Journal of Contaminant Hydrology 131 (2012) 54–63

Contents lists available at SciVerse ScienceDirect

Journal of Contaminant Hydrology

j ourna l homepage: www.e lsev ie r .com/ locate / jconhyd

Assessing the Cr(VI) reduction efficiency of a permeable reactive barrierusing Cr isotope measurements and 2D reactive transport modeling

Christoph Wanner a,⁎, Sonja Zink b, Urs Eggenberger a, Urs Mäder a

a Institute of Geological Sciences, University of Bern, Switzerlandb Institute of Mineralogy, Leibniz University of Hannover, Germany

a r t i c l e i n f o

⁎ Corresponding author.E-mail address: wanner@geo.unibe.ch (C. Wanner

0169-7722/$ – see front matter © 2012 Elsevier B.V. Adoi:10.1016/j.jconhyd.2012.01.007

a b s t r a c t

Article history:Received 29 July 2011Received in revised form 7 January 2012Accepted 13 January 2012Available online 27 January 2012

In Thun, Switzerland, a permeable reactive barrier (PRB) for Cr(VI) reduction by gray cast ironwas installed in May 2008. The PRB is composed of a double array of vertical piles containingiron shavings and gravel. The aquifer in Thun is almost saturated with dissolved oxygen andthe groundwater flow velocities are ca. 10–15 m/day. Two years after PRB installation Cr(VI)concentrations still permanently exceed the Swiss threshold value for contaminated sitesdownstream of the barrier at selected localities.Groundwater δ53/52CrSRM979 measurements were used to track Cr(VI) reduction induced by thePRB. δ53/52CrSRM979 values of two samples downstream of the PRB showed a clear fractionationtowards more positive values compared to four samples from the hotspot, which is clear evi-dence of Cr(VI) reduction induced by the PRB. Another downstream sample did not show ashift to more positive δ53/52CrSRM979 values. Because this latter location correlates with thehighest downstream Cr(VI) concentration it is proposed that a part of the Cr(VI) plume isbypassing the barrier. Using a Rayleigh fractionation model a minimum present-day overallCr(VI) reduction efficiency of ca. 15% was estimated. A series of 2D model simulations, includ-ing the fractionation of Cr isotopes, confirm that only a PRB bypass of parts of the Cr(VI) plumecan lead to the observed values. Additionally, the simulations revealed that the proposed bypassoccurs due to an insufficient permeability of the individual PRB piles.It is concluded that with this type of PRB a complete and long-lasting Cr(VI) reduction isextremely difficult to achieve for Cr(VI) contaminations located in nearly oxygen and calciumcarbonate saturated aquifer in a regime of high groundwater velocities. Additional remediationaction would limit the environmental impact and allow to reach target concentrations.

© 2012 Elsevier B.V. All rights reserved.

Keywords:Permeable reactive barrier (PRB)Stable Cr isotopesTracking of Cr(VI) reductionReactive transport modeling

1. Introduction

Chromium occurs in two stable oxidation states in crustalrocks. The highly toxic and very soluble oxidized form Cr(VI)first of all derives from anthropogenic activities such as leathertanning, wood impregnation, galvanization of metal surfacesand cement clinker. Cr(VI) occurs as the chromate oxyanionsCrO4

2−-, HCrO4− and Cr2O7

2− and can lead to health problemssuch as lung cancer and dermatitis (Kotas and Stasicka,2000). Naturally occurring chromium occurs mostly in the re-

).

ll rights reserved.

duced form Cr(III), which is an essential nutrient, less soluble,adsorbs strongly on solid surfaces and co-precipitates withFe(III) hydroxides (Davis and Olsen, 1995; Rai et al., 1987).Exceptions are some rare naturally occurring Cr(VI) mineralssuch as the lead chromate crocoite (PbCrO4) (Frost, 2004)and Cr(VI) bearing groundwater systems naturally occurringin arid regions (Izbicki et al., 2008).

Permeable reactive barriers (PRB) are cost effective, passiveremediation systems and have become a common remediationstrategy for the cleanup of Cr(VI) contaminated sites usingme-tallic iron Fe0 as the reactivematerial (Naftz et al., 2002; Ponderet al., 2000; Powell et al., 1995). In contact with Cr(VI) contam-inated groundwater Fe0 is readily oxidized to Fe2+ or Fe3+

55C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

whereas Cr(VI) is reduced to Cr(III). With iron present Cr(III)precipitates as FexCr1-x(OH)3 (Davis and Olsen, 1995). Becausecontinuous PRB installations did show technical problemsconcerning the maintenance of barrier permeability (Burmeieret al., 2006; Simon et al., 2003) a new design was tested inWillisau, Switzerland (Flury et al., 2009a). This new PRB designconsisted of a double array of vertical piles containing amixtureof gray cast iron shavings and gravel. Pile installation compriseddrilling and excavation activities where subsoil material wasreplaced by the reactive mixture described above. Encouragedby promising results from the Willisau site (Flury et al.,2009a) a PRB of the same design was constructed at anotherCr(VI) contaminated site in Thun, Switzerland in May 2008.

Assessment of Cr(VI) reduction efficiency of a PRB is notstraightforward because simple Cr(VI) concentration mea-surements do not distinguish between decreasing Cr(VI) con-centration due to reduction or due to the effects of dilutionand/or dispersion, respectively. Cr(VI) reduction by itselfcan be tracked by measuring shifts in Cr isotopic ratios(Blowes, 2002) because reduction of Cr(VI) favors the lighterof the four stable Cr isotopes (50Cr, 52Cr, 53Cr and 54Cr) to beaccumulated in reduced form. Accordingly, if a Cr(VI) load isreduced to Cr(III) remnant Cr(VI) becomes successivelyheavier (Ellis et al., 2002). Fractionation of Cr isotopes canbe described with a Rayleigh-type fractionation modelbecause produced Cr(III) is removed from a Cr(VI) solutionby precipitation and adsorption. Furthermore, Cr(III) spe-cies do not undergo rapid isotopic exchange with Cr(VI)species (Zink et al., 2010). Field scale Cr(VI) reduction hasbeen tracked for several natural systems applying Cr iso-tope measurements (Berna et al., 2010; Ellis et al., 2002;Izbicki et al., 2008; Wanner et al., 2011b). To our knowledgeno study has yet tracked Cr(VI) reduction efficiency of aPRB by measuring Cr isotopes. Gibson et al. (2011) recentlydescribed and modeled sulfur isotope fractionation occur-ring in a PRB system where reduced conditions for minewaste water treatment were obtained by adding organiccarbon.

