Electrochemical and ColumnInvestigation of Iron-MediatedReductive Dechlorination ofTrichloroethylene andPerchloroethyleneJ A M E S F A R R E L L , * N I K O S M E L I T A S ,M A R K K A S O N , A N D T I E L I

Department of Chemical and Environmental Engineering,University of Arizona, Tucson, Arizona 85721

This research investigated the long-term performance ofzero-valent iron aggregates for reductive dechlorination oftrichloroethylene (TCE) and perchloroethylene (PCE). Theeffects of elapsed time, mass transfer limitations, and influenthalocarbon concentration on reductive dechlorinationrates were investigated using groundwater obtained froma field site contaminated with chlorinated organiccompounds. Over the first 300 days of operation, reactionrates for TCE and PCE gradually increased due toincreasing porosity of the iron aggregates. Although therewas microbial growth in the column, biological activitydid not measurably contribute to reductive dechlorination.Dechlorination rates were pseudo-first-order in reactantconcentration for submillimolar halocarbon concentrations.TCE concentrations near aqueous saturation resulted inpassivation of the iron surfaces and deviation from first-order reaction kinetics. However, this passivation was slowlyreversible upon lowering the influent TCE concentration.Tafel polarization diagrams for an electrode constructedfrom the iron aggregates indicated that corrosion of theaggregates was anodically controlled. At all halocarbonconcentrations, aggregate oxidation by water accounted formore than 80% of the corrosion. Throughout the courseof the 3-yr column investigation, reaction rates for TCE were2-3 times faster than those for PCE. However, currentmeasurements with the aggregate electrode indicated thatdirect PCE reduction was faster than that for TCE. Thisdisparity between amperometrically measured reaction ratesand those measured in the column reactor indicatedthat halocarbon reduction may occur via direct electrontransfer or may occur indirectly through reaction with atomichydrogen adsorbed to the iron. Comparison of aggregatecorrosion rates with those of fresh iron suggested that anodiccontrol of corrosion leads to predominance of theindirect reduction mechanism. The faster reaction rate forTCE under anodically controlled conditions can thereforebe attributed to its faster rate of indirect reduction as comparedto PCE.

IntroductionIn recent years, there has been considerable interest in theuse of zero-valent iron for passive remediation of groundwater

contaminated with chlorinated organic compounds andredox active metals (1-12). Installation considerations oftenlimit zero-valent iron barriers to the remediation of shallowcontaminant plumes in sedimentary aquifers. However, anew application has recently been developed that enableszero-valent iron to be used for remediation of chlorinatedorganic compounds in fractured bedrock (13). This newapplication involves injection of zero-valent iron proppantsinto fractured bedrock using hydraulic fracturing techniques.

The goals of this study were to investigate the long-termeffectiveness of zero-valent iron proppants for reductivedechlorination and to determine the mechanisms controllingreaction rates on aged iron surfaces. Because the ironproppants were designed to be injected into bedrockcontaining pure phase halocarbons, the effectiveness of theproppants for reductive dechlorination at halocarbon con-centrations near aqueous saturation were also investigated.Column experiments were conducted over a 1-yr period inorder to determine the effect of elapsed time, mass transferlimitations, and influent halocarbon concentration on reac-tion rates of trichloroethylene (TCE) and perchloroethylene(PCE). The mechanisms responsible for the results observedduring the first year of operation were then investigated byvarying the column operating conditions over the next 2 yearsand by electrochemical investigation of halocarbon reductionand iron corrosion.

Reaction rates for one and two carbon chlorinatedcompounds are highly variable among different investigators(14). This variability can be attributed to differences inexperimental conditions and to differences in iron type andspecific surface area. Among all the rate affecting factors, thetype of iron appears to be the most significant determinantof reaction rates. Even when reaction rates are normalizedfor specific surface area and are measured under the sameexperimental conditions, rates for a particular compoundmay vary over 3 orders of magnitude between different typesof iron (15).

Early investigators recognized that the condition of theiron surfaces has a profound impact on reaction rates, andearly batch testing was often performed on acid-washed ironfilings in order to achieve a repeatable initial surface condition(1). Several investigations have observed that reaction ratesare initially first order in reactant concentration but deviatefrom first-order behavior with increasing elapsed time (2,16). Declining reaction rates with time have been attributedto increasing mass transfer limitations through surface-associated oxide layers (2) and to anodic control of ironcorrosion (16). X-ray diffraction analyses from field andcolumn tests indicate that the passivating layer on iron filingsconsists of an inner layer of magnetite (Fe3O4) with an outercoating of maghemite (γ-Fe2O3) (16, 17).

There are both direct and indirect mechanisms forhalocarbon reduction by zero-valent metals. Direct reductionmay occur by electron tunneling or via formation of achemisorption complex of the organic compound withsurface metal atoms (18). Electron tunneling may occur tohalocarbons that are physically adsorbed at the metal surfaceor to halocarbons that are separated from the surface byadsorbed water or corrosion products (19, 20). Indirectreduction of organic compounds involves atomic hydrogen.Atomic hydrogen adsorbed at the metal surface may reduceorganic compounds through the formation of chemisorbedhydride complexes (18). This mechanism is fast on metalswith low hydrogen overpotentials, such as platinum andpalladium, but is much slower on metals with high hydrogenoverpotentials, such as iron (18). Indirect evidence has also

* Corresponding author phone: (520)621-2465; fax: (520)621-6048;e-mail: farrell@engr.arizona.edu.

Environ. Sci. Technol. 2000, 34, 2549-2556

10.1021/es991135b CCC: $19.00 2000 American Chemical Society VOL. 34, NO. 12, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2549Published on Web 05/05/2000

been presented that Fe2+ may contribute to halocarbonreduction (21). However, two investigations into the effectof Fe2+-chelating agents on halocarbon reduction ratesindicated that the effect of available Fe2+ on reaction ratesis small (22, 23).

