All chemicals used in these studies were of analytical reagent (AR) grade without any additional purification. Quinoline, pyridine, ferrous sulfate heptahydrate, ethanol, NaOH and H₂SO₄, were obtained from SD fine chemicals, India. Hydroxylamine hydrochloride, sodium borohydride, 1,10-phenanthroline were purchased from Sigma Aldrich. 30–33% hydrochloric acid, dichloromethane, acetic acid and 30% hydrogen peroxide were procured from Rankem, India. All analytical standards were prepared from Millipore water.

Sodium borohydride method was used for the synthesis of nanoscale zero-valent iron incorporated on granular activated carbon (20% nFe⁰/GAC) [14]. The commercial grade GAC was washed with distilled water to remove impurities and soaked in hot distilled water at 100°C to increase the activation of pores. Soaked activated carbon filtered and dried in an oven at 120°C for overnight. 9.956 g of FeSO₄.7H₂O was used as an iron precursor mixed with ethanol and water solution (80 mL ethanol + 20 mL water) with continuous stirring for 10 min. After complete dissolution of the iron precursor in the solution, 8 g of prepared GAC was added to iron precursor solution, in order to disperse the ion species uniformly over the GAC pores, the solution was kept for 20 min in an ultrasonic shaker. Gosu et al. [18] reported the iron reduction by borohydride is followed.

In the above equation, Eq. (1), ${Fe}_s^{2+}/GAC$ indicates iron species incorporated in the GAC framework and ${Fe}_s^0/GAC$ indicates as nanoscale zero-valent iron species incorporated in the GAC framework.

The strong reducing agent NaBH₄ was dissolved in 100 mL ultra-pure deionized water based on stoichiometric quantity and further added into the sonicated solution with continuous stirring. The black solid particles immediately appeared during the dropwise addition of the reducing agent, which indicates the formation of Fe⁰ on granular activated carbon. The complete solution filtered by using a vacuum filtration unit and the filtrate washed with ethanol for two to three times to remove water. At 50°C, the synthesized nFe⁰/GAC were dried and designated as 20% nFe⁰/GAC or nZVI/GAC.

The experiments were carried out in a three-necked round bottom glass reactor equipped with a reflux system to condense the vapor samples and to minimize the experimental error. The reactor was kept inside the oil bath to maintain the uniform desired temperature. The whole setup was mounted on a magnetic stirrer with a hot plate (2MLH, REMI). The temperature of the reaction mixture was increased using a proportional–integral–derivative controller. The reaction mixture was uniformly maintained with a magnetic stirrer at 300 rpm. For each experimental run, a 100 mL reaction mixture initial pH was adjusted (2–10 pH) by using 0.1 N HCl and 0.1 N NaOH solution and then charged into the reactor. After that, the reaction mixture heated to 60°C, with the help of oil bath, and required quantities of catalyst dose (nFe⁰/GAC) and oxidant (hydrogen peroxide) were added to the reaction mixture. During the reaction, pH was not controlled. After completion of reaction 5 h, the sample in the reactor was filtered using 0.45 μm syringe membrane filter (PTFE-2545, MOXCARE), and the filtrate samples were analyzed with a TOC analyzer.

Mineralization of pyridine and quinoline was quantified with the help of total organic carbon (TOC) conversion. TOC was quantified by catalytic oxidation of organic compounds into CO₂, the formed CO₂ quantified using non-dispersive infrared (NDIR) detector by TOC-VCPH-analyzer (Shimadzu 5500A, Germany). The instrument measure the total carbon by combustion of the sample at 700°C over a Pt catalyst bed and total inorganic carbon was measured by treating the sample with 25% phosphoric acid. TOC obtained by subtracting total inorganic carbon from total carbon. TOC values present the average of at least two measurements, in some cases samples were measured three times by injecting three times which is evaluated with the TOC apparatus. For the quantification of Fe in the aqueous solution, all the experimental samples were digested and further iron reduced to the ferrous state using hydroxylamine hydrochloride. The 1,10-phenanthroline was used as a ligand that reacts with metal (Fe) to form a strongly coloured complex. The coloured complex was measured at intensity 510 nm with the help of UV visible spectrophotometer (HACH, DR 5000, USA) [17]. Each sample was measured triplicate and results are presented with ± 5% deviation from the average value.

Micromeritics ASAP 2020 instrument was used to determine the surface area and pore volume with the help of N²-adsorption and desorption isotherm. Samples were pretreated by degassing the samples at 150°C under the vacuum of 10–3 torr for 6 h. Micropore volume was calculated using the t-plot and surface area was calculated using the Brunauer-Emmett-Teller (BET) equation by assuming that all pores in the sample are cylindrical and parallel.

