Content uploaded by Korneel Rabaey
Author content
All content in this area was uploaded by Korneel Rabaey on Aug 14, 2014
Content may be subject to copyright.
Continuous microbial fuel cells convert carbohydrates to electricity
K. Rabaey1, W. Ossieur1, M. Verhaege² and W. Verstraete1
1 Laboratory for Microbial Ecology and Technology (LabMET), Ghent University, Coupure Links 653, Ghent,
Belgium
2 Laboratory for Non-ferrous Metallurgy, Ghent University,
Abstract Microbial fuel cells which are operated in continuous mode are more suitable for practical
applications than fed batch ones. Three influent types containing carbohydrates were tested, i.e. a glucose
medium, a plant extract and artificial wastewater. The anode reactor compartment yielding the best results
was a packed bed reactor containing graphite granules. While in non-mediated batch systems power
outputs up to 479 W m-3 of anode compartment could be attained; in continuous mode the power outputs
were limited to 49 W m-3. Cyclic voltammetry was performed to determine the potential of the in-situ
synthesized bacterial redox mediators. Addition of mediators with a potential similar to the bacterial potential
did not significantly alter the MFC power output, indicating a limited influence of soluble mediators for
continuous microbial fuel cells. Maximum coulombic and energy conversion efficiencies were, for the
continuous microbial fuel cell operating on plant extract at a loading rate of 1 kg COD m-3 of anode
compartment per day, 50.3% and 26.0% respectively.
Keywords biofuel cell, glucose, sucrose, flow through, wastewater, plant sap
Introduction
Bacteria gain metabolic energy by transferring electrons from an electron donor, such as glucose, to
an electron acceptor, such as oxygen. The larger the potential difference between donor and
acceptor, the larger the gain for the bacterium. In a microbial fuel cell (MFC), bacteria do not
directly transfer their produced electrons to their characteristic terminal electron acceptor but these
electrons are diverted towards an electrode (anode). The electrons are subsequently conducted over
a resistance or power user towards a cathode and thus, bacterial energy is directly converted to
electrical energy (Rao et al., 1976). To close the cycle, protons migrate through a proton exchange
membrane from anode to cathode (Figure 1).
Three main types are commonly
distinguished: photo-autotrophic type
biofuel cells (Tsujimura et al., 2001),
heterotrophic type biofuel cells, (Cooney et
al., 1996) and sediment biofuel cells (Bond
et al., 2002). Biofuel cells have
characteristics similar to traditional power
sources, as well as to anaerobic reactors.
They can on the one hand be described by
electrochemical parameters such as power
density (W m-2 electrode surface or per m3
anode compartment), electrical current
output and cell voltage and on the other
hand by biological parameters such as the
substrate loading rate (kg m-3 d-1) (Rabaey
et al., 2003).
Glucose
ANODE CAT HODE
PROTON EXCHANGE
MEMBRANE
H+
O2
H+
O2
H2O
H2O
CO2
Figure 1 Working principle of a microbial fuel cell
Direct electron transfer from bacteria to an electrode is hampered by overpotentials, which can
be described as transfer resistances (Bard and Faulkner, 2001). In order to reduce these resistances,
the specific surface of the electrodes needs to be increased, and/or redox mediators need addition to
the solution. A redox mediator is a compound that can be reversibly oxidized or reduced. Bacteria
can use redox mediators to deposit their electrons onto a substrate they cannot directly reduce
(Hernandez et al., 2004). However, the addition of these mediators to the solution can represent a
considerable cost, and their presence is usually not desirable for the effluent due to either colour or
environmental effects.
During a long term batch enrichment process for anodophylic consortia, the power output of a
microbial fuel cell increased to 4.31 W m-² anodic surface, corresponding to 479 W m-3 reactor
(Rabaey et al., 2004). Analysis of the bacterial community structure and identification of the
bacteria suggested the evolution of the microbial community towards species that self-mediate this
electron transfer. Analysis using cyclic voltammetry furthermore demonstrated that this self-
mediating effect is dependent on two processes: (i) the expression of membrane-bound electron
transfer components and (ii) the production of soluble redox mediators. An example of such a
mediator is pyocyanin (Mavrodi et al., 2001), produced by an isolated Pseudomonas aeruginosa
species from the fuel cell. Using the two processes in parallel, bacterial consortia could transfer
electrons towards an electrode with a maximum electron transfer efficiency up to 89%, calculated
as coulombs gained as electricity per coulombs added as carbon source. This corresponds to an
energetic efficiency up to 79%, calculated as joules gained as electricity per joules added as carbon
source.