Rates of reductive dechlorination for one and two carboncompounds have been correlated with several halocarbonproperties, including the equilibrium potentials for one andtwo electron-transfer reactions (24, 25). Generally, iron-mediated dechlorination rates for homologous compoundsincrease with increasing degree of halogenation. For example,recent articles using data compiled from a variety of publishedsources have reported that PCE reaction rates are ap-proximately 5 times faster than those for TCE, which in turnare 3-60 times faster than those for the three dichloro-ethylene isomers (14, 24). However, this trend in reactivityhas not been observed in all studies. For example, Burris etal. (5) reported that reaction rates for TCE and PCE weresimilar in short-term batch testing. This disparity frompreviously reported trends was attributed to possible masstransfer limitations and greater adsorption of PCE to non-reactive sites. Increasing dechlorination rates with decreasingdegree of halogenation have been observed for chlorinatedethenes in bimetallic iron-palladium systems (26) and forreduction by supported palladium catalysts (27, 28).

Electrochemical Analysis. The current-voltage relation-ship for redox reactions involving a corroding iron electrodecan be described by a form of the Butler-Volmer equation(29):

where i is the net current, icorr is the corrosion current, E isthe electrode potential, Ecorr is the free corrosion potential,and âc and âa are the cathodic and anodic Tafel slopes,respectively. Under open circuit conditions where theelectrode potential is equal to Ecorr, i is equal to zero, and thecorrosion rate of the electrode is equal to icorr. At electrodepotentials sufficiently above or below Ecorr, eq 1 indicatesthat a plot of log |i| versus E will be linear. The linear regionsof the polarization profiles have slopes equal to âa and âc andare known as the anodic and cathodic Tafel regions. The icorr

under open circuit conditions can be determined graphicallyby extrapolating the anodic and cathodic Tafel slopes to theirpoint of intersection at Ecorr (19).

Under open circuit conditions, the overall rate of ironcorrosion may be limited by either the anodic or cathodichalf-cell reactions. For corrosion of iron by aqueous TCE inanaerobic solutions, the anodic reaction is given by

while the cathodic reactions when ethene is the product maybe expressed as

If the aggregate rate of the anodic reactions is slower thanthe potential rate of the cathodic reactions, corrosion of theiron is anodically controlled (30). Under anodically controlledconditions, the rate at which oxidants are able to acceptelectrons is faster than the rate at which the iron is able torelease electrons. In contrast, cathodically controlled cor-rosion occurs when the rate at which oxidants are able toaccept electrons is slower than the rate at which the iron isable to release electrons. The relative magnitude of âa andâc indicate whether the anodic or cathodic reactions arelimiting the overall rate of iron corrosion. For example, when

âc > âa, corrosion of the iron is anodically controlled. Thecondition of the iron surfaces and the polarizing ability ofpotential oxidants determine whether iron corrosion isanodically or cathodically controlled.

Materials and MethodsIron Column Experiments Pseudo-steady-state columnexperiments were used to measure reaction rates anddetermine reaction byproducts of TCE and PCE transforma-tion. The iron reactants used in this investigation consistedof 300-1000 µm diameter, spherical iron aggregates obtainedfrom Cercona of America (Dayton, OH). The iron aggregatesare a commercially available material produced for use asreactive proppants for in-situ remediation of halocarbons infractured bedrock. The iron aggregates were composed ofsmaller (<100 µm diameter) iron particles cemented togetherwith an aluminosilicate binder, which comprised 2% of theparticle mass. The specific surface area and internal porosityof the iron aggregates were measured before and after theinvestigation by Micromeritics (Norcross, GA) using nitrogenBET analysis (31) and mercury porosimetry.

The iron reactants were packed in a 60 cm long by 4.8 cmi.d. glass column (Kontes, Vineland, NJ). The column wasadapted by installing three intracolumn sampling ports sealedwith silicone rubber septa at 15, 30, and 45 cm from thecolumn inlet. After 250 days, port 3 (at 45 cm) becamedamaged and could no longer be sampled. The column voidvolume and axial dispersivity (R) were determined using abromide tracer solution. Properties of the column and thezero-valent iron reactants are listed in Table 1.

For the first 426 days of the investigation, the column wasoperated using groundwater obtained from a leach fieldassociated with a former aerospace facility in southernCalifornia. The groundwater obtained from the site wascontaminated with a variety of chlorinated hydrocarbons atconcentrations up to 180 mg/L (13). The site groundwateralso contained a variety of inorganic contaminants, includingnitrate at concentrations up to 45 mg/L. Carbonate was addedto the water at a level of 3 mM, and dissolved oxygen wasremoved by purging the feedwater reservoir with a mixtureof nitrogen and carbon dioxide.

Other details of the experimental setup are given in apreviously reported investigation (16). Reactant concentra-tion profiles were measured by withdrawing 100-µL samplesfrom the inlet, outlet, and three intracolumn sampling ports.The 100-µL aqueous samples were injected into 1 g of pentaneand analyzed by injection into a Hewlett-Packard 5890 gaschromatograph (GC) equipped with an electron capturedetector and autosampler. Analyses of nonchlorinated reac-tion byproducts were performed using a molecular sievechromatography column (Alltech, Deerfield, IL) and a flameionization detector connected to a SRI Instruments GC(Torrance, CA). Trace reaction byproducts were identifiedusing a Finnigan (San Jose, CA) gas chromatograph/massspectrometer. Anion and dissolved iron concentrations weredetermined by the Soil, Water, and Plant Analysis Laboratoryat the University of Arizona.

i ) icorr[e-âc(E-Ecorr) - eâa(E-Ecorr)] (1)

Fe f Fe2+ + 2e- (2)

H2O + H+ + 2e- f H2 + OH - (3)

C2HCl3 + 3H+ + 6e- f C2H4 + 3Cl- (4)

