RSM-driven enhancement of catalytic nitrate reduction: Mechanism, kinetics, and strategic N₂ selectivity control

2.1. Materials

The following chemicals were employed in this research: sodium nitrate (NaNO₃, AR, 99.0%), palladium chloride (PdCl₂, AR, 99.9%), cupric chloride dihydrate (CuCl₂·2H₂O, AR, 99%), hydrochloric acid (HCl, AR, 36.5-38.0%). Iron powder (Fe⁰, AR, 98%, 5μm) served as the reductant for catalytic denitrification. The following materials were utilized to prepare the catalyst carriers via wet impregnation [22]: γ-aluminum oxide (γ-Al₂O₃, AR, 99.9%,<50nm), diatomite(AR, 99%, 4.5-5mm), graphene (AR, 99%, 2nm), kaolin (AR, 99%, 2.5μm), silica gel (AR, 99%, 3-5mm) and manganese dioxide (MnO_2, AR, 91%, 20μm). All chemicals used in this study were sourced from Sinopharm Chemical Reagent Co., Ltd., China.

2.2. Experimental design

Batch experiments were carried out in a 1 L plexiglass reactor. 300 mL synthetic solution (NaNO₃, 20 mg L^-1) was pumped into the reactor with the addition of 3 g L^-1 of Fe⁰ and Pd-Cu catalyst. A 400 rpm agitation velocity was provided by the magnetic stirrer to enhance the overall mass transfer in the reactor. To maintain an appropriate solution pH for catalytic denitrification, automatic titration was set.

Samples were periodically collected to measure the concentration of nitrate-nitrogen (NO₃^--N), nitrite-nitrogen (NO₂^--N), ammonium-nitrogen (NH₄⁺-N), and total nitrogen (TN) after filtration of a 0.45 μm membrane. NO₃^-, NO₂^-, and TN were tested with ion chromatography, while NH₄⁺ was measured by Nessler’s reagent spectrophotometry. N₂ selectivity and nitrate removal were calculated as Eq (1,2) [20,23]:

Where ${\text{C}}_{\text{0}}$ is the initial concentration of NO₃^- (mg L^-1), $C_{N{O}_3^{-}-N}$, $C_{N{H}_4^{+}-N},$ and $C_{N{O}_2^{-}-N}$ are the concentrations of NO₃^-, NH₄⁺-N, and NO₂^--N after catalytic reaction (mg L^-1), assuming that the NO_x produced in catalytic denitrification was negligible [23]. To ensure data accuracy, all measurements were performed in triplicate with subsequent error analysis.

2.3. Statistical analysis

Pd-Cu/graphene was used as the catalyst in this study. Scanning electron microscope (SEM), X-ray diffraction (XRD), and Energy dispersive spectroscopy (EDS) analyses on the catalyst were conducted (Figure 1). EDS analysis implies that Pd and Cu loaded on graphene were detected. The XRD pattern of Pd/graphene shows three sharp Pd peaks, while the XRD spectra of Pd-Cu/graphene show no distinguished characteristic peaks of Pd, Cu, or their Pd-Cu combination. This phenomenon has been due to the addition of promoter metal-Cu, which may improve the dispersion of Pd or change the crystallinity of metal Pd on graphene.

Close

Figure 1.

(a) SEM image and (b) EDS analysis of pristine graphene; (c) EDS analysis and (d) XRD pattern of Pd-Cu nanoparticles supported on graphene (5 wt% Pd).

Export to PPT

RSM was employed to analyze the interaction between independent variables and their response variable using Minitab statistical software (version 19). A Box-Behnken Design (BBD) was implemented for the experimental setup. As shown in Table 1, the effects of four independent variables (X) on the predicted response variable (Y), specifically the N₂ selectivity of catalytic denitrification, have been investigated. These variables included pH (X₁), reaction time (X₂), Pd:Cu mass ratio (X₃), and Fe⁰ dosage (X₄). Each factor was tested at three coded levels (-1, 0, +1), determined based on preliminary experimental results. All experiments adhered to the BBD matrix outlined in Table 1. The relationship between the independent variables (X) and the response variable (Y) was evaluated using the following second-order quadratic polynomial model Eq 3 [24]:

Where Y represents the predicted dependent variable, X_i and X_j denote independent variables. The coefficients λ, λ_i, λ_ii, and λ_ij correspond to the constant term, linear effects, quadratic effects, and interaction effects, respectively.

