Statistical investigation of pivotal physical and chemical factors on the performance of ceramic-based microbial fuel cells

The electricity generation of the ceramic-based microbial fuel cells

The electricity generation performance of MFCs is the most critical approach that must be studied carefully. The OCV, V and current (mA) generation (R_ext = 100 Ω) trends of the MFCs during the time are depicted in Figure 1A and B, respectively. The number in the parenthesis after the name of MFCs represents the replications of each experiment. Each experiment was replicated at least two times and prolonged for about 350 h. However, since the OCV and current generation values became relatively stabled after about 140 h, the period between 0 and 140 h has been depicted here.

Figure 1:

The electricity generation of the MFCs: (A) the open-circuit voltage (OCV), (B) the current generation, (C) the voltage versus the current density during the polarization test, and (D) the power density versus current density during the polarization test.

As can be seen, the voltage and current values of all the MFCs with the 9 mm thick ceramics (MFC-Y-9, MFC-R-9) are increased rapidly during the first 40 h of experiments, when the exo-electrogenic bacteria are in the growth phase, and the electricity generation is enhanced by the formation of biofilms over the electrodes, consequently. Thereinafter, the electricity generation of the MFCs did not change significantly and almost reached stable values, except for a few occasional losses owing to the probable sludge clogging. However, the voltage and current generation have been postponed between 40 and 60 h for MFC-Y-3, and between 20 and 40 h for MFC-Y-6, then increased gradually afterward. The excess diffusion of oxygen through the thinner ceramics towards the anode chamber, which interferes with the growth of anaerobic exo-electrogenic bacteria, could be responsible for the delay.

Comparatively, MFC-R-9 produced higher current values (2.80 ± 0.58 mA) than MFC-Y-9 (1.91 ± 0.3 mA). Presumably, due to the better oxygen barrier capability of the denser Red ceramic than the more porous Yellow ceramic, which provides an acceptable anoxic environment in the anode chamber of MFC-R-9. Although increasing the ceramic porosity improves the proton conductivity of the separator, on the other hand, it could eventuate in the extra oxygen permeation from the cathode to the anode compartment, subsequently interfering with the growth of anaerobic microorganisms in the anode chamber. Moreover, the CE of MFC-R-9 (38.44 ± 7.61%) is also higher than the value of MFC-Y-9 (27.58 ± 4.2%).

Furthermore, the average values of the CCV, V and the power density (PD, mW/m²) versus the current density of the MFCs during the polarization test are displayed in Figure 1C and D. The voltages of the MFCs are subjected to the polarization losses including the activation, ohmic, and mass transfer losses. As depicted in Figure 1C, a rapid voltage drop is occurred at the first section of the polarization curve, i.e., for the low current density part which is attributed to the activation losses. Thereafter, a nearly linear decrease in voltage at the middle part of the polarization curve can be observed owing to the ohmic overpotential in the MFC systems, which is attributed to the resistance of internal parts such as the electrodes, current collectors, and electrolytes. Finally, another rapid voltage drop is taking place at a high-current density region as a result of faster anodic substrate oxidation than the rate of electron transportation to the electrode surface, assigned as the mass transfer overpotential. As can be seen in Figure 1C, MFC-R-9 has higher activation overpotential than MFC-Y-9. Subsequently, the ohmic losses is rather similar for these two ceramic-based MFCs, but MFC-Y-9 shows a relatively higher mass transfer losses at the high-current densities compared to MFC-R-9. Moreover, the comparison of the maximum power and current densities in Figure 1D reveals the higher electricity generation performance using 9 mm thick Yellow ceramic (72.54 ± 10.42 mW/m², 462.22 ± 21.37 mA/m²) than the values obtained by the Red ceramic with the same thickness (45.94 ± 5.67 mW/m², 315.1 ± 27.63 mA/m²). Though MFC-Y-6 has lower OCV and takes more time to reach stable conditions, it generates a higher PD than MFC-R-9, owing to the lesser ohmic resistance of this MFC. However, MFC-R-9 has outperformed the thinnest Yellow ceramic (MFC-Y-3) because of uncontrolled oxygen diffusion toward the anode chamber.

