Luận án tiến sĩ: Nonequilibrium transport of heavy metals in soils and its influence on soil remediation

Nonequilibrium Transport of Heavy Metals in Soils and Its Influence on Soil Remediation Daniel Chiu Wa TSANG Department of Civil Engineering The Hong Kong University of Science and Tech

Introduction 1

Background - 5S Ăn SH HT Khen 1

Heavy metal contamination is one of the most common problems constraining cleanup at hazardous waste sites around the world For instance, more than 60% of the sites on the U.S Environmental Protection Agency National Priority List and over 1,400,000 sites in Western Europe were contaminated with heavy metals Metal most often encountered include lead, chromium, copper, zinc, arsenic, and cadmium (Peters 1999; Sun et al 2001) Excavation of polluted soil without remediation is the most common remedial practice up to early 1990’s, but this does not offer a permanent solution (Papassiopi et al 1999) There are two main types of treatment for metal-contaminated soils: technologies that leave the metal in the soil (solidification/stabilization and vitrification), and technologies that remove the heavy metals from the soil (in-situ soil flushing and ex-situ soil washing) As solidification/stabilization becomes increasingly costly due to limited landfill space and processing costs, soil remediation that can give a permanent solution to the contamination problem is urged, allowing recycling of treated soil and future land-use options (Peters 1999)

In general, in-situ technologies are more economical and safer than ex-situ technologies because excavation of contaminated soil is avoided, requiring smaller amount of facilities and reducing the potential for human exposure to contaminants (Davis and Sigh 1995; Reed et al 1996) An effective cleanup by in-situ flushing is built on the understanding of pollutant transport However, the traditional equilibrium transport model has been found unsuccessful in describing the actual movement of pollutants in soils While the equilibrium transport assumptions may be valid for large time and space scales, it becomes questionable for in-situ remediation of contaminated field soils since the time and space scales are often limited (Wagenet and Chen 1998) A great challenge to remediation comes along with the failure of the majority of in-situ systems to reduce the contaminant levels to health-based standards within the expected time due to tailing behavior; and rebound of solution concentration often occurs after cessation of pumping (Figure 1.1) (Mackay and Cherry 1989; USEPA 1992a, 1996) These observations have aroused an extensive attention on nonequilibrium transport, which probably results from rate-limited (i.e slow) sorption/desorption (Brusseau and Rao 1989; Weber et al 1991; Selim 1992; Scheidegger and Sparks 1996; Luthy et al 1997)

FIGURE 1.1 Contaminant concentration versus pumping volume showing tailing and rebound effects (From USEPA 1996)

Investigation of the nonequilibrium transport in soils is a multi-disciplinary endeavor The influence of each factor should be investigated as isolated as possible (Brusseau 1998a) There is a need for research on how the remediation operation (e.g flushing velocity) and contaminated site conditions (e.g subsurface temperature, presence of multi-metals, metal loading levels) influence the rate-limited reactions and the resultant nonequilibrium transport In addition, since the parameters of transport modeling probably lump the several closely correlated factors together,

2 complementary experiments may be essential to enrich the technical know-how about transport behavior

In order to facilitate the removal of the rate-limited and/or resistant contaminants from soils, addition of chemical reagents into flushing solution has been suggested as chemical enhancements The reagents include oxidant/reductant, strong acids or chelating agents (for heavy metals), cosolvents or surfactants (for hydrophobic organics) (USEPA 1992b) Thereafter, an increasing number of fundamental studies and pilot-scale applications have been implemented (GWRTAC 1997, 1998) The rationale underlying the use of chemicals is that rapid extraction of concentrated solutions is beneficial to remediation, which is justified by shorter operation time, lower facility and maintenance cost, and the ease of treating smaller volume of concentrated wastes But it has been shown that the extraction kinetics by chemical agents, EDTA for example, is time-dependent on the order of hours or days (Bermond and Ghestem 2001; Wasay et al 2001; Tandy et al 2004) Therefore, the chemical-enhanced extraction is still probably rate-limited on the transport timescale of in-situ flushing, which would in turn lead to nonequilibrium transport behavior However, the significance of nonequilibrium transport during chemical-enhanced flushing has gone nearly unheeded to date

Heavy metal transport in four soil samples was investigated using column experiments in which Cd and/or Cu were injected The experimental data, plotted as breakthrough curves (BTCs), were analyzed with linear/nonlinear equilibrium and nonequilibrium transport models Additional experiments including soil characterization, batch kinetics and isotherms, sequential extraction, flow interruption, and non-sorbing solute transport were also implemented Besides, ethylenediaminetetraacetic acid (EDTA), as a reference EDTA-flushing of Cu-contaminated soils was evaluated in terms of the extraction effectiveness and the coupled impacts on soils A transport model was developed to simulate the observed nonequilibrium transport of metals in the course of EDTA-flushing

This study attempts to improve the understanding on nonequilibrium transport of heavy metals in soils under different contamination or remediation conditions To achieve this, the objectives can be specifically underlined as follows: e to study the disparity of equilibrium and nonequilibrium transport behavior at different pore-water velocities and temperatures; e to evaluate the competitive effects on metal sorption kinetics and corresponding transport behavior; e to identify the relative contribution of sorption nonlinearity and rate-limited sorption to the nonideal transport behavior; e to examine the metal extraction effectiveness and soil dissolution issues during EDTA-flushing; and ¢ to model and simulate the nonequilibrium transport behavior of metal complexes resulting from EDTA-promoted extraction and dissolution

This thesis has nine chapters, beginning with an introductory chapter The purpose of Chapter 2 is to provide a comprehensive and critical review of the current state of knowledge on heavy metal sorption and transport in soils Heavy metal contamination and soil properties are briefly reviewed (Section 2.1) Sorption mechanisms and the influential factors are discussed (Section 2.2), followed by conceptual and mathematical development of equilibrium and nonequilibrium solute transport (Section 2.3) This chapter ends with pointing out the key issues to be addressed in this research (Section 2.4)

Chapter 3 outlines the experiments and analysis approach employed in the subsequent chapters, including soil characterization, solution preparation, batch experiments, column experiments, sequential extraction, and transport modeling It provides sufficient details to allow the work to be reproduced elsewhere The methods that have been well characterized and discussed are referred to the corresponding references, while the necessary modifications are described

Chapters 4-6 investigate the relationship between nonequilibrium transport and subsurface conditions including flushing rate, temperature (Chapter 4), presence of competitive metal (Chapter 5), and metal loading (Chapter 6) An approach incorporating multiple experiments (batch, sequential extraction and column) is proposed to aid transport modeling and interpretation Chapters 7&8 put emphasis on nonequilibrium transport of metals during chemical-enhanced flushing Extraction effectiveness and soil dissolution issues are assessed under different contamination conditions (Chapter 7) The corresponding transport behavior of each metal is monitored and simulated with a model development taking second-order kinetics of extraction and dissolution into account (Chapter 8) This thesis concludes with Chapter 9 which highlights the most important findings and accordingly suggests relevant research directions in future.

Cadmium (Cd) and copper (Cu) are of interest in this study, since they are included in the list of priority metals because of their potential hazard to human health (McBride 1994) A large number of field sites have been contaminated with Cd (Neale et al 1997; Seuntjens et al 2001; Collins et al 2003; Voegelin et al 2003) and/or Cu (McBride et al 1998; Atanassova 1999; Thayalakumaran et al 2003) over the past decades by industrial wastes, land application of sewage sludge, atmospheric deposition of smelter dust, and military operation Metals can be present in soil as either sorbed or particulate (or elemental) contaminants Sorbed metals are typically concentrated in the fines fraction of soils, while discrete metal-mineral phases would be present when metal contamination of soils is very high, i.e thousands of mg/kg, at near neutral or alkaline conditions (Davis and Sigh 1995; Neale et al 1997; Davis and Hotha 1998)

The metal bioavailability and toxicity are usually assumed to be controlled by the buffered activity of the free ions in solution, while biological uptake of metals as metal complexes cannot be precluded (Sauve et al 2000a) Cu is more toxic to plants (also termed phyto-toxic) than animals whereas Cd is particularly toxic to higher animals (McBride 1994)

Cd has no essential biological function or redox reaction and bioaccumulates in the food web The toxicity of Cd is believed to be caused by competition with Zn for metal-binding sites in proteins (Karlsson et al 2005) Acute exposure to Cd leads to nausea, vomiting, salivation, diarrhea, and muscular cramps and even to fatal cases Chronic exposure to Cd causes dysfunction of kidney as well as lung, bone, cardiovascular system, liver, and reproductive system damage (Alloway 1995; Appel and Ma 2002)

Cu is a ubiquitous metal ion in soil and aquatic environments, and at background levels, it poses no serious threat to the biota and vegetation It is a necessary element for a great number of biochemical processes Elevated levels of Cu do not inflict significant adverse effects on human health but are detrimental to the environment that plant growth becomes visually retarded (Alloway 1995; Flogeac et al 2004)

The loading limits established in the USEPA 503 Regulations are 1500 and 39 kg/ha for Cu and Cd, respectively (McBride et al 1997) Therefore, commonly reported contaminations of these metals range from tens to thousands of mg/kg (Alloway 1995; McBride et al 1998; Thayalakumaran et al 2003; Voegelin et al 2005) Instead, contaminated soils require remediation if the total Cd content exceeds the intervention level of 12 mg/kg; or if the soil Cu concentration is above the intervention level of 190 mg/kg, according to Dutch Soil Standard In addition, in view of the high toxicity of Cd to human health, the maximum contaminant levels of

Cd and Cu in drinking water are 0.005 and 1.3 mg/L, according to U.S Environmental Protection Agency and World Health Organization.

