Long-term performance evaluation of zero-valent iron amended permeable reactive barriers for groundwater remediation – A mechanistic approach

Abstract

Permeable reactive barriers (PRBs) are used for groundwater remediation at contaminated sites worldwide. This technology has been efficient at appropriate sites for treating organic and inorganic contaminants using zero-valent iron (ZVI) as a reductant and as a reactive material. Continued development of the technology over the years suggests that a robust understanding of PRB performance and the mechanisms involved is still lacking. Conflicting information in the scientific literature downplays the critical role of ZVI corrosion in the remediation of various organic and inorganic pollutants. Additionally, there is a lack of information on how different mechanisms act in tandem to affect ZVI-groundwater systems through time. In this review paper, we describe the underlying mechanisms of PRB performance and remove isolated misconceptions. We discuss the primary mechanisms of ZVI transformation and aging in PRBs and the role of iron corrosion products. We review numerous sites to reinforce our understanding of the interactions between groundwater contaminants and ZVI and the authigenic minerals that form within PRBs. Our findings show that ZVI corrosion products and mineral precipitates play critical roles in the long-term performance of PRBs by influencing the reactivity of ZVI. Pore occlusion by mineral precipitates occurs at the influent side of PRBs and is enhanced by dissolved oxygen and groundwater rich in dissolved solids and high alkalinity, which negatively impacts hydraulic conductivity, allowing contaminants to potentially bypass the treatment zone. Further development of site characterization tools and models is needed to support effective PRB designs for groundwater remediation.

1. Introduction

Various industrial and manufacturing processes involve the application, transport, and storage of hazardous chemicals, including heavy metals, pesticides, fumigants, and volatile organic compounds (VOCs). Unintentional and unregulated discharge of these chemicals results in their release to the environment. These chemicals leach and diffuse away from the entry point, contaminating deeper soil horizons and eventually the underlying aquifer. Effective treatment of contaminated soil and groundwater requires access to the contaminants and the availability of an appropriate remediation technology. According to the Federal Remediation Technologies Roundtable screening matrix (FRTR, 2022), in the United States, various full-scale commercially available, in-situ physical/chemical treatment technologies exist that are effective against halogenated and non-halogenated volatile organic compounds, fuels, inorganics, radionuclides, munitions, and emerging contaminants. These technologies include in-situ chemical oxidation/reduction, soil vapor extraction, thermal treatment, air sparging, permeable reactive barriers (PRBs), monitored natural attenuation and enhanced monitored attenuation, in-situ activated carbon, large-diameter auger mixing, and amendment injection (ITRC, 1998).

Performance evaluations of these technologies reveal various limitations. For example, some ex-situ remediation technologies, such as pump-and-treat, have high operational costs, high energy consumption, and restricted applicability (McGuire et al., 2006). Other in-situ technologies such as enhanced bioremediation, chemical oxidation/reduction, activated carbon injection, thermal treatment, and in-situ solidification often result in additional waste streams, incomplete contaminant degradation, and/or end up displacing contaminants away from the treatment zone (Henry et al., 2002; Brooks et al., 2004). Microbial degradation is slow due to metabolic activities, can be limited by the availability of carbon substrate, and often does not fully treat contaminants of concern, thereby allowing metabolites to persist (ITRC, 1998; Wiedemeier et al., 1998; Lapat-Polasko et al., 2001; Henry et al., 2002; Popovich et al., 2018). Nevertheless, microbial degradation is a mechanism of monitored natural attenuation at some sites with long histories of active remediation where the extent of contamination is fully delineated, the plume has stabilized, and years of monitoring show concentration levels that do not warrant active remediation (Henry et al., 2002).

PRBs can be a preferred remedial approach for the long-term treatment of a broad range of contaminants (ITRC, 2005). The PRB technology is an in-situ remediation approach employed for contaminated groundwater. The installation of a PRB begins with the excavation and removal of soil that is replaced by a permeable, contaminant-reactive material. The treatment zone intercepts contaminated groundwater flowing under the natural hydraulic gradient and reactions occur that degrade or arrest the transport of contaminants, then treated water flows out the other side (Puls et al., 1999; Obiri-Nyarko et al., 2014). Based on site conditions, such as groundwater chemistry and hydrological characteristics, the expected performance of PRBs can be predicted. Site-specific data can be used to conduct life-cycle analysis of PRBs (Henderson and Demond, 2007). Under ideal conditions such as pH and redox conditions within the PRB, sufficient residence time coupled with low total dissolved solids and negligible oxygen in the influent groundwater can lead to sustained PRB performance that surpasses the lifetime of contaminant plumes (Wilkin et al., 2014).

In 1991, the first field application of the PRB technology was demonstrated at the Canadian Forces Base in Borden, Ontario, where a permeable wall of granular zero-valent iron (ZVI) was entrenched to treat chlorinated organics (Gillham and O’Hannesin, 2009). In 2002, the United States Environmental Protection Agency (EPA) adopted PRBs as a standard remediation technology for the long-term treatment of polluted groundwater. Since then, PRBs have been used at over 90 sites in the United States and over 200 sites worldwide, of which 120 PRBs are iron-based (ITRC, 2005; Henderson and Demond, 2007; Chen et al., 2011; Fu et al., 2014). Most of these PRBs were installed for groundwater remediation of chlorinated hydrocarbons (Weber et al., 2013). The PRBs are generally effective; however, challenging groundwater conditions and resulting in-situ changes have led to their varied performance and shorter than expected service life in some cases (Henderson and Demond, 2007).

This review is focused on PRBs that utilized ZVI as the primary reactive medium. Other reviews (e.g., Gupta and Fox, 1999; Li et al., 2005; Thiruvenkatachari et al., 2008; Faisal et al., 2017) present some of the factors that influence the performance of PRBs; however, these studies were limited to reports of relatively young PRBs with less than ten years of service life. Lacking is a comprehensive review of authigenic mineral formation and transformation observed in the subsurface associated with PRBs, hydrological, and original design factors that contributed to successful long-term remediation. Here we present a long-term evaluation of PRBs that employed ZVI as a reactive medium. We report mechanisms associated with their performance by reviewing over 20 years of field reports and evaluating performance related to biogeochemical processes occurring in the subsurface. Finally, we report on the water chemistry, design, and site characterization factors contributing to successful remediation in ZVI-PRB systems to make improved best management recommendations for future use of this technology.

2. Groundwater contaminants and their interactions with ZVI in PRBs

2.1. ZVI amended PRBs

Numerous reactive media and their mixtures (Al, Zn, limestone, zeolite, pyrite, Sn, Mg, steel wool, Pd/C, Ni/Fe, Cu, brass, sulfur, sulfide, vermiculite, corrin, biotite, Pd/Zn, Pd/Fe, organic substrate, and anaerobic bacteria) have been investigated for specific contaminants for their use in PRBs. Nevertheless, iron (also referred to as Fe⁰, Fe(0), or ZVI) remains the preferred reactive medium used in the majority of the field-scale and laboratory investigations (Powell et al., 1998). Current research often emphasizes nano-scale ZVI (nZVI) in benchtop studies for contaminant removal; however, practical field applications of nZVI remain limited. The selection of reactive materials employed in PRBs is based on the geochemical reaction pathway, which renders the dissolved groundwater contaminants less harmful and eventually degrades and/or stabilizes them (Blowes et al., 2000). The geochemical reactions often include a redox couple that targets the solubility and molecular structure of the contaminant. Redox coupling can be a natural or engineered process in which an electron donor and acceptor are paired, such that electron transfer occurs between reactants, yielding change in oxidation states (Thiruvenkatachari et al., 2008). Since many contaminants are redox-sensitive, this approach can effectively overcome many challenges associated with traditional remediation techniques.