TABLE 1. Column and Iron Properties

empty bed vol (mL) 1085mass of iron aggregates (g) 3639initial aggregate specific surface area (m2/g) 3.7final aggregate specific surface area (m2/g) 5.3initial aggregate internal pore vol (mL/g) 0.073final aggregate internal pore vol (mL/g) 0.121initial packed bed pore vol (mL) 647final packed bed pore vol (mL) 552initial axial dispersivity (cm) 0.22final axial dispersivity (cm) 2.2

2550 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 12, 2000

Over the course of the 3-yr investigation, the column wasoperated over a flow rate range of 0.04-5 mL/min. For thefirst 30 days, the column was operated with only TCE in thefeed stream at an influent concentration of 0.3 mM. After 30days, the feed stream was supplemented with PCE at aconcentration of 0.3 mM. Between 309 and 427 days, theinfluent TCE concentration was increased to 7.8 mM andPCE was removed from the feed stream. After 427 days, thesite groundwater was replaced with ultrapure deionized (DI)water (18 MΩ-cm resistivity) in order to eliminate potentialnutrients for microbiological activity. Between 427 and 507days, 0.40 M lithium chloride was added to the feedwater toinhibit microbiological activity (32). Hexachloroethane (HCA)was added to the feed stream for a short period to investigatemass transfer limitations associated with adsorption-retardedaqueous diffusion through porous iron corrosion products.A summary of the operating conditions is listed in Table 2.

Amperometric Experiments. The effect of halocarbonconcentration on iron corrosion rates and surface passivationwas investigated by measuring the icorr and Ecorr for anelectrode constructed from 0.75 g of the iron aggregatescemented together with a conducting epoxy. These experi-ments were performed by placing the aggregate electrode inanaerobic 10 mM CaSO4 solutions of varying TCE or PCEconcentration. The corrosion currents and free corrosionpotentials were measured by analysis of Tafel diagramsproduced by polarizing the electrode (50 mV with respectto its open circuit potential (30). The polarization experimentswere performed using an EG&G (Oak Ridge, TN) model 273Ascanning potentiostat and M270 software. For comparativepurposes, corrosion currents were also determined for ironwire electrodes in 10 mM CaSO4 solutions of varying TCE orPCE concentration. The iron wire electrodes were obtainedfrom Aesar (Ward Hill, MA) and consisted of 2.8 cm long and1.2 mm diameter iron wire of 99.9% purity.

Chronoamperometric experiments were performed withthe aggregate and iron wire electrodes in a custom glass cellwith a nitrogen-purged headspace. The test solutions werepurged with nitrogen gas containing TCE or PCE at 50% ofvapor phase saturation. Once the solutions were equilibratedwith the halocarbon vapors, the solution purge was termi-nated, and the electrode potential was stepped to -50 mVwith respect to its open circuit potential. The resulting currentwas recorded using the potentiostat and associated software.

Adsorption Experiments. Adsorption of TCE, PCE, andHCA onto simulated iron corrosion products was measuredin order to evaluate adsorption-retarded diffusion througha porous layer of iron corrosion products. The simulatediron corrosion products were produced by oxidizing a sampleof the iron aggregates at 650 °C under an air atmosphere for5 days. Halocarbon adsorption to the oxidized aggregateswas then measured using a previously described batch testingprocedure (33). Adsorbed/aqueous phase distribution coef-ficients (Kd) for each compound were determined from theslope of the adsorption isotherms.

Results and DiscussionUpon commencement of operation, 90 days was required toestablish reactant concentration profiles that conformed toa first-order reaction rate model. Once first-order conditions

were established, the reactant concentrations (C) along thelength of the column could be described by

where Co is the influent reactant concentration, x is thecoordinate of position, u is the interstitial fluid velocity, Dis the dispersion coefficient, and k1 is the first-order reactionrate constant. For a reaction that is first order in reactantconcentration, dispersion will decrease the extent of reactionas compared to that in a plug flow reactor (34). However, atall flow rates tested, the reaction rate constants determinedusing a plug flow model and those incorporating the effectsof dispersion differed by less than 2%, indicating thatdispersion effects on reactant conversion were negligible (34).Under these conditions, the dechlorination rate may becharacterized by a single parameter, such as the disappear-ance half-life (t1/2), as determined by t1/2 ) 0.693/k1. However,before 90 days elapsed, the observed reaction rate constantdecreased with distance into the column. Therefore, in thesecases (i.e., where the regression coefficient for the first-ordermodel was less than 0.80), the reported t1/2 values have beencalculated from only the influent and effluent (Ce) reactantconcentrations, according to t1/2 ) 0.693θ/ln(Co/Ce), whereθ is the hydraulic residence time. These t1/2 values arerepresentative of the integrated reaction rates along the lengthof the column and will hereafter be referred to as effectivehalf-lives. For all flow rates tested between 0.04 and 2.0 mL/min, the first-order reaction rate constants for TCE and PCEwere independent of the fluid velocity. This indicates thathydrodynamic boundary layer mass transfer limitations hada negligible effect on the reported half-life values.

Dissolved and colloidal iron concentrations in the columneffluent were below the detection limit of 1 µM. However,in the absence of iron precipitation, effluent iron concentra-tions should have been more than 3 orders of magnitudegreater, based on rates of halocarbon reduction. This indicatesthat almost all of the corroding iron remained in the column.Despite precipitation of the corrosion products, the initialand final pore volume data in Table 1 show only a 14% lossin column pore volume over the course of the investigation.For consistency in calculating the t1/2 values, all θ valueswere based on the initial pore volume of the column.

