Porous Substoichiometric TiO2 Anodes as Reactive Electrochemical

May 20, 2013 - anodic reactive electrochemical membrane (REM) for water ... REM consists of a porous substoichiometric titanium dioxide (Ti4O7) tubula...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/est

Porous Substoichiometric TiO2 Anodes as Reactive Electrochemical Membranes for Water Treatment Amr M. Zaky† and Brian P. Chaplin†,* †

Department of Civil and Environmental Engineering and Villanova Center for the Advancement of Sustainable Engineering, Villanova University, Villanova, Pennsylvania 19085, United States S Supporting Information *

ABSTRACT: This research investigates the characterization and testing of an anodic reactive electrochemical membrane (REM) for water treatment. The REM consists of a porous substoichiometric titanium dioxide (Ti4O7) tubular, ceramic electrode operated in cross-flow filtration mode. Advection-enhanced mass transfer rates, on the order of a 10-fold increase, are obtained when the REM is operated in filtration-mode, relative to a traditional flow-through mode. Oxidation experiments with model organic compounds showed that the REM was active for both direct oxidation reactions and formation of hydroxyl radicals (OH•). Electrochemical impedance spectroscopy data interpreted by transmission line modeling determined that the electro-active surface area was 619 times the nominal geometric surface area. Results from filtration-mode experiments with p-methoxyphenol indicate that compound removal occurred by electro-assisted adsorption and subsequent oxidation. Electro-assisted adsorption was the primary removal mechanism at potentials where OH• did not form. At higher potentials (>2.0 V), where OH• concentrations were significant, p-methoxyphenol removal occurred by a combination of electro-assisted adsorption and OH• oxidation. These removal mechanisms resulted in 99.9% p-methoxyphenol removal in the permeate, with calculated current efficiencies >73% at applied current densities of 0.5−1.0 mA cm−2. These results illustrate the extreme promise of the REM for water treatment.



R → (R•)+ + e−

INTRODUCTION The integration of membrane and advanced oxidation process (AOP) technologies into a single reactive membrane unit has many potential applications in water treatment and could be useful in increasing the efficiency of water recycling. Current approaches include depositing a TiO2 photocatalyst on a membrane surface,1−9 and using conductive porous electrodes.10−12 Using a TiO2 photocatalyst on a membrane surface requires UV light to oxidize water to hydroxyl radicals (OH•).1−9 While these studies were successful at the labscale, the placement of UV lamps around membranes would result in potentially high capital cost and a large reactor footprint. Additionally, the high absorption of UV in water also increases the cost of photocatalytic treatment. The development of a reactive electrochemical membrane (REM) is a potentially transformative technology that integrates physical separation with electrochemical oxidation. Electrochemical oxidation is an emerging AOP, where water is oxidized to OH• on the anode surface (eq 1). H 2O → OH • + H+ + e−

Direct oxidation can degrade recalcitrant compounds that are unreactive toward OH• (e.g., fluorinated organics).14 The key to the development of a REM is the use of porous electrode materials that are active for both reactions 1 and 2. A promising electrode material is porous Ti4O7, which is known as substoichiometric titanium dioxide. The conversion of TiO2 to Ti4O7 is accomplished at temperatures above 900 °C and under a H2 atmosphere.15,16 The conversion of TiO2 to Ti4O7 has been confirmed by XRD and electrical conductivity measurements.15,16 The conductivity of Ti4O7 (166 Ω−1cm−1) is many orders of magnitude greater than TiO2 (∼10−9 Ω−1cm−1).15,16 The water treatment potential of Ti4O7 electrodes is promising, although few studies on this material exist.17−23 Studies have shown that Ti4O7 electrodes behave as inactive electrodes and thus produce physisorbed OH• via water oxidation (eq 1),17,23 and are also active for direct electron transfer reactions (eq 2).22 In contrast, other inactive anodes, such as boron-doped diamond electrodes are expensive and thin films coated onto a porous electrically conductive

(1)

These OH• react unselectively with a wide range of recalcitrant organics, often at diffusion-limited rates.13 An additional mechanism during anodic oxidation is direct oxidation, where an electron is transferred from the contaminant (R) to the anode (eq 2). © XXXX American Chemical Society

