3.1 Analysis of the system phosphorus removal effect
3.1.1 Total phosphorus removal efficiency at different pollutant concentrations
In this group of experiments, four different influent phosphorus concentrations were used, namely 1, 2, 3, and 5 mg/L. The COD concentrations were 230, 460, 690, and 1000 mg/L, respectively, the total hydraulic retention time was 5 days, and sampling took place once every 24 h. After the samples had been filtered, the total phosphorus concentration in the top outlet solution was determined and the effects of the magnesia and garnet treatments at different influent pollutant concentrations were compared. The garnet device was referred to as device I, and the magnesia device was called device II.
Figure 2 shows that when the phosphorus concentration of the influent water was 1, 2, 3, and 5 mg/L the total phosphorus removal rate of device I reached 92.4%, 87.35%, 82%, and 78%, respectively, and the phosphorus removal rate for device II reached 80.3%, 77.9%, 75%, and 67%, respectively. It can be seen that the treatment effect of device I is stronger than that of device II. Furthermore, as the hydraulic retention time increases, the simulated domestic sewage treatment efficiency at the different influent phosphorus concentrations also shows an upward trend. When the hydraulic retention time is 1 day, the removal efficiency shows a rapid increase and the removal rates of the two devices reach about half of the total removal rate. The removal rate increased rapidly from the second day to the third day, but then the removal rate only increased slightly. This suggests that total phosphorus removal by the system mainly depends on filtration and adsorption by the matrix. The removal rate increase slowed as the matrix became saturated after a certain point. The total phosphorus removal rate also decreased as the influent pollutant concentration increased. The sewage treatment effect was optimal when the influent pollutant was 1 mg/L. In general, an increase in influent pollutant concentration and long-term operation of the system could easily saturate the matrix. However, this system overflows with phosphate ions and then re-adsorbs and degrades them.
3.1.2 Comparison between the phosphorus removal effects of the CW-MFC and CW systems
After the completion of the second phase of the experiment, the microbial fuel cell was disconnected and the external resistor, the cathode, and the anode were removed, which made the system into a separate CW system that still used the operation mode of water inlet and outlet from the bottom. The water phosphorus concentrations were 1, 2, 3, and 5 mg/L, the total hydraulic retention time was 5 days, and the system was sampled every 24 h. After filtration, the total phosphorus concentration in the uppermost water outlet solution was determined and compared with the phosphorus removal effects by the CW-MFC system.
Figure 3 is a comparison of the phosphorus removal effect between the CW-MFC system based on the two substrates and the CW system. It can be seen that, for both substrates, phosphorus removal by the CW-MFC system is better than that of the CW system. The difference between the two systems was smallest when the influent concentration was 1 mg/L. Furthermore, removal by the garnet based system is better than that of the magnesia-based system. This shows that the type of MFC system used has an impact on the removal of total phosphorus. In addition, electrolysis by the system and the action of microorganisms, such as phosphorus accumulating bacteria and phosphorus phagophores, also have an impact on the removal of phosphorus. The phosphorus removal effect of the garnet matrix is stronger than that of the magnesia matrix. The decrease in the treatment effect at the higher pollutant concentrations was probably due to the matrices becoming saturated.
3.2 Analysis of the COD removal effect
3.2.1 COD removal rate at different pollutant concentrations
There were four different influent COD concentrations in this experiment: 230, 460, 690, and 1000 mg/L. The total hydraulic retention time was 5 days, sampling took place every 24 h, and the samples were filtered before they were analyzed. The total phosphorus concentration in the uppermost water outlet solution was compared with the COD removal effect due magnesia and garnet at different influent pollutant concentrations as the hydraulic retention time increased. As above, the device with garnet as the matrix is device I, and the device with magnesia as the matrix is device II.
When the influent COD concentration was 230, 460, 690, and 1000 mg/L, the COD removal rates were 91.56%, 88.73%, 80% for device I and 73%, and 93.74%, 94.21%, 83.5%, and 85.1% for device II, respectively. The treatment effect of device II was stronger than that of device I. Between 0 h and 48 h HRT, the treatment efficiency of the simulated domestic sewage device showed a rapid upward trend at all influent COD concentrations and the removal rate of the two devices approximately approached the optimal removal rate. Between 48 and 72 h, the COD concentration in the devices suddenly increased and the removal rate decreased. After 72 h, the removal rate showed a linear increase again, finally achieving the optimal removal rate at 120 h. The devices also reached approximately their optimal treatment rate when the lowest influent COD concentration was added during the first stage. As the concentration of the influent pollutants increased, the treatment efficiency gradually decreased. This suggests that the different removal effects shown by the devices may be because the magnesia consumes more organic matter in the anodic oxidation-reduction reaction. The fluctuation in the removal effect of the systems shows that the optimal system reaction time is 48 h, and new pollutants will appear in the system after 48 h.
