Pyridine, hydrochloric acid, sodium hydroxide and other chemicals are analytical reagents and purchased from Aladdin Industrial Corporation (Shanghai, China). Natural shale in this study is obtained from Yiye-1 shale gas well in Yichang, Hubei province of China at the burial depth of 1835.0 m. The shale sample is crushed with a mechanical rock crusher, ground in a ceramic mill, and then sieved to the desired particle size fractions. The sample of the particle size 177 μm is washed using magnetic stirrer with the distilled water to remove any dissolved salts. After filtration, the washed sample is dried at 110˚C overnight in an electric oven (Tianjin Taisite Instrument Co., Ltd., 101-1AB). The dry shale sample is kept in a desiccator over silica gel desiccant until use in characterization and adsorption experiments.

The X-ray diffraction (XRD) technique is used to obtain the semi-qualitative mineralogical compositions of the shale samples. The samples are mechanically crushed and ground to size of <200 mesh. XRD data are collected at room temperature with a PANalytical X’Pert PRO X-ray diffractometer (PANalytical Co., Almelo, Netherlands). About 30 mg of powdered solid samples are kept in a quartz block and pressed onto the quartz block using a glass slide to obtain a uniform distribution. The 2θ Bragg angles are scanned over a range of 5 - 85 at a scanning speed of 6.25/min, using graphite monochromated Cu-K_α radiation source and a nickel filter. The tube current and the tube voltage is 30 mA and 40 kV, respectively. The experiment parameters are as followed: DS = 1/4, SS = 1/8, RS = 5.0 mm. The computer automates the data collection and data reduction steps of the analysis. To obtain a semiquantitative measurement of the mineral components of a given sample, the maximum intensity of each identified mineral is measured and compared to the standard intensity obtained from a pure mineral sample.

The FT-IR spectra are obtained in a diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) mode using the micro sampling cup of a Spectra-Tech diffuse reflectance accessory against a KBr background on a Nicolet 750 FT-IR spectrometer (Thermo Nicolet Corporation, Waltham, MA, DTGS detector; Nichrome source; KBr beam splitter). A small amount of the shale samples is finely ground, mixed with dried potassium bromide (Merck, Sharp and Dhome Ltd., Shanghai, China, spectroscopic grade) in the ratio of 1:100, and pressed into pellets for a spectrum test. The FT-IR spectra are collected and manipulated using the Thermo Electron software, OMNIC 7.0, supplied from the manufacturer of the spectrometer by averaging 32 scans at a spectral resolution of 4 cm⁻¹ under a dried nitrogen flow (10 cm³/min) condition. All spectra are smoothed using the “automatic smooth” function of the above software, which uses the Savitsky-Golay algorithm (95-point moving second-degree polynomial). After that, the baseline is corrected using the “automatic baseline correct” and the spectra scale is normalized with the “normalize scale” function. The spectroscopic region from 4,000 cm⁻¹ to 400 cm⁻¹ is used for the hierarchical cluster analysis.

Morphology of the shale samples is observed by a field emission-scanning electron microscope (FE-SEM, Model MIRA3 -XMU, TESCAN Czech Republic) operated at an acceleration voltage of 15 kV. The samples are observed under the conditions of 10.0 k and 50.0 k magnifications.

In batch adsorption experiments, accurately weighed shale samples (0.1 - 1.5 g) and 30 mL of the aqueous solution containing different initial concentrations of pyridine (2.0 - 50 mg/L) are put into a glass-stoppered flask. The flask is darkly brown-colored to prevent photoxidation. The pH of the solution is adjusted with HCl or NaOH solution by using a pH meter (Model PB-10, Shanghai Cany Precision Instrument Co., Ltd, China). The flask is subsequently capped with the glass stopper and shaken in a temperature-controlled water bath shaker (Model SHA-B, Tianjin Saidlis Experimental Analytical Instrument Factory, China) for 2 h at the temperature range of 25˚C - 45˚C. At preset contact times, the concentrations of pyridine in the solution are analyzed for residual concentration of pyridine at λ_max = 256 nm [7] using a 722N UV-vis spectrophotometer (Shanghai AuCy Scientific Instrument Co., Ltd., China). The amount adsorbed is calculated from the concentration of the pyridine solution before and after the adsorption experiments using a Beer’s law plot to interpolate concentrations. Pyridine uptake at equilibrium, q_e (mg/g), is calculated by the equation, q_e = (C₀ − C_e) V/W. The C₀ and C_e (mg/L) are the concentrations of pyridine at initial and at equilibrium, respectively. V is the volume of the solution (L) and W is the mass of dry adsorbent used (g). The calibration curves between absorbance and the concentration of the pyridine solution are established. The calibration plot of absorbance versus concentration of pyridine shows a linear variation. The standard curve is given as y = 31.895x + 0.002, R² = 0.9994, where y is the absorbance and x is the pyridine concentration.