(2)

Received: March 23, 2013 Revised: May 17, 2013 Accepted: May 20, 2013

A

dx.doi.org/10.1021/es401287e | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Environmental Science & Technology

Article

electrolyte was chosen because it is nonreactive under anodic and cathodic conditions.33,34 Experiments with oxalic acid and Fe(CN)64‑ were conducted with a feed concentration of 5 mM and solution volume of 1 L. The feed and permeate solutions were 100% recycled for 30 60 min without an applied current. After this time, a current density between 0.203.50 mA cm−2 was applied to the REM cell, and the permeate was no longer recycled and was collected in a separate reservoir. Experiments with p-methoxyphenol were performed for 120 min at a concentration of 1 mM and a solution volume of 0.5 L. Both the feed and permeate streams were 100% recycled to the feedwater reservoir. A constant current density (01.0 mA cm−2) was applied to the cell after 5 min of filtration. For all experiments, the electrode potential and permeate flux were monitored. Samples from the feed and permeate were collected and analyzed for organic compound concentration and chemical oxygen demand (COD). After each experiment, the REM was cleaned to remove and quantify adsorbed compounds, as described in the SI. Previous research has suggested that Ti4O7 electrodes can be oxidized to TiO2 during anodic polarization leading to passivation.18,20,35 However, repeatable results were observed for duplicate experiments and measured oxidation rates did not decline as a function of electrode use, indicating that passivation did not occur. Analytical Methods. Concentrations of p-methoxyphenol and p-benzoquinone were measured using HPLC with a C18 column and a photodiode array detector (wavelength = 314 nm) (Shimadzu SPD-M20A). The mobile phase was methanol:water (50:50). Concentrations of oxalic acid were measured using ion chromatography with a conductivity detector (Dionex DX-600). The concentrations of Fe(CN)64− and Fe(CN)63− were determined by UV−vis spectrophotometry (Spectronic Genesys 5) at wavelengths of 303 and 420 nm, respectively. COD concentrations were determined by Hach method 8000. Electrochemical Impedance Spectroscopy (EIS) and Transmission Line Modeling (TLM). Potentiostatic EIS measurements were performed on the REM cell shown in Figure S-1 of the SI, to characterize the electro-active surface area. EIS measurements were made at potentials of 0.81.0 V with an amplitude of ±10 mV and over a frequency range of 0.1 to 1 × 103 Hz. Measurements were made in a 1 M NaClO4 electrolyte and under identical hydrodynamic conditions as cross-flow filtration experiments (Q = 36 L h−1 and J = 50 L m−2 h−1). A TLM36 was fit to the EIS data using Gamry Echem Analyst V 6.0.1 software. A description of the TLM and fits to experimental data are provided in the SI (Figure S-2 and Table S-1). The TLM contains separate variables that account for the resistance to charge transfer (rct,o (ohm)) and double-layer capacitance (Cdl,o (F)) of the outer electrode surface and the resistance to charge transfer (rct,p (ohm)) and double-layer capacitance (Cdl,p (F)) of the inner REM pores. Decoupling the impedance allows a quantitative estimate of the electro-active surface area at these locations. The outer electro-active surface area (Ao) is the region of the REM that does not experience a significant potential gradient, unlike the surface area of the inner pores (Ap), where potential gradients are significant.37 Both Ao and Ap were determined using previous methods.38 Values obtained for Cdl,i were divided by the average doublelayer capacitance of oxide electrodes (60 μF cm−2).39 The

membrane are needed to make them cost-effective. Additional materials used for electrochemical filters include carbon nanotubes (CNTs)24,25 and carbon membranes (graphite) coated with electro-active catalysts.10,11 CNTs have good electrical conductivity but evidence does not exist that they produce OH•. Graphite anodes are not electrochemically stable, as the carbon is eventually oxidized to CO2,26−29 which would result in material deterioration and catalyst leaching. In this study, a porous tubular Ti4O7 electrode is used in filtration-mode in order to create a REM. The purpose of this study was to characterize the physical and electrochemical properties of the REM. The porosity and electro-active surface area of the REM are characterized using scanning electron microscopy, Hg porosimetry, and electrochemical impedance spectroscopy. Specific probe molecules (oxalic acid, pbenzoquinone, and Fe(CN)64−) are used to assess the potential of the REM for direct oxidation and OH• generation. The oxidation efficiency of the REM is assessed for p-methoxyphenol removal, which is a recalcitrant organic contaminant found in industrial wastewater.30 Experimental results are used to elucidate a preliminary mechanism for p-methoxyphenol removal with the REM.



