Effect of internal and external resistances on desalination in microbial desalination cell

Global freshwater is under severe stress due to increasing urbanization and anthropogenic pollution. Increasing climate change is predicted to raise water stress by up to 40% by 2030 (Connor 2015; Jafary et al. 2018a). Due to the gradual decline in sources, freshwater extraction has shifted to new alternatives, such as reusing treated wastewater or desalination of seawater. Conversely, 80% of the world's wastewater is discharged into the environment without proper treatment, making the freshwater more polluted (Connor et al. 2017). The reason behind this practice is the use of expensive and energy-intensive technologies for wastewater treatment and seawater desalination (Ragab et al. 2019). Conventional desalination techniques like reverse osmosis, multi-stage flash, and multi-effect distillation mostly rely on energy-intensive processes representing a high footprint of greenhouse gas (GHG) emission as 6.7 kg CO₂ eq/m³ of water (Cornejo et al. 2014; Alhimali et al. 2019). Therefore, the exploration of alternative energy-efficient techniques would be the solution for sustainable wastewater treatment and desalination (Jafary et al. 2018a).

Recently, the microbial desalination cell (MDC) has emerged as a green mechanism of bio-electrochemical desalination, in which microorganisms extract the chemical energy bound in wastewater to achieve bio-electrochemical desalination (Al-Mamun et al. 2018). In MDCs, electro-active bacteria (EAB) form a layer of biofilm attached to the surface of the conductive anode to perform oxidation of soluble organics from wastewater (municipal/industrial/synthetic) (Al Lawati et al. 2019). The oxidation process generates protons and extracellular electrons (e⁻) that are transmitted to the anode either through bacterial membrane-bound cytochrome c, conductive nano-wires (e-pili) or extracellular enzymes (Ping et al. 2015; Al-Mamun et al. 2017a, 2017b). The metabolically generated e⁻ then travel through the outer circuit to the cathode, where they are retrieved by electron acceptors (commonly oxygen/reductive pollutants) to accomplish the reduction reactions (Logan 2009; Feng et al. 2018; Jafary et al. 2018a). The movement of the metabolically produced e⁻ is spontaneous due to the electromotive force (E^′0 = +ve) generated by the anodic and cathodic redox reactions (He & Angenent 2006). However, the electrochemical desalination (removal of salt ions, i.e., Na⁺ and Cl⁻) takes place when seawater is separated by a pair of ion exchange membranes (i.e., anion exchange membrane (AEM) and cation exchange membrane (CEM)) placed between the anode and cathode (Cao et al. 2009; Al-Mamun et al. 2018). The transport of salt ions takes place through two driving forces, such as (1) electrochemical attraction to the counter ions and (2) concentration gradient of salt ions across the chambers (Mehanna et al. 2010). The flow of electrons generated by EAB causes the transport of salt ions across the membranes to balance the generated charges, i.e., H⁺ in the anode and OH⁻ in the cathode (Qu et al. 2012).

