yeast ratio in yeast microbial fuel cell using response surface methodology approach

Journal of Power Sources 402 (2018) 402–412

Contents lists available at ScienceDirect

Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour

Optimization of glucose concentration and glucose/yeast ratio in yeast microbial fuel cell using response surface methodology approach

T

Marcelinus Christwardanaa,1, Domenico Frattinia,1, Grazia Accardob, Sung Pil Yoonb, Yongchai Kwona,∗ a b

Graduate School of Energy and Environment, Seoul National University of Science and Technology 232 Gongneung-ro, Nowon-gu, Seoul, 01811, Republic of Korea Fuel Cell Research Center, KIST - Korea Institute of Science and Technology, Hwarang-ro 14-gil 5, Seongbuk-gu, Seoul, 02792, Republic of Korea

H I GH L IG H T S

G R A P H I C A L A B S T R A C T

chamber yeast microbial fuel • Single cell with a novel anode is optimized. Surface Methodology was • Response used to tune 2 operational parameters. of experimental response with • Dataset or without methylene blue was created.

shape of response surface shows • The that a non-trivial optimum point exists.

optimal point was experimentally • The checked and error was only 3–5%.

A R T I C LE I N FO

A B S T R A C T

Keywords: ANOVA Methylene blue OCV Optimization Power density Yeast

In this work the influence of two practical parameters, i.e. glucose concentration and glucose/yeast ratio, on performance of yeast-based microbial fuel cells (yeast-MFC) is investigated. The novel carbon felt pretreated with polyethylenimine is adopted as anode in open-air single chamber yeast-MFCs. The combination of the two parameters is optimized using response surface methodology with statistical approach. The optional presence of methylene blue as mediator is also included for comparison. Experimental dataset is initially built as reference and 4 mathematical equations are derived to predict the response regarding open circuit voltage (OCV) and maximum power density (MPD). By varying glucose concentration and glucose/yeast ratio, computed response surfaces show different responses are obtained and an optimum point exists within the range investigated. Finally, the optimized combinations for yeast-MFCs with/without mediator are predicted and response is verified in real experiment. The model tends to slightly overestimate the response, but accuracy is within confident range for both OCV and MPD. In fact, MPD obtained for the optimized yeast-MFC without mediator is 340.9 mW m−2, 3.2% lower than model, while it is 374.4 mW m−2, 5% lower than model, for the case including mediator. The discrepancy of OCV prediction is below 3%, making the approach reliable.

1. Introduction Microbial fuel cell (MFC) is a bio-device of actual interest in many

application fields like energy recovery, biosynthesis of secondary nutrients, wastewater treatment, residual biomass valorization, medicine and environmental sensing [1]. They are essentially based on the

∗

Corresponding author. E-mail address: [email protected] (Y. Kwon). 1 These authors contributed equally to this work. https://doi.org/10.1016/j.jpowsour.2018.09.068 Received 24 June 2018; Received in revised form 11 September 2018; Accepted 19 September 2018 0378-7753/ © 2018 Elsevier B.V. All rights reserved.

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

CF-PEI, methylene blue (MB) and yeast cells was studied [48]. The combination of yeast-MFC, CF-PEI and RSM is studied for the first time. The influence of two major parameters on the performance response of a single chamber open-air yeast-MFC is experimentally determined and then modeled according to RSM to optimize the parameters and getting the highest possible performance within a suitable range for real conditions. More than two parameters could be studied in RSM but, to have a complete and comprehensive visual representation of the response surfaces, two is the optimal number of tunable parameters to study the response of each output. Glucose concentration and glucose to yeast ratio (G/Y) are considered as the tunable parameters, and open circuit voltage (OCV) and maximum power density (MPD) are the responses modeled for the yeast-MFC. Other parameters as pH and temperature can be considered as tunable parameters, but in real situations they act more like external factors rather than real manipulable parameters. The optional adoption of MB as redox mediator is also included to demonstrate the feasibility and reliability of the RSM approach.

capability of some microorganisms to oxidize organic compounds at the anode and to extract protons and electrons from the substrate at a relatively high rate [2,3], depending on several environmental conditions and on the interactions between living cells and electrode surface [4–8]. The most important concept of MFCs is to find out the best electron transfer pathway from microorganism to anode. The electron transfer can occur as the direct transfer by the immediate delivery of electrons from cellular membrane to anode, or as the mediated transfer by the intermediate exogenous redox species that can shuttle electrons from cellular membrane, via anodic medium, to anode [9,10]. Therefore, the discovery of a highly biocompatible anode environment and a suitable 3D electrode material for enhancing the colonization of anode by the microorganisms is of primary interest [11]. Particularly, many efforts have been made for developing suitable and cheap 3D electrode materials based on carbon felt (CF) [12–14], graphite [15], graphitic carbon [7,16,17], nanostructured carbon [18–20]. Among these materials, CF is very attractive because it possesses low density (∼0.14 g cm−3), high porosity (84–95%), high surface area (0.09–0.6 m2 g−1), flexible pore size (from 100 nm to 50 μm) accessible to microbes, substrates and anolyte [12] and the cost is lower than carbon nanotubes of graphene (i.e. 70 $·kg−1 vs 100–1000 $·kg−1) and more effective if considering the actual performance of MFCs [21]. This material has a porous structure, from macro to nano level, that can be further decorated with precious/non-precious metal particles or modified with different amino-terminated molecules to increase adhesion and biocompatibility [8,22–24], thus becoming the cheapest, most effective and commercially reliable anode material for MFCs. Hidalgo et al. [13] proposed a layer-by-layer procedure to depose polyaniline on CF and tripled the power density with respect to the pristine CF. Zhu et al. [25] obtained similar increase by treating CF with nitric acid and ethylenediamine, while Han et al. [26] applied CFs treated with HCl in a photoelectrocatalytic microbial fuel cell and measured improved current generation and tolerance to low pH. Zou et al. [27] functionalized CF with nanoporous Mo2C particles to introduce a robust interface between carbon fibers and endogenous flavins of microorganisms to promote the direct transfer of electrons. In this regard, the development of performing anode materials for MFCs is an important issue [28,29]. However, the development is a general topic that is not related to the specific optimization of anodic environment for a precise microorganism. Some relevant environmental parameters to consider in anode side, when operating a MFC with a specific inoculum, are temperature, pH, concentration of the substrate [6,30,31], presence and concentration of inhibiting species or not-degradable compounds [32], ratio between initial biomass and the useful substrate. The effects of these parameters can determine the biological activity of microorganism and the type of current extraction [33]. Recently suggested numerical and statistical models have considered the biological, chemical and operational factors [34]. Also, the factors are coupled with the geometry and architecture of MFC reactors [35], and the response and performance of MFCs can be tuned and regulated. As an efficient and practical approach to determine the effects of these parameters, the response surface methodology (RSM) can be proposed. The use of RSM was already applied in some biotechnology processes [36–39] and even in MFCs, experimental modeling and design using the RSM has been used for maximizing power output [40]. In spite of that, more studies are still needed to consolidate this method, to enhance the understanding of empirical phenomena involved in MFCs and to predict and manipulate the performance of MFCs [3,41–43]. These kinds of study can contribute to increase the market share of this technology and to realize scaled-up systems [44]. The novelty of this work is the focus on yeast-MFCs [45,46], with Saccharomyces Cerevisiae as biocatalyst and glucose as model substrate, and the follow-up of two previous works in which firstly a CF pretreated with polyethylenimine (CF-PEI) demonstrated higher capacity to attach yeast cells [47] and secondly the electron transfer mechanisms between

