Electrochemical Oxidation of Phenolic Compounds at Boron-DopedDiamond Anodes: Structure−Reactivity RelationshipsYi Jiang,*,†,∥ Xiuping Zhu,‡,∥ and Xuan Xing§,∥

†Department of Energy, Environmental and Chemical Engineering, Washington University in St. Louis, St. Louis, Missouri 63130,United States‡Department of Civil and Environmental Engineering, Louisiana State University, Baton Rouge, Louisiana 70803, United States§College of Life and Environmental Sciences, Minzu University of China, Beijing 100081, China∥Department of Environmental Engineering, Peking University, The Key Laboratory of Water and Sediment Sciences,Ministry of Education, Beijing 100871, China

*S Supporting Information

ABSTRACT: Electrochemical oxidation of phenolic compounds using boron-doped diamond (BDD) anodes has been shown as an effective approach toremove these contaminants from water. However, the understanding of the reac-tion mechanisms of substituted phenolic compounds at the BDD anode remainsincomplete. In the present work, we investigated the electrochemical oxidation of12 representative phenolic compounds (with varied substitution groups (e.g.,−CH3, −OCH3, −NH2, −Cl, −OH, −COOH, −NO2, −CHO) and positions(-ortho, -meta, and -para)) at the BDD anode. Our analysis shows that unlikeprevious studies the two parameters, the Hammett constants of the substituentsand the highest atomic charge on the aromatic ring, fail to adequately describe thereaction rate change when the chemical structures become complicated (i.e., withincreased steric effects). Instead, a quantitative structure−property relationship(QSPR) was established with 26 molecular descriptors and using a partial least-squares regression approach. The QSPR analysis shows that the energy gap between the lowest unoccupied molecular orbital andthe highest occupied molecular orbital, ELUMO − EHOMO, which reflects the chemical stability of a molecule, is the predominantmolecular descriptor determining the reaction rate constant. Furthermore, the predicated rate constants agree well with theobserved ones. The findings are consistent with previous studies of SnO2 anodes, suggesting that chemical structural parameterssuch as the molecular orbital energies are critical to consider when elucidating and predicating the electrochemical reactivity ofphenolic compounds at these nonactive anodes.

■ INTRODUCTION

Phenolic compounds are produced in various industrial pro-cesses, including dyes, textiles, coking, pharmaceuticals, andpesticides, among others.1−5 While many phenolic compoundsare listed in the Toxic Pollutant List by the US EPA, they cannotbe effectively removed from wastewaters through conventionalbiological processes due to their biological recalcitrance. Toaddress this challenge, electrochemical advanced oxidation tech-nology has been developed and demonstrated.6−10 With thistechnology, various radicals, especially hydroxyl radicals thathave strong oxidizing power, are produced to react with pollutants.The critical component in this electrochemical system is theelectrode, with some of the most popular ones being Pt, RuO2,IrO2, SnO2, PbO2, and boron-doped diamond (BDD). Inparticular, electrochemical oxidation at the BDD anode, due toits remarkable mineralization ability, wide potential window,strong anticorrosion stability, and low surface adsorption, hasattracted significant attention and been extensively studied.11−14

BDD films are grown by chemical vapor deposition from avariety of carbon-containing precursors, whereby doping boron

(with a B/C ratio of about 10−5 to 10−3) creates a p-type, moreconductive semiconducting character to diamond.15 This featurerenders BDD both indirect and direct oxidation reaction regimesunder different applied potentials.16 The direct oxidation occursin the potential region before oxygen evolution (i.e., waterstability), and the indirect oxidation takes place in the potentialregion of oxygen evolution (i.e., water decomposition), whichmainly involves reactions of electrogenerated hydroxyl radi-cals.16,17 The indirect oxidation pathways have been proven to bevery effective in degrading organics into CO2 and H2O, includingphenolic ones.6,13,16,18−20

One area of research focus has been understanding thereaction mechanisms of substituted phenolic compounds at theBDD anode. This has been partially achieved through comparingthe reactions of structurally similar compounds at the anode,regarding reaction rates and pathways. Flox et al. revealed

Received: March 20, 2017Revised: May 9, 2017Published: May 11, 2017

Article

pubs.acs.org/JPCA

© 2017 American Chemical Society 4326 DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

that o-, m-, and p-cresol have a similar minerization rate atthe BDD anode, suggesting the negligible role of the positionof the methyl substitution.21 We, however, found that formonosubstituted nitrophenols the degradation rate followed:o-nitrophenol > p-nitrophenol > m-nitrophenol > phenol.19

