Enhanced water desalination efficiency in an air-cathode stackedmicrobial electrodeionization cell (SMEDIC)

Noura A. Shehab a, Gary L. Amy a, Bruce E. Logan b, Pascal E. Saikaly a,n

a Water Desalination and Reuse Center, Division of Biological and Environmental Sciences and Engineering, King Abdullah University of Science andTechnology, Thuwal 23955-6900, Saudi Arabiab Department of Civil and Environmental Engineering, 212 Sackett Building, The Pennsylvania State University, University Park, PA 16802, USA

a r t i c l e i n f o

Article history:Received 20 March 2014Received in revised form26 May 2014Accepted 29 June 2014Available online 8 July 2014

Keywords:Microbial desalination cellIon exchange resinElectrodeionizationBrackish water desalinationSeawater desalination

a b s t r a c t

A microbial desalination cell was developed that contained a stack of membranes packed with ionexchange resins between the membranes to reduce ohmic resistances and improve performance. Thisnew configuration, called a stacked microbial electro-deionization cell (SMEDIC), was compared to acontrol reactor (SMDC) lacking the resins. The SMEDICþS reactors contained both a spacer and1.470.2 mL of ion exchange resin (IER) per membrane channel, while the spacer was omitted in theSMEDIC-S reactors and so a larger volume of resin (2.470.2 mL) was used. The overall extent ofdesalination using the SMEDIC with a moderate (brackish water) salt concentration (13 g/L) was 90–94%,compared to only 60% for the SMDC after 7 fed-batch cycles of the anode. At a higher (seawater) saltconcentration of 35 g/L, the extent of desalination reached 61–72% (after 10 cycles) for the SMEDIC,compared to 43% for the SMDC. The improved performance was shown to be due to the reduction inohmic resistances, which were 130 Ω (SMEDIC-S) and 180 Ω (SMEDICþS) at the high salt concentration,compared to 210Ω without resin (SMDC). These results show that IERs can improve performance ofstacked membranes for both moderate and high initial salt concentrations.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

Water scarcity is a major global challenge, and it is predictedthat by 2025 two-thirds of the world's population will be living inwater-stressed countries [1]. Desalination of seawater and brack-ish water can be used to alleviate water stress, and it is estimatedthat the number of seawater desalination facilities will increasesubstantially over the next 10 years. Current commercial desa-lination technologies, including electrodialysis (ED), electro-deionization (EDI), thermal desalination, and reverse osmosis(RO) are energy intensive and therefore there is great interest inother technologies that use less electrical energy [2,3].

Microbial desalination cells (MDCs) have recently drawn atten-tion as a low-energy method of water desalination. The simplestMDC is a microbial fuel cell (MFC) that is modified to contain amiddle chamber for desalination, by using two ion exchangemembranes between the anode and cathode chamber [4]. In theanode chamber organic matter is oxidized by exoelectrogenicbacteria [5], with the electrons released to the circuit and protonsinto solution. Electrons from the anode flow to the cathode wherethey combine with protons and oxygen to form water [4,5].

The production of protons at the anode and consumption ofprotons at the cathode drives desalination of saltwater in themiddle chamber, as salt ions in the saline water in the middlechamber migrate through the cation and anion exchange mem-branes to balance charge [6]. The performance of MDC is limitedby several factors including the microbial community compositionon the anode, electrode materials, pH imbalances, and internalresistance [7,8,11,13]. The internal resistance has several differentcomponents including solution and membrane (ohmic), chargetransfer, contact, and mass transfer resistances. The low ionicconductivity of less saline waters (1–10 g/L) can substantiallyincrease ohmic resistances, particularly as the water becomesprogressively desalinated [4–6,9,10].

Several methods have been proposed to reduce internal resis-tance and enhance desalination performance and rates of MDCs.Instead of using only a pair of ion exchange membranes, a stack ofmembranes can be placed between the electrodes [10] similar tostack configurations used for conventional ED. However, a largespacing between the electrodes can produce a high solutionresistance. In one early study, only 1.5 cell pairs could be used inthe stacked-MDC at the voltage generated, due to the wide spacingbetween the electrodes (1 cm) which produced a high ohmicresistance (18 Ω per membrane pair) [10]. Performance wasimproved by reducing the chamber width to that of the thinspacers used to separate the membranes (1.3 mm), enabling the

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/memsci

Journal of Membrane Science

http://dx.doi.org/10.1016/j.memsci.2014.06.0580376-7388/& 2014 Elsevier B.V. All rights reserved.

n Corresponding author. Tel.: þ966 5 44700129.E-mail address: pascal.saikaly@kaust.edu.sa (P.E. Saikaly).