2. Materials and methods 2.1. MFC configuration and measurements For the MFC configuration, Schott's glass half H-type bottles (Adams and Chittenden, USA) were used as MFC reactor, while the CF electrodes have 7 cm2 of geometric area. A PEI treated CF acted as anodic electrode while untreated CF acted as cathodic electrode. Carbon felt (XF30A–3.5T) was purchased from Toyobo Co. (Osaka, Japan) and treated using 5 mg mL−1 of 50% w/v polyethylenimine (PEI) solution (Sigma Aldrich, St. Louis, USA) for 3 h at room temperature, then washed with de-ionized (DI) water until a neutral pH was detected in the water to prevent other bonding or polymerization. Finally, the modified CF was vacuum dried in oven at 80 °C for 12 h before being used for single cell tests. Nafion 117 membranes were treated with 3% w/w H2O2, 0.5 M H2SO4, and DI water. It was used as the separator and placed in between anode chamber and open-air cathode. Commercial yeast from Saccharomyces cerevisiae (Sigma Aldrich, St. Louis, USA) was used as biocatalyst for bioelectricity generation. The yeast was prepared in the batch anodic chamber using modified yeast extract-peptone-D-glucose (YPD) medium [16,47] that consists of 5 mg mL−1 of yeast extract, 2.5 mg mL−1 of peptone, and various concentration of D-glucose and yeast amount, following the design of experiments and the imposed glucose/yeast ratio, i.e. mg of glucose/mg of yeast. All nutrients were prepared in 0.1 M PBS (pH 7.4) and 140 mL of yeast in YPD medium was fed to the anode chamber, while Methylene Blue (MB) was optionally fed into the anode solution with the final concentration of 0.1 mM. A magnetic stirrer was then placed in anode chamber to keep anolyte in homogenous state and a magnetic hot plate was used to keep temperature in range between 25 and 27 °C, while an orifice valve in the cap of anode chamber was used to evacuate eventual gases. On the other side, environmental air was used as oxidizer to the open-air cathode and provides oxygen as terminal acceptor. The complete experimental setup is shown in Fig. 1. The yeast-MFC was operated for 72 h until stable OCV was reached, while a WonaTech Zive SP-2 potentiostat (Seoul, Korea) was used for electrochemical measurements. The potentiostat was also connected to a Frequency Response Analyzer (FRA). By coupling the frequency response analyzer with the potentiostat, the power output was analyzed as a product of the current and potential. Polarization curves were measured with a scan rate of 10 mV s−1 from the OCV until the voltage reached 0 V. 2.2. Design of experiments The optimization study was conducted to verify the influence of two major parameters, i.e. glucose concentration and glucose/yeast ratio G/ 403

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Table 1 Design of Experiments of yeast-MFC in absence or presence of MB as mediator. Run

Glucose (mg·mL−1)

G/Y Ratio (−)

1 2 3 4 5 6 7 8 9

5 5 5 10 10 10 15 15 15

1 5 10 1 5 10 1 5 10

of experimental runs is shown in Table 1. These conditions are the same for both mediatorless and mediated yeast-MFCs. The range for glucose concentration was selected according to some references [1,4,5,17,47,48,50,51] related to yeast and sugar. Concentrations lower than 5 mg mL−1 are not considered favorable for yeast growth, whereas very high concentrations of substrate are not effective and should be avoided [13]. To verify the validity of this approach, a control experiment with the optimized conditions determined in the analysis was performed and the accuracy between model and experimental results was estimated.

Fig. 1. Schematic of yeast MFC in (a) absence and (b) presence of Methylene Blue as mediator.

Y, which are related to the composition of the anodic solution and to the electricity generation. In fact, the absolute concentration of glucose substrate sets the maximum amount of available chemical energy to be converted and hence the theoretical endurance and autonomy of a batch MFC device before refilling or replacement of anodic solution to maintain current generation, while the G/Y ratio between substrate and yeast biocatalyst is important to modulate substrate consumption rate, metabolism and growth of yeast depending on the anodic environment in order to find the most efficient combination to utilize as much substrate as possible, i.e. to obtain a high current density. Another significant aspect of these parameters is that they can be easily imposed by users when preparing MFC devices for operation and therefore the study of their variation is important for MFCs management and operation because, in case of low performances, glucose concentration and G/Y ratio can be restored and kept to the optimal level. The two input factors were varied in three levels, i.e. glucose concentration 5, 10, and 15 mg mL−1 and G/Y 1, 5, and 10 w/w, selected as independent variable, while Open Circuit Voltage (OCV) and Maximum Power Density (MPD) are used as response variables to determine the combined effects of the inputs. The optional addition of a mediator, i.e. Methylene Blue 0.1 mM, was also considered as additional case to compare the influence of glucose concentration and G/Y between mediatorless and mediated MFC systems, and therefore its concentration is not an independent variable, but imposed for comparison purpose. The focus of this work is to investigate the correlation of glucose concentration and glucose/yeast ratio in yeast-MFC by using Response Surface Method (RSM). RSM is an efficient method to obtain an accurate combination of optimized conditions with fewer experiments. The design of experiments for mediatorless and mediated systems was determined by the Box-Behnken methodology [49], the statistical software Minitab 17 (Pennsylvania, USA) was used to model the empirical equations and plot the response surface while Analysis of Variance (ANOVA) was used to process the statistical parameters as well. The One-way ANOVA was specifically used to determine the response of a dependent continuous variable (i.e. OCV or MPD) from a group of unrelated and independent variables (i.e. glucose concentration and G/Y) in independent observations (i.e. unique experiments). Therefore, a 32 design with single replication and a total of 18 unique runs (32 × 2) were produced, 9 runs were MFCs without MB and other 9 runs were MFCs in mediated system with MB. The complete set

3. Results and discussion 3.1. Effect of glucose concentration and glucose/yeast ratio on OCV The OCV represents the highest voltage that can be produced in a MFC, i.e. when the device is disconnected, and is a useful value to design connection of more devices in series or parallel, evaluate polarization phenomena by comparing with thermodynamic relationships, deduce the correct acclimation and adaptation of different bacteria, cultures [16,52] or the growth cycle of the biocatalyst [15,47]. As mentioned before, the glucose concentration, 5, 10 and 15 mg mL−1 were selected in compliance with some concentrations that can be found in residual wastewaters including simple sugars [53,54]. The G/Y ratio was little investigated in yeast-MFC previously, but it becomes increasingly important as factor because the concentrations of substrate and biocatalyst have a strong influence on the performance of yeastMFCs and scale-up. To build the response surfaces and carry out the sensitivity analysis, the complete dataset obtained from the 18 runs as response to inputs variation is summarized in Table 2. From Table 2, the OCV extracted from yeast-MFC in the absence of MB was from 0.455 to 0.711 V, while in the presence of MB, the OCV range was from 0.429 to 0.584 V. This dataset was firstly compared with theoretical value representing the ideal condition. Obviously, real MFCs can only be as close as possible to ideal conditions without achieving them completely. For example, for the glucose oxidation reaction (GOR) occurring in anode without mediator (C6H12O6 + 6H2O → 6CO2 + 24H+ + 24e−), effective potential can be expressed as: 0′ 0 Ean = Ean −

[C6 H12 O6] RT ln 24F [CO2][H+]24

(1)

0′ 0 With the Ean of −0.014 V, Ean was −0.428 V in pH 7 [16]. The OCVs measured with the adoption of MB as mediator may induce the lower 0′ value due to the more positive value of Ean [55,56]. However, the values found are compatible with the literature [13,50]. In a similar way, for the oxygen reduction reaction (ORR) occurring in cathode (1/2O2 + 2H+ + 2e− → H2O), effective potential is expressed as: 0 ′ = Ecat − Ecat

404

RT 1 ln 2F [O2]1/2 [H+]2

(2)