Furthermore, Canizares et al. showed that the release of chloro-or nitro-group was the first step in the degradation of nitro- orchloro-substituted phenols.18 We further confirmed that therelease of p-substituted groups from the aromatic ring wasthe rate-limiting step.16 Overall, these studies have shown thatthe substituents and positions would have a major impact on theelectrochemical reactivity and subsequent reaction pathways, andelectrophilic attack of hydroxyl radicals is the main reactionmechanism at the BDD anode.A quantitative understanding has been attempted by cor-

relating the reaction rate constants (or reaction rates) with theHammett constants (σ) of the substituents.16,19,20 The Hammettconstants, representing the electron-donating or -withdrawingcharacter of the substituents, were shown to have an overall linearrelationship with the reaction rates of simple phenolic struc-tures, such as the p-substituted phenols16 and nitro-substitutedphenols.19 Furthermore, to elucidate the reaction pathways, weidentified the carbon atom(s) on the aromatic ring with highatomic charge (electron-rich) as the active site(s) for theelectrophilic attack of hydroxyl radicals produced at the BDDanode.19,20 This hypothesis has been confirmed by analyzing thereaction intermediates.19,20 These studies were among the firstones to explain reaction rates and pathways based on available/calculated structural parameters at the BDD anode. However,this knowledge has been very limited due to relatively simplechemical structures (mostly monosubstituted phenols). Forphenolic compounds with more complicated chemical struc-tures, where the steric effect plays a significant role, such infor-mation remains incomplete.In this work, we investigated the electrochemical oxida-

tion of 12 substituted phenolic compounds at the BDD anode(Figure 1). Their degradation kinetics were fitted by pseudo-first-order kinetics and correlated with two structural parameters, theHammett constants of the substituents and the highest atomiccharge on the aromatic ring. Followed by that, we calculated thequantum-chemical molecular descriptors using a semiempiricalapproach (PM3) and performed a quantitative structure−property relationship analysis. The study shows that eitherthe Hammett constants or the carbon atomic charge cannotadequately correlate with the electrochemical reactivity due totheir fragmentary role in describing the whole chemical structure.

The QSPR analysis, on the contrary, identifies the energy gapELUMO − EHOMO to be the most important molecular descriptorthat determines the reaction rate constant. The predicated rateconstants agree well with the observed ones, suggesting theapplicability of the structure−reactivity model in explaining andpredicting the rate constants. This study deepens our under-standing on how chemical structures affect the electrochemicalreactivity of phenolic compounds at nonactive anode such as theBDD anode.

■ MATERIALS AND METHODSChemicals. Twelve phenolic compounds, including phenol

(Ph), o-nitrophenol (o-NO2), m-nitrophenol (m-NO2), p-nitro-phenol (p-NO2), resorcinol (m-OH), p-hydroxybezonic acid(p-COOH), p-hydroxybezaldehyde (p-CHO), p-cresol (p-CH3),p-methoxyphenol (p-OCH3), p-aminophenol (p-NH2), p-chlor-oresorcinol (p-Cl + m-OH), and nitrocatechol (p-NO2 +o-OH) were purchased from Beijing Chemical. All chemicalswere analytical grade and used without further purification.Solutions were prepared using deionized Milli-Q water(Millipore). The BDD electrode was purchased from CON-DIAS, Germany.

Bulk Electrolysis. For bulk electrolysis, a BDD electrodewith a working geometric area of 4 cm2 was used as the anode,while a stainless-steel plate with the same area was used as thecathode. The distance between the two electrodes was set to be10 mm. The experimental conditions followed those in ourprevious work.19 The bulk electrolysis was performed in a beakerreactor under galvanostatic conditions, with a current density of20 mA cm−2 and at ambient temperature (25 °C). The initialconcentrations of the phenol solutions were 1 mM, with 0.2 MNa2SO4 as the supporting electrolyte. The pH was adjusted to be11. The 250 mL solution was stirred by a magnetic stirring barduring electrolysis to promote mass transfer. Samples werecollected at each time interval (40 min) and stored at 4 °C ifimmediate analysis was not performed. All of the experimentswere carried out in duplicate. Before experiments started, BDDelectrode was subjected to ultrasound for 5 min to removecontaminants and then washed with deionized water; stainless-steel cathode was polished and then washed with deionizedwater.

