Altered Transport of Lindane Caused by the Retention of Natural Particles in Saturated Porous Media St´ephane K. Ngueleu, Peter Grathwohl, Olaf A. Cirpka PII: DOI: Reference:

S0169-7722(14)00062-X doi: 10.1016/j.jconhyd.2014.05.002 CONHYD 2995

To appear in:

Journal of Contaminant Hydrology

Received date: Revised date: Accepted date:

30 October 2013 5 May 2014 5 May 2014

Please cite this article as: Ngueleu, St´ephane K., Grathwohl, Peter, Cirpka, Olaf A., Altered Transport of Lindane Caused by the Retention of Natural Particles in Saturated Porous Media, Journal of Contaminant Hydrology (2014), doi: 10.1016/j.jconhyd.2014.05.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Altered Transport of Lindane Caused by the Retention of Natural Particles in Saturated Porous Media

IP

T

Stéphane K. Ngueleu, Peter Grathwohl, Olaf A. Cirpka*

SC R

Center for Applied Geoscience, University of Tübingen, Hölderlinstrasse 12, 72074 Tübingen, Germany

NU

*Corresponding author E-mail: [email protected]; Tel: +49-(0)7071-29 78928;

AC

CE P

TE

D

MA

Fax: +49-(0)7071-29 5059.

1

ACCEPTED MANUSCRIPT Abstract Attachment and straining of colloidal particles in porous media result in their

T

reversible and irreversible retention. The retained particles may either increase the retention of

IP

hydrophobic pollutants by sorption onto the particles, or enhance pollutant transport when

SC R

particles, loaded with the pollutants, are remobilized. The present study examines the effects of retained particles on the transport of the hydrophobic pesticide lindane (gammahexachlorocyclohexane) in saturated porous media. The lignite particles used have median

NU

diameters of about 3 µm, 1 µm, 0.8 µm, and 0.2 µm, respectively. Laboratory column

MA

experiments were analyzed by numerical modeling in order to identify and understand the processes involved in the transport of the particles and of lindane. Four scenarios were

D

considered in which the solution containing lindane is injected either during or after the

TE

elution of the particles. The results show that lignite particles retained in a sandy porous medium alter the transport of the invading lindane. Particle retention was high in all scenarios

CE P

and increased with increasing particle size. Remobilization of particles occurred due to a change in solution chemistry, and continuous particle detachment was observed over time.

AC

Numerical modeling of particle transport suggests that both reversible attachment and irreversible straining affected the transport of the particles. Lindane was retarded in all scenarios due to the strong particle retention in conjunction with the sorption of lindane onto the sand and onto retained particles, and the limited number of mobile particles carrying lindane. Moreover, it was found that intra-particle diffusion limited adsorption/desorption of lindane onto/from both limestone fragments of the sand and lignite particles. We assume that retention of lindane is reversible even though lindane recovery was incomplete over the duration of the experiments. The analysis of the effluent concentration suggests that retained particles loaded with lindane may become a secondary source of lindane. Models describing the transport of lindane fitted the experimental data very well and indicated the specific contribution of retained particles to the retardation of lindane. Since the properties of lignite 2

ACCEPTED MANUSCRIPT also known as brown coal are similar to those of biochar, the results of the present study could be extended to the potential effects of biochar on lindane and other contaminants in soils,

T

which would include both their retention and their enhanced transport. However, while the

IP

transport mechanisms of lindane are similar in water-unsaturated soils and saturated porous

SC R

media considered here, the behavior of particles is more complex, requiring additional

AC

CE P

TE

D

MA

NU

studies.

3

ACCEPTED MANUSCRIPT 1

Introduction Natural colloidal particles are ubiquitous in subsurface environments. Their

T

concentration in groundwater can range from 108 to 1017 particles per liter of water (Kim,

IP

1989; Kim, 1991). The evidence that colloidal particles can facilitate the transport of

SC R

pollutants in porous media has been shown by many studies (e.g., Fang et al., 2011; Kersting et al., 1999; McCarthy and Zachara, 1989; McDowell-Boyer et al., 1986; Ngueleu et al., 2013; Yin et al., 2010). A potential consequence of colloidal transport on pollutant behavior is

NU

that the mobility of the pollutant may be enhanced leading to shorter travel times, larger

MA

transport distances, or higher concentrations of the pollutant in the effluent, as compared to the case where no particles are involved. Trapped colloids, however, may cause the opposite –

D

retardation – because of their high sorption capacities.

TE

Transport of colloid-sized particles through porous media has been studied mainly using the colloid filtration theory (CFT) (e.g., Friedlander, 1958; Hunt et al., 1987; O'Melia,

CE P

1980), which describes the processes of colloid deposition onto the grains of porous media such as attachment, filtration, and size exclusion (Happel, 1958; Nicholson and Petropoulos,

AC

1968; Nicholson and Petropoulos, 1971; Nicholson and Petropoulos, 1973; Nicholson and Petropoulos, 1975; Payatakes et al., 1974; Rajagopalan and Tien, 1976). Specific models may include trapping of particles at water-air interfaces in unsaturated soils (Simunek et al., 2006) or ionic-strength effects on particle coagulation, disintegration, attachment and detachment (Tosco and Sethi, 2009; Tosco et al., 2009). Many studies have shown the necessity to understand the deposition and transport of colloids in underground systems for the assessment and prediction of colloid-facilitated transport of pollutants (e.g., DeNovio et al., 2004; Grolimund et al., 2007; Grolimund et al., 1996; Kretzschmar et al., 1999; McCarthy and McKay, 2004; Saiers and Hornberger, 1996; Zhuang et al., 2003; Zhuang et al., 2009). They contributed in giving the fundamental

4

ACCEPTED MANUSCRIPT information on the transport of colloidal particles and their role in pollutant spreading through the subsurface.

T

Transport and deposition of engineered nanoparticles could be well predicted from

IP

their zeta-potential, whereas this was not the case for natural particles (Hydutsky et al., 2007;

SC R

Wang and Keller, 2009). Size exclusion, restricted sampling of the velocity profile within the pores, or transport along water-air interfaces can lead to particle transport faster than the mean velocity of water (e.g., Panfilov et al., 2008). Several studies have highlighted the effects of

NU

ionic strength, pH, and colloid size on colloid transport (Bradford et al., 2006; Pelley and

MA

Tufenkji, 2008; Roy and Dzombak, 1997; Saleh et al., 2008; Zvikelsky and Weisbrod, 2006). Controlled laboratory experiments on the transport of colloidal particles and associated chemical agents were done with latex particles (Pelley and Tufenkji, 2008; Wan

TE

D

and Wilson, 1994b; Zvikelsky and Weisbrod, 2006; Zvikelsky et al., 2008), C60 fullerenes (Lecoanet et al., 2004; Wang et al., 2008; Xie et al., 2008), (surface-modified) zero-valent

CE P

iron nanoparticles (Saleh et al., 2008; Tiraferri and Sethi, 2009), as well as natural particles (Bold et al., 2003; Ngueleu et al., 2013). Optical detection was achieved by using fluorescent

AC

particle surfaces (Burkhardt et al., 2008; Zvikelsky and Weisbrod, 2006), or by matching the optical refraction index of the porous-matrix material and the fluid (Lachhab et al., 2008; Moroni et al., 2007).

The effect of dissolved particles on the transport of organic pollutants was illustrated by Magee et al. (1991) who showed that dissolved organic matter derived from soil enhanced the transport of phenanthrene in sandy porous media. Enhanced transport of phenanthrene by colloids released from natural sand was also shown by Roy and Dzombak (1998). The effect of groundwater flow velocity on the transport of natural lignite and activated carbon particles and trichloroethylene was studied by Bold et al. (2003) who showed simultaneous decrease of particles and trichloroethylene concentrations with the velocity. The latter results was in agreement with filtration theory which states that the decrease of flow velocity increases 5

ACCEPTED MANUSCRIPT particle interception and sedimentation to the immobile matrix (McDowell-Boyer et al., 1986). Cheng et al. (2005) showed that nano-C60 particles contained in soil affected

T

naphthalene transport by contributing to its retention. The contribution of colloid-facilitated

SC R

Jonge et al. (2000) for particles with diameter > 0.24 µm.

IP

transport to the transport of glyphosate in undisturbed topsoil columns was shown by De

The following conditions are necessary for colloid-facilitated transport of pollutants to be relevant: 1) colloid retention in the porous medium is weak or reversible; 2) the mobility of

NU

colloids is larger than that of the pollutant; 3) the adsorption of the pollutant onto the colloids

MA

is stronger than adsorption onto the matrix of the porous medium (Magee et al., 1991; McCarthy and Zachara, 1989; Mills et al., 1991; Roy and Dzombak, 1998). The latter highly depends on the size of the particles as it has been shown that particle retention increases with

TE

D

particle size (Hahn et al., 2004; Pelley and Tufenkji, 2008). Previous studies have shown that in water-saturated porous media, particles retained in

CE P

the pore space can be remobilized as a result of change in solution chemistry, an example being the reduction of ionic strength which causes an increase of inter-particle repulsion

AC

(Lenhart and Saiers, 2003; McDowell-Boyer, 1992; Roy and Dzombak, 1996; Roy and Dzombak, 1997; Ryan and Gschwend, 1994a; Ryan and Gschwend, 1994b; Tosco et al., 2009). In polluted porous aquifers, this remobilization of particles can contribute to pollutant removal and was suggested as groundwater remediation technique, especially for pollutants that would be difficult to remove by conventional pump-and-treat (McCarthy, 1993). However, Bradford et al. (2002) showed that so-called straining, that is, the irreversible trapping of particles in pore throats, can limit this type of remobilization. Large particles are more susceptible to straining than small ones. The second mechanism of particle retention in porous media is the reversible attachment of the particles onto the surfaces of the matrix. Understanding the retention processes of the particles is a prerequisite to study particlefacilitated transport. 6

ACCEPTED MANUSCRIPT Particle-facilitated transport has been studied for various organic and inorganic pollutants. In the present study, the specific case of lindane is investigated. Lindane (gamma-

T

hexachlorocyclohexane) is a representative hydrophobic insecticide which has been banned in

IP

2009 but still occurs as diffuse pollution globally and is still in use in many developing

SC R

countries, such as Bangladesh, Nepal, Sri Lanka, and India, among others (Bashir et al., 2013; IPEN, 2006; Pozo et al., 2011). Chen and Strevett (2001) investigated the impact of biocolloids (bacterial extracellular polymers) on the transport of lindane. Ngueleu et al.

NU

(2013) focused on the transport of lindane in the presence of organic colloidal particles. A

MA

major result of the latter study was that lignite particles with a size > 0.45 µm were strongly retained in the sandy porous media used, thus limiting the accelerated transport of lindane. The present study extends the work of Ngueleu et al. (2013) to improve our understanding on

TE

D

how the transport of lindane is altered by particle retention, which may be usefull for remediation of groundwater contaminated with lindane.

CE P

Several studies on particle-facilitated transport in saturated porous media have been based on simultaneous injection of particles and pollutants. Some studies investigated a

AC

scenario where sorption of the pollutant in a porous medium is followed by infiltration (Sun et al., 2010) or mobilization of particles (Roy and Dzombak, 1997), resulting in particlefacilitated transport of pollutants. In this study, by contrast, the injection of particles precedes the injection of lindane in order to study the effect of particle retention on the transport of lindane. The questions to be addressed in this study are: To which extent does the retention of clean particles previously transported through the system lead to increased retardation or even irreversible retention of invading lindane? And to which extent does the retention of previously transported particles loaded with lindane lead to secondary lindane sources? Thus, in contrast to Ngueleu et al. (2013) where lignite particles and lindane were jointly injected into water-saturated sand columns, the injection of lignite particles and lindane is not 7

ACCEPTED MANUSCRIPT synchronous. For the interpretation of the experimental data, we simulate the experiments by numerical modeling to understand processes relevant for particle and lindane transport and

Materials and Methods

SC R

2

IP

T

quantify parameters relevant for the retention of both lignite particles and lindane.

