Updated Reaction Pathway for Dichloramine Decomposition: Formation of Reactive Nitrogen Species and N-Nitrosodimethylamine

Abstract

The N-nitrosodimethylamine (NDMA) formation pathway in chloraminated drinking water remains unresolved. In pH 7–10 waters amended with 10 μM total dimethylamine and 800 μeq Cl₂·L⁻¹ dichloramine (NHCl₂), NDMA, nitrous oxide (N₂O), dissolved oxygen (DO), NHCl₂, and monochloramine (NH₂Cl) were kinetically quantified. NHCl₂, N₂O, and DO profiles indicated that reactive nitrogen species (RNS) formed during NHCl₂ decomposition, including nitroxyl/nitroxyl anion (HNO/NO⁻) and peroxynitrous acid/peroxynitrite anion (ONOOH/ONOO⁻). Experiments with uric acid (a ONOOH/ONOO⁻ scavenger) implicated ONOOH/ONOO⁻ as a central node for NDMA formation, which were further supported by the concomitant N-nitro-dimethylamine formation. A kinetic model accurately simulated NHCl₂, NH₂Cl, NDMA, and DO concentrations and included (1) the unified model of chloramine chemistry revised with HNO as a direct product of NHCl₂ hydrolysis; (2) HNO/NO⁻ then reacting with (i) HNO to form N₂O, (ii) DO to form ONOOH/ONOO⁻, or (iii) NHCl₂ or NH₂Cl to form nitrogen gas; and (3) NDMA formation via ONOOH/ONOO⁻ or their decomposition products reacting with (i) dimethylamine (DMA) and/or (ii) chlorinated unsymmetrical dimethylhydrazine (UDMH-Cl), the product of NHCl₂ and DMA. Overall, updated NHCl₂ decomposition pathways are proposed, yielding (1) RNS via ${\text{NHCl}}_2\to \text{HNO}/{\text{NO}}^{-}\stackrel{{\text{O}}_2}{\to }\text{ONOOH}/{\text{ONOO}}^{-}$ and (2) NDMA via $\text{ONOOH}/{\text{ONOO}}^{-}\stackrel{\text{UDMH}-\text{Cl}\text{or}\text{DMA}}{\to }\text{NDMA}$.

Graphical Abstract

INTRODUCTION

Chloramines are commonly used drinking water disinfectants in the United States.¹ While chloramination is associated with a decreased formation of regulated trihalomethanes and haloacetic acids, N-nitrosodimethylamine (NDMA) is a concern² given its formation at toxicologically relevant levels.³ Some NDMA precursors are composed of secondary amines such as dimethylamine (DMA, (CH₃)₂NH)⁴ as well as tertiary amines and quaternary ammonium compounds with DMA functional groups. These precursors include pharmaceuticals and personal-care products,⁵ veterinary antibiotics,⁶ anion exchange resins,⁷ and unidentified components of pipeline materials.⁸ In the conversion of DMA to NDMA, the nitrogen and oxygen in the added NO group have been attributed to chloramines⁹ and dissolved oxygen (DO), respectively.^10–12

Early studies with DMA as a model precursor postulated a two-step NDMA formation pathway in which monochloramine (NH₂Cl) and DMA reacted to form unsymmetrical dimethylhydrazine (UDMH), followed by a NH₂Cl and UDMH reaction to form NDMA.^13,14 Schreiber and Mitch¹⁰ cast doubt on this pathway by (i) showing that NDMA yields with NH₂Cl and DMA were about 100 times greater than those with NH₂Cl and UDMH and (ii) noting that the rate constant for UDMH formation¹⁵ (0.081 M⁻¹·s⁻¹) was about 100 times lower than needed (6.4 M⁻¹·s⁻¹) to simulate their NDMA data.¹³ This led to a revision and replacement of the NDMA–chloramine formation pathway¹⁰ in which dichloramine (NHCl₂) was proposed as the primary reactant. This reaction pathway from Schreiber and Mitch¹⁰ (SM) combines (i) 14 reactions from the unified (UF) model of chloramine chemistry¹⁶ (Table S1, U1–U14) with (ii) 8 reactions that culminate in NDMA formation (Table S2, P1–P8) and associated acid–base chemistry (Table S3), referred to herein as the UF + SM model. This model is shown as Pathway A1 of Scheme 1 where, for clarity, only UF model reactions U3–U9 and NDMA formation reactions P5, P7, and P8 are presented. For NDMA formation, the first step is a reaction between NHCl₂ and DMA (P5) to form chlorinated unsymmetrical dimethylhydrazine (UDMH-Cl).

Scheme 1. NDMA Formation Stemming from NHCl₂ Decomposition in the Presence of DMA; See ^{Table S1} (U1–U14) for UF Model Reactions^a.

^aA1: currently accepted NHCl₂–DMA reaction pathway represented by the UF + SM model;¹⁰ see Table S2 for P1–P8 of the UF + SM model (replaced with A2 in the UF + RNS model). B1: proposed RNS scheme for NHCl₂ hydrolysis to HNO and subsequent formation of N₂O, ONOOH, and other products; see Tables S3 and S4 for the relevant equilibrium reactions and principal RNS model reactions, respectively. In the presence of DMA (R7), ONOOH (shown) and/or radicals formed from ONOOH/ONOO⁻ decomposition (not shown) react with DMA to form NDMA. A2: proposed replacement of A1 in which ONOOH (shown) and/or radicals formed from ONOOH/ONOO⁻ decomposition (not shown) react with UDMH-Cl to form NDMA. The UF + RNS model is composed of A2 and B1, the latter of which includes the 117 reactions in the ONOOH/ONOO⁻ model of Kirsch et al.¹⁷ with only the principal end products shown for simplicity. Reactions [U#], [P#], [E#], and [R#] correspond to those listed in Tables S1–S4, respectively. Blue arrows and text denote the previously published reactions for which their rate constants were modified to fit the NH₂Cl, NHCl₂, DO, and NDMA kinetic data at pH 7–10 (see Table S9), and magenta arrows and text denote the new reactions and rate constants added to the UF + RNS model to fit these same data (see Table 1).