Chemical Analysis. The concentration of these phenols wasmeasured by high-performance liquid chromatography (HPLC)with a ZORBAX SB-C18 column and a DAD detector (AgilentHP1100).19 The mobile phase was methanol/water (50:50), andthe flow rate was 1.0mLmin−1. The wavelength for UV detection

Figure 1. Chemical structures of studied phenols with carbon atom charge (eV, marked in purple) calculated by the PM3 method.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4327

was 314 nm for p-nitrophenol, 210 nm for p-nitrocatechol, and280 nm for other phenols.Molecular Descriptors Calculation. Unlike the ab initio

molecular orbital calculations, semiempirical methods startwith the general form of ab initio Hartree−Fock calculationsbut make approximations for the various Integrals, thus beingmore convenient yet with satisfactory results. Here we used thePM3 semiempirical method (Parameterized Model number 3),by which electrons are included explicitly and many of thedetailed integrals are replaced by empirical parameters.22 Thequantum-chemical calculation was performed with the softwareWinMopac (Ver. 7.20, Fujitsu), with the following keywords:PM3, EF, ESP, POLAR, DIPOLE, BONDS, ENPART, PRECISE,and NOINTER. In total, 26 descriptors were calculated as shownin Table 1. The detailed calculation results are provided in theSupporting Information.Partial Least-Squares Regression Analysis. Partial least-

squares (PLS) regression is a widely used technique that gen-eralizes and combines features from principal component anal-ysis and multiple regression. PLS regression is particularly usefulwhen the matrix of independent variables has larger sizes thanobservations. This technique has been used in a number ofprevious studies to construct a quantitative structure−propertyrelationship between rate constants (observed Y variables) andmolecular descriptors of organic compounds.23−25 The generalunderlying model of multivariate PLS can be described as follows

= +X TP ET (1)

= +Y UQ FT (2)

where X is an n × m matrix of predicators, Y is an n × p matrixof responses, T and U are n × l matrices that are projections ofX and Y, respectively (factor matrix), P and Q are m × l andp × l orthogonal loading matrices, and matrices E and F are theerror terms. The decompositions of X and Y are made so as tomaximize the covariance between T and U. The PLS regressionanalysis was performed using software OriginPro 2016, with anin-built leave-one-out cross-validation method (OriginLabCorporation).

■ RESULTS AND DISCUSSION

Bulk Electrolysis. Figure 2a shows that the concentrations ofthe 12 phenols decreased with the electrolysis time, but todifferent degrees. Among all phenols, p-NO2 + o−OH (p-nitro-catechol) degraded at the fastest rate, while p-COOH degradedat the slowest rate. At the end of the 4 h electrolysis, o-NO2and p-NO2 + o-OH were completely removed, while p-COOHhad the most residue. The remaining quantity of the phenolsfollowed this order: p-COOH (0.49 mM) > p-CH3 (0.45 mM) >p-OCH3 (0.44 mM) > Ph (0.42 mM) > p-CHO = m-OH(0.35 mM) > p-NH2 (0.31 mM) > p-Cl + m-OH (0.13 mM) >m-NO2 (0.11 mM) > p-NO2 (0.08 mM) > o-NO2 (0.02 mM) >p-NO2 + o-OH (0 mM). The degradation of all phenols wasfitted to pseudo-first-order kinetics (Figure 2b), and the apparentreaction rates were obtained (Table 2). The degradation meetspseudo-first-order kinetics (R2 = 0.950 to 0.999), revealing amass-transfer-controlled process. The pseudo-first-order kineticsare also consistent with previous observations of electrochem-ical degradation of nitrogen-heterocyclic compounds at theBDD anode20 and phenols at the SnO2 anode.

24 The calculatedreaction rate constants show that the fastest rate constant

Table 1. Molecular Descriptors Calculated by the PM3 Methoda

nomenclature molecular descriptors

Mw molecular weightu dipole momentα average molecular polarizabilityHOF heat of formationEE electron energyTE total energyCCR core−core repulsion energyq+ most positive net atom chargeq− most negative atom chargeELUMO energy of the lowest unoccupied molecular orbitalEHOMO energy of the highest occupied molecular orbitalELUMO − EHOMO absolute hardness(ELUMO − EHOMO)

2

ELUMO + EHOMO electron negativityBO bond order; for phenol, the weakest C−H was selected; for phenols with more than two substituents, the weakest C−X was selected (X is the

substituent atom)qx net atomic charges on the substituent atomqc atomic charge of the carbon atoms connected with the substituent atomEE1 electron−electron repulsion energy of one-center term for the substituent atomEN1 electron−nuclear attraction energy of the one-center term for the substituent atomJ resonance energy of the two-center term for the carbon-substituent atom bondK exchange energy of the two-center term for the carbon-substituent atom bondEE2 electron−electron repulsion energy of the two-center term for the carbon-substituent atom bondEN2 electron−nuclear repulsion energy of the two-center term for the carbon-substituent atom bondNN2 nuclear−nuclear repulsion energy of the two-center term of the carbon-substituent atom bondC Coulombic interaction energy of the two-center term for the carbon-substituent atom bond, C = EE2 + EN2 + NN2TE2 total of electron and nuclear energies of the two-center term for the carbon-substituent atom bond

aDetailed calculation results are provided in the Supporting Information.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4328

