Chlorinated Solvent Transformation by Palladized Zerovalent Iron

Jun 11, 2013 - Palladized nanoscale zerovalent iron (Pd/NZVI) has been utilized for source zone control, yet the reductant responsible for pollutant ...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/est

Chlorinated Solvent Transformation by Palladized Zerovalent Iron: Mechanistic Insights from Reductant Loading Studies and Solvent Kinetic Isotope Effects Yang Xie† and David M. Cwiertny*,‡ †

Department of Chemical and Environmental Engineering, University of California, A242 Bourns Hall, Riverside, California 92521, United States ‡ Department of Civil and Environmental Engineering, University of Iowa, 4105 Seamans Center, Iowa City, Iowa 52242, United States S Supporting Information *

ABSTRACT: Palladized nanoscale zerovalent iron (Pd/ NZVI) has been utilized for source zone control, yet the reductant responsible for pollutant transformation and the optimal conditions for subsurface application remain poorly understood. Here, trends in Pd/Fe reactivity toward 1,1,1,2tetrachloroethane (1,1,1,2-TeCA) and cis-dichloroethene (cisDCE) were compared in H2O and D2O batch systems as a function of pH, chlorinated solvent concentration, Pd surface loading, Pd/Fe mass loading, Pd/Fe aging time, and zerovalent iron [Fe(0)] particle size. For Pd/NZVI, the solvent kinetic isotope effect [i.e., kobs(H2O)/kobs(D2O) or SKIE] for 1,1,1,2-TeCA and cis-DCE reduction increased substantially with Pd loading and Pd/NZVI concentration, evidence that multiple pathways exist for chlorinated solvent reduction. At low Pd loadings and Pd/NZVI concentrations with relatively small SKIEs (less than ∼5), we propose that modest reactivity enhancements (≤10fold) reflect more efficient electron transfer to 1,1,1,2-TeCA from Fe(0) facilitated by Pd nanodeposits. Much larger SKIEs (e.g., exceeding 100 for cis-DCE) imply the involvement of atomic hydrogen in more reactive systems with high Pd loadings and Pd/ NZVI concentrations. Generally, evidence of SKIEs supporting a dominant role for atomic hydrogen was not observed for Pd/Fe prepared from micrometer-sized Fe(0), or for any size of nonpalladized Fe(0). During anaerobic aging of Pd/NZVI, decreases in the SKIE for 1,1,1,2-TeCA reduction suggest that atomic hydrogen will contribute to reactivity for only approximately 1 week after application.



INTRODUCTION Since its inception,1 application2−6 of palladized nanoscale zerovalent iron (Pd/NZVI) for source zone control has grown in popularity. Pd/Fe is typically the most reactive bimetallic reductant.7,8 As such, it can achieve greater rates of pollutant removal1,7,9 and more fully dehalogenated products7,10 than zerovalent iron [Fe(0)] and other bimetallic formulations. Despite the increase in its frequency of application, prior work with Pd/Fe reveals a relatively broad spectrum of performance. A common metric for Pd/Fe reactivity is the extent to which it enhances pollutant transformation relative to Fe(0), a value typically obtained by normalizing the pseudofirst-order rate constant (kobs values) for pollutant reduction by Pd/Fe [kobs(Pd/Fe)] to the corresponding kobs value from an identical Fe(0) system [kobs(Fe)]. In a series of studies with Pd/Fe, we previously observed rate enhancement factors (hereafter REFs) anywhere between 27 and 1000.11 Some of this variation could be explained; increasing the amount and coverage of Pd, which is widely believed to be the reduction site on the bimetallic surface, increased REFs from 2 to 10 for 1,1,1trichloroethane (1,1,1-TCA) reduction. Subsequently,11 however, these same Pd/Fe formulations enhanced cis-dichloro© 2013 American Chemical Society

ethylene (cis-DCE) reduction by more than 1000-fold. Moreover, cis-DCE exhibited a far greater rate of change in REF per unit of deposited Pd mass on the Fe(0) surface than 1,1,1-TCA. Clearly, factors other than Pd loading, including, but not limited to, the nature of the target oxidant (e.g., alkyl vs vinyl polyhalide), also influence reactivity. Variations in REFs would also be introduced if the mechanism of pollutant transformation, specifically the nature of the reductant species, differs between Fe(0) and Pd/Fe systems. Two pathways for reactivity enhancement by Pd/Fe have gained popular acceptance. One hypothesis2,12−14 involves the formation of a galvanic couple, in which iron functions as the anode where oxidation of Fe(0) to Fe(II) occurs while Pd serves as the cathode at which pollutant reduction takes place. In this couple, Fe(0) becomes more easily oxidized, in turn increasing the rate of electron transfer to the pollutant at the Pd surface. Others have postulated7,8,13,15−19 that Pd functions as a Received: Revised: Accepted: Published: 7940

April 5, 2013 June 11, 2013 June 11, 2013 June 11, 2013 dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

