ate

w pulpere reantributportionMicha

ups, in continuous mode, with hydraulic retention times from 6 to 48 h and

phenson et al., 2012). In recent years, an increasing number ofstudies have aimed to adapt the emergent technology of microbialfuel cells (MFCs) to the treatment of efuents. MFCs would offerthe great advantage of abating organic matter in anaerobicconditions, thus saving the cost of aeration. In addition, they couldtransform a part of the chemical energy contained in the organicmatter directly into electricity (Logan and Rabaey, 2012).

only very modeste Table 1 in Sup-chemical oxygenthe low vatrochemicoval. Und

tinuous ow in an air cathode MFC, Liu et al. obtained CEwith the efuent from a treatment plant (Liu et al., 2004);et al. found maximal CE of 27% with domestic wastewater (Chenget al., 2006). Several other studies using raw efuents have exhib-ited very low Coulombic efciencies, between 1% and 5% (Younghoand Logan, 2010; Wen et al., 2010; Lee et al., 2010; Kassongo andTogo, 2011; Lefebvre et al., 2013). It is thus widely agreed that asubstantial fraction of organic matter is oxidized in alternativeways such as aerobic respiration or anaerobic pathways due tothe presence of dissolved electron acceptors (nitrate, sulphate)

Corresponding author. Tel.: +33 534323673.

Bioresource Technology 149 (2013) 117125

Contents lists availab

T

elsE-mail address: alain.bergel@ensiacet.fr (A. Bergel).cern of modern society and is responsible for the consumption oflarge amounts of energy in developed countries. Various technolo-gies are implemented, which generally require high electricityinput, particularly for the aeration of the large tanks used inaerobic treatments (Youngho and Logan, 2010), and result in theproduction of substantial amounts of sludge (Li et al., 2007; Ste-

paper mill efuents so far and they have shownperformance in current and power generation (seplementary data). Abatements of the efuentsdemand (COD) generally seem satisfactory, butCoulombic efciencies (CE) indicate that the eleccess makes only a minor contribution to COD rem0960-8524/$ - see front matter 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biortech.2013.09.025lues ofal pro-er con-of 12%Cheng 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The treatment of urban and industrial efuents is a major con-

Paper-making is a water-intensive industrial sector thatrequires high volumes of water and produces high volumes of con-centrated efuents. Nevertheless, few MFC studies have addressedAccepted 4 September 2013Available online 12 September 2013

Keywords:BioanodeMicrobial anodePulp mill efuentMicrobial fuel cell

inlet COD from 500 to 5200 mg/L. Current densities around 4 A/m2 were obtained and voltammetrycurves indicated that 6 A/m2 could be reached at +0.1 V/SCE. A theoretical model was designed, whichallowed the effects of HRT and COD to be distinguished in the complex experimental data obtained withconcomitant variations of the two parameters. COD removal due to the electrochemical process was pro-portional to the hydraulic retention time and obeyed a MichaelisMenten law with respect to the COD ofthe outlet ow, with a Michaelis constant KCOD of 400 mg/L. An inhibition effect occurred above inlet CODof around 3000 mg/L.Received 14 June 2013Received in revised form 2 September 2013

polarization at 0.3 V/SCEterized in 3-electrode set-h i g h l i g h t s

Microbial bioanodes were formed in ra In well-controlled conditions 4 A/m2 w A theoretical model determined the co Electrochemical COD removal was pro Electrochemical COD removal obeyed a

a r t i c l e i n f o

Article history:and paper mill efuent.ched at 0.3 V/ECS.ion of the bioanode to the COD removal.al to HRT.elisMenten law with the COD outlet.

a b s t r a c t

Microbial bioanodes were formed in pulp and paper efuent on graphite plate electrodes under constant, without any addition of nutriment or substrate. The bioanodes were charac-aCentre Technique du Papier, BP 251, 38044 Grenoble Cedex 9, Franceb Laboratoire de Gnie Chimique (LGC), CNRS-Universit de Toulouse (INPT), BP 84234, 31432 Toulouse, FranceExperimental and theoretical characterizformed in pulp and paper mill efuent inconditions

Stephanie F. Ketep a,b, Eric Fourest a, Alain Bergel b,

Bioresource

journal homepage: www.ion of microbial bioanodeslectrochemically controlled

le at ScienceDirect

echnology

evier .com/locate /bior tech

chemical conditions. Most studies using raw industrial efuentshave been carried out in whole MFC devices so far. MFCs are quite

Techcomplex systems in which various processes occur and can interact(anode and cathode kinetics, diffusion and migration in differentmedia including membrane, etc.) and multiply the number of pos-sible causes of rate limitation. In contrast, in the present study, anelectroanalysis cell including a 3-electrode set-up was used to rig-orously control the potential of the microbial bioanode and toavoid any rate limitation not related to the bioanode kinetics. Inparticular, the high internal resistance of MFC due to the lowconductivity of efuents was overcome by locating the referenceelectrode very close to the bioanode surface. A small electrodesurface area (10 cm2) was used in a large volume of efuent(0.5 L) to favour a uniform current distribution and to maintainstable substrate concentration in the reactor. Furthermore, smoothgraphite plate electrodes were used to avoid morphologicalenhancing effects. This cell design was established according toconventional electroanalytical rules in order to characterize theintrinsic bioanode behaviour in electrochemically rigorous condi-tions. Obviously, the same design would not be suitable for anindustrial reactor, in which the electrode surface area must be aslarge as possible, the electrode morphology must favour local cur-rent heterogeneities and the use of a reference electrode cannot becontemplated. Similarly, nitrogen was sparged continuously in theelectroanalysis cell to avoid aerobic side-oxidations, while it wouldnot be useful at the industrial scale.