Journal of Membrane Science 469 (2014) 364–370

use of a five-cell pair membrane stack that substantially improvedMDC performance with a high initial salt concentrations (35 g/L).However, the desalination performance of this 5-cell stack wasreduced at a low salt concentration of 6 g/L due to high ohmicresistance of the lower conductivity solution [11]. Omitting spacerscan improve performance [12], although this can result in thedeformation of the membranes and adversely affect flow throughthe chamber. Another approach is to use ion-exchange resins(IERs) to increase the ion conductivity in the solution betweenthe membranes [6,9]. This IER approach was shown to enhance thedesalination of lower salinity solutions (0.7–10 g/L) in a single-desalination-chambered MDC due to the reduction in the ohmicresistance [6,9].

The objective of this study was to further improve the perfor-mance of the 5-cell stack MDC developed by Kim and Logan [11]by using IERs in the solution chambers in a stack of ion exchangemembranes, for both high and moderate salt concentration solu-tions. This configuration is referred to as a stacked microbialelectro-deionization cell (SMEDIC), due to the expected enhanceddeionization effect of the IERs on performance. The desalinationperformance of the SMEDIC was examined for solutions with twodifferent salt concentrations (13 g/L and 35 g/L) in order todemonstrate that improved performance with the IERs was notlimited to using only lower conductivity solutions. The perfor-mance of the system was compared to a control reactor, lackingIERs, referred to as a stacked MDC (SMDC). The performance of theSMEDIC was evaluated in terms of the volumes of IERs used andthe presence or absence of spacers, in terms of desalinationefficiency (extent of desalination), electrical power production,and internal resistance.

2. Experimental

2.1. Construction and operation of SMEDIC and SMDC

The anode (30 mL; empty bed volume) and cathode (18 mL;empty bed volume) chambers were made from polycarbonatecylindrical chambers with a cross-sectional area of 7 cm2 followinga previous design [11,13]. The anode was a graphite fiber brush2.7 cm in diameter and 2.3 cm long, and it was heat treated beforeuse (Mill-Rose Lab Inc., USA) [14]. The air cathode containedplatinum nanoparticle catalysts on the water side (3.5 mg Pt) witha Nafion binder, and four polytetrafluoroethylene diffusion layerson the air side [15]. The desalination chamber in both the SMEDICand SMDC reactors contained a five-cell pair ED stack (10 totalcells) built between the anode and cathode chambers (Fig. 1A). TheED stack was constructed with 5 cation-(CEM) and 6 anion-(AEM)exchange membranes (Selemion CMV and AMV, Asahi glass,Japan). The membranes (�0.1 mm thick) were pretreated bystorage in a 0.6 M NaCl solution for 24 h, and then rinsed withdeionized water. Silicon gaskets (�1.3 mm thickness) were usedbetween the membranes to create a water tight seal and provide aflow path across the membranes [11]. Polyethylene mesh spacers(4 cm�0.5 cm; 1 mm thickness) were used to maintain cellthickness in the SMDC. SMEDIC reactors were tested in twoconfigurations (with and without spacers) which required theuse of different amounts of resin. When spacers were used, thereactors were packed with 1.470.2 mL of IERs (SMEDICþS).When spacers were omitted (SMEDIC-S), additional resin wasused (2.470.2 mL) in order to completely fill the chamber andmaintain a constant chamber size. The diluate or desalinatedwater volumes were 144 mL (SMDC), 85 mL (SMEDICþS), and82 mL (SMEDIC-S).

The anion IER used was a strong base resin type with a totalexchange capacity of 1.1 eq/L (DOWEX MONOSPHERE 550A (OH),

DOW Chemicals, USA). The cation IER used was a strong acid resintype with a total exchange capacity of 2.0 eq/L (DOWEX MONO-SPHERE 650C (H), DOW Chemicals, USA). Cation- and anion-exchange resins were mixed at a ratio of 1:1.6 (v/v) based on theirdifferent exchange capacities, and these mixed-beds wereemployed between adjacent AEM and CEM pairs.

The anodes were initially acclimated in MFCs until the peakvoltage was stable at around 550 mV for three reproducible cycles(external resistance of 1000Ω) [10]. The MFCs were inoculatedwith primary clarifier effluent and operated in a fed-batch modewith acetate as the main carbon and energy source. After acclima-tion, the anodes were transferred to the different desalinationreactors, and operated at a lower external resistance (10Ω) toimprove power production [10].