Fertiizer industrial ; Raina t Dust input so wastes ` ' ẹ " —>~ Food —eeHuman beings " i ,

Sorbed and ca desorption : immobilization Organic phase exchangeable 3¢ ior - — * Organic matter vw sorption 7 mineralization * microorganisms

Solid phases Gas phase minerais ‡ Poo, Po,

FIGURE 2.1 Heavy metal dynamics in soil environment (From Naidu et al 1997)

Both soil properties and soil solution composition determine the dynamic equilibrium between metal in solution and on the solid phase (Figure 2.1) Precipitation appears to be the predominant process in the presence of anions such as S*, COzŸ, OH and PO¿Ÿ and when the metal concentration is high Sorption of Cd at soil mineral surfaces may occur by both specific and nonspecific processes Understanding the mechanisms involved in the retention and mobility of heavy metals is a necessary precursor for the determination of the metal distribution in soil profile, environmental impact and regulatory measures of metal contamination (Naidu et al 1997) The interactions between metals and soils also determine the bioavailability and transport of metals in soils Therefore, the nature of soil components, metal sorption mechanisms, and corresponding influential factors are discussed in the following sections.

Thesis OrganizafiOn - Ăn HH HH tờ 4

This thesis has nine chapters, beginning with an introductory chapter The purpose of Chapter 2 is to provide a comprehensive and critical review of the current state of knowledge on heavy metal sorption and transport in soils Heavy metal contamination and soil properties are briefly reviewed (Section 2.1) Sorption mechanisms and the influential factors are discussed (Section 2.2), followed by conceptual and mathematical development of equilibrium and nonequilibrium solute transport (Section 2.3) This chapter ends with pointing out the key issues to be addressed in this research (Section 2.4)

Chapter 3 outlines the experiments and analysis approach employed in the subsequent chapters, including soil characterization, solution preparation, batch experiments, column experiments, sequential extraction, and transport modeling It provides sufficient details to allow the work to be reproduced elsewhere The methods that have been well characterized and discussed are referred to the corresponding references, while the necessary modifications are described

Chapters 4-6 investigate the relationship between nonequilibrium transport and subsurface conditions including flushing rate, temperature (Chapter 4), presence of competitive metal (Chapter 5), and metal loading (Chapter 6) An approach incorporating multiple experiments (batch, sequential extraction and column) is proposed to aid transport modeling and interpretation Chapters 7&8 put emphasis on nonequilibrium transport of metals during chemical-enhanced flushing Extraction effectiveness and soil dissolution issues are assessed under different contamination conditions (Chapter 7) The corresponding transport behavior of each metal is monitored and simulated with a model development taking second-order kinetics of extraction and dissolution into account (Chapter 8) This thesis concludes with Chapter 9 which highlights the most important findings and accordingly suggests relevant research directions in future.

Heavy Metals and SOiÌS - 2S nọ 6

Cadmium (Cd) and copper (Cu) are of interest in this study, since they are included in the list of priority metals because of their potential hazard to human health (McBride 1994) A large number of field sites have been contaminated with Cd (Neale et al 1997; Seuntjens et al 2001; Collins et al 2003; Voegelin et al 2003) and/or Cu (McBride et al 1998; Atanassova 1999; Thayalakumaran et al 2003) over the past decades by industrial wastes, land application of sewage sludge, atmospheric deposition of smelter dust, and military operation Metals can be present in soil as either sorbed or particulate (or elemental) contaminants Sorbed metals are typically concentrated in the fines fraction of soils, while discrete metal-mineral phases would be present when metal contamination of soils is very high, i.e thousands of mg/kg, at near neutral or alkaline conditions (Davis and Sigh 1995; Neale et al 1997; Davis and Hotha 1998)

The metal bioavailability and toxicity are usually assumed to be controlled by the buffered activity of the free ions in solution, while biological uptake of metals as metal complexes cannot be precluded (Sauve et al 2000a) Cu is more toxic to plants (also termed phyto-toxic) than animals whereas Cd is particularly toxic to higher animals (McBride 1994)

Cd has no essential biological function or redox reaction and bioaccumulates in the food web The toxicity of Cd is believed to be caused by competition with Zn for metal-binding sites in proteins (Karlsson et al 2005) Acute exposure to Cd leads to nausea, vomiting, salivation, diarrhea, and muscular cramps and even to fatal cases Chronic exposure to Cd causes dysfunction of kidney as well as lung, bone, cardiovascular system, liver, and reproductive system damage (Alloway 1995; Appel and Ma 2002)

Cu is a ubiquitous metal ion in soil and aquatic environments, and at background levels, it poses no serious threat to the biota and vegetation It is a necessary element for a great number of biochemical processes Elevated levels of Cu do not inflict significant adverse effects on human health but are detrimental to the environment that plant growth becomes visually retarded (Alloway 1995; Flogeac et al 2004)

The loading limits established in the USEPA 503 Regulations are 1500 and 39 kg/ha for Cu and Cd, respectively (McBride et al 1997) Therefore, commonly reported contaminations of these metals range from tens to thousands of mg/kg (Alloway 1995; McBride et al 1998; Thayalakumaran et al 2003; Voegelin et al 2005) Instead, contaminated soils require remediation if the total Cd content exceeds the intervention level of 12 mg/kg; or if the soil Cu concentration is above the intervention level of 190 mg/kg, according to Dutch Soil Standard In addition, in view of the high toxicity of Cd to human health, the maximum contaminant levels of

Cd and Cu in drinking water are 0.005 and 1.3 mg/L, according to U.S Environmental Protection Agency and World Health Organization.

Fertiizer industrial ; Raina t Dust input so wastes ` ' ẹ " —>~ Food —eeHuman beings " i ,

Sorbed and ca desorption : immobilization Organic phase exchangeable 3¢ ior - — * Organic matter vw sorption 7 mineralization * microorganisms

Solid phases Gas phase minerais ‡ Poo, Po,

FIGURE 2.1 Heavy metal dynamics in soil environment (From Naidu et al 1997)

Both soil properties and soil solution composition determine the dynamic equilibrium between metal in solution and on the solid phase (Figure 2.1) Precipitation appears to be the predominant process in the presence of anions such as S*, COzŸ, OH and PO¿Ÿ and when the metal concentration is high Sorption of Cd at soil mineral surfaces may occur by both specific and nonspecific processes Understanding the mechanisms involved in the retention and mobility of heavy metals is a necessary precursor for the determination of the metal distribution in soil profile, environmental impact and regulatory measures of metal contamination (Naidu et al 1997) The interactions between metals and soils also determine the bioavailability and transport of metals in soils Therefore, the nature of soil components, metal sorption mechanisms, and corresponding influential factors are discussed in the following sections.

Soils are extremely complex materials reflecting the variability of the parent rock material and organic residues from which they form The soil components can be categorized into soil minerals (primary silicates, clay minerals, oxides/hydroxides, carbonates and sulfates) and soil organic matter (unaltered debris and humus) (Sposito 1989) Of soil minerals, the principal primary minerals (i.e rock-forming minerals) are primary silicates including quartz, feldspars, micas, etc; the secondary minerals (i.e products of weathering of primary minerals) include layer silicate clays, oxides, and non-crystalline aluminosilicates The secondary minerals are predominant in the clay fraction (McBride 1994) The chemical reactivity of soils is mainly attributed to secondary minerals and decomposed organic matter (i.e humus) on the grounds of their high surface area and chemical structures (McBride 1994) Thus, a brief introduction of these components is provided in this section

Clay minerals (with particle size < 2um) usually refer to aluminosilicates that predominate in the clay fractions of soils, which are made up of tetrahedral and octahedral sheet silicates that are bonded together by the sharing of oxygen ions (Sposito 1989) Their contribution to soil chemical properties results from their comparatively large surface area and negative surface charge They are classified by the number of tetrahedral and octahedral sheets combined and the kinds of isomorphic cation substitution The most common types of clay mineral include: Kaolinite comprising one silica sheet and one gibbsite sheet referred to as a 1:1 clay mineral, which has very limited isomorphic substitution and swelling in water; smectite (principally montmorillonite in soils), vermiculite and illite, which comprise two silica sheets and one gibbsite sheet (2:1) and each represents a reasonably well-defined range of chemical compositions.

