Electrochemical Nutrient Recovery Enables Ammonia ToxicityControl and Biogas Desulfurization in Anaerobic DigestionJoachim Desloover, Jo De Vrieze, Maarten Van de Vijver, Jacky Mortelmans, Rene ́ Rozendal,and Korneel Rabaey*

Laboratory of Microbial Ecology and Technology (LabMET), Ghent University, Coupure Links 653, B-9000 Gent, Belgium

*S Supporting Information

ABSTRACT: Organic waste streams can be valorized and reduced in volumewith anaerobic digestion (AD). An often-encountered key issue however is thehigh ammonium (NH4

+) content of certain waste streams. Ammonia (NH3), inequilibrium with NH4

+, is a toxic compound to the methanogenic community,which limits the organic loading rate and endangers process stability. Anelectrochemical system (ES) linked to a digester could, besides recovering thisnutrient, decrease NH3 toxicity through electrochemical extraction. Therefore,two digesters with and without ES attached in the recirculation loop wereoperated to test whether the ES could control NH3 toxicity. During periods ofhigh ammonium loading rates, the methane (CH4) production of the ES-coupledreactor was up to 4.5 times higher compared to the control, which could beexplained through simultaneous NH4

+ extraction and electrochemical pH control.A nitrogen flux of 47 g N m−2 membrane d−1 could be obtained in the ES-coupled reactor, resulting in a current and removalefficiency of 38 ± 5% and 28 ± 2%, respectively, at an electrochemical power input of 17 ± 2 kWh kg−1 N. The anode alsooxidized sulfide, resulting in a significantly lower H2S emission via the biogas. Lastly, limited methanogenic community dynamicspointed to a nonselective influence of the different operational conditions.

■ INTRODUCTION

Anaerobic digestion (AD) is a key technology for stabilizationand valorization of organic waste streams.1,2 In short, thistechnology comprises a stepwise conversion of low-valueorganic compounds into biogas, a mixture of mainly methane(CH4) and carbon dioxide (CO2). Methane is an energy carrierand can be valorized through, for example, a combined heat andpower unit, delivering electricity and heat. Next to biogas, ADalso produces a nutrient-rich digestate that can be applied as afertilizer in agriculture.3 Despite numerous advantages of theAD process, instability is an often-encountered problem thatcan lead to complete failure of the reactor. One of the keycompounds causing instability is ammonia (NH3), especiallywhen treating nitrogen-rich waste streams.4,5 Ammonia is a cellmembrane-permeable molecule for which methanogens,executing the final step in AD, have a low tolerance.6 Moreover,acetoclastic methanogenesis is, in general, more susceptible toinhibition than hydrogenotrophic methanogenesis.6 Thisexposes digesters to a risk of process instability and limits theloading and thus biogas production rate.4 To avoid this,operators usually feed the digester at a lower and safer loadingrate, acclimate the biomass, or co-digest with carbon-richsubstrates to maintain a suitable carbon to nitrogen ratio.5

Other more advanced but also more expensive approaches arestruvite precipitation, anammox, and the use of zeolites.5

In this study, we present an alternative to control NH3toxicity and to maximize resource recovery from AD bycoupling an electrochemical system (ES) to an anaerobic

digester. An ES has the attractive feature that an oxidationprocess (anode) is separated from a reduction process(cathode), typically by an ion selective membrane. By applyinga current to this system, membrane electrolysis can take placeduring which ions can be extracted from anode to cathode orvice versa.7,8 In the context of anaerobic digestion, one can sendthe digestate through an anode compartment, enablingrecovery of valuable nutrients such as ammonium (NH4

+)and potassium (K+). Hence, by combining AD with EStechnology, nutrients can be harvested while simultaneouslylowering the risk for ammonia toxicity. We have recentlydemonstrated the proof of concept of an ES for nutrientrecovery from liquid waste streams.8

Here, we studied the direct coupling of an ES to an upflowanaerobic sludge blanket (UASB) reactor treating molasses. Weinvestigated whether the placement of an anode in therecirculation line had any negative effects on the digester andwhether the ES could stabilize AD performance when exposedto toxic NH3 concentrations. Lastly, we also investigated inwhat manner the ES can affect the quality of the biogasgenerated.

Received: October 1, 2014Revised: December 13, 2014Accepted: December 17, 2014

Article

pubs.acs.org/est

© XXXX American Chemical Society A DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

■ MATERIAL AND METHODSExperimental Setup. Two cylindrical UASB reactors (2.3

L glass reactor with effective volume of 2 L) were constructed,serving as test and control reactors (Figure 1). These reactors