Over the first 309 days, reaction rates for both TCE andPCE gradually increased, as shown by the data in Figure 1.The increasing reaction rate with elapsed time cannot beattributed to pH effects since the pH in the column wasapproximately constant after 30 days, as illustrated by thedata in Figure 2. The increase in reaction rates can beattributed to an increase in reactant-accessible iron surfacearea. Because the iron reactants were aggregates cementedtogether with a nonreactive binder, oxidation of the ironresulted in increased intraparticle porosity, as shown by theinitial and final intraparticle pore volume data in Table 1.The increase in particle porosity resulted in an increase inaccessible surface area over the course of the investigation,as shown in Table 1 by the initial and final specific surfacearea measurements.

Ethene was the major reaction product and accountedfor more than 80% of the TCE and PCE that were transformed.

TABLE 2. Column Operating Conditions

days elapsed 0-30 30-309 309-427 427-507 507-1063 1063-1111water type site site site DI + LiCl DI DITCE concn (mM) 0.3 0.3 7.8 0.3 0.3 0.3PCE concn (mM) 0.3 0.015 0.015 0.015HCA concn (mM) 0.01

C ) Co exp[xu - xu2 + 4Dk1

2D ] (5)

VOL. 34, NO. 12, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2551

Ethane and acetylene were the next most abundant byprod-ucts, and the sum of all chlorinated products accounted forless than 4% of the total TCE and PCE disappearance. Thesereaction products suggest that the â-elimination pathwayproposed by Roberts et al. (35, 36) was the primary dechlo-rination pathway. Trace amounts of vinyl chloride and thethree dichloroethylene isomers suggest that dechlorinationalso occurred via the sequential hydrogenolysis pathway (35,36).

The data in Figure 1 show that reaction rates for TCEwere 2-3 times faster than those for PCE. This is surprisinggiven that data from several batch and column studiesindicate that PCE transformation rates are typically 5 timesfaster than those for TCE (11, 14). In a recent papersummarizing results from a number of laboratory investiga-tions, Tratnyek et al. (14) reported representative half-lifevalues for TCE and PCE of 110 and 20 min, respectively, forsystems with 1 m2 of iron surface area/mL of solution volume.Using the initially measured iron-specific surface area of 3.7m2/g, the reaction rates observed at 309 days correspond tohalf-life values of >10 000 and >22 000 min, respectively, forTCE and PCE when normalized to a 1 m2/mL surface areato solution volume ratio. Although some of the measuredsurface area may be due to the nonreactive binder, thenormalized reaction rates observed in this study are morethan 2 orders of magnitude slower than most previouslyreported values.

Iron Surface Passivation. Because the iron aggregateswere intended for use in fractured bedrock containing bulkphase halocarbons, it was important to determine if reaction

rates would remain first order at TCE concentrations nearaqueous saturation. To answer this question, the TCEconcentration was increased from 0.3 to 7.8 mM (∼92% ofaqueous saturation) between 309 and 427 days. As shown bythe nonlinear concentration profiles for TCE in Figure 3 at408 and 426 days, reaction rates were no longer first orderin TCE concentration. This type of deviation from first-orderreaction kinetics is often attributed to saturation of reactivesites and has been described using a combined zero- andfirst-order kinetic model (11, 37). The combined model isgiven by (37)

where k0 is the zero-order rate constant, and k1 is thepreviously defined first-order rate constant. This model wasfit to the data at 408 and 426 days using the Solver functionin Excel (38) to optimize the value for k0 by minimizing thesum of the squared errors between the model and the data.Comparison of the model to the data in Figure 3 indicatesthat the combined model adequately describes the TCEconcentration profiles for an influent TCE concentration of7.8 mM. However, as described below, this model does notmechanistically describe halocarbon reduction kinetics and,therefore, is not useful for extrapolation to other conditions.For example, the combined model predicts an effluent TCEconcentration of 5.4 × 10-5 µM at 507 days, whereas theactual concentration was 2.8 µM.

The combined model in eq 6 assumes that the numberand reactivity of cathodic sites on the surface of the iron isfixed (37). However, the number of reactive sites for TCEreduction may actually decrease with increasing concentra-tion. This can be attributed to the effect of Ecorr on the rateof halocarbon reduction and to passivation of the ironsurfaces. According to the Bultler-Volmer equation, the rateconstant for TCE reduction by corroding iron should bedependent on Ecorr according to (19, 29)

where k is the standard rate constant, which is dependenton only chemical factors; F is the Faraday constant; R is thegas constant; T is the temperature; E°′ is the formal potentialfor the redox reaction; and Rb and Ra are the transfer coefficientsfor the reduction and oxidation reactions, respectively. Thetransfer coefficients depend on the number of electrons

FIGURE 1. t1/2 values with 95% confidence intervals for TCE andPCE as a function of elapsed time of operation. At 280 days, the flowrate was increased from 0.075 to 0.15 mL/min. The increased t1/2 forPCE at 302 days indicates that more than 7 pore vol was requiredto resume pseudo-steady-state operation after the increase in flowrate. The faster equilibration of TCE as compared to PCE can beattributed to its faster reaction rate and smaller adsorption coefficient(Kd).

FIGURE 2. Column pH profiles at different elapsed times of operation.

FIGURE 3. Concentration profiles for TCE at an influent concentrationof 0.3 mM and a flow rate of 0.075 mL/min (276 and 280 days) ascompared with profiles at a flow rate of 0.15 mL/min (302 and 309days). Also shown are profiles for an influent TCE concentrationof 7.8 mM and flow rates of 0.060 mL/min (408 days) and 0.068 mL/min (426 days). For TCE Co ) 0.3 mM, the data were fit to a first-ordermodel (eq 5, k1 ) 0.0014 min-1); for Co ) 7.8 mM, the data were fitto the combined model (eq 6, k0 ) 0.0016 mM/min).

dCdt

) -k0C

k0/k1 + C(6)

k1 ) k[e-RbF(Ecorr-E°′)/RT - eRaF(Ecorr-E°′)/RT] (7)

2552 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 12, 2000

transferred before and after the rate-determining step andwhether the rate-determining step is limited by chemical- orpotential-dependent factors (19). Under the conditions ofthis study, the E°′ for TCE reduction to chloroacetylene is∼420 mV (SHE) (35). Therefore, the second bracketed termin eq 7, which represents the rate of the reverse oxidationreaction, will be negligibly small as compared to the firstterm. Therefore, the effect of Ecorr on the TCE reduction rateconstant can be described by

As shown in Figure 4, the Ecorr of the iron increases withincreasing TCE concentration. Therefore, if the rate of TCEreduction is limited by the rate of electron transfer (i.e., Rb *0), the rate constant for TCE reduction should decrease withincreasing concentration. This effect may explain part of thedeviation from first-order reaction kinetics observed at 408and 426 days and may also explain the decrease in k0 valueswith increasing TCE concentration in a previously reportedstudy (37).