Clay minerals contain both constant and variable sources of surface charge with the significance of each source being aluminosilicate group dependent (Sparks 1996) The layer charge decreases in the order illite > vermiculite > smectite >> kaolinite, because of the extent of isomorphic substitution within the 2:1 layer mineral lattice (resulting in permanent negative charge) and abundance of hydroxyl groups at the edges (aluminol (AJOH) and silanol (SiOH) groups) (resulting in variable charge)

In addition, structurally disordered (i.e noncrystalline) aluminosilicates, formed from volcanic ash, are known collectively as allophane Although its structure is variable, allophane is thought to consist of 1:1 aluminosilicate layer in which AI? occupies both octahedral and tetrahedral sites Allophane does not seem to possess permanent charge and it is often found in association with clay minerals of kaolinite group More detailed information can be found in literature regarding their composition and structures (Sposito 1989; Stumm 1992; McBride 1994) and identification using X-ray diffraction analysis and differential thermal analysis (Tan

While the layer silicates are dominant in soils that have not been subjected to intense of prolonged weathering (e.g soils in glaciated or arid regions), large areas of the earth’s surface (e.g tropics or subtropics) are characterized by ancient soils, of which the mineral fractions are typically composed of nonsilicate minerals, namely oxide minerals (a collective term including oxides, hydroxides, and oxyhydroxides) The

Fe, Al and Mn oxides are the most important The oxides are structurally simpler than the layer silicates, consisting of hexagonal or cubic close-packed O* and/or OH” anions with Fe**, Al?*, Mn** or Mn** residing in octahedral sites They are not inclined to develop structural charge as a result of isomorphic substitution, but they develop limited variable charge (McBride 1994)

Among the Fe oxides, goethite (crystalline, common in temperate region) is the most often found in soils regardless of climatic region Under oxic conditions and inhibition of crystallization by ligands, ferrihydrite (poorly crystalline, common in temperate region) may precipitate Ferrihydrite can transform either to hematite (crystalline, common in hot climates) that ultimately transform to goethite, or to goethite directly Goethite is responsible for yellow to yellowish-brown colors, while hematite (second most abundant Fe oxides) is the reason for red colors of soils The most commonly found Al and Mn oxides are gibbsite (crystalline, common in humid tropics) and birnessite, respectively (Sposito 1989; Tan 1993; McBride 1994) The amount of oxides is usually related to the extent of weathering process

In partnership with the above fractions, soil organic matter has an extremely important influence on the chemical and physical properties of soils although they occupy just a few percent of most mineral soils, except peat and organic soil (Alloway 1995; Sparks 1996) Organic matter refers to the nonliving organic materials that are plant, animal, and microbial residues, classified as unaltered debris, and to the transformed products of decomposition of dead plant and animal matter, termed as humus Much of the reactivity of organic matter is attributed to humus Humus can be further divided into humic and nonhumic substances All the recognizable and identifiable organic compounds, such as polysaccharides, lignins, and polypeptides, are categorized as nonhumic substances The remaining amorphous, highly transformed, darkly colored materials are humic substances

It is generally accepted that humic substances contain several major functional groups: carboxyl (R-COOH), phenolic OH (aromatic-OH), alcoholic OH (R—CH2—OH), and carbonyl (R~C=O-—R/H/OR) (Xia et al 1997a; Sparks 2003), and thus are variable-charge materials The negative charge of the humic substances results from deprotonation of acidic functional groups (Christl and Kretzschmar

na e

Materials and Methods 71 côn 9c on ố ố e

Four soil samples were taken from 25-50 cm below ground surface at Hong Kong University of Science and Technology (UST & UST2 soils), Tai Mo Shan (TMS soil), and Clearwater Bay (CWB soil) in Hong Kong The UST, TMS and CWB soils were used in Chapters 4&6, whereas UST2 soil was used in Chapters 5, 7&8 (referred to “the soil” in the text of these chapters) The soil samples were air-dried and passed through a 2-mm sieve, which are the most reactive fraction of soils The particle size distribution, obtained by sieving and hydrometer methods, was used to determine the mass % of sand, silt and clay The soil pH was measured at a 1:2 soil-to-water ratio The organic carbon contents of the soils were analyzed using a total organic carbon analyzer (Shimadzu TOC-SO00A) with infrared spectrometer after combustion in a furnace (total carbon) and acidification (total inorganic carbon) The BET surface area was measured by nitrogen gas adsorption (Micromeritics ASAP2010) The cation exchange capacity (CEC) of the soils was determined by NH4,-Na exchange (USEPA method 9081) The clay fractions were separated by pipette method, of which the most abundant minerals were determined by X-ray diffraction (XRD) analysis (PW 1825, Philips) Total heavy metal contents in the soils were measured by X-ray fluorescence (XRF) spectrometry Amorphous Fe and

Al oxides were determined using ammonium oxalate extraction and the total amount of Fe oxide determined by DCB (dithionite-carbonate-bicarbonate) extraction Details of the above methods for soil characterization are referred to Methods of Soil Analysis: Parts 1, 3 and 4 (Klute 1986; Sparks 1996; Dane and Topp 2002) The soil characteristics of the fours soil samples are summarized in Table 3.1 cL

'Á1Jaulo122đS (;Ti[X) 22ue2sa1on|J Ấ1-X tSISÁJEut ((TIX) UOI2EJIID Ấ61-X :98ut2x2 #N-THN ‘pomou Lag uondiospe ussoniy “UOQIED STUBSIOUT - UOGIED [e}0} = UOGIED 21UEổLO '19ZÁ[EUt UOQ12 216810 [E10], :OIB1 12JÊAA-O1-[IOS Z:] É :SDOU]U 12†2UIOIĐÁU pư 8u1A2IS b0/0/L06/6/6 Đty0/91/8/1/V66 PVO/TE/OS7/09 €0/S66/661/601 310105 91111965 911uIO65 31uoš5I b[c0 69T'0 ore 0 8cF0 UI6T 06/0 0 381) t0 té LOTT vO'L 6c ty90 Icc0 Ly'l 9190 ev GS OL 9L UIEO'] Ae) uIEƠT uI£ŒT Apues weoy Apues ct/cWSC 02/LtCy 91/00/y9 6/91/SL aM) SIAL c1SnA Lsa 3 (% SSEI)

WJ/8.2/[Y/2đ ¡I8 Ae › (3 Đa) 2/12 p (¡3,) 831V 328J1nS 310đ0.121JAJ p (r3;U) 891V 928110 1T s (%) u0q1e2) 2qu681@ ạ Hđ [I0S u0I)821JISSE[“)

The background solution (or background electrolyte) was prepared by dissolving 0.03 M NaNO; into ultra-pure water (quality > 17.5 MQ) under acidic condition (pH

5, adjusted by dilute HNO3) in which hydroxide and carbonate concentrations were negligible Cd, Cu, and Br solutions were prepared by dissolving analytical-grade Cd(NO3)2, Cu(NO3)2, and NaBr, respectively in the background electrolyte NO3 ions were unlikely to form stable complexes with metals so that all the dissolved metals were assumed to remain as free metal ions The slightly acidic condition is prevalent in regions with high annual rainfall (e.g tropical regions) and industrial sites (Section 2.1) It was reported that no precipitation of Cu and Cd would occur until pH>6-6.5 (Appel and Ma 2002; Flogeac et al 2004) Control experiments conducted in the absence of soil showed no loss of metals from the solution over 7 days In Chapters 7&8, EDTA solution was prepared by dissolving analytical-grade disodium salt of EDTA (Na2EDTA) into solution of NaNO3, maintaining a constant ionic strength of 0.03 M and pH 5

The metal concentrations in the solution were measured by Atomic Absorption Spectrometry (AAS) (Hitachi Z-8200 or Varian SpectrAA 220FS), of which the detection limits of Cd and Cu are 0.02 and 0.03 mg L", respectively Br concentration was determined by Ion Chromatography (IC) (Dionex DX-500), of which the detection limit of Br ¡is 0.01 mg Lˆ” In view of possible dissolution of soil minerals and organic matter, these parameters were also monitored The concentrations of Fe, Al, and Ca were analyzed by AAS, of which the detection limits are 0.06, 0.3, and 0.01 mg L”, respectively The concentration of dissolved organic carbon was measured by Total Organic Carbon (TOC) analyzer (Shimadzu TOC-5000A) The amount of dissolved organic matter was reflected by measuring

