely on theoretical stoichiometry or generic dosing charts from chemical manufacturers for ranges of water temperatures and pH, with little information on how the guidance was produced (e.g., water quality and hydrodynamic conditions). Accordingly, the main goal of this study was to provide quantitative guidelines for quenching O₃ with the three commonly used QAs (Ca-Thio, Na-Bi, H₂O₂). Sodium metabisulfite (Na-Metabi), potassium metabisulfite (K-Metabi), ascorbic acid (AscAcid), sodium sulfite (Na-S), and sodium thiosulfate (Na-Thio) are reducing agents used to quench other oxidants (e.g., chlorine) and were also included in the systematic evaluation for O₃ quenching. The specific objectives of this work were to (1) determine the mass and molar ratios of quenched O₃ to applied QA (qO₃:QA) under ambient conditions in Colorado River water (CRW) and compare to the theoretical stoichiometry ((qO₃:QA)_th), (2) determine the effect of temperature and pH on O₃ quenching efficiency, (3) use computational fluid dynamics to demonstrate the effect of QA diffuser design and contactor hydraulics on quenching efficacy at full-scale, and (4) provide practical guidelines and information for O₃ quenching applications.

Materials and methods

Terminology

Key experimental and theoretical terms used through the manuscript are defined in Table 1.

Table 1. Key terms used throughout the manuscript and their definitions.

Term	Description
QA	Quenching agent
qO3	Quenched ozone
(qO3: QA)th	Theoretical molar/mass ratio of dissolved ozone residual concentration quenched (qO3) to quenching agent concentration applied (QA), based on stoichiometry
(qO3: QA)exp-30s	Experimental molar/mass ratio of dissolved ozone residual concentration quenched (qO3) to quenching agent concentration applied (QA) after 30 seconds of reaction time
(qO3:QA)exp-max	Experimental molar/mass ratio of dissolved ozone residual concentration quenched (qO3) to quenching agent concentration applied (QA) after the reaction is completed
qTimeth	Theoretical time to complete quenching, based on ozone reaction rate constants and initial O3 and QA concentrations
qTimeexp-stoi	Experimental time to reach the stoichiometric ratio, (qO3:QA)th
qTimeexp-max	Experimental time to reach quenching completion, to the (qO3:QA)exp-max

QA selection and reagents

QAs were selected based on their reaction rates with O₃ and hydroxyl radicals. Most QAs are reductants that participate in a redox reaction by donating electrons to an oxidant (Kristiana et al. 2014; Langlais, Reckhow and Brink 2019). As shown in Table 2, O₃ gains two electrons from the QA during the quenching reactions. The theoretical ratio of quenched O₃ to applied QA (qO₃:QA)_th was calculated from the stoichiometric equations (Table 2). Under ideal conditions, the (qO₃:QA)_th of the QAs evaluated in this study are expected to have molar ratios between 1:1 and 4:1 and mass ratios between 1:0.35 and 1:3.59 (Table 2). With the exception of H₂O₂, all quenching agents are Group I compounds, with O₃ reaction rate constants (k_O3) in the range of 10⁶ to 10⁸ M⁻¹s⁻¹ (Table 2) (Lee et al. 2013); meaning that under ideal conditions, quenching will occur on the order of seconds or faster (K. L. Rakness 2005).

Table 2. Key theoretical values associated with O₃ and QA reactions, including theoretical quenched ozone to QA ratios (qO₃:QA)_th, chemical formula and molar mass of each QA represented as the parent compound and as the active compound, theoretical O₃ + QA reaction rate constants (k_O3), hydroxyl radical reaction rate constants (k_OH•), and half reactions and overall quenching redox reactions.

	Parent Compound Chemical Formula (Molar Mass)	(qO3 :QA)th (Mass Ratio), QA as parent	Active Compound Chemical Formula (Molar Mass)	(qO3 :QA)th (Mass Ratio), QA as active	Theoretical Ozone Reaction Rate (kO3)^a	Theoretical Hydroxyl Radical Reaction Rate (kOH•)^b
Quenching Agent	(mg/mmol)	mol:mol (mg:mg)	(mg/mmol)	mol:mol (mg:mg)	M^−1s^−1	M^−1s^−1	Half Reactions and Redox Equation
Calcium Thiosulfate	CaS2O3 (152.2)	4 : 1 (1 : 0.79)	S2O3^2- (112.1)	4 : 1 (1 : 0.58)	0.7–2.2 × 10^8	7.0 × 10^9	4O3 +8 H^+ +8e^−→ 4O2 +4 H2O
							Ca^2+ + S2O3^2- + 5 H2O → Ca^2+ + 2SO4^2- + 10 H^± +8e^−
							4O3+CaS2O3+H2O → 4O2 + Ca^2+ +2SO4^2- +2 H^+
Sodium Thiosulfate	Na2S2O3 (158.1)	4 : 1 (1 : 0.82)	S2O3^2- (112.1)	4 : 1 (1 : 0.58)	0.7–2.2 × 10^8	7.0 × 10^9	4O3 +8 H^+ +8e^− → 4O2 +4 H2O
							2Na^± +S2 O3^−2 + 5 H2O → 2Na^± + 2SO4^2- + 10 H^± +8e^−
							4O3 + Na2S2O3 + H2O → 4O2 +2Na^+ +2SO4^2- +2 H^+
Hydrogen Peroxide	H2O2 (34.0)	2 : 1 (1 : 0.35)	H2O2 (34.0)	2 : 1 (1 : 0.35)	9.6 × 10^6 (HO2^−) <0.01 (H2O2)	7.5 × 10^9 (HO2^−) 2.7 × 107 (H2O2)	H2O2 +2O3 → 2∙OH +3O2 ^c
Sodium Metabisulfite	Na2S2O5 (190.1)	2 : 1 (1 : 1.98)	HSO3^− (82.0)	1 : 1 (1 : 1.71)	0.4-2 × 10^6	4.5 × 10^9	2O3 +4 H^+ +4e^− → 2O2 +2 H2O
							Na2S2O5 +2 H2O → 2Na^±+ 2HSO3^− +2 H2O → 2Na^±+2SO4^2-+6 H^± + 4e^−
							2O3 + Na2S2O5 → 2Na^+ +2SO4^2- +2 H^+ +2O2
Potassium Metabisulfite	K2S2O5 (222.3)	2 : 1 (1 : 2.32)	HSO3^− (82.0)	1 : 1 (1 : 1.71)	0.4–2 × 10^6	4.5 × 10^9	2O3 +4 H^+ +4e^− → 2O2 +2 H2O
							K2S2O5 +2 H2O → 2K^±+ 2HSO3^− + 2 H2O→ 2K^±+ 2SO4^2- +6 H^± + 4e^−
							2O3 + K2S2O5→ 2K^+ +2SO4^2- +2 H^+ +2O2
Sodium Bisulfite	NaHSO3 (104.1)	1 : 1 (1 : 2.17)	HSO3^− (82.0)	1 : 1 (1 : 1.71)	0.4–2 × 10^6	4.5 × 10^9	O3 +2 H^+ +2e^− → O2 + H2O
							Na^±+ HSO3^−+ H2O→ Na^±+ SO4^2- + 3 H^± + 2e^−
							O3 + Na^+ + HSO3^− → Na^+ + SO4^2-+ H^+ + O2
Ascorbic Acid	C6H8O6 (176.1)	1 : 1 (1 : 3.59)	C6H8O6 (176.1)	1 : 1 (1 : 3.59)	6 × 10^7	1.0 × 10^10	O3 +2 H^+ +2e^− → O2 + H2O
							C6H8O6 → C6H6O6 +2 H^± + 2e^−
							O3 + C6H8O6→ C6H6O6+O2+ H2O
Sodium Sulfite	Na2SO3 (126.0)	1 : 1 (1 : 2.63)	SO3^2- (80.1)	1 : 1 (1 : 1.67)	1.3–1.5 × 10^9	5.5 × 10^9	O3 +2 H^+ +2e^− → O2 + H2O
							2Na^±+ SO3^2-+ H2O → 2Na^±+SO4^2- +2 H^±+ 2e^−
							O3 +2Na^+ + SO3^−2 → O2 +2Na^+ SO4^2-

^aReaction rate does not account for dissolution or dissociation. Rates were obtained from (von Sonntag and von Gunten 2015) unless otherwise specified.

^bRates were obtained from •OH Rate Constants in (Buxton et al. 1988).

^cPeroxone chemistry is complex and includes multiple intermediate steps, the overall stoichiometry is given here, for more details see (Fischbacher et al. 2013; Sein et al. 2008; Staehelin and Hoigne 1982; Staehelin and Hoigne 1985; von Sonntag 2008; von Sonntag and von Gunten 2015).