2.1 Porous Medium and Particles

NU

The sand used was collected from a glacio-fluvial sedimentary deposit located in Tübingen-Hirschau (South-West Germany). The original sediment consisted of 70% sand and

MA

30% gravel (DIN 18123), and the mean grain diameter d 50 was determined to be 0.001 m. The middle sand fraction with a grain size ranging from 0.25 mm to 0.80 mm was sieved out

D

and used as porous medium for the experiments.

TE

The sand was treated as recommended by Wan and Wilson (1994a). However, no boiling with nitric acid was performed and drying in the oven was done at 60°C as the goal of

CE P

the treatment was the removal of fine particles from sand surfaces while keeping the mineralogy and organic-carbon content of the remaining grains close to natural conditions.

AC

Basically, the sand was first washed in a 1.0% sodium diphosphate solution to detach fine particles from sand surfaces, and then sonicated for 20 minutes. The sonicated sand was batch-washed with distilled water until the pH reached that of distilled water. Afterwards it was sonicated in distilled water for 20 minutes, then batch-washed again with distilled water more than five times. It was finally dried in the oven at 60°C for 24 hours and stored in a glass container. The properties of the sand are listed in Table 1 from which it is interpreted that the sand has two types of particles, namely (1) quartz particles because of the particle density showing that quartz dominates in the sand, and (2) carbonate particles because of the high weight percent of CaCO3. Previous studies indicated that the minerals are mainly calcite, a small fraction of dolomite, quartz and feldspars (Kleineidam et al., 1999a). 8

ACCEPTED MANUSCRIPT Table 1: Properties of the sand and lignite particles. # Organic carbon content foc #Particle density ρ s [weight%] [g cm-3] Sand 0.06 2.66 Lignite particles 60.5 1.32

CaCO3 [weight%] 38.4 1.2

#

IP

T

Ngueleu et al. (2013)

SC R

Lignite particles were used as colloids because they are a typical, very strong natural sorbent, ensuring effective uptake of lindane (Kleineidam et al., 2002). Natural lignite was obtained from a lignite pit in Lausitz (East Germany) and dry pulverized with a zirconium

NU

oxide planet ball mill (Laborette, Fritsch) for about 40 minutes. The pulverization produced a

MA

lignite powder with a ≥ 80% mass fraction of grains smaller than 63 µm (Kleineidam et al., 1999b). The properties of the pulverized lignite are also listed in Table 1. The specific surface

D

area of lignite particles produced in a similar way was determined to be 1.78 m2 g-1 by

CE P

2.2 Chemicals

TE

Kleineidam et al. (2002).

Lindane, C6H6Cl6, was obtained at 99.5% purity in crystal form from Riedel de Haeen

AC

Chemicals, Germany. Its solid and subcooled liquid solubilities in water are 10 mg l-1 (IARC, 1979) and 56.7 mg l-1 (Schwarzenbach et al., 1993), respectively. Its aqueous diffusion coefficient Daq at 20°C was determined to be 5.2×10-10 m2 s-1 (Ngueleu et al., 2013). A stock solution of lindane was prepared by using methanol as solvent with a lindane concentration of 1.5 g l-1. The analysis of lindane in water samples extracted with cyclohexane was carried out by gas chromatography - mass spectrometry (GC-MS) using pentachlorobenzene as internal standard. The detection limit for lindane in water was about 0.005 mg l-1 and the average relative measurement error was in the order of 10%. Sodium chloride was used as tracer for conservative transport experiments with the purpose of determining the porosity of the porous medium. A concentration of 30 mg l-1 was

9

ACCEPTED MANUSCRIPT prepared with millipore water for each experiment. The calculated aqueous diffusion

T

coefficient Daq of chloride after Worch (1993) is 1.58×10-9 m2 s-1 at 20°C.

IP

2.3 Column Experiments

SC R

The experimental setup (Figure 1) was similar to the one used by Ngueleu et al. (2013). The column was set up in a manner that minimizes sorption of the compounds onto

NU

the equipment. For this reason, glass columns (length: 15 cm, inner diameter: 2.4 cm) and stainless steel tubing were used. To avoid artifacts due to sorption of lindane onto the polymer

MA

tubing, the peristaltic pump was installed downstream of the column. The flow rate in all column experiments was 8.33×10-4 ml s-1. It was controlled by the peristaltic pump but

D

induced by the head difference between the elevated reservoir containing the influent solution

TE

and the column. The column experiments were run in a temperature-controlled room maintained at 20°C in the dark.

CE P

The influent solution flowed through a stainless steel tube, upward through the column to assure saturated conditions, and through a second stainless steel tube. The effluent solution

AC

was sampled for lindane at the first T-valve (v1), placed before the peristaltic pump, and for lignite at the collector (COL). In order to ensure uniform flow through the column while reducing the dead volume of the system, we placed quartz frits (average mesh size: 205 µm) inside the caps of the column exits. These quartz frits also served as support for the sand directly at the column edges. For volume related calculations, we subtracted the entire dead volume (the volume of the tubing and the space inside the caps of the column) from the total volume (volume of the tubing, the column and the space inside the caps of the column) to obtain the inner, sand filled volume of the column. A lindane-tracer test of the empty system showed that sorption of lindane onto the column material was negligible. 10

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

MA

Figure 1: Experimental setup of the column experiment. I: influent solution, C: Glass column, v1: valve for effluent sampling, v2: on/off-valve, ST: sampling tray, COL: collector for particles. Arrows show the flow direction.

D

The column was packed with the dry sand in approximately 5 layers with vigorous

TE

shaking between placing of each layer. Subsequently the column was slowly percolated with

CE P

millipore water from the bottom to the top in order to avoid entrapment of air. Steady-state flow was applied in all experiments. Injection of particles and lindane was achieved by following the (continuous) step-input method.

AC

Except for the filtered sample (denoted as LIG2) containing particles smaller than 0.45 µm, we use the same lignite particles as Ngueleu et al. (2013). The filtered sample was excluded because these small particles behaved like a conservative tracer in the columns, whereas the present study focuses on the effects of larger, retained particles on the transport of lindane. The two types of lignite particles chosen are denoted hereafter as LIG0 and LIG1, respectively. Basically, the raw sample LIG0 was obtained by well mixing between 0.8 g and 1.6 g of pulverized lignite in 200 ml of millipore water, whereas the sample with finer particles LIG1 was obtained by letting particles of the raw sample settle over six hours and extracting the supernatant. Because the median particle diameters (d50) were not identical for

11

ACCEPTED MANUSCRIPT each batch of particles, subscripts a, b, c and d were used to differentiate the denotations of particles (Table 2).

T

The pH of the solutions containing lignite was 5.7 and similar to that of millipore

IP

water which was used as background solution and influent solution in the elution phase. The

SC R

pH of the influent solutions containing lignite was measured during mixing of the solutions at time intervals of thirty minutes to one hour for at least three hours for LIG0 (the raw sample) and after six-hour waiting time of settlement for LIG1 (the fine sample) to check for any

NU

change of pH prior to the beginning of the experiment. The influent solutions were closed

MA

with aluminum foil to reduce the effect of CO2 in the air on the pH. The stability of the pH was confirmed by running a column experiment where only millipore water was continuously injected through the sandy porous medium, and the measurements performed in the influent

TE

D

and effluent solutions showed that the pH did not change. Carbonate minerals in the sand used consist of minerals derived from limestone and

CE P

are mainly calcite minerals (Kleineidam et al., 1999a; Schüth, 1994). Dissolution rates of carbonate minerals including calcite (CaCO3, ZnCO3, MnCO3, FeCO3, and MgCO3) at 25 °C

AC

and at varying pH were determined in previous studies showing that they decrease with increasing pH (Duckworth and Martin, 2004; Sjöberg and Rickard, 1984). In general, at acidic pH, a sharp decrease of the dissolution rates of carbonate minerals occur and become stable at circumneutral pH. For calcite, the pH value at which a stable value of the dissolution rate is reached is 5.5 and the stability comes from the fact that beyond this value, the dissolution rate is independent of the H+ concentration (Sjöberg and Rickard, 1984). The values of the dissolution rates of carbonate minerals at circumneutral pH (>5.5 to about 8) was found to be very small as they range between 10-5.5 and 10-9.1 mol m-2 s-1 (Duckworth and Martin, 2004; Sjöberg and Rickard, 1984). In the present study, the fact that the pH did not change during the column experiment using only millipore water suggests that the dissolution of carbonate minerals of the sand was negligible (as pH = 5.7). 12

ACCEPTED MANUSCRIPT The ionic strength of the influent solutions was not measured. Only a qualitative assumption could be made about the possible difference between the solutions used. It was

T

assumed that influent solutions containing lignite had a higher ionic strength than millipore

IP

water used in the elution phase (justification in section 3). This may be relevant to interpret

SC R

possible particle aggregation and immobilization (e.g., attachment) at high ionic strength, and remobilization at low ionic strength (Roy and Dzombak, 1996). Straining also becomes important if particle aggregation is significant.

NU

Four scenarios were defined to study the effect of particles retained in porous media

MA

on the transport of lindane. These scenarios were organized in three to four phases listed in Table 2.

AC

CE P

TE

D

Table 2: Phases of the experimental scenarios considered in this study. Scenario Phase number Pore volumes Column Influent 1 6.2 1578.90×10-6 kg l-1 of LIG1a (¶d50 = 0.2 µm) 2 5.9 Millipore water A 3 7.6 1.66 mg l-1 of lindane 4 7.8 Millipore water 1 6.1 7961.00×10-6 kg l-1 of LIG0b (¶d50 = 1 µm) 2 5.9 Millipore water B 3 12.8 1.56 mg l-1 of lindane 4 7.9 Millipore water 1 6.1 4057.55×10-6 kg l-1 of LIG0c (¶d50 = 3 µm) C 2 12.7 0.76 mg l-1 of lindane 3 9.4 Millipore water 725.00×10-6 kg l-1 of LIG1d (¶d50 = 0.8 µm) 1 6.4 0.67 mg l-1 of lindane 2 8.3 Millipore water D 3 13.3 0.69 mg l-1 of lindane 4 9.2 Millipore water ¶

Here d50 is the median diameter of lignite particles that was determined based on the number of these particles in the influent solution.

Scenarios A and B differ only by the size of the particles. In these scenarios, the injection of clean lignite particles is successively followed by the injection of millipore water, the injection of a lindane solution, and finally the elution of lindane with millipore water.

13

ACCEPTED MANUSCRIPT Scenario C considers a situation where clean LIG0 particles are injected first, immediately followed by the injection of lindane; and finally lindane is eluted with millipore

T

water. In contrast to scenario B, where LIG0 particles are also used, the elution of particles

IP

and the injection of lindane overlapped.