The electroanalysis cells were fed with efuents at differentCOD concentrations supplied in continuous mode with differenthydraulic retention times (HRT). To remain as close as possible toindustrial conditions, the efuent was used as collected from thetreatment unit, without any input of nutrients or addition of sub-strate. This was an original approach with respect to current liter-ature. Less than 30% of the studies dealing with MFCs haveoperated in continuous mode and most were conducted with puresubstrates (Zhang et al., 2013).

A theoretical approach was also developed to extract conclu-sions from experimental data that varied in a complex mannerdue to the evolution of the raw efuent during long-lasting exper-iments. As far as we know, this study was the rst attempt to usewell-controlled electrochemical conditions when characterizingbioanodes operating in continuous mode in a raw efuent.

2. Experimental section

2.1. Efuent

Efuent from a recycled paper producing mill was collected atthe outlet of the anaerobic efuent treatment step with COD rang-ing from 1400 to 1650 mg/L, pH of 7.15 0.15 and conductivity of3.3 ms/cm. It was stored at 4 C until used. This efuent was usedas sole inoculum and substrate source without addition of anynutrients or trace minerals. Lower COD values were obtained by(Min et al., 2005; You et al., 2006; Lu et al., 2009). In parallel, fer-mentation processes cannot be ruled out and biomass growth hasalso been suggested to account for part of COD removal (Min andLogan, 2004; Liu et al., 2004). The formation of microbial bioanodesin raw efuents is consequently hard to investigate because of thelarge part of COD removal that is not related to electrochemicalreactions.

The purpose of the present study was to identify the electro-chemical performance that microbial bioanodes formed in pulpand paper efuents could reach, using well controlled electro-

118 S.F. Ketep et al. / Bioresourcediluting the raw efuent with fresh water. Higher COD values werereached by mixing efuents collected at the outlet and the inlet ofthe anaerobic process (inlet COD up to 5200 mg/L).2.2. Bioanode formation and continuous reactor operation

Working electrodes were at 2 cm 5 cm 0.5 cm graphiteplates (Goodfellow) electrically connected via screwed titaniumwires (Alfa Aesar). Current densities were calculated with respectto the 10 cm2 projected surface area. Before use, the graphite elec-trodes were cleaned by 1 h in 1.0 M HCl, 20 min rinsing with dis-tilled water, 1 h in 1.0 M NaOH and nal thorough rinsing withdistilled water. Auxiliary electrodes were 90% Platinum10% Irid-ium grids (Heraeus). Potentials were controlled versus a saturatedcalomel electrode (SCE; potential +0.24 V/SHE; Radiometer, Copen-hagen) using a multi-channel potentiostat (Bio-Logic SA). Allexperiments were achieved under constant applied potential(chronoamperometry, CA) at 0.3 V/SCE. The applied potentialwas periodically suspended to record cyclic voltammetry (CV)curves at 1 mV/s. The reactors were maintained at 25 C in a tem-perature-controlled room and were continuously sparged with N2to maintain anaerobic conditions. Replicates were systematicallyperformed as discussed below.

The bioanodes were rstly formed for around 1 week in batchmode, i.e. with the reactor lled with 500 mL raw efuent in theabsence of any ow. The reactors were then continuously fed withthe raw efuent thanks to a peristaltic pump. The ow rate wasvaried from of 250 to 2000 mL/day by steps of around a week each,which gave hydraulic retention time (HRT) values from 48 to 6 h,respectively. During the experiments, the tank, in which the efu-ent was stored, was kept at 4 C under nitrogen.

Samples were collected daily from the inuents (input) andefuents (output) of the reactors to measure their COD accordingto the ISO 15705 standard micro-method using Merck reagentsand a WTW Photolab 6000 spectrophotometer.

