Article pubs.acs.org/est
Mechanistic Insight into the Reactivity of Chlorine-Derived Radicals in the Aqueous-Phase UV−Chlorine Advanced Oxidation Process: Quantum Mechanical Calculations Daisuke Minakata,* Divya Kamath, and Shaye Maetzold Department of Civil and Environmental Engineering, Michigan Technological University, 1400 Townsend Drive, Houghton, Michigan 49931, United States S Supporting Information *
ABSTRACT: The combined ultraviolet (UV) and free chlorine (UV−chlorine) advanced oxidation process that produces highly reactive hydroxyl radicals (HO•) and chlorine radicals (Cl•) is an attractive alternative to UV alone or chlorination for disinfection because of the destruction of a wide variety of organic compounds. However, concerns about the potential formation of chlorinated transformation products require an understanding of the radical-induced elementary reaction mechanisms and their reaction-rate constants. While many studies have revealed the reactivity of oxygenated radicals, the reaction mechanisms of chlorine-derived radicals have not been elucidated due to the data scarcity and discrepancies among experimental observations. We found a linear free-energy relationship quantum mechanically calculated free energies of reaction and the literature-reported experimentally measured reaction rate constants, kexp, for 22 chlorine-derived inorganic radical reactions in the UV−chlorine process. This relationship highlights the discrepancy among literature-reported rate constants and aids in the determination of the rate constant using quantum mechanical calculations. We also found linear correlations between the theoretically calculated free energies of activation and kexp for 31 reactions of Cl• with organic compounds. The correlation suggests that H-abstraction and Cl-adduct formation are the major reaction mechanisms. This is the first comprehensive study on chlorine-derived radical reactions, and it provides mechanistic insight into the reaction mechanisms for the development of an elementary reaction-based kinetic model.
■
HOCl/OCl− further react with Cl• to produce dichlorine radicals (Cl 2•−, 2.13 V) that may react with organic compound(s). The reactions of HO• and Cl• with HOCl/ OCl− produce oxychlorine radicals (ClO•, 1.39 V), whose reactivity with organic compounds is not well-understood.5,7,11 Consequently, the involvement of various chlorine species raises serious concerns about the potential formation of toxic degradation by-products, such as chlorinated by-products.12,13 Given that there are many organic compounds in commercial use and production,14 the development of a kinetic model that can predict the fate of degraded organic compounds is needed for the preliminary design and screening of a number of these compounds. This type of predictive kinetic model requires three components: (1) the identification of reaction pathways; (2) the prediction of reaction rate constants, and (3) the numerical solution of ordinary differential equations (ODEs) for each species. Once the ODEs are numerically solved, the time-consequent concentration profiles of each species can be predicted.
INTRODUCTION Ultraviolet radiation combined with free chlorine (UV− chlorine) that produces highly reactive hydroxyl radicals (HO•) and chlorine radicals (Cl•) at ambient temperature and atmospheric pressure is one of the attractive aqueous-phase advanced oxidation processes (AOPs).1−3 UV−chlorine can serve as an alternative AOP to UV with hydrogen peroxide because of the utilization of the existing chlorination process or as an alternative to the UV process alone for the disinfection of pathogens and the destruction of dissolved organic compounds by both HO• and Cl•.4 Previous studies indicated the different degree of contribution of initial reaction of Cl• with a target compound to the overall degradation.5−7 In addition, chlorine residue after the UV−chlorine AOP can be used as a secondary disinfectant, rendering this process suitable for the application of drinking water treatment as well as wastewater reclamation for direct potable reuse. In the aqueous-phase UV−chlorine AOP, photolysis of hypochlorous acid/hypochlorite ions (HOCl/OCl−) generates HO• and Cl• radicals,8,9 which react with target organic compound(s) as well as with HOCl−OCl−. These highly reactive HO• (2.73 V versus SHE)10 and Cl• (2.43 V) radicals transform organic compounds into various degradation byproducts via complicated radical-involved reactions. In addition, chloride ions (Cl−) generated by the reaction of Cl• with © 2017 American Chemical Society
Received: Revised: Accepted: Published: 6918
January 26, 2017 May 1, 2017 May 25, 2017 May 25, 2017 DOI: 10.1021/acs.est.7b00507 Environ. Sci. Technol. 2017, 51, 6918−6926
Article
Environmental Science & Technology
Table 1. Elementary Reactions in the UV−Chlorine AOP with kexp Values in the Literature and Our Calculated ΔGreact aq,calc, kchem, and kD Values no.