(p-NO2 + o-OH, k = 0.0200 min−1) can be almost one order ofmagnitude higher than the slowest (p-COOH, k = 0.0029 min−1).Take phenol as a benchmark; almost all have higher reaction rateconstants, except p-COOH, p-CH3, and p-OCH3.Correlations between Reaction Constants and Ham-

mett Constants. The Hammett constant was first developedfrom linear free-energy relationships (LFERs) as a quantitativemeasurement of the effect of structural moieties on the electroniccharacter of a given aromatic system.26 A positive value of theHammett constant indicates an electron-withdrawing group,

while a negative value indicates an electron-donating group.However, because of its nature in describing the electroniceffects, the Hammett constants usually can only be applied topara- and meta-substitution positions, where the steric effects areminimal. Overall, the phenols with electron-withdrawing sub-stituent (e.g., −NO2 and −Cl) have higher reaction rates thanthose with electron-donating substituents (e.g., −CH3 and−OCH3 groups). We here correlated the reaction rate constantsand Hammett constants of the para- or meta-substituted phenolsthrough the Hammett equation26

ρ σ= Σ⎛⎝⎜

⎞⎠⎟

kk

log ( )H (3)

where k and kH are the reaction rate constants of substituted andunsubstituted compounds respectively, ρ is the susceptibilityfactor, and ∑σ is the sum of the Hammett constants.For all meta- and para-substituted phenols, there exists a poor

correlation between the rate constants and the Hammett con-stants (R2 = 0.40, Figure 3a). However, after excluding two

outlier data points for p-COOH and p-NH2, there appears to be alinear relationship between the rate constants and the Hammettconstants (R2 = 0.92, Figure 3b). The rate constant for p-COOHappeared lower and the rate constant of p-NH2 was higher thanwhat was expected from the Hammett constants. For thedegradation of p-COOH, 1,4-hydroquinone, 1,4-benzoquinone,and 3,4-dihydroxybenzoic acid were observed as intermediateswhen the hydroxyl radical was the predominant oxidizing

Figure 2. (a) Evolution of phenols concentration with time duringelectrochemical oxidation at the BDD anode. (b) Linear relationshipbetween ln(C0/C) and electrolysis time.

Table 2. Pseudo-First-Order Kinetics of ElectrochemicalDegradation of Phenols at BDD Anode

phenolsk

(min−1)pseudo-first-order

equationt1/2

(min) R2

Ph 0.0036 ln(C0/Ct) = 0.0036t 193 0.997o-NO2 0.0146 ln(C0/Ct) = 0.0146t 47 0.985m-NO2 0.0102 ln(C0/Ct) = 0.0102t 68 0.950p-NO2 0.0093 ln(C0/Ct) = 0.0093t 75 0.957p-NO2 + o-OH 0.0200 ln(C0/Ct) = 0.0200t 35 0.991m-OH 0.0046 ln(C0/Ct) = 0.0046t 151 0.979p-COOH 0.0029 ln(C0/Ct) = 0.0029t 239 0.999p-CHO 0.0042 ln(C0/Ct) = 0.0042t 165 0.989p-CH3 0.0034 ln(C0/Ct) = 0.0034t 204 0.994p-OCH3 0.0034 ln(C0/Ct) = 0.0034t 204 0.998p-NH2 0.0045 ln(C0/Ct) = 0.0045t 154 0.972p-Cl + m-OH 0.0087 ln(C0/Ct) = 0.0087t 80 0.990

Figure 3. Correlations between reaction rate constants and Hammettconstants of meta- and para-substituted phenols: (a) includes all ninephenols and (b) includes all phenols but p-NH2 and p-COOH.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4329

species present.27 The slow reaction rate for p-COOH was likelydue to the production of polymeric materials from those inter-mediates that rapidly decreased electrode activity (electrodefouling).28,29 On the contrary, for the electrochemical oxidationof p-NH2, a concurrent hydrolysis reaction producing hydro-quinone was observed before,30 which resulted in a fast dis-appearance rate for p-NH2 in this case.The limitation of using Hammett constants is the difficulty to

describe the ortho-substitution, where the steric effect comesinto play. As a result, o-NO2 and p-NO2 + o-OH were not able tobe taken into analysis. Furthermore, the Hammett constants lackthe consideration for intramolecular interaction (e.g., the for-mation of hydrogen bond bridges, tautomerism, and salinebridge),24 thus further limiting its applicability and accuracy.Correlations between Reaction Rate Constants and