However, batch type microbial fuel cells are not suitable for applications such as wastewater
treatment, since mediators and bacteria will flow out of the system. Therefore, the anode
compartment needs to use an efficient biofilm on a large surface anode. Furthermore, the
characteristics of the electrode need to be adapted to the biofilm and the used application.
In this study, continuous flow microbial fuel cells were designed operating on several types of
influents. After optimization of reactor parameters, the electrochemical characteristics of the
bacteria were determined in order to define suitable mediators.
Materials and methods
Reactor setup
The microbial fuel cells were made of plexi glass elements that were bolted together as described
previously (Rabaey et al., 2003). The inner volume of the anode compartment was 0.250 l, the
liquid volume varied between 0.080 and 0.220 l depending on the reactor configuration. Three
anode types were used: (i) plain graphite anodes (8 x 4 x 0.6 cm) (Morgan, Belgium); (ii) granular
anodes (3 mm average diameter) (Le Carbon, Belgium) without baffles (0.080 l of liquid volume);
(iii) granular anodes, identical to previous set-up but with baffles (Figure 2) (0.080 l of liquid
volume).
ANODE
Figure 2 Overview of the used set-up for continuous microbial fuel cells
The 250 ml cathode compartment contained a 100 mM phosphate buffered 50 mM potassium
hexacyanoferrate solution (VWR, Belgium). Several types of cathode electrodes were used: (i)
plain graphite cathodes (8 x 4 x 0.6 cm) (Morgan, Belgium); (ii) graphite felt electrodes (8 x 4 x 0.4
cm) (Alfa Aesar, Germany); (iii) granular cathodes (identical to anode granules), without baffles.
The influent for the anode was provided at a loading rate of 1 g l-1 d-1, corresponding with a COD
concentration between 0.111 g l-1 (0.080 ml anode liquid volume) and 0.305 g l-1 (0.220 ml anode
liquid volume) for an influent volume of 0.720 l feed per day, corresponding to a hydraulic
residence time between 2.7 and 7.3 h. A recirculation of 8 l h-1 was foreseen to provide sufficient
anode turbidity. Measurements of the power output were performed using an Agilent HP 34970
data acquisition unit. Every 30 seconds, a full channel scan was performed and the data stored.
External system resistance R was maximum 100 Ω, current was deducible as I = V x R-1 = Q x t-1
(Eq.1) with I the current (A), V the voltage (V), Q the charge (C) and t time (s). Power output of the
cells was calculated as P(W) = I x V (Eq. 2). Energy production can then be expressed as E(J) = P x
t (Eq. 3).
Two different inocula were initially tested: a sediment sample (Scheldt estuary, Belgium) grown in
anaerobic nutrient broth prior to addition to the reactor, and a culture obtained from a microbial fuel
cell on acetate, previously operated at LabMET.
Influent preparation
Three influent types were prepared. The basic medium was M9 with a composition per litre: 6 g
Na2HPO4; 1 g NH4Cl; 0.5 g NaCl; 0.2465 g MgSO4.7H2O; 3 g KH2PO4; 14.7 g CaCl2. To this
medium, a carbon source (glucose/maple syrup) was added to attain a loading rate of 1 g COD l-1
anode liquid volume per day. The artificial wastewater was prepared according to Jang et al. (Jang
et al., 2003), and contained per litre 0.56 g (NH4)2SO4; 0.2 g MgSO4.7H2O; 0.42 g NaHCO3; 1 mg
FeCl3.6H2O; 20 mg MnSO4.H2O; 14.7 CaCl2; 6.8 g KH2PO4. This so-called artificial wastewater
was supplied with glucose to obtain a reactor loading rate of 1 g COD l-1 anode liquid volume per
day. To all influents, 1 ml of a trace element solution was added, containing per litre: FeSO4.7H2O
1g; ZnCl2 70 mg; MnCl2.4H2O 100 mg; H3BO3 6 mg; CaCl2.6H2O 130 mg; CuCl2.2H2O 2 mg;
NiCl2.6H2O 24 mg; Na2Mo4.2H2O 36 mg; CoCl2.6H2O 238 mg.