(ρm) ranged from 0.03 to 0.6 g/L for Pd/NZVI and from 0.6 to 5.7 g/L for Pd/ZVI. The initial 1,1,1,2-TeCA concentration ranged from 3 to 2000 μM (∼0.5−335 mg/L). The reactor pH was controlled via 25 mM HEPES (pH 7−8) or MES (pH 6) buffer in 25 mM NaCl. Details on reactor construction, operation, and sampling are provided in the Supporting Information and in our prior work.21,22 Measurement of Solvent Kinetic Isotope Effects. Reactivity studies were conducted in parallel with H2O and D2O. The 25 mM HEPES/NaCl solutions were prepared in D2O using appropriate pH meter corrections (pD = pH meter reading + 0.40)23 for differences in the pH and pD scales. SKIEs were measured primarily as a function of Pd loading on NZVI and ZVI and over a range of ρm values. In systems with variable ρm values, initial concentrations of 1,1,1,2-TeCA or cisDCE were chosen so that a constant molar ratio of chlorinated solvent to iron-based reductant of 0.02 was employed in all systems. SKIEs were quantified from the reactivity difference between reactors [i.e., kobs(H2O)/kobs(D2O)] prepared under otherwise identical conditions. Pd/NZVI Aging Studies. Aging studies with NZVI and 0.35% Pd/NZVI were conducted over 30 days following anaerobic aging procedures for suspensions (2 g/L) described previously21,22 and detailed in the Supporting Information. Periodically during aging, a portion of the suspension was withdrawn and diluted with fresh buffer prepared in H2O to a final concentration of 0.22 g/L. In parallel, another suspension sample was diluted to the same ρm with buffer prepared in D2O, taking care to remove residual H2O via pipet prior to D2O addition. The suspension reactivity was then quantified toward 1,1,1,2-TeCA. Analytical Methods. The mass of deposited Pd was determined via inductively coupled plasma mass spectrometry analysis conducted at the University of Iowa State Hygienic Laboratory, with results indicating essentially complete deposition (≥99%) of the available Pd mass during displacement plating. Aqueous concentrations of 1,1,1,2-TeCA and cisDCE were determined via gas chromatography with an electron capture detector. Dissolved Fe(II) concentrations were quantified in select systems using the 1,10-phenanthroline method,24,25 while the morphology of aged Pd/NZVI was examined via transmission electron microscopy (TEM). Details on all analytical methods and instrumentation are provided in the Supporting Information and in our earlier works.21,22

hydrogenation catalyst where adsorbed or absorbed atomic hydrogen species associated with the Pd surface are responsible for pollutant transformation. Indirect evidence supporting both pathways exists, although the questions of which is dominant for Pd/Fe and how their relative contributions may change with system conditions remain unanswered. When Pd/Fe reactivity is being assessed, values of kobs(Pd/ Fe) and kobs(Fe) used to quantify REFs are typically measured in identical experimental systems. Accordingly, it is often assumed that these values are independent of system variables such as reductant loading (ρm in grams per liter), target pollutant concentration, and pH. Such assumptions are rarely tested, however, and, most importantly, will certainly fail if different reductant species are involved in Fe(0) and Pd/Fe systems. This in turn would introduce large variations in reported REFs for Pd/Fe systems and confound our ability to optimize their in situ performance at the field scale based on laboratory studies. Here, we explore REFs achieved in Pd/Fe reductant systems over a broad range of conditions. Pd/Fe reductants were prepared via displacement plating using a commercially available NZVI powder (hereafter Pd/NZVI). In reactivity experiments with 1,1,1,2-tetrachloroethane (1,1,1,2-TeCA) and cis-dichloroethene (cis-DCE) as model chlorinated solvents, we examined how REFs for Pd/NZVI changed over a broad range of Pd loadings [0.05−0.35% (w/w) Pd], solution pH values (6−8), reductant mass loadings (ρm values of 0.03−5.7 g/L), chlorinated solvent concentrations (3−2000 μM), and suspension aging times (0−30 days). Additional reactivity studies utilized Pd/Fe reductants prepared from micrometersized Fisher electrolytic iron (hereafter Pd/ZVI) to assess the influence of Fe(0) particle size on bimetal reactivity. Practically, this experimental matrix helps to identify the optimal reductant characteristics and application conditions for Pd/Fe performance. New mechanistic insights regarding pollutant transformation in Fe(0) and Pd/Fe systems were also obtained through solvent kinetic isotope effects (SKIEs), measured by comparing rate constants of 1,1,1,2-TeCA and cis-DCE reduction in D2O- and H2O-based buffer systems [i.e., kobs(H2O)/kobs(D2O)]. When a solvent is replaced with an isotopically substituted analogue, SKIEs occur either when the solvent is a reactant or when there are significant interactions between the solvent and transition structures generated via reaction.20 SKIEs were measured for a subset of the aforementioned reductant and system variables to explore conditions promoting atomic hydrogen involvement during chlorinated solvent reduction. Specifically, larger SKIEs indicative of solvent molecule involvement during reaction were thus interpreted as evidence that atomic hydrogen derived from water reduction participated in organohalide transformation.



RESULTS AND DISCUSSION Influence of Solid Loading on Pd/Fe Reactivity toward 1,1,1,2-TeCA. With all reductants, 1,1,1,2-TeCA was transformed exclusively to 1,1-DCE (i.e., reductive βelimination), with loss following exponential decay. Values of kobs for 1,1,1,2-TeCA reduction as a function of reductant loading (ρm) are shown in Figure 1 for reductants prepared from Fisher electrolytic iron (hereafter ZVI) and iron nanoparticles from Nanostructured and Amorphous Materials (hereafter NZVI). Values of kobs are shown for both unamended iron [hereafter Fe(0)] and palladized iron (hereafter Pd/Fe) with 1% Pd (w/w) on ZVI and 0.05 or 0.35% Pd (w/w) on NZVI. As anticipated for ZVI7 and NZVI,9,26 Pd/Fe was more reactive than Fe(0) and reactivity increased with Pd loading. However, the magnitude of REF in Pd/Fe systems varied not only between ZVI and NZVI but also as a function of ρm. In ZVI systems (Figure 1a), kobs values exhibited a nonlinear



MATERIALS AND METHODS Reagents. A complete reagent list is provided in the Supporting Information. Pd/Fe Synthesis. Pd/Fe reductants were prepared on the basis of procedures modified from previous work.7,11 NZVI was plated with Pd loadings of 0.05 and 0.35% (w/w) Pd, whereas micrometer scale ZVI was plated with 1% (w/w) Pd. Detailed protocols are provided in the Supporting Information. Reactivity toward Chlorinated Solvents. The reactivity of all Fe(0) and Pd/Fe reductants was examined toward 1,1,1,2TeCA and cis-DCE in batch systems. Reductant solid loadings 7941

