Materials

Sodium pyrophosphate (Na₄P₂O₇), dibasic sodium phosphate (Na₂HPO₄), sodium dihydrogen phosphate (NaH₂PO₄) and SHMP (Na₆P₆O₁₈) were all procured from Shanghai Aladdin Biochemical Technology Co., Ltd (Shanghai, China). Corundum (Al₂O₃) was purchased from Shanghai Macklin Biochemical Co., Ltd (Shanghai, China). All of the materials were of analytical grade and were used exactly as received. MB (C₁₆H₁₈ClN₃S·3H₂O) of indicator grade was purchased from Xilong Scientific Co., Ltd (Guangzhou, China). Standard Pal samples (GBW(E)070055) with a purity of 85 wt.% were purchased from the Institute of Soil Science, Chinese Academy of Sciences (Nanjing, China). The raw Pal was acquired from Gansu Western Attapulgite Research and Application Institute (Baiyin, China) and passed through a 200 mesh sieve before use.

Pal purification by the S-RLFR separation strategy

The Na⁺ in SHMP can effectively replace Mg²⁺ or Ca²⁺ for complex flocculation. HMP^– can be adsorbed onto the surface or edge of Pal, which not only increases the negative surface charge but also promotes steric hindrance between clay fibres and clay bundles. Therefore, three sodium hydrogen phosphate salts (NaH₂PO₄, Na₂HPO₄ and Na₄P₂O₇), which have similar elemental structures to SHMP, were used to investigate the effect of anions with different charges on the surface potential and dispersion of Pal.

The modified Pal (mPal) consisted of 20.0 g of raw Pal dispersed in 200 mL of NaH₂PO₄, Na₂HPO₄, Na₄P₂O₇ or SHMP solutions at the indicated concentrations (2,4,6 and 8 wt.%), and this was then stirred for 0–8 h at room temperature.

For the homogenized Pal (hPal), the mPal suspension was homogenized in the RLFR at 2000–4000 rpm for 5 min and subjected to ultrasonication for 30 min.

For the purified Pal (pPal), the hPal suspension was subjected to a separation process that included several centrifugations at 500–4000 rpm to remove impurities. Finally, the Pal samples were collected by centrifugation at 10 000 rpm for 5 min.

The unhomogenized Pal (uPal) was prepared in a similar manner but was obtained before the homogenization step.

For further experiments, all samples were dried at 60℃ and ground for subsequent use. Upon the purification of Pal, the effects of dispersant dosage, modification time, RLFR rotation speed, shearing time and slit width were determined. The specific contents of Pal and pPal are compared in the Supplementary Material.

Pal purity determination

Pal purity was determined using an internal standard method (Hubbard & Snyder, Reference Hubbard and Snyder1988). Corundum and silica were chosen as the internal standard substance and filler material, respectively. Standard Pal samples of various purities (15–85 wt.%) were prepared by blending the standard Pal sample of 85 wt.% purity and silica with 50 wt.% corundum. The specific sample compositions are shown in Table S1. The X-ray diffraction (XRD) traces of the samples were collected, and the integral intensity ratios of Pal (8.495°2θ) and corundum (26.749°2θ) were calculated and plotted against the purity of Pal to obtain the standard curve.

Adsorption studies

We prepared 100 mg L^–1 MB solutions using deionized water and adjusted to pH 6.5 with 0.1 M HCl and NaOH. Pal samples of 25 mg were mixed with 50 mL of MB solution in a shaker at 25℃ for 12 h. The adsorbent was filtered through a 0.45 μm membrane after absorption, and the residual MB concentration was immediately determined using an ultraviolet–visible light (UV–Vis) spectrophotometer at λ_MB = 664 nm (Xiong et al., Reference Xiong, Li and Hu2020). The MB concentration difference after adsorption was used to calculate the amount of MB adsorbed by Pal. The amount of MB adsorbed at time t, q_t (mg g^–1), was calculated using Equation 1:

(1)\begin{equation}{q_t}\, = \,\frac{{({C_0} - {C_t}) \times V}}{m}\,\,\,\,\,\,\,\,\,\end{equation}

where C ₀ and C_t are the MB concentrations (mg L⁻¹) at the start and at time t (min), respectively, V is the volume of the solution (L) and m is the mass of Pal (g).

The MB adsorption isotherm was determined by mixing 25 mg Pal samples with a series of varying concentrations of 50.0 mL MB solution ranging from 50 to 800 mg L^–1 in Erlenmeyer flasks and after shaking the flasks for 12 h. The equilibrium concentration of MB after 12 h of shaking was determined using a UV–Vis spectrophotometer.