SC R

Scenario D considers a situation where LIG1 particles in equilibrium with lindane were first injected and eluted, followed by the injection and elution of only lindane. Equilibrium between the LIG1 particles and lindane was obtained by mixing them for 24

NU

hours in a temperature-controlled room maintained at 20°C before the injection; the liquid-to-

MA

solid ratio was about 1353 l kg-1 (230 ml of liquid and about 0.17 g of media). 24-hour shaking time was previously determined to be sufficient to reach equilibrium for small sized particles (Chen and Strevett, 2001). Ngueleu et al. (2013) showed that the sorption of lindane

TE

D

onto lignite approximately follows a Freundlich isotherm with a Freundlich distribution coefficient KFr of 707 ± 18 mg

11/n Fr

l1/n Fr kg 1 and a Freundlich exponent 1/nFr of 0.72 ±

CE P

0.02. These parameter values were used as initial guesses in modeling. A laser particle analyzer (Mastersizer 2000) was used to measure the general particle

AC

size distributions of the LIG0 and LIG1 particles used in this study (Figure 2). The size of the LIG1 particles ranged between 0.1 µm and 2188 µm, with only very few particles > 26 µm (Figure 2; scenarios A and D). The size of the LIG0 particles ranged between 0.8 µm and 1096 µm but with only a few particles > 138 µm (Figure 2; scenarios B and C). Table 2 lists their median diameters (d50).

14

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

D

Figure 2: Size distribution of lignite particles LIG0 and LIG1 measured in column influent with a laser particle analyzer (Mastersizer 2000).

TE

Prior to each non-conservative transport experiment using lindane and/or particles, we

CE P

performed a conservative transport experiment with sodium chloride as tracer. We sampled the effluent at defined time intervals and measured the concentration of chloride by IonChromatography (Dionex DX-120, Germany). Figure 3 shows the breakthrough curves

AC

(BTCs) of chloride. As a sufficiently low dispersivity was expected, these BTCs of chloride were fitted by nonlinear least-squares fitting to the following analytical expression (for conditions of validity see Ogata and Banks, 1961):

 x  vt  C 1  erfc  , Cin 2  4 Dt 

(1)

where C [mg l-1] and Cin [mg l-1] are the aqueous-phase and inflow concentrations, respectively, x [m] the spatial coordinate; v [m s-1] the seepage velocity; t [s] the time, and D [m2 s-1] the hydrodynamic dispersion coefficient. All variables with their units are also summarized in the list of abbreviations in the Appendix.

15

ACCEPTED MANUSCRIPT The fitting parameters of equation 1 were the seepage velocity v and the hydrodynamic dispersion coefficient D, whose values were determined to be 5.7×10-6 m s-1

T

and 1.2×10-8 m2 s-1, respectively. The porosity was afterward calculated by dividing the

IP

specific discharge q (flow rate divided by the cross sectional area of the column) by v, giving

SC R

a value of 0.32. The calculated dry bulk density of the sand was then 1.8 kg l-1. The column Péclet number could also be calculated using v, the column length L, and D (product of v and L, divided by D), giving a value of 71.2.

NU

After complete breakthrough, sodium chloride was totally eluted with millipore water

MA

from the column, requiring the same number of pore volumes to flush it out as to saturate the

AC

CE P

TE

D

porous medium, which confirms that sodium chloride behaved like a conservative tracer.

Figure 3: Transport of chloride and best model fit. Scenarios A (injection of fine particles (LIG1a), followed by millipore water and lindane), B (injection of raw particle mixture (LIG0b), followed by millipore water and lindane), C (injection of raw particle mixture (LIG0c), immediately followed by lindane), and D (injection of fine particles (LIG1d) loaded with lindane, followed by millipore water and lindane) refer to particle/lindane tests following the conservative-tracer tests.

In experiments with lindane and/or particles, influent solutions were continuously mixed throughout the experiments with glass magnet stirrers in order to keep them homogeneous. For the transport of lignite particles alone, about 5 ml of effluent samples were 16

ACCEPTED MANUSCRIPT collected at the collector (COL) between defined time intervals. Lignite in these samples was quantified by measuring its turbidity in Nephelometric Turbidity Units (NTU) using a HACH

T

2100 Turbidimeter. A linear relationship between turbidity and mass concentration of lignite

IP

was determined and allowed the conversion of turbidity measurements into mass

SC R

concentrations (1 mg/l ≡ 5.7 NTU; turbidity ranged from 0 to 2144 NTU). In experiments with lindane alone, 2-3 ml of the effluent were sampled at the first valve (v1) between defined time intervals. Subsequently, 1 ml was used to measure the

NU

concentration of lindane. These 1-ml samples were injected into 5-ml vials containing 5 µl of

MA

pentachlorobenzene (500 mg l-1) and 1 ml of cyclohexane. A vortex shaker was used to vigorously shake the mixture for 1 min. The mixture was later left at rest for more than 5 min to allow the aqueous and organic phases to separate. The upper phase of the mixture

TE

D

containing cyclohexane, lindane and pentachlorobenzene was finally transferred into 0.3-ml automatic sampler vials for lindane analysis by GC-MS.

CE P

In experiments involving both lignite and lindane, sampling was done as described above. In this case, the total concentration of lindane (mass of lindane, both dissolved and

AC

particle-bound, per volume of water) was measured as explained above for lindane alone and the concentration of lignite was measured via turbidity measurements as explained above. Singular column experiments were run because Ngueleu et al. (2013), who used the same experimental setup, demonstrated high reproducibility of their BTCs at similar flow velocities (8.33×10-4 ml s-1). Only the average relative measurement error of the GC-MS of 10% was considered here in all BTCs of lindane.

2.4

Model Formulation Two one-dimensional models were used in this study. The first one denoted hereafter

as M1 couples the transport of lignite particles and lindane by following the approach of Ngueleu et al. (2013), which is illustrated in Figure 4. Sorption kinetics in this model is 17

ACCEPTED MANUSCRIPT considered to follow the linear-driving-force model for the interactions of both between

MA

NU

SC R

IP

T

lindane and the sand, and lindane and the lignite particles.

TE

D

Figure 4: Graphical illustration of particle-facilitated transport of sorbing compounds in porous media (Ngueleu et al., 2013).

CE P

The underlying assumptions of model M1 are that (1) the porous medium is water saturated, (2) size exclusion is neglected, (3) the velocity of the particles is identical to that of water, (4) chemical or biological transformations are neglected, and (5) mass-transfer fluxes

AC

of lindane between water and kinetic sorption sites are proportional to the deviation between the actual sorbing-phase concentration and that in sorption equilibrium. These assumptions result in the following system of equations (Ngueleu et al., 2013; Simunek et al., 2006):

ρ C  2C C D 2 v   b r1  r2  r6  r7   r5 , t x n x 2 C p  Cp C p ρ D v   b r3  r4 , 2 t x n x  C p S mp   2 C p S mp   C p S mp  ρ D  v   b r3 p  r4 p   r5 , 2 t x n x  S str S strp   r3 p  r7 , t  S p S ip   r4 p  r6 , t

Req

in which 18

(a) (b) (c) (d) (e)

(2)

ACCEPTED MANUSCRIPT ρb K d ,eq , n

(3)

r1 

S1 n  k1 C  k 2 S1 , t ρb

r2 

S 2 n  k3 C  k4 S2 , t ρb

T IP



  k strC p , 

r4  k ap

n C p  k dp S p , ρb

n r3 p  ρb

 d 50  X   d 50



  k str C p S mp , 

(7)

(8)



(10)



(11)

D

(9)



CE P

r6  k p S p K Fr C 1 / nFr  S p S ip ,



(6)

n C p S mp  k dp S p S ip , ρb

r5  k p C p K Fr C 1 / nFr  C p S mp ,



(5)

TE

r4 p  k ap

SC R

 d 50  X   d 50

MA

n r3  ρb

(4)

NU

Req  1 



r7  k p S str K Fr C 1 / nFr  S str S strp .

AC

(12)

The new variables in the above equations are Req [-] the retardation factor due to equilibrium sorption onto the sand matrix;  b [kg l-1] the bulk density of the sand matrix; n [] the porosity of the sand matrix; Cp [kg l-1] the concentration of mobile particles; Smp [mg kg1

], Sip [mg kg-1], and Sstrp [mg kg-1] the concentrations of lindane sorbed onto mobile,

reversibly attached, and irreversibly strained particles, respectively; Sp [kg kg-1] and Sstr [kg kg-1] the concentrations of reversibly attached and irreversibly strained particles, respectively; r1 [mg kg-1 s-1] and r2 [mg kg-1 s-1] the mass transfer rates of dissolved lindane onto kinetic

sites 1 and 2 of the matrix, respectively; r3 [kg kg-1 s-1] and r4 [kg kg-1 s-1] the straining and

19

ACCEPTED MANUSCRIPT net attachment rates of particles; r3 p [mg kg-1 s-1] and r4 p [mg kg-1 s-1] the mass transfer rates of lindane due to particle straining and to net particle attachment, respectively; r5 [mg ls-1], r6 [mg kg-1 s-1], and r7 [mg kg-1 s-1] the mass transfer rates of dissolved lindane onto

IP

T

1

SC R

mobile, reversibly attached, and irreversibly strained particles, respectively; K d ,eq [l kg-1] the linear distribution coefficient onto equilibrium sites of the sand matrix; S1 [mg kg-1] and S2 [mg kg-1] the concentrations of lindane sorbed onto kinetic sites 1 and 2 of the matrix,

NU

respectively; k1 [s-1] and k3 [s-1] the rate coefficients of adsorption of lindane onto matrix sites

MA

1 and 2 of the sand, respectively; k2 [s-1] and k4 [s-1] the rate coefficients of desorption of lindane from matrix sites 1 and 2 of the sand, respectively; X [m] the distance from the porous medium inlet used to limit the straining effect; β [-] is the fitting parameter that controls the

TE

D

shape of the particle spatial distribution; kap [s-1] and kdp [s-1] the first-order particle attachment and detachment coefficients, respectively; and k p [s-1] the first-order kinetic rate

CE P

coefficient of lindane adsorption/desorption onto/from particles. Equation 2 describes (a) the transport of lindane not bound to particles, (b) the

AC

transport of the particles, (c) the transport of lindane sorbed onto mobile particles, (d) the mass balance of lindane sorbed onto irreversibly strained particles, and (e) the mass balance of lindane sorbed onto reversibly attached particles, respectively. Like in Ngueleu et al. (2013), the sorption kinetics of lindane onto the sand is assumed to follow linear three-site sorption with one site in local equilibrium and the two others (site 1 and site 2) undergoing kinetic sorption. While local equilibrium is expressed by introduction of the equilibrium retardation factor Req , kinetic sorption is expressed by the mass transfer rates r1 and r2. The sorption of lindane onto the particles at equilibrium is expressed by the mass transfer rates r5, r6, and r7, and a common coefficient k p was used to define these mass transfer rates as Ngueleu et al. (2013) showed that a common first-order rate coefficient could 20

ACCEPTED MANUSCRIPT be enough to express the adsorption/desorption of lindane onto/from the particles of the same size. k p is a kinetic rate coefficient that does not affect equilibrium. The straining distance X

T

was set to 0.05 m after observing that lignite accumulated mainly over the first 5 cm of the

IP

column.

