A series of proof-of-principle studies was initiated to evaluate the soil remediation technology, phytoimmobilization, for application at the TNX Outfall Delta (TNX OD) operable unit. Phytoimmobilization involves two steps. The first step is entitled phytoextraction, and it takes place mostly during the spring and summer. During this step the plants extract contaminants from the sediment into the roots and then translocate the contaminants to the above-ground plant parts. The second step is referred to as sequestration and it takes place largely during the autumn and winter when annual plants senesce or perennial trees drop their leafs. This step involves the immobilization of the contaminant once it leaches from the fallen leaf.

Leaf litter at the site was found to contain measurable concentrations of the constituents of concern (COCs; actinium, cobalt, chromium, mercury, lead, radium, thorium and uranium). Equally important, the leaf litter at the site was found to have a large annual biomass flux, 2000 kg/ha/yr. As part of a preliminary survey of the indigenous plants, it was discovered that a fern, Woodwardia areolata, contained exceptionally high contaminant concentrations. The species was particularly effective at accumulating cobalt and chromium. It also removed high concentrations of lead, uranium and thorium, and low concentrations of mercury. A greenhouse study is underway that will further quantify contaminant accumulations by the ferns.

In separate laboratory studies, sequestering agents were evaluated. Pyrite, a sulfide mineral, was found to have a distribution coefficient (Kd value) of 20,000 mL/g for mercury. Hydroxyapatite, a phosphate source, was able to remove large amounts of cobalt (Kd = 7700 mL/g), europium (an analogue for actinium; Kd = 720,000), lead (Kd = 138,000 mL/g), and uranium (Kd = 282,000 mL/g). Clinoptilolite, a zeolite cation exchange mineral, effectively removed barium (an analogue for radium; Kd = 6200 mL/g). A field demonstration of the various sequestering agent has been set up at the TNX OD.

The manner in which phytoimmobilization might be deployed at the site is to use the existing trees, which generates a lot of leaf-litter biomass with moderate contaminant concentrations, and to plant ferns, which generate less biomass, but greater contaminant concentrations. The sequestering agents should be made up of a combination of hydroxyapatite, clinoptilolite, and a sulfide source. Pieces of geomat, a layered material in which the sequestering agent is held between two geofabrics, should be used to emplacement the sequestering agents around the existing trees.

This proposed approach to remediating the ecologically sensitive wetland has a number of attributes. First, this approach uses existing natural geocycling processes and simply interrupts these processes by accumulating the contaminants in the geomat. Secondly, this approach should have minimal impact on the ecologically sensitive wetland at the TNX OD site; the geomat may be removed or left in place, depending on the risk. Finally, the technology should greatly reduce the cost of waste disposal by creating a concentrated waste in the sequestering agent.

The constituents of concern (COCs) in these studies were Ac, Ba, Co, Cr, Hg, Pb, Ra, Th, and U. Evaluation of the technology was to be made relevant to the TNX OD operable unit. To reduce analytical cost, it was decided by SRTC and ER personnel to restrict analyses to ICP-MS and cold-vapor fluorescence spectroscopy, avoiding costly radiochemical analyses. Actinium and radium concentrations were below detection by the ICP-MS. Thus, Eu and Ce, both trivalent cations, were used as analogues for Ac biogeochemical behavior, and Ba, a divalent cation, was used as an analogue for Ra biogeochemical behavior. Europium, Ce, and Ra were easily detected by the ICP-MS. The cold-vapor fluorescence spectroscopy was used for Hg analyses.

The project was broken down into 7 tasks, the six experiments described in the Objective section and a Modeling task. The status of each task is presented in Table 1.

Unexpected delays in the project occurred as a result of the August breakdown of the ICP-MS instruments at SREL and SRTC. A second and final version of this document will be issued 11/30/00 that will include the remaining data not reported in this first version.

There were seven tasks in this project: the six experiments depicted in Figure 7 and the Linear Kinetic Reservoir Modeling. The Materials and Methods of each of these tasks will be described below.