Two concentrations of synthetic salt solution were tested: 13 g/L,representing brackish water; and 35 g/L, representing seawater.The salt solution was continuously pumped (from the bottom tothe top of the chamber) into the diluate and concentrate cells inthe ED stack at a rate of 0.1 mL/min (144 mL/d). The anode andcathode chambers were operated in a fed-batch mode overmultiple batch cycles. When current decreased below 0.20 mA,the catholyte solution was replaced with fresh synthetic saltsolution (13 g/L or 35 g/L NaCl) and the anolyte solution wasreplaced with fresh medium consisting of 1 g/L sodium acetatein a phosphate buffer (9.16 g/L Na2HPO4; 4.9 g/L NaH2PO4–H2O;0.62 g/L NH4Cl; 0.26 g/L KCl) with minerals and vitamins [11,16].The desalinated effluent (diluate solution) was recycled throughthe ED stack over several fed-batch cycles of anode operation. Thediluate solution was continuously recycled using a 200 mL saltwater reservoir, while the concentrate solution was not recycled.The diluate solution flowed serially from the cathode side throughevery diluate cell, and the concentrate solution flowed co-currently through the concentrate cells (Fig. 1A). All reactors wereoperated in duplicate, at 2772 1C.

2.2. Analyses and calculations

Power density and polarization curves were generated using apotentiostat (VMP3 Multichannel Workstation, Biologic ScienceInstruments, USA) at 30 1C. Current was scanned from 0 mA to3 mA, with each current step held for 15 min to reach steadyconditions. Power densities (mW/m2) were normalized by cathodeprojected working area (7 cm2).

Influent and effluent conductivities and pH for the diluate,concentrate, anolyte and catholyte solutions were measured usinga conductivity-pH meter (Seven Multi, Mettler-Toledo Interna-tional Inc., USA). The salinity was estimated from conductivitymeasurements using an in situ conductivity conversion as pre-viously outlined by Bennett [17], and assuming the conductivitymeasured was due only to NaCl. Salinity reduction was calculatedbased on the influent and effluent conductivities. Current effi-ciency was determined as the ratio of ionic separation of NaCl tothe total number of electrons passed through the circuit, as

η¼ FðcDinvDin�cDoutvDoutÞ

Ncp∑Ridt

ð1Þ

where F is Faraday's constant, c the molar concentration of NaCl inthe diluate, v the volume of the diluate, Ncp the number of cellpairs in the ED stack, and i the current generated in the reactor.The subscript “in” indicates conditions at the beginning of thecycle, “out” the end of the cycle, and the superscript “D” indicatesdiluate [11].

At the end of each desalination cycle, the total desalination rate(TDR) was calculated to evaluate desalination performance

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370 365

according to

TDR ¼ ðCi�CeÞ=t ð2Þwhere Ci and Ce are the initial and final concentrations (g/L) of saltin the desalination chamber over the desalination period, t (h).

The chemical oxygen demand (COD) was measured for influentand effluent anolyte solutions using standard methods (Hach Co.,USA). The coulombic efficiency (CE) was calculated based on thetotal COD removed and the number of coulombs collected duringthe cycle as previously described [18].

2.3. Electrochemical impedance spectroscopy

Ohmic and charge transfer resistances were determined usingGalvanostatic electrochemical impedance spectroscopy (GEIS)with a potentiostat (VMP3 Multichannel Workstation, BiologicScience Instruments, USA). Current was set at 1.5 mA, and thefrequency ranged from 1 MHz to 50 MHz, with a sinusoidalperturbation of 0.1 mA. The ohmic resistances of the cell wereobtained from Nyquist plots, based on the intercept of the curvewith the X-axis, and the charge transfer resistance was obtainedusing circle fit software from the diameter of the first (highfrequency) circle [19,20].

3. Results and discussion

3.1. The effect of IERs on salinity reduction and desalination rates

At an influent salt concentration of 13 g/L NaCl (brackish water),the SMEDIC-S reactor containing the IERs achieved 94% desalination,compared to 90% for the SMEDICþS containing spacers, and theSMDC control (Fig. 2a). This extent of desalination required 7 fed-batch cycles (replacement of the anode and cathode solutions), for atotal of 56 h for reactors containing IERs (SMEDIC and SMEDICþS),compared to a longer cycle time of 77 h for the SMDC. The differencein cycle times was based on the time required for oxidation oforganic matter in the anode chamber, as reflected by the timeneeded for a reduction in current generation (o0.2 mA).

The use of a higher initial salt concentration (35 g/L) resulted inless desalination due to the greater amount of charge that needed

to be transferred. However, the use of IERs still improved perfor-mance even with these relatively high initial solution conductiv-ities. For the same 7 cycles, the SMEDIC-S reached 50%desalination, compared to 43% for the SMEDICþS, and 30% forthe SMDC (Fig. 2b). Higher salt removals were achieved by using10 cycles (80 h total time for reactors with IERs, compared to 110 h

Fig. 1. (a) Schematic of bench scale SMEDIC with parallel continuously recycled flow through the 5 cell pair ED stack, anion exchange membrane (AEM), and cation exchangemembrane (CEM). (b) Photograph of the reactor in operation.