SC R

We used the finite volume method with upwind differentiation for spatial discretization, and defined nine state variables, namely the concentrations of dissolved

NU

lindane, mobile particles, and lindane on mobile particles as mobile compounds, and the kinetically sorbed lindane concentration on matrix site 1, the kinetically sorbed lindane

MA

concentration on matrix site 2, the concentration of strained particles, the concentration of attached particles, and the concentrations of lindane on strained particles and on attached

D

particles as immobile compounds. A fully implicit time integration method was applied to

TE

solve the system of equations 2a-e. The program is written as MATLAB application and the boundary conditions of Dirichlet and Neumann were applied at the column inlet (x=0) and

CE P

outlet (x=L), respectively (Ngueleu et al., 2013):

C x

AC

C C ( x  0, t )   in 0  0t;

xL

if 0  t  tin C ; C p ( x  0, t )   p ,in otherwise 0

C p x

 0t .

if 0  t  tin , otherwise

(13)

(14)

xL

where tin [s] is the time of injection, Cp,in [kg l-1] is the inflow concentration of the particles, and L [m] is the column length. The transport of lindane in the absence of particles was experimentally studied by Ngueleu et al. (2013) who fitted the BTC and determined the values of the parameters K d ,eq , k1 , k2 , k 3 and k 4 (Table 3). The same values were used in the present study since lindane

remains in contact with the same sand as well as with the particles during its transport. The remaining fitting parameters were k ap , k dp , k str , β, and k p . The Melder-Nead simplex 21

ACCEPTED MANUSCRIPT algorithm implemented in the MATLAB function fminsearch (Lagarias et al., 1998) was used to fit the parameters.

SC R

IP

T

Table 3: Fitted model parameters of lindane transport without particles through sand (Ngueleu et al., 2013). K d ,eq k3 RMSE k1 k2 k4 -1 -1 -1 -1 -1 [-] [l kg ] [s ] [s ] [s ] [s ] -6 -6 -6 -6 0.07 8.09×10 8.01×10 2.14×10 1.04×10 0.04

The first model could not fit the BTC of lindane in scenario B, leading to errors in the

NU

tails. This is an indication that the first-order mass-transfer kinetics for sorption assumed in model M1 was too simplistic. We thus used a second model, denoted hereafter as M2, to

MA

model the BTC of lindane in scenario B where strained LIG0b particles became part of the sandy porous medium. The model M2 couples equilibrium sorption and spherical intra-

D

particle diffusion. This modeling approach was previously used for example by Bold et al.

TE

(2003), Bold et al. (2005), and Wu and Gschwend (1988).

CE P

The choice of using intra-particle diffusion in scenario B was justified by a delayed arrival of the BTC of lindane at the column outlet compared to other scenarios. The numerical code of the model M2 is incorporated in the program SMART (Finkel et al., 1998) which was

AC

used for the simulations. SMART enabled to define a heterogeneous porous medium made of sand and lignite by defining their corresponding mass fraction in the column. This model uses Fick’s 2nd Law in radial coordinates to describe such transport:

  2 C 2 C  C  Da  2  , t r r   r

(15)

in which Da [m2 s-1] is the apparent diffusion coefficient and r [m] is the radial distance from the center of the particle. The model M2 considers intra-particle diffusion of lindane in lignite particles and the expression of Da is as follows (Rügner et al., 1999):

22

ACCEPTED MANUSCRIPT Da 

  

Daq p



(16)

K Fr 1 / nFr C 1 / nFr 1  1m

T

in which the new variables are ε [-] the intra-particle porosity,  p [kg l-1] the bulk density of

IP

the particle, and m [-] the empirical exponent of tortuosity.

SC R

The numerical code of SMART couples advection to equations 15 and 16. In SMART, sorption/desorption kinetics are described as intra-particle diffusion of chemicals inside the particles of the porous medium (Finkel et al., 2007). Therefore, sorption kinetics of lindane

NU

onto the sand that were described by the rate coefficients k1, k2, k3 and k4 (Table 3) could be

MA

added to the model by considering it as intra-particle diffusion. This is practically done in SMART by assigning a value smaller than 1 to the fraction of equilibrium sorption. As for the

D

equilibrium sorption of lindane onto the sand, the K d ,eq value of Table 3 was used.

TE

In order to compute Da, m was used as fitting parameter. Additional fitting parameters

CE P

were the mass fraction of each lithological component (sand and lignite) in the column and the fraction of equilibrium sorption. Intra-particle diffusion often is approximated by the calculation of a first order rate coefficient ( k diff ,calc ) defined e.g. as follows (Wu and

AC

Gschwend, 1988):

k diff ,calc  22.7

Da  d 50     2 

(17)

2

If solute adsorption/desorption onto/from the particle is limited by intra-particle diffusion, k diff ,calc should be similar to k p of model M1.

23

ACCEPTED MANUSCRIPT 3

Results and Discussion The effluent concentrations of lignite particles and lindane were normalized by their

T

corresponding inflow concentrations. The resulting normalized concentrations were plotted as

IP

a function of pore volumes in Figures 5-9. In the following, we first present the breakthrough

SC R

curves and model-based interpretation of each scenario, followed by a comparison of lindane BTCs in the absence (Ngueleu et al., 2013) and presence of retained particles.

NU

3.1 Scenario A: Injection of Fine Particles (LIG1a), Followed by Millipore

MA

Water and Lindane

The BTCs of both LIG1a particles and lindane are shown in Figure 5. The analysis of

D

the experimental data shows that LIG1a particles were strongly retained in the sand during

TE

phase 1. LIG1a particles arrived at the column outlet after about 1 pore volume and their concentration increased until about 2 pore volumes where a relatively stable value close to

CE P

20% of the inflow concentration was reached. At the beginning of the elution phase (phase 2), a sudden increase in the concentration

AC

of LIG1a particles occurred. We assume that this increase resulted from a lower ionic strength of the millipore water than the solution containing lignite, which was injected in phase 1. This observation is in agreement with previous studies (e.g., McDowell-Boyer, 1992; Roy and Dzombak, 1996) who observed that remobilization of colloidal particles in porous media occurred when they were exposed to a lower ionic strength solution at the same flow rate. A comprehensive interpretation is that particle remobilization was caused by the destabilization of the bonds between the particles themselves (disaggregation) and between the particles and the sand of the porous medium. Directly after the sudden increase of particle concentration, the descending limb of the particle elution curve followed. Although a fast decrease of the concentration is noticeable, a 24

ACCEPTED MANUSCRIPT value of zero was not reached. On the contrary, long tailing due to continuous particle detachment from the sand was observed. The concentration of LIG1a particles was not

T

measured after phase 2 in this scenario but we still observed that effluent samples in phases 3

IP

and 4 had a light brown color, indicating the presence of lignite particles at very small

SC R

concentrations. The latter could be addressed as dissolved organic carbon (DOC). In phase 3, the mean arrival time (at C/Cin = 0.5) of lindane corresponds to about 1.7 pore volumes after the beginning of lindane injection. Compared to conservative tracers with

NU

a mean arrival time of 1 pore volume, lindane was retarded due to sorption onto the sand and

MA

the retained particles. Moreover, although the concentration increased fast, the normalized concentration of lindane did not reach a value of 1 despite continuous injection over a long period of time. Ngueleu et al. (2013) attributed this effect to slow sorption kinetics onto the

TE

D

sand.

In phase 4, the concentration of lindane dropped but did not reach a value of zero. The

CE P

tail of the lindane elution curve was a result of continuous desorption of lindane from the sand and the retained particles.

AC

The BTCs of LIG1a particles and lindane were simulated with the model M1 and the results are also shown in Figure 5. Trials to fit these BTCs with the model M2 were not successful. The very good fits between the experimental data and the results of the model M1 are well appreciated by the small value of the normalized root mean square error (RMSE) listed in Tables 4 and 5. Note that the term “normalized” is used here because the RMSE was determined using the normalized concentrations.

25

TE D

MA N

US

CR

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 5: Transport of LIG1a particles (d50 = 0.2 µm) and lindane in scenario A. Open symbols represent normalized lignite concentrations, whereas black symbols represent normalized total (dissolved and particle-bound) lindane concentrations.

26

ACCEPTED MANUSCRIPT

NU

SC R

IP

T

Table 4: Fitted model parameters of particle transport in all scenarios. Scenario kap kdp kstr (particle size) [s-1] [s-1] [s-1] A 4.59×10-5 4.51×10-5 2.19×10-4 (LIG1a, d50 = 0.2 µm) B 1.06×10-4 6.51×10-5 2.79×10-4 (LIG0b, d50 = 1 µm) C 1.06×10-4 6.51×10-5 2.79×10-4 (LIG0c, d50 = 3 µm) D 4.59×10-5 4.51×10-5 2.19×10-4 (LIG1d, d50 = 0.8 µm)

β [-]

RMSE [-]

0.32

0.08

0.40

0.09

0.25

0.08

0.36

0.06

TE

D

MA

Table 5: Fitted model parameters of particle-bound lindane transport (kp) and diffusion rate coefficients (*kdiff,calc) for lignite particles in scenarios A, C and D. kp KFr 1/nFr RMSE Scenario -1 1  1/n 1/n  1 [s ] [-] [-] (particle size) mg Fr l Fr kg A 1.87×10-7 707 0.72 0.08 (LIG1a, d50 = 0.2 µm) C 6.10×10-7 707 0.72 0.08 (LIG0c, d50 = 3 µm) 707 0.72 0.14 D 1.02×10-4 (LIG1d, d50 = 0.8 µm) 420 0.72 0.08

calculated kdiff,calc [s-1] 2.78×10-3 1.16×10-5 1.83×10-4 3.08×10-4

AC

CE P

*kdiff,calc was calculated by equation 17. Da needed for kdiff,calc was calculated by equation 16 and by using the average concentration of lindane in the tail of the breakthrough curve (scenario A: average is 0.07 mg l-1 and range is 0.04-0.10 mg l1 ; scenario C: average is 0.05 mg l-1 and range is 0.04-0.06 mg l-1). A value of 0.0102 for ε was obtained from Wang and Grathwohl (2009) who also used lignite, after which, the bulk density of the particle was determined to be 1.31 kg l-1 by using the formula ρp = (1-ε)×particle density. A value of 2.75 was used for m based on the results of scenario B (section 3.2).

The values of the fitting parameters for the transport of LIG1a particles are identical to those determined by Ngueleu et al. (2013) for particles of the same size, except for β whose value is somewhat smaller (0.32 rather than 0.36, see Table 4). We believe this to be an effect of a slightly smaller maximum concentration than observed by Ngueleu et al. (2013). Particle straining was stronger than particle attachment. In addition to straining and attachment, the assumed higher ionic strength in the injection phase (phase 1), when compared to the elution phase (phase 2), may also have contributed to particle retention (Ben-Moshe et al., 2010). It is also possible that aggregation occurred due to the higher ionic strength in phase 1, increasing particle size which in turn increased particle retention by straining. 27

ACCEPTED MANUSCRIPT Concerning the transport of lindane, the fitted first-order kinetic rate coefficient kp of lindane adsorption/desorption onto/from LIG1a particles, as well as the Freundlich parameters

T

KFr and 1/nFr (Table 5), are identical to those determined by Ngueleu et al. (2013) although

IP

lindane and the particles were injected independently from each other.

SC R

In order to check whether intra-particle diffusion occurred in LIG1a particles and limited lindane adsorption/desorption to/from these particles, the diffusion rate coefficient kdiff,calc was calculated using the tail of the BTC similarly to Wang and Grathwohl (2009). It

NU

can be seen in Table 5 that the value of kdiff,calc is four orders of magnitude bigger than that of

MA

kp. This difference leads to the conclusion that adsorption/desorption of lindane onto/from LIG1a particles was not limited by intra-particle diffusion. Instead, the tailing observed in the BTC of lindane originated from slow desorption from the sand which most likely is limited by

TE

D

intraparticle diffusion in porous limestone particles. Overall sorption of lindane onto the sedimentary rock fragments was more pronounced than onto the trapped LIG1a particles.