A reducing agent such as ZVI can facilitate rapid treatment of groundwater contaminants by reducing solubility and forcing mineral precipitation, as in the case of chromium, or even reductive dechlorination of chlorinated hydrocarbons (Henderson and Demond, 2007). Dehalogenation is a surface-mediated process that occurs through the electron transfer from iron to the chlorinated hydrocarbon at the iron surface (Powell et al., 1998). ZVI has often been employed as a reductant for the removal of chlorinated organics (Wilkin et al., 2014) and many of the oxyanion contaminants, including arsenic, chromium, lead, selenium (Blowes et al., 2000; Morrison et al., 2003; Wilkin et al., 2005), pesticides (Sayles et al., 1997), and explosive residues (Serrano-González et al., 2018) in the treatment of contaminated soil and groundwater. For example, the reduction of Cr(VI) by ZVI (Eq. (1)) and its subsequent in-situ precipitation as Cr(III) oxyhydroxide or chromium-iron hydroxide (Eq. (2)) occurs through the following reaction sequence (Powell et al., 1998; Blowes et al., 2000):

$${\text{CrO}}_{4\text{(aq)}}^{2-}+{\text{Fe}}_{(\text{s})}^0+8{\text{H}}_3{\text{O}}_{(\text{aq})}^{+}\to {\text{Fe}}_{(\text{aq})}^{3+}+{\text{Cr}}_{(\text{aq})}^{3+}+12{\text{H}}_2{\text{O}}_{(\text{l})}$$ (1)

$$\begin{array}{l}(1-x){\text{Fe}}_{(\text{aq})}^{3+}+(x){\text{Cr}}_{(\text{aq})}^{3+}+5{\text{H}}_2{\text{O}}_{(\text{l})}\\ \to {\text{Fe}}_{(1-x)}{\text{Cr}}_{\text{x}}{\text{OOH}}_{(\text{s})}+3{\text{H}}_3{\text{O}}_{(\text{aq})}^{+}\end{array}$$ (2)

While the above reactions show that chromium is removed in the PRB as Cr(III), field studies show that the solids in which these precipitates are incorporated may not necessarily be iron (oxy)hydroxides (Wilkin et al., 2005). Besides ZVI, other reactive minerals such as iron sulfides and ferrous iron-bearing oxyhydroxides also reduce Cr(VI). Even Fe²⁺ released during the corrosion of ZVI contributes to reduction reactions (Feng et al., 2015). The oxidation of iron applies a strong electromotive force capable of reducing many contaminants, and the rust formed does not adversely affect water quality (Fu et al., 2014). For example, it has been shown that reductive precipitation of ${\text{UO}}_2^{2+}$, ${\text{MoO}}_4^{2-}$, ${\text{TcO}}_4^{-}$, and ${\text{CrO}}_4^{2-}$ increased between the couple redox half-reaction of interest and the Fe⁰/Fe²⁺ couple (Cantrell et al., 1995). Hence, ZVI has become a preferred material for treating contaminated soil and groundwater, where contaminant reduction is the remedial pathway (Noubactep, 2010). Indeed, oxyhydroxide surfaces of the residual iron serve as sinks for arsenate, phosphate, chromate, and many other contaminants in their oxyanionic forms by chemisorbing them and removing them from groundwater, providing long-term increased benefit from the initial application of ZVI (Zachara et al., 1987; Furukawa et al., 2002).

2.2. Nature of groundwater contaminants and their interaction with ZVI

Some pollutants are naturally present in soil and geologic materials and can also be introduced to groundwater from mining and ore refinement. Others are manufactured and enter soil and groundwater through various anthropogenic activities, including waste disposal, commercial operations, manufacturing, and agriculture. Because of the different chemical characteristics of groundwater pollutants, their removal mechanisms in PRBs differ significantly. Groundwater contaminants such as chlorinated organics are removed in PRBs through the coupling of oxidation of ZVI and reductive dechlorination. Here, the reductive dechlorination of trichloroethylene is presented following the reaction (Powell et al., 1998):

$${\text{Fe}}_{(\text{s})}^0+{\text{CHCl}}_{3(\text{aq})}+{\text{H}}_3{\text{O}}_{(\text{aq})}^{+}{\text{Fe}}_{(\text{aq})}^{2+}+{\text{CH}}_2{\text{Cl}}_{2(\text{aq})}+{\text{Cl}}_{(\text{aq})}^{-}+{\text{H}}_2{\text{O}}_{(\text{l})}$$ (3)

A brief list of some of the redox-susceptible toxic pollutants treated by electrochemical couples and effectively mitigated by ZVI-PRBs is presented in Table 1. The preponderance and diversity of contaminants demonstrate how natural and anthropogenic activities affect the subsurface environment. Ingestion of contaminated groundwater may pose a substantial health risk as evidenced by extremely low Maximum Contaminant Level (MCL) values, justifying the need for contaminant removal and remediation.

Table 1.

Selected redox sensitive contaminants, their Maximum Contaminant Level (MCL), and removal mechanisms by ZVI.

Class	Use	Contaminant	MCL (mg/L)	Removal Mechanism
Metals	Mining, steel, electronics, and many other uses	Cadmium	0.005	Reduction, Co-precipitation,
		Chromium	0.1	Adsorption
		Copper	1.3†
		Lead	0.015†
		Mercury	0.002
		Molybdenum	*
		Nickel	*
		Zinc	*
Non-metals	Agriculture	Nitrate	10	Reduction
	Agriculture	Nitrite	1	Biotic removal
	Smelting	Selenium	0.05	Reduction, Co-precipitation,
Metalloid	Fire retardant	Antimony	0.006	Adsorption
	Various	Arsenic	0.01
Organic compounds	Explosives	Trinitrotoluene (2,4,6-)	*	Reduction
		Hexahydro-1,3,5-trinitro-1,3,5-triazine	*	Adsorption
	Pesticides	Atrazine	0.003	Reduction
		Dalapon	0.2
		2,4-dichlorophenoxyacetic acid	0.07
		Dichlorobenzene p-	0.075
	Degreaser	Dichloromethane	0.005
	Dry cleaning	Tetrachloroethylene	0.005
	Solvent	Trichloroethylene	0.005
	Various	Monochlorobenzenes	0.1
	Power	Polychlorinated biphenyls (PCBs)	0.0005
Radionuclides	Power & weapon	Technetium	*	Reduction, Co-precipitation
	Power & weapon	Uranium	0.03

^*No MCL provided by US EPA; health advisories exist with reference doses and drinking water equivalencies in the part per billion per day intake level in the 2018 Edition of the Drinking Water Standards and Health Advisories Tab.