2.3. Calculations

Only the denitions of the conventional parameters are recalledin this section. The modelling work is described in Section 3.2.1.Hydraulic retention time (HRT, s) is the ratio of the solution vol-ume in the reactor VS (0.5 103 m3) to the volumetric ow rate( _V ; m3=s):

HRT VS_V

1

The total COD removal (DCOD, mg/L or g/m3) was dened as thedifference between the input and output COD values:DCOD CODinput CODoutput 2and the total COD removal yield (%) was:

YCOD DCODCODinput 100 3

The average YCOD values reported in the tables and gures werecalculated from 5 to 7 daily measurements of CODinput andCODoutput. Experimental errors on COD abatements were calculatedas follows:

2uDCO 2 1DCOinput

u2

DCOinput u2DCOoutput

DCOinputDCOoutput2DCO2input

u2DCOoutput

vuut ; 4

where u = 30 mg/L is the experimental error on each COD measure-ment related to the measurement kit. In all cases, experimental er-rors on YCOD were of 56%. Coulombic efciency (CE) was calculatedas the charge passed through the circuit (Q), calculated by integrat-ing the current over a given period of time, to the charge that couldbe produced by complete oxidation of the efuent supplied duringthe same period (Qmax):

nology 149 (2013) 117125CE QQmax

5

In continuous mode, assuming stable values of the concentra-tion in the inlet and outlet ows, the theoretical charge is:

Qmax nF _VDtCODinput CODoutput=MO2 6where n = 4 is the number of electrons exchanged per mol of oxy-gen, F = 96 485 C/mol e is the Faraday constant, Dt is the periodof time considered (s), COD is the chemical oxygen demand in theinput and output ows (mg/L or g/m3),MO2 32 g=mol is the molarmass of oxygen, so that:

CE 32F4

Q

Dt _V DCOD 7

3. Experimental results and modelling

3.1. Effect of the hydraulic retention time (HRT)

3.1.1. Evolution of current density with timeThe reactor was kept in batch mode for 5 days to form the bio-

anode from the raw efuent without addition of any complement

or substrate. It was then switched to the continuous mode, whereit was fed with raw efuent at constant ow rate. The current in-creased as soon as the ow was started and presented a brief peakreaching 6.4 A/m2, which was hard to explain (Fig. 1A). During therst period, an attempt was made to stabilize HRT around 48 h butit varied between 40 and 50 h because ow rate stability was tech-nically difcult to ensure at the very low value of 250 mL/day. Dur-ing this rst phase at HRT around 48 h, and except for the initialhigh current peak, the bioanode produced an average current den-sity of 3.5 0.5 A/m2. Then, decreasing HRT to 24, 12 and 6 h bysuccessive periods of 1 week each led to identical current densitiesof 4.5 0.5 A/m2.

The performance of the microbial anode was also characterizedover the whole range of potential by recording cyclic voltammetry(1 mV/s). It was veried that CV recorded at the beginning of theexperiment with the clean electrode (day 0) did not give any cur-rent (data not shown), conrming that the oxidation currents ob-served in the subsequent CVs were related to the formation of anelectroactive biolm. Fig. 1B reports a typical CV obtained duringreactor operation (day 18) and the CV recorded at the end of the

S.F. Ketep et al. / Bioresource Technology 149 (2013) 117125 119Fig. 1. Microbial bioanode formed and operated in continuous mode under constant povalues. The rst 5 days corresponded to the bioanode formation in batch mode; (B) cyclarization at 0.3 V/SCE. (A) Current density as a function of time at different HRTlic voltammetry (1 mV/s) at day 18 and at the end of the experiment (day 29).

experiment. The nal CV showed greater hysteresis, indicating thatthe 29-day old bioanode produced larger capacitive currents. Overthe course of the experiment, microbial cells producedexopolymeric substances resulting in a nal compact biolm(Supplementary data, Fig. 1), which accumulated ions and othercompounds coming from the efuent. The interfacial capacity con-sequently increased with ageing as shown by the CV record. Wehave already observed this phenomenon during ageing of othermicrobial bioanodes formed in raw industrial efuent (data to bepublished).

For each CV, the current density measured at 0.3 V/SCE wasof the same order of magnitude as the current density obtainedduring chronoamperometry (forward scans: 4.3 A/m2 at day 18and 2.8 A/m2 at day 29). The scan rate of 1 mV/s was consequentlylow enough to characterize the stationary behaviour of the elec-trodes. Maximum current densities of 6 A/m2 obtained above0.2 V/SCE. CV showed that the potential of 0.3 V/SCE that waschosen here did not ensure the highest possible current density.5 A/m2 could be reached at 0.2 V/SCE and close to 6 A/m2 at

120 S.F. Ketep et al. / Bioresource Tech0.0 V/SCE. Nevertheless, when the objective is to design efcientmicrobial electrochemical systems, a bioanode must ensure thehighest possible current density at the lowest possible potential.Here the applied potential of 0.3 V/SCE ensured currents of75% of the maximum current and was low enough for the bioan-ode to be efcient in a microbial electrochemical system. The va-lue of 0.3 V/SCE was consequently kept for the entire studybecause it represented a satisfactory compromise between thetwo opposing purposes: increasing current while decreasing thepotential.

The experiment was replicated twice. One replicate producedidentical stable current densities around 4.5 0.5 A/m2, the otherslightly lower values around 3 0.5 A/m2 (Table 1). These currentdensities were considerably higher than the values reported inthe literature so far for MFCs using raw efuents; for instance itwas 3 times the highest current obtained with paper-making efu-ent (1.6 A/m2 fromMathuriya and Sharma, 2009a, see details in Ta-ble 1 in Supplementary data). Moreover, these current densitieswere obtained with at non-porous electrodes. The current densitywas consequently not enhanced by any porous or three-dimen-sional effect. As expected, the electrochemically well-controlledexperimental conditions used here (3-electrode system, small elec-trode surface area vs. solution volume) succeeded in enhancing thecurrent densities generated by the bioanodes. It has to be con-cluded that current densities reported so far in the literature with

Table 1Total COD removal yield, Coulombic efciency (CE), and COD removal related toelectricity production (eDCOD) at different HRT values. (A) From experiment reportedin Fig. 1; (B) and (C) replicates with different efuent samples. Each HRT wasmaintained from 3 to 7 days.

