- Email: [email protected]
Dynamic polarisation reveals differential steady-state stabilisation and capacitive-like behaviour in microbial fuel cells
Sustainable Energy Technologies and Assessments 5 (2014) 1–6
Contents lists available at ScienceDirect
Sustainable Energy Technologies and Assessments journal homepage: www.elsevier.com/locate/seta
Original Research Article
Dynamic polarisation reveals differential steady-state stabilisation and capacitive-like behaviour in microbial fuel cells Pablo Ledezma a,⇑, Nicolas Degrenne b, Pascal Bevilacqua c, François Buret b, Bruno Allard c, John Greenman d, Ioannis Ieropoulos a a
Bristol Robotics Laboratory, Universities of Bristol and of the West of England, Frenchay Campus, Bristol BS34 8QZ, UK Laboratoire Ampère, Ecole centrale de Lyon, Université de Lyon, Ecully 69134, France Laboratoire Ampère, INSA de Lyon, Université de Lyon, Villeurbanne 69100, France d Department of Applied Sciences, University of the West of England, Bristol BS16 1QY, UK b c
a r t i c l e
i n f o
Article history: Received 28 July 2013 Revised 14 October 2013 Accepted 25 October 2013
Keywords: Microbial fuel cells External resistors Polarisation curve Time constant Steady-states
a b s t r a c t In this paper we present several preliminary results produced with a purposely-designed external-resistor (Rext) sweeping tool for microbial fuel cells (MFCs). Fast sampling rates show that MFCs exhibit differential steady-state stabilisation behaviours depending on Rext, with consequences for time constant (tc) selection. At high Rext (35 kO), it is demonstrated that a tc P 10 min avoids underestimation not overestimation, whilst at low Rext (100 O) 5 min are sufficient, suggesting that sweeps with variable tc are possible. However, within the maximum power transfer range (2.5 kO), steady-states are only observed at 20 min tc but with a smaller confidence interval, questioning whether the polarisation technique is suitable to estimate maximum power transfer. Finally, a strategy towards the exploitation of a capacitive-like behaviour in MFCs is proposed, tapping into P10 min periods with up to 50% higher current and energy transfer that could prove important for MFC-powered applications. Ó 2013 Elsevier Ltd. All rights reserved.
Introduction Microbial fuel cells (MFCs) are an exciting technology with a growing scientific interest and research dynamism due to the potential of sustainable electricity production from organic waste and, notably, wastewater. The latter is either released untreated into the environment – with severe environmental consequences – or treated through expensive and energy intensive processes. MFCs offer the possibility of transforming such wastewater treatment plants into energy- and carbon-neutral systems or even the longer-term potential of net energy production, similar to novel anaerobic digestion plants but operating at higher efficiencies and requiring less investment [1]. One of the most commonly used tools for characterising the behaviour of MFC systems is known as the polarisation technique. It is a process whereby the fuel cell is subjected to a set number of external loads (Rext of known resistance in O) over a constant period of time (time constant, tc), resulting in the production of an electric current (in A) at a corresponding cell voltage (in V). The graphical representation of the obtained results or Polarisation curve (voltage vs current) provides considerable information about the tested systems, particularly about the associated activation, ⇑ Corresponding author. Tel.: +44 (0) 117 32 86350; fax: +44 (0) 117 32 83960. E-mail address: [email protected] (P. Ledezma). 2213-1388/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.seta.2013.10.008
ohmic and mass transfer losses [2]. Moreover, from the derivation of Joule’s law, the electrical power (in W) can be calculated at the different exhibited current levels and plotted as a Power curve (power vs current), which again reveals information about the behaviour and stability of the system predominantly in comparison with a mathematical model. According to Jacobi’s law, an ideal fuel cell produces a parabolic Power curve whereby the maximum power transfer (MPT) point occurs at the mid points of current and voltage [3]. This protocol has become almost ubiquitous for determining the maximum attainable power from MFCs. A debate is currently ongoing with regards to this technique and in particular to the value (in O) and quantity of external resistors employed, as well as the time constant (tc) in terms of accurately determining the real, sustainable MPT by MFCs. This investigation addresses these concerns, presenting a more accurate polarisation technique that aims to question the use of constant or long time intervals for the evaluation of MFC performance and whether the technique is suitable for the estimation of MPT. Materials and methods MFC assembly and operation Eight 2-chamber MFCs (20 mL/side) made of four different materials [4] and of properties previously described [5] were
2
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
inoculated with activated sewage sludge and then operated in continuous mode, under identical conditions, for at least 18 months. During this period, the MFCs were fed with acetate TYE (5 mM C2H3NaO2, 0.1% w/v Tryptone, 0.05% w/v Yeast Extract) at a low flow rate of 120 lL min 1, with tap water (D.O. avg. 6.5 mg L 1; Conductivity avg. 680 lS cm 1) as the catholyte at a flow rate of 19.35 mL min 1. Our research focuses on the utilisation of MFC systems as power sources for practical applications e.g. robots [6]. For such purposes, we have already demonstrated that a suitable strategy to achieve the necessary voltages and currents is to construct MFC collectives (stacks) with combinations of series/parallel connections [6] and that we are able to maintain the operation of such stacks at MPT point with high efficiency [7], both necessary elements for the stable and predictable operation of applications. More recently, we have also shown that a stacking architecture that balances the internal resistances of embedded cells avoids cell voltage reversal even under fuel starvation [8] and that this strategy can be used for scaled-up reactors with up to 20 MFCs in series [9]. We have utilised the latter approach to energise applications previously considered impossible with MFCs e.g. a multi-channel peristaltic pump [9] and a mobile phone [10]. However, a significant amount of research is still needed into stacking at the fundamental levels – 2 to 4 MFCs in different configurations – in order to elucidate techniques that can enhance their collective electrical output without reversal. In concordance with these objectives, the experiments hereby presented involve four stacks of two MFCs connected in series (namely stacks A, B, C & D). Prior to stacking, all cells were individually verified for output under continuous load and polarisation conditions and no significant differences were found between MFCs fabricated in the same material (data not shown). The electrical output of each stack (voltage) was monitored/controlled in real-time with the tool described below.