Calcium thiosulfate was obtained from Best Sulfur Products (Fresno, CA) as Captor 30% w/v. Sodium thiosulfate (CAS: 7772-98-7), hydrogen peroxide (7722-84-1), ascorbic acid (50-81-7), sodium sulfite (7757-83-7), sodium bisulfite (7631-90-5), sodium metabisulfite (7681-57-4), and potassium metabisulfite (16731-55-8) were all obtained from Fisher and were anhydrous and of Certified ACS reagent grade or higher. Quenching reagents were volumetrically diluted in ultrapure water (>17 MΩ cm) to produce QA stocks in the 0.3 to 2.0 mg/mL concentration range.

Batch ozonation and quenching experiments

All experiments were performed using Colorado River water (CRW), collected from a raw water sample tap at the Alfred Merritt Smith Water Treatment Facility (AMSWTF), whose intake is in Lake Mead (Boulder City, Nevada). The CRW was characterized by pH (7.9), turbidity (0.6 NTU), conductivity (850–980 µS/cm), total alkalinity (140–144 mg/L as CaCO₃), and TOC (2.4–2.6 mg/L) and was stored at 4°C until use. To adjust water temperature, water was placed in refrigeration at 4 or 15°C overnight until immediately before use or left at ambient room temperature overnight for 23°C conditions. Water temperature was checked before and at the end of each experiment to ensure temperature did not substantially change (±2℃), and when necessary, the batch reactor was chilled to maintain temperature. Sodium hydroxide and hydrochloric acid were used to adjust pH of water samples, as appropriate. The pH of water was measured before and after experiments (including after addition of quenching agent) and pH did not change substantially (<0.5) in any of the experiments.

Batch ozonation methods were used to prepare ozonated samples (Hoigné and Bader 1994; Wert et al. 2011). Briefly, gaseous O₃ (260–280 g/Nm³, (Cataldo 2019)) was generated via a Plasma Block (PlasmaTechnics, Racine, WI) fed by high-purity oxygen. The O₃ gas was diffused in a 4 L reactor containing chilled (4°C) ultrapure water to produce a high concentration O₃ stock (60–100 mg/L as DO₃). The O₃ stock concentration used to spike samples was measured for each batch of samples, and concentration varied by up to 5%. O₃ stock was spiked into 1 L stirred batch reactors containing CRW to achieve an initial O₃ concentration of 2.5–3.5 mg/L (target level depending on water quality condition being tested). Due to variation in O₃ stock the applied ozone dose varied slightly from sample to sample; for each condition, the applied O₃ dose of quenching experiment samples were within <1–12% of the applied O₃ dose of the corresponding control experiment. The effects of this variability in applied O₃ dose on O₃ decay and quenching ratio results were evaluated and were found not to be substantial. Control experiments (no QA added) at the different temperatures and pHs were conducted in this manner and immediately prior to and on the same day as corresponding quenching experiments. Quenching experiments were conducted similar to control experiments, except quenching agent was spiked into the ozonated sample at a time point (1.5–5 mins) selected to target when DO₃ was 0.6–1.0 mg/L. The volume of QA spiked into the sample was based on stoichiometry, targeted at quenching 25% of the expected O₃ residual at that time point, which was based on control tests. The volume of QA stock spiked was generally low (<0.5% total reactor volume) and thus did not have a measurable dilution effect on ozone residual. A subset of experiments spiking QAs at different applied ratios (e.g., 25% vs. 50% of the stoichiometric ratio) were conducted. Insignificant differences were observed in most cases and thus experiments discussed throughout the manuscript were conducted at the lower applied QA ratio (25%) condition. For further discussion see Supplemental Information Text S2, Figure S1.

DO₃ residuals were measured using the indigo trisulfonate method (Standard Method 4500-O₃) (APHA; AWWA; WEF 1998; K. L. Rakness et al. 2010) at 30s after O₃ spiking and subsequently at 30 s to 1 min intervals until complete O₃ depletion (<0.02 mg/L DO₃). The QA was spiked immediately after a DO₃ residual measurement, and DO₃ was measured again 30s after QA was spiked and then every 30 s–1 min until DO₃ was depleted.

Data analysis

The theoretical O₃ quenched to QA (qO₃:QA)_th values are derived from stoichiometry as described in a previous section. The theoretical time to quenching completion (qTime_th) was found by kinetic modeling at standard conditions, further described in the Supplemental Information, Text S3 and Table S2.

Derivation of key experimental values are summarized by Figure 1, which describes three example outcomes in terms of efficiency and quenching speed. Change in O₃ residual (∆[O₃]) over each time interval during control and quenching experiments was determined using equation 1[1] $\mathrm{\Delta }{\left[\mathrm{O}3\right]}_{\mathit{t}1\to \mathit{t}2}={\left[\mathit{O}3\right]}_{\mathit{t}1}-{\left[\mathit{O}3\right]}_{\mathit{t}2}$[1] and defined as the decrease in the O₃ residual from time point 1 (e.g., 30 s) to time point 2 (e.g., 1 min).

Figure 1. Graphical (top row) and tabular (bottom row) examples of how the quenched ozone (qO₃) was determined, for use in subsequent ratio calculations (qO₃:QA)_exp, and identification of key quenching timepoints (qTime_exp) on experimental ozone decay curves (i.e., ozone residual measurements overtime). Three example outcomes are shown where (a) calcium thiosulfate was a fast (i.e., qTime_exp-max ≤30s) and inefficent QA (i.e., (qO₃:QA)_exp-max <85% of the (qO₃:QA)_th), (b) sodium metabisulfite was a fast (i.e., qTime_exp-stoi ≤30s) and efficent QA (i.e., (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th), and (c) ascorbic acid was a fast (i.e., qTime_exp-stoi ≤30s) and greater than ideal (i.e., (qO₃:QA)_exp-max >; (qO₃:QA)_th by more than 15%) QA. Each graph shows results from both a control test (purple points) and an experimental test with quenching (green points). Note, the x-axis starts at 2 minutes and that time point zero (t₀) is when the QA is added, in these examples at 3 minutes. The calculations for determining the qO₃ and its relation to the theoretical stoichiometry ((qO₃:QA)_th) are given below each graph. The quenching time where (qO₃:QA)_th was achieved is noted as qTime_exp-stoi and the time to reach quenching completion is noted as qTime_exp-max.

[1] $\mathrm{\Delta }{\left[\mathrm{O}3\right]}_{\mathit{t}1\to \mathit{t}2}={\left[\mathit{O}3\right]}_{\mathit{t}1}-{\left[\mathit{O}3\right]}_{\mathit{t}2}$[1]

In the control samples, the decrease in DO₃ residual is due to O₃ reacting with DOC and other water quality constituents. In the experimental samples, where a QA was applied, the decrease in DO₃ is due to both reactions with water matrix constituents and O₃ quenched by the QA. As described by equation 2[2] $\mathit{q}{O}_3={\left[O_3\right]}_{\mathit{exp}}-{\left[O_3\right]}_{\mathit{ctrl}}$[2] and shown in Figure 1, the decrease in O₃ due to quenching only (qO₃) was found at each paired time interval by the difference in ∆[O₃] for the experimental sample and the control sample.

[2] $\mathit{q}{O}_3={\left[O_3\right]}_{\mathit{exp}}-{\left[O_3\right]}_{\mathit{ctrl}}$[2]

Since all QAs, except H₂O₂, were expected to react with O₃ on the order of seconds or less, the qO₃ was calculated from the O₃ residual measurements immediately prior to QA application and 30 s after QA application ((qO₃)_exp-30s). The (qO₃)_exp-30s was then used to calculate the O₃ quenched to applied QA molar ratio at 30 s ((qO₃:QA)_exp-30s). For most of the QAs, (qO₃:QA)_exp-30s < (qO₃:QA)_th; thus, the cumulative qO₃ was determined by adding together the qO₃ for each subsequent time interval. The time at which the cumulative (qO₃:QA)_exp= (qO₃:QA)_th, was noted as qTime_exp-stoi. The (qO₃:QA)_exp-max was defined as the maximum cumulative (qO₃:QA)_exp. The endpoint of the time interval at which (qO₃:QA)_exp-max was reached, was considered as the time to quenching completion (qTime_exp-max). All (qO₃:QA)_exp values were calculated using both molar units (mol:mol) and mass units (mg:mg), and concentrations for QAs represented in units “as the active compound” and “as parent compound,” indicated in Table 2.

Defining relative QA efficiency and quenching speed

When conditions are ideal, theoretically the QAs would react with O₃ instantaneously, with the exception of H₂O₂, and at 100% efficiency, i.e., with qTime_th and at (qO₃:QA)_th shown in Table 2. In application, the efficiency and quenching rate may be affected by several factors including water quality conditions and other constituents in the water (e.g., NOM, pH, temperature, inhibitors, initiators, promotors). These effects will vary with the different QAs and with different waters. In this study, QAs with (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th, were considered “efficient.” QAs with (qO₃:QA)_exp-max at <85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less), were considered “inefficient.” If (qO₃:QA)_exp-max > (qO₃:QA)_th by more than 15%, then the QA efficiency was considered “greater than ideal.”

