Bioelectrochemistry 106 (2015) 105–114
Contents lists available at ScienceDirect
Bioelectrochemistry journal homepage: www.elsevier.com/locate/bioelechem
Chemometrical assessment of the electrical parameters obtained by long-term operating freshwater sediment microbial fuel cells Mario Mitov a,⁎, Ivo Bardarov a, Petko Mandjukov a, Yolina Hubenova b a b
Department of Chemistry, South-West University “Neofit Rilski”, 66 Ivan Mihajlov Str., 2700 Blagoevgrad, Bulgaria Department of Biochemistry and Microbiology, “Paisii Hilendarski” University of Plovdiv, 24 Tzar Asen Str., 4000 Plovdiv, Bulgaria
a r t i c l e
i n f o
Article history: Received 31 October 2014 Received in revised form 7 May 2015 Accepted 22 May 2015 Available online 3 June 2015 Keywords: Sediment microbial fuel cells Freshwater sediments Statistics Sustainability Autonomous power generation
a b s t r a c t The electrical parameters of nine freshwater sediment microbial fuel cells (SMFCs) were monitored for a period of over 20 months. The developed SMFCs, divided into three groups, were started up and continuously operated under different constant loads (100, 510 and 1100 Ω) for 2.5 months. At this stage of the experiment, the highest power density values, reaching 1.2 ± 0.2 mW/m2, were achieved by the SMFCs loaded with 510 Ω. The maximum power obtained at periodical polarization during the rest period, however, ranged between 26.2 ± 2.8 and 35.3 ± 2.8 mW/m2, strongly depending on the internal cell resistance. The statistical evaluation of data derived from the polarization curves shows that after 300 days of operation all examined SMFCs reached a steady-state and the system might be assumed as homoscedastic. The estimated values of standard and expanded uncertainties of the electric parameters indicate a high repeatability and reproducibility of the SMFCs' performance. Results obtained in subsequent discharge–recovery cycles reveal the opportunity for practical application of studied SMFCs as autonomous power sources. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Organic-rich freshwater and marine sediments can be considered as an abundant source of renewable energy. In 2001, Reimers et al. [1] introduced the concept for harvesting energy from marine sediment– water interface, based on the principles of microbial fuel cells (MFCs). The sediment microbial fuel cells (SMFCs) mimic the natural processes in the sediments, where bacteria, capable of transferring electrons extracellularly (called exoelectrogenic bacteria), couple the degradation of the sedimentary organic matter with the reduction of available metal oxides like Fe2O3 and MnO2 or sulfates as final electron acceptors. In the same manner, the exoelectrogenic bacteria transfer electrons to the SMFC-anode, buried in the anoxic sediment layer, using it as an alternative electron acceptor. The decreasing oxygen concentration gradient over the depth of water and sediment columns creates the necessary potential difference, which drives the electrons through an external electric circuit to the cathode and thus an electric current is generated. This simplifies significantly the SMFC construction, eliminating the necessity of purging the anodic compartment with inert gas and the use of a cation-exchange membrane to separate both compartments, and reduces the operational costs in comparison to the other MFC types. Numerous papers and reviews provide detailed information ⁎ Corresponding author at: Innovative Center for Eco Energy Technologies at SouthWest University, Blagoevgrad, 66 Ivan Mihajlov Str., 2700 Blagoevgrad, Bulgaria. E-mail addresses:
[email protected] (M. Mitov),
[email protected] (Y. Hubenova).
http://dx.doi.org/10.1016/j.bioelechem.2015.05.017 1567-5394/© 2015 Elsevier B.V. All rights reserved.
about identified exoelectrogenic bacteria [2–9], possible reactions [4, 10–12] and used electrodes and constructions [13–16] of SMFCs. Most SMFCs have been explored in marine environments, but recently the freshwater SMFCs have attracted an increasing attention [17–19]. Because of the high salt concentration, seawater possesses a much higher electrical conductivity than river water (∼ 50,000 vs. ∼ 500 μS/cm at 20 °C). For this reason, seawater SMFCs normally produce greater electric power than river-water ones due to the lower electrolyte resistance. In addition, the salt water enhances “virtual” corrosion at the cathode, which is beneficial for the cathode performance [20]. Typical maximum power densities obtained with SMFCs vary in the range of 10–20 mW/m2 [12], however, both Tender et al. [2] and Reimers et al. [21] have reported a maximum power density higher than 30 mW/m2 in marine environments. Based on long-term experiments in different environments, potential applications of SMFCs as sustainable power sources for small electrical devices, operating in remote areas, for enhanced biodegradation of excess organic matter and contaminants in aquatic sediments and wetlands have been demonstrated and proposed [10,22–25]. The main drawback for the wide practical application of SMFCs (as well as the other MFCs types) is their low power generation and output voltage. Studies toward finding appropriate approaches for scaling-up the technology have demonstrated that the output voltage cannot be increased by stacking multiple SMFCs in series because all electrodes are ionically connected in the same electrolyte solution [23]. Moreover, the performance of the stack is limited by the worst performing unit because of the voltage reversal [26,27]. On the other hand, the parallel connection
106
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
of SMFCs is equivalent to enlarging the electrode surface area. However, the amount of energy generated by MFCs is not a linear function of their size, thus the power density decreases with the increased electrode surface area [28,29]. Recently, Ewing et al. [30] proposed a promising approach, demonstrating that SMFC power can be scaled up electronically by using individually operated SMFCs connected to a power management system, which electrically isolates the anodes and cathodes. These findings put an attention on the importance of the individual performance of SMFCs, constructed and operated at identical conditions. It is important to be mentioned that most of the studies in the field provide data collected from few number of replicates (3, 2 or even 1), which restricts the correct evaluation of the system reliability. Moreover, the uncertainty evaluation of measurements is missing and only mean values with standard deviations are usually presented. When studying bioelectrochemical systems, a quite large dispersion of the investigated parameters is often observed. This is due to the fact that these systems are very complex, composed of many elements, each of which exerting a significant influence on the values of the generated electrical energy. The presence of a biological component (live microbial communities) additionally complicates the task for obtaining reproducible results. The main goal of the present study is to evaluate the stability and robustness of freshwater SMFCs and to estimate their reliability as autonomous power sources. For this purpose, nine identical SMFCs were constructed and operated for over 20 months. Taking into consideration that the resistance of the external circuit in an MFC directly influences the anode potential and the resultant bioavailability of anodic exoelectrogenic bacteria [31–33], the constructed SMFCs were initially operated under different constant loads for a period of 2.5 months in order to explore the effects of external resistances on their start-up and power output. The load resistance, by which highest power outputs were achieved during this period, was applied to all nine SMFCs in the subsequent long-term experiments. Various chemometrical approaches were applied to check the time needed for SMFC synchronization, temporal stability of the electric parameters and the uncertainties of the obtained values. 