A
B
Dynamic polarisation method with novel Resistorstat The constructed stacks were connected to an 8-channel automated Resistorstat tool, whose development and architecture is described in detail by Degrenne et al. [7]. This system was programmed to perform sweeps of external resistor (Rext) values, ranging decreasingly from 38.5 kX to 4 X (35 values with avg. 18.5% resistance reduction per step, so as to make the reduction rates constant), at different time constant (tc) intervals (1, 5, 10 and 20 min). The cells were kept in open-circuit mode for 1 h prior to testing. Their voltage was recorded every 30 s in order to monitor the dynamic response of MFCs to changes in Rext over the extent of each tc step. Calculations The resulting voltage output readings were automatically processed by the Resistorstat system, calculating current (I) and power (P) output per MFC stack, as previously described [7]. For the sake of clarity, exemplars of the most representative results are shown, which illustrate responses that were consistently observed in the more than 25 polarisation sweeps performed per stack.
Results and discussion Novel dynamic polarisation and power curves The data produced with the technique hereby presented result in a new form of polarisation (Fig. 1A) and power curves (Fig. 1B), distinguished by the presence of spikes. These seem to indicate a
Fig. 1. (A) Polarisation curve of a 2-MFC stack connected in series with a 30 s sampling time resolution and a 5 min time constant. The modelling lines (in gray) highlight the fact that the spikes converge, in a linear manner, in concordance with the external resistor employed. (B) Power curve of the same 2-MFC stack. The modelling lines (in gray) exhibit the parabolic convergence behaviour of the spikes. In both cases, blue lines indicate the typical polarisation and power curves that would be obtained from examined systems, showing the difference with dynamic curves.
plateauing tendency in the output or convergence for each Rext tested. Although the variation observed can be considered low, both curves (see Fig. 1A and B) illustrate that the aforementioned behaviour is different between points tested in terms of the orientation/ width of the spikes depending on the zone in the curves. Simple modelling revealed that the orientation of the spikes in the polarisation curve can be explained by a linear relationship depending on Rext, each spike aligning with the Cartesian equation V = Rext I of Ohm’s law (grey modelling lines, Fig. 1A). In a similar manner, the spikes’ orientation in the power curve (see Fig. 1B) can be explained by the square function P = Rext I2. The latter formula is commonly employed in electrical engineering to determine power losses in an electrical transmission, so the spikes could be considered as an indicator of the power that is lost (or in the power curves, initially overestimated) whilst the system stabilises into a new steady-state (stable electrical production over time). Since it was observed that these spikes appeared substantially different throughout the curves (see Fig. 1), more detailed examinations of the converging behaviour at different sections of the polarisation range were undertaken and are presented in the next section.
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
Dynamic differential polarisation behaviour at different Rext Three standard resistor values 35 kO, 2.5 kO and 100 O were selected for further detailed examination from the polarisation sweeps. These were chosen because they represent considerably different areas of the dynamic ranges. In the case of polarisation, 35 kO represents the activation, 2.5 kO the ohmic and 100 O the mass-transfer losses regions. In terms of power, 2.5 kO is a Rext value close to the maximum power transfer (MPT) point as determined by the polarisation sweeps themselves, whereas 35 kO and 100 O are good exemplars of suboptimal power on either side of MPT. Behaviour at 35 kO In this case of 5 min tc (see 5 min curves, Fig. 2A), the voltage drops initially after change of resistor (1st min) as expected, but then increases with no clear stabilisation until the next Rext step. This increase is rather unexpected because, if the stack was in steady-state (i.e. stable electrical output) then a decrease in the load (in this case from 38.5 to 35 kO) would result in a concomitant drop of the working voltage, as per Ohm’s law. It has been said that a short tc is not convenient for MFCs because it leads to an overestimation of the voltage and therefore the current/power produced [11,12]. The results in Fig. 2A show that a short tc (65 min) at such large Rext levels leads to in fact to an underestimation of the actual voltage levels that can be produced. Again by Ohm’s law, if the voltage increases at a fixed load, then the current increases as well; therefore, the current production can be underestimated, and not overestimated, by using short time constants at high Rext. A larger tc of 10 min (see 10 min curves, Fig. 2A and inset) exhibits an ideal plateau situation whereby constant output levels, indicative of steady-state, are obtained after just 30 seconds (99.95% voltage/current similarity between 0.5 and 10.0 min sampling points); such a tc can therefore be deemed sufficient for the presented MFCs at this level of Rext. Furthermore, linear-regression modelling of the 5 min steps (see blue dotted lines, Fig. 2A and inset) indicated that the final output levels after 10 min could be attained given sufficient time, even if the previous Rext step was set to a tc of 5 min, suggesting that perhaps tc could be selectively varied. This possibility is further discussed in the next sections. Behaviour at 100 O Detailed analysis of more than 25 polarisation sweeps showed that at 100 O, a different, more stable scenario unfolds (see Fig. 2B). After 30 s of stabilisation, the decrease in voltage (and therefore current) output averaged 2.53 ± 0.67% (0.720 ± 0.291 mV) with a 5 min tc, whilst with tc of 10 and 20 min, the drops averaged 2.52% and 2.97% (0.702 ± 0.127 and 0.722 ± 0.192 mV) in that order. More importantly, the output decrease after 5 min for the two longer tc sweeps only averaged 1.15% and 1.99% (0.312 ± 0.054 and 0.478 ± 0.138 mV) respectively. Given this (and the overall levels of similarity), it becomes clear that tc longer than 5 min are not necessary for estimating the output levels within an acceptable 97.5% confidence band at low Rext such as 100 O. This is in contrast to the results presented in 3.2.1 (tc > 5 min needed), but effectively opens the possibility of designing a polarisation sweep with differential tc, which could (i) increase the accuracy of polarisation data, (ii) help to reduce the overall testing time and (iii) increase the number of external loads tested. A reduction of testing times by differential tc could help avoid substrate depletion, particularly in batch mode systems; it is conventional within the field that substrate concentrations should not be reduced during the tests, because this would lead to uneven performance and non-reproducible results [13]. On the other hand, it has been
3