In this study we present Cr isotope measurements fromthe PRB site in Thun that were used to identify Cr(VI) reduc-tion and to estimate its extent. In addition to Cr(VI) concen-tration and Cr isotope measurements the Cr(VI) reductionefficiency of the PRB was assessed by performing a series ofreactive transport model simulations. Reactive transportmodeling has become an essential tool in many earth sciencedisciplines and is used for unraveling complex interactionsbetween coupled rock–water interaction processes or as aforecasting tool (Steefel et al., 2005). Several studies includemodeling of Cr(VI) treatment with PRBs (Jeen et al., 2007b;Mayer et al., 2001; Wanner et al., 2011a). The simulationsdescribed in this paper were performed using the computercode CrunchFlow (Steefel, 2009) and the focus was on testinghypotheses explaining the observed insufficient PRB Cr(VI) re-duction by including the simulation of Cr isotope fractionation.

2. Site description

Thun is located 30 km south of Bern, the capital of Swit-zerland. Recent investigations revealed the mineralogicalcomposition and geological setup as well as the history ofthe industrial site in detail (GIG consulting company,

personal communication): The subsoil is contaminated withheavy metals and chlorinated hydrocarbons due to metalprocessing activities of the former company “MetallwerkeSelve & Co”, active between 1895 and 1990. The contamina-tion covers an area of approximately 10,000 m2, but legalgroundwater threshold values are only exceeded for Cr(VI)concentrations. The Cr(VI) contamination has its origin inacid cleaning of copper dominated metal surfaces (e.g., cop-per wire) with chromic acid (H2CrO4). From former em-ployees it is known that chromic acid residues wereperiodically disposed into the subsurface. The gravel aquiferwas cemented by the interaction with these residues. Thiscemented rock (“industrial rock”) now forms the contamina-tion hotspot. It consists of silicate and carbonate aquifer grav-el that is cemented by a fine-grained greenish matrixcontaining ca. 50 wt.% copper, 2 wt.% chromium and 1 wt%zinc. Malachite (Cu(CO3)(OH)2) was identified as the domi-nant Cu bearingmineral by X-ray powder diffraction analysis.None of the known Cr(VI) bearing minerals were identified.Deionized water extraction of pieces of the “industrial rock”resulted in a Cr(VI) concentration of 8 mg/L confirming thatchromium occurs to a significant part as Cr(VI). The sourcezone covers an area of ca. 30 m2, reaches a maximum depthof 7 m and is inducing a groundwater Cr(VI) plume with amaximum measured groundwater Cr(VI) concentration of4 mg/L (Fig. 1). The plume is contained in a carbonate domi-nated gravel aquifer where the average groundwater table is2–4 m below surface and reaching a depth of 30–50 m. Fine-grained sediment lenses (silty sands) are rare and thehydraulic conductivity in the top 20 m is on the order of10−2 m/s. Groundwater pH values are buffered by the pres-ence of carbonate minerals such as calcite and dolomiteresulting in values close to 8. Average linear groundwatervelocities can reach up to 15 m per day. The hydrological sit-uation is strongly influenced by the river Aare, which isflowing by less than 100 m north of the contamination hot-spot and is recharging the aquifer. The proximity of theriver leads to a groundwater composition that is almost sat-urated with respect to oxygen and that varies seasonally intemperature.

Remediation by excavation of the contamination hotspotwas hindered by the fact that it is located beneath the edgeof an existing building, called Halle 6, which is of historicalvalue and protected. An excavation without destruction ofthe building was not possible. Therefore a PRB was con-structed in May 2008 approximately 30 m downstreamfrom the contamination hotspot (Fig. 1). The barrier designwas the same as used in Willisau, Switzerland, reported pre-viously (Flury et al., 2009a, 2009b). Both PRBs are composedof a double array of vertical piles (Fig. 1, insert) containingiron shavings and gravel with an iron to gravel weight ratioof 3:1. The PRB in Thun consists of 62 individual piles of1.3 m in diameter and up to 13 m in depth, containing atotal of 352 tons of iron. The iron shavings consist of graycast iron and are by-products of cast iron shaping processes.The type of shavings used was selected based on previousbatch and column experiments (Wanner et al., 2011a). Thedouble array of piles setup was chosen based on transportsimulations revealing that the risk of a barrier bypass is lesspronounced when compared to a continuous PRB (Köhler,2003).

Fig. 1. Overview of the Thun site corresponding to the situation before PRB installation (GIG, personal communication): The Cr(VI) plume is illustrated withCr(VI) concentration contour lines (1.0, 0.5, 0.1 and 0.01 mg/L). Black arrows show the corresponding groundwater flow direction. Also illustrated are thelocations of observation boreholes (dots) and the one of the PRB. The insert illustrates the general setup of the PRB providing the detailed geometry of 13 outof 62 piles. The dashed area corresponds to the domain for which the 2D model simulations were performed.

56 C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

3. Field study: methods

Groundwater is regularly sampled by a geological consult-ing company and Cr(VI) concentrations are measured by acommercial laboratory. For measuring Cr isotopes additionalgroundwater samples were collected from wells within thegroundwater Cr(VI) plume using a GRUNDFOS MP1 pumpin July 2009, 13 month after PRB installation. AqueousCr(VI) concentrations were measured instantaneously withthe DPC method using a Merck Pharo 100 photometer(US-EPA, 1995). Four groundwater samples originated fromthe contamination hotspot and three samples form down-stream of the PRB were collected. Depending on the ground-water Cr(VI) concentration up to 0.75 L per sample wascollected. It was assured that the minimum amount of chro-miumwas 2.5 μg leading to individual chromium isotope con-centrations sufficient for analysis. After collecting, sampleswere filtered at 0.45 μm, acidified with 2 mL of a 65% HNO3

solution and filled in polyethylene bottles. Samples werestored at 4 °C for 4 weeks prior to measuring Cr isotopes.

The sample preparation and the measurement of Cr isotopeswas performed at the Institute of Mineralogy at the LeibnizUniversity in Hannover with a Thermo Finnigan Neptune MC-ICPmass spectrometer using a double spike technique describedin detail by Schoenberg et al. (2008) and Zink et al. (2010). Inshort, groundwater samples were evaporated in open PFA vialson a hot plate at 120 °C in order to get high CrTotal concentrationsformeasurements. Sampleswere taken up in dilute hydrochloricacid, where Cr mostly occurs as Cr(III). Cr(III) was then oxidizedto Cr(VI) using a 0.02 mol L−1 (NH4)2S2O8 solution. Prior tomeasurement, the anion exchange chromatography techniquedescribed by Frei and Rosing (2005) was performed to separatechromate ions from other dissolved species.