The effectiveness of the model obtained from the software have been evaluated using analysis of variance (ANOVA). The statistical significance of each independent variable in the model was validated by the p-value (p<0.05). Furthermore, the adequacy, accuracy, and reliability of the model in illustrating the relationship between the independent variables and the response variables were determined by the coefficient of determination (R²) and the lack-of-fit test [24]. Additionally, 3D response surface and contour plots were employed to analyze the effects of independent variables on the response variable.

3.1. Optimization of reaction conditions for nitrate reduction catalysis

3.1.2. Analysis of variance

Table 2 presents the ANOVA for catalytic performance (N₂ selectivity), which was employed to evaluate the significance and adequacy of the proposed model. The optimal response model was selected based on p-values (<0.05) and regression coefficients with values approaching 1 [25].

Table 2. Analysis of variance.

Source	DF	Adj SS	Adj MSS	F-value	p-value	Statistic significance
Model	14	596.935	42.638	8.13	0.000	**
Linear	4	138.167	34.542	6.59	0.005
X1	1	65.333	65.333	12.46	0.004	**
X2	1	24.083	24.083	4.59	0.053
X3	1	36.750	36.750	7.01	0.021
X4	1	12.000	12.000	2.29	0.156
Square	4	447.519	111.880	21.34	0.000
X12	1	436.009	436.009	83.16	0.000	**
X22	1	45.370	45.370	8.65	0.012	*
X32	1	25.037	25.037	4.78	0.049	*
X42	1	17.120	17.120	3.27	0.096
2-way Interaction	6	11.250	1.875	0.36	0.892
X1 X2	1	1.000	1.000	0.19	0.670
X1 X3	1	1.000	1.000	0.19	0.670
X1 X4	1	9.000	9.000	1.72	0.215
X2 X3	1	0.250	0.250	0.05	0.831
X2 X4	1	0.000	0.000	0.00	1.000
X3X4	1	0.000	0.000	0.00	1.000
Error	12	62.917	5.243
Total	26	659.852
Lack-of-Fit	10	60.917	6.092	6.09	0.149	Insignificant
Pure error	2	2.000	1.000

* Significant, p<0.05 **; Extremely significant, p<0.01

The F-value for the catalytic performance model was 8.13, accompanied by a low probability value (p<0.0001), indicating the model’s extreme statistical significance (p<0.01) [26]. Among the model terms, squared terms (X₂², X₃²) exhibited significant effects, while the main factor X₁ and its squared term (X₁²) demonstrated a highly significant influence. The lack-of-fit test yielded a p-value of 0.149 (>0.05), confirming the model’s adequacy [27]. The coefficient of determination (R² = 96.47%) demonstrated that 96.47% of the variance in N₂ conversion could be explained by the experimental data, underscoring the model’s predictive validity [28]. Collectively, these results validate the model’s accuracy and reliability for optimizing operational conditions in catalytic denitrification processes.

Furthermore, to validate the proposed model, residual plots were examined, as shown in Figure 2. It is widely accepted that randomness and unpredictability constitute fundamental attributes of any valid regression model. Analysis of residual plots enables an accurate assessment of whether the observational errors (residuals) align with random errors. The residuals should be centered around zero across the range of fitted values, have been illustrated in Figures 2(b) and (d). The random errors generating these residuals are assumed to follow a normal distribution. In other words, the residuals should exhibit a symmetric pattern with constant dispersion throughout their range, as evidenced in Figure 2(a) and (c).

Close

Figure 2.

Residual plots for N₂ selectivity (a) Normal probability plot; (b) Residuals versus fitted values; (c) Histogram of the residuals; (d) Residuals versus order of the data..