suggested that an increased number of Rext points are needed on both sides of the peak power area to correctly estimate performance [11], something that a Resistorstat programmed with differential tc can easily achieve. Polarisation sweeps could thus be designed to selectively investigate particular areas of interest (e.g. MPT) or where problems tend to arise e.g. power overshoots, whilst decreasing the time invested in areas of perhaps less interest (e.g. high-voltage/low-current). In a previous study [14], very large tc (up to 70 h) were selected in order to guarantee stable current production at each step and allegedly avoid overshoots. Because of the considerable time invested for each step, only a very limited number of external resistors were employed (<10), limiting the information obtained. However, with such large tc it is possible that the overshoot effect — which is transitory [15–17] and has been demonstrated to last down to seconds [12] — was overlooked, rather than non-occurring, simply because it was not sampled (1 sample every 30 min). Hence, very large tc are not a valid mechanism through which overshoots can be avoided. Moreover, it has been discussed that tc P hours can lead to changes in the growth and development of the biofilms [18] and it has also been mentioned that tc P days can affect ecological steady-states [12], effectively resulting in a different bioelectrochemical system before and after each Rext step. Such changes are highly undesirable for reproducibility and comparative purposes. The results hereby presented demonstrate that large tc are not needed to characterise the activation and mass-transfer limitation areas of a Microbial Fuel Cell system. When considering the aforementioned problems, it becomes apparent that large time constants should rather be avoided. Behaviour at 2.5 kO Differentially to the previous two cases, voltage stabilisation at 2.5 kX Rext was observed only at the relatively larger tc of 20 min, despite the presence of good reproducibility for shorter time constants (see Stack C 1 & 2, Fig. 2C). With this large tc (see Stack A 20 min, Fig. 2C), the observed drops in voltage/current averaged 2.94% after 30 s on a 2.5 kO load, whilst these were larger with the shorter tc of 5 and 10 min (average 3.29% and 3.99% respectively), effectively indicating that steady-state conditions could not be reached with the same level of confidence (when compared to P97.5% similarity for 100 O and 35 kO values) regardless of the time constant utilised, but particularly for those shorter than 20 min. The inability to reach steady-state (as in other cases at similar tc) can be explained in terms of power losses (PL), which in the context of power curves translate into the amplitude of power that is dropped (the overestimation after Rext change) until steady-state is reached. As previously discussed, PL = Rext I2, so when the currents or the external load are low (cases presented in 3.2.1 and 3.2.2, respectively), the effect is not as significant, allowing for steady-state in short tc. However, at or around the MPT point, where both current and Rext are relatively high, the phenomenon is more marked, explaining why there are usually larger overestimations and steady-states cannot be reached within the same timeframes. Further research is needed, with the same precision provided by the Resistorstat, to determine whether even longer tc are key in this particular near-MPT region to correctly estimate the capabilities of the tested systems. However, excessive elongation of time constants can lead to the aforementioned unwanted effects of system variation, thus questioning whether the technique is actually suitable for determination of steady-state performance around the MPT area. Other techniques based on real-time tracking are perhaps equally or even more suited to address this issue and have additionally shown to improve performance [7,19–21].
4
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
A
B
C
Fig. 2. (A) Examples of 2-MFC stacks’ converging polarisation behaviour at 35 kO Rext with 5 and 10 min tc. Inset: corresponding current production of Stack D at 5 and 10 min tc. Blue dotted lines: modelling of the convergence between 5 min and a 10 min steps. (B) Examples of 2-MFC stacks’ polarisation behaviour at 100 O Rext with 5, 10 and 20 min tc. Inset: corresponding current production for the same stacks/time constants. (C) Examples of 2-MFC stacks’ converging polarisation behaviour at 2.5 KO Rext with 5, 10 and 20 min tc. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
Polarisation sweeps are nevertheless very useful to understand the behaviour of a tested MFC over the broader range of voltage/current/power outputs and to elucidate specific problems highlighted by the occurrence of unexpected behaviours such as power overshoots. Capacitive-like boost behaviour For a much shorter tc of 1 min, the 2-MFC stacks constantly underperformed in terms of MPT. Though the latter was usually lower in magnitude (avg. 20.13 ± 4.37% inferior to values observed with a 10 min tc), a consistently higher maximum current production (MCP) was observed repeatedly (up to +50.89%, see Fig. 3A; average +29.16 ± 17.74%) for P12 min, confirming that this observation is not artefactual. This is similar to a capacitive behaviour, where energy transfer at low current levels 6MPT is restricted by the high Rext values, but is then more effectively released at lower Rext, perhaps as a result of fast resistor switching.
5
We have recently demonstrated that MFC output, when tested under polarisation, is heavily dependent upon the previous activity of the tested cells [22], with considerable hysteresis observed when performing consecutive sweeps with different time constants. The capacitive-like boost phenomenon also seems to depend on this, and so the magnitude of the boost observed is thought to change depending on the level of activity of the MFCs prior to the temporary output improvement. This is evidenced by the differences in extra current production shown in Fig. 3A: the first sweep returns a lower MCP (but a higher MPT) whilst the second sweep results in a lower MPT, but with a higher MCP, thus decreasing the observable percentage boost when compared to the 1 min tc sweep (+50.89 vs +40.47%, respectively). These differences are further evidenced in Fig. 3B, where consecutive polarisation sweeps that did not lead to capacitive boost (tc: 10 min then 5 min) followed by one that did (tc: 1 min) result in significantly-different levels of total energy transmission (standard area-under-the-curve analysis): 285.26, 207.95 and
Fig. 3. (A) Power curves of consecutive sweeps (10 min, 5 min and 1 min tc) for a 2-MFC stack connected in series using the Resistorstat. The maximum overperformance in terms of current production (vs 10 min tc) is indicated and highlighted in orange; the difference with a 5 min tc is also indicated. (B) Polarisation curves for the same sweeps. The legend indicates the total energy transmission observed from each sweep, based on area-under-the-curve analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
6
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
312.23 mJ s 1 respectively. Accordingly, the energy transmission boost can be as little as +9.45% if the MFCs have been inactive (cells were kept in open-circuit prior to first sweep) or as large as +50.15% if subject to intensive activity (difference between the second and last sweeps). To date, only three other studies have observed this phenomenon in microbial fuel cells, highlighting the need further research. The study of Uría and colleagues with S. oneidensis MR-1 probably presented the first indications of capacitive behaviour in individual MFCs [23]; although the experiments hereby presented confirm (in mixed-culture MFCs) their observation of superior performance after open-circuit time, results herewith also demonstrate that there is no need to periodically break the electrical circuit (as in Uría et al.) to benefit from a capacitive boost. It is also demonstrated that the capacitive behaviour can be exploited without expensive or specialised electrode materials – as attempted by other studies [24,25] – and that therefore this phenomenon is fundamentally lying with the biofilms as previously suggested [23]. Although it can be hypothesised that the boost is unsustainable for mid to long-term scales (Phours), quick Rext changes demonstrate a possible method to exploit otherwise-untapped extra energy. A potential gain of 50% in current production and energy transmission, as hereby demonstrated, represents a massive incentive needing exploration. Further research is therefore needed to understand the biological and electric implications and limitations of the boost phenomenon and to determine how MFC-stacks can be reversibly driven into such high-energy states without affecting their long-term performance. We envisage that the boost could have an even more significant impact if several stacks connected together were driven into this transitional high-energy state. This is of particular interest for dynamic MFC-powered applications in robotics, where e.g. an ‘escape’ actuation may be suddenly required.