The objective of this greenhouse study was to determine the contaminant concentrations that would accumulate in the above ground portion of ferns (Woodwardia areolata) and Bermuda grass grown in TNX OD contaminated soils. This study was conducted at the Savannah River Ecology Laboratory’s greenhouse facility. There was one uncontaminated and two contaminated soils (collected from coordinates B-5 and C-5 Figure 8). The fern was selected because, during a preliminary sampling, it was discovered that this plant species accumulated large concentrations of several of the targeted contaminants. The Bermuda grass was selected because its seeds are readily available, it will grow well in a wide range of growing conditions, and it is a monocot that may provide interesting insight into the mechanism of plant uptake.

The experimental factorial was 3 soil types, 2 plants, 4 replicates, and 2 fertilizer regimes. The two fertilizer regimes were with and without 25-kg/ha 10-10-10, N-P-K fertilizer added. In addition to these treatments there were 8 controls without plants: 2 fertilizer regimes x 2 soil types x 2 replicates. The uncontaminated soil provided a negative control. Two-kg of soil was added to each pot. Each soil-containing pot had one large hole on the bottom to permit excess water to leave the root zone. Each pot was placed in a larger pot without holes to contain contaminated water. Ferns were collected from a non-contaminated portion of the TNX OD and transplanted into the pots. The Bermuda grass was seeded directly into the pots. The plants were watered every work day and the water that would collect in the outside pot would be reintroduced to the soil each work day.

For the first sampling, approximately 50-g (wet) of plant material was collected. The sample was digested and then the samples were analyzed for contaminant concentrations. A second sampling is anticipated in early spring next year.

The objective of this experiment was to determine the rate that contaminants leached from leaves. The plant material used in this study was a fern collected from an uncontaminated area of the TNX OD and from coordinate A-5 (Figure 9), within the contaminated area. Five grams of each fern material were placed in separate dialysis bags and then placed in 200-mL of uncontaminated surface water collected from near the TNX-OD. These tests were conducted in duplicate. This set up was placed in a platform shaker and sampled periodically over a two-month period. The 10-ml aliquots were then submitted for analyses by ICP-MS and cold-vapor atomic fluorescence spectroscopy.

The objective of this laboratory study was to conduct a survey of potential sequestering agent materials. The following sequestering agents were tested:

• metallic iron (Fe),
• North Carolina apatite (NCA),
• hydroxylapatite (HA),
• zeolite (clinoptilolite -ZC and phillipsite-ZP),
• Fe oxide waste (Fe rich^TM waste byproduct from an industrial process that generates TiO₂ pigment; E.I. Du Pont de Nemours, Wilmington, DE),
• gypsum, and
• pyrite.

Due to the different geochemistry of the studied elements, three separate laboratory experiments were conducted: a Cr experiment; a Hg experiment; and a Ba (used as an analog for radium), Co, Pb, and U experiment. All experiments were conducted in centrifuge tubes for a period of one week. Each treatment had three replicates. All three experiments were conducted with two background solutions: distilled water and rainwater with dissolved organic carbon (decomposed leaf-litter). Rainwater was collected in Aiken, SC, and leaf-litter was collected at a non-contaminated area of TNX. Approximately, 0.3-g of each sequestering agent was shaken with 30-mL of spike solution. The spike solution contained 1-mg L^-1 Cr(VI), 2-mg L^-1 Hg(II), and 50-m g L^-1 of Ba, Co, Pb and U. The aqueous phases were analyzed for Cr, Ba, Co, Pb, and U by ICP-MS and for Hg by Cold-Vapor Atomic Fluorescence Spectrometer. The partitioning of the metals to the various sequestering agents was quantified by a distribution coefficient, Kd.

The objective of this study was to field test some of the better sequestering agents identified during the laboratory trials described above in Section 3.4.