Export to PPT

3.1.3. Response surface analyses

The interactive effects of combining independent variables and response variables can be tested using a 3D response surface, from which the optimal operational conditions of various processes can be predicted [27]. Figure 3 illustrates the graphical representation of the response variable (N₂ selectivity) against two distinct independent factors in catalytic denitrification. As depicted in the 3D response surface plot of Figure 1(a), N₂ selectivity (Y) exhibited an upward trend with increasing pH (X₁) and reaction time (X₂), reaching a peak before subsequently declining. This peak corresponds to the maximum N₂ selectivity (71%) observed at a pH of 5.2 and a reaction time of 127 min. Similar trends were observed for N₂ selectivity across other independent variables. The 3D response surface can be transformed into a 2D contour plot to visualize the interactive effects of two random independent variables on N₂ selectivity [29]. As shown in Figure 3, this contour plot consists of multiple elliptical zones with distinct color gradients, representing different ranges of N₂ selectivity. The central point of the smallest ellipse corresponds to the peak value observed in the 3D graph. Furthermore, the contour plot’s shape provides insights into the degree of interaction between variables: a more pronounced elliptical form indicates stronger synergistic effects.

Close

Figure 3.

3D response surface plots (left) and contour plots (right) between two factors (a) pH and time; (b) pH and Pd:Cu mass ratio; (c) pH and catalyst dosage; (d) time and Pd:Cu mass ratio; (e) time and catalyst dosage; (f) Pd:Cu mass ratio and catalyst dosage.

Export to PPT

Based on these response surface analyses, optimal interactive effects between any two variables were identified at predicted maximum values. Canonical analysis of the response surface was therefore conducted using the previously established quadratic regression model to determine optimal operational conditions. The predicted maximum N₂ selectivity (Y) reached 71% under the following parameters: pH 5.2, reaction time 127 mi, Pd:Cu mass ratio 3.4, and catalyst dosage 3.2 g L^-1. Experimental validation under these predicted conditions yielded a measured N₂ selectivity of 71.6%, demonstrating strong consistency between theoretical predictions and practical outcomes.

3.2. Catalytic nitrate reduction mechanism

Catalytic denitrification is a heterogeneous process involving three sequential steps: (1) adsorption: nitrate ions (NO₃^-) adsorb onto bimetallic active sites. (2) reduction: adsorbed nitrate undergoes redox reactions facilitated by electron transfer. (3) desorption/diffusion: reaction products (NH₄⁺, NO₂⁻, N₂) detach from the catalyst surface and diffuse into the solution.

Under weakly acidic conditions (pH 5.2), Fe⁰ serves as the primary reductant, donating electrons to drive redox reactions. Hydrogen ions (H⁺) in the solution may combine with electrons from Fe⁰ to generate atomic hydrogen (H), which could participate in nitrate reduction [23]. This process produces nitrogen species such as NO₂^-, NH, NH₄⁺, and N₂ [30].

The catalytic system comprises active components and a carrier, with bimetallic catalysts demonstrating superior efficiency over monometallic systems [31]. Pd-Cu catalysts are particularly effective due to their synergistic roles: (1) copper (Cu) sites: preferential adsorption of NO₃^- occurs here, forming NO₂^- and CuO via initial reduction [32]. (2) palladium (Pd) sites: subsequent reduction of NO₂^- to NO, NH, NH₄⁺, N₂, or other nitrogen species is catalyzed [33]. (3) regeneration: CuO is regenerated by atomic hydrogen (H*) from the solution, enabling continuous catalytic activity [34]. Figure 4 illustrates this reaction mechanism.

Close

Figure 4.

Mechanism of catalytic nitrate reduction.

Export to PPT

The carrier’s physicochemical properties (pore structure, surface area, chemical composition, and stability) critically affect catalytic performance by enhancing dispersion of active metals (Pd, Cu), facilitating nitrate adsorption, intermediate migration, and product desorption [35]. Among tested carriers (SiO₂, γ-Al₂O₃, diatomite), Pd-Cu/graphene exhibits the highest efficiency [14]. This is attributed to graphene’s high specific surface area for active site exposure, porous structure enabling mass transfer, chemical/thermal stability under reaction conditions, and strong electron transfer capacity, enhancing redox kinetics.