An efficient or greater than ideal QA was considered “fast” to react when a qTime_exp-stoi ≤ 30 s. In either case, the qTime_exp-max may also be at ≤ 30 s or the quenching reaction may be “extended” if the reaction continues beyond 30 s. Inefficient QAs may also be considered “fast” if qTime_exp-max ≤ 30 s, despite not having a qTime_exp-stoi and (qO₃:QA)_exp-max < 85% (qO₃:QA)_th. A QA may be considered “moderately fast” if qTime_exp-stoi = 1 min, or in the case of an inefficient QA when qTime_exp-max = 1 min. A QA was considered “slow” if qTime_exp-stoi > 1 min, or in the case of an inefficient QA when qTime_exp-max > 1 min.

Unit expression

Mass and molar ratios values are expressed in terms of “as parent compound” and/or “as active compound/component” throughout the manuscript. This is important to note as different individuals will utilize different unit expressions depending on the intended use and their perspective. When discussing results in terms of O₃ and quenching chemistry, units are expressed in terms of “as active compound” in the main text because it enabled the efficiency comparison of multiple QAs that have the same active component but different parent compound. Typically, practitioners will use units in terms of “as parent compound,” and thus, in the practical application sections, units are expressed as such to enable direct use. In many cases in the literature and in discussion, the unit terms are not fully expressed, and it is unclear whether mg/L or mol/L or mg:mg or mol:mol values are in terms of “as parent” or “as active component.” One might assume “as parent compound” is being used if units are not specifically expressed in these terms, but this may not always be the case and can result in negative practical outcomes. The mass requirements values expressed “as parent” are 17–58% higher than the mass requirement values expressed “as active compound,” depending on the QA. Thus, if the wrong unit expression is assumed then QA could be greatly over or underdosed. The unit expressions used in the main text are indicated, particularly in the figure and table captions. All results are provided in both “as parent compound” and “as active compound” unit expressions in the supplemental information, Tables S5-S8.

Computational fluid dynamic analysis

A Revit model was developed to create the model domain for the CFD analysis. The CFD model domain consists of a single O₃ contactor at AMSWTF from the contactor inlet gates to the outlet gates. The average contactor train flow was 37.5 mgd with peak and low flows of 75 mgd and 20.6 mgd, respectively. AMSWTF is a 600 mgd facility where O₃ is applied to raw untreated water as the first step in treatment. The AMSWTF O₃ contactor design is comprised of 12 Cells with the Ca-Thio applied in the outlet of Cell 10, above or adjacent to the baffle (Figure S2). AMSWTF currently employs Ca-Thio for quenching residual O₃ to prevent off-gassing and safeguard downstream equipment from oxidative damage. The QA injection points were included in the model where the chemical discharges into the primary flow. The flow distribution to the individual orifices in the chemical diffuser piping was evaluated with analytical calculations and specified in the CFD model, further described in (Table S3–S4 and Text S4).

The CFD models used were developed using Ansys 2022 R2. This software package includes Space Claim Direct Modeler to develop the model geometry, Fluent Meshing for the meshing generation, Ansys Fluent for model processing and post-processing, and CFD-Post for additional post-processing. A three-dimensional, single-phase flow (water only) model was developed for each flow condition with a fixed water surface elevation. The species transport model was used to trace the Ca-Thio (QA), which solves the conservation equations describing convection and diffusion without chemical reactions. Three designs for stainless steel diffuser manifolds, that span the width of the contactor, were modeled. Further detail on CFD model methodology, the domain, and boundary conditions can be found in Supplemental Information Text S4 and Tables S3- Table S4.

Results and discussion

Quenching efficiency at ambient conditions (23°C, pH 7.9)

Figure 2a and Table 3 shows, under ambient conditions (23°C and pH 7.9), Ca-Thio, Na-Thio, and K-Metabi quenched less O₃ than expected based on stoichiometry ((qO₃:QA)_exp-max <85% (qO₃:QA)_th); although, Ca-Thio and K-Metabi were fast to react, reaching quenching completion (qTime_exp-max) in ≤ 30 s (Figure 2b, Table 3). This fast reaction was expected as all three of these QAs are Group I Compounds, having high reaction rates with O₃ (k_O3 >10⁶-10⁹ M.s⁻¹) (Lee et al. 2013; von Sonntag and von Gunten 2015). Na-Thio was slow to react, with qTime_exp-max = 1.5 min (Figure 2b, Table 3). Interestingly, both Ca-Thio and Na-Thio quenched at the same (qO₃:QA)_exp-max of 3 moles O₃:1 mole QA, which is 25% less O₃ than expected (Figure 2a, Table S5). In both O₃ quenching reactions, the thiosulfate anion (S₂O₃^2-) is the active compound responsible for quenching O₃, and the cation (Ca²⁺, 2 Na⁺) is what differentiates the two QAs. K-Metabi also quenched 25% less than expected overall (Figure 2a). The active component of K-Metabi is bisulfite (HSO₃⁻), which is the same active component for Na-Metabi and Na-Bi, yet Na-Metabi and Na-Bi did quench to (qO₃:QA)_th, with the K-Metabi (qO₃:QA)_exp-max being 19% and 35% lower than Na-Metabi and Na-Bi, respectively. Burns et al. conducted bench-scale testing with ozonated surface water from Vancouver (pH 7.5) and evaluated multiple QAs, including Ca-Thio. They found 60% higher mass concentration of Ca-Thio was required than indicated by (qO₃:QA)_th (Burns et al. 2008).

Figure 2. Experimental quenching results at the ambient water quality conditions (Colorado River Water at 23°C and pH 7.9) including (a) molar quenching efficiency and (b) time to reach quenching stoichiometry (qTime_exp-stoi) and quenching completion (qTime_exp-max). The molar ratio for both the parent and active compound are given for Na-Metabi and K-Metabi (e.g., HSO₃⁻ vs. Na-Metabi). For all other QAs, the molar ratios results are the same whether represented as parent and active compound. Initial DO₃ dose = 1.6–1.9 mg/L, QAs were dosed at 13–50% stochiometric mass ratio when residual DO₃ = 0.87–1.05 mg/L. An “Inefficient” QA was defined by (qO₃:QA)_exp-max at < 85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less), an “Efficient” QA was defined by (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th, and a “Greater than Ideal” QA is defined by (qO₃:QA)_exp-max > (qO₃:QA)_th by more than 15%. A “fast” QA was defined by qTime_exp-stoi ≤30s, a “moderately fast” QA was defined by qTime_exp-stoi = 1 min, and a “slow” QA was defined by qTime_exp-stoi >1 min. The quenching reaction was considered “extended” if the reaction continued beyond 30s. Note the black dashed lines in (b) represents the qTime_th, with the qTime_th for H₂O₂ above the scale at 8.5 min.

Table 3. The theoretical (qO₃:QA)_th and experimentally derived ozone to quenching agent molar and mass ratios (given as parent and as active) at 30s (qO₃:QA)_exp-30s and maximum quenching time (qO₃:QA)_exp-max) in Colorado River Water at ambient conditions (23°C and pH 7.9), the theoretical time to quenching completion (qTime_th) and experimental quenching time to reach theoretical stoichiometry (qTime_exp-stoi) and quenching completion (qTime_exp-max.).

