applied clay science · containing 78% and 96% of clay-sized (b5 mm) material, respectively. the...

Applied Clay Science 86 (2013) 83–98

Contents lists available at ScienceDirect

Applied Clay Science

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

Research paper

Improving membrane performance via bentonitepolymer nanocomposite

Gretchen L. Bohnhoff a,b, Charles D. Shackelford c,⁎a Department of Civil & Environmental Engineering, 1 University Plaza, University of Wisconsin-Platteville, Platteville, WI 53818-3099, USAb Department of Civil and Environmental Engineering, Colorado State University, USAc Department of Civil and Environmental Engineering, 1372 Campus Delivery, Colorado State University, Fort Collins, CO 80523-1372, USA

⁎ Corresponding author. Tel.: +1 970 491 5051; fax: +E-mail addresses: [email protected] (G.L. Bohnh

[email protected] (C.D. Shackelford).

0169-1317/$ – see front matter © 2013 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.clay.2013.09.017

a b s t r a c t
a r t i c l e i n f o
Article history:Received 24 March 2012Received in revised form 17 August 2013Accepted 11 September 2013Available online xxxx

Keywords:BentoniteClay membranesClay polymer nanocomposite (CPN)Geosynthetic clay linerMembrane efficiencySemipermeable membrane

Traditional (non-treated) sodium bentonite has been shown to exhibit substantial semipermeable membranebehavior, or the ability to restrict the migration of solutes. However, partial or complete degradation of themembrane behavior due to diffusion of invading salt cations into the sodium bentonite also has been observed.In this study, a polyacrylate modified bentonite, referred to as a bentonite polymer nanocomposite, or BPN, wasevaluated as a potential substitute for sodium bentonite for the purpose of providing increased resistance to saltdegradation of membrane behavior. The membrane behavior of the BPN was measured in the laboratory byestablishing differences in salt concentrations ranging from 4.7 to 54 mM KCl across specimens contained ineither rigid-wall or flexible-wall cells under closed-system boundary conditions. The measured membraneefficiency coefficients, ω, for the BPN specimens were greater than those previously reported for specimenscontaining sodium bentonite under similar testing conditions. For example, the measured ω values for the BPNspecimens ranged from 109% to 433% of those previously reported for the specimens containing traditionalsodium bentonite, depending on the initial porosity or initial effective stress of the specimen, the concentrationof KCl, and the type of test cell (rigid vs.flexible), and despite the BPN specimens not beingflushed of soluble saltsprior to membrane testing as in the case of the specimens containing sodium bentonite. Thus, the BPN exhibitedsubstantially improved membrane behavior under conditions that presumably were more adverse with respectto soluble salts than those previously reported for specimens containing a sodium bentonite. However, thepotential role that any excess lowmolecular weight polymer in the BPNmay have played in affecting the resultsis uncertain, such that additional testing of the BPN is warranted to better understand the long-term behaviorof the BPN.

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Bentonite is commonly used to control liquid flow and contaminanttransport for a variety of hydraulic containment applications, includinggroundwater cutoff walls, barriers for waste containment (e.g., landfills,wastewater ponds, manure lagoons, nuclear storage, etc.), secondarycontainment in tank farms, and seals in monitoring and water supplywells (Christman et al., 2002; Estornell and Daniel, 1992; Evans, 1993;Kajita, 1997; Katsumi et al., 2008; Smith et al., 2003). In nearly allof these applications, traditional sodium bentonite is used, meaningthat sodium is the predominant cation associated with the exchangecomplex of the bentonite particles. The preference for sodium bentonitestems from desirable engineering properties, such as low hydraulicconductivity (k) to water (typically k b 10−10 m/s), and the existenceof semipermeable membrane behavior, or the restriction of solutes

1 970 491 7727.off),

ghts reserved.

during migration which gives rise to hyperfiltration, chemico-osmosis,and reduced diffusive solute mass flux (Malusis et al., 2003).

Unfortunately, sodium bentonite is thermodynamically unstable inenvironments where multivalent cations are present or predominant(Sposito, 1989), including most naturally occurring pore waters inearthenmaterials. Under such conditions, multivalent cations graduallyreplace monovalent cations originally dominating the exchange com-plex, thereby reducing or eliminating osmotic swelling of the sodiumbentonite and the ability of sodium bentonite to function effectively,even if the sodium bentonite is prehydrated (Kolstad et al., 2004; Leeand Shackelford, 2005b; Vasko et al., 2001). For example, several fieldstudies have shown that Ca2+-for-Na+ exchange can result in reducedswelling capability of the sodium bentonite contained in geosyntheticclay liners (GCLs) used in hydraulic containment applications uponhydration, and ultimately to poor hydraulic performance (ATU, 1992;Benson et al., 2007; Egloffstein, 2001; James et al., 1997; Jo et al., 2001,2004; Meer and Benson, 2007; Shackelford et al., 2000). The Ca2+ isderived from surrounding soils, and migrates into the GCL usually inresponse to hydraulic (suction) and/or chemical (diffusive) gradients.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.clay.2013.09.017&domain=pdf

http://dx.doi.org/10.1016/j.clay.2013.09.017

mailto:[email protected]

mailto:[email protected]

http://dx.doi.org/10.1016/j.clay.2013.09.017

http://www.sciencedirect.com/science/journal/01691317

Table 2Index properties of the conventional sodium bentonite used to create the bentonitepolymer nanocomposite (BPN) and the BPN evaluated in this study.

Property Standard Average value or type [no. trials]

Sodiumbentonite

Bentonite polymernanocomposite (BPN)

Specific gravity ASTM D 854 2.76 2.67Soil classification ASTM D 2487 CH [3] CH [3]a

Percent clay (%) ASTM D 422 78 [3] 96 [3]Liquid limit, LL (%) ASTM D 4318 422 [3] 255 [2]Plasticity index, PI (%) ASTM D 4318 384 [3] NASwell index, SI (mL/2 g) ASTM D 5890 22 [3] 73 [5]Solution retention capacity,SRC (mL/1 g)

b 7.3 [3] 11.0 [6]

NA=Not applicable.a Based on grain-size distribution.b Based on procedures in Lee and Shackelford (2005c) with centrifuge speed at

3000 rpm.

Table 1Chemical properties of liquids used in study.

Liquid Concentration pH Solution electrical conductivity,ECo (mS/m), @ 25 °C

(mM) (mg/L)

De-ionized water 0 0 7.35 b0.1KCl solutions 4.7 350 5.31 71

9.3 690 5.20 13920 1500 5.28 27654 4000 5.24 661

84 G.L. Bohnhoff, C.D. Shackelford / Applied Clay Science 86 (2013) 83–98

Recent laboratory studies have illustrated the deleterious effectsof long-term cation exchange on both k and membrane behavior(Egloffstein, 2001; Jo et al., 2001, 2005; Kolstad et al., 2004; Lee andShackelford, 2005a,b; Lin and Benson, 2000; Shackelford and Lee,2003). For example, in terms ofmembrane behavior, a direct correlationbetween diffusion of invading cations into bentonites and either partialor complete destruction of membrane behavior has been shown(Malusis and Shackelford, 2002a; Shackelford and Lee, 2003). Thesefield and laboratory data paint an unsettling picture regarding thelong-term effectiveness of bentonites used for hydraulic containmentapplications. However, bentonites can be chemically modified to makethe bentonite properties more compatible with the surroundingenvironment (e.g., Katsumi et al., 2008). For example, anionic polymersare added to bentonite to attain rheological properties needed fordrilling fluids (Heller and Keren, 2003). Organobentonites also havebeen used for applications where an organophylic material is neededfor adsorption or containment of organic compounds (e.g., Lo et al.,1996; Smith et al., 2003).

The objective of this study was to evaluate the potential use of apolyacrylate modified bentonite, referred to as a bentonite polymernanocomposite, or BPN, for improving membrane behavior in thepresence of simple salt solutions relative to traditional sodium bentonitethat is commonly used in hydraulic containment barriers, such as GCLs(e.g., Kang and Shackelford, 2009; Malusis and Shackelford, 2002a),compacted soil–bentonite liners (Kang and Shackelford, 2010), andsoil–bentonite backfills in vertical cutoff walls (Evans et al., 2008;Henning et al., 2006; Yeo et al., 2005). The BPN evaluated in this studyis thought to be fundamentally different from other polymer-modifiedbentonites in that the organic molecules are polymerized afterintercalation, resulting in the polymer bonding with sodium ions initiallyin the interlayer region and the surface of the bentonites, a processwhichis hypothesized to result in a stable structure where the layers arepropped open in the state attained by osmotic swelling (Bohnhoff,2012; Scalia, 2012). In contrast, conventional anionic polymer-modifiedbentonites employ long-chain anionic polymers (e.g., anionicpolyacrylimides) that are electrostatically associated with the positivelycharged edges of the bentonite platelets (e.g., Boels and van der Wal,1999; Heller and Keren, 2003; Pavlidou and Papaspyrides, 2008).

2. Materials and methods

2.1. Liquids

The liquids used in this study included de-ionized water (DIW) andsolutions of DIW with potassium chloride (KCl) (certified A.C.S.; FisherScientific, Fair Lawn, NJ). The KCl solutions were used as circulatingliquids in the membrane tests to allow for direct comparison of resultsfrom this study with those previously reported for tests conducted onGCLs (Kang and Shackelford, 2009; Malusis and Shackelford, 2002a).Solutionswere prepared and stored in 20-L carboys (Nalgene®; ThermoFisher Scientific, Rochester, NY). The pH and electrical conductivity,EC, of the solutions were measured using a pH meter (Accumet®AB15meter; Fisher Scientific Co., Pittsburgh, PA) and an EC probe (150A+Conductivity meter; Thermo Orion, Beverly, MA), respectively. Ionchromatography (Dionex® 4000i IC Module, Dionex Co., Sunnyvale,CA)was used tomeasure chloride (Cl−) concentrations, and inductivelycoupled plasma-atomic emission spectrometry (IRIS® Advantage/1000ICAP Spectrometer, Thermo Jarrel Ash Co., Franklin, MA) was used tomeasure potassium (K+) concentrations. The measured EC and pH ofthe KCl solutions are given in Table 1.

2.2. BPN

The BPN evaluated in this study was prepared by ColloidEnvironmental Technologies Co. (CETCO, Hoffman Estates, IL). TheBPN was produced with polyacrylic acid (PAA) and a conventional

sodium bentonite using methods similar to those used for the pro-duction of polymer nanocomposites (e.g., Muzny et al., 1996). First, amonomer solution was prepared by dissolving acrylic acid in waterfollowed by neutralization with sodium hydroxide and addition of aninitiator, sodium persulfate. Then, the sodium bentonite was added tothe monomer solution in concentrations ranging from 30 to 50% bymass of the slurry to form a bentonite–monomer slurry. Polymerizationwas initiated by raising the temperature of the bentonite–monomerslurry above the decomposition temperature of the initiator moleculecausing the initiator molecule to decompose into free radicals. The freeradicals react with the acrylic acid monomer to formmore free radicals,which in turn react with additionalmonomer to proliferate the polymerchain. Following polymerization, the BPN was oven dried at 105 °C,milled, and screened by CETCO (Scalia et al., 2011).

Based on the grain-size distributions from a hydrometer analysis(ASTM D 422), 100% of the sodium bentonite and the BPN werefine grained (b0.075 mm) with the sodium bentonite and the BPNcontaining 78% and 96% of clay-sized (b5 mm) material, respectively.The sodium bentonite and the BPN classified as a highly plastic clay(or CH according to the Unified Soil Classification System (ASTM D2487)). Mineralogical analyses conducted by Mineralogy, Inc., Tulsa,OK, using X-ray diffraction indicated that the sodium bentonite wascomprised of 77% montmorillonite, 15% quartz, 4% plagioclase feldspar,3% illite/mica, and 1 % calcite, whereas BPN was comprised of 76%montmorillonite, 15% quartz, 7% plagioclase feldspar, and 2% illite/mica. Thus, despite polymerization, the mineralogy of the BPN wasclose to that of the sodium bentonite.