Fig. 2. Salinity reduction after recirculating the diluate solution over (a) 7 (13 g/LNaCl, 23.2 mS/cm) and (b) 10 (35 g/L NaCl, 54.1 mS/cm) fed-batch cycles of anodeoperation.

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370366

for the SMDC), with 72% desalination for the SMEDIC-S, comparedto 61% for the SMEDICþS and 43% for the SMDC reactors (Fig. 2b).

The improved performance of the reactors at the higher saltconcentrations suggested that there may have been an additionalbenefit of osmotic desalination (from the stack to the anode) thatcontributed to the improved rate of desalination. Therefore, abioticSMEDIC-S reactors were operated in an open-circuit mode toevaluate desalination (i.e., without electricity generation). Nosignificant variations in conductivity or pH in the anode chamberwere observed when the reactors were operated in open-circuitmode, indicating that the reduction in salinity was mainly due tocurrent generation.

The desalination rates were higher in the SMEDIC reactors thanthe other reactors due primarily to the greater extent of desalina-tion, but also in terms of time needed to complete the fed-batchcycle (Fig. 3). The desalination rate was greatest using the highconcentration of seawater, with a TDR of 0.33370.005 g/L-h forthe SMEDIC-S, followed by SMEDICþS (0.28670.045 g/L-h) andSMDC (0.14970.005 g/L-h) (Fig. 3). At the lower salt concentra-tion, the TDR with the SMEDIC-S was �30% less (0.23370.005 g/L-h) than that obtained at the higher salt concentration, withlower rates in the other two systems.

The desalination rates obtained here are better than mostpreviously reported results, when comparisons are made on thebasis of the initial salt concentrations and the volume of waterneeded to accomplish desalination. In one early MDC study, a TDRof 0.3 g/L-h was achieved using an initial concentration of 35 g/LNaCl, but 133 L of water was used to desalinate 1 L of water and aferricyanide catholyte was used [4]. In another study, the TDR wasmuch lower (0.0252 g/L-h) than rates obtained here, at an inter-mediate initial salt concentration (20 g/L NaCl) compared to thoseused here, and a much larger volume of water was needed toaccomplish desalination (47 L per 1 L of desalinated water) [10].These rates with the SMEDIC reactors were higher than thosepreviously obtained using IERs in an MDC of �0.17 g/L-h at aninitial salinity of 10 g/L NaCl [6] (Table 1). Part of the reason for thislower rate in the previous study was likely the lower initial salinityof 10 g/L, compared to 13 or 35 g/L used here. The study by Daviset al. [21] reported a higher TDR (0.92 g/L-h) at 35 g/L NaCl thanthat achieved here, and only1.4 L was used to desalinate 1 L ofwater, but there was only a 26% reduction in salinity.

The volume of water used in these different studies to accom-plish desalination is clearly an important factor in performance,although direct comparisons are difficult due to differences in theinitial salt concentrations and various volumes of electrolytesolutions used to desalinate the water (Table 1). For studies withan initial salt concentration of 35 g/L, 72% desalination wasachieved here using the SMEDIC-S configuration, with 2.7 L of

water required per liter of desalinated water produced. While thepercentage of salt removed in tests has reached 98% [11] and 100% [22],21–60 L of water were needed per 1 L of desalinated water. At alower initial salt concentration of 10 g/L of NaCl, 99% salt removalwas previously obtained [23]; however, 14 L was used per liter ofdesalinated water (Table 1). At a comparable but slightly higherinitial salt concentration of 13 g/L, 94% desalination was achievedin using the SMEDIC-S with only 2.7 L used per 1 L ofdesalinated water

There were changes in the volumes of the diluent and con-centrate over the duration of the experiment due to osmosis andelectro-osmosis (transport of water molecules moving with thetransported ion). For example, at the end of 7 fed-batch cycles, theconcentrate volume increased by �80 mL (�44%) for the SMDCreactor at low and high salt concentrations, compared to �46 mL(�68%) for the SMEDICþS and �45 mL (�69%) for the SMEDIC-S.These percentage changes are similar to those reported by Kimand Logan [11], who found that the concentrate increased by�19.4% under conditions with a starting salt concentration of35 g/L NaCl, and total desalination time for 12 h of a fed-batchcycle. In that study, it was reported that electroosmosis wasresponsible for 65% of the increased volume of the concentrateeffluent in MDC stack, with the remainder assumed to be due toosmosis. Davis et al. [21] suggested that osmotic losses could bebetter controlled by stopping the fed-batch cycle sooner, so that ahigher average current would be maintained over the fed-batch cycle.