3.3. Approaches to improve N₂ selectivity

Based on the aforementioned research findings, catalytic denitrification has demonstrated effectiveness in nitrate removal from water systems, though N₂ selectivity remains a critical area for improvement. To address this, targeted research was conducted to enhance N₂ selectivity in catalytic denitrification processes.

3.3.1. Catalyst pretreatment

The catalytic performance of pretreated catalysts was investigated, focusing on two strategies: (1) composite carrier replacement: substituting the single-component catalyst carrier with a composite carrier to optimize reaction efficiency; (2) acid washing treatment: implementing acid washing pretreatment on the catalyst carrier to improve surface reactivity. The operational conditions were: 20 mg L^-1 NaNO₃, 3 g L^-1 Fe⁰, 3.2 g L^-1 catalyst, 127 min reaction time, 5.2 pH, and 3.4:1 Pd:Cu mass ratio, Pd 5%.

The catalytic performance of various Pd-Cu catalysts supported on composite carriers is shown in Figure 5(a). The composite carrier was prepared by combining γ-Al₂O₃ and diatomite at a mass ratio of 2:1. Among all the catalysts tested, Pd-Cu/γ-Al₂O₃-diatomite achieved the highest N₂ selectivity (68%), outperforming those using single-component carriers: Pd-Cu/Al₂O₃ (66%) and Pd–Cu/diatomite (57%). This improved performance can be attributed to the synergistic effect of the composite support, which provides more favorable physical structure and surface properties that enhance key catalytic steps, including nitrate adsorption, decomposition, migration, and transformation. The composite carrier effectively mitigates the limitations associated with each support, thereby optimizing the surface reactivity and electron transfer efficiency during the catalytic process [23].

Close

Figure 5.

(a,b) N₂ selectivity of catalysts with different carriers and their acid-washed counterparts (A:graphene, B:diatomite, C:γ-Al₂O₃, D:MnO₂, E:silica gel, F:kaolin, G:γ-Al₂O₃-diatomite, H:γ-Al₂O₃-silica gel, I:γ-Al₂O₃-kaolin, J:γ-Al₂O₃-MnO₂, K:diatomite-silica gel, L:diatomite-kaolin, M:Diatomite-MnO₂, N:kaolin-silica gel, O:kaolin-MnO₂, and P:silica gel-MnO₂).

Export to PPT

Figure 5(b) compares the N₂ selectivity of catalysts with different carriers before and after treatment with 1 mol L^-1 HCl. Acid washing significantly improved the performance of most catalysts, particularly Pd-Cu/diatomite, whose N₂ selectivity increased from 56% to 62% (a 6% rise). This may be due to the following reasons: (1) removal of surface-adsorbed impurities, increasing specific surface area and pore volume; (2) promotion of CuO reduction to metallic Cu, enhancing active site availability; (3) dissolving metal oxides (like CaO, MgO) within the carrier, enhancing its surface activity; and (4) creating new Si-O-H functional groups that generate active hydrogen, thereby improving N₂ selectivity during catalysis [23]. However, for some catalysts, acid picking can lead to a degradation in catalytic performance, as evidenced by the case of Pd-Cu/γ-Al₂O₃. After acid washing, the N₂ selectivity decreased from 67% to 64%. Characterization of the catalyst before and after treatment indicated that its specific surface area was reduced from 305 m² g^-1 to 245 m² g^-1. This significant loss in surface area is likely responsible for the slight decrease in catalytic efficiency.

3.3.2. Surface modification of the catalyst’s active components with sodium bis-2-ethylhexyl sulphosuccinate (AOT)

It is widely acknowledged that the catalytic denitrification process takes place on the bimetallic active sites (Pd and Cu), which participate in the catalytic reduction of NO_X^- (either NO₃^- or NO₂^-). To investigate the effect of adding AOT to different active components (Pd, Cu, and Pd-Cu alloy) on catalytic performance, Pd_AOT-Cu, Pd-Cu_AOT, and (Pd-Cu)_AOT catalysts were made. Their catalytic performances are shown in Figure 6. The experimental operational parameters were as follows: a 20mg L^-1 NaNO₃ solution, 3 g L^-1 of Fe⁰, 3.2 g L^-1 of the catalyst, a reaction time of 127 h, a pH value of 5.2, a Pd:Cu mass ratio of 3.4:1, and a Pd content of 5%. It was discovered that all Pd-Cu_AOT catalysts exhibited a slight decrease in nitrate removal, while the N₂ selectivity remained at the same level compared to their original counterparts. This can be attributed to the negative effect of AOT on the conversion of nitrate to nitrite triggered by Cu (as shown in Figure 7a). However, for all (Pd-Cu)_AOT catalysts, both the N₂ selectivity and nitrate removal significantly declined, which was due to the shielding of the Cu active sites by AOT (Figure 7b) [36].