Some recently investigated MDC configurations not only achieved desalination, but also showed great potential for removing pollutants like sulfate (SO₄²⁻) (Jafary et al. 2018a), nitrate (NO₃⁻) (Zhang & Angelidaki 2013), copper (An et al. 2014), and petrochemical organics (Ashwaniy & Perumalsamy 2017). While, some other MDC configurations were dedicated to recovering electricity (Kalankesh et al. 2019), chemical resources (Liu et al. 2014) and algal biomass (Kokabian & Gude 2015) with simultaneous desalination. Some researchers have studied the feasibility of scaling-up MDC by applying different reactor configurations (Jafary et al. 2020a, 2020b; Wang et al. 2020), while others have sought to improve parameters, such as the mechanisms of ion transport (Ge et al. 2014; Ping et al. 2016; Lin et al. 2017; Alhimali et al. 2019), and capacitive desalination (Forrestal et al. 2015; Hou & Ma 2017; Santoro et al. 2017a). Advanced materials were also investigated concurrently to improve the low electron-carrying capacity (Santoro et al. 2017b; Anusha et al. 2018; Zuo et al. 2018) and enhance the performance of EAB biofilm (Liang et al. 2016; Ebrahimi et al. 2017a). However, the MDC is currently facing the challenges of a low rate of desalination and power generation, which are restricting its practical application (Liang et al. 2016). Therefore, understanding the factors that influence the bio-electrochemical process in MDCs is essential, and requires in-depth exploration (Jingyu et al. 2017). In any BES, external and internal resistances are two crucial operating factors that can determine performance. External resistance (R_ext) regulates the flow of e⁻ at the outer circuit using Ohm's law (V=I·R_ext) and subsequently influences the cell potential (V) and the current (I) (Rismani-Yazdi et al. 2011). Therefore, power (W=I²·R_ext) is also affected. In contrast, internal resistance (R_int) depends on the inherent configuration of BES, which is considered as the lost potential to transport electrons and protons within the BES. The R_int could be high if the resistances induced by electrodes, membranes, electrolytes, structural assemblies and biofilms were high (Rismani-Yazdi et al. 2008; Lawson et al. 2020). However, R_int can be reduced by several ways such as eliminating chambers (Jafary et al. 2020b), using ion-exchange resins (Liu et al. 2019a), achieving least internal distance between the membrane and electrode (Chen et al. 2016), fabricating electrodes with efficient electro-conductive materials (Liang et al. 2016; Elawwad et al. 2020), etc. Application of the mitigation measures could effectively reduce the R_int of MDCs and help to produce the higher power or desalination outcomes. From the electrochemical findings above, it is clear that R_ext could be more easily controlled at the outer circuit than R_int as it is an intrinsic property of the BES. Therefore R_int in a BES should be taken into account before any operation, as high R_int could be detrimental for electrochemical performance (Winfield et al. 2010).

In MDCs, electrons flow through the outer circuit mainly depends on the applied R_ext, while the capacity of ion (Na⁺, H⁺, Ca²⁺, Cl⁻, OH⁻, etc.) transfer across the chambers mainly depends on the system's R_int (Rahman et al. 2021). The best performing MDC needs to have the least R_int for minimum electrochemical loss to optimize R_ext for power generation, treatment or desalination. Therefore, it is very crucial to observe the synergistic effect of R_ext and R_int. A recent investigation observed the effect of R_ext on the electricity generation in an MDC containing high R_int (i.e., 228–420 Ω) and using variable substrate concentrations (e.g., 500, 1,500, and 3,000 mg/L of chemical oxygen demand (COD)). The study found an insignificant change in electricity generation (2 mA at 10 Ω and 1 mA at 500 Ω) due to high R_int across all three substrate concentrations (Ragab et al. 2019). Similarly, two other studies observed the effect of R_ext for pollution (i.e., NO₃⁻ and petroleum refinery wastes) removal in MDC without paying attention to the system's R_int (Zhang & Angelidaki 2013; Sevda & Abu-Reesh 2017a). It was observed that most of the MDC studies ignored the R_int of the system before applying the optimal R_ext for the best performing conditions. As a result, the previous MDC studies with high R_int did not perform satisfactorily. Therefore, there is an urgent need to manufacture MDCs with less R_int for efficient electron transfer before optimizing operational R_ext. The effect of high and low R_ext in MDC configured with high R_int and reconfigured with low R_int needs to be investigated to show the contrast and possibility to achieve the best performing condtions. Such a comparative study was rarely carried out to determine the synergistic effect of R_ext and R_int on MDC performance. Considering the above facts, it is essential to observe the effect of R_ext on desalination and current generation performance of MDCs with variable R_int. Therefore, this study aimed to investigate the effect of variable R_ext on electrochemical performances (i.e., electrochemical desalination, current generation, COD removal and columbic efficiency) using two extreme configurations for MDCs, one with high and another with low R_int. The two extreme R_ints in the MDCs were achieved by applying the different modes of air supply to the cathode (i.e., pumping diffused air into the catholyte results in a high R_int, while the cathode exposed directly to air results in a low R_int). The air cathode has been reported for a high oxygen reduction reaction and a low cathodic overpotential (Chen et al. 2018), which already been proven as improved electron transfer mechanisms that subsequently leads to a low R_int. Further, the study aimed to select the optimal range of R_ext for the best electrochemical and desalination performance in investigated MDC with a low R_int. This study will be useful to understand the effect of resistance on bioelectrical desalination and its optimization mechanism.

