sorption of trichloroethene onto stylolites

Ž .Journal of Contaminant Hydrology 40 1999 1–23www.elsevier.comrlocaterjconhyd

Sorption of trichloroethene onto stylolites

Vera W. Langer a,) ,1, Kent S. Novakowski b, Allan D. Woodbury a

a Department of CiÕil and Geological Engineering, 344 Engineering Building, UniÕersity of Manitoba,Winnipeg, MB, Canada R3T 5V6

b National Water Research Institute, Burlington, ON, Canada

Received 22 October 1998; accepted 8 April 1999

Abstract

Batch and double reservoir diffusion cell experiments are used to investigate sorption oftrichloroethene onto stylolites. Stylolites are common features in carbonate rock formations, andmight contain high amounts of organic matter. Due to the hydrophobic character of TCE, itstransport in fractured carbonate aquifers could be significantly affected due to these aforemen-tioned features. No research has been carried out to evaluate the impact of stylolites on organicpollutant transport. The main objectives of this experimental research are to verify TCE sorptiononto stylolites, and to derive sorption and diffusion parameters describing the soluterrockinteraction. Test results show that stylolites from the Lockport Formation in Southern Ontario,Canada contain significant amounts of organic carbon. Discrepancies are noted between carbonanalyzer data and estimates from batch experiments and these might be due to TCE sorption alsoonto a clay mineral phase in stylolites or due to selective sampling. Adsorption and desorptionbehavior of TCE is investigated in specially designed double reservoir diffusion cells made out ofstainless steel and Teflon. Three semi-analytical solutions for one-dimensional, reactive tracermigration through a porous medium are derived and used to evaluate TCE time–concentrationprofiles. Experimental data can best be modeled using a kinetic Langmuir sorption formulationwith a maximum sorption capacity of 1.3 to 4.6 mgrg and a kinetic sorption constant of 4=10y7

to 5=10y7 lrmg sy1. TCE desorption into the exit reservoir is found to be a very slow kineticprocess. No retardation is observed during TCE migration through a clay and organic matter freedolostone sample. TCE seems not to interact with calcareous mineral phases and moves conserva-

Ž .tively. Bromide diffusion curves yield geometry factors g for dolostone ranging between 0.05and 0.13. From this study it can be concluded that TCE sorption is of importance when modelingTCE migration in fractured, stylolitic limestone aquifers where diffusion into the rock matrix takesplace. Temporal TCE storage in rock matrix stylolites and fracture wall stylolites has to be taken

) Corresponding author. Fax: q41-31-6314843; E-mail: [email protected] Present address: Rock Water Interaction Group, University of Bern, Dept. of Geol. and Min. and Petr.,

Baltzerstr. 1, 3012 Bern, Switzerland.

0169-7722r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0169-7722 99 00042-X

( )V.W. Langer et al.rJournal of Contaminant Hydrology 40 1999 1–232

into consideration when evaluating actions of remediation at organic spill sites. q 1999 ElsevierScience B.V. All rights reserved.

Keywords: Trichloroethene; Sorption; Stylolite; Diffusion; Organic carbon; Analytical solution

1. Introduction

Ž .Trichloroethene TCE is a non-polar organic compound belonging to the group ofŽ .highly volatile chlorinated hydrocarbons HVCH . It is highly toxic and is even at low

Žconcentrations a serious risk to human health Bedient et al., 1994; U.S. Environmental.Protection Agency, 1994 . HVCHs are often considered to be immiscible with water but

Ž .nevertheless have significant water solubility 0.15 to 8000 mgrl . The US EPAŽ . Ž .Drinking Water Standard for TCE is 5 ppb 0.005 mgrl Spitz and Moreno, 1996 .

Ž .Note its water solubility is 1100 mgrl at 258C Verschueren, 1983 , which is muchhigher than the aforementioned water standard. Therefore, small amounts of organic fuelleaking into our drinking water supply can represent a serious health risk and pollute alarge water volume.

In many carbonate aquifers around the world, groundwater has become contaminatedby volatile organic compounds, through industrial activity or waste disposal practices. Inmost carbonate formations, the transport of pollutants to local water supplies or surfacewaters occurs primarily through a network of bedding plane and vertical fractures. It hasbeen well established in the literature that sorption and matrix diffusion results in

Žsignificant retardation of pollutants in fractured rock formations Freeze and Cherry,.1979; McKay et al., 1993a,b; Novakowski and Lapcevic, 1994 . The amount of mass

that will be stored temporary in the rock matrix will largely depend on the soluterrockinteractions. It has been shown that in water, dissolved non-polar organic compounds

Žwill adsorb to the surface of organic particles present in geological material Karickhoff.et al., 1979 .

The organic content in carbonate aquifers might be concentrated in stylolite layers.Stylolites are common pressure dissolution features and are found in carbonate forma-tions throughout the world. Organic particles and other residual phases can become

Ž .concentrated in these thin 0.1 to 5 mm , serrated dark layers during their evolution.Because stylolites are often associated with the presents of fractures, TCE sorption ontothe organic fraction of stylolites may provide for a significant retardation mechanism.No studies to date have recognized the importance of stylolites on the transport of

Ž .volatile, chlorinated hydrocarbons such as TCE in fractured carbonates. The impact ofstylolitization on hydrogeological issues, like contaminant transport, is unknown but ispotentially significant. Because sorption of hydrophobic contaminants onto organicmatter is thought to be reversible, stylolites might act as contaminant sinks duringadsorption and as contaminant sources during desorption.

Before attempting to simulate and predict the fate of organic pollutants in carbonateaquifers it is important to find parameters and mathematical expressions describing thesoluterrock interactions accurately. It is common practice to utilize batch and diffusionexperiments for these purposes. Diffusion experiments have been employed to investi-gate the effective diffusion coefficient, retardation factor, and effective porosity for

( )V.W. Langer et al.rJournal of Contaminant Hydrology 40 1999 1–23 3

Ž .tracers Shackelford, 1991; Bickerton, 1993; Novakowski and van der Kamp, 1996 , andŽ .batch experiments have been used to determine the distribution coefficient K , thed

Ž U . Ž .fracture wall distribution coefficient K , the specific surface area h , and the fractiondŽ . Žof organic carbon f when assuming linear-Freundlich sorption Karickhoff et al.,oc

.1979; Chiou et al., 1981; Giger et al., 1983; Hassett and Banwart, 1989 . In order toevaluate the effective diffusion coefficient for a tracer, its free-water molecular diffusion

Ž . Ž .coefficient D and the rock matrix tortuosity factor g have to be known. For0Ž .geological materials, g ranges commonly from 0.01 to 0.5 Freeze and Cherry, 1979 .