Next, UDMH-Cl reacts with DO to form NDMA (P8) or with NHCl₂ to form other products (P7). To curb NDMA formation, Pathway A1 emphasizes minimizing the NHCl₂ Concentration¹⁸ or removing and/or transforming DMA-like precursors.¹⁹ However, P8 (UDMH-Cl + DO) is spin-forbidden and DO consumption was not previously kinetically validated,¹⁰ prompting others to investigate other potential reaction pathways. Subsequent studies have found that NDMA yields from NHCl₂ and NH₂Cl were dependent on the pH and precursor type. For example, in waters containing ranitidine, a high-yielding NDMA precursor containing a DMA functional group, NDMA yields at pH 8 were lower with NHCl₂ (ca. 47% yield) compared to NH₂Cl (ca. 80% yield).¹² However, at pH 7, Huang et al.²⁰ found greater NDMA with NHCl₂ compared to NH₂Cl with four N,N-dimethyl-α-arylamine precursors, including ranitidine. They computationally rationalized the contrary findings, attributing these to high NHCl₂ doses. Selbes et al.²¹ found that NH₂Cl preferentially reacted with precursors containing electron-withdrawing groups and NHCl₂ with precursors containing electron-donating groups.

Radical species in chloramine systems may also be relevant in NDMA formation. Spahr et al.¹¹ studied NH₂Cl and ranitidine reaction kinetics at pH 8 and showed that radical scavengers (e.g., ABTS and Trolox) shut down NDMA formation, concluding that aminyl radicals (from NH₂Cl oxidation of organic amines) and N-peroxyl radicals (from aminyl radicals reacting with DO) were part of the NDMA formation pathway. The hypothesized source of the radical was an electron transfer reaction of an amine with NH₂Cl but was not demonstrated experimentally.

A long-standing unknown in the UF model that remained in the UF + SM model is the identity of the reactive intermediate (I) from NHCl₂ hydrolysis¹⁶ (Table S1, U7). I has been hypothesized to be the reactive nitrogen species (RNS) nitroxyl (HNO),²² but this has not been experimentally proven. HNO and its conjugate base, nitroxyl anion (NO⁻), exist in a slow equilibrium because of the spin-forbidden oxygen transition from singlet state HNO to triplet state NO⁻.²³ HNO/NO⁻ are difficult to measure kinetically because they are short lived and unlikely to accumulate. However, HNO/NO⁻ may react with HNO to form N₂O,²⁴ a stable end product, via R1 and R2 → R3 (see Scheme 1). Real-time N₂O measurements can be made nondestructively using a N₂O microelectrode;²⁵ therefore, N₂O kinetic measurements would serve as a total nitroxyl (HNO plus NO⁻) marker. In competition with N₂O formation, DO may react in a pH-dependent manner with HNO or NO^−26,27 via R5 and R2 → R4 (see Scheme 1) to form peroxynitrous acid/peroxynitrite anion (ONOOH/ONOO⁻), which are also RNS. A microelectrode can be used to measure DO in real time,¹¹ and DO consumption would support HNO/NO⁻ and ONOOH/ONOO⁻ formation.

Once formed, ONOOH/ONOO⁻ are unstable and known to decompose through a complex series of 117 reactions¹⁷ to nitrite (NO₂⁻) and nitrate (NO₃⁻), which would both further serve as markers for ONOOH/ONOO⁻ formation. With DMA present and competing with NO₂⁻ and NO₃⁻ formation, ONOOH/ONOO⁻ may directly react with DMA or decompose to radical species that may react with DMA, forming NDMA and another suspected carcinogen, N-nitrodimethylamine (DMNO).^28,29 Uric acid is a known scavenger of ONOOH/ONOO⁻ and its decomposition products,³⁰ including the free radicals NO₂^● and CO₃^●− that form by rapid decay of ONOOCO₂⁻ and O₂NOCO₂⁻, which are short-lived intermediates generated by the reaction of ONOOH/ONOO⁻ with CO₂.³¹ While uric acid is non-selective and can also scavenge reactive oxygen species such as hydroxyl radical and superoxide radical, Schreiber and Mitch¹⁰ ruled out these species in the NDMA formation pathway through scavenging experiments with superoxide dismutase and tert-butanol (see Supporting Information S0.2 for details). Experiments with uric acid would allow for the evaluation of ONOOH/ONOO⁻ (i) presence from NHCl₂ decomposition and (ii) importance in NDMA formation. Taken together, identifying I and understanding its fate during NHCl₂ decomposition may lead to an updated pathway for RNS and NDMA formation during NHCl₂ decomposition and/or a revised interpretation of the UF + SM model.

Based on the previous discussion, the objective of the current study was to evaluate revisions to the chemistry of NHCl₂ decomposition and associated NDMA formation with DMA present. To accomplish this, Pathway A1 was first evaluated through kinetic model simulations to assess the accuracy of NDMA formation kinetics in the UF + SM model at pH 7–10. Next, experiments (pH 7–10) were completed with and without 10 μM total DMA (TOTDMA, DMA plus dimethylammonium cation, DMAH⁺) and 800 μeq Cl₂·L⁻¹ NHCl₂, measuring NH₂Cl, NHCl₂, N₂O, DO, and NDMA kinetically and NO₂⁻, NO₃⁻, and DMNO at 4 h. NH₂Cl, NHCl₂, and N₂O data allowed the evaluation of the hypothesis that I is HNO during NHCl₂ decomposition, and DO, NO₂⁻, NO₃⁻, and NDMA data allowed the evaluation of ONOOH/ONOO⁻ formation and its potential relevance in NDMA formation. Furthermore, under drinking water conditions, DMNO most likely forms through a reaction of DMA with radicals that form during ONOOH/ONOO⁻ decomposition.^17,32 Therefore, DMNO formation would further support a pathway containing ONOOH/ONOO⁻. Additional experiments with uric acid, an ONOOH/ONOO⁻ scavenger,³⁰ allowed the evaluation of ONOOH/ONOO⁻ formation and its importance in NDMA formation. Concurrently with the kinetic experiments, a kinetic model was used to evaluate proposed revisions to NHCl₂ decomposition and NDMA formation chemistry, including the addition of HNO/NO⁻, DO, and ONOOH/ONOO⁻. Overall, the current study’s objectives were to advance fundamental chloramine chemistry by evaluating the identity of I during NHCl₂ decomposition, evaluate a pathway for RNS formation and propagation through DO consumption, and demonstrate the formation of toxicologically relevant end products such as NDMA through RNS-mediated pathways.