MATERIALS AND METHODS Reagents. Chemicals were reagent-grade and obtained from Fisher Scientific and Sigma-Aldrich. All chemicals were used as received. All solutions were made from Milli-Q ultrapure water (18.2 MΩ cm at 21 °C). Reactive Electrochemical Membrane. The REM was constructed from a tubular Ebonex electrode, with outer and inner diameters of 28 and 20 mm, respectively (Vector Corrosion Technologies, Inc.). Ebonex is a substoichiometric TiO2 material consisting primarily of the Ti4O7 suboxide.16 Physical Characterization. The REM was analyzed by scanning electron microscopy (SEM), and Hg porosimetry. Porosimetry analysis was performed by Micrometric Analytical Services (Norcross, GA) according to method ISO 15901-1.31 A Hitachi S-4800 cold field emission SEM was used to investigate the REM morphology and surficial pore size. Cross-flow Filtration Setup. The cross-flow filtration unit is shown in the Supporting Information, SI, Figure S-1. The Ebonex electrode (anode) was utilized in filtration mode, and a 3.18 mm diameter 316 stainless steel rod (cathode) was arranged in the center of the REM. The anode and cathode were 9.8 mm apart (Figure S-1 of the SI). Feed water was pumped through the REM at a constant cross-flow rate (Q = 36 L h−1), using a bench analog drive gear pump (Cole Parmer) at constant back-pressure (10 psi). A constant current density (03.5 mA cm−2) was applied to the REM cell using a programmable direct current (DC) power supply (Proteck P6035). Potentials were measured versus a Ag/AgCl reference electrode (Warner Instruments LLC), located 1 mm from the REM surface, using a Gamry Reference 600 potentiostat/ galvanostat. All potentials were reported versus the standard hydrogen electrode (SHE). Cross-flow Filtration Experiments. The mass transfer rate constant (km) to the REM surface was characterized using the limiting current density approach,32 as described in the SI. Duplicate cross-flow filtration experiments were performed using the reactor shown in Figure S-1 of the SI, at a temperature of 21 ± 2 °C. Individual organic compounds (oxalic acid, Fe(CN)64−, p-methoxyphenol) were added to a 10 mM NaClO4 supporting electrolyte. The NaClO4 supporting B

dx.doi.org/10.1021/es401287e | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Environmental Science & Technology

Article

roughness factor was determined by dividing (Ao + As) by the geometric surface area. Kinetic Analysis. The removal rates of compound i from the permeate (rp,i) was calculated by the following:

rP,i = J(C F,i − C P,i)

(3)

where J is the permeate flux (L m−2 h−1), and CF,i and CP,i are the concentrations of the feed and permeate, respectively (mmole L−1). The percent removal (Ri) in the permeate was calculated as follows: Ri =

(C F,i − C P,i) C F,i

*100 (4)

The current efficiency of the permeate (CEp,i (%)), current directed toward oxidation of the substrate that passes through the REM, was calculated by eq 5 CE P,i =

JFz(C F,i − C P,i) j

*100

(5) −2

where j is the applied current density (A m ). The current efficiency related to the oxidation of compounds recycled in the feed (CEF,i (%)) was calculated by eq 6. CE F,i =

FzrF,i j

*100

(6)

where rF,i is the removal rate in the feed determined by linear regression of concentration profiles with time. The overall current efficiency (CEO,i (%)) was determined by summation of CEP,i and CEF,i. The overall current efficiency for COD removal (CECOD (%)) was calculated by eq 7. CECOD =

FV[CODox, f ] 8jAΔt

*100

(7)

where CODox,f is the concentration of COD oxidized during a given experiment (g L−1), V is the solution volume (L), Δt is the time of the experiment (s), and 8 is a dimensionless conversion factor.