^†level, regulated by treatment technique. If more than 10% of samples exceed the action level, then the utility will take additional steps.

3. Mechanisms of ZVI interactions with groundwater and implications

3.1. Reactions and transformations of iron in the PRB

The corrosion of ZVI drives reduction reactions that mitigate various contaminants; however, interactions with solutes in groundwater influence these reactions and affect the performance and reliability of the PRB in remediating pollutants (Zachara et al., 1987; Matheson and Tratnyek, 1994; Gillham and O’Hannesin, 2009). Table 2 presents several pathways of iron and standard reduction potentials for families of contaminants based on specific treatment pathways, which render these pollutants to less harmful forms. Irrespective of the types of contaminants – i.e., chlorinated organics or heavy metals, the interaction of these contaminants with ZVI depends on groundwater chemistry and reactions that occur between the reactive surfaces within the PRB and the contaminants of concern (Henderson and Demond, 2007). Standard reduction potentials presented in Table 2 reveal that reduction of these contaminants is thermodynamically favorable and driven by the oxidation of zero and divalent iron to the trivalent, ferric form. Reactions between iron and contaminants are not exclusive to free cations and dissolved iron species as various ferrous iron-containing minerals like magnetite, green rusts, sulfides, and other minerals are also redox-active and participate in redox transformations of contaminants.

Table 2.

Corrosion of iron and redox reactions of contaminants. E⁰ (V) is the standard reduction potential of the associated reaction in volts.

		Oxidant	Reductant	E0 (V)	References
Iron corrosion		Fe3+ + 3e−	Fe0	−0.036	Haynes et al., 2015
		Fe2+ + 2e−	Fe0	−0.447
		Fe3O4 + 8H+ + 8e−	3 Fe0 + 4H2O	−0.085	Wang et al., 2013
		3Fe2O3 + 2H+ + 2e−	2Fe3O4 + H2O	0.220
		Fe3+ + e−	Fe2+	0.770	Haynes et al., 2015
		2Fe2+ + 2H2O	2Fe0 + O2 + 4H+	−1.676
		Fe2+ + H2 + 2OH−	Fe0 + 2H2O	0.381
Contaminant	Nitrate	${\text{NO}}_3^{-}+2{\text{H}}^{+}+e^{-}$	NO2 + H2O	0.800	Haynes et al., 2015
		${\text{NO}}_3^{-}+4{\text{H}}^{+}+3e^{-}$	NO + 2H2O	0.958
		${\text{NO}}_3^{-}+3{\text{H}}^{+}+2e^{-}$	HNO2 + H2O	0.750
	Arsenic	${\text{AsO}}_4^{3-}+2{\text{H}}_2\text{O}+2e^{-}$	${\text{AsO}}_2^{-}+4{\text{OH}}^{-}$	−0.710
		H3ASO4 + 2H+ + 2e−	HAsO2 + 2H2O	0.560
		As2O3 + 6H+ + 6e−	2As0 + 3H2O	0.240
	Cadmium	Cd2+ + 2e−	Cd0	−0.430
	Chromium	${\text{CrO}}_4^{2-}+4{\text{H}}_2\text{O}+3e^{-}$	Cr(OH)3 + 5OH−	−0.120
	Copper	Cu+ + e−	Cu0	0.520
		Cu2+ + e−	Cu+	0.160
	Manganese	MnO2 + 4H+ + 2e−	Mn2+ + 2H2O	1.210
	Mercury	HgO + H2O + 2e−	Hg0 + 2OH−	0.098
		${\text{Hg}}_2^{2+}+2e^{-}$	2Hg0	0.800
		Hg2+ + 2e−	Hg0	0.850
		2Hg2+ + 2e−	${\text{Hg}}_2^{2+}$	0.910
	Molybdenum	H2MoO4 + 2H+ + 2e−	MoO2 + 2H2O	0.650
	Nickel	Ni2+ + 2e−	Ni0	−0.230	Haynes et al., 2015
	Selenium	${\text{SeO}}_3^{2-}+3{\text{H}}_2\text{O}+4e^{-}$	Se0 + 6OH−	−0.350
	Technetium	${\text{TcO}}_4^{-}+4{\text{H}}^{+}+3e^{-}$	TcO2 + 2H2O	0.740
	Uranium	${\text{UO}}_2^{+2}+e^{-}$	${\text{UO}}_2^{+}$	0.062
	Zinc	Zn2+ + 2e−	Zn0	−0.760
	Nitroaromatics	Ar-NO2 + e−	$\text{Ar}-{\text{NO}}_2^{-}$	−0.250 < E0<−0.590	Salter-Blanc et al., 2015; Uchimiya et al., 2010
	Chlorinated hydrocarbons	R-Cl	R’ + Cl−	0.380 < E0<0.630	Arnold et al., 1999; Roberts et al., 1996

The reaction between contaminants and ZVI is a surface-mediated process that relies on electron transport from anodic sites on the reduced iron surface to the cathode that receives the electron. Hence, the reaction between contaminants and ZVI occurs at the interface between the solid ZVI particle and waterborne contaminant that sorbs to its surface (Yong and Schoonen, 2000). If a redox reaction is thermodynamically favorable (E⁰ > −0.44 V), electron transfer occurs; however, the rate of transfer is regulated by the electrical conductivity of ZVI and contaminant concentration (Lawrinenko et al., 2017a). Groundwater contaminant concentrations vary over time and throughout the season as transport is influenced by solubility, rainfall events, depletion, and dilution through the aquifer. The electrical conductivity of ZVI complements contaminant concentrations in reaction kinetics which is influenced by the chemical environment of iron in ZVI, composition, crystallinity, and the mineral assemblages that form with the oxidation of iron and how these influence electron transport from the corroding ZVI particle to its exterior surface.

In natural systems, iron corrosion yields a heterogeneous mixture of iron minerals that form at the corroding surface of a ZVI particle. Additionally, dissolved Fe⁺³, Fe⁺², their hydrated species, and the complexes they form with other groundwater anions may associate with some contaminants and other dissolved cations and anions to form various complex minerals through adsorption and co-precipitation (Wilkin et al., 2005; Beaulieu and Ramirez, 2013; Kumar et al., 2015). The various corrosion products that form with the aging of iron in PRBs influence contaminant transport and play a significant role in long-term remediation, particularly with the chemisorption and co-precipitation of many heavy metals, non-metals, and metalloid oxyanions (Vodyanitskii, 2010). Our purpose in relating corrosion and biogeochemical processes in this manuscript is to highlight to the reader some of the significant transformations of iron that occur in the PRB to present a general understanding of these processes and emphasize the biological, chemical, and physical processes that collectively drive the changes occurring in PRBs over time. Selective reviews of these processes are also provided by Thiruvenkatachari et al. (2008), Vodyanitskii (2010), and references therein.