UV absorbance at a wavelength of 254 nm, using UV/visible spectrophotometer (Milton Roy Spectronic 3000 Array) The interferences (i.e absorbance) by 50 mg L" Cu, Cd, Fe, Al, and Ca were 0.053, 0.043, 2.365, 0.082, and 0.1 18, respectively

Batch sorption experiments were performed to determine the sorption kinetics and equilibrium in the soils in Chapters 4—6, while batch extraction experiments were conducted to measure the EDTA-extractable amounts of metals of the soils in Chapter 8 Prior to the batch sorption experiments, it was necessary to saturate and condition the soil to the desired pH and ionic strength The soils were washed with the background solution for at least three times, each over a period of 24 h, until soil solution pH was close to 5 and ionic strength was maintained at 0.03M A soil-to-solution ratio of 1:10 (240.005 g soil in 20 mL solution) was used, except that Chapter 5 employed ratios of 1:10 and 1:50 because competitive effects depend on solid concentration The polypropylene centrifuge tubes containing the soil suspension were shaken end-over-end at 25 rpm for different equilibration periods, then centrifuged at 3500 rpm for 10 min and filtered using 0.45-m membrane filter (if necessary) The metal concentrations in the supernatants were measured by AAS All batch experiments were triplicated The sorbed amount was calculated from the initial concentration, equilibrium concentration, volume of solution, and mass of soil

In Chapter 4, to investigate the temperature dependence of sorption, a total of eight

Cd concentrations ranging from 1 to 50 mg L' were used to construct sorption isotherms at different temperatures The pH values measured at the end of the experiments ranged between 5.0 and 5.4 Preliminary kinetic experiments using an initial concentration of 2 mg L'! Cd at 21 °C showed that the apparent sorption equilibrium was attained in 24 hours, which has been conventionally adopted for batch equilibrium experiments (Ptacek and Gillham 1992) Thus, the tubes with soil suspension were shaken for 24 h at 10, 21 and 35 +0.5 °C in a climate chamber The distribution coefficient was obtained from linear isotherm

In Chapter 5, batch kinetics study was performed to probe into the relationship between competitive sorption and sorption kinetics Single and binary solutions consisted of 10 or 100 mg L” Cu and Cd were prepared The experiments were carried out at 10 °C (at which nonequilibrium condition is more prominent, as shown in Chapter 4) using solid concentrations of 100 g L and 20 5 LỶ (.e 2g:20mL, and 0.5g:25mL), in order to illustrate the effect of metal to soil ratios on competition The tubes with soil suspension were tumbled for a total of 14 reaction times ranging from

1 min to 7 days The centrifuged soil samples were then studied by sequential extraction (details in Section 3.4)

In Chapter 6, batch kinetics and sorption isotherm were studied so as to provide information about the relative importance of rate-limited sorption and nonlinear sorption at a range of metal concentrations Batch kinetics experiments, using four

Cd concentrations in the range of 1.124-112.4 mg L' (i.e 10°-107 M) were carried out using 14 reaction times ranging from 1 min to 7 days Batch sorption experiments were performed with a series of 16 Cd concentrations, resulting in equilibrium concentrations lying in the same concentration region (10° to 10° M) The equilibration time was 24 h, which was sufficient for reaching apparent equilibrium as determined from kinetic studies The plots of sorbed concentration against equilibrium concentration were fitted with linear and Freundlich (i.e nonlinear) isotherms Considering soils contain heterogeneous sorption sites, Freundlich isotherm is the most widely used isotherm function for many contaminants of interest (Selim and Amacher 1997; Hu and Brusseau 1998)

In Chapter 8, the extractable amounts (mmol kg”) of metals in the soils, which were used as input parameters for later modeling, were estimated by 3-day batch extraction experiments A 15 g of soil contaminated with Cu (low loading, high loading, and aged) was extracted by 0.5 L of 10” M EDTA solution or by 1 L of 10°

M EDTA solution These ratios approximately corresponded to the ratios of soil amount in columns to total flushing volume The extractable amount of Cu was largely decided by contamination conditions while those of Fe, Al, and Ca were virtually dependent on EDTA concentration

Column experiments, also known as miscible displacement experiments, are the most commonly used method to investigate solute transport in soils A summary of column experiments is presented in Table 3.2 A 20% of.total soil columns were randomly replicated The observed phenomenon was qualitatively the same and the overal] data variation was within 5% Thus, the transport data were reproducible

PVC columns of internal diameter 3.6 cm and length 10 cm were individually packed with the four soils in small increments After each increment, the soils were compacted with a piston of slightly smaller size than the column, in order to obtain uniform and homogeneous soil columns (Figure 3.1) O-rings were used to seal the end plates against the interior of the columns Porous stainless steel plates were attached at both ends of columns to promote a radial distribution of the influent solution and reduce hydrodynamic dispersion of the moving solution at or near the column exit Filter papers (0.45-um nominal pore size) were placed at the ends of the soil columns to ensure the effluent free of turbidity For UST & UST2 soils, the measured bulk soil density and accordingly calculated porosity were 1.508 g cm™ and 0.433, respectively These values were 1.352 g cm” and 0.495 for the TMS soil columns, and 1.242 g cm® and 0.532 for the CWB soil columns, respectively

The soil columns were slowly saturated (~2 cm h”) with upward flowing background electrolyte (0.03M NaNOs, pH 5 by dilute HNO;) from the bottom using peristaltic pumps (Figure 3.1) The water saturation was gravimetrically checked A stable effluent pH and ionic strength was maintained after flushing of about 100 pore volumes The effluent concentrations of Fe, Al, Ca and dissolved organic carbon were not detectable after initial equilibration period, indicating negligible dissolution of mineral oxides or soil organic matter during the course of column experiments

FIGURE 3.1 Soil column (top) and setup for a series of column experiments (bottom)

Influences of Pore-Water Velocity and Temperature

Nonequilibrium transport could severely stall off soil remediation process For instance, a recent model simulation demonstrated that the predicted cleanup time of the nonequilibrium model could be five-fold longer than what is expected under the ideal conditions assumed in equilibrium model (Opdyke and Loehr 2002) As reviewed in Sections 2.3.4 and 2.4.1, the factors that probably affect the residence time or the reaction time are crucial for controlling the degree of nonequilibrium transport Pore-water velocity and temperature are thus worth investigation Higher pore-water velocity is expected to reduce the residence time, whereas higher temperature is supposed to increase the reaction rate

In this chapter, the disparity of transport behavior of Cd at a range of pore-water velocity in the three soils of different clay contents (sandy loam, loam, and clay loam) was first evaluated (Section 4.3-4.5) The equilibrium transport model incorporating with effective dispersion coefficient and nonequilibrium transport model were applied to describe the breakthrough curves (BTCs) of column experiments Then, the temperature dependence of Cd sorption and transport behavior in soils was investigated (Section 4.6-4.8) The breakthrough curves (BTCs) were analyzed with the transport models to evaluate the temperature dependence of Cd transport behavior The Cd sorption at different temperatures was also determined in batch experiments

A non-sorptive bromide tracer was individually introduced to the soil columns to identify the appropriateness of the two-site nonequilibrium model and to determine the column hydraulic properties, such as Peclet numbers and dispersion coefficients This approach was similarly employed in the subsequent chapters Figure 4.1a shows the Br BTCs at the highest pore-water velocity; Figure 4.1b illustrates the Br BTCs at the lowest temperature The Br BTCs at other conditions were similar The retardation factors of the Br BTCs computed by first moment (Eq 3.1) were 0.97 — 1.09 for UST soil, 1.05 — 1.15 for TMS soil, and 1.06 — 1.18 for CWB soil, respectively The Br BTCs at all pore-water velocities were symmetrical in shape and well depicted by equilibrium model simulation (goodness-of-fit r’>0.98), and the nonequilibrium model produced the same prediction Therefore, an immobile water region did not exist even at the highest pore-water velocity, and the interpretation of the two-site nonequilibrium model rather than the two-region nonequilibrium model would be adequate to analyze the Cd BTCs (Sparks 1999) The values of Peclet number and dispersion coefficient are summarized with other parameters of transport models in later sections (Sections 4.3&4.6)

06 4 J SH oO lu `" 4 ` kề 2 ` toà

0.0 AK on — T T Pore Volume T T So

FIGURE 4.1 Breakthrough curves of Br in UST, TMS and CWB soils: (a) at 58.3—

82.2 cm h" and 21 °C; (b) at 6.56-7.16 cm h and 10 °C (G UST soil; © TMS soil;

A CWB soil; _ represents the corresponding simulations of equilibrium model)