Passivation of the iron surfaces also contributed to thedeviation from first-order reaction kinetics at 408 and 426days. Two lines of evidence confirm that the iron surfacesbecame passivated by exposure to the 7.8 mM TCE con-centration. Upon lowering the influent TCE concentrationback to 0.3 mM, the t1/2 for TCE did not recover to its previousvalue at 309 days, as indicated in Table 3. This indicates thatthe reactivity of the iron surfaces declined as a result ofexposure to the 7.8 mM TCE concentration. However, thispassivation was slowly reversible, and both the TCE and PCEreaction rates gradually recovered to their values beforepassivation, as indicated by t1/2 data in Table 3 at 592 days.

Corrosion current measurements for an electrode con-structed from the iron aggregates also indicated that the ironsurfaces became passivated by exposure to high TCEconcentrations. As shown in Figure 4, the corrosion rate of

the electrode was measured as a function of the aqueousTCE concentration in a 10 mM CaSO4 electrolyte solution.Up to a concentration of 6 mM, the iron corrosion rateincreases in an approximately linear fashion with TCEconcentration. However, TCE concentrations greater than 6mM increase the corrosion current above the passivationthreshold and lead to a decreased rate of iron corrosion. Infact, the iron corrosion rate in the 8 mM solution was lessthan that in the blank electrolyte solution. This indicatesthat the number of reactive sites for dechlorination actuallydeclines for TCE concentrations above 6 mM. The data inFigure 4 also show that the Ecorr of the iron is not a goodindicator of the degree of passivation, since it increasesmonotonically with TCE concentration, despite the decreasein corrosion current.

The corrosion current data indicate that oxidation by TCEcontributes only minimally to the overall rate of iron aggregatecorrosion. The small effect of TCE on the corrosion rate maybe attributed to the fact that corrosion of the iron aggregateswas anodically controlled. Tafel polarization diagrams forthe iron aggregate electrode in solutions with TCE and PCEeach at 50% of aqueous saturation are shown in Figure 5.The anodic Tafel slopes (âa) of 0.016 decade/mV as comparedto cathodic slopes (âc) of 0.032 decade/mV indicate thatcorrosion of the aggregates is anodically controlled (30).Polarization experiments in blank 10 mM CaSO4 solutionswithout added halocarbons indicate that corrosion of theaggregates by water alone is also anodically controlled. Anodiccontrol of iron corrosion can be attributed to adherentcorrosion products on the surfaces of the iron. Thesecorrosion products reduce the rate at which Fe2+ producedat anodic sites on the iron surface may enter the solution.This results in concentration polarization at anodic sites andreduces the thermodynamic favorability for further ironoxidation (30).

Microbiological Effects. The increasing reaction rate overthe first 309 days could potentially be attributed to an

TABLE 3. Summary of t1/2 Values and 95% Confidence Intervals for Different Operating Conditions and Elapsed Times

days elapsed operating conditions TCE half-life (min) PCE half-life (min)

309 before passivation/TCE Co ) 0.3 mM 481 ( 36 1076 ( 93426 TCE Co ) 7.8 mM 2528a ( 1817 ndb

507 post-passivation/TCE Co ) 0.3 mM 1528a ( 255 4537 ( 549565 partially recovered from passivation 949a ( 274 1935 ( 534592 fully recovered from passivation 443 ( 70 1031 ( 163886 pre-autoclaving 478 ( 22 685 ( 58933 post-autoclaving 273 ( 24 514 ( 70954 post-methanol disinfection 278 ( 14 500 ( 22

a Effective half-life: t1/2 ) 0.693θ/ln (Co/Ce). b nd, not determined.

FIGURE 4. Corrosion currents for the iron aggregate electrode in10 mM CaSO4 electrolyte solutions containing TCE. Also shown arethe free corrosion potentials of the iron aggregate electrode withrespect to the standard hydrogen electrode (SHE).

k1 ) ke-RbF(Ecorr-E°′)/RT (8)

FIGURE 5. Tafel polarization profiles for the iron aggregate electrodein 10 mM CaSO4 electrolyte solutions containing TCE or PCE presentat 50% of aqueous saturation.

VOL. 34, NO. 12, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2553

increasing contribution over time of microbiological reduc-tive dechlorination. The presence of microbial growth in thecolumn was confirmed by cell counts of influent and effluentwater samples. The site groundwater had cell counts ofapproximately 106/mL, while cell counts in the columneffluent were approximately 2 orders of magnitude greater.However, despite the presence of microbial growth, subse-quent testing described below indicated that there was nosignificant biological reductive dechlorination of either TCEor PCE.

Several changes were made to the column operatingconditions to investigate the contribution of biologicalreductive dechlorination. To eliminate nutrients for microbialactivity, the feedwater was changed from the site groundwaterto deionized water after 427 days. Also, LiCl was added as amicrobial inhibitor between 427 and 507 days. As shown bythe half-life data in Table 3, the effect of adding the LiCl tothe column was inconclusive since the t1/2 values for TCEdecreased at the same rate during periods when LiCl waspresent (between 427 and 507 days) and when LiCl was notpresent (between 507 and 592 days).

