Al₂O₃ encapsulated by calcium alginate as composite for efficient removal of phosphate from aqueous solutions: batch and column studies

In recent decades, eutrophication has become a focused global issue because it not only causes great damage to the ecological environment, but also induces immeasurable social and economic losses (Li et al. 2021). The Excessive phosphorus (P) entering into the water body is considered as one of the important reasons to cause this serious environmental problem worldwide (Wang et al. 2016; Mallin & Cahoon 2020). It seems a feasible strategy to prevent the eutrophication via the efficient control of P concentration in the aqueous solution (Fan et al. 2021). Many countries have adopted an updated strict discharge standard of total P in wastewater with the range of 0.5–1 mg/L (Kalmykova & Karlfeldt Fedje 2013; Preisner et al. 2020).

In this context, various treatment processes have been extensively explored to reduce and prevent P influx into aquatic ecosystems, such as chemical precipitation (Rokidi & Koutsoukos 2012; Liu et al. 2020), biological treatment (Zhao et al. 2021a, 2021b), ion exchange (Nur et al. 2014; Pan et al. 2014; Chen et al. 2015), adsorption (Chen et al. 2020; Makita et al. 2020; Oginni et al. 2020; Zhu et al. 2020a, 2020b; Gu et al. 2021), electrodialysis (Cai et al. 2020), microfiltration and reverse osmosis methods (Zhao et al. 2021a, 2021b). Among these treatments, adsorption is undoubtedly considered as one of the most attractive ways for P removal due to its low cost and stable removal capability, and simplicity of operation and maintenance (Zhu et al. 2016; Wang et al. 2018; Wen et al. 2021). The adsorption performance of P can greatly depend on the properties of the adsorbents. Currently, the most commonly used adsorbents for phosphorus removal are resin-supported composite materials (Mucalo et al. 1995; Nur et al. 2014) and metal oxides (Koilraj & Sasaki 2017; Ahmad et al. 2020; Feng et al. 2020; Song et al. 2020; Sun et al. 2020; Zhu et al. 2020a, 2020b), in which the resin-based composites are too expensive. Furthermore, as ion-exchange resins are capable of a wide-range of removal of almost all other anions, it is impractical to use conventional ion-exchange resins to selectively remove phosphate, which reduces the adsorption efficiency and exchange capacity of the resins, and these resins are highly sensitive to pH (Ames & Dean 1970). Comparably, the most studied metal oxides are iron oxides, aluminum oxides (Al₂O₃), zirconium oxides (Feng et al. 2020) and lanthanum (hydrogen) oxides (Koilraj & Sasaki 2017). Among them, Al₂O₃ has attracted more and more attention due to its low cost and high adsorption affinity to phosphate (Liu et al. 2019; Cui et al. 2020; Li et al. 2022). Yee (1966) was one of the first researchers to propose the selective removal of phosphate by activated alumina. The results demonstrated that the removal of phosphate is not only efficient but also economical. Urano et al. (1992) showed the feasibility of using activated alumina to remove bioprocessed domestic sewage. However, although activated alumina exhibited satisfactory effectiveness on P adsorption, it usually presents as nano or micro powder particles, resulting in difficult separation from water after treatment. Furthermore, practical applications such as the industrial wastewater treatment process require a continuous flow through a fixed-bed column, and the least pressure drop of the water coming through the column were always expected. In this situation, due to small particle size, activated alumina materials were not suitable to be sufficiently packed and stabilized inside the column. Therefore, it is urgent to fabricate activated alumina on bulk supporters or modification of the surface to accord with the requirements of the column testing.

In this study, Al₂O₃ was encapsulated by calcium alginate (CA) to synthesize an Al₂O₃/CA composite, considering that sodium alginate (SA) is easy to form a water-insoluble calcium alginate gel with polyvalent metal ions (such as calcium chloride). Then, the synthesized Al₂O₃/CA was characterized and examined for the efficiency to adsorb phosphate from water in batch experiments and continuous column studies. In batch adsorption experiments, the effects of concentrations of phosphorous, reaction temperature, pH of solution and coexisting ions on the adsorption efficiency were studied, while in the dynamic adsorption, the effects of proportions of material, bed height, flow rate and influent phosphate concentration on the adsorption efficiency were investigated.