The study was established in coordinate H-5 at the test site (Figure 9). Twenty-nine mesocosms were set up (Table 2). The mesocosms were 30-cm high and had a diameter of 15-cm. They were cut from PVC tubing. Within each mesocosm, the following layers were placed, starting from the bottom: 2-cm of sand (to act as a spacer so that the geomat does not come into contact with the contaminated soil), and 1.5-cm geomat. The geomat was made by cutting 16-cm diameter circles out of a geofabric (AMOCO Style 5412, Atlanta, GA). The edges of two geofabric circles were sewed together, leaving a 3-cm opening. Through this opening 400-g of sequestering agent was added. The opening was then sewed together. A screen was placed on top of the mesocosm to minimize the occurrence of leaves falling into the mesocosm, but at the same time permitting rain to enter.

Ferns and leaf litter will be collected this autumn from coordinates A-5 and B-5. Each material will be homogenized, ground into ~1-cm pieces, and then added to the mesocosms. At this time (time = 0), a subsample of the leaves will be analyzed for contaminant concentrations.

After 3 months, a subsample of the leaves and geomat material will be digested and analyzed for contaminant concentrations. Also at this time, it will be determined whether to continue the study as is, or to remove the remaining plant tissue and add new fresh material from the same homogenized sample used at the start of the experiment. If a great deal leaf decay has taken place, than the 9old leaves will be removed and more fresh leaves will be added.

Because the biomass flux will be known for both the leaf litter and the fern, the amount of plant material added to the mesocosms will be translated into terms of "number of years worth of plant material." This arrangement will permit us to evaluate tens of years worth of leaf litter in one year.

The objective of this study was to determine the relative availability of the various contaminants and to determine distribution coefficients, Kd values, that could be used in the Linear Kinetic Reservoir Model. The laboratory portion of this work was completed as part of a previous study (Kaplan et al. 1999). The site-specific field data taken from this report and that will be used in our modeling is presented below as part of the Results and Discussion of the Linear Kinetic Reservoir Model.

The linear kinetic reservoir model is based on the well-established geocycling model that was first proposed by Lasaga (1980). Computer code to execute the model has been written and preliminary input values have been selected. These input values are presented in the Results and Discussion Section. A full description of the application of this model for describing the phytoimmobilization at the TNX OD will be included in the final report.

Sixteen sets of soil, herbaceous, and leaf litter samples were collected for this survey. Plant samples were totally digested. It was important to collect soil samples along with the plant samples because, low concentrations in the plant tissue could be attributed to either low soil contaminant concentrations or low plant uptake rates. The soils were digested to provide a measure of the total amount of each contaminant in the soil and were also extracted with a DTPA solution to provide an index of the plant available fraction. For phytoimmobilization, a measure of the plant available fraction in the soil is important and likely represents only a fraction of the total. The percentage of DTPA extractable compared to the total fraction will likely vary with element. Strongly sorbing contaminants, such as lead and mercury, are likely to have appreciably less DTPA extractable concentrations than total concentration. Conversely, chromium, a weakly sorbing anion, is likely to have a relatively large DTPA-extractable fraction.

As discussed in the Status section, very little of the analytical results from the field survey data have been completed. All the analytical data for the total digestion and DTPA extraction is completed but was delivered to us too late to be included in this report. The biomass annual flux data and some plant and soil analytical results, albeit quite limited, have been completed and will be presented here.

In the first stage of phytoimmobilization, phytoextraction, herbaceous and woody plants are used to remove contaminants from soil. In this technology, plants are not harvested but fall and decompose on the sequestering agent. In this study, the main focus was on evaluating the ability of indigenous plants to take-up soil contaminants. This study permitted us to screen plant species for more in-depth future studies.

Selected properties of an uncontaminated soil collected from coordinate BGCH05 and a contaminated soil collected from coordinate A-5 are presented in Table 3. Both soils are acidic, contain moderate levels of organic matter and have a sandy texture. The cation exchange capacity (CEC) values are typical of this area. Also, reported in Table 3 are anion exchange capacity (AEC) values. This parameter is like CEC except it is for anions, and it has been shown to increase substantially in SRS soils under increasingly acidic conditions (Kaplan et al. 2000). More detailed properties and the geochemistry of the TNX OD soil are reported by Kaplan and Serkiz (1999).