CE P

Aside the simulated BTCs of LIG1a particles and total lindane, Figure 5 also contains the computed BTCs of dissolved lindane and particle-bound lindane. It can be seen that

AC

transport of dissolved lindane contributed to the total mass flux of lindane as its BTC did not differ from that of total lindane, whereas particle-bound transport of lindane did not occur as its BTC remained on the line C/Cin = 0. On the whole, retained LIG1a particles did neither irreversibly retain lindane nor highly retard it. The tailing of lindane BTC came from slow desorption of lindane from the sand. Section 3.5 which compares all BTCs with that of lindane in the absence of lignite particles also contains information supporting this statement.

28

ACCEPTED MANUSCRIPT 3.2 Scenario B: Injection of Raw Particle Mixture (LIG0b), Followed by Millipore Water and Lindane

T

The BTCs of LIG0b particles and lindane are shown in Figure 6. Similarly to the

IP

transport of LIG1a particles in scenario A, LIG0b particles were strongly retained within the

SC R

sand during phase 1. They were even more retained than LIG1a particles because it took more time to reach the highest concentration observed, corresponding to about 20% of the inflow concentration. This can be explained with the larger particle diameters of LIG0b in

NU

comparison to LIG1a, which is in agreement with previous studies showing that particle

MA

retention increases with particle size (e.g., Bradford et al., 2002; Hahn et al., 2004; Litton and Olson, 1996; Pelley and Tufenkji, 2008).

D

It is also noticeable that the concentration of LIG0b particles suddenly increased even

TE

more at the beginning of the elution phase (phase 2) as it was the case for LIG1a particles, most likely caused by the change in solution chemistry. The descending limb also had a steep

CE P

slope and a long tail due to the slow detachment of particles from the sand. Measurement of LIG0b particles was not done during phases 3 and 4 but it was possible to see that the effluent

AC

samples had a somewhat light brown color which was characteristic of lignite particles at a very small concentration. Like in scenario A, the latter could be addressed as DOC. In phase 3, the mean arrival time of lindane corresponds to about 3.5 pore volumes after the beginning of lindane injection. Obviously, lindane was more retarded in the presence of LIG0b particles than in that of LIG1a particles. This is due to the fact that LIG0b particles were more retained in the sand and could thus sorb more lindane than LIG1a particles. The maximum normalized concentration after several pore volumes was also smaller than 1, suggesting that slow sorption kinetics in the sand was limiting transport of lindane. In phase 4, lindane was desorbed from the sand and the particles. The decrease in concentration was fast before slowing down at small concentration values, indicating kinetic desorption effects with eventually intra-particle diffusion. 29

TE D

MA N

US

CR

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 6: Transport of LIG0b particles (d50 = 1 µm) and lindane in scenario B. Open symbols represent normalized lignite concentrations, whereas black symbols represent normalized total (dissolved and particle-bound) lindane concentrations.

30

ACCEPTED MANUSCRIPT The model M1 did not reproduce good fits when simulating the BTC of lindane, only the BTC of LIG0b particles could be simulated using equation 2b (Figure 6). A good model

T

fit for the BTC of lindane was obtained using the model M2 whose results are shown in

IP

Figure 6 and Table 6.

SC R

Concerning the transport of LIG0b particles, the values of the fitting parameters are identical to those determined by Ngueleu et al. (2013) (Table 4). Attachment and straining were both involved in particle retention, with straining having a slightly stronger effect.

NU

Similarly to LIG1a particles, the assumed ionic strength in the injection phase (phase 1) may

MA

also have contributed to the retention of LIG0b particles.

Sorption of lindane onto the sand and retained LIG0b particles combined with intraparticle diffusion in both carbonate particles of the sand and retained LIG0b particles were

TE

D

responsible of the delayed arrival of lindane. When compared to m and foc of the sand (Table 6), the values of m’ and foc’ suggest an increase of the organic carbon content of the sandy

CE P

porous medium due to the retention of LIG0b particles, favoring the sorption of lindane. On the whole, transport of lindane was retarded more strongly in the presence of

AC

retained LIG0b particles than in the presence of the finer particle fraction of LIG1a. Section 3.5 contains information on calculated percentages of retained particles, showing that a higher percentage of retained LIG0b particles than retained LIG1a particles was responsible of more lindane retardation.

31

ACCEPTED MANUSCRIPT Table 6: Model parameters for the simulation of lindane transport in scenario B with the model M2 and calculated parameters for the heterogeneous porous medium ( m'    f m i mi and f oc '    f m i  f oc i ). m’ and foc’ are the empirical exponent and the organic carbon content of the sand

matrix containing retained lignite particles, respectively.

0.07

-

-

707

[mg 11/nFr

Fraction of equilibrium sorption [-]

Mass fraction fm [%]

foc (weight%)

*ε [-]

*m [-]

-

0.8

99.92

0.06

0.0311

2

0.72

0.001

0.08

60.5

0.0102

2.75

1/nFr l 1/nFr kg 1 ] [-]

T

KFr

IP

Sand particles Lignite particles

Kd,eq [l kg-1]

SC R

Lithological components

m’ [-]

foc’ (weight%)

2

0.11

NU

*ε and m for the sand were obtained from Schüth (1994) where the sand is named as Neckar sand. ε for lignite was obtained from Wang and Grathwohl (2009).

3.3 Scenario C: Injection of Raw Particle Mixture (LIG0c), Immediately

MA

Followed by Lindane

In this scenario, the injection of millipore water between the particle and lindane

D

injections was skipped, facilitating overlap of the particle and lindane containing solutions at

TE

the replacement front. The BTCs of both LIG0c particles and lindane are shown in Figure 7.

CE P

During phases 1 and 2, it can be seen that the transport of LIG0c particles was similar to that of LIG0b in scenario B with the only difference that before the elution phase, the biggest

AC

measured concentration was smaller, namely about 10% of the inflow concentration. Elution with lindane-loaded millipore water also induced remobilization of retained particles. The BTC of lindane appeared during the elution (or remobilization) of LIG0c particles, giving the possibility to lindane to sorb onto mobile particles. The concentration of lindane increased quickly while there were still enough mobile particles present before slowing down after about 3 pore volumes after the beginning of lindane injection at very small concentrations of mobile particles. We interpret these findings such that strong sorption of lindane onto LIG0c particles occurred and that the increase of lindane concentration was induced by particle-facilitated transport. It can also be seen that the mean arrival time of lindane corresponds to about 2.2 pore volumes since the beginning of lindane injection. Lindane was thus less retarded than in 32

ACCEPTED MANUSCRIPT scenario B and this earlier arrival of lindane is due to its particle-facilitated transport. However, only about 75% of the inflow concentration of lindane was measured at the column

T

outlet even though the injection of lindane lasted for several pore volumes. This observation

IP

indicates that, aside slow sorption kinetics onto the sand, the trapping of lignite particles also

SC R

affected the transport of lindane.

During the elution phase of lindane (phase 3), desorption of lindane from the sand and the trapped lignite particles occurred simultaneously. Similarly to scenario B, a fast decrease

NU

of the concentration was observed, followed by a long tail of the BTC characteristic of slow

MA

desorption.

In contrast to model M2, model M1 reproduced the experimental BTCs very well (Figure 7a) and the RMSEs are very small (Tables 4 and 5). The fitted model parameters of

TE

D

LIG0c particles are similar to those of scenario B with the only exception that β was decreased from 0.40 to 0.25 (Table 4) to account for the small maximum concentration. We

CE P

explain this by the diameter of the particles which is bigger than that of particles used in scenario B, supporting the statement that particle retention increases with particle size

AC

(Bradford et al., 2002; Pelley and Tufenkji, 2008). In addition, the assumed ionic strength of the solution containing lignite may have contributed to the retention of LIG0c particles in the injection phase (phase 1).

33

AC

CE P

TE D

MA N

US

CR

IP

T

ACCEPTED MANUSCRIPT

Figure 7: Transport of LIG0c particles (d50 = 3 µm) and lindane in scenario C. Open symbols in (a) represent normalized lignite concentrations, whereas black symbols represent normalized total (dissolved and particle-bound) lindane concentrations. (b) is a zooming plot of (a) at small C/Cin values. 34

ACCEPTED MANUSCRIPT The model reproduced the experimental data of lindane well, especially in the ascending and descending limbs of the BTC. However, the exact shape of the BTC at the

T

plateau did not agree: the measured concentrations showed oscillations that could not be

IP

explained by the model. These oscillations may be caused by the mobilization of LIG0c

SC R

particles carrying lindane. The fitted kinetic rate coefficient kp is two orders of magnitude smaller than the diffusion rate coefficient kdiff,calc estimated for the lignite particles (Table 5). Again, intraparticle diffusion in the limestone fragments of the sand is responsible for slow

NU

desorption kinetics and dominated the tailing of lindane BTC.

MA

Transport of dissolved lindane contributed the most to the total mass flux of lindane. Unlike in scenarios A and B, particle-bound transport also contributed to the total mass flux and this is well visible in Figure 7b by the model BTC of particle-bound transport. Lindane

TE

D

was less retarded than in scenario B because of particle facilitated transport of lindane.

CE P

3.4 Scenario D: Injection of Fine Particles (LIG1d) Loaded with Lindane, Followed by Millipore Water and Lindane

AC

Figure 8 shows the BTCs of both LIG1d particles and lindane for scenario D, in which the lignite particles were pre-equilibrated with lindane prior to injection. After flushing the column with millipore water, a solution was injected that contained lindane. The analysis of the experimental data during phases 1 and 2 shows that the transport of LIG1d particles and lindane were very similar. It is evident that the transport of lindane was driven by the particles since lindane was in equilibrium with them prior to their injection. One can also notice that the arrival time of lindane at the column outlet was identical to that of the particles. During phase 2, the slope of the descending limb of the elution curve of lindane was gentle, reflecting the slow desorption of lindane from the sand and the trapped particles. A long tail was observable that even extended into phase 3. The particle concentration remained very small until the end of the experiment. 35

ACCEPTED MANUSCRIPT The mean arrival time of lindane during phase 3 corresponds to about 2.3 pore volumes. When comparing to scenario A, where LIG1 particles where also used, a secondary

T

source was not obtained. Although the concentration of lindane seemed to increase fast, the

IP

slope of the ascending limb decreased after about 2.3 pore volumes. The decrease of the slope

SC R

is a result of the decreasing concentration of mobile particles. Particles in equilibrium with lindane in phases 1 and 2 had a higher affinity with the new lindane of phase 3. As a consequence, particle retention also led to lindane sorption. A stable value of about 95% of

NU

the inflow concentration of lindane was later reached as a result of slow sorption kinetics of

MA

lindane onto the sand and LIG1d particles. During phase 4, elution with millipore water led to

AC

CE P

TE

D

first fast followed by slow lindane desorption after several pore volumes.

36

TE D

MA N

US

CR

IP

T

ACCEPTED MANUSCRIPT

11/n Fr

l1/n Fr kg 1 ] for the black and red breakthrough curves, respectively.

AC

KFr [mg

CE P

Figure 8: Transport of LIG1d particles (d50 = 0.8 µm) and lindane in scenario D. Open symbols represent normalized lignite concentrations, whereas black symbols represent normalized total (dissolved and particle-bound) lindane concentrations. Values of 707 and 420 were assigned to

37

ACCEPTED MANUSCRIPT No good model fit was obtained with the model M2, whereas the model M1 was able to reproduce the experimental data well (Figure 8), which is also substantiated by small

T

RMSEs (Tables 4 and 5). The parameter values describing the transport of LIG1d particles

IP

are all identical to those determined by Ngueleu et al. (2013) for particles with a d50 of 0.2 µm

SC R

which were prepared in the same manner as those of the current study. Straining dominates over reversible attachment. Similarly to the other scenarios, the assumed ionic strength of the solution containing lignite may have contributed to the retention of LIG1d particles in the

NU

injection phase (phase 1), which also reduces particle-facilitated transport of lindane in phase

MA

1.