3.2.2 Comparison of the COD removal effect between the CW-MFC system and CW system
After completion of the second phase of the experiment, the microbial fuel cell was removed and the connection between the external resistor and the cathode and anode was discontinued so that the system became a separate constructed wetland system. However, it still used the water inlet and water outlet operation mode. A total of four different COD concentrations (230, 460, 690, and 1000 mg/L) were used, the total hydraulic retention time was 5 days, and samples were taken every 24 h. After filtration, the COD concentration in the top outlet solution was measured and the COD removal effects of CW and CW-MFC systems were compared.
Figure 5 shows a comparison of the COD removal effects of the CW and CW-MFC systems. It can be seen that both systems have good COD removal effects. The COD removal effect of the CW-MFC system with either the garnet of magnesia matrix was better than that of the CW system. The COD removal rate of the CW and CW-MFC systems with the garnet matrix was lower than that of the magnesia-based systems. The reason for the overall high removal rate by the CW-MFC system compared to the CW system may be that the redox reaction in the MFC system consumes organic matter and converts it into electricity. When the system generates electricity, an electrochemical reaction takes place to remove pollutants by electrolysis. The stable electricity generation system promotes the growth and reproduction of microorganisms, which subsequently leads to an increase in pollutant degradation. These results also show that the MFC system has a positive effect on COD removal.
3.3 Power generation performance of the CW-MFC system
The system operation is divided into five stages. The two CW-MFCs based on magnesia or garnet were operated at the same time. The first stage was the establishment and trial operation of the experimental device. The cathode and anode were connected with an external 1000 Ω resistor. Simulated domestic sewage without phosphorus was introduced and the voltage was monitored. During the second stage, the two microbial fuel cell systems were operated together. The pollutant concentration was 1 mg/L total phosphorus and the COD concentration was 230 mg/L. The pollutant concentration during the third stage was 2 mg/L total phosphorus and 460 mg/L COD, the pollutants concentration during the fourth stage was 3 mg/L total phosphorus and 690 mg/L COD, and the pollutant concentration during the fifth stage was 5 mg/L total phosphorus and 1000 mg/L COD.
It can be seen from Fig. 6 that the highest output voltage and stable voltage of the garnet device was greater than the magnesia device. During the third stage, the total phosphorus concentration was 2mg/L and COD concentration was 460 mg/L, and both devices reached their maximum output voltage. The maximum stable power generation voltage of the garnet device was 500 mV, and the maximum stable power generation voltage of the magnesia device was 290 mV. Temperature, the concentration of dissolved oxygen at the cathode, and the COD load at the anode all affected the voltage level.
In order to test the power generation performance of the CW-MFCs, the best power generation stage for the magnesia and garnet systems, namely the third stage, was used to conduct current density and power density tests. After the systems were stable, a 10–5000 Ω external resistor between the anode and cathode was connected and the output voltage at both ends of the external resistor was measured to obtain the system polarization curve and power density curve.
It can be seen from Fig. 7 that when the open circuit voltage of the garnet (a) matrix was 0.75 mV and the internal resistance was 245 Ω, the maximum power density was 0.48 W/m3, and the maximum current density was 2.1 A/m3. The open circuit voltage of the magnesia (b) device was 0.53 mV, the internal resistance was 450 Ω, and the greatest power density was 0.33W/m3. The results show that the open circuit voltage was determined by the internal resistance. The smaller the internal resistance, the greater the open circuit voltage value and the system power generation capacity.
3.4 Microbial community structure in wetlands
(1) Community structure characteristics of soil microorganisms in wetland sediments
At the phylum level, there were nine dominant microbial groups with an average relative abundance of > 0.1% and their relative abundance accounted for 85.8–86.6% of the total microbial community. The D10 layer mainly consisted of Proteobacteria (38.5%), Cyanobacteria (25.1%), Bacteroidetes (12.4%), and Gemmatimonadetes (3.7%), which accounted for 79% of the total microbial community. Five phyla changed significantly between the D30 layer and the D10 layer. These were Proteobacteria (32.2%), Cyanobacteria (21.2%), Bacteroidetes (7.3%), Gemmatimonadetes (8.7%), and Acidobacteria (5.7%). There were three phyla whose relative decreased in the D30 layer compared to the D10 layer (Fig. 8a): Proteobacteria, Bacteroidetes, and Cyanobacteria, but their change in relative abundances were positive at 6.30%, 5.10%, and 3.90%, respectively (Fig. 8b). The phyla whose relative abundances increased the most between the D10 and D30 layers were Gemmatimonadetes, Firmicutes, Acidobacteria, Actinobacteria, and Nitrospirae, but their change in relative abundance was − 5.00%, − 2.20%, − 4.40%, − 0.60%, and − 2.60%, respectively (Fig. 8b).
