MFC-Cascade Stacksmaximise COD reduction and avoid voltage reversal under adverse conditions

Pablo Ledezma1, John Greenman2 and Ioannis Ieropoulos1*

1Bristol Robotics Laboratory, Universities of Bristol and of the West of England, Frenchay Campus, Bristol BS34 8QZ, U.K. 2 Department of Applied Sciences, University of the West of England, Bristol BS16 1QY, U.K.

* Corresponding author:

Telephone: +44 (0) 117 32 86318

Fax: +44 (0) 117 32 83960

Abstract

This study presents a 6-unit continuous flow Microbial Fuel Cell (MFC) stack organised as a vertical cascade and tested under different electrical configurations. Under parallel conditions, the stacked-system demonstrated stable operation despite the substrate imbalances inherent to cascading and higher power and current densities than individual MFCs (positive stacking effect), whilst the cascading dynamic allowed for a cumulative COD reduction of >95% in less than 5.7h, equivalent to a very competitive 7.97 kgCOD m-3 d-1. When changed to a series configuration, the stack exhibited considerable losses (cross-conductance) unless correct fluidic/electrical insulation of the units was applied; with this problem resolved, the stack in series also exhibited electrical stability and performance superior to individual cells. In both electrical configurations, the 6MFC system was purposefully deprived of feed repeatedly for up to 15 days, and in no case resulting in significant losses post-starvation. The combined results of 14 months of experimentation hereby presented, demonstrate that cascade-stacking of small units can result in enhanced electricity production (vs single large units) and treatment rates without the need for expensive catalysts. Moreover, it is also demonstrated that MFC operation (even when the continuous flow is stopped) in series configuration, substrate imbalance and starvation does not necessarily result in cell-voltage reversal.

Keywords

Microbial fuel cells; stacking; wastewater treatment; voltage reversal

Introduction

Microbial Fuel Cells (MFCs) have been receiving increased interest in the last three decades, since the production of electricity, from a wide variety of organic wastes, was demonstratedwithout the need for artificial mediators (Habermann & Pommer, 1991). MFCs are thus an attractive alternative for effluent treatment (Clauwaert et al., 2008) and despite the technological progress, it is clear that current bench top MFCs (particularly individual systems), cannot cope with the volumes and flow rates needed to viably operate in a wastewater treatment plant. Scaled-up systems are therefore needed to allow for efficient and high treatment of such large volumes, but also to be able to produce sufficient electricity so that their operation for wastewater treatment can be made autonomous (Clauwaert et al., 2008; Winfield et al., 2011).

However, scaling-up via increasing reactor sizes carries the inherent challenge of increased internal resistances (Rint), since this is volume-dependent (Dekker et al., 2009; Ieropoulos et al., 2008). This is probably one of many factors that could explain why large volume MFCs have not demonstrated high-power output levels as smaller bench systems have. In fact, power density has been demonstrated to be higher at the smaller scale (Ieropoulos et al., 2008), meaning that practical (useful) levels of power from MFCs are more likely to be generated via miniaturisation and multiplication, rather than enlarging single reactors (Fornero et al., 2010; Ieropoulos et al., 2008; Ieropoulos et al., 2010a; Kim et al., 2011).In terms of MFC unit stacking of various sizes, there have been numerous reports in the literature,(Aelterman et al., 2006; Dekker et al., 2009; Ieropoulos et al., 2003; Ieropoulos et al., 2005; Ieropoulos et al., 2010a; Ieropoulos et al., 2010b; Melhuish et al., 2006; Oh & Logan, 2007; Shimoyama et al., 2008; Shin et al., 2006; Wang & Han, 2009; Wilkinson, 2000; Zhuang & Zhou, 2009; Zhuang et al., 2012), however much less attention has been paid on the cascade element of multiple MFC units in a stack operating under continuous flow conditions (Aelterman 2006; Ieropoulos 2008, Galvez 2009, Pinto 2010, Winfield 2011) and how the losses from direct fluidic paths oppose electron flow (Ieropoulos 2008, Zhuang & Zou 2009; Shuang et al 2012). The aim of this paper is therefore to address the issues of fluidic conductance in a continuous flow cascade system and demonstrate improved operation both in terms of power output as well as treatment efficiency.