Conclusions The present study highlights the behaviour of MFCs when confronted with different Rext and tc. Results show that there is differential convergence depending on the zone of the polarisation/ power curves, making the use of differential time constants possible. It is also demonstrated that for the activation- and mass-transfer-losses zones, large tc are not needed to obtain confident steady-states as previously suggested. Furthermore, it is shown that the polarisation technique is perhaps not suitable to estimate MPT. Finally, the capacitive-like boost phenomenon, a transitory regime with up to 50% current/energy transfer gain, is presented; this high-energy state could prove important for future MFC-powered applications.
Acknowledgements The authors would like to thank Sam Coupland of the Bristol Robotics Laboratory for his help in the design and fabrication of the MFC parts. Pablo Ledezma is supported by the National Science and Technology Council of Mexico (CONACYT) Ref. 206298. Ioannis Ieropoulos is supported by the Engineering and Physical Sciences Research Council of the UK (EPSRC) CAF EP/I004653/1.
References [1] Pant D, Singh A, Van Bogaert G, Gallego YA, Diels L, Vanbroekhoven K. An introduction to the life cycle assessment (LCA) of bioelectrochemical systems (BES) for sustainable energy and product generation: relevance and key aspects. Renew Sustain Energy Rev 2011;15:1305–13. [2] Clauwaert P, Aelterman P, Pham T, De Schamphelaire L, Carballa M, Rabaey K, Verstraete W. Minimizing losses in bio-electrochemical systems: the road to applications. Appl Microbiol Biotechnol 2008;79:901–13. [3] Dicks JLaA. Fuel cell systems explained, 2nd ed. New York: John Wiley & Sons; 2003. [4] Ledezma P, Ieropoulos I, Greenman J. Comparative analysis of different polymer materials for the construction of microbial fuel cell stacks. J Biotechnol 2010;150:143. [5] Ledezma P, Greenman J, Ieropoulos I. Maximising electricity production by controlling the biofilm specific growth rate in microbial fuel cells. Bioresour Technol 2012;118:615–8. [6] Ieropoulos I, Greenman J, Melhuish C, Horsfield I. In: Fellermann H, Dörr M, Hanczyc M, Laursen L, Maurer S, Merkle D, Monnard P, Stoy K, Rasmussen S, editors. Twelfth international conference on the synthesis and simulation of living systems. Odense Denmark: Massachusetts Institute of Technology Press; 2010. p. 733–40. [7] Degrenne N, Buret F, Allard B, Bevilacqua P. Electrical energy generation from a large number of microbial fuel cells operating at maximum power point electrical load. J Power Sources 2012;205:188–93. [8] Ledezma P, Greenman J, Greenman I. MFC-Cascade Stacks maximise COD reduction and avoid voltage reversal under adverse conditions. Bioresour Technol 2013;134:158–65. [9] Ledezma P, Stinchcombe A, Greenman J, Ieropoulos I. The first self-sustainable microbial fuel cell stack. Phys Chem Chem Phys 2013;15:2278–81. [10] Ieropoulos I, Ledezma P, Stinchcombe A, Papaharalabos G, Melhuish C, Greenman J. Waste to real energy: the first MFC powered mobile phone, Phys Chem Chem Phys; 2013. [11] Logan BE. Essential data and techniques for conducting microbial fuel cell and other types of bioelectrochemical system experiments. ChemSusChem 2012;5:988–94. [12] Winfield J, Ieropoulos I, Greenman J, Dennis J. The overshoot phenomenon as a function of internal resistance in microbial fuel cells. Bioelectrochemistry 2011;81:22–7. [13] Ieropoulos IA, Winfield J, Greenman J, Melhuish C. Small scale microbial fuel cells and different ways of reporting output. ECS Trans 2010;28:1–9. [14] Watson VJ, Logan BE. Analysis of polarization methods for elimination of power overshoot in microbial fuel cells. Electrochem Commun 2011;13:54–6. [15] Nien P-C, Lee C-Y, Ho K-C, Adav SS, Liu L, Wang A, Ren N, Lee D-J. Power overshoot in two-chambered microbial fuel cell (MFC). Bioresour Technol 2011;102:4742–6. [16] Ieropoulos I, Winfield J, Greenman J. Effects of flow-rate, inoculum and time on the internal resistance of microbial fuel cells. Bioresour Technol 2010;101:3520–5. [17] Nam J-Y, Kim H-W, Lim K-H, Shin H-S, Logan BE. Variation of power generation at different buffer types and conductivities in single chamber microbial fuel cells. Biosens Bioelectron 2010;25:1155–9. [18] Cheng KY, Cord-Ruwisch R, Ho G. A new approach for in situ cyclic voltammetry of a microbial fuel cell biofilm without using a potentiostat. Bioelectrochemistry 2009;74:227–31. [19] Woodward L, Perrier M, Srinivasan B, Pinto RP, Tartakovsky B. Comparison of real-time methods for maximizing power output in microbial fuel cells. AIChE J 2010;56:2742–50. [20] Pinto RP, Srinivasan B, Guiot SR, Tartakovsky B. The effect of real-time external resistance optimization on microbial fuel cell performance. Water Res 2011;45:1571–8. [21] Premier GC, Kim JR, Michie I, Dinsdale RM, Guwy AJ. Automatic control of load increases power and efficiency in a microbial fuel cell. J Power Sources 2011;196:2013–9. [22] Degrenne N, Ledezma P, Bevilacqua P, Buret F, Allard B, Greenman J, Ieropoulos IA. Bi-directional electrical characterisation of microbial fuel cell. Bioresour Technol 2013;128:769–73. [23] Uría N, Muñoz Berbel X, Sánchez O, Muñoz FX, Mas J. Transient storage of electrical charge in biofilms of shewanella oneidensis MR-1 growing in a microbial fuel cell. Environ Sci Technol 2011;45:10250–6. [24] Peng X, Yu H, Wang X, Zhou Q, Zhang S, Geng L, Sun J, Cai Z. Enhanced performance and capacitance behavior of anode by rolling Fe3O4 into activated carbon in microbial fuel cells. Bioresour Technol 2012;121:450–3. [25] Deeke A, Sleutels THJA, Hamelers HVM, Buisman CJN. Capacitive bioanodes enable renewable energy storage in microbial fuel cells. Environ Sci Technol 2012;46:3554–60.