Carbon Atom Charges. The electrophilic attack of hydroxylradicals was revealed to be the predominant reaction mechanismat the BDD anode.12,16,19 Accordingly, the electron densitydistribution is expected to have an impact on the reaction rateand the production of intermediates. We have previouslyrevealed that through the calculation of electron densitydistribution, the reaction active sites were identified as thosewith abundant carbon atom charge, which was confirmed bysubsequent intermediate analysis using HPLC and GC−MS.19,20

Through theoretical calculation by WinMOPAC, the highestcarbon atom charges on the aromatic ring were obtained(Figure 1). Correlations between the reaction constants andhighest carbon atom charge on the aromatic ring are presented inFigure 4. From Figure 4a, for the seven p-substituted phenolsexcept p-NH2 and p-COOH, the reaction rate constant has alinear relationship with the highest (most negative) carbon atomcharge (R2 = 0.860). Furthermore, for meta- and para-substitutedphenols, the linear relationship remained (R2 = 0.765), althoughthere appears to be an appreciable deviation for m-NO2. Whenconsidering all 12 phenols, the correlation coefficient continuedto decrease to 0.673. Overall, the trend of decreasing correlationcoefficients indicates that the highest atom charge does play asignificant role in determining the electrochemical degradationrate of substituted phenols (R2 = 0.673−0.860), but the stericeffect becomes significant in more complicated structures (e.g.,multiple substituted phenols). As a result, when complicatedby both significant electronic and steric effects, it is difficult touse one single parameter, highest carbon atom charge on thearomatic ring, to explain the differences of electrochemicalactivities of different substituted phenols at the BDD anode.More sophisticated methods, such as establishing a quantitativestructure−property relationship that considers parameters derivedfrom whole molecules, are needed to offer a better explanation ofthe electrochemical reactivity.Quantitative Structure−Property Relationship Analysis.

The PLS analysis has shown two principle componentslead to the minimum root mean of the prediction error sum ofsquares (PRESS, min = 0.6372). The two principle componentscontribute to 61.9 and 87.3% of cumulative X and Y variance,respectively. The VIP (variable importance in the projection)values were used to reduce the number of variables inthe model (Table S2). Usually, a descriptor (or X variable)with the VIP value smaller than 1 can be considered as unimpor-tant and thus be excluded from the further QSRP model-ing.24,31 From Table S2, molecular descriptors with VIP > 1,namely, Mw, u, EE, TE, CCR, q

+, q−, ELUMO, (ELUMO − EHOMO),(ELUMO − EHOMO)

2, (ELUMO + EHOMO), qx, qc, and C, were more

important than others and were used to construct the model.

We obtained the relationship between ln(k) and the structuraldescriptors

= − + + + ×

+ × + × +

− − − −

− − − +

+ − −

−

− − +

−

k M u

q

q E E

E E E E

q q C

ln( ) 2.2263 0.0025 0.0435 2.4686 10 EE

1.3175 10 TE 3.0262 10 CCR 0.1021

0.2675 0.0639 0.3000( E )

0.0166( ) 0.0197( )

0.0733 0.2769 0.0224x c

w5

4 5

LUMO LUMO HOMO

LUMO HOMO2

LUMO HOMO

(4)

Figure 4.Correlations between reaction constants and highest carbonatom charge on the aromatic ring: (a) p-substituted phenol,(b) meta- and para-substituted phenol, and (c) all 12 substitutedphenols.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4330

From the standardized coefficients in Table 3, the energy gapbetween the lowest unoccupied orbital and the highest occupiedorbital, ELUMO − EHOMO, is the most predominant moleculardescriptor in determining the electrochemical reactivity, followedby the dipole moment u and the Coulombic interaction energy ofthe two-center term for the carbon-substituent atom bond (C).According to the molecular orbital theory, EHOMO reflects the