Ž .The main objectives of this experimental and analytical study are i to verify TCEŽ .sorption onto stylolites, ii to estimate the amount of organic carbon present in

Ž .stylolites, iii to determine sorption and diffusion parameters describing the soluterrockŽ .interaction, and iv to derive semi-analytical solutions describing kinetic mass transport

behavior in the rock matrix. Assuming linear-Freundlich sorption, batch experiments areemployed to estimate the amount of organic carbon in stylolites, and to evaluate thefracture wall distribution coefficient for stylolite lined fractures. In addition, diffusionexperiments are conducted in horizontal, double-reservoir diffusion cells, which allowsthe simultaneous observation of adsorption and desorption behavior of reactive com-pounds. In a horizontal diffusion cell, the tracer is forced to migrate orthogonallythrough the linear stylolite layers. Analytical solutions for three different sorptionmodels are derived in Laplace space and inverted using a numerical inversion algorithm.The models differ in their finite or infinite sorption capacity and kinetic sorptionbehavior. Each model is tested to explain and fit the experimental data. The geometricfactor g for the rock matrix is determined using time–concentration profiles of theconservative tracer bromide. All rock samples are obtained from the Lockport Forma-tion, which is a fractured dolostone aquifer with numerous, parallel-bedding stylolites at

Ž .Smithville Ontario .

2. Stylolites in the Lockport Formation

The term ‘stylolite’ is generic and does not imply a specific composition. Stylolitiza-tion takes place under pressure dissolution due to loading or tectonically related stresses.Carbonates are mobilized under high pressure, whereas relatively insoluble residualphases like silica minerals, ore minerals and organic matter are concentrated on internaldiscontinuities. Stylolites are classified following the geometric appearance of the line of

Ž .discontinuity and their orientation Park and Schot, 1968 . In the Lockport Formation,Ž .stylolites have a suture-like geometry with a large range in amplitude 0.1 to 10 mm .

The relationship to bedding planes is dominantly horizontal, with occasional stylolitesinclined or in interconnected network form. In the Lincoln quarry near Smithville,individual stylolites extend laterally to over 100 m.

Drill core observations associate two types of fractures with stylolitization; horizontalunloading fractures and vertical cracks. Vertical cracks are found most predominantly inthe upper 25 m of the dolostone aquifer at Smithville. Horizontal unloading fractures are

Ž .perpendicular to the maximum paleo-stresses and can be grouped into 1 coring induced


Ž .mechanical breaks and 2 open bedding plane fractures. Unloading induced fracturingoccurs in association with stylolites having low amplitude, leaving the stylolite horizon-tally divided into two parts. Statistical evaluations from drill cores obtained at theSmithville site verify a correlation between highly fractured zones and zones with

Ž .numerous stylolites Radcliffe, 1995 . The highest stylolite density of 2.9 stylolitesrm isfound in the upper portion of the Lockport Formation.

The organic content of stylolites and of the ‘pure’ rock matrix from the LockportFormation was determined with a carbon analyzer on crushed and powdered material.As a result the organic carbon fraction in stylolites ranges between 0.01 and 5.26% and

Žin the rock matrix between 0.0 and 0.29% unpublished data; Hilverda and Langer,.1997; personal communication . These are significant amounts.

3. TCE sorption studies

3.1. Sampling procedure

To conduct the batch and diffusion experiments, stylolitic dolostone samples aretaken from the Lockport Formation in the vicinity of Smithville. For batch experiments,11 stylolitic and 11 control dolostone samples without large, visible stylolites are cutfrom drill core. The complete drill core is stored in water basins to ensure saturation ofthe rock matrix at all times. The stylolitic dolostone samples are cut adjacent to potentialbedding plane fractures so that one side exposed a stylolitic, dark gray to black surface.The control samples are cut next to the stylolitic dolostone samples. This samplingprocedure assures that control samples are from the same depth and similar in chemicalcomposition. It is desirable that the control samples should contain no organic carbon,but some of the controls show also dark lines and spots indicating possible organic

Žmatter organic matter is also finely distributed throughout the rock matrix and therefore.not visible . The volume of each sample is measured beforehand, whereas the dry solid

mass and porosity are calculated from weight loss by heating the samples in the oven at1058C after equilibrium concentrations are determined. For the diffusion experiments,

Ž .six thin 0.8 to 1.0 cm , water-saturated dolostone rock samples are cut parallel to thebedding. One sample contains no stylolite or other visible organic matter and is verylight colored, whereas the other samples contain one or more stylolites. The stylolites

Ž . Ž . Ž .embrace a large variety of amplitude 0 to 5 mm and thickness 0.5 to 3 mm Fig. 1 .The total porosity of the rock samples ranges from 2 to 8%. Samples are taken fromdifferent depth covering the upper 19 m of the Lockport Formation.

3.2. Batch experiments

The batch experiments are performed in special 50 ml glass vials capped with aTeflonw-coated silicon septum in order to minimize volatile TCE loss. The glass vialsand caps are EPA-certified for experimental use with volatile, organic compounds. Therock samples are put into the wide mouth glass vials and covered with 25 ml de-ionized


Fig. 1. Stylolitic dolostone samples from diffusion cell experiments.


H O. A 500 ppb initial TCE solution is obtained by spiking the 25 ml H O with 10 ml2 2Ž .TCE stock solution 1250 grl in methanol. After spiking, the glass vials are turned over

to prevent any gas leakage through the caps. TCE concentrations in the solution areŽ .monitored using the Head Space Method Garbarini and Lion, 1985 , where 50 ml gas is

extracted with a gas tight syringe from the head space and analyzed using a Photovacw

10 s Plus gas chromatograph. To account for volatile loss, all TCE concentrations arecalibrated using three blanks. The samples are monitored over 19 days to assureequilibrium is reached. Throughout the experimental period less than 5% of the initialTCE mass are lost in the blank containers.