3.2. Corrosion of ZVI and contaminant reduction – Mechanisms and misconceptions

The oxidation of ZVI to ferrous iron provides the driving force to reduce many redox-sensitive contaminants (Fe⁰ → Fe²⁺ + 2e⁻, 0.44 E⁰(V); Table 2). This corrosion step is responsible for the strongly negative oxidation–reduction potential (ORP) measurements commonly reported in PRBs throughout the literature, including EPA and Interstate Technology and Regulatory Council (ITRC) reports. Conflicting opinions have emerged in the literature, which question the role of the two-electron transfer process in contaminant remediation. For example, recent reports by Noubactep (2014, 2015, 2016) appear to discount the strong reducing force of ZVI oxidation in remediation; instead, these authors attributed contaminant removal to co-precipitation, size exclusion, and adsorption involving authigenic ferrous minerals (Noubactep, 2015). Interestingly, in 2009, the author acknowledged the role of ZVI corrosion and the necessity for electron transfer from its exterior passivated surface to the contaminant (Noubactep, 2009). However, recent publications demonstrate an incomplete understanding of the fundamental mode of action of ZVI used for in-situ remediation at contaminated sites (Nadège and Noubactep, 2013; Noubactep, 2014, 2015, 2016). While it is true that co-precipitation, chemisorption, electrostatic physisorption, and size exclusion are all processes that occur, the extent to which these processes play a role in groundwater remediation depends on the interactions between the contaminant and iron species, which is highly pH-dependent. The seminal work of Cutler (1987), Gillham and O’Hannesin (1994), Matheson and Tratnyek (1994) provides evidence for direct electron transfer and the corrosion of ZVI in contact with chlorinated hydrocarbons; particularly the reductive dechlorination that occurs in steel sampling vessels and steel pipe lacking moisture or oxygen. By ignoring the role of electron transfer from the ZVI matrix within a particle to the contaminant, the arguments of Noubactep and colleagues appear to focus on the role of electron transfer from ferrous species and neglect the fact that the low Eh, production of dihydrogen, and oxidation of ZVI suggest that the corrosion of ZVI is what continues to drive reduction reactions for many years after application. We seek to correct this misconception and better inform the reader by discussing ZVI corrosion and the biogeochemical processes that occur as ZVI ages in groundwater.

Corrosion and oxidation of ZVI form Fe⁺² and Fe⁺³, yielding polymeric hydroxides and oxyhydroxides with Fe⁺³ and Fe(OH)₂ with Fe⁺² (O’Grady, 1980; Kamimura et al., 2002). Ferrous iron species are soluble in water and diffuse away from ZVI particles; a portion of the Fe⁺² is electrostatically adsorbed to the surface as both positive and negative pH-dependent charge sites are present on various iron minerals (Kosmulski, 2009). While ferrous iron behaves as a reductant, the kinetics of aqueous Fe⁺² redox couples with chlorinated hydrocarbons are several orders of magnitude slower than surface-bound and otherwise structurally incorporated Fe⁺² cations in iron minerals (Elsner et al., 2004; Maithreepala and Doong, 2004). Laboratory studies of acid-washed ZVI to remove surface rust demonstrated that reductive dechlorination of TCE occurred more rapidly than with non-acid-washed ZVI (Liu et al., 2006). These authors also showed that increasing amounts of aqueous Fe⁺² slowed dechlorination.

Similarly, Doong and Wu (1992) demonstrated that dechlorination of carbon tetrachloride by aqueous Fe⁺² only occurred below Eh −360 mV, consistent with the oxidation of ZVI to Fe⁺² and that rates of dechlorination of different chlorinated hydrocarbons varied. They also showed that free Fe⁺² alone had no reactivity toward 1,1,1-trichloroethane or tetrachloroethylene. The range of dechlorination rates reported by these authors is explainable by the spread of reduction potentials of chlorinated organics presented in Table 2, which reflect the differences in C-Cl bond strengths. The standard reduction potential for reduction of Fe⁺³ to Fe⁺² is +0.77 V, far too high for the reduction of many contaminants presented in Table 2 and well above the largely negative Eh values observed at many PRBs (ITRC, 2011; Wilkin et al., 2003). Accelerated ZVI corrosion by sodium chloride pre-treatment similarly demonstrated enhanced dechlorination of TCE by ZVI as surface pitting increased the number of electron transfer sites and, hence, the reactivity of ZVI (Golpagar et al., 1997). In a neat system comprised of deionized, deoxygenated water using standard reagents, Culpepper et al. (2018) reported negligible dechlorination of PCE and TCE by magnetite. Similar results were also observed with Fe (OH)₂ unless Fe⁺² was added. These results are consistent with Huang and Zhang (2005), who reported that Fe⁺² sorbed onto corroded ZVI, which developed a lepidocrocite-magnetite surface coating, promoted nitrate reduction; a lack of sorbed Fe²⁺ decreased reactivity and contaminant removal. The literature collectively reported that free Fe⁺² reacts slowly with some contaminants having reduction potentials above that of the transition of Fe⁺³ to Fe⁺² and plays a critical role in co-precipitation and some redox reactions. However, Fe⁺² is not responsible for contaminant reduction for reactions requiring Eh below +0.77 V. Instead, the oxidation of ZVI provides the driving force as manifested by largely negative field Eh values and evidence of reduction reactions that are not achievable by the oxidation of Fe⁺² to Fe⁺³.

Plating reactions also occur on ZVI surfaces and provide further evidence for electron transfer from ZVI to reduceable contaminants. Standard reduction potentials reported in Table 2 demonstrate the thermodynamic favorability of arsenic reduction in concert with the oxidation of iron species. Reduction of arsenite to metallic arsenic (As⁰) on ZVI has been documented by Klas and Kirk (2013) under acidic and anaerobic conditions. At alkaline pH, no reduction was observed. Similar findings were also reported by Ramos et al. (2009), in which zerovalent arsenic formed on ZVI surfaces from both arsenite and arsenate after reaction in water. Arsenite reduction by nZVI to metallic arsenic and even anionic arsenic (As⁻¹) in arsenopyrite (FeAsS) has also been reported (Yan et al., 2012). These findings suggest that electron transfer from core α-Fe in ZVI continues through ferrous iron atoms either sorbed or structurally incorporated in magnetite, green rust species, and ferrous sulfides attached to ZVI surfaces which are commonly found in PRBs such as the Oakridge Y-12, Elizabeth City, and Portsmouth sites (Roh et al., 2000; Furukawa et al., 2002; Liang et al., 2005).

Additionally, the plating of zerovalent arsenic onto ZVI surfaces at low pH can eventually lead to surface accumulation of several weight percent of arsenic relative to the iron. Indeed, detection of accumulated arsenic onto ZVI by energy dispersive spectroscopy has provided evidence for plated arsenic to at least several nanometers in thickness given the need for the depth of interaction volume for the As Kα and As K_β transitions at 10.539 and 11.721 keV, respectively (Manning et al., 2002; Ramos et al., 2009). While the oxidation of Fe²⁺ to Fe³⁺ drives the reduction of arsenic species (Table 2), the continued accumulation of arsenic plated onto ZVI surfaces after exposure to ZVI requires continued electron transfer from core Fe⁰ through the lattice of ZVI atoms through surface conducting minerals where they are cathodically received by the frontier orbitals of reducible sorbates since any surface Fe²⁺ would be oxidized to electrically insulating Fe³⁺ surface minerals. Hence, the reduction–oxidation couple between arsenic and iron drives corrosion of ZVI and is driven by the transfer of electrons from Fe⁰ within ZVI particles. The literature also abundantly reports plating of copper (Li and Zhang, 2007), mercury (Nam et al., 2017), selenium (Gibson et al., 2012), and silver (Li and Zhang, 2007), resulting in the formation of thin films of these elements on ZVI surfaces in reduced states. Some plating reactions (Table 2) can only be achieved by the oxidation of Fe⁰ (Eh = −0.4 4 V) and not by the oxidation of Fe²⁺ (Eh = +0.77 V). Deposited zerovalent metals on ZVI surfaces require the conduction of electrons from core Fe⁰ as electrons from surface Fe atoms would not result in any significant contaminant accumulation and plating. Therefore, the oxidation of Fe⁰ provides the driving force for many contaminant reductions.