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

It is obvious that the determination coefficient increases as the number of experiment increases. However, a reliable correlation among OCV, glucose concentration and glucose/yeast ratio was derived and the confidence is reliable [59]. To further elaborate this dataset, the ANOVA analysis was performed and the effects of glucose concentration and glucose/yeast ratio on the OCV in mediatorless yeast-MFC were investigated. The analysis for the case without MB is reported in Table 3. Surprisingly, the linear terms of X1 and X2 alone, as well as the quadratic terms, had a negligible effect on the OCV (P-value > 0.05), but an interaction between them (X1·X2) affected significantly the OCV (0.005 < P < 0.05). In a single experiment, these variables are independent and can be set separately, obtaining a certain value of OCV. However, their appropriate combination can induce a different evolution of OCV because they together directly influences the growth and metabolism of yeast. Therefore, whatever glucose concentration or glucose/yeast ratio is, a certain value of OCV is obtained but an optimum combination of absolute glucose concentration and relative glucose/yeast ratio exists to get the highest OCV. Some literatures about yeast-MFC reported that the common value of OCV was around 0.4–0.5 V for different glucose concentrations, without the consideration of glucose/yeast ratio [13,17,50]. Table 4 represents the ANOVA results showing the effects of glucose concentration and glucose/yeast ratio on OCV of the yeast-MFC operated in the presence of MB. In this case, the linear term of X1 had closely affected the OCV (0.005 < P < 0.05), while the other linear term (X2) and quadratic terms of X1 and X2 as well as interaction between the two inputs (X1·X2) had a negligible impact on the OCV (P > 0.05). In addition, the coefficient of X1 was higher than that of X2 in equation (2), irrespective of MB. It means that glucose concentration is a little bit (due to the low absolute values of the coefficients) more influent than G/Y ratio on the OCV, because the mediated electron transfer (MET) induced by MB is more decisive than the direct electron transfer (DET) induced by yeast cells. The yeast cells attached to the anode can transfer electrons only by the DET and so the growth and adaptation of yeast to the anode surface is crucial. On the other hand, the MET occurs by both the yeast cells attached to the electrode and floating in the medium. Therefore, the colonization of supporting electrode is probably of secondary importance, whereas the absolute amount of glucose, i.e. source of electrons, is more critical because if the amount of available fuel increases, the potential amount of electrons needed for establishing OCV and electrical current also increases. Two and three-dimensional surfaces representing the combined interaction of variables in the yeast-MFCs are presented in Fig. 3. In the Fig. 3a, there are four interesting findings related to the four corners of response surface describing the behavior of mediatorless yeast-MFC:

Table 2 Experimental results of runs under different glucose concentrations and glucose/yeast ratios. Run

Parameters Glucose (mg·mL−1)

Actual Results G/Y Ratio (−)

OCV (V)

MPD (mW·m−2)

1 5 10 1 5 10 1 5 10

0.455 0.512 0.654 0.665 0.642 0.614 0.711 0.596 0.585

270.484 290.327 288.291 368.035 328.802 316.215 333.892 316.672 303.707

1 5 10 1 5 10 1 5 10

0.489 0.571 0.584 0.568 0.553 0.514 0.458 0.448 0.429

340.648 342.163 355.006 409.055 350.699 346.484 386.065 335.626 319.556

Without MB 1 2 3 4 5 6 7 8 9

5 5 5 10 10 10 15 15 15

With MB 1 2 3 4 5 6 7 8 9

5 5 5 10 10 10 15 15 15

0′ 0 With the Ecat of 1.229 V, Ecat was 0.805 V at O2 partial pressure (pO2) of 0.2 and pH 7 [9]. From the result above, the OCV of yeast-MFC was estimated as 1.233 V in ideal condition but it is known that a real MFC can achieve a maximum cell potential of 0.3–0.7 V [10], the cell potential being limited by the low capacity (electrical) of the single chamber configuration [22] and other practical reasons. One possibility of the low OCV is the real pO2 at open air cathode. Since environmental air is acted as oxidizer in CF of cathode side and the pO2 is approx. 0.2, if the diffusion of air through the felt is not free and complete, the real 0′ pO2 maybe lower and that will affect the Ecat in equation (2). Environmental humidity also has an important effect on the amount of electron accepted in open air cathode as reported in Refs. [57,58]. Therefore, the optimization of empirical parameters is needed. From the regression of data in Table 2, second-order polynomial equations for the OCV of yeast-MFC in both “with MB” and “without MB” were obtained as follow:

OCVwo MB = 0.16 + 7.2·10−2X1 + 2.01·10−2X2 − 2.19·10−3X12 + 1.51·10−3X22 − 3.59·10−3X1 X2

(3)

OCVw MB = 0.365 + 3.57·10−2X1 + 2.34·10−2X2 − 1.94·10−3X12 + 0.86·10−3X22 − 1.35·10−3X1 X2

i. Low glucose concentration and low G/Y ratio: the mediatorless yeast-MFCs has low OCV of 0.4–0.5 V. However, it rapidly increases when both glucose concentration and glucose/yeast ratio increases. ii. High glucose concentration and low G/Y ratio: when glucose/yeast ratio is very low, the OCV of mediatorless yeast-MFCs has a high value of 0.7 V and then slightly decrease to 0.6 V as the glucose concentration increases. iii. High glucose concentration and high G/Y ratio: in this corner of the response surface, the OCV of mediatorless yeast-MFCs is again lower than 0.6 V because a maximum is encountered when both glucose concentration and glucose/yeast ratio increases. High glucose/yeast ratio and high glucose concentration means that the yeast concentration is less than that of glucose and every yeast cell is wellsurrounded by the substrate, hence metabolism is slower and there is no need for a fast growth and consumption of glucose. iv. Low glucose concentration and high G/Y ratio: in this condition, the OCV of mediatorless yeast-MFCs linearly increases and it reaches 0.6–0.65 V.

(4)

Where OCV is in V, X1 and X2 was the coded value of glucose concentration and glucose/yeast ratio, respectively. Fig. 2 represents the correlation of experimental and predicted values in yeast-MFCs in absence and presence of MB. According to Liu and Tzeng [37], when the result of determination coefficient (R2) is approx. 0.9 to 1, the polynomial equation is accurate enough to be capable to distinguish the effects of independent factors. The R2 obtained from the linear fitting of predicted OCV vs. actual OCV was 0.9225 and 0.8926 for the yeast-MFCs in the absence and the presence of MB. Considering the intrinsic fluctuation of OCVs experienced in the biological systems [59], the squared residual obtained in this study with a limited number of experiments already implied a satisfying degree of confidence between model (regressed) and real experiments. Based on that, in single repetition, the R2 value of mediatorless yeast-MFC was > 0.9, whereas the R2 was 0.9 for mediated yeast-MFCs. 405

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Fig. 2. Actual vs Predicted values for OCV of yeast-MFC in (a) absence and (b) presence of Methylene Blue as mediator.

active biofilm-like layer is developed on the anode. Therefore, the conclusion of this analysis is that the highest OCV is obtained when the proper amount of yeast is added to the optimal, e.g. not the highest, glucose concentration, when the glucose/yeast ratio is near one. The response surface in Fig. 3b was very different and suggested that the addition of MB completely changed the behavior and evolution of OCV. After MB was introduced in the system, there were several differences with respect to the mediatorless yeast-MFC:

Table 3 ANOVA results of glucose concentration and glucose/yeast ratio effect on OCV in absence of Methylene Blue as mediator. OCV

DF

SS

MS

F-value

P-value

Without Mediator Model Linear X1 - Glucose (mg·mL−1) X2 - G/Y Ratio (−) Quadratic X12 X22 2-way interaction X1·X2 Error Total R2