Isotopic data are reported as 53Cr/52Cr ratio in terms ofthe δ-notation relative to the certified Cr isotope standardNIST SRM 979 and are given in per mill

δ53=52CrSRM979¼53Cr=52Cr

� �sample

53Cr=52Cr� �

SRM979

0@

1A�1 ð1Þ

where (53Cr/52Cr)sample is the measured 53Cr/52Cr ratio of thesample and (53Cr/ 52Cr)SRM979 is the known 53Cr/52Cr ratio(=0.11339) of the certified standard. The δ-notation allowsreporting Cr isotopic composition with respect to a referencematerial. Positive values mean that a sample is isotopicallyheavier than the standard material and negative values areobtained for isotopically lighter samples. The analyticaluncertainties of Cr isotope measurements performed by thelaboratory in Hannover were investigated extensively bymeasuring a series of NIST SRM 3112a standard samples aswell as replicates of rock samples and resulted in a value ofless than±0.05‰ (Schoenberg et al., 2008).

4. Field study: results and discussion

4.1. Groundwater Cr(VI) concentration

Evolution of Cr(VI) concentrations since the PRB installa-tion in May 2008 are illustrated in Fig. 2 for five differentobservation boreholes. One of the two boreholes just down-stream of the PRB (KB08/02) showed Cr(VI) concentrationsbelow detection limit (0.003 mg/L) for the entire time period.Cr(VI) concentrations measured on samples originating fromthe second borehole located downgradient and close to thePRB (KB08/03) resulted in concentrations that stronglyexceed the Swiss threshold value, which is 0.01 mg/L (SwissConfederation, 1998). Interestingly, at KB08/03 the Cr(VI)concentration strongly increased immediately after the in-stallation. 100 days after PRB installation the concentrationwas more than 10 times higher than the one measured2 weeks before installation. No data exist for the pre PRBtime because KB08/02 and KB08/03 were only drilled justshortly before PRB installation. Cr(VI) concentrations origi-nating from boreholes further downstream (KB01/05 andKB07/26) also periodically exceeded the Swiss thresholdvalue. Compared to the pre-PRB time no significant changein Cr(VI) concentration was observed for these latter twoboreholes. Cr(VI) concentration measurements of samplescollected at a dozen other locations resulted in Cr(VI)concentrations that were below detection.

Fig. 2. Evolution of Cr(VI) concentrations since PRB installation in May 2008for 5 observation boreholes located in the Cr(VI) hotspot (KB06/17) anddownstream of the PRB (KB08/02, KB08/03, KB09/06 and KB07/26). Notethe logarithmic y-axis.

57C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

4.2. δ53/52CrSRM979 measurements

Spatial distribution of measured δ53/52CrSRM979 values andCr(VI) concentrations are illustrated in Fig. 3. δ53/52CrSRM979

values determined for the four hotspot samples were withina range of 0.63–1.14‰ indicating an isotopically heteroge-neous Cr(VI) source. Interestingly, the values were muchheavier than the δ53/52CrSRM979 values of naturally occurringchromium and other reported anthropogenic Cr(VI) sources,which are all close to 0‰ (Berna et al., 2010; Ellis et al., 2002;Schoenberg et al., 2008). The observation may be explainedby a partial Cr(VI) reduction during spillage of the chromicacid residues leading to a fractionation of δ53/52CrSRM979

values towards positive values. The exact mechanismsfor such a reduction, however, remains unclear because

Fig. 3. Spatial distribution of measured Cr(VI) concentrations (±0.005 mg/L) andexplained by a partial PRB bypass either occurring at the southeast edge (flow pathby Cr(VI) reduction of another part of the Cr(VI) plume (flow path c). Cr(VI) loads tand in the vicinity of boreholes KB01/05 and KB07/26.

the aquifer does not contain significant amounts of knownCr(VI) reducing agents.

From the three downstream samples only the samplescollected from KB01/05 and KB07/26 (δ53/52CrSRM979=1.42and 1.66‰) showed a clear fractionation towards morepositive values compared to the average δ53/52CrSRM979

value of the four Cr(VI) hotspot samples (0.87‰±0.26‰,0.26‰=standard deviation). This observation can either beexplained by a partial reduction of the Cr(VI) plume inducedby the PRB or by mixing of the plume with another Cr(VI) loaddownstream of the main Cr(VI)-hotspot having a δ53/52CrSRM979

value >0.87‰. Since such a secondary plume can be excluded(GIG, personal communication), these measurements con-firmed that at least a part of the Cr(VI) plume is reducedwithinthe PRB even though Cr(VI) concentration measurements atthese two boreholes did not show a clear indication for Cr(VI)reduction. The sample originating from KB08/03 resulted ina δ53/52CrSRM979 value of 0.92‰ that is almost equal to theaverage of the hotspot samples. Interestingly, the absence offractionation at KB08/03 correlateswith high Cr(VI) concentra-tions (Fig. 2). Even though KB08/03 is located immediatelydownstream of the PRB (Fig. 1) this correlation provides evi-dence that most of the Cr(VI) load at KB08/03 was not affectedby the PRB. A possible explanation is that this Cr(VI) loadbypassed the PRB in between the individual piles (Fig. 3, flowpath b). However, the absence of detectable Cr(VI) at boreholeKB08/02 (Fig. 2) reveals that such a bypass does not occuralong the entire PRB. It is more likely that aquifer heterogene-ities induced a bypass in between piles in the vicinity of bore-hole KB08/03 only. Alternatively, it can be proposed that theCr(VI) load sampled at KB08/03 was mostly bypassing thePRB at the southeast edge on its transport path from the hot-spot towards this observation borehole (Fig. 3, flow path a).For both bypass scenarios, Cr isotope fractionation observedat KB01/05 and KB07/26 is the result of mixing a Cr(VI) loadthat was partially reduced during its transport through thePRB and a Cr(VI) load that was bypassing the PRB and wasnot reduced at all (Fig. 3).

