Adsorption of methyl orange on ZnO supported by seawater-modified red mud

Nowadays, large quantities of dye effluents produced from textile, paper, printing, leather, plastics, leather tanning industrial activities are released into environment, leading to harmful effects on the aquatic organisms and human health (Gupta & Saleh 2013; Dil et al. 2016). It is reported that around one thousand tons of dyes are discharged in to the environment only as textile waste annually (Ajmal et al. 2014). Therefore, it is urgent to find efficient ways to treat these toxic dye effluents efficiently and economically. A wide range of technologies including adsorption, biological treatment, flotation, ozone oxidation, electrochemical, photodegradation, and chemical coagulation have been used to treat dye-rich wastewaters. Among the various treatment technologies, adsorption is attracting attention owing to its high efficiency and low cost. However, the adsorption performance of the adsorbents and the constant supply of materials determine whether the adsorption process will be successful or not (Ghaedi et al. 2012). Dyes can be classified as anionic, cationic and non-ionic (Fernandes et al. 2020). Methyl orange (MO) is a kind of anionic dye on account of the deprotonation of sulfonate on MO molecule in the aqueous solution with a complex chemical formula of C₁₄H₁₄N₃SO₃Na shown in Figure 1 (Azam et al. 2020). MO is commonly present in effluent discharges from the textile, printing, paper manufacturing, food, and pharmaceutical industries and research laboratories (Haque et al. 2010). In this work, MO was used as a target organic pollutant representative of anionic dyes to explore the dye removal capacity of adsorbents. The selection of MO is of typical in the treatment of anionic dye effluents because MO is toxic and difficult to be adsorbed (Fernandes et al. 2020).

In recent years, metal oxide nanoparticles have been commonly used as adsorbents to treat dye wastewater due to their high surface area and strong physical and chemical effects on the adsorption of dye effluents. Zinc oxide (ZnO) is a kind of common metal oxide nanoparticle as adsorbent with low cost, high chemical stability, harmlessness to the environment, and high regeneration capacity (Kataria et al. 2016). It also has a positive charge at neutral pH with an isoelectric point (IEP) in the range from 9.3 to 9.8, indicating its adsorption potential for anionic dye MO as the MO molecules, which are negative in aqueous solution were easily adsorbed onto positive surfaces under the action of electrostatic force (Kosmulski 2004). Several methods have been used to synthesize ZnO, such as precipitation method (Rodrıguez-Paéz et al. 2001), indirect electrodeposition method (Kanade et al. 2006), sol-gel assisted hydrothermal method (Zhao et al. 2014), ultrasonication (Ma et al. 2015), sonochemical method (Chen et al. 2001), and mechanochemical method (Ao et al. 2006). From these, the precipitation method was selected to synthesize ZnO in this study for its advantage of simple operation. In addition, the precursor of Zn(OH)₂ with layered structure was obtained by precipitation method, and anions (NO₃⁻, Ac⁻, SO₄²⁻) would be inserted into the layer and released during calcination, resulting in larger pores of ZnO (Hou et al. 2012). However, ZnO nanoparticles (NPs) tend to aggregate in aqueous solution and are difficult to separate and recycle due to their high colloid stability (Keller et al. 2010; Zhang et al. 2018). Nowadays, there are two approaches to solve this problem and improve the adsorption capacity of ZnO. On one hand, methods to selectively control the size, shape, and organization of materials have been studied recently, such as flower-shaped ZnO NPs (Hou et al. 2012), hierarchical hollow structure ZnO (Lan et al. 2014), hierarchical porous ZnO microspheres (Lei et al. 2017), 3D hierarchical PbS/ZnO microspheres (Liu et al. 2020), ZnO hollow microspheres (Wang et al. 2013), etc. On the other hand, porous bodies with a larger surface area are used to load with ZnO NPs, which may change the adsorbent morphology and increase the available adsorption sites of adsorbent (Madan et al. 2019).