2. Materials and methods 2.1. Sample collection Both sediments and water necessary for construction of SMFCs were collected in February 2013 from river Struma in the south-west region of Bulgaria (GPS coordinates: 41.990354, 23.067501). The choice of the point for sample collection was based on the results obtained in previous 18 months of experiments, aiming at verification the possibility for long-term electricity generation of SMFCs by utilization of river sediments and soil. The higher electrical outputs achieved with the river sediments [34] determined their use in the present study. The sediment samples were collected from a depth of 10 to 15 cm. The deep black color and the specific odor of the sediments indicated the presence of sulfides. Prior introducing into the fuel cells, the sediments were cleaned off hard bodies as stones and branches, sieved on site using 1/2 mesh stainless steel sieve and gently homogenized to preserve the existing natural bacterial flora. The water, taken from the same place in the river, was stored in an airtight container. 2.2. Construction of SMFCs and experimental setup Nine identical SMFCs were constructed using cylindrical polyethylene vessels (height 23 cm, diameter 9 cm). The vessels were filled with sediment on the site of sampling within 15 min of the sample collection in order the preserve the integrity of the sediment as much as possible and minimize the effects of aeration by atmospheric air. The sediment layer was ca. 15 cm in height and had a wet weight of ca. 1200 g. Plane graphite disks (diameter 8 cm, thickness 1.5 cm; GES
Co., apparent density 1.68 g/cm3, porosity 24%, electrical resistance 6.0 μΩ·m) were used as both cathodes and anodes. Insolated copper wires (3 mm2) were used to connect the electrodes with the external electrical circuit. Conductive graphite epoxy was applied to create low-resistance connections between electrodes and the bare end of wires. Non-conductive epoxy was used to protect the electrical connections from water. The resistance of such prepared electrodes ranged between 0.1 and 0.3 Ω and did not exceed 1.0 Ω at the end of the experiment. The anodes were embedded in the sediment layer 5 cm above the bottom of the vessel. A 4 cm layer of collected water (ca. 200 ml) was poured above the sediments and the cathodes were situated few millimeters under the water surface. The distance between anodes and cathodes was 14 cm. Each of the fully assembled SMFCs had a total weight of ca. 1600 g. The constructed freshwater SMFCs have been operated autotrophically for more than 20 months and only river water was added periodically for compensation of the evaporation. All fuel cells were kept at identical temperature and indirect sunlight irradiation, without any treatment or modifications. The experiments were divided into two stages. At the first stage, the nine SMFCs, further denoted as SMFC n (where n = 1 ÷ 9 is the number of each cell), were separated in three groups, each consisting of three cells. The SMFCs of each group were loaded with different external resistors as follows: the first group (SMFC 1, SMFC 2 and SMFC 3) with 100 Ω, the second group (SMFC 4, SMFC 5 and SMFC 6) with 510 Ω and the third group (SMFC 7, SMFC 8 and SMFC 9) with 1100 Ω. The cells were kept under constant load for two and a half months and their terminal voltage was periodically measured in order to analyze the behavior under the different loads. After the initial stage of the experiment, the resistance loads were removed and all nine cells were left to recover. During the second period, the fuel cells operated in a repeatable mode of 15 h under constant load of 510 Ω, followed by 72 h of recovery. The voltage was measured using a data logger and the generated current was estimated applying the Ohm's law. Polarization curves were obtained after each recovery mode, varying the external resistance from 100 kΩ to 10 Ω by using resistance decade box. The load resistances were changed through 2 units at each decade. The cell voltage, U, was recorded using digital multimeter DMM2700 (Keithley Instruments Inc., US). The current was calculated according to equation I/mA = U/R and the current density according to j/(mA/m2) = I/S, where U/mV is the measured cell voltage, R/Ω — the external resistance, and S/m2 is the geometric area of the anode. The linear slope of the polarization curves was used for evaluation of the cell internal resistance, R int . The power density, P/(mW/m2), was calculated as a product of the current density and the cell voltage. The open circuit voltage, OCV, and the short circuit current, Isc, were measured before each polarization. Using the data obtained, the fill factor, FF, was estimated by the following equation: FF ¼ P max =OCV Isc :
ð1Þ
At the end of the long-term experiment, the SMFCs were run through several subsequent discharge–recovery cycles. Each cycle consisted of 10 min discharge by 510 Ω load resistor followed by 10 min recovery at open circuit. In parallel, a SMFC was directly connected to a fully discharged 2F ultracapacitor (Panasonic Electric Double Layer Capacitor) and the time needed for its charging to pre-determined voltage (0.5 V) was monitored. 3. Calculations An evaluation of the SMFC performance and sustainability was done by processing the collected data using different chemometrical approaches. In order to ensure a proper selection of methods providing reliable statistical data treatment, the distributions of the measured
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
MAD ¼ medianjðxi −medianðxÞÞj
Expanded uncertainty (half-width of the confidence interval for the most probable value) is evaluated as: U E ¼ kU c
ð5Þ
where k is a coverage factor (typically k = 2, corresponding to confidence level of 95%). 4. Results and discussion 4.1. Performance of the SMFCs under different constant loads Immediately after assembling, the SMFCs were started up in a constant load operation mode. The nine constructed SMFCs, divided randomly into three series, were loaded by switching on a constant resistance (100, 510 or 1100 Ω) in the external electric circuit in order to establish the influence of the load resistance on their further performance. The variation of the terminal voltage, recorded during the SMFC operation under permanent load, is presented in Fig. 1. In parallel with the measurement of cell voltage, the ambient temperature in the vicinity of the surveyed SMFCs was also monitored (Fig. 1b, inner graph).