3.3. Effect of groundwater chemistry on iron transformations

Groundwater contains numerous dissolved substances, including inorganic cations, anions, and dissolved organic matter (Fig. 1). Many of these substances are present due to the natural biogeochemical cycling of soil materials resulting in spatial and depth-wise variations in groundwater chemistry throughout the field. Dissolved oxygen promotes iron oxidation and forms ferric oxyhydroxide rusts, including lepidocrocite, goethite, and less crystalline polymorphs such as ferrihydrite (Antunes et al., 2003; Zhao et al., 2016). In deeper, anoxic soil horizons, the lack of oxygen promotes a more reducing environment where microbes utilize ${\text{NO}}_3^{-}$, Mn⁺⁴, Fe⁺³, ${\text{SO}}_4^{2-}$, and CO₂ as electron acceptors to fulfill their metabolic needs (Rivett et al., 2008). Reduction of Fe⁺³ to Fe⁺² allows for magnetite formation and the precipitation of green rusts, which have anion exchange properties. Sulfate reduction by microbes yields dissolved sulfide (H₂S or HS⁻), which coordinates with Fe⁺² to form ferrous sulfides such as mackinawite (FeS), marcasite and pyrite (FeS₂), greigite (Fe₃S₄), and non-stoichiometric variants (Fig. 1; Butler and Hayes, 2000; Weerasooriya and Dharmasena, 2001; He et al., 2010). Iron sulfides play a critical role in the long-term sequestration of contaminants, including arsenic, chromium, and zinc as these minerals are redox active and can therefore reduce some contaminants and also sequester certain metals through co-precipitation (Wilkin et al., 2005; Liu and Lo, 2010; Kumar et al., 2015). These transformations occur with native iron present in soil and direct the diagenetic transformations of iron in PRBs as ZVI corrodes over time (Phillips et al., 2010).

Fig. 1.

Groundwater geochemical and hydrological parameters that are closely linked to long-term performance of ZVI PRBs. Ideal ranges are provided to support the selection of the PRB technology based on site characteristics.

Oxidation of ZVI is influenced by water chemistry: dissolved oxygen promotes corrosion, as do some of the common anions found in groundwater (Fig. 1). In water production, the charge ratio of chloride and sulfate to bicarbonate is often monitored since values greater than 0.2 promote corrosion of iron pipes. This is known as the Larson ratio. Hence chloride and sulfate-rich groundwaters can lead to rapid corrosion and surface passivation of ZVI in PRBs. ZVI also reduces nitrate, leading to the decreased service life of some PRBs (Henderson and Demond, 2007), and has been particularly shown to impact permeability and hydraulic conductivity at the Oak Ridge Y-12 site (Roh et al., 2000).

Anaerobic corrosion of ZVI yields significant amounts of dihydrogen and increases water alkalinity (Reardon, 2005). Hydrogen gas can be utilized as an electron donor by microbes; however, rapid release of this gas can also decrease hydraulic conductivity due to pore plugging, as observed in a PRB in Durango, CO deployed for treatment of uranium (Morrison et al., 2002, 2003). Field sampling in PRBs consistently has shown elevated pH as much as 11 (but generally 9.2 to 10.5) due to hydroxide generation (Li et al., 2005; Wilkin et al., 2014), promoting other mineral assemblages of carbonates of iron, calcium, and magnesium (Wilkin and Puls, 2003; Wilkin et al., 2003). These mineral precipitates can clog PRBs causing groundwater to bypass rather than flow through the barrier, leading to contaminant transport around the treatment system (ITRC, 2011). The formation of authigenic minerals on ZVI surfaces blocks the surface from contact with groundwater in a process described as passivation.

Moreover, the types of minerals and their electrical conductivity determine whether the ZVI surface continues to donate electrons for reactions with contaminants or becomes quenched (Noubactep, 2009). Rapid corrosion leads to surface passivation that decreases ZVI reactivity (Kim et al., 2014). As corrosion products and minerals accumulate on the surface, the reactivity of ZVI decreases as the diffusion path for the contaminant increases (Noubactep, 2010). Corrosion products are observed as relatively darker regions that surround the brighter ZVI particles in SEM images such as in Fig. 2; ZVI phases have higher average atomic mass (Z ~ 57 g/mol) vs iron oxides, sulfides, and other precipitates that collectively have lower mean atomic mass which have deeper interaction volume with the electron beam and hence lower electron yield to the detector. Surface passivation has been a contributor to decreased reactivity of ZVI (Farrell et al., 1999), decreased porosity in some PRBs (Wilkin et al., 2014), and the decreased service life of PRBs associated with contaminant breakthrough in some field cases (Henderson and Demond, 2007).

Fig. 2.

SEM micrograph of core extract from Elizabeth City, North Carolina (USA).

Surface passivation can also occur by the adsorption of natural organic matter, resulting in decreased reactivity of ZVI with contaminants (Liu and Lo, 2010; Kim et al., 2014); however, sorption of organic matter is expected at pH below the point of zero net charge of iron oxyhydroxides which ranges from 5 to 8, above which, electrostatic repulsion is expected between negatively charged iron oxyhydroxide surfaces and anionic carboxylate moieties of dissolved organic matter. Given the high in-wall pH reported in many PRBs (Li et al., 2005; Wilkin et al., 2014), sorption of organic matter and, hence, surface passivation of ZVI caused by surface conditioning is likely marginal in-wall and likely limited to the influent side of PRBs.

3.4. Iron corrosion, properties that govern reactivity, and effect of groundwater chemistry

Iron corrosion is described as a shell-to-core process, initiated at the exterior surfaces and defect sites in the crystalline lattice of ZVI particles, proceeding toward the inner region (Shavel et al., 2007). The pathways of iron oxidation presented in Table 2 demonstrate that ZVI can readily oxidize to ferric and ferrous iron; however, the latter two-electron transfer process is more thermodynamically favored and occurs as the initial step in the oxidation of ZVI followed by the subsequent oxidation of Fe⁺² to Fe⁺³ (Sarin et al., 2004; Wang et al., 2021). In the presence of water, the two-electron transfer process has recently been shown to be a combination of two one-electron oxidations with water that forms an HFeOH intermediate, ultimately yielding Fe(OH)₂ (Filip et al., 2014). Water is the dominant electron acceptor in aqueous environments, especially groundwater in PRBs. Dissolved oxygen, when available, also accepts electrons to facilitate iron corrosion. Moreover, contaminants and other redox-active species in groundwater also accept electrons, cathodically promoting the corrosion of ZVI when thermodynamically feasible (Gillham and O’Hannesin, 1994; Puls et al., 1999; Huang and Zhang, 2005).