Cyclic voltammetry
For the voltammetry analysis, 15 ml samples were taken from microbial fuel cells prior to the daily
feeding, inserted in a test vial and flushed with nitrogen gas prior to the measurement. The cyclic
voltammetry was performed as described previously (Park et al., 2001). A potentiostat (Princeton
Applied Research, USA, model 263a), branched to a PC was used (Princeton Applied Research,
USA, SoftCorr III), at a scan rate of 50 mV s-1 in the potential range of -450 to 900 mV. The
working electrode was a 5 cm² graphite rod, cleaned in ethanol and deionised water prior to use, the
counter electrode was a platinum wire and an Ag/AgCl electrode (BAS, USA, MF-2052) was used
as reference. All three were inserted in the test vial avoiding any contact between the electrodes.
The mediators chosen for addition to the MFC were Resorufin (Fluka), Indigo carmine (Sigma),
Safranine O (Sigma) and New Methylene Blue (UCB).
Analysis
Samples were filtered through a syringe 0.22µm filter unit (Millex, USA). For analysis of the
volatile fatty acids (VFA), an extraction in diethyl ether was performed (Greenberg et al., 1992).
The samples were analysed with a capillary FID (flame ionization detector) gas chromatograph, GC
8000 Carlo Erba Instruments (Wigan, UK), connected to a computer. The column used was an
Alltech (Deerfield, USA) EC-1000 (30 m, I.D.: 0.32 mm, df: 0.25 µm). The temperature was
controlled at 135 °C for the isotherm oven and 200 °C for the detector and the injector. Nitrogen
gas was used as the carrier gas at 3 ml min-1. Samples were diluted 10 times and glucose assessed
using ion chromatography (Dionex, Carbopac1 column with borate and amino trap, PAD ED40)
(Groussac et al., 2000). Gas chromatography (Intersmat IGC 120 MB) was used for determination
of CO2 and CH4 in the headspace (Tsujimura et al., 2001). H2 was measured using a Microtox
exhaled hydrogen monitor (GMI, Germany), detection limit 5ppmv (Greenberg et al., 1992). For
the gas samples, 5 ml of anode headspace was obtained, after which the biofuel cell was flushed
with nitrogen gas. pH electrodes (Metrohm, Switzerland) were installed to monitor the
compartment pH. COD analysis was performed using the dichromate method (Greenberg et al.,
1992).
Results
All loading rates and power output results are expressed as W m-3 of anode liquid volume to enable
comparison between reactors that have a different content of electrode material per unit of reactor
volume.
Influence of the initial inoculum
The reactors inoculated with the bacterial culture obtained from the sediment, demonstrated an
average voltage of 184 ± 71 mV versus 133 ± 21 mV obtained with the former biofuel cell culture.
Despite this non-significant difference between both inocula, all further tests were performed using
the sediment inoculum.
Microbial fuel cells on glucose
A stable power output could be obtained within a short period of time for the reactors fed with
glucose. Initially the power output was, on average, 5.7 W m-3. After an interruption of the feeding
for one day due to a technical problem, the power output increased gradually during a three week
period to a daily average of 37 W m-3. As verified by ion chromatography, glucose was always
completely used. No accumulation of volatile fatty acids (VFA) was observed, the effluent
concentrations gradually decreased to on average 52 ± 45 mg VFA l-1. Acetate was generally the
only VFA present in significant quantities. The corresponding current generated represented up to
49.0 ± 0.3 % of the added COD (Table 1). This represented an energetic efficiency of 20 % (Table
2). Declines were noted in the power output, which could be related to a visual deterioration of the
influent. Indeed, when the latter became infected by acidifiers, the power output decreased.