At the genus level, the > 0.1% relative abundance groups were classified as the dominant genera. The main dominant groups in the D10 layer were Nitrospira (2.8%), Rhodobacter (2.4%), Sphingomonas (7.2%), Tabrizicola (2.1%), Nevskia (1.1%), and Devosia (1.3%) (Fig. 9a). The main dominant groups in the D30 layer were Nitrospira (3.4%), Rhodobacter (1.2%), Sphingomonas (2.1%), Haliangium (1.7%), and H16 (1.6%). Nitrospira, Haliangium, H16, Opitutus, and Pseudomonas (Fig. 9a), whose relative abundances increased by 0.60%, 0.99%, 0.64%, 6.38%, and 0.36%, respectively. The genera that showed the most obvious decreases in relative abundance were Rhodobacter, Sphingomonas, Tabrizicola, Devosia, and Porphyrobacter, which decreased by 1.20%, 5.10%, 1.21%, 0.46%, and 0.34%, respectively. (Fig. 9b).
As shown in Fig. 10, at the wetland electrode, there were nine dominant microbial groups with an average relative abundance of > 1% at the phylum level, and their relative abundance accounted for 75.0–77.9% of the total microbial community. At the electrodes, Proteobacteria (34.5%), Cyanobacteria (19.0%), Bacteroidetes (6.3%) and Gemmatimonadetes (5.7%) were the main groups present, and these groups accounted for 77.1% of the total microbial community.
At the wetland electrode, the groups with an average relative abundance of > 0.1% at the genus level were classified as the dominant genera. The dominant genera groups were Nitrospira (2.4%), Rhodobacter (1.50%), Sphingomonas (4.30%), Tabrizicola (1.25%), Nevskia (0.76%), and Devosia (1.13%).
An analysis and comparison of the microbial community structure at the phylum level and genus level in the D10 and D30 layers of the wetland sediment soil showed that Proteobacteria and Bacteroidetes became more dominant at the phylum level, and the dominant microorganisms accounted for 85.8–86.6% of the total microbial community. At the genus level, Haliangium and Opitutus accounted for 23.29–24.23% of the microbial community. When the amount of microorganisms in the sediment and the electrode were compared, the microbial population at the electrode was significantly lower than that in the sediment, but the migration of microorganisms became the main reason for the increase in electricity generation efficiency. The results suggest that when the CW-MFC system was operational, the significant changes in these microorganisms affected the removal of organic matter in the system and the electrical generation efficiency.
3.5 Analysis of the phosphorus removal mechanism used by the two matrices
3.5.1 Scanning electron microscopy
Figures 12 and 14 show the original state of the magnesia and garnet surface structure, respectively. It can be seen that magnesia and garnet surface are relatively rough, and the magnesia surface has a typical crystal structure. There are a large number of small protrusions on the magnesia matrix surface, whereas the garnet surface has a large number of lamellar structures and fine pores, which make it more suitable for microbial adhesion. Figures 13 and 15 are electron microscopic images of the magnesia and garnet surfaces after sewage water treatment. It can be seen that dense biofilms have formed on the surfaces of both the magnesia and garnet matrices and that the biofilm layer on the garnet matrix surface is greater than that on the magnesia surface. This may be because the surface of the garnet matrix has more and finer pores that induce the formation of biofilm on the surface of matrix. This biofilm plays an important role in the treatment of pollutants and enhances the effect of phosphorus degradation.
3.5.2 XRD detection and analysis
An XRD analysis was used to further explore the dephosphorization mechanism used by the matrices, and the object images of the matrices before and after the reaction were analyzed. The XRD images are shown in Figs. 16 and 17.