Statistical optimization results

MFCs are highly complicated systems that could be influenced by several parameters, including the architecture design, membranes, electrodes, catalysts, the influent substrate, and anodic microbial consortia. Moreover, the MFC performance could be investigated regarding the distinct and complementary viewpoints like boosting the electricity generation performance or enhancing the wastewater treatment capability of the MFC system. In the present study, the combination of the chemical (ceramic composition) and physical (thickness and porosity) specifications of different ceramic-ware membranes have been statistically examined to find the main influencing features of ceramics on the overall performance of the MFCs. Table 1 shows the critical factors and their experimental levels in this study. Three levels of 3, 6, and 9 mm have been adjusted for the thickness of four types of commercial unglazed ceramics (with different porosity and chemical composition) which created four levels for the porosity and the chemical composition components of the ceramics. The RSM has been applied for this statistical study. However, the RSM analysis did not perform based on the commonly designed experiments like the central composite (CCD), Box–Behnken, or D-Optimal designs. Therein, the utilized ceramic membranes have been adopted from the available unglazed ceramic biscuits in the market and were not hand-made. Hence, the levels for the porosity and the chemical composition factors have been specified based on the existing characteristics of the commercial ceramic biscuits. Therefore, the Historical data interface from the response surface method has been selected that provides an option for adjusting the levels of RSM factors to the available experimental data.

Besides, the maximum power density (MPD) and the CE of the MFCs have been adopted as the responses of the statistical investigation. As regards, the current generation and wastewater treatment features of the MFCs are inherently considered in the CE and power generation performance, and thereby they cannot be perceived as independent responses. Hence, two distinct ANOVA have been done separately for the MPD and the CE of the MFCs as the main responses of the present study. The experimental values of the independent factors and responses of the present study are exhibited in Table 2. All the experiments were performed at least two times and their average values are reported here.

The first response of the statistical analysis: maximum power density (MPD)

The results of ANOVA for the MPD of the MFCs are presented in Table 3. The Model F-value of 840.95 implies that the model is statistically significant. There is a chance of 0.01% that this large quantity of “Model F-Value” could take place due to noise. Moreover, considering the “P-value” of less than 0.05 for the variables of Thickness (A), Porosity (B), SiO₂ (C), and Al₂O₃ (D) contents of ceramic as well as the interaction terms of AB and AC, reveals their viable significance on the electrical performance of the ceramic-based MFCs. The P-value of 0.8736 (P-value>0.05) for the “Lack of fit” parameter, which compares the residual and pure errors from the replicated experimental design points, implies the non-significant lack of fit relative to the pure error. Furthermore, a highly close value of the coefficient of determination (R² = 0.9986) to the unit and the reasonable agreement of the “Adjusted R²” (0.9974) and the “Predicted R²” (0.9563) corroborated the acceptable correlation of the observed and predicted values. Further, the “Adequate Precision” of 80.489, implies an adequate signal-to-noise ratio. Overall, the obtained model can reliably be employed to navigate the design space. Finally, the following mathematical expressions were produced from the statistical analysis to exhibit the correlation between the MPD and the significant factors of the analysis, in terms of the codded (Eq. (2)) and the actual values of parameters (Eq. (3)):

and

where t, p, Si, and Al represent the thickness (mm), porosity (%), SiO₂ (w/w%), and Al₂O₃ (w/w%) contents of ceramic membranes, respectively.

The normality assumption is estimated by constructing a normal probability plot of the residuals. Here, the linearity of the normal probability plot for the MPD confirms an acceptable normal distribution of errors in this statistical investigation (Figure 2A). Moreover, a close agreement between the experimental results and the predicted values by the model indicates the adequacy of the model equation for the prediction of MPD values in the design criteria (Figure 2C). Also, including the internally studentized values between the lateral cutoff values of 3 and −3 shows no outlier in the residual values (Figure 2B). The Box-Cox plot shows the natural logarithm (ln) of the residual sum of square (SS) against lambda, with the current value of 1.0 and the optimum value equal to 1.4 (Figure 2D). Considering the lambda value is placed in the confidence interval region (the minimum and maximum confidence interval values of 0.52 and 1.89, respectively) that is very close to the optimum value, the data do not require any transformation based on the Box-Cox plot.

Figure 2:

The diagnosis plots for the first model response of MPD: (A) the normal probability plot, (B) the internally studentized residuals versus the predicted values, (C) the predicted versus actual values, (D) the Box-Cox plot.