Index properties of the sodium bentonite and the BPN are includedin Table 2. The Atterberg limits of the sodium bentonite and the BPNalso were measured according to ASTM D 4318. The measured liquidlimit, LL, was 255, which is significantly lower than the LL of 422for the sodium bentonite that was used to create the BPN as well asother traditional sodium bentonites, with values of LL that typicallyvary from 400 to 700 (e.g., Ito, 2006; Kenney et al., 1992; Lee andShackelford, 2005b). The values of LL for other treated bentonites also

85G.L. Bohnhoff, C.D. Shackelford / Applied Clay Science 86 (2013) 83–98

have been reported as being lower relative to those of traditionalsodium bentonites. For example, Di Emidio (2010) reported a LL of375 for dense pre-hydrated GCL, which is a GCL containing sodiumbentonite that is prehydrated with sodium-carboxylmethylcelluloseand methanol and densified via calendaring. In contrast, the valuesof the LL for HYPER clay, an anionic polymer modified bentonite,and multi-swellable bentonite are reported by Di Emidio (2010) tobe similar to those of traditional bentonites (e.g., 650 for HYPERclay and 554 for multi-swellable bentonite). McRory and Ashmawy(2005) reported increases and decreases in the LL of five polymertreated clays with the value of LL depending on the type of polymer(e.g., cationic, anionic), theweight percent of the polymer, andwhetheror not the clay was heated. For example, LL for a bentonite modifiedwith 1% cationic polymer was 291.

In contrast to the LL, the plastic limit, PL, of the BPN could not bedetermined. As shown in Fig. 1, the consistency of the BPN during thetest for PLwas similar to “silly putty” at low gravimetric water contents(b150%), and the soil “worm” did not exhibit the same cracking patternthat a traditional sodium bentonite would exhibit as the water contentapproaches the PL. The initial polymer content of the BPNwas calculatedto be 28.5% by mass (Scalia, 2012). Therefore, this “silly putty” behavioris likely a result of the superabsorbent polymer in the BPN causing theBPN to act more like a polymer than a typical clay. As a result, theplasticity index, PI, of the BPN could not be determined.

Tests to measure both the swell index, SI, and the solution retentioncapacity, SRC, using DIW also were conducted. The SI tests wereconducted in general accordance with ASTM D 5890. Briefly, 2 g ofoven dried BPN were dusted over the test solution in 0.1-g increments.The SI (mL/2 g) was monitored initially after 16 h and subsequentlyevery 4h thereafter up to a total elapsed time of 48h, at which swellinghad ceased. The SRC tests were conducted following the proceduresdescribed in Lee and Shackelford (2005c), except that the centrifugespeed was set at 3000 rpm versus the 5000 rpm used by Lee andShackelford (2005c) to allow for an increased sensitivity of the resultingmeasurements.

The resulting values of SI and SRC for the BPN were 73mL/2 g and11 mL/g, respectively. These values are noticeably higher than thosetypically reported for traditional sodium bentonite (e.g., Lee andShackelford, 2005c). For example, values of SI and SRC for traditionalsodium bentonite typically range from 25 to 36 mL/2 g and from 5.7to 7.7 mL/g, respectively (Benson and Meer, 2009; Bohnhoff, 2012;Katsumi et al., 2007; Kenney et al., 1992; Kolstad et al., 2004; Leeand Shackelford, 2005b). Also, the values for the SI and SRC of thesodium bentonite that was used to create the BPN were 22mL/2 g and

Fig. 1. Photograph of plastic limit “worms” for traditional (untreated) bentonite andbentonite polymer nanocomposite (BPN).

7.3mL/g, respectively. The greater swelling behavior of the BPN com-pared to the sodium bentonite is in contrast to the LL results, and likelyis attributable to the contribution of the swelling resulting from thesuperabsorbent polymer portion of the BPN (Bohnhoff, 2012; Scalia,2012; Scalia et al., 2011).

2.3. Membrane testing

A schematic of the overall testing apparatus is shown in Fig. 2a, andschematics of the rigid-wall and the flexible-wall cells used in the studyare shown in Fig. 2b and c, respectively. The rigid-wall cell is similar tothat described by Malusis et al. (2001), and the flexible-wall cell issimilar to that described by Kang and Shackelford (2009). The rigid-wall cell is used to test specimens under constant-volume conditionsthroughout the test. As explained by Kang and Shackelford (2009), theflexible-wall cell also maintains constant specimen volume during thestages of the test devoted tomeasuringmembrane behavior, as requiredby the constant-volume assumption inherent in the equation used tocalculate the membrane efficiency (described subsequently). However,some volume change generally occurs during the brief (b5min) liquidrefilling and sampling periods that are required periodically betweenthe longer (typically either 24 or 48 h) membrane testing periods.Also, the flexible-wall cell allows for back-pressure saturation of thespecimen prior to membrane testing and for control of the stressconditions imposed on the specimen (Kang and Shackelford, 2009).

As shown in Fig. 2a, either a rigid-wall or a flexible-wall cell wasconnected via stainless steel tubing to a hydraulic control systemconsisting of a flow-pump with syringes (actuators) used to circulatechemical solutions across both the top and bottom boundaries of thespecimen contained in the cell at the same, constant displacement rate(Kang and Shackelford, 2009;Malusis et al., 2001). Stainless steel tubingwas used to minimize volume change within the testing apparatus andto provide chemical resistance to electrolyte solutions. Two in-linepressure transducers (model Nos. PX181-100G5V or PX209-015G10V,Omega Engineering Inc., Stamford, CT) were used to measure thepressures existing in the boundary liquids, and a differential pressuretransducer (model DP15-64, Validyne Engineering Corp., Northridge,CA) was used to measure the generation of pressure difference acrossthe specimen due to specimen membrane behavior. These pressureswere recorded via a data acquisition (DAQ) system consisting of acircuit board (SCB-68, National Instruments, Austin, TX), a DAQ device(National Instruments, Austin, TX), and the LabVIEW software (NationalInstruments, Austin, TX). Refilling and sampling reservoirs wereused to replenish the syringes with fresh chemical solutions or DIWand to collect circulation outflows for measurement of chemicalparameters (e.g., solute concentrations, electrical conductivity) duringthe aforementioned replenishment and sampling periods at the end ofeach circulation cycle.

The rigid-wall cell (Fig. 2b) consists of an acrylic cylinder with a71.0-mm (2.8-in) inner diameter, a top piston, and a base pedestal.The acrylic cylinder slides over the base pedestal and the top pistonslides into the acrylic cylinder. O-rings on both the base pedestal andtop piston and vacuum grease provide a tight seal between the acryliccylinder and the base pedestal and top piston. The top piston isused to control the thickness (i.e., volume) of the specimen and can belocked in place during testing. The base pedestal and top piston havethree ports. The inflow and outflow ports allow circulation of liquids(e.g., electrolyte solutions) through the porous plastic disks (GenPoreporous sheet TO-6, General Polymer Corp., Reading, PA) located alongthe top and bottom boundaries of the specimen, whereas thirdports located in the center of the top piston and the base pedestalprovide connections to the individual in-line pressure transducersfor measurement of the boundary liquid pressures at the top andbottom of the specimen, or uTop and uBottom, respectively, and the dif-ferential pressure transducer used to measure the pressure difference,ΔP (=uBottom−uTop), across the specimen.

(a)

CirculationInflow, Cob

Circulation Outflow,Ct < Cot

O-rings

Specimen

Base Pedastal

Top Piston

Acrylic Cylinder

Circulation Inflow, Cot

Circulation Outflow,Cb > Cob

AcrylicSpacer

Porous PlasticDisks

Geotextiles

To pressure transducers


AcrylicSpacer

Specimen

CirculationInflow, Cob

Circulation Outflow,Cb > Cob


Circulation Outflow,Ct < Cot

Circulation Inflow, Cot

Top Plate

BottomPlate

AcrylicCylinder

Flexible (Polymer) Membrane

O-rings

Top Cap

AcrylicSpacers

GeotextilesPorous PlasticDisks

Cell Water Accumulator

Cell Water

Applied Confining Stress, c

Air

Water

(b) (c)

Fig. 2. Schematic drawings of: (a) membrane testing apparatus; (b) rigid-wall cell (after Malusis et al., 2001); and (c) flexible-wall cell (after Kang and Shackelford, 2009).


The flexible-wall cell (Fig. 2c) is a triaxial cell with special top andbottom caps as described in Kang and Shackelford (2009). The cellconsists of an acrylic outer cylinder, top and bottom plates, and topand bottom caps. Each top and bottom cap has three ports, i.e., oneport to allow for a circulation inflow, one port to allow for circulationoutflow, and a third port to allow for measurement of the boundaryliquid pressures (i.e., uTop and uBottom) and the pressure difference acrossthe specimen (i.e., ΔP). The bottom cap is attached to a bottom plate.

For the flexible-wall tests, a test specimen with a diameter of102 mm (4.0 in) was placed above a heat-bonded, 140 g/m2, non-woven geotextile on the bottom cap followed by a top cap and a flexible(polymer) membrane that is sealed with o-rings. The flexible mem-brane separated the specimen from the cell water. The outer cylinderwas filled with de-aired water, which may be pressurized to applya confining pressure (i.e., total stress) to the specimen. A cell-water

accumulatorwas connected to the cellwater tomonitor volume change.Backpressure was applied to the specimen through the inflow andoutflow ports in the top and bottom caps.

The flow-pump system is essentially the same as that described byMalusis et al. (2001) and consists of a dual carriage syringe pump(Model 940 or 944, Harvard Apparatus, Holliston, MA), two stainlesssteel actuators (syringes), and stainless steel tubing. Stainless steelcomponents are used to minimize corrosion and apparatus volumechange. The syringes displace liquids at the same, constant rate intothe top and bottom caps of the cell via the inflow port, the liquidscirculate through the porous plastic disks, and then exit via the outflowports. In addition, infusion and withdrawal of the liquids occur atconstant rates, such that the incremental volume of liquid that entersthe cell, ΔVin, is equal to the incremental volume of liquid that exitsthe cell, ΔVout. As a result, there is no volume change in the circulation


system during circulation, such that there can be no flow of liquidthrough the specimen during circulation. Thus, the circulation systemrepresents a closed system, i.e., at least during the periods of the testdevoted to measurement of the membrane behavior of the specimen.

Operation of the flow-pump system consists of a circulation(i.e., testing) phase and a sampling and refilling phase. During the testingphase, the syringes are open to the cell, but closed to the reservoirs byopening valves 1 and 3 and closing valves 6 and 7 (see Fig. 2a). Thesyringes circulate liquids (DIW or electrolyte solutions) through thetop and bottom of the specimens, such that a constant concentrationdifference (ΔC) is maintained across the specimen. If the specimenbehaves as a semipermeable membrane, then a pressure difference,ΔP, is generated in response to the ΔC to counteract the tendencyfor chemico-osmotic flow, which is prevented from occurring dueto maintaining a constant volume during circulation. This pressuredifference is measured both directly by the differential transducer andindirectly as the difference in pressures measured by the two in-linepressure transducers.

The circulation rate of the liquid is adjustable by changing thedisplacement rate for the syringes of the flow pump. The possibledisplacement rates for the pumps and syringes that were used inthis study varied from 9.2 × 10–12 to 4.7 × 10–7 m3/s. The circulationrate was adjusted to mimic “perfectly flushing” boundary conditionssuch that a steady-state ΔP was achieved during the circulation period,i.e., the time between the sampling and refilling periods (e.g., seeMalusis et al., 2001). The circulation rate used in this study was2.3 × 10–10 m3/s, which amounted to a circulation of approximately40 mL every two days, after which the syringes were sampled andrefilled before the start of a new circulation cycle.