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

suggesting that the reductant surface becomes more reactive as solid loading increases. As a practical consequence of these relationships, the magnitude of rate enhancement provided by Pd/Fe is not constant but depends strongly on ρm. For example, the REF for Pd/ZVI relative to ZVI ranges from 3 to 10 over the range of ρm values investigated (up to 5.7 g/L), while the range in REFs is far greater for Pd/NZVI (from 2 to 300 up to 0.56 g/L). This dependence of REF on ρm likely explains some of the scatter in previously reported REFs for Pd/Fe. Fundamentally, the nonlinear dependence of kobs on ρm indicates some system variable limits the reactive surface area to a greater extent at high ρm values in ZVI and NZVI systems. In contrast, the near first-order dependence of kobs on ρm for 1% Pd/ZVI and 0.05% Pd/NZVI indicates that Pd deposition overcomes this limitation, producing a bimetallic reductant surface more uniformly active relative to Fe(0). Furthermore, such behavior implies that the deposited Pd is indeed the site of 1,1,1,2-TeCA reduction. For 0.35% Pd/NZVI, the unique parabolic dependence between kobs and ρm provides evidence that multiple mechanisms for 1,1,1,2-TeCA reduction are at play in Pd/NZVI systems and that their relative contribution depends on both the Pd coverage and Pd/NZVI mass loading. Trends in Mass-Normalized Rate Constants for 1,1,1,2-TeCA Reduction by NZVI. Mass normalized rate constants (km = kobs/ρm) are frequently reported for Fe(0)based reductant systems.28 We calculated km values for 1,1,1,2TeCA reduction [km(1,1,1,2-TeCA)] for NZVI, 0.05% Pd/ NZVI, and 0.35% Pd/NZVI, which are reported in Figure 2 as a function of the molar ratio of 1,1,1,2-TeCA to NZVI-based

Figure 1. Values of kobs for 1,1,1,2-TeCA reduction as a function of solid loading (ρm) in systems with (a) ZVI and 1% Pd/ZVI and (b) NZVI, 0.05% Pd/NZVI, and 0.35% Pd/NZVI. Reactions were conducted at pH 8 (25 mM HEPES and 25 mM NaCl). Lines indicate the functional relationship between kobs and ρm determined from regression analyses of log−log plots. Uncertainties for kobs represent 95% confidence intervals from linear regression analyses used in their determination. Reactor-to-reactor variability is shown for NZVI, for which data were acquired in duplicate experiments.

dependence on ρm (kobs ∝ ρm0.4±0.1 from the slope of log kobs vs log ρm plots), consistent with previous results with this same ZVI batch.27 In contrast, addition of 1% Pd produced kobs values exhibiting a nearly first-order dependence on ρm (kobs ∝ ρm0.9±0.2), as expected for surface reactions. Unexpectedly in NZVI systems, kobs values were essentially independent of solid loading at pH 8.0 [kobs ∝ ρm0.1±0.1 (Figure S1 of the Supporting Information)], whereas addition of 0.05% Pd produced a relationship (kobs ∝ ρm1.2±0.2) equivalent to a first-order dependence. Interestingly, 0.35% Pd/NZVI produced a reaction order in ρm greater than unity (1.6 ± 0.2),

Figure 2. Mass-normalized rate constants for 1,1,1,2-TeCA reduction [km(1,1,1,2-TeCA)] as a function of the molar ratio of initial 1,1,1,2TeCA to iron-based reductant concentration. Data are shown for NZVI (triangles), 0.05% Pd/NZVI (squares), and 0.35% Pd/NZVI (circles). Empty symbols represent km(1,1,1,2-TeCA) values measured experimentally by holding the initial concentration of 1,1,1,2-TeCA constant (175 μM) and varying ρm (0.03−0.6 g/L). Filled symbols represent km(1,1,1,2-TeCA) values measured experimentally by holding ρm constant (0.05 g/L) and varying the initial 1,1,1,2-TeCA concentration (3−2000 μM). All experiments were conducted at pH 8 (25 mM HEPES and 25 mM NaCl). 7942

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

a bimetallic reductant surface more resistant to passivation in high-ρm systems. Because rates of Fe(II) production were equivalent in our NZVI and Pd/NZVI systems (Figure S2 of the Supporting Information), the extent of insoluble corrosion product formation should be comparable for both reductants. However, these corrosion products appear to be less effective at passivating the 0.05% Pd/NZVI surface relative to NZVI, at least over the time scales of these experiments. Finally, because kobs values increased in a parabolic fashion with respect to ρm in 0.35% Pd/NZVI systems, values of km(1,1,1,2-TeCA) decrease as the molar ratio of 1,1,1,2-TeCA to Pd/NZVI increases. Values of km(1,1,1,2-TeCA) were in agreement regardless of whether the initial 1,1,1,2-TeCA concentration or ρm was varied experimentally. Consequently, the reactivity of 0.35% Pd/NZVI is greatest with excess Pd/ NZVI (as in Figure 1), but it is also notably diminished in systems with high initial 1,1,1,2-TeCA concentrations. From a practical perspective, data in Figure 2 reveal that approaching equimolar concentrations of 1,1,1,2-TeCA and NZVI-based reductant yields roughly equivalent km(1,1,1,2TeCA) values for NZVI and Pd/NZVI. We acknowledge that under such conditions assumptions of the pseudo-first-order model used to quantify km(1,1,1,2-TeCA) fail, but comparison of 1,1,1,2-TeCA concentration profiles over time (Figure S3 of the Supporting Information) clearly reveals the equivalent reactivity of Pd/NZVI and NZVI as the molar ratio of 1,1,1,2TeCA to reductant approaches unity. Thus, little, if any, reactivity enhancement is provided by Pd/NZVI when it is not present in excess relative to its target oxidant (i.e., 1,1,1,2TeCA). Possible explanations include limitations in reactive surface sites at such high solvent concentrations or that Pd/ NZVI may be more susceptible to passivation when exposed to high concentrations of 1,1,1,2-TeCA, a relatively strong oxidant. Regardless, the reactivity enhancement that makes Pd/NZVI attractive for field application may be lost at sites where target oxidant concentrations are high (e.g., near saturation). Solvent Kinetic Isotope Effects for 1,1,1,2-TeCA Reduction by NZVI and Pd/NZVI. Figure S4 of the Supporting Information shows plots of ln[1,1,1,2-TeCA] versus time in NZVI, 0.05% Pd/NZVI, and 0.35% Pd/NZVI systems prepared in H2O and D2O at pH 8. Data are shown for two different ρm values (0.03 and 0.5 g/L) for each reductant. No significant SKIE was observed in NZVI systems over the ρm explored (i.e., 1,1,1,2-TeCA decay was essentially equivalent in H2O and D2O). For 0.05% Pd/NZVI, although 0.03 g/L suspensions also did not exhibit an appreciable SKIE, the higher reductant loading (0.5 g/L) exhibited a clear reactivity difference between H2O and D2O systems, producing a SKIE of 4. For 0.35% Pd/NZVI, both low and high ρm values produced obvious SKIEs, but the magnitude was considerably greater (SKIE ∼ 20) at 0.5 g/L than at 0.03 g/L (SKIE ∼ 2). These trends in SKIE imply that different processes contribute to 1,1,1,2-TeCA reduction by Pd/NZVI relative to NZVI. If the same reductant and mechanism were at play, then SKIEs should be comparable across all systems and invariant with system conditions. Rather, observations of increasing SKIEs with Pd loading and ρm suggest that (i) multiple mechanisms or reactive entities are responsible for 1,1,1,2TeCA reduction by Pd/NZVI, (ii) some of these pathways are not otherwise available in unamended NZVI systems, and (iii) system variables influence the relative contribution of these parallel reduction pathways in Pd/NZVI suspensions.