However, when influent was replaced a rapid (generally less than one hour time) return to the
original voltage was noted.
Table 1 Overview of COD conversion towards VFA and current in continuous MFCs operated at a
loading rate of 1 g COD l-1 anode compartment per day. The calculations were made on the basis
of average daily current and VFA output at the beginning and the end of the experimental period
Reactor
feeding
CODin CODout as current CODout as VFA CODout
residual
CODout total Recovery as
current
Balance
recovery
(kg m-3d-1) (kg m-3d-1) (kg m-3d-1) (kg m-3d-1) (kg m-3d-1) (%) (%)
Glucose Initial 1.000 0.165 ± 0.036 0.863 ± 0.578 n.d. 1.028 ± 0.579* 16.5 ± 3.6 103 ± 58
End 1.000 0.490 ± 0.003 0.096 ± 0.057 0.449 ± 0.068 0.939 ± 0.068♦ 49.0 ± 0.3 94 ± 7
Maple syrup Initial 1.000 0.164 ± 0.014 0.684 ± 0.104 n.d. 0.848 ± 0.105* 16.4 ± 1.4 85 ± 11
End 1.000 0.541 ± 0.013 0.233 ± 0.117 0.306 ± 0.086 0.847 ± 0.086♦ 54.1 ± 1.3 85 ± 9
Wastewater Initial 1.000 0.171 ± 0.013 0.827 ± 0.302 n.d. 0.998 ± 0.302* 17.1 ± 1.3 100 ± 30
End 1.000 0.391 ± 0.009 0.058 ± 0.029 0.679 ± 0.090 1.070 ± 0.090♦ 39.1 ± 0.9 107 ± 9
COD calculated on the basis of current and VFAout-COD (*) or total residual CODout (♦)
n.d. not determined
Microbial fuel cells on maple syrup
In parallel with the microbial fuel cells on glucose, other reactors were fed with M9 medium
supplemented with maple syrup as carbon source. Initial power output was of the order of 4 W m-3,
corresponding with an energy production of 0.09 kWh m-3 d-1. This output gradually increased to
49 W m-3, corresponding with an energy production of 1.2 kWh m-3 d-1. This implied a maximum
cell potential of 625 mV and a current of 6.3 mA (Table 1 and 2). No accumulation of volatile fatty
acids was observed, and concentrations decreased during the test period to 17 ± 11 mg VFA l-1. As
in the glucose fed reactors, mainly acetate was present in the effluent
Microbial fuel cells on artificial wastewater
A third series of microbial fuel cells was operated with artificial wastewater (Jang et al., 2003),
with glucose as carbon source. No large differences exist between this medium and the glucose
medium used as described previously. However, the power output of these MFCs was lower than
the power obtained for the other two substrates, namely 24.7 W m-3 after one month of operation
(Table 1 and 2). As in the previous cases, no accumulation of VFA was observed, since effluent
concentrations gradually decreased to about 16 ± 9 mg l-1. The VFA was mainly composed of
acetate. There is clearly an influence of the basic medium onto the reactor performance.
Influence of the electrode structure
To investigate the influence of the cathode structure onto the biofuel cell performance, two reactors
were installed with either a graphite felt or plain graphite cathode. No significant differences in
power output were noted between the graphite felt and the plain graphite cathode. During the first
50 hours of operation, the power output was, on average, 8.8 ± 0.4 W m-3 and 8.0 ± 0.6 W m-3 for
the graphite felt and the plain graphite respectively. Maximum power output of the reactors was
15.9 W m-3 and 15.2 W m-3 for the graphite felt and the plain graphite respectively.
When graphite granules were used for the cathode, power output was initially of the same level
of the other cathodes tested, but the output rapidly decreased. A decolourization of the catholyte
was observed, but this did not imply decreasing iron concentrations in solution. Presumably, the
granular matrix did not allow sufficient proton transport to obtain sufficient water formation.
Several different types of anode were used during the experimental period. Changing the anode
from plate shaped to granular caused a two fold increase of the MFC voltage. The further addition
of baffles to the anode compartment, in order to force a flow through the granular bed, allowed a
further increase of the voltage to the values indicated previously.