During the brief (b5 min) sampling and refilling periods, thesyringes were closed to the cell and opened to the reservoirs by closingvalves 1 and 3 and opening valves 6 and 7 (see Fig. 2a), and thedirections of the syringe displacementwere reversed. The front (inflow)chambers of the syringeswere refilledwith fresh circulation liquid fromthe refilling reservoirs, whereas the circulation outflow liquids fromthe back chambers of the syringes were emptied into the samplingreservoirs. The outflow liquidswere collected and stored for subsequentanalysis (e.g., EC and solute concentration, C).

The membrane tests consisted of two stages, a permeation stageand a chemico-osmotic testing stage. During the permeation stage,specimens were permeated for periods from weeks to months inorder to saturate the specimens. Despite those durations of permeation,little effluent was collected due to the affinity of the BPN for water, thelow k (1–2 × 10−11 m/s) of the BPN, and the presence of excess lowmolecularweight polymer clogging the system (Scalia, 2012). Therefore,the BPN specimens tested in this study essentially representedspecimens that were not flushed of soluble salts prior to membranetesting. This precondition of the BPN specimens was unlike that forthe majority of previous studies evaluating the membrane behavior ofGCL specimens containing traditional sodium bentonite, whereby thespecimens purposely were permeated with DIW to reduce the solublesalt content in the pores of the sodiumbentonite and, therefore, enhancethe overall potential for significant membrane behavior (Malusis andShackelford, 2002a; Malusis et al., 2001; Shackelford and Lee, 2003).Thus, in terms of flushing of soluble salts, the precondition of the BPNspecimens tested in this study was less favorable towards the existenceand magnitude of semipermeable membrane behavior than that forpreviously tested GCLs. However, the potential role any excess lowmolecular weight polymer in the BPN may have played in affecting theresults is uncertain, but likely was minimal at most, i.e., based on theresults discussed subsequently (e.g., see Section 3.3).

Rigid-wall specimens were prepared and permeated directly in therigid-wall membrane cells, whereas flexible-wall specimens werepermeated in separate flexible-wall permeameters prior to beingtransferred to the flexible-wall membrane cell for the second stageof the test. The required mass of dry granular BPN was measured on a

precision balance (Sartorius GP 5202, Goettingen, Germany). The dryBPN then was poured into the mold (rigid-wall cell) or membrane(flexible-wall cell) and leveled off using a plastic top cap to create aflat surface. The dry mass per unit area of the BPN specimens wascontrolled to be 4.8 kg/m2, a typical value for traditional sodiumbentonite contained in a GCL (Scalia, 2012). Theflexible-wall specimenswere permeated under 172 kPa (25 psi) backpressure at an averageeffective stress of 34.5 kPa (5 psi) using the falling headwater-risingtailwater method (ASTM D 5084). Specimens of BPN were permeatedfrom the bottom to the top by simultaneously increasing the headwaterpressure by 17.2kPa (2.5psi) to 189.2kPa (27.5psi) and decreasing thetailwater pressure by 17.2kPa (2.5psi) to 154.8 (22.5psi). The thicknessof the BPN specimens varied from 7 to 11 mm resulting in hydraulicgradients ranging from approximately 500 to 300, respectively. Asdescribed by Shackelford et al. (2000), these relatively highmagnitudesof hydraulic gradients are typical for permeation of GCLs due to the lowk values typically associated with GCLs permeated with water or dilutechemical solutions (i.e., ≤3 × 10−11 m/s), and should not have asignificant influence on the resulting measured k values.

After the permeation stage was complete, the specimens weretransferred to the membrane testing apparatus for measurement ofmembrane behavior. First, the baseline pressure difference, ΔP, wasestablished by simultaneously circulating DIW across both the top andbottom boundaries of the specimen (i.e., Cot = Cob = 0). Then, thecirculating liquid on the top boundary was switched to KCl solutions(i.e., Cot N 0), and the generated pressure difference was recordedcontinuously.

2.4. Measurement of membrane efficiency

Under closed-system boundary conditions such as those imposed inthis study, the membrane efficiency coefficient, ω, is defined as follows(Groenevelt and Elrick, 1976; Malusis et al., 2001):

ω ¼ ΔPΔπ

ð1Þ

where ΔP (b0, since the positive x-direction is assumed downwardfrom the top of the specimen) is the measured chemico-osmoticpressure difference induced across the specimen as a result ofprohibiting chemico-osmotic flux of liquid, and Δπ (b0) is thetheoretical chemico-osmotic pressure difference across an “ideal”semipermeable membrane (i.e., ω = 1) subjected to an applieddifference in solute (electrolyte) concentration (e.g., Olsen et al.,1990). The value of Δπ can be calculated in accordance with the van'tHoff expression in terms of either the source concentrations of KClin the circulation inflows across the top and bottom of the specimen,or the average of the boundary salt concentrations (Malusis et al.,2001). The membrane efficiency coefficient in terms of the source KClconcentrations, designated as ω0, is given as follows (e.g., Kang andShackelford, 2011):

ω0 ¼ ΔPΔπ

��o¼ ΔP

Δπo¼ ΔP

νRTΔCo¼ ΔP

νRT Cob−Cotð Þ ¼ΔP

−νRTCotð2Þ

where ν is the number of ions per molecule of the salt (=2 for KCl), R isthe universal gas constant [8.314 J mol−1 K−1], T is the absolutetemperature (293K in this study corresponding to 20 °C), and Cot (N0)and Cob (= 0) are the initial concentrations of KCl (M) in thesource solutions introduced across the top and bottom specimenboundaries, respectively. In terms of average KCl concentrations, themembrane efficiency coefficient, ωave, is given as follows (e.g., Kangand Shackelford, 2011):

ωave ¼ΔPΔπ

��ave

¼ ΔPΔπave

¼ ΔPνRTΔCave

¼ ΔP

νRT Cb;ave−Ct;ave

� � ð3Þ


where Ct,ave and Cb,ave are the average KCl concentrations across the topand bottom of the specimen boundaries defined as follows:

Ct;ave ¼Cot þ Ct

2; Cb;ave ¼

Cob þ Cb

2ð4Þ

and Cb and Ct are the measured KCl concentrations via calibration withEC in the circulation outflows from the bottom and top of the specimenboundaries, respectively. Since Ct,avebCot and Cb,ave≥Cob, themagnitudeof Δπo will be greater than that of Δπave such that, for the samemeasured value of ΔP,ω0bωave. Thus, membrane efficiency coefficientsbased on source salt concentrations typically are more conservative(lower) than those based on average salt concentrations. However, inthe limit as ω → 1, solutes can neither enter nor exit the specimen,such that Ct,ave → Cot, Cb,ave → Cob, and ωave → ω0 (Kang andShackelford, 2009).

2.5. Testing program

Multiple-stage membrane tests were conducted in this study bysequentially circulating four electrolyte solutions containing 4.7, 9.3,20, and 54mM KCl across the top boundary of the specimen for eachtest, while simultaneously circulating DIW across the bottom boundaryof the specimen. After steady-state conditions were achieved for eachstage of the test, i.e., afterΔP and ECwere constant, the source circulatingsolution was replaced with the subsequent solution containing a higherconcentration of KCl, and this process was repeated until all four stageswere completed. This procedure resulted in test durations ranging

1

10

100

1000

ECTop

ECBottom

Ele

ctric

al C

ondu

ctiv

ity, E

C (

mS

/m)

Time, t (d)

Time, t (wk)

(a) RW1

ECo (9.3 mM KCl)

ECo (4.7 mM KCl)

ECo (20 mM KCl)

ECo (54 mM KCl)

20 mM KClDIW

54 mM KCl

9.3 mM KCl

Reduced circulation

rate

4.7 mM KCl

54 mM KCl

1

10

100

1000

ECTop

ECBottom

Ele

ctric

al C

ondu

ctiv

ity, E

C (

mS

/m)

Time, t (d)

Time, t (wk)

DIW

4.7 mM KCl

9.3 mM KCl

20 mM KCl

(c) FW1

ECo (9.3 mM KCl)

ECo (54 mM KCl)

ECo (20 mM KCl)

ECo (4.7 mM KCl)

-14 0 14 28 42 56 70 84 98

-14 0 14 28 42 56 70 84 98

-2 0 2 4 6 8 10 12 14

-2 0 2 4 6 8 10 12 14

Fig. 3.Measured electrical conductivity across top and bottom boundaries of BPN specim

from 96 to 154 d. The two rigid-wall membrane tests designated asRW1 and RW2 corresponded to specimen thicknesses, L, of 16.7 mmand 8.6 mm and specimen total porosities, n, of 0.92 and 0.80,respectively. The two flexible-wall membrane tests designated as FW1and FW2 corresponded to initial effective stresses, σ′, of 34.5 kPa(5 psi) and 103 kPa (15 psi), respectively, and initial values of nimmediately after consolidation, but prior to membrane testing, of0.95 and 0.84, respectively.

3. Results

3.1. Boundary electrical conductivity values

The measured EC of the circulation outflows versus time for therigid-wall and flexible-wall membrane tests are presented in Fig. 3.The EC of the outflows from both the top and the bottom of thespecimen increased as the solute concentration of the circulation liquidincreased. The increase in EC at the bottom of the specimen can beattributed initially to diffusion of any remnant ions contained withinthe pores of the BPN specimens and eventually to diffusion of Cl− andK+ completely through the specimens from the top boundary.

The differences in electrical conductivity, ΔEC, between the steady-state EC values measured in the circulation outflows across the topand the bottom of the specimens, or ECt and ECb, respectively, relativeto the respective limiting values of EC in the circulation inflows, i.e.,ΔECt=ECt−ECot and ΔECb=ECb−ECob, are plotted versus the averagesalt concentration across the specimen, Cave, in Fig. 4. The values of ΔECfor the top of the specimen are negative, because ECt b ECot due to

1

10

100

1000

ECTop

ECBottom

Ele

ctric

al C

ondu

ctiv

ity, E

C (

mS

/m)

Time, t (d)

Time, t (wk)

DIW4.7 mM

KCl

9.3 mM KCl

20 mM KCl

54 mM KCl

ECo (54 mM KCl)

ECo (20 mM KCl)

ECo (9.3 mM KCl)

ECo (4.7 mM KCl)

(b) RW2

1

10

100

1000

-14 0 14 28 42 56 70 84 98 112 126 140 154

-14 0 14 28 42 56 70 84 98 112 126

-2 0 2 4 6 8 10 12 14 16 18 20 22

-2 0 2 4 6 8 10 12 14 16 18

ECTop

ECBottom

Ele

ctric

al C

ondu

ctiv

ity, E

C (

mS

/m)

Time, t (d)

Time, t (wk)

DIW4.7 mM

KCl

Lost cell pressure

9.3 mM KCl

20 mM KCl

54 mM KCl

(d) FW2 ECo (54 mM KCl)

ECo (20 mM KCl)

ECo (9.3 mM KCl)

ECo

(4.7 mM KCl)

ens during membrane testing in (a,b) rigid-wall cells and (c,d) flexible-wall cells.

-300

-200

-100

0

100

200

300

1 10 100

RW1RW2FW1FW2

ΔEC

= E

C -

EC

o (m

S/m

)

Average Source ConcentrationAcross Specimen, Cave(mM)

Bottom

Top

Fig. 4. The change in electrical conductivity, ΔEC, of non-flushed BPN specimens duringmembrane testing in rigid-wall cells (RW1, RW2) and flexible-wall cells (FW1, FW2) asa function of average of the source salt (KCl) concentration across the specimen.


diffusion of K+ and Cl− ions into the specimen at the top, whereas thevalues ofΔEC for thebase of the specimenarepositive, becauseECbNECobdue to diffusion of any remnant ions initially in the pore waters of thespecimens and eventually including K+ and Cl− ions from the bottomof the specimens into the circulation outflows. As shown in Fig. 4, themagnitude of ΔEC (positive or negative) increases with an increase inCave for all four tests due to the increase in concentration difference,ΔC, across the specimens during the sequential circulation of higherconcentration KCl solutions in the multistage tests. This trend ofincreasing ΔEC with increasing Cave is consistent with increased solutediffusion through the specimen due to an increasing concentrationgradient across the specimen with increasing Cave.

