Environ. Sci. Technol. 2009, 43, 1153–1159
Coupling Centrifuge Modeling and Laser Ablation Inductively Coupled Plasma Mass Spectrometry To Determine Contaminant Retardation in Clays W E N D Y T I M M S , * ,†,‡ M . J I M H E N D R Y , † JASON MUISE,† AND ROBERT KERRICH† Department of Geological Sciences, University of Saskatchewan, Saskatoon, Saskatchewan, S7N 5E2, Canada, and Water Research Laboratory, School of Civil and Environmental Engineering, University of New South Wales, Manly Vale, NSW 2015, Australia
Received July 24, 2008. Revised manuscript received November 29, 2008. Accepted December 10, 2008.
are allowed to accumulate in the closed system, batch mixing significantly decreases the solid/liquid (S/L) ratio and may increase surface area available for reaction, and phase separation and sampling are operator-dependent and are not always uniform (4). Allowing species to accumulate in the closed system changes the equilibrium status, may slow the rate of reaction, and requires reverse reactions be taken into account. In an attempt to overcome the limitations of these existing techniques, the objective of the current study was to develop and test a novel approach to quantify retardation of reactive solutes in clay-rich media by combining centrifuge modeling with laser ablation coupled with inductively coupled plasma mass spectroscopy (LA-ICP-MS). The solute transport experiments were conducted on core samples from a wellcharacterized, thick, plastic, clay-rich till deposit located in southern Saskatchewan, Canada. The retardation factor (Rd) describes partitioning of a contaminant between aqueous and solid phases according to the following relationship (5):
(
Rd ) 1 +
Quantifying the retardation (Rd) of reactive solutes as they migrate through low-permeability clay-rich media is difficult, thus motivating this study to assess the viability of combining centrifuge modeling and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) techniques. An influent solution containing Cl-, trace metals, and lanthanide species flowed at 1.0 mL · h-1 through an undisturbed clay-rich core sample (33 mm diameter × 50 mm long) mounted in a UFA Beckman centrifuge operating at 3000 rpm (N factor ) 876g). During the 87 day experiment, the hydraulic conductivity of the core was 3.4 × 10-10 m · s-1. Effluent breakthrough data indicate the Rd of Tl to be 10; incomplete breakthrough (non-steady-state) data for 145Nd and 171Yb suggest Rd values of .75 and .85, respectively. At the completion of the transport experiment, longitudinal sections of the core solid were analyzed for 145Nd and 171Yb using a Cetac laser ablation system coupled with an ICP-MS. The longitudinal core sections yielded Rd values of >10 000 for 145Nd and 171Yb. This study demonstrates coupling these techniques can provide Rd values for a wide range of reactive solutes with relatively rapid testing of smallscale, low hydraulic conductivity core samples.