reductant (i.e., [1,1,1,2-TeCA]/[reductant]). Two methods were used to experimentally determine values of km(1,1,1,2TeCA). First, we varied ρm (0.03−0.56 g/L) and held the initial concentration of 1,1,1,2-TeCA constant (175 μM). Alternatively, the initial 1,1,1,2-TeCA concentration was varied (3.5−2000 μM) in systems with a constant ρm (0.05 g/L). For NZVI, km(1,1,1,2-TeCA) values differed depending on the method in which they were determined because of the nonlinear dependence of kobs(1,1,1,2-TeCA) on ρm. For experiments varying ρm but using a constant initial 1,1,1,2TeCA concentration (empty symbols), km(1,1,1,2-TeCA) decreased with a decreasing molar ratio of 1,1,1,2-TeCA to NZVI [i.e., km(1,1,1,2-TeCA) decreased with an increase in ρm]. In contrast, km(1,1,1,2-TeCA) values from systems with varying 1,1,1,2-TeCA concentrations and a constant ρm (filled symbols) were nearly constant over the range of molar ratios explored. Consequently, at low 1,1,1,2-TeCA to NZVI molar ratios, km(1,1,1,2-TeCA) values varied by a factor of 20 depending on how they were measured. These differences in k m (1,1,1,2-TeCA) indicate that conditions unique to systems with high ρm values must limit NZVI reactivity. We thus attribute the diminished reactivity at high ρm values to the disproportionate formation of corrosion products in such systems. In closed systems with high NZVI concentrations, solubility limits for ferrous [Fe(II)] and ferric [Fe(III)] iron solid phases will be achieved quickly, in turn causing greater levels of accumulation of passivating corrosion products on the NZVI particle surface. This scenario is also likely responsible for the nonlinear relationship between kobs and ρm in ZVI systems observed here (see Figure 1a) and elsewhere (see ref 27 and references therein). In fact, the near independence between kobs and ρm in NZVI systems is presumably caused by the corrosion rates reported for these nanomaterials that are orders of magnitude greater than those of larger Fe(0) particles.29−32 Evidence supporting greater NZVI passivation in systems with high ρm values was obtained by examining the influence of pH on the relationship between kobs and ρm. Figure S1 of the Supporting Information shows that kobs values exhibit a greater, albeit still nonlinear, dependence on NZVI loading at pH 6, at which the solubility of Fe(II) is orders of magnitude greater than that at pH 8. This also agrees with our earlier work with Fisher electrolytic iron27 in which the relationship between kobs for 1,1,1,-trichloroethane reduction and ρm became essentially first-order at low pH. Thus, as pH decreases, constraints on Fe(II) solubility at high ρm values are relieved, slowing corrosion product accumulation on the NZVI surface and in turn producing a more linear, nearly first-order dependence of kobs on ρm. Trends in Mass-Normalized Rate Constants for 1,1,1,2-TeCA Reduction by Pd/NZVI. For 0.05% Pd/ NZVI, relatively constant km(1,1,1,2-TeCA) values were measured over the entire range of molar ratios investigated. Moreover, km(1,1,1,2-TeCA) values in 0.05% Pd/NZVI systems were essentially equivalent whether data were collected experimentally by varying ρm or the initial 1,1,1,2-TeCA concentration. Notably, the agreement in km(1,1,1,2-TeCA) values holds even at low 1,1,1,2-TeCA to 0.05% Pd/NZVI ratios, at which km(1,1,1,2-TeCA) values for NZVI systems diverged considerably [e.g., compare km(1,1,1,2-TeCA) values for NZVI and 0.05% Pd/NZVI at molar ratios of 0.02−0.03]. The consistency of km(1,1,1,2-TeCA) values measured in 0.05% Pd/NZVI systems suggests that Pd deposition produces 7943