Scanning electron microscopy (SEM-501 Phillips, Holland) employed to analyze the surface morphology of synthesized catalyst by operating at acceleration voltage 15–25 kV and magnification values up to 40000X. Before SEM analysis, samples were pretreated by gold coating with sputter coater thereby enhances the conductivity of the samples.

In this study, three-level and five-factorial BBD design were employed to optimize the process variables resulting in maximum TOC removal and minimum Iron leaching. BBD estimates the critical operating condition by second-order multivariate. BBD design consist of variable combinations at the center and at middle points of the edges by rotatable quadratic design [15].

The design composed of 3 levels which were coded as −1 (low), 0 (central point or middle) and 1 (high). The process variables are initial pH (A), hydrogen peroxide dose (B), catalyst dose (C), Py concentration (C_Py) (D) and Qn concentration (C_Qn) (E) shown in Table 1. The ranges of the process variables are selected based on the preliminary experiments. The statistical analysis was performed using Design-Expert^™ software. The corresponding design matrix consists of 46 experiments with center point. The experimental design matrix by the BBD is tabulated in Table 2. To fit the mathematic model, analysis of variance (ANOVA) and multiple regression analysis were estimated. The Experimental data was analyzed with multiple regressions in order to fit the data in the second-order polynomial equation (Eq. (2)). It can be written as

Where Y is the response and A, B, C, D and E are process variables. A², B², C², D² and E² are square of process variables, AB, AC, AD, AE, BC, BD, BE, CD and CE are interaction effect of process variables. β₀ and ɛ are constants. β_1, β₂, β₃, β₄ and β₅ are linear coefficients β₁₂, β₁₃, β₁₄, β₁₅, β₂₃, β₂₄, β₂₅, β₃₄ and β₃₅ are interaction coefficients.

Surface area and textural properties of GAC and 20% nFe⁰/GAC was studied by N₂ adsorption-desorption isotherms. Shape and behavior of the nitrogen adsorption-desorption isotherms of GAC and 20% nFe⁰/GAC are illustrated in Fig. 1. According to IUPAC classification, the GAC and 20% nFe⁰/GAC exhibits the type IV hysteresis based on nitrogen adsorption-desorption isotherms with the relative pressures (p/p₀) in the range of 0.2 to 0.8 that may be attributed to the characteristic of mesoporous structure. In addition, the observations from the nitrogen adsorption-desorption loop that the adsorption branches of isotherm were horizontal and parallel to the desorption branches it indicates the possibility of the existence of slit pores in dominant form [19]. Passe-Coutrin et al. [20] reported that activated carbon exhibits type IV hysteresis due to the presence of ink-bottle type pore with slit type pore structure in the activated carbon framework. Table 3 depicted the textural properties such as surface area, pore volume and average pore diameter of the samples of bare GAC and nFe⁰/GAC. Moreover, nFe⁰ impregnated onto GAC decreased the corresponding BET surface area from 273 m²/g to 69 m²/g. In addition, the pore volume also decreased from 0.162 to 0.041 cm³/g. The above observation is due to the blocking of pores (growth of nFe⁰) of GAC by excess loading of nFe⁰ [18, 10].

Fig. 2 indicates the SEM image of GAC and 20% nFe⁰/GAC. The GAC images illustrate the porous nature and 20% nFe⁰/GAC shows the formation of the uniform film on the surface of GAC. At very high loading (above 10% weight) of active species (Iron) onto the porous support, the channels of porous material (GAC) get saturated with multilayer adsorption. Further, indicates the blocking of the channel which leads to the decrease in the surface area of the material. SEM image of 20% nFe⁰/GAC is good agreement with the results of the BET surface area. For bare GAC has a surface area of 273 m²/g and after loading of active metal the surface area was decreased drastically to 69 m²/g due to blocking of pores with an excess quantity of active metal. Subbaramaiah et al. [6] studied the different percentage (5%, 10% and 20%) of active metal (Cu) loading on SBA-15. At higher loading (20% Cu/SBA-15) of active metal the surface area was decreased drastically from 650 m²/g to 313 m²/g because of blocking of pores. The morphology of nFe⁰/GAC has been found as a uniform film, which is similar to trends reported in other studies [21].

The optimum process parameters such as initial pH, catalyst dose, oxidant dose, the concentration of Py and Qn determined by BBD analysis. TOC removal and Fe leaching were used as response value to perform the analysis using RSM. The BBD design matrix for real values together with experimental and predicted values in terms of percentage removal of TOC and total Fe leaching are tabulated in Table 2. The results of BBD were used to perform the analysis of variance (ANOVA) using Expert Design software (trial version). A polynomial regression model was found to be the best suitable fit between the response variables and input process variables. The calculated regression model was a function of initial pH (A), hydrogen peroxide dose (B), catalyst dose (C), Py conc. (D), and Qn conc. (E). The best fit of the responses with coded factors are given below (Eq. (3) and (4)).