δ53/52CrSRM979 values of samples taken in July 2009. The measurements area) or in between piles (flow path b) and a PRB flow-through accompanied

hat bypassed the PRB mix with (partially) treated Cr(VI) loads downgradient

58 C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

For the mixing scenarios described above it is not possibleto calculate the exact extent of Cr(VI) reduction using a sim-ple closed system Rayleigh fractionation model. One issueoccurs because high δ53/52CrSRM979 values induced by Cr(VI)reduction are masked by mixing with highly concentratedCr(VI) loads that did not undergo any Cr(VI) reduction.Furthermore, a complete Cr(VI) reduction possibly occurringalong specific flow paths does not induce any Cr isotopefractionation downstream of the PRB. Following this argu-mentation, the extent of Cr(VI) reduction of the Cr(VI)loads observed at KB01/05 and KB07/26 would be underesti-mated if a closed system Rayleigh-type fractionation modelwould be applied. Nevertheless, we used such a model tocalculate the minimum extent of Cr(VI) reduction inducedby the entire PRB system assuming that the Cr(VI) loadsobserved at KB01/05 and KB07/26 represents a mixture ofthe proposed flow paths (Fig. 3). The calculation of minimumdegrees of Cr(VI) reduction bears the advantage of beingmore robust as a conservative estimate when dealing withthe authorities (Grandel and Dahmke, 2008). The calculationwas performed using Eq. (2)

δ53=52CrSRM979 ¼�

δ53=52CrSRM979INI þ 1000� �

f ða−1Þ�−1000

ð2Þwhere δ53/52CrSRM979INI and δ53/52CrSRM979 refer to the aver-age of the upstream δ53/52CrSRM979 values (0.87±0.26‰)and δ53/52CrSRM979 values of samples from KB07/26 (1.66‰)and KB01/05 (1.42‰), respectively,α refers to the isotopic frac-tionation factor, f is the remaining Cr(VI) fraction and f−1represents the reduced Cr(VI) fraction, accordingly. Reportedlaboratory generated, intrinsic α for Cr(VI) reduction are inthe range of 0.9950–0.9976. Berna et al. (2010) observedisotopic fractionation factors between 0.9969 and 0.9976 forCr(VI) reduction occurring in a batch experiment by using sub-soil material containing green rust phases. Sikora et al. (2008)reported α between 0.9955 and 0.9960 for a Cr(VI) reductionmicrobially stimulated by Shewanella oneidensis. Ellis et al.(2002) performed reduction experiments with magnetiteslurry and fine-grained sediments with total organic carboncontents of 0.3–0.9 wt.%, resulting in an average α of 0.9966.Finally, Zink et al. (2010) reported α=0.995 for Cr(VI) reduc-tion induced by H2O2. For our calculation αwas set to themax-imum published value for naturally occurring abiotic Cr(VI)reducing agents, which is 0.9966 (Ellis et al., 2002). This selec-tion results in minimum Cr(VI) reduction efficiencies (Eq. 2).Furthermore, it is in accordance with the argumentation ofGrandel and Dahmke (2008). Taking into account the standarddeviation of the average hotspot δ53/52CrSRM979 value, the calcu-lations resulted in a minimum PRB Cr(VI) reduction efficiencyof 14–27% for the load observed at KB7/26 and 8–21% for theone at KB01/05, respectively. The fact that these values arelower than the corresponding relative Cr(VI) concentration var-iations (Fig. 2) illustrates why Cr(VI) concentrations did not re-spond to the PRB installation and emphasize that Cr isotopemeasurements are a much more powerful tool for demonstrat-ing Cr(VI) reduction. Additionally, the low values indicate thatthe decrease in Cr(VI) concentration observed along the flowpath (Fig. 3) is first of all caused by the effects of dilution whencompared to the effects of a PRB induced Cr(VI) reduction.Adsorption of Cr(VI) on ferric hydroxides forms an additional

attenuation process possibly limiting Cr(VI) concentrations.At typical field site pH-values on the order of 8, however, theCrO4

2− anion being the dominant Cr(VI) species (Kotas andStasicka, 2000) shows a minor adsorption tendency only(Dzombak and Morel, 1990). Accordingly, Cr(VI) adsorption isassumed to only play a minor role in controlling the fate ofCr(VI).

The correlation of high Cr(VI) concentrations and the ab-sence of Cr isotope fractionation at KB08/03 as well as theCr(VI) concentration below detection limit at KB08/02(Fig. 2) indicate that any kind of PRB bypass is the primarycause for the observed low Cr(VI) reduction efficiencies atKB07/26 and KB01/05. The bypass was likely induced by so-called skin effects that are known phenomena during drillingactivities (Henebry and Robbins, 2000). Drilling the PRB pilesinto the aquifer has possibly sheared and compacted theinterface between the piles and the aquifer reducing thepermeability of the PRB piles. Accordingly, such skin effectcould have resulted in a limitation of the groundwater flowthrough the entire PRB system even though the permeabilityof the piles is theoretically at least 3 times higher than thesurrounding aquifer (Köhler, 2003). Another possible cause isthe process of internal erosion due to a higher porosity of thepiles compared to the surrounding aquifer. In this case fine-grained aquifer particles were transported into the pore spaceat the interface between PRB and aquifer and could have limit-ed the PRB permeability.

Besides a possible PRB bypass additional factors may ex-plain high Cr(VI) concentrations observed downstream ofthe PRB: (i) it is well known that metallic iron is subject toa strong decline in reactivity due to passivation with iron hy-droxides and carbonates (Flury et al., 2009b; Jeen et al.,2007a; Zhang and Gillham, 2005), which could have inhib-ited a full treatment of the Cr(VI) plume after a short timeperiod. A column experiment performed at the planningstage of the PRB in Thun revealed that the loss in reactivitycould lead to a Cr(VI) breakthrough already after 60 daysof PRB operation (Wanner et al., 2011a). (ii) In wintertimethe minimum groundwater temperature is ca. 5 °C. At thistemperature the iron corrosion rate is much lower thanat the maximum groundwater temperature of ca. 20 °C(Del Campo et al., 2008). The same is true for oxygen solu-bility, which was confirmed by measuring maximum O2

concentrations of 11.6 mg/L during wintertime. Both fac-tors (aqueous O2 conc., low iron corrosion rate) result indecreasing Cr(VI) reduction rates with decreasing temper-atures and this could also contribute to a low Cr(VI) reduc-tion efficiency.

5. Reactive transport modeling

To test the various hypotheses explaining the low PRBCr(VI) reduction efficiency of the PRB four 2D reactive trans-port model simulations were run using the code Crunchflow(Steefel, 2009). The primary purpose of our reactive trans-port modeling work was to quantitatively evaluate the sce-narios proposed in Section 4. The simulations were notintended to serve as accurate predictive simulations of theactual site because most of the hydrological constraints arenot very well known.