The second response of the statistical analysis: coulombic efficiency (CE)

The results of ANOVA analysis for the second response, i.e., CE of the MFCs, demonstrates the significant effect (P-value<0.05) of Thickness (A), Porosity (B), SiO₂ (C), and Al₂O₃ (D) content of ceramic separators. As demonstrated by Table 4, none of the interactions between the parameters has any significant influence on the CE of the MFCs. The F-value of 30.90 accompanied by the P-value of lower than 0.0001 has implied the model significance. In other words, there is only a 0.01% probability that the high “Model F-Value” could occur due to noise. Moreover, the P-value of 0.1774 for the “Lack of Fit” exhibits the not significant lack of fit relative to the pure error, and there is a 17.74% chance that the F-value of 2.44 could happen as a result of noise. Also, the highly adjacent values of R² (0.9321) and Adjusted R² (0.9019) to the unit, accompanied by a reasonable agreement of Adjusted R² with the predicted R² (0.8590), demonstrate a good correlation of the actual and predicted values obtained by the model. Furthermore, the large value of adequate precision (16.238) indicates a competent signal-to-noise ratio.

Finally, the following mathematical correlations were derived to predict the relationship between the CE of MFCs and the significant influencing parameters of “Thickness” (A), “Porosity” (B), “SiO₂” (C), and “ Al₂O₃” (D) contents of ceramic membranes, regarding the coded (Eq. (4)) and the factual factors (Eq. (5)):

where t, p, Si, and Al, abbreviations respectively show the thickness (mm), porosity (%), SiO₂ (w/w%), and Al₂O₃ (w/w%) contents of ceramic separators.

Furthermore, the diagnosis plots for the CE response of the present study are inquired to ensure the sufficiency of model correlation. Firstly, the almost linear normal probability plot authenticates the assumption of normally distributed errors (Figure 3A). Besides, the internally studentized residuals have values between 3 and −3 (Figure 3B), and reasonable conformity of the experimental values with the predicted ones indicates the admissible correlation of the model in the design criteria (Figure 3C). The Box-Cox plot for the equation model of CE demonstrates that the current value of lambda (equal to 1.0) is adjacent to the optimum value of 0.87, and is included in the confidence interval region (between the −0.14 and 1.7), thereby no transformation is needed for the data (Figure 3D).

Figure 3:

The diagnosis plots for the second model response of CE: (A) the normal probability plot, (B) the internally studentized residuals versus the predicted values, (C) the predicted versus actual values, (D) the Box-Cox plot.

Wastewater treatment efficiencies

The wastewater treatment performance of the MFCs has been estimated based on the BOD and COD removal ratios. The influent wastewater was similar for all the MFCs, with the average influent COD and BOD of 1200 ± 95 and 540 ± 24 mg/l, respectively. From Table 5, the highest BOD and COD removal efficiencies are acquired using the yellow ceramic membrane (more than 85% removal of both the COD and BOD). Subsequently, the dense floor ceramics, including the Red ceramic and UGFC, are in the second place by more than 80% removal of COD and BOD. Finally, the most porous ceramic membrane, i.e., UGWC, acquired the lowest wastewater treatment efficiency amongst all the ceramic membranes.

Moreover, the treatment proficiency was more or less enhanced by decreasing the wall thickness of the ceramic membranes, particularly for the UGWC and the yellow ceramics. This augmentation of substrate removal efficiency may be related to the addition of aerobic cathodic degradation to the anaerobic mechanism of substrate elimination due to the higher permission of substrates through the thinner ceramic separators. In this respect, Winfield et al. (2013) reported that increasing the moisture transmission to the cathode surface via the clayware membranes was consequently amended the wastewater treatment proficiency of air-cathode MFCs.

The investigation of the effect of prominent factors on the performance of MFCs

The thickness of ceramic separators

The thickness of ceramic membrane has been investigated in several MFC studies (Salar-García et al. 2019; Winfield et al. 2016; Yousefi, Mohebbi-Kalhori, and Samimi 2017). The higher thickness of ceramic resulted in the increment of the distance between the anode and the cathode electrodes, thereby increasing the ohmic losses of MFCs. However, some researchers reported that the thicker ceramic membrane did not necessarily cause lower electrical productivity in the MFC systems. In this respect, Winfield et al. (2013) indicated that the MFC with the thicker earthenware separator outperformed the thinner terracotta membranes owing to the greater water absorption capacity, which can significantly assist the proton conductivity of the earthenware membrane. Similarly, the investigation of the performance of ceramic-based MFCs using the EIS analysis confirmed that raising the thickness of ceramic separators has not an inevitably adverse effect on the electricity generation yield of MFC systems. Although the utilization of the thicker ceramic increased the ohmic resistance, the charge transfer resistance was decreased compared to the values obtained by thinner ceramics, probably owing to the superior oxygen barrier capability and hence the better accommodation of the anodic biofilm in the case of the thicker ceramic membrane (Yousefi, Mohebbi-Kalhori, and Samimi 2019).