HRT(h)

CODinput(mg/L)

CODoutput(mg/L)

J average(A/m2)

CODremovalyield (%)

CE eDCOD(mg/L)

A48 1623 1257 3.5 23 25 9124 1443 1145 4.5 21 20 6012 1389 1097 4.5 21 12 356 1284 1046 4.5 19 8 19

B41 1260 978 3.3 22 16 4524 1080 813 3.0 25 12 3214 1300 1009 3.5 22 4 12

C

26 1300 1000 5.0 23 18 5418 1000 800 4.5 20 19 388 1300 900 5.0 31 6 24MFC set-ups were signicantly lower than the maximum currentthe bioanodes could reach, because of limitation due to the devices.The results obtained with MFC set-ups probably underestimatedthe current that microbial bioanodes could ensure in raw industrialefuents.

3.1.2. COD removal and Coulombic efciencyDuring the whole duration of the experiment (1 month) the

COD of the inlet efuent decreased from 1623 to 1284 mg/L be-cause of spontaneous consumption of the organic matter it con-tained, in spite of rigorous storage conditions under nitrogenand at 4 C. Consequently, it was necessary to measure COD inboth the inlet and the outlet ows for each measurement to ob-tain an accurate evaluation of the performance of the reactors.Average COD removal and CE are reported in Table 1 for eachapproximately 1 week period. The total COD removal yield (YCOD)remained constant around 21% 2% and did not decrease signi-cantly when the retention time was reduced. Two replicates gavesimilar COD removals, which were not signicantly affected bythe HRT value. This experimental observation was different fromthose reported by similar studies but carried out with MFC set-ups, which stipulated that COD removal was proportional toHRT (Liu et al., 2004; Min and Logan, 2004; You et al., 2006; DiLorenzo et al., 2010; Huang et al., 2009). A likely explanation isthat a part (2030%) of the organic matter in the efuent usedhere was easy to oxidize, while the other part was recalcitrant.The easily-oxidizable part was oxidized so fast that even the low-est HRT was sufcient to achieve its complete consumption. Incontrast, the recalcitrant part was not attacked even at the lon-gest HRT. This assumption remains to be conrmed. Actually, thiskind of efuent is generally 80% biodegradable in conventionalaerobic conditions.

Coulombic efciencies increased from 8% to 25% when HRT wasincreased from 6 to 48 h in reactor A. The replicates conrmed thesame trend, with CEs rising from 4% to 16% when HRT increasedfrom 14 to 41 h. Similar relationships have already been observedin several works performed with MFCs (Liu et al., 2004; Huang andLogan, 2008a; Di Lorenzo et al., 2010). Huang et al., using a singlechamber MFC fed with paper efuents, have obtained CE from 13%to 39% with HRT increasing from 2 to 25 h. (Huang et al., 2009). TheCoulombic efciency represents the ratio of the organic matter oxi-dized by the bioanode with respect to the total COD abatement. CEvalues always less than 25% indicated that the electrochemicalreactions were responsible for a minor part of the total COD re-moval. Low CEs, of the same order of magnitude or even less, havecommonly been reported in the literature for MFCs using real rawefuents (see Section 1).

Control experiments with non-polarized electrodes were per-formed strictly in parallel for some of the electrochemical reactors,i.e. at exactly the same time, with the same HRTs and fed with thesame efuent (see Table 2 in Supplementary data for example). Inthe control reactors, COD removals were generally a few percentlower than those of the corresponding polarized reactors but, ina few cases, the control gave higher COD removal. Unfortunately,only a few similar control experiments with non-polarized elec-trodes have been reported in the literature (Liu et al., 2004; Huangand Logan, 2008a; Huang et al., 2009). When available, they alsoshow similar small differences between non-polarized controlsand polarized reactors.

The low CE obtained under polarization and the high CODremovals observed in the non-polarized control experimentsconrmed that the main part of the total COD removal wasnot related to the electrochemical assistance. The spontaneous

nology 149 (2013) 117125COD decrease was hard to control because of the complex com-position of the organic matter contained in real industrialefuents.

Tech3.2. Modelling the effect of the hydraulic retention time (HRT)including COD variations

3.2.1. TheoryRaw efuents underwent spontaneous COD abatement not re-

lated to electrochemical pathways. CE expressed with respect tothe total COD removal (Eq. (8)) was consequently not the most rel-evant parameter because it expressed the small electrochemicalcontribution under study with respect to a large COD removal thatwas hard to grasp and not related to the electrochemical process.To investigate the real impact of the electrochemical pathwayson raw industrial efuent, a theoretical approach is proposed here,based only on the COD abatement related to the electrochemicalreactions.