a
100
80
U / mV
electrical parameters were checked by “Kernel density plot method” [35,36]. The Kernel density estimation is a standard non-parametric procedure to evaluate and present the probability density function of a random variable based on finite samples. The general concept of Kernel density plot is the replacement of each point by certain distribution (typically normal) with assigned dispersion. Combining the distributions for all points in the sample and scaling to unit area, an estimation of the population distribution can be obtained [36]. In order to evaluate similarity between cells in time, a robust paired sign test was applied (see Supplementary data, sub-section I.1.). Applying paired test, every element of the vector containing values of maximum power or OCV from different days for one SMFC is compared with the corresponding element from the similar vector for other cell. Since the studied system is complex and involving living organisms, some bias between responses of different cells naturally occurs. To eliminate the influence of such bias, initially normalization of the data by scaling to range between 0 and 1 was performed. The scaling procedure was done as described in Supplementary data, Section I.2. The normalized parameters for the cells with similar behavior should not be significantly different according to pairwise comparison. The significance of the temporal tendencies in the variations of studied parameters was examined by linear trend analysis (see Supplementary data, I.3.). Before the calculation of the slopes of the regression lines (electrical parameter vs. time) and evaluation of their significance, the homoscedasticity of every data set was checked using Hartley's test, combined with a robust Siegel–Tukey's test for comparison of the dispersions (see Supplementary data, I.4.). The estimation of the most probable values of studied electrical parameters and their corresponding uncertainties was done by means of the robust statistics. Median of the data set is one of the robust measures for most probable value of random data set. The spread of the data points around the most probable value was characterized by the median of absolute deviations (MAD)
107
60
40
ð2Þ 20
where xi is the measured value for the individual SMFC; median(x) is the median of all values from the measurement cycle. MAD is a robust estimation of the dispersion of the data set, however, it is not compatible with the standard deviation, i.e. it is not directly a standard uncertainty. In order to transform MAD in a standard uncertainty, normalization was done according to Eq. (3): MADN ¼ 1:4826MAD
0 0
10
20
30
40
60
70
80
t/d
b
100
ð3Þ
20
vffiffiffiffiffiffiffiffiffiffiffiffiffi u n X 1:25 u t UC ¼ u2i n i¼1
ð4Þ
80
10 5
U / mV
where MADN is the normalized median of absolute deviations. Factor 1.4826 provides compatibility of the robust estimation of the dispersion with the standard deviation and can be used as a standard uncertainty. Since the data is obtained within single measurement cycle, MADN is in fact the repeatability of the measured parameter. Reproducibility (as a standard combined uncertainty within a period of 405 days) was evaluated on the basis of uncertainties obtained in each single measurement cycle (repeatability data) [37]. Combined uncertainty was calculated according to Eq. (4):
T/°C
15
0 0
60
10
20
30
40
60
70
80
t/d
40 20 0 0
10
20
30
40
60
70
80
t/d where UC is the combined uncertainty determined in long period of time (reproducibility); ui is the uncertainty in single measurement cycle (repeatability), n is number of points and 1.25 is standardization factor. It should be noted that Eq. (4) is suitable in the case of robust evaluation of the uncertainties ui.
Fig. 1. a) Variation of the cell voltage at the first stage of the experiment, when the SMFCs were operated under constant load: squares — SMFC 2 (100 Ω); circles — SMFC 6 (510 Ω); triangles — SMFC 8 (1100 Ω); b) Performance of SMFC 4 and SMFC 5 loaded with the same resistance (510 Ω). Inner graph presents the changes of the ambient temperature during the experiment.
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
2.5 months operation under constant load, stable steady-state open circuit voltage values for all nine cells were achieved for 72 h. Then, all SMFCs were discharged for 15 h under the same load of 510 Ω, determined as optimal in respect to the power generation in the previous experiment. This value of the external resistance is also closest to the determined internal resistances at this stage of the experiment, which coincides with the theoretical prediction that maximum power outputs are obtained when the load resistance is equal to the internal resistance of the system. The described recovery–discharge cycle was repeated five times, after which the SMFCs were operated at open circuit and only polarization curves were periodically taken. The change of the OCV and the terminal voltage of selected SMFC of each group (based on their division for the initial stage of experiment) during recovery and discharge mode, respectively, is shown in Fig. 2. Excluding the first recovery, the OCV attained steady-state values (Fig. 2a), ranging between 520 and 600 mV for all tested samples (see Supplementary data, Fig. S2a), after about 20 h. At discharge, the terminal voltage steeply decreased in the first 20 min and reached stationary values after approximately 300 min (Fig. 2b and Suppl. Fig. S2b). No substantial differences between the values achieved by all explored SMFCs were observed, which shows that the different load resistances applied during the SMFC startup period have no substantial influence on their further performance. Polarization and power curves, obtained with the examined SMFCs, are shown in Fig. 3 and Suppl. Fig. S3. The achieved power density
800
a 600
OCV / mV
A gradual increase of the terminal voltage was observed for all tested SMFCs in the initial period after their startup. The voltage of each individual SMFC reached absolute maximum between the 8th and 14th days, after that it began to fall and fluctuated till the end of this experiment. Despite the differences between the readings, some general trends in the voltage changes can be distinguished. As might be expected, the highest voltage values were achieved for the SMFCs loaded with 1100 Ω and the lowest — for those with 100 Ω resistance. The magnitude of the generated current, however, increased inversely proportional to the value of the load resistance, while the maximum power values were reached with 510 Ω load (Supplementary data, Fig. S1b,c). Comparable electric outputs were obtained for the samples loaded with the same resistance (Fig. 1b). Despite the lowest temperatures at the beginning of the experiment, the greatest increase in the electric outputs was observed exactly during this period. In order to check the possible temperature dependence of the measured electrical parameters, correlation analysis of power vs. temperature at the moment of measurement was performed. The results (R2 within range from 0.01 to 0.25) clearly indicate that the temperature effect cannot be assumed as significant. In this stage the SMFC behavior seems to be controlled by other factors possibly related to the adaptation of the microbial communities and especially of the exoelectrogenic strains to the new conditions in the bioelectrochemical system and their re-distribution. Taking into account that the explored SMFCs operated autotrophically without addition of any substrates, the initial rapid increase of voltage can be attributed to the presence of compounds in the organic-rich river sediments that are easily oxidized by electro-active bacteria. The typical sediments contain 0.4–2.2% of organic matter [38] and the power density could be enhanced in sediments with higher organic contents [1,39]. When the organic substances were used up, the power outputs began to decrease. Meanwhile, a re-distribution of the indigenous microbial communities, documented by appearance of different colored spots along the height of the sediment layer (Supplementary data, Fig. S5), took place in the system. Similar to the classical Winogradsky column [40], this rearrangement is driven by the existing oxygen and sulfide concentration gradients in the sediment depth. These two gradients, acting in opposite directions, create a continuous range of habitats selective for a variety of microorganisms. The availability of appropriate nutrients and terminal electron acceptors (ferric ions, sulfates, nitrates, oxygen) for each microbial phylotype as well as the synergistic interaction between photosynthetic microorganisms and heterotrophic bacteria [41,42] determine the specific colonization of the column. Typically, the lower anaerobic parts of the column are colonized by sulfurreducing bacteria as Clostridium and Desulfovibrio [40], green and purple sulfur bacteria, while the aerobic surface of the column is occupied by photoautotrophic oxygenic cyanobacteria. The switching on of an external resistance(s) in SMFC, however, suggests an additional option to the exoelectrogenic bacteria to colonize and use the anoxic anode as an alternative electron acceptor. In fact, the current generated by the examined SMFCs reflects namely the metabolic activity of the existing exoelectrogenic strains and is proportional to the extracellular electron flux, transferred to the anode. The degree to which the direct and mediated mechanisms of extracellular electron transfer (EET) are differentially responsible for power production in SMFCs, as well as the extent to which these mechanisms are found within microbial communities, remains unknown. In more general aspect, the observed fluctuations in the recorded electric outputs represent the dynamics of the complex processes during autotrophic evolution of the microbial community in SMFC.