Corrosion of ZVI and reduction of contaminants are regulated by electron transport. While ZVI is electrically conductive, the minerals that form via oxidation and precipitation of Fe⁺² and Fe⁺³ with available anions are typically less electrically conductive. Some species are semi-conductive (Fe⁺² bearing minerals), and others are non-conductive (Fe⁺³ oxides, hydroxides, carbonates, and others) (Wang et al., 2021). Magnetite has been demonstrated to be the initial corrosion product of ZVI, followed by complete oxidation to maghemite and other ferric oxides (Gilbert et al., 2010). Initially, magnetite forms on ZVI surfaces or as hollow oxide crystals exsolved from solid solution with α-iron. Formation continues with the diffusion of source oxygen (dioxygen, water) due to Kirkendall-driven electron deficiencies and cation diffusion between magnetite and ZVI phases (Cabot et al., 2007; Wang et al., 2009). Magnetite is a semi-conductor that conducts electrons when energy levels of the frontier orbitals of a sorbate and the highest occupied molecular orbital (HOMO) of the solid surface are similar. Further, the band gap, the difference in energy between the bottom of the conduction band (lowest unoccupied molecular orbital) and the top of the valence band (HOMO) of the solid’s surface also governs electrical conductivity. Smaller band gaps lead to increased conductivity, such as those found in magnetite or large band gaps of insulating materials – a characteristic of the ferric oxides (Cornell and Schwertmann, 2003). The hybrid character of molecular orbitals also influences conductivity, where greater d vs p character in oxide minerals leads to smaller band gaps and increased electrical conductivity (Yong and Schoonen, 2000). For these reasons, the accumulation of ferric and other non-conducting minerals on ZVI surfaces leads to passivation and decreased reactivity of ZVI in the PRB. This mechanism is evidenced from the coatings of magnetite, green rust, aragonite, calcite, mackinawite, greigite, and lepidocrocite observed on the ZVI surfaces at the PRB in Elizabeth City, NC, as observed in Fig. 2 (Furukawa et al., 2002). These findings clearly reveal surface passivation and its effect on the reactivity of ZVI surfaces.

Sulfide minerals commonly develop in PRBs and participate in reduction reactions with contaminants (Wilkin et al., 2003). They form with the assemblage of Fe⁺² with sulfide anions generated by the metabolism of sulfate by sulfate-reducing bacteria (SRB) (Odom and Singleton, 1993). Iron sulfides often exhibit greater reducing power than their oxide counterparts (Yong and Schoonen, 2000). The electronic environment of iron sulfides is typically low-spin due to increased crystal field splitting that comes with the octahedral coordination of Fe⁺² to disulfide $\left({\text{S}}_2^{-2}\right)$ (Arumugam et al., 2019), decreasing the band gap relative to oxides (Vaughan and Craig, 1978). Sulfur also has 3d orbitals. Though these orbitals are poorly occupied, they do indeed influence reactivity and electrical conductivity of sulfur compounds (Meyer, 1976; Reed and Weinhold, 1986). Further, the larger size of sulfide versus oxide makes the sulfide anion more polarizable which also influences reactivity. For these reasons, greigite (FeS•Fe₂S₃) is conductive while magnetite (FeO•Fe₂O₃) is semi-conductive. Likewise, the increased sulfur to iron ratio of pyrrhotites (Fe_1-xS) has been shown to decrease electrical resistivity of the semi-conductor (Wang and Salveson, 2006). This may explain the results of Butler and Hayes (2000), where the reactivity of aged iron sulfides with TCE was dramatically enhanced when iron sulfides were treated with bisulfide. Here, sulfide may have acted as an n-dopant in aged ferrous sulfide by replacing ${\text{S}}_2^{-2}$ anions with S⁻² anions, narrowing the band gap of the iron sulfide surface to promote electron transfer.

Iron silicates also form during the corrosion of ZVI. In a column study (Kohn et al., 2005), silica was observed to accumulate on ZVI grains and reduced the overall reactivity of TCE with ZVI; however, silicates, particularly fayalite, have been demonstrated to increase the electrical conductivity of ZVI (Bradley et al., 1962; Hinze et al., 1981) and the corrosion rate of ZVI (Lawrinenko et al., 2017b). Decreased reactivity between ZVI and TCE may be due to rapid surface corrosion that leads to passivation, which may occur in silicarich groundwater (Fig. 1; Kohn et al., 2005). Dissolved silica (Si (OH)₄) is present in most groundwater systems at concentrations near the solubility of quartz (~0.4 mM), but higher concentrations are possible, especially in clay-rich aquifers or aquifers with long groundwater residence time.

Ultimately, the transformations of iron that occur with its corrosion lead to myriad authigenic mineral transformations. The formation of these new minerals is governed by groundwater chemistry. These nascent minerals coat the surface of the corroding ZVI particle, influencing its reactivity by controlling electron transport from the core of the particle to its exterior. In general, the interaction of various subsurface parameters such as geochemical, microbial, and hydraulic processes with ZVI embedded PRBs is complex (Weber et al., 2013). There exists a continued need to understand how the aging of iron in groundwater influences the reactivity and long-term performance of ZVI based on groundwater chemistry, corrosion products, and other minerals that form in place, and how these new surfaces influence contaminant transport to the conductive surface, sorption, and electron transfer between the aged iron particle and the contaminant.

Not all ZVI corrodes equally or at the same rate, and ZVI from different sources exhibits different reaction rates with contaminants and corrosion characteristics (Johnson et al., 1996; Reardon, 2005). The ZVI utilized in PRBs is recycled steel from metal fabrication: turnings and fines that are further processed and graded to specific mesh sizes (Puls et al., 1999; Landis et al., 2001). Johnson et al. (1996) identified that purity of the iron, specific surface area, and particle size all influence performance. They conducted kinetic studies on various chlorinated hydrocarbon contaminants using different ZVI sources and demonstrated that half-lives of these contaminants ranged over an order of magnitude and that different sources of ZVI varied greatly in performance. Likewise, Powell et al. (1998) reported a range of half-lives of TCE in both kinetic and column studies in which ZVI from several commercial sources was evaluated. Carbon steel has been found to corrode faster than ductile iron (Song et al., 2017). Larger particle size often corrodes slower (Comba et al., 2010); however, the crystallinity or relative long-range to short-range order in the lattice structure and the presence of defect sites in ZVI also influences corrosion rates. Graphite and carbides of iron form during iron smelting and steel production. While graphite is relatively inert, it is electrically conductive and conducts electrons from one edge of a graphite phase to another under environmentally relevant temperatures (Buerschaper, 2004). Carbides readily oxidize to organic acids, contributing to defect formation and an increased number of sites for oxidation (Deng et al., 1997; Cvetković et al., 2018). The reactivity of iron carbides and the presence of graphite inclusions in ZVI reasonably suggest that the carbon content of steel used in PRBs contributes to its reactivity with contaminants which influences the performance of the PRB over time. As such, there is a need to better characterize ZVI to understand the cause-and-effect relationship between its structural chemistry and physical properties on performance in a PRB to improve long-term performance.