The electrochemically removed COD (eDCOD) was dened asthe amount of organic matter (COD) that was oxidized by the bio-anode in a given period of time. The experimental value eDCODexp,(mg/L or, identically, g/m3) was calculated from the electricalcharge extracted from the system by the bioanode:

eDCODexp AVS324F

Z tDtt

Jdt 8

where A is the electrode surface area (m2), VS is the solution volume(m3), 32 is the molar mass of oxygen (g/mol), n = 4 is the number ofelectrons exchanged per mol of oxygen oxidized (mol1), F = 96485is the Faraday constant (C/mol), J is the current density (A/m2) and tis the time (s).

The theoretical value eDCODtheo was obtained through the massbalance on the continuous reactor involving only the electrochem-ical reaction:

_VeDCODtheo A J 324F 9

Dividing each part by _V and using the denition of HTR (Eq. (1))led to the expression for the theoretical electrochemical COD re-moval (eDCODtheo):

eDCODtheo 324FAVS

HRT J 10

A conventional MichaelisMenten-type law was introduced toexpress the variation of the current density with COD:

J JmaxCOD

KCOD COD ; 11

where Jmax (A/m2) is the maximum current that could be providedby the bioanode at the applied potential in optimal COD conditions,and KCOD the MichaelisMenten equivalent constant in terms ofCOD (g/m3 or mg/L). Such a relationship between substrate concen-tration (mM) and current density has been established for acetate(Kato Marcus et al., 2007; Hamelers et al., 2011) and has also beenshown to be suitable for use with COD for complex substrates (Minand Logan, 2004). Actually, this equation has generally been re-ferred as a Monod-type law, but we prefer to use the MichaelisMenten qualication because the phenomenon is controlled bythe enzymatic mechanisms that control substrate oxidation ratherthan being related to microbial growth for which the Monod lawwas established.

Finally, according to the well-stirred hypothesis, the COD con-centrations of the outlet ow and inside the reactor were assumedto be equal. This hypothesis was justied here by the continuousstirring ensured by nitrogen sparging. The electrode was exposedto a COD concentration equal to those of the outlet ow, whichgave the expression for J:

S.F. Ketep et al. / BioresourceJ JmaxCODoutput

KCOD CODoutput 12and led to the theoretical expression for eDCODtheo:

eDCODtheo 324FAVS

HRT JmaxCODoutput

KCOD CODoutput : 13

This expression gives the COD removal due to the electrochem-ical process as a function of HRT and the COD in the outlet ow.The function eDCODtheo is not biased by the large part of COD thatis consumed by non-electrochemical routes and gives an accurateestimation of the amount of COD consumed by the electrochemicalprocess.

Eq. (13) conrm that the efciency of a bioanode is proportionalto the electrode surface area/solution volume ratio, as is the casefor any conventional electrochemical process. Here the volumewas deliberately chosen high (0.5 L) with respect to the electrodesurface area (10 cm2), giving a small A/VS ratio of 2 m2/m3. Gener-ally, studies carried out with MFC set-ups have involved larger A/VSratios: 8.2 m2/m3 (Mathuriya and Sharma, 2009a) 10.7 m2/m3 (Liet al., 2007) or 18.2 m2/m3 (Huang et al., 2009). The choice madehere succeeded in promoting high current densities as targeted(see Section 1) but it was detrimental to the electrochemical ef-ciency for organic removal. Eq. (13), which can be transformed to:

e DCODtheoJ

Constant AVS

HRT 14

expresses the opposition between the two targets: promoting largeelectrochemical COD abatement is favoured by high values of the A/VS ratio, while targeting high current densities requires low A/VSvalues.

3.2.2. Theoretical tting of the experimental dataThe theoretical eDCODtheo equation (Eq. (13)) was tted to the

experimental values extracted from the charge (eDCODexp) byadjusting the two parameters Jmax and KCOD (Fig. 2). The values ofJmax = 6 A/m2 led to satisfactory tting of experiment A, exceptfor the rst point recorded at HRT = 48 h. Jmax = 6 A/m2 was per-fectly realistic value with respect to the Jmax given by the CV curves(Fig. 1). The rst phase of experiment A at HRT = 48 h gave lowercurrent (3.5 A/m2) than the subsequent phases (4.5 A/m2) (Fig. 1).The Jmax parameter was consequently decreased in the same pro-portion (4.7 A/m2 instead of 6 A/m2) for the point at HRT 48 h. Thisadaptation led to a very satisfactory t of the whole experimentand it also showed the consistency of the model with the experi-mental behaviour. For experiment B, the best t was obtained withJmax of 3 A/m2, which was also consistent with the lower currentdensities obtained in this experiment. Experiment C was ttedwith Jmax = 6 A/m2 identically to experiment A. The KCOD constantremained unchanged, equal to 400 mg/L for all experiments.