'Í' R(đ-1)]/Ađ = ỉ Áq pateIn9[#O, :{1-)/(-Äg) = qÁq pereino[e2, :€ Ê-ÊÊ € sbe 3upsn Japoui ưnIqrimbeuou tyA pazundQ p 9918029 uos12dsIp 9AI22119 uỉ1q0 01 /J'Z2991Ê sbe 8ursn jepoui unuqr[inbe yua poztuundOQ > :SISÁTEUE 39UIOUI [#1OđU19) 151IJ p9Z1[Eu1ou Áq p1e[n2[Ê2) q -OđSutT) 12201) 1 2AId1OS-uou U101J p9u1E)QQ) „

IZÊ0 690 0/0 10/0 L'L6 Sb 012 91Z €8 QMO ZI 00I0 88/0 9970 6/0 6€ỳ ELE 6l S6l = 067 QMO Il 6110 /I80 P80 7280 bcÊ tỳ 9'0 LI6 — 681 đAO2 0I 900 Ê/80 IIE0 7980 I€8 vie IV's SIT /E9 đAA2 6 I0 6990 1650 990 0ẽ ey 8IÊ vol — 919 SAL 8 6810 80/0 0/90 91/0 90€ 66 ztI SZ = OVE SAL L 0600 O180 /0E0 SI80 66Ê LE ức 066 €02 SAL 9 0600 Lr60 6/00 060 869 6TÊ 6€ vel = EL SAL  9470 6450 Z6II ĐĐ€0 esl €8 cyc See TB isn ỳ I0 €690 t9II 8650 V6L Z'€8 0 Srl — 9% isn € 6/00 LL90 86/0 1890 C€b CLL LzI LŨ — tt 18a ể Œ00 /SL0 990 09/0 601 60L L€ SLI ETB isn I (Wo „I1 g9 od Wad wd ¿”3Œ ,dj (09A jog 3dxg uunHqijinbeuoN uunUqrimPba

S[IOS U† 10dSuE1, p2 101 S[2pOJN 0InJaqr[inbouojN 2)5-0A4,, pưẽ tưnHg![tnbg 001519511 9AI991ƒđ J0 S1912u1616đ 10dSuE1T, Jˆp ATAVL

4.3 Transport Behavior at Different Pore-Water Velocities

Figures 4.2 — 4.4 display the BTCs for Cd transport in UST, TMS, and CWB soils at four different pore-water velocities The equilibrium model simulation did not match with the BTCs in UST soil at the entire range of pore-water velocities, but it fairly described the BTCs in TMS and CWB soils at low pore-water velocity Apparently, increasing the pore-water velocities induced an earlier breakthrough and longer tailing of Cd transport in all the three soils, which were better captured by nonequilibrium model simulation Hence, the errors associated with equilibrium model simulation were magnified at higher pore-water velocities While the delineation of equilibrium model simulation appeared to be inaccurate, the curve fittings of Cd BTCs with the nonequilibrium model and the effective dispersion equilibrium model were excellent with a goodness-of-fit (r”) over 0.98 The values of the parameters in the nonequilibrium model and effective dispersion equilibrium model are summarized in Table 4.1

Although there was only a slight discrepancy between the simulations by the nonequilibrium model and the effective dispersion equilibrium model, the nonequilibrium transport model appeared to provide a more appropriate mechanistic interpretation The salient rises in effluent concentration after flow interruption corroborated the significance of nonequilibrium transport behavior (Brusseau et al

1997) The degree of perturbation was enlarged with increasing pore-water velocity (Figures 4.2a&d, 4.3a&d, 4.4a&d), suggesting an accordingly increasing extent of nonequilibrium It is probably because the shorter residence times at higher pore-water velocities restrain the likeliness of the attainment of equilibrium sorption

In addition, Figures 4.2 — 4.4 reveal different shapes and positions of BTCs, while the BTCs of different pore-water velocities should overlap if the transport is truly under an equilibrium condition (Eick et al 1990; Gerritse 1996) Instead, the effective dispersion equilibrium approach could superficially well describe the BTCs because the effective dispersion coefficient incorporated the asymmetry due to nonequilibrium transport behavior This approach however _ significantly | overestimated the dispersion coefficient, by comparing the values of D with those from Br tracer tests (Table 4.1), which were in line with previous studies (Goltz and Roberts 1986; Seuntjens et al 2001) The overestimation of dispersion coefficient ranged from 2.4- to 6.3-fold in UST soil, 1.3- to 4.4-fold in TMS soil, 1.6- to 3.6-fold in CWB soil, respectively The degree of overestimation was in accordance with the extent of nonequilibrium, which was larger with increasing pore-water velocity Such dispersion coefficients obtained from the effective dispersion equilibrium model was believed to be unrealistic Previous studies also found similar overestimation (Fesch et al 1998; Pang and Close 1999), which was believed to indicate the degree of nonideality of the BTCs (Sugita and Gillham 1995) Therefore, the nonequilibrium model is considered to be more fundamentally accurate to explain the Cd transport behavior in soils

It is important to note that the degree of the impact of pore-water velocity on Cd transport varied with different soils Of the three soils used in this study, UST soil exhibited the most substantial degree of nonequilibrium as well as the greatest extent of nonequilibrium enhancement in response to increasing pore-water velocity, while TMS and CWB soils exhibited less The degree of nonequilibrium Cd transport in the three soils exhibited a positive correlation with the corresponding retardation Previous findings also showed that the degree of nonequilibrium transport appeared to be larger for greater retardation of the solute (Brusseau et al 1991; Thorbjarnarson and Mackay 1994) The transport of Cd in UST soil manifested the most prominent retardation, while the transport in the other two soils was less retarded Thus, the impact of pore-water velocity and the corresponding nonequilibrium behavior appears to depend on the strength of interaction between Cd and the soil

The characteristics of the three soils probably play an important role in determining the Cd transport behavior The dominant clay mineral in the three soils was kaolinite, on which sorption would not be rate-limited by diffusion into the interlayer of 2:1 layer clay minerals, such as vermiculite (Jardine et al 1988; Eick et al 1990) The soil with the higher clay content (CWB>TMS>UST) possesses the higher external surface area (Table 3.1), but the retardation, on the contrary, was largest in UST soil and nearly the same in TMS and CWB soils (Table 4.1) Therefore, a relatively greater proportion of Cd sorption is postulated to be on the external surface of clay mineral in TMS and CWB soils than in UST soil, which is less easily subject to rate-limitation that brings about nonequilibrium transport behavior However, the difference in the characteristics of soil organic matter may also be critical and requires further investigation

Có (uoe[nus [ứpou1 uin1q1inbo 001512đSID 9AI)29112 -~'~' “Uu0IEInu1Is [2poui uunHdt[nbe -^~~^ uoip[nuIs [apou1 uinl1gijinbauou ——2 :#1Bp [81u9uIHSdX9e) ,ự 02 Z Z8 (P) :Ă.J 09 9'€Ê (9) :Ă 9 ÿˆÿÊ () ớ.0 09 €['8 (8) :JI0S LĐf1 0 P2 J0 seA1no YSnompyearg 7p AMMO 20A 210 2un[oA ạoqJ 08t 0b 00y 09C 0C 087 0Vế 002 091L 0L O08 0ÿ 0 00y 09 0C O82 OF 00 091L 02L 0Đ OO 0

2„TJOA 310đ nI0A 3101 00W 09V 0 O8% OFZ 00 09L 0l 09 OF 0 0t 08 0í 00 091 071 08 0t 0 1 hư 1 1 1 00 r 10 F0 r £0 Ft0 F90 r90 + LO t 80 t 60 + OT

(uonne†nuIs [opoui unH1[nbo uoISI2dSIP ứAI221J2 -~'~' ~ UOIIBIRUIS JapouI uunHqI[inbe v6 tUOoII8InuIS [2pou1 uInH1g1Inbauou ———— :#1ep J81ueuedxee) , 192 9'19 (p) :, 019 06c (9) :, 02 £'0 (Q) :, 012 ££'¿ (E) :[I0S SW,L, tỊ P2 JO S2A1n2 ysnompyeag Ep TUNA

: L0 : r SOG , E90 L0 L80 F60 peddos Aol,T 30A 3101 oor 08 09 OV 07 0 1 L L 1 L š * (uoneinus Japoui ưnHqi[inba uoJs12dSIp 2AI29JJ9 '~~' ~ *IoIqpịnuns [apoui uInHqr[inba E1Ep [8)u9UI19đdX@) , 02 €'§€ (P) :¡ 19 0 6 (9) :¡.ú 02 6'§[ (Q) :, 019 /£'9 (8) :[IOS {MD UT pO jo SoAino ysnompyeag py AND

1 1 L L L 00 F F0 [ ¿c0 t Ê0 t FO + $0 t 90 t LO t 80 ce +60 t OT tuOI†PJnu1s [ứpoui uunHqi[mnbauou ————