The desalination of the saline water in the ED stack resultedin an increase in the salinity of the anolyte. The conductivityin the SMDC reactors increased from 12.2870.04 to 16.570.11 mS/cm at the end of the cycle and it increased slightly more(from 12.2870.04 to 18.470.01 mS/cm) in the SMEDICþS andSMEDIC-S reactors. The increase in anolyte conductivity was dueto the increased concentration of chloride ions that migrated fromthe desalination chamber, although the final pH also affected thechange in conductivity as well as a lower pH increases solutionconductivity [5,10,11,24].

The increase in salinity at the anode due to the transfer of Cl�

ions from the stack to the anode chamber to balance charge couldbe harmful to exoelectrogenic bacteria [5]. One way to avoid thiseffect of Cl� ions, is to use a bipolar membrane between the anode

Fig. 3. Total desalination rate (TDR) after recirculating the diluate solution over 7(13 g/L NaCl, 23.2 mS/cm) and 10 (35 g/L NaCl, 54.1 mS/cm) fed-batch cycles ofanode operation.

Table 1Comparison of desalination performance with other MDC studies. Velec is the totalvolume of electrolyte used in the anode and cathode chambers to desalinate 1 L ofwater to the extent shown.

NaCl (g/L) Velect (L) Salinity reduction (%) TDR (g/L�h) Ref.

35 133 93 1.3 [4]35 21 98 NA [11]35 3 44 NA [11]35 3 (SMDC)a 30 0.116 This study35 3(SMDC)b 43 0.149 This study35 2.7 (SMDIC-S)a 51 0.28 This study35 2.7(SMDC-S)b 72 0.233 This study35 2.5 (SMDICþS)a 43 0.25 This study35 2.5(SMDCþS)b 61 0.286 This study35 1.4 26 0.92 [21]30 60 100 NA [22]20 47 80 0.0252 [10]20 3 63 NA [13]20 2 37 NA [5]13 3 (SMDC)a 60 0.112 This study13 2.7 (SMDIC-S)a 94 0.333 This study13 2.5 (SMDICþS)a 90 0.224 This study10 14 99 NA [23]10 3 58 0.17 [6]

a Salt removals were achieved by using 7 cycles of fed-batch anode.b Salt removals were achieved by using 10 cycles of fed-batch anode.

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370 367

chamber and the adjacent membrane chamber, as done for amicrobial electrolysis cell [27]. However, the use of this membranewould effectively eliminate any useful power production. Alter-natively, the anode bacteria can be gradually acclimated to highersalt concentrations, avoiding decreases in power production athigher anode conductivities [21].

3.2. pH, COD removal and CE

The pH of desalinated effluent (diluate) was stable (6.370.4)for all reactors and was slightly lower than the influent pH(7.470.2). The concentrate effluent pH increased for all reactortypes (SMEDIC-S: 9.370.2; SMEDICþS: 10.770.2; SMDC:11.170.2). The increase in pH was likely due to the transport ofhydroxyl ions from the cathode chamber into the concentratestream [11].

The decrease in anolyte pH over an anolyte fed-batch cycle washigher in the SMDC reactors (from 7.170.04 to 5.270.3) com-pared to the SMEDICþS and SMEDIC-S reactors (from 7.170.04 to6.4370.01) possibly due the longer cycle time (11 h vs. 8 h in theSMDIC reactors). The decrease in anolyte pH was due to accumu-lation of protons generated from acetate degradation by exoelec-trogenic bacteria [11]. In contrast, the catholyte pH increasedsharply from 7.570.3 to 11.270.02 (SMEDICþS and SMEDIC-S)and 12.470.3 (SMDC) over a fed-batch cycle. To avoid a high pH atthe cathode, the catholyte solution was replaced by fresh syntheticsalt solution (13 g/L or 35 g/L NaCl) at the end of each fed-batchcycle. This increase in pH at the cathode was due to the consump-tion of protons for oxygen reduction. The pH imbalance betweenthe anode (5.2–6.4) and cathode (11.2–12.4) reported in this studyis consistent with previous stacked MDC studies [9,11].