Introduction Clay-rich geologic deposits are often selected for waste disposal because they have a low hydraulic conductivity (K) that slows the migration of contaminants to underlying regional aquifers. Furthermore, these fine-grained geologic media are commonly used as waste-containment barriers, including applications to high-level radioactive waste. The geochemical retardation (e.g., sorption and exchange reactions between the solutes and the matrix media) of some solutes in these deposits has been intractable to quantify using classical column experiments due to very slow groundwater velocities (typically 15 m depth in the unoxidized till at the test site (17). Solution GW1 was created using ultrapure 1154
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 4, 2009
TABLE 1. Chemistry of Influent Solutions GW1 influent solution parameter Ca Mg Na Cl SO4 DOC pH ionic strength
units
concn
mmol · L-1 9000 mmol · L-1 7000 mmol · L-1 6000 5 mmol · L-1 mmol · L-1 22 000 mmol · L-1 1.1 7.1 M 0.008
GW1s spikes (added to GW1) atomic concn parameter weight (mmol · L-1) Zr La Nd Nd Yb Lu Hf Tl Th
90 139 145 146 171 175 178 205 232
0.02 0.44 0.32 0.10 0.26 0.32 0.03 0.21 0.14
deionized water, NaHSO4(s), MgSO4(s), and CaSO4(s). NH4OH and HNO3 were added to stabilize the pH at 7.1 ( 0.3 (Table 1). Added to GW1 to create the influent spike solution GW1s were 0.005 M Cl- (added as NH4Cl), lanthanides and trace metals at concentrations of 2-60 µg · L-1 (Table 1) from Inorganic Ventures Inc. (1000 µg · mL-1 standards), and enriched isotope spikes of 145Nd (78.7%), 171Yb (83.4%); the solution was stabilized at pH 7.0 ((0.4). All influent solutions were stored in 4 L Nalgene containers at 15 °C. The UFA centrifuge was developed specifically for unsaturated/saturated flow studies of porous media (18). The Beckman centrifuge on which the UFA system is based (UFAJ6-M1) includes a rotating seal assembly that allows influent solution to flow constantly through a microdispersal system and solid samples to an effluent collection system (Supporting Information Figure S1). Two cores are run simultaneously to maintain balance in the centrifuge, and this capacity has been used to test the repeatability of the centrifuge method (8). In this study, however, results are reported for only one core as a different influent was used in the other core centrifuged at the same time. The target centrifuge speed was designed to maximize fluid flux through the core material with the low-permeability clay till determining the steadystate fluid flux that was possible. The centrifuge was accelerated to 700 rpm at which time the nonspiked influent porewater (GW1) was introduced to the top of the core sample and then accelerated to the experimental operational speed of 3000 rpm. Influent was introduced using an intravenous (IV) pump (AVI 280 RT, from 3M) at 1.0% ( 2% mL · h-1. The core was flushed with GW1 until the chemistry (i.e., pH, Cl-) of the effluent was stable, after which time (approximately 336 h) the spiked solution GW1s was introduced. Effluent Sampling and Analysis. Effluent samples were collected by decelerating the centrifuge and sampling following the protocol in section S1 of the Supporting Information. After collection, effluent samples were immediately filtered with a 0.45 µm cellulose acetate membrane filter (Advantec MFS, Inc.). This standard filter was assumed to not impact the concentration of trace metals and lanthanides given no sorption was reported using this method during detailed analysis of lanthanides at the study site (19). Furthermore, Reszat and Hendry (20) show colloids larger than 2 nm (10-9 m) are completely excluded from movement in this clay till due to straining, suggesting the 0.45 µm (10-6 m) filter paper should not affect our results. The filtrate was split into two aliquots and analyzed for pH and chloride using microelectrodes as described by Timms and Hendry (21). Section S2 of the Supporting Information provides more details on the sample volumes, calibration procedures, and precision of this methodology. ICP-MS Analysis. A Micromass platform hexapole ICP-MS was used for analysis of acidified effluent samples from centrifugation experiments and was also used coupled with a laser ablation system for analysis of solid core samples.