elementary reaction
1
HOCl + Cl• → •OCl + HCl HOCl + Cl• → Cl2 + HO− HOCl + Cl• → HOCl(Cl•) HOCl + HO• → •OCl + H2O HOCl + HO• → HO• + HOCl HOCl + HO• → HOCl(HO•) OCl− + Cl• → •OCl + Cl− OCl− + HO• → •OCl + HO− Cl− + Cl• → Cl2•− Cl− + Cl2 → Cl3− Cl2•− → Cl• + Cl− Cl• + Cl2•− → Cl2 + Cl− Cl2•− + Cl2•− → Cl2 + 2Cl− (or Cl3− + Cl−) Cl2•− + HO• → HOCl + Cl− Cl3− → Cl2 + Cl− HO• + Cl− → •OHCl− Cl• + H2O → •OHCl− + H+ Cl• + OH− → •OHCl− Cl2•− + OH− → •OHCl− + Cl− •OHCl− → HO• + Cl− •OHCl− + H+ → Cl• + H2O •OHCl− + Cl− → Cl2•− + OH− Cl• + H2O → •Cl(H2O) Cl2•− + H2O → •Cl(H2O) + Cl− •Cl(H2O) + Cl− → Cl2•− + H2O •OCl + •OCl + H2O → HClO + HClO2
2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 a
kexp, M−1 s−1
ref
3 × 10
42
8.46 × 104−2.0 × 109
3, 18, 19
9
ΔGactaq,calc, kcal/mol 14.7 3.2 4.6 11.8
ΔGreactaq,calc, kcal/mol −5.9 −132.6 4.6 −20.8
8.4
0
7.2
5.7
kchem, M−1 s−1
kD, M−1 s−1
4.2 × 10
1.1 × 1010
8.5 × 104−2.4 × 109
8.5 × 104−1.2 × 1010
1.1 × 1010 1.2 × 1010
9
8.2 × 109 8.8 × 109
42 43
−23.1 −5.2
3.3 × 1010 3.4 × 1010
6.5−8.5 × 109 2.0 × 104 5.2 × 104−1.4 × 105 (2.1 ± 0.05) × 109 (0.9−9.0) × 109
−13.1 0.22 13.1 −37.5 −24.4 (−24.2)
1.6 2.0 5.2 2.6 1.0
× × × × ×
1010 104 104−1.4 × 105 109 109
1.1 1.0 1.2 1.1 1.0
× × × × ×
1010 1010 1010 1010 1010
1 × 109 1.1 × 105 (4.3 ± 0.4) × 109 (1.6 ± 0.2) × 105 1.8 × 1010 4.5 × 107
42, 44 45 44, 46 47 46, 48, 49 50 45 50 46 42 51
−34.5 −0.22 2.99 (−2.4)a
1.1 1.1 6.8 1.6 1.8 4.5
× × × × × ×
109 105 109 105 1010 107
1.2 1.2 1.2 1.2 1.1 1.0
× × × × × ×
1010 1010 1010 1010 1010 1010
(6.1 ± 0.8) × 109 (2.6 ± 0.6) × 1010 1.0 × 105
46 46 51
−2.99
2.5 × 105 1.3 × 103
52 52
5 × 109
52
2.5−7.5 × 109
42, 43
−14.9 −1.8
1.0 × 105
1.0 × 1010 1.0 × 1010 1.0 × 1010
3.8 16.9
2.5 × 105 1.3 × 103
1.3 × 1010 1.2 × 1010
−16.9
9.9 × 109
1.0 × 1010
3.3−27.0 × 109
1.1 × 1010
1.84
−8.1
Value with two explicit water molecules.