had an internal diameter of 5.4 cm and a total height of 900 cm.For the test reactor, an ES was coupled to the UASB forextraction of cations. This was done by inserting the anodecompartment (5 cm × 20 cm × 2 cm) of the ES in therecirculation loop of the UASB reactor. The anode compart-ment was separated from the cathode compartment (5 cm × 20cm × 2 cm3) by a cation exchange membrane (CEM,Membranes International, U.S.A.). The anode electrode usedwas an IrOx coated titanium mesh electrode 9 (12g m−2 of Ir/Ta = 65/35), with a projected surface area of 5 cm × 20 cm(Magneto Special Anodes, The Netherlands), while thecathode electrode was a stainless steel mesh (5 × 20 cm2,mesh width 564 μm Solana, Belgium). Both electrodes wereplaced close to the CEM and separated by a polytetrafluoro-ethylene (PTFE) spacer with a projected surface area of 5 cm ×20 cm (turbulence promoter mesh, Electrocell, Denmark) toavoid direct contact. At the anode, water was oxidized tooxygen and protons, while at the cathode, water was reduced tohydrogen gas and hydroxyl ions. The ES was controlledgalvanostatically by a VSP multipotentiostat (Biologic, France).The control reactor was also coupled to an ES in therecirculation loop. The ES was equipped with a CEM, butelectrodes were omitted. Hence, the ES of the control setupwas operated in open circuit (no anode and cathode), meaningthat no current could be applied to the system and onlydiffusion driven processes could take place.Reactor Operation. The experiment was conducted under

mesophilic conditions (34 ± 1 °C). The UASB reactors wereinoculated with granular sludge from a full-scale UASB reactor(Brewery Van Steenberge, Belgium) and diluted with tap waterto obtain an initial sludge concentration of 10 g of volatilesuspended solids (VSS) L−1. The UASBs were fed every twohours with tap water-diluted molasses according to the desiredloading rate and were operated at a hydraulic retention time(HRT) of 2 days. The characteristics of the diluted molasses toobtain a desired loading rate of 5 g COD L−1 d−1 are shown inTable 1 (raw composition molasses, Table S1, SupportingInformation). Furthermore, an internal recirculation rate wasapplied over the UASB and anode compartment of 2 L h−1 tomaintain an upflow velocity of 1 m h−1 in the digester. In orderto maintain the same conductivity throughout the experimentalperiod, 4.14 g L−1 NaCl was initially added to the feed of both

UASBs to compensate for any later addition of NH4Cl, whenthe performance was investigated under high nitrogen loadingconditions. Both reactors were pH controlled (DulcometerD1C, Prominent, Germany) with 1 M NaOH. The electro-chemical process parameters of the ES were defined andcalculated according to Desloover et al.8

The cathode compartments of the control and test setupswere fed continuously with 6.4 g L−1 NaCl at 1 L d−1 (HRT of4.8 h) with an internal recirculation rate of 2 L h−1.

Experimental Plan. The experimental plan comprised fourmain phases (Table 2). During Phase I, the organic loading ratewas gradually increased from 1 to 5 g COD L−1 d−1 (Phase Ia).After stable operation, the pH was stepwise (0.25 pH units perweek) increased from 7 to 8 (Phase Ib) to shift the NH4

+/NH3equilibrium more to the direction of NH3 (ratio NH3/NH4

+ =0.11 at pH 8 and 34 °C). The free ammonia fraction wascalculated according to Anthonisen et al.10 Next, the ES of thetest setup was switched on at an applied current density of 10 Am−2 (relative to projected membrane surface area) toinvestigate the impact during low nitrogen loading. At day120, the UASB of the test setup crashed due to clogging andsubsequent malfunction of the pH controller. Hence, both thetest and control reactors were cleaned, and the biomass of thecontrol UASB was split over the test and control setups toinitiate a second start-up (Phase IIIa) during which the organicloading rate and pH were maintained at 5 g COD L−1 d−1 and8, respectively. Also, the ES of the test setup was stepwiseincreased from 5 to 10 A m−2. After a steady-state period(Phase IIIb), the effect of the ES was investigated underperiodically increased nitrogen loading conditions (Phase IV).Therefore, the effect of an operational ES on the test setup wasinvestigated during a period of increased nitrogen loading(Phase IVa), as well as a period during which the extra-addednitrogen was again removed from the feed (Phase IVb). Next,this operational procedure was repeated during a period wherethe ES of the test setup was switched off (Phases IVc and IVd).Finally, after an adaptation period where the ES was switchedon again (Phase IVe), the effect of the ES was investigatedduring additional nitrogen loading up to 2 g N L−1 (Phases IVfand IVg) and where we also allowed the electrochemical cell tocontrol the pH by taking advantage of the acidifying anodereaction.

Chemical Analysis. Liquid samples of the influent andeffluent streams as well as gaseous samples from the headspacewere taken three times a week. Liquid samples were filtered(0.22 μm) and stored at 4 °C until further analysis.

Figure 1. Schematical overview of the experimental setup.

Table 1. Composition Tap Water-Diluted Molasses ToObtain Desired Loading Rate of 5 g COD L−1 d−1.

parameter value unit

pH 5.7 ± 0.3conductivity 8.5 ± 1 mS cm−1

COD 11.7 ± 2.2 g L−1

Kj-N 439 ± 19 mg L−1

TAN 16 mg N L−1

SO42‑ 254 mg L−1

T-P 80 mg P L−1

Cl− 156 mg L−1

TS 11 g L−1

VS 9 g L−1

VSS 0.8 g L−1

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

B

Volatile suspended solids (VSS), Kjeldahl nitrogen (Kj-N),ammonium (NH4

+), chemical oxygen demand (COD), pH,and conductivity were analyzed according to standardmethods.11 Volatile fatty acids (VFA) were, after extraction indiethyl ether, analyzed with a DB-FFAP 12-3232 column (30 m× 0.32 mm × 0.25 μm; Agilent, Belgium) and a flameionization detector (FID) gas chromatograph (GC-2014,Shimadzu, The Netherlands). The gas phase composition wasanalyzed with a compact GC (Global Analyzer Solutions,Breda, The Netherlands). The GC was equipped with twochannels. In channel 1, a Porabond precolumn and Molsieve 5Acolumn were used for CH4, O2, H2, and N2 measurement, andin channel 2 a Rt-Q-bond precolumn and column were used forCO2, N2O, and H2S analysis. Concentrations of gases weredetermined by means of a thermal conductivity detector andwere reported at STP (standard temperature and pressure)conditions. Biogas production was measured with an in-housemanufactured calibrated gas counter.Molecular Analysis. Total DNA was extracted from the