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

up to ∼100, while the second displays larger SKIEs with a much steeper rate of increase once REFs exceed 100. We believe the regime of very large REFs (≥100) and SKIEs (≥5) corresponds to conditions under which some form of atomic hydrogen (i.e., either adsorbed or absorbed H) is directly involved in 1,1,1,2-TeCA reduction. As in Figure S4 of the Supporting Information, the largest SKIEs were observed at high Pd loadings and high ρm values. In closed batch systems, these conditions promote the formation and accumulation of atomic hydrogen, a product of iron corrosion and intermediate in H2 evolution, at the reductant particle surface. We note that previously Schrick et al.13 reported an SKIE of 14 for trichloroethene (TCE) reduction by Ni/NZVI, which they also interpreted as evidence of the involvement of atomic hydrogen. In this high-reactivity regime, ours are among the largest solvent kinetic isotope effects ever reported in the literature. We believe this is likely due to an additive isotope effect that accrues via sequential steps in 1,1,1,2-TeCA reduction where H2O is preferred over D2O. Possible reaction steps involving the solvent include formation of atomic H(D) via solvent reduction, potential uptake of H(D) into the Pd lattice to promote hydride (deuteride) formation, and subsequent reaction of the active form of H(D) with 1,1,1,2-TeCA. If the switch from H2O to D2O slows each of these processes to some degree, it is not difficult to envision a rather large SKIE for 1,1,1,2-TeCA reduction. For example, Figure S5 of the Supporting Information shows that Fe(II) dissolution rates arising from solvent reduction are dramatically slower in D2O than in H2O. Accordingly, if solvent reduction to form atomic H(D) is the necessary first step in the reaction chain, then 1,1,1,2-TeCA reduction would be expected to be significantly slower in D2O. For the regime of small to negligible SKIEs, which corresponds to REFs up to at least 10, an alternative process must enhance 1,1,1,2-TeCA reduction without the involvement of atomic H. As shown in Figure 4 and consistent with previous reports,14 displacement plating produces numerous 3−4 nm Pd particles deposited nearly uniformly on the NZVI surface at both 0.05 and 0.35% Pd. We propose these Pd nanodeposits simply represent more efficient sites for electron transfer from the underlying Fe(0) to 1,1,1,2-TeCA at the solution interface than the surface oxides that coat NZVI particles at pH 8. While it is tempting to attribute reactivity enhancement in this lowSKIE regime to galvanic couple formation, we reiterate that Pd deposition does not significantly enhance Fe(0) corrosion (Figure S2 of the Supporting Information). Instead, we note that correlations between kobs(1,1,1,2-TeCA) and Fe(II) dissolution rates, while decidedly nonlinear for NZVI, exhibit a first-order relationship for Pd/NZVI systems within this lowSKIE regime (Figure S6 of the Supporting Information). This first-order dependence implies that a greater fraction of the electrons generated by Fe(0) oxidation to Fe(II) are transferred to 1,1,1,2-TeCA in Pd/NZVI systems, presumably because electron transfer is less efficient when it occurs through mixedvalent oxide layers present on NZVI. This scenario is consistent with Figure 1b, in which Pd deposition makes the NZVI surface more uniformly active at moderate Pd loadings. Accordingly, fresh Pd/NZVI appears to be less prone to the loss of reactivity from surface oxides, at least initially, particularly at high ρm values where surface oxides most extensively accumulate on NZVI in our batch systems.

Figure 3 more closely examines the relationship between measured SKIEs and corresponding REFs observed in different

Figure 3. Relationship between the solvent kinetic isotope effect [SKIE = kobs(H2O)/kobs(D2O)] and reactivity enhancement factor [REF = kobs(Pd/Fe)/kobs(Fe)] for reduction of (a) 1,1,1,2-TeCA and (b) cis-DCE. In panel b, data for 1,1,1,2-TeCA are provided for comparison (within the dashed box). Data were collected at different ρm values in various Fe(0) and Pd/Fe reductant systems, as indicated. The majority were collected at pH 8, although data in panel a are also presented for NZVI to illustrate the SKIE associated with the rate increase achieved with a decrease in pH from 8 to 6 (empty green circles). For pH 6 data, the REF was calculated by normalizing the kobs(1,1,1,2-TeCA) value at pH 6 to the corresponding value measured in an otherwise identical suspension at pH 8.

Pd/NZVI systems. Data are shown for 0.05 and 0.35% Pd/ NZVI reacted over a range of ρm values at pH 8. Generally, SKIEs increase with REF. However, there are two distinct regimes for SKIEs over the range of REFs achieved; the first is characterized by small SKIEs that slightly increase with REFs 7944

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

Figure 4. Representative TEM images of (a) NZVI, (b) 0.05% Pd/NZVI, and (c) 0.35% Pd/NZVI. Images are shown for freshly prepared NZVI and Pd/NZVI and particles aged for 40 days in 2 g/L suspensions at pH 8. Arrows indicate regions on freshly prepared 0.05% Pd/NZVI where Pd was not evident. The diameter of Pd deposits (dPd, determined by measuring the size of at least 100 particles via TEM) is provided for 0.05 and 0.35% Pd/NZVI particles.

sorption onto the septa used to seal reactors. This observation is in contrast to that with 1,1,1,2-TeCA, for which SKIEs were not observed for NZVI. In Pd/NZVI systems, rather large SKIEs were observed for cis-DCE reduction. As with 1,1,1,2-TeCA, SKIEs were not constant but increased with an increase in Pd loading and Pd/ NZVI concentration (Figure S9 of the Supporting Information). Using the initial kobs(cis-DCE) measured in NZVI systems [i.e., the first 60 min of decay (Figure S8 of the Supporting Information)], SKIEs for cis-DCE reduction by Pd/ NZVI were quantified and assessed as a function of REF. Figure 3b (see also Figure S10 of the Supporting Information) shows that SKIEs for cis-DCE follow trends observed for 1,1,1,2TeCA as a function of REF, with low SKIEs corresponding to relatively low REFs and SKIEs increasing rapidly, ultimately exceeding 100, as REFs approached 1000. The general agreement for 1,1,1,2-TeCA and cis-DCE in Figure 3b suggests that the same reduction mechanisms are operative for both compounds; as previously hypothesized, low SKIEs likely correspond to cis-DCE reduction primarily via electron transfer through Pd from Fe(0), whereas larger SKIEs are consistent with atomic hydrogen being the primary reductant. However, measured SKIEs at a specific Pd loading and ρm were consistently considerably greater for cis-DCE than for 1,1,1,2-TeCA. For example, 0.5 g/L suspensions of 0.35% Pd/NZVI yielded SKIEs for 1,1,1,2-TeCA and cis-DCE of 20 and 100, respectively, despite comparable REFs. In otherwise identical experimental systems, where we assume concentrations of surface-associated atomic hydrogen are equivalent, this may simply reflect a greater affinity for the reaction of atomic hydrogen at the sp2-hybridized centers of cis-DCE.