Although HO•-induced initial degradation pathways of organic compounds have been studied over the past several decades,15,16 and predicting the degradation by-products for small-molecular-weight aliphatic compounds and alkenes has become feasible,17 the reaction mechanisms in the UV− chlorine AOP are not well-understood. For example, literaturereported experimentally obtained reaction rate constants (kexp) for the reaction of HO• with HOCl vary from 104 to 109 M−1 s−1.3,18,19 This rate constant significantly affects the calculations of the concentration profile of subsequent by-products because HOCl is dominantly present near a neutral pH (i.e., the pKa of HOCl is 7.5).20 Among approximately 30 kexp studies on Cl•,21 for saturated aliphatic compounds, some claim that a single electron transfer (SET) is the major Cl• reaction, but others find that hydrogen(H)-atom abstraction from a carbon− hydrogen (C−H) bond is the major reaction.22−25 These inconsistencies cause difficulties in defining the reaction products (e.g., carbon-centered radical versus alkoxyl radical) of each elementary reaction (e.g., carbon-centered radical from H abstraction versus radical cation or Cl adduct from SET). Furthermore, while HO• predominantly attacks the α position of organic compounds,26 it is unclear whether chlorine-derived radicals selectively react with specific sites of organic
compounds. This reactivity can affect the branching ratio of reaction pathways and the estimation of the by-products. Ab initio quantum mechanical (QM) calculations are robust techniques for identifing elementary reaction pathways, and they can be used to predict aqueous-phase reaction rate constants. Successful applications of these calculations include an identification of elementary reaction steps and determining rate constants for HO• reactions with phthalate27 and ibuprofen,28 as well as those for the formation of Nnitrosodimethylamine during ozonation.29 These studies, in general, calculated the aqueous-phase enthalpies and free energies of reaction to identify the thermodynamically preferable reaction pathways (e.g., ozone)30 and free energies of activation for reaction rate constants (e.g., HO•).31−33 While the use of ab initio QM methods can be applied for H abstraction by Cl•, the lack of applicable theory and experimental evidence limit the QM application to SET reactions. The Marcus theory34 was used to calculate the aqueous-phase free energies of reaction for SET reactions, but significantly higher values were reported compared with those for H abstraction and HO• addition due to the uncertainty of the reorganization of water molecules.35 6919
DOI: 10.1021/acs.est.7b00507 Environ. Sci. Technol. 2017, 51, 6918−6926
Article
Environmental Science & Technology In this study, we use ab initio QM calculations to identify the dominant reaction mechanisms for chlorine-derived radicals to ultimately develop an elementary reaction-based UV−chlorine kinetic model. Our hypothesis is that for chlorine-derived radical reactions, the theoretically calculated aqueous-phase free energies of reaction (i.e., ΔGreact aq,calc) are linearly correlated with kexp based on a linear free-energy relationship (LFER),35 and the theoretically calculated aqueous-phase free energies of activation (i.e., ΔGact aq,calc) are linearly correlated with kexp based on Eyring’s transition-state theory (TST)36 under the same reaction mechanism. Once these correlations are developed, we can predict the reaction rate constants for reactions that have not been investigated experimentally; these results will be used to solve ODEs for predicting the formation of by-products.
■
MATERIALS AND METHODS All of the ab initio molecular orbital and density functional theory (DFT)-based QM calculations were performed with the Gaussian 09 revision D.02 program37 using the Michigan Tech high-performance cluster “Superior”. The electronic structures of the molecules and radicals in the ground and transition states were optimized at the level of B3LYP/6-31G(2df,p) implemented in Gaussian-4 theory (G4)38 in both the gaseous and the aqueous phases. The aqueous-phase calculations were performed using an implicit polarizable continuum model (universal solvation model, SMD).39 Previously, we verified the combination of G4 with the SMD model by successful application to other aqueous-phase radical-involved reactions.40 The Supporting Information provides the detailed calculations of the transition-state search, the aqueous-phase free energies of activation and reaction, and the associated computational methods.
Figure 1. Linear free-energy relationship between ln kchem and ΔGreact aq,calc for chlorine-derived inorganic radical reactions. Diamond: reactions with HOCl/OCl−; circle: Cl• and Cl2•− involving reactions; square: • OHCl− involving reactions; and triangle: ClO• reaction. The vertical bar results from the range of the reported kexp values in the literature. Note that data points 6 and 11 are not included in the LFER analysis due to the different reaction mechanism.