Contents lists available at ScienceDirect
Sustainable Energy Technologies and Assessments journal homepage: www.elsevier.com/locate/seta
Original Research Article
Dynamic polarisation reveals differential steady-state stabilisation and capacitive-like behaviour in microbial fuel cells Pablo Ledezma a,⇑, Nicolas Degrenne b, Pascal Bevilacqua c, François Buret b, Bruno Allard c, John Greenman d, Ioannis Ieropoulos a a
Bristol Robotics Laboratory, Universities of Bristol and of the West of England, Frenchay Campus, Bristol BS34 8QZ, UK Laboratoire Ampère, Ecole centrale de Lyon, Université de Lyon, Ecully 69134, France Laboratoire Ampère, INSA de Lyon, Université de Lyon, Villeurbanne 69100, France d Department of Applied Sciences, University of the West of England, Bristol BS16 1QY, UK b c
a r t i c l e
i n f o
Article history: Received 28 July 2013 Revised 14 October 2013 Accepted 25 October 2013
Keywords: Microbial fuel cells External resistors Polarisation curve Time constant Steady-states
a b s t r a c t In this paper we present several preliminary results produced with a purposely-designed external-resistor (Rext) sweeping tool for microbial fuel cells (MFCs). Fast sampling rates show that MFCs exhibit differential steady-state stabilisation behaviours depending on Rext, with consequences for time constant (tc) selection. At high Rext (35 kO), it is demonstrated that a tc P 10 min avoids underestimation not overestimation, whilst at low Rext (100 O) 5 min are sufficient, suggesting that sweeps with variable tc are possible. However, within the maximum power transfer range (2.5 kO), steady-states are only observed at 20 min tc but with a smaller confidence interval, questioning whether the polarisation technique is suitable to estimate maximum power transfer. Finally, a strategy towards the exploitation of a capacitive-like behaviour in MFCs is proposed, tapping into P10 min periods with up to 50% higher current and energy transfer that could prove important for MFC-powered applications. Ó 2013 Elsevier Ltd. All rights reserved.
Introduction Microbial fuel cells (MFCs) are an exciting technology with a growing scientific interest and research dynamism due to the potential of sustainable electricity production from organic waste and, notably, wastewater. The latter is either released untreated into the environment – with severe environmental consequences – or treated through expensive and energy intensive processes. MFCs offer the possibility of transforming such wastewater treatment plants into energy- and carbon-neutral systems or even the longer-term potential of net energy production, similar to novel anaerobic digestion plants but operating at higher efficiencies and requiring less investment [1]. One of the most commonly used tools for characterising the behaviour of MFC systems is known as the polarisation technique. It is a process whereby the fuel cell is subjected to a set number of external loads (Rext of known resistance in O) over a constant period of time (time constant, tc), resulting in the production of an electric current (in A) at a corresponding cell voltage (in V). The graphical representation of the obtained results or Polarisation curve (voltage vs current) provides considerable information about the tested systems, particularly about the associated activation, ⇑ Corresponding author. Tel.: +44 (0) 117 32 86350; fax: +44 (0) 117 32 83960. E-mail address: [email protected] (P. Ledezma). 2213-1388/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.seta.2013.10.008
ohmic and mass transfer losses [2]. Moreover, from the derivation of Joule’s law, the electrical power (in W) can be calculated at the different exhibited current levels and plotted as a Power curve (power vs current), which again reveals information about the behaviour and stability of the system predominantly in comparison with a mathematical model. According to Jacobi’s law, an ideal fuel cell produces a parabolic Power curve whereby the maximum power transfer (MPT) point occurs at the mid points of current and voltage [3]. This protocol has become almost ubiquitous for determining the maximum attainable power from MFCs. A debate is currently ongoing with regards to this technique and in particular to the value (in O) and quantity of external resistors employed, as well as the time constant (tc) in terms of accurately determining the real, sustainable MPT by MFCs. This investigation addresses these concerns, presenting a more accurate polarisation technique that aims to question the use of constant or long time intervals for the evaluation of MFC performance and whether the technique is suitable for the estimation of MPT. Materials and methods MFC assembly and operation Eight 2-chamber MFCs (20 mL/side) made of four different materials [4] and of properties previously described [5] were
2
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
inoculated with activated sewage sludge and then operated in continuous mode, under identical conditions, for at least 18 months. During this period, the MFCs were fed with acetate TYE (5 mM C2H3NaO2, 0.1% w/v Tryptone, 0.05% w/v Yeast Extract) at a low flow rate of 120 lL min 1, with tap water (D.O. avg. 6.5 mg L 1; Conductivity avg. 680 lS cm 1) as the catholyte at a flow rate of 19.35 mL min 1. Our research focuses on the utilisation of MFC systems as power sources for practical applications e.g. robots [6]. For such purposes, we have already demonstrated that a suitable strategy to achieve the necessary voltages and currents is to construct MFC collectives (stacks) with combinations of series/parallel connections [6] and that we are able to maintain the operation of such stacks at MPT point with high efficiency [7], both necessary elements for the stable and predictable operation of applications. More recently, we have also shown that a stacking architecture that balances the internal resistances of embedded cells avoids cell voltage reversal even under fuel starvation [8] and that this strategy can be used for scaled-up reactors with up to 20 MFCs in series [9]. We have utilised the latter approach to energise applications previously considered impossible with MFCs e.g. a multi-channel peristaltic pump [9] and a mobile phone [10]. However, a significant amount of research is still needed into stacking at the fundamental levels – 2 to 4 MFCs in different configurations – in order to elucidate techniques that can enhance their collective electrical output without reversal. In concordance with these objectives, the experiments hereby presented involve four stacks of two MFCs connected in series (namely stacks A, B, C & D). Prior to stacking, all cells were individually verified for output under continuous load and polarisation conditions and no significant differences were found between MFCs fabricated in the same material (data not shown). The electrical output of each stack (voltage) was monitored/controlled in real-time with the tool described below.
A
B
Dynamic polarisation method with novel Resistorstat The constructed stacks were connected to an 8-channel automated Resistorstat tool, whose development and architecture is described in detail by Degrenne et al. [7]. This system was programmed to perform sweeps of external resistor (Rext) values, ranging decreasingly from 38.5 kX to 4 X (35 values with avg. 18.5% resistance reduction per step, so as to make the reduction rates constant), at different time constant (tc) intervals (1, 5, 10 and 20 min). The cells were kept in open-circuit mode for 1 h prior to testing. Their voltage was recorded every 30 s in order to monitor the dynamic response of MFCs to changes in Rext over the extent of each tc step. Calculations The resulting voltage output readings were automatically processed by the Resistorstat system, calculating current (I) and power (P) output per MFC stack, as previously described [7]. For the sake of clarity, exemplars of the most representative results are shown, which illustrate responses that were consistently observed in the more than 25 polarisation sweeps performed per stack.