4.4 Sorption Rate Coefficient and Fraction of Instantaneous

Sorption at Different Pore-Water Velocities

The values of sorption rate coefficient (&) and fraction of instantaneous sorption (F) are related to the degree of nonequilibrium transport behavior in different soils In general, there were smaller values of œ and F for UST soil than for TMS and CWB soils (Table 4.1), reflecting the larger degree of nonequilibrium in UST soil Figure 4.5a displays that the sorption rate coefficient (&) is positively related with pore-water velocity in the three soils, consistent with the previous findings (Kookana et al 1993; Maraqa et al 1999; Maraqa 2001; Seuntjens et al 2001; Pang et al 2002) The value of & is not expected to vary with pore-water velocity if nonequilibrium transport arises from slow chemical sorption kinetics, which should be independent of residence time Therefore, the finding implies that the nonequilibrium transport of

Cd results from sorption being rate-limited by a physical process, such as diffusion into the soil matrix (Sparks 1999) The velocity dependence of the sorption rate coefficient is ascribed to the time-scale effect (Brusseau 1992), since the diffusivity of Cd and the diffusion path length into the soil matrix are not likely to vary with pore-water velocity It should be noted that the first-order kinetic equation in nonequilibrium transport model is apparent rather than elementary, and the sorption rate coefficient is a time-averaged value calculated over the residence time Thus, the increase of pore-water velocity, which reduces the residence time available for diffusion, increases the sorption rate coefficient

The fraction of instantaneous sorption (F) decreased with increasing pore-water velocity (Figure 4.5b) Kookana et al (1993) also found a negative relationship between the value of F and pore-water velocity using the two-site nonequilibrium model The dependence of the value of F indicates that there is a continuous distribution of sorption times for attaining equilibrium The sorption sites for Cd could probably be continuously distributed in different soil domains that are not equally effectively accessible at a given pore-water velocity, and thus, the access to the distributed sorption sites requires different diffusion times Increasing pore-water velocity provides shorter residence time for diffusion to the less readily accessible sorption sites, resulting in smaller fraction of sorption that can be regarded as instantaneous

4.5 Retardation of Transport at Different Pore-Water Velocities

EDTA-Flushing Effectiveness and EmpaCÉS . ceesss<ssssssse 141 Noi on

A large number of sites contaminated by heavy metals have been reported around the world, where copper (Cu) is one of the heavy metals most often encountered and its strong sorption (probably specific adsorption or surface precipitation) on soils pose an obstacle for effective cleanup (Sections 2.1.1, 2.2.1 and 2.2.5) From the previous results, rate-limited Cu sorption is little affected by competitive metal cations and the resultant nonequilibrium transport appears to be even more severe at high flushing velocities, low temperatures, and low contamination levels (Chapters 4-6) Chemical enhancement is, therefore, a convincing option to facilitate the rate-limited and/or resistant removal of Cu EDTA (ethylenediaminetetraacetic acid) is one of the most widely used chelating agents for chemical-enhanced soil washing/flushing (Section 2.4.4) However, a vast body of literature has focused on comparing remediation capabilities of different chemical agents (e.g NTA, oxalate, citrate) and strong acids (e.g HCl, HNO3, H2SO,4) EDTA was shown to have the cleanup power comparable to strong acids while supposedly induce less damage than strong acids Instead, the actual impacts brought by EDTA on soils have not been evaluated in detail over the course of EDTA-flushing

In this part of study, the soils (UST2 soil) were contaminated and remediated in acidic condition, since the homogeneous metal precipitates that would form at high pH should be more effectively dissolved by using strong acids than EDTA The acidic condition is most appropriate for an effective use of EDTA-flushing The soils loaded with two Cu concentration levels, fresh and aged contamination were flushed with EDTA solution of various concentrations (102 to 102 M) at two pore-water velocities The effectiveness of soil flushing with EDTA was assessed in terms of the

Cu extraction efficiency, pore volume of flushing, and EDTA consumption

Sequential extraction scheme was applied to understand the Cu distribution in soils and the fractions resistant to EDTA extraction The associated impacts were also monitored Soil dissolution of Fe/Al oxides, Ca carbonates and soil organic matter was assessed along with EDTA-flushing

7.2 Copper Distribution in Soils before EDTA Flushing

Based on sequential extraction results, the distributions of Cu in soils with low and high Cu loadings and aged contamination are summarized in Table 7.1 A slightly larger fraction of total sorption was observed to be strongly sorbed at low Cu loading, particularly in organic matter fraction, because heavy metals are preferentially sorbed by high-energy binding sites The Cu distributions in soils of fresh and aged contamination were quite different Aging of the contaminated soils resulted in the redistribution of sorbed Cu from exchangeable and carbonate fractions to oxide and organic matter fractions This redistribution of Cu from weakly to strongly bound fractions in soils agreed with the results of Han and Banin (1997) Several mechanisms may be accountable, including surface diffusion to high-energy binding sites, micropore diffusion into soil matrix, and formation of surface precipitates, of which confirmation requires in-situ spectroscopic techniques The effluent pH was lowered to values between 3.9 and 4.4 after aging, indicating a greater extent of specific sorption of Cu, probably via inner-sphere complexation with oxides and organic matter (Christl et al 2001; Scheinost et al 2001), which releases H* ions er

“sasayjuored u OIJtIAứp pIEpu01S uOIJSSẩIP 9A#AAOI2IJN

9) or" on er (% S70) ( 87'0) (% 610) '{%¿£) IOH Pue lenpisoy % 61 £ % 98'T % 002 '(%8y) đH '(%00) 'ONH

2c 68 “t € 'ấ*ONH (A/A) ( t0) (% 61) (% 08'0) % 0 tL HNfOOfHO Wẹ Z€_ 19IEUI 921u6810 0 punog % 6 Vĩ % OTT % vl z Hd) 'ONH + *O#H % 0€

(% £90) (% 88'0) (% 690) 1 Q22HO (ny) % ẤnuI[[#ls&1o % 6'91 % €P'6 % 8Œ L c0 [2H-HOfHN N00 “0149 %9PEO 9 0) pưnog

(% €t'0) (% y€'0) (% /00) 2s 0 “q6 9£uOq169 0) punog % 6 6 % L'EE % 9€ '€Hd“OOfH2£N NI — putpaxe|duio ÁJje2AA

(% y0) ( 101) (% 8€'T) "- ẤL 2qee8uvuoxa % Œ9€ % 0't %0 :¿ Hđ “I25N NW I

UOI16UIUIE1UuO~) poỞV 3uIpeo[ n2 qZ!H đuIp£o[ n2 MOT HOHIPROS Pue UOI2E1J [E]9[A uoInJOS 8un2011xq

S[I0S u uonnqI1s† JoddoZ pue swayss uonoenxg [enuonbas J'¿ W1AVL

During EDTA-flushing, the effluent pH ranged between 4.5 and 5.3 in freshly contaminated soil columns, while it was in the range of 3.9-4.4 initially and eventually raised to about 5 for aged soil columns EDTA extraction efficiency was suggested to be independent of pH so long as there is free EDTA remaining in solution (Bermond and Ghestem 2001; Tandy et al 2004), which was justified for 10” and 10” M EDTA-flushing (by comparing molar concentrations of metals and EDTA) in this study Figure 7.1 illustrates the cumulative removal of Cu from soil columns by EDTA-flushing As shown, the extraction effectiveness varied with EDTA concentrations In general, for all contamination conditions, there was a slight enhancement on Cu removal by 10” M EDTA compared with background solution alone It required an extensive flushing (>400 pore volumes, not shown) for both 10*

M EDTA and background solution to obtain a marginal increase in Cu removal Thus, flushing with 10* M EDTA solution was ineffective

FADER EOWK OK DORE 328V ROL ORR, PH Xt

90 4 en art 7 TH ae) kkk a +f

FIGURE 7.1 Cumulative removal of Cu using various EDTA concentrations from soils of: (a) low Cu loading; (b) high Cu loading; (c) aged contamination ( O 0.03

M NaNO;; @ 107M EDTA, 4 10° MEDTA; @ 107M EDTA; + 10? MEDTA with slow flushing; x 10? M EDTA with slow flushing)

The removal efficiency increased dramatically to above 85 % by using 10° and 107

M EDTA A gentler slope at later stage of flushing reflected that the extraction was less efficient with decreasing amount of remaining Cu in soils Excessive addition of EDTA to Cu in soils, i.e molar ratio greater than 1, was always needed for an effective removal (Table 7.2) This may be attributed to Cu extraction kinetics that limits complete utilization of free EDTA or competition of cations in soils that reduces available amount of EDTA for Cu Compared with 10° M EDTA flushing, 10° M EDTA required fewer pore volumes but a higher molar ratio of total addition of EDTA to Cu in soils to achieve the same extent of removal (Table 7.2) A larger volume of flushing solution of 10° M EDTA flushing would increase the post-treatment cost and prolong the treatment duration, while a larger excess of EDTA consumption in 10? M EDTA flushing would increase the chemical cost A further concern about their respective impacts on soils has to be taken into account