Two similar three-chamber MDCs were fabricated using polycarbonate materials. The inner sizes of the anode and cathode chambers were 4 cm × 4 cm × 3 cm, corresponding to a 48 mL volume of anode and cathode chambers. The dimensions of the desalination chamber were 4 cm × 4 cm × 2 cm, holding 32 mL of salt solution. The ratio of desalination to anode/cathode volume was 0.67. The desalination chamber was inserted between the anode and cathode chambers, separated from the anode by an AEM (AMI-7001, Membranes International Inc.) and the cathode by a cation exchange membrane (CEM, CMI-7000, Membranes International Inc.). Both the AEM and CEM were exactly the same sizes at 16 cm². The membranes were prepared by submerging in 5% NaCl solution overnight for sufficient hydration and expansion, followed by rinsing in deionized water before use. Membranes were placed at the junction of chambers using silicon gaskets as sealing materials.

A carbon fiber brush (Mill-Rose Lab Inc., USA) was used as an anode electrode. The brush was cleaned with pure acetone solution followed by sintering at 450 °C for 30 minutes before use (Liu et al. 2015). In the first stage of experiments, both MDCs were operated as air-pumped MDC (APMDC) by pumping air into the catholyte during the first four months of operation. Carbon cloth coated with 0.5 gm Pt/cm² (Sainergy Fuel Cell Pvt. Ltd) was used as cathode electrodes and was submerged in catholyte at this stage (Figure 1(a)). In the second stage of experiments, the air-pumped cathodes of APMDCs were replaced by commercial air-breathing cathodes, known as an air cathode MDC (ACMDC) and operated for the next three months. The air cathodes were made of wet proofing carbon cloth with 0.5 mg/cm² of Pt loading on the catholyte exposed side and four polytetrafluoroethylene (PTFE) layers on the air-exposed side (Figure 1(b)). Titanium sheet was used as a current collector in all experiments. The schematic and experimental setups of the MDCs are illustrated in Figure 1.

Anodic mixed culture was taken from an MDC, which was already operated for three months with municipal wastewater (MWW) collected from Haya Water (Muscat, Oman). At first, APMDCs were operated with R_ext = 1,000 Ω to acclimate exoelectrogenic biofilm for sufficient voltage generation (Jafary et al. 2020b). After producing repetitive voltage in the range 200–300 mV, the actual operation started with the APMDCs. The anode was operated anaerobically with synthetic wastewater (anolyte) consisting of NH₄Cl: 0.5 g/L, MgSO₄.7H₂O: 0.1 g/L, KCl: 0.13 g/L, CaCl₂.2H₂O: 0.1 g/L, sodium acetate: 2 g/L, and 10 mL/L of Wolfe's mineral and vitamin solution. Phosphate-buffered solution (100 mM, pH = 7.0) was used as a catholyte. The anolyte was refreshed in a cycle when the voltage dropped to 10% of the maximum voltage due to substrate consumption by biofilm (referred to as anolyte refreshment cycle), while the catholyte was refreshed at the beginning of every new cycle. During the operation, anolyte and catholyte were recirculated over 1-L media bottles using a peristaltic pump (BT100-1 L, Boading Longer Precision Pump Co. Ltd, China) in batch recirculation mode (Jafary et al. 2017b). In APMDCs, an aquarium pump was used to purge air into the catholyte to support the oxygen reduction reaction, while there was no need to purge air into the cathode of ACMDCs. A 35 g/L NaCl salt in deionized water was used as an influent in the desalination chambers at the beginning of all operations and refreshed before each operation. For each operation, the complete removal (∼100%) of NaCl was considered as a full operation/cycle time. The full cycles were replicated twice for the repetitive pattern of the operation. In APMDC studies, APMDCs were operated with R_ext of 1 Ω (very low R_ext) and 1,000 Ω (very high R_ext), which were designated as APMDC-1 and APMDC-1000, respectively. In the ACMDC studies, the ACMDCs were also operated with R_ext of 1 Ω and 1,000 Ω and named as ACMDC-1 and ACMDC-1000, respectively, followed by applying other external resistances; 5, 10, 20, 69, 100, 150, 200, 400, 500 Ω. The corresponding desalination and operating current generation were measured and reported accordingly.

