*Article* **Application of Response Surface Methodology for Preparation of ZnAC2/CAC Adsorbents for Hydrogen Sulfide (H2S) Capture**

**Nurul Noramelya Zulkefli <sup>1</sup> , Mohd Shahbudin Masdar 1,2,\* , Wan Nor Roslam Wan Isahak <sup>1</sup> , Siti Nur Hatika Abu Bakar <sup>1</sup> , Hassimi Abu Hasan 1,3 and Nabilah Mohd Sofian <sup>2</sup>**


**Abstract:** Hydrogen sulfide (H2S) should be removed in the early stage of biogas purification as it may affect biogas production and cause environmental and catalyst toxicity. The adsorption of H2S gas by using activated carbon as a catalyst has been explored as a possible technology to remove H2S in the biogas industry. In this study, we investigated the optimal catalytic preparation conditions of the H2S adsorbent by using the RSM methodology and the Box–Behnken experimental design. The H2S catalyst was synthesized by impregnating commercial activated carbon (