The pH imbalance is an inherent problem in MDCs. At theanode chamber organic matter is oxidized by exoelectrogenicbacteria while protons are released into solution, lowering theanolyte pH. The consumption of protons for oxygen reduction atthe air cathode increases the catholyte pH. This pH imbalance ismore pronounced in MDCs compared to MFCs because the pre-sence of AEM at the anode side limits the transport of protons, andthe presence of CEM at the cathode side limits the transport ofOH� ions. The pH decrease in the anode chamber could beharmful to the activity of exoelectrogenic bacteria. Severalapproaches have been used to reduce the inherent pH problemsin MDCs including the use of larger electrolyte solution volumes[4,10], recirculating the electrolyte between the cathode andanode chambers [25,26], or inserting a bipolar membrane nextto the anode chamber [27].

The COD removal in the SMEDIC-S was 8873%, which wassimilar to that of the SMEDICþS (9072%) but much higher thanthat of the SMDC (8072%). The CEs ranged from 45% for the SMDCand SMEDICþS, to 61% for the SMEDIC-S. These values were lowerthan what was reported (average CE of 80%) by Kim and Logan [11]using a SMDC-type of configuration, compared to this study withIERs. In addition to oxygen that could be leaking into the system,loss of acetate could occur due to diffusion through the anionexchange membrane placed adjacent to the anode chamber, orfrom methanogenesis. Zhang et al. [9] obtained a much lower CE(o10%) in an MDC packed with IERs, even with potassiumferricyanide as the catholyte, and attributed this low CE to lossof acetate via methanogensis.

3.3. Current efficiency and power generation

Current efficiencies, which relate the ions transported fordesalination to coulombs transferred, were all 490%. The averagecurrent efficiency for the SMEDIC-S was 9973%, compared to9777% for the SMEDICþS and 9373% for the SDMC. These

results demonstrate that the ion exchange membranes effectivelytransported sodium or chloride counter ions with current genera-tion, with little back-diffusion of these ions. Other counter ionscould back-diffuse across the ion exchange membranes, such asacetate or hydroxyl ions, are not accounted for in Eq. (1).

The maximum power density observed at moderate salt con-centration (13 g/L NaCl) was higher in the SMEDIC-S (0.6570.04 W/m2) compared to SMEDICþS (0.4970.03 W/m2) and theSMDC (0.3870.05 W/m2) (Fig. 4a). Similar but higher maximumpower densities were obtained using the higher initial salt con-centration (35 g/L NaCl) where the maximum power density washigher in the SMEDIC-S reactors (0.8370.04 W/m2) (Fig. 4b).

3.4. Internal and ohmic resistances

The ohmic resistances (membranes and solution resistances)and charge transfer resistances of the whole cell calculated usingGalvanostatic EIS, decreased in the order expected based on thechanges in power production for the different reactor configura-tions. For the 35 g/L test conditions, the SMEDIC-S reactors withthe highest power density had the lowest resistances, with 130 Ωfor the ohmic resistance, and 60 Ω for the total charge transferresistance, for a total of 190 Ω (Fig. 5b). These resistances increasedby �39% for SMEDICþS with the spacers and IERs, and by �62%for the SMDC with only the spacers (Fig. 5b). The same trend wasobserved at the lower salt concentration (Fig. 5a) although, in allcases, the resistances were slightly larger for each reactor config-uration due to the lower conductivity of the salt solution. Therewas no apparent trend with the total internal resistance calculatedfrom the slopes of the polarization data. At the 13 g/L initial saltcondition, the internal resistances increased in the order 17275 Ω(SMEDICþS), 20372 Ω (SMEDIC-S), and 204715 Ω (SMDC),while at the 35 g/L salt concentration they were 18972 Ω(SMEDICþS), 19776 Ω (SMDC), and 20471 Ω (SMEDIC-S). Thereason for the inconsistent ordering of these resistances from theslopes of the polarization data with the maximum power densitiesis not known, but the changes in ohmic resistance based on the

Fig. 4. Power densities for (a) low (13 g/L NaCl, 23.2 mS/cm) and (b) high (35 g/LNaCl, 54.1 mS/cm) salt concentrations.

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370368

GEIS analysis provided a clear indication for the changes in reactorperformance with the different configurations.

Spacers are needed to provide mixing and an even distributionof flow across the membranes in electrodialysis stacks [12].A recent study showed that spacer-less reverse electrodialysisstacks resulted in non-uniform flow, and this in turn caused anincrease in membrane resistance by 21 Ω as the flow rate wasincreased from 4 mm s�1 to 18 mm s�1. In constrast, flow rate didnot have an effect on membrane resistance in RED stacks withspacers. In the current study spacers were used between each pairof ion exchange membranes in order to provide uniform flow ofthe feed and prevent deformation of the membranes. However,when 2.470.2 mL of ion exchange resins were used the spacerswere omitted in order to completely fill the chamber and maintaina constant chamber size. Omitting the spacers resulted in animprovement in performance because spacers are non-conductive,and therefore their use increases ohmic resistance (membranesand solution resistances) by hindering conduction of ions throughthe IEM [12]. The effect of the spacers on ohmic resistance isconsistent with a previous study using a reverse electrodialysisstack, where the resistance due to the spacers was estimated to be50 Ω [12]. Based on the GEIS analysis here, there was a reductionin ohmic resistances of 85 Ω (moderate salt concentration) and80 Ω (high salt concentration) for SMEDIC-S compared to SMDC.