Results and discussion Novel dynamic polarisation and power curves The data produced with the technique hereby presented result in a new form of polarisation (Fig. 1A) and power curves (Fig. 1B), distinguished by the presence of spikes. These seem to indicate a
Fig. 1. (A) Polarisation curve of a 2-MFC stack connected in series with a 30 s sampling time resolution and a 5 min time constant. The modelling lines (in gray) highlight the fact that the spikes converge, in a linear manner, in concordance with the external resistor employed. (B) Power curve of the same 2-MFC stack. The modelling lines (in gray) exhibit the parabolic convergence behaviour of the spikes. In both cases, blue lines indicate the typical polarisation and power curves that would be obtained from examined systems, showing the difference with dynamic curves.
plateauing tendency in the output or convergence for each Rext tested. Although the variation observed can be considered low, both curves (see Fig. 1A and B) illustrate that the aforementioned behaviour is different between points tested in terms of the orientation/ width of the spikes depending on the zone in the curves. Simple modelling revealed that the orientation of the spikes in the polarisation curve can be explained by a linear relationship depending on Rext, each spike aligning with the Cartesian equation V = Rext I of Ohm’s law (grey modelling lines, Fig. 1A). In a similar manner, the spikes’ orientation in the power curve (see Fig. 1B) can be explained by the square function P = Rext I2. The latter formula is commonly employed in electrical engineering to determine power losses in an electrical transmission, so the spikes could be considered as an indicator of the power that is lost (or in the power curves, initially overestimated) whilst the system stabilises into a new steady-state (stable electrical production over time). Since it was observed that these spikes appeared substantially different throughout the curves (see Fig. 1), more detailed examinations of the converging behaviour at different sections of the polarisation range were undertaken and are presented in the next section.
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
Dynamic differential polarisation behaviour at different Rext Three standard resistor values 35 kO, 2.5 kO and 100 O were selected for further detailed examination from the polarisation sweeps. These were chosen because they represent considerably different areas of the dynamic ranges. In the case of polarisation, 35 kO represents the activation, 2.5 kO the ohmic and 100 O the mass-transfer losses regions. In terms of power, 2.5 kO is a Rext value close to the maximum power transfer (MPT) point as determined by the polarisation sweeps themselves, whereas 35 kO and 100 O are good exemplars of suboptimal power on either side of MPT. Behaviour at 35 kO In this case of 5 min tc (see 5 min curves, Fig. 2A), the voltage drops initially after change of resistor (1st min) as expected, but then increases with no clear stabilisation until the next Rext step. This increase is rather unexpected because, if the stack was in steady-state (i.e. stable electrical output) then a decrease in the load (in this case from 38.5 to 35 kO) would result in a concomitant drop of the working voltage, as per Ohm’s law. It has been said that a short tc is not convenient for MFCs because it leads to an overestimation of the voltage and therefore the current/power produced [11,12]. The results in Fig. 2A show that a short tc (65 min) at such large Rext levels leads to in fact to an underestimation of the actual voltage levels that can be produced. Again by Ohm’s law, if the voltage increases at a fixed load, then the current increases as well; therefore, the current production can be underestimated, and not overestimated, by using short time constants at high Rext. A larger tc of 10 min (see 10 min curves, Fig. 2A and inset) exhibits an ideal plateau situation whereby constant output levels, indicative of steady-state, are obtained after just 30 seconds (99.95% voltage/current similarity between 0.5 and 10.0 min sampling points); such a tc can therefore be deemed sufficient for the presented MFCs at this level of Rext. Furthermore, linear-regression modelling of the 5 min steps (see blue dotted lines, Fig. 2A and inset) indicated that the final output levels after 10 min could be attained given sufficient time, even if the previous Rext step was set to a tc of 5 min, suggesting that perhaps tc could be selectively varied. This possibility is further discussed in the next sections. Behaviour at 100 O Detailed analysis of more than 25 polarisation sweeps showed that at 100 O, a different, more stable scenario unfolds (see Fig. 2B). After 30 s of stabilisation, the decrease in voltage (and therefore current) output averaged 2.53 ± 0.67% (0.720 ± 0.291 mV) with a 5 min tc, whilst with tc of 10 and 20 min, the drops averaged 2.52% and 2.97% (0.702 ± 0.127 and 0.722 ± 0.192 mV) in that order. More importantly, the output decrease after 5 min for the two longer tc sweeps only averaged 1.15% and 1.99% (0.312 ± 0.054 and 0.478 ± 0.138 mV) respectively. Given this (and the overall levels of similarity), it becomes clear that tc longer than 5 min are not necessary for estimating the output levels within an acceptable 97.5% confidence band at low Rext such as 100 O. This is in contrast to the results presented in 3.2.1 (tc > 5 min needed), but effectively opens the possibility of designing a polarisation sweep with differential tc, which could (i) increase the accuracy of polarisation data, (ii) help to reduce the overall testing time and (iii) increase the number of external loads tested. A reduction of testing times by differential tc could help avoid substrate depletion, particularly in batch mode systems; it is conventional within the field that substrate concentrations should not be reduced during the tests, because this would lead to uneven performance and non-reproducible results [13]. On the other hand, it has been
3
suggested that an increased number of Rext points are needed on both sides of the peak power area to correctly estimate performance [11], something that a Resistorstat programmed with differential tc can easily achieve. Polarisation sweeps could thus be designed to selectively investigate particular areas of interest (e.g. MPT) or where problems tend to arise e.g. power overshoots, whilst decreasing the time invested in areas of perhaps less interest (e.g. high-voltage/low-current). In a previous study [14], very large tc (up to 70 h) were selected in order to guarantee stable current production at each step and allegedly avoid overshoots. Because of the considerable time invested for each step, only a very limited number of external resistors were employed (<10), limiting the information obtained. However, with such large tc it is possible that the overshoot effect — which is transitory [15–17] and has been demonstrated to last down to seconds [12] — was overlooked, rather than non-occurring, simply because it was not sampled (1 sample every 30 min). Hence, very large tc are not a valid mechanism through which overshoots can be avoided. Moreover, it has been discussed that tc P hours can lead to changes in the growth and development of the biofilms [18] and it has also been mentioned that tc P days can affect ecological steady-states [12], effectively resulting in a different bioelectrochemical system before and after each Rext step. Such changes are highly undesirable for reproducibility and comparative purposes. The results hereby presented demonstrate that large tc are not needed to