within this error from the LFER. Indeed, many radical-involved reactions such as addition, SET, and H abstraction do not follow the thermodynamic law, and kinetics may overrun thermodynamics.30 However, our findings indicate that ΔGreact aq,calc can be used to predict the rate constants. The LFER can also be used to evaluate the series of kexp values systematically and provide critical feedback for future experimental studies. In the following section, we discuss each chlorine-derived inorganic radical reaction. Reactions with Hypochlorous Acid and Hypochlorite Ions. While Cl• reacts with both HOCl and OCl− rapidly (i.e., 3.0 × 109 M−1 s−1 in reaction 1 and 8.2 × 109 M−1 s−1 in reaction 3),42 there is disagreement among the kexp values for the reaction of HO• with HOCl (8.5 × 104,3 1.1−1.4 × 108,19 and 2.0 × 109 M−1 s−118 in reaction 2). First, we calculated ΔGact aq,calc for the H-abstraction reaction from an O−H bond of HOCl by Cl• and HO• as 14.7 and 11.8 kcal/mol, respectively. The addition of one to three explicit water molecules did not act significantly change the ΔGact aq,calc values. The calculated ΔGaq,calc 8 value is apparently too large for the observed kexp of 10 ∼109 M−1 s−1 for Cl• because 109 M−1 s−1 of kexp corresponds to ∼2 kcal/mol of ΔGact aq,calc. This result allows the estimation of the kexp value of 8.5 × 104 M−1 s−1 to correspond with the ΔGact aq,calc • value. Second, the ΔGact aq,calc values for Cl abstraction by Cl and HO• are 3.2 and 8.4 kcal/mol, respectively. However, while the gaseous-phase Cl abstraction with a product of Cl2 was reported,53 the experimentally observed product, ClO•, in the aqueous phase is different from Cl2, thus eliminating the possibility of Cl abstraction. Third, we obtained ΔGact aq,calc values of 4.6 and 7.2 kcal/mol for the formation of the Cl or HO adduct, respectively, between each radical and the oxygen atom of HOCl. However, our theoretically calculated ΔGreact aq,calc values are positive, and the adduct formation reactions are not thermodynamically favorable. Finally, the ΔGreact aq,calc value for a SET between HOCl and Cl• was calculated for the products of ClO• and HCl, and this energy is found to be negative (i.e., −5.2 kcal/mol). Consequently, Cl• reacts with HOCl via a
■
RESULTS AND DISCUSSION Reactivity of Chlorine-Derived Radicals with Inorganic Radicals. Table 1 summarizes each elementary reaction in the UV-chlorine AOP and the literature-reported kexp as well as our calculated ΔGreact aq,calc values at 25 °C and the chemical reaction rate constant (kchem) obtained from the relationship with the diffusion reaction rate constant (kD) from kexp (eq 1): kexp =
kD × kchem kD + kchem
(1)
ΔGreact aq,calc
We find a LFER between ln kchem and (Figure 1). With the decrease in the ΔGreact aq,calc values below −20 kcal/mol, the kchem values became constant at approximately (1.0−2.0) × 1010 M−1 s−1 because of the limitation of the diffusion contribution and the exothermic reactions caused by the highly hydrolyzed reaction products in the aqueous phase. Our calculated kD values using Smoluchowski’s equation41 (see the Supporting Information) are approximately 1.1 × 1010 M−1 s−1, and the upper kexp value is approximately 8.0 × 109 M−1 s−1; therefore, the kchem value can be estimated as 2.9 × 1010 M−1 s−1, which is close to the observed kchem values in Figure 1. In general, the G4 method provides 0.83 kcal/mol of average absolute deviation from the experimentally obtained gaseous-phase energy values.38 The SMD model at the various levels of DFT methods with a 6-31G* basis set provides mean unsigned errors of 0.6−1.0 kcal/mol in solvation free energies for neutral compounds and 4 kcal/mol on average for ions.39 Based on the estimated computational error, ± 5.0 kcal/mol of ΔGreact aq,calc, corresponding to ±2.8 of lnkchem, it is found that 64% of kexp is 6920
DOI: 10.1021/acs.est.7b00507 Environ. Sci. Technol. 2017, 51, 6918−6926
Article
Environmental Science & Technology