MATERIALS AND METHODS

The following are detailed in the Supporting Information (SI): (i) reagent preparation and washing procedures (Supporting Information S1.1); (ii) chloramine preparation (Supporting Information S1.2); and (iii) chloramine (Supporting Information S1.3), NO₂⁻ and NO₃⁻ (Supporting Information S1.4), and NDMA and DMNO (Supporting Information S1.5) quantification. This section’s remainder details the kinetic experiments, analytical techniques, and kinetic parameter estimation methodology.

Chloramine Formation/Quenching, TOTDMA Addition, and NDMA/DMNO Extraction.

NH₂Cl stock solutions were freshly prepared before each experiment³³ at ca. 4 mM to make NHCl₂ stock solutions as detailed in the Supporting Information (see S1.2). NHCl₂ stock solutions were made by decreasing the pH of the NH₂Cl stock solutions to 3.7–4.0 with 4 N H₂SO₄ and aging for 45 to 60 min until all the NH₂Cl was transformed to NHCl₂. Because NHCl₂ formation consumes protons, H₂SO₄ was added throughout the aging process to maintain the pH between 3.7 and 4.0.

NHCl₂ concentrations were quantified (Shimadzu UV 2450 spectrophotometer) following Schreiber and Mitch¹⁸ by deconvoluting the 245 and 295 nm absorbance spectra. Within 15 min of complete conversion to NHCl₂, the NHCl₂ solutions were diluted to 0.4 mM NHCl₂ (800 μeq Cl₂·L⁻¹) with pH-adjusted buffer to the desired pH, which was time zero. For the experiments in which NDMA was measured, this buffer also contained TOTDMA.

The starting experiment concentrations were 40 mM for the phosphate (pH 7 and 8), borate (pH 9), or carbonate (pH 10) buffers and 10 μM for TOTDMA (pH 7–10). The initial TOTDMA and NHCl₂ concentrations were twice those used by Schreiber and Mitch,¹⁰ which were done to exceed the detection limits for N₂O formation and DO consumption at pH 7 and prolong the temporally changing periods at pH 8 and 9.

Following the desired reaction times (0.25, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, and 4.0 h), waters were analyzed for NHCl₂; NH₂Cl; and, following the quenching of chloramine species, NDMA. DMNO was measured at 4 h only. For NDMA and DMNO, 10 mL aliquots were quenched for chloramines using 0.5 g of a dry quenching mix (1.8 g ascorbic acid, 1 g KH₂PO₄, and 39 g Na₂HPO₄) and shaken vigorously to ensure complete dissolution within a few seconds, minimizing localized concentration gradients.^10,12,34,35 This quenching mix acted as a pH buffer and salting out agent to improve the recoveries of NDMA and DMNO. Next, 1 mL of 100 μg·L⁻¹ d₆-NDMA was added and samples were immediately extracted with dichloromethane (10:1 water/dichloromethane volume ratio) using a back-and-forth shaker table at high speed for 15 min. Following a 5-min quiescent settling period, dichloromethane was decanted with a Pasteur pipette and stored for NDMA and DMNO analysis.

N₂O and DO Microelectrode Measurements.

N₂O and DO were measured using microelectrodes from Unisense that have manufacturer-reported response times of less than 45 and 10 s, respectively, and were calibrated before each experiment. Per the manufacturer’s recommendation, N₂O standards were made from dilutions of N₂O saturated solution and had a 1.0 μM-N limit of quantification that produced signal-to-noise ratios greater than 10 for all experiments. The DO sensor used a two-point calibration: (i) atmospheric air saturated water and (ii) zero DO condition, achieved by scavenging DO with 0.1 M sodium ascorbate in 0.1 M NaOH.

Kinetic Parameter Estimation Methodology.

To estimate parameters and their standard errors, the secant parameter estimation function in AQUASIM³⁶ was configured to minimize the weighted residual sum of squares (WRSS) between the measured and simulated results using the square of the measured value as the weight, resulting in a dimensionless WRSS.²⁵ This procedure was used to prevent greater concentrations from biasing the parameter estimates.³⁷

RESULTS AND DISCUSSION

Preliminary Kinetic Evaluation of the UF + SM Model (Pathway A1).

Pathway A1 was originally developed and validated from end-point (e.g., 2 h reaction) NDMA measurements but was not validated with detailed kinetic measurements (pH 7–10) of reactants (e.g., DO or NHCl₂), intermediates (e.g., UDMH-Cl), or products other than NDMA.¹⁰ If incomplete or not robust, viewing NDMA formation as Pathway A1 may not only limit the development of NDMA control strategies but fail to account for the production of RNS in chloramine systems and formation of RNS-mediated end products of human health concern.²⁸ The Supporting Information (see S2.1 and Figures S1 and S2) contains a detailed discussion of Pathway A1 kinetics that revealed two results subsequently questioned by the current experimental work: (1) DO consumption during NHCl₂ decomposition was nominal and only occurs in the presence of TOTDMA, and (2) at pH 9 and 10, NDMA formation continues after complete NHCl₂ decomposition.