TABLE 7.2 Pore Volume, Molar Ratio of EDTA to Cu, and Residual Organic Carbon in Soils

Molar ratio of total Residual Organic

Pore Volume* addition of EDTA Carbon in soils to Cu in soils * (%y°

107 M (slow flushing) 11.9 2.17 0.98 ® Required for Cu removal of 85 %, except 10* M EDTA concentration;

> Measured at the end of experiments; initial soil organic carbon content was 1.47 %

The ultimate removal % was similar for soils of low and high Cu loadings, probably because the Cu distribution did not change significantly On the other hand, about

4-7 % less in the ultimate removal was found for aged contaminated soils (Figure 7.1c), which was in line with previous findings (Gao et al 2003) A larger amount of

Cu in soils was resistant to EDTA extraction probably due to the redistribution of Cu from weakly to strongly sorbed fractions upon aging (Table 7.1) Results of sequential extraction of aged contaminated soils after flushing (Figure 7.2) revealed that EDTA-flushing was able to extract Cu from all fractions, particularly effective by 107 and 10? M EDTA There were two probable reasons for EDTA extraction of strongly sorbed Cu First, complexation between Cu and EDTA appears to be thermodynamically favorable that overcomes some relatively less strong sorption interaction between Cu and the soils Second, the extraction is in part attributed to EDTA-enhanced dissolution of oxides and organic matter of the soils, which releases

Cu previously bound to oxide and organic matter fractions and breaks down soil aggregation structure that further releases occluded Cu from residual fraction Soil dissolution was discussed in detail subsequently

Exchangeable Carbonate Oxide Organic Matter Residual

FIGURE 7.2 Distribution of Cu in aged soils after flushing (results are the average of triplicates and error bars represent standard deviation)

The EDTA extraction was found to be kinetically limited at the employed flushing rate because a significant increase in effluent concentration was observed after flow interruption The extraction kinetics was dependent on EDTA concentration since the slope of removal pattern was steeper at higher EDTA concentration (Figure 7.1) A comparison of Figures 7.1a and 7.1b revealed a steeper slope for low Cu loading that less pore volumes of flushing was required (Table 7.2) The faster extraction kinetics probably resulted from a higher EDTA:Cu molar ratio in soils of low metal loading (i.e larger excess of EDTA and lower utilization of EDTA), which has been observed in soil washing (Tandy et al 2004; Di Palma and Ferrantelli 2005) Therefore, the Cu extraction kinetics is a function of EDTA concentration as well as metal loading

Based on the recognition of kinetic limitation of EDTA extraction, it is expected that the ultimate removal would increase with slow flushing Although the increase of ultimate removal was marginal (Figure 7.1c), less flushing volume and EDTA consumption were needed (Table 7.2) The extraction from oxide, organic matter and residual fractions were also greater with slow flushing (Figure 7.2), demonstrating a more complete Cu removal

These results suggest the prime importance of the concentration of EDTA solution in determining the effectiveness of soil flushing Unlike soil washing, in which the molar ratio between EDTA and metal in soils can be manipulated at a desired level by changing EDTA concentration or soil-to-solution ratio to give the same extraction efficiency (Kim et al 2003), the extraction kinetics and efficiency, which in turn determine the molar ratio of EDTA to Cu, of soil flushing are highly dependent on the concentration of EDTA

7.4 Iron, Aluminium, and Calcium Dissolution

Figures 7.3 — 7.5 illustrates the solubilization of Fe, Al, and Ca from soils along with flushing Negligible amount of these cations were solubilized by flushing with background solution, except that Ca was dissolved immediately along with flushing in aged contaminated soils (Figure 7.5c), which resulted from proton-promoted dissolution as pH was brought down to 3.9-4.4 by aging process Soil dissolution of

Fe, Al, and Ca was substantially increased in the presence of EDTA The emergence of dissolution was later than Cu extraction, showing that EDTA complexation with sorbed Cu was prior to cations in oxide and carbonate structure Among the soil cations, Fe was dissolved first, followed by Al, and Ca was the last (comparing Figures 7.3a&b&c, etc) This order of dissolution conformed to the magnitude of stability constants of metal-EDTA complexes (i.e log K, pe= 26.5 > log K, a;= 17.6

> log Ks,ca= 11.5, according to McBride (1994)) Thus, both metal lability in soils and thermodynamics of EDTA complexation with metals have to be considered to account for metal extraction and soil dissolution The cumulative dissolved amount of each cation reached levels of hundreds of mg kg", which was a significant amount compared with the levels of Cu contamination Besides, the dissolution curves did not level off till the end of the experiments, suggesting that solubilization of these cations from soils would continue with further flushing Thus, the importance of mineral dissolution during EDTA-flushing should not be disregarded

The kinetics of dissolution was dependent on EDTA concentration that a higher EDTA concentration led to a steeper slope and earlier emergence of dissolution The extent of dissolution also increased with the concentration of EDTA (Figures 7.3 — 7.5) All these support that the mineral dissolution is a result of EDTA-promoted dissolution, which can be described by a fast adsorption of free or complexed EDTA onto specific surface sites (via surface complexation) followed by a rate-limiting metal detachment from the oxide structure (Stumm 1992; Nowack and Sigg 1997; Sposito 2004) Fe and Al dissolution was time-dependent as confirmed by flow interruption, corroborating with previous results (Mayes et al 2000) Thus, slow flushing that allows longer reaction time enhanced soil dissolution, of which Al increased to a greater extent than Fe (Figures 7.5a&b), reflecting a faster overall kinetics of Al dissolution The greater soil dissolution appears to account for the more complete Cu extraction, especially from strongly bound fractions (Figures

7.1c&7.2) However, this may not be favorable as the soil structure and properties would be more severely affected

Transport of Metal-EDTA Complexes 160

From the results of Chapter 7, it has been shown that EDTA-flushing of contaminated soils enhanced Cu extraction but also induced mineral dissolution simultaneously, which would further dissolve soil organic matter and destabilize or disintegrate the soil structure Such mineral dissolution has been suggested to be delineated as a result of EDTA-promoted dissolution (i.e ligand-promoted dissolution), which is initiated by a fast adsorption of free or complexed EDTA onto specific surface sites via surface complexation and followed by a rate-limiting metal detachment from the soil structure (reviewed in Section 2.2.4) The transport of mobilized metal-EDTA complexes can induce metal re-adsorption and further mineral dissolution, or remobilization of sorbed metals of an extended soil region

To date, only few studies have attempted to describe the transport of metal-EDTA complexes with modeling efforts As briefly reviewed in Section 2.4.5, previously proposed models do not fit in simulating the respective transport of each metal resulting from extraction (Cu) and dissolution (Fe, Al, and Ca) during remediation of copper-contaminated soils (UST2 soil) by EDTA-flushing Therefore, this study extends the previous transport models to account for simultaneous EDTA-promoted extraction of the target metal and dissolution of soil minerals, which are probably rate-limited The model simulations were compared with experimental data The ultimate objective is to improve the understanding of the transport of metal-EDTA complexes (i.e Cu, Fe, Al, and Ca) during EDTA-flushing with appropriate modeling The experimental conditions were the same as in Chapter 7

In order to simulate the transport behavior of metal-EDTA complexes with minimum requirement of parameters, this model only takes into account the most influential chemical reactions (complexation) between metals and EDTA _ during EDTA-flushing They include the extraction of sorbed Cu and dissolution of Fe and

Al oxides and Ca carbonates, all are assumed to be solely promoted by free EDTA (H,2EDTA” is the dominant species across the pH range of this study) forming 1:1 metal-EDTA complexes This assumption was justified by preliminary experiments showing negligible Fe, Al, and Ca dissolution and obviously less significant Cu extraction by flushing in the absence of EDTA

>SOCu* + HzEDTA” — CuEDTA” + >SOH + H* (8.1) Fe(OH)3 + H2EDTA* + H* > FeEDTA’ + 3H>O (8.2)

For simplicity of modeling the transport behavior, the possible reactions (remobilization/dissolution) mediated by metal-EDTA complexes are assumed negligible compared with the reactions mediated by free EDTA (Eqs 8.1 — 8.4), which are considered the most kinetically significant reactions Previous studies demonstrated that rate coefficients for Fe-oxide dissolution kinetics were in the order of free EDTA > AIEDTA’, CaEDTA” > CuEDTA” (Nowack and Sigg 1997), and Al-oxide dissolution kinetics was much faster when mediated by free EDTA than CaEDTA” (Friedly et al 2002) Thus, the most kinetically significant reactions are those induced by free EDTA Besides, for step input of EDTA-flushing, the dominance of the reactions mediated by free EDTA (Eqs 8.1 — 8.4) can be justified by mass balance calculation The sum of molar concentrations of all cations (Cu, Fe,