Quenching Agent Parent Compound (Active Component)	(qO3:QA)th	(qO3:QA)exp-30s	(qO3:QA)exp-max	Experimental Quenching Efficiency		qTimeth	qTimeexp-stoi/ qTimeexp-max
	Molar Ratio (Mass Ratio-Parent) [Mass Ratio-Active]	Molar Ratio (Mass Ratio-Parent) [Mass Ratio-Active]	Molar Ratio (Mass Ratio-Parent) [Mass Ratio-Active]
				30 sec	Complete
Calcium Thiosulfate CaS2O3 (S2O3^2-)	4 : 1 (1 : 0.79 as parent) [1 : 0.58 as active]	3.04 : 1	3.04 : 1	Inefficient*	Inefficient*	<1s	>10 min/ <30s
		(1 : 1.04)	(1 : 1.04)
		[1 : 0.77]	[1 : 0.77]
Sodium Thiosulfate Na2S2O3 (S2O3^2-)	4 : 1 (1 : 0.82 as parent) [1 : 0.58 as active]	2.72 : 1	3.02 : 1	Inefficient*	Inefficient*	<1s	>6 min/ 1.5 min
		(1 : 1.21)	(1 : 1.09)
		[1 : 0.86]	[1 : 0.77]
Hydrogen Peroxide (H2O2)	2 : 1 ([1 : 0.35])	1.28 : 1 ([1 : 0.56])	1.89 : 1 ([1 : 0.37])	Inefficient*	Efficient**	24s (pH 8.5) ‡ 8.5 min (pH 7.9) ‡ 9.3 min (pH 7.5) ‡	1 min/ 1.5 min
Sodium Metabisulfite Na2S2O5 (HSO3^−)	2 : 1 as parent (1 : 1.98 as parent) 1:1 as active [1 : 1.69 as active]	1.86 : 1 (1 : 2.13) 0.93 : 1 [1 : 1.82]	1.86 : 1 (1 : 2.13) 0.93 : 1 [1 : 1.82]	Efficient**	Efficient**	<2s	<30s/ <30s
Potassium Metabisulfite K2S2O5 (HSO3^−)	2 : 1 as parent (1 : 2.32 as parent) 1:1 as active [1 : 1.69 as active]	1.54 : 1 (1 : 3.00) 0.77 : 1 [1 : 2.19]	1.54 : 1 (1 : 3.00) 0.77 : 1 [1 : 2.19]	Inefficient*	Inefficient*	<2s	>6 min/ <30s
Sodium Bisulfite NaHSO3 (HSO3^−)	1 : 1 (1 : 2.17 as parent) [1 : 1.69 as active]	1.01 : 1 (1 : 2.13) [1 : 1.66]	1.09 : 1 (1 : 2.00) [1 : 1.56]	Efficient**	Efficient**	<1s	<30s/ 1 min
Ascorbic Acid (C6H8O6)	1 : 1 ([1 : 3.59])	1.68 : 1 ([1 : 2.19])	1.89 : 1 ([1 : 1.94])	Greater than Ideal***	Greater than Ideal***	<1s	<30s/
Sodium Sulfite Na2SO3 (SO3^2-)	1 : 1 (1 : 2.63 as parent) [1 : 1.67 as active]	0.91 : 1 (1 : 2.89) [1 : 1.86]	1.02 : 1 (1 : 2.57) [1 : 1.63]	Efficient**	Efficient**	<1s	<30s/ 1.5 min

*“Inefficient” was defined as QAs with (qO₃:QA)_exp-max at < 85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less).

** “Efficient” is defined as a QA with (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th,

***“Greater than Ideal” is defined as a QA with (qO₃:QA)_exp-max >; (qO₃:QA)_th by more than 15%,

‡ The theoretical time to quenching completion was calculated for H₂O₂ using k_obs = 9.6 × 10⁶ x 10^(pH-11.8) and using k_O3’s for the other QAs (Sein et al. 2008; von Sonntag and von Gunten 2015).

Though less common than Ca-Thio and H₂O₂, Na-Bi is currently used as an O₃ QA at select full-scale facilities. Na-Bi quenched O₃ efficiently and fast, i.e., reached (qO₃:QA)_th with a qTime_exp-stoi ≤ 30 s (Figure 2). Na-Bi had an extended reaction time, continuing to react in the first minute after being added to solution (qTime_exp-max = 1 min), quenching 10% more O₃ cumulatively than predicted by stoichiometry (Figure 2), although still within a ± 15% error margin. The Burns et al. study also evaluated Na-Bi and found it was more efficient than Ca-Thio requiring only slight overdosing (12–17% by mass as parent compound) (Burns et al. 2008). Another study evaluated Na-Bi for the Water Disinfection Plant in Incline Village, Nevada (Richey et al. 2000). This study determined that more Na-Bi was required than the stoichiometric amount, 35% (w/w as parent compound) more, but water quality conditions (i.e., pH and temperature) were not stated and thus it is difficult to directly compare results (Richey et al. 2000).

H₂O₂ quenched > 85% (qO₃:QA)_th moderately fast with a qTime_exp-stoi = 1 min (Figure 2b). H₂O₂ had an extended reaction with the completion of O₃ quenching occurring at a qTime_exp-max = 1.5 min (Figure 2b); which was almost 6 times faster than expected, based on the low k_O3 at ambient pH (qTime_th = 8.5 min, Table 2). Hydroxyl radicals (∙OH) are a product of the H₂O₂ + O₃ reaction, and ∙OH may accelerate the decrease in O₃ residual via the hydroxyl radical chain reaction (Sein et al. 2007; Staehelin and Hoigne 1982; von Sonntag and von Gunten 2015). H₂O₂ is defined as a promotor through its conversion of ∙OH to superoxide radicals ($O_2^{\bullet -}$and $\mathit{H}{O}_2^{\bullet }$) which act as chain reaction carriers and further quench the residual O₃ Gardoni et al., (2012). Cumulatively, H₂O₂ quenched O₃ efficiently at 95% of the (qO₃:QA)_th (i.e., (qO₃:QA)_exp-max = 1.9:1.0, Figure 2a). The complex reaction between H₂O₂ and O₃ has been well studied and is known as the peroxone process (Sein et al. 2007; Staehelin and Hoigne 1982; von Sonntag and von Gunten 2015). The yield of ∙OH, through reaction of O₃ with H₂O₂ (Table 1), is purportedly only ~ 50% due to the second elementary step where the HO₅⁻ adduct (formed in the first elementary reaction step) may decompose in two different ways, with each decomposition reaction having similar rates (Merényi et al. 2010). O₃ and H₂O₂ are consumed in the first elementary step to form HO₅⁻ and thus O₃ depletion by H₂O₂ is still expected to occur at the (qO₃:QA)_th, despite the low ∙OH reaction yield. Previous tests evaluating H₂O₂ as a QA have described it as slow to react, requiring 3–5 min, but presumably this descriptor is based on the low reaction rate at typical water pH (Table 2) (Langlais, Reckhow and Brink 2019; Richey et al. 2000). Both Burns et al. and Richey et al. found H₂O₂ required the most overdosing of the three common QAs (H₂O₂, Ca-Thio, Na-Bi) at 78% and 140% higher than the (qO₃:QA)_th (Burns et al. 2008; Richey et al. 2000).

Three of the unconventional O₃ QAs assessed, Na-Metabi, AscAcid, and Na-S, all quenched fast and efficiently (i.e., qTime_exp-stoi ≤ 30 s, (qO₃:QA)_exp-30s and (qO₃:QA)_exp-max ≥85% (qO₃:QA)_th (Figure 2a)). Na-Metabi reached quenching completion the fastest of the three with a qTime_exp-stoi and qTime_exp-max ≤30 s (Figure 2b). Although Na-S reached the stochiometric ratio (±15%) within 30 s, the quenching reaction was extended with qTime_exp-max = 1.5 min. AscAcid was the only QA that showed greater than ideal quenching efficiency. Almost twice (1.9×) as much O₃ was cumulatively quenched by AscAcid than predicted by the stoichiometric ratio of 1 mole of O₃:1 mole of AscAcid (Figure 2a), which was achieved by a slightly extended reaction time of qTime_exp-max = 1 min (Figure 2). These results suggest that that the complexity of the AscAcid + O₃ reaction may not be entirely reflected by the single reaction presented in Table 1. The reaction between O₃ and AscAcid produces dehydroascorbic acid (DHA) which can be reduced back to AscAcid spontaneously under neutral and alkaline pH and by thiols such as cysteine (Jung and Wells 1998); thus, furthering the O₃ quenching ability of AscAcid (Deutsch 1998; Yin et al. 2022). DHA can also rapidly hydrolyze to 2,3-diketogulonic acid, which can be oxidized by O₃ to >50 product compounds theorized to also be reducing agents (Deutsch 1998).

Effect of temperature

During the control tests, the first order O₃ decay rate increased linearly with temperature (Figure S3), indicating accelerated reactions between O₃ and NOM. At 23°C, O₃ decay is nearly 2 times quicker (1.73×) than at 4°C. These results are similar to those presented by Bromley and Wert which found that at pH 7.5 and 8.5, an increase in temperature from 12°C to 22°C decreased the O₃ half-life in CRW by 1.5–2.9 (Bromley and Wert 2002). Abada et al.observed an O₃ half-life more than two times greater (2.2×) at 26°C compared to 15°C in CRW (Abada, Atkinson and Wert 2024). Shin et al. found that the O₃ decay was most significantly impacted by the seasonal variation in water temperature (6–26℃) where the minimum (0.027 min⁻¹) and maximum (0.066 min⁻¹) decay constants were observed when the water was 6°C and 26°C, respectively; a difference of 2.4 (Shin, Hidayat and Lee 2016).