Concerning the transport of lindane, the Freundlich coefficient K Fr was changed to 11/n Fr

l1/n Fr kg 1 in order to get good model fits (red BTCs in Figure 8). This latter

D

420 mg

TE

value is close to the lower bound of 689 mg

11/n Fr

l1/n Fr kg 1 of the equilibrium K Fr and

CE P

suggests that less sorption of lindane onto the particles occurred. It is however noticeable that in both cases (707 and 420), the fitted kinetic rate coefficient kp is in the same order of magnitude with the calculated diffusion rate coefficient kdiff,calc (Table 5), which indicates that

diffusion.

AC

adsorption/desorption of lindane onto/from LIG1d particles was limited by intra-particle

Analysis of the red BTCs shows that particle-bound lindane transport was dominated by dissolved lindane transport during phases 1 and 2. Afterward in phases 3 and 4, lindane was exclusively transported as dissolved lindane because the concentration of LIG1d particles was already very low. Considering the arrival time, the retention of particles loaded with lindane did not lead to a secondary source of lindane in the phase following particle elution. However, based on the concentration, accumulation of lindane prior to the second injection of lindane could result

38

ACCEPTED MANUSCRIPT to its significant release at late times. Section 3.5 contains information supporting the latter

T

interpretation.

IP

3.5 Comparison of Lindane Breakthrough Curves between the Scenarios

SC R

Figure 9a-b shows the BTCs of total lindane in the absence and presence of particles as obtained in all scenarios. In the absence of particles (Ngueleu et al., 2013), retardation of lindane occurred due to sorption onto the sand only and the mean arrival time corresponds to

NU

about 1.5 pore volumes. It is now clear to see that the BTCs of lindane in the absence of

MA

particles and in scenario A (injection of fine particles (LIG1a), followed by millipore water and lindane) are similar. This concludes that sorption of lindane in scenario A was more onto

D

the sand than onto the particles. Nonetheless sorption onto the particles may be responsible

AC

CE P

TE

for the smaller maximum concentration.

Figure 9: Summary of the breakthrough curves of total (dissolved and particle-bound) lindane during the two last phases of all scenarios. (a): injection phase; (b): elution phase. The scenario with “no particles present” is from Ngueleu et al. (2013).

Concerning the BTC of scenario C (injection of raw particle mixture (LIG0c), immediately followed by lindane), mobile particles at the beginning of lindane injection led to

39

ACCEPTED MANUSCRIPT temporary particle-facilitated transport of lindane. The relevance of the latter transport mechanism subsided later because of the limited concentration of mobile particles.

T

When compared to scenario A, it may be expected in scenario D (injection of fine

IP

particles (LIG1d) loaded with lindane, followed by millipore water and lindane) that the

SC R

release of lindane previously accumulated in the first two phases could just be a matter of time. At last, the BTC of scenario B (injection of raw particle mixture (LIG0b), followed by millipore water and lindane) is the most retarded one due to strong sorption of lindane onto

NU

LIG0b particles combined to intra-particle diffusion in these particles.

MA

The inflow concentration (indicated by a normalized concentration of 1) was not reached in the outflow in any scenario because of kinetic sorption of lindane coupled to the limited availability of mobile particles carrying lindane. We calculated the percentages of

TE

D

lindane recovered at the column outlet for the two last phases of each scenario by integrating the time series of experimental concentrations over the pore volumes. Since the number of

CE P

pore volumes in the injection phase is not the same for all scenarios, the percentage of lindane recovered was determined and then normalized by the percentage of an ideal tracer

AC

transported under identical conditions in order to allow the comparison of the BTCs. These calculations yielded the following results: 70.7% for scenario C, 86.6% for scenario B, 88.3% for scenario A, and 89.4% for scenario D. These results are well reflected in Figure 9a-b. It is also worth saying that a secondary source of lindane can be expected in scenario D as it has the highest percentage of lindane recovery. When applying the same percentage calculations to the experimental concentrations of lignite particles and then subtracting the results from 100%, the percentages of particles retained in the sand are as follows: 69.3% for scenario D, 78.7% for scenario A, 83.3% for scenario B, and 90.4% for scenario C (Table 7). This sequence is the opposite of that of lindane, indicating that the increase of lignite particle retention increased lindane sorption.

40

ACCEPTED MANUSCRIPT Scenarios A and B were different only by the size of the particles as d50 values were 0.2 µm and 1 µm in scenarios A and B, respectively. It is noticeable in Table 7 that the

T

percentage and mass of particle retained in scenario B are bigger than that in scenario A. This

IP

observation confirms the statement made in section 3.2 that more retention of LIG0b particles

SC R

than LIG1a particles was responsible of more lindane sorption, leading to more lindane retardation.

Note that in the increasing order, the sequences of the percentage and mass of particle

NU

retained are D < A < B < C and D < A < C < B, respectively (Table 7). The difference

MA

between B and C is because the mass of particle retained depends on the mass of particle initially injected (Table 7). However the relevant information remains that particle retention increases with particle size.

TE

D

We calculated the equilibrium retardation factor (Rp) that would be expected from lindane sorption only onto retained lignite particles. The values of Rp are shown in Table 7

CE P

and depend on the mass of lignite particle retained. If the column was filled with only lignite, that is f m  100% , the expected value of Rp would be at least 3000, indicating that lignite and

AC

other similar carbonaceous particles can strongly retain hydrophobic pollutants at equilibrium. The large retardation factor of scenario B is also a consequence of the large mass fraction of particle retained (fm) which results from strong particle retention, as well as the fact that the mass or concentration of particle initially injected was high.

41

ACCEPTED MANUSCRIPT

#

retardation factor expected from lindane sorption only onto

mp = Percentage of particle retained × mp,in / 100.

§

mp m p  ms



Mass fraction of particle retained (fm) [%] 0.2 0.1 1.1 0.6

T

Mass of particle retained (mp) [g] 0.24 0.12 1.32 0.73

Rp [-]

8.0 5.4 39.5 29.8

where the mass of the sand ms used for the calculations was 115 g.

K Fr C 1 / nFr  ' b Rp 1  fm C n

where the new bulk density

11/n Fr

1/n Fr

1

 'b 

m p  ms

NU

fm 



#

IP

Mass of particle initially injected (mp,in) [g] 0.31 0.17 1.59 0.81

SC R

Table 7: Calculated equilibrium retained lignite particles (§Rp). Lignite particle Percentage size and scenario of particle retained [%] 0.2 µm, scenario A 78.7 0.8 µm, scenario D 69.3 1 µm, scenario B 83.3 3 µm, scenario C 90.4

V

, V being the volume of the column equal

Conclusions

D

4

MA

l kg and 1/nFr of 0.72 were used for the calculations. Breakthrough to 0.068 l. KFr of 707 mg concentrations that were used for the calculations were 1.46 mg l-1, 0.65 mg l-1, 1.48 mg l-1 and 0.50 mg l-1 for scenario A, D, B and C, respectively.

TE

In this study we used two types of lignite particles, denoted as LIG0 with median

CE P

diameters of about 1 µm (LIG0b) and 3 µm (LIG0c), and LIG1 with median diameters of about 0.2 µm (LIG1a) and 0.8 µm (LIG1d). We performed experiments in which first these particles were injected, followed by a lindane-containing solution in order to see how retained

AC

particles affect the breakthrough of lindane. The results demonstrated that lignite particles retained in the sandy porous medium can alter the transport of invading lindane. In all scenarios, particle retention was high and increased with increasing particle size. Remobilization of these particles with millipore water was characterized by a sudden increase of particle concentration due to change in solution chemistry. The elution curve had a long tail as a result of continuous particle detachment over time. Numerical modeling of particle transport showed that most likely both reversible attachment and irreversible straining controlled the transport of particles. Alteration of lindane transport by retained particles was characterized by evaluating the retardation of the lindane breakthrough through determining the number of pore volumes 42

ACCEPTED MANUSCRIPT corresponding to the mean arrival time of lindane. The mean arrival time was the time needed to reach half of the inflow concentration. This analysis revealed that the mean arrival time of

T

lindane in the absence of particles corresponded to about 1.5 pore volumes (Ngueleu et al.,

IP

2013); whereas in the presence of retained particles, it corresponded to about 1.7 pore

SC R

volumes in scenario A (injection of LIG1a particles with median diameter of 0.2 µm, followed by millipore water and lindane), about 3.5 pore volumes in scenario B (injection of LIG0b particles with median diameter of 1 µm, followed by millipore water and lindane),

NU

about 2.2 pore volumes in scenario C (injection of LIG0c particles with median diameter of 3

MA

µm, immediately followed by lindane), and about 2.3 pore volumes in scenario D (injection of LIG1d particles with median diameter of 0.8 µm loaded with lindane, followed by millipore water and lindane). These results showed that retardation of lindane occurred in all scenarios

TE

D

as the number of pore volumes of a non-retarded tracer is 1. Major causes of retardation of lindane were: the high particle retention, the sorption of

CE P

lindane onto the sand and onto retained particles, and the limited concentration of mobile particles carrying lindane. Particle retention most likely did not lead to irreversible retention

AC

of lindane even though lindane recovery was incomplete over the duration of the experiments. When lindane-loaded particles of LIG1d were retained in the porous medium, no secondary source of lindane was observed based on the arrival time. However, based on the concentration and the percentage of lindane recovered, a secondary source can be expected at late times. This experiment highlights the necessity to check the contamination history of aquifers in which particle-facilitated transport may be expected. Entrapped particles alter the transport behavior of fresh contaminants in different ways when they themselves are contaminated or not. We could simulate lindane transport with models which fitted the experimental data well. The numerical models could be used to quantify the contribution of retained particles to the retardation of lindane. The derived models of particle-associated lindane and dissolved 43

ACCEPTED MANUSCRIPT lindane suggest that dissolved lindane transport dominated over particle-bound transport in all scenarios.

T

While adsorption of lindane onto particles contributed to its retardation and desorption

IP

to its release into the aqueous phase, it was found that intra-particle diffusion was a limiting

SC R

process to lindane adsorption/desorption onto/from both limestone fragments in the sand and lignite particles. Intra-particle diffusion in lignite particles retained in the sand was explained by the similarity between the calculated diffusion rate coefficient (kdiff,calc) and the first-order

NU

kinetic rate coefficient (kp) of lindane adsorption/desorption onto/from lignite particles of the

MA

same size. In general, fitted kp values depended on the preloading of the sand particles and the time scales involved.

The results of this study may be of relevance in the proposed application of biochar

TE

D

for soil amendment. Known as a carbon-rich solid material, biochar may be added to poorly fertile soils in order to improve their functions. However, the current study shows that biochar

CE P

whose properties are similar to lignite (brown coal) may also alter the transport of lindane and other contaminants in soils. Lignite particles are the product of slow geological thermal

AC

alteration of organic matter at different thermal conditions and various types of biochars exist which may be considered similar to lignite. In fact peat bogs also contain high amounts of charcoals (biochars) because of a high frequency of wildfires there. Overall, biochar could then retain contaminants like in scenario B, but their mobilization may also enhance contaminant transport like in scenario C. Furthermore, biochar may act as a potential secondary source of sorbing contaminant like in scenario D. However, further studies considering unsaturated conditions are needed to elucidate the conditions for straining of organic particles in unsaturated zones and its effects on particle-modified transport of sorbing contaminants, whereas the present study was restricted to water-saturated groundwater conditions.