Figure 16 shows that the main component of magnesia is MgO. After dephosphorization, Mg3(PO4)2 appears at 27.5°, and MgHPO4 appears at 55.3°. Magnesium oxide dissolves in water and releases Mg2+ and OH− and the ions diffuse outward. The phosphate ions in the water are released to the surface of the substrate to react with magnesium ions on the surface of the magnesia to form a magnesium phosphate precipitate. The large numbers of hydrogen phosphate ions in the alkaline water react with magnesium ions to form a magnesium hydrogen phosphate precipitate. Therefore, it is probable that the magnesia phosphorus removal mechanisms are mainly adsorption and reactions between ions.
The composition of the garnet matrix is relatively complex because it contains FeO, MgO, and other components. The removal of phosphorus by the garnet matrix mainly depends on a complex adsorption process and an ion exchange reaction that is similar to magnesia. It can be seen that after the reaction, Mg4(PO4)2OH appears at 33.5°, AlPO4(H2O)1.5 appears at 66.7°, and CaPO4 appears at 31.6°. Magnesium, aluminum, and calcium ions are all present in garnet. The main components diffuse into the water from the surface of the substrate and easily react with the phosphate and hydrogen phosphate ions that are free in the sewage to form a precipitate. This precipitate then becomes attached to the surface of the substrate.
3.5.3 Isothermal adsorption
The isothermal adsorption curve can explain the relationship between the absorbate equilibrium concentration and equilibrium adsorption in solution at a certain temperature. In this experiment, two commonly used isothermal adsorption models were selected for investigation, which were the Langmuir and Freundlich isothermal adsorption models. The complex adsorption process means that it is not presently known what the exact adsorption mechanism is. This experiment assumed that the adsorbent surface was uniform, the adsorption process took place in the monolayer, absorption takes place uniformly across the matrix surface, and the maximum adsorption amount is reached after the surface adsorbent is saturated.
Tables 5 and 6 show the linear fitting regression equation parameters for the three substrates and the two models:
Table 5
Langmuir isotherm adsorption curve results
Substrate
|
KL
|
qm
|
R2
|
1/KLqm
|
Linear regression equation
|
cordierite
|
0.018
|
0.29
|
0.9999
|
182.6
|
1/qe = 182.6/Ce + 3.448
|
magnesia
|
0.16
|
2.328
|
0.9891
|
0.0713
|
1/qe = 0.0713/Ce + 0.429
|
garnet
|
0.15
|
1.675
|
0.9371
|
3.98
|
1/qe = 3.98/Ce + 0.597
|
Table 6
Freundlich isotherm adsorption curve results
Substrate
|
KF
|
1/n
|
R2
|
Linear regression equation
|
cordierite
|
0.006
|
0.96
|
0.9996
|
Lnqe = 0.96lnCe – 2.813
|
magnesia
|
1.928
|
0.48
|
0.9662
|
Lnqe = 0.48lnCe + 0.66
|
garnet
|
1.294
|
0.68
|
0.9114
|
Lnqe = 0.68lnCe + 0.258
|
As shown in the Figs. 18–20 and Tables 5 and 6, cordierite, magnesia, and garnet all have good correlations when the Langmuir isothermal adsorption model is applied, and the corresponding coefficients R2 are 0.9999, 0.9891, and 0.9371, respectively. The results suggest that phosphorus is absorbed by the monolayer on the cordierite, magnesia, and garnet surfaces. The maximum theoretical adsorption capacities were 0.29, 2.328, and 1.675 mg/g, respectively. The KL values were 0.018, 0.16, and 0.15 on average, and in the 10–70 mg/L concentration range, the KL values for magnesia and garnet were much higher than for cordierite. In general, the adsorption capacity order for phosphorus was magnesia > garnet > cordierite. Therefore, the magnesia absorption capacity represents the theoretical maximum adsorption capacity.
In the Freundlich isothermal adsorption model, cordierite, magnesia, and garnet also show good linear relationships, and the corresponding R2 coefficients were 0.9996, 0.9662, and 0.9114, respectively. The results suggest that the cordierite, magnesia, and garnet adsorption process is inter-molecular chemisorption. The KF values were 0.006, 1.928, and 1.294 respectively, and the descending order for absorption was magnesia > garnet > cordierite. The higher the KF value, the better the adsorption performance. The magnesia and garnet adsorption performances were similar, and much higher than that of cordierite. At the same time, the smaller the value of 1/n, the easier the adsorption reaction is.
Cordierite, magnesia, and garnet conform to both the Langmuir and Freundlich equations. This suggests that the phosphorus absorption mechanisms for the three substrates are mono-molecular physical adsorption and multi-molecular chemical adsorption at the same time. Magnesia showed the best phosphorus adsorption capacity and cordierite showed the worst adsorption capacity in this experiment.