A class-100 clean room was used for the preparation of all associated standards and samples. Calibration used standard additions, internal standardization, and external calibration (22-24). Rhodium (2 µg · L-1) was used as an internal standard; three external standards were created to confirm sample concentration and check for interference of metal oxides (22). A standard spiked with natural isotope abundances of Nd and Yb was used to quantify the mass bias of the hexapole ICP-MS. Standards and a HNO3 blank were analyzed every 8-10 samples. Effluent Analysis Using Cumulative Mass Ratio. Retardation (Rd) in permeable media is defined by the time at which the relative effluent concentration attains a value of 0.5 relative to a conservative species (e.g., ref 25). However, for slowly permeable media such as clay-rich deposits, Rd is better represented by the area above the solute breakthrough curve (e.g., refs 25 and 26). Rather than methods relying on solute concentration, the present study adopted the cumulative mass ratio (CMR) as described in section S3 of the Supporting Information. Flow Measurement. A form of Darcy’s Law that incorporates centrifugal force (9, 27-29) was applied to calculate K for materials subjected to accelerated gravity. In practice, K was determined for the UFA centrifuge with quasi-constant head, using eq 4 (18, 28) with centrifugal force driving fluid through the sample: Kp )
Q 8928 A F r(ω)2
(4)
w
where Kp is hydraulic conductivity (m · s-1), Q is fluid discharge (L · s-1), ω is angular velocity (rpm), A is sample area (m2), r is radial distance (m), and Fw is fluid density (kg · m-3). For example, K ) 1.5 × 10-11 m · s-1 for Q ) 4.2 × 10-9 L · s-1 (0.015 mL · h-1), A ) 0.00085 m2, r ) 0.087 m (center of core where r0 ) axis of rotation), Fw ) 1005 kg · m-3, and ω ) 1840 rpm. A constant flow and K value indicates conservation of mass is attained. Solid Analysis. At the completion of the centrifuge experiment the cores were placed in a freezer for about 4 h to partially freeze the outer 2-4 mm. This ensured the sample remained competent during its extrusion from the sample holder, as confirmed by pin markers showing no significant ((0.1 mm) difference in sample dimensions. The extruded sample was immediately divided with a dissecting scalpel into four axisymmetric sections (parallel to flow). One section was split into five equal axial lengths (8-10 mm) perpendicular to the direction of flow. The gravimetric water content (GWC) was determined for each subsection using the method in ASTM (30). The volumetric water content (VWC) was calculated by dividing the GWC by the volume of the sample. The cores required specialized preparation to minimize the effects of plucking and gouging of samples during LA analysis. Half of the core sample as well as a pristine core sample collected from the same depth (10.3-10.8 m BG) were freeze-dried to remove all water. The samples were then impregnated with Epofix (Struers Inc.), mounted on glass slides (25 × 33 mm), and polished. No measurable change was observed in sample size during this procedure. The core sample was split in half perpendicular to the flow direction to allow use of the maximum surface area of the laser ablation sample stage (900 mm2). Laser Ablation and ICP-MS. A Cetac LSX-200+ UV laser ablation (LA) system coupled with hexapole ICP-MS was used to analyze mounted sections of the centrifuge test core and the pristine core. The Cetac LSX-200+ LA system used in this study is a solid-state ultraviolet multimode Nd yttrium aluminum garnet (Nd:YAG) laser, where the YAG rod was doped with Nd3+ ions. A 266 nm UV wavelength output was produced by a harmonic generator in the laser system. The laser is focused by specialized beam delivery optics on a
FIGURE 1. Chloride breakthrough curve (closed symbols) and cumulative mass ratios (CMR; open symbols) vs PV calculated from VWCavg. sample spot of 5-250 µm in diameter, with ablated material carried to the ICP-MS by argon gas in Tygon tubing. An ICP-MS technique including internal and external calibrations was used as described earlier. The laser pulse energy was controlled using an optical attenuator to maintain thermal equilibrium and output stability. In this experiment, LA and ICP-MS were used to determine the adsorption sites for the metals and, where possible, their distribution along the flow paths in the impregnated and mounted core samples. Two transects were laser ablated parallel to the core axis and, hence, to the water flow path through the cores. Both transects were 200 µm wide and approximately 10 mm apart. The rate of ablation was 30 µm · s-1. As a standard with a similar matrix to that of the clay samples was not available, raw count data after the centrifugation experiment were compared directly to raw counts of the pristine core.