59C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

5.1. Model setup

The PRB system was modeled for three out of four simula-tions (see Sections 5.3.1–5.3.3) using a 2D domain with anextent of 29.1×43.3 m (Fig. 1). Table 1 gives an overview ofthe model input parameters. Groundwater flow was assumedto be at steady state with a Darcy flux of 7.5 m/day in west-ward direction, which is in accordance to the general hydro-logical conceptual site model (Fig. 1). A Darcy flux of 7.5 m/day corresponds to the maximum expected average lineargroundwater velocity within the PRB of ca. 15 m/day givena porosity of the PRB piles of ca. 50%. The model was dividedinto 3 domains defining the Cr(VI) source area, the unconta-minated aquifer and a PRB treatment zone (Fig. 4). Forthe simulations 10203 grid cells with cell dimensions of0.1625×1 m outside the PRB domain and 0.1625×0.2167 mwithin the PRB domain were defined. The mineralogicalcomposition of the aquifer domain was specified as 45 vol%calcite and 22.5 vol% quartz corresponding to XRD analysisperformed by Wanner et al. (2011a). The composition ofthe individual PRB piles was defined as 48.6 vol% quartz and5.4 vol% Fe0 simulating the mixture that was filled into thepiles. The Cr(VI) contamination was specified within an

Table 1Model input parameters.

Parameter Unit Aquiferdomain

Barrierdomain

Cr(VI)hotspot

Darcy fluxa m/day 7.5 7.5 7.5porosity – 0.33 0.46 1longitudinaldispersivity

m 0.3 0.3 0.3

Flow parameters transversedispersivity

m 0.03 0.03 0.03

intrinsicpermeabilityb

m2 2.3·10-9 1.15·10-9 2.3·10-9

Pressuregradient b

Pa/m 45

quartz vol % 22.5 48.6 0Mineralogy calcite vol % 45 0 0

Fe0 vol % 0 5.4 0

pH – 8.0 8.2 8.1temperature °C 25 or 5 25 or 5 25 or 5Na+ mg/L 4.6 1.6 1.6K+ mg/L 1.1 2.0 11.0Mg2+ mg/L 5.0 4.8 5.2Ca2+ mg/L 49.8 44.6 51.4

Initialconcentrations

Cl− mg/L 1.3 3.3 2.0SO4

2− mg/L 45.1 38.5 45.7NO3

− mg/L 2.3 b0.01 2.7HCO3

− mg/L 126.9 114.7 128.9Cr(VI) mg/L b0.002 b0.002 2.0O2(aq) mg/L 7.5c 2.0 7.5c

Fe2+ mg/L b0.01 2.0 b0.01

a steady state, fixed value for simulations presented in Sections 5.3.1–5.3.3.Using Crunchflow it is not required to specify intrinsic permeability and initialpressure heads for constant velocity problems.

b permeability and pressure gradient along model domain (Δx=7.5 m)only specified for simulation presented in Section 5.3.4. For other simula-tions this value was not used because a steady state flow field was specified(see a).

c For the simulation presented in Section 5.3.2. O2 concentration was setto 11.6 mg/L corresponding to the maximum value measured duringwintertime.

Fig. 4. Results of model run performed to simulate the theoretical PRB Cr(VI)breakthrough using the reaction kinetics obtained from column experiments(Wanner et al., 2011a): a illustrates the model setup in terms of porosityshowing the Cr(VI) source area in red, the PRB domain in yellow and theaquifer domain in blue. Obtained temporal Cr(VI) evolution are shown for100 days (b), 150 days (c) and 200 days (d).

area of 5.2×1 m as an unlimited amount of a Cr(VI) bearingsolution with a fixed Cr(VI) concentration of 2 mg/L corre-sponding to the average of the observed hotspot groundwa-ter Cr(VI) concentrations (KB06/16, KB06/17 and KB07/22).Initial groundwater concentrations of the aquifer and PRB do-mains were specified according to major cations and anionsconcentrations reported by Wanner et al. (2011a). Longitudi-nal dispersivity was set to 0.3 m according to the results of atracer test performed in a column experiment (Wanner et al.,2011a). Note that the true value for the site is possibly muchlarger because the dispersivity is known to be a scale depen-dent parameter (Appelo and Postma, 2005). However, thevalue derived from the column experiment was the onlyknown value and the purpose of the model was not to oper-ate as a prediction tool. Transverse dispersivity was set to0.03 m in order to obtain enough dispersive dilution to simu-late the expected Cr(VI) concentrations at the PRB inflow ofca. 1 mg/L (Fig. 1).

For the fourth simulation (see Section 5.3.4.) slightly dif-ferent model input parameters were used. The simulationwas only performed for a small part of the PRB (2.8×7.5 m)

60 C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

containing one and a half piles. This model was divided into476 cells with cell dimensions of 0.1625×0.2167 m upstreamand within the PRB and 0.1625×0.5 m downstream of thePRB. Due to the small model domain, Cr(VI) concentrationof the Cr(VI) source was limited to 1 mg/L accounting for di-lution occurring between the true Cr(VI) source and thismodel domain. Furthermore, the flow field was not fixedto a constant velocity but it was calculated by specifyinga pressure difference of 338.9 Pa along the model domain(Δx=7.5 m) together with intrinsic permeability values forthe entire model domain. Specifying a pressure gradient of338.9/7.5=45 Pa/m correspond to a hydraulic head gradientof 4.6‰. The permeability of the PRB piles was set to1.15×10−9 m2 or half the value of the permeability of thesurrounding aquifer, which was defined as 2.3×10-9 m2. Byspecifying pressure gradients and intrinsic permeabilityvalues CrunchFlow calculates the flow field according to Dar-cy's Law

v ¼ − kμ ⋅

∂P∂x ð3Þ

where k (m2) refers to the intrinsic aquifer permeability, μ(Pa⋅s) is the dynamic fluid viscosity, and ∂P/∂x (Pa/m) refersto the fluid pressure gradient along the flow path. Remainingmodel input parameters were defined as for the other threesimulations (Table 1).

5.2. Reaction network

Reaction networks for reactive transport modeling of Fe0

systems do have to account for the fact that Fe0 reactivity isusually declining with time (Jeen et al., 2007a; Vikesland etal., 2002). Jeen et al. (2007b) recently described an approachthat allowed incorporating the decline in Fe0 reactivity intothe reaction network of Mayer et al. (2001). This wasachieved by defining an expression for the reactive Fe0 sur-face area being dependent on the amount of aragonite pre-cipitation causing a surface passivation of the reactive Fe0

surfaces. In contrast to the column experiment simulated byJeen et al. (2007b), Fe0 is used in a fully oxidized groundwa-ter setting at the Thun site. A column experiment performedat the planning stage of the PRB showed that a dramaticchange of redox conditions is occurring along the flow pathacross the PRB (Wanner et al., 2011a). To simulate thischange in redox conditions as well as the observed Cr(VI)breakthrough after 60 days of operation a different approachcompared to the Jeen et al. (2007b) passivation model wasproposed for such conditions (Wanner et al., 2011a). Themain difference was the formulation of variably reactive Fe0

fractions and the specification of kinetic intra-aqueous reac-tions allowing to model iron cycling phenomena. The detaildescription of the reaction network and a full comparisonwith the passivation model of Jeen et al. (2007b) is given inWanner et al. (2011a). Wanner et al. (2011a) also describethe explicit calibration of the proposed reaction networkbased on the results of the column experiment simulatingthe PRB at the planning stage successfully. Accordingly, thekinetic parameters were also used for the 2D reactive trans-port model simulations presented here.