400
200
0 0
To evaluate the effect of different loads on the development and further performance of explored BESs, at day 73 the external resistors were disconnected and the SMFCs were left to recover at open circuit. After
2000
3000
4000
5000
800
b 600
400
200
0 0
4.2. Performance of the SMFCs at periodical polarization
1000
t / min
U / mV
108
100
200
300
400
500
600
700
800
t / min Fig. 2. a) Recovery of the OCV after switching off the load resistance; b) Change of the SMFC terminal voltage during discharge under load of 510 Ω: squares — SMFC 2, circles — SMFC 6, triangles — SMFC 8.
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
1000
30
109
1000
a 25
15 400 10
-2
OCV / mV
U / mV
20 600
900 P / mW m
800
800 700 600
200
5
0
500
0 0
20
40
60
80
100
120
400
-2
j / mA m
100
200
300
Fig. 3. Polarization and power curves obtained with SMFC 2 (squares), SMFC 6 (circles) and SMFC 8 (triangles) at day 561.
550
555
560
555
560
555
560
t/d 40
b
-2
30
Pmax / mW m
values are comparable with those reported by other authors [2,42], when the SMFCs were not additionally fed with organic substances. Zhao et al. [42] reported that the maximum power density increased from 15 mW/m2 to 43 mW/m2 and 57 mW/m2 when the SMFC was fed with glucose and acetate, respectively, which demonstrates the importance of the nutrient availability for the power generation. The variations of OCV, maximal power density and internal resistances, derived from the polarization and corresponding power curves, over time of the long-term experiment are presented in Fig. 4 and Suppl. Fig. S4. Despite the differences in the electrical output values obtained with the individual SMFCs, common tendencies in their development are observed. These tendencies are statistically evaluated and commented in details in Section 4.3.
10
0
4.3. Statistical evaluation
4.3.1. System stability evaluation The first question to be solved was to check the type of distribution of the data obtained from different SMFCs at one measurement procedure. In order to check the distributions and their changes during the observation period, the Kernel density plot method [36] was employed. Results for OCV and maximum power density are presented in Fig. 5. Both parameters represented are showing asymmetric distributions with clear bimodal structure at earlier phases of the experiment. Since these are the basic electrical parameters, similar distributions are typical for all parameters studied. The obtained density plots clearly indicate that distributions of the parameter cannot be assumed as normal ones.
100
200
300
550
t /d 700
c 600
Rint / Ω
Considering potential practical application of the SMFCs, evaluation of their stability and robustness as well as of the uncertainties of electrical parameters becomes crucially important. Participation of the living organisms in the process makes statistical treatment a challenging task. Response of the microorganisms is a typical example for deviations from the normal distribution. In such a case, the application of robust statistical methods could be the reasonable solution. The experiments were carried out using nine SMFCs working in a similar environment. The duration of the electrical parameters measurement procedure is practically disparagingly low in comparison with the observation period. Thus, the proper estimation of the dispersion between cells within single measurement procedure might be assumed as a combined uncertainty of type A [43,44] taking into account all possible uncertainty sources. Using the above noted chemometrical methods, the data for five parameters — OCV, maximum power density (Pmax), short circuit current density (jsc), internal cell resistance (Rint.) and fill factor (FF), was analyzed.
20
500
400
300 100
200
300
550
t /d Fig. 4. Variation of: a) OCV; b) maximum power density; c) internal resistance of SMFC 3 (squares), SMFC 6 (circles) and SMFC 9 (triangles) over time.
In such a case, standardly applied Gaussian statistics becomes unreliable and thus, robust (distribution free) statistics should be applied for parameter evaluation and further data treatment. It should be also noted that application of Gaussian statistics provides reliable results only in the case of normally distributed data. In difference, the robust methods
110
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
0,024
concluding that limited number of cells can be accepted as similar, comparing the changes of the studied parameters in time. However, after elimination of the first 5 points, all differences between SMFC parameters become insignificant. Thus, after approximately 150 days all cells are entering a stable phase of their development and start to respond in a similar way, keeping some difference between absolute values of the measured parameters. The data for this phase can be used for evaluation of uncertainty for the electric parameters of SMFCs.
a
Relative frequency
0,020 0,016 0,012 0,008 0,004 0,000 -200
0
200
400
600
800
1000
1200
OCV / mV 0,012
b
Relative frequency
0,010 0,008 0,006 0,004 0,002 0,000 -10
0
10
20
Pmax / mW m
30
40
50
-2
Fig. 5. Kernel density plots representing the distribution of: a) OCV; b) maximum power density data from all nine SMFCs for representative days (solid line — day 75; short dash line — day 88; short dash dot line — day 156; dash dot line — day 229; dash dot dot line — day 561) of the experimental period.