Contaminant concentrations in soils are presented in Table 4. Concentrations of Ag, Cr, Co, Cu, Hg, Pb, Th, and U-238 are appreciably greater in the A-5 soil than in the background soil. For U-238, there is a four order-of-magnitude difference between the concentrations in these two soils. We are not interested in the concentration of Ba and Ce per se, but are interested in these elements only in so far that they can be used as analogues for Ra and Ac, respectively.

Table 5 contains the mass of leaf litter collected from 40 sampling baskets located in the contaminated area and 6 sampling baskets located in the uncontaminated area of the study site. However, the sampling baskets were placed after the deciduous leaves at the site had started to fall, consequently, they do not provide a measure of the annual leaf fall. Due to radiological safety concerns, we were not permitted to collect the leaves beneath the baskets that had already fallen in the contaminated area. However, we were able to collect and weigh the leaves beneath the collection baskets located in the uncontaminated area; this permitted calculation of the annual biomass flux (Table 6). Based on these 6 samples, the annual biomass flux is:

• 2669 ± 296 kg/ha/yr for the wetland soils, and
• 2000 ± 374 kg/ha/yr for the upland soils.

Only 24% and 33% of the total leaf litter that fell during the fall of 1999 was collected in the baskets located in the wetlands and uplands, respectively. For the fall of 2000, all 46 baskets will be in place at the start of the leaf-falling season and should provide additional data to estimate the biomass annual flux.

The material collected in 21 leaf litter baskets was separated by species; the percent of the mass of each species is presented in Table 7 and Table 8. A summary of the relative abundance of the dominant species is present in Table 9. Water oak, tupelo, baldcypress and loblolly pine account for 43% of the total leaf litter mass.

As part of a preliminary study, five plants and unsorted leaf litter collected from coordinate A-5 were analyzed for targeted contaminant concentrations (Table 10). In addition to reporting contaminant concentrations in the plants, the data was normalized for differences in soil contaminant concentrations by calculating concentration ratios (CR = mg kg^-1 dry plant / mg kg^-1 dry soil). One herbaceous plant, a fern (Woodwardia areolata), had a high propensity to uptake all targeted contaminants, especially U, Th, and Hg. Also important to note is that the leaf litter contained appreciable concentrations of the targeted contaminants. This is notable because the biomass of the leaf litter is large in this area, and will always account for a majority of the annual biomass, even if a monoculture of a hyperaccumulater was to be introduced to the site for remediation purposes. The leaf litter had high concentration of Co (17 mg/kg) and also a high CR (4.59). The high Co concentration and CR in the leaf litter needs to be further evaluated.

The Plant Uptake Experiment is a greenhouse study that is growing Bermuda grass and a fern, Woodwardia areolata, found in the preliminary field survey to take up high concentrations of the contaminants. The plants are growing in 2 soils collected from the TNX OD contaminated area. The first of two harvests has been completed. The second harvest will be conducted later next year. No analytical data is presently available.

The Leaf Leaching Experiment is designed to quantify the rate that contaminants leach from various plant materials. The two plant materials being evaluated Woodwardia areolata and unsorted leaf litter, were both collected from coordinate A-5. The experiments were conducted in duplicate. No analytical data is presently available.

In the second stage of phytoimmobilization, sequestration, contaminants from decomposed plant material would not partition directly to soil, but instead, sorb onto the sequestering agent, or geomat. Geomats are made from minerals, such as apatite, zeolite, metallic iron, and others. Ground-up minerals can be placed inside of geotextile material (e.g., AMOCO Style 5412).

Mixtures of metallic iron with other sequestering agents were evaluated because it is likely that more than one material will be required to sequester all eight contaminants. The addition of North Carolina apatite, hydroxyapatite, and clinoptilolite (a type of zeolite) greatly decreased the Cr (VI)-removal effectiveness of the metallic iron. The pH of the equilibrium solutions at the end of the contact time between the sequestering agents and the aqueous Cr(VI) did not vary greatly between treatments. As expected, the pH of the metallic iron, Fe(0), treatment increased. This can be attributed to the metallic iron reducing water, thereby creating hydroxides as shown in Equation 2.