Red mud is the waste of the alumina production process with a large specific surface area and adsorption capacity for anionic dyes (Namasivayam & Arasi 1997). Some alumina refineries have used seawater to neutralize red mud by precipitation of hydrotalcite, which can increase the specific surface area of red mud. Palmer & Frost (2009) found that the process of treating red mud with seawater facilitated a significant reduction in dissolved metal concentrations, which could improve the safety and reliability of adsorption because the metals in red mud would not cause secondary pollution to water again. In addition, neutralization of red mud by seawater (SRM) could increase the settling rates due to agglomerate consolidation, which is conducive to the separation of adsorbent and solution (Palmer et al. 2009). At present, studies on SRM have been reported, while few studies have been conducted on the applications of SRM on the adsorption of dye. Therefore, the present work aims to use SRM as substrates and ZnO NPs were deposited on their surfaces to synthesize adsorbents of SRM-ZnO and RM-ZnO. In contrast with RM-ZnO, the prepared SRM-ZnO exhibited stronger adsorption capacity of MO in aqueous solution. SRM-ZnO was characterized via scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), and X-ray diffraction (XRD) analysis. In addition, the influences of pH, dosage, and anions on the adsorption of SRM-ZnO on MO were investigated to evaluate the adsorption performance and analyze the mechanism of the adsorption.

Red mud was provided from an electrolytic aluminium factory in China, and seawater was obtained from Qiangdao, China. The major chemical composition of red mud is presented in Table 1 and the main parameters of raw seawater are presented in Table 2 (Song et al. 2011). There may be complexation of DOC with Zn²⁺ during the preparation of adsorptions, which could increase the solubility of Zn²⁺ and prevent the conversion of Zn²⁺ from dissolved fraction into particulate fractions. However, it can be seen from Table 2 that the value of COD_Mn is low, so the value of DOC is also low and has little influence on the morphological transformation of Zn²⁺ (Shafer et al. 1999; Chakraborty & Chakrabarti 2006). In addition, the other components in seawater has little negative impact on the preparation of adsorbent and raw seawater can be used directly in subsequent test. Zinc acetate dihydrate (Zn(Ac)₂·2H₂O), sodium hydroxide (NaOH), hydrochloric acid (HCl), and MO were purchased from Sinopharm Chemical Reagent Co. LTD. Purified water (UNIQUE-R20 purification system, Xiamen, China) was used in the experiments.

The synthesis method of ZnO supported by seawater modified red mud (SRM-ZnO) is as follows: a weight of 2 g of red mud sifted through 100 mesh was dispersed into 250 mL of seawater and left to age overnight to synthesize SRM. After that, 0.4 g of Zn(Ac)₂·2H₂O was dissolved in 5 mL purified water to produce a solution of Zn(Ac)₂. The prepared Zn(Ac)₂ solution was added into the mixture as described above under stirring on a magnetic stirrer to obtain a homogenous deposition of Zn²⁺ on the surface of the solid. Then the pH of the solution was adjusted to above 10 to precipitate the Zn²⁺. After being stirred for 1 h at room temperature, the precipitates were washed with deionized water, then dried at 80 °C and calcinated at 400 °C for 2 h. After the above steps, SRM-ZnO was fabricated for the adsorption experiments.

To further understand the adsorption mechanism of SRM-ZnO, we also synthesized ZnO supported by red mud (RM-ZnO), pure ZnO and SRM for reference. In a similar way, for the preparation of RM-ZnO, 2 g of sifted red mud was left in 250 ml purified water overnight and the solution with red mud was processed as above to obtain RM-ZnO. For the preparation of ZnO, a solution of 0.1 M Zn(Ac)₂ was added dropwise to 0.5 M of NaOH solution on the magnetic stirrer. After stirring for an hour, the white precipitates of Zn(OH)₂ formed were separated from the liquid, washed with deionized water, and then dried at 80 °C and calcinated at 400 °C for 2 h to obtain pure ZnO. For the preparation of SRM, 2 g of red mud was added into 250 mL of seawater overnight, and then the solids were calcinated at 400 °C for 2 h. The schematic diagram for synthesis process of SRM-ZnO, RM-ZnO, ZnO, and SRM is shown in Figure 2.

A series of experiments were conducted in the batch form to investigate the effect of different factors in the adsorption of MO on adsorbents. All experiments were carried out using glass sample bottles. In a typical procedure, the desired dosage of adsorbent was dispersed into 25 ml MO solution under different condition such as pH, initial concentration of dyes, contact time, and temperature. The solution was vibrated in the dark at 180 rpm for the predefined time and then centrifuged for solid−liquid separation at 4,000 rpm for 5 min. The absorbency of MO over each sample in the supernatant was monitored using a UV-Vis spectrophotometer at a maximum wavelength of 463 nm. It was used to determine the residual concentration of MO. To achieve statistical soundness, experiments were replicated at least three times and the data were average values.