Characterization

An XRF-1800 X-ray fluorescence (XRF) spectroscopy device (Shimadzu, Japan) was used to determine the Pal composition. A ZEN3600 Nano-ZS model Zetasizer (Malvern, UK) was used to measure zeta-potential. The crystalline structure and purity of Pal were determined using an XRD device (XRD-6000 diffractometer, Shimadzu, Japan) with a copper target (λ = 0.1542 nm), two-axis goniometers, a graphite monochromator and a scintillation detector at a voltage of 40 kV and a current of 30 mA. The XRD traces were recorded at the scan range of 3–70° with a scanning rate of 2°2θ min^–1, a divergence slit of 1°, a scattering slit of 1° and a receiving slit of 0.15 mm. Fourier-transform infrared (FTIR) spectra were obtained using a Vector 22 spectrophotometer (Bruker, Germany) with KBr pellets in the 4000–400 cm^–1 range. The specific surface areas and pore properties were calculated from N₂ adsorption–desorption isothermal curves recorded at 77 K on a 3QDS-MP-30 specific surface aperture analyser (Quantachrome, USA). A 2501PC UV–Vis spectrophotometer (Shimadzu, Japan) was used to measure the concentration of MB.

Characterization of raw Pal

XRF was applied to determine the composition of raw Pal with a purity of 5.74%, and the specific contents are presented in Table S2. It can be observed that the raw Pal is primarily composed of SiO₂, Al₂O₃, MgO and CaO. The extremely high SiO₂ content indicates that the raw Pal contains abundant quartz impurities. When considering the overlapping peaks in the XRD trace of the raw Pal, Jade analysis was employed to examine this trace using numerous standard cards. Figure 2 illustrates that, apart from Pal (PDF#31-0783), various impurities such as mica (PDF#02-0227), quartz (PDF#85-0865), dolomite (PDF#75-1654), clinochlore (PDF#29-0701), muscovite (PDF#74-1392) and feldspar (PDF#89-8572) are present in the raw Pal sample, indicating the low grade of the raw Pal.

Figure 2. XRD traces of the raw Pal and its impurities.

Effect of modification

Figure S1 displays the standard XRD traces of the Pal samples with varying purity levels, and a linear purity standard curve (y = 2.1528x + 0.1656, R ² = 0.9968) was established for the purity analysis of Pal (Fig. S2), where x and y represent Pal purity and the integral intensity ratios of Pal (8.495°2θ) and corundum (26.749°2θ). The efficacy of various phosphate dispersants on modification was tested using raw Pal with a purity of 5.74%. As shown in Fig. 3a, uPal and hPal exhibit more negative zeta-potential values compared with the raw clay, illustrating that the introduction of dispersants can promote the ion exchange and dispersion of Pal. NaH₂PO₄, Na₂HPO₄ and Na₂HPO₄ only had a relatively limited effect on the surface charge enhancement of Pal. However, both Na₄P₂O₇- and SHMP-modified Pal exhibited comparably high negative surface charges. Therefore, hPal modified with Na₄P₂O₇ and SHMP was selected for impurity removal research. The pPal modified with SHMP was of greater purity than the pPal modified with Na₄P₂O₇ as dispersants containing high-valence anions tend to increase the repulsive forces on Pal (Xu et al., Reference Xu, Wang and Wang2013). As a result, SHMP was selected as the preferred dispersant for subsequent modification.

Figure 3. (a) Influence of various dispersants on the zeta-potential values of uPal and hPal. (b) Zeta-potential variation tendency of hPal and pPal with SHMP modification time. (c) Effect of modification time on the purity of pPal. (d) Dispersant-assisted purification mechanism of Pal.