Al, and Ca) in effluent was smaller than injected molar concentration of EDTA (10° or 107 M) all the time, which means free EDTA is always present in the flushing solution

In view of previous observation of a slow metal extraction kinetics following an initial fast release in batch studies (Bermond and Ghestem 2001; Wasay et al 2001;

Tandy et al 2004) and the reported time scale of EDTA-promoted dissolution (Nowack and Sigg 1997; Sposito 2004), Cu extraction and Fe/Al/Ca dissolution were described with kinetic terms, in turn incorporated into advection-dispersion equations (ADEs) This model extends the proposed framework of Kedziorek et al (1998)

For the transport of free EDTA, several kinetic sink terms were included in the ADE to account for the formation of metal-EDTA complexes of Cu, Fe, Al, and Ca, respectively aC, =D 9?C E_y aC, E + aC M ,Cu + dC ME dC M,AI + dC M,Ca 8.5 or dz? dz or ot ot or &3) where v is pore-water velocity (cm h); D ¡s hydrodynamic dispersion coefficient (cm? h”); Cg is the free EDTA concentration in solution (mmol L"); Cm,j G = Cu, Fe,

Al, Ca, respectively) are the extractable concentrations of metals in soils (mmol LẺ; and dCy, //ot are the kinetic sink terms resulting from reactions 8.1 — 8.4 for free EDTA transport

To simulate the transport of metal-EDTA complexes, respective kinetic source term was included in the ADE of each metal to account for the formation of metal-EDTA complexes

R,—*! = p—_™ i 13 azz SM/_— a 9 "Mi (=(Cu, Fe, Al, Ca ) ( 8.6-8.9 ) where Cem, ; are the concentrations of complexed metals in solution (mmol LỆ; R; are the apparent retardation factors (dimensionless) for each metal-EDTA complex; and dCw, ;/ot are the kinetic source terms for respective metal-EDTA complex transport resulting from reactions 8.1 — 8.4

The kinetic sink/source terms (0Cw, ;/ot), which represent Eqs 8.1 — 8.4 to account for the formation of metal-EDTA complexes in the ADEs (Egs 8.5 — 8.9), can be described with second-order kinetics (i.e first-order with respect to free EDTA concentration in solution and first-order with respect to extractable concentration of metals in soils) In addition to the consideration of competitive EDTA-promoted dissolution of Fe, Al, and Ca, the formulation differs from previous one (Kedziorek et al 1998) in a way that the extractable level of metals (mmol L"), instead of the extractable fraction of metals, is taken into account dCy ; ot =-k,C,Cy,,; G = Cu, Fe, Al, Ca, respectively) (8.10—§.13) where k; (L mol s?) are the kinetic rate coefficients of EDTA-promoted extraction (i.e Cu) or dissolution (i.e Fe, Al, and Ca) Despite involvement of hydrogen ion concentration in Eqs 8.1 — 8.4, it is assumed relatively constant and incorporated in the rate coefficients (Friedly et al 2002)

The simulations of transport of metal-EDTA complexes were performed using a finite difference method (Appendix C) to solve Eqs 8.5 ~ 8.9 (advection and sink/source terms were solved explicitly with upstream weighting; dispersion term was solved implicitly), with appropriate initial and boundary conditions Initial values of Cm, ; G = Cu, Fe, Al, Ca) were calculated by dividing the product of batch-measured extractable amounts (mmol kg!) and bulk density (kg L") by column porosity (dimensionless) The kinetic rate coefficients kj were adjusted to fit the experimental data of each metal-EDTA complex transport

For reactions 8.1 — 8.4 initiated by free EDTA, known as EDTA-promoted extraction and dissolution, the kinetic rate law is believed to represent the rate-limiting detachment of metal-EDTA complex from the surface, which depends on the concentration of the prior activated surface complex formation (Nowack and Sigg 1997; Sposito 2004) In addition to the above second-order formulations (Eqs 8.10 — 8.13), the kinetic expression may be considered as pseudo-first-order with respect to the concentration of EDTA, as employed by Friedly et al (2002), which implicitly assumed that the amounts of extractable metals in soils are comparatively in excess Since XRF results showed relatively abundant amounts of Fe, Al, and Ca compared with Cu in soils, the use of first-order kinetic terms for Fe, Al, and Ca dissolution was also analyzed

The hydrodynamic dispersion coefficient was determined by modeling the Br transport and assumed to be constant irrespective of metal-EDTA complexes, because preliminary modeling efforts revealed the simulation is relatively insensitive to dispersion coefficient On the other hand, knowing that the transport behaviors and elution time of different heavy metals are probably different, as observed in previous studies (Kedziorek and Bourg 2000; Sun et al 2001), an apparent retardation factor is assigned to each metal-EDTA complex so as to describe the corresponding retarded breakthrough The values of R were empirically determined by dividing the pore volume of respective heavy metal peaks by that of Br peak Preliminary simulations without apparent retardation (i.e R=1 for all complexes) failed to match with experimental BTCs

8.3 Simulation of Transport of Metal-EDTA Complexes

EDTA-flushing significantly enhanced the Cu extraction while induced simultaneous transport of metal-EDTA complexes of Fe, Al and Ca from soils under different contamination and operation conditions Figure 8.1 illustrates the observed and simulated transport of Cu alone (i.e Eqs 8.1, 8.5, 8.6, 8.10 only) The Cu peaks emerged at about 2-3 pore volumes, closely together with the EDTA and bromide peaks, indicating that Cu-EDTA complex was not obviously retarded and thus the apparent retardation factor for Cu transport was set equal to unity for all simulations

An increase of effluent concentration upon 16-h flow interruption confirmed that Cu extraction was time-dependent, and therefore justified the use of kinetic expression rather than equilibrium in modeling

The extractable amount of Cu was exhausted with continuous EDTA-flushing and the extraction efficiency decreased after the peaks Flushing with lower EDTA concentration resulted in smaller peak concentration and larger pore volume of flushing, indicative of the dependence of extraction kinetics on EDTA concentration The Cu BTCs of slow flushing resembled that of normal flushing (Figure 8.1c), reflecting similar overall kinetics of EDTA-promoted Cu extraction at the studied range of pore-water velocity Thus, slow flushing that allows a longer residence time for the rate-limited extraction was not as effective as expected The fitted values of the second-order kinetic rate coefficients were between 9.0x10° and 5.1x107 L moI! s' (Table 8.1), which may have varied with different contamination conditions but compelling inferences cannot be drawn from modeling results alone Compared with previously reported first-order rate coefficients (on the order of 10° s"') (Kedziorek et al 1998; Thayalakumaran et al 2003), the values calculated by multiplying the second-order kinetic rate coefficients by metal concentrations in soils of this study were larger (on the order of 10“— 10° s’, Table 8.3) This is probably because of the relatively high metal loadings (4.989-15.784 mmol kg”) with predominant fractions of weakly sorbed Cu in soils

The simultaneous transport of metal-EDTA complexes is illustrated in Figures 8.2&8.3 Perturbation of Fe and Al concentrations after flow interruption (Figures 8.2b, c, d), in line with previous findings (Jardine and Taylor 1995; Mayes et al

Conclusions and Recommendations 181

The most important findings of this study are recapped below

The symmetrical Br breakthrough curves indicated the absence of immobile water region at the applied pore-water velocities The equilibrium model could not describe the breakthrough curves of Cd, particularly at high pore-water velocities The shorter residence times at higher pore-water velocities resulted in a more significant early breakthrough and long tailing, reflecting a higher degree of nonequilibrium Although both the two-site nonequilibrium model and the effective dispersion equilibrium model could describe the transport behavior of Cd, the latter appeared to be inappropriate because it overestimated the dispersion coefficients as high as 6.3-fold compared with the coefficients obtained from the Br breakthrough curves Thus, the nonequilibrium model was used to analyze the results With increasing pore-water velocity, the sorption rate coefficient increased suggesting that diffusion was responsible for the rate-limited sorption, and the fraction of instantaneous sorption decreased inferring a range of diffusion times to the sorption sites The retardation also increased, probably because the soils were exposed to a greater mass of Cd at higher pore-water velocities over the same time span

Nonequilibrium transport behavior was observed in breakthrough curves at

10 and 21 °C, while the transport behavior at 35 °C was in general agreement with the prediction of the equilibrium transport model The durations of tailing were also reduced with increasing temperatures, indicating a lesser extent of nonequilibrium The shift from nonequilibrium to equilibrium transport behavior with increasing temperature was probably