In this scenario, deposited Pd nanoparticles essentially act in a manner analogous to processes that minimize the impact of passive surface oxides on NZVI reactivity. Traditionally, this is accomplished by acid washing or working at lower pH values where oxide solubility is greater. Accordingly, we examined the rate increase and corresponding SKIE achieved when the pH of the NZVI suspension decreased from 8 to 6 [i.e., REFpH = kobs(pH 6)/kobs(pH 8)]. For the several ρm values investigated at pH 6, this pH decrease produced an up to 6-fold increase in NZVI reactivity relative to reactivity in otherwise identical systems at pH 8 (Figure 3a). In contrast, SKIEs did not increase significantly above unity in pH 6 systems. Thus, thinning the surface oxide layer on NZVI by lowering the suspension pH yields REFs and SKIEs comparable to those measured for Pd/NZVI systems with low Pd concentrations and low ρm values. Finally, despite strong evidence of multiple mechanisms by which 1,1,1,2-TeCA reduction is enhanced by Pd/NZVI, we reiterate there was no observable change in reduction product formation across these reactivity regimes. Reduction of 1,1,1,2TeCA consistently yielded 1,1-DCE via reductive β-elimination in both NZVI and all Pd/NZVI systems (Figure S7 of the Supporting Information). The only difference for Pd/NZVI was that 1,1-DCE was subsequently, rapidly degraded to ethane and ethylene, transformations not observed in NZVI systems. Solvent Kinetic Isotope Effects for cis-DCE Reduction by NZVI and Pd/NZVI. For cis-DCE reduction by NZVI, a SKIE was observed but not quantifiable. As shown in Figure S8 of the Supporting Information for a 0.6 g/L NZVI suspension, cis-DCE reduction occurred in H2O, but cis-DCE loss in D2O was essentially equivalent to that observed in NZVI-free controls, where modest cis-DCE decay (∼1%/h) occurred via 7945

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

Solvent Kinetic Isotope Effects in ZVI Systems. For comparison to NZVI-based reductants, a limited number of experiments explored SKIEs for 1,1,1,2-TeCA and cis-DCE reduction in micrometer-sized ZVI and Pd/ZVI systems. For 1,1,1,2-TeCA reduction by ZVI, equivalent rates were observed in H2O and D2O (Figure S11a of the Supporting Information), consistent with results for NZVI. For 1% Pd/ZVI, which enhanced 1,1,1,2-TeCA reduction by roughly 10-fold (Figure S11b of the Supporting Information), an essentially negligible SKIE of 1.6 was observed, consistent with trends in SKIEs and REFs for Pd/NZVI (see Figure 3a). Notably, however, Cwiertny et al.11 previously quantified that this REF of 10 is a maximum for alkyl polyhalide reduction by this particular Pd/ ZVI formulation. Thus, Pd/ZVI reactivity toward 1,1,1,2-TeCA appears to be constrained to the low-SKIE regime in Figure 3a, implying that atomic hydrogen plays a negligible role in these systems. As reported previously,11 the rate of cis-DCE reduction by this lot of Fisher electrolytic iron was too slow to furnish a measurable rate constant; thus, SKIEs could not be measured in ZVI suspensions. However, in systems with 1% Pd/ZVI, cisDCE was rapidly degraded, exhibiting a half-life of 6 min (Figure S12 of the Supporting Information). Using estimates of the rate constant for cis-DCE reduction by ZVI described by Cwiertny et al.,7 this corresponds to a REF of at least 3000, a value far exceeding the REF for 1% Pd/ZVI reduction of 1,1,1,2-TeCA and any REFs measured with Pd/NZVI. Despite this very large REF, a significant but relatively small SKIE of 5.3 was measured via comparison of cis-DCE decay in H2O and D2O (Figure S12 of the Supporting Information). In Figure 3b, this SKIE and this REF deviate considerably from the behavior of NZVI-based reductants, suggesting that different molecularlevel processes likely govern cis-DCE reduction in ZVI-based reductant systems, although the same reduction products (ethane and ethane) were observed in each case. Specifically, we interpret the intermediate SKIE obtained at such a high REF as evidence of a diminished role for atomic hydrogen relative to other available pathways for promoting cis-DCE reduction in Pd/ZVI systems. Influence of Pd/NZVI Aging on the Mechanisms for 1,1,1,2-TeCA Reduction. Our results support multiple mechanisms for chlorinated solvent reduction, involving either direct electron transfer or reaction with atomic hydrogen, in Pd/NZVI systems. To evaluate if the relative importance of these mechanisms evolves over time during Pd/NZVI application, we measured SKIEs for NZVI and 0.35% Pd/ NZVI particles aged for up to 1 month at pH 8. As shown in Figure 5, the reactivity of 0.35% Pd/NZVI measured in both H2O and D2O solutions decreased nearly 50-fold over 2 weeks, consistent with previous reports,14,33 but then remained essentially constant for the remainder of the month. In contrast, the reactivity of NZVI, while initially far lower than that of 0.35% Pd/NZVI, exhibited only modest decreases over 30 days. As anticipated from earlier results, negligible SKIEs were observed in NZVI systems over the duration of aging. In contrast, significant SKIEs (≥5) were measured for 0.35% Pd/ NZVI during the first 5 days of aging, but SKIEs decreased over time, ultimately approaching the value observed for aged NZVI. Notably, the decrease in the SKIE for Pd/NZVI coincides with the time scales of greatest reactivity loss. On the basis of measured SKIEs, atomic hydrogen species appear to contribute to Pd/NZVI reactivity only over 1 week.

Figure 5. Values of kobs for 1,1,1,2-TeCA reduction vs aging time for NZVI (blue) and 0.35% Pd/NZVI (red) in H2O (filled symbols) and D2O (empty symbols). Corresponding SKIEs are provided adjacent to data points. Suspensions (2 g/L) were aged and reacted at pH 8 (25 mM HEPES and 25 mM NaCl). Reactors contained a ρm of 0.22 g/L and an initial 1,1,1,2,-TeCA concentration of 75 μM.