All chemical reagents used were of analytical grade without further purification in this study. The activated alumina was purchased from West Asia Chemical Reagent Co. Ltd (Shandong, China). The phosphate stock solution (500 mg/L-P) was prepared by dissolving 2.196 g of dry potassium dihydrogen phosphate (KH₂PO₄) in 1,000 mL of deionized water. The concentration of phosphate was varied from 1 to 100 mg L⁻¹ by diluting the stock solution with pH kept at 7.0 using HCl and NaOH solution. 10% (m/v) ascorbic acid solution was obtained by dissolving 10 g of ascorbic acid in 100 mL of deionized water. The molybdate solution was prepared as follow: dissolve 13 g of ammonium molybdate in 100 mL of deionized water and 0.35 g of potassium antimony tartrate in 100 mL of water, respectively, then slowly add ammonium molybdate solution into 300 mL of (1:1) sulfuric acid, and finally add tartrate potassium antimony solution and mix completely. The quartz sand (25–50 mesh) was soaked in 1.0 mol/L HCl for 24 h and then rinsed with deionized water until the pH of the rinse water was close to neutral and then dried in a 110 °C oven.

0.2–2 g of SA was dissolved in 100 mL warm deionized water completely to form a transparent solution. After the solution cooled to room temperature, 10 g of activated alumina powder was added and mixed completely. 10 mL of CaCl₂ solution (20 wt%) was then dropwise added with continuous stirring. After precipitated completely, the product was filtrated and washed several times with deionized water and acetone, respectively, and then dried in a vacuum oven at 60 °C for 24 hours. The obtained Al₂O₃/CA composites with SA: Al₂O₃ molar ratio of 1:50, 1:25, 1:17, 1:10 and 1:5 were named as Al₂O₃/CA-1, Al₂O₃/CA-2, Al₂O₃/CA-3, Al₂O₃/CA-4 and Al₂O₃/CA-5, respectively.

An UV-Vis spectrophotometer (Shimadzu UV-2450, Japan) was used to determine the concentrations of phosphate by the molybdenum blue spectrophotometric method. The specific surface area was determined using an ASAP2020M + C type specific surface area analyzer States (Mack ASAP2020M + C, USA). The crystal structure and material composition were analyzed using the X'pert diffractometer (Panaka X'Pert Pro, Netherlands) with copper radiation (CuK_α, λ = 1.54 Å) at room temperature. The pyrolysis characteristics was carried out by thermogravimetric analysis (TGA, Netzsch TG 209 F3, Germany). All Fourier transform infrared (FTIR) spectra were obtained by a thermo spectrophotometer (Nicolet iS10, USA). The N₂ adsorption/desorption curves and pore characteristics of the sample was determined using a BET analyzer (Costech Instruments Sorptometer Kelvin 1042).

0.5 g of adsorbents (Al₂O₃ or Al₂O₃/CA) were added into a 500 mL conical flask with various concentrations of phosphorus solution (1, 2.5, 5, 10, 25, 50, 100 mg/L-P). For adsorption kinetics, the initial phosphate concentration was 2.5 mg/L. In addition, batch experiments were performed to investigate the effect of co-existing Na₂SO₄, NaCl, NaNO₃ or Na₂CO₃ on phosphate adsorption at molar ratio of 5:1, 10:1, 20:1, 40:1 and 60:1. The experimental solutions were placed in a thermostatic shaker under 200 rpm at 30 °C for 24 hours. At the scheduled sampling time, liquid samples were taken and filtered with 0.22 μm filter and stored at 4 °C before analysis.

The adsorption of phosphate on Al₂O₃/CA in a continuous fixed-bed column was conducted in a transparent glass column (inner diameter 2.0 cm, length 40 cm). The bottom and top parts of the glass column was filled with quartz sand (≈9.5 cm) to produce a constant flow. A certain amount of Al₂O₃/CA (2.0 g, 4.0 g, or 6.0 g) was filled into the columns, corresponding to a certain bed height (5.2 cm, 8.0 cm, or 12.5 cm). Therefore, the adsorption zone used in the packed column was composed of Al₂O₃/CA and quartz sand at a mixture ratio of 1:5. The influent solutions containing different phosphorus concentrations (2.5, 5, 10 mg/L-P), were pumped through the glass column at various rates (3, 5, 10 mL/min) using peristaltic pumps. The samples of effluent were collected with 10 mL plastic tubes and the concentration were determined. The phosphate adsorption breakthrough curve was obtained by plotting the ratio of the C_t/C₀ (C₀ and C_t are the phosphate concentration in the influent and effluent, respectively) as a function of time. Dynamic and batch adsorption tests were all carried out in duplicate, and the mean values were presented.