3.2.3. Model validation on a large range of inlet COD concentrationsThe objective of the following experiments was to explore the

model robustness on a larger range of COD values obtained bydiluting the inlet efuent. Two reactors (D and E) were run in par-allel and fed with the same efuent (duplicates). As previously, thebioanodes were rstly formed by 5 day polarization in batch. Thereactors were then switched to continuous mode with an HRT of24 h. The two reactors exhibited slight differences in currents butshowed identical global behaviours (Fig. 3). As in the previousexperiments, there was an initial unstable current phase for afew days after the reactor had been set in continuous mode. There-after the reactors produced quite stable currents (from day 12 today 18), which slowly decreased because of the slow spontaneousdecrease of the inlet COD concentration. After day 18, the HRT was

nology 149 (2013) 117125 121decreased by periods of around 1 week and the COD inlet was con-comitantly reduced from 1650 to 500 mg/L by diluting the efuentwith fresh water, resulting in a progressive decrease of the current

Tech40

60

80

100

120

e-

COD

(mg/

L)

A experimentalA modelB experimental B modelC experimentalC model

122 S.F. Ketep et al. / Bioresourceproduced. When COD was nally increased to 900 and then1500 mg/L in the last two phases (after day 35), the bioanodesrecovered current values of the same order of magnitude as inthe starting phase, showing a nice stability of the current for morethan 6 weeks.

The theoretical values eDCODtheo were tted to the experimen-tal data with Jmax = 4.75 A/m2 and KCOD = 400 mg/L for bothduplicates (Fig. 4A). The model gave an accurate representationof the experiments, with errors between theory and experimentof the same order of magnitude as the experimental deviationsbetween the duplicates. It can be concluded that Eq. (13) gave arobust and accurate description of the reactor behaviours in a largerange of HRT (648 h) and inlet COD values (5001650 mg/L).Among the two adjustable parameters, the KCOD value did not varyfrom one experiment to the other, showing that it was probably

0

20

0 5 10 15 20

HR

Fig. 2. Comparison of the experimental (full diamonds) and theoretical (empty diamondreported in Table 2.

Fig. 3. Current density produced under constant polarization at 0.3 V/SCE during continD and E). The rst 5 days corresponded to bioanode formation in batch mode.nology 149 (2013) 117125related to the nature of the efuent and to the properties of themicrobial population it contained. The model nally had onlyone parameter to be adjusted from one experiment to the other,Jmax, the value of which varied consistently with the current re-corded experimentally.

3.3. Discussion on the coupled experimental-model approach

All experiments were tted with a single value of KCOD of400 mg/L. A similar relationship used to t the power suppliedby an MFC fed by domestic efuents has led to KCOD of the sameorder of magnitude (461 and 719 mg/L; (Min and Logan, 2004)).Bioanodes formed with acetate (Hamelers et al., 2011) gave Kacetateof 2.2 mM in acetate concentration, which corresponds to 180 mg/L in COD. The present study conrmed the wide applicability of a

25 30 35 40 45 50

T (h)s) COD removals related to the electrochemical reaction for the three experiments

uous operation at different HRTs with two identical reactors run in parallel (reactors

800t C

Tech0

20

40

60

80

100

120

0 200 400 600Inle

e- D

COD

(m

g/L)

ExperimentalModel

A

120

140B

S.F. Ketep et al. / BioresourceMichaelisMenten law for microbial bioanodes, whatever thesubstrate nature. It should be kept in mind that the COD value tobe taken into account is the value to which the electrode is ex-posed. In the case of continuous operation under the hypothesisof a well-stirred reactor, the COD to be considered is the value ofthe outlet ow.

Studies that have investigated the effect of COD on MFC perfor-mance have generally observed lower CE for higher COD, in CODranges up to around 1000 mg/L (Min et al., 2005; Huang et al.,2009; Lee et al., 2010; Wen et al., 2010; Lefebvre et al., 2013). Dif-ferent hypotheses have been proposed depending on the feedingmode of the MFC. Under fed batch, various works using syntheticxylose efuent (Huang and Logan, 2008b) or industrial paper efu-ent (Huang and Logan, 2008a; Huang et al. 2009) have stated thatthe accumulation of intermediates in the reactor may be responsi-ble for lowering CEs at high COD. Nevertheless, this possible effectis ruled out when the MFC is operated in continuous mode. Mostauthors attribute CE lowering with COD increase to the complexityof the organic matter present in the efuent but with no obviousexplanation, except for Nam et al. (2010) who explained that thecomplex substrates present in the wastewater induced a highinternal resistance, which limited power generation. The decreaseof CE would consequently be due to a limitation that is not relatedto the bioanode itself (Nam et al., 2010). This assumption explainswhy CE decrease with COD was not observed here, because

0

20

40

60

80

100

0 1000 2000Inlet C

e-C

OD (m

g/L)

Fig. 4. Comparison of the experimental (full diamonds and triangles) and theoretical (emfunction of the COD of the inlet ow; (A) reactors D and E with different HRT (see Table 2the COD range to high values.1000 1200 1400 1600 1800OD (mg/L)

F experimental F model

nology 149 (2013) 117125 123experiments carried out in 3-electrode set-ups are not affectedby the internal resistance of the cell.