are reliable for practically all types of distributions, including the normal one. After a period of stabilization, the data distribution becomes more narrow and the most probable value is shifted. However, in all cases it remains asymmetric, indicating that application of robust statistics would be preferable. The initial scattered values are consequence of the discharge cycles under different loads and possibly consequent reorganization of the microbial community. The data for OCV and maximum power density obtained with tested SMFCs for a period of approximately twenty months is presented in Fig. 4a, b and Suppl. Fig. S4a,b. All cells are showing significant differences in the values of both parameters in time. There are two obviously different periods. The first one starts after the discharges under different constant load procedures and shows significant difference in responses of the different fuel cells. After a period of approximately 150 days the system reaches more stable condition and the further changes of the parameters of different SMFCs in time seems to be better synchronized. The cells start to respond in more similar and predictable way. From a practical point of view, it is important to evaluate the time needed for such stabilization as well as to check if some trends exist in the stable state. The paired sign test was used as an objective criterion to evaluate the time needed for synchronization of SMFC behavior (see Supplementary data, Section I.1.). The test was initially applied to the entire data set including values measured at different time points. The results allow
4.3.2. Uncertainty evaluation Evaluation of uncertainty is an important question related to the potential practical applications of the SMFCs. In the case of homoscedasticity of the data set, the most probable values of the electric parameters studied, eventual trend in time and reproducibility (intermediate precision) based on long-time functioning of the fuel cells would be possible to be evaluated. The uncertainty, evaluated in this way, might be attributed to all subsequent measurements of the corresponding electric parameters, unless dramatic changes in the SMFCs status occur. 4.3.2.1. Evaluation of uncertainty from single measurement cycle (repeatability). As could be concluded from Fig. 5, there are no reasons the hypothesis for normal distribution of the measured parameters to be accepted. Thus, the most probable value and corresponding uncertainty of the data from each single measurement cycle were evaluated by means of robust statistics. The median of the data set was selected as a measure for the most probable value and the normalized median of absolute deviations (MADN) was accepted as an estimation of the standard combined uncertainty. The expanded uncertainty was calculated using coverage factor k = 2, corresponding to confidence level of 95%. The results obtained for all five parameters (OCV, Pmax, jsc, Rint. and FF) in the stable period of SMFCs are presented in Fig. 6. The estimated values of the median and MADN were compared with those of robust mean and robust standard deviation, calculated according to Huber's H15 algorithm (Algorithm A [37,45]), and were found to be identical. Homoscedasticity of the system was checked comparing the highest and the lowest values for repeatability obtained. The concept is based on widely used Hartley's homoscedasticity test, however, combined with a robust Siegel–Tukey's test for comparison of the dispersions (see Supplementary data, Section I.4.). For all studied parameters the difference in dispersions was found to be insignificant at confidence level of 0.95, i.e. the system might be assumed as homoscedastic. 4.3.2.2. Reproducibility evaluation. Reproducibility was evaluated on the basis of all single measurement cycles with corresponding repeatability data according to Eq. (4). The obtained results for all parameters are presented in Table 1. After the synchronization of the SMFC performance, counted from the measurements at day 156, a clear significant ascending trend (0.264 mV/day; R2 = 0.9754) of the OCV values is observed (Fig. 6a). The high correlation coefficient, combined with rather low uncertainties in all measurement cycles allows suggesting OCV as the most stable electric parameter of the SMFCs. After 20 months of operation, the OCV for all examined SMFCs reached stable values ranging between 800 and 900 mV (see also Suppl. Fig S4a). In contrary, the most unstable parameter is the internal resistance Rint. A clear maximum in the Rint graphics (Figs. 4c and 6d) is observed at day 300, after which a steady decrease of resistance (trend of − 0.1818 Ω/day; R2 = 0.7883) takes place. The lower, but significant, correlation coefficient can be explained by a low number of points used for trend evaluation, on the one hand, and the quite “noisy” data on the other one. Except OCV, all other electrical parameters are strongly influenced by the variations of internal resistance. As could be expected, the maximum power (Fig. 6b) and short circuit current (Fig. 6c) show opposite changes compared to those of Rint, as well as high uncertainty
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
111
50
1000
b
a 40
Pmax / mW m
OCV / mV
-2
800
600
400
30
20
10
200
0 100
200
300
550
555
560
0 100
565
200
300
c
800
80
560
565
555
560
565
d
400
200
40
0 100
555
600
120
Rint / Ω
jsc / mA m
-2
160
550
t/d
t/d
200
300
550
555
560
0 100
565
200
300
550
t/d
t/d 0,4
e
FF
0,3
0,2
0,1
0,0 100
200
300
550
555
560
565
t/d Fig. 6. Most probable values and corresponding uncertainties of the data from each single measurement cycle: a) OCV; b) maximum power density; c) short circuit current density; d) internal cell resistance; e) fill factor.
level at all points. Stability of the OCV can be explained with the relative independence of this parameter from Rint. Possibly, the observed maximum in the Rint and corresponding minimum in Pmax and jsc values indicate another important point in the SMFC development. As far as the resistance of the sediment determines
in the highest extent the internal resistance of the SMFCs [46], it could be considered that the mass transport of available biodegradable organic substances to the anode and of reaction products away from the anode [47] is the main limiting factor in these bioelectrochemical systems. In the closed, self-contained recycling ecosystem of SMFC the
Table 1 Reproducibility data for electrical parameters studied: open circuit voltage (OCV), maximum power density (Pmax), short circuit current density (jsc), internal resistance (Rint) and fill factor (FF). Parameter
Combined uncertainty
Value at the first point — day 156 (k = 2)
Value at the last point — day 561 (k = 2)
OCV/mV Pmax/mW m−2 jsc/mA m−2 Rint/Ω FF
11 1.4 3.5 24 0.0041
805 ± 22 35.3 ± 2.8 142.3 ± 7.0 366 ± 48 0.3014 ± 0.0082
918 ± 22 26.2 ± 2.8 108.3 ± 7.0 597 ± 48 0.2780 ± 0.0082
112
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
3,5
1000
3,0
800 2,5
U / mV
Q/C
2,0 1,5
600
400
1,0
200
0,5 0,0
0 0
50
100
150
200
250
300
0
20
40
t / min
60
80
100
120
t / min
Fig. 7. The quantity of electricity, Q, derived from SMFC 2, SMFC 6 and SMFC 8 during discharge at 510 Ω load.