Langmuir and Freundlich equations were used to analysis the experimental data for adsorption isotherms, which were described as follows (Wang et al. 2018):

In addition, commonly used adsorption kinetic models were used to fit in the experimental data (Huo et al. 2018).

The breakthrough curve is a significant parameter during the dynamic adsorption of the continuous fixed-bed column, which reflects the adsorption kinetics, adsorption equilibrium, and mass transfer mechanism (Netpradit et al. 2004). When the effluent phosphorous concentration (c_t) reaches 10% of the influent phosphorous concentration (c_t/c₀ = 0.1), the corresponding time and volume are regarded as the breakthrough time (t_b) and the treated volume at breakthrough time (V_b). Similarly, the corresponding time and volume are regarded as the exhaustion time (t_s) and the treated volume at exhaustion time (V_s) when c_t/c₀ = 0.9.

As shown in Figure 1, there was no significant difference observed on the XRD patterns of the serial Al₂O₃/CA composites. Obviously, the peaks at 2θ = 14.46°, 2θ = 37.28°, 2θ = 42.61° and 2θ = 67.03° existed on all the materials, although the intensity might be slightly different. These results showed that the modification of calcium alginate did not change the crystalline phase of activated alumina. The XRD analysis suggested that activated alumina is a mixture of AlOOH (PDF No. 01-072-0359) and Al₂O₃ (PDF No. 00-013-0373). The intensity of the diffraction peaks in the spectra is relatively weak, indicating a weak crystallinity of the synthesized Al₂O₃/CA. There is no obvious difference between the diffraction patterns of serial Al₂O₃/CA and pure activated alumina. It was worthwhile to mention that the peak of AlPO₄ (JCPDS No. 00-010-0423) at 2θ = 26.43° was observed after phosphate adsorption. Therefore, it can be confirmed that the adsorption of phosphate by Al₂O₃/CA led to the formation of aluminum phosphate precipitate.

The thermogravimetric analysis of three different proportions of Al₂O₃/CA (SA and Al₂O₃ molar ratios of 1:5, 1:17 and 1:50) in an air atmosphere at a ramp rate of 10 °C/min from 40 °C to 800 °C is shown in Fig. S1. As shown in Fig. S1a, the mass loss increased with the increase of the amount of calcium alginate. As for Al₂O₃/CA with a ratio of 1:17, the TG analysis (Fig. S1b) is mainly divided into four stages: (1) the first stage occurs at 4–170 °C. The weight loss in this process can be attributed to the evaporation of bound water in the material and the breakage of some glycoside bonds; (2) the second stage occurs at 170 °C to 280 °C where the alginic acid of the outer layer cleaved into a more stable intermediate product; (3) the third stage begins with a significant weight loss process from 280 °C, which includes further cracking of the intermediate product and partial carbonation; and (4) the fourth stage falls in the range of 550 °C to 800 °C, which is the process for the further oxidation of calcium oxide.

Fig. S2 presents the specific surface area (SSA) and pore size of Al₂O₃ and Al₂O₃/CA-3 with a ratio of 1:17. The results showed that pore sizes of Al₂O₃ and Al₂O₃/CA-3 are mostly distributed below 10 nm in Table S1, suggesting both are mesoporous materials. The average pore size of Al₂O₃ and Al₂O₃/CA-3 was 6.28 nm and 4.67 nm, respectively. The BET measured SSA of Al₂O₃ and Al₂O₃/CA-3 were 37.31 m²/g and 51.73 m²/g, respectively. The result demonstrated that the modification with CA on Al₂O₃ decreased the pore size but increased the SSA. It was worth noting that the characteristic peak of Ca-O appears at 1,070 cm⁻¹ as shown in Figure 2, which confirms that CA is successfully coated on the surface of Al₂O₃. In addition, it could be inferred from the FTIR spectrum of the adsorbed material that there is an obvious antisymmetric stretching characteristic peak of at 1,050–1,100 cm⁻¹, which means the effective adsorption of phosphate by the material. Figure 3 presents SEM images of Al₂O₃/CA showing inhomogeneous particles encapsulated with tangled network structure. The size was increased compared with aluminum oxide in Figure 3(b), which means that CA is successfully coated on the surface of aluminum oxide.