For example, consider the following form of Fick's first law for steady-state diffusive solute mass flux, Jd, in saturated soil (Shackelford andDaniel, 1991):

Jd ¼ −nD� ΔCL

ð5Þ

where n is the porosity of the soil, D⁎ is the effective diffusion coefficientof the diffusing chemical species, and ΔC is the difference in theconcentration of chemical species across a specimen of length L. Asdefined by Shackelford and Daniel (1991), D⁎ is the product of theaqueous-phase diffusion coefficient for a specific chemical species, Do,and the apparent tortuosity factor, τa, or D⁎=τaDo, where 0b τa≤1. Asfurther defined by Malusis and Shackelford (2002b), τa equals theproduct of the matrix tortuosity factor, τm (0b τm≤1), representing theportion of the apparent tortuosity factor attributed to the geometry ofthe interconnected pore structure, and a restrictive tortuosity factor,τr (0bτr≤1), resulting from other factors that can affect solute diffusionthrough soil, such as anion exclusion and increased viscosity of the liquidadjacent to clay particles. In the case where semipermeable membranebehavior is evident, Manassero and Dominijanni (2003) propose that τrbe taken as an inverse linear function of the membrane efficiencycoefficient, ω, or τr = 1 − ω. This correlation is supported by theexperimental results reported by Malusis and Shackelford (2002a) andMalusis et al. (2013), as well as by the theoretical model proposed byDominijanni and Manassero (2012). Thus, as ω approaches unitycorresponding to a perfect membrane, τr, D⁎ and Jd all approach zero asrequired by definition for ideal or perfect membrane behavior.

Of all the factors affecting Jd, n and L should be relatively constant fora given test specimen. In terms of D⁎, results of numerous tests haveclearly shown that ω decreases approximately semi-log linearly withincreasing Cave (e.g., Shackelford et al., 2003). Thus, Jd also would be

expected to increase with increasing Cave, since D⁎ is expected toincrease with decreasing ω. Finally, when Cave is defined with respectto the initial source concentrations, such that Cave=(Cot+ Cob) / 2 andCob is zero (i.e., DIW), ΔC (= Cob− Cot) in Eq. (5) is directly related toCave, or ΔC = –Cot = 2Cave. Thus, as Cave increases, ΔC increases and,therefore, Jd increases.

Based on the aforementioned considerations, as ΔC across thespecimen increases from 4.7 to 54 mM during a given test, Jd acrossthe top and bottom boundaries increases into and out from thespecimen, respectively, resulting in a parallel effect on the magnitudeof theΔEC. In addition, theΔEC is lower for the testswith lower porosity(i.e., n= 0.80 for RW2 vs. n= 0.92 for RW1) and for the test with ahigher initial effective stress, σ′ (i.e., σ′=103 kPa (15 psi) for FW2 vs.σ′=34.5 kPa (5 psi) FW1), also corresponding to a lower initial valueof n (i.e., n=0.84 for FW2 vs. n=0.95 for FW1). Finally, the ΔEC valuesfor FW1 and FW2 are greater than those for RW1 and RW2, which isconsistent with the lower membrane efficiencies (higher D⁎ values)measured for the specimens in the flexible-wall cells, as discussedsubsequently.

3.2. Chemico-osmotic pressures

The measured chemico-osmotic pressure differences (–ΔP N 0) arepresented in Fig. 5. The baseline differential pressure, −ΔP0, inducedduring the circulation of DIW along both boundaries (i.e., Cot=Cob=0)varied from 6.7 kPa for RW2 to −5.0 kPa for FW2. As described inMalusis et al. (2001), possible reasons for a nonzero baseline differentialpressure include slight differences in the hydraulic resistance of theporous plastic disks along the top and bottom of the specimen, whichwould result in different head losseswithin the disks between the pointswhere the circulation liquids enter the disks and the middle of the diskwhere the pressure difference is measured, and slightly different flowrates across the top and bottom of the specimen due to slight differencesin themachining of the syringes. Therefore, corrected or effective valuesof the differential pressure,−ΔPe [=−ΔPss−(−ΔP0)], as summarizedin Table 3, were used to calculate ω0 and ωave.

The values of−ΔPe for test RW1 (n=0.92) increased from 18.4kPaduring the circulation of 4.7mM KCl across the top to the specimen to59.0 kPa during the circulation of 54 mM KCl (see Fig. 5a). The −ΔPefor RW2 (n = 0.80) was higher than that for the RW1, in that the−ΔPe for RW2 increased from 21.3 kPa during the circulation of4.7 mM KCl across the top of the specimen to 59.0 kPa during thecirculation of 54mM KCl for RW2 (see Fig. 5b). The steady-state −ΔPefor RW1 and RW2 during the circulation of 54mM KCl was lower thanthe aforementioned peak value of −ΔPe as a result of post-peakdegradation resulting from diffusion of solutes into and subsequentlythrough the specimen (e.g., maximum −ΔPe = 59.0 kPa versussteady-state −ΔPe = 44.3 kPa for RW1). This time-dependent, post-peak degradation of −ΔPe, especially at higher KCl concentrations,is consistent with the same trend previously reported by others(e.g., Malusis and Shackelford, 2002a), and has been attributed to pro-gressive compression of the diffuse double layer surrounding individualclay particles due to increasing salt concentration within the poresresulting from diffusion of KCl from the upper boundary (Fritz, 1986;Malusis and Shackelford, 2002a; Mazzieri et al., 2010; Shackelford andLee, 2003; Shackelford et al., 2003).

The values of−ΔPe for the flexible-wall membrane tests were lowerthan those for the rigid-wall tests, with −ΔPe for FW1 increasing from12.4kPa during circulation of 4.7mM KCl to 14.6kPa during circulationof 20mM KCl (see Fig. 5c), and−ΔPe for FW2 increasing from 10.3kPaduring circulation of 4.7mMKCl to 35.7kPa during circulation of 20mMKCl (see Fig. 5d). This trendof a slightly lower−ΔPe for theflexible-wallversus rigid-wall membrane tests also was observed by Kang andShackelford (2009) for their tests performed using a specimen of aGCL, and was attributed to the different stress conditions and volumecontrol conditions in the two types of cells.

-13.8

0

13.8

27.6

41.4

55.2

69

-2

0

2

4

6

8

10

Che

mic

o-O

smot

ic P

ress

ure

Diff

eren

ce, -

ΔP (

kPa)

Time, t (d)

Chem

ico-Osm

otic Pressure

Difference, -Δ

P (psi)DIW

Time, t (wk)

4.7 mM 9.3 mM 20 mM

Flushed air bubble from line

KCl Solutions

54 mM

Reduced circulation rate

(a)

RW1n = 0.92

-13.8

0

13.8

27.6

41.4

55.2

69

-2

0

2

4

6

8

10

Che

mic

o-O

smot

ic P

ress

ure

Diff

eren

ce, -

ΔP (

kPa)

Time, t (d)

Chem

ico-Osm

otic Pressure

Difference, -Δ

P (psi)

DIW

Time, t (wk)

4.7 mM

(b)

9.3 mM 20 mM 54 mM

Flushed lines

RW2n = 0.80

KCL Solutions

-13.8

-6.9

0

6.9

13.8

20.7

27.6

34.5

41.4

Differential Pressure, - ΔP-Δu = u

Top - u

Bottom

-2

-1

0

1

2

3

4

5

6

Che

mic

o-O

smot

ic P

ress

ure

Diff

eren

ce, -

ΔP &

-Δu

(kP

a)

Time, t (d)

Chem

ico-Osm

otic Pressure

Difference, -Δ

P &

-Δu (psi)

DIW

Time, t (wk)

4.7 mM 9.3 mM 20 mMKCl Solutions

54 mM

(c)

FW1σ

c = 207 kPa (30 psi)

ubp

= 172 kPa (25 psi)

σ' = 34.5 kPa (5 psi)

-13.8

0

13.8

27.6

41.4

55.2Differential Pressure, -ΔP-Δu = u

Top - u

Bottom

-2

0

2

4

6

8

-14 -7 0 7 14 21 28 35 42 49 56 63 70 77 84 91 98

-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-14 0 14 28 42 56 70 84 98 112 126 140 154 168

-2 0 2 4 6 8 10 12 14 16 18 20 22 24

-14 -7 0 7 14 21 28 35 42 49 56 63 70 77 84 91 98

-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-14 -7 0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126

-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Che

mic

o-O

smot

ic P

ress

ure

Diff

eren

ce, -

ΔP &

-Δu

(kP

a)

Time, t (d)

Chem

ico-Osm

otic Pressure

Difference, -Δ

P &

-Δu (psi)DIW

Time, t (wk)

4.7 mM 9.3 mM 20 mM

σc = 172 kPa (25 psi)

ubp

= 69 kPa (10 psi)

σ' = 103 kPa (15 psi)Lost cell pressure

KCl Solutions

DAQ down

54 mM

(d)

FW2

Fig. 5. Measured chemico-osmotic pressure differences across BPN specimens tested in rigid-wall cells (a,b) and flexible-wall cells (c,d) (n = initial specimen porosity; σc, ubp, andσ′ = confining stress, back pressure, and initial effective stress before membrane testing, respectively).


Another possible explanation for the difference in membrane per-formance of the BPN specimens in the rigid-wall cells versus theflexible-wall cells may be the circumstances under which the BPNspecimens were initially exposed to water. In the rigid-wall cell, theBPN specimens were permeated with water in the confined (fixedvolume) cell that restricted swelling of the super absorbent polymerportion of the BPN and limited flushing of the excess low molecularweight polymer from the system, resulting in smaller hydraulically activepores (i.e., pores are smaller andmore pores are clogged with excess low

molecular weight polymer that exists in the BPN), as described by Scalia(2012). In contrast, the BPN specimens in the flexible-wall cells werepermeated with water under an essentially constant, average value forσ′ of 34.5 kPa (5 psi) or 103 kPa (15 psi), but the specimens were ableto swell due to the propensity for montmorillonite and the super-absorbent polymer portion of the BPN to absorb water. This swellingresulted in larger hydraulically active pores and additional flushing ofthe excess lowmolecularweight polymer from the systemand, therefore,less clogging of pores prior to the membrane testing stage of the test.

Table 3Results of membrane testing on BPN using rigid-wall (RW) and flexible-wall (FW) cells.

Test designation Specimenporosity, n

Initial effective stress,σ′ [kPa (psi)]

Source KClconcentration,Cot (mM)

Maximum chemico-osmotic pressuredifference (kPa)

Effective (measured)chemico-osmoticpressure difference,−ΔPe (kPa)

Membrane efficiencycoefficient @steady-state

−Δπoa −Δπave

b ω0a ωave

b

RW1 0.92 NA 4.7 22.9 21.9 18.4 0.80 0.849.3 45.3 43.6 30.1 0.65 0.69

20 97.4 83.2 41.6 0.43 0.5054 263 206 44.3 0.17 0.21

RW2 0.80 NA 4.7 22.9 22.8 21.3 0.88 0.889.3 45.3 44.5 34.9 0.73 0.76

20 97.4 88.0 47.5 0.46 0.5154 263 213 54.3 0.20 0.25

FW1 0.94–0.95

34.5 (5) 4.7 22.9 19.3 12.4 0.55 0.639.3 45.3 32.2 14.6 0.32 0.45

20 97.4 60.7 13.4 0.14 0.2254 263 150 9.9 0.04 0.07

FW2 0.78–0.84

103 (15)34.5 (5)

4.7 22.9 22.5 10.3 0.46 0.459.3 45.3 37.6 17.3 0.38 0.46

20 97.4 70.0 24.2 0.25 0.3554 263 165 26 0.10 0.15

a Based on difference in initial (source) KCl concentrations.b Based on difference in average, boundary KCl concentrations.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RW1

RW2 FW1 FW2

0

10

20

30

40

50

60

70

80

90

100M

embr

ane

Effi

cien

cy C

oeffi

cien

t, ω

Average Source Concentration Across Specimen, Cave (mM)

Mem

brane Efficiency (%

)

?