The concomitant decrease in reactivity and SKIEs over this interval likely arises from the growth of surface oxides that bury the deposited Pd, similar to aging effects reported elsewhere,14,17 and in turn limits the production, accumulation, and solution access of atomic hydrogen at the Pd surface. Such a scenario is supported by TEM images (Figure 4), in which Pd deposits initially visible on the NZVI surface were no longer apparent after 40 days of aging, coated by oxides that accumulated on the reductant surface over time. Accordingly, we attribute the long-term reactivity enhancement observed for 0.35% Pd/NZVI to small amounts of Pd left exposed to solution that represent sites for electron transfer more ideal than the surface of the thickening iron oxide film. Given prior reports that zerovalent metal deposits can enhance the reducing capacity of mixed-valent iron oxides,34 we also cannot rule out the possibility that the Pd nanodeposits enhance electron conduction through the oxide that develops on the NZVI surface over longer aging times. Notably, electron transfer from Fe(0) through a conductive surface oxide is also the most probable mechanism responsible for 1,1,1,2-TeCA reduction in aged NZVI systems, consistent with their negligible SKIEs during aging. A schematic summarizing the proposed mechanism and aging pathways as a function of Pd loading is shown in Figure S13 of the Supporting Information. Implications for Pd/Fe Use for Subsurface Cleanup. Here we show that during application of Pd/NZVI, multiple, parallel mechanisms for chlorinated solvent transformation are potentially at play. Further, the nature of the reductant particle (e.g., Pd loading), the type of pollutant target (e.g., alkyl vs vinyl polyhalide), and characteristics of the subsurface treatment system (e.g., Pd/NZVI concentration) will dictate the relative importance of each mechanism. Practically, the mechanism involving atomic hydrogen is most desirable, as it coincides with the greatest rates of pollutant transformation. The participation of atomic hydrogen will be favored at early 7946

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology



operation time scales in treatment zones utilizing high Pd loadings and Pd/NZVI subsurface concentrations. All indications herein support atomic hydrogen participating most extensively in Pd/NZVI systems and to a lesser extent, if at all, for unpalladized NZVI and reductants prepared from larger ZVI particles. The role of atomic hydrogen is most pronounced in Pd/NZVI systems presumably from the enhanced rates of corrosion exhibited by NZVI relative to larger commercially available iron particles.29−32 Many of our results, however, raise questions about the value of using Pd/Fe in field scale applications. To optimize performance, system attributes long thought to be disadvantageous from the standpoint of economics (e.g., high Pd loading and high ρm will increase material costs) must be embraced. For example, with identical sizes of Pd deposits on 0.05 and 0.35% Pd/NZVI measured via TEM analysis (Figure 4), the total Pd mass in the system dictates accumulation of hydrogen. Further, our results show that at high chlorinated solvent to reductant ratios, which would be typical of most saturated contaminant plumes, there is little reactivity difference between Pd/NZVI and NZVI, potentially indicating surface site limitation in Pd/ NZVI systems that would negate any benefit of their use. A final point of concern relates to system conditions required for atomic hydrogen formation. All experiments herein were conducted in closed batch systems, which allow Fe(0) corrosion products including atomic or molecular hydrogen to accumulate. Indeed, high SKIEs indicative of atomic hydrogen participation were observed in only Pd/NZVI systems (e.g., high ρm) in which corrosion products can accumulate most extensively. During field applications, it remains to be seen if the levels of atomic hydrogen needed to attain optimal reactivity can be achieved under more dynamic, flow-through conditions, which will likely alter the driving force for hydrogen accumulation on or within Pd. Thus, while our results shed new light on the possible mechanisms driving contaminant transformation at Pd/Fe surfaces, future efforts must demonstrate that the favorable reactivity of bimetals that can be achieved in the laboratory can also be achieved and sustained in the field. To date, the limited availability of performance data from existing Pd/NZVI field scale installations precludes such an analysis.