44

ACCEPTED MANUSCRIPT 5

Acknowledgements The authors gratefully thank the German Federal Ministry of Education and Research

T

(BMBF) for supporting this research through the program International Postgraduate Studies

IP

in Water Technologies (IPSWaT), and Bertrand Ligouis for providing natural lignite. We are

SC R

also thankful to Julia Knapp and Christina Haberer for their valuable assistance during the review process.

References

NU

6

Stable

Isotope

MA

Bashir, S., Fischer, A., Nijenhuis, I. and Richnow, H.-H., 2013. Enantioselective Carbon Fractionation

of

Hexachlorocyclohexane

during

Aerobic

D

Biodegradation by Sphingobium spp. Environmental Science & Technology, 47(20):

TE

11432-11439.

Ben-Moshe, T., Dror, I. and Berkowitz, B., 2010. Transport of metal oxide nanoparticles in

CE P

saturated porous media. Chemosphere, 81(3): 387-393. Bold, S., Kraft, S., Grathwohl, P. and Liedl, R., 2003. Sorption/desorption kinetics of

AC

contaminants on mobile particles: Modeling and experimental evidence. Water Resour. Res., 39(12): 1329. Bold, S., Liedl, R. and Grathwohl, P., 2005. Enhancement of solute spreading in soils due to particle-facilitated transport and preferential flow.

In: Nützmann, G., Viotti, P.,

Aagard, P. (eds.) Reactive transport in soil and groundwater, 157 - 169, Springer Verlag. Bradford, S.A., Simunek, J., Bettahar, M., van Genuchten, M.T. and Yates, S.R., 2006. Significance of straining in colloid deposition: Evidence and implications. Water Resources Research, 42(12): W12S15.

45

ACCEPTED MANUSCRIPT Bradford, S.A., Yates, S.R., Bettahar, M. and Simunek, J., 2002. Physical factors affecting the transport and fate of colloids in saturated porous media. Water Resour. Res., 38(12):

T

1327.

IP

Burkhardt, M., Kasteel, R., Vanderborght, J. and Vereecken, H., 2008. Field study on colloid

SC R

transport using fluorescent microspheres. European Journal of Soil Science, 59(1): 8293.

Chen, G. and Strevett, K.A., 2001. Impact of bacterial extracellular polymers on lindane

NU

transport. Environmental Engineering Science, 18(3): 191-203.

MA

Cheng, X., Kan, A.T. and Tomson, M.B., 2005. Study of C60 transport in porous media and the effect of sorbed C60 on naphthalene transport. Journal of Materials Research, 20(12): 3244-3254.

TE

D

De Jonge, H., De Jonge, L.W. and Jacobsen, O.H., 2000. [14C]Glyphosate transport in undisturbed topsoil columns. Pest Management Science, 56(10): 909-915.

CE P

DeNovio, N.M., Saiers, J.E. and Ryan, J.N., 2004. Colloid Movement in Unsaturated Porous Media: Recent Advances and Future Directions. Vadose Zone J., 3(2): 338-351.

AC

Duckworth, Q.W. and Martin, S.T., 2004. Dissolution rates and pit morphologies of rhombohedral carbonate minerals. American Mineralogist, 89(4): 554-563. Fang, J. et al., 2011. Transport of copper as affected by titania nanoparticles in soil columns. Environmental Pollution, 159(5): 1248-1256. Finkel, M., Liedl, R. and Teutsch, G., 1998. Modelling surfactant-enhanced remediation of polycyclic aromatic hydrocarbons. Environmental Modelling & Software, 14(2–3): 203-211. Finkel, M., Liedl, R. and Teutsch, G., 2007. Modelling Reactive Transport of Organic Solutes in Groundwater with a Lagrangian Streamtube Approach. In: F. Deutsche (Editor), Geochemical Processes, pp. 115-134.

46

ACCEPTED MANUSCRIPT Friedlander, S.K., 1958. Theory of Aerosol Filtration. Industrial & Engineering Chemistry, 50(8): 1161-1164.

T

Grolimund, D., Barmettler, K. and Borkovec, M., 2007. Colloid Facilitated Transport in

IP

Natural Porous Media: Fundamental Phenomena and Modelling, pp. 3-27.

SC R

Grolimund, D., Borkovec, M., Barmettler, K. and Sticher, H., 1996. Colloid-Facilitated Transport of Strongly Sorbing Contaminants in Natural Porous Media:  A Laboratory Column Study. Environmental Science & Technology, 30(10): 3118-3123.

NU

Hahn, M.W., Abadzic, D. and O'Melia, C.R., 2004. Aquasols:  On the Role of Secondary

MA

Minima†. Environmental Science & Technology, 38(22): 5915-5924. Happel, J., 1958. Viscous flow in multiparticle systems: Slow motion of fluids relative to beds of spherical particles. AIChE Journal, 4(2): 197-201.

TE

D

Hunt, J.R., McDowell-Boyer, L.M. and Sitar, N., 1987. Colloid migration in porous media. An analysis of mechanisms. Published in: Tsang, C.F., 1987. Coupled processes

CE P

associated with nuclear waste repositories, pp. 453-472, 795 p. Hydutsky, B.W., Mack, E.J., Beckerman, B.B., Skluzacek, J.M. and Mallouk, T.E., 2007.

AC

Optimization of nano- and microiron transport through sand columns using polyelectrolyte mixtures. Environmental Science & Technology, 41(18): 6418-6424. IARC, 1979. IARC Monographs on the evaluation of the carcinogenic risk of chemicals to humans. Some halogenated hydrocarbons. International Agency for Research on Cancer, Lyon. 20: 195. IPEN, 2006. Country situation on Persistent Organic Pollutants (POPs) in India. International POPs Elimination Network. India, March 2006, 57 pp. Kersting, A.B. et al., 1999. Migration of plutonium in ground water at the Nevada Test Site. Nature, 397(6714): 56-59.

47

ACCEPTED MANUSCRIPT Kim, J.I., 1989. Geochemistry of Actinides and Fission Products in Natural Aquifer Systems. In: CEC Project MIRAGE, Second Summary Progress Report, EUR 12229 EN, ed. B.

T

Come, Brussels, p. 1-72.

IP

Kim, J.I., 1991. Actinide colloid generation in groundwater. Radiochimica Acta, 52-3: 71-81.

SC R

Kleineidam, S., Rügner, H. and Grathwohl, P., 1999a. Impact of grain scale heterogeneity on slow sorption kinetics. Environmental Toxicology and Chemistry, 18(8): 1673-1678. Kleineidam, S., Rügner, H., Ligouis, B. and Grathwohl, P., 1999b. Organic Matter Facies and

NU

Equilibrium Sorption of Phenanthrene. Environmental Science & Technology, 33(10):

MA

1637-1644.

Kleineidam, S., Schüth, C. and Grathwohl, P., 2002. Solubility-Normalized Combined Adsorption-Partitioning Sorption Isotherms for Organic Pollutants. Environmental

TE

D

Science & Technology, 36(21): 4689-4697. Kretzschmar, R., Borkovec, M., Grolimund, D. and Elimelech, M., 1999. Mobile subsurface

CE P

colloids and their role in contaminant transport. Adv. Agron. 66, 121–193. Lachhab, A., Zhang, Y.-K. and Muste, M.V.I., 2008. Particle Tracking Experiments in Match-

AC

Index-Refraction Porous Media. Ground Water, 46(6): 865-872. Lagarias, J.C., Reeds, J.A., Wright, M.H. and Wright, P.E., 1998. Convergence properties of the Nelder-Mead simplex method in low dimensions. Siam Journal on Optimization, 9(1): 112-147. Lecoanet, H.F., Bottero, J.-Y. and Wiesner, M.R., 2004. Laboratory Assessment of the Mobility of Nanomaterials in Porous Media. Environmental Science & Technology, 38(19): 5164-5169. Lenhart, J.J. and Saiers, J.E., 2003. Colloid mobilization in water-saturated porous media under transient chemical conditions. Environmental Science & Technology, 37(12): 2780-2787.

48

ACCEPTED MANUSCRIPT Litton, G.M. and Olson, T.M., 1996. Particle size effects on colloid deposition kinetics: evidence

of

secondary

minimum

deposition.

Colloids

and

Surfaces

A:

T

Physicochemical and Engineering Aspects, 107: 273-283.

IP

Magee, B.R., Lion, L.W. and Lemley, A.T., 1991. Transport of dissolved organic

SC R

macromolecules and their effect on the transport of phenanthrene in porous media. Environmental Science and Technology, 25(2): 323-331.

McCarthy, J., 1993. Manipulation of Groundwater Colloids for Environmental Restoration.

NU

Taylor & Francis.

MA

McCarthy, J.F. and McKay, L.D., 2004. Colloid Transport in the Subsurface: Past, Present, and Future Challenges. Vadose Zone J, 3(2): 326-337. McCarthy, J.F. and Zachara, J.M., 1989. Subsurface transport of contaminants-mobile

TE

D

colloids in the subsurface environment may alter the transport of contaminants. Environ. Sci. Technol. 23, 496–502.

CE P

McDowell-Boyer, L.M., 1992. Chemical Mobilization of Micron-sized Particles in Saturated Porous-Media under Steady Flow Conditions Environmental Science & Technology,

AC

26(3): 586-593.

McDowell-Boyer, L.M., Hunt, J.R. and Sitar, N., 1986. Particle transport through porous media. Water Resour. Res., 22(13): 1901-1921. Mills, W.B., Liu, S. and Fong, F.K., 1991. Literature-review and model (comet) for colloid metals transport in porous-media. Ground Water, 29(2): 199-208. Moroni, M., Kleinfelter, N. and Cushman, J.H., 2007. Analysis of dispersion in porous media via matched-index particle tracking velocimetry experiments. Advances in Water Resources, 30(1): 1-15. Ngueleu, S.K., Grathwohl, P. and Cirpka, O.A., 2013. Effect of natural particles on the transport of lindane in saturated porous media: Laboratory experiments and modelbased analysis. Journal of Contaminant Hydrology, 149: 13-26. 49

ACCEPTED MANUSCRIPT Nicholson, D. and Petropoulos, J.H., 1968. Capillary models for porous media I. Two-phase flow in a serial model. Journal of Physics D: Applied Physics, 1(10): 1379-1385.

T

Nicholson, D. and Petropoulos, J.H., 1971. Capillary Models for Porous Media: III. Two-

SC R

of Physics D: Applied Physics, 4(2): 181-189.

IP

phase flow in a three-dimensional network with Gaussian radius distribution. Journal

Nicholson, D. and Petropoulos, J.H., 1973. Capillary models for porous media: IV. Flow properties of parallel and serial capillary models with various radius distributions.

NU

Journal of Physics D: Applied Physics, 6(14): 1737-1744.

MA

Nicholson, D. and Petropoulos, J.H., 1975. Capillary models for porous media. V. Flow properties of random networks with various radius distributions. Journal of Physics D: Applied Physics, 8(12): 1430-1440.

TE

D

O'Melia, C., 1980. ES&T Features: Aquasols: the behavior of small particles in aquatic systems. Environmental Science & Technology, 14(9): 1052-1060.

CE P

Ogata, A. and Banks, R.B., 1961. A solution of the differential equation of longitudinal dispersion in porous media, United States Geological Survey, Washington, D.C.