RESULTS AND DISCUSSION Physical Characterization. The morphology of the REM surface was characterized using SEM and shows surficial pore sizes ranging between approximately 16 μm (Figure 1a). The pore structure of the REM was characterized more thoroughly using Hg porosimetry (Figures 1b,c). The REM exhibits a bimodal pore size distribution with ∼90% of the measured surface area associated with pores 1.3 × 106 L m−2 h−1 (3.6 × 10−1 m s−1) for an 880-μm electro-active depth. Rates of electron transfer may ultimately limit the applied membrane flux. Surface area normalized-rate constants for the one-electron oxidation of Fe(CN)64‑ of 10−3 m s−1 have been reported on various electrode materials.43−46 Since the oxidation of Fe(CN)64− is a facile outer sphere electron transfer reaction, J = 1 × 10−3 m s−1 (∼3000 L m−2 h−1) is considered an upper bound for membrane flux. The oxidation of other substrates may dictate lower fluxes. For example, a rate constant of 850 L m−2 h−1 for the oxidation of Nnitrosodimethylamine on boron-doped diamond electrodes has been reported.47 However, in applied settings, the REM will likely be advection-limited, as typical microfiltration permeate fluxes are 30170 L m−2 h−1.48 For subsequent analysis in this study, the mass transfer rate (rm,i) to the REM surface was calculated for species i using eq 8.

Assuming that only the REM micropores are electro-active, than the total electro-active surface area is approximately 22% of the microporous surface area, which corresponds to an 880μm electro-active depth. The true electro-active depth will fall between theses two extremes (60880 μm). Mass Transfer Determination. The mass transfer rate constant (km) was calculated as a function of the permeate flux (J) according to Eq S-1 (SI). The relationship between km and J are shown in Figure 2. Results indicate that for low values of J

rm,i = JC F,i = k mC F,i

(8)

Electrochemical Reactivity Characterization. Model organic probe molecules were used to assess the reactivity of the REM. The key physical and chemical properties of these molecules are listed in the SI (Table S-2), and include benzoquinone (BQ), oxalic acid (OA), Fe(CN)64−, and pmethoxyphenol (p-MP). BQ reacts readily with OH• (kOH•,p‑BQ = 1.2 × 109 M−1s−1),13 but is highly resistant to direct oxidation.23 Thus, BQ was used as an OH• probe. Although compounds with nitrone groups are commonly used for the detection of OH• in solution (e.g., N,Ndimethy-p-nitrosoaniline and 5,5-dimethyl-1-pyrroline-N-oxide (DMPO)), these compounds are not definitive OH• probes during electrochemical oxidation experiments.23 Nitrones are not resistant to direct oxidation and can react at the anode to form radical cations at lower potentials than OH• are formed.49−51 These radical cations are susceptible to nucleophilic attack by water to produce compounds that potentially yield false positive detections of OH•.23 Since BQ is reduced on the cathode to hydroquinone, and the REM setup is configured as an undivided electrochemical cell, BQ oxidation was conducted in a divided batch cell, where the anode and cathode were separated by a Nafion membrane. Results indicate that BQ was rapidly oxidized by the REM (SI Figure S-3), which are consistent with previous reports that Ti4O7 anodes generate OH•.17,23 Control experiments in the absence of an applied current confirmed that BQ did not adsorb to the REM surface, and organic compounds were not detected during the post experiment cleaning protocol. The oxidation rate of OA with OH• is slow (kOH•,p‑OA = 1.4 × 106 M−1s−1),13 and studies on various electrode materials have shown that OA reacts predominately via direct oxidation at potentials less than that necessary for significant water oxidation.52,53 Thus OA is used as a direct oxidation probe, that undergoes a 2e− direct oxidation mechanism,53 as shown in eq 9:

Figure 2. Relationship between measured normalized membrane flux (J) and calculated mass transfer rate constant (km). Error bars represent 95% confidence intervals. Solid line represents linear regression. Dashed line represents km for diffusion in the absence of a membrane flux, determined by Sherwood number correlation (Details provided in the SI).