The effect of modification time was investigated, and relationships between surface charge and Pal purity were established. Notably, the zeta-potential values were found to be unaffected by the Pal concentration within the range of 0.25–2.00 g L^–1 (Fig. S3). As depicted in Fig. 3b, the negative potential increases as the modification time increases from 0 to 3 h. The zeta-potential values of hPal and pPal remained essentially unchanged when the modification time was extended to 3–6 h. However, a decrease in negative potential was observed after a modification time of 6 h. Compared with the hPal samples, the pPal samples exhibited more negative zeta-potential values due to their greater purity. As the pPal with a modification time of 4 h had the greatest purity (Fig. 3c), this following modification was carried out for 4 h. Figure 3d depicts a schematic diagram of the purification mechanism. The presence of negatively charged Pal resulted from the isomorphic substitution of metal ions, structural defects and the appearance of surface residual valence bonds (Krekeler & Guggenheim, Reference Krekeler and Guggenheim2008). To maintain the charge balance, exchangeable cations were present either in the channels of the Pal crystals or on the crystal surface, constructing the double electron layer on the surface of Pal (Xie et al., Reference Xie, Zhang, Mu, Tang and Zhang2020). As the modification time increased from 0 to 3 h, increasing numbers of negatively charged phosphates from the dispersant penetrated the diffusion layer of the double electron layer, resulting in a gradually more negative zeta-potential. With the longer modification time (3–6 h), the ion exchange between phosphates and Pal reached an equilibrium, resulting in relatively stable zeta-potential values. When the modification time exceeded 6 h, more ions infiltrated into the double electron layer, causing an expansion of its thickness, accompanied by a reduction in zeta-potential.

Effect of SHMP dosage

To investigate the optimal dosage of SHMP, raw Pal with a purity of 5.74% was modified by SHMP with various mass fractions of 2, 4, 6 and 8 wt.%. Subsequently, one-step centrifugation was performed on the modified clay samples (Fig. 4a). At very low dispersant concentrations (2 wt.%) uniformly dispersed samples cannot form. pPal modified with 6 wt.% SHMP exhibits a relatively low purity compared to those modified with 4 and 8 wt.% SHMP. This trend aligns with the trend in surface potential, indicating a positive effect of repulsive force on purification. As pPal modified with 4 wt.% SHMP has comparable purity but a higher surface area than the sample modified with 8 wt.% SHMP (Figs S4 & S5 & Table S3), an SHMP dosage of 4 wt.% was selected for modification. After modification and corresponding purification, the pPal samples demonstrated a greater proportion of mesopores in comparison to raw Pal, indicating the effectiveness of the purification process on the removing of impurities from (i.e. dredging) the pores of Pal, enhancing the specific surface area and eliminating embedded impurities. These improvements are advantageous as they enable high-value applications of Pal in adsorption-related fields. Additionally, we conducted an investigation into the phase structure of pPal (Fig. 4b) and raw Pal (Fig. 2) in order to characterize the impacts of the purification process. It can be observed from the XRD trace of pPal (Fig. 4b) that impurities such as quartz, clinozeolite and dolomite were significantly reduced. Nevertheless, the characteristic peaks of Pal remained weak because the associated minerals could not be completely separated by one-step centrifugation. Figure 4c depicts the FTIR spectra of standard Pal, pPal and raw Pal. The stretching vibrations of (Al)O–H and (Mg)O–H bonds are responsible for the broad bands at 3617–3543 cm^–1 (Yousefi et al., Reference Yousefi, Ghasemi and Tajally2020; Sarma et al., Reference Sarma, Rajkhowa and Mahiuddin2022). The characteristic bands located at 3425 and 3225 cm^–1 are relevant to the stretching bands of dis-coordinated and coordinated H–O–H, respectively. Additionally, the peak located at 1630–1685 cm^–1 corresponds to the bending vibration of H–O–H (Vasylieva et al., Reference Vasylieva, Doroshenko, Vaskivskyi, Chernolevska and Pogorelov2018; Kaviya et al., Reference Kaviya, Ramakrishnan, Mohamed, Ramakrishnan, Gimbun and Veerabadran2021). The absorption band at 1441 cm^–1 is ascribed to the stretching and bending vibration of C=O bonds in carbonate. The symmetric and asymmetric stretching vibration peaks of Si–O and Si–O–Si are situated at 1033 and 794 cm^–1, respectively. The absorption band at 528 cm^–1 is attributed to the Si–O–Si bending vibrations (Hospodarova et al., Reference Hospodarova, Singovszka and Stevulova2018). These distinctive peaks were well preserved in the pPal samples, confirming that the Pal structure was intact. The carbonate asymmetric stretching vibration peak at 1441 cm^–1 almost vanished with the treatment of 4 and 8 wt.% SHMP, indicating that the carbonate in raw Pal can be removed with a single centrifugation step.

Figure 4. (a) Influence of SHMP dosage on the purity and zeta-potential of Pal. (b) XRD traces of pPal with various amounts of SHMP. (c) FTIR spectra of standard Pal, pPal and raw Pal.