(a)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

1 10 100

RW1

RW2

FW1 FW2

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ωav

e


Mem

brane Efficiency (%

)

?

(b)

ο

Fig. 6. Steady-state membrane efficiency coefficients of BPN specimens duringmembranetesting in rigid-wall cells (RW1, RW2) and flexible-wall cells (FW1, FW2) as a functionof average source salt (KCl) concentration across the specimen: (a) ω values based oninitial (source) boundary concentrations, ω0; (b) ω values based on average boundaryconcentrations, ωave.


3.3. Membrane efficiency coefficients

The values of ω0 and ωave for the membrane tests based on Eqs. (2)and (3), respectively, are shown as a function of Cave (=ΔCo/2) in Fig. 6.With the exception of the values of ω0 and ωave for test FW2 with the4.7 mM KCl solution, the results are consistent with previous results(e.g., Kang and Shackelford, 2009; Malusis and Shackelford, 2002a) inthat (1) the values of ωave are generally greater than the values of ω0

for a given value of Cave, (2) the values of both ω0 and ωave decreaseapproximately semi-log linearly with increasing Cave, (3) the values ofω0 or ωave from the rigid-wall tests are greater than those from theflexible-wall tests, and (4) the values of ω0 or ωave are greater forspecimens at higher initial effective stresses, σ′, and specimens withlower initial porosities, n.

The values of ω0 and ωave for test FW2 with the 4.7 mM KClsolution are anomalous due to an unexpected loss of system pressure14d into the circulation of 4.7mM KCl resulting in loss of the appliedcell pressure. Upon restoration of the system pressure, both the in-lineand differential pressures returned to the same values prior to lossof system pressure. However, the calculated ω0 and ωave were lowerthan the expected values, in that the values of ω0 and ωave fortest FW2 were expected to be greater than the same respectivevalues for test FW1, because test FW2 was conducted at a highereffective stress than test FW1 [103 kPa (15 psi) vs. 34.5 kPa (5 psi),respectively].

Finally, the approximate semi-log linear decrease in ω0 and ωave

with increasing Cave, which as previously noted is consistent with theresults of several other membrane studies involving traditional sodiumbentonite and KCl solutions, suggests that the mechanisms of themembrane behavior for the BPN specimens are consistent with thosepreviously associated with traditional sodium bentonite, i.e., physico-chemical interactions between the salts in the pores of the diffusedouble layers (DDL) surrounding the BPN particles, such that increasein concentration results in progressively greater collapse of the DDL,progressively larger pores, and subsequent decrease in observedmembrane behavior (e.g., see Kang and Shackelford, 2011; Malusisand Shackelford, 2002a). If the mechanism for solute restrictionhad been physical clogging of the pores by excess polymer thatwas not flushed from the specimens during membrane testing, thenthis trend of decreasing ω0 and ωave with increasing log Cave wouldnot necessarily have been expected, i.e., unless the physico-chemical


behavior of the polymer is the same as that of sodium bentonite.Instead, physical pore clogging by excess polymer likely would haveresulted in an observed membrane behavior that was more inde-pendent of the Cave. Thus, the similarity in the membrane behavior ofthe BPN specimens with that previously shown for sodium bentoniteunder the same conditions suggest that the mechanism responsiblefor the membrane behavior observed for the BPN specimens in thisstudy is that same as that previously reported for specimens of sodiumbentonite.

-3

-2

-1

0

1

2

3

4

Incr

emen

tal V

olum

e C

hang

e, -

ΔV (

mL)

Time, t (d)

Time, t (wk)

4.7 mM

KCl Solutions

DIW 54 mM20 mM9.3 mM

-20

-10

0

10

20

30

40

50

Cum

ulat

ive

Vol

ume

Cha

nge,

Σ(-

ΔV)

(mL)

Time, t (d)

Time, t (wk)

(c) FW1

KCl Solutions

20 mM 9.3 mM 4.7 mM DIW 54 mM

-50

-40

-30

-20

-10

0

10

20

30

40

50

Cum

ulat

ive

Vol

umet

ric S

trai

n, Σ

(-ΔV

/Vo)

(%

)

Time, t (d)

Time, t (wk)

(e) FW1 Vo = 191 mL

54 mMDIW4.7 mM 9.3 mM 20 mM

KCl Solutions

-14 0 14 28 42 56 70 84 98

-2 0 2 4 6 8 10 12 14

-14 0 14 28 42 56 70 84 98

-2 0 2 4 6 8 10 12 1

-14 0 14 28 42 56 70 84 98

-2 0 2 4 6 8 10 12 14

(a) FW1

Fig. 7. Incremental volume change (a,b), cumulative volume change (c,d) and cumulative volumcell: (a,c,e) FW1; (b,d,f) FW2. (Note: Vo=initial specimen volume).

3.4. Volume changes

As described in Kang and Shackelford (2009), although drainagewas not allowed to occur during the circulation periods of the mem-brane tests conducted using the flexible-wall cells, drainage from thespecimen did occur during the brief (b5 min) refilling and samplingperiods of the test when the valves to the reservoirs and the cell (valves1, 3, 6 and 7 in Fig. 2a) were simultaneously and temporarily opened tore-establish the backpressure. Kang and Shackelford (2009) attributed

-3

-2

-1

0

1

2

3

4

Incr

emen

tal V

olum

e C

hang

e, -

ΔV (

mL)

Time, t (d)

Time, t (wk)

Lost cell pressure

4.7 mM

KCl Solutions(b) FW2

20 mM 9.3 mM DIW 54 mM

-20

-10

0

10

20

30

40

50

Cum

ulat

ive

Vol

ume

Cha

nge,

Σ(-

ΔV)

(mL)

Time, t (d)

Time, t (wk)

(d) FW2

Lost cell pressure

KCl Solutions

20 mM 9.3 mM 4.7 mM DIW 54 mM

-50

-40

-30

-20

-10

0

10

20

30

40

50

-14 0 14 28 42 56 70 84 98 112 126

-2 0 2 4 6 8 10 12 14 16 18

-14 0 14 28 42 56 70 84 98 112 126

-2 0 2 4 6 8 10 12 14 16 18

-14 0 14 28 42 56 70 84 98 112 126

-2 0 2 4 6 8 10 12 14 16 18

Cum

ulat

ive

Vol

umet

ric S

trai

n, Σ

(-ΔV

/Vo)

(%

)

Time, t (d)

Time, t (wk)

(f) FW2 Vo = 75 mL

Lost cellpressure

KCl Solutions

20 mM9.3 mM4.7 mMDIW 54 mM

etric strain (e,f) versus time duringmembrane testing of BPN specimens in a flexible-wall


the drainage to an increment in effective stress resulting from physico-chemical interactions between the pore water in the bentonite and theindividual particles of bentonite. Theoretically, an increase in saltconcentration in the pore water results in decrease in the repulsiveelectrical forces relative to the adsorptive forces between individual soilparticles (i.e., so-called R − A effect), such that the effective stress inthe soil also increases resulting in compression of the soil (i.e., provideddrainage is allowed). Such an increase in salt (KCl) concentration inthe pore water is consistent with diffusion of the KCl into the specimensas previously described. This phenomenon has been referred to asosmotic consolidation (e.g., Mitchell, 1976; Barbour and Fredlund,1989; Di Maio, 1996).

Accordingly, volume changes were recorded for themembrane testsconducted in the flexible-wall cell using the cell-water accumulatorattached to the flexible-wall cell (see Fig. 2c). The resulting measuredvalues of incremental volume change, −ΔV, cumulative volumechange, Σ(−ΔV), and cumulative volumetric strain, ∑(–ΔV) / Vo, arepresented in Fig. 7 and the resulting differences in bulk specimenporosities are indicated in Table 4. The ΔV values generally werenegative, indicating that the volume of the specimen decreasedduring the test (i.e., a final volume less than the initial volume). Themagnitude of −ΔV increased with increasing concentration of KCl. ForFW1 with σ′ = 34.5 kPa (5 psi), −ΔV was less than ±0.5 mL duringthe circulation of DIW and 4.7mM KCl, whereas −ΔV was as much as3.2mL during the circulation of 54mM KCl (Fig. 7a). The Σ(−ΔV) was46.3 mL (Fig. 7c), which corresponds with a cumulative volumetricstrain, Σ(−ΔV) / Vo, of 25% (see Fig. 7e). The incremental volumetricstrains (not shown) for FW1 were less than −0.5% during thecirculation of DIW, 4.7 mM KCl, and 9.3 mM KCl, but increased to asmuch as 1.7% during the circulation of 54mM KCl, which is consistentwith an increasing R−A effect with an increasing salt concentration.

In contrast, for FW2withσ′=103kPa (15psi),−ΔVwas asmuch as−2 mL during circulation of DIW and 4.7 mM KCl, but −ΔV wasless than 1.5 mL during circulation of 54 mM KCl (Fig. 7b). Duringthe circulation of both DIW and the 4.7 mM KCl solution after theloss of cell pressure, the specimen swelled likely as a result of the affinityfor water of the superabsorbent polymer portion of the BPN. However,during the remainder of the test, the specimen compressed, i.e., volumeof the specimen decreased, which is similar to the results for FW1 andconsistent with the R − A effect with increasing salt concentrations.The Σ(−ΔV) was 20.7 mL (Fig. 7d), which corresponds to a value forΣ(−ΔV) / Vo of 27% (see Fig. 7f). The incremental volumetric strains(not shown) were as much as −2.5% during the circulation of DIWand 4.7mM KCl after the loss of cell pressure, were generally less than1.0% during the circulation of 9.3 mM KCl and 20 mM KCl, and thenincreased to as much as 2.0% during the circulation of 54mM KCl.

3.5. Limits on membrane behavior

As previously described, an approximately semi-log linearly rela-tionship between ω and Cave has been shown to exist on the basis ofnumerous experimental results (e.g., see Shackelford et al., 2003). This

Table 4Bulk porosities of BPN specimens during membrane testing in flexible-wall (FW) cells.

Stage of test Test designation

FW1[σ′=34.5 kPa (5 psi)]

FW2[σ′=103 kPa (15 psi)]

After consolidation 0.95 0.84After DIW circulation 0.95 0.87After 4.7mM KCl circulation 0.95 0.84After 9.3mM KCl circulation 0.95 0.84After 20mM KCl circulation 0.95 0.82After 54mM KCl circulation 0.94 0.78

Note: σ′ = initial effective stress in specimen after consolidation but prior to DIWcirculation

relationship is illustrated schematically in Fig. 8, where the slope ofrelationship is defined as the membrane index, Iω, and the limitingvalues of Cave are defined as the threshold concentration, Cave,ω,corresponding to the concentration below which membrane behavioroccurs (i.e., ω N 0 for Cave b Cave,ω), and the perfect membrane concen-tration, Cave,pm, corresponding to the concentration below which thematerial behaves as a perfect membrane (i.e., ω=1 for Cave b Cave,pm).As depicted in Fig. 8, the trend between ω versus Cave in reality likelybecomes non-semi-log linear as the concentrations approach Cave,pmand Cave,ω. However, the exact functional forms of this non-semi-loglinear behavior are unknown, such that semi-log linear regressionscan be used as a first approximation (Shackelford et al., 2003).