various solvated states of the •OHCl− species and to determine whether these reactions adhere to the LFER. The reaction of HO• with Cl− to produce a covalently bonded adduct, Cl−−•OH, was found to be endothermic with a ΔGreact aq,calc value of 2.99 kcal/mol, which is inconsistent with the LFER. We added explicit water molecule(s) to hydrolyze Cl• and found that while the ΔGreact aq,calc value with one water molecule remained positive (i.e., 3.7 kcal/mol), the ΔGreact aq,calc values became negative with two (i.e., −1.87 kcal/mol) and three explicit water molecules (i.e., −2.4 kcal/mol). The adduct with two or three explicit water molecules has a charge of −0.95 on Cl and 7% of the spin-density distribution on Cl with the hemibonded structure, which is consistent with the ESR observation.59,60 With this ΔGreact aq,calc value, reaction 12 adheres to the LFER. Our observations with explicit water molecules are consistent with previous theoretical studies56−58 and highlight the importance of including explicit water molecules in calculating the thermodynamic properties of Cl−−•OH. For reaction 19, we observed the formation of a covalently bonded adduct •Cl(H2O) in the absence or presence of explicit water molecules. Regardless of the addition of explicit water molecules, the ΔGreact aq,calc values were positive but still adhered to the LFER because of the observed smaller kexp (i.e., ∼105 M−1 s−1). We propose that reaction 19 undergoes hydration in which water molecules form new covalent bonds to •Cl. Reactivity of Chlorine-Derived Radicals with Organic Compounds. Chlorine Radicals. We calculated the ΔGact aq,calc and ΔGreact aq,calc values for each elementary reaction with 31 organic compounds, including 9 alcohols, 5 aldehydes and ketones, 6 carboxylic acids and carboxylates, 6 esters and ethers, 2 haloalkanes, 2 alkenes and 1 aromatic compound (Table S1). Overall, we find linear correlations for H abstraction from a C− H bond and Cl-adduct formation between ln kexp and ΔGact aq,calc, respectively (i.e., ln kexp = −0.50 ΔGact aq,calc + 20.53 for H abstraction from 23 organic compounds and ln kexp = −0.95 ΔGact aq,calc + 23.43 for Cl adduct formation from 14 alcohols and carboxylic compounds) (Figures 2 and 3). The linear correlation results from Eyring’s TST based on quantum and statistical mechanics using partition functions to quantize energy states of reactants and activated intermediates.36 Note that the conventional TST requires an accuracy of ±0.4 kcal/ mol of ΔGact aq,calc to predict kexp within a difference of a factor of 2 from the experimental observations, which is analogous to the experimental accuracy of kinetic measurements. Accordingly, the direct calculation of k based on ΔGact aq,calc values is not feasible because the estimated computational error resulting from the corresponding QM method exceeds the required accuracy. Based on the estimated computational error, ±5.0 kcal/mol of ΔGact aq,calc, corresponding to ±2.4 of ln kchem, it was found that 96% of kexp is within this error from the linear correlation for H abstraction. For Cl-adduct formation, based on the ±5.0 kcal/mol of ΔGact aq,calc, corresponding to ±4.8 of lnkchem, it was found that 100% of kexp is within the error from the linear correlation. Therefore, the linear correlations developed here are valid within the estimated computational error and can be used to predict k for compounds that have not been examined experimentally. Note that we did not observe the LFER based on our calculated ΔGreact aq,calc for the reactions with organic compounds, most likely because kinetics overrun thermodynamics for these reactions. In the following section, we discuss each Cl• reaction with organic compounds. Alcohols. A total of nine compounds containing alcohol functional groups were investigated in this study. A total of 99%
SET. In this case, the kexp value adheres to the LFER with • ΔGact aq,calc, as shown in Figure 1. In contrast, HO does not react with an oxygen or chlorine atom of HOCl via SET.15 The investigations above lead to the conclusion that H abstraction from the O−H bond of HOCl by HO• is more reasonable, as this mechanism is consistent with the observed product (i.e., ClO•). Furthermore, our estimation of kexp for HO• is supported by the Curtin−Hammett principle54,55 describing that dominant species (i.e., HOCl) in equilibrium with OCl− at pH below pKa has lower reactivity with an identical reactant (i.e., HO•) compared to that for minor species (OCl−). For the reaction with OCl−, the ΔGreact aq,calc values for SET are −23.1 kcal/mol for the reaction with Cl• producing •ClO and Cl− and −5.2 kcal/mol for the HO• reaction producing •OCl and HO−; therefore, SET for these reactions is thermodynamically favorable. This reaction mechanism is verified by the LFER and is consistent with the experimental observation.42,43 Reactions with Chlorine-Derived Radicals. Cl• reacts with Cl− to