3.3. Diffusion experiments

The double reservoir diffusion cells are constructed of stainless steel and Teflonw toassure no sorption of dissolved organic tracers onto the apparatus itself. Two stainlesssteel tubes are separated by the rock sample and are cone shaped towards the ends to

Ž .prevent air entrapment during filling with de-aired, de-ionized water Fig. 2 . Thisdesign leaves only a small opening at either end, which are closed off by a Swagelogw

fitting with a Teflonw-coated silicon septa in the cap. Through the septa, both reservoirscan be sampled with a syringe without opening the ends. Stainless steel tubes and rocksamples are hold in place by a Teflonw coat, which closes tight around both reservoirs.This was accomplished by heating the Teflonw liner, which causes it to shrink. Sixdouble reservoir diffusion cells and one blank single reservoir cell are build. The blankis capped at one side with a stainless steel plate and used to account for volatile TCEloss. The source reservoir in all cells has a volume of 20 ml. The exit reservoir is buildwith a volume of 20 ml for two cells and 209 ml for 4 cells. Having a 10-fold larger exitreservoir than source reservoir accelerates the drop in tracer concentration in the source

Fig. 2. Design of double reservoir diffusion cell.


reservoir, but on the other hand, significantly prolongs equilibrium times and causesanalytical problems due to very low tracer concentrations in the exit reservoir.

Ž y.Bromide Br is chosen as a conservative tracer because it does not naturally occurin the dolostone aquifer and therefore the background concentration in the saturated rocksample can be assumed to be zero. Bromide is thought to be non-reactive with

Žcalcareous minerals like calcite and dolomite as well as with organic matter Smart and. Ž .Laidlaw, 1977 . The free-water diffusion coefficient D for a diluted potassium0

y10 2 Ž .bromide electrolyte at 258C is 20.1=10 m rs Stokes, 1950 . Although D is a0

function of concentration, the experimental method is utilized over a sufficiently smallconcentration range so that D can be considered constant.0

TCE, in contrast, is a volatile organic compound with a water solubility of 1100mgrl at 258C. It reacts with organic matter in the dolostone samples by adsorption, asshown in the batch experiments. The TCE free-water diffusion coefficient is thought tobe lower than bromide due to its larger molecule size, but is found to be similar.

The source reservoirs of all double diffusion cells are filled with a 309 ppm Bry

solution. After the source reservoirs are closed off, 20 ml liquid are extracted with asyringe through the septa and all diffusion cells are than spiked with 2=10 ml TCEstock solution. The stock solution is composed of 1000 mgrl TCE in methanol. Thisprocedure results in an initial TCE concentration of 1000 ppb in all source reservoirs.The exit reservoirs remains filled with de-ionized water.

Over a period of 6 to 7 months, 150-ml liquid samples are taken out of the sourcereservoirs with a gas tight syringe and analyzed for Bry and TCE. The liquid samplesare injected in special, gas tight, 300 ml glass vials with a Teflonw coated silicon septain its cap. After equilibration of liquid and gas phase, 50 ml gas phase from the glass

Ž .vial is sampled and analyzed for TCE with a gas chromatograph head space analysis .In the first three sample rounds a Photovacw is used, in later rounds a 5890A HewlettPackard gas chromatograph with a sensitivity below 1 ppb is used. For each sampleround three TCE standards are prepared and analyzed for calibration. Subsequently, allTCE concentrations are calibrated through the TCE concentration in the blank diffusioncell to account for volatile loss. Throughout the experimental period about 10% TCE arelost in the blank diffusion cell.

After TCE analyses, 100 ml liquid from the glass vial are diluted with 4 mlde-ionized water for Bry analyses with an ion chromatograph. Standards of 3, 5, and 10ppm are used for calibration.

4. Analytical models for diffusion cell

To classify TCE sorption behavior onto stylolites into linear or nonlinear, reversibleor irreversible, instantaneous or kinetic behavior, analytical solutions for three different

Ž .one-dimensional sorption models are derived in Laplace space see Appendix A .Ž .Solutions are obtained using an inversion algorithm by De Hoog et al. 1982 . Computed

time–concentration profiles obtained from the analytical solutions are used to evaluatecontaminant concentrations measured in source and exit reservoirs of the diffusion cells.


Optimal sorption parameters are determined by calibration of the experimental andcomputed data. The % error is defined as

N2

C yCŽ .Ý comp. exp1% ees )100. 1Ž .N

2) CŽ .Ý exp1

In a horizontal diffusion cell, advective flow and longitudinal dispersivity are zero.Therefore, a concentration gradient is the only driving force, and is given by Fick’s lawfor a conservative tracer. Adding a sorption term the governing equation becomes:

EC E2 C r ECUb

sD y 0FxGL 2Ž .2Et u EtEx

with

DsD g 3Ž .0

where D: effective diffusion coefficient, D : free-water diffusion coefficient, g : geo-0Ž 3. Umetric factor related to tortuosity, C: solute concentration in the pore fluid MrL , C :

Ž . Ž . Ž 3.solute sorbed onto rock mass MrM , u : porosity LrL , r : rock bulk density MrL ,bŽ y3 y1.and ECrEt: change in concentration with time M l T . Note that the x direction

extends from 0 to L, where L is the rock sample thickness. For many dilute organiccontaminants, the sorption process onto organic matter in the porous rock sample is

Ž .thought to be linear and reversible Karickhoff et al., 1979 . Assuming instantaneousequilibrium, the linear Freundich sorption isotherm can be applied with CU sCK ,d

Ž . Ž .where K is the organic contaminant distribution coefficient. Transport Eqs. 4 and 5dŽ . Ž .are derived Fetter, 1993 , see also Table 1; models I and II :

EC E2 C r E K CŽ .b dsD y 4Ž .2Et u EtEx

or

EC D E2 C E2 Cs sD 5Ž .A2 2Et R Ex E x

withr b

Rs1q Kdu

Non-equilibrium or kinetic sorption describes the interaction of the organic solute withthe rock matrix where adsorption and desorption processes have different rates. Irre-

Ž .versible and reversible, linear, first-order kinetic sorption Fetter, 1993 is modeled using

ECU

Us k C y k C 6Ž .2 3Et Ž . Ž .sorption desorption

()

V.W

.Langer

etal.r

JournalofC

ontaminantH

ydrology40

19991

–23

9

Table 1Transport equations for conservative and reactive tracers

Model Tracer Transport equation Sorption isotherm Parameters evaluated2 2Ž . Ž . ŽŽ . Ž ..Conservative Bromide EC r Et s D E C r Ex D: effective diffusion

Ž .model I with Ds D g coefficient0

g : geometry factorU2 2Ž . Ž . ŽŽ . Ž ..Linear, reversible TCE EC r Et s D E C r Ex C sCK D : apparent diffusiond A