electron-donating ability of one molecule when it interacts withothers, and electrons will be more easily shared with higherEHOMO. On the contrary, ELUMO denotes the electron acceptingability. Taken together, the energy gap ELUMO − EHOMO showsthe energy needed for one electron to migrate from the highestoccupied orbital to the lowest unoccupied orbital. As a result,ELUMO− EHOMO reflects the chemical stability of a molecule, withhigher value showing more stability. The ELUMO − EHOMO is alsodefined as twice the chemical hardness, with hard moleculesresisting electron transfer or rearrangement and thus being lessreactive.32 The term therefore has a negative impact on thereaction rate constant, as shown in the equation. In our previousstudy, we have been able to identify the role of the delocalizationenergy of a molecule on the degradation rate of nitrogen-heterocyclic compounds, where the delocalization energy alsoreflects the chemical stability.20 Furthermore, this crucial role ofELUMO − EHOMO has also been discovered in the photolysis ofchlorinated biphenyls, in which the radical reaction with homo-lysis of the C−Cl bond was the main reaction mechanism.25

The dipole moment u is due to nonuniform distributionsof positive and negative charges on the various atoms. As anindicator of polarity, the dipole moment reflects the electronicforces that each substituent is applying on the phenyl ring togenerate a particular electronic architecture that is suitable forelectrochemical oxidation.24 The higher the dipole moment, themore uneven the electron distribution. As a result, the moleculesare more prone to electrophilic or nucleophilic attacks. TheCoulombic interaction energy of the two-center term for thecarbon-substituent atom bond (C = EE2 + EN2 +NN2) includesthree terms: the electron−electron repulsion energy of the two-center term for the carbon-substituent atom bond (EE2), theelectron−nuclear repulsion energy of the two-center term for thecarbon-substituent atom bond (EN2), and the nuclear−nuclearrepulsion energy of the two-center term of the carbon-sub-stituent atom bond (NN2). The higher the interaction energy,the lower the reaction rate. The inclusion of the C term in the

model indicates the degradation of substituted phenols inelectrochemical process is associated with the destruction ofcarbon-substituent atom (N or Cl) bond.23 This was consistentwith our previous experimental observations that the release ofthe substituent groups is likely the first step in degradation ofp-substituted phenols.16

Overall, our QSPRmodel has been able to identify a similar setof molecular descriptors as a previous study did,24 which inves-tigated the electrochemical degradation of phenols at theSnO2 anode. The set of molecular descriptors includes ELUMO,EHOMO, and dipole moment, among others. This similarity canbe attributed to similar reaction mechanisms of BDD and SnO2anodes, where hydroxyl radicals are the dominant reactionspecies.12 This implicates that our models may be applicable tothe electrochemical reactions at other nonactive electrodeswhere hydroxyl radicals are predominant. Furthermore, Figure 5

shows the predicated rate constants (calculated from eq 4) agreewell with the observed ones, suggesting the applicability of thestructure−reactivity model in explaining and predicting the rateconstants.

■ CONCLUSIONSOur results show that both the single use of Hammett constantsand the highest carbon atom charge on the aromatic ring cannotaccurately predicate the electrochemical reaction rate constantsof phenolic compounds with complicated chemical structures atthe BDD anode. They fail because only a fragment contributionwas used to account for the observed response. A quantitativestructure−property relationship was established between quan-tum chemical descriptors and the reaction rate constants. Thegeneration of molecular descriptors from quantum molecularmechanics calculations was made considering the wholemolecule, thus having considerably improved the accuracy ofthe model. Our model shows that the energy gap between thelowest unoccupied molecular orbital and the highest occupiedmolecular orbital, ELUMO − EHOMO, is the most importantmolecular descriptor affecting the electrochemical reactivity ofphenols at the BDD anode. The results further confirm thatelectrophilic attack by hydroxyl radicals is the main reactionmechanism at the BDD anode.Furthermore, as discussed in our previous work,16,19 the

removal kinetics of the phenolic substrates was found to bedifferent from that of the chemical oxygen demand (COD),symbolizing different production rates of intermediates. For

Table 3. Variable Importance in the Projection (VIP) andRegression Coefficients from the PLS Analysis (with VIP > 1Shown)

descriptor VIP coefficient standardized coefficient

Mw 1.1573 0.0025 0.07042u 1.3586 0.0435 0.10498EE 1.1788 0.0000 0.05627TE 1.1917 0.0001 0.06222CCR 1.1752 0.0000 0.05491q+ 1.3224 0.1021 0.08446q− 1.2938 −0.2675 −0.07231ELUMO 1.2777 −0.0639 −0.06288(ELUMO − EHOMO) 1.2625 −0.3000 −0.13435(ELUMO − EHOMO)