produce Cl2•− reversibly. The ΔGreact aq,calc for reaction 5 is −13.1 kcal/mol and adheres to the LFER. Because Cl2•− is a stable radical, the reverse reaction 7 is substantially slow (kexp ∼ 105 M−1 s−1). Cl2•− is an intermediate species as Cl2•− that subsequently reacts with other radicals and undergoes unimolecular decay. Cl• also reacts with Cl2•− very rapidly to produce molecular chlorine, Cl2, via SET with a ΔGreact aq,calc value of −37.5 kcal/mol (reaction 8). Molecular chlorine Cl2 is also produced from the disproportionation reaction of Cl2•− (reaction 9). A subsequent Cl−Cl bond rupture at different sites forms the different products, such as Cl2 or Cl3−. Cl3− produces Cl2 and Cl− via a slightly exothermic reaction with a value of −0.22 kcal/mol for ΔGreact aq,calc (reaction 11). Molecular chlorine also reacts with Cl− to produce stable Cl3− (reaction 6). Because the reactions 6 and 11 do not undergo radicalinvolved single electron transfer, we did not include these in the correlation analysis for LFER. Because of the nature of Cl2•−, the reaction of Cl2•− with HO• produces stable HOCl (reaction 10) with a negative ΔGreact aq,calc value (i.e., −34.5 kcal/mol). Finally, we investigated the reactivity of ClO•. In our preliminary nonsteady-state kinetic simulation of the UV− chlorine system that followed the changes in pH, a significant amount of •ClO was produced (10−4∼10−5 mol/L under typical low-pressure UV lamps with 4−5 mgCl2/L in DI water at an approximately neutral pH initially). Limited information is available for the reactivity of ClO• with other species.5,7,11 The disproportionation reaction of ClO• was reported as (2.5−7.5) × 10 9 M −1 s −1 (reaction 22). 42,43 We identified a thermodynamically stable pathway via an adduct of Cl−O··· O−Cl with −8.1 kcal/mol of ΔGreact aq,calc, which adheres to the LFER. We propose that O2 and Cl• are the primary reaction products and postulate that this Cl• is rapidly hydrolyzed to produce •Cl(H2O), as shown experimentally.42 The majority of chlorine-derived radical reactions (reactions 5−11 and reaction 22) in the UV−chlorine AOP undergoes either radical combination or SET, and these reactions are found to adhere to the LFER in Figure 1. Formation of •OHCl− and •Cl(H2O). The formation of • OHCl− from the reaction of HO• with Cl− (reaction 12) has been studied experimentally46 and theoretically56−58 due to the intriguing nature of the reactions. While the majority of chlorine-derived reactions with other inorganic radials adheres to the LFER, the reactions involving •OHCl− do not apparently adhere to this relationship. Our attempt here is to identify the thermodynamic properties for reactions 12 by examining 6921
DOI: 10.1021/acs.est.7b00507 Environ. Sci. Technol. 2017, 51, 6918−6926
Article
Environmental Science & Technology
of formaldehyde,61 55% of acetaldehyde,62 and 0.1% of acetone are hydrolyzed to form geminal diols in the aqueous-phase; therefore, hydrated forms of these compounds were included in the alcohol group. The calculated ΔGact aq,calc values for H abstraction from a C−H bond ranged from −4.5 to 5.3 kcal/ mol. The experiment-based ΔGact aq,exp values calculated from the Arrhenius kinetic parameters are 3.6 ± 1.0−5.3 ± 1.2 kcal/mol. Notably, both the theoretical calculation and the experimental values had the highest ΔGact aq,exp for methanol. The estimated k based on the linear correlation for methanol was 6.0 × 107 M−1 s−1, which was more than 1 order of magnitude smaller than the value of kexp. The kexp value for H abstraction from C−H bond at β-position of ethanol also indicated the similar trend. Therefore, we did not include these data points in the correlation. Except for methanol, we observe a linear correlation between kexp and ΔGact aq,calc obtained from the H abstraction of all tested alcohols (Figure 2). Regardless of the addition of one to three explicit water molecule(s), no significant changes were observed in the ΔGact aq,calc values for methanol. This finding is probably because methanol does not undergo H abstraction by Cl• but by another reaction mechanism. For H abstraction from the oxygen−hydrogen (O−H) bond of alcohol functional groups, the ΔGact aq,calc values ranged from 5.4 to 12.0 kcal/mol, indicating the insignificance of this reaction. The calculated ΔGreact aq,calc values for SET are positive and range from 23.6 to 27.1 kcal/mol for all the tested alcohols. Consequently, our theoretical calculations