O₃ decay is more rapid at warmer temperatures due to the accelerated reactions with NOM and autodecomposition that increases the ∙OH yield. An increase in ∙OH yield from 0.28 to 0.68 was demonstrated for the same lake water with a temperature increase from 6°C to 26°C (Shin, Hidayat and Lee 2016). It has been posed that the temperature-dependent enhanced ∙OH formation is due to the more favorable solvent-cage effects during O₃ decomposition initiation reactions (Shin, Hidayat and Lee 2016). As the temperature increases, the water has greater fluidity as a solvent and thus more free radicals can escape from the ozone-adduct decomposition cage (Shin, Hidayat and Lee 2016).

O₃ quenching results demonstrated that temperature also affected efficiency and kinetics of reactions between O₃ and QAs, Figure 3. As the temperature increased, the ratio of O₃ quenched immediately, i.e., (qO₃:QA)_exp-30s, increased for all QAs, but this trend did not hold for the cumulative O₃ quenched over time, (qO₃:QA)_exp-max (Figure 3a). The (qO₃:QA)_exp-max appears to be consistent at either the two lower temperatures (e.g., Ca-Thio and Na-Thio) or the two higher temperatures (e.g., H₂O₂ and AscAcid), which may reflect different points of diminishing temperature effects for the different QAs. Surprisingly, with Ca-Thio, Na-Thio, H₂O₂, and AscAcid, more O₃ was quenched over the full quenching time at 4°C than 23°C. This increase in (qO₃:QA)_exp-max at the lower temperature appears to be due to the extended reaction time (before O₃ was fully depleted), with qTime_exp-max = 16 min at 4°C versus qTime_max = 1–1.5 min at 23°C (Figure 3b).

Figure 3. The effect of temperature on (a) molar quenching efficiency and (b) time to reach quenching stoichiometry (qTime_exp-stoi) and quenching completion (qTime_exp-max). All molar ratio results are given for the active compound (e.g., HSO₃⁻ not Na-Metabi or K-Metabi). Initial DO₃ dose = 1.4–2.3 mg/L, the QAs were dosed at 13–81% stochiometric mass ratio when the DO₃ residual was = 0.79–1.45 mg/L into Colorado River Water at pH 7.9. “NA” indicates that an experiment was not conducted for that QA at the corresponding water quality condition. An “Inefficient” QA was defined by (qO₃:QA)_exp-max at < 85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less), an “Efficient” QA was defined by (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th, and a “Greater than Ideal” QA is defined by (qO₃:QA)_exp-max > (qO₃:QA)_th by more than 15%. A “fast” QA was defined by qTime_exp-stoi ≤ 30 s, a “moderately fast” QA was defined by qTime_exp-stoi = 1 min, and a “slow” QA was defined by qTime_exp-stoi >1 min. The quenching reaction was considered “extended” if the reaction continued beyond 30 s. Note the black dashed lines in (b) represents the qTime_th.

Although Ca-Thio and Na-Thio never achieved (qO₃:QA)_th (±15%) at 23°C, they did at 4°C and 15°C (Figure 3a). Ca-Thio and Na-Thio were slow to react at all temperatures tested, but the speed of quenching did increase with temperature resulting in the qTime_exp-stoi decreasing from 6 min to 1.5 min at 4°C and 15°C, respectively. At 4°C, a small increase in (qO₃:QA)_exp was achieved between qTime_exp-stoi = 6 min and qTime_exp-max = 16 min, but this increase was not substantial for Ca-Thio (4%). For Na-Thio, there was a 30% increase in (qO₃:QA)_exp-max achieved between 6 min and 16 min, yet (qO₃:QA)_exp-max = (qO₃:QA)_th ±15%. While greater O₃ was quenched and (qO₃:QA)_th was achieved, it is important to note that either a 6 or 16 min quenching time may not be realistic nor practical at full-scale.

At 4°C, Na-Metabi was unable to quench to (qO₃:QA)_th ±15% in 30 s, instead requiring a qTime_exp-stoi = 1.5 min; yet at 23°C, the qTime_exp-stoi was reduced to 30 s (Figure 3a). Whereas K-Metabi was inefficient at both temperatures and was the only QA that didn’t achieve (qO₃:QA)_th at any temperature, quenching 25% less O₃ than expected (qO₃:QA_max = 1.5:1, Figure 3a). For both QAs, the (qO₃:QA)_exp-max was consistent at high and low temperatures; however, in both cases the kinetics were considered slow at 4°C (qTime_exp-stoi/max > 1 min) and fast at 23°C (qTime_exp-stoi/max <30s) (Figure 3b). The similar results between Na-Metabi and K-Metabi is consistent with their chemical similarities, with bisulfite as their active compound (HSO₃⁻). Similarly, Na-Bi also has HSO₃⁻ as the active component. Na-Bi achieved (qO₃:QA)_th and was fast (qTime_exp-stoi < 30 s) at both 4°C and 23°C. Na-Bi had a slightly extended reaction time at 23°C (qTime_exp-max = 1 min) and achieved a slightly higher (qO₃:QA)_exp-max, though the difference was insignificant (<15%).

AscAcid was also a fast QA at all temperatures (qTime_exp-stoi < 30 s). AscAcid had a greater than ideal efficiency, achieving >115% (qO₃:QA)_th in ≤ 30 s (Figure 3a), and quenching was extended at all temperatures but with a decreasing trend in qTime_exp-max (Figure 3b). The greatest (qO₃:QA)_exp-max was achieved at 4°C, quenching 3.6× more O₃ than predicted based on stoichiometry. Similarly, H₂O₂ quenched nearly two times (1.7×) more O₃ than expected ((qO₃:QA)_exp-max = 3.4:1.0) at 4°C. Shin et al. observed decreasing ∙OH yield and subsequent exposure at colder temperatures (6°C vs. 26℃) and suggested NOM moieties are less reactive at lower temperatures (Shin, Hidayat and Lee 2016). At 4°C, the O₃ decomposition via the hydroxyl radical chain reactions also slows (Staehelin and Hoigne 1982; Staehelin and Hoigne 1985). Considering this, the greater than ideal quenching efficiency observed for AscAcid and H₂O₂ may then be due to less reactions between O₃ and ∙OH or QAs and NOM which allowed for extended reactions between O₃, the QA, and their promotor-induced by-products (Gardoni, Vailati and Canziani 2012).

Note, other studies on O₃ quenching evaluated the effect of temperature but did not distinguish changes in QA dosing requirements due to changes in O₃ demand-decay versus quenching efficiency/kinetics. The (qO₃:QA)_exp-max values at all temperatures are provided in Supplemental Information Table S5 in molar units both as-parent compound and as-active compound.

Effect of pH

Results from the control tests (without QA) showed an accelerated decrease in O₃ residual over time (O₃ decay rate) as pH increased, with the half-life of O₃ 2× higher at pH 7.5 (1.9 min) than at pH 8.5 (0.93 min). These results are consistent with those from Abada et al. (2024), where pilot tests with CRW at pH 7.5–7.8 and 26°C, gave the same O₃ half-life of 1.9 min. As pH increases, O₃ decay is accelerated due to the hydroxide-induced decomposition reactions and subsequent increase in ∙OH production (Elovitz, von Gunten and Kaiser 2000; Gardoni, Vailati and Canziani 2012; Staehelin and Hoigne 1982). Alkaline conditions also enhance the reactions between O₃ and NOM due to deprotonation of the phenol and acid moieties (Elovitz, von Gunten and Kaiser 2000).

Figure 4a shows that the efficiency of most quenching reactions evaluated were also pH dependent. Ca-Thio and K-Metabi were the only two QAs that quenched < 85% (qO₃:QA)_th over the range of pHs evaluated, at 50–75% and 72–77% of the (qO₃:QA)_th, respectively. Ca-Thio did show some pH dependency where (qO₃:QA)_exp-max increased from 2:1 to 3:1 with a pH increase in CRW of 7.5 to 7.9, but remained at 3:1 for pH = 8.5. As mentioned in the temperature results section, Burns et al. also demonstrated the need to overdose Ca-Thio, similarly observing a (qO₃:QA)_exp-max = 2.4:1 (Burns et al. 2008). Similar to Ca-Thio, efficiency of quenching with Na-Thio displayed pH dependence but to a lesser extent with (qO₃:QA)_exp-max ranging from 3.0:1 to 3.8:1. Na-Thio quenched < 85% (qO₃:QA)_th at pH 7.5 and 7.9, but did quench to (qO₃:QA)_th at pH 8.5. Similarly, Bromley and Wert found that Ca-Thio and Na-Thio could be dosed at the stochiometric ratios but concluded a safety factor of 10–20% overdosing should be maintained (Bromley and Wert 2002). The lower quenching efficiency at lower pH could be attributed to the speciation of the active component, the thiosulfate ion (S₂O₃^2-). Takizawa, Okuwaki and Okabe (1973) showed that under neutral pH, ~40% of the S₂O₃^2- is oxidized by O₃ to sulfite (SO₃^2-) which is further oxidized by O₃ to sulfate (SO₄^2-) but many by-products were also produced including tri- and tetrathionates (S₃O₆^2-, S₄O₆^2-) and hydrogen sulfide/sulfur dioxide (H₂S/SO₂ gas). Yet, in an alkaline solution, S₂O₃^2- is oxidized by O₃ to sulfite SO₃^2- which has a higher k_O3 then thiosulfate. The sulfite SO₃^2- is then fully oxidized to sulfate (SO₄^2-), with none of the other previously mentioned byproducts (Takizawa, Okuwaki and Okabe 1973). The observed difference in O₃ + S₂O₃^2- products highlights the effect pH has on the reaction chemistry and why quenching with thiosulfate-based QAs may be more efficient at higher pHs.