Applicability of the UF Model for Chloramine Profiles.

Under the conditions of the current experiments (decomposition of ca. 800 μeq Cl₂·L⁻¹ NHCl₂ at pH 7–10 with and without 10 μM TOTDMA addition), the UF model reactions control chloramine concentrations in Scheme 1. As detailed in the Supporting Information (S1.3 and S2.2), the UF model accurately simulated time-course profiles of total chlorine (Figure S3), and NH₂Cl and NHCl₂ (Figures S4 and S5c,d). The slight mismatch of NH₂Cl at pH 10 (Figure S4d) was not surprising because the original UF model^38,39 was validated at pH 6–9. These results illustrated the applicability of the UF model to simulate chloramine concentrations under the selected conditions in this study.

Limitations of NDMA Simulations with the UF + SM Model (Pathway A1).

NDMA formation was measured kinetically in waters containing 10 μM TOTDMA and dosed with ca. 800 μeq Cl₂·L⁻¹ NHCl₂ at pH 7–10. Figure S6a (see Supporting Information S2.3) shows that the measured NDMA profiles at pH 7 and 8 were adequately simulated by the UF + SM model over the 4 h experiment. However, deficiencies in the UF + SM model were apparent at pH 9 and 10 (Figure S6b). At pH 9, the UF + SM model underpredicted the measured NDMA formation throughout the 4 h test by a factor of about 2.5. At pH 10, the UF + SM model underpredicted NDMA formation during the first hour, an indication that the underlying reaction kinetics were not robust. The measured NDMA profile was fully developed by the first sampling point at 0.25 h, whereas the UF + SM model simulated the increasing NDMA formation through ca. 2.0 h. These deficiencies along with the collected experimental data motivated the assessment of two additional pathways in Scheme 1 that are subsequently discussed: (1) Pathway B1 and (2) Pathway A2 as a replacement for Pathway A1.

Proposed Pathways B1 and A2.

Pathway B1 (see Scheme 1) combines (i) the UF model (Table S1), assuming HNO as I in U7-U9, and (ii) reactions involving DO, HNO/NO⁻, and other relevant RNS (Table S4, Supporting Information S2.4). RNS formation is initiated by NHCl₂ hydrolysis to HNO (U7) and NO⁻ (U7 → R2). HNO/NO⁻ may then react with (i) NH₂Cl or NHCl₂ to form nitrogen gas (N₂) via U8–U10, which is expected to be the major end product in the UF model;⁴⁰ (ii) HNO to form N₂O via R1 and/or R2 → R3 as a minor end product relative to N₂; or (iii) DO to form ONOOH/ONOO⁻ via R4 and/or R2 → R5, which subsequently may react with DMA to form NDMA via R7. Otherwise, ONOOH/ONOO⁻ decomposition results in NO₂⁻ and NO₃⁻ formation¹⁷ as other minor end products relative to N₂. Pathway A2 revises Pathway A1, which included the spin-forbidden incorporation of DO to form NDMA (P8). In the proposed Pathway A2, NDMA is formed through a reaction with ONOOH (R8, see Scheme 1), which is pH-dependent because the pK_a of 6.8 for ONOOH/ONOO⁻ (see Table S3) is relevant between pH 7 and 10; the pK_a for UDMH-Cl was estimated as 1.5,⁴¹ and thus, its concentration is not affected by the acid–base speciation between pH 7 and 10. Experimental data and the associated kinetic modeling were subsequently used to evaluate the proposed Pathways B1 and A2.

N₂O Formation from NHCl₂ Decomposition.

To initiate Pathway B1, NHCl₂ hydrolysis (Scheme 1, U7) is hypothesized to form HNO. Subsequently, N₂O may form from HNO (R1) or NO⁻ (R2 → R3).²⁴ Because N₂O is a stable end product, its detection serves as a HNO/NO⁻ formation marker and would support Pathway B1.⁴² Experiments were conducted with ca. 800 μeq Cl₂·L⁻¹ NHCl₂ and 10 μM TOTDMA at pH 7–10. Figure 1 shows that N₂O formed at pH 7–10 with pH-dependent kinetics and yields, and Figure S7 (see Supporting Information S2.5) shows that these N₂O profiles inversely tracked NHCl₂ decomposition. The N₂O profiles, therefore, qualitatively support HNO/NO⁻ formation from NHCl₂ hydrolysis (Table S1, U7), which is required to initiate Pathway B1 (see Scheme 1).

Figure 1.

NDMA (black, primary y axis), N₂O (magenta, primary y axis), and DO (green, secondary y axis) profiles in waters dosed with ca. 800 μeq Cl₂·L⁻¹ NHCl₂ and containing 10 μM TOTDMA buffered at pH (a) 7, (b) 8, (c) 9, and (d) 10. Points are measured values, lines are UF + RNS model simulations, and shaded areas are simulations encompassing 1 standard error in the estimated parameters (see Table 1). Table S9 contains the weighted residual sum of squares (WRSS) and the corresponding average weighted residual sum of squares (AWRSS) for data sets used in parameter estimation for the UF + RNS model (e.g., NDMA and DO). In addition, Table S10 contains the WRSS and AWRSS for the N₂O data sets to provide an indication of how well the UF + RNS model simulated these data, which were not used during parameter estimation. Notes: N₂O data were divided by 10 for scaling purposes (e.g., at pH 10, N₂O stabilized at about 13 μM-N), and the limit of quantitation was 1 μM-N, which corresponded to a signal-to-noise ratio greater than 10; at pH 10, the gap in DO and N₂O microelectrode data between 2.6 and 3.2 h was due to a lost computer connection.

DO Consumption from NHCl₂ Decomposition.