3.5. The role of subsurface biota in remediation

Remediation of contaminated groundwater and soil is a concerted, dynamic process that involves combined physicochemical and biotically mediated mechanisms. Some contaminants can be naturally attenuated by microbes via processes in which they consume or convert the contaminant to less toxic forms. Microbes utilize various substrates to fulfill their metabolic needs. In aerobic environments, organic materials are oxidized to carbon dioxide while dioxygen is reduced. During this process, many organic compounds are wholly or partially mineralized; however, not all microbes have the same ability to digest carbon, nor are microbes capable of robustly utilizing every form of carbon in their metabolic processes as the enzymes produced by microbes are stereospecific and made to cleave certain bonds.

Further, enzymatic activity is influenced by soil chemical factors such as pH (Sharma et al., 2016). For example, certain fungi, ascomycetes, and basidiomycetes, can oxidize phenolic materials originating from lignin in plant litter while other fungi cannot (Claus and Filip, 1990). Other microbial taxa present in soil share similar trophic constraints (Sinsabaugh, 2010). In the absence of oxygen, diverse facultative microbes utilize nitrate, manganate, ferric iron, sulfate, and even hydrogen as electron acceptors to drive their metabolic processes. Some microbes can reduce arsenic (Ahmann et al., 1994; Escudero et al., 2013) and uranium (Stylo et al., 2015) in metabolism rendering these contaminants to be either more or less mobile in the environment. Oligotrophic conditions develop with the loss of the ‘easy food’, thereby driving relative abundances of different soil organisms where those capable of processing the available substrate survive while others do not. For this reason, SRB are more abundant in sulfate-rich groundwater (Leloup et al., 2009; Popovich et al., 2018), particularly in anaerobic environments present in deep subsurface horizons within and downgradient of PRBs, where competition and microbial abundance is driven by available substrate (Gu et al., 2002). Comprehensive reviews of soil microbial processes, including adaptation to heavy metals, are provided by Colombo et al. (2013) and Giller et al. (2009).

The efficacy of microbial attenuation of contaminants in groundwater varies with the contaminant. It is driven by microbes’ ability to thrive in the presence of contaminants, thermodynamic favorability of metabolically utilizing the contaminant, and environmental factors such as pH, redox, and temperature. Chemical stability of C-Cl bonds and aromatic rings are examples of moieties difficult for most microbes to mineralize, hence the persistence of many of the chlorinated hydrocarbons (Lenczewski et al., 2003; Pant and Pant, 2010), benzene, toluene, ethylbenzene, and xylene (BTEX), and related compounds in soil and groundwater (Seagren and Becker, 2002). A review of the transformations of halogenated organic compounds is presented by Vogel et al. (1987).

Bioattenuation is enhanced by the co-metabolism of organic carbon present with the contaminant. Addition of labile carbon such as methanol, glucose, acetate, and humic substances to groundwater contaminated with cis-1,2-dichloroethene (DCE) has been demonstrated to augment the dechlorination rate of DCE (Doong and Wu, 1995). Co-metabolism of toluene with DCE also increased the degradation of DCE and vinyl chloride. For example, at a PRB located at the Denver Air Force Base, it was observed that the addition of DCE led to the decreased half-lives of benzene and toluene present at this site (Schäfer and Bouwer, 2000). Cometabolism has been demonstrated to support the bioattenuation of benzene in anaerobic environments in which benzene is known to persist (Atlas, 1981). Vogt et al. (2011) suggested a symbiotic relationship as one of the pathways of co-metabolism in which one organism creates the substrate needed by another organism capable of mineralizing the aromatic ring as the other organism by itself is incapable of mineralizing the aromatic ring. The terminal electron acceptor plays a crucial role in driving the mineralization of organic contaminants and is promoted by syntropy between different organisms present in the aquifer. Sulfate (Anderson and Lovley, 2000), nitrate (Majora et al., 1988), manganese (Villatoro-Monzón et al., 2003), ferric iron (Botton and Parsons, 2006), and even graphitic carbon (Zhang et al., 2010) all serve as electron acceptors in the oxidation and mineralization of benzene in the environment. These studies demonstrate that the aromatic ring typical in many contaminants originating from petroleum, explosives, and certain pesticides can be bioattenuated and that enhanced bioattenuation through the provision of specific substrates may promote remediation under some circumstances.

A perusal of the literature demonstrates that various contaminants are labile to microbes; however, the extent to which bioattenuation occurs varies with the chemistry of the contaminant and groundwater chemical and physical factors. Sometimes bioattenuation is sufficient to naturally mitigate contaminants in soil and groundwater; however, the persistence and long half-lives of many contaminants attest to the need for more aggressive physicochemical measures for which PRBs have been designed. In PRBs, the accumulation of microbial biomass will decrease porosity and could potentially decrease hydraulic conductivity (Wiedemeier et al., 1998); however, this microbial fouling has not been identified as contributing factor to decreased performance (Vogan et al., 1999). Instead, surface passivation and accumulation of other minerals have been identified as key processes responsible for the decreased performance, as previously discussed. Microbial activity in most cases appears to improve system performance via direct (contaminant degradation) and indirect (sulfate reduction leading to FeS formation) processes.

4. Role of PRB design on long-term performance

4.1. Design considerations for PRBs

In the construction of PRBs, native aquifer material is excavated and replaced with alternate porous media having higher hydraulic conductivity than the surrounding aquifer. Early PRBs were primarily comprised of granular iron, which is generally greater than 95% ZVI; however, more recent installations employ mixtures of granular iron mixed with sand, gravel, pumice to avoid influent plugging of the PRB (Birke et al., 2007; Moraci and Calabrò, 2010), and organic matter such as compost or peat to provide substrate to enhance microbial activity (Flury et al., 2009; Ludwig et al., 2009; Jeen et al., 2014). Site characterization is performed before construction using appropriate hydrological models to understand groundwater flow and kinetic studies to determine the reactivity of the ZVI selected for use. Column studies are performed to simulate PRBs on a pilot scale using site or simulated water to estimate kinetic parameters necessary to determine the thickness of the PRB. A safety factor of four has been used for many PRBs to increase assurance of successful remediation; however, safety factors have ranged from less than one to greater than ten, often determined by budget and construction constraints at specific sites (Muegge and Hadley, 2009).

Several physical designs of PRBs exist, including the continuous wall, caisson, funnel-and-gate, and staggered hole systems (Gavaskar et al., 1998). Continuous walls require the complete replacement of aquifer solids with reactive porous media. The caisson is a tank-like reactive cell that can be excavated and re-filled when the reactive medium is exhausted (Conca et al., 2003). Funnel-and-gate systems channel groundwater into the PRB using steel sheeting or compacted clay. Funnels must have substantially lower permeability than the surrounding aquifer to conduct groundwater through the reactive zone. The funnel-and-gate systems use less reactive material than other physical designs of PRBs; however, there is a possibility of contaminated water bypassing the gate. Specifically, if funnel structures are not driven into the aquitard (bedrock, argillic horizon, or other soil material of relatively lower hydraulic conductivity) then bypass is possible or likely. These systems also become increasingly vulnerable to flows bypassing the funnel walls with substantial rain events. The staggered hole design is a series of wells drilled into the subsurface. Compared to other systems, this approach can reach greater depths; however, some contaminant bypass may be encountered under conditions of increased groundwater flow. Early PRBs were mostly continuous wall systems; however, the choice of a design depends on results of site characterization, field challenges during construction, and cost (ITRC, 2005, 2011).