Figure 4. The effect of pH on (a) molar quenching efficiency and (b) time to reach quenching stoichiometry (qTime_exp-stoi) and quenching completion (qTime_exp-max). All molar ratio results are given for the active compound (e.g., HSO₃⁻ not Na-Metabi or K-Metabi). Initial DO₃ dose = 1.3–1.9 mg/L, the QAs were dosed at 6–68% stochiometric mass ratio when the DO₃ residual = 0.47–1.32 mg/L into Alfred Merritt Smith Water Treatment Facility raw water at 23°C. “NA” indicates that an experiment was not conducted for that QA at the corresponding water quality condition. An “Inefficient” QA was defined by (qO₃:QA)_exp-max at < 85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less), an “Efficient” QA was defined by (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th, and a “Greater than Ideal” QA is defined by (qO₃:QA)_exp-max > (qO₃:QA)_th by more than 15%. A “fast” QA was defined by qTime_exp-stoi ≤30s, a “moderately fast” QA was defined by qTime_exp-stoi = 1 min, and a “slow” QA was defined by qTime_exp-stoi >1 min. The quenching reaction was considered “extended” if the reaction continued beyond 30s. Note the black dashed lines in (b) represents the qTime_th, with the qTime_th for H₂O₂ above the scale for pH 7.9 at 8.5 min and pH 7.5 at 9.3 min.

Time to O₃ quenching completion by Ca-Thio did not substantially change with pH, with qTime_exp-max being ≤ 30 s–1 min at all three pHs (Figure 4b); however, Na-Thio was slower to quench O₃ with increasing pH, qTime_exp-max of ≤ 30 s at pH 7.5, 1.5 min at pH 7.9, and 3 min at pH 8.5 (Figure 4b). Bromley and Wert’s experiments also indicated that the Ca-Thio and Na-Thio quenching reactions were completed within 30 s, with the O₃ decay rate returning to that of the control tests 30 s after quenching (Bromley and Wert 2002).

Efficiency of quenching by H₂O₂ and AscAcid had an apparent pH dependence, with (qO₃:QA)_exp-max increasing with pH (Figure 4a). These results demonstrate alkaline conditions as more favorable for O₃-QA reactions with H₂O₂ and AscAcid, as well as the potential secondary reactions discussed previously. AscAcid quenched > 85% (qO₃:QA)_th within 30 s at all 3 pH conditions with qTime_exp-stoi ≤ 30s and qTime_exp-max ≤ 30 s–1 min. H₂O₂, however, required more time at pH 7.5 and 7.9 to reach (qO₃:QA)_th ±15% with a qTime_exp-stoi of 2 min at pH 7.5 and 1 min at pH 7.9 (Figure 4b). Additional time to reach quenching completion was also needed for H₂O₂ with qTime_exp-max = 4 min, 1.5 min, and 1 min at pH = 7.5, 7.9, and 8.5, respectively. The slower quenching kinetics of H₂O₂ have previously been reported⁴ (Bromley and Wert 2002) where O₃ was quenched < 0.1 mg/L (with [O₃]_i = 0.3–0.8 mg/L) within 2 min at pH 7.5 and within 1–1.5 min at pH 8.5. The pH dependence of the O₃-H₂O₂ reaction rate (k_O3) is described by equation 5 (von Sonntag and von Gunten 2015).

[3] $k_{\mathit{obs}}=9.6\times {10}^6\times {10}^{\left(\mathit{pH}-11.8\right)}$[3]

The closer the pH gets to the H₂O₂ pK_a = 11.8 (Sein et al. 2008), the greater the proportion of H₂O₂ is deprotonated and exists as the more O₃ reactive HO₂⁻ species (von Sonntag and von Gunten 2015; Langlais, Reckhow, and Brink 2019). HO₂⁻ has a greater k_O3 (9.6 × 10⁶ M⁻¹s⁻¹) (Sein et al. 2007) than H₂O₂ (<0.01 M⁻¹s⁻¹) (Gardoni, Vailati and Canziani 2012; Staehelin and Hoigne 1982; von Sonntag and von Gunten 2015; Langlais, Reckhow, and Brink 2019). The pattern observed in the experimental results for H₂O₂ are consistent with the patterns shown by the kinetic modeling, which utilized equation 5. Quenching was predicted to be fastest at pH 8.5 and slower at pHs 7.9 and 7.5; however, under all conditions the modeling predicted quenching to be >2–9× slower (at 52s, 8 min, and 9 min, respectively) than was observed in the experiments. This accelerated decrease in O₃ residual is likely due to the secondary reactions between byproducts (e.g., •OH) and O₃, as discussed in previous section.

AscAcid quenched more O₃ than expected for all 3 pH conditions with (qO₃:QA)_exp-max increasing as pH increased (130% (qO₃:QA)_th at pH 7.5, 170% (qO₃:QA)_th at pH 7.9, and 220% (qO₃:QA)_th at pH 8.5, Figure 4a). As described in the previous section, the reaction between O₃ and AscAcid could produce >50 product compounds (e.g., DHA, DKG) through a series of reducing, hydrolyzing, and oxidizing reactions. These product compounds may be more reactive with O₃ at higher pH (Chang et al. 2021) which could explain the increasing (qO₃:QA)_exp-max and how AscAcid quenched 2.2× greater O₃ than the stoichiometric ratio at pH 8.5. The O₃ quenching reactions by AscAcid occurred rapidly for all 3 pH conditions with qTime_exp-stoi ≤ 30 s and qTime_exp-max < 1 min (Figure 4b), somewhat consistent with the kinetic modeling which theorized the O₃-AscAcid reaction to occur within 1 s.

Results showed that Na-S, Na-Bi, and Na-Metabi were efficient, quenching the expected amount of O₃ ((qO₃:QA)_th ±15%) with qTime_exp-stoi ≤ 30 s; though when reactions continued past qTime_exp-stoi, both Na-S and Na-Bi quenched > (qO₃:QA)_th at pH 7.5 (Figure 4). The (qO₃:QA)_exp-max for these three QAs was affected by the change in pH from 7.5 to 7.9, with less O₃ quenched at the higher pH. Bisulfite (HSO₃⁻) is a weak acid (pK_a = 6.97), with the conjugate base of sulfite (SO₃^2-); thus ultimately, the active component of Na-S, Na-Bi, Na-Metabi, and K-Metabi is mostly SO₃^2- under the experimental pH conditions. SO₃^2- is a more favorable species for O₃ quenching than HSO₃⁻ with a k_O3 ~3–4 orders of magnitude higher (Langlais, Reckhow, and Brink 2019). K-Metabi was the only sulfite-based QA that quenched less O₃ than expected, (qO₃:QA)_exp-max was 26–31% lower than (qO₃:QA)_th at both pH 7.5 and pH 7.9.

Effect of diffuser on QA dispersion and downstream hydraulic mixing

Hydraulics, quenching chemical diffuser design, and diffuser location can affect QA mixing and dispersion and thus affect overall quenching efficacy in a system. Due to the non-ideality of full-scale flow through systems such as water treatment plants and specifically O₃ contactors, heterogenous mixing and subsequent non-uniformity to the molecular level is expected (Crittenden et al. 2012). Residence time distribution (RTD) is an important concept when designing O₃ quenching systems, as it reflects the various paths and characterizes variability of contact times the O₃ and/or QA molecules have in the reactor (Crittenden et al. 2012). For many years, tracer testing and RTD analysis has been performed to characterize reactor ideality with single quantitative values e.g., T₁₀ or T₁₀/T (K. L. Rakness et al. 2005; USEPA 1991; USEPA 2010; USEPA 2020). Typical full-scale ozone contactors are in the average (T₁₀/T = 0.5) to superior (T₁₀/T = 0.7) baffling classifications, meaning there is going to be a level of mixing non-ideality (Bellamy 1995; Roustan et al. 1993; Teefy 1996). The development of computational fluid dynamic (CFD) analysis has allowed reactors to be designed and/or optimized without assuming uniformity and in some cases without tracer testing and providing a means of more fully and visually characterizing mixing (non-)ideality. CFD is the science of predicting fluid flow by solving the governing equations that represent the three fundamental principles of fluid flow: the conservation of mass, momentum, and energy. Under specified conditions, a CFD model simulates flow fields and particle transport within the model domain, solving the Reynolds-averaged, Navier-Stokes equations that describe the flux of mass and momentum within a fixed domain subject to specified boundary conditions. To illustrate the effect QA diffusion can have, a CFD analysis was performed for a full-scale O₃ contactor with a Ca-Thio quenching system (Figure S6), comparing three Ca-Thio diffuser designs:

• 1” diameter stainless steel diffuser placed parallel but slightly above the top of the baffle on the upstream side, with twenty-two ¼-inch diameter orifices oriented parallel and counter-current to flow, spaced 2-feet on center. The diluted Ca-Thio stock flow rate was 17 gpm.