Fe⁰ + H₂O = Fe²⁺ + OH^- + 1/2H_2(gas) (2)

Table 11. Chromium distribution coefficients (Kd) for several sequestering agents in a SRS surface water.

Evaluation of Hg sequestering agents was conducted in uncontaminated SRS surface water containing 6 mg/L total organic carbon, and a leaf litter leachate containing >100 mg/L total organic carbon. The Kd values in the presence of dissolved organic matter tended to be lower than when the organic matter was not present (Table 12). Thus, it appears the presence of the strongly complexing organic ligands present in plant leachate moderately decrease the removal effectiveness with these sequestering materials.

Pyrite had a Kd value ~ 20,000 mL/g and metallic iron had a Kd value of >1000 mL/g (Table 12). The reduction in soluble Hg concentrations in pyrite (FeS₂) treated solutions is not surprising because the pyrite can release sulfides, which can then combine with Hg(II) to form the sparingly soluble mercury-sulfide species, cinnabar, HgS (Bodek et al. 1988). The relatively low Kd for the gypsum (CaSO₄) addition may be attributed to the fact that the test suspension was oxic, therefore the sulfate in the gypsum was never reduced to sulfide, the form of sulfur that forms the sparingly soluble mercury precipitate. In a soil system where microbes are present and where reducing conditions exist, it is very likely that gypsum will be a better sequestering agent for mercury. Thus, evaluation of gypsum for this application needs to be conducted under conditions more similar to those of its final application, namely in the presence of soils that may convert the sulfate to sulfide.

The larger the redox potential, the more oxidized the system is; conversely the smaller the redox potential, the more reduced the system is. At pH 5, the approximate pH of the TNX OD soils, the expected redox range is between 50 (reduced) and 650 mV (oxidized). In the metallic iron system, the pH rose to 10.2 and the oxidation-reduction potential (or redox) dropped to 110 mV (Table 12). The rather large metallic-iron reduction potential is appreciably larger than expected; values of ~ 200 are common. Faust and Osman (1981) reported similar reductions in soluble Hg concentrations in experimental systems when pH values increased and Eh values decreased. As described above, the elevated pH in the metallic iron system can be attributed to the water being reduced to hydrogen gas and hydroxide ions. The low pH of the pyrite addition can be attributed to the oxidation of pyrite, which is a well known acidifying process that is a widespread occurrence at coal mining sites.

In the third experiment, six elements were added to the solution at concentrations of approximately 50 m g/L. After one week the concentrations of each tested element were significantly reduced by almost all the amendments (Table 13).

Hydroxylapatite and North Carolina Apatite were the most effective at reducing Eu, Pb, and U concentrations in the solutions and K_d values in these treatments were very high (Table 13). Other researchers also reported the high effectiveness of apatite in remediation of Pb or U contaminated soils through a precipitation mechanism (Knox et al. 2000). Both zeolites were effective for most elements, however, not for uranium. Metallic iron and Fe oxide (Fe-rich ^TM, waste byproduct) removed the greatest mass of the targeted constituents from solution. The metallic iron could be attractive as a geomat material because it was effective in immobilizing Cr, Hg, Co, Ba, Eu, Pb and U; however, one problem with metallic iron is that it has a limited life span before it oxidizes, rusts, and looses its reductive capacity. Thus, if metallic iron was to be included in a geomat, then the material would have to be removed before the metallic iron completely rusted.

The first part of this project, setting up in the field, was completed. Results will become available as soon as next spring.

Table 15. Preliminary Input Values for the Linear Kinetic Reservoir Model of Phytoimmobilization (Additional input in previous table).

The manner in which phytoimmobilization should be deployed at the site is to use both the natural leaf litter, which generates a lot of biomass with moderate levels of contaminants, and plant ferns to phytoextract the contaminants near the soil surface, where the contaminants are most concentrated. The geomat should be made up of a combination of hydroxyapatite, clinoptilolite, and a sulfide source. The geomat should be fabricated by placing the sequestering agent in between two pieces of geotextile material, and then placed in pieces around the existing trees.