The FT-IR spectra of the shale sample are given in Figure 1. The broad and strong band in the range of 3140 - 3620 cm⁻¹ is assigned to the –OH stretching vibration in hydroxyl groups of constitution water and interlayer water of clay minerals [12]. The peak at 1410 cm⁻¹ is attributed to the stretching vibration of C-O in the carbonate anion ( CO 3 2 − ) of calcite [13]. The intense band in the range of 1000 - 1100 cm⁻¹ is ascribed to the Si-O vibration in quartz [14]. The peaks at 515 - 795 cm⁻¹ indicate the vibration of Si-O, Al-O-H and Si-O-Al [14].

X-ray diffraction patterns of the shale sample are exhibited in Figure 2. Quantitative mineralogical analyses by XRD (Figure 2) reveal that the typical shale sample primarily consists of illite (45.94 wt%), quartz (38.05 wt%), calcite (11.28 wt%), chlorite (3.04 wt%) and sericite (1.68 wt%).

The SEM micrograph in Figure 3 is employed to evaluate the morphological structure of the shale. Figure 3 shows that the prepared the shale sample is characterized by various visible fractures and pore structures with different diameters, which may be derived from naturally and/or artificially occurring disintegration of the shale matrix. The well-developed porosity of the shale may favor pyridine removal during the actual effluent treatment process.

The study of adsorption kinetics can provide valuable insights into the uptake rate of sorbents, the adsorption mechanisms as well as the proper design and model for the adsorption process. The experimental kinetic data of adsorption of pyridines onto the shale are analyzed by pseudo-first order, pseudo-second order, Elovich equation, intra-particle diffusion and two-constant rate equation [14], respectively. The fitting results are listed in Table 1. The pseudo first order model is the earliest developed equation that explains reversible equilibrium between adsorbate and adsorbent, assuming the rate in which the sites becoming occupied is exactly equal to number of vacant sites [14]. The pseudo-first-order rate equation is generally described by the following equation:

ln ( q e − q t ) = ln q e − k 1 t (1)

where q_e and q_t are the amounts of pyridine adsorbed (mg/g) at equilibrium and at time t (min), respectively. k₁ is the rate constant of adsorption (min⁻¹) and calculated from the plots of ln(q_e − q_t) versus t for different concentrations of pyridine. The results of k₁ and correlation coefficients (R²) are shown in Table 1.

Pseudo-second-order kinetic model is based on the assumption that the rate-determining step may be a chemical sorption involving valence forces through sharing or exchange of electrons between adsorbent and adsorbate. The pseudo-second-order kinetic model is expressed as follows:

t q t = 1 k 2 q e 2 + t q e (2)

where k₂ is the pseudo-second-order rate constant (g∙mg⁻¹∙min⁻¹). Values of k₂ and q_e are calculated from the intercept and slope of the plots of t/q versus t. The results of k₂ and correlation coefficients (R²) are shown in Table 1.

The Elovich equation is often used to interpret the kinetics of sorption and successfully describe the predominantly chemical sorption on highly heterogeneous sorbents. It is based on kinetic principle assuming that the sorption sites increase exponentially with sorption, which implies a multilayer sorption and each layer exhibits a different activation energy for chemisorption. The Elovich equation can be written, in its integrated form, as follows:

q t = 1 β ln ( α β ) + 1 β ln t (3)

where, α is the initial adsorption rate (mg∙g⁻¹∙min⁻¹) and β is the desorption constant (g∙mg⁻¹). The term 1/β indicates the number of sites available for adsorption. A plot between q_t versus ln(t) yields a linear relationship with a slope of (1/β) and an intercept of (1/β)ln(αβ). The values of α, β and correlation coefficients are shown in Table 1.

