Two sets of lab-scale corridor-style MFC-A²/O reactors in parallel running were set up, labeled A and B. The components of two identical reactors mainly including: sewage tank, MFC-A²/O bioreactor, secondary settling tank, sludge storage tank. And three of them were made of plexiglass.

MFC-A²/O bioreactor was mainly composed of anaerobic zone, anoxic zone and aerobic zone. In addition, there were three baffles among them and a gap located at the bottom of anoxic zone, so that the wastewater can flow into the oxic zone. In front of the anaerobic zone, there were two inlets. At end of the oxiczone, there were two outlets containing effluent and reflux. We installed a card slot between anaerobic and anoxic zone, and the projected area could be adjusted by adding baffles. The sewage and sludge were thoroughly mixed by magnetic stirrers, which were actuated by arotated magnet module driven by a DC electric motor.

Two sets of MFC-A²/O reactors (A and B) used artificial sewage to start up, and were all inoculated with residual sludge taken from a Beijing sewage factory. The artificial wastewater contains (per liter) [9]: 0.02 g starch, 0.19 g acetic acid, 0.158 g peptone, 0.04 g beef extract, 0.0347 g NH₄Cl, 0.284 g (NH₄)₂SO₄, 0.06 g Na₂CO₃, 0.0293 g KH₂PO₄. All chemicals were of analytical grade and used as received, and all aqueous solutions were prepared with purified water.

The sewage and reflux sludge in a ratio of 1:1 by peristaltic pumped into tank, both the inflow rate were about 10 mL/min as well as the inner recycling flow rate. HRT was kept at 16.7 h. The reactor was maintained at room temperature. After adding the sludge, we adopted the way of continuous water running. The initial mixed liquor suspended solids (MLSS) was 2400 mg/L. In the whole experiment, the MLSS was maintained at 2000 - 3000 mg/L by discharging the excess sludge from the returned sludge tank.

In the start-up stage, two devices were the conventional A²/O reactor which didn’t embed the MFC. When the pollutant removal of the reactor became stable by continuously monitoring the concentration of the COD, TN and TP, the devices entered into the operational stage that MFCs were embedded to construct the MFC-A²/O reactor. The anodes of two devices were made of carbon brushes (diameter 10 mm, length 300 mm, TC12K35, Taiwan Formosa) in anaerobic zone. Adding carbon cloth (20 mm × 20 mm × 2, HP330 Shanghai Heseng, China, labeled group A) and brushes (diameter 10 mm, length 300 mm,TC12K35, Taiwan Formosa, labeled group B) respectively to the anoxic zone. The electrodes were both pre-treated with ammonia and connected to a closed loop through the external wire and resistance box (0 - 99999.9 Ω, ZX21, Shanghai Guoguang, China), to form a complete MFC-A²/O reactor.

Total nitrogen (TN) was measured by an alkaline potassiumpersulfate digestion-UV spectrophotometric method, and total phosphorus (TP), the chemical oxygen demand (COD), and MLSS were measured following standard methods [10].

The output voltage of the MFC was monitored every 10 s bydata acquisition system (USB1608FS, Measurement Computing Corporation, USA; Tracer DAQ pro software).

The polarization curve was obtained by changing the circuit external resistance when the MFC achieved steady state.

After nearly a month of debugging time, the effluent of reactors achieved steady state and the removal efficiencies of pollutant also reached a certain level. In the start-upstage, the average removal efficiencies of COD, TN and TP were shown in Table 1. The data indicated that the conditions of were relatively stable and demonstrated the successful startup for two sets of devices.

The reactors entered the operation stage after the successful startup, and two kinds of electrodes were installed correspondingly in the anaerobic and anoxic zone and thus to set up complete MFC-A²/O reactors. The average removal efficiencies of pollutants were shown in Table 1.

Comparing start-up with operation stage, we could see the influence of embedded MFC on effluent quality in A²/O process. Before embedding the MFC, the COD removal efficiencies of group A and B were not very stable, ranging from 80% to 97%. After 27 days, the removal of A and B were 95.35% and 94.80%, respectively. At 28th, two groups were embedded the MFC, and then the removal efficiencies of A and B were 97.21% and 96.10%. In Table 1, we found that, before embedding the MFC in A and B, the average removal efficiencies of COD were 91.95% ± 6.08% and 91.51% ± 4.40%. After equipped with electrode, the average removal efficiencies of COD were rose up to 94.96% ± 2.56% and 94.33% ± 2.15%, respectively. The average removal efficiencies

R^*: The average removal efficiency; ¹―start-up stage; ²―operating stage.

increased about three percentages, and that became more stable. It was mainly due to the embedded MFC used anaerobic zone as the anode chamber and a part of organic matters in swage could beconsumed by electricigens to generate electricity, thus to strengthen the COD removal efficiencies [11].

Before 15 days, the TN removal efficiencies of two groups were fluctuant, which from 60% to 90%. After 15 days, the efficiencies gradually became stable and improved. At 27th, the TN removal efficiencies of A and B were 90.30% and 91.95%, respectively. To embed the MFC in 28 days, the efficiencies increased to 96.50% and 97.19%, respectively. Compared to the average removal efficiency, we found that two reactors were only 80.23% ± 12.51% and 88.13% ± 9.87% before embedding MFC. After embedding MFC, the efficiencies of A and B changed to 87.69% ± 9.41% and 86.73% ± 9.70%, and also the TN removal efficiencies were stable at the higher level for a period of time. It was shown that embedding MFC in A²/O had certain reinforcement on denitrification. Analysis was that the denitrifying bacteria of bio-cathode could strengthen the denitrification process, so as to enhance the whole reactor denitrification efficiency [12]. Because of the changes in electrode operation parameters such as the conditions and agingsludge at end of the trial, the reactor denitrification performance would decrease as a whole.

From Table 1, we concluded that the TP removal efficiencies of two reactors were relatively low. The analysis may be associated with a higher TP concentration of the inflow sewage and the microbial community structure of the sludge. Because of the slow growth of the activated sludge, we didn’t discharge regularly excess sludge in order to maintain sufficient concentration of sludge in the reactor. It also might be one of the reasons that phosphorus removal efficiencies were not high. Before embedding MFC, we found that TP removal efficiencies were less than 50% on average and showed a downward trend in detecting the quality of the effluent. While the TP removal efficiencies had been on an upward trend after embedding MFCs, the average removal efficiencies were about 60%. We also found phosphorus removal efficiencies had improved when MFC embedded in A²/O process. The reason was partly because the bio-cathodes could ease competition for carbon between phosphorus and denitrification in a certain extent, and were helpful for anaerobic phosphorus’ release [13]. What’s more, the electrode materials of reactor could be seen as the larger surface area of the filler which provided a better living environment for PAOs, thus to reinforce the phosphorus removal process [14].