Table 2 Energy output of continuous MFCs on diverse substrate, without addition of redox
mediators. The end period voltage indicated is the highest daily averaged voltage obtained.
Reactor feeding Energyin Biofuel cell voltage Energyout Energy recovery
(kJ m-3 d-1) (mV) (kJ m-3 d-1) (%)
Glucose Initial period 15875 213 490 3.1
End period 15875 545 3208 20.2
Maple syrup Initial period 16200 172 320 2.0
End period 16200 625 4220 26.0
Wastewater Initial period 15875 201 436 2.7
End period 15875 445 2139 13.5
Cyclic voltammetry for mediator selection
Cyclic voltammetry was performed onto the bacterial culture derived from the glucose fed
microbial fuel cells. Oxidation and reduction peaks were found, indicating electrochemical activity
at a potential of approximately -50 mV versus standard hydrogen electrode. Redox mediators,
having a standard redox potential near the measured activity potential of the bacteria, were applied
at a concentration of 50 µM. Upon addition of the mediator to the anode, a fast decolourisation was
noted, indicating reduction of the mediator. No significant effects of the mediators onto the MFC
power output were noted in any of the cases. These results differ from the previous findings
obtained for batch systems (Choi et al., 2003; Park and Zeikus, 1999; Roller et al., 1984).
Discussion
Removal of COD by microbial fuel cells
Up to 50 % of the COD present in the influent was removed as electricity (Table 1). The remainder
of the COD was either present in the effluent as acetate, or not detected. The COD which is lacking
is most likely hydrogen gas, which was not measured using these continuous reactors. The fact that
the acetate present in the effluent is reasonably low, demonstrates the capability of the microbial
fuel cell to biodegrade substrates such as glucose further down than simple fermentation to acetate
and other VFA. This opens possibilities for microbial fuel cells to be used for COD removal.
A discrepancy exists between the coulombic efficiencies obtained (Table 1) and the energetic
efficiencies obtained (Table 2). Generally, for wastewater treatment, COD removal is the prime
parameter of importance. This COD removal can be expressed as coulombic efficiency. However,
the added value of the produced power can be as important in terms of economic feasibility of the
WWTP. To obtain a high energetic efficiency, both voltage and current need to be of sufficient
magnitude. This implies that the resistance over the microbial fuel cell cannot be too high, neither
too low. Thus far, almost all studies expressed microbial fuel cell output by mA m-² or W m-² of
anode. In order to compare this technology with other existing technologies, the calculations should
be remade based on W m-3 of anode liquid volume. This calculation should clarify whether the
loading rates used and the power outputs observed have significance for practice. Furthermore, this
would enable a comparison of the different studies performed.
Studies on microbial fuel cells have thus far always applied large resistances of over 500 Ω,
with the exception of two publications by our group (Rabaey et al., 2003, 2004). This generated
large cell potentials, but low currents. Hence, total power output of the microbial fuel cells was
low. Further research is needed to determine the optimal resistance over the MFC generating the
largest power. This optimal resistance will more than likely be depending on the application, the
inoculum, the type of reactor, the substrate and the potential losses within the microbial fuel cell.
Applicability of the used substrates
The substrates used were chosen as model components towards practical applications. The data
obtained indicates energy recoveries up to 26% at the moderate loading rate of 1 kg m-3 d
-1.
Microbial fuel cells for wastewater treatment, based on this study, would carry several advantages:
(i) COD is removed (ii) useful power can be obtained (iii) no aeration is needed (iv) no complicated
equipment for processing of biogases is needed. These advantages compete with the disadvantage
of a larger process complexity of the reactor in comparison to traditional wastewater treatment.
Importance of redox mediators
While in batch systems, the improvement of the power output through the addition of redox
mediators was repeatedly reported (Park and Zeikus, 1999; Roller et al., 1984), for the continuous
microbial fuel cells no significant effect was observed. Clearly, the addition of soluble mediators to
the anode is not a desirable strategy for continuous MFCs. Hence, the possibility of immobilizing
the mediator onto either active biomass or onto/into electrode material should be further elaborated.