Fig. 8. Variation of voltage at subsequent cycles discharge–recovery, applied to SMFC 2 and SMFC 6 on day 615 after their start-up.
adaptation of microorganisms to the gradient conditions includes their ability for a full value assimilation of the available substrates and their circulation. Each waste product secreted by one bacterial type is consumed by the appropriate bacteria from other type (for example H2S released by the sulfate reducers is used by the anaerobic sulfur bacteria). Thus, the observed increase of the internal resistance before stabilization of the system (Fig. 4c and Suppl. Fig. S4c) is probably related to enhancing transport hindrances for delivery of appropriate substrates for exoelectrogenic bacteria in the anode vicinity and/or retarded mass transport of waste products away from the anode. The stable electrical outputs achieved after day 300 could be assigned to establishment of biological steady-state in the system (when the redistribution of the microbial community is accomplished and the rates of bacterial reproduction and death become similar) and equilibration of the mass transport rates of substances to and from the anode. The measured anodic potentials at the end the experimental window reached negative values of ca. − 480 mV (vs. Ag/AgCl), which additionally supported the role of the anodic bacterial biofilm to the electrical outputs. At the present, it is not clear if the time for synchronization of the cells and parameters' stabilization is influenced by the previous treatment of the cells. However, the low combined uncertainty (Table 1), evaluated on the basis of set of nine fuel cells over a period of ca. 400 days indicates significant stability making SMFCs a quite promising potential future energy source.
load, however, requires over 20 h for total recovery of the system (Fig. 2a), which is not acceptable for sustained application. For this reason, we explored the SMFCs' behavior at consecutive short cycles, consisting of 10 min discharge followed by 10 min recovery. The obtained results, plotted in Fig. 8, demonstrate a wellreproducible performance of the tested SMFCs. Moreover, the values of terminal voltage and OCV at the end of discharge and recover mode, respectively, have reached approximately the same levels after the third discharge–recovery cycle. In addition, the SMFCs were connected to 2F ultracapacitor for a direct charging test. The highest charging voltage that an ultracapacitor can achieve is determined by the voltage of the SMFCs [48]. In our case, the charging voltage could reach as high as 0.9 V, but it took a substantial time to get this value. Fig. 9 representing the charging progress shows that about 1.5 h was needed to charge the 2F ultracapacitor to 0.5 V. This time could be reduced if larger surface electrodes are used or several SMFCs are connected in parallel. Moreover, the explored SMFCs could directly supply various low power consumers. A toy with moving parts was chosen for such demonstration (see Video data). The original power source (a solar cell) was removed and the toy was directly supplied by a SMFC. Based on the results obtained, the potentials for practical application of the examined SMFCs are opened.
1,0 4.4. Potentials for practical application
U / mV
0,8 The data evaluation has shown that the explored SMFCs possess high durability and reliability at long-term autotrophic operation. In contrary to the traditional batteries, they are able to sustain stable electrical outputs under permanent load without exhaustion. Even after operation at short circuit for several hours, the tested SMFCs totally recover and at subsequent polarization they are capable to generate currents at the levels achieved in the previous periods. However, the steady-state voltage values under load are rather low, which has put the reasonable question about the possibilities for their practical application. Analyzing in details the discharge curves, presented in Fig. 2b, it was found that the time for establishment of a steady-state is about 5 h. The quantity of electricity, Q/C, derived from the SMFCs, was estimated by integration of the curves for different times of discharge and plotted in Fig. 7. The re-calculations show that 57.9 ± 2.5% of the electricity generated before attainment of steady state is obtained in the first 2 h of discharge. The longer operation (N10 h) under
0,6
0,4
0,2
0,0 0
20
40
60
t / min Fig. 9. Charging of a 2F ultracapacitor by SMFC 8.
80
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114
5. Conclusions Chemometrical assessment of SMFCs' electrical outputs is performed for the first time using robust statistic approach. The statistical evaluation of data (OCV, power and current density, internal resistance) collected from nine freshwater SMFCs, operating autotrophically for over twenty months, shows that their performance becomes homoscedastic after reaching a steady-state. The high repeatability and reproducibility of the evaluated parameters reveal the perspectives for development of sustainable autonomous power sources based on freshwater SMFCs. Acknowledgements This study was supported by the National Science Fund of the Ministry of Education and Science of Bulgaria through the contract E02/14/ 2014. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.bioelechem.2015.05.017. References [1] C.E. Reimers, L.M. Tender, S. Fertig, W. Wang, Harvesting energy from the marine sediment-water interface, Environ. Sci. Technol. 35 (2001) 192–195. [2] L.M. Tender, C.E. Reimers, H.A. Stecher III, D.E. Holmes, D.R. Bond, D.A. Lowy, K. Pilobello, S.J. Fertig, D.R. Lovley, Harnessing microbially generated power on the seafloor, Nat. Biotechnol. 