Heterogeneous distribution of hydraulic conductivity contributes to the non-uniform flow of groundwater and enlargement of contaminant plumes. Rahman et al. (2005) reported that “at the current state of knowledge, it is not clear to which extent microheterogeneities are important for solute mixing.” These authors and others (Rolle et al., 2009; De Barros and Nowak, 2010) later demonstrated that the presence of high permeability inclusions within aquifer compartments of relatively lower permeability can cause a channeling effect on water flow which results in transverse spreading of the contaminant plume and the potential for bypass of in-situ remediation systems. Xu et al. (2018) examined steady-state dilution and reactive mixing in saturated porous media and reported that numerical approaches used to understand contaminant fate and transport need continuous improvement. These authors emphasized that soil heterogeneity strongly influences the macro-dispersion of contaminants in groundwater. They also noted a lack of large-scale experiments to accurately assess the effect of soil heterogeneity on large-scale dilution and mixing of contaminants in the subsurface. Thus, contaminants bypassing some PRBs may be attributed to the unexpected spreading of the plume related to design and installation based on the state of science at the time of construction. In two particular PRBs, the Department of Energy’s Kansas City plant (Laase et al., 2000) and the Lake City Army Ammunitions Plant (ITRC, 2005), groundwater was found to pass around the PRB and not directly through it. Such failure is likely due to inadequate characterization of hydraulic flow, attributed to an inadequate number of measurement points (ITRC, 2005). In conclusion, future installations need to employ more descriptive groundwater flow models that account for variable flow, contaminant spreading, and variation in hydraulic conductivities that contribute to bypass. Additionally, more field experiments are necessary to improve the reliability of groundwater flow models.

4.2. Importance of site characterization: hydraulic changes

In the design of a PRB, besides general convection–dispersion properties of the soil and the PRB material, hydraulic properties such as porosity and hydraulic conductivity are equally important. In time, the physical and chemical makeup of the PRB changes with progressive mineral accumulation. These changes are primarily governed by water chemistry (Fig. 1). Surface passivation and the deposition of authigenic minerals alter the corrosion characteristics of ZVI and also block reactive sites, occlude pores, and eventually alter the transport of water and contaminants through the PRB (Li et al., 2005).

Mineral deposition also potentially alters the hydraulic properties of the PRB. Particularly, inorganic carbon deposition has been shown to cause substantial accumulation of aragonite, calcite, siderite, chukanovite, other ferrous carbonates, and carbonate green rusts in the PRB, particularly at the influent side (Mackenzie et al., 1999; Wilkin et al., 2003; Henderson and Demond, 2007). An illustrative example of such deposition is presented in Fig. 2, which shows acicular crystal growth, presumably aragonite, platy crystalline structures, and amorphous iron oxyhydroxides that surround a ZVI particle extracted from the fine core fraction of a PRB from the Elizabeth City site in North Carolina (Furukawa et al., 2002). In a quantitative analysis of ZVI PRBs reported by Henderson and Demond (2007), high influent pH, internal Eh, and high concentrations of chloride, nitrate, and alkalinity were associated with reduced PRB performance. Elevated pH and alkalinity permit the precipitation of carbonate minerals. High chloride and sulfate concentrations promote rapid corrosion and surface passivation, resulting in increased Eh within the PRB.

The oxidation of ZVI in the PRB also leads to mineral deposition and pore-clogging. The density of ZVI is 7.7 g/cm³, while the density of ferrihydrite (3.8 g/cm³), magnetite (5.2 g/cm³), mackinawite (4.2 g/cm³), and other iron precipitates is much lower. These differences lead to a molar volume change of iron minerals, causing pore filling and occlusion of ZVI from groundwater (Wilkin and Puls, 2003). This pore filling leads to increased macropore flow and decreased groundwater residence time in the PRB (Kamolpornwijit et al., 2003) and has contributed to zones of decreased hydraulic conductivity after 20 years in the Elizabeth City PRB; however, it did not affect the bulk hydraulic conductivity (Wilkin et al., 2014, 2019). Earlier PRBs that employed only ZVI as fill material were prone to pore filling at the influent side of the PRB; however, PRBs which contained sand mixed with ZVI were not reported to have this problem (Li and Liu, 2022).

4.3. Knowledge gaps and future perspectives for PRBs

While PRBs have been in operation for more than 30 years, the understanding of their biogeochemical mechanisms continues to develop, owing to the variation in ZVI used in PRBs, groundwater chemistry, hydraulic factors, and the interactions between ZVI, its corrosion products, and contaminants. This review paper provides baseline mechanistic information supported by selected examples from field studies that demonstrate how the performance of aging ZVI-amended PRBs continues to change over time. The corrodibility of ZVI, mineral deposition, and the effect of authigenic minerals on ZVI reactivity influence the long-term performance of PRBs. While other studies have discounted the role of iron corrosion, analysis of field studies strongly refutes such claims and appears to reemphasize the importance of iron corrosion as a primary control on the long-term performance of PRBs. Successful remediation using PRBs requires careful characterization of groundwater flow, seasonality, and a better understanding of groundwater chemistry and its influence on the reactivity of ZVI with contaminants. Dissolved oxygen drives rapid corrosion and mineral deposition and is a leading cause of contaminant bypass and breakthrough. We recommend pretreatment upgradient of PRBs using organic matter to reduce dissolved oxygen in groundwater that enters the PRB to extend service life. Increased travel paths through wider PRBs and co-mixing ZVI with sand would increase the performance of PRBs when groundwater is high in alkalinity and dissolved solids by deterring plugging at the influent side of a PRB. Recent modelling work demonstrated that multiple PRBs in series should more reliably remove contaminants. Some of these design improvement options may not always be feasible due to construction constraints; therefore, caisson-type PRBs may be a better option where ZVI could at least be agitated to break up cemented aggregates. There exists a need to characterize ZVI better to select the most effective material for long-term performance as ZVI reactivity is influenced by its chemical makeup. Finally, there is still a need for improvement in models used for site characterization and field studies to better understand groundwater flow, as some PRBs were not as effective as they could have been due to their orientation not aligning properly with the groundwater flow direction.

This review paper comprehensively documented the numerous mechanisms associated with the operation of PRBs. This exhaustive mechanistic review is intended to rectify prevailing misconceptions about the nature of contaminants and their interaction with the ZVI-amended PRBs. We documented the role of iron corrosion products and their effect on overall reactivity and contaminant reduction to examine past misconceptions that negated the role of iron corrosion. We developed this insight by carefully reviewing the performance of successful ZVI-amended PRBs over the last twenty-plus years. We supported our understanding by documenting the interaction of contaminants with the PRBs and taking into consideration groundwater geochemistry, presence of subsurface biota, design, and the overall ZVI reactions and transformation. By carefully discussing numerous mechanisms, we demonstrated that ZVI corrosion products and mineral precipitates play a critical role in the remediation of various organic and inorganic contaminants and influence the long-term performance of ZVI PRBs.