between 9.014.9 L m−2 h−1 (2.54.1 × 10−6 m s−1) transport to the REM surface is controlled by diffusion, as measured km values are similar to the km estimate determined by the empirical Sherwood correlation for flow in a pipe (dashed line in Figure 2). A linear relationship is observed between J and km, for values of J between 30102 L m−2 h−1 (0.82.8 × 10−5 m s−1) (R2 = 0.994), and a slope close to unity (slope = 1.02). The linear relationship indicates that increasing the membrane flux results in advection-enhanced transport of Fe(CN)64− to the anode surface. An approximate 10-fold increase of km (0.32.6 × 10−5 m s−1) was obtained by increasing J from 9 to 102 L m−2 h−1. The slope close to unity indicates that advection controls mass transport to the REM surface, and the diffusion boundary layer present in typical cross-flow parallel plate electrode reactors was eliminated. The flow in the REM was characterized as laminar (Re = 1270). By contrast, using the Seider-Tate correlation for flow in a pipe, it is necessary to invoke turbulent conditions (Re > 20,000) to obtain km > 1.0 × 10−5 m s−1.40 Our results show greater mass transfer rate enhancements when compared to other studies. Results from studies utilizing electrochemical carbon nanotube flow-through reactors showed an approximate 2- to 6-fold increase in km relative to their respective batch systems,41,42 where km values as high as 1.7 × 10−5 m s−1 have been reported.41 Due to pressure drop through the 4 mm thick REM, values for J > 102 L m−2 h−1 (2.8 × 10−5 m s−1) were not tested. The time scales of advection (tA) and diffusion (tD) were calculated to confirm that advection was limiting the mass transport process (SI). Calculations were performed assuming electroactive depths of 60 and 880 μm. For J = 100 L m−2 h−1 it was determined that tA = 1.2 and 18 s for electro-active depths of 60 and 880 μm, respectively. For both cases tD = 723 μs. Diffusion limitations will occur at J > 8.3 × 104 L m−2 h−1 (2.3 × 10−2 m

C2H 2O4 → 2CO2 + 2H+ + 2e−

(9)

The reactivity of the REM for 5 mM of OA was characterized as a function of current density (03.50 mA cm−2), using the REM in Figure S-1 of the SI. Results from the experiments are shown in Figure 3 and summarized in Table 1a. The data for the first 60 min of the experiment were conducted without an applied current, and the similarity between permeate and feed concentrations indicates that adsorption of OA to the REM D

dx.doi.org/10.1021/es401287e | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Environmental Science & Technology

Article

transported through the membrane. Conditions in the REM leading to fluid mixing are discussed in the SI. The overall current efficiency (CEo.OA) of the anodic oxidation of OA remained high under all current densities tested, and was between 82.5 ± 5.05% and 108 ± 11% (Table 1a). These high CEo,OA values illustrate the extreme promise of the REM for water treatment. Additionally, OA typically accumulates during other AOPs, due to low reactivity with OH•,13 further illustrating the benefits of the REM. Reactivity of the REM was also assessed using Fe(CN)64−, as it reacts with OH• under mass transfer limited rates (kOH•,Fe(CN)64− = 1.1 × 1010 M−1s−1),13 undergoes direct oxidation via a facile 1 e− outer sphere electron transfer reaction at low potentials (eq 11), is nonadsorbing, and its sole product (Fe(CN)63−) is resistant to both OH• and direct oxidation reactions. Thus oxidation of Fe(CN)64− was used to assess reactivity of the REM without complications arising from daughter products interfering with parent compound oxidation. Fe(CN)64 − → Fe(CN)36 − + e−

(10)