Accordingly, semi-log linear regressions to the data presented inFig. 6 are shown in Fig. 9 along with the corresponding coefficientsof determination, r2, and the resulting regression coefficients aresummarized in Table 5. The r2 values indicate that the regressionsare reasonable (r2≥0.94) over the range of measured data with some-what better regressions based on the results of the rigid-wall tests(i.e., r2 ≥ 0.99) versus those for the flexible-wall tests (i.e., r2 ≥ 0.94).The threshold concentrations, Cave,ω, range from 25mM for FW1 basedon ω0 to 68mM for RW2 based on ωave. The values for the membraneindex, Iω, ranged from 0.60 to 0.67 for tests RW1 based on ωave andRW2 based on ω0, respectively, whereas the Iω values ranged from0.38 to 0.55 for tests FW2 based on ω0 and FW1 based on ωave,respectively. In general, the threshold concentrations based on thetwo flexible-wall tests were lower than those based on the two rigid-wall tests. The higher threshold concentrations for the rigid-wall testsindicate that the BPNunder these test conditionswill exhibitmembranebehavior at higher Cave which is advantageous in practical applications.This increased performance of BPN in the rigid-wall cells relative to BPNin the flexible-wall cells cannot be attributed to the porosity alone,because the porosities of the RW and FW tests were similar (RW1similar to FW1 and RW2 similar to FW2; see Table 5).

4. Discussion

4.1. Comparison of the limits to membrane behavior

For comparison to the limits of membrane behavior for the BPNspecimens as summarized in Table 5, the results of semi-log linear

Fig. 8. Schematic illustration of limiting concentrations and slope of approximate semi-logarithmic relationship between membrane efficiency, ω, and average boundary saltconcentration (redrawn after Shackelford et al., 2003). [Note:ωref=referencemembraneefficiency coefficient at log Cave = 0; Cave,ω = threshold concentration corresponding toω = 0; Cave,pm = perfect membrane concentration corresponding to ω = 1; Iω = themembrane index].

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RW1RW2FW1FW2

0

10

20

30

40

50

60

70

80

90

100M

embr

ane

Effi

cien

cy C

oeffi

cien

t, ω

Average Source ConcentrationAcross Specimen, Cave (mM)

Mem

brane Efficiency (%

)

?

(a)r2 = 0.99

r2 = 1.0

r2 = 0.94

r2 = 1.0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

1 10 100

RW1RW2FW1FW2

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ωav

e


Mem

brane Efficiency (%

)

?

(b)r2 = 0.99

r2 = 0.99

r2 = 0.98

r2 = 0.99

ο

Fig. 9. Semi-log linear regressions of membrane efficiency coefficients for BPN specimensduring membrane testing in rigid-wall cells (RW1, RW2) and flexible-wall cells (FW1,FW2) as a function of average salt (KCl) concentration difference across the specimen:(a) ω values based on initial (source) boundary concentrations, ω0; (b) ω values basedon average boundary concentrations, ωave.


regressions to the measured values of ω versus logarithm of Cavereported by Kang and Shackelford (2009) and Malusis and Shackelford(2002a)for traditionalGCL containing sodiumbentonite are summarizedin Table 6. The threshold concentrations, Cave,ω, for the BPN specimenswere greater than those for the GCL specimens tested in both rigid-wall cells and flexible-wall cells. For example, the highest Cave,ωdetermined for a BPN specimen in a rigid-wall cell of 67.7 mM is 1.41times greater than the value of 48.0mM for the GCL specimen, whereasthe highest value for Cave,ω of 62.2mM determined for a BPN specimen

Table 5Results of regression analyses for the BPN specimens tested in this study (see Fig. 9).

Test designation Specimenporosity, n

Initial effective stress,σ′ [kPa (psi)]

Definition of membefficiency coefficien

RW1 0.92 NA ω0

ωave

RW2 0.80 NA ω0

ωave

FW1 0.94–0.95 34.5 (5) ω0

ωave

FW2 0.78–0.84 103 (15) ω0

ωave

a ω0=membrane efficiency coefficient based on initial (source) salt concentrations; ωave=b ωref=referencemembrane efficiency coefficient corresponding to log Cave=0; Iω=theme

in a flexible-wall cell with σ′=103kPa (15 psi) is 2.38 times the valueof 26.1 mM reported for the GCL specimen tested under similar stressconditions.

As previously mentioned, the higher Cave,ω values for the BPNcompared to the GCL suggest that the BPN would presumably exhibitmembrane behavior at higher concentrations. The superior membraneperformance of the BPN compared to a GCL is likely due, in part, to thesuperabsorbent polymer portion of the BPN resulting in higher swelland smaller hydraulically active pores resulting in enhancedmembranebehavior (Scalia, 2012). Also, clogging of any hydraulically active poresby excess low molecular weight polymer also may have contributedto the enhanced membrane behavior of the BPN relative to that of aGCL containing traditional bentonite (Scalia, 2012). However, thelong-term sustainability of the possible contribution of any excess lowmolecular weight polymer to the enhanced membrane behavior of theBPN is unknown. Additional studies on BPN specimens that have beenflushed to remove the excess lowmolecularweight polymermay answerthis issue, although available evidence (e.g., Scalia, 2012) indicates thatseveral years likely will be required to flush any excess polymer priorto membrane testing, i.e., assuming flushing by permeation.

4.2. Comparison of membrane efficiency coefficients

The measured values of ω from this study are compared withthose previously reported for tests conducted using GCLs (Kang andShackelford, 2009; Malusis and Shackelford, 2002a) in Fig. 10. Forthe rigid-wall membrane tests performed using the BPN, values ofboth ω0 and ωave are higher than those reported by Malusis andShackelford (2002a) for a GCL that was 8-mm thick (i.e., L = 8 mm)with a lower porosity of 0.74 (see Fig. 10a,b). Also, values of both ω0

and ωave for the flexible-wall membrane tests conducted in this studyare higher than those reported by Kang and Shackelford (2009) for aGCL under similar effective confining stresses of 34.5 kPa (5 psi) and103kPa (15psi) (see Fig. 10c,d). In addition, unlike the GCL specimensthat were flushed (leached) of soluble salts by permeation with DIWprior to membrane testing to enhance the potential for significantmembrane efficiency, the higher values of ω0 and ωave for the testsperformed in this study using the BPN were achieved without flushing(leaching) of soluble salts from the BPN specimens prior to membranetesting. Thus, the BPN specimens not only reflected greater membraneefficiencies than previously reported for specimens of a GCL containinga traditional bentonite, but also did so presumably under more disad-vantageous initial conditions in terms of soluble salts. However, thelong-term membrane behavior of a flushed BPN specimen may betemporarily variable as the excess low molecular weight polymer maybe removed from the material with time resulting in a decrease inthe observed membrane behavior. Thus, even though the trends inthe observed membrane behaviors with increasing KCl concentrationsfor the BPN specimens shown in Fig. 10 are similar to those for theGCL, suggesting that the mechanisms responsible for the membrane

raneta

Regression results:ω=ωref− Iω · log Cave

bLimiting average concentrations,Cave (mM)

ωref Iω r2 @ ω=0, Cave,ω @ ω=1,Cave,pm

1.0 0.62 1.0 50 1.21.1 0.60 1.0 61 1.41.1 0.67 0.99 52 1.61.1 0.62 0.99 68 1.70.68 0.49 0.94 25 0.220.81 0.55 0.98 30 0.460.62 0.38 1.0 47 0.100.75 0.42 0.99 62 0.25

membrane efficiency coefficient based on average boundary salt concentrations.mbrane index as defined by Shackelford et al. (2003) and illustrated schematically in Fig. 8.

Table 6Results of regression analyses shown based on data from the literature for geosynthetic clay liners.

Type of cell Porosity, n Initial effective stress,σ′ [kPa (psi)]

Definition of membraneefficiency coefficienta

Regression results for:ω=ωref− Iω · log Cave

bLimiting averageconcentrations, Cave (mM)

ωref Iω r2 @ω=0,Cave,ω

@ω=1,Cave,pm

Rigid-wallc 0.74 NA ω0 0.81 0.50 1.0 44 0.42ωave 0.83 0.49 1.0 48 0.45

Flexible-walld 0.79–0.81 34.5 (5) ω0 0.49 0.39 0.92 18 0.049ωave 0.65 0.51 0.94 19 0.21

Flexible-walld 0.79–0.80 103 (15) ω0 0.57 0.41 0.94 24 0.092ωave 0.68 0.48 0.95 26 0.22

a ω0=membrane efficiency coefficient based on initial (source) salt concentrations;ωave =membrane efficiency coefficient based on average boundary salt concentrations.

b ωref=referencemembrane efficiency coefficient corresponding to log Cave=0; Iω=themembrane index as defined by Shackelford et al. (2003) and illustrated schematically in Fig. 8.c From Malusis and Shackelford (2002a).d From Kang and Shackelford (2011).


behaviors in both the BPN and the GCl are also similar, additional long-term tests on flushed BPN specimens may be required to more fullyunderstand the behavior of the BPN under these conditions.

The ratios of the values ofω for the BPN specimens measured in thisstudy to the previously reported values of ω for the GCL specimens, orRω, BPN/GCL, are shown in Fig. 11. Because of slight differences betweenthe values of Cave used in the present study for the BPN specimensrelative to those used in the previous studies for the GCL specimens,the values of ω used to calculate the values of Rω, BPN/GCL shown in

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

BPN (n = 0.92)BPN (n = 0.80)GCL (n = 0.74)

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ω


Mem

brane Efficiency (%

)

(a) (

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

(c)

BPN (σ' = 34.5 kPa (5 psi))BPN (σ' = 103 kPa (15 psi))GCL (σ' = 34.5 kPa (5 psi))GCL (σ' = 103 kPa (15 psi))

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ω


Mem

brane Efficiency (%

)

?

(

οο

Fig. 10. Steady-state membrane efficiency coefficients of non-flushed BPN specimens versusShackelford, 2002a) or (c,d) flexible-wall cells (from Kang and Shackelford, 2010) as a functio(source) boundary concentrations, ωo; (b,d) ω values based on average boundary concentratio

Fig. 11 were based on the regressed semi-log linear expressions givenin Tables 4 and 5. The resulting values of Rω, BPN/GCL should be reasonablyaccurate given the high values of r2 for the regressions as previouslynoted. Several observations are apparent based on the results shownin Fig. 11.

First, regardless of type of cell (i.e., rigid-wall vs. flexible-wall) or thebasis for calculatingω (i.e.,ω0 vs.ωave), the values of Rω, BPN/GCL increasewith increasing Cave. This trend occurs as a result of the trend ofdecreasingωwith increasing Cave (e.g., Fig. 10). Thus, the improvement

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

BPN (n = 0.92)BPN (n = 0.80)GCL (n = 0.74)

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ωav

e


Mem

brane Efficiency (%

)

b)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 10 100

d)

BPN (σ' = 34.5 kPa (5 psi))BPN (σ' = 103 kPa (15 psi))GCL (σ' = 34.5 kPa (5 psi))GCL (σ' = 103 kPa (15 psi))

0

10

20

30

40

50

60

70

80

90

100

Mem

bran

e E

ffici

ency

Coe

ffici

ent,

ωav

e


Mem

brane Efficiency (%

)

?

those of flushed GCL specimens tested in either (a,b) rigid-wall cells (from Malusis andn of average salt (KCl) concentration across the specimen: (a,c) ω values based on initialns, ωave. (Note: n= initial specimen porosity; σ′= initial effective stress).