X-ray diffraction (XRD) patterns were conducted from Smartlab 9 kw XRD of Nippon Science Corporation using Cu Kα radiation at 45 kV and 200 mA with a scanning range of 5°–80° and a step size of 0.02°. The morphology observation and elemental composition of the prepared adsorbents were detected with a scanning electron microscope (Zeiss Sigma300, SEM) with an acceleration voltage of 5 kV and an energy dispersive spectroscopy (Ultim Max 40, Oxford, EDS). The samples were previously loaded with a thin gold layer using a fine coater (ETD-2000).

As can be seen in the pattern of SRM-ZnO, the intensity of these diffraction peaks concerning the hydrotalcite is low, and even some diffraction peaks corresponding to hydrotalcite cannot be detected. The phenomenon can be explained by the following reasons, (i) The hydrotalcite compounds were destroyed and converted to Al₂O₃ and MgO during calcination of 400 °C so that there were less hydrotalcite compounds in SRM-ZnO in solid state (Orthman et al. 2003); (ii) The concentration of OH⁻ and CO₃²⁻ decreased during the formation of hydrotalcite, which led to the fact that the pH dropped from 11 to 9 and it was difficult to dissolve Al from red mud into Al(OH)₄⁻ in an alkalescent environment. Therefore, the crystallinity of hydrotalcite was poor and overshadowed by the sharper and more crystalline mineralogical phases in red mud (Palmer & Frost 2009).

To evaluate the textual morphology of surfaces of adsorbents, the micrographic images of RM-ZnO and SRM-ZnO were depicted by the scanning electron microscope (Zeiss Sigma300). Furthermore, the elemental compositions of these two adsorbents were measured by EDS analysis. Typical SEM images of RM-ZnO and SRM-ZnO and the EDS spectra of SRM-ZnO are shown in Figure 3(b) and 3(c). Besides, Table 3 shows the weight percentage of Mg, Ca²⁺, K⁺ and Zn elements in corresponding zones of RM-ZnO and SRM-ZnO, respectively. It can be seen clearly in Figure 3(b) and 3(c) that the surfaces of both RM-ZnO and SRM-ZnO are uneven, while the surface of SRM-ZnO is coarser with higher pore density and heterogeneity. What is noteworthy is that the raised parts of SRM-ZnO surface are connected by reticular aggregates, as shown in the red curve in Figure 3(b), which may be related to the addition of RM to seawater and these reticular aggregates may be hydrotalcite compounds. According to relevant literatures, the fine mineral particles flocculated into large agglomerates by Mg²⁺, which also formed electrostatic bridges in seawater and acted as nucleation sites for the precipitation of hydrotalcite (Hanahan et al. 2004; Palmer et al. 2009). This finding can be confirmed by the higher content of Mg element in SRM-ZnO than RM-ZnO in Table 3. In addition, the EDS spectra taken from the surface of SRM-ZnO verify the presence of Zn element, and the distribution of Zn similar to Si, which comes from RM, suggests the even distribution of ZnO. However, the typical flake structure of ZnO could hardly be observed in SEM images of RM-ZnO and SRM-ZnO, confirming that the aggregation was prevented by RM and SRM. As can be seen from Table 3 that the weight percentage of the Zn element on SRM-ZnO is higher than RM-ZnO, which may account for the formation of hydrotalcite that can provide more adsorption sites for the loaded ZnO.