The model predicted a straightforward proportionality betweeneDCOD and HRT, which was not obvious from the raw experimen-tal data alone. For instance, such proportionality was not directlyvisible in Fig. 2 or when the data from Fig. 4A were plotted as afunction of HRT (data not shown). Actually, the points reportedin Fig. 2 corresponded to different inlet COD values ranging from1000 to 1623 mg/L because of non-controlled drift of the efuentand the efuent variability depending on sampling time at theindustrial site. Data reported in Fig. 4A show an even larger varia-tion because of deliberate dilutions. It is not possible to compen-sate the COD drift of the efuent experimentally by addingacetate when the choice has been made to work with raw indus-trial efuent only. In this fairly cumbersome experimental frame-work, the model offered a powerful tool to extract soundconclusions from complex experimental data. For each reactor,the model proved to t the experimental points satisfactorily withvalues of Jmax that were consistent with the experimentalobservations. From the set of ve reactors, fed with differentefuent samples, it could thus be concluded that Eq. (13) wasrespected, meaning that the bioanodes oxidized amounts oforganic matter that increased proportionally to HRT. This propor-tional increase was qualied by the concentration of COD thatremained in the reactor (CODoutput) affecting the current through

3000 4000 5000 6000OD (mg/L)

G experimentalG model

pty diamonds or triangles) COD removals due to the electrochemical reaction as ain Supplementary data); (B) reactors F and G with constant HRT of 24 h, extending

hypothesis was not acceptable here because the high COD efuents

centration of 35 mM of acetate.

vious conclusions that have been extracted fromMFC experiments.

did not produce electricity. It was consequently proposed to

Techa MichaelisMenten law. From a practical point of view, Eq. (13)gave a rst tool for optimizing the reactor design (A, VS) and gavethe operating parameter (HRT) as a function of the bioanode char-acteristics (Jmax, KCOD) depending on the target: favouring COD re-moval or favouring electricity production.

3.4. Using the model to investigate the effect of high inlet CODconcentrations

The initial phase of bioanode formation in batch was avoided inthis section by using a bioanode that had already been used in aprevious reactor. The bioanode was extracted from the initial reac-tor after 75 days of operation in continuous mode and transferredinto a new reactor (reactor F) without any particular care (in par-ticular transfer was made in air). The reactor was then immediatelyput under continuous feed with raw efuent. HRT was maintainedat 24 h in all experiments and only the inlet COD was varied. Thecurrent density started to increase immediately and reached amaximum of 3.8 A/m2 after only 1 week. The inlet COD was variedfrom 500 to 5200 mg/L. Each concentration level was maintainedfor 1 or 2 weeks. The maximum current densities, between 2.6and 3.8 A/m2, were obtained with COD from 1200 to 2800 mg/Land currents were signicantly reduced with the lowest and high-est values of the inlet COD. The experiment replicated in identicalconditions (reactor G) gave similar results. The experimental andtheoretical COD removals due to the electrochemical reaction(eDCOD) were calculated as previously and tting was done withJmax = 4.75 A/m2 and KCOD = 400 mg/L for both reactors (Fig. 4B).

The model tted the experimental eDCOD obtained at low CODinlet well for both reactors but a surprisingly high electrochemicalCOD abatement was obtained for each reactor at the inlet CODs of1460 and 1300 mg/L. These two measurements were obtainedwhen the efuents in the storage tank were replaced by freshly col-lected efuents. In the previous experiments (AE) a fresh efuentwas used at the beginning of the experiment and was rst used for1 week in batch mode to form the bioanode. In contrast, in reactorsF and G, the fresh efuent was not used at the beginning but onlyafter 2 weeks of continuous operation. The possible effect of thefresh efuent was consequently not observed in the previous reac-tors because of the initial batch phase, while here both reactorsshowed a considerably higher electrochemical COD abatementwith fresh efuents. Freshly collected efuents containing easilybiodegradable compounds were more sensitive to the treatmentby the microbial bioanodes.

Chemical analysis of the freshly collected efuent revealed thata signicant part of the COD was composed of volatile fatty acids,mainly including acetic acid (200 mg/L), propionic acid (98 mg/L)and butyric acid (36 mg/L). Trace amounts (

Huang, L., Logan, B., 2008a. Electricity generation and treatment of paper recyclingwastewater using a microbial fuel cell. Appl. Microbiol. Biotechnol. 80, 349355.

Huang, L., Logan, B., 2008b. Electricity production from xylose in fed-batch andcontinuous-ow microbial fuel cells. Appl. Microbiol. Biotechnol. 80, 655664.

Huang, L., Cheng, S., Rezaei, F., Logan, B., 2009. Reducing organic loads inwastewater efuents from paper recycling plants using microbial fuel cells.Environ. Technol. 30, 499504.

Kassongo, J., Togo, C.A., 2011. Evaluation of full-strength paper mill efuent forelectricity generation in mediator-less microbial fuel cells. Afr. J. Biotechnol. 10,1556415570.

Kato Marcus, A., Torres, C.I., Rittmann, B.E., 2007. Conduction-based modeling of thebiolm anode of a microbial fuel cell. Biotechnol. Bioeng. 98, 11711182.