Once HNO/NO⁻ forms, DO consumption is required to continue Pathway B1, forming ONOOH (Scheme 1, R5) or ONOO⁻ (Scheme 1, R4). Therefore, DO should be consumed regardless of DMA presence and serves as an additional ONOOH/ONOO⁻ (and HNO/NO⁻) marker. Experiments conducted with ca. 800 μeq Cl₂·L⁻¹ NHCl₂ confirmed that DO consumption occurred at pH 7–10 in the absence of DMA (Figure S8, Supporting Information S2.6). Furthermore, DO profiles tracked with NHCl₂ decomposition, indicating that DO consumption was linked to NHCl₂ decomposition and supported DO consumption in Pathway B1.

Peroxynitrite Scavenging Decreased NDMA Formation.

To further support Pathway B1 by (i) providing evidence for ONOOH/ONOO⁻ formation during NHCl₂ decomposition and (ii) showing that NDMA formation proceeded through ONOOH/ONOO⁻, ONOOH/ONOO⁻ scavenging experiments were conducted with uric acid. To establish the maximum uric acid concentration that could be used, initial control experiments were conducted with uric acid and NHCl₂. Figure S9 (see Supporting Information S2.7) shows the impact of uric acid during NHCl₂ decomposition at pH 8–10. Profiles at pH 7 were not provided because of the aforementioned interference with the indophenol method (see Supporting Information S1.3). At pH 9 (Figure S9a,b and Supporting Information S2.7), a detailed uric acid control experiment was conducted to establish the upper uric acid dose, and uric acid doses up to and including 160 μM had minimal impact on NHCl₂ decomposition.

However, a uric acid dose of 200 μM resulted in greater NHCl₂ concentrations that deviated from other data starting at about 0.4 h (Figure S9b). Therefore, a target uric acid dose of about 160 μM was set as the upper limit and was checked at pH 8 and 10. Uric acid doses of 160 and 170 μM did not impact NHCl₂ decomposition at pH 8 (Figure S9c) and 10 (Figure S9d), respectively. Therefore, Figure 2 shows NDMA yields at uric acid doses of 160 or 170 μM and less.

Figure 2.

NDMA formation at 4 h versus uric acid dose in waters amended with 10 μM TOTDMA and ca. 800 μeq Cl₂·L⁻¹ NHCl₂ at pH 7, 8, 9, and 10. Uric acid is a known ONOOH/ONOO⁻ scavenger.³⁰

Figure 2 shows that uric acid, a ONOOH/ONOO⁻ scavenger,³⁰ decreased NDMA formation at pH 7–10. At pH 7, NDMA decreased from about 0.3 μM to below the limit of quantitation (0.05 μM) at uric doses of 140 and 170 μM. This supported that (i) ONOOH/ONOO⁻ is associated with NHCl₂ decomposition and (ii) the vast majority, if not all, of the NDMA formation occurred through a ONOOH/ONOO⁻ mediated pathway at pH 7. At pH 9, where NDMA formation was maximal, NDMA decreased from about 3.5 to 1.7 μM as the uric acid dose was increased from 0 to 160 μM. At pH 10, a decrease in NDMA formation was observed only after the uric dose was increased from 140 to 170 μM. On balance, the results in Figure 2 implicated ONOOH/ONOO⁻ as (i) a downstream product of NHCl₂ decomposition and (ii) the central node in NDMA formation during NHCl₂ decomposition, supporting Pathway B1 and contradicting Pathway A1. The pH 7 results (Figure 2) support Pathways A2 and B1 over A1 in which all NDMA formation proceeds through ONOOH/ONOO⁻. This contention is further supported in the remaining subsections.

NDMA and DMNO Yields from NHCl₂ Decomposition.

Further support for ONOOH/ONOO⁻ formation is NDMA formation concomitantly with DMNO. Experiments conducted with ca. 800 μeq Cl₂·L⁻¹ NHCl₂ and 10 μM TOTDMA at pH 7–10 showed NDMA and DMNO formation, with DMNO yields of 0.8–1.3% of NDMA regardless of pH (Table S5, Supporting Information S2.8). Masuda et al.²⁸ and Uppu et al.³² showed that reactions between ONOOH/ONOO⁻ and/or their decomposition products with a secondary amine produced N-nitrosamines and N-nitramines through nitrosation and nitration pathways, respectively. The presence of carbonate has been shown to alter the prevalence of nitrosation and nitration pathways. A study using the precursor morpholine showed that low levels of carbonate relative to ONOOH/ONOO⁻ could catalyze morpholine nitrosation, but high carbonate levels inhibited nitrosation in favor of N-nitramine formation.³² Because carbonate buffer was used only at pH 10 in this work, experiments were conducted at pH 10 with varying amounts (10–42 mM) of total carbonate (TOTCO₃). Results in Table S5 showed that NDMA yields decreased by about a factor of 2 with increasing TOTCO₃. DMNO yields also decreased with increasing TOTCO₃, but the percentage of DMNO relative to NDMA remained constant (1.1%). The consistent ratios between DMNO and NDMA at pH 7–10 (0.8–1.3%) further support the revision and replacement of Pathway A1 with A2 in which all NDMA formation occurs through ONOOH/ONOO⁻. Further work is needed to elucidate the role of carbonate species and buffer type relative to the fate of ONOOH/ONOO⁻ decomposition products during NHCl₂ decomposition. However, the decreased NDMA and DMNO yields with increased TOTCO₃ suggest a common source, and DMNO formation itself directly implicates ONOOH/ONOO⁻ formation in Pathway B1.

NO₂⁻ and NO₃⁻ Formation from NHCl₂ Decomposition.