Figure 4 shows the differences in adsorption effects of MO by different adsorbents prepared above. The adsorption capacity of MO onto SRM is limited, indicating that the hydrotalcite compounds on SRM have poor efficiency of adsorption on MO. In addition, the removal rate of ZnO is 35.28% lower than that of RM-ZnO and even 73.18% lower than that of SRM-ZnO, indicating that ZnO plays an important role in the adsorption of RM-ZnO and SRM-ZnO. And according to the result from XRD, the size of ZnO particles on RM-ZnO and SRM-ZnO is smaller than that of ZnO, so the specific areas of RM-ZnO and SRM-ZnO are higher. RM and SRM might be appropriate for providing the suitable conditions for the growth of ZnO and hindering the aggregation of ZnO particles. The SRM-ZnO is observed distinctly to have the best effect on the adsorption of MO and get a removal ratio of 84%, indicating that the SRM-ZnO can be used to remove MO dye as adsorbent. Hydrotalcite formed in the process of seawater immersion was found in SRM-ZnO by XRD as above. The hydrotalcite compounds lost their structure and converted to Al₂O₃ and MgO during calcination of 400 °C. However, the destroyed hydrotalcite compounds on the SRM-ZnO could restore by adsorbing H₂O molecules through hydrogen bonding in the aqueous solution, which is known as the ‘memory effect’ (Palmer et al. 2009). ZnO plays a major role in the adsorption of MO onto SRM-ZnO and the hydrotalcite compounds play a secondary role that remove MO through adsorption onto its external surfaces and intercalation into its hierarchical pore network. In addition, the hydrotalcite compounds provide a larger load area for ZnO due to their hierarchical pore networks. Besides, Rath et al. found that the incorporation of Mg²⁺ Ca²⁺ and K⁺ from seawater into the ZnO lattice could lead to a smaller crystallite size of ZnO NPs on SRM-ZnO, which may increase the specific surface area of SRM-ZnO (Hameed et al. 2013; Li et al. 2014; Rath et al. 2019). Figure 3(j) and 3(k) and Table 3 show that elements Mg and K on SRM-ZnO is higher than that on RM-ZnO. The difference of Ca from Mg and K may due to that the uneven distribution of Ca in RM can lead to the measurement error of Ca. However, the effect of these metal ions on the crystallite size of ZnO on SRM-ZnO is difficult to be reflected by XRD due to the low crystallization and small content of ZnO on SRM-ZnO. According to the analysis above, the schematic diagrams of RM-ZnO and SRM-ZnO in solution are shown in Figure 5.

The influence of pH on adsorption of MO was from pH 4 to 12, while maintaining the other parameters constant. The results depicted in Figure 6 show that the percentage of MO removal remains constant up to pH 9. The sulfonate group in the MO molecule is deprotonated to -SO₃⁻, resulting in the MO molecule being negatively charged in aqueous solution. In the meanwhile, the adsorbent has a positive surface since the oxygen atoms in hydrotalcite and ZnO are protonated, which generates enhanced electrostatic attraction between MO and adsorbent. However, the removal of MO decreases at pH 10–12. On the one hand, at higher pH, more negatively charged sites existed on the surface of adsorbent, which is adverse for the adsorption of MO due to electrostatic repulsion. On the other hand, there are more OH⁻ ions that compete with anionic dye for the finite adsorption sites in the alkaline aqueous solution. It could also be observed from Figure 6 that the removal of MO decreases slightly at pH 4. Three major reasons could account for this phenomenon: (i) There is increased protonation of nitrogen on MO, which can weaken the electronegativity of sulfonic acid groups on MO and affect the electrostatic interactions between adsorbent and MO (Dirksen et al. 2002); (ii) Part dissolution of ZnO NPs may take place in acidic solution, reducing adsorption sites (Zafar et al. 2019); (iii) A high concentration of H⁺ can prevent strong ionization of Na⁺ of MO, which results in a decline in adsorption of MO. It could be obviously found that it still has adsorption capacity at pH ≥ 10, indicating that the electrostatic mechanism is not the only mechanism in the adsorption of MO. SRM-ZnO may also adsorb MO through hydrogen bonding or other interaction (Al-Degs et al. 2008; Zhang et al. 2014). Because the percentage of MO removal remains constant expect for low acidic pH and high alkaline pH, initial pH was chosen for future experiments.

According to the results, an excess of adsorbent can lead to a low utility of adsorption active sites. To reduce the loss of material, the subsequent study was carried out with 1 g/L of SRM-ZnO amount as the optimal dosage.

The possibility of intraparticle diffusion is explained by the intraparticle diffusion model based on the intraparticle mass transfer mechanism. As shown in Figure 8(c), multilinearity is observed throughout the plots of q_t versus t^1/2, indicating that more than one step takes place in the adsorption. The slope of the initial linear portions is larger than that of the second portions at different concentrations, indicating the adsorption rate decreases with time. The first linear region is attributed to the diffusion of MO molecules to the external surface of SRM-ZnO, which is controlled primarily by boundary layer diffusion. The concentration of MO in the aqueous solution is higher than that in the solid particles in this region, providing an impetus to promote the adsorption process (Vimonses et al. 2009). The adsorption rate in the second linear region slows down, which is related to intraparticle diffusion within the pores of SRM-ZnO (Tajizadegan et al. 2016).