sludge samples using the protocol of Vilchez-Vargas et al.12

DNA quality and quantity of the extracts was analyzed bymeans of a 1% agarose gel and a Nanodrop ND-1000spectrophotometer (Isogen Life Science, IJsselstein, TheNetherlands). Triplicate samples of a 100-fold dilution of theDNA samples were prepared to reach a final DNAconcentration between 1 and 10 ng μL−1.Real-time PCR (qPCR) was performed on a StepOnePlus

Real-Time PCR System (Applied Biosystems, Carlsbad, CA,U.S.A.). The reaction mixture of 15 μL was prepared using theGoTaq qPCR Master Mix (Promega, Madison, WI, U.S.A.) andconsisted of 10 μL of GoTaq PCR Master Mix, 3.5 μL ofnuclease-free water, and 0.75 μL of each primer (finalconcentration of 375 nM), to which 5 μL of template DNAwas added. The qPCR program was performed in a two-stepthermal cycling procedure that consisted of a predenaturationstep of 10 min at 94 °C, followed by 40 cycles of 15 s at 94 °Cand 1 min at 60 °C for total bacteria, using the P338F andP518r primers, as described by Ovreas et al.13 The qPCRprogram for the methanogenic orders Methanobacteriales andthe families Methanosaetaceae and Methanosarcinaceae consistedof a predenaturation step of 10 min at 94 °C, followed by 40cycles of 10 s at 94 °C and 1 min at 60 °C. For quantification ofthe Methanomicrobiales order an annealing temperature of 63°C was used. The primers for the methanogenic ordersMethanomicrobiales and Methanobacteriales and the familiesMethanosaetaceae and Methanosarcinaceae were described by Yuet al.14 Real-time PCR quality was evaluated by means of the

different parameters obtained during analysis with theStepOnePlus software V2.3 (Table S2, Supporting Informa-tion).

Statistical Analysis. All statistical data analysis wereperformed with the statistical software R, version 3.0.2. forWindows. In the case of normally distributed data sets thatwere homoscedastic the regular t test was applied. In the casewhere the data was heteroscedastic, a Welch-modified t test wasused. In the case where the data was not normally distributed,the Wilcoxon Rank Sum (Mann−Whitney U) test was applied.

■ RESULTS AND DISCUSSIONES Has a Temporal Effect on Digester during Low

Nitrogen Loading Conditions. After the start-up phase(Phase Ia), the average CH4 production rate of the test andcontrol reactors during Phase Ib was 949 ± 90 and 950 ± 134mL CH4 L

−1 d−1, respectively (Figure 2, Table S3, Supporting

Information). Hence, gradually adapting the pH from 7 to 8 didnot have a notable effect on methane production. Moreover, byoperating at a high pH, a CH4 content up to 83% could bereached in both the test and control reactors (Table S3,Supporting Information).When the ES of the test setup was switched on at the start of

Phase II, an initial decrease in CH4 production rate of about

Table 2. Overview of Experimental Plan

phase operation period (d)

Ia start-up 1 1−30Ib gradual increase pH 7→8 30−70II switch on ES test setup 70−120IIIa start-up 2 + switch on ES test setup 130−210IIIb steady state 210−225IVa add 1 g N L−1 to feed 225−238IVb remove additional 1 g N L−1 from feed 238−266IVc switch off ES test setup + add 1 g N L−1 to feed 266−275IVd remove additional 1 g N L−1 from feed 275−287IVe switch on ES test setup 287−303IVf add 1 g N L−1 to feed 303−313IVg add 2 g N L−1 to feed + electrochemical acidification (pH 8→7) 313−340

Figure 2. CH4 production in function of time of the test and controlsetup during Phases Ia−II.

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

C

20% could be observed. The temporary negative impact on theperformance of the microbial community was probably due to ashock effect caused by the instantaneous oxygen and protonproduction by the anode reaction, in combination with a higherNaOH dosage to counteract acidification (Table S3, SupportingInformation). The produced oxygen represented 7% of theCOD loading rate at 10 A m−2 but was consumed because noO2 could be detected in the biogas. Moreover, the conductivityof the test (19.3 ± 2.7 mS cm−1) and control (21.1 ± 3.2 mScm−1) reactors were not significantly different (Table S3,Supporting Information, p > 0.05), due to membraneelectrolysis.After 40 days of operation, the test reactor recovered and

reached again the performance of the control reactor (Figure2).Electrochemical NH3 Toxicity Control during High

Nitrogen Loading Conditions. After the crash of the testreactor and a second start-up (Phase IIIa), both reactors againreached equal performance (Phase IIIb, p > 0.05), with a CH4production rate of 925 ± 94 and 848 ± 66 mL CH4 L