Lee, C.-Y., Chen, J.-H., Cai, Y.-Y., 2010. Bioelectricity generation and organic removalin microbial fuel cells used for treatment of wastewater from sh-market. J.Environ. Eng. Manage. 20, 173180.

Lefebvre, O., Tan, Z., Shen, Y., Ng, H.Y., 2013. Optimization of a microbial fuel cell forwastewater treatment using recycled scrap metals as a cost-effective cathodematerial. Bioresour. Technol. 127, 158164.

Li, Z., Yao, L., Kong, L., Liu, H., 2007. Electricity generation using a bafed microbialfuel cell convenient for stacking. Bioresour. Technol. 99, 16501655.

Liu, H., Ramanathan, R., Logan, B., 2004. Production of electricity during wastewatertreatment using a single chamber microbial fuel cell. Environ. Sci. Technol. 38,22812285.

Liu, H., Cheng, S., Logan, B., 2005. Production of electricity from acetate or butyrateusing a single-chamber microbial fuel cell. Environ. Sci. Technol. 39, 658662.

Logan, B.E., Rabaey, K., 2012. Conversion of wastes into bioelectricity and chemicalsby using microbial electrochemical technologies. Science 337, 686690.

Lu, N., Zhou, S.-G., Zhuang, L., Zhang, J.-T., Ni, J.-R., 2009. Electricity generation fromstarch processing wastewater using microbial fuel cell technology. Biochem.Eng. J. 43, 246251.

Mathuriya, A.S., Sharma, V.N., 2009. Bioelectricity production from variouswastewaters through microbial fuel cell technology. J. Biochem. Technol. 2(1), 133137.

Min, B., Logan, B.E., 2004. Continuous electricity generation from domesticwastewater and organic substrates in a at plate microbial fuel cell. Environ.Sci. Technol. 38, 58095814.

Min, B., Kim, J., Oh, S., Regan, J.M., Logan, B.E., 2005. Electricity generation fromswine wastewater using microbial fuel cells. Water Res. 39, 49614968.

Nam, J.-Y., Kim, H.-W., Lim, K.-H., Shin, H.-S., 2010. Effects of organic loading rateson the continuous electricity generation from fermented wastewater using asingle-chamber microbial fuel cell. Bioresour. Technol. 101, 3337.

Pocaznoi, D., Erable, B., Etcheverry, L., Delia, M.-L., Bergel, A., 2012. Towards anengineering-oriented strategy for building microbial anodes for microbial fuelcells. Phys. Chem. Chem. Phys. 14, 1333213343.

Rabaey, K., Clauwaert, P., Aelterman, P., Verstraete, W., 2005. Tubular microbialfuel cells for efcient electricity generation. Environ. Sci. Technol. 39, 80778082.

Sharma, Y., Li, B., 2010. The variation of power generation with organic substrates insingle-chamber microbial fuel cells (SCMFCs). Bioresour. Technol. 101, 18441850.

Stephenson, R., Mahmood, T., Elliot, A., OConnor, B., Eskicioglu, C., Saha, M.,Ericksen, B., 2012. How microsludge and anaerobic digestion or aerobicstabilization of waste activated sludge can save on sludge management costs.J. Sci. Technol. For. Prod. Processes 2, 2631.

Wen, Q., Wu, Y., Zhao, L., Sun, Q., 2010. Production of electricity from the treatmentof continuous brewery wastewater using a microbial fuel cell. Fuel 89, 13811385.

You, S.J., Zhao, Q.L., Jiang, J.Q., Zhang, J.N., 2006. Treatment of domestic wastewaterwith simultaneous electricity generation in microbial fuel cell under continuousoperation. Chem. Biochem. Eng. 20, 407412.

Youngho, A., Logan, B., 2010. Effectiveness of domestic wastewater treatment usingmicrobial fuel cells at ambient and mesophilic temperatures. Bioresour.Technol. 101, 469475.

Zhang, F., Ge, Z., Grimaud, J., Hurst, J., He, Z., 2013. Long-term performance of liter-scale microbial fuel cells treating primary efuent installed in a municipalwastewater treatment facility. Environ. Sci. Technol. 47, 49414948.

S.F. Ketep et al. / Bioresource Technology 149 (2013) 117125 125

Experimental and theoretical characterization of microbial bioanodes formed in pulp and paper mill effluent in electrochemically controlled conditions1 Introduction2 Experimental section2.1 Effluent2.2 Bioanode formation and continuous reactor operation2.3 Calculations

3 Experimental results and modelling3.1 Effect of the hydraulic retention time (HRT)3.1.1 Evolution of current density with time3.1.2 COD removal and Coulombic efficiency

3.2 Modelling the effect of the hydraulic retention time (HRT) including COD variations3.2.1 Theory3.2.2 Theoretical fitting of the experimental data3.2.3 Model validation on a large range of inlet COD concentrations

3.3 Discussion on the coupled experimental-model approach3.4 Using the model to investigate the effect of high inlet COD concentrations

4 ConclusionsAcknowledgementAppendix A Supplementary dataReferences