The Al₂O₃/CA with a ratio of 1:17 was chosen for adsorption kinetics and affecting factors studies based on previous studies and cost considerations. Figure 4(a) presents the adsorption kinetics of phosphate on Al₂O₃/CA and Al₂O₃ with an initial phosphate concentration of 2.5 mg/L. During the initial 6 h of the reaction, the adsorption rate of phosphate was quite fast and the adsorption capacity increases rapidly. And then the adsorption rate slowed down and eventually reached the adsorption equilibrium. The adsorption rate and phosphate removal efficiency of Al₂O₃/CA were higher than those of Al₂O₃. Additionally, the experimental data were fit by the pseudo-first-order kinetics model and pseudo-second-order model in Table S2. The correlation coefficients, R², exhibited that the pseudo-second-order model fits a bit better for the kinetics data, implying that the rate-limiting step may be the chemical process (Liu et al. 2019).

Figure 4(a) included three processes that can be explained as follows: before 180 min (step 1), adsorption rate was quite fast due to the abundant adsorption sites of material. From 180 min to 6 h (step 2), the adsorption rate progressively slowed down, probably due to the effect of intraparticle diffusion. Finally, after 6–14 h (step 3) the adsorbent surface was gradually saturated and adsorption reached equilibrium (Yee 1966; Urano et al. 1992).

To further compare the adsorption capacity of on Al₂O₃ and Al₂O₃/CA, the adsorption isotherms were conducted with initial phosphate concentration in the range of 1–100 mg/L. As shown in the Figure 4(b), the adsorption capacity tended to increase with increasing equilibrium concentration. The Langmuir and Freundlich adsorption models were used to fit the experimental data. As demonstrated in Figure 4(b), the experimental data were fitted better with the Freundlich model. Table S3 summarized the model parameters, suggesting the adsorption isotherm was well expressed by the Freundlich model (R² = 0.9933) compared with the Langmuir model (R² = 0.7564). The values of K_F are 2.79 and 4.69, while the values of 1/n are 0.13 and 0.18, for Al₂O₃ and Al₂O₃/CA, respectively. Al₂O₃/CA composites with a ratio of 1:17 were selected to treat simulated phosphorus containing wastewater with an initial concentration of 2.5 mg/L at pH = 6.50. Thermodynamic studies were carried out at 288, 303 and 318 K. The results of the calculations in Table S4 showed that ΔH and ΔS are positive, indicating that the adsorption process is a heat-absorbing entropy process, while ΔG is negative, indicating that the process can proceed spontaneously.

The effect of temperature on the adsorption of phosphate by Al₂O₃/CA was investigated by batch experiments. The results showed that the adsorption capacity increased with rising temperature in the range of 288–318 K, indicating that the reaction is endothermic (Hano et al. 1997). Moreover, the adsorption isotherms of on Al₂O₃/CA were consistent under three different temperatures, fitting the Freundlich isotherm model (Fig. S3 and Table S5). This behavior indicates the favorability of synthesized composites for phosphate sorption at higher temperatures. In Table S5, the Freundlich model (R² = 0.977–0.993) provided a better description than the Langmuir model (0.72–0.747), which implies the adsorption could be either monolayer or multilayer adsorption on the surfaces and the adsorption process was a heterogeneous process (Huo et al. 2019). The value of 1/n was between 0.17 and 0.18, indicating that it is easy for the phosphorus to be adsorbed by the material.

pH is a significant water chemistry factor affecting the adsorption process (Ames & Dean 1970). In this study, the influence of pH on phosphate adsorption by Al₂O₃/CA was investigated under different pH ranging from 2 to 12 (Figure 5). The phosphate removal efficiency measured was close to 100% in the pH range of 2–7, indicating no obvious pH effect on the adsorption of on Al₂O₃/CA. When pH increased to the range of 8–10, the removal efficiency decreased slightly. However, with continuous increase of pH up to 12, the removal efficiency decreased greatly to only 6.04%. The reason can be ascribed to the instability of composite in the strong alkaline solution and the competitive adsorption between OH⁻ and (Fu et al. 2018). Although the isoelectric point of the material decreased to 2.7 from 9.8 after coating CA (Fig. S4), which was not conducive to the adsorption of phosphate, it had no great effect on the adsorption of phosphate when the pH was between 2 and 8. This indicated that the removal mechanism of phosphate may not be electrostatic adsorption or physical adsorption.