• 3” diameter stainless steel diffuser placed parallel but slightly above the top of the baffle on the upstream side, with forty-five ¼-inch diameter orifices oriented down, spaced 1-feet on center. The diluted Ca-Thio stock flow was 70 gpm.

• Two 3” diameter stainless steel diffusers placed parallel to each other and slightly below the top of the baffle on the upstream side, with eighty-nine ¼-inch diameter orifices total, 44 orifices on one diffuser and 45 on the other, orifices oriented down, spaced 1-feet on center on each diffuser with orifices staggered from the other diffuser. The diluted Ca-Thio stock flow was 200 gpm.

Results of the CFD analysis are given in Figure 5. When the smaller 1” diameter pipe diffuser (22 orifices) was used, there was poor mixing performance with a maldistribution of Ca-Thio across the diffuser orifices being maintained to the end of the contactor (Figure 5a). There were substantial portions of ozonated water in the contactor and effluent gate that are less than 40% of the ideal Ca-Thio mass fraction (dark blue contours) and low in concentration (light blue contours) due to the poor dispersion and mixing with the 1” diffuser design. This means, if O₃ residual was still present, there’s a risk that a portion of O₃ would not have been quenched. Figure 5b,c show that the other two diffuser designs, with the larger 3” pipe and more orifices (45 and 89), produced relatively homogenous distribution at the Ca-Thio target concentration (green contours) at the effluent gate due to the increased velocity differential between the pipe and the orifices. The design with the single 3” diffuser pipe produced a more even distribution of Ca-Thio consistently and with less areas without Ca-Thio at the underbaffle between cells 11 and 12 than the double 3” diffuser pipe design. The improved performance of the single diffuser head may be due to the difference in location of the diffusers, i.e., slightly above the over baffle for the single diffuser vs. slightly below the over baffle for the double diffuser and suggest that more of the ozonated water was in contact with Ca-Thio for a longer contact time (not just at the effluent gate). Ultimately, within the level of accuracy of the model, there was no attributable difference in mixing performance between the single and double 3” diffuser designs for the relevant location, the outlet of Cell 12. The single 3” diffuser with 45 orifices was recommended for the quenching diffuser design at this full-scale plant because it required one less diffuser pipe and half the flow rate, thus lower pumping requirements. This diffuser design was also assessed under varying water flow scenarios, maximum, average, and minimum flow, to ensure that Ca-Thio was well mixed under the range of hydrodynamics that may be encountered. This QA mixing examination shows that the diffuser design and hydrodynamics are important, additional factors that should be considered when designing and/or troubleshooting an O₃ quenching system. Further, these results demonstrate CFD analysis is a powerful tool that can provide insight into QA diffusion and mixing. Given the lack of existing guidance on QA diffuser design, mixing of QA may be a major contributor to O₃ residual carryover in many full-scale systems. Analysis of QA diffusion, such as this CFD analysis, may identify this often-overlooked aspect of an O₃ quenching system.

Figure 5. Computational fluid dynamic (CFD) analysis results showing the Ca-Thio distribution in an example ozone contactor that resulted from three diffuser designs: (a) 1” diameter pipe diffuser with orifices every 2 ft resulting in 22 orifices, (b) 3” diameter pipe diffuser with orifices every 1 ft resulting in 45 orifices, (c) two 3” diameter pipe diffusers with 44 orifices on one diffuser and 45 orifices on the other 1 ft apart and staggered resulting in 89 orifices. Colors represent Ca-Thio mass fraction of the total fluid mass. Green contours represent the ideal mass fraction (completely mixed), warmer colored contours represent higher than ideal mass fraction, and colder colored contours represent less than ideal mass fraction. The upper (bright red) and lower (dark blue) bounds of the color scale illustrate that the fluid is more than 40% or less than 40% of Ca-Thio ideal mass fraction, respectively. Ca-Thio injection at the under baffle between cells 11 and 12 is represented in the middle of each image and the Ca-Thio distribution at the two effluent gates leaving cell 12 and entering the combined effluent channel is represented on the right of each image. Figures were generated for SNWA by CDM Smith.

Practical implications

Mass ratio

Quenching efficiency results presented earlier, in pH and temperature sections, were used to calculate the mass of QA required to quench 1 mg of O₃ during an up to 5 min contact time, shown in Figure 6. The mass ratio can be used directly by practitioners and is a practical metric to compare dosing requirements for each QA under these various conditions. The mass results in Figure 6 are given for the parent compound, referring to the total mass of chemical required for practitioners. All experimental mass ratio dosing requirements, for each condition, represented in units both as parent and as active compounds, are compiled in the Supplemental Information Table S7 and S8. As discussed in previous sections, AscAcid, H₂O₂, Na-S, Na-Bi, and Na-Metabi were either efficient, quenching the stoichiometric ratio (qO₃:QA)_th ±15%, or quenched more O₃ than predicted by stoichiometry under all tested conditions. Ca-Thio and Na-Thio quenched less O₃ than predicted by stoichiometry for most conditions (Figure 6).

Figure 6. The mass ratio requirements (qO₃:QA)_exp-5 min for each QA, assuming 5 minutes of quenching time is available, at various (a) temperature conditions and (b) pH conditions. All results are given in mass units as the parent compound (e.g., Na-Metabi not HSO₃⁻). The QAs were dosed at 6–81% stochiometric mass ratio with an O₃ dose = 1.3–2.3 mg/L into Alfred Merritt Smith Water Treatment Facility raw water. “NA” indicates that an experiment was not conducted for that QA at the corresponding water quality condition. An “Inefficient” QA was defined by (qO₃:QA)_exp-max at < 85% of the (qO₃:QA)_th, meaning 15% less O₃ was quenched than expected (or even less), an “Efficient” QA was defined by (qO₃:QA)_exp-max within ± 15% of the (qO₃:QA)_th, and a “Greater than Ideal” QA is defined by (qO₃:QA)_exp-max > (qO₃:QA)_th by more than 15%. A “fast” QA was defined by qTime_exp-stoi ≤30s, a “moderately fast” QA was defined by qTime_exp-stoi = 1 min, and a “slow” QA was defined by qTime_exp-stoi >1 min. The quenching reaction was considered “extended” if the reaction continued beyond 30s.

In practical considerations, however, a higher mass of AscAcid and the sulfite-based QAs are required to quench an equivalent mass of O₃ when compared to Ca-Thio or Na-Thio, under most conditions. The higher mass requirements are due to the higher molecular weights of these QAs and lower (qO₃:QA)_th ratios. K-Metabi had the highest molecular weight and was the most inefficient QA, requiring 3–3.2 mg to quench 1 mg O₃; which is 3 times higher than the Ca-Thio and Na-Thio dosing requirements. At ambient conditions, Na-Metabi and Na-Bi required the same mass ratio of 2 mg ± 15%, to quench 1 mg of O₃; while Na-S required slightly more, 2.6 mg (Figure 6), and again, all 3 of these QAs had 2–2.6× higher mass requirements than Ca-Thio or Na-Thio.

In terms of water quality effects, pH and water temperature did not have as apparent of an effect on mass requirements (5 min contact time) as observed with molar ratios. For the thiosulfate-based reagents, pH had the greatest effect with pH 7.5 requiring 60% greater Ca-Thio mass than pH 7.9 and 8.5 (Figure 6a,b). AscAcid requirements were the most affected by pH and temperature conditions. The mass requirement increased by 41% when temperature increased from 4°C to 15°C and 23°C, and the mass requirement decreased by ~50% when pH increased from 7.5 to pH 7.9. The trend with pH for Na-S and Na-Bi was opposite of this, with mass requirements increasing by ~60% for Na-S when pH was increased from pH 7.5 to 7.9.