The SMEDIC-S reactor with no spacers (but more IERs) had a lowerohmic resistance of 50–56Ω than the SMEDICþS reactors thatcontained spacers (but lower IERs) at high and low salt concen-tration, respectively. These results indicate that IERs could be usedto keep the membranes separated, but produce lower internalresistances and higher power densities compared to reactors withspacers, thus negating the need for spacers. However, non-uniform flow might limit the use of spacer-less SMEDIC systemsto low flow rate conditions [12]. In the current study the flow ratewas �1 mm s�1. Future studies should address the effect of flowrate on the performance of spacer-less SMEDIC systems.

4. Conclusions

The use of IERs (without spacers) in a stacked, air-cathodemicrobial desalination reactors improved desalination perfor-mance relative to systems containing IERs and spacers, or onlyspacers. The rate of desalination was increased due to the reduc-tion in membrane stack resistance with both a high (35 g/L) andmoderate (13 g/L) initial salt concentration. The MEDIC-S systemachieved a desalination rate of 0.33370.005 g/L-h to 0.23370.005 g/L-h, with 72–94% total desalination of the water using2.7 L per 1 L of desalinated water with both initial salt concentra-tions. In contrast, the SMDC reactor, lacking IERs but with spacers,showed a desalination rate of 0.14970.005 g/L-h to 0.11270.005 g/L-h, with 43–60% desalination with 3 L needed per 1 L ofdistilled water. The improved performance was due primarily to areduction in the membrane resistance. These results show thatadding IERs and eliminating the spacer between the ion exchangemembranes is a useful approach to improving desalination perfor-mance even at high initial salt concentrations typical of seawater.

Acknowledgments

This work was sponsored by a PhD fellowship, a Global ResearchPartnership-Collaborative Fellows Award (GRP-CF-2011-14-S),KAUST Award KUS-I1-003-13 and discretionary investigator fundsat King Abdullah University of Science and Technology (KAUST).We would like to thank Dr. Fang Zhang, from Penn State Uni-versity, for helping with analysis of the components of the internalresistances.

References

[1] R.F. Service, Desalination freshens up, Science 313 (2006) 1088–1090.[2] Y. Zhou, R.S.J. Tol, Evaluating the costs of desalination and water transport,

Water Resour. Res. 41 (2005) 1–10.[3] K. Christen, Environmental costs of desalination, Environ. Sci. Technol. 41

(2007) (5579–5579).[4] X. Cao, X. Huang, P. Liang, K. Xiao, Y. Zhou, X. Zhang, B.E. Logan, A new method

for water desalination using microbial desalination cells, Environ. Sci. Technol.43 (2009) 7148–7152.

[5] M. Mehanna, P. Kiely, D. Call, B.E. Logan, Microbial electrodialysis cell forsimultaneous water desalination and hydrogen gas production, Environ. Sci.Technol. 44 (2010) 9578–9583.

[6] A. Morel, K. Zuo, X. Xia, J. Wei, X. Luo, P. Liang, X. Huang, Microbial desalinationcells packed with ion-exchange resin to enhance water desalination rate,Bioreso. Technol. 118 (2012) 43–48.

[7] R.A. Rozendal, T.H. Sleutels, H.V. Hamelers, C.J. Buisman, Effect of the type ofion exchange membrane on performance, ion transport, and pH in biocata-lyzed electrolysis of wastewater, Water Sci. Technol. 57 (2008) 1757–1762.

[8] M. Zhou, M. Chi, J. Luo, H. He, T. Jin, An overview of electrode materials inmicrobial fuel cells, J. Power Sour 196 (2011) 4427–4435.

[9] F. Zhang, M. Chen, Y. Zhang, R.J. Zeng, Microbial desalination cells with ionexchange resin packed to enhance desalination at low salt concentration, J.Membr. Sci. 417 (2012) 28–33.

[10] X. Chen, X. Xia, P. Liang, X. Cao, H. Sun, X. Huang, Stacked microbialdesalination cells to enhance water desalination efficiency, Environ. Sci.Technol. 45 (2011) 2465–2470.