−1 d−1 inthe test and control reactors, respectively (Figure 3A, Table 3).When the nitrogen loading was increased by adding an

additional 1 g N L−1 to the feed (Phase IVa), a 43% decrease inthe CH4 production rate could be observed for the controlreactor, while the test reactor was able to maintain its

performance (Figure 3A). Furthermore, the decline in CH4production of the control reactor coincided with theaccumulation of VFA up to 2700 mg COD L−1, whereas noVFA could be detected in the test reactor (Figure 3B). VFAaccumulation is a strong sign of methanogen inhibition, and asdescribed in other studies, this was most probably caused by thehigh ammonium content in combination with a high pH.6,15

The fact that the test reactor outperforms the control reactorcan thus be explained by an on average 23% lower ammoniumlevel in the test reactor caused by membrane electrolysis(Figure 3C, Table 3). By omitting the additional nitrogen fromthe feed (Phase IVb), the control reactor was able to partiallyrecover over a period of 30 days (Figure 3A).Repeating this procedure with a nonworking ES of the test

reactor (Phases IVc and IVd) resulted in a decrease inperformance of both the test and control reactors (Figure 3A).The ammonium levels in both reactors were identical (Figure3C), and also, this time not only VFA accumulation could beobserved in the control reactor, but also in the test reactor(Figure 3B). These findings prove that ammonium extractionby the ES was essential to maintain constant increased methaneproduction values under high nitrogen loading conditions.Most likely, the corresponding average NH3 concentrationduring Phase IVa in the test reactor (94 mg N L−1) did notreach a level that inhibited the methanogenic community,

Figure 3. CH4 production (A), VFA concentration (B), NH4+ concentration (C), and H2S content in the biogas (D) in function of time of the test

and control setups during Phases IIIa−IVg. The labels presented in this figure account for all graphs.

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

D

whereas inhibition might have occurred in the control reactorwere a higher average NH3 concentration was present (118 mgN L−1). Indeed, reported inhibitory NH3 concentrations arewithin a range of 80−150 mg N L−1 and are dependent on theoperational conditions and degree of adaptation.4,5 Removal ofthe extra-added NH4

+ from the feed (Phase IVd) caused

instantaneous recovery of the test reactor, while the controlreactor seemed to reach an inhibited steady state,4 which wason average 43% lower in performance compared to the testreactor.Gradually switching on the ES of the test reactor from 5 to

10 A m−2 caused again a temporary negative effect, comparable

Table 3. Overview of Parameters in Test and Control Setups during Phases IIIa−IVca

Phase IIIa (n = 27) Phase IIIb (n = 6) Phase Iva (n = 6) Phase IVb (n = 12) Phase IVc (n = 4)

parameter test control test control test control test control test control

UASB (with anodecompartment)

CH4 production(mL CH4 L

−1d−1)618 ± 173 626 ± 155 925 ± 94 848 ± 66 890 ± 51 571 ± 121 940 ± 64 740 ± 63 749 ± 147 551 ± 107

efficiencyb (%) 36 ± 10 36 ± 9 52 ± 2 50 ± 2 54 ± 3 34 ± 8 56 ± 5 44 ± 4 45 ± 10 34 ± 7

sCOD effluent(g COD L−1)

ND ND 0.96 ± 0.26 2.04 ± 0.37 0.97 ± 0.02 2.82 ± 0.40 0.93 ± 0.12 3.71 ± 1.21 3.30 ± 0.99 4.45 ± 0.75

total VFA(mg COD L−1)

ND ND BDL 630 ± 306 BDL 1939 ± 704 BDL 1983 ± 429 1575 ± 817 3119 ± 730

acetate (mg L−1) ND ND BDL 249 ± 95 BDL 1218 ± 459 BDL 1299 ± 293 1133 ± 645 2519 ± 669

propionate (mg L−1) ND ND BDL 370 ± 203 BDL 670 ± 259 BDL 609 ± 163 408 ± 147 510 ± 71

TAN (mg N L−1) ND ND 250 ± 30 294 ± 20 823 ± 116 1069 ± 51 302 ± 54 376 ± 69 1001 ± 199 1001 ± 195

conductivity (mS cm−1) 20.1 ± 2.1 19.7 ± 2.8 16.0 ± 0.3 17.5 ± 0.5 18.3 ± 1.3 17.8 ± 0.5 18.6 ± 1.9 20.0 ± 1.3 22.3 ± 1.3 21.6 ± 1.2

pH (−) 7.9 ± 0.1 7.9 ± 0.1 7.9 ± 0.1 8.0 ± 0.1 7.9 ± 0.1 8.0 ± 0.1 8.0 ± 0.2 7.9 ± 0.1 8.0 ± 0.1 8.0 ± 0.1

NaOH dosage (mL d−1) 106 ± 30 81 ± 10 142 ± 28 73 ± 5 166 ± 16 83 ± 4 155 ± 15 80 ± 12 95 ± 17 86 ± 10