AscAcid, despite its greater than ideal quenching efficiency at all conditions, has the second highest molecular weight of QAs evaluated, resulting in mass requirements comparable to Na-Metabi and Na-Bi at ambient conditions. The primary exception was when the water was colder (4℃); the mass requirement after 5 min for Ca-Thio, Na-Thio, H₂O₂, Na-Metabi, K-Metabi, and Na-Bi remained roughly unchanged yet the mass requirement for AscAcid decreased significantly to within 20% of the required mass for Ca-Thio and Na-Thio. Therefore, for utilities in consistently cold climates, AscAcid may be an advantageous choice due to its quenching efficiency and nontoxic nature.

Under all tested conditions, H₂O₂ had the lowest mass requirements overall, in the range of 0.32–0.39 mg of H₂O₂ to quench 1 mg of O₃ (Figure 6). For utilities with seasonal changes in water temperature or a fluctuating pH (±0.5 pH units), H₂O₂ demonstrated the most consistent quenching efficiency and lowest mass requirements.

Application benefits and barriers

The choice of QA will be determined by a number of factors relating both to quenching efficiency and water chemistry, as well as practical application concerns. As described in the introduction, handling (e.g., worker/storage safety), chemical form (i.e., liquid or powder, stock concentration), required storage and monitoring requirements, toxicity concerns, potential downstream effects, local or regional supply reliability, compliance requirements (e.g., NSF 60 certification), and operating cost will all be considered in a utility’s choice of QA. Table 4 describes the chemical form (thus QA dose delivery), supply chain (defined by number of manufacturers), and hazard classifications (described by NFPA Codes) for each of the QAs evaluated in this study.

Table 4. Dosing (chemical form), supply chain (number of NSF 60 certified manufacturers worldwide and in US), and handling (represented by National Fire Protection Association (NFPA) classifications) considerations for each quenching agent.

Quenching Agent	Chemical Form	NSF 60 Certified Manufacturers		NFPA Classifications
		Worldwide	U.S.	Health	Flammability	Instability	Physical Hazards
Calcium Thiosulfate	30–50% in water	9	6	1	1	0	N/A
→ CAPTOR®	24% in water	4	2	1	0	0	N/A
Sodium Thiosulfate	30% in water	2	2	1	0	0	N/A
Hydrogen Peroxide	7–70% in water	51	36	3	0	2	OX
Sodium Metabisulfite	Dry chemical	10	3	3	0	1	N/A
Potassium Metabisulfite	Dry chemical	0	0	3	0	0	N/A
Ascorbic Acid	Dry chemical	1	1	1	1	1	N/A
Sodium Sulfite	Dry chemical, 15% in water	2	1	1	0	1	N/A
Sodium Bisulfite	10–44% in water	63	47	3	0	2	N/A

Ca-Thio is a desirable QA due to its non-corrosive nature and chemical stability, meaning Ca-Thio storage does not require insulation, special transport/delivery, or accounting for off-gas concerns. It is safe for worker handling, and it also does not have environmental toxicity (Hardison 2000; Richey et al. 2000). Little information is available that details the handling and use of Na-Thio in bulk. It has a low molecular weight and is relatively stable and non-reactive, likely with similar requirements as Ca-Thio (Esseco USA 2013). The sulfite-based QAs require more handling related concerns and have greater mass ratio dosing requirements for quenching O₃. Na-S and Na-Bi have stochiometric (qO₃:QA)_th molar ratios of 1:1, while Na-Metabi and K-Metabi have ratios of 2:1; therefore, they each require more chemical to achieve similar quenching as Ca-Thio and Na-Thio (4:1). Na-S is nontoxic and non-corrosive but has a high freezing temperature, and therefore requires special temperature control and storage design when used in colder climates. Na-Bi is corrosive, has a high freezing temperature, and is a dermal irritant requiring special handling (Richey et al. 2000). Richey et al. reported that Na-Bi released SO₂ gas and required care in handling due to crystallization at temperatures < 60℉, which were all reasons Na-Bi was not selected as the QA for use at full-scale in the case study (Richey et al. 2000).

H₂O₂ is chemically reactive and will decay over time; thus, it needs to be delivered more frequently than required for other QAs (Richey et al. 2000). It has a low molecular weight and a 2:1 stoichiometric molar ratio, requiring the lowest mass concentration of any of the tested QAs. H₂O₂ also reacts with O₃ to create hydroxyl radicals and superoxides (AOP) which have the added benefit of oxidizing taste and odor compounds and other recalcitrant chemicals of concern that are not O₃ reactive (Coffey 2000; Richey et al. 2000).

AscAcid is a naturally produced (vitamin c), non-corrosive, biologically benign or even beneficial compound with high O₃ and hydroxyl radical reaction rates (Peterka 1998). AscAcid has a high molecular weight, requiring the highest mass concentration, based on stoichiometry, of the tested QAs. Although at the tested conditions, AscAcid required similar if not less than the sulfite-based QAs. With few evaluations of AscAcid as an O₃ QA, if any besides this study, there is a lack of example use cases particularly at higher application scales.

Compliance is an important consideration. In the US, it is required that all chemicals used in water treatment be NSF 60 certified for the approved application, in most cases. NSF certification is obtained by the supplier and thus for some chemicals there may be a limited number of suppliers. The number of manufacturers and ability to acquire these chemicals from multiple suppliers is paramount for redundancy and supply flexibility, which is particularly important for critical infrastructure such as water utilities. Chemicals with limited manufactures and distributors leave utilities, who rely on the chemical for operation, vulnerable to sudden changes in demand, supply chain disruption, and chemical price determinations/changes. Utilities likely would have no option but to pay inflated costs, reduce water production, or make drastic and sudden changes to ozonation or quenching processes in emergency situations. Table 4 shows the number of NSF60 certified worldwide and U.S. manufacturers for each QA. Na-Bi and H₂O₂ have >50 manufactures worldwide and >35 in the U.S. due to their historical use in treatment and as O₃ QAs. Ca-Thio has <10 suppliers in the world. CAPTOR® which is the patented Ca-Thio QA solution sold to water treatment plants, has only 2 manufactures in the U.S. Na-Thio and Na-S also only have 2 manufactures available, while AscAcid is limited to a single manufacturer. K-Metabi is the only QA tested that is a not an NSF60 certified chemical for drinking water application; therefore, its use is disallowed in the US until certification is obtained or if exceptions by regulating agencies are granted.

Conclusions

• H₂O₂, which has the benefit of producing AOP conditions for taste and odor/micropollutant removal, has minimal handling concerns and the lowest mass requirements (e.g., 1 mg O₃: 0.32–0.39 mg H₂O₂). H₂O₂ may be an ideal QA for utilities that experience seasonal water quality changes as it was the least variable, in terms of efficiency, with temperature and pH (e.g., only 8% difference in mass requirement at pH 7.5, 7.9, and 8.5).

• Ca-Thio and Na-Thio demonstrated lower quenching efficiency compared to other QAs, yet still had lower mass requirements and less handling concerns.

• The sulfite-based QAs overall were efficient but had higher mass requirements and more extensive handling and storage concerns compared to other QAs.

• AscAcid had greater than ideal quenching efficiency and minimal handling concerns yet had high mass requirements. AscAcid may be a viable alternative QA and most beneficial in colder water temperatures where mass requirements (based on 5 min contact time) were comparatively low.

• Temperature had significant effects on initial quenching efficiency (within 30 s), with less quenching initially occurring at 4°C compared to 23°C. With extended contact time, more cumulative O₃ quenching occurred at 4°C compared to 23°C; which allowed Na-Thio, Ca-Thio, and Na-Metabi to achieve (qO₃:QA)_th ±15%. Extended contact time may not always be practical though and quenching speed vs. range of contact times expected should be considered.

• At 4°C, AscAcid and H₂O₂ were able to achieve high quenching efficiency (>115% (qO₃:QA)_th) within 5 min, and if longer contact time is available, could achieve even greater quenching efficiency.

• pH had less effects on quenching efficiency compared to temperature fluctuation, except with Na-S and Na-Bi, which had decreased quenching efficiency.

• Hydrodynamics, including QA diffuser design, were shown as critical factors to consider in assessing quenching at full-scale. CFD analysis was demonstrated to be a powerful tool to enable this analysis.

• Supply chain limitations should be considered when choosing a QA and was generally assessed by the number of NSF certified manufacturers. The alternative QAs AscAcid, Na-S, and K-Metabi had the least number of NSF certified manufacturers. Captor (Ca-Thio) and Na-Thio also had a very low number of US and Worldwide NSF certified manufacturers, compared to H₂O₂ and Na-Bi.

Acknowledgments

We would like to acknowledge Emily Christensen and Gabriela Portillo-Chamul for their assistance in experimental preparation and sample collection. We would like to acknowledge CDM Smith for production of the CFD figures and review, including Chris Schulz, Ben Finnegan, and the CFD team led by Carrie Knatz, consisting of: Stephanie Mericka, Sri Pathapati, Sudheer Reddy, and Mark Allen.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/01919512.2024.2366239.