Ž .ŽŽ Ž ..equilibrium Freundlich y r ru E K C coefficientb dŽ . Ž .. Ž .sorption model II r Et 4

2 2Ž . Ž . Ž .ŽŽ . Ž .. Ž .or EC r Et s DrR E C r Ex 5 R: retardation coefficientŽ .with Rs1q r ru K K : distribution coefficientb d d

Ž .and D s DrRAU U2 2Ž . Ž . ŽŽ . Ž .. Ž .Irreversible and reversible, TCE EC r Et s D E C r Ex EC rEt s k Cy k C k ,k : kinetic rate constants2 3 2 3

UŽ .Ž . Ž .first-order kinetic sorption y r ru k Cy k C 7 k ™0 irreversibleb 2 3 3Ž .model III

U U2 2Ž . Ž . ŽŽ . Ž .. Ž . Ž . Ž .Nonlinear, irreversible, TCE EC r Et s D E C r Ex EC r Et s k C C yC k : kinetic rate constant4 max 4UŽ . Ž . Ž .kinetic Langmuir sorption y r ru k C C yC 9 C : maximum sorption capacityb 4 max max

Ž .model IV

Formulations after Fetter, 1993.


Ž .where k , k are rate constants model III . The governing transport equation is:2 3

EC E2 C r b UsD y k Cyk C 7Ž . Ž .2 32Et uEx

Contaminant sorption might also be limited by the capacity of the organic material toadsorb. This nonlinear behavior can be modeled with the Langmuir sorption isothermŽ .model IV . The kinetic version is given by

ECU

U Usk C C yC yk C , 8Ž . Ž .4 max 5Et

Žwhere k and k are rate constants and C is the maximum sorption capacity Fetter,4 5 max. Ž .1993 . Assuming that the desorption rate is near zero k ™0 , contaminant transport is5

Ž .formulated through Eq. 9 .

E C E2 C r b UsD y k C C yC 9Ž . Ž .4 max2E t uE xŽ . Ž . Ž .To solve Eqs. 5 , 7 and 9 for contaminant migration in a horizontal diffusion cell the

Laplace transform is applied. After obtaining a solution in Laplace space a semi-analyti-Ž .cal approach is used to solve for C in the source or exit reservoir Bickerton, 1993 .

5. Data analyses and results from batch experiments and diffusion experiments

TCE loss due to gas leakage andror degradation is eliminated from the results bycalibration with control samples containing the same initial TCE concentration but norock material. Due to the differences in organic content and rock mass in the samples,TCE concentrations drop at different time rates in batch and diffusion experiments.

In batch experiments, sample 65E-ST3-20 shows the highest sorption capacity forŽ .TCE Fig. 3 . The largest amount of sorption takes place in the first two days. The

concentration decreases to 0.26 CrC , whereas in the following 17 days the concentra-0

tion decline slows down and reaches equilibrium at 0.11 CrC . In all other samples0

Fig. 3. TCE time–concentration profiles from batch experiments for stylolitic dolostone and dolostone rockmatrix.


most of the sorption is completed after about eight days, and 50% from the total amountof sorption takes place in the first five days. The result from a batch experiment on

Žground, pure stylolitic matter is used for comparison unpublished data; Hilverda, 1997;.personal communication . The pure, ground stylolite material sorbed 75% of the starting

Ž . Ž . 3 Ž .500 ppb 0.25 CrC in less than a day 20 h . The five cm 11 g pure, grounded0

Fig. 4. Experimental and computed TCE and bromide time–concentration profiles for source and exitreservoirs. Sample 65E-ST18-1 is a light dolostone sample with presumably no organic matter and clay.Sample 65E-ST11-3 and 65E-ST1-17 are stylolitic dolostone samples. Model I simulates conservative masstransport, and model II accounts for linear, reversible Freundlich sorption during mass transport.


stylolite material interacted much faster with TCE than most stylolitic dolostonesamples. This behavior is expected because the grounded material exposed a largersurface area to the dissolved TCE. In the rock sample batch experiments, an initial massof 12.5 mg TCE is injected. The stylolite mass is calculated to range from 1.7 to 5.2 g,and the TCE mass sorbed lays between 2.8 and 11.3 mg. The pure, ground stylolitesample had a mass of 11 g. The initial TCE mass injected was 17.5 mg, and the TCEmass sorbed was 14.6 mg.

Diffusion experiments with stylolitic dolostone samples results in TCE time–con-centration data, from source reservoirs, that are below their respective bromide data.With sample 65E-ST11-3 the normalized CrC concentration decreases to about 0.20

Ž .and with sample 65E-ST1-17 to about 0.4 after 180 days Fig. 4 . This values areconsistent with the different exit reservoir sizes, which cause different concentrationgradients. In both diffusion cell exit reservoirs, the TCE concentrations remain low with0.05 CrC after 180 days, whereas Bry concentrations increase to the expected0

amounts as concentrations in the source reservoirs decrease. TCE and Bry soluteconcentrations in the control diffusion cell 65E-ST18-1 are almost identical. This is truefor source and exit reservoirs.

6. Interpretations

6.1. Batch experiments

Assuming a linear, equilibrium sorption isotherm, the TCE distribution coefficient foreach sample is calculated using the ratio C rC . The organic carbon partition coeffi-eq s

Ž 3 .cient K l rM is computed from the empirical, log–linear regression equationocŽ .established by Karickhoff et al. 1979 , and an octanol–water partition coefficient for

Ž .TCE of log K s2.29 Giger et al., 1983 .ow

log K sy0.21q log K Karickhoff et al., 1979Ž .oc ow

The above relationship was established from batch experiments with different chlori-Ž .nated hydrocarbons. The fraction of organic carbon f is estimated using the linearoc

relationship K sK f , where K for TCE is calculated from the above equation.d oc oc oc

Two main assumption are made using this simplified approach. First, K is indepen-oc

dent of the type of organic matter, and second, adsorption onto organic matter is the3 Žonly soluterrock interaction. As a result, K ranges from 0.6 to 7.8 cm rg averaged

. 3 Ž .2.6 in the stylolitic dolostone samples and between 0.4 and 2.0 cm rg average 0.9 inŽ .the dolostone control samples Table 2 . The data indicate clearly that the stylolitic rock

matrix contains organic matter of significant amount contributing to TCE sorption. Thefraction of organic carbon for all 22 rock samples ranges between 0.0032 and 0.066.Estimating the thickness of the stylolite layer with 0.5 to 1.5 mm on the rock samplesurface and assuming a density of 2.2 grcm3, the stylolite mass is calculated for eachsample. The stylolite mass from the 11 samples ranges between 1.7 and 5.2 g. Taking