Complexity is further increased when considering the peripherals required for the operation of MFC stacks (Ieropoulos et al., 2009). According to a life-cycle evaluation of MFCs, which has found that fabrication costs should be brought to a minimum (i.e. no expensive catalysts) (Pant et al., 2011), it is clear that the inclusion of additional electronic components in the operation e.g. DC-DC converters(with their inherent losses) can only increase system costs and thus reduce the competitiveness of the technology. Accordingly, the most inexpensive way to achieve working voltages and currents is through a combination of series and parallel configurations. Yet this is not without problems, with cell-voltage reversal being one of the most commonly reported phenomena for stacks in a series configuration. Probably observed for the first time in MFCs by Aelterman et al.(2006), this incident was then further investigated by Oh & Logan (2007), with unbalanced substrate concentrations and starvation reportedly being the main causes of reversal. Further investigation of MFC stacks in series has since been largely neglected with a few exceptions (Dekker et al., 2009; Ieropoulos et al., 2008; Kim et al., 2012; Pinto et al., 2010). During this time the consensus has been that series connection, under continuous flow, almost always implies cell voltage reversal (Kim et al., 2011; Lefebvre et al., 2011), despite the fact that in the initial studies, this was not the case (Ieropoulos et al., 2008; Oh & Logan, 2007). Since wastewater treatment systems operate in continuous or semi-continuous mode, perhaps there should be no hesitation in using, at least partially, series connections within the stacks to increase voltage to usable levels. Moreover, Pinto et al.(2010) have demonstrated that MFCs with anolyte staging connected fluidically in series, are capable of larger treatment capacity than the same cells in (fluidic) parallel configuration. Enhanced treatment is of particular importance, especially since MFCs might find their first commercial application in wastewater treatment, as currently the market value of the electricity produced at low performance levels is considerably lower than the added value of the environmental service of treating the waste (Fornero et al., 2010).

The present investigation follows on from all these observations and previously demonstrated stable stack operation (Ieropoulos et al., 2010b) and proposes a simplified, inexpensive solution to stacking for maximising treatment without parasitic cross-conduction or voltage reversal losses.

Materials andMethods

  1. Construction of the MFC-Cascade system

A cascade-type MFC system was modelled using Solid Edge v20.0 UGS PLM Software (Siemens, Germany). The design incorporated both an upflow siphon-type tubing system (embedded) that allowed continuous flow whist maintaining a constant internal electrolyte volume, in addition to a dripping mechanism that created a fluidic isolation and thus electrical insulation between MFC units of the cascade. In such a system, the fluid flow was mostly gravity driven, with the need to pump only to the fuel cell at the top.

The MFC units were also designed to allow for continuous flow of nutrients even in the event of channel blocking or the absence of cells in the cascade (for e.g. maintenance). The design was then analysed for structural reliability and mouldability with the ProtoQuote® analysis system (ProtoMold Injection Moulding Service, Shropshire, UK) and modified to guarantee that the design could be fabricated by Rapid prototyping techniques as well as Injection Moulding, since the latter can dramatically reduce the costs of fabrication. Once finalised, the design was used to make, by means of Fused Deposition Modelling (FDM Titan, Laserlines, Bedford, UK), packs of 3 identical 2-chamber MFCs (each triplet in a containing unit, see Fig. 1) with 6mL internal volume for each half-cell in ABS (Acrylonitrile Butadiene Styrene) plastic. The electrodes, both anode and cathode, were made of 180 cm2 of 20 g.m-2 plain carbon fibre veil (PRF Composites, Dorset, UK) without any added precious metals or catalysts; the current collector for each electrode was made of plain NiCr wire (6 cm long, ø0.45 mm; SWC, Essex, UK). A cation-exchange-membrane with a surface area of 6 cm2 (VWR, Leicestershire, UK) was placed between the two chambers. The original inoculum was activated sewage sludge provided by the local water utility company (Wessex Water Services Ltd). Two single MFCs with identical dimensions were set up at the same time and operated independently as controls.

Taking advantage of the precision in the fabrication technique, along with the predictable hygroscopic swelling of the parts (avg. +0.399% according to the manufacturer), the MFCs were watertight without the need for screws, fixtures or sealants thus reducing assembly time, complexity, cost and weight of the stacks. Consequently, the utilised design conveyed highly practical “plug-and-play” modularity, whereby each MFC could be easily added to a stack or removed for e.g. maintenance, both in a matter of seconds and without affecting the downstream flow for the rest of the stack.