characterise the activation and mass-transfer limitation areas of a Microbial Fuel Cell system. When considering the aforementioned problems, it becomes apparent that large time constants should rather be avoided. Behaviour at 2.5 kO Differentially to the previous two cases, voltage stabilisation at 2.5 kX Rext was observed only at the relatively larger tc of 20 min, despite the presence of good reproducibility for shorter time constants (see Stack C 1 & 2, Fig. 2C). With this large tc (see Stack A 20 min, Fig. 2C), the observed drops in voltage/current averaged 2.94% after 30 s on a 2.5 kO load, whilst these were larger with the shorter tc of 5 and 10 min (average 3.29% and 3.99% respectively), effectively indicating that steady-state conditions could not be reached with the same level of confidence (when compared to P97.5% similarity for 100 O and 35 kO values) regardless of the time constant utilised, but particularly for those shorter than 20 min. The inability to reach steady-state (as in other cases at similar tc) can be explained in terms of power losses (PL), which in the context of power curves translate into the amplitude of power that is dropped (the overestimation after Rext change) until steady-state is reached. As previously discussed, PL = Rext I2, so when the currents or the external load are low (cases presented in 3.2.1 and 3.2.2, respectively), the effect is not as significant, allowing for steady-state in short tc. However, at or around the MPT point, where both current and Rext are relatively high, the phenomenon is more marked, explaining why there are usually larger overestimations and steady-states cannot be reached within the same timeframes. Further research is needed, with the same precision provided by the Resistorstat, to determine whether even longer tc are key in this particular near-MPT region to correctly estimate the capabilities of the tested systems. However, excessive elongation of time constants can lead to the aforementioned unwanted effects of system variation, thus questioning whether the technique is actually suitable for determination of steady-state performance around the MPT area. Other techniques based on real-time tracking are perhaps equally or even more suited to address this issue and have additionally shown to improve performance [7,19–21].
4
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
A
B
C
Fig. 2. (A) Examples of 2-MFC stacks’ converging polarisation behaviour at 35 kO Rext with 5 and 10 min tc. Inset: corresponding current production of Stack D at 5 and 10 min tc. Blue dotted lines: modelling of the convergence between 5 min and a 10 min steps. (B) Examples of 2-MFC stacks’ polarisation behaviour at 100 O Rext with 5, 10 and 20 min tc. Inset: corresponding current production for the same stacks/time constants. (C) Examples of 2-MFC stacks’ converging polarisation behaviour at 2.5 KO Rext with 5, 10 and 20 min tc. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
Polarisation sweeps are nevertheless very useful to understand the behaviour of a tested MFC over the broader range of voltage/current/power outputs and to elucidate specific problems highlighted by the occurrence of unexpected behaviours such as power overshoots. Capacitive-like boost behaviour For a much shorter tc of 1 min, the 2-MFC stacks constantly underperformed in terms of MPT. Though the latter was usually lower in magnitude (avg. 20.13 ± 4.37% inferior to values observed with a 10 min tc), a consistently higher maximum current production (MCP) was observed repeatedly (up to +50.89%, see Fig. 3A; average +29.16 ± 17.74%) for P12 min, confirming that this observation is not artefactual. This is similar to a capacitive behaviour, where energy transfer at low current levels 6MPT is restricted by the high Rext values, but is then more effectively released at lower Rext, perhaps as a result of fast resistor switching.
5
We have recently demonstrated that MFC output, when tested under polarisation, is heavily dependent upon the previous activity of the tested cells [22], with considerable hysteresis observed when performing consecutive sweeps with different time constants. The capacitive-like boost phenomenon also seems to depend on this, and so the magnitude of the boost observed is thought to change depending on the level of activity of the MFCs prior to the temporary output improvement. This is evidenced by the differences in extra current production shown in Fig. 3A: the first sweep returns a lower MCP (but a higher MPT) whilst the second sweep results in a lower MPT, but with a higher MCP, thus decreasing the observable percentage boost when compared to the 1 min tc sweep (+50.89 vs +40.47%, respectively). These differences are further evidenced in Fig. 3B, where consecutive polarisation sweeps that did not lead to capacitive boost (tc: 10 min then 5 min) followed by one that did (tc: 1 min) result in significantly-different levels of total energy transmission (standard area-under-the-curve analysis): 285.26, 207.95 and
Fig. 3. (A) Power curves of consecutive sweeps (10 min, 5 min and 1 min tc) for a 2-MFC stack connected in series using the Resistorstat. The maximum overperformance in terms of current production (vs 10 min tc) is indicated and highlighted in orange; the difference with a 5 min tc is also indicated. (B) Polarisation curves for the same sweeps. The legend indicates the total energy transmission observed from each sweep, based on area-under-the-curve analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
6
P. Ledezma et al. / Sustainable Energy Technologies and Assessments 5 (2014) 1–6
312.23 mJ s 1 respectively. Accordingly, the energy transmission boost can be as little as +9.45% if the MFCs have been inactive (cells were kept in open-circuit prior to first sweep) or as large as +50.15% if subject to intensive activity (difference between the second and last sweeps). To date, only three other studies have observed this phenomenon in microbial fuel cells, highlighting the need further research. The study of Uría and colleagues with S. oneidensis MR-1 probably presented the first indications of capacitive behaviour in individual MFCs [23]; although the experiments hereby presented confirm (in mixed-culture MFCs) their observation of superior performance after open-circuit time, results herewith also demonstrate that there is no need to periodically break the electrical circuit (as in Uría et al.) to benefit from a capacitive boost. It is also demonstrated that the capacitive behaviour can be exploited without expensive or specialised electrode materials – as attempted by other studies [24,25] – and that therefore this phenomenon is fundamentally lying with the biofilms as previously suggested [23]. Although it can be hypothesised that the boost is unsustainable for mid to long-term scales (Phours), quick Rext changes demonstrate a possible method to exploit otherwise-untapped extra energy. A potential gain of 50% in current production and energy transmission, as hereby demonstrated, represents a massive incentive needing exploration. Further research is therefore needed to understand the biological and electric implications and limitations of the boost phenomenon and to determine how MFC-stacks can be reversibly driven into such high-energy states without affecting their long-term performance. We envisage that the boost could have an even more significant impact if several stacks connected together were driven into this transitional high-energy state. This is of particular interest for dynamic MFC-powered applications in robotics, where e.g. an ‘escape’ actuation may be suddenly required.