To determine whether the recovery in reaction ratesbetween 427 and 592 days was due to gradual recovery of abiological reduction mechanism, the column was removedfrom operation and autoclaved for 60 min at 121 °C and 2bar after 921 days. As indicated by the t1/2 data in Table 3 at886 and 933 days, the TCE and PCE t1/2 values after theautoclaving procedure were faster than those before auto-claving. This increase in reaction rates can be attributed toloss of corrosion products from the iron surfaces due toexposure to oxygen. A previous investigation has shown thatexposure of passivated iron surfaces to air leads to exfoliationof corrosion products from the iron surfaces and a con-comitant reversal of passivation (16). This same phenomenonoccurred in this investigation, and visible rust was observedin the effluent water upon resaturating the column after theautoclaving procedure. However, prior to this time, ironconcentrations in the column effluent were always belowthe detection limit of 1 µM.

After 937 days, one additional test was performed tovalidate the absence of biological reductive dechlorination.Between 937 and 940 days, the column was flushed withpure methanol in order to eliminate any microbial activity.As shown by the t1/2 data in Table 3, the TCE and PCE reactionrates were statistically identical before and after the methanoltreatment. This confirmed that biological reductive dechlo-rination did not significantly contribute to TCE and PCEtransformation.

Mass Transfer and Adsorption Effects. The faster TCEversus PCE transformation observed in this investigationcould potentially be due to greater mass transfer limitationsfor PCE. The independence of the t1/2 values with flow rateindicates that diffusion through the hydrodynamic filmboundary layer at the exterior surface of the iron aggregateswas fast as compared to the rate of reaction. However,because the iron aggregates were covered with a porous oxidelayer, reaction rates at the iron surface may have been limitedby adsorption retarded diffusion through iron corrosionproducts. Although the aqueous diffusion coefficient for PCEis only 10% smaller than that for TCE (39), greater adsorptionof PCE on the oxide pore surfaces could contribute to a greaterdifference in intraoxide diffusion rates. To investigate thispossibility, the adsorbed/aqueous phase distribution coef-ficients (Kd) for TCE and PCE adsorption onto the ovenoxidized aggregates were determined. The Kd of 0.18 mL/gfor PCE confirmed that PCE adsorbed more strongly thanTCE, which had a Kd of 0.15 mL/g. This indicates thatadsorption retarded diffusion through a porous oxide layerwould be slower for PCE as compared to TCE.

To investigate possible mass transfer limitations associ-ated with the oxide film, HCA was added to the feed streamafter 1063 days. The hypothesis behind adding HCA was thatif adsorption retarded diffusion through a porous oxide layerwas limiting reaction rates, the reaction rate for HCA shouldbe slower than that for TCE, since HCA adsorbed morestrongly than TCE, as indicated by its Kd of 1.0 mL/g. However,the measured HCA half-life of less than 23 min, comparedto a TCE half-life of approximately 250 min, indicated thatintraoxide diffusional limitations were not limiting reactionrates in the column.

Direct and Indirect Reduction. Although the fasterreaction rate of TCE versus PCE is not consistent with mostother investigations of halocarbon reduction by zero-valentiron, it is consistent with previously published data forhalocarbon reduction by palladium catalysts and by an iron-palladium bimetallic system (26-28). Since hydrocarbonreduction by palladium catalysts is believed to occur viareaction with atomic hydrogen, the faster reduction of TCEobserved in this investigation suggests that the predominantreaction mechanism may involve indirect reduction byatomic hydrogen. A recent investigation of halocarbonreduction by iron cathodes has reported that the predominantmechanism for TCE reduction is indirect and involves atomichydrogen (40).

For reduction by zero-valent iron, comparisons of am-perometrically measured reaction rates with those measuredanalytically can be used to assess the relative rates of directand indirect reduction. Chronoamperometry (CA) and cor-rosion current measurements can be used to measure ratesof direct halocarbon reduction but cannot detect indirectreduction by atomic hydrogen. For example, by comparingthe current in a blank electrolyte solution to that measuredin a solution containing TCE or PCE, the rate of directhalocarbon reduction can be determined. CA experimentsperformed with the iron aggregate electrode show that therewas greater current going toward direct PCE reduction ascompared to that for TCE. Figure 6 compares CA profiles forthe aggregate electrode in a blank electrolyte solution andin electrolyte solutions containing TCE or PCE, each presentat 50% of aqueous saturation. At early elapsed time beforethe electrode currents became mass transfer limited, thecurrent in the PCE containing solution was greater than thatin the TCE solution. This indicates that direct reduction ofPCE is faster than direct TCE reduction. There is greatercurrent in the PCE solution despite the fact that the PCEconcentration of 0.45 mM is ∼9 times lower than the TCEconcentration of 4 mM.

Corrosion current measurements for the proppant elec-trode also indicate that direct PCE reduction is faster than

FIGURE 6. Chronoamperometric profiles for the iron aggregateelectrode in 10 mM CaSO4 electrolyte solutions without addedhalocarbons and with TCE or PCE present at 50% of aqueoussaturation. The profiles were generated by polarizing the electrodeto -453 mV (SHE).

2554 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 12, 2000

that for TCE. For example, the presence of PCE at aconcentration of 0.91 mM increased the corrosion rate ofthe aggregate electrode by 4.0 µA as compared to the blankelectrolyte. However, as shown in Figure 4, this sameconcentration of TCE increased the corrosion rate of theelectrode by only 0.44 µA. Assuming a six-electron transferfor TCE and an eight-electron transfer for PCE, this suggeststhat the rate constant for direct PCE reduction was 6.8 timesgreater than that for TCE. Thus, the CA and corrosion currentmeasurements indicate that direct PCE reduction by the ironaggregates is faster than direct TCE reduction. However, asshown in the column experiments, the analytically measuredrate constant for TCE is approximately twice as fast as thatfor PCE. This disparity between the current measurementsand analytically measured reaction rates suggests thathalocarbon reduction may also occur indirectly via reactionwith atomic hydrogen adsorbed to the iron.