Power and polarization tests were carried out using linear sweep voltammetry (Squidstat Plus 1132, Admiral Instruments, USA) recording at a scan rate of 5 mV/s. The tests were conducted under open circuit conditions (OCV). R_int was calculated from the linear slope of the polarization curve (Logan et al. 2006).

At the beginning of the operation, both APMDCs were operated in open circuit (OC) mode for four days. The maximum voltages of 1,001 mV for APMDC-1 and 927 mV for APMDC-1000 were generated in OC mode, respectively. At 4-days of OC operation, an average salt removal of 1.71 g/L was observed in both systems showing their identical operation and salt removal due to the concentration gradient among the chambers (Zhang & He 2012; Yang et al. 2015). Thereafter, the external resistances of 1 and 1,000 Ω were applied to initiate the biofilm enrichment stage. The electrical performance of APMDC-1 and APMDC-1000 were monitored for three months of enrichment to ensure sufficient biofilm formation and stable biocatalytic activity of the anodes.

The desalination performance of APMDCs was then investigated over full desalination cycles (>99% of salt removal). Figure 2(a) shows the trends of the current generation in APMDC-1 and APMDC-1000 over their full desalination cycles. The average current densities of 400 ± 3 mA/m² (0.64 mA) and 200 ± 2 mA/m² (0.32 mA) were observed for APMDC-1 and APMDC-1000, respectively. The maximum generated current densities were 940 mA/m² (1.5 mA) and 440 mA/m² (0.704 mA) in APMDC-1 and APMDC-1000, respectively. The high initial values of current for both APMDCs were attributed to the higher conductivity of salt in the desalination chamber at the beginning of the cycles (Jafary et al. 2017a). APMDC-1 showed almost two times the current production compared with APMDC-1000. The higher electrical performance of APMDC-1 could be explained by the larger amount of electrons, which were transferred through exoelectrogenic microorganisms to the anode at lower external resistance (Ragab et al. 2019). The results showed a small fluctuation of the current profiles in both the APMDCs. This fluctuation in the current production was mainly due to the low concentration of dissolved oxygen and its improper distribution in the catholyte of APMDC (Srivastava et al. 2017).

At the end of operation, both APMDCs were set to OC mode and polarization tests were conducted to obtain polarization and power density behavior of the systems as shown in Figure 2(b). The peak voltages of APMDC-1 and APMDC-1000 were 1,010 and 1,020 mV, respectively. The maximum power density for APMDC-1 was 930 mW/m² vs 880 mW/m² for APMDC-1000. Higher electrical performance was expected from APMDC-1 due to the huge difference between the applied external resistance of 1 and 1,000 Ω. To evaluate this aspect, internal resistance for both systems was then calculated from the linear slope of the polarization curve. The R_int of 200.5 and 295 Ω were measured for APMDC-1 and APMDC-1000, respectively. The small differences in R_int were in line with small differences in the power production of both APMDCs, indicating the superior effect of the internal resistance over the external resistance on APMDCs performance. Moreover, high internal resistance could also justify the low current generation of APMDC-1 and APMDC-1000, i.e., 0.64 and 0.32 mA, respectively.