The oxidation of Fe(CN) 6 4− was also tested at a concentration of 5 mM and current densities of 0.23.5 mA cm−2. Results are summarized in Table 1a, and individual experiments are shown in the SI (Figures S-4S-7). Both the percent removal of Fe(CN)64− in the permeate (RFe(CN)4− ) and 6 ) increased linearly removal rates in the permeate (rp,Fe(CN)4− 6 with current densities between 0.2 and 1.0 mA cm−2 (R2 = 0.991 and 0.980, respectively). Measured rp,Fe(CN)4− values at 6 these current densities were limited by the applied current, and it was observed that the current efficiency of the permeate stream CEp,Fe(CN)4− decreased from 105 ± 5.4% to 27.1 ± 1.3% 6 as the current density increased from 0.20 to 1.00 mA cm−2, respectively. At a current density of 3.50 mA cm−2 complete removal of Fe(CN)64− was observed in the permeate, and removal was limited by mass-transport (Table 1a). Removal rates of FeCN64− were also observed in the recycled feed stream rF,Fe(CN)4− , and increased as a function of current 6 density (Table 1a). The removal of Fe(CN)64− in the feed was once again attributed to oxidative processes that occurred due to fluid mixing (SI), as Fe(CN)64− is not amenable to cathodic reduction, and adsorption was not observed in the absence of an applied current or during postexperiment cleaning procedures. The values for CEo,Fe(CN)4− decreased as a function of current 6 density (Table 1a), which was attributed to O2 evolution becoming more favorable at higher potentials. The high reactivity of Fe(CN)64− coupled to the increase in overpotential with increasing current density, causes a shift in the reaction zone of Fe(CN)64− toward the outer REM surface. The measured open circuit potential of the 10 mM NaClO4 and 5 mM Fe(CN)64− electrolyte solution was 0.25 V (SI Table S-2), which is much lower than potentials measured at the anode surface (Table 1a). Since water oxidation is a more sluggish electron transfer reaction than Fe(CN)64− and has a higher redox potential (Eo = 1.23 V), the oxidation of water is expected to occur over a greater depth into the REM than Fe(CN)64−.54 These processes result in a decreased CEo,Fe(CN)4− 6 as a function of increased current density. Prior research has shown that the water oxidation reaction was active on nearly 100% of the surface area of a porous Ti/IrO2 electrode, while

Figure 3. Concentration profiles of oxalic acid in the feed and permeate solutions of REM cross-flow filtration experiments. The feed (solution which passed through the inner diameter of the REM) and permeate (solution which flowed through the REM pores) were recycled for the first 60 min of the experiment and a current was not applied to the cell. At 60 min, a current density of (a) 0.5 mA cm−2, (b) 1.0 mA cm−2, and (c) 3.5 mA cm−2 was applied to the cell, and the permeate was no longer recycled.

surface did not occur (Figure 3). Adsorbed compounds were likewise not detected during the cleaning protocol performed at the end of each experiment. Removal of OA in the permeate (ROA) increased linearly with current density (R2 = 0.982) and reached 98.6 ± 2.2% at 3.5 mA cm−2 (Table 1a). The removal rates in the permeate stream (rp,OA) also increased linearly with current density (R2 = 0.999) (Table 1a). Comparing the calculated mass-transfer rates (rm,OA) to values for rp,OA, indicate that OA removal was limited by the applied current for current densities between 0.2 and 1.0 mA cm−2 and reached the mass transfer limit at a current density of 3.5 mA cm−2 (Table 1a). The current efficiency for the permeate (CEp,OA) was 108 ± 11% at a current density of 0.2 mA cm−2 and decreased to 59.3 ± 3.9% at a current density of 3.5 mA cm−2 (Table 1a). Removal of OA was also observed in the recycled feed stream (rF,OA), and increased linearly as a function of current density (R2 = 0.999) (Table 1a). The removal of OA in the feed was attributed to oxidative processes, as adsorption was not observed in the absence of an applied current or during postexperiment cleaning procedures, and reduction of OA was not observed in cathodic control experiments. The removal of OA in the feed was attributed to mixing of water inside of the REM during cross-flow operation, which allowed a portion of the water to make contact with the REM surface but it was not E

dx.doi.org/10.1021/es401287e | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

F

0.00 0.20 0.50 1.00

current densitya (mA cm−2)

0.20 0.50 1.00 3.50 0.20 0.50 1.00 3.50

current densitya (mA cm−2)