Table 2TCE distribution coefficients for pure stylolite, stylolitic dolostone and dolostone rock matrix

Type Sample mass TCE distribution coefficient Organic carbon3Ž . Ž . Ž .g K cm rg f %d oc

bŽ .Stylolite pure 11 17.61 15 1Ž .Stylolitic dolostone 10.55–37.41 0.63–7.83 0.61–6.51 11

a Ž .Stylolite layer 1.7–5.2 4.09–54.70 3.4–45.5 11Ž .Dolostone matrix 14.64–34.14 0.38–2.01 0.32–1.67 11

a Estimating the stylolite mass in each sample and assuming an organic carbon background value equal to thedolostine matrix value, K and f are calculated for stylolite layers.d ocb Number in parentheses indicate the number of samples.

the fraction of organic carbon from the dolostone control sample as background value,Ž .f for the stylolite mass part is calculated Table 2 . The average f for the styloliteoc oc

Ž 3 . Žlayers is 0.1875 with an average K of 22.54 cm rg compared to a f of 0.15 withd oc3 .K of 17.61 cm rg found for the pure, ground stylolite sample in trial test-1. A wided

range in K for the different stylolite layers is expected due to the large variation ind

their composition. The dolostone matrix without stylolites has an average amount ofŽ .0.8% f s0.008 organic carbon.oc

As stated above, fraction of organic carbon in stylolites are estimated on theassumption that TCE adsorbs only onto organic carbon. This cannot be verified at thispoint. Stylolites might also contain a large fraction of clay minerals which mightcontribute towards TCE sorption. This would result in lower f estimates; however, itoc

would not diminish the important finding of TCE sorption onto stylolites.The fracture wall distribution coefficient is calculated for each stylolite surface by

knowing the concentration of TCE in solution and the mass of TCE adsorbed per unitarea of fracture surface.

mass of solute adsorbed per unit area of fracture surface MrL2Ž .UK sd 3concentration of solute in solution MrLŽ .

Therefore, knowing K U and K , the specific surface area h for TCE sorption isd d

evaluated and ranges between 1.9 and 8.2 cm2rg. The average specific surface area forTCE sorption onto stylolite fracture wall layer is 4.872 cm2rg.

6.2. Estimation of retardation factors

Ž .Assuming linear-reversible equilibrium sorption, retardation factors R are estimatedto range between 100 and 1000 for stylolite layers and between 10 and 50 for the rock

Ž .matrix based on the range of TCE distribution coefficients and log K s2.29 Fig. 5 .ow

From 546 core sample measurements the average total porosity of the rock matrix in theLockport Formation is 6.81%, and the bulk density is 2.51 grcm3. Estimation of thestylolitic fracture wall retardation factor of 30 to 400 is based on the range of TCE

Ž .specific surface area data and an assumed fracture half aperture width b of 0.02 cm.K U

dR s1q 10Ž .f b


Fig. 5. Assuming linear, reversible equilibrium sorption, retardation factors for dolostone rock matrix andstylolite layers are estimated on the basis of the TCE octanol–water partition coefficient.

6.3. Diffusion experiments

Curve fitting by calibration between analytical- and experimental data is used toevaluate the time–concentration data in source and exit reservoirs for both tracers.Decreases in solute concentration with time are strictly due to diffusion and sorption.

Ž .For the conservative tracer bromide the semi-analytical solution from Bickerton 1993is used. Knowing the initial concentrations in the source reservoir, porous medium andexit reservoir, and estimating the porosity and the diffusion coefficient, concentration–time data for the solute in source- and exit reservoir are calculated. Porosity values forthe rock samples are determined after diffusion experiments are terminated. Bromidetime–concentration profiles are used to evaluate the rock sample’s tortuosity related

Ž .geometry factor g model I; Table 1 . The % error between computed and experimentaldata ranges from 2% to 7%. Curve fitting results in effective diffusion coefficients for

y y10 2 y10 2 Ž .Br ranging from 0.9=10 m rs to 2.6=10 m rs Table 3 . These values yielddolostone geometry factors between 0.045 and 0.13. TCE time–concentration data areevaluated employing the derived sorption models, and assuming that the free-waterdiffusion coefficient and the geometry factor for Bry and TCE are identical.

First, all time–concentration profiles from source and exit reservoirs are analyzedŽ . Ž .assuming reversible, linear, equilibrium sorption model II for TCE Fig. 5 . Experi-

mental Bry and TCE concentrations in source and exit reservoir of control sample65E-ST18-1 are best modeled with an effective diffusion coefficient of 2.61=10y10

m2rs and Rs1. This confirms that the free-water diffusion coefficient for TCE is verysimilar to the one for Bry and no adsorption or reaction takes place in the organiccarbon free dolostone sample. Evaluating the TCE diffusion data from stylolitic dolo-stone samples, an apparent diffusion coefficient is found, which ranges between

y10 y10 2 Ž .5.0=10 and 12.1=10 m rs Table 3 . Separating retardation factors from theapparent diffusion coefficient, assuming D equals Dy, results in unacceptableTCE Br

()

V.W

.Langer

etal.r

JournalofC

ontaminantH

ydrology40

19991

–23

15

Table 3Results from diffusion experiments using models I and II

w xSample Tracer Porosity Effective diffusion coefficient g Apparent diffusion coefficient Retardation factorU U2 2Ž . Ž . Ž .% D s D g m rs D s D rR m rs Rs D g rD0 A 0 A

model Iq II: source reserÕoirsy y1065E-ST27-4 Br 5.1 0.9=10 0.045

y1 0TCE 5.1 0.045 5.4=10 0.17y y1065E-ST18-2 Br 3.9 2.0=10 0.1

y1 0TCE 3.9 0.1 5.6=10 0.36y y1065E-ST30-2 Br 2.0 2.0=10 0.1

y1 0TCE 2.0 0.1 5.0=10 0.4y y1065E-ST11-3 Br 2.0 2.2=10 0.11

y1 0TCE 2.0 0.11 12.1=10 0.18y y1065E-ST1-17 Br 7.0 1.6=10 0.08

y1 0TCE 7.0 0.08 5.0=10 0.32y y10Ž .65E-ST18-1 control Br 8.0 2.6=10 0.13

y1 0TCE 8.0 0.13 2.6=10 1

model Iq II: exit reserÕoirsy y1065E-R1-17 Br 7.0 1.6=10 0.08

y1 2TCE 7.0 0.08 1.0=10 160y y1065E-R11-3 Br 2.0 2.2=10 0.11

y1 2TCE 2.0 0.11 2.0=10 110y y10Ž .65E-R18-1 control Br 8.0 2.6=10 0.13

y1 0TCE 8.0 0.13 2.6=10 1


values between 0.2 and 0.4. Therefore, model II is not physically meaningful and mustbe rejected. However, the low TCE time–concentration data from exit reservoirs can beevaluated with model II, and result in retardation factors of 110 to 160. It is apparentthat concentration profiles from the source and exit reservoirs cannot be modeledtogether when assuming reversible, linear, equilibrium sorption for TCE.