To run the simulations at various temperatures T, reactionrate constants k were calculated based on transition statetheory (Lasaga, 1984) using the Arrhenius relationship

k ¼ k25−EaR

1T− 1

298:15ð Þ½ � ð4Þ

where k25 is the defined reaction rate constants at 25 °C spec-ified by Wanner et al. (2011a), Ea is the reaction's activationenergy and R refers to the ideal gas constant. The activationenergies of the various Fe0 fractions defined in the reactionnetwork were not empirically determined. Therefore Ea wasset to an arbitrary value of 15 kcal/mol for all Fe0 fractions.This value corresponds to the default value used by Crunch-Flow (Steefel, 2009) for reactions with unknown activationenergy.

The 2D model simulations additionally include the simu-lation of Cr isotope fractionation. The fractionation was incor-porated into the reaction network using the methoddescribed by Van Breukelen et al. (2005) to model carbonisotope fractionation. The Cr isotopes 52Cr and 53Cr were de-fined in the CrunchFlow database as individual CrO4

2- specieswith the same log K values as the bulk CrO4

2− species. Addi-tionally reaction rate constants for the reduction of the twoCr(VI) isotope species were defined according to Eqs. (5)and (6)

52k ¼ k � f Cr52 ð5Þ53k ¼ k � f Cr53 � α ð6Þ

where 52k and 53k refer to the reaction rate constants ofCr(VI) reduction reactions for the two Cr(VI) isotope species,k is the corresponding reaction rate constant for the bulkCr(VI) species (Wanner et al., 2011a), fCr52 (=0.83789) andfCr53 (=0.09501) are the bulk Earth fractions of 52Cr and53Cr of total Cr (Shields et al., 1966) and α refers to the isotopicfractionation factor, which was set to 0.9966 according to Elliset al. (2002). The initial δ53/52CrSRM979 of the Cr(VI) sourcewas set to 0.87‰ corresponding to the average value of the 4hotspot samples (Fig. 3).

5.3. Simulated scenarios: results and discussion

5.3.1. Simulation of high Cr(VI) concentrations due to Fe0

reactivity declineThe first model run simulates the timing of the theoretical

Cr(VI) breakthrough assuming that the decrease of Fe0

reactivity occurs as observed in the column experiment(Wanner et al., 2011a). Furthermore, the temperature wasset to that measured during the column experiments(25 °C) in order to fully upscale the experiment. SimulatedCr(VI) concentrations are illustrated in Fig. 4 for varioustimes after PRB installation. The model results reveal thatthe PRB is able to reduce the entire expected Cr(VI) load forapproximately 140 days with the defined model parameters(e.g., flow field, reaction kinetics). The simulated gradualbreakthrough is in contrast to the instantaneous Cr(VI)breakthrough observed at KB08/03 (Fig. 2). The time lagbetween modeled and observed Cr(VI) breakthrough pointsout that the observed breakthrough not only occurred dueto a decline in reactivity of the iron shavings. Hence, the

Fig. 5. Results of model run testing the hypothesis if the observed early Cr(VI) breakthrough (Fig. 2) was due to low groundwater temperatures. a and b illustratespatial Cr(VI) concentrations 150 days after PRB installation if temperatures were constantly at 25 °C (a) or at 5 °C (b), respectively. c shows the temporal evo-lution of the Cr(VI) concentration at borehole KB08/03 showing that low temperatures only has a small effect on the time of Cr(VI) breakthrough.

61C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

simulation is in good agreement with Cr isotope and Cr(VI)concentration measurements (Fig. 3) favoring the hypothesisthat part of the Cr(VI) plume was bypassing the PRB.

5.3.2. Simulation of high Cr(VI) concentrations due to lowgroundwater temperatures

This model run simulates the effect of the observed lowtemperatures during wintertime (ca. 5 °C) on the timing ofthe predicted Cr(VI) breakthrough. Fig. 5 compares the simu-lated Cr(VI) concentrations to the one obtained for the previ-ous simulation (at 25 °C). The simulation resulted in a Cr(VI)breakthrough that occurred approximately 20 days earliercompared to the simulation performed at 25 °C. Accordingly,low groundwater temperatures at wintertime but also aver-age annual temperatures (12 °C) below the ones observedduring the column experiment are not considered as an

Fig. 6. Results of model run simulating the proposed mixing of a Cr(VI) load bypasmodel setup in terms of porosity showing the Cr(VI) source area in red, the PRB dstate Cr(VI) concentrations and δ53/52CrSRM979 values illustrating that Cr(VI) reducttwo Cr(VI) loads are mixed downstream of the PRB leading to decreasing Cr(VI) co

important factor why high Cr(VI) concentrations down-stream of the PRB were observed instantaneously after PRBinstallation (Fig. 2).

5.3.3. Simulation of high Cr(VI) concentrations due to PRBbypass at barrier edge

The third model run simulates the effect of the proposedPRB bypass possibly occurring at the southeast PRB edgeusing quite a large simplification in terms of the hydrologicalflow field. Compared to the first two simulations the setup forthis model run was only differing by the fact that the 13 pileslocated at the barrier edge were removed and the Cr(VI)source was defined for an area of 6.8 m instead of 5.2 m asillustrated in Fig. 6a.

Simulated steady state Cr(VI) concentrations andδ53/52CrSRM979 values (Fig. 6b–c.) showed the same trend as

sing the PRB and a load that is partially reduced by the PRB. a illustrates theomain in yellow and the aquifer domain in blue. b and c present the steadyion is accompanied by a fractionation of Cr isotopes. Also shown is that thencentrations and δ53/52CrSRM979 values.

62 C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

the field observations: Cr(VI) concentration is decreasingwhere the Cr(VI) load is flowing through the PRB. HereCr(VI) reduction is accompanied by a fractionation of Cr iso-topes resulting in more positive δ53/52CrSRM979 values. At theedge of the PRB the bypass of the Cr(VI) load led to highCr(VI) concentrations downstream of the PRB and did notaffect δ53/52CrSRM979 values. The two Cr(VI) loads are mixedfurther downstream due to transverse dispersion leading todecreasing Cr(VI) concentrations and δ53/52CrSRM979 values.The successful simulation of the observed Cr(VI) concentrationand δ53/52CrSRM979 trend lends support for our conceptual PRBbypass model proposing a bypass at the barrier edge (Fig. 3,flow path a).