2 1.2558 −0.0166 −0.13313(ELUMO + EHOMO) 1.1558 −0.0197 −0.03532qx 1.2895 0.0733 0.07288qc 1.2466 −0.2769 −0.08771C = EE2 + EN2 + NN2 1.0596 −0.0224 −0.10408

Figure 5. Comparison between observed and predicated values ofreaction rate constants.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4331

example, the degradation kinetics of phenols with electron-withdrawing groups such as −NO2 (e.g., p-NO2, p-CHO,m-NO2, o-NO2) is usually faster than that of phenol (Ph), butthey accumulate more intermediates than phenol during theelectrolysis.16,19 Efforts will be needed to further correlate suchstructure−reactivity models with the production of intermedi-ates and final products and thus to identify/prevent potentialrisks associated with the intermediates and final products.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.jpca.7b02630.

Detailed values of the molecular structure parameters,variable importance in the projection (VIP) and regressioncoefficients from the PLS analysis. (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: jiang.y@wustl.edu. Phone: +1-314-562-1830.ORCIDYi Jiang: 0000-0002-4193-0330NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work was supported by National Natural ScienceFoundation of China (Grant nos. 20877001 and 51409285).

■ REFERENCES(1) Kiwi, J.; Pulgarin, C.; Peringer, P. Effect of Fenton and Photo-Fenton reactions on the degradation and biodegradability of 2-nitrophenols and 4-nitrophenols in water-treatment. Appl. Catal., B1994, 3 (4), 335−350.(2) Chen, D. W.; Ray, A. K. Photodegradation kinetics of 4-nitrophenol in TiO2 suspension.Water Res. 1998, 32 (11), 3223−3234.(3) Parra, S.; Olivero, J.; Pacheco, L.; Pulgarin, C. Structural propertiesand photoreactivity relationships of substituted phenols in TiO2suspensions. Appl. Catal., B 2003, 43 (3), 293−301.(4) Canizares, P.; Garcia-Gomez, J.; Saez, C.; Rodrigo, M. A.Electrochemical oxidation of several chlorophenols on diamondelectrodes: Part II. Influence of waste characteristics and operatingconditions. J. Appl. Electrochem. 2004, 34 (1), 87−94.(5) Zhou, M.; Lei, L. The role of activated carbon on the removal of p-nitrophenol in an integrated three-phase electrochemical reactor.Chemosphere 2006, 65 (7), 1197−1203.(6) Li, H.; Zhu, X.; Jiang, Y.; Ni, J. Comparative electrochemicaldegradation of phthalic acid esters using boron-doped diamond and Ptanodes. Chemosphere 2010, 80 (8), 845−851.(7) Zhu, X.; Ni, J.; Li, H.; Jiang, Y.; Xing, X.; Borthwick, A. G. L. Effectsof ultrasound on electrochemical oxidation mechanisms of p-substitutedphenols at BDD and PbO2 anodes. Electrochim. Acta 2010, 55 (20),5569−5575.(8) Zhu, X.; Ni, J.; Xing, X.; Li, H.; Jiang, Y. Synergies betweenelectrochemical oxidation and activated carbon adsorption in three-dimensional boron-doped diamond anode system. Electrochim. Acta2011, 56 (3), 1270−1274.(9) Zhu, X.; Ni, J.; Wei, J.; Xing, X.; Li, H.; Jiang, Y. Scale-up of BDDanode system for electrochemical oxidation of phenol simulatedwastewater in continuous mode. J. Hazard. Mater. 2010, 184 (1−3),493−498.(10) Chaplin, B. P. Critical review of electrochemical advancedoxidation processes for water treatment applications. Environ. Sci.:Processes Impacts 2014, 16 (6), 1182−1203.