Effects of RLFR parameters

The RLFR is a type of ultra-fine pulverizing equipment including a rotor and a stator (Li et al., Reference Li, He, Tian, Huang, Wang and Li2021). When the Pal sample passes through the slit between the high-speed rotating rotor and stator, it is subjected to a strong shearing force, frictional force, high-frequency vibration and a high-speed vortex, all of which effectively disperse and disaggregate the Pal. As a result, the rotational speed, slit width and cycle number of the RLFR all play crucial roles in the dissociation and purification of Pal. As the rotation speed increased from 2000 to 4000 rpm, the purity of pPal initially exhibited an upward trend followed by a subsequent decline, as shown in Fig. 5a, because the shearing force is directly proportional to the rotation speed (Krekeler & Guggenheim, Reference Krekeler and Guggenheim2008). Thus, rotational speeds below 3500 rpm are insufficient for homogenizing the Pal, whereas rotation speeds exceeding 3500 rpm will disrupt the crystal structure of the Pal. Slit width and shearing force are inversely correlated (Brand et al., Reference Brand, Silberbauer and Kulozik2016). When the slit width is less than 690 μm, the rod structure is destroyed during repeated shearing, whereas a slit width larger than 690 μm fails to generate sufficient homogenizing forces (Fig. 5b). The relationship between cycle number and pPal purity is depicted in Fig. 5c. As the cycle number increased from 0 to 6, there was an initial increase in pPal purity, followed by a gradual decrease as the cycle number further increased from 6 to 50. This decline can be attributed to the mechanical stress-induced breakage of Pal crystal rods (Wang et al., Reference Wang, Ren, Lian, Wang, Zhao and Liu2021). To summarize, a rotational speed of 3500 rpm, a slid width of 690 μm and a cycle number of 6 yielded the best homogenization performance.

Figure 5. The effect of (a) rotational speed, (b) slit width and (c) cycle number of the RLFR on the purity of pPal.

Effect of centrifugation on the separation of impurities

The purity of pPal was further elevated based on the density difference between the impurities and Pal. By applying differential centrifugation, the impurities in the homogenized solution can be separated by optimizing the centrifugation speed (S _C) and centrifugation number (N _C; corresponding samples are labelled as S_C-N_C). The XRD traces of the impurities collected at S _C = 500–4000 rpm and N _C = 1 are shown in Fig. 6a. Although the main impurities of mica, quartz, dolomite, clinochlore, muscovite and feldspar could be separated to varying degrees, these impurities appear in all XRD traces in Fig. 6a, suggesting that separation of these impurities remained unattainable. The XRD traces of the impurities obtained at S _C = 500–4000 rpm and N _C = 3 are presented in Fig. 6b. The characteristic peaks of dolomite and feldspar disappeared when S _C ≥ 2000 rpm, reflecting that dolomite and feldspar had been successfully removed. Clinochlore, quartz and feldspar could be removed at S _C ≥ 3000 rpm and S _C ≥ 3500 rpm. Figure 6c depicts the impurities collected at S _C = 500–4000 rpm and N _C > 3. When the centrifugation was conducted at S _C = 3000 rpm and N _C = 4, Pal separation commenced, with the emergence of distinctive peaks fixed at 8.495°. However, at S _C < 3000 rpm and N _C > 4, the impurities could not be completely separated. The impurity separation rules are summarized in Fig. 6e, and suitable separation conditions can be chosen based on the Pal type.

Figure 6. XRD traces of the impurities separated under various centrifugal speeds (a) once, (b) three times and (c) more than three times. (d) XRD traces of the pPal with high purity and standard Pal. (e) Determined separation rules for low-grade Pal. The blue box highlights the optimal purification parameters.

Red Pal (RPal) was purified as an illustrative example based on the separation rule (Fig. 6d). At S _C = 2000 rpm, the purity of purified RPal (pRPal) increased from 5.74% to 60% and 70% after purification at N _C = 3 and N _C = 6, respectively. Furthermore, a differential centrifugation method was developed by combining a low-speed centrifugation (S _C = 2000 rpm, N _C = 3) and a high-speed centrifugation (S _C = 3500 rpm, N _C = 3) to effectively remove impurities with similar density and structure to Pal, resulting in a pRPal with an exceptional purity of as high as 80%. Against the reference of standard Pal, the XRD traces of these three pPal samples demonstrated good consistency and consistent characteristic peaks.