The Tafel diagrams for the iron proppant electrode inFigure 5 indicate that corrosion of the iron was anodicallycontrolled. Under this condition, the rate of direct halocarbonreduction was limited by the rate at which the iron was ableto release additional electrons. In contrast to the anodicallycontrolled shortage of electrons available for direct halo-carbon reduction, there may be an excess of atomic hydrogenfor indirect reduction since more than 80% of the aggregatecorrosion was due to reduction of water at submillimolarhalocarbon concentrations. Under this scenario, the fasterreaction rate of TCE can be attributed to its faster rate ofindirect reduction compared to that for PCE. Faster rates ofTCE versus PCE reduction by atomic hydrogen in catalyticsystems (26-28) is consistent with this hypothesis.

The fact that most other investigators have observed fasterPCE versus TCE reaction rates may be attributed to the factthat iron corrosion rates were not anodically controlled inthose investigations. In the absence of dissolved oxygen, ironcorrosion by water is normally cathodically controlled (30).Even in solutions containing high concentrations of TCE,corrosion of fresh iron was cathodically controlled, asindicated by the data in Figure 7a,b for corrosion of a freshiron wire in solutions containing TCE. At each aqueous TCEconcentration, the cathodic Tafel slope for the iron wire wassmaller than the anodic Tafel slope, as shown in Figure 7b.Under these conditions, iron corrosion rates were limited bythe rate at which oxidants were able to accept electrons.Thus, the faster PCE versus TCE reaction rates observed inother studies can likely be attributed to cathodically con-trolled corrosion leading to faster rates of direct PCEreduction.

For the fresh iron wire, direct reduction of PCE is fasterthan that for TCE. For example, the data in Figure 7a showthat, at equivalent aqueous concentrations, the corrosioncurrent from the iron wire going toward PCE reduction wasgreater than that going toward reduction of TCE. Assuminga six-electron transfer for TCE and an eight-electron transferfor PCE, the corrosion currents in Figure 7a at concentrationsof 0.91 mM indicate that the rate constant for direct PCEreduction was a factor of 4.3 times greater than that for TCE.However, the chloride balance from the experiment in Figure7a indicates that the total rate constant for PCE reductionby the iron wire was only a factor of 2.8 times greater thanthat for TCE. This indicates that there is also faster indirectreduction of TCE by the iron wire. These results suggest thatthe relative rates of TCE and PCE reduction by zero-valentiron depend on the relative contributions of the direct andindirect reduction pathways.

Data from this study does show that zero-valent ironsystems are capable of maintaining their reactivity over along period of operation. However, high halocarbon con-centrations may contribute to iron surface passivation. At

submillimolar concentrations, corrosion current measure-ments for both 99.9% pure iron and the iron aggregatesindicate that oxidation by TCE contributes only minimallyto the overall rate of corrosion. Therefore, the iron con-sumption rate in reactive barrier remedial systems may beonly weakly dependent on the influent halocarbon concen-tration.

The finding of anodically controlled corrosion of agediron surfaces is not unique to the iron aggregates used in thisinvestigation. In a previously reported study, corrosion ofMasters Builders’ Supply iron and 99.9% pure iron coatedwith magnetite was also anodically controlled (16). The fasterreduction of TCE as compared to PCE in anodically controlledsystems is consistent with results from a long-term field testof a zero-valent reactive barrier (41). In that study, reactionrates for TCE were approximately 2 times faster than thosefor PCE after 5 years. This suggests that after a long periodof operation reaction rates become anodically controlled,even at low halocarbon concentrations. Therefore, reactionrates measured with fresh iron particles may not yield realisticlong-term rates of halocarbon reduction.

AcknowledgmentsThe authors thank Don Marcus at Emcon, Rich Helferich atCercona of America, and undergraduate research assistantsMichael Thomas and David Harman. This work was sup-ported by the United States Environmental Protection Agency(U.S. EPA) Office of Research and Development and EmconAssociates. Although the research described in this paperhas been funded in part by the U.S. EPA through grantR-825223-01-0 to J.F., it has not been subjected to the Agency’srequired peer and policy review and therefore does notnecessarily reflect the views of the Agency, and no officialendorsement should be inferred.

FIGURE 7. (a) Corrosion currents for the iron wire electrode in 10mM CaSO4 electrolyte solutions containing TCE or PCE; (b) anodicand cathodic Tafel slopes for the iron wire electrode in 10 mMCaSO4 electrolyte solutions containing TCE.

VOL. 34, NO. 12, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2555

Literature Cited(1) Matheson, L. J.; Tratnyek, P. G. Environ. Sci. Technol. 1994, 28,

2045.(2) Gillham, R. W.; O’Hannesin, S. F. Ground Water 1994, 32, 958.(3) Cantrell, K. J.; Kaplan, D. I.; Weitsma, T. W. J. Hazard. Mater.

1995, 42, 201.(4) Agrawal, A.; Tratnyek, P. G. Environ. Sci. Technol. 1996, 30, 153.(5) Burris, D. R.; Campbell, T. J.; Manoranjan, V. S. Environ. Sci.

Technol. 1995, 29, 2850.(6) Bostick, W. D.; Shoemaker, J. L.; Osborne, P. E.; Evans-Brown,

B. Treatment and Disposal Options for a Heavy Metals WasteContaining Soluble Technetium-99; ACS Symposium Series 422;American Chemical Society: Washington, DC, 1990; pp 345-367.

(7) Gillham, R. W.; O’Hannesin. S. F. Metal-Catalyzed AbioticDegradation of Halogenated Organic Compounds. Presentedat the IAH Conference Modern Trends in Hydrogeology,Hamilton, Ontario, Canada, May 10-13, 1992.