5 2 1 1 2 1 1 1 1 3 8 0.92

0.046399 0.010732 0.010651 0.000081 0.007837 0.006013 0.001824 0.026184 0.026184 0.003904 0.050304

0.009280 0.005366 0.010651 0.000081 0.003919 0.006013 0.001824 0.026184 0.026184 0.001301

7.13 4.12 8.18 0.06 3.01 4.62 1.4 20.12 20.12

0.068 0.138 0.065 0.819 0.192 0.121 0.322 0.021 0.021

i. The average OCV of the experiments is lower and the surface is quite flat. The OCV decreases from 0.52-0.55 V to 0.4–0.45 V when glucose concentration increases from 5 to 15 mg mL−1 in all glucose/yeast ratios. ii. Increasing glucose/yeast ratio makes the concentration of yeast in MFC system very low and the MET of MB is not effective. The shift from DET to MET, by adding the MB, does not increase OCV. iii. The optimum point is near the corner at low glucose concentration and high G/Y ratio. In this situation, glucose and yeast have low concentrations indicating that there are no needs for high amounts of substrate and biocatalyst but MB redox behavior shifts the value 0′ of Ean to more negative potential direction, although the yeast concentration is low (and the glucose/yeast ratio is high).

DF = Degrees of Freedom; SS=Sum-of-Square; MS = Mean Sum-of-Square; Fvalue = Fisher's test value; P-value = Probability test value. Table 4 ANOVA results of glucose concentration and glucose/yeast ratio effect on OCV in presence of Methylene Blue as mediator. OCV

DF

SS

MS

F-value

P-value

With Mediator Model Linear X1 - Glucose (mg·mL−1) X2 - G/Y Ratio (−) Quadratic X12 X22 2-way interaction X1·X2 Error Total R2

5 2 1 1 2 1 1 1 1 3 8 0.89

0.024912 0.016605 0.016581 0.000024 0.005295 0.004705 0.000591 0.003692 0.003692 0.003 0.027912

0.004982 0.008303 0.016581 0.000024 0.002648 0.004703 0.000591 0.003692 0.003692 0.001

4.98 8.30 16.58 0.02 2.65 4.70 0.59 3.69 3.69

0.108 0.060 0.027 0.887 0.217 0.119 0.498 0.150 0.150

The third evidence is a hint to suppose that the establishment of OCV in mediated systems may be governed from another parameter related to the absolute and relative concentrations of mediator. The optimization of mediator is behind the scope of this work and therefore is not analyzed here in details but it could be an interesting follow up for future researches based on current results because a drawback in the use of mediator is the biocompatibility with biocatalyst which limit the concentration of mediators. However, it is accepted that when a mediator is employed in MFCs, the current generation should be better because the charge delivery to anode, and other transport phenomena inside the anode solution, are faster and facilitated whereas OCV is a “static” parameter indicating the initial equilibrium condition for polarization. Based on that, the highest OCV could be obtained when glucose concentration was 5 mg mL−1 and glucose/yeast ratio was 10. Generally, OCV of yeast-MFCs in the presence of MB is lower than yeast-MFC in the absence of MB due to the effect of MB redox potential 0′ on the effective value of Ean , which is reduced.

DF = Degrees of Freedom; SS=Sum-of-Square; MS = Mean Sum-of-Square; Fvalue = Fisher's test value; P-value = Probability test value.

The behavior of yeast-MFCs is not trivial and difficult to predict because the results show that the substrate/biocatalyst ratio and the absolute concentration of substrate both influence performance and an optimum point exists. A low glucose/yeast ratio means that only small portion of yeast cells are close to glucose molecules: the different availability of substrate for the yeast can greatly alter the metabolism and growth. A low concentration of glucose prevents the growth of yeast in the medium because the substrate is not readily available for all cells. In contrast, when yeast and glucose concentrations are balanced, yeast will grow well and can lead to a stable OCV meaning that an

3.2. Effect of glucose concentration and glucose/yeast ratio on maximum power density The second response studied in this work is the correlation among glucose concentration, glucose/yeast ratio and the maximum power 406

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Fig. 3. Response surfaces of glucose concentration and glucose/yeast ratio effects on OCV in absence (a) and (b) presence of Methylene Blue as mediator.

Fig. 4. Actual vs Predicted values for MPD of yeast-MFC in (a) absence and (b) presence of Methylene Blue as mediator.

mediator [41,60,61]. However, even if these solutions are adopted and cell design is optimized, they do not ensure the highest MPD because the yeast-MFCs are strongly limited by the electron transfer mechanism, i.e. the interaction between anode, biocatalyst and substrate. Apart from the physical properties of the supporting anode (already reported in previous works [12,14,47,48]), these interactions can be studied and empirically optimized with the runs in Table 2. From Table 2, a specular analysis to that of OCV was also performed to obtain the MPD correlations as the function of glucose concentration and glucose/yeast ratio. Second-order polynomial equations for the MPD of yeast-MFC in “with MB” and “without MB” cases were obtained on the basis of values of Table 2:

density of yeast-MFCs. The MPD range obtained from MFCs are very extensive, commonly corresponding to 40–4000 mW m−2 and the average value of 900–1400 mW m−2 because this depends on the microbe strain and type of substrate [24,32]. This value is relatively higher when it is compared with that of yeast-based biofuel cells, which usually produce up to 600–700 mW m−2 with the use of ferricyanide as terminal electron acceptor, or 400–500 mW m−2 with the use of pure oxygen or air at cathode [5,51]. The main reasons for low power density in yeast-MFCs are the internal resistance of the reactor and the efficiency of the electron transfer mechanism between anode and biocatalyst. There are many ways to reduce the internal resistance to improve the MPD: (i) reducing the distance between electrode, (ii) reducing electrolyte resistance by increasing ion concentration in electrolyte, (iii) increasing surface area of electrode, (iv) changing the membrane to one with lower resistivity, and (v) adding a redox

MPDwo MB = 141.3 + 36.0X1 + 1.34X2 − 1.485X12 + 0.135X22 − 0.520X1 X2 (5) 407

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Table 5 ANOVA results of glucose concentration and glucose/yeast ratio effect on MPD in absence of Methylene Blue as mediator.

Table 6 ANOVA results of glucose concentration and glucose/yeast ratio effect on MPD in presence of Methylene Blue as mediator.

MPD

DF

SS

MS

F-value

P-value

MPD

DF

SS

MS

F-value

P-value

Without Mediator Model Linear X1 - Glucose (mg·mL−1) X2 - G/Y Ratio (−) Quadratic X12 X22 2-way interaction X1·X2 Error Total R2

5 2 1 1 2 1 1 1 1 3 8 0.86

5841.5 2436.7 1749.8 686.9 2770.5 2756.1 14.4 550.5 550.5 915.9 6757.4

1168.3 1218.4 1749.8 686.9 1385.3 2756.1 14.4 550.5 550.5 305.3

3.83 3.99 5.73 2.25 4.54 9.03 0.05 1.80 1.80

0.149 0.143 0.096 0.231 0.124 0.057 0.842 0.272 0.272

With Mediator Model Linear X1 - Glucose (mg·mL−1) X2 - G/Y Ratio (−) Quadratic X12 X22 2-way interaction X1·X2 Error Total R2

5 2 1 1 2 1 1 1 1 3 8 0.89

5332.11 2193.71 0.17 2193.54 1690.14 988.82 701.31 1594.02 1594.02 637.91 5970.02

1066.42 1096.85 0.17 2193.54 845.07 988.82 701.31 1594.02 1594.02 212.64

5.02 5.16 0.00 10.32 3.97 4.65 3.3 7.5 7.5

0.107 0.107 0.980 0.049 0.143 0.120 0.167 0.071 0.071

DF = Degrees of Freedom; SS=Sum-of-Square; MS = Mean Sum-of-Square. F-value = Fisher's test value; P-value = Probability test value.