5.3.4. Simulation of PRB bypass due to low permeable PRB pilesThis model run simulates the effects of the proposed low

permeable individual PRB piles on the groundwater flowfield and on the Cr(VI) concentrations, respectively. Theobtained groundwater flow velocities of model domains notaffected by the PRB were in the range of 16 m/day corre-sponding to the assumed maximum average linear velocitiesof 15 m/day. Obtained steady state Cr(VI) concentrations(Fig. 7b) clearly demonstrate that the Cr(VI) load is not flow-ing through the PRB piles to a great extent if the permeabilityof the piles are reduced by factor of two. It is assumed thatdrilling activities can induce such a relatively small perme-ability decline suggesting that a PRB bypass in betweenpiles is a valid hypothesis (Fig. 3, flow path b). However,due to the large amount of piles present at the field scale

Fig. 7. Results ofmodel run testing the hypothesis if the observed early Cr(VI)breakthrough (Fig. 2) was due to low permeability within piles: a illustratesthe model setup in terms of porosity showing the Cr(VI) source area in redthe PRB domain in green and the aquifer domain in blue. b illustratesobtained steady state spatial Cr(VI) distribution showing that a Cr(VI) break-through occurs by specifying piles that are less permeable (factor 2) than thesurrounding aquifer. The obtained hydrological flow field is qualitativelyillustrated in terms of black arrows showing that groundwater is forced toflow around the piles.

,

(62 piles) it is more likely that the permeability of the entirePRB domain is reduced, which in turn favors the barrier edgebypass hypothesis (Fig. 3, flow path a).

In the overall context, both bypass simulations (Figs. 6and 7) correspond well with the field observations and con-firm that a partial PRB bypass is limiting the overall Cr(VI) re-duction efficiency. The exact bypass mechanism, however,remains unclear unless further investigations are performed.

6. Conclusions

Cr(VI) concentration measurements clearly show that thePRB in Thun is not operating as desired. Additional remedia-tion action is therefore suggested to be performed to limit theenvironmental impact and to reach the required Cr(VI) con-centrations downstream of the PRB. Cr isotope measure-ments allowed to identify that high downstream Cr(VI)concentrations are mainly caused by a barrier bypass. Reac-tive transport model simulations suggested that the bypassis occurring due to a limited permeability of the PRB system.In contrast, column experiments that were performed re-cently revealed that the proper PRB operation is limited bythe decline in iron reactivity. Accordingly, it is inferred thatthe hydrological conditions at the field site may form a limi-tation for the success of engineered remediation systemseven though they were successfully tested at the laboratoryscale. In terms of the PRB setup used at the Thun site, howev-er, it is assumed that even under a perfect flow-through asuccessful long-term operation is very difficult to achievefor fast flow conditions and nearly oxygen and calcium car-bonate saturated aquifer conditions.

The study confirms that Cr isotope measurements are aunique and powerful tool for tracking Cr(VI) reductionwhen compared to traditional Cr(VI) concentration measure-ments. However, it is inferred that for complex hydrologicalsituations (e.g., mixing of different flow paths), the exactamount of Cr(VI) reduction is difficult to estimate unlessthe flow field is fully understood. The study also illustrateshow reactive transport modeling is used as a tool for datainterpretation in terms of a quantitative evaluation of site hy-potheses. As an additional benefit, modeling offers the possi-bility of evaluating future remediation scenarios and theirconsequences with respect to environmental impact andthe remaining contamination over time. Such models do,however, require numerous site-specific parameters and anunderstanding of the governing water–rock interaction pro-cesses and of the flow field. Additional laboratory experi-ments such as column experiments or sequential leachingtechniques allow to calibrate reaction rates for processes. Inthe Thun case, direct constraints on iron reactivity obtainedfrom column experiments helped to tightly constrain theprocesses at the site scale. Modeling is also cost-efficientwhen compared to additional piezometer installations ortrenching, and particularly when costly excavations and sub-sequent on-site or off-site treatment and deposition can beminimized.

Acknowledgments

We thank the geological consulting companies Geotest AGand Schenker Korner & Partner GmbH (operating together as

63C. Wanner et al. / Journal of Contaminant Hydrology 131 (2012) 54–63

GIG Selve-Areal) for giving us access to confidential siteinformation and for their financial support. Constructivediscussions with Franz Schenker, Juergen Abrecht, RonnySchoenberg and Carl Steefel are highly appreciated. The manu-script was improved thanks to comments of two anonymousreviewers. Theprojectwas financially supported by theGermanResearch Foundation (DFG) grant SCHO1071/3-1 to RonnySchoenberg.

References

Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution,2nd edition. Balkema Amsterdam, The Netherlands.

Berna, E.C., Johnson, T.M., Makdisi, R.S., Basui, A., 2010. Cr stable isotopes asindicators of Cr(VI) reduction in groundwater: a detailed time-seriesstudy of a point-source plume. Environmental Science and Technology44, 1043–1048.

Blowes, D., 2002. Tracking hexavalent Cr in groundwater. Science 295,2024–2025.

Burmeier, H., Birke, V., Ebert, M., Finkel, M., Rosenau, D., Schad, H., 2006.Anwendung von durchströmten Reinigungswänden zur Sanierung vonAltlasten. Universität Lüneburg, Fakultät III (Umwelt und Technik),Suderburg, Germany.

Davis, A., Olsen, R.L., 1995. The geochemistry of chromium migration andremediation in the subsurface. Ground Water 33, 759–768.

Del Campo, L., Perez-Saez, R.B., Tello, M.J., 2008. Iron oxidation kinetics studyby using infrared spectral emissivity measurements below 570 degreesC°. Corrosion Science 50, 194–199.

Dzombak, D.A., Morel, F.M.M., 1990. Surface complexation modeling: hydrousferric oxides. John Wiley & Sons, New York, USA.

Ellis, A.S., Johnson, T.M., Bullen, T.D., 2002. Chromium isotopes and the fateof hexavalent chromium in the environment. Science 295, 2060–2062.

Flury, B., Eggenberger, U., Mäder, U., 2009a. First results of operating andmonitoring an innovative design of a permeable reactive barrier forthe remediation of chromate contaminated groundwater. Applied Geo-chemistry 24, 687–696.

Flury, B., Frommer, J., Eggenberger, U., Mäder, U., Nachtegaal, M., Kretzschmar,R., 2009b. Assessment of long-term performance and chromate reductionmechanisms in a field scale permeable reactive barrier. EnvironmentalScience and Technology 43, 6786–6792.