(8) Senzaki, T.; Kumagai, Y. Removal of Organochloro Compoundsin Wastewater by Reductive Treatment: Reductive Degradationof 1,1,2,2-Tetrachloroethane with Iron Powder; NASA TT-21250;NASA STI Program, Scientific and Technical Information; 1988;translation from Japanese, December 1992.

(9) Senzaki, T.; Kumagai, Y. Removal of Organic Chlorine ChemicalCompounds by Use of Some Reduction Processes, ProcessingTrichloroethylene with Iron Powder; NASA TT-21249; NASA STIProgram, Scientific and Technical Information; 1989; translationfrom Japanese, December 1992.

(10) O’Hannesin, S. F.; Gillham, R. W. A Permeable Reaction Wallfor In Situ Degradation of Halogenated Organic Compounds.Presented at the 45th Canadian Geotechnical Conference,Toronto, Ontario, Canada, October 25-28, 1992.

(11) Johnson, T. L.; Scherer, M. M.; Tratnyek, P. G. Environ. Sci.Technol. 1996, 30, 2634.

(12) Schreier, C. G.; Reinhard, M. Chemosphere 1994, 29, 1743.(13) Marcus, D. L.; Farrell, J. Reactant Sand-Fracturing Pilot Test

Results. Presented at the Battelle Conference on Remediationof Chlorinated and Recalcitrant Compounds, Monterey, CA,May 18-21, 1998.

(14) Tratnyek, P. G.; Johnson, T. L.; Scherer, M. M.; Eykholt, G. R.Ground Water Monit. Rev. 1997, Fall, 108.

(15) Su, C.; Puls, R. W. Environ. Sci. Technol. 1999, 33, 163.(16) Farrell, J.; Kason, M.; Melitas, N.; Li, T. Environ. Sci. Technol.

2000, 34, 514.(17) Sivavec, T. M.; Mackenzie, P. D.; Horney, D. P.; Baghel, S. S.

Redox-Active Media for Permeable Reactive Barriers. In Pro-ceedings of the International Containment Technology Confer-ence, St. Petersburg, FL, February 9-12, 1997; pp 753-759.

(18) Brewster, J. H. J. Am. Chem. Soc. 1954, 76, 6361.(19) Bockris, J. O’M.; Reddy, A. K. Modern Electrochemistry, Vol. 2;

Plenum Press: New York, 1970.

(20) Balko, B.; Tratnyek, P. G. J. Phys. Chem. B 1998, 102, 1459.(21) Charlet, L.; Liger, E.; Gerasimo, P. J. J. Environ. Eng. 1998, 124,

25.(22) Scherer, M. M.; Balko, B. B.; Tratnyek, P. G. The Role of Oxides

in Reduction Reactions at the Metal-Water Interface. InMineral-Water Interfacial Reactions: Kinetics and Mechanisms;Sparks, D. L., Grundi, T. J., Eds.; ACS Symposium Series 715;American Chemical Society: Washington, DC, 1998; pp 301-332.

(23) Deng, B.; Burris, D. R.; Campbell, T. J. Environ. Sci. Technol.1999, 33, 2651-2656.

(24) Scherer, M. M.; Balko, B. B.; Gallagher, D. A.; Tratnyek, P. G.Environ. Sci. Technol. 1998, 32, 3026.

(25) Arnold, W. A.; Roberts, A. L. Environ. Sci. Technol. 1998, 32,3017.

(26) Kim, Y. H.; Carraway, E. R. Abstracts of Papers, 217th AmericanChemical Society Meeting, Anaheim, CA; American ChemicalSociety: Washington, DC, 1999; Vol. 39, Abstr. 351.

(27) Lowry, G.; Roberts, P.; McNab, W.; Reinhard, M. Abstracts ofPapers, 213th American Chemical Society Meeting, San Fran-cisco; American Chemical Society: Washington, DC, 1997; Vol.37, Abstr. 171.

(28) Lowry, G. V.; Reinhard, M. Environ. Sci. Technol. 1999, 33, 1905.(29) Bard, A. J.; Faulkner, L. R.; Electrochemical Methods; John Wiley

and Sons: New York, 1980.(30) Uhlig, H. H.; Revie, R. W. Corrosion and Corrosion Control, 3rd

ed.; John Wiley and Sons: New York, 1985.(31) Brunauer, S.; Emmett, P. H.; Teller, E. J. Am. Chem. Soc. 1938,

60, 309.(32) Cox, L. J.; Dooley, D. Beumer, R. Food Microbiol. 1990, 7, 311.(33) Farrell, J.; Grassian, D.; Jones, M. Environ. Sci. Technol. 1999,

33, 1237.(34) Levenspiel, O. Chemical Reaction Engineering, 2nd ed.; John

Wiley and Sons: New York, 1972.(35) Roberts, A. L.; Totten, L. A.; Arnold, W. A.; Burris, D. R.; Campbell,

T. J. Environ Sci. Technol. 1996, 30, 2654.(36) Campbell, T. J.; Burris, D. R.; Roberts, A. L.; Wells, J. R. Environ.

Toxicol. Chem. 1997, 16, 625.(37) Wust, W. F.; Kober, R.; Schlicker, O.; Dahmke, A. Environ. Sci.

Technol. 1999, 33, 4304.(38) Excel, Version 5.0; Microsoft, Inc.: Redmond, WA, 1994.(39) Wilke, C. R.; Chang, P. AIChE J. 1955, 1, 264.(40) Li, T.; Farrell, J. Environ. Sci. Technol. 2000, 34, 173.(41) O’Hannesin, S. F.; Gillham, R. W. Ground Water 1998, 36, 164.

Received for review October 4, 1999. Revised manuscriptreceived March 6, 2000. Accepted March 29, 2000.

ES991135B

2556 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 12, 2000