20 (2002) 821–825. [3] D.R. Bond, D.E. Holmes, L.M. Tender, D.R. Lovley, Electrode-reducing microorganisms that harvest energy from marine sediments, Science 295 (2002) 483–485. [4] D.E. Holmes, D.R. Bond, R.A. O'Neil, C.E. Reimers, L.R. Tender, D.R. Lovley, Microbial communities associated with electrodes harvesting electricity from a variety of aquatic sediments, Microb. Ecol. 48 (2004) 178–190. [5] D.R. Bond, D.R. Lovley, Electricity production by Geobacter sulfurreducens attached to electrodes, Appl. Environ. Microbiol. 69 (2003) 1548–1555. [6] K. Venkateswaran, D.P. Moser, M.E. Dollhopf, D.P. Lies, D.A. Saffarini, B.J. MacGregor, D.B. Ringelberg, D.C. White, M. Nisxhijirna, H. Sano, J. Burghardt, E. Stackebrandt, K.H. Nealson, Polyphasic taxonomy of the genus Shewanella and description of Shewanella oneidensis sp. nov. Int. J. Syst. Bacteriol. 49 (1999) 705–724. [7] B.H. Kim, H.J. Kim, M.S. Hyun, D.S. Park, Direct electrode reaction of Fe(III) reducing bacterium, Shewanella putrefizciens, J. Microbiol. Biotechnol. 9 (1999) 127–131. [8] H.J. Kim, H.S. Park, M.S. Hyun, I.S. Chang, M. Kim, B.H. Kim, A mediator-less microbial fuel cell using a metal reducing bacterium, Shewanella putrefaciens, Enzyme Microb. Technol. 30 (2002) 145–152. [9] H.S. Park, B.H. Kim, H.S. Kim, H.J. Kim, G.T. Kim, M. Kim, I.S. Chang, Y.K. Park, H.I. Chang, A novel electrochemically active and Fe(III) reducing bacterium phylogenetically related to Clostridium butyricum isolated from a microbial fuel cell, Anaerobe 7 (2001) 297–306. [10] L. De Schamphelaire, P. Boeckx, N. Boon, K. Rabaey, W. Verstraete, Outlook for benefits of sediment microbial fuel cells with two bio-electrodes, J. Microbial. Biotechnol. 1 (2008) 446–462. [11] N. Ryckelynck, H.A. Stecher, C.E. Reimers, Understanding the anodic mechanism of a seafloor fuel cell: interactions between geochemistry and microbial activity, Biogeochem. 76 (2005) 113–139. [12] D.A. Lowy, L.M. Tender, J.G. Zeikus, D.H. Park, D.R. Lovley, Harvesting energy from the marine sediment-water interface II: kinetic activity of anode materials, Biosens. Bioelectron. 21 (2006) 2058–2063. [13] K. Scott, I. Cotlarciuc, D. Hall, J.B. Lakeman, D. Browning, Power from marine sediment fuel cells: the influence of anode material, J. Appl. Electrochem. 38 (2008) 1313–1319. [14] K. Scott, I. Cotlarciuc, I. Head, K.P. Katuri, D. Hall, J.B. Lakeman, D. Browning, Fuel cell power generation from marine sediments: investigation of cathode materials, J. Chem. Technol. Biotechnol. 83 (2008) 1244–1254. [15] C. Dumas, A. Mollica, D. Feron, R. Basseguy, L. Etcheverry, A. Bergel, Marine microbial fuel cell: use of stainless steel electrodes as anode and cathode materials, Electrochim. Acta 53 (2007) 468–473. [16] Z. He, H. Shao, L.T. Angenent, Increased power production from a sediment microbial fuel cell with a rotating cathode, Biosens. Bioelectron. 22 (2007) 3252–3255. [17] Y.-L. Zhou, Y. Yang, M. Chen, Z.-W. Zhao, H.-L. Jiang, To improve the performance of sediment microbial fuel cell through amending colloidal iron oxyhydroxide into freshwater sediments, Bioresour. Technol. 159 (2014) 232–239. [18] T.-S. Song, Z.-S. Yan, Z.-W. Zhao, H.-L. Jiang, Construction and operation of freshwater sediment microbial fuel cell for electricity generation, Bioprocess Biosyst. Eng. 34 (2011) 621–627.
113
[19] T. Song, W. Tan, X. Wu, C.C. Zhou, Effect of graphite felt and activated carbon fiber felt on performance of freshwater sediment microbial fuel cell, J. Chem. Technol. Biotechnol. 87 (2012) 1436–1440. [20] A. Bergel, D. Feron, A. Mollica, Catalysis of oxygen reduction in PEM fuel cell by seawater biofilm, Electrochem. Commun. 7 (2005) 900–904. [21] C.E. Reimers, P. Girguis, H.A. Stecher III, L.M. Tender, N. Ryckelynck, P. Whaling, Microbial fuel cell energy from an ocean cold seep, Geobiology 4 (2006) 123–136. [22] L.M. Tender, S.A. Gray, E. Groveman, D.A. Lowy, P. Kauffman, J. Melhado, R.C. Tyce, D. Flynn, R. Petrecca, J. Dobarro, The first demonstration of a microbial fuel cell as a viable power supply: powering a meteorological buoy, J. Power Sources 179 (2008) 571–575. [23] C. Donovan, A. Dewan, H. Peng, D. Heo, H. Beyenal, Power management system for a 2.5 W remote sensor powered by a sediment microbial fuel cell, J. Power Sources 196 (2011) 1171–1177. [24] Y. Yang, Z. Lu, X. Lin, C. Xia, G. Sun, Y. Lian, M. Xu, Enhancing the bioremediation by harvesting electricity from the heavily contaminated sediments, Bioresour. Technol. 179 (2015) 615–618. [25] A. Wang, H. Cheng, N. Ren, D. Cui, N. Lin, W. Wu, Sediment microbial fuel cell with floating biocathode for organic removal and energy recovery, Front. Environ. Sci. Eng. 6 (2012) 569–574. [26] S. Oh, B. Logan, Voltage reversal during microbial fuel cell stack operation, J. Power Sources 167 (2007) 11–17. [27] Y. Zhang, I, Angelidaki, self-stacked submersible microbial fuel cell (SSMFC) for improved remote power generation from lake sediments, Biosens. Bioelectron. 35 (2012) 265–270. [28] J.-D. Park, Z. Ren, Hysteresis-controller-based energy harvesting scheme for microbial fuel cells with parallel operation capability, IEEE T. Energy Conver. 27 (2012) 715–724. [29] L. Hsu, B. Chadwick, J. Kagan, R. Thacher, A. Wotawa-Bergena, K. Richter, Scale up considerations for sediment microbial fuel cells, RSC Adv. 3 (2013) 15947–15954. [30] T. Ewing, P.T. Ha, J.T. Babauta, N.T. Tang, D. Heo, H. Beyenal, Scale-up of sediment microbial fuel cells, J. Power Sources 272 (2014) 311–319. [31] Z. Ren, H. Yan, W. Wang, M.M. Mench, J.M. Regan, Characterization of microbial fuel cells at microbially and electrochemically meaningful time scales, Environ. Sci. Technol. 45 (2011) 2435–2441. [32] C.I. Torres, R. Krajmalnik-Brown, P. Parameswaran, A.K. Marcus, G. Wanger, Y.A. Gorby, B.E. Rittmann, Selecting anode-respiring bacteria based on anode potential: phylogenetic, electrochemical, and microscopic characterization, Environ. Sci. Technol. 