AC

Panfilov, M., Panfilova, I. and Stepanyants, Y., 2008. Mechanisms of particle transport acceleration in porous media. Transport in Porous Media, 74(1): 49-71. Payatakes, A.C., Tien, C. and Turian, R.M., 1974. Trajectory calculation of particle deposition in deep bed filtration: Part I. Model formulation. AIChE Journal, 20(5): 889-900. Pelley, A.J. and Tufenkji, N., 2008. Effect of particle size and natural organic matter on the migration of nano- and microscale latex particles in saturated porous media. Journal of Colloid and Interface Science, 321(1): 74-83. Pozo, K. et al., 2011. Assessing seasonal and spatial trends of persistent organic pollutants (POPs) in Indian agricultural regions using PUF disk passive air samplers. Environmental Pollution, 159(2): 646-653. 50

ACCEPTED MANUSCRIPT Rajagopalan, R. and Tien, C., 1976. Trajectory analysis of deep-bed filtration with the spherein-cell porous media model. AIChE Journal, 22(3): 523-533.

T

Roy, S.B. and Dzombak, D.A., 1996. Colloid release and transport processes in natural and

IP

model porous media. Colloids and Surfaces A: Physicochemical and Engineering

SC R

Aspects, 107(0): 245-262.

Roy, S.B. and Dzombak, D.A., 1997. Chemical factors influencing colloid-facilitated transport of contaminants in porous media. Environmental Science & Technology,

NU

31(3): 656-664.

MA

Roy, S.B. and Dzombak, D.A., 1998. Sorption nonequilibrium effects on colloid-enhanced transport of hydrophobic organic compounds in porous media. Journal of Contaminant Hydrology, 30(1–2): 179-200.

TE

D

Rügner, H., Kleineidam, S. and Grathwohl, P., 1999. Long Term Sorption Kinetics of Phenanthrene in Aquifer Materials. Environmental Science & Technology, 33(10):

CE P

1645-1651.

Ryan, J.N. and Gschwend, P.M., 1994a. Effect of Solution Chemistry on Clay Colloid

AC

Release from an Iron Oxide-Coated Aquifer Sand. Environmental Science & Technology, 28(9): 1717-1726. Ryan, J.N. and Gschwend, P.M., 1994b. Effects of Ionic Strength and Flow Rate on Colloid Release: Relating Kinetics to Intersurface Potential Energy. Journal of Colloid and Interface Science, 164(1): 21-34. Saiers, J.E. and Hornberger, G.M., 1996. The Role of Colloidal Kaolinite in the Transport of Cesium through Laboratory Sand Columns. Water Resour. Res., 32(1): 33-41. Saleh, N. et al., 2008. Ionic Strength and Composition Affect the Mobility of SurfaceModified Fe0 Nanoparticles in Water-Saturated Sand Columns. Environmental Science & Technology, 42(9): 3349-3355.

51

ACCEPTED MANUSCRIPT Schüth, C., 1994. Sorptionskinetik und Transportverhalten von polyzyklischen aromatischen Kohlenwasserstoffen (PAK) im Grundwasser - Laborversuche (Sorption kinetics and

T

transport behavior of polycyclic aromatic hydrocarbons (PAH) in groundwater -

IP

Laboratory experiments). Tübingen Geowissenschaftliche Arbeiten (TGA), C19

SC R

(Universität Tübingen). 80 pp.

Schwarzenbach, R.P., Gschwend, P.M. and Imboden, D.M., 1993. Environmental organic chemistry. J. Wiley, 681 pp.

NU

Simunek, J., He, C.M., Pang, L.P. and Bradford, S.A., 2006. Colloid-facilitated solute

MA

transport in variably saturated porous media: Numerical model and experimental verification. Vadose Zone Journal, 5(3): 1035-1047. Sjöberg, E.L. and Rickard, D.T., 1984. Calcite dissolution kinetics: Surface speciation and the

TE

D

origin of the variable pH dependence. Chemical Geology, 42(1–4): 119-136. Sun, H.M. et al., 2010. Kaolinite and Lead in Saturated Porous Media: Facilitated and

CE P

Impeded Transport. Journal of Environmental Engineering-Asce, 136(11): 1305-1308. Tiraferri, A. and Sethi, R., 2009. Enhanced transport of zerovalent iron nanoparticles in

AC

saturated porous media by guar gum. Journal of Nanoparticle Research, 11(3): 635645.

Tosco, T. and Sethi, R., 2009. MNM1D: A Numerical Code for Colloid Transport in Porous Media: Implementation and Validation. American Journal of Environmental Sciences, 5(4): 516-524. Tosco, T., Tiraferri, A. and Sethi, R., 2009. Ionic Strength Dependent Transport of Microparticles in Saturated Porous Media: Modeling Mobilization and Immobilization Phenomena under Transient Chemical Conditions. Environmental Science & Technology, 43(12): 4425-4431. Wan, J. and Wilson, J.L., 1994a. Colloid transport in unsaturated porous media. Water Resour. Res., 30(4): 857-864. 52

ACCEPTED MANUSCRIPT Wan, J.M. and Wilson, J.L., 1994b. Visualization of the role of the gas-water interface on the fate and transport of colloids in porous-media. Water Resources Research, 30(1): 11-

T

23.

IP

Wang, G. and Grathwohl, P., 2009. Activation energies of phenanthrene desorption from

SC R

carbonaceous materials: Column studies. Journal of Hydrology, 369(3-4): 234-240. Wang, P. and Keller, A.A., 2009. Natural and Engineered Nano and Colloidal Transport: Role of Zeta Potential in Prediction of Particle Deposition. Langmuir, 25(12): 6856-6862.

NU

Wang, Y., Li, Y. and Pennell, K.D., 2008. Influence of electrolyte species and concentration

MA

on the aggregation and transport of fullerene nanoparticles in quartz sands. Environmental Toxicology and Chemistry, 27(9): 1860-1867. Worch, E., 1993. Eine neue Gleichung zur Berechnung von Diffusionskoeffizenten geloester

TE

D

Stoffe. Wasser, 81: 289-297.

Wu, S.-C. and Gschwend, P.M., 1988. Numerical modeling of sorption kinetics of organic

1383.

CE P

compounds to soil and sediment particles. Water Resources Research, 24(8): 1373-

AC

Xie, B., Xu, Z., Guo, W. and Li, Q., 2008. Impact of Natural Organic Matter on the Physicochemical Properties of Aqueous C60 Nanoparticles. Environmental Science & Technology, 42(8): 2853-2859. Yin, X. et al., 2010. Colloid-facilitated Pb transport in two shooting-range soils in Florida. Journal of Hazardous Materials, 177(1-3): 620-625. Zhuang, J., Flury, M. and Jin, Y., 2003. Colloid-Facilitated Cs Transport through WaterSaturated Hanford Sediment and Ottawa Sand. Environmental Science & Technology, 37(21): 4905-4911. Zhuang, J., Tyner, J.S. and Perfect, E., 2009. Colloid transport and remobilization in porous media during infiltration and drainage. Journal of Hydrology, 377(1-2): 112-119.

53

ACCEPTED MANUSCRIPT Zvikelsky, O. and Weisbrod, N., 2006. Impact of particle size on colloid transport in discrete fractures. Water Resources Research, 42(12).

T

Zvikelsky, O., Weisbrod, N. and Dody, A., 2008. A comparison of clay colloid and artificial

IP

microsphere transport in natural discrete fractures. Journal of Colloid and Interface

NU

SC R

Science, 323(2): 286-292.

Summary of Abbreviations

foc’

k1 k2 k3 k4 kap kdiff,calc kdp KFr kp kstr Kd,eq

CE P

TE

D

Meaning Aqueous-phase concentration of the tracer or lindane Inflow concentration of the tracer or lindane Concentration of mobile particles Inflow concentration of the particles Median grain size Hydrodynamic dispersion coefficient Apparent diffusion coefficient Aqueous diffusion coefficient Mass fraction Organic carbon content Organic carbon content of the sand matrix containing retained lignite particles Rate coefficient of adsorption of lindane onto matrix sites 1 of the sand Rate coefficient of desorption of lindane from matrix sites 1 of the sand Rate coefficient of adsorption of lindane onto matrix sites 2 of the sand Rate coefficient of desorption of lindane from matrix sites 2 of the sand First-order particle attachment coefficient Calculated diffusion rate coefficient First-order particle detachment coefficient Freundlich distribution coefficient First-order kinetic rate coefficient of lindane adsorption/desorption onto/from particles First-order particle straining coefficient Linear distribution coefficient onto equilibrium sites of the

AC

Symbol C Cin Cp Cp,in d50 D Da Daq fm foc

MA

Appendix

54

Unit mg l-1 mg l-1 kg l-1 kg l-1 m m2 s-1 m2 s-1 m2 s-1 % weight% weight% s-1 s-1 s-1 s-1 s-1 s-1 s-1

mg11/n Fr l1/n Fr kg 1 s-1 s-1 l kg-1

ACCEPTED MANUSCRIPT

r3

sand matrix Column length Empirical exponent of tortuosity Empirical exponent of the sand matrix containing retained lignite particles Mass of particle retained Mass of particle initially injected Mass of the sand Porosity of the sand matrix Darcy velocity Radial distance from the center of the particle Mass transfer rate of dissolved lindane onto kinetic site 1 of the matrix Mass transfer rate of dissolved lindane onto kinetic site 2 of the matrix Straining rate of particles

r4

Net attachment rate of particles

r5

Mass transfer rate of dissolved lindane onto mobile particles Mass transfer rate of dissolved lindane onto reversibly attached particles Mass transfer rate of dissolved lindane onto irreversibly strained particles Mass transfer rate of lindane due to particle straining

mg l-1 s-1

Mass transfer rate of lindane due to net particle attachment Retardation factor due to equilibrium sorption onto the sand matrix Equilibrium retardation factor expected from lindane sorption only onto retained lignite particles Concentration of lindane sorbed onto kinetic site 1 of the matrix Concentration of lindane sorbed onto kinetic site 2 of the matrix Concentration of reversibly attached particles Concentration of lindane sorbed onto reversibly attached particles Concentration of lindane sorbed onto mobile particles Concentration of irreversibly strained particles Concentration of lindane sorbed onto irreversibly strained particles Time Time of injection Seepage velocity Volume of the column Spatial coordinate Distance from the porous medium inlet used to limit the straining effect Freundlich exponent

mg kg-1 s-1

r6 r7 r3 p r4 p Req Rp S1 S2 Sp Sip Smp Sstr Sstrp t tin v V x X 1/nFr

[-] g g g m s-1 m

T

IP

SC R

NU

MA

r2

D

r1

TE

mp mp,in ms n q r

CE P

m'

AC

L m

55

mg kg-1 s-1 mg kg-1 s-1 kg kg-1 s-1 kg kg-1 s-1

mg kg-1 s-1 mg kg-1 s-1 mg kg-1 s-1

mg kg-1 mg kg-1 kg kg-1 mg kg-1 mg kg-1 kg kg-1 mg kg-1 s s m s-1 l m m -

ACCEPTED MANUSCRIPT kg l-1 kg l-1

T

IP

SC R NU MA

p

D

 b'

TE

b

CE P

ε

Fitting parameter that controls the shape of the particle spatial distribution Intra-particle porosity Bulk density of the sand matrix Bulk density of the sand matrix containing retained lignite particles Bulk density of the particle

AC

β

56

kg l-1

ACCEPTED MANUSCRIPT Highlights Attachment and straining affect the transport of lignite particles.



Retained lignite particles alter the transport of invading lindane.



Sorption and intra-particle diffusion affects lindane transport.



Increase of lignite particle retention increases lindane sorption.



Sorption and spherical intra-particle diffusion models reproduce well the results.

AC

CE P

TE

D

MA

NU

SC R

IP

T



57

Altered transport of lindane caused by the retention of natural particles in saturated porous media.

Attachment and straining of colloidal particles in porous media result in their reversible and irreversible retention. The retained particles may eith...
1MB Sizes 0 Downloads 3 Views