Close

Figure 6.

Catalytic performance of catalysts with different metallic active components((a)PdAOT-Cu (b)Pd-CuAOT (c) (Pd-Cu)AOT) (A:graphene, B:diatomite, C:γ-Al₂O₃, D:MnO₂, E:silica gel, F:kaolin, G:γ-Al₂O₃-diatomite, H:γ-Al₂O₃-silica gel, I:γ-Al₂O₃-kaolin, J:γ-Al₂O₃-MnO₂, K:diatomite-silica gel, L:diatomite-kaolin, M:Diatomite-MnO₂, N:kaolin-silica gel, O:kaolin-MnO₂, and P:silica gel-MnO₂).

Export to PPT

Close

Figure 7.

(a) Mechanism of catalytic denitrification with Pd-Cu_AOT catalyst, (b) (Pd-Cu)_AOT catalyst, and (c) Pd_AOT -Cu catalyst.

Export to PPT

Moreover, in the case of the Pd_AOT - Cu catalysts, the N₂ selectivity dramatically increased by 4%-8%, while the nitrate removal showed an opposite trend. This phenomenon can be explained by the reaction mechanism of catalytic denitrification depicted in Figure 7(c). Firstly, the addition of AOT to Pd might shield the Pd active sites on the catalyst, thereby inhibiting the production of NH₄⁺ and promoting N₂ selectivity. Secondly, it restricted the transfer of active H to CuO, resulting in a decrease in nitrate removal [37].

3.4. Kinetic analysis of catalytic nitrate reduction

For in-depth investigation of the reaction kinetics in catalytic nitrate reduction, a comprehensive study was conducted. Comparative catalytic effects of different catalysts (N₂ selectivity) at varying time points have been illustrated in Figure 8.

Close

Figure 8.

The time-dependent catalytic performance of various Pd-Cu catalysts (A:graphene, B:diatomite, C:γ-Al₂O₃, D:MnO₂, E:silica gel, F:kaolin, G:γ-Al₂O₃-diatomite, H:γ-Al₂O₃-silica gel, I:γ-Al₂O₃-kaolin, J:γ-Al₂O₃-MnO₂, K:diatomite-silica gel, L:diatomite-kaolin, M:Diatomite-MnO₂, N:kaolin-silica gel, O:kaolin-MnO₂, and P:silica gel-MnO₂).

Export to PPT

As shown in Figure 9, the graphene-supported catalyst exhibits superior catalytic performance among the tested catalysts, achieving the N₂ selectivity exceeding 70%. This is followed by Pd-Cu/diatomite-γ-A₂O₃ and Pd-Cu/γ-A₂O₃ with N₂ selectivity of 68% and 66%, respectively. Other catalysts demonstrated substantially lower performance (<60%), while Pd-Cu/MnO₂ showed the poorest activity with only 43% N₂ selectivity. These performance variations originate from differences in physicochemical properties among the catalysts, including surface area, porous structure, chemical composition, and electron conductivity, which critically influence their catalytic efficacy.

Close

Figure 9.

Fitted line with 95% confidence and prediction intervals.