To explain the effect of resistance on APMDC in more detail, the cathode as the main limiting factor of power in BESs should be considered (Jafary et al. 2018b). In the air-pumped cathode type, oxygen imposes limitations due to slow dissolving kinetics in liquid through air pumping. Hence, the O₂ reduction becomes constrained and causes increments in cathodic overpotential and R_int (Mateo et al. 2015). This will lead to a reduction in electron flow and eventually the low performance of APMDC (Lyon et al. 2010). That is why the overall performances of APMDCs were quite low regardless of the value of R_ext applied in the system. Therefore, the results of the study at this stage directed the experiments towards the reducing system R_int before studying the impact of external resistance on MDC performance.

COD removal is a function of substrate consumption by the system and CE is the ratio of experimentally generated current to the theoretically recoverable current by EABs. Both of these terms may vary directly (as stated in the equations) and indirectly with respect to various parameters, e.g., applied external resistance, configuration, etc. (Katuri et al. 2011). Table 1 presents the COD removal and CE for APMDC-1 and APMDC-1000.

COD removal of 53.76% and 83.56% and CE of 4.62% and 1.79% were recorded for APMDC-1 and APMDC-1000, respectively. Higher COD removal at higher external resistance has also been reported in previous studies. Ragab et al. (2019) reported approximately 78 and 89% COD removal when a R_ext of 10 Ω and 500 Ω were applied in MDC circuits, respectively. Sevda & Abu-Reesh (2017b) also reported high COD removal (∼84%) under high R_ext conditions. Rismani-Yazdi et al. (2011) related the anodic bacterial composition to the applied R_ext. A high R_ext (∼1,000 Ω) resulted in low anode potential due to low colonization of EABs on the anode. Low rate of anaerobic respiration due to less population of exoelectrochemically active microorganisms at higher applied external resistance could shift respiration (electron transfer activity) to fermentation, i.e., higher substrate consumption and lower electron transfer. Hence, the higher external resistance will lead to high COD removal and low CE overall. Moreover, enhanced production of extracellular polymeric substances (EPS) might occur through anodophilic biofilm on the anode surface at low R_ext. The produced EPS can stimulate greater extracellular electron transfer, which is attributed to the higher current generation (Xiao et al. 2017). Hence, it can justify higher CE in APMDC-1 compared with APMDC-1000. Furthermore, low cathodic redox potential due to cathodic limitation (as discussed earlier in the section on Electricity generation) is another reason for low CE in APMDC-1000 (Ebrahimi et al. 2017a).

The CTEs for APMDC-1 and APMDC-1000 were 161.63 and 241.32%, respectively, as presented in Table 1. Theoretically, 100% CTE corresponds to the full contribution of actual generated electrons on salt removal. As the CET values were above 100% in APMDC-1 and APMDC-1000, this implied the contribution of other transport phenomena than the electrical gradient in desalination; i.e., concentration gradient (Chen et al. 2011; Jafary et al. 2020a). Hence, the relatively lower CTE value of 161.63% at low R_ext (1 Ω) compared with a CTE of 241% at high R_ext (1,000 Ω) could show the higher contribution of electrical gradient and lower contribution of concentration gradient for desalination in APMDC-1 compared with APMDC-1000 (Chen et al. 2011).