For each response, the adequacy of the model was justified with the analysis of variance (ANOVA) and the results are shown in Table S1(a) and (b), that the liability of the model was extremely significant. In the obtained regression models (Eq. (3) and (4)), the positive sign indicates the positive effect on the removal of TOC and total iron leaching, respectively. Similarly, negative coefficients signify the negative effect on the removal of TOC, and total iron leaching, respectively. The “Prob > F” is the probability that all the variation in the results are due to random error [22]. When “Prob > F” is less than 0.05, terms in the model have a significant effect on response listed in the Table S1(a) and (b). The significant model terms for TOC removal are A, B, C, E, AB, AC, AD, AE, BC, BD, BE, CD, DE, A², B², C², D² and E². While, for total Fe leaching A, B, C, E, AB, AC, AD, AE, BE, CD, CE, DE, A², B², C², D² and E² are significant model terms. In addition, values of “Prob > F” greater than 0.1 specifies the insignificant model terms. The F-values for TOC and total Fe leaching are 1594.82 and 75.89, which suggested that the terms in the model have a significant effect on response. The F-values could be larger due to noise. And the desirable adequate precision is always greater than four, and for TOC and total iron leaching it was found to be 198.018 and 38.75, respectively. Further affirmed the convenience of the model to navigate the design space. The P value represents the lack of fit which corresponds to 1.00 entail the insignificant as compared to the net error. Further regression model provides better knowledge in connection with five factors and their response [23].

The diagnostic plot of actual versus predicted and normal probability plots are acquired from the adequacy of the model. These plots provide information between actual values and predicted values and also assist to evaluate the model fitness. From Fig. 3(b), the data points between the actual value and predicted values, located closed to a diametrical line attributed to good concurrence with the fitted model [24]. In addition, the prediction of the adequacy of the model can be determined by residual normality, Fig. 3(a) depicted the studentized residuals of a normal probability plot. No deviation was observed between observed residual and expected values [23, 25]. Two models namely i) the sequential model sum of squares, and ii) sequential model summary statistics were assessed to evaluate the adequacy of the test models (Table S2). Based on Table S2, the quadratic model is the best fit among other models. Furthermore, the R² and adjusted R² are closely related to each other which indicate a good correlation between the experimental and predicted values.

Simultaneous mineralization of Py and Qn by catalytic wet peroxidation process was carried out in a batch study in the presence of the heterogeneous catalyst (nFe⁰/GAC) using hydrogen peroxide as oxidant. In this process nanoscale zero-valent iron utilizes the available dissolved oxygen converted into hydrogen peroxide and Fe²⁺. The formed hydrogen peroxide, Fe²⁺, and externally added hydrogen peroxide lead to generation of hydroxyl radical and Fe³⁺, the generated hydroxyl radical has high oxidation potential and is responsible for the oxidation/mineralization of organic compound (NHCs) (R) present in the reaction mixture. The generated hydroxyl radicals are non-selectively degrade the organic compound (NHCs) (R) into intermediate compounds (R*) and further oxidation of these intermediate compounds (R*) into simple harmless compounds which are illustrated in below reaction scheme (Eqs. (5), (6) and (7)) [9, 18].

Where: R: Organic compound (NHCs); R^*: Intermediate compounds

In order to gain the better understanding of the effects of independent variables and their interaction on TOC removal and total Fe leaching, 3D surface response plots and 2D contour plots were constructed based on the quadratic model. Moreover, 2D contour plots provide a straightforward assessment of the effects of experimental factors on the responses. In addition, an elliptical or saddle nature of the contour plots indicates the Interaction between the corresponding variables are significant [26]. Furthermore, each contour line designates the infinite number of arrangement of the selected process variables. The displayed plots reasonably interact with the process variables. All plots are nonlinear with selected process variables. Detailed effects of the process variable are discussed in the following section.

In the present study, multi-response optimization by desirability function approach was used for the optimization of the process variables by Derringer’s desirability function. As shown in Table S3, the desired goal was selected for each process variable and response. The each goal was changed with respect to our desired goals. Further, 10 optimum points are generated through numerical simulation. Among those, the best optimum conditions are initial solution pH of 3.5, hydrogen peroxide dose of 0.34 mmol, catalyst dose of 0.55 g/L, Py concentration of 200 mg/L and Qn concentration of 200 mg/L, at which TOC removal of 84% has suggested with the desirability of 0.7. The experiments are conducted in duplicate to validate the model (Table S4). The predicted values nearly close to experimental values obtained from optimization analysis (Table S4).