In the present study, the statistical results indicate the significance of ceramic thickness on power production (P-value = 0.0006) and CE (P-value = 0.0126) of the MFCs. However, a comparison of the P-values for all the influencing parameters cleared that the importance of thickness (A) is lower than the other parameters, including the porosity (B), the SiO₂ (C), and Al₂O₃ (D) contents of the ceramic membranes (Tables 3 and 4).

Figure 4 exhibits the thickness effect of different investigated ceramics on the MPD and CE of MFCs. As can be seen, the performance of different ceramic types has not similar trend regarding their thicknesses. The MPD and CE of MFC-Y (ceramic porosity of 28.9%) are amended from 43.09 to 56.91 mW/m², and from 11.1 to 27.6%, respectively. Similarly, for the most porous unglazed wall ceramic (UGWC, porosity of 30.45%), both MPD and CE responses are enhanced from 225.07 to 321.11 mW/m² and from 51% to 68%, respectively, by increasing the ceramic thickness from 3 to 9 mm. Likewise, the current densities of these extensively porous ceramics are boosted, from 359.91 to 473.26 mA/m² for MFC-Y and from 1067.7 to 1433.89 mA/m² for MFC-UGWC, by thickening of the ceramic membranes. While for the denser unglazed floor ceramic (UGFC, porosity of 11%), both the MPD and CE values have been reduced from 106.89 mW/m² and 29% to 57.65 mW/m² and 18.3%, respectively, by elevating the ceramic thickness from 3 to 6 mm. These differences in the MFC performance alteration regarding the ceramic thickness are probably related to the different oxygen blockage capabilities of ceramics with distinct porosities. In such a way that elevating the ceramic thickness of highly porous ceramics limited the oxygen transmission through the membrane, hence the MFC performance was improved as a result of providing a more anoxic anodic compartment. Whereas in the case of the denser ceramic separator with tolerable oxygen barrier ability, the overall performance of the MFC has been diminished owing to raising the electrode’s distance and ohmic resistance by amplifying the thickness.

Figure 4:

The effect of ceramic thickness on the maximum power density (MPD (mW/m²), columns in the chart, left axis) and the coulombic efficiency (CE [%], dashed lines in the chart, right axis) for three different ceramic types (Yellow ceramic, MFC-Y, Red columns, and lines), unglazed wall ceramic (UGWC, blue columns, and purple lines), and unglazed floor ceramic (UGFC, green columns, and lines)).

The porosity of ceramic separators

The selection of appropriate porosity for the ceramic separator is of great importance for the efficient performance of MFCs. Though the employment of a highly porous ceramic membrane may result in excess oxygen and substrate leakage between the MFC chambers, it could ensure an adequate proton transfer through the membrane via the vehicular mechanism.

Though a highly porous ceramic separator may result in excess oxygen and substrate leakage between the MFC chambers, having a ceramic with sufficient porosity can ensure an adequate proton transfer through the membrane via the vehicular mechanism (Yousefi, Mohebbi-Kalhori, and Samimi 2017). In this context, Ghadge and Ghangrekar (2015) reported that the very porous clayware separator produced an inadequate CE and volumetric PD owing to the uncontrolled oxygen and substrate crossovers between the MFC chambers. In contrast, Pasternak et al., who studied the performance of four different ceramics as the separator of MFCs, observed the lowest power productivity for the most compact membrane (Alumina, porosity <1%). Besides the insufficient proton conductivity of the Alumina membrane, it has an extremely low porosity which hinders the water hydration of the air-cathode surface, and subsequently, the overall MFC performance was declined (Pasternak, Greenman, and Ieropoulos 2016).

In the current study, the results of the statistical analysis confirmed the significant effect of porosity on the power production (P-value<0.0001) and the CE (P-value<0.0001) of MFCs (Tables 3 and 4). Figure 5 depicts the effect of different porosities of ceramics on the MFC performance. The value of MPD is marginally enhanced by increasing the porosity from 21.45% (for the Red ceramic) to 28.94% (for the Yellow ceramic), subsequently, it has been ascended sharply for higher porosity of 30.45% (for UGWC). The CE of the MFCs has a rather similar trend, except for the minor decrease for the yellow ceramic (28.94%). However, it must be considered that these different porosities are attributed to the various ceramics with dissimilar chemical compositions and specifications. Therefore, the overperformance of the UGWC compared to the other ceramics may ascribe to its different chemical composition and proton conductivity, not to its higher porosity.