APMDC-1 and APMDC-1000 were operated with 35 g/L of salt corresponding to 53.2 mS/cm of electrical conductivity in the desalination chamber. Figure 3 shows the trends of desalination for APMDC-1 and APMDC-1000. More than 99% of desalination was achieved over 20.27 and 27.56 days of operation in APMDC-1 and APMDC-1000, respectively. Relative faster desalination was observed for APMDC-1 during the first three days. During this period, 30% of salt was removed with an average desalination rate of 3.49 g/L/d in APMDC-1. It took 4 days for the removal of the same amount of salt (30%) in APMDC-1000 with an average desalination rate of 2.64 g/L/d. The remaining 70% (24.5 g/L) of salt was removed in 17.27 and 23.56 days for APMDC-1 and APMDC-1000, respectively. It was extensively reported that high salt concentration in the desalination chamber (>10 g/L) promoted a higher desalination rate (especially at the beginning of the cycle) by achieving the benefits of maximum concentration and electrical gradients (Ebrahimi et al. 2017b; Jafary et al. 2018a). However, the desalination rate dropped as the concentration of salt in the middle chamber was reduced due to low osmotic pressure from salt to adjacent chambers, decreasing the effect of concentration gradient on desalination (Liu et al. 2019b). A higher electrical gradient obtained at lower external resistance could be the main reason behind shorter desalination time at APMDC-1 compared with APMDC-1000 (Jafary et al. 2018a). The observed average daily salt removal was 1.72 ± 0.2151 and 1.26 ± 0.1486 g/L/d, which corresponded to the desalination rate of 2.52 and 1.82 mg/h for APMDC-1 and APMDC-1000, respectively. Another reason for the sluggish desalination rate at this stage was that increments of electrical conductivity in adjacent chambers promoted back diffusion of ions towards the desalination chamber (Ramírez-Moreno et al. 2019).

It was importantly noticed that, despite the observed differences in performance of the APMDCs at low and high R_ext, the bio-electrochemical and desalination performances of both systems were still too low to make the comparison point reliable and valid. The high internal resistance was hypothesized as the reason for neutralizing the impact of external resistance on APMDC-1 and APMDC-1000 performances. In other words, for APMDC with high internal resistance, there might be no effective range of R_ext to enhance the performance of desalination in the system. Therefore, the second stage of experiments was designed to study the impact of R_ext on a low R_int MDC. Hence, the APMDCs were converted to air cathode MDCs (ACMDCs) with lower internal resistance (Lawson et al. 2020) and similar R_ext studies were done.

The APMDCs were converted to ACMDCs by some simple modifications in the cathode and the impact of 1 and 1,000 Ω R_ext was investigated (Figure 1(b) and 1(d)). Figure 4(a) shows the operating current densities for both ACMDCs all through the desalination cycles. Fast and sharp increments of the current generation in ACMDC against the slow and lagged increase of current in APMDC showed how the external resistance could control the electron transfer in the ACMDCs (Lyon et al. 2010; Pinto et al. 2011). The average and maximum operating current of ACMDC-1 were 2,595 ± 9.22 mA/m² (4.15 mA) and 2,840 mA/m² (4.54 mA), respectively. ACMDC-1000 showed 395 ± 1.5 (0.63 mA) and 465 mA/m² (0.74 mA) as the average and maximum operating current, respectively. ACMDC-1 generated 6.5 times higher current than ACMDC-1000. This result showed that changing the air-pumped system to an open-air cathode could reduce the cathodic limitation by improving redox reaction kinetics (Chen et al. 2018).

At the end of the operation, both the ACMDCs were set on OC mode and polarization tests were conducted to obtain polarization and power density behavior of the systems as shown in Figure 4(b). The peak voltages of ACMDC-1 and ACMDC-1000 were 926 and 919 mV, respectively. The maximum power density for ACMDC-1 was 1,819 mW/m² vs 1,771 mW/m² for ACMDC-1000. Moreover, the maximum current density observed for ACMDC-1 and ACMDC-1,000 was 4,500 mA/m² and 4,200 mA/m², respectively. A quick comparison between APMDC and ACMDC showed a two-fold enhancement in the electrical performance of the system. However, the internal resistance changed more significantly in ACMDCs compared with the APMDCs. The internal resistances of 67 and 77 Ω were calculated for ACMDC-1 and ACMDC-1000, respectively, showing more than three-times reduction compared with APMDCs. Internal resistance varies with resistance induced by anode, cathode and electrolytes. In APMDCs and ACMDCs, the anode and electrolytes were identical (brush anode, 35 g/L of salt and PBS catholyte) but not the cathode, which caused the noticeable decrease in R_int in the ACMDCs (Lawson et al. 2020).