NO₂⁻ and NO₃⁻ form rapidly from ONOOH/ONOO⁻ decomposition¹⁷ via the reaction of radical species and hydrolysis, respectively. Therefore, NO₂⁻ and NO₃⁻ are markers for ONOOH/ONOO⁻ formation. As noted in the Supporting Information S1.4 and S2.9, it was only possible to accurately quantify NO₂⁻ and NO₃⁻ after NHCl₂ had decomposed to less than about 50 μeq Cl₂·L⁻¹ because of a positive interference produced on NO₂⁻ by the quenching of NHCl₂ with thiosulfate. Figure S10 shows that quenching 50 μeq Cl₂·L⁻¹ NHCl₂ produced about 1 μM-N NO₂⁻. Therefore, NO₂⁻ and NO₃⁻ are reported in Table S6 (Supporting Information S2.9) as yields only at pH 8–10 in the absence and presence of 10 μM TOTDMA. Without TOTDMA, NO₂⁻ production was greater at pH 10 (26 μM) compared to pH 8 and 9 (both 10 μM) and NO₃⁻ production was similar at all three pH levels (14–16 μM); therefore, the percentage of NO₂⁻ relative to NO₂⁻ plus NO₃⁻ increased with pH, in agreement with the ONOOH/ONOO⁻ reaction scheme validated by Kirsch et al.,¹⁷ albeit at 37 °C (e.g., internal temperature of the human body). The addition of 10 μM TOTDMA resulted in decreased sums of NO₂⁻ plus NO₃⁻ (Table S6) by 8 μM at pH 8, 14 μM at pH 9, and 6 μM at pH 10. These decreases are consistent with the ONOOH/ONOO⁻ conversion to other intermediates and minor end products, including NDMA (see Scheme 1) that is maximal at pH 9 (see Figure 1). The DO formation from the production of NO₂⁻ is detailed in the Supporting Information (S2.10, see also Table S7).

UF + RNS Model Implementation.

To assess the RNS-mediated NDMA formation pathways (see Scheme 1, B1 and A2) and to help interpret the experimental data, a kinetic model (UF + RNS) was implemented in AQUASIM.³⁶ The UF + RNS model included the UF model¹⁶ (Table S1, U1–U14), assuming that HNO was I formed by NHCl₂ hydrolysis (U7); the two previously described RNS-related pathways, B1 and A2; and the Kirsch et al.¹⁷ model for ONOOH/ONOO⁻ decomposition to NO₂⁻ and NO₃⁻. Table S8 (see S2.11) shows that the current AQUASIM implementation of the Kirsch et al.¹⁷ model was within 0.1–3.5% of their digitized NO₂⁻ and NO₃⁻ concentrations between pH 7 and 10. Pathway A2 was included in lieu of Pathway A1 from the UF + SM model. Subsequently, the implemented UF + RNS model was used along with the experimental data (see Supporting Information S2.11, Table S9) to estimate revisions to three existing (k_u7, k_u8, and k_p5) and three new (k_r7A, k_r7B, and k_r8) parameters implemented in the UF + RNS model (Table 1).

Table 1.

Revised Reactions and Rate Constants Implemented in the UF + RNS Model^j

#	reaction stoichiometrya	rate expressiona	rate constant (M−1·s−1) unless otherwise noted
			published	this work (UF + RNS model) ± SE
U7	${\text{NHCl}}_2+{\text{H}}_2\text{O}\stackrel{k_{u}7}{\to }\text{HNO}+2{\text{H}}^{+}+2{\text{Cl}}^{-}$	ku7[NHCl2][OH−]	b110	186 ± 6
U8	$\text{HNO}+{\text{NHCl}}_2\stackrel{k_{u8}}{\to }\text{HOCl}+\text{products}$c	ku8[HNO][NHCl2]	d2.7 × 104	(8.2 ± 0.8) × 104
P5	${\text{NHCl}}_2+{\left({\text{CH}}_3\right)}_2\text{NH}\stackrel{k_{P5}}{\to }{\left({\text{CH}}_3\right)}_2\text{NNHCl}+{\text{H}}^{+}+{\text{Cl}}^{-}$	kp5[NHCl2][(CH3)2NH]	e52	28 ± 8
R7	$\text{ONOOH}+{\left({\text{CH}}_3\right)}_2\text{NNHCl}\stackrel{k_{r8}}{\to }{\left({\text{CH}}_3\right)}_2\text{NNO}+\text{products}$f	kr7[ONOOH][(CH3)2NH]	NA	g$k_{\text{r}7}=k_{\text{r}7\text{A}}\left[\exp \left(-\frac{k_{\text{r}7\text{B}}}{\left[{\text{H}}^{+}\right]}\right)\right]$ kr7A = (2.1 ± 0.4) × 107 M−1 ·s−1 kr7B = (4.4 ± 0.3) × 10−10 M
R8	$\text{ONOOH}+{\left({\text{CH}}_3\right)}_2\text{NH}\stackrel{k_{r}7}{\to }{\left({\text{CH}}_3\right)}_2\text{NNO}+\text{products}$h	kr8[ONOOH][(CH3)2NNHCl]	NA	i(1.3 ± 0.8) × 107

^aUnidentified intermediate, I, of NHCl₂ hydrolysis was assumed to be HNO in the UF + RNS model.

^bJafvert and Valentine.¹⁶

^cMay include N₂, Cl⁻, H⁺, and other unidentified reaction products.

^dJafvert and Valentine.³⁹

^eSchreiber and Mitch.¹⁰

^fMay include H₂O₂⁴³ and other unidentified reaction products.

^gEmpirical formulation currently only applicable at pH 7–10, indicating that ONOOH/ONOO⁻ decomposition products may also react with (CH₃)₂NH to form NDMA.

^hMay include H⁺, Cl⁻, NO₂⁻, and other unidentified reaction products.

ⁱONOOH/ONOO⁻ decomposition products may also react with (CH₃)₂NNHCl to form NDMA. NA: not applicable because these reactions are reported for the first time in this work.

^jEstimated rate constants provided with their standard error (SE, N = 528 total data points).