CH4 (%) 82 ± 4 84 ± 3 92 ± 2 94 ± 1 89 ± 1 93 ± 1 91 ± 7 91 ± 3 89 ± 1 89 ± 2

H2S (%) 0.25 ± 0.15 0.27 ± 0.06 0.05 ± 0.07 0.22 ± 0.03 BDL 0.22 ± 0.04 0.05 ± 0.08 0.19 ± 0.09 0.51 ± 0.10 0.23 ± 0.01

Cathode

TAN (mg N L−1) ND ND 42 ± 21 10 ± 6 220 ± 20 38 ± 12 60 ± 15 23 ± 8 24 ± 11 44 ± 9

conductivity (mS cm−1) 17.0 ± 5.2 14.1 ± 3.8 22.4 ± 5.5 9.4 ± 2.7 20.6 ± 2.0 12.0 ± 1.1 24.0 ± 3.3 12.2 ± 1.4 12.3 ± 1.5 12.6 ± 2.3

pH (−) 9.9 ± 2.2 7.9 ± 0.3 12.5 ± 0.1 7.8 ± 0.2 12.1 ± 0.1 7.6 ± 0.1 12.4 ± 0.1 7.7 ± 0.2 8.5 ± 0.3 7.9 ± 0.1

Electrochemical

N fluxc (g N m−2 d−1) ND ND 4 ± 2 1 ± 1 19 ± 2 3 ± 1 5 ± 1 2 ± 1 2 ± 1 3 ± 0

NH4+ current efficiency

(CE,%)ND ND 3 ± 1 NA 15 ± 1 NA 4 ± 1 NA NA NA

NH4+ removal efficiency

(RE,%)ND ND 14 ± 6 3 ± 2 22 ± 3 3 ± 1 17 ± 1 6 ± 2 3 ± 1 4 ± 0

cell voltage (V) ND ND 3.58 ± 0.65 NA 4.72 ± 0.91 NA 4.72 ± 0.91 NA NA NA

energy input(kWh kg−1 N)

ND ND 307 ± 182 NA 60 ± 6 NA 251 ± 37 NA NA NA

an = number of data points. ND: not determined. BDL: below detection limit. NA: not applicable. bCOD to CH4 conversion efficiency at STPconditions and relative to the amount of COD fed to the reactor. cRelative to the projected membrane surface area.

Figure 4. Relative abundance of methanogens (Methanomicrobiales,Methanobacteriales,Methanosarcinacea, andMethanosaetaceae) in the test (T) andcontrol (C) reactors throughout the experimental period.

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

E

to Phase II. However, this time the recovery took only 16 days(Phase IVe). This shows that gradual increase of the currentdensity is necessary to allow adaptation of the microbialcommunity to the new conditions. The limited degree ofchange in the methanogenic community (Figure 4) indicatesthat adaptation of the microbial community took place throughphysiological responses rather than microbial communitycomposition variation.During the final two phases (Phases IVf and IVg), the

nitrogen loading was again increased by addition of 1 (PhaseIVf) and 2 (Phase IVg) g N L−1 to the feed. In contrast toPhase IVa, the CH4 production rate of the test reactor startedto decrease dramatically (Figure 3A), which can be explainedby the fact that the methanogenic community was also stressedin the test reactor because residual VFA were present (Figure3B). A further increase in the nitrogen loading (additional 2 gN L−1) led to a minimum CH4 production rate of 322 mL ofCH4 L

−1 d−1 at day 317, which was equal to the performance ofthe control reactor (Figure 3A). Clearly, the extraction ofammonium by the ES was insufficient to decrease the NH4

+

and hence also the NH3 concentration below a toxic level(Figure 3C).Therefore, from day 317 onward, not only the extraction

capacity of the ES was utilized but also the ability to acidify andhence control the pH of the reactor through the use of theacidifying anode oxidation reaction. Until day 317, thegenerated protons by the ES were counteracted by NaOHaddition in order to operate the test and control setups at thesame pH. As a result, NaOH addition to the test reactor wasalmost double of the control reactor whenever the ES wasswitched on (Tables 3 and 4). This, however, did not generatea significant (p > 0.05) difference in conductivity between bothsetups, as the base addition to the test reactor was compensated

by membrane extraction of the ES. By steadily decreasing thebase dosage of the test reactor down to the level of the controlreactor, the pH of the test reactor evolved to 7.1 (Table 4), andas such, the acidifying effect of the ES could be investigated.The test reactor completely recovered and reached a CH4production rate of 856 ± 38 mL of CH4 L

−1 d−1 during the lastseven days of operation, while the control reactor remained at a4.5 times lower CH4 production rate of 192 ± 10 mL of CH4L−1 d−1. This coincided with a steep decrease in VFA belowdetection limit in the test reactor and VFA accumulation up to7500 mg L−1 in the control reactor. From this, we can concludethat, next to electrochemical NH4

+ extraction, also electro-chemical pH control is a powerful tool allowing efficient NH3toxicity control in this study. Moreover, electrochemical NH4

+

extraction generates added value as it allows for recovery of thisnutrient here under the form of a H2/NH3 gas mixture. Thisgas mixture can easily be separated via, for example, ammoniacondensation, thus delivering a concentrated liquid ammoniumstream and a purified H2 gas stream for injection in theanaerobic digester.