To explore other factors that can control the adsorption rate, the adsorption mechanism was investigated in terms of pseudo-first-order kinetics, pseudo-second-order kineticsm and Elovich kinetics. The linear curves of the pseudo-first-order, pseudo-second-order, and Elovich kinetics are shown in Figure 8(a), 8(b), and 8(d) and the relevant kinetic parameters are shown in Table 4. From the results of Table 4, the values of R² from the pseudo-second order kinetics are highest, indicating that the adsorption mechanism may maintain chemisorption (Ho 2006). However, the values of R² from the pseudo-second-order kinetics are also close to 1 and the values of q_{e, cal} display a good fit with experimentally observed values of q_{e, exp}, except when the concentration is 10 mg/L, reflecting that physisorption may be one of the adsorption mechanisms. The nonconformity of values at the concentration of 10 mg/L may account for the fact that it has reached final equilibrium stage until 8 h the pseudo-first-order kinetics is normally valid over the early stage of the adsorption process. Hence, the physisorption is primary in the initial stage of the adsorption process, which is controlled chiefly by boundary layer diffusion, electrostatic effect and other physical interactions, like hydrogen bonds. And the latter stage of the adsorption process is mainly dominated by chemisorption with the control by intraparticle diffusion and chemical interaction primarily (Wanyonyi et al. 2014).

To further analyze and comprehend the mechanism, adsorption isotherms were investigated in terms of Langmuir, Freundlich, and Temkin isotherm models. The values of constants and coefficients have been summarized in Table 5. According to the results, the adsorption capacity q_e and equilibrium constants K_L, K_F, K_T, and B decrease with the rising temperature, indicating that lower temperature facilitates the adsorption and the adsorption is an exothermal process. This phenomenon may account for the fact that high temperature can promote the desorption and escape of MO molecules from the surface of SRM-ZnO (Sujana et al. 1998). It is also possible that adsorption can be influenced by the thermal motion of disordered molecules with the rising temperature (Yuan et al. 2015). However, as can be seen in Figure 9, the adsorption capacities at 303 K are close to that at 313 K, indicating that the inhibition effect of increasing temperature on adsorption is limited. The difference of R² represented in Table 5 shows that the Langmuir and Temkin isotherm models can elucidate the isotherm of adsorption. Furthermore, it can be seen intuitively in Figure 9 that the Temkin isotherm has more facility to display a good fit with experimental values compared with Langmuir isotherm model. These findings suggest that the adsorption perfectly complies with the Temkin isotherm model. From the Langmuir model, R_L values were calculated and found to be from 0.03 to 0.68 in the range of 0–1, demonstrating that the adsorption is favourable at all three temperatures. The 1/n values of the Freundlich model were observed to be in the range of 0–1, indicating that the adsorption process is under favorable adsorption conditions.

Negative values of ΔG^θ show that the adsorption is spontaneous. The absolute values of ΔG^θ decrease with the rising temperature, indicating that the adsorption is an exothermal process with the reason that the value of ΔG^θ represents the magnitude of adsorption force. In addition, the absolute values of ΔG^θ fall in the range of 0–20 kJ·mol⁻¹ for physisorption, rather than the range of 80–400 kJ·mol⁻¹ for chemisorption, indicating that the adsorption is dominated by the physisorption process (Zhu et al. 2011). The values of ΔH^θ < 0 suggest that the adsorption is an exothermal process once again, and the released energy inhibits the adsorption of MO onto SRM-ZnO. Besides, the absolute value of ΔH^θ at 57.03 kJ·mol⁻¹ does not reach the 60 kJ·mol⁻¹ at which the bond breaks, confirming that the adsorption of MO onto SRM-ZnO can be defined as a physisorption process (Wang et al. 2019). The decrease of the standard entropy S^θ listed in Table 6 may accounts for the fact that the three-dimensional motion of MO molecules dispersed freely in the aqueous solution changed into the two-dimensional motion of MO molecules adsorbed on the surface of SRM-ZnO.