0

50

100

150

200

250

300

SMEDIC-S SMEDIC+S SMDC

Res

ista

nce

(Ohm

)Charge transferOhmic resistance

0

50

100

150

200

250

300

SMEDIC-S SMEDIC+S SMDC

Res

ista

nce

(Ohm

)

Fig. 5. Internal resistance (Ohmic resistance for solutionþmembranes and chargetransfer) determined by GEIS for (a) low (13 g/L NaCl, 23.2 mS/cm) and (b) high(35 g/L NaCl, 54.1 mS/cm) salt concentrations.

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370 369

[11] Y. Kim, B.E. Logan, Series assembly of microbial desalination cells containingstacked electrodialysis cells for partial or complete seawater desalination,Environ. Sci. Technol. 45 (2011) 5840–5845.

[12] M.C. Hatzell, B.E. Logan, Evaluation of flow fields on bubble removal andsystem performance in an ammonium bicarbonate reverse electrodialysisstack, J. Membr. Sci. 10 (2013) 449–455.

[13] M. Mehanna, T. Saito, J. Yan, M. Hickner, X. Cao, X. Huang, B.E. Logan, Usingmicrobial desalination cells to reduce water salinity prior to reverse osmosis,Energy. Environ. Sci. 3 (2010) 1114–1120.

[14] B.E. Logan, S. Cheng, V. Watson, G. Estadt, Graphite fiber brush anodes forincreased power production in air-cathode microbial fuel cells, Environ. Sci.Technol. 41 (2007) 3341–3346.

[15] S. Cheng, H. Liu, B.E. Logan, Increased performance of single-chamber micro-bial fuel cells using an improved cathode structure, Electrochem. Commun. 8(2006) 489–494.

[16] D.R. Lovley, E.J.P. Phillips, Novel mode of microbial energy metabolism:organic carbon oxidation coupled to dissimilatory reduction of iron ormanganese, Appl. Environ. Microbiol. 54 (1988) 1472–1480.

[17] A.S. Bennett, Conversion of in situ measurements of conductivity to salinity,Deep-Sea. Res 23 (1976) 157–1659.

[18] B.E. Logan, B. Hamelers, R. Rozendal, U. Schroder, J. Keller, S. Freguia,P. Aelterman, W. Verstraete, K. Rabaey, Microbial fuel cells: methodologyand technology, Environ. Sci. Technol. 40 (2006) 5181–5192.

[19] Z. He, F. Mansfeld, Exploring the use of electrochemical impedance spectro-scopy (EIS) in microbial fuel cell studies, Energy Environ. Sci. 2 (2009)215–219.

[20] A.J. Hutchinson, J.C. Tokash, B.E. Logan, Analysis of carbon fiber brush loadingin anodes on startup and performance of microbial fuel cells, J. Power Sour.196 (2011) 9213–9219.

[21] R.J. Davis, Y. Kim, B.E. Logan, Increasing desalination by mitigating anolyte pHimbalance using catholyte effluent addition in a multi-anode bench scalemicrobial desalination cell, ACS Sustain. Chem. Eng. 1 (2013) 1200–1206.

[22] K.S. Jacobson, D.M. Drew, Z. He, Efficient salt removal in a continuouslyoperated upflow microbial desalination cell with an air cathode, Bioresour.Technol. 102 (2011) 376–380.

[23] H. Luo, P.E. Jenkins, Z. Ren, Concurrent desalination and hydrogen generationusing microbial electrolysis and desalination cells, Environ. Sci. Technol. 45(2011) 340–344.

[24] P.B. Spoor, W.R.T. Veen, L.J.J. Janssen, Electrodeionization 1: migration of nickelions absorbed in a rigid, macroporous cation-exchange resin, J. Appl. Electro-chem. 31 (2001) 523–530.

[25] Y. Qu, Y. Feng, X. Wang, J. Liu, J. Lv, W. He, B.E. Logan, Simultaneous waterdesalination and electricity generation in a microbial desalination cell withelectrolyte recirculation for pH control, Bioresour. Technol. 106 (2012) 89–94.

[26] X. Chen, P. Liang, Z. Wei, X. Zhang, X. Huang, Sustainable water desalinationand electricity generation in a separator coupled stacked microbial desalina-tion cell with buffer free electrolyte circulation, Bioresour. Technol. 119 (2012)88–93.

[27] S. Chen, G. Liu, R. Zhang, B. Qin, Y. Luo, Development of the microbialelectrolysis desalination and chemical-production cell for desalination as wellas acid and alkali productions, Environ. Sci. Technol. 46 (2012) 2467–2472.

N.A. Shehab et al. / Journal of Membrane Science 469 (2014) 364–370370