The intra-particle diffusion model is used to identify the diffusion mechanism. This kinetic model considers the rate limiting step of transport of adsorbates from bulk into adsorbents and is expressed as follows:

q t = I + K id t (4)

where k_id is the intra-particle diffusion rate constant (mg∙g⁻¹∙min^−0.5) and I is the intercept of vertical axis (mg∙g⁻¹). According to Equation (4), a plot of q_t versus t_1/2 should be a straight line with a slope k_id and intercept I when adsorption mechanism follows the intra-particle diffusion process. The values of k_id and correlation coefficients are listed in Table 1.

The two-constant model is a useful empirical equation. With two fitting parameters this model is quite flexible to predict the extraction kinetics, but note that this equation has not been rigorously derived or explained. The two-constant kinetic model is expressed as follows:

ln q t = ln a + b ln t (5)

where a is the initial desorption rate constant (mg∙g⁻¹∙min⁻¹) and b is the desorption rate coefficient [(mg∙g⁻¹)⁻¹]. The values of a and b and correlation coefficients are listed in Table 1. The fitting curves and calculated parameters of the four adsorption kinetics models at different temperatures are shown in Figure 4 and Table 1, respectively.

Since each fitting equation has different assumptions, the correlation coefficient (R²) represents the correlation between the equation and the adsorption curves. It is generally believed that when the correlation coefficient R² of the fitting equation reaches ≥ 0.99, the adsorption curves conform to the fitting equation and the adsorption process satisfies the assumption of the fitting equation. Fitting the adsorption curve can also contribute to the prediction of adsorption equilibrium and adsorption capacity. The correlation coefficients of the pseudo-first-order model, two-constant model, intra-particle diffusion model and Elovich kinetic model are found to be less than 0.99. Furthermore, the adsorption capacities calculated using these fitting models greatly deviate from the experiment data. By contrast, the correlation coefficients of the pseudo-second-order kinetic model are more than 0.99, suggesting that the adsorption of pyridine on shale matches well with the pseudo-second-order kinetic model. Our kinetic results for the adsorption of pyridine by shale are consistent with some previous studies using bamboo charcoal [15], mesoporous silica [16] and polymeric adsorbents [3] [17] [18].

Some isotherm models are presented to describe the adsorption process such as Langmuir, Freundlich, Tempkin, and Redlich-Peterson. A survey of adsorption isotherms by Awad et al. (2019) reveals that Langmuir, Freundlich and Temkin are more frequently employed by previous researchers to investigate the simulation experiments. Based on the statistical result of Awad et al. (2019), we tentatively examine the present experimental data using Equations (6), (7) and (8) as followed.

Langmuir equation:

C e q e = C e q max + 1 K L × q max (6)

Freundlich equation:

lg q e = lg C e n + lg K F (7)

Tempkin isotherm is represented by the following equation:

q e = R T b T ln K T + R T b T ln C e (8)

where C_e stands for equilibrium concentration of pyridines (mg/L), q_e denotes the equilibrium adsorption capacity (mg/g), q_max refers to maximum adsorption capacity (mg/g), K_L designates Langmuir constant (L/mg), K_F represents the Freundlich constant (mg/g)(mg/L)^(−1/n), K_T signifies the equilibrium binding constant corresponding to the maximum binding energy, and b_T is equilibrium Temkin binding constant, related to the heat of adsorption. The fitting curves of the Langmuir, Freundlich and Temkin isotherms were exhibited in Figure 5.

The related model parameters and the corresponding regression coefficients are listed in Table 2. The Freundlich model with the highest R² values (Table 2) demonstrates its best describing the whole process for pyridine removal in comparison with the Langmuir and Temkin models.

It should be noted that some previous researchers have carried out laboratory experiments to examine the performance of natural and/or synthetic materials as adsorbents for pyridine removal (Table 3). From Table 3 it could be concluded that the Langmuir and Freundlich models are regularly utilized to test the isothermal experimental data.

Our investigation reveals that the removal of pyridine by shale from aqueous solution is much better demonstrated by the Freundlich adsorption isotherm model than the Langmuir model. This result coincides with former studies employing minerals widespread in sedimentary basins including natural apatite [24], kaolinite [25], and montmorillonite [25]. Our preliminary study shows that natural shale exhibits a limited adsorption capacity (of an order of magnitude of 1 mg/g) in Table 2. Nevertheless, unmodified shale may be potentially applied to treat wastewaters containing trace pyridines. Should the adsorption capacity of the natural shale be further improved by different modifications in the next wok, natural shale is undoubtedly one of the promising candidates for pyridine bearing effluents treatment since it is environment-friendly, cost-effectively and rich in nature.