Conclusions The present study highlights the behaviour of MFCs when confronted with different Rext and tc. Results show that there is differential convergence depending on the zone of the polarisation/ power curves, making the use of differential time constants possible. It is also demonstrated that for the activation- and mass-transfer-losses zones, large tc are not needed to obtain confident steady-states as previously suggested. Furthermore, it is shown that the polarisation technique is perhaps not suitable to estimate MPT. Finally, the capacitive-like boost phenomenon, a transitory regime with up to 50% current/energy transfer gain, is presented; this high-energy state could prove important for future MFC-powered applications.
Acknowledgements The authors would like to thank Sam Coupland of the Bristol Robotics Laboratory for his help in the design and fabrication of the MFC parts. Pablo Ledezma is supported by the National Science and Technology Council of Mexico (CONACYT) Ref. 206298. Ioannis Ieropoulos is supported by the Engineering and Physical Sciences Research Council of the UK (EPSRC) CAF EP/I004653/1.
References [1] Pant D, Singh A, Van Bogaert G, Gallego YA, Diels L, Vanbroekhoven K. An introduction to the life cycle assessment (LCA) of bioelectrochemical systems (BES) for sustainable energy and product generation: relevance and key aspects. Renew Sustain Energy Rev 2011;15:1305–13. [2] Clauwaert P, Aelterman P, Pham T, De Schamphelaire L, Carballa M, Rabaey K, Verstraete W. Minimizing losses in bio-electrochemical systems: the road to applications. Appl Microbiol Biotechnol 2008;79:901–13. [3] Dicks JLaA. Fuel cell systems explained, 2nd ed. New York: John Wiley & Sons; 2003. [4] Ledezma P, Ieropoulos I, Greenman J. Comparative analysis of different polymer materials for the construction of microbial fuel cell stacks. J Biotechnol 2010;150:143. [5] Ledezma P, Greenman J, Ieropoulos I. Maximising electricity production by controlling the biofilm specific growth rate in microbial fuel cells. Bioresour Technol 2012;118:615–8. [6] Ieropoulos I, Greenman J, Melhuish C, Horsfield I. In: Fellermann H, Dörr M, Hanczyc M, Laursen L, Maurer S, Merkle D, Monnard P, Stoy K, Rasmussen S, editors. Twelfth international conference on the synthesis and simulation of living systems. Odense Denmark: Massachusetts Institute of Technology Press; 2010. p. 733–40. [7] Degrenne N, Buret F, Allard B, Bevilacqua P. Electrical energy generation from a large number of microbial fuel cells operating at maximum power point electrical load. J Power Sources 2012;205:188–93. [8] Ledezma P, Greenman J, Greenman I. MFC-Cascade Stacks maximise COD reduction and avoid voltage reversal under adverse conditions. Bioresour Technol 2013;134:158–65. [9] Ledezma P, Stinchcombe A, Greenman J, Ieropoulos I. The first self-sustainable microbial fuel cell stack. Phys Chem Chem Phys 2013;15:2278–81. [10] Ieropoulos I, Ledezma P, Stinchcombe A, Papaharalabos G, Melhuish C, Greenman J. Waste to real energy: the first MFC powered mobile phone, Phys Chem Chem Phys; 2013. [11] Logan BE. Essential data and techniques for conducting microbial fuel cell and other types of bioelectrochemical system experiments. ChemSusChem 2012;5:988–94. [12] Winfield J, Ieropoulos I, Greenman J, Dennis J. The overshoot phenomenon as a function of internal resistance in microbial fuel cells. Bioelectrochemistry 2011;81:22–7. [13] Ieropoulos IA, Winfield J, Greenman J, Melhuish C. Small scale microbial fuel cells and different ways of reporting output. ECS Trans 2010;28:1–9. [14] Watson VJ, Logan BE. Analysis of polarization methods for elimination of power overshoot in microbial fuel cells. Electrochem Commun 2011;13:54–6. [15] Nien P-C, Lee C-Y, Ho K-C, Adav SS, Liu L, Wang A, Ren N, Lee D-J. Power overshoot in two-chambered microbial fuel cell (MFC). Bioresour Technol 2011;102:4742–6. [16] Ieropoulos I, Winfield J, Greenman J. Effects of flow-rate, inoculum and time on the internal resistance of microbial fuel cells. Bioresour Technol 2010;101:3520–5. [17] Nam J-Y, Kim H-W, Lim K-H, Shin H-S, Logan BE. Variation of power generation at different buffer types and conductivities in single chamber microbial fuel cells. Biosens Bioelectron 2010;25:1155–9. [18] Cheng KY, Cord-Ruwisch R, Ho G. A new approach for in situ cyclic voltammetry of a microbial fuel cell biofilm without using a potentiostat. Bioelectrochemistry 2009;74:227–31. [19] Woodward L, Perrier M, Srinivasan B, Pinto RP, Tartakovsky B. Comparison of real-time methods for maximizing power output in microbial fuel cells. AIChE J 2010;56:2742–50. [20] Pinto RP, Srinivasan B, Guiot SR, Tartakovsky B. The effect of real-time external resistance optimization on microbial fuel cell performance. Water Res 2011;45:1571–8. [21] Premier GC, Kim JR, Michie I, Dinsdale RM, Guwy AJ. Automatic control of load increases power and efficiency in a microbial fuel cell. J Power Sources 2011;196:2013–9. [22] Degrenne N, Ledezma P, Bevilacqua P, Buret F, Allard B, Greenman J, Ieropoulos IA. Bi-directional electrical characterisation of microbial fuel cell. Bioresour Technol 2013;128:769–73. [23] Uría N, Muñoz Berbel X, Sánchez O, Muñoz FX, Mas J. Transient storage of electrical charge in biofilms of shewanella oneidensis MR-1 growing in a microbial fuel cell. Environ Sci Technol 2011;45:10250–6. [24] Peng X, Yu H, Wang X, Zhou Q, Zhang S, Geng L, Sun J, Cai Z. Enhanced performance and capacitance behavior of anode by rolling Fe3O4 into activated carbon in microbial fuel cells. Bioresour Technol 2012;121:450–3. [25] Deeke A, Sleutels THJA, Hamelers HVM, Buisman CJN. Capacitive bioanodes enable renewable energy storage in microbial fuel cells. Environ Sci Technol 2012;46:3554–60.