Ž .A kinetic, irreversible sorption model model III is much more appropriate todescribe the observed behavior. This model leads to attenuation of the solute in the

Fig. 6. Experimental and computed TCE and bromide time–concentration profiles for source reservoirs. Rocksamples are stylolitic dolostone. Model III simulates mass transport with irreversible, first-order kineticsorption, and model IV accounts for nonlinear, irreversible kinetic Langmuir sorption during mass transport.

()

V.W

.Langer

etal.r

JournalofC

ontaminantH

ydrology40

19991

–23

17

Table 4Results from diffusion experiments using model IV

Model IV: source reservoirsa w xSample Tracer Porosity Effective diffusion coefficient g Maximum adsorption capacity Kinetic rate constant

U 2 y1Ž . Ž . Ž . Ž .% D s D g m rs C mgrg k lrmg s0 max 3

y10 y765E-ST27-4 TCE 5.1 0.9=10 0.045 3.18 5=10y1 0 y765E-ST18-2 TCE 3.9 2.0=10 0.1 2.72 5=10y1 0 y765E-ST30-2 TCE 2.0 2.0=10 0.1 3.2 4=10y1 0 y765E-ST11-3 TCE 2.0 2.2=10 0.11 4.6 5=10y1 0 y765E-ST1-17 TCE 7.0 1.6=10 0.08 1.3 4=10y1 0Ž .65E-ST18-1 control TCE 8.0 2.6=10 0.13 0 0

a y10 Ž 2 .D s20.1)10 m rs .0 TCE


stylolite layers. In the given time frame the TCE desorption rate is zero, meaningk s0. The forward, first order kinetic constant k is found to lie between 3.15 and 9.463 2

cmrg yeary1. The steep drop in TCE concentration during the first 30 days suggeststhat the adsorption process is nonlinear and therefore concentration-dependent. How-

Žever, TCE source reservoir concentrations in the two smaller diffusion cells samples.65E-ST1-17 and 65E-ST27-4 seem to have reached equilibrium, which cannot be

Ž .explained with a linear, irreversible sorption model Fig. 6 . In fact, a nonlinear, kineticŽ .Langmuir type model IV adsorption behavior with a zero desorption rate provides an

even better correlation between experimental and analytical data for source and exitreservoirs. The % error decreased by about 10% ranging from 16 to 26% using a linear,

Ž .irreversible sorption model model III to about 6 to 15% using a kinetic LangmuirŽ .sorption model model IV . The decrease in TCE concentrations, in the source reser-

voirs, depends on the rock sample sorption capacity. Because the organic content in allsamples is assumed to be different, it is not surprised to observe a range of sorptioncapacities. Sample 65E-ST11-3 with the highest, visible organic content has also thehighest TCE sorption capacity with 4.6 mgrg and sample 65E-ST1-17 with the thinnest

Ž .stylolite layer has the lowest capacity with 1.3 mgrg Table 4 . A forward kinetic rateŽ . y7 y7 y1constant k ranging between 4=10 and 5=10 lrmg s and a backward4

Ž .kinetic rate constant k near zero fits the observed concentration profiles in source and5

exit reservoirs best.

7. Discussion and conclusions

Using the semi-analytical solution for conservative tracers and knowing that thefree-water diffusion coefficient for bromide in diluted systems is 20.1=10y10 m2rs,the tortuosity related geometry factor g for dolostone samples ranges between 0.045 and

Ž0.13. This is in good agreement with published data for geological materials Freeze and.Cherry, 1979 . Bromide, as a conservative tracer, is used for comparison with TCE.

Time–concentration profiles from both tracers show that when migrating through a clayŽ .and organic-matter free dolostone sample 65E-ST18-1 TCE moves conservatively. It

can be concluded that no interactions between TCE and calcareous mineral phases takeŽ .place. Benker et al. 1997 showed in a TCE-bromide tracer field experiment, conducted

in a sand aquifer with less than 1% silt and clay and virtually no organic matter present,that TCE moves conservatively. TCE, as a non-polar organic compound, does not sorbonto quartz mineral surfaces. Note that the organic carbon fraction determined by carbonanalyzer in stylolites is between 0.01 and 5.26% and in the rock matrix between 0.0 and0.29%. Estimates of the organic carbon content in stylolites from batch adsorptionexperiments yields values between 3.4% and 45.5%. The organic carbon in the rockmatrix is estimated with 1.67%. This discrepancy might be explained by TCE adsorptionnot only onto organic matter but also onto clay minerals in stylolites and rock matrix inbatch experiments or simply by selective sampling. The small number of samples forcarbon analyses might not represent the large compositional range of stylolites.

Batch experiments and diffusion studies in this research clearly demonstrate thecapability of stylolites to adsorb significant amounts of dissolved TCE. In the experi-


mental time frame of six months no significant amounts of desorbed TCE are measur-able. Only 1.7% TCE from the initial mass injected into the source reservoir is detected

Ž .in the exit reservoir. This supports experimental results from Culver et al. 1997 whereonly 5% of the initially sorbed mass from batch experiments had been desorbed after

Ž600 h. The rate of desorption might also depend on long term TCE exposure Grathwohl. Žand Reinhard, 1993; Culver et al., 1997 . With large advective flows 24000 times the

.void volume , 72% of the initial soil TCE was removed in desorption column experi-Ž .ments from soils with 0.13% organic carbon Pavlostathis and Jaglal, 1991 . It seems

that advective flow has a large influence on the time dependent desorption behavior ofTCE. The difficulty of finding an appropriate distribution coefficient for TCE due totime dependent changes in mass transfer has to be pointed out. TCE distributioncoefficients for stylolites from our adsorption batch experiments are found to range from

3 Ž4.1 to 54.7 cm rg, which correlates with published data Pavlostathis and Jaglal, 1991;. Ž 3 .Zytner, 1992; Culver et al., 1997 . Unusually low 1 cm rg TCE distribution coeffi-

Ž .cients for fine to medium grained soil were reported from Picatinny Arsenal NJ withŽ .1.04% organic material Sahoo and Smith, 1997 . Retardation factors for stylolite layers

are estimated from TCE distribution coefficients and range between 100 and 1000.Comparison between bromide and TCE concentration profiles in diffusion cell exitreservoirs yield retardation factors of 110 to 160 assuming linear, reversible equilibriumsorption.