  1. Operation

The stacks and control MFCs were continuously supplied with sterile 5mM Acetate TYE anolyte (5mM C2H3NaO2, 0.1% w/v Tryptone, 0.05% w/v Yeast Extract; pH 6.8 unbuffered, 680 ± 20 μS.cm-1) at 120 µL.min-1 by a multi-channel peristaltic pump (Watson-Marlow Pumps Group, Cornwall, UK), equivalent to an organic loading rate (OLR) of 8.366 kgCOD.m3.day-1 and a hydraulic retention time (HRT) of 0.63h (per MFC in the cascade). In additional experiments, the anolyte was replaced by an artificial wastewater solution as formulated by Aldrovandi et al. (2009), without buffering. The catholyte was fresh tap water in all cases.

The external load (Rext) connected to each MFC or Stack was varied over time in an attempt to achieve impedance-matching (Rext = Rint), which according to Jacobi’s law, is the condition under which maximum power is transferred to the external load, a.k.a the Maximum Power Transfer (MPT) point. By repeatedly testing the MFCs/Stacks using polarisation — standard technique with a range of external resistors (30kΩ to 50Ω) and 5min steps with a variable resistor box (ELC, Annecy, France) — it was possible to observe which Rext values produced MPT. These same values (which decreased over time, indicating a reduction of Rint as previously described by Ieropoulos et al. 2010b) were then utilised as the load in the normal operation of the cells, leading to continuous MPT over time.

Additional experiments involved the connection of the MFCs in series and parallel configurations; the resulting stacks were tested for stability by different cycles of starvation brought about by stopping and re-starting the anolyte pump for different periods of time. All tests were carried out at ambient temperature in an attempt to simulate a pragmatic natural environment for a period of 14 months.

Polarisation and power curves for stacks and single units were generated under the same continuous flow by connecting the MFCs to load values ranging from 50000 to 5Ω (at 3min intervals, found sufficient to allow the MFCs to reach quasi-steady-state for each resistance value).

  1. Data collection and analysis

3. 1 Electrical output

Electrical output was measured in real time in the form of voltage with a PicoLog ADC-16 interface (Pico Technology, Cambridgeshire, UK) connected to a computer. Current production was calculated using Ohm’s law by which the current I = V/R , where V is the measured voltage and R is the load. Electrical power production was determined using the derivation of Joule’s law where the power P = V x I. Power and current densities were normalised based on total anodic compartment volume (6cm3, multiplied by the number of MFCs when stacked).

3. 2 COD measurements

The total COD of samples was determined according to the potassium dichromate oxidation method with a kit from VWR (COD MR test system, Lutterworth, UK) and using a compatible photometer (Aquagem, Jenway, Chelmsford, UK) and interference filter module (COD MR IM, Jenway, Chelmsford, UK). Samples were collected at the outlet port of each MFC and compared with the supplied sterile input anolyte (Acetate TYE or Artificial wastewater).

Results and Discussion

  1. Individual operation

Originally, a single triplet (3MFCs; see yellow triplet in Fig.1) and two single reference MFCs were inoculated and operated until a steady-state (based on voltage output) was observed for a period of at least 7 days; this took approximately 3 weeks in the case of the triplet. It was soon realised that 3MFCs were not sufficient for full nutrient utilisation/treatment (data not shown). Hence, a second triplet was placed immediately below thus making a cascade of 6 MFCs, but still fed only from the drip at the top of the cascade. However, the new triplet was not inoculated with activated sludge but just left to be colonised by the cells shedding off from the MFCs placed above them.

In less than 24h, a clear electrical potential rose in these newly added MFCs (data not shown). The system was tested continuously for a further 3 weeks until steady states were reached for all 6 MFCs. The power curves obtained at the end of this period (seeFig. 2) indicate a pattern, whereby the fuel cell placed at the top of the cascade (MFC1) produced the maximum power and current densities, and the fuel cells downstream produced lower values, in a linear relationship with their position in the cascade; a significant correlation was actually found between the maximum power density (MPD) points for each MFC (see insert table on Fig. 2) and the corresponding current (CCD) with regards to the actual positioning (linear regression: r2= 0.9296; p= 0.0019). Similarly, the cells presented differential open circuit voltages (OCVs) in linearproportionality (r2= 0.9472; p= 0.0011), with MFC1’s OCV = 584.76 mV and 576.70, 567.22, 528.47, 511.97 and 503.32 mV respectively for MFCs 2-6. Nevertheless, MPT for all cells was found to occur at very similar values (avg. 3.27 ± 0.21 kΩ) for all cells, indicating near-identical Rint. The latter could be expected because all MFCs were of identical in construction, so the differences in OCV can be explained by the cumulative degradation of the nutrient supply, brought about by feeding the stack only at the top cell; as the anolyte flows from one cell to the other, electrons are extracted by the bacteria, thus changing the anolyte’s redox potential with direct consequences for the anodic potential of the cells. Furthermore, but to a lesser extent, it is also possible that a cumulative reduction in the catholyte’s dissolved oxygen also contributed to this effect.