DF = Degrees of Freedom; SS=Sum-of-Square; MS = Mean Sum-of-Square; Fvalue = Fisher's test value; P-value = Probability test value.

MPDw MB = 269.7 + 22.62X1 − 5.72X2 − 0.889X12 + 0.938X22 − 0.885X1 X2

concentration of glucose increased. There are three noticeable things:

(6) i. When glucose concentration increased in all G/Y ratios, MPD increased until a certain point and then decreased again. This confirms again that the two parameters investigated can effectively determine the power production of yeast-MFCs. When yeast consumed glucose in optimal concentration and relative ratio (moderate to low values), it grew well, and produce more electricity to electrode. ii. Until a certain point, increased glucose concentration and limited amount of yeast biocatalyst (i.e. high G/Y ratios) will inhibit grow of yeast although the initial yeast concentration had abundant availability of glucose because the internal resistance of anode chamber become high since glucose is non-conductive material and MPD decreased. iii. At low glucose concentration, the increase in G/Y ratio didn't affect the MPD suggesting that there is a minimum MPD value determined only by the glucose concentration.

Where MPD is expressed in mW·m−2, X1 and X2 were glucose concentration and glucose/yeast ratio, respectively. According to Fig. 4, the R2s obtained from linear fitting between actual and predicted MPD value from equations above are a little bit worse than that obtained for OCV. However, the obtained R2s were 0.8566 and 0.8925 for yeastMFC in “with MB” and “without MB” cases. MPD is more sensitive than OCV to many parameters, such as temperature, humidity, pH, electrolyte solution, etc. This is the reason why the value of R2 is slightly less than one for the “without MB” case [33,57]. ANOVA results for the effects of glucose concentration and glucose/ yeast ratio on the MPD in the “without MB” case are shown in Table 5. From Table 5 it is inferred that the linear terms X1 and X2, as well as the quadratic term X22 and the 2-way interaction term X1·X2 had negligible effect on MPD (P > 0.05). In contrast, the quadratic term X12 (glucose concentration) had a P value of 0.057 that can have a statistically significant effect on MPD. This result is opposed to that obtained in previous section for OCV because the linear combination of the two varied parameters is the governing term of equation (3) whereas maximum MPD is governed by the square of glucose concentration. Glucose is the substrate, i.e. the source of the chemical energy that is extracted to be converted in electrical energy, and a higher power density is obtained when a high concentration of the substrate is used in the yeast-MFC because the amount of energy, that can be potentially extracted, was increased. However, it can be concluded that the combination of glucose concentration and G/Y ratio to maximize OCV and MPD conveys in two different regions of the response surface. Finally, the ANOVA results for yeast-MFCs using the MB mediator are listed in Table 6. By using the same rationale, similar considerations are drawn on the relative influence of the single terms in equation (6). Briefly, in yeast-MFCs with MB as mediator, only the linear term of X2, i.e. G/Y ratio, had the most significant effect on MPD because its Pvalue is almost 0.05. The 2-way interaction term, i.e. the linear combination X1·X2, may also give a significant contribution in regulating the MPD. Although X2 had significant effect on MPD, compared to X1, the negative value of X2 coefficient will reduce the MPD response value, suggesting that the maximum was located in the low G/Y region of the surface. Even in this case, the ANOVA results in Tables 4 and 6 indicated that the maximum OCV and MPD are obtained in two different regions of the surface and therefore a compromise to optimize both should be determined. The two response surfaces analytically described by equations (5) and (6) are computed and plotted in Fig. 5. Fig. 5a showed two and three-dimensional contour of glucose concentration and G/Y ratio effect on MPD of mediatorless yeast-MFC. The response surface was considerably curved and stepped as the

The change in absolute glucose concentration and the G/Y ratio is critical to manipulate the growth of yeast because a medium with concentrated glucose will inhibit cells growth and slow down the metabolic activity, hence electron generation, as reported in Refs. [30,31]. Especially in absence of a mediator, the direct electron transfer depended only by yeast cells immobilized in the surface of electrode to form a layer which increased electrode conductivity, compatible with other studies [62,63]. Finally, the highest MPD could be obtained when glucose concentration was 12 mg mL−1 and G/Y ratio was 1, with an MPD equal to 354.7 mW m−2. Contrarily, the response surface in Fig. 5b had a saddle shape, a lower curvature, completely different from Fig. 5a. Even for the MPD, the addition of MB as mediator changed the type of response of yeastMFCs due to the different electron transfer mechanism induced. As a consequence, by varying glucose concentration and G/Y ratio a different response is obtained and the RSM approach is fundamental to model and understand the behavior for the design, control and operation of these systems. Three qualitative considerations can be drawn from Fig. 5b: i. Generally, MPD is higher in mediated yeast-MFCs because the mediated electron transfer mechanism by MB is more efficient then direct mechanism. In low G/Y ratio (approx. 1–3), MPD of mediated yeast-MFCs increased up to a quasi-plateau value when glucose concentration was also increased. In this region, yeast is capable to fully utilize glucose for electron generation and produced high MPD. ii. On the contrary, when G/Y ratio was high (approx. 8–10), the MPD 408

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Fig. 5. Plot and contour of Glucose concentration and glucose/yeast ratio effect on MPD in (a) absence and (b) presence of Methylene Blue as mediator.

the electricity generation that supports it. iii. Surprisingly, at low concentration of glucose (approx. 5–5.5 mg mL−1), MPD decreased until a certain point and then slightly increased again when glucose/yeast ratio was increased. This is in accordance to the ANOVA analysis and the leading role of G/Y variable in equation (6) and the secondary role of the interaction term which is more evident al low substrate concentration.

Table 7 Optimization results and predicted values for yeast-MFCs in absence and presence of Methylene Blue as mediator.

Glucose concentration (mg·mL G/Y ratio (−) Predicted OCV (V) Predicted MPD (mW·m−2)

−1

)

Without MB

With MB

13.18 1 0.703 352.2

10.19 1 0.536 394.1

In conclusion, the highest MPD that could be obtained with MB was 413.59 mW m−2 when glucose concentration was 12.7 mg mL−1 and G/Y ratio was 1. Therefore, MPD of yeast-MFCs in presence of MB is higher than that of yeast-MFC without MB, due to effect of MB as mediator enhancing MPD and lowering internal resistance [61]. From the results, it can be deduced that the optimal conditions, i.e. the value of glucose concentration and G/Y ratio, do not depend specifically on the presence of mediator, i.e. the two parameters investigated are independent, and the artificial mediator added is just and external aid to the electron transfer, without an intrinsic contribution to electrons generation but only a faster delivery and transport of them from the yeast to the anode. In fact, the optimal G/Y is same while in presence of MB, less concentrated glucose is necessary to achieve the optimum point, because MB can harvest electrons even from floating yeast cells increasing the electron transfer efficiency and therefore exploiting more the substrate, extracting more electrons to be given to the electrode, and the process is faster and extended to all the yeast cells in the anode chamber, i.e. not only to the cells attached directly to the supporting anode. Fig. 6. Polarization and power curves of optimized in yeast-MFC in absence and presence of MB as mediator.