Batch experiments on stylolitic fracture surfaces reveal TCE specific surface areas of1.9 to 8.2 cm2rg, which seems unusually high when compared to specific surface areas

Ž 2 . Ž 2 .of peat moss 0.4 m rg and GAC 1300 m rg , a commercial product with 74.1%organic carbon.

Ž .Zytner 1992 was able to model TCE adsorption and desorption onto differentorganic soils with an organic carbon content ranging from 1.0% to 49.4% using anonlinear, reversible Freundlich isotherm. TCE time–concentration profiles from ourdiffusion study does not support that finding. An interesting conclusion from Zytner’sŽ .1992 batch experiments was that not only the adsorption potential increases withincreasing organic carbon content, but also the soils retention potential. Due to unknownorganic carbon content in the stylolite samples used in the diffusion cells this potential

Ž .could not be evaluated. The nonlinear, kinetic Langmuir sorption model model IVprovides significantly improved fits of experimental TCE data with computed concentra-

Ž .tions when compared to the reversible, linear equilibrium sorption model model II andŽ .the irreversible, kinetic sorption model model III . There appears to be a maximum

sorption capacity of organic contaminant onto stylolites. The rate of mass transferdecreases as more solute adsorbs onto organic surfaces and time–concentration profilesflatten out. In other words, the TCE distribution coefficient is not constant. Styloliticrock samples, containing different amounts of organic material, show a maximum TCEsorption capacity of 1.3 to 4.6 mgrg. The time dependent adsorption behavior is also

Žseen in the batch experiment profiles. Depending on the stylolite rock sample its.organic carbon content , equilibration times ranges between 8 and 17 days.

The importance of stylolites on TCE transport through dolostone aquifers is apparent.An average of 55 horizontal stylolite layers with an average thickness of 0.5 mm per

Ž .stylolite in the upper Lockport Formation Eramosa Member can adsorb 78.7 to 278.3 g


TCE per square meter surface area. A dissolved TCE plume that extends over 5000 m2

Žcould lead to TCE attenuation in stylolites of 393.25 to 1391.5 kg 270 to 955 l pure.TCE assuming that all stylolites are reached by the TCE plume and the maximum

possible amount of TCE is adsorbed. Any bedrock remediation plan has to take intoaccount that TCE desorption from stylolites back into the pore volume of rock matrixand into fractures is a very slow kinetic process. Instead of trying to retrieve dissolvedTCE by pumping, enhancing in situ biodegradation might be in many cases the betteroption. Incorporating the determined physical and hydrogeological parameters into alarger scale model and a stylolite sensitivity analyses are the next steps in the ongoingresearch project.

Acknowledgements

Most of the experimental work was carried out at the Canadian Center for InlandWaters in Burlington and funded through the Bedrock Remediation Project Smithville.Special thanks go to Kelly Miller and Lavinia Zanini for technical advise and help withchemical analyses. Analytical work continued at the University of Manitoba, and thanksare extended to Prof. Racz and Mr. Sarna from the Department of Soil Science and toMr. Stainton from the Freshwater Institute. The authors like to acknowledge suggestionsand comments from Prof. Benker and an anonymous reviewer, which helped to improvethe paper significantly.

Appendix A. Application of Laplace transform

Ž .The Laplace transform of f C is defined as

`

c x , p s exp ypt C x ,t d t 11Ž . Ž . Ž . Ž .H0

with p: Laplace variable; c: solution of C in Laplace space.Ž . Ž . Ž .Eqs. 5 , 7 and 9 transform to

2D d cpc x , p s 12Ž . Ž .2R d x

U2d c r r Cb bpc x , p sD y k cq k 13Ž . Ž .2 32 u u pd x

and

2d c r b Upc x , p sD y k c C yC , respectively. 14Ž . Ž . Ž .4 max2 ud x


Using the following initial and boundary conditions the particular solution of theŽ . Ž . Ž .Laplace transform Eq. 12 , 13 and 14 are found:

initial conditions:

C x ,0 s0 0FxFL in the porous mediumŽ .C 0 sC in the source reservoirŽ .s 0

C 0 s0 in the exit reservoirŽ .e

boundary conditions:

C t sC 0,tŽ . Ž .s

C t sC L,tŽ . Ž .e

Ž . Ž .The solution in Laplace space of Eqs. 12 – 14 is formulated as:

x' 'C pyj Rp exp Rp ylŽ .0 e ž /ž /'Dc x , p sŽ .

D

x' 'C pqj Rp exp Rp lyŽ .0 e ž /ž /'Dy 15Ž .

D

with

' ' 'Ds pyj Rp pyj Rp exp yl RpŽ . Ž . Ž .s e

' ' 'y pqj Rp pqj Rp exp l RpŽ . Ž . Ž .s e

and

Ž .c x , p

U Us C s C

Ž . Ž . Ž .C y exp y pqt l pyj pqt y pqj pqt' ' 'Ž . Ž . Ž .0 e ež / ž /Ž . Ž .pqt pqts

D

=Ž .pqt

exp x(ž /D

U Us C s C

Ž . Ž . Ž .y C y exp pqt l pqj pqt q pyj pqt' ' 'Ž . Ž . Ž .0 e ež / ž /Ž . Ž .pqt pqtq

D

=UŽ .pqt s C

exp y x q 16Ž .( ž /ž / Ž .D p pqt


with'A D u

j ss Vs

'A D uj se Ve

Lls 'D

r bts k2

ur b

ss k3u

' ' 'Ds pyj pqt pyj pqt exp yl pqtŽ . Ž . Ž .s e

' ' 'y pqj pqt pqj pqt exp l pqtŽ . Ž . Ž .s e

andc x , pŽ .

xU UC pyj pqt C yC exp pqt C yC yl( (Ž . Ž .ž /0 e max max ž /ž /'D

sD

xU UC pqj pqt C yC exp pqt C yC ly( (Ž . Ž .ž /0 e max max ž /ž /'D

yD

17Ž .with

U UDs pyj pqt C yC pyj pqt C yC( (Ž . Ž .ž / ž /s max e max

U U=exp yl pqt C yC y pqj pqt C yC( (Ž . Ž .ž / ž /max s max

U U= pqj pqt C yC exp l pqt C yC( (Ž . Ž .ž / ž /e max max

where, A: rock sample surface, C and C : concentration in the source and exits e

reservoir, respectively, and V and V : volume of the source and exit reservoir,s e

respectively. The solution in real space was obtained using a numerical inversionŽ . Ž U .algorithm De Hoog et al., 1982 . The sorbed concentration C was updated after each

time step.