Figure 5:

The effect of ceramic porosity on the maximum power density (MPD, the columns in the chart, left axis) and the coulombic efficiency (CE [%], red dashed lines in the chart, right axis); different porosities are attributed to distinct ceramic types: the Red ceramic (MFC-R, 21.45%), the Yellow ceramic (MFC-Y, 28.94%), and the unglazed wall ceramic (UGWC, 30.45%).

The chemical composition of ceramic membranes

The chemical composition, type, and amount of clay minerals that have been used for ceramic production have an inevitable effect on the applicability of ceramics as the membrane of MFCs. The total amount of exchangeable cations in the soil, i.e., the cation exchange capacity (CEC), is directly correlated with the proton conductivity of ceramic membranes. The results of statistical analysis indicate that the concentration of SiO₂ (C) and Al₂O₃ (D) has a significant effect (P-value<0.0001) on the power generation and the CE of the MFCs. While, the concentration of the other components, including Fe₂O₃, K₂O, CaO, MgO, and TiO₂, does not recognize as significant parameters (Tables 3 and 4). Therefore, the relative importance of the silica and aluminum contents on the proton conduction rate of ceramics, consequently, on the overall performance of MFCs, is revealed.

In this respect, Ghadge et al. (2014) observed that the ceramic made of the red soil (rich in aluminum and silica) generated a higher volumetric PD compared to the black soil clayware (containing more calcium, iron, and magnesium). Pasternak, Greenman, and Ieropoulos (2016) have investigated the applicability of four distinct ceramics as the MFC separators. They obtained higher power proficiency for the Earthenware and pyrophyllite ceramics which contained more silica concentrations. Moreover, Raychaudhuri et al. employed the modified clayware ceramic membranes with different silica contents. They gained not only a higher volumetric PD (60.4%) and CE (48.5%) for the modified ceramics (30% silica) compared to the bare clayware membrane, but also the COD and phenol removal efficiencies had inevitable improvements, subsequently. They concluded that the addition of hygroscopic silica modifiers enhanced the hydration properties and thereby the proton conductivity of soil which has a critical role in lowering the internal resistance and boosting the power productivity of the MFCs (Raychaudhuri, Sahoo, and Behera 2021). Cheraghipour et al. studied the effect of SiO₂ addition into two samples of acid-leached and un-processed native soil implemented for ceramic fabrication. The augmentation of non-leached Kalporgan soil with 30% SiO₂ eventuated an extreme decrease in internal resistance from 977.36 to 115.96 Ω, boosting the power density by 15 folds and improving the CE from 27 to 79%. However, considering the supereminent increase in the micro-porosity of the ceramic during the acid leaching modification, the addition of silica did not show a drastic enhancement of the MFC performance in the case of the acid-leached soil (Cheraghipoor et al. 2021).

The effect of ceramic composition on the MPD and CE of the MFCs with different thicknesses of 9 mm and 6 mm are represented in Figure 6A and B. Seemingly, the MPD and CE values did not alter significantly (P-value<0.05) by the increase of SiO₂ content of ceramics from 65.70 to 68.46%, but further increasing the SiO₂ weight percent to 72.46%, obviously enhanced the MPD and the CE of the MFCs for both the 9 and 6 mm thick ceramics. The investigation of the slope of polarization curves for the MFCs with different SiO₂ contents (Figure 1C and D) revealed that increasing the silica content of ceramics resulted in the decrement of internal resistance (Table 2), probably owing to the enhancement of proton conductivity. The internal resistance of 635.2 Ω for the Red ceramic (MFC-R, 9 mm) with the lowest silica content (65.7 w/w%) has been reduced to the value of 559.3 Ω for the Yellow ceramic (MFC-Y, 9 mm) with 68.46 w/w% of silica. Further increment of SiO₂ content to 72.46 w/w% for the UGWC eventuated in the lowest internal resistance of 100.6 Ω among the 9 mm thick ceramics. This reduction of internal resistance may attribute to the elevated proton conductivity of ceramic separators because of a higher amount of SiO₂ in the ceramic composition. However, the definite influence of the silica content of ceramics on the internal resistance constituents, i.e., the ohmic, the charge transfer, and the diffusional impedances, could be determined using EIS analysis.

Figure 6:

The effect of SiO₂ and Al₂O₃ contents of ceramic membranes on the maximum power density (MPD (mW/m²), the blue columns in the charts, left axis) and the coulombic efficiency (CE [%], red dashed lines in the charts, right axis); (A) Thickness = 9 mm, (B) Thickness = 6 mm.