3.3. Implementation of optimal conditions from RSM in real experiment decreased when more glucose was added. High G/Y ratios means that in a batch system more substrate is available and this is positive in prospect of longer operation but the yeast concentration is far lower inhibiting growth and activity of the biocatalyst and therefore

The optimal operating conditions derived from the dataset were verified in real experiments to test the accuracy of the approach and to check the best performances for biofuel cells running with or without mediator. After running the ANOVA and RSM analysis of the dataset in 409

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

Fig. 7. Comparison between modeled and experimental OCV/MPD in yeast-MFC: a) absence and b) presence of MB as mediator.

the first time. Even with a single repetition, a high degree of accuracy was obtained in any case. The model tends to slightly overestimate the response, but with a larger dataset and repeated trials, it could be easily refined to get better predictions, opening many interesting possibilities for further works on these devices. The RSM approach is a valid tool to quickly standardize complex and mutable yeast-MFCs and can characterize designs and architectures adopting new components. In fact, the response sensibility of mediated yeast-MFCs is less scattered over the response surface due to the saddle-shape, while in the mediatorless case a sudden change in glucose concentration can reduce dramatically MPD. In prospect of long term operations, being aware of the peculiar conduct of these systems, the optimal concentration of glucose and G/Y ratio derived from the model can be used to correctly schedule the periodic replacement of substrate to amend and keep MPD as high as possible. Nevertheless, the use of this simple RSM approach can help in analysis of other variables, e.g. when shifting from pure substrates (glucose, fructose, acetate, etc.), to wastes or wastewater. Chemical composition or pH can be studied with a limited number of batch experiments, as it was performed in this work, and empirical formulas can be derived to predict in advance and govern the behavior of yeastMFCs. These results can be very important to make progress and spread interest for this class of MFC.

Table 2, the software was used for the optimization task. The highest OCV and MPD values were selected as target of the optimization. The optimizer tool of the statistic software was used to determine the optimal glucose concentration and G/Y ratio based on RSM. From optimization results, the optimal glucose concentration and G/Y ratio are listed in Table 7. From these predictions, the OCV value of mediatorless yeast-MFC should be higher than yeast-MFC with presence of MB as mediator. Consequently, yeast-MFC with presence of MB should produce more electricity and current and so higher MPD than mediatorless system due to the lower internal resistance and faster electron transfer. These conditions were used to conduct two real experiments, with and without MB, and polarization and power curves are shown in Fig. 6. Fig. 6 clearly shows that the experiments confirm qualitatively the prediction of RSM. In absence of MB, OCV is sensibly higher but the MPD and current density that can be obtained are lower when compared to the case with addition of mediator. The reason is that, in closed circuit condition, the presence of the mediator in the anolyte medium contributed to reduce the internal resistance of the anode chamber as well as to harvest more electrons from floating yeast cells, i.e. not attached directly to the supporting electrode. OCV and MPD were compared with predicted values from the RSM to verify the accuracy. The comparison is shown in Fig. 7. Using the optimal parameter reported in Table 7, for the mediatorless yeast-MFC (Fig. 7a) the OCV of real yeast-MFC was 0.688 V, slightly lower than model (approx. 2.1%), while the MPD value was 340.9 mW m−2, 3.2% lower than model, showing an excellent degree of accuracy compared to model predictions. Similarly, OCV and MPD value of yeast-MFC in presence of MB as mediator were 0.521 V and 374.4 mW m−2, respectively, 2.8% and 5% lower than the predicted value from model, as shown Fig. 7b. Even if the experiments were conducted in a confined environment, a deviation from model is always expected because OCV and MPD can be affected by many environmental conditions and also because R2 < 1 after data regression. However, these results were encouraging because the model was obtained from a limited number of experiments with single repetition and it is well-known that with repeated experiments a further refinement is possible. Moreover, by using the analytical equations (3)–(6) derived from the model, an engineering track of the actual glucose concentration and G/Y in yeast-MFC is possible during operations for monitoring purposes and to correct deviations from the optimal condition. Similarly, by using the RSM approach and an extended range for the two parameters investigated, the validity range and the accuracy of the equations could be further refined.

Declaration of interest The Authors declare no conflicts of interests. Acknowledgement Dr. Domenico Frattini was supported by the Korea Research Fellowship through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT of Republic of Korea (No. 2017H1D3A1A01013887), by the National Research Foundation of Korea (NRF) and the Ministry of the Ministry of Education (MOE) (No. 2018R1D1A1A09036711) and by Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry, and Energy (MOTIE) of the Republic of Korea (No. 20184030202230). References [1] C. Santoro, C. Arbizzani, B. Erable, I. Ieropoulos, Microbial fuel cells: from fundamentals to applications. A review, J. Power Sources 356 (2017) 225–244, https:// doi.org/10.1016/j.jpowsour.2017.03.109. [2] S.-X. Teng, Z.-H. Tong, W.-W. Li, S.-G. Wang, G.-P. Sheng, X.-Y. Shi, X.-W. Liu, H.Q. Yu, Electricity generation from mixed volatile fatty acids using microbial fuel cells, Appl. Microbiol. Biotechnol. 87 (2010) 2365–2372, https://doi.org/10.1007/ s00253-010-2746-5. [3] D. Frattini, G. Falcucci, M. Minutillo, C. Ferone, R. Cioffi, E. Jannelli, On the effect of different configurations in air-cathode MFCs fed by composite food waste for energy harvesting, Chem. Eng. Trans. 49 (2016) 85–90, https://doi.org/10.3303/ CET1649015.

4. Conclusions The novel application of RSM to yeast-MFCs, adopting a CF-PEI anode, is reliably demonstrated and deeply discussed in this work for 410