Electrochemical Nutrient Recovery. The concept ofelectrochemical nutrient recovery has been demonstratedpreviously, both with an electrochemical 8 as with abioelectrochemical system,16,17 whether or not in combinationwith power consumption or production. This concept was alsoapplied in this study, especially during the final phase (PhaseIVg). In this phase, the need for a high ammoniumconcentration in order to obtain a high extraction efficiencywas again shown (Figure 4C). Here, NH4

+ could be extracted ata flux of 47 ± 6 g N m−2 membrane d−1, resulting in a removaland current efficiency of 36 ± 6% and 27 ± 3%, respectively(Table 4). In terms of electrochemical power input, theammonium could be extracted at 17 ± 2 kWh kg−1 N. This is

Table 4. Overview of Parameters in Test and Control Setups during Phases IVd−IVga

Phase IVd (n = 5) Phase IVe (n = 7) Phase IVf (n = 4) Phase IVg (n = 12)

parameter test control test control test control test control

UASB (with anode compartment)CH4 production (mL CH4 L

−1d−1) 973 ± 87 537 ± 39 798 ± 136 507 ± 33 786 ± 139 458 ± 22 644 ± 188 252 ± 87efficiencyb (%) 58 ± 8 33 ± 4 50 ± 9 32 ± 3 48 ± 7 30 ± 3 40 ± 11 16 ± 5sCOD effluent (g COD L−1) 2.1 ± 0.7 4.7 ± 0.2 2.7 ± 0.8 4.8 ± 0.3 3.5 ± 1.4 5.7 ± 0.8 3.1 ± 1.0 7.0 ± 0.4total VFA (mg COD L−1) 780 ± 529 3239 ± 281 1019 ± 386 3735 ± 210 3052 ± 1493 3291 ± 686 1459 ± 1041 6252 ± 1000acetate (mg L−1) 504 ± 326 2611 ± 237 840 ± 291 2834 ± 140 2465 ± 1096 2552 ± 528 1204 ± 876 5168 ± 878propionate (mg L−1) 260 ± 180 545 ± 51 86 ± 43 442 ± 84 233 ± 182 312 ± 96 78 ± 53 440 ± 35TAN (mg N L−1) 504 ± 128 482 ± 121 284 ± 16 341 ± 26 703 ± 198 1040 ± 120 1527 ± 92 1889 ± 113conductivity (mS cm−1) 21.7 ± 0.6 23.6 ± 0.9 21.5 ± 2.6 24.4 ± 2.3 17.7 ± 0.9 21.3 ± 3.6 23.6 ± 1.8 27.0 ± 2.1pH (−) 8.1 ± 0.1 8.0 ± 0.1 8.0 ± 0.1 8.0 ± 0.1 8.0 ± 0.1 7.9 ± 0.1 7.1 ± 0.2 7.9 ± 0.1NaOH dosage (mL d−1) 72 ± 15 82 ± 5 140 ± 43 94 ± 24 179 ± 26 88 ± 11 109 ± 39 82 ± 5CH4 (%) 89 ± 0 90 ± 1 89 ± 2 91 ± 1 93 ± 2 90 ± 1 64 ± 11 90 ± 3H2S (%) 0.27 ± 0.10 0.22 ± 0.02 0.11 ± 0.03 0.19 ± 0.03 0.08 ± 0.02 0.18 ± 0.02 0.12 ± 0.13 0.19 ± 0.02CathodeTAN (mg N L−1) 28 ± 5 44 ± 8 109 ± 15 45 ± 9 296 ± 51 99 ± 15 600 ± 101 155 ± 19conductivity (mS cm−1) 12.1 ± 0.3 13.6 ± 3.2 30.1 ± 3.4 18.6 ± 4.5 25.6 ± 4.3 16.6 ± 5.0 18.6 ± 2.6 14.1 ± 1.3pH (−) 8.3 ± 0.2 7.9 ± 0.2 12.6 ± 0.2 8.1 ± 0.2 12.5 ± 0.3 7.8 ± 0.2 12.4 ± 0.2 7.9 ± 0.1ElectrochemicalN fluxc (g N m−2 d−1) 2 ± 0 3 ± 1 9 ± 1 3 ± 1 23 ± 3 7 ± 2 45 ± 8 13 ± 1NH4

+ current efficiency (CE,%) NA NA 9 ± 2 NA 18 ± 2 NA 36 ± 6 NANH4

+ removal efficiency (RE,%) 5 ± 1 8 ± 1 29 ± 5 11 ± 3 27 ± 4 9 ± 2 27 ± 3 8 ± 1cell voltage (V) NA NA 1.69 ± 0.13 NA 3.40 ± 0.25 NA 3.25 ± 0.10 NAenergy input (kWh kg−1 N) NA NA 39 ± 9 NA 36 ± 4 NA 18 ± 3 NAan = number of data points. ND: not determined. BDL: below detection limit. NA: not applicable. bCOD to CH4 conversion efficiency at STPconditions and relative to the amount of COD fed to the reactor. cRelative to the projected membrane surface area.