In the natural water, varieties of anions usually co-exist with ions, mainly including Cl⁻, ⁠, and ⁠. In this study, their effects on the phosphate adsorption were investigated with the concentration of 5–60 times of the initial concentration of phosphate (Figure 6). The results showed that the presence of ⁠, Cl− and did not affect the removal efficiency of at the experimental concentrations. However, the existence of did cause a decrease of removal efficiency of on Al₂O₃/CA with increase of the concentration of ⁠, suggesting that may form complexes on the surface of Al₂O₃/CA via chemisorption and the presence of will compete with for adsorption sites, thereby affecting the adsorption of ⁠.

Each Al₂O₃/CA composite with same weight was mixed with quartz sand at a ratio of 1:5, respectively, and filled in a column at room temperature. The influent with an initial phosphorus concentration at 2.5 mg/L was fed into the column with a flow rate of 3 mL/min. The effluent samples were collected into a 10 mL centrifuge tube for the measurement of concentration after adsorption. As shown in Figure 7(a) and Table 1, when m_SA:m_Al2O3 ratios were 1:50, 1:25, 1:17, 1:10 and 1:5, the breakthrough times (t_b) were 300 min, 130 min, 90, 60 and 80 min with the corresponding exhaustion time (t_s) of 9,033 min, 4,262 min, 3,470 min, 3,485 min and 3,380 min, respectively. The breakthrough time and exhaustion time were not significantly changed with increasing dosage of sodium alginate, except the ratio of 1:50. The reason could be attributed to that there is a limit to the amount of CA that can be carried on the surface of activated alumina, even if the amount of CA is increased, that is the CA coated by activated alumina has reached saturation (Ames & Dean 1970; Tanada et al. 2003). Considering the performance and cost of sodium alginate, thus, the Al₂O₃/CA composite with a ratio of 1:17 was used for the rest of experiments.

In this study, 2, 4 and 6 g Al₂O₃/CA was uniformly mixed with quartz sand at a ratio of 1:5, which was then filled in the column, corresponding the bed height at 5.2 cm, 8 cm or 12.5 cm, respectively. The influent with an initial phosphorus concentration at 2.5 mg/L was fed into the column with a flow rate of 3 mL/min. In Figure 7(b), the breakthrough time were 60, 300 and 930 min with corresponding exhaustion time of 1,489 min, 4,947 min and 7,050 min, respectively. The increase of bed height led to the longer breakthrough time from 60 to 930 min and EBCT increased from 5.44 min to 13.08 min as shown in Table 1. These results suggest that increased adsorption column bed height, just as increase the amount of adsorbent, is favorite to the removal of (Long et al. 2011). When the bed height raised from 5.2 cm to 12.5 cm, the length of the MTZ increased, suggesting more adsorption sites, and consequently enhanced the phosphate adsorption capacity on Al₂O₃/CA from 1.40 mg/g to 4.67 mg/g. However, the more adsorbent fills it, the more the cost applied. Therefore, 2 g Al₂O₃/CA corresponding the bed height at 5.2 cm was used for the rest of experiments.

The effect of flow rate on phosphate adsorption of on Al₂O₃/CA was examined with various flow rates of 3, 5 and 10 mL/min, respective. The column was operated with an initial phosphate concentration of 2.5 mg/L and a bed height of 5.2 cm. As depicted in Figure 7(c), increase of flow rates significantly shortened the breakthrough and exhaustion times. With an increase in flow rate from 3 to 10 mL/min, the breakthrough time was reduced from 90 to 20 min while the EBCT was reduced from 5.44 min to 1.63 min in Table 1. The contribution to the difference is that larger volume was passed through the column at higher flow rates, more phosphate adsorbed on the active sites of Al₂O₃/CA and reached saturation quickly.

The main reason for this result is that most phosphate ions occupy the adsorption site of Al₂O₃/CA when a large volume of wastewater passes through the bed at a high flow rate, making the adsorbent reach saturation quickly and become unable to treat more wastewater. At a lower flow rate, it means that the wastewater has more residence time in the adsorption column, which makes the phosphate ion have a longer contact time with Al₂O₃/CA. Therefore, at a lower flow rate, the adsorption column has a higher adsorption capacity. These conclusions are consistent with the findings of Awual (Awual & Jyo 2011) and Paudyal et al. (2013).