References

Bedient, P.B., Rifai, H.S., Newell, C.J., 1994. Ground Water Contamination: Transport and Remediation.Prentice Hall PTR, New Jersey.

Benker, E., Davis, G.B., Barry, A.D., 1997. Factors controlling the distribution and transport of trichloroetheneŽ .in a sandy aquifer-hydrogeology and results of an in situ transport experiment. J. Hydrogeology 202 1–4 ,

315–340.


Bickerton, G.S., 1993. A Semi-Analytical Model for Solute Diffusion Trough Rock Samples of FiniteThickness. Unpublished BS Thesis, Univ. of Waterloo, Ontario, Canada.

Chiou, C.T., Peters, L.J., Freed, V.H., 1981. Soil–water equilibria for nonionic organic compounds. ScienceŽ .213 8 , 683–684, Washington.

Culver, T.B., Hallisey, S.P., Sahoo, D., Deitsch, J.J., Smith, J.A., 1997. Modeling the desorption of organiccontaminants from long-term contaminated soil using distributed mass transport rates. Environ. Sci.

Ž .Technol. 31 6 , 1581–1587.De Hoog, F.R., Knight, J.H., Stokes, A.N., 1982. An improved method for numerical inversion of Laplace

Ž .transforms, SIAM. Journal on Scientific and Statistical Computing 3 3 , 357–366.Fetter, C.W., 1993. Contaminant Hydrogeology, Macmillan, New York.Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice-Hall, Englewood Cliffs, NJ.Garbarini, D.R., Lion, L.W., 1985. Evaluation of sorptive partitioning of nonionic pollutants in closed systems

by headspace analysis. Environ. Sci. Technol. 19, 1122–1128.Giger, W., Schwarzenbach, R.P., Hoehn, E., Schellenberg, K., Schneider, J.K., Wasmer, H.R., Westall, J.,

Zobrist, J., 1983. Das verhalten organischer wasserinhaltsstoffe bei der grundwasserbildung und imŽ .grundwasser. Gas-Wasser-Abwasser 63 9 , 517–531, Zurich.¨

Grathwohl, P., Reinhard, M., 1993. Desorption of trichloroethylene in aquifer material. Rate limitation at theŽ .grain scale. Environ. Sci. Technol. 36 1 , 89–108.

Hassett, J.J., Banwart, W.L., 1989. The sorption of nonpolar organics by soils and sediments. Soil ScienceŽ .Society of America SSSA 22, 31–44, special publication.

Karickhoff, S.W., Brown, D.S., Scott, T.A., 1979. Sorption of hydrophobic pollutants on natural sediments.Ž .Water Res. 13 3 , 241–248, Oxford Pergamon Press.

McKay, L.D., Cherry, J.A., Gillham, R.W., 1993a. Field experiments in fractured clay till: 1. HydraulicŽ .conductivity and fracture aperture. Water Resour. Res. 29 4 , 1149–1162.

McKay, L.D., Gillham, R.W., Cherry, J.A., 1993b. Field experiments in fractured clay till: 2. Solute andŽ .colloid transport. Water Resour. Res. 29 12 , 3879–3890.

Novakowski, K.S., Lapcevic, P.A., 1994. Field measurement of radial solute transport in fractured rock. WaterŽ .Resour. Res. 30 1 , 37–44.

Novakowski, K.S., van der Kamp, G., 1996. The radial diffusion method: 2. A semianalytical model for theŽ .determination of effective diffusion coefficients, porosity, and adsorption. Water Resour. Res. 32 6 ,

1823–1830.Ž .Park, W., Schot, E.H., 1968. Stylolitization in carbonate rocks. In: Muller, G., Friedman, G.M. Eds. , Recent

Developments in Carbonate Sedimentology in Central Europe. Springer, New York, NY, 66–74.Pavlostathis, S.G., Jaglal, K., 1991. Desorption behavior of trichloroethylene in contaminated soil. Environ.

Ž .Sci. Technol. 25 2 , 274–279.Radcliffe, A.J., Analysis of Stylolitization and Fracturing in Rock Core from the Smithville Phase IV Site.

Unpublished thesis, Univ. of Waterloo, 1995.Sahoo, D., Smith, J.A., 1997. Enhanced trichloroethene desorption from long-term contaminated soil using

Ž .Triton X-100 and pH increases. Environ. Sci. Technol. 31 7 , 1910–1915.Shackelford, C.D., 1991. Laboratory diffusion testing for waste disposal — a review. J. Contam. Hydrol. 7,

177–217.Smart, P.L., Laidlaw, I.M.S., 1977. An evaluation of some fluorescent dyes for water tracing. Water Resour.

Ž .Res. 13 1 , 15–33.Spitz, K.-H., Moreno, J., 1996. A Practical Guide to Groundwater and Solute Transport Modeling. Wiley-In-

terscience, New York.Stokes, R.H., 1950. The diffusion coefficients of eight uni-univalent electrolytes in aqueous solution. Am.

Chem. Soc. J. 72, 2243–2247.U.S. Environmental Protection Agency. Technical background document to support rulemaking pursuant to the

Ž .Clean Air Act — Section 122 g . Ranking of pollutants with respect to hazard to human health.EPA-450r3-92-010. Emissions Standard Division, Office of Air Quality Planning and Standards, ResearchTriangle Park, NC, 1994.

Verschueren, K., 1983. Handbook of Environmental Data on Organic Chemicals, 2nd edn. Van NostrandReinhold, New York, 1310 pp.

Zytner, R.G., 1992. Adsorption–desorption of trichloroethylene in granular media. Water Air Soil Pollut. 65Ž .3–4 , 245–255.

sorption of trichloroethene onto stylolites

Documents