Purification of raw Pal from various mining sites

To verify the feasibility of the S-RLFR separation strategy, three kinds of raw Pal from disparate mining sites were collected. RPal, black Pal (BPal) and off-white Pal (OPal) are ion-rich, carbon-rich and aluminium-rich Pals with purities of 5.74%, 11.60% and 7.60%, respectively (Fig. S6). After primary centrifugation at S _C = 2000 rpm and N _C = 1, the purity of the corresponding pRPal, pBPal and pOPal samples increased to 33.4%, 35.1% and 38.1%, respectively, representing increases of 5.8, 5.0 and 3.0 times, respectively, and demonstrating the universality of the S-RLFR separation strategy (Fig. 7a). As shown in Fig. 7b, the content of Pal in raw clays was quite low, and the main impurities were all quartz and mica. In contrast, the XRD peaks of quartz and feldspar were barely discernible in the pPal traces, while the intensity of the Pal peaks was significantly increased (Fig. 7c). The purification effect was studied further in terms of structure and morphology (Fig. S7). After the S-RLFR separation strategy, quartz blocks were separated from the raw clay, while wrapped Pal crystal rods in the original bundles were released and underwent a reduction in length due to the high-speed shearing force exerted by the RFLR. Given their analogous layered silicate structure (Smirnov et al., Reference Smirnov, Deryagina, Afanasieva, Rudmin and Gursky2020), there was still a large number of flake-shaped mica and muscovite particles in the pPal, which is consistent with the XRD trace.

Figure 7. (a) Purity of RPal, BPal and OPal before and after the S-RLFR separation strategy. XRD traces of RPal, BPal and OPal (b) before and (c) after the S-RLFR separation strategy. (d) N₂ adsorption–desorption isotherms, (e) pore diameter distributions and (f) MB adsorption abilities of pRPal, pBPal and pOPal.

The low-temperature nitrogen adsorption–desorption curves of the pPal conformed to type-IV curves with H3 hysteresis loops (Fig. 7d), and the corresponding specific surface area, pore volume and average pore diameter values are presented in Table S4. The specific surface area and internal pore diameter of pPal ranged from 121.8 to 167.0 m² g^–1 and from 2 to 30 nm, respectively (Fig. 7e). Micropores were formed due to Pal’s close packing, while mesopores were formed as a result of Pal aggregation (Al-Futaisi et al., Reference Al-Futaisi, Jamrah and Al-Hanai2007). pPal’s adsorption performance for MB was further investigated, and the results are shown in Fig. 7f. Compared to the raw Pal, pPal demonstrated significantly greater adsorption capacity for MB, which was directly proportional to its purity and specific surface area, and the pOPal, with the greatest purity, showed a maximum equilibrium adsorption capacity of 189.9 mg g^–1.

Adsorption properties of Pal for MB

To evaluate the effect of purity on adsorption properties, raw Pal and two types of pPal centrifugated using the S-RLFR separation strategy, labelled as Pal, pPal-1 and pPal-2 with purities of 10%, 25% and 32%, respectively, were adapted to adsorb MB. The adsorption kinetics were evaluated using the pseudo-first-order (Lagergren, Reference Lagergren1898) and pseudo-second-order models (Ho & McKay, Reference Ho and McKay1999) expressed by Equations 2 & 3:

(2)\begin{equation}\ln \left( {{q_{\text{e}}} - {q_t}} \right) = \ln {q_{\text{e}}} - {{\text{k}}_1}t\,\,\,\,\end{equation}

(3)\begin{equation}\frac{t}{{{q_t}}}{ } = { }\frac{1}{{{{\text{k}}_2}q_{\text{e}}^2}}{ } + { }\frac{t}{{{q_{\text{e}}}}}{\text{ }}\end{equation}

where q _e denotes the equilibrium adsorption quantity (mg g^–1), q_t denotes the removal capacity (mg g^–1) at time t, and k₁ (h^–1) and k₂ (h^–1) represent the pseudo-first-order and pseudo-second-order adsorption rate constants, respectively. Figure 8a, b & Fig. S8 demonstrate the non-linear and linear fitting results of pseudo-first-order and pseudo-second-order models towards MB, with the fitting parameters listed in Table 1 & Table S5. According to the results, the fitting correlation coefficient (R ²) of the pseudo-second-order kinetic model is greater than that of the pseudo-first-order kinetic model. In addition, the adsorption amount fitted by the pseudo-second-order kinetic model (q _e,cal) is closer to that obtained by the experiment (q _e,exp), suggesting that the adsorption behaviour of both raw Pal and pPal towards MB is attributable to chemical adsorption.