Results and Discussion Hydrogeological Characteristics. The VWC measurements on core samples ranged from 23% to 26% with a mean of 25% (n ) 10). The mean was slightly less than values reported by Shaw and Hendry (16) (range of 27-31%, mean 29%; n ) 10) for this till using the same test method. Measured bulk was 2170 kg · m-3, and calculated porosity was 21.8%. The pore volume (PV) in the core sample, based on the mean VWC, was 10.5 mL. Given a flow rate of 1.0 mL · h-1 for 87 days, approximately 199 PVs (167 PVs for the spiked influent) were eluted from the base of the core with a fluid residence time of about 10.5 h. A Darcy flow rate for the influent solution of >3.25 × 10-7 m · s-1 (>10.25 m · year-1) was calculated by dividing the flow rate of 2.78 × 10-10 m3 · s-1 (1.0 mL · h-1) by a cross-sectional area of 1.48 × 10-6 m · s-1 was derived by dividing the Darcy flow rate by the porosity (0.22). As predicted by scaling laws, the flow rate during centrifugation was N times greater than if a hydraulic gradient of 1 was applied to the core samples at 1g. For example, the Darcy flow rate of 3.25 × 10-7 m · s-1 (10.25 m · year-1) would be equivalent to 0.012 m · year-1 in the prototype at 1g (VM/ VP ) N ) 876). However, the hydraulic gradient of 1.0 assumed in the prototype was ∼70 times greater than the in situ hydraulic gradient of 0.014. Consequently, the velocity during centrifugation was in the order of 105 times faster than the in situ porewater flux of 0.5-0.8 m per 10 000 year. The geometric bulk mean Kp for this core sample was 3.4 × 10-10 m · s-1 (n ) 44). This values is consistent with other laboratory and field measured K values for this unoxidized VOL. 43, NO. 4, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1155
FIGURE 2. Effluent plotted as (A) C/C0 and (B) CMR. Cl and reactive solute Tl exhibit complete breakthrough, with Rd values of 5 and 12, respectively. Transport of Zr, Hf, and Th is highly retarded. Various lanthanide species exhibit partial breakthrough. The estimation of minimum Rd values for partial breakthrough of 145Nd and 171Yb are shown in panel B with a 1:1 line extended from the last effluent point to the X-axis. till at this depth (in the order of 10-10 m · s-1) (16, 31). The similarity between these K data suggests solute transport through the cores occurs primarily via matrix flow, although the possibility of discrete flow paths cannot be ruled out. Cl- Effluent Concentrations. Dissolved Cl- is commonly used as a conservative tracer in groundwater studies because it typically exhibits a lack of adsorption in its anionic form. Influent GW1 was used until stable background values of 0.005 M were achieved at about 336 h (about 32 PV) prior to introducing GW1s. The Cl- effluent breakthrough curve (Figure 1) is expressed as relative concentration (C/C0) and CMR. The Rd value for Cl- was close to 1.0, as expected for a conservative solute. Some evidence indicated the occurrence of early chloride breakthrough, although more precise identification of breakthrough was not possible due to the minimum volume (2.0 mL) of effluent samples required for analysis. The possibility of early breakthrough could be 1156
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 4, 2009
attributed to discrete flow paths or the effects of anion exclusion and water bound within enclosed pores. Trace Metals and Lanthanides. Speciation of the spiked influent solution was calculated using PHREEQC and the Lawrence Livermore National Laboratory database (32). The speciation calculations are presented in Supporting Information Table S1 and discussed in section S4 of the Supporting Information. Speciation calculations indicate that trace metals and lanthanides are stable in the dissolved phase as hydroxide, sulfate, and dicarbonato complexes. CMR curves were determined for Cl-, trace metals, transition metals, and lanthanides, including 145Nd and 171Yb (Figure 2). No breakthrough during the test period (i.e., C/C0 < 0.01) was detected for Th, Hf, or Zr, suggesting their Rd values are at least an order of magnitude greater than Rd values for lanthanides. The effluent concentrations of Tl exhibited complete breakthrough (i.e., C/C0 ) 1.0; Figure 2).