Frei, R., Rosing, M.T., 2005. Search for traces of the late heavy bombardmenton Earth—results from high precision chromium isotopes. Earth andPlanetary Science Letters 236, 28–40.

Frost, R.L., 2004. Raman microscopy of selected chromate minerals. Journalof Raman Spectroscopy 35, 153–158.

Gibson, B.D.G.B.D., Amos, R.T., Blowes, D.W., 2011. (34)S/(32)S fractionationduring sulfate reduction in groundwater treatment systems: reactive trans-port modeling. Environmental Science and Technology 45, 2863–2870.

Grandel, S., Dahmke, A., 2008. Natürliche Schadstoffminderung bei LCKW-kontaminierten Standorten Methoden, Empfehlungen und Hinweise zurUntersuchung und Beurteilung. Christian-Albrechts-Universität zu Kiel.

Henebry, B.J., Robbins, G.A., 2000. Reducing the influence of skin effects onhydraulic conductivity determinations in multilevel samplers installedwith direct push methods. Ground Water 38, 882–886.

Izbicki, J.A., Ball, J.W., Bullen, T.D., Sutley, S.J., 2008. Chromium, chromiumisotopes and selected trace elements, westernMojaveDesert, USA. AppliedGeochemistry 23, 1325–1352.

Jeen, S.W., Jambor, J.L., Blowes, D.W., Gillham, R.W., 2007a. Precipitates ongranular iron in solutions containing calcium carbonate with Trichlor-oethene and hexavalent chromium. Environmental Science and Technology41, 1989–1994.

Jeen, S.W., Mayer, K.U., Gillham, R.W., Blowes, D.W., 2007b. Reactive transportmodeling of trichloroethene treatment with declining reactivity of iron.Environmental Science and Technology 41, 1432–1438.

Köhler, S. 2003. Auswahl und Einsatz von eisenhaltigen Füllmaterialien fürpermeable reaktive Barrieren (PRB) zur in situ Grundwassersanierungam Beispiel eines Chromatschadenfalls. PhD Thesis. ETH Zürich, ETHZürich.

Kotas, J., Stasicka, Z., 2000. Chromium occurrence in the environment andmethods of its speciation. Environmental Pollution 107, 263–283.

Lasaga, A.C., 1984. Chemical kinetics of water–rock interactions. Journal ofGeophysical Research 89, 4009–4025.

Mayer, K.U., Blowes, D.W., Frind, E.O., 2001. Reactive transport modeling ofan in situ reactive barrier for the treatment of hexavalent chromiumand trichloroethylene in groundwater. Water Resources Research 37,3091–3103.

Naftz, D.L., Morrison, S.J., Davies, J.A., Fuller, C.C., 2002. Handbook of GroundwaterRemediation Using Permeable Reactive Barriers. Elsevier Science, San Diego,USA.

Ponder, S.M., Darab, J.G., Mallouk, T.E., 2000. Remediation of Cr(VI) andPb(II) aqueous solutions using supported, nanoscale zero-valent iron.Environmental Science and Technology 34, 2564–2569.

Powell, R.M., Puls, R.W., Hightower, S.K., Sabatini, D.A., 1995. Coupled ironcorrosion and chromate reduction:mechanisms for subsurface remediation.Environmental Science and Technology 29, 1913–1922.

Rai, D., Sass, B.M., Moore, D.A., 1987. Chromium(III) hydrolysis constantsand solubility of chromium(III) hydroxide. Inorganic Chemistry 26,345–349.

Schoenberg, R., Zink, S., Staubwasser, M., von Blanckenburg, F., 2008. Thestable Cr isotope inventory of solid earth reservoirs determined by doublespike MC-ICP-MS. Chemical Geology 249, 294–306.

Shields, W.R., Murphy, T.J., Catanzaro, E.J., Garner, E.L., 1966. Absolute isotopicabundance ratios and atomic weight of a reference sample of chromium.Journal of Research of the National Bureau of Standards Section Physicsand Chemistry. A 70 193-&.

Sikora, E.R., Johnson, T.M., Bullen, T.D., 2008. Microbial mass-dependent frac-tionation of chromium isotopes. Geochimica et Cosmochimica Acta 72,3631–3641.

Simon, F.G., Meggyes, T., Czurda, K., Roehl, K.E., 2003. Langzeitbeständigkeitvon permeablen reaktiven Wänden zur Sanierung kontaminierterGrundwässer. Altlasten Spektrum 6, 285–290.

Steefel, C.I., 2009. CrunchFlow. Software for modeling multicomponent reactiveflow and transport. User's manual. Berkeley, USA.

Steefel, C.I., DePaolo, D.J., Lichtner, P.C., 2005. Reactive transport modeling:an essential tool and a new research approach for the Earth sciences.Earth and Planetary Science Letters 240, 539–558.

Swiss Confederation, 1998. Verordnung über die Sanierung von belastetenStandorten (Altlasten-Verordnung AltlV)Public Law 814.680 http://www.admin.ch/ch/d/sr/8/814.680.de.pd 1998.

US-EPA, 1995. SW-847 Method 7198A. Test methods for evaluating solidwaste, physical/chemical methods, SW-847, 3rd ed. U.S. EPA, Washington,USA. Rev. 3.

Van Breukelen, B.M., Hunkeler, D., Volkering, F., 2005. Quantification of se-quential chlorinated ethene degradation by use of a reactive transportmodel incorporating isotope fractionation. Environmental Science andTechnology 39, 4189–4197.

Vikesland, P.J., Klausen, J., Zimmermann, H., Roberts, A.L., Ball, W.P., 2002.Longevity of granular iron in groundwater treatment processes: changesin solute transport properties over time. Journal of Contaminant Hydrology64, 3–33.

Wanner, C., Eggenberger, U., Mäder, U., 2011a. Reactive transport modelingof Cr(VI) treatment under fast flow conditions. Applied Geochemistry26, 1513–1523.

Wanner, C., Eggenberger, U., Zink, S., Kurz, D., Mäder, U., 2011b. A chromate-contaminated site in southern Switzerland—Part 1: site characterizationand the use of Cr isotopes to delineate fate and transport. Applied Geo-chemistry. doi:10.1016/j.apgeochem.2011.11.009.

Zhang, Y., Gillham, R.W., 2005. Effects of gas generation and precipitates onperformance of Fe° PRBs. Ground Water 43, 113–121.

Zink, S., Schoenberg, R., Staubwasser, M., 2010. Isotopic fractionation and reac-tion kinetics between Cr(III) and Cr(VI) in aqueous media. Geochimica etCosmochimica Acta 74, 5729–5745.