0 42.6 ± 5.7 39.5 ± 1.0 47.7 ± 3.3

1.9 3.8 2.7 3.7

± ± ± ±

0.73c 1.67 1.88 2.13 51.1 45.8 37.2 43.8

normalized rate constant (kN,p‑MP) (L m−2 h−1)

23 12 15 50 14 3.6 4.8 9.0

mass transfer rate constant (km)d (L m−2 h−1)

360 384 366 372 162 150 174 78.0

40.2 78.0 138 390 78 90 102 66

0 92.9 ± 6.8 99.9 ± 0.12 99.9 ± 0.17

± ± ± ± ± ± ± ±

100, 3.71, 6.98, 5.04,

100 10.8 0.00 3.05

p-MP remaininge (%)

4.0 2.6 1.0 26b 4.1 1.3 4.9 28b

removal rate (rP) (eq 3) (mmole h−1 m−2)

removal in permeate (Ri) (eq 4) (%)

p-MP

± ± ± ± ± ± ± ±

4.7 2.4 3.0 10 2.9 0.7 1.0 1.8

± ± ± ± ± ± ± ±

72 77 73 74 32 30 35 15

mass transfer rate (rm) (eq 8) (mmole h−1 m−2)

membrane flux (J) (L m−2 h−1)

measured anode potential (V vs SHE)

1.68 1.84 2.07 2.90 1.68 1.84 2.07 2.97

measured anode potential (V vs SHE) ± ± ± ± ± ± ± ± 11 2.8 5.0 3.9b 5.4 0.7 1.3 0.021b

0 14.4 ± 0.91 14.4 ± 1.5 14.7 ± 0.71

normalized rate constant (kN,COD) (L m−2 h−1)

108 80.1 74.3 59.3 105 49.1 27.1 4.83

current efficiency (CEp) (eq 5) (%)

permeate

± ± ± ± ± ± ± ± 0.76 0.68 1.5 2.2 0.87 0.61 0.93 9.2

0 30.1 ± 3.1 24.5 ± 9.8 25.6 ± 3.5

removal in permeate (Ri) (eq 4) (%)

11.6 19.9 43.2 98.6 48.9 65.9 70.9 112

percent removal (Ri) (eq 4) (%)

COD

± 0.56 ± 8.4 ± 5.9

± 3.2 ± 21 ± 53

100, 39.8, 41.3, 39.9,

100 41.2 33.5 34.7

COD remaininge (%)

0 13.8 53.4 246 0 11.4 39.6 60.0

0 14.7 8.19 37.4 0 6.25 10.7 4.80

0,0 47.4, 44.7 22.2, 26.4 6.05, 5.67

0,0 12.8, 14.1 36.5, 40.1 54.1, 59.6

COD oxidationg (%)

± 0.31 ± 0.52 ± 0.47

± 3.5 ± 3.3 ± 8.0

current efficiency (CEF) (eq 6) (%)

adsorbed CODf (%)

removal rate (rF) (mmole h−1 m−2)

feed

± ± ± ± ± ± ± ±

11 2.8 5.0 3.9 5.4 0.68 1.3 0.43

86.8, 88.8 99.0, 109 73.3, 80.8

efficiency (CECOD)h (eq 7) (%)

108 94.8 82.5 96.7 105 55.4 38.4 9.19

overall current efficiency (CE0) (%)

Current density calculated using nominal geometric surface area of inner REM wall. bCalculated based on initial removal rate. cOpen circuit potential (OCP). dDetermined by measured membrane flux (J). eFinal concentration at the end of the experiment (values for duplicate experiments). fFinal adsorbed mass remaining at the end of the experiment (values for duplicate experiments). gCalculated by eq 11 hCumulative current efficiency for experiment (values for duplicate experiments).

a

p-MP

compound

(b)

Fe(CN)4− 6

OA

compound

(a)

Table 1. Summary of Results for REM Filtration Experiments with (a) Oxalic Acid (OA) and Fe(CN)64− and (b) p-Methoxyphenol (MP)

Environmental Science & Technology Article

dx.doi.org/10.1021/es401287e | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Environmental Science & Technology

Article

the more kinetically favorable Fe2+ oxidation reaction occurred over