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

F

comparable to the results obtained in our previous study wherewe obtained a power input of 13 kWh kg−1 N at a currentdensity of 10 A m−2 and 2 g N L−1 with digestate from amunicipal solid waste digester.8 Simultaneously with the NH3,K+ is removed from the digestate, and upon extraction of thelatter, the energy input could be spread over those twoproducts. Ammonium is driven to a solids-free catholyte withhigh pH (Tables 3 and 4) and converted into volatile NH3. TheNH3 can then be recovered from this stream through strippingand absorption technology8 but was not subject of this study.Extra Asset: Electrochemical Remediation of H2S. An

interesting side observation during this study was thesignificantly lower H2S content in the biogas of the test reactorwhen the ES was switched on (Figure 3D, p < 0.05). Underthese conditions, the H2S concentration in the biogas of the testreactor was often below the detection limit (0.01%, Tables 3and 4), except for the period after the crash and cleaning of thereactors (Phase IIIa) when the ES was switched on. However, alot of variation was observed due to the different phases. Assuch, H2S was higher during the period after the crash andcleaning of the reactors (Phase IIIa) and at the beginning ofPhase IVg, when the pH shifted from 8 to 7. In contrast, theH2S concentration in the control setup was on average 0.20 ±0.06% over the entire experimental period. The overall lowerH2S content in the biogas of the test reactor when the ES wasswitched on was most likely caused by direct or indirectelectrochemical oxidation of the dissolved sulfide species in thevicinity of the anode electrode. Electrochemical sulfide removalfrom domestic wastewater was recently studied in detail at Ir/Ta-coated MMO-coated titanium anodes.18 The main mech-anism was indirect sulfide oxidation to elemental sulfur,thiosulphate, and sulfate by the in situ produced oxygen atthe anode. Most likely, this also took place in our setup, as asimilar anode electrode was used. Next to plain electrochemicaloxidation, also bioelectrochemical oxidation could take place asmicroorganisms were growing in the anode compartment.19 Ayellow deposition was observed on the anode, indicating theproduction of elemental sulfur. However, due to the complexmixed liquor, it was not possible to analyze this.The presence of H2S constitutes a severe problem to any

biogas conversion technology as it can cause corrosion.20

Hence, electrochemical H2S removal is a valuable asset next toNH3 toxicity control and nutrient recovery. In combinationwith a typically high reactor pH, the biogas has a low CO2content as well, leading to a highly methane enriched biogas.Microbial Community Findings. Microbial community

analysis was carried out to evaluate whether a differentiatingimpact could be observed due to the presence of the ES andthrough time. Real-time PCR analysis of the total bacteria andsummation of the different methanogenic groups resulted in anoverall coverage of the microbial community by themethanogens between 1% and 5% only (Figure S1, SupportingInformation). This is in contrast to their crucial role in theanaerobic digestion process and anticipated activity, yet is inagreement with other lab- and full-scale AD installations inwhich the methanogenic community covered no more than 5%of the total microbial community.21−23 Analysis of the maindifferent methanogenic populations in AD demonstrated thepresence of Methanobacteriales, Methanomicrobiales, and Meth-anosaetaceae for at least 10% of the methanogenic communityin each sample, irrespective of the treatment or time point(Figure 4). The assumed dominance of Methanosaetaceae overMethanosarcinaceae (below detection limit in every sample) as

the main acetoclastic methanogens relates to their filamentousstructure and therefore crucial role in anaerobic granuleformation.24 The persistence of the Methanobacteriales andMethanomicrobiales order points out that not only acetoclasticbut also hydrogenotrophic methanogenesis, whether or not incombination with syntrophic acetate oxidation, could takeplace. The overall limited dynamics of the methanogeniccommunity composition and abundance indicates that themethanogenic community was not selectively influenced by theoperational conditions in the rapid succession of the differentphases.

■ ASSOCIATED CONTENT*S Supporting InformationInformation as mentioned in the text. This material is availablefree of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Phone: +32 (0)9 264 59 76. Fax: +32 (0)9 264 62 48. E-mail:Korneel.Rabaey@ugent.be. Web page: www.labmet.Ugent.be.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSJ.D. is supported by an Advanced Grant of the IndustrialResearch Fund at Ghent University (F2012/IOF-Advanced/094). K.R. is supported by a European Research Council StarterGrant ELECTROTALK. We acknowledge Tim Lacoere for thegraphical abstract and Frederiek-Maarten Kerckhof fordeveloping the R-script and statistical analysis.

■ REFERENCES(1) Mata-Alvarez, J.; Mace, S.; Llabres, P. Anaerobic digestion oforganic solid wastes. An overview of research achievements andperspectives. Bioresour. Technol. 2000, 74 (1), 3−16.(2) Verstraete, W.; Morgan-Sagastume, F.; Aiyuk, S.; Waweru, M.;Rabaey, K.; Lissens, G. Anaerobic digestion as a core technology insustainable management of organic matter. Water Sci. Technol. 2005,52 (1−2), 59−66.(3) Tambone, F.; Scaglia, B.; D’Imporzano, G.; Schievano, A.; Orzi,V.; Salati, S.; Adani, F. Assessing amendment and fertilizing propertiesof digestates from anaerobic digestion through a comparative studywith digested sludge and compost. Chemosphere 2010, 81 (5), 577−583.(4) Angelidaki, I.; Ahring, B. K. Thermophilic anaerobic digestion oflivestock waste − The effect of ammonia. Appl. Microbiol. Biotechnol.1993, 38 (4), 560−564.(5) Rajagopal, R.; Masse, D. I.; Singh, G. A critical review oninhibition of anaerobic digestion process by excess ammonia. Bioresour.Technol. 2013, 143, 632−641.(6) Chen, Y.; Cheng, J. J.; Creamer, K. S. Inhibition of anaerobicdigestion process: A review. Bioresour. Technol. 2008, 99 (10), 4044−4064.(7) Andersen, S. J.; Hennebel, T.; Gildemyn, S.; Coma, M.;Desloover, J.; Berton, J.; Tsukamoto, J.; Stevens, C.; Rabaey, K.Electrolytic membrane extraction enables production of fine chemicalsfrom biorefinery sidestreams. Environ. Sci. Technol. 2014, 48 (12),7135−7142.(8) Desloover, J.; Woldeyohannis, A. A.; Verstraete, W.; Boon, N.;Rabaey, K. Electrochemical resource recovery from digestate toprevent ammonia toxicity during anaerobic digestion. Environ. Sci.Technol. 2012, 46 (21), 12209−12216.(9) Bagastyo, A. Y.; Radjenovic, J.; Mu, Y.; Rozendal, R. A.; Batstone,D. J.; Rabaey, K. Electrochemical oxidation of reverse osmosis