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

[4] M. Christwardana, Y. Kwon, Yeast and carbon nanotube based biocatalyst developed by synergetic effects of covalent bonding and hydrophobic interaction for performance enhancement of membraneless microbial fuel cell, Bioresour. Technol. 225 (2017) 175–182, https://doi.org/10.1016/j.biortech.2016.11.051. [5] Y.V. Hubenova, R.S. Rashkov, V.D. Buchvarov, M.H. Arnaudova, S.M. Babanova, M.Y. Mitov, Improvement of Yeast−Biofuel cell output by electrode modifications, Ind. Eng. Chem. Res. 50 (2011) 557–564, https://doi.org/10.1021/ie1000949. [6] Q. Jia, L. Wei, H. Han, J. Shen, Factors that influence the performance of twochamber microbial fuel cell, Int. J. Hydrogen Energy 39 (2014) 13687–13693 https://doi.org/10.1016/j.ijhydene.2014.04.023. [7] H. Jiang, L. Yang, W. Deng, Y. Tan, Q. Xie, Macroporous graphitic carbon foam decorated with polydopamine as a high-performance anode for microbial fuel cell, J. Power Sources 363 (2017) 27–33, https://doi.org/10.1016/j.jpowsour.2017.07. 064. [8] H. Wei, X.-S. Wu, L. Zou, G.-Y. Wen, D.-Y. Liu, Y. Qiao, Amine-terminated ionic liquid functionalized carbon nanotubes for enhanced interfacial electron transfer of Shewanella putrefaciens anode in microbial fuel cells, J. Power Sources 315 (2016) 192–198 https://doi.org/10.1016/j.jpowsour.2016.03.033. [9] B.E. Logan, B. Hamelers, R. Rozendal, U. Schröder, J. Keller, S. Freguia, P. Aelterman, W. Verstraete, K. Rabaey, Microbial fuel cells: methodology and technology †, Environ. Sci. Technol. 40 (2006) 5181–5192, https://doi.org/10. 1021/es0605016. [10] B.E. Logan, J.M. Regan, Microbial fuel cells—challenges and applications, Environ. Sci. Technol. 40 (2006) 5172–5180, https://doi.org/10.1021/es0627592. [11] C. Zhao, P. Gai, R. Song, Y. Chen, J. Zhang, J.-J. Zhu, Nanostructured materialbased biofuel cells: recent advances and future prospects, Chem. Soc. Rev. 46 (2017) 1545–1564, https://doi.org/10.1039/C6CS00044D. [12] T.X. Huong Le, M. Bechelany, M. Cretin, Carbon felt based-electrodes for energy and environmental applications: a review, Carbon N. Y. 122 (2017) 564–591, https://doi.org/10.1016/j.carbon.2017.06.078. [13] D. Hidalgo, T. Tommasi, S. Bocchini, A. Chiolerio, A. Chiodoni, I. Mazzarino, B. Ruggeri, Surface modification of commercial carbon felt used as anode for Microbial Fuel Cells, Energy 99 (2016) 193–201, https://doi.org/10.1016/j.energy. 2016.01.039. [14] J. González-García, P. Bonete, E. Expósito, V. Montiel, A. Aldaz, R. TorregrosaMaciá, Characterization of a carbon felt electrode: structural and physical properties, J. Mater. Chem. 9 (1999) 419–426, https://doi.org/10.1039/a805823g. [15] D. Frattini, G. Accardo, C. Ferone, R. Cioffi, Fabrication and characterization of graphite-cement composites for microbial fuel cells applications, Mater. Res. Bull. 88 (2017) 188–199, https://doi.org/10.1016/j.materresbull.2016.12.037. [16] B.E. Logan, Microbial Fuel Cells, John Wiley & Sons, Inc., Hoboken, NJ, USA, 2007, https://doi.org/10.1002/9780470258590. [17] R. Rossi, M. Cavina, L. Setti, Characterization of electron transfer mechanism in mediated microbial fuel cell by entrapped electron mediator in saccharomyces cerevisiae, Chem. Eng. Trans. 49 (2016) 559–564, https://doi.org/10.3303/ CET1649094. [18] Y. Chung, M. Christwardana, D.C. Tannia, K.J. Kim, Y. Kwon, Biocatalyst including porous enzyme cluster composite immobilized by two-step crosslinking and its utilization as enzymatic biofuel cell, J. Power Sources 360 (2017) 172–179, https:// doi.org/10.1016/j.jpowsour.2017.06.012. [19] J.J.L. Guzman, M.O. Pehlivaner Kara, M.W. Frey, L.T. Angenent, Performance of electro-spun carbon nanofiber electrodes with conductive poly(3,4-ethylenedioxythiophene) coatings in bioelectrochemical systems, J. Power Sources 356 (2017) 331–337, https://doi.org/10.1016/j.jpowsour.2017.03.133. [20] A. Zebda, C. Gondran, A. Le Goff, M. Holzinger, P. Cinquin, S. Cosnier, Mediatorless high-power glucose biofuel cells based on compressed carbon nanotube-enzyme electrodes, Nat. Commun. 2 (2011) 370, https://doi.org/10.1038/ncomms1365. [21] A. ElMekawy, H.M. Hegab, D. Losic, C.P. Saint, D. Pant, Applications of graphene in microbial fuel cells: the gap between promise and reality, Renew. Sustain. Energy Rev. 72 (2017) 1389–1403, https://doi.org/10.1016/j.rser.2016.10.044. [22] Y. Xie, Z. Ma, H. Song, H. Wang, P. Xu, Improving the performance of microbial fuel cells by reducing the inherent resistivity of carbon fiber brush anodes, J. Power Sources 348 (2017) 193–200, https://doi.org/10.1016/j.jpowsour.2017.02.083. [23] J.M. Sonawane, S.A. Patil, P.C. Ghosh, S.B. Adeloju, Low-cost stainless-steel wool anodes modified with polyaniline and polypyrrole for high-performance microbial fuel cells, J. Power Sources 379 (2018) 103–114, https://doi.org/10.1016/j. jpowsour.2018.01.001. [24] F.-L. Ng, S.-M. Phang, V. Periasamy, K. Yunus, A.C. Fisher, Enhancement of power output by using alginate immobilized algae in biophotovoltaic devices, Sci. Rep. 7 (2017) 16237, https://doi.org/10.1038/s41598-017-16530-y. [25] N. Zhu, X. Chen, T. Zhang, P. Wu, P. Li, J. Wu, Improved performance of membrane free single-chamber air-cathode microbial fuel cells with nitric acid and ethylenediamine surface modified activated carbon fiber felt anodes, Bioresour. Technol. 102 (2011) 422–426, https://doi.org/10.1016/j.biortech.2010.06.046. [26] H.X. Han, C. Shi, L. Yuan, G.P. Sheng, Enhancement of methyl orange degradation and power generation in a photoelectrocatalytic microbial fuel cell, Appl. Energy 204 (2017) 382–389, https://doi.org/10.1016/j.apenergy.2017.07.032. [27] L. Zou, Z. Lu, Y. Huang, Z. Long, Y. Qiao, Nanoporous Mo2C functionalized 3D carbon architecture anode for boosting flavins mediated interfacial bioelectrocatalysis in microbial fuel cells, J. Power Sources 359 (2017) 549–555, https:// doi.org/10.1016/j.jpowsour.2017.05.101. [28] H. Li, B. Liao, J. Xiong, X. Zhou, H. Zhi, X. Liu, X. Li, W. Li, Power output of microbial fuel cell emphasizing interaction of anodic binder with bacteria, J. Power Sources 379 (2018) 115–122, https://doi.org/10.1016/j.jpowsour.2018.01.040. [29] T. Nam, S. Son, B. Koo, H.V. Hoa Tran, J.R. Kim, Y. Choi, S.P. Jung, Comparative evaluation of performance and electrochemistry of microbial fuel cells with

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53] [54]