FIGURE 3. Laser ablation axial concentration profiles for core solid. Isotopic count ratios were corrected using the mean isotopic ratio of the pristine sample. The Rd values for 145Nd and 171Yb were >10 000. With the use of the CMR technique, the Rd of Tl was calculated by dividing the sorbed distance from the top of determined to be 10. The lanthanide concentrations inthe core by the distance the conservative tracer Cl- would have traveled (11 m) during the experiment. The Rd for 145Nd creased more gradually than Tl and did not exhibit complete was therefore >13 750 (11 000 mm/0.8 mm). These lanbreakthrough (Figure 2). By extending a 1:1 line from the last thanide Rd values are considered minimums because retardata point for the CMR versus T data, the incomplete dation could be underestimated for a field site if centrifuge breakthrough data for 145Nd and 171Yb suggest Rd values for 145 Nd and 171Yb of .75 and .85, respectively. modeling increases dispersion and decreases breakthrough The relative retardation of individual lanthanide species time (8). could be identified despite incomplete breakthrough (Figure The calculated Rd values in this study were of a similar order of magnitude to batch test Rd values of >50 000 for 2). Increasing retardation corresponded with increasing actinide (Am and Eu) sorption on saprolite (2), but greater atomic mass (Rd of Yb > Nd > La); this fractionation of lanthanides due to sorption of heavier species during reactive than Rd values of ∼400 to ∼10 000 for sorption of Yb and other lanthanides to fine-grained alluvium from Yucca transport is consistent with Coppin et al. (33), who report Mountain (3). The heterogeneous alluvial material from Yucca fractionation at higher ionic strengths (0.5 M compared to Mountain exhibited higher Rd values from batch testing than ∼0.01 M in this study). The fractionation at higher ionic pure clay, and light lanthanides were sorbed by most strengths occurs via a surface complexation reaction that materials in preference to heavy lanthanides. Clearly, very depends on the electrons in the outer shell and accounts for high retardation values are possible for these species; distinctive fractionation behavior for different species. In however, a review of the conditions necessary for such this case, increasing retardation for isotopes (173Yb > 171Yb > 146Nd > 145Nd) was also observed. The opposite fractionsignificant sorption of species would be warranted across a ation trend has been reported for other alluvial materials (3) range of hydrogeochemical conditions. and for in situ porewater from this unoxidized till (19). Equilibrium Geochemical Reactions. The likelihood of Benedict et al. (3) and Johannesson and Hendry (19) observed local chemical equilibrium during the centrifugation experigreater sorption of lanthanides with relatively low atomic ment was determined considering the nondimensional mass, a phenomenon attributed to the nature of carbonate Damkohler number, Da (34): and sulfate complexes. (5) Da ) KfbL/V LA-ICP-MS Analysis. At the completion of the centrifuge -1 where Kfb is the reaction rate (s ) for first-order exchange, experiment, laser ablation was conducted along the axial L is 50 mm, and V is 1.48 × 10-6 m · s-1. A Da of >1.0 indicates profile of the sectioned core, and ionic count ratios were local chemical equilibrium can been assumed. Batch kinetic compared with the pristine core (Figure 3). Rd values calculated from the axial profiles for 145Nd and 171Yb reflect experiments by Aja (35) show ∼75% of sorption equilibrium the distance the isotopes migrated through the core from its for Nd within 15 min of contact with bentonite. The high upper surface. Axial concentration profiles exhibited very kinetic rate is attributed to high sorption affinity. Similarly, high concentrations near the upper surface of the core. The Benedict et al. (3) report fast kinetics (within 20-40 min) for solute fronts for 145Nd and 171Yb advanced 0.8 and 1.4 mm, rare earth element sorption onto alluvial materials, including respectively, within 87 days, with very high, relatively stable fine particles; the kinetic rate slowed and stabilized in about concentrations observed behind the solute fronts. Multiple 2-6 h. peaks at other locations along the core axis (e.g., 17 mm for Assuming reaction times of between 15 min and 6 h from 145 Nd and 171Yb) were narrow (