As shown in Table 1, the average COD removal of 51 and 83%, CE of 34 and 2% and CTE of 96 and 195% were obtained for ACMDC-1 and ACMDC-1000, respectively. Liu et al. (2019b) reported 82% of COD removal in the air cathode MDC under 1,000 Ω of R_ext, which was comparable with this study. As the low R_ext stimulated the dominance of exoelectrogenic anodophilic biofilm, it might be the reason for the higher CE and lower COD removal in ACMDC-1 compared with ACMDC-1000 (as discussed in the section on COD removal, CE and CTE). Also, the high CTE value in ACMDC-1 (95.74%) showed that the electrical gradient efficiently contributed to salt removal. Conversely, CTE above 100% obtained in ACMDC-1000 (195.19%) showed a significant influence of other ionic movements because of poor electron flow conditions in ACMDC-1000.

The desalination performance of ACMDCs is shown in Figure 5. More than 99% desalination was achieved over 5.18 and 16.74 days of operation in APMDC-1 and APMDC-1000, respectively. The low electrolyte resistance at high salt ionic conductivity improved the electrochemical performance at the beginning of the desalination cycle in ACMDC-1 and ACMDC-1000 (Yang et al. 2015). At the beginning of the desalination, the salt concentration difference between the chambers provided a strong concentration gradient, which accompanied the electrical gradient for faster desalination. However, as the concentration changed over time due to desalination, the concentration gradient gradually became active against the electrical gradient, resulting in a decreasing desalination rate, back diffusion and increasing internal resistance (Liu et al. 2019b; Jafary et al. 2020a). An MDC with low initial overpotential and electrically efficient performance could only reduce these negative impacts. Then, the high internal resistance of APMDC could justify its low desalination performance regardless of the external resistance applied. Therefore, both external and internal resistances would affect the MDC performance over a desalination cycle. For ACMDCs, the system conditions were advantageous from the perspective of internal resistance, while external resistance significantly affected the ACMDC performance. Therefore, as the salt concentration of the desalination chamber decreased, the desalination rate drastically dropped for ACMDC-1000, while no specific change was observed in ACMDC-1. In very low R_ext (1 Ω), the redox potential of the anode became favorable for the biofilms to transfer electrons at a high electron flow rate (Lyon et al. 2010). The average daily salt removal and corresponding desalination rate for ACMDC-1 were 6.76 ± 1.05 g/L/d and 9.85 mg/h versus 2.09 ± 0.086 g/L/d and 3.05 mg/h for ACMDC-1000, respectively. An air cathode MDC reported by Liu et al. (2019b) showed 4.48 mg/h desalination rate with 1,000 Ω of R_ext, which was comparable with ACMDC-1000 (Liu et al. 2019b).

Since the APMDC was quite indifferent to the changes in external resistance, ACMDC was selected to evaluate MDC reflection for the changes in R_ext. To observe the influence of R_ext on the desalination rate, the ACMDC electrical and desalination performance under different regions of external resistance were obtained (Figure 6); R_ext ≪ R_int, R_ext < R_int, R_ext ≈ R_int, R_ext > R_int, R_ext ≫ R_int. Applying 5 and 10 Ω R_ext (R_ext ≪ R_int) resulted in a quite similar desalination rate (10.42, 10.27 mg/h in turn) to that obtained at 1 Ω (9.85 mg/h). As the applied R_ext increased to 20 Ω (R_ext < R_int), the desalination rate was slightly affected. At R_ext ≈ R_int (69 Ω), the desalination rate dropped to 8.59 mg/h, showing a 17% decrease compared with its maximum value. As R_ext was increased to 100, 150 and 200 Ω (R_ext > R_int) the desalination rate decreased to 7.64, 7.35, 7.14 mg/h, respectively. At R_ext ≫ R_int, the drop in the desalination rate was significant. Increasing the R_ext to 300, 400 and 500 Ω resulted in only a 5.84, 4.66, 4.21 mg/h desalination rate, respectively, showing a 44, 55 and 59% decrease in desalination rate compared with its maximum value. The electrical performance of ACMDC under the above-mentioned applied R_ext followed a similar pattern as shown in Figure 6(a). High R_ext was reported to hinder metabolic activity and growth of exoelectrogenic organisms while hampering electron transfer in the microbial fuel cells (Pinto et al. 2011). The results of the previous study could also impose a similar impact on MDCs due to similarities in anodic and cathodic mechanisms. Higher current generation at lower external resistance (Zhang et al. 2015) facilitated desalination in ACMDCs at R_ext ≪ R_int, R_ext < R_int, R_ext ≈ R_int regions. It is worth highlighting that desalination rates were quite satisfactory in these regions in ACMDC with low internal resistance.