The effect of influent phosphate concentration on the adsorption of phosphorus on Al₂O₃/CA was explored with various concentrations of 2.48, 5.96 and 12.73 mg/L, respectively. The column was operated with a flow rate of 3 mL/min and a bed height of 5.2 cm. The results showed that the increase of influent concentration shortened the breakthrough and exhaustion time of phosphate adsorption in the columns (Figure 7(d)). From Table 1, the breakthrough time were 90, 60 and 30 min for initial concentrations of 2.48, 5.96 and 12.73 mg/L, with corresponding exhaustion time of 3,470 min, 1,489 and 660 min, respectively, indicating that the adsorption process becomes progressively faster and shorter with the increasing phosphate concentration. The differences could be attributed to following reasons: (1) the higher concentration, the more mass of phosphate adsorbed on Al₂O₃/CA during the same period, which led to faster breakthrough and exhaustion; (2) the higher concentration, the higher mass transfer coefficient or diffusion coefficient, which provided a greater driving force to overcome mass transfer resistance, hence the adsorption achieved saturation more quickly (Hano et al. 1997).

As the amount of CA, the influent flow rate and concentration of P increased, the saturation concentration N₀ decreased and the kinetic constant K_AB increased, as shown in the Table 2. Conversely, as the bed height increased from 5.2 cm to 12.5 cm, N₀ increased from 0.11 to 0.36 mg/L and K_AB decreased from 18.5 to 2.0 L/mg. The R² values fitted by the Adams-Bohart model were all below 0.9, indicating that the model could not describe the dynamic sorption behavior of Al₂O₃/CA on very well.

The Thomas rate constant, K_Th, increased with increasing calcium alginate amount and influent flow rate, and decreased with increasing bed height, as shown in the Table 3. Thomas dynamic adsorption capacity q₀ decreased with increasing calcium alginate volume and influent flow rate, and increased with increasing bed height. Therefore, it is more conducive to the dynamic adsorption of by Al₂O₃/CA under the conditions of low flow rate and high bed height.

The τ value predicted by the Yoon-Nelson model was very close to the actual experimental τ value in Table 4. With the increase of the dosage of calcium alginate, influent flow rate and influent phosphate concentration, the values of k_YN increased while the values of τ decreased. Additionally, the opposite trend was observed with the increasing of bed height. This may be due to the rapid saturation of the Al₂O₃/CA at higher flow rates and initial concentrations. The R² of Yoon-Nelson model was mostly above 0.9, indicating that the Yoon-Nelson model is more suitable to describe the dynamic adsorption behavior of on Al₂O₃/CA than Adams-Bohart model and Thomas model. This provides a certain reference value for industrial application.

In this study, we synthesized and explored the Al₂O₃/CA composite as effective adsorbent to remove phosphate from water. Compared to Al₂O₃, Al₂O₃/CA composite showed the advantages with larger specific surface area (increasing from 37.31 m²/g to 51.73 m²/g) and better adsorption capacity of phosphate, and easier settle and separation from water. Within 12 h, the removal rate for 2.5 mg/L of phosphorus reached 100%. In addition, Al₂O₃/CA composite demonstrate efficient removal of phosphorous from water in the continuous flow through fixed-bed column, which overcome the problem associated with active alumina fine powder because of its larger particle size. Moreover, the adsorption performance in the column was significantly influenced by the amount of outer calcium alginate, bed height of adsorbents, influent flow rate, and influent phosphate concentration. The optimal conditions for Al₂O₃/CA dynamic adsorption of were low concentration, low influent flow rate and high bed height (2.48 mg/L, 3 mL/min and 12.5 cm). In summary, this study provides a new effective adsorptive material to remove phosphate from water in continuous flow system, which also enhance the solid separation after treatment.

This work is financially supported by the Science and Technology Major Project of the Bureau of Science and Technology of Xiamen (Grant No. 3502Z20191012), the Key Projects of Enterprise and University Cooperation in Fujian (Grant No. 2018Y4010), Quanzhou City Sciences & Technology Program of China (Grant No. 2018C082R) and Subsidized Project for Postgraduates’ Innovative Fund in Scientific Research of Huaqiao University. Ming-Lai Fu also acknowledge support by the Start-up Foundation from Huaqiao University (20BS109).

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