411

different anode structures and materials, Int. J. Hydrogen Energy 42 (2017) 27677–27684 https://doi.org/10.1016/j.ijhydene.2017.07.180. A.A. Ghoreyshi, T. Jafary, G.D. Najafpour, F. Haghparast, Effect of type and concentration of substrate on power generation in a dual chambered microbial fuel cell, World Renew. Energy Congr. - Sweden; 8-13 May; 2011, Linkoping, Sweden, 2011, pp. 1174–1181, , https://doi.org/10.3384/ecp110571174. T. Jafary, A.A. Ghoreyshi, G.D. Najafpour, S. Fatemi, M. Rahimnejad, Investigation on performance of microbial fuel cells based on carbon sources and kinetic models, Int. J. Energy Res. 37 (2013) 1539–1549, https://doi.org/10.1002/er.2994. W. Yang, K.-Y. Kim, P.E. Saikaly, B.E. Logan, The impact of new cathode materials relative to baseline performance of microbial fuel cells all with the same architecture and solution chemistry, Energy Environ. Sci. 10 (2017) 1025–1033, https:// doi.org/10.1039/C7EE00910K. M. Alaraj, S. Feng, T.M. Roane, J.-D. Park, Effect of power shape on energy extraction from microbial fuel cell, J. Power Sources 366 (2017) 86–92, https://doi. org/10.1016/J.JPOWSOUR.2017.09.002. C. Xia, D. Zhang, W. Pedrycz, Y. Zhu, Y. Guo, Models for microbial fuel cells: a critical review, J. Power Sources 373 (2018) 119–131, https://doi.org/10.1016/j. jpowsour.2017.11.001. C. Picioreanu, M.C.M. van Loosdrecht, T.P. Curtis, K. Scott, Model based evaluation of the effect of pH and electrode geometry on microbial fuel cell performance, Bioelectrochemistry 78 (2010) 8–24, https://doi.org/10.1016/j.bioelechem.2009. 04.009. C.-L. Cheng, G.-B. Hong, Optimization of extraction process for bioactive compounds from Litsea cubeba fruits, Kor. J. Chem. Eng. 35 (2018) 187–194, https:// doi.org/10.1007/s11814-017-0251-6. B.-L. Liu, Y.-M. Tzeng, Optimization of growth medium for the production of spores from Bacillus thuringiensis using response surface methodology, Bioprocess Eng. 18 (1998) 413–418, https://doi.org/10.1007/PL00008999. M. Mohadesi, B. Aghel, M.H. Khademi, S. Sahraei, Optimization of biodiesel production process in a continuous microchannel using response surface methodology, Kor. J. Chem. Eng. 34 (2017) 1013–1020, https://doi.org/10.1007/s11814-0160342-9. H.N. Park, C.-W. Cho, H.A. Choi, S.W. Won, Polyethylenimine-coated polysulfone/ bacterial biomass composite fiber as a biosorbent for the removal of anionic dyes: optimization of manufacturing conditions using response surface methodology, Kor. J. Chem. Eng. 34 (2017) 2519–2526, https://doi.org/10.1007/s11814-017-0149-3. H. Sun, S. Luo, R. Jin, Z. He, Ensemble engineering and statistical modeling for parameter calibration towards optimal design of microbial fuel cells, J. Power Sources 356 (2017) 288–298, https://doi.org/10.1016/j.jpowsour.2017.02.051. J.R. Kim, S. Cheng, S.-E. Oh, B.E. Logan, Power generation using different cation, anion, and ultrafiltration membranes in microbial fuel cells, Environ. Sci. Technol. 41 (2007) 1004–1009, https://doi.org/10.1021/es062202m. S.V. Mohan, K. Chandrasekhar, Solid phase microbial fuel cell (SMFC) for harnessing bioelectricity from composite food waste fermentation: influence of electrode assembly and buffering capacity, Bioresour. Technol. 102 (2011) 7077–7085, https://doi.org/10.1016/j.biortech.2011.04.039. S. Venkata Mohan, G. Velvizhi, J. Annie Modestra, S. Srikanth, Microbial fuel cell: critical factors regulating bio-catalyzed electrochemical process and recent advancements, Renew. Sustain. Energy Rev. 40 (2014) 779–797, https://doi.org/10. 1016/j.rser.2014.07.109. J. Greenman, I.A. Ieropoulos, Allometric scaling of microbial fuel cells and stacks: the lifeform case for scale-up, J. Power Sources 356 (2017) 365–370, https://doi. org/10.1016/j.jpowsour.2017.04.033. O. Schaetzle, F. Barrière, K. Baronian, Bacteria and yeasts as catalysts in microbial fuel cells: electron transfer from micro-organisms to electrodes for green electricity, Energy Environ. Sci. 1 (2008) 607, https://doi.org/10.1039/b810642h. Y. Hubenova, M. Mitov, Extracellular electron transfer in yeast-based biofuel cells: a review, Bioelectrochemistry 106 (2015) 177–185, https://doi.org/10.1016/j. bioelechem.2015.04.001. M. Christwardana, D. Frattini, G. Accardo, S.P. Yoon, Y. Kwon, Early-stage performance evaluation of flowing microbial fuel cells using chemically treated carbon felt and yeast biocatalyst, Appl. Energy 222 (2018) 369–382, https://doi.org/10. 1016/j.apenergy.2018.03.193. M. Christwardana, D. Frattini, G. Accardo, S.P. Yoon, Y. Kwon, Effects of methylene blue and methyl red mediators on performance of yeast based microbial fuel cells adopting polyethylenimine coated carbon felt as anode, J. Power Sources 396 (2018) 1–11, https://doi.org/10.1016/j.jpowsour.2018.06.005. M. Rahimi, P. Valeh-e-Sheyda, H. Rashidi, Statistical optimization of curcumin nanosuspension through liquid anti-solvent precipitation (LASP) process in a microfluidic platform: Box-Behnken design approach, Kor. J. Chem. Eng. 34 (2017) 3017–3027, https://doi.org/10.1007/s11814-017-0201-3. R. Rossi, A. Fedrigucci, L. Setti, Characterization of electron mediated microbial fuel cell by Saccharomyces Cerevisiae, Chem. Eng. Trans. 43 (2015) 337–342, https://doi.org/10.3303/CET1543057. S.V. Raghavulu, R.K. Goud, P.N. Sarma, S.V. Mohan, Saccharomyces cerevisiae as anodic biocatalyst for power generation in biofuel cell: influence of redox condition and substrate load, Bioresour. Technol. 102 (2011) 2751–2757, https://doi.org/10. 1016/j.biortech.2010.11.048. B.E. Logan, Essential data and techniques for conducting microbial fuel cell and other types of bioelectrochemical system experiments, ChemSusChem 5 (2012) 988–994, https://doi.org/10.1002/cssc.201100604. V.G. Gude, Wastewater treatment in microbial fuel cells - an overview, J. Clean. Prod. 122 (2016) 287–307, https://doi.org/10.1016/j.jclepro.2016.02.022. P. Pandey, V.N. Shinde, R.L. Deopurkar, S.P. Kale, S.A. Patil, D. Pant, Recent advances in the use of different substrates in microbial fuel cells toward wastewater

Journal of Power Sources 402 (2018) 402–412

M. Christwardana et al.

[55]

[56]

[57]

[58]

13695–13702, https://doi.org/10.1016/j.ijhydene.2011.07.116. [59] H.L. Weng, D.J. Lee, Performance of sulfate reducing bacteria-microbial fuel cells: Reproducibility, J. Taiwan Inst. Chem. Eng. 56 (2015) 148–153, https://doi.org/10. 1016/j.jtice.2015.04.028. [60] A. ElMekawy, H.M. Hegab, X. Dominguez-Benetton, D. Pant, Internal resistance of microfluidic microbial fuel cell: challenges and potential opportunities, Bioresour. Technol. 142 (2013) 672–682, https://doi.org/10.1016/j.biortech.2013.05.061. [61] A. Ramanavicius, A. Ramanaviciene, Hemoproteins in design of biofuel cells, Fuel Cell. 9 (2009) 25–36, https://doi.org/10.1002/fuce.200800052. [62] T. Kim, J. Kang, J.-H. Lee, J. Yoon, Influence of attached bacteria and biofilm on double-layer capacitance during biofilm monitoring by electrochemical impedance spectroscopy, Water Res. 45 (2011) 4615–4622, https://doi.org/10.1016/j.watres. 2011.06.010. [63] R.P. Ramasamy, Z. Ren, M.M. Mench, J.M. Regan, Impact of initial biofilm growth on the anode impedance of microbial fuel cells, Biotechnol. Bioengineered 101 (2008) 101–108, https://doi.org/10.1002/bit.21878.

treatment and simultaneous energy recovery, Appl. Energy 168 (2016) 706–723, https://doi.org/10.1016/j.apenergy.2016.01.056. G. Zaitseva, Y. Gushikem, E.S. Ribeiro, S.S. Rosatto, Electrochemical property of methylene blue redox dye immobilized on porous silica–zirconia–antimonia mixed oxide, Electrochim. Acta 47 (2002) 1469–1474, https://doi.org/10.1016/S00134686(01)00870-2. C.A. Pessoa, Y. Gushikem, L.T. Kubota, L. Gorton, Preliminary electrochemical study of phenothiazines and phenoxazines immobilized on zirconium phosphate, J. Electroanal. Chem. 431 (1997) 23–27 https://doi.org/10.1016/S0022-0728(97) 00168-X. Y. Ahn, F. Zhang, B.E. Logan, Air humidity and water pressure effects on the performance of air-cathode microbial fuel cell cathodes, J. Power Sources 247 (2014) 655–659, https://doi.org/10.1016/J.JPOWSOUR.2013.08.084. K.-H. Oh, W.-K. Kim, K.A. Sung, M.-J. Choo, K.-W. Nam, J.W. Choi, J.-K. Park, A hydrophobic blend binder for anti-water flooding of cathode catalyst layers in polymer electrolyte membrane fuel cells, Int. J. Hydrogen Energy 36 (2011)

412