Acknowledgments
The authors wish to thank Joris De Backer for his contribution to operating the microbial fuel cells,
and Nico Boon and Geert Lissens for their useful comments.
References
Bard, A.J. and Faulkner, L.R. (2001). Electrochemical methods: fundamentals and applications. 2nd ed, John Wiley &
Sons, New York.
Bond, D.R., Holmes, D.E., Tender, L.M. and Lovley, D.R. (2002). Electrode-reducing microorganisms that harvest
energy from marine sediments. Science, 295(5554), 483-485.
Choi, Y., Kim, N., Kim, S. and Jung, S. (2003). Dynamic behaviors of redox mediators within the hydrophobic layers
as an important factor for effective microbial fuel cell operation. B Kor Chem Soc, 24(4), 437-440.
Cooney, M.J., Roschi, E., Marison, I.W., Comninellis, C. and vonStockar, U. (1996). Physiologic studies with the
sulfate-reducing bacterium Desulfovibrio desulfuricans: Evaluation for use in a biofuel cell. Enzyme Microb.
Technol., 18(5), 358-365.
Greenberg, A., Clesceri, L.S. and Eaton, A.D. (1992). Standard Methods for the Examination of Water and
Wastewater.18th ed, American public health association, Washington.
Groussac, E., Ortiz, M. and Francois, J. (2000). Improved protocols for quantitative determination of metabolites from
biological samples using high performance ionic-exchange chromatography with conductimetric and pulsed
amperometric detection. Enzyme Microb. Technol., 26(9-10), 715-723.
Hernandez, M.E., Kappler, A. and Newman, D.K. (2004). Phenazines and other redox-active antibiotics promote
microbial mineral reduction. Appl. Environ. Microbiol., 70(2), 921-928.
Jang, J.K., Pham, T.H., Chang, I.S., et al (2003). Construction and operation of a novel mediator- and membrane-less
microbial fuel cell. Process Biochemistry, In Press.
Mavrodi, D.V., Bonsall, R.F., Delaney, S.M., Soule, M.J., Phillips, G. and Thomashow, L.S. (2001). Functional
analysis of genes for biosynthesis of pyocyanin and phenazine-1-carboxamide from Pseudomonas aeruginosa
PAO1. J. Bacteriol., 183(21), 6454-6465.
Park, D.H. and Zeikus, J.G. (1999). Utilization of electrically reduced neutral red by Actinobacillus succinogenes:
Physiological function of neutral red in membrane-driven fumarate reduction and energy conservation. J.
Bacteriol., 181(8), 2403-2410.
Park, H.S., Kim, B.H., Kim, H.S., et al (2001). A novel electrochemically active and Fe(III)-reducing bacterium
phylogenetically related to Clostridium butyricum isolated from a microbial fuel cell. Anaerobe, 7(6), 297-
306.
Rabaey, K., Boon, N., Siciliano, S.D., Verhaege, M. and Verstraete, W. (2004). Biofuel cells select for microbial
consortia that self-mediate electron transfer. Appl. Environ. Microbiol., In press.
Rabaey, K., Lissens, G., Siciliano, S.D. and Verstraete, W. (2003). A microbial fuel cell capable of converting glucose
to electricity at high rate and efficiency. Biotechnol. Lett., 25(18), 1531-1535.
Rao, J.R., Richter, G.J., Vonsturm, F. and Weidlich, E. (1976). Performance of glucose electrodes and characteristics
of different biofuel cell constructions. Bioelectrochem. Bioenerg., 3(1), 139-150.
Roller, S.D., Bennetto, H.P., Delaney, G.M., Mason, J.R., Stirling, J.L. and Thurston, C.F. (1984). Electron-transfer
coupling in microbial fuel-cells .1. Comparison of redox-mediator reduction rates and respiratory rates of
bacteria. J. Chem. Tech. Biotech. B- Biotechnology, 34(1), 3-12.
Tsujimura, S., Wadano, A., Kano, K. and Ikeda, T. (2001). Photosynthetic bioelectrochemical cell utilizing
cyanobacteria and water-generating oxidase. Enzyme Microb. Technol., 29(4-5), 225-231.