Figure 8. (a) Pseudo-first-order and pseudo-second-order kinetic non-linear fitting of Pal, pPal-1 and pPal-2 towards MB. (b) Pseudo-second-order kinetic linear fitting of Pal, pPal-1 and pPal-2 towards MB. (c) Langmuir and Freundlich isotherm non-linear fitting of Pal, pPal-1 and pPal-2 towards MB. (d) Langmuir isotherm linear fitting of Pal, pPal-1 and pPal-2 towards MB.

Table 1. Pseudo-first-order and pseudo-second-order kinetic linear fitting results of Pal, pPal-1 and pPal-2 towards MB.

The Langmuir (Langmuir, Reference Langmuir1918) and Freundlich (Hall et al., Reference Hall, Eagleton, Acrivos and Vermeulen1966) isotherms are applicable to adsorption on homogeneous and heterogeneous surfaces separately, as shown in Equations 4 & 5:

(4)\begin{equation}\frac{{{C_{\text{e}}}}}{{{q_{\text{e}}}}}{ } = { }\frac{{{C_{\text{e}}}}}{{{q_{{\text{max}}}}}}{ } + { }\frac{1}{{{{\text{K}}_{\text{L}}}{q_{{\text{max}}}}}}{\text{ }}\end{equation}

(5)\begin{equation}\ln {q_{\text{e}}} = \ln {{\text{K}}_{\text{F}}} - { }\frac{1}{n}{\text{ ln}}{C_{\text{e}}}{\text{ }}\end{equation}

where q _max represents the maximum adsorption quantity, C _e is the equilibrium MB concentration, n is the adsorption intensity, and K_L (L mg⁻¹) and K_F (mg g⁻¹) are the Langmuir and Freundlich parameters, respectively. The non-linear fitting results of the two isotherm models are plotted in Fig. 8c, and the linear fitting results of the models are shown in Fig. 8d & Fig. S9, respectively. The corresponding fitting parameters are summarized in Table 2 & Table S6. The Langmuir adsorption model demonstrates a higher correlation coefficient (R ²) for MB adsorption onto Pal, indicating that the adsorption of MB onto Pal is more likely to be in a monolayer. The adsorption performance of pPal for MB was compared with that of Pal adsorbents reported in the literature. As shown in Table 3, the maximum adsorption of MB by Pal and Pal’s composite materials is located at 50.8–197.8 mg g^–1, whereas the maximum adsorption of MB by pPal purified using the S-RLFR method in this work reached an adsorption capacity of up to 256.4 mg g^–1. The adsorption capacity was effectively increased by the simultaneous dispersion and purification introduced by the S-RLFR method using Pal with only 32% purity. This result suggests that the S-RLFR separation strategy is an effective method for increasing the value of low-grade Pal.

Table 2. Langmuir and Freundlich isotherm linear fitting results of Pal, pPal-1 and pPal-2 towards MB.

Table 3. Comparison of the MB adsorption capacities of various Pal adsorbents.

Based on the analysis of the above findings, it can be inferred that the increases in the dispersion and purity of Pal contribute to its improved adsorption capacity for MB. Therefore, the process of adsorption of MB by Pal may be specified as follows: firstly, the introduction of dispersants can effectively change the surface charge of Pal and increase the repulsive forces between Pal particles (Wang & Wang, Reference Wang and Wang2016), inhibiting the aggregation of Pal (Fig. 3a,b). Subsequently, the high-speed shear, impact and dispersive forces provided by the RLFR further improve the dispersion of Pal (Fig. S10). The highly dispersed Pal nanorods have a relatively large surface area and high surface activity (Figs S3 & S4 & Table S3), which facilitate the adsorption process of dye molecules. In addition, the increase in dispersion facilitates the purification process of Pal. As can be seen from the comparison of the results in Fig. 3b,c, the surfaces of Pal with varying degrees of purification carry varying degrees of negative charge. This may be due to the ion-exchange process that brings the phosphate into the double electron layer (Figs S11 & S12 & Tables S2 & S7). Due to electrostatic interactions, negatively charged Pal is then more favourable for the adsorption of positively charged MB molecules (Duan et al., Reference Duan, Zhu, Zhang, Lu and Wang2023). In addition, the M–OH on the surface of Pal can form intermolecular hydrogen bonds between the N and S of MB, which further facilitates the adsorption of MB (Fig. 4c).