Table 2 summarizes the performance of different three-chambered MDCs operated under various external and internal resistances in the literature. As clearly comparable in Table 2, the studies done by Ragab et al. (2019), Gholizadeh et al. (2017), and APMDC in the present study, the MDCs with high internal resistance demonstrated very low desalination performance regardless of the value of applied R_ext. Nevertheless, the desalination performance for the MDCs with low internal resistance depended on the amount of external resistance. The desalination performance was high at low external resistance (Ma & Hou 2019) and was low at high external resistance (Liu et al. 2019b; Zhang et al. 2019), which was in total agreement with the results of this study. Ma & Hou (2019) fabricated a low internal resistance MDC by using 3D carbon nanotube-sponge anode and potassium ferricyanide as catholyte. They could achieve a high rate of desalination as 16.5 mg/h by applying a low R_ext of 0.5 Ω. On the other hand, the other three-chamber MDC with low R_int fabricated by Zhang et al. (2019) resulted in a low desalination rate since it was operated under high R_ext (1,000 Ω). Liu et al. (2019b) operated an air cathode three-chamber MDC, which was very close to ACMDC in terms of design and internal resistance. Although the system had very low internal resistance, the reported desalination rate was low due to applying high external resistance in the system (1,000 Ω). Therefore, the dominant effect of external resistance on desalination performance could be noticeable if MDC internal resistance was low enough to engage electrical generation of the system in desalination, but not overcoming the system overpotential. In this study, air-pumped MDCs were unable to show any performance variation because of their high internal resistance. However, when the internal resistance of MDC was reduced by cathode modification, the MDC showed significant improvements in electrical and desalination performance. Both configurations of the MDC used high initial salt concentration (35 g/L), which contributed higher conductivity in the desalination chamber and caused a significant concentration gradient across the chambers. That is why, the influence of low initial salt concentrations and corresponding internal resistance should be investigated to observe the effect of external resistance on MDC performance. Future studies are required to develop a model to optimize the values of internal and external resistance with respect to the desired desalination rate. The model can help researchers to fabricate MDCs with optimum internal resistance (by considering the factors which affect the internal resistance) and to operate them under optimum external resistance to reach the expected desalination rate for real applications.

The effect of external resistance on two MDCs with different values of internal resistance (i.e., low and high) was studied (APMDC and ACMDC). The results suggested that fabricating an MDC with low internal resistance was the key factor prior to adjusting the external resistance. For ACMDCs with low internal resistance, three regions of external resistance (R_ext ≪ R_int, R_ext < R_int, R_ext ≈ R_int) resulted in the performance as high desalination rate and electrical power generation of the system. However, the performance of the APMDCs with high internal resistance was too low regardless of the applied external resistance.

The authors wish to extend their appreciation to Sultan Qaboos University (SQU), Muscat, Oman, for the financial support through His Majesty's Trust Fund (SR/ENG/CAED/17/01).

All relevant data are included in the paper or its Supplementary Information.