43 (2009) 9519–9524. [33] D. Molognoni, S. Puig, M.D. Balaguer, A. Liberale, A.G. Capodaglio, A. Callegari, J. Colprim, Reducing start-up time and minimizing energy losses of microbial fuel cells using maximum power point tracking strategy, J. Power Sources 269 (2014) 403–411. [34] I. Bardarov, Y. Hubenova, M. Mitov, Sediment microbial fuel cell utilizing river sediments and soil, Bulg. Chem. Commun. 45A (2013) 223–226. [35] Representing data distributions with kernel density estimates, Analytical Methods Committee Technical Brief No. 4, Royal Society of Chemistry, 2006. (http://www. rsc.org/images/data-distributions-kernel-density-technical-brief-4_tcm18-214836. pdf). [36] B.W. Silverman, Density estimation for statistics and data analysis, Monographs on Statistics and Applied Probability, Chapman and Hall, London, 1986. [37] ISO 5725, Accuracy (trueness and precision) of measurement methods and results. [38] F. Rezaei, T.L. Richard, R.A. Brennan, B.E. Logan, Substrate-enhanced microbial fuel cells for improved remote power generation from sediment-based system, Environ. Sci. Technol. 41 (2007) 4053–4058. [39] S.P. Jung, M.-H. Yoon, S.-M. Lee, S.-E. Oh, H. Kang, J.-K. Yang, Power generation and anode bacterial community compositions of sediment fuel cells differing in anode materials and carbon sources, Int. J. Electrochem. Sci. 9 (2014) 315–326. [40] B. Rogan, M. Lemke, M. Levandowsky, T. Gorrell, Exploring the sulfur nutrient cycle using the Winogradsky column, Am. Biol. Teach. 67 (2005) 348–356. [41] Z. He, J.J. Kan, F. Mansfeld, L.T. Angenent, K.H. Nealson, Self-sustained phototrophic microbial fuel cells based on the synergistic cooperation between photosynthetic microorganisms and heterotrophic bacteria, Environ. Sci. Technol. 43 (2009) 1648–1654. [42] J. Zhao, X.-F. Li, Y.-P. Ren, X.-H. Wang, C. Jian, Electricity generation from Taihu lake cyanobacteria by sediment microbial fuel cells, J. Chem. Technol. Biotechnol. 87 (2012) 1567–1573. [43] International Vocabulary of Metrology — Basic and General Concepts and Associated Terms, Third Ed., Joint Committee on Guides for Metrology (JCGM), 2012. (http:// www.bipm.org/en/publications/guides/vim.html). [44] Evaluation of Measurement Data — Guide to the Expression of Uncertainty in Measurement (GUM), Joint Committee for Guides in Metrology (JCGM), 2008.(http:// www.bipm.org/en/publications/guides/#gum). [45] Robust Statistics: A Method of Coping With Outliers, Analytical Methods Committee Technical Brief No. 6, Royal Society of Chemistry, 2001. (http://www.rsc.org/images/robust-statistics-technical-brief-6_tcm18-214850.pdf). [46] A. Pietrelli, A. Micangeli, V. Ferrara, A. Raffi, Wireless sensor network powered by a terrestrial microbial fuel cell as a sustainable land monitoring energy system, Sustainability 6 (2014) 7263–7275. [47] M.E. Nielsen, C.E. Reimers, H.A. Stecher, Enhanced power from chambered benthic microbial fuel cells, Environ. Sci. Technol. 41 (2007) 7895–7900. [48] F. Zhang, L. Tian, Z. He, Powering a wireless temperature sensor using sediment microbial fuel cells with vertical arrangement of electrodes, J. Power Sources 196 (2011) 9568–9573.
114
M. Mitov et al. / Bioelectrochemistry 106 (2015) 105–114 Mario Mitov graduates as a chemical engineer in “Electrochemical productions and power sources” at the University of Chemical Engineering and Metallurgy, Sofia, Bulgaria, in 1985. He begins his professional carrier in the Department of Chemistry at South-West University, Blagoevgrad, Bulgaria in 1987. Prof. Mitov is a Head of the Innovative Center for Eco Energy Technologies at South-West University. Currently, Prof. Mitov delivers lectures on General and Inorganic Chemistry, Physicochemistry and Electrochemistry. His research interests are focused on characterization of nanomaterials as potential electrodes for rechargeable batteries and fuel cells and investigation of bioelectrochemical systems such as MFCs, MECs and DPPFCs.
Petko Mandjukov obtained his Magister's degree in Inorganic and Analytical chemistry in 1985. In 1996 defends his Ph.D. in analytical chemistry at the “St. Kliment Ohridski” University of Sofia. Since 2002 he is an Associated Professor in Analytical chemistry at the South-West UniversityBlagoevgrad. Since 2011, Assoc. Prof. Petko Mandjukov is an expert at the International Atomic Energy Agency (IAEA). His research interests are in the field of metrology in chemistry, chemometrics, statistics, analytical chemistry, atomic spectroscopy, environmental monitoring.
Ivo Bardarov is a 26 year old young researcher. He obtained Bachelor's degree in Physics from the South-West University of Bulgaria in 2012. In the following was granted with a Master's degree in Metrology in Chemistry by the same university. Currently, Ivo Bardarov is a PhD student at the SouthWest University. His research work is focused on the performance and improvement of different materials used as electrodes in Sediment Microbial Fuel Cells, especially graphite and nanoporous carbon.
Yolina Hubenova receives M.Sc. in “Biotechnology-Gene and cell engineering” at “St. Kliment Ohridski” University of Sofia. In 1999 became a specialist in “Medical biology”. In 2005 Yolina Hubenova is awarded by the University of Bonn with the degree Dr.rer.nat. in Neurobiochemistry. In 2013 she is granted with the degree “Doctor of Sciences” from the “St. Kliment Ohridski” University. Currently she is working at University of Plovdiv, Bulgaria. DSc. Hubenova delivers lectures on Ecological Biochemistry, Clinic Biochemistry and Protein Engineering. Her research interests are in the field of bioelectrochemical system development and investigation of the extracellular electron transfer in biofuel cells.