1

2

3

4

5 Rigid-Wall Tests

BPN @ n = 0.92BPN @ n = 0.80BPN @ n = 0.92BPN @ n = 0.80

100

200

300

400

500R

ω, B

PN

/GC

L


Closed: ωo

Open: ωave

Rω

, BP

N/G

CL (%

)

1

2

3

4

5

100

200

300

400

500

0 5 10 15 20 25 30 35

0 5 10 15 20

(b)

BPN @ σ' = 34.5 kPaBPN @ σ' = 103 kPaBPN @ σ' = 34.5 kPaBPN @ σ' = 103 kPa

Rω

, BP

N/G

CL

Closed: ωo

Open: ωave


Rω

, BP

N/G

CL (%

)

(a)

Flexible-Wall Tests

Fig. 11. Ratios of membrane efficiency coefficients a of non-flushed BPN specimens testedin this study to those of previously tested flushed specimens of a GCL, Rω, BPN/GCL, as afunction of the average source salt (KCl) concentration across the specimen: (a) ω valuesbased on rigid-wall tests for BPN specimens at given initial specimenporosity,n, relative toGCL specimen at n = 0.74; (b) ω values based on flexible-wall tests for BPN and GCLspecimens at the same initial effective stress, σ′.


inmembrane performance offered by the BPN is greater with increasingCave, i.e., up to the extent that membrane behavior exists (i.e., forCave b Cave,ω). Second, regardless of type of cell, the difference betweenthe values of Rω, BPN/GCL based on ω0 versus those based on ωave is lessthan 6% for Cave less than 20 mM KCl in the case of tests conductedusing rigid-wall cells (Fig. 11a) and less than 8% for Cave less than10 mM KCl in the case of tests conducted using flexible-wall cells(Fig. 11b). The greater differences in the case of the flexible-wall testresults can be attributed, in part, to the greater divergence from linearityin the trends of ω versus logarithm of Cave, especially for values of Cavegreater than 10 mM (e.g., compare Fig. 10a,b vs. Fig. 10c,d). Finally,a lower porosity, n, for the BPN specimens tested in rigid-wall cellscorrelates with greater values of Rω, BPN/GCL (see Fig. 11a), which isconsistent with expected behavior. However, this relative difference iscontradicted in the case of the BPN specimens tested in the flexible-wall specimens,where a lower initial effective stress,σ′ (i.e., higher initialporosity) correlates with greater values of Rω, BPN/GCL (see Fig. 11b). Thedifference in trends likely results, in part, from the different confiningconditions offered by the two types of cells and the resulting volume

changes that occur during the tests conducted using flexible-wall cells,as previously described (e.g., see Fig. 7).

Overall, the BPN specimens exhibited substantially improvedmembrane performance relative to the GCL specimens, with values ofRω, BPN/GCL ranging from 128 to 294% for tests conducted using rigid-wall cells and from 109 to 433% for tests conducted using flexible-wallcells. There are at least two possible mechanistic explanations for theimproved performance of the BPN compared to the sodium bentonitein a GCL (e.g., Scalia, 2012; Scalia et al., 2011). First, the superabsorbentpolymer portion of the BPN swells in low EC solutions, resulting inosmotic swelling and smaller hydraulically active pores than those inthe sodium bentonite typically found in GCLs. Second, an excess oflow molecular weight polymer in the BPN during the manufacturingprocess resulting from incomplete polymerization (chain termination)likely results in clogging the interconnected pores of the BPN.Regardless of the exact mechanisms, the BPN evaluated in this studyexhibited substantially improved membrane behavior relative to theGCL containing traditional sodium bentonite.

4.3. Comparison of volume changes

The extent of the volume changes that occurred with the BPNspecimens in this study was far greater than that previously reportedfor membrane tests in flexible-wall cells using GCL containingtraditional bentonites (Kang and Shackelford, 2009, 2011). For example,the cumulative volume change, Σ(−ΔV), for a traditional GCL tested atσ′=34.5 kPa (5 psi) as reported by Kang and Shackelford (2011) was5.98 mL, whereas, the Σ(−ΔV) for BPN tested at the same effectivestress in this studywas 46.3mL, or almost eight times greater. However,similar to the traditional GCL, the Σ(−ΔV) for the BPN decreased withincreased σ′ [Σ(−ΔV) = 20.7 mL versus 46.3 mL for σ′ = 103 kPa(15psi) and σ′=34.5kPa (5psi), respectively].

The volume changes observed for the traditional GCLwere attributedto compression of the diffuse double layers as the concentration ofthe salts in the pore water increased (Kang and Shackelford, 2009).In addition to this possibility, the increased volume change of the BPNmay be attributed, in part, to the unique properties of the BPN. Theinitial volume, Vo, of the BPN specimen was greater than that of theGCL (e.g., 191 mL versus 75.5 mL, respectively) for tests conducted atσ′=34.5 kPa (5psi), because of the greater swelling of the BPN versusa traditional GCL (SI= 73 mL/2 g vs. 25 to 35 mL/2 g). This increasedswelling for the BPN is a result of the superabsorbent polymer portionof the BPN (Scalia et al., 2011), which likely caused greater volumechanges throughout the membrane test compared to that of a GCL.

The swelling capacity of the superabsorbent polymer portion of theBPN is affected by the applied load (i.e., σ′) and the salt concentration inthe pores (Buchholz andGraham, 1998). In addition, for superabsorbentpolymers, the effect of the salt concentration in the pores on theswelling capacity decreases with increasing load (Buchholz andGraham, 1998). These properties of superabsorbent polymer correlatewith the observed volume changes of the BPN. For example, for thetest conducted at the higher σ′ [i.e., σ′= 103 kPa (15 psi) for FW 2],the specimen swelled less resulting in a lower Vo (Vo = 75 mL vs.191mL for FW2 and FW1, respectively), and consequently less volumechange throughout the test (i.e., Σ(−ΔV)=20.7mL versus 46.3mL forFW2 vs. FW1, respectively). In addition, the volume change of the BPNdecreased with increasing σ′. For example, the value of −ΔV was lessthan 1.5 mL during circulation of 54 mM KCl for FW2 [σ′ = 103 kPa(15 psi)], whereas that for FW1 [σ′=34.5 kPa (5 psi)] was as much as3.2mL during the circulation of 54mM KCl.

4.4. Effect of effective stress and boundary salt concentration

The ratio of themembrane efficiency coefficient at any σ′, relative tothat atσ′ of 34.5kPa (5psi), or Rω,σ′ [=ω@σ′/ω@σ′=34.5kPa (5psi)],is plotted as a function of Cave for the BPN and a GCL (Kang and

0

3

6

9

12

15(a)

BPN [σ' = 103 kPa (15 psi)]

GCL [σ' = 103 kPa (15 psi)]

GCL [σ' = 172 kPa (25 psi)]

GCL [σ' = 241 kPa (35 psi)]

0

300

600

900

1200

1500R

ω, σ

'


Rω

, σ' (%

)

0

3

6

9

12

15

1 10 100

1 10 100

(b)

BPN [σ' = 103 kPa (15 psi)]

GCL [σ' = 103 kPa (15 psi)]

GCL [σ' = 172 kPa (25 psi)]

GCL [σ' = 241 kPa (35 psi)]

0

300

600

900

1200

1500

Rω

, σ'


Rω

, σ' (%

)

Fig. 12. Ratio of the membrane efficiency coefficient, ω, at any σ′ to ω at σ′= 34.5 kPa(5 psi), Rω,σ′, as a function of the average salt (KCl) concentration across the specimen,Cave: (a) ω based on initial (source) KCl concentrations, ω0 and (b) ω based on averageKCl concentrations, ωave. (Note: σ′=initial effective stress).


Shackelford, 2011) in Fig. 12. In general, the values of Rω,σ′ are greaterthan unity as would be expected on the basis of studies that haveshown that an increase in σ′ results in a higher membrane efficiency(Kang and Shackelford, 2011). In addition, the effect of increasing σ′is greater with higher average source concentrations across thespecimen. However, this increase Rω,σ′ is lower for the BPN relativeto that of the GCL at similar σ′, especially for higher concentrations(e.g., Cave = 27 mM KCl or Cot = 54 mM KCl). As described in Kangand Shackelford (2011), the pore sizes decrease with increasing σ′,which tends to offset the effect of an increase in Cot on the membraneefficiency. The lower increase in values of Rω,σ′ for the BPN specimensrelative to that of the GCL suggests that the BPN is less sensitive to σ′than the GCL, likely due, in part, to the differences in the compressibilityof the BPN relative to that for the GCL. As previously discussed, thecompressibility of the BPN likely is greater than that of the GCL as aresult of the superabsorbent polymer portion of the BPN.

5. Conclusions

The results of a laboratory experimental program to deter-mine the existence of semipermeable membrane behavior for a

polyacrylatemodified bentonitematerial knownas a bentonite polymernanocomposite, or BPN, were presented. The results indicated that theBPN exhibited membrane behavior when exposed to KCl with initialconcentration differences ranging from 4.7 to 54 mM. A comparisonof the measured membrane behavior for the BPN with that previouslyreported for sodium bentonite contained in a GCL showed that theBPN exhibited greater membrane behavior.

For example, the membrane efficiency coefficient,ω, for a specimenof the BPN with a porosity, n, of 0.92 contained within a rigid-wall cellwas 0.43 based on an initial concentration difference of 20 mM KCl,whereas ω for a GCL specimen with a lower (more favorable) porosityof 0.74 also contained within a rigid-wall cell and subjected to thesame initial concentration difference was only 0.30. Similarly, the ωfor a specimen of the BPN contained within a flexible-wall cell at aninitial effective stress of 34.5 kPa (5 psi) and subjected to an initialconcentration difference of 9.3mM KCl was 0.32, whereas ω for a GCLspecimen also contained within a flexible-wall cell at the same initialeffective stress but subjected to a slightly lower (more favorable)concentration difference of 8.7 mM KCl was only 0.20. Overall, themeasured ω values for the BPN specimens ranged from 109% to 433%of those previously reported for the specimens containing traditionalsodium bentonite, depending on the initial porosity or initial effectivestress of the specimen, the concentration of KCl, and the type of testcell (rigid vs. flexible), despite the BPN specimens not being flushed ofsoluble salts prior to membrane testing as in the case of the specimenscontaining sodium bentonite. Finally, the BPN exhibited membranebehavior over a wider range in concentrations than those for the GCL.Thus, relative to values of ω previously reported for GCL specimenscontaining a traditional sodium bentonite, the BPN evaluated in thisstudy exhibited improved membrane behavior under conditions thatpresumably were more adverse with respect to soluble salts to thosefor the GCL specimens. However, additional testing of the BPN iswarranted to better understand the long-term behavior of the BPNand to investigate the sustainability of the membrane behavior withother solutions containing divalent cations such as CaCl2.

Acknowledgment

Financial support for this project, which was a collaboration amongColorado State University (CSU), the University of Wisconsin-Madison(UW-Madison), and Colloid Environmental Technologies Co. (CETCO),was provided by the U.S. National Science Foundation (NSF), Arlington,VA under Grant No. CMMI-0757815. The opinions expressed in thispaper are solely those of the authors and are not necessarily consistentwith the policies or opinions of the NSF. The authors thank CETCO forproviding the bentonites used in this study, the assistance of all thecollaborators including Craig Benson, Tuncer Edil, and Joe Scalia atUW-Madison andMike Donovan and Jerry Darlington of CETCO, KristinSample-Lord and Catherine Soo Jung Hong of CSU for assistance withthe membrane tests, and Thomas Borch of CSU for the use of the ionchromatograph for anion concentration analyses.

References

ATU, 1992. ClayMax sodium bentonite liner found degraded at New Jersey site.Aboveground Tank Update, 3(2) 1 (17–18).

Barbour, S.L., Fredlund, D.G., 1989. Mechanisms of osmotic flow and volume change inclay soils. Can. Geotech. J. 26 (4), 551–562.

Benson, C.H., Meer, S.R., 2009. Relative abundance of monovalent and divalent cations andthe impact of desiccation on geosynthetic clay liners. J. Geotech. Geoenviron. 135 (3),349–358.