Figure 10(a) shows the influence of different anions on adsorption of MO onto SRM-ZnO. To reduce the influence of cations and ion strength on the adsorption, sodium salt was adopted with an anion concentration of 20 mg·L⁻¹. It can be seen in Figure 10(a) that the adsorption can be affected by PO₄³⁻ and CO₃²⁻, while the other anions (NO₃⁻, SO₄²⁻, F⁻, Cl⁻) have little effect on the adsorption of MO. The competition between PO₄³⁻ ions and anionic dye for occupying the adsorption sites on SRM-ZnO can disfavour the adsorption of MO dye with the reason that hydrotalcite compounds on the surface of SRM-ZnO have the adsorption capacity of PO₄³⁻ (Hanahan et al. 2004). In addition, a mechanism of anion exchange and a ranking of affinity for intercalation exist in the interlayer of hydrotalcite and the CO₃²⁻ has an exceeding preference for intercalation and prevents the exchange with other anions, so that the intercalation of CO₃²⁻ can reduce the number of adsorption sites on SRM-ZnO (Hanahan et al. 2004; Palmer et al. 2009). As can be seen in Figure 10(b) that the sorption percentage for MO decreases from 81.66% to 22.65% with an increasing concentration of CO₃²⁻, indicating that intercalation may exist in the adsorption of MO onto SRM-ZnO. However, according to the poor adsorption effect of MO onto SRM showed in Figure 4, intercalation is not the main mechanism in the adsorption. The effect of ionic effect on the adsorption was studied by changing the concentration of NO₃⁻ which can reduce the impact of pH changes on adsorption. As shown in Figure 10(c), the removal rate of MO decreases in the lower ionic strength range, and then goes up with increasing ionic strength. The decreased adsorption at lower ionic strength may be attributed to the competition between NO₃⁻ ions and MO molecules and the increased adsorption at higher ionic strength may be explained by the hydrophobic and the aggregation of MO molecules (Liu et al. 2001; Al-Degs et al. 2008). However, the small change in removal rate indicates the poor effect of ionic strength on the adsorption of MO onto SRM-ZnO.

The mainly adsorption mechanism of MO by SRM-ZnO is shown in Figure 11. It can be seen from above that the adsorption of MO by SRM-ZnO is mainly dominated by physisorption, and ZnO plays the most important role in adsorption. In addition, we can find that the hydrotalcites on SRM still have a small adsorption capacity of MO from the adsorption results of MO by SRM. Oxygen atoms on ZnO and hydrotalcites tend to protonate in solution and become positive charged, which could adsorb negative MO molecules by electrostatic attraction. Besides, the hydrotalcites on the SRM have reticular structure from SEM image. The hydrotalcites not only provide a larger load area for ZnO, but also remove some of the MO molecules by intercalation, which can be seen from the effect of CO₃²⁻ on adsorption.

SRM-ZnO exhibits higher adsorption capacity of MO than RM-ZnO, SRM, ZnO, and RM alone. The hydrotalcite compounds synthesized by seawater on the surface of SRM-ZnO not only improves adsorbing capacity, but also increases the surface area of red mud where more ZnO NPs could be loaded. In addition, the alkali metal ions from seawater may lead to smaller radius of ZnO atoms on SRM-ZnO, increasing the surface area of the adsorbent. The adsorption process is dominated by physisorption and supplemented with chemical adsorption. The adsorption process could be affected by PO₄³⁻ and CO₃²⁻. Besides, the ionic strength has poor effect on the adsorption and can be explained by the hydrophobicity and aggregation of MO molecules.

SRM-ZnO provides an efficient way to recycle red mud. As shown in Table 3, the content of Zn element in SRM-ZnO is only 4.19%, indicating that there is little ZnO on SRM-ZnO, and the most of the rest is SRM. Therefore, SRM-ZnO can use little Zn element and a lot of red mud. In addition, the use of SRM is economic and can reduce the harm of red mud to the environment, and ZnO also has little harm to the environment. According to the results, SRM-ZnO is an alternative adsorbent for the removal of MO from aqueous solution and is suitable for the preparation of efficient adsorption by red mud in coastal red mud dumps.

This study was funded by the Major Science and Technology Innovation Project of Shandong Province (Grant No. 2018YFJH0902) and the Taishan Scholar Foundation of Shandong Province (Grant No. tsqn201909126).

The authors declare that there is no conflict of interest regarding the publication of this paper.

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