^#: reaction number corresponding to Scheme 1; Tables S1–S4 detail published reactions and rate constants

Rate constants for U7–U10 were empirical in the UF model,¹⁶ formulated to match chloramine species concentrations only; therefore, they were initially all considered for re-estimation in the current work. Preliminary analysis indicated that k_u9 and k_u10 were not sensitive to the conditions of this study (i.e., preformed NHCl₂); therefore, k_u9 and k_u10 were excluded from the final parameter estimation procedure. Table 1 shows the revised rate constants ± 1 standard error for k_u7 (from 110 to 186 ± 6 M⁻¹·s⁻¹) and k_u8 (from 2.7 × 10⁴ to (8.2 ± 0.8) × 10⁴ M⁻¹·s⁻¹).

R7 was included in the UF + RNS model as a new NDMA-formation reaction between ONOOH and DMA (Scheme 1, Pathway B1), justified by (i) the scavenger results (Figure 2) that indicated that ONOOH/ONOO⁻ was a central node in NDMA formation and (ii) the RNS literature that showed that ONOOH/ONOO⁻ and/or their decomposition products (i.e., radicals) reacted with DMA to form NDMA.²⁸ Figure 1 indicated that NDMA formation was maximal at pH 9, which is close to the average of the pK_a values listed in Table S3 for ONOOH/ONOO⁻ (pK_a = 6.8) and DMAH⁺/DMA (pK_a = 10.73). Generally, the reaction rate between an acid of one species and conjugate base of another is maximal at a pH equal to the average of their pK_a values, which is pH 8.8 in this case. The UF + SM model used DMA (i.e., the uncharged base form) in the NDMA formation pathway.¹⁰ Therefore, if a direct reaction occurred involving DMA (base form), ONOOH (acid form) is the likely candidate. The estimated rate constant for R7 (see Table 1 and Table S11) was determined to be acid catalyzed, $k_{\text{r}7}=k_{\text{r}7\text{A}}\left[\exp \left(-\frac{k_{\text{r}7\text{B}}}{\left[{\text{H}}^{+}\right]}\right)\right]$ with k_r7A = (2.1 ± 0.4) × 10⁷ M⁻¹·s⁻¹ and k_r7B = (4.4 ± 0.3) × 10⁻¹⁰ M, an empirical formulation that warrants a future research effort to further develop a mechanistic revision of the UF + RNS model.

Although our NDMA data could also be simulated by a UF + RNS model version using only Pathways A1 (e.g., the UF + SM model) and B1, our experimental data did not support Pathway A1 because NDMA formation was shut down at pH 7 in the ONOOH/ONOO⁻ scavenging experiments (Figure 2). Therefore, Pathway A1 was revised and Reaction P8 (UDMH-Cl + DO) was the likely candidate for this revision because it is spin-forbidden and DO consumption was not previously kinetically validated.¹⁰ This prompted the assessment of Pathway A2 in which P8 was replaced by R8, the reaction of UDMH-Cl and ONOOH to form NDMA with an estimated rate constant of (1.3 ± 0.8) × 10⁷ M⁻¹·s⁻¹ (Table 1, k_r8). The current implementation of R8 is also empirical because it is not known if it is a direct reaction with ONOOH/ONOO⁻ and/or their decomposition products,³² but it was assumed that either could be simulated based on the ONOOH concentration. Rate constants for reactions P5 and P7 were also initially considered for revision, but P7 proved insensitive and was therefore not re-estimated. The final parameter estimation resulted in a decreased rate constant estimate for P5 from 52 to 28 ± 8 M⁻¹·s⁻¹, a logical result given that the UF + SM model did not consider Pathway B1.

UF + RNS Model Well Simulated NHCl₂, NH₂Cl, DO, and NDMA at pH 7–10.

The revised rate constants for U7 and U8 were validated by comparing simulated UF + RNS and UF model free chlorine, NH₂Cl, and NHCl₂ concentrations for 20 additional experimental conditions (Exp. 1–15 and 23–27) from Jafvert³⁸ that are shown in Figures S11–S14, S16d, and S17 (see Supporting Information S2.12 and Table S12 for WRSS and average WRSS (AWRSS) comparisons). In addition, the NH₂Cl and NHCl₂ data sets used in parameter estimation (Table S9) are also included in Table S12. Compared to the UF model, the simulated free chlorine, NH₂Cl, and NHCl₂ concentrations with the UF + RNS model had similar or lower summed totals of WRSS and AWRSS (see Table S12), indicating that the UF + RNS model simulated these data comparably to the UF model. These results showed that the revised estimates to the rate constants for U7 and U8 in the UF + RNS model did not compromise the simulations of chloramine species concentrations over a wide range of conditions.

Figure 1 shows the UF + RNS model simulations and measured time-course profiles of DO and NDMA at pH 7–10. The kinetics and yields of these measured data were well simulated by the UF + RNS model at all pH levels tested. Notably, at pH 9 and 10 (Figure 1c,d), the simulated DO and NDMA profiles matched the measured data in terms of kinetics and yields throughout the 4 h time-course.

The UF + RNS model was used to determine the relative importance of the two NDMA formation pathways (i.e., Pathways A2 and B1). Figure 3 summarizes the percentage of the simulated NDMA that formed via R7 (DMA + ONOOH, Pathway B1) and R8 (UDMH-Cl + ONOOH, Pathway A2) at 0.25 h (Figure 3a) and 4 h (Figure 3b) along with the UF + RNS model simulated and measured NDMA concentrations.

Figure 3.

Percentage of UF + RNS model simulated NDMA formed via R7 and R8 (primary y axis) and measured and UF + RNS model simulated NDMA concentrations (secondary y axis) at pH 7, 8, 9, and 10 at (a) 0.25 h and (b) 4 h.