Benson, C.H., Thorstad, P.A., Jo, H.Y., Rock, S.A., 2007. Hydraulic performance of geosyntheticclay liners in a landfill final cover. J. Geotech. Geoenviron. 133 (7), 814–827.

Boels, D., van der Wal, K., 1999. Trisoplast®: new developments in soil protection.Cagliari, Italy, Sardinia 99, CISA, Padova, Italy, pp. 77–84.

Bohnhoff, G.L., 2012. Membrane Behavior, Diffusion, and Compatibility of a PolymerizedBentonite for Containment Barrier Applications. (PhD Dissertation) Colorado StateUniversity, Fort Collins, Colorado, USA.

Buchholz, F., Graham, A., 1998. Modern Superabsorbent Polymer Technology. JohnWileyand Sons, New York.

http://refhub.elsevier.com/S0169-1317(13)00330-X/rf0285

















Christman, M., Benson, C.H., Edil, T.B., 2002. Geophysical evaluation of annular well seals.Groundw. Monit. Remediat. 22 (3), 104–112.

Di Emidio, G., 2010. Hydraulic and Chemico-Osmotic Performance of Polymer TreatedClay. (PhD Dissertation) University of Ghent, Ghent, Belgium.

Di Maio, C., 1996. Exposure of bentonite to salt solution: osmotic and mechanical effects.Geotechnique 46 (4), 695–707.

Dominijanni, A., Manassero, M., 2012. Modelling the swelling and osmotic properties ofclay soils. Part II: the physical approach. Int. J. Eng. Sci. 51, 51–73.

Egloffstein, T., 2001. Natural bentonites — influence of the ion exchange and partialdesiccation on permeability and self-healing capacity of bentonites used in GCLs.Geotext. Geomembr. 19 (7), 427–444.

Estornell, P., Daniel, D.E., 1992. Hydraulic conductivity of 3 geosynthetic clay liners.J. Geotech. Eng. 118 (10), 1592–1606.

Evans, J.C., 1993. Vertical cutoff walls. In: Daniel, D.E. (Ed.), Geotechnical Practice forWaste Disposal. Chapman & Hall, London, 430–454.

Evans, J.C., Shackelford, C.D., Yeo, S.S., Henning, J., 2008. Membrane behavior of soil–bentonite slurry–trench cutoff walls. Soil Sediment Contam. 17 (4), 316–322.

Fritz, S.J., 1986. Ideality of clay membranes in osmotic processes — a review. Clay ClayMiner. 34 (2), 214–223.

Groenevelt, P.H., Elrick, D.E., 1976. Coupling phenomena in saturate homo-ionicmontmorillonite. 2. Theoretical. Soil Sci. Soc. Am. J. 40 (6), 820–823.

Heller, H., Keren, R., 2003. Anionic polyacrylamide polymer adsorption by pyrophylliteand montmorillonite. Clay Clay Miner. 51 (3), 334–339.

Henning, J.T., Evans, J.C., Shackelford, C.D., 2006. Membrane behavior of two backfillsfrom field-constructed soil–bentonite cutoff walls. J. Geotech. Geoenviron. 132 (10),1243–1249.

Ito, H., 2006. Compaction properties of granular bentonites. Appl. Clay Sci. 31 (1–2),47–55.

James, A.N., Fullerton, D., Drake, R., 1997. Field performance of GCL under ion exchangeconditions. J. Geotech. Geoenviron. 123 (10), 897–901.

Jo, H.Y., Katsumi, T., Benson, C.H., Edil, T.B., 2001. Hydraulic conductivity and swellingof nonprehydrated GCLs permeated with single-species salt solutions. J. Geotech.Geoenviron. 127 (7), 557–567.

Jo, H.Y., Benson, C.H., Edil, T.B., 2004. Hydraulic conductivity and cation exchange innon-prehydrated and prehydrated bentonite permeated with weak inorganic saltsolutions. Clay Clay Miner. 52 (6), 661–679.

Jo, H.Y., Benson, C.H., Shackelford, C.D., Lee, J.M., Edil, T.B., 2005. Long-term hydraulicconductivity of a geosynthetic clay liner permeated with inorganic salt solutions.J. Geotech. Geoenviron. 131 (4), 405–417.

Kajita, L.S., 1997. An improved contaminant resistant clay for environmental clay linerapplications. Clay Clay Miner. 45 (5), 609–617.

Kang, J.B., Shackelford, C.D., 2009. Clay membrane testing using a flexible-wall cell underclosed-system boundary conditions. Appl. Clay Sci. 44 (1–2), 43–58.

Kang, J.B., Shackelford, C.D., 2010. Membrane behavior of compacted clay liners.J. Geotech. Geoenviron. 136 (10), 1368–1382.

Kang, J.B., Shackelford, C.D., 2011. Consolidation enhanced membrane behavior of ageosynthetic clay liner. Geotext. Geomembr. 29 (6), 544–556.

Katsumi, T., Ishimori, H., Ogawa, A., Yoshikawa, K., Hanamoto, K., Fukagawa, R., 2007.Hydraulic conductivity of nonprehydrated geosynthetic clay liners permeated withinorganic solutions and waste leachates. Soils Found. 47 (1), 79–96.

Katsumi, T., Ishimori, H., Onikata, M., Fukagawa, R., 2008. Long-term barrier performanceof modified bentonite materials against sodium and calcium permeant solutions.Geotext. Geomembr. 26 (1), 14–30.

Kenney, T.C., Vanveen, W.A., Swallow, M.A., Sungaila, M.A., 1992. Hydraulic conductivityof compacted bentonite sand mixtures. Can. Geotech. J. 29 (3), 364–374.

Kolstad, D.C., Benson, C.H., Edil, T.B., 2004. Hydraulic conductivity and swell ofnonprehydrated geosynthetic clay liners permeated with multispecies inorganicsolutions. J. Geotech. Geoenviron. 130 (12), 1236–1249.

Lee, J.M., Shackelford, C.D., 2005a. Concentration dependency of the prehydration effectfor a geosynthetic clay liner. Soils Found. 45 (4), 27–41.

Lee, J.M., Shackelford, C.D., 2005b. Impact of bentonite quality on hydraulic conductivityof geosynthetic clay liners. J. Geotech. Geoenviron. 131 (1), 64–77.

Lee, J.M., Shackelford, C.D., 2005c. Solution retention capacity as an alternative to theswell index test for sodium bentonite. Geotech. Test. J. 28 (1), 61–70.

Lin, L.C., Benson, C.H., 2000. Effect ofwet–dry cycling on swelling andhydraulic conductivityof GCLs. J. Geotech. Geoenviron. 126 (1), 40–49.

Lo, I.M.C., Liljestrand, H.M., Khim, J., Shimizu, Y., 1996. Clay liner materials for landdisposal of hazardous non-metal wastes. Water Sci. Technol. 33 (8), 71–77.

Malusis, M.A., Shackelford, C.D., 2002a. Chemico-osmotic efficiency of a geosynthetic clayliner. J. Geotech. Geoenviron. 128 (2), 97–106.

Malusis,M.A., Shackelford, C.D., 2002b. Coupling effects during steady-state solute diffusionthrough a semipermeable clay membrane. Environ. Sci. Technol. 36 (6), 1312–1319.

Malusis, M.A., Shackelford, C.D., Olsen, H.W., 2001. A laboratory apparatus to measurechemico-osmotic efficiency coefficients for clay soils. Geotech. Test. J. 24 (3),229–242.

Malusis, M.A., Shackelford, C.D., Olsen, H.W., 2003. Flow and transport through claymembrane barriers. Eng. Geol. 70 (3–4), 235–248.

Malusis, M.A., Kang, J.-B., Shackelford, C.D., 2013. Influence of membrane behavior onsolute diffusion through GCLs. In: Manassero, M., Dominijanni, A., Foti, S., Musso, G.(Eds.), Coupled Phenomena in Environmental Geotechnics. CRC Press/Balkema.Taylor & Francis Group, London, 267–274.

Manassero, M., Dominijanni, A., 2003. Modelling the osmosis effect on solute migrationthrough porous media. Geotechnique 53 (5), 481–492.

Mazzieri, F., Di Emidio, G., Van Impe, P.O., 2010. Diffusion of calcium chloride in amodified bentonite: impact on osmotic efficiency and hydraulic conductivity. ClayClay Miner. 58 (3), 351–363.

McRory, J., Ashmawy, A., 2005. Polymer treatment of bentonite clay for contaminantresistant barriers. Geotechnical Special Publication 142, Waste Containment andRemediation.ASCE, Reston, Virgina, USA 3399–3409.

Meer, S.R., Benson, C.H., 2007. Hydraulic conductivity of geosynthetic clay liners exhumedfrom landfill final covers. J. Geotech. Geoenviron. 133 (5), 550–563.

Mitchell, J.K., 1976. Fundamentals of Soil Behavior. John Wiley & Sons, New York.Muzny, C.D., Butler, B.D., Hanley, H.J.M., Tsvetkov, F., Peiffer, D.G., 1996. Clay platelet

dispersion in a polymer matrix. Mater. Lett. 28 (4–6), 379–384.Olsen, H.W., Yearsley, E.N., Nelson, K.R., 1990. Chemico-osmosis versus diffusion-osmosis.

Transportation Research Board, Washington, DC, USA 15–22.Pavlidou, S., Papaspyrides, C.D., 2008. A reviewonpolymer-layered silicate nanocomposites.

Prog. Polym. Sci. 33 (12), 1119–1198.Scalia, J., 2012. Bentonite–Polymer Composites for Containment Applications. (PhD

Dissertation) University of Wisconsin, Madison, Wisconsin, USA.Scalia, J., Benson, C.H., Edil, T.B., Bohnhoff, G.L., Shackelford, C.D., 2011. Geosynthetic clay

liners containing bentonite polymer nanocomposites. GeoFrontiers 2011: Advancesin Geotechnical Engineering.ASCE, Reston, VA 1961–1970.

Shackelford, C.D., Daniel, D.E., 1991. Diffusion in saturated soil. 1. Background. J. Geotech.Eng. 117 (3), 467–484.

Shackelford, C.D., Lee, J.M., 2003. The destructive role of diffusion on clay membranebehavior. Clay Clay Miner. 51 (2), 186–196.

Shackelford, C.D., Benson, C.H., Katsumi, T., Edil, T.B., Lin, L., 2000. Evaluating the hydraulicconductivity of GCLs permeated with non-standard liquids. Geotext. Geomembr. 18(2–4), 133–161.

Shackelford, C.D., Malusis, M.A., Olsen, H.W., 2003. Clay membrane behavior forgeoenvironmental containment. In: Culligan, P.J., Einstein, H.H., Whittle, A.J. (Eds.),Soil and Rock America Conference 2003, Proceedings of the joint 12th PanamericanConference on Soil Mechanics and Geotechnical Engineering and the 39th U.S. RockMechanics Symposium, vol. 1. Verlag Glückauf GMBH, Essen, Germany, 767–774.

Smith, J.A., Bartelt-Hunt, S.L., Burns, S.E., 2003. Sorption and permeability of gasolinehydrocarbons in organobentonite porous media. J. Hazard. Mater. 96 (1), 91–97.

Sposito, G., 1989. The Chemistry of Soils. Oxford University Press, New York.Vasko, S., Jo, H.Y., Benson, C.H., Edil, T.B., Katsumi, T., 2001. Hydraulic conductivity of

partially prehydrated geosynthetic clay liners permeated with aqueous calciumchloride solutions. GeosyntheticsIndustrial Fabrics Assoc. Int., St. Paul, Minnesota,USA 685–699.

Yeo, S.S., Shackelford, C.D., Evans, J.C., 2005. Membrane behavior of model soil–bentonitebackfills. J. Geotech. Geoenviron. 131 (4), 418–429.




























































































































applied clay science · containing 78% and 96% of clay-sized (b5 mm) material, respectively. the...

Documents