REFERENCES

(1) Wang, C.-B.; Zhang, W.-x. Synthesizing nanoscale iron particles for rapid and complete dechlorination of TCE and PCBs. Environ. Sci. Technol. 1997, 31, 2154−2156. (2) Elliott, D. W.; Zhang, W.-x. Field assessment of nanoscale bimetallic particles for groundwater treatment. Environ. Sci. Technol. 2001, 35, 4922−4926. (3) Korte, N. E.; Zutman, J. L.; Schlosser, R. M.; Liang, L.; Gu, B.; Fernando, Q. Field application of palladized iron for the dechlorination of trichloroethene. Waste Manage. 2000, 20, 687−694. (4) Karn, B.; Kuiken, T.; Otto, M. Nanotechnology and in situ remediation: A review of the benefits and potential risks. Environ. Health Perspect. 2009, 117, 1823−1831. (5) Mueller, N. C.; Braun, J.; Bruns, J.; Cernik, M.; Rissing, P.; Rickerby, D.; Nowack, B. Application of nanoscale zero valent iron (NZVI) for groundwater remdiation in Europe. Environ. Sci. Pollut. Res. 2012, 19, 550−558. (6) Nanotechnology for Site Remediation: Fact Sheet; EPA 542-F08-009; U.S. Environmental Protection Agency: Washington, DC, 2008 (http://www.epa.gov/tio/download/remed/542-f-08-009.pdf; accessed March 24, 2013). (7) Cwiertny, D. M.; Bransfield, S. J.; Livi, K. J. T.; Fairbrother, D. H.; Roberts, A. L. Exploring the influence of granular iron additives on 1,1,1-trichloroethane reduction. Environ. Sci. Technol. 2006, 40, 6837− 6843. (8) Lin, C. J.; Lo, S. L.; Liou, Y. H. Dechlorination of trichloroethylene in aqueous solution by noble metal-modified iron. J. Hazard. Mater. 2004, 116, 219−228. (9) Lien, H.-L.; Zhang, W.-X. Nanoscale Pd/Fe bimetallic particles: Catalytic effects of palladium on hydrodechlorination. Appl. Catal., B 2007, 77, 110−116. (10) Lien, H.-L.; Zhang, W.-x. Nanoscale iron particles for complete reduction of chlorinated ethenes. Colloids Surf., A 2001, 191, 97−105. (11) Cwiertny, D. M.; Bransfield, S. J.; Roberts, A. L. Influence of the oxidizing species on the reactivity of iron-based bimetallic reductants. Environ. Sci. Technol. 2007, 41, 3734−3740. (12) Xu, Y.; Zhang, W.-x. Subcolloidal Fe/Ag particles for reductive dehalogenation of chlorinated benzenes. Ind. Eng. Chem. 2000, 39, 2238−2244. (13) Schrick, B.; Blough, J. L.; Jones, A. D.; Mallouk, T. E. Hydrodechlorination of trichloroethylene to hydrocarbons using bimetallic nickel iron nanoparticles. Chem. Mater. 2002, 14, 5140− 5147. (14) Yan, W.; Herzing, A. A.; Li, X.-q.; Kiely, C. J.; Zhang, W.-x. Structural evolution of Pd-doped nanoscale zero-valent iron (nZVI) in aqueous media and implications for particle aging and reactivity. Environ. Sci. Technol. 2010, 44, 4288−4294. (15) Cheng, I. F.; Fernando, Q.; Korte, N. Electrochemical dechlorination of 4-chlorophenol to phenol. Environ. Sci. Technol. 1997, 31, 2443−2443. (16) Fang, Y.; Al-Abed, S. R. Correlation of 2-chlorobiphenyl dechlorination by Fe/Pd with iron corrosion at different pH. Environ. Sci. Technol. 2008, 42, 6942−6948. (17) Chun, C. L.; Baer, D. R.; Matson, D. W.; Amonette, J. E.; Penn, R. L. Characterization and reactivity of iron nanoparticles prepared with added Cu, Pd, and Ni. Environ. Sci. Technol. 2010, 44, 5079− 5085. (18) Bransfield, S. J.; Cwiertny, D. M.; Livi, K.; Fairbrother, D. H. Influence of transition metal additives and temperature on the rate of organohalide reduction by granular iron: Implications for reaction mechanisms. Appl. Catal., B 2007, 76, 348−356. (19) Li, T.; Farrell, J. Reductive dechlorination of trichloroethene and carbon tetrachloride using iron and palladized-iron cathodes. Environ. Sci. Technol. 2000, 34, 173−179. (20) Carroll, F. A. Perspectives on Structure and Mechanism in Organic Chemistry; Brooks Cole: New York, 1998. (21) Xie, Y.; Cwiertny, D. M. Use of dithionite to extend the reactive lifetime of nanoscale zero-valent iron treatment systems. Environ. Sci. Technol. 2010, 44, 8649−8655.

ASSOCIATED CONTENT

* Supporting Information S

More detailed materials and methods and results of characterization and reactivity studies with Pd/Fe reductants. This material is available free of charge via the Internet at http:// pubs.acs.org.



Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (319) 335-1401. Fax: (319) 335-5660. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We acknowledge the Center for Advanced Microscopy and Microanalysis at the University of California (Riverside, CA) for assistance with TEM image collection. 7947

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948

Environmental Science & Technology

Article

(22) Xie, Y.; Cwiertny, D. M. Influence of anionic cosolutes and pH on nanoscale zerovalent iron longevity: Time scales and mechanisms of reactivity loss toward 1,1,1,2-tetrachloroethane and Cr(VI). Environ. Sci. Technol. 2012, 46, 8365−8373. (23) Perrin, D. D.; Dempsey, B. Buffers for pH and Metal Ion Control; John Wiley and Sons: New York, 1974. (24) Stucki, J. W.; Anderson, W. L. The quantitative assay of minerals for Fe2+ and Fe3+ using 1,10-phenanthroline. I. Sources of variability. Soil Sci. Soc. Am. J. 1981, 45. (25) Stucki, J. W. The quantitative assay of minerals for Fe2+ and Fe3+ using 1,10-phenanthroline. II. A photochemical method. Soil Sci. Soc. Am. J. 1981, 45, 638−641. (26) Wei, J.; Xu, X.; Liu, Y.; Wang, D. Catalytic hydrodechlorination of 2,4-dichlorophenol over nanoscale Pd/Fe: Reaction pathway and some experimental parameters. Water Res. 2006, 40, 348−354. (27) Cwiertny, D. M.; Roberts, A. L. On the nonlinear relationship between kobs and reductant mass loading in iron batch systems. Environ. Sci. Technol. 2005, 39, 8948−8957. (28) Nurmi, J. T.; Trantyek, P. G.; Baer, D. R.; Amonette, J. E.; Pecher, K.; Wang, C.; Linehan, J. C.; Matson, D. W.; Penn, R. L.; Driessen, M. D. Characterization and properites of metallic iron nanoparticles. Environ. Sci. Technol. 2005, 39, 1221−1230. (29) Reardon, E. J.; Fagan, R.; Vogan, J. L.; Przepiora, A. Anaerobic corrosion reaction kinetics of nanosized iron. Environ. Sci. Technol. 2008, 42, 2420−2425. (30) Reardon, E. J. Anaerobic corrosion of granular iron: Measurement and interpretation of hydrogen evolution rates. Environ. Sci. Technol. 1995, 29, 2936−2945. (31) Reardon, E. J. Zerovalent irons: Styles of corrosion and inorganic control on hydrogen pressure buildup. Environ. Sci. Technol. 2005, 39, 7311−7317. (32) Chen, K.-F.; Li, S.; Zhang, W.-x. Renewable hydrogen generation by bimetallic zero valent iron nanoparticles. Chem. Eng. J. 2011, 170, 562−567. (33) Zhu, B.-W.; Lim, T.-T. Catalytic reduction of chlorobenzenes with Pd/Fe nanoparticles: Reactive sites, catalyst stability, particle aging, and regeneration. Environ. Sci. Technol. 2007, 41, 7523−7529. (34) O’loughlin, E. J.; Kemner, K. M.; Burris, D. R. Effects of AgI, AuIII, and CuII on the reductive dechlorination of carbon tetrachoride by green rust. Environ. Sci. Technol. 2003, 37, 2905−2912.

7948

dx.doi.org/10.1021/es401481a | Environ. Sci. Technol. 2013, 47, 7940−7948