indicate that SET by Cl• is not thermodynamically favorable. Next, we investigated the formation of a Cl adduct by the reaction of Cl• with the oxygen of alcohol functional groups. This reaction was studied because ab initio QM calculations do not allow the calculation of ΔGact aq,calc for a specific site via SET, although Gilbert proposed SET for these sites.22 The ΔGact aq,calc values range from 2.3 to 3.8 kcal/mol, indicating that the formation of the Cl adduct competes with H atom abstraction reactions; accordingly, another linear correlation is found (Figure 3). The investigation here reveals that Cl• forms an adduct with the OH of alcohols, and subsequently, SET is thought to occurs. Note that the data point for methanol is consistent with a linear correlation, and Cl-adduct formation is the dominant mechanism for methanol. To investigate the selectivity of Cl• for H abstraction, we compared the ΔGact aq,calc values for longer chain alcohols. The values at the α and β positions of ethanol are −2.4 and ΔGact aq,calc 4.8 kcal/mol, respectively. This observation agrees with the experimentally obtained partial rate constants for ethanol (1.5 × 109 M−1 s−1 at the α position and 7.5 × 108 M−1 s−1 at the β position).22 A similar trend was observed for 1-propanol and 2butanol (Table S2). In general, Cl• reactions have an earlier transition state, with a lower occurrence of bond breaking and a contribution from the resonance stabilization of H abstraction at the α position in the transition state.22 This fact explains why the calculated bond length of carbon−hydrogen is significantly shorter than that of the hydrogen−chlorine bond at the transition states. Because of this phenomenon, Gilbert proposed the nonselective reactivity of Cl•;22 however, our calculations do not support this concept. Except for methanol, the ΔGact aq,calc values at the α positions for the other eight alcohols are much smaller (i.e., −5.3−2.0 kcal/mol). Our observation of the selective Cl• reactivity for H abstraction can be explained with the bond dissociation energies (BDE) of the C−H bonds in alcohols at various positions,25 although the BDE are measured in the gaseous phase. For example, the BDE
Figure 2. Linear correlation between ln kexp and ΔGact aq,calc for H abstraction by Cl• from a C−H bond. The vertical bar results from the range of the reported kexp values in the literature. Labels: 2, ethanol (αposition); 3, 1-propanol; 4, 2-propanol; 5, 2-butanol; 6, tert-butanol (C−H bond) and 6′ tert-butanol (O−H bond); 7, hydrated formaldehyde; 8, acetaldehyde; 9, hydrated acetaldehyde; 10, propionaldehyde; 11, acetone; 14, 2-butanone; 17, acetate; 18, acetic acid; 19, propionic acid; 20, isobutyric acid; 21, methylformate; 22, ethylformate; 23, methyl acetate; 24, ethylaceate; 25, diethyl ether; 26, methyl-tert-butyl ether; and 28, dichloromethane. Note that compounds are not included from 1, methanol; 12, hydrated acetone; 13, chloroacetone; 15, formate, and 16, formic acid due to either different mechanism or uncertainty.
Figure 3. Linear correlation between ln kexp and ΔGact aq,calc for Cl adduct formation of alcohol and carboxylic compounds. The vertical bar results from the range of the reported kexp values in the literature. Labels: 1, methanol; 2, ethanol; 3, 1-propanol; 4, 2-propanol; 5, 2butanol; 6, tert-butanol; 7, hydrated formaldehyde; 9, hydrated acetaldehyde; 15, formate; 16, formic acid; 17, acetate; 18, acetic acid; 19, propion acid, and 20, isobutyric acid.
6922
DOI: 10.1021/acs.est.7b00507 Environ. Sci. Technol. 2017, 51, 6918−6926
Article
Environmental Science & Technology of C−H at the α and β positions in ethanol are 95.8 and 100.7 kcal/mol, respectively.63 A similar trend can be seen for the BDE of 2-propanol (i.e., 91.7 and 94.3 kcal/mol at the α and β positions, respectively). act values for H Aldehydes and Ketones. The ΔGaq,calc abstraction for a total of 5 aldehydes and ketones range from 0.17 to 7.1 kcal/mol, which are significantly larger values than those for alcohols. These values are consistent with the kexp (i.e., 106∼108 M−1 s−1). In general, H abstraction involves a considerable charge separation between the negatively charged chlorine atom and the positively charged hydrogen that is abstracted from a C−H bond in the transition state.22 As a result, a higher ΔGact aq is required for the reaction to occur. A pair of reported kexp values for acetone have a large discrepancy (i.e.,