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

G

concentrate on mixed metal oxide (MMO) titanium coated electrodes.Water Res. 2011, 45 (16), 4951−4959.(10) Anthonisen, A. C.; Loehr, R. C.; Prakasam, T. B. S.; Srinath, E.G. Inhibition of nitrification by ammonia and nitrous acid. J. WaterPollut. Control Fed. 1976, 48 (5), 835−852.(11) Greenberg, A. E. Standard Methods for the Examination of Waterand Wastewater; American Public Health Association: WashingtonDC, 1992.(12) Vilchez-Vargas, R.; Geffers, R.; Suarez-Diez, M.; Conte, I.;Waliczek, A.; Kaser, V. S.; Kralova, M.; Junca, H.; Pieper, D. H.Analysis of the microbial gene landscape and transcriptome foraromatic pollutants and alkane degradation using a novel internallycalibrated microarray system. Environ. Microbiol. 2013, 15 (4), 1016−1039.(13) Ovreas, L.; Forney, L.; Daae, F. L.; Torsvik, V. Distribution ofbacterioplankton in meromictic Lake Saelenvannet, as determined bydenaturing gradient gel electrophoresis of PCR-amplified genefragments coding for 16S rRNA. Appl. Environ. Microbiol. 1997, 63(9), 3367−3373.(14) Yu, Y.; Lee, C.; Kim, J.; Hwang, S. Group-specific primer andprobe sets to detect methanogenic communities using quantitativereal-time polymerase chain reaction. Biotechnol. Bioeng. 2005, 89 (6),670−679.(15) De Vrieze, J.; De Lathouwer, L.; Verstraete, W.; Boon, N. High-rate iron-rich activated sludge as stabilizing agent for the anaerobicdigestion of kitchen waste. Water Res. 2013, 47 (11), 3732−3741.(16) Kuntke, P.; Sleutels, T. H. J. A.; Saakes, M.; Buisman, C. J. N.Hydrogen production and ammonium recovery from urine by amicrobial electrolysis cell. Int. J. Hydrogen Energy 2014, 39 (10), 4771−4778.(17) Gajda, I.; Greenman, J.; Melhuish, C.; Santoro, C.; Li, B.;Cristiani, P.; Ieropoulos, I. Water formation at the cathode and sodiumrecovery using microbial fuel cells (MFCs). Sustainable Energy Technol.Assess. 2014, 7 (0), 187−194.(18) Pikaar, I.; Rozendal, R. A.; Yuan, Z. G.; Keller, J.; Rabaey, K.Electrochemical sulfide oxidation from domestic wastewater usingmixed metal-coated titanium electrodes. Water Res. 2011, 45 (17),5381−5388.(19) Dutta, P. K.; Keller, J.; Yuan, Z. G.; Rozendal, R. A.; Rabaey, K.Role of sulfur duringacetate oxidation in biological anodes. Environ. Sci.Technol. 2009, 43 (10), 3839−3845.(20) Abatzoglou, N.; Boivin, S. A review of biogas purificationprocesses. Biofuel. Bioprod. Bior. 2009, 3 (1), 42−71.(21) Sundberg, C.; Al-Soud, W. A.; Larsson, M.; Alm, E.; Yekta, S. S.;Svensson, B. H.; Sørensen, S. J.; Karlsson, A. 454 pyrosequencinganalyses of bacterial and archaeal richness in 21 full-scale biogasdigesters. FEMS Microbiol. Ecol. 2013, 85 (3), 612−626.(22) Vanwonterghem, I.; Jensen, P. D.; Ho, D. P.; Batstone, D. J.;Tyson, G. W. Linking microbial community structure, interactions andfunction in anaerobic digesters using new molecular techniques. Curr.Opin. Biotechnol. 2014, 27 (0), 55−64.(23) Zakrzewski, M.; Goesmann, A.; Jaenicke, S.; Junemann, S.;Eikmeyer, F.; Szczepanowski, R.; Abu Al-Soud, W.; Sorensen, S.;Puhler, A.; Schluter, A. Profiling of the metabolically active communityfrom a production-scale biogas plant by means of high-throughputmetatranscriptome sequencing. J. Biotechnol. 2012, 158 (4), 248−258.(24) Hulshoff Pol, L. W.; Lopes, S. I. D.; Lettinga, G.; Lens, P. N. L.Anaerobic sludge granulation. Water Res. 2004, 38 (6), 1376−1389.

Environmental Science & Technology Article

DOI: 10.1021/es504811aEnviron. Sci. Technol. XXXX, XXX, XXX−XXX

H