At pH 7, R8 was the predominate NDMA formation pathway (70–90%), although the measured NDMA was lower at pH 7 compared to the other pH levels. At pH 8, R8 accounted for 55–59% of the NDMA formation, illustrating that Pathways A2 and B1 were both important contributors. At pH 9, where NDMA formation was maximal, R7 was the predominant NDMA formation pathway at 0.25 and 4 h (81–83%). At pH 10, R7 accounted for 52% of the NDMA formation at 0.25 and 4 h, illustrating the importance of the A2 and B1 pathways. The results in Figures 1 and 3 support the UF + RNS model because it captured the kinetics and yields of the measured DO and NDMA profiles. Pathway B1 (see Scheme 1) was important at pH 7–10 and, together with Pathway A2, indicated that all the measured NDMA could be accounted for through ONOOH/ONOO⁻, stemming from NHCl₂ decomposition.

Current UF + RNS Model Limitations and Future Research Needs.

The UF + RNS model Pathways B1 and A2 each currently use an empirical NDMA formation reaction, R7 and R8, respectively. Because of the complexity of the current reaction scheme, the mechanistic determination of R7 and R8 is beyond the scope of the current research and is an avenue of future research. Potential cross-interactions remain to be investigated between the chloramine species and known decomposition products of ONOOH/ONOO⁻, including hydrogen peroxide and hydroxyl radical.^44,45 In addition, deviations between the measured NO₂⁻ and NO₃⁻ formation and UF + RNS model simulated values (see Table S13 in Supporting Information S2.13) indicate the need for kinetic N₂, NO₂⁻, and NO₃⁻ data to validate and/or revise the Kirsch et al.¹⁷ ONOOH/ONOO⁻ decomposition model for drinking water conditions and temperatures. Finally, the UF + RNS model overpredicted the N₂O data at pH 10 (Figure 1d) but underpredicted N₂O at pH 7 to 9 (Figure 1a–c), with the corresponding WRSS and AWRSS shown in Table S10, illustrating the need to measure N₂, NO₂⁻, and NO₃⁻ kinetically in future work to facilitate further advancement of the UF + RNS model.

Mechanistic Considerations.

This study presented multiple lines of evidence to demonstrate that NHCl₂ hydrolysis resulted in HNO formation (Scheme 1, U7), the so-called unidentified reactive intermediate (I) in the UF model.¹⁶ Per Pathway B1, HNO/NO⁻ reacts through three competing pathways, which include two minor pathways: (i) N₂O formation (see Figure 1) via R1 and/or R2 → R3 and (ii) reaction with DO (see Figure 1) to form ONOOH/ONOO⁻ via R5 and/or R2 → R4, and one major pathway: (iii) N₂ formation via U8–U10 that has been previously quantified during NH₂Cl decomposition⁴⁰ and can now be used to investigate nitrogen mass balances during NHCl₂ decomposition along with the advancements made in this work.

The two minor pathways are important new additions to NHCl₂ decomposition chemistry because they help explain the formation of NO₂⁻, NO₃⁻, NDMA, and DMNO. Table S13 shows that the measured values for N₂O, NO₂⁻, and NO₃⁻ account for ca. 4–6% of the nitrogen originally present in NHCl₂ and total free ammonia (NH₃ + NH₄⁺). Although complete nitrogen mass balances have not been tracked during NHCl₂ decomposition, Saunier⁴⁶ estimated that NO₃⁻ yields accounted for 8–11% of the total nitrogen. Future work should include N₂, NO₂⁻, NO₃⁻, and total free ammonia kinetic measurements along with the products kinetically measured in this study.

Measurements of NHCl₂, NH₂Cl, N₂O, DO, NO₂⁻, NO₃⁻, NDMA, and DMNO implicate ONOOH/ONOO⁻ as a proposed unstable intermediate formed during NHCl₂ decomposition, which was supported by experiments with uric acid, a ONOOH/ONOO⁻ scavenger.³⁰ NDMA decreased with increasing uric acid at pH 7–10 (Figure 2), and the shutdown of NDMA formation at pH 7 spurred the replacement of P8 in the UF + SM model with R8 in the UF + RNS model (Table 1), corresponding to the replacement of Pathway A1 with A2 in Scheme 1. Together with R7, the UF + RNS model accurately captured the kinetics and yields of DO and NDMA at pH 7–10 (Figure 1), indicating that all NDMA formation (Figure 3) could be accounted for through ONOOH/ONOO⁻ formation, stemming from NHCl₂ decomposition.

Implications.

A logical extension of the current results is that scavenging HNO/NO⁻ and/or ONOOH/ONOO⁻ may curb NDMA formation in chloramine systems. NHCl₂ does not accumulate to measurable concentrations at pH 9 and greater because its rate of decomposition is greater than its formation,¹⁶ but if total free ammonia is in excess as is the case in chloramine systems, chloramine decomposition occurs through NHCl₂; therefore, HNO/NO⁻ and ONOOH/ONOO⁻ will form regardless of the pH. Scavenging of these RNS and/or strategies to promote chloramine stability would presumably lead to less NDMA formation.

Future research should extend the UF + RNS model presented here. Specifically, a further mechanistic revision of the UF + RNS model is needed, focusing on U7–U10,¹⁶ R7, R8, and ONOOH/ONOO⁻ decomposition chemistry¹⁷ to well simulate the measured profiles of the minor products (N₂O, NO₂⁻, NO₃⁻) and the major products (N₂ and total free ammonia) while maintaining simulations of chloramine species (Figures S4, S7, S15, and S16), DO (Figure 1), and NDMA (Figure 1).

Further experimental work should consider additional HNO/NO⁻ formation pathways independent of NHCl₂, including the reaction between NH₂Cl and hydroxylamine (NH₂OH) generated from ammonia-oxidizing bacteria.^25,47 The reaction of NH₂Cl with NH₂OH implicated HNO/NO⁻ and ONOOH/ONOO⁻ production,²⁵ providing a potential mechanism to evaluate and potentially explain the enhanced NDMA formation observed during nitrification episodes.⁴⁸