Leaf litter at the site was found to contain measurable concentrations of the constituents of concern (COCs; actinium, cobalt, chromium, mercury, lead, radium, thorium and uranium). Equally important, the leaf litter at the site was found to have a large annual biomass flux, 2000 kg/ha/yr. As part of a preliminary survey of the indigenous plants, it was discovered that a fern, Woodwardia areolata, contained exceptionally high contaminant concentrations. The species was particularly effective at accumulating cobalt and chromium. It also removed high concentrations of lead, uranium and thorium, and low concentrations of mercury. A greenhouse study is underway that will further quantify contaminant accumulations by the ferns.

In separate laboratory studies, sequestering agents were evaluated. Pyrite, a sulfide mineral, was found to have a distribution coefficient (Kd value) of 20,000 mL/g for mercury. Hydroxyapatite, a phosphate source, was able to remove large amounts of cobalt (Kd = 7700 mL/g), europium (an analogue for actinium; Kd = 720,000), lead (Kd = 138,000 mL/g), and uranium (Kd = 282,000 mL/g). Clinoptilolite, a zeolite cation exchange mineral, effectively removed barium (an analogue for radium; Kd = 6200 mL/g). A field demonstration of the various sequestering agent has been set up at the TNX OD.

This proposed approach to remediating the ecologically sensitive wetland has a number of attributes. First, this approach uses existing natural geocycling processes and simply interrupts these processes by accumulating the contaminants in the geomat. Secondly, this approach should have minimal impact on the ecologically sensitive wetland at the TNX OD site; the geomat may be removed or left in place, depending on the risk. Finally, the technology should greatly reduce the cost of waste disposal by creating a concentrated waste in the sequestering agent.

The information contained in this report was developed during the course of work under Contract No. DE-AC09-96SR18500 with the U.S. Department of Energy. Dr. Rebecca Sharitz and Mr. Bruce Allen (Savannah River Ecology Laboratory, Aiken SC) contributed greatly to this research but were unable to review the document before it was issued. They will be co-authors on the next version of this document.

7.0 References

Bodek, W.J. Lymn, W.F. Reehel, D.H. Rosenblott, B.I. Walton, R.A. Conway, Environmental Inorganic Chemistry: Properties, Processes, and Estimation Methods, New York, Pergamon Press, (1988).

Cantrell, K. J., and D. I. Kaplan. 1998. "Sorptive Barriers for Groundwater Remediation." pp. 4663 - 4677. In Robert A. Meyers (ed.), Encyclopedia of Environmental Analysis and Remediation, John Wiley & Sons, Inc., New York.

Cantrell, K. J., D.I. Kaplan, D.I., T. W. Wietsma, J. Hazardous Materials. 42, 201-212, (1995).

Faust, S. D. In: Chemistry of Natural Waters, Ann Arbor Science Publishers, Inc., (1981).

Kaplan, D. I., S. V. Mattigod, K. E. Parker, and G. Iversen. I-129 Test and Research to Support Disposal Decisions. WSRC-TR-2000-00283, Rev. 0, Westinghouse Savannah River Company, Aiken, SC (2000).

Knox, A. S, J.C. Seaman, M.J. Mench, J. Vangronsveld, 2000, Remediation of Metal- and Radionuclide-Contaminated Soils by In Situ Stabilization In: Environmental Restoration of Metals Contaminated Soils, Ed. IK Iskandar, CRC Press, Boca Raton, FL., (2000).

Lasaga, A. C. 1980. "The Kinetic Treatment of Geochemical Cycles." Geochimica et Cosmochimica Acta. 44, p 815 (1980).

WSRC. RFI/RI with BRA for the TNX Outfall Delta, Lower Discharge Gully and Swamp Operable Unit. WSRC-RP-98-4158, Rev. 0. Westinghouse Savannah River Company, Aiken, SC. (1999).