(11) Panizza, M.; Cerisola, G. Application of diamond electrodes toelectrochemical processes. Electrochim. Acta 2005, 51 (2), 191−199.(12) Zhu, X.; Tong, M.; Shi, S.; Zhao, H.; Ni, J. Essential explanation ofthe strong mineralization performance of boron-doped diamondelectrodes. Environ. Sci. Technol. 2008, 42 (13), 4914−4920.(13) Chaplin, B. P.; Schrader, G.; Farrell, J. Electrochemicaldestruction of N-nitrosodimethylamine in reverse osmosis concentratesusing boron-doped diamond film electrodes. Environ. Sci. Technol. 2010,44 (11), 4264−4269.(14) Jawando, W.; Gayen, P.; Chaplin, B. P. The effects of surfaceoxidation and fluorination of boron-doped diamond anodes onperchlorate formation and organic compound oxidation. Electrochim.Acta 2015, 174, 1067−1078.(15) McCreery, R. L. Advanced carbon electrode materials formolecular electrochemistry. Chem. Rev. 2008, 108 (7), 2646−2687.(16) Zhu, X.; Shi, S.; Wei, J.; Lv, F.; Zhao, H.; Kong, J.; He, Q.; Ni, J.Electrochemical oxidation characteristics of p-substituted phenols usinga boron-doped diamond electrode. Environ. Sci. Technol. 2007, 41 (18),6541−6546.(17) Iniesta, J.; Michaud, P. A.; Panizza, M.; Cerisola, G.; Aldaz, A.;Comninellis, C. Electrochemical oxidation of phenol at boron-dopeddiamond electrode. Electrochim. Acta 2001, 46 (23), 3573−3578.(18) Canizares, P.; Lobato, J.; Paz, R.; Rodrigo, M. A.; Saez, C.Electrochemical oxidation of phenolic wastes with boron-dopeddiamond anodes. Water Res. 2005, 39 (12), 2687−2703.(19) Jiang, Y.; Zhu, X.; Li, H.; Ni, J. Effect of nitro substituent onelectrochemical oxidation of phenols at boron-doped diamond anodes.Chemosphere 2010, 78 (9), 1093−1099.(20) Xing, X.; Zhu, X.; Li, H.; Jiang, Y.; Ni, J. Electrochemical oxidationof nitrogen-heterocyclic compounds at boron-doped diamondelectrode. Chemosphere 2012, 86 (4), 368−375.(21) Flox, C.; Arias, C.; Brillas, E.; Savall, A.; Groenen-Serrano, K.Electrochemical incineration of cresols: A comparative study betweenPbO2 and boron-doped diamond anodes. Chemosphere 2009, 74 (10),1340−1347.(22) Dewar, M. J.; Healy, E. F.; Holder, A. J.; Yuan, Y. C. Comments ona comparison of AM1 with the recently developed PM3 method. J.Comput. Chem. 1990, 11 (4), 541−542.(23) Yuan, S.; Xiao, M.; Zheng, G.; Tian, M.; Lu, X. Quantitativestructure-property relationship studies on electrochemical degradationof substituted phenols using a support vector machine. SAR QSAREnviron. Res. 2006, 17 (5), 473−481.(24) Tian, M.; Thind, S. S.; Simko, M.; Gao, F.; Chen, A. QuantitativeStructure−Reactivity Study of Electrochemical Oxidation of PhenolicCompounds at the SnO2−Based Electrode. J. Phys. Chem. A 2012, 116(11), 2927−2934.(25) Li, X.; Fang, L.; Huang, J.; Yu, G. Photolysis of mono-throughdeca-chlorinated biphenyls by ultraviolet irradiation in n-hexane andquantitative structure-property relationship analysis. J. Environ. Sci.2008, 20 (6), 753−759.(26) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M.Environmental Organic Chemistry; John Wiley & Sons: 2005.(27) Duesterberg, C. K.;Waite, T. D. Kinetic modeling of the oxidationof p-hydroxybenzoic acid by Fenton’s reagent: implications of the role ofquinones in the redox cycling of iron. Environ. Sci. Technol. 2007, 41(11), 4103−4110.(28) Canizares, P.; García-Gomez, J.; Saez, C.; Rodrigo, M. A.Electrochemical oxidation of several chlorophenols on diamondelectrodes - Part I. Reaction mechanism. J. Appl. Electrochem. 2003, 33(10), 917−927.(29) Ferreira, M.; Varela, H.; Torresi, R. M.; Tremiliosi-Filho, G.Electrode passivation caused by polymerization of different phenoliccompounds. Electrochim. Acta 2006, 52 (2), 434−442.(30) Salavagione, H. J.; Arias, J.; Garces, P.; Morallon, E.; Barbero, C.;Vazquez, J. L. Spectroelectrochemical study of the oxidation ofaminophenols on platinum electrode in acid medium. J. Electroanal.Chem. 2004, 565 (2), 375−383.(31) Li, L.; Xie, S.; Cai, H.; Bai, X.; Xue, Z. Quantitative structure−property relationships for octanol−water partition coefficients of

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4332

polybrominated diphenyl ethers. Chemosphere 2008, 72 (10), 1602−1606.(32) Pearson, R. G. Absolute electronegativity and hardness correlatedwith molecular orbital theory. Proc. Natl. Acad. Sci. U. S. A. 1986, 83(22), 8440−8441.

The Journal of Physical Chemistry A Article

DOI: 10.1021/acs.jpca.7b02630J. Phys. Chem. A 2017, 121, 4326−4333

4333