  1. Stacked operation

2.1Parallel connections: positive stacking-effect

After estimating the individual performance of each cell in the stack, the cells were connected in parallel and left to stabilise; as expected the OCVs for both the stack and the individual MFCs were very similar (ca. 560mV). After a week of stacked operation, a polarisation sweep was performed, with the results shown on Table 1a. As can be seen,the 6MFC stack was capable of producing approx. 7 times more power than individual MFCs, implying a superior power density from the pack. In a similar fashion, the MCP was 8.33-fold larger than the single units’ production. After another ~8 months of operation in parallel (seeTable 1b), both individual and stacked systems exhibited improvement, but the superior performance of the stack was maintained with respect to the reference cells (7.46 W m-3and 91.11 A m-3 vs 6.83 W m-3 and 85 A m-3), implying that this effect is permanently maintained throughout biofilm development stages.The benefits of stacking – when this is done optimally – are thought to be a consequence of impedance matching (see above), whereby the plurality of units tends to uniformity in terms of Rint. If the majority of the cells perform well then the negative effects of the underperforming minority are compensated for, and in the longer term their performance is “assisted” by, the stronger majority. If the opposite case occurs, then the overall performance is brought down (Ieropoulos et al., 2010c). The importance of Rint balancing is further explored in section 2.5.

2.2Parallel connections: cumulative COD reduction to high standards

2.2.1COD reduction for Acetate-based anolyte

The COD-reduction was measured both for the individual reference cells and for each cell of the stack operating in parallel, with the latter being the preferred configuration due to higher currents obtained. The ΔCOD levels achieved were based on the COD of the sterile Acetate TYE anolyte (1982.5 ± 2.12 mg L-1). Determinations were performed until the COD-reduction values became stable (approx. 3 months after connection in parallel; see Fig. 3“5mM Acetate + TYE 1”). A second set of determinations were performed one month later (see Fig. 3“5mM Acetate + TYE 2”), but no significant difference was found. The average ΔCOD for the top-placed MFC1 (cascade) was 591 mg L-1, whereas the ΔCOD for a non-stacked MFC was on average 542 mgCOD L-1. The difference observed can be explained by the positive stacking-effect previously discussed: operation at higher current densities means enhanced abstraction of electrons from the carbon source, naturally translating into better COD reduction rates than in individual units.

The average ΔCOD for the 6MFC-cascade stack was found to be 1753.5 mg L-1or 88.45%, producing an effluent of 229 ± 8.48 mgCOD L-1. The latter, although favourable, was nevertheless of insufficient quality, particularly with reference to the European Union urban wastewater treatment directive (EU WWTD), which requires COD reduction to a minimum of 125 mgCOD L-1(EEC, 1991). As ΔCOD was found to be cumulative in the cascade system, it became clear that adding further units below the stacked system would assist in reaching acceptable levels. However, due to the impracticality of setting up a second 6-MFC stack — with comparable levels of biofilm maturity — within a reasonable timeframe, the addition of a second 6MFC stack immediately below was not immediately feasible. Recirculation of the effluent was also not suitable due to the long HRTs employed (degradation of the substrate by planktonic bacteria would have continued regardless, so the feedstock would have been inconsistent, with continuously decreasing COD over time). Further degradation capabilities were thus ‘simulated’ i.e. estimated by switching the feedstock, at the top of the cascade, to a concentration of ~230 mgCOD L-1 (obtained by the dilution of 116mL of 5mM Acetate TYE into 1L of distilled water). This may have not been the most ideal simulation scenario, since it eliminated the bacterial component from the feed, however it was one practical way of guaranteeing constant COD feeding for prolonged periods of time at the levels exhibited by the original stack’s effluent. The combined results of COD degradation for the 6MFCs at ~2 gCOD.L-1 and then 6MFCs at ~230 mgCOD L-1are presented inFig. 3.