Export to PPT

Additionally, using Pd-Cu/graphene as the catalyst, the time-dependent N₂ yield was examined, with the corresponding fitting results presented in Figure 8, in which both the 95% confidence interval and the 95% prediction interval are demarcated. The linear regression equation is expressed as y= 0.29x + 27.98, with an R² value of 0.9485. This value, being very close to 1, indicates an excellent fit between the regression line and experimental data points, demonstrating that the linear model effectively describes the trend of N₂ selectivity over time. The red band in the figure represents the 95% confidence interval, signifying that we can be 95% confident that the true population regression line lies within this band. The narrow width of this confidence interval indicates high precision in the estimation of the regression line. The light pink band denotes the 95% prediction interval. This interval predicts that 95% of future new observations (i.e., individual N₂ selectivity measurements) are expected to fall within this range. The relatively wider prediction interval reflects greater uncertainty in forecasting individual future data points. The central solid red line represents the fitted regression line [38]. The experimental data points (black circles) are distributed closely along this line, with the majority lying within the 95% confidence interval. This distribution further validates the effectiveness of the linear model. Analysis of the above figure reveals a distinct linear increasing trend in N₂ selectivity with increasing reaction time. This trend indicates that the catalyst progressively enhances its selectivity towards N₂ as the denitrification reaction proceeds over time.

Kinetic analysis establishes that catalytic nitrate reduction proceeds via first-order reaction kinetics with respect to nitrate concentration [23]. The reaction rate for this catalytic process can be calculated using the following equations (5-8) [39].

Where k₁, k₂, and k₃ are the rate constants for the reduction of NO₃^- to NO₂^-, N₂, and NH₄⁺, respectively; and k₄ and k₅ are the rate constants for the reduction of NO₂^- to NH₄⁺ and N₂. The first-order kinetic parameters for different catalysts have been summarized in Table 3.

As indicated in Table 3, catalytic denitrification reaction rates varied significantly across different catalysts. Pd-Cu/MnO₂ exhibited a relatively low reaction rate (k = 0.0041), while Pd-Cu/diatomite-γ-Al₂O₃ demonstrated the highest rate (0.0087). This enhanced performance is likely attributable to the unique physicochemical properties of the composite carrier (diatomite-γ-Al₂O₃). Diatomite features a cribriform, porous structure with a specific surface area of 273 m² g^-1. Similarly, γ-Al₂O₃ is a nanoporous material with strong adsorption capacity and a specific surface area of 293 m² g^-1. When combined at a 2:1 mass ratio, the composite achieves a specific surface area of 352 m² g^-1. The synergistic effects of diatomite’s stable framework and γ-Al₂O₃’s nanostructure provide abundant active sites and facilitate efficient electron transfer during nitrate reduction. Notably, for all catalysts, the sum of the pathway-specific rate constants (k₁+k₂+k₃) closely approximated the total reaction rate constant (k). This alignment verifies that catalytic denitrification follows a surface-mediated multistep reaction pathway.

3.5 Application strategies for catalytic denitrification technology

Currently, this technology remains at the laboratory research stage. The following critical issues must be addressed for practical application:

• (1)

  Reactor design. Reactors such as fluidized bed reactors or suspended catalytic systems can be employed to ensure effective mass transfer within the system. To prevent catalyst loss, fixed-bed reactors may also be utilized, where Fe⁰ and catalysts are homogeneously mixed and packed as stationary fillers, allowing wastewater to pass through the bed for reaction.

• (2)

  Process enhancement and intelligent control. Given the diversity of actual wastewater constituents and the complexity of external environmental conditions, integrating the catalytic system with other treatment processes as a tertiary denitrification unit is recommended to achieve improved effluent quality. For instance, preliminary physical and biological treatment units can be installed to effectively remove suspended solids (SS), organic matter, nitrogen, and phosphorus. Downstream adsorption units-using specific adsorbents such as activated carbon-can be applied to treat by-products like ammonia nitrogen and Fe²⁺/Fe³⁺ generated during catalysis. Intelligent control systems can be developed to dynamically adjust key parameters-such as Fe⁰ and catalyst dosage, hydraulic retention time, and pH-based on real-time influent water quality [40].

• (3)

  Catalyst recovery and regeneration. Magnetic separation (e.g., by incorporating magnetic components into the catalyst) or membrane filtration units can be considered to enable catalyst recycling. Substances present in water bodies, such as natural organic matter, sulfides, and heavy metals may adsorb onto active sites, leading to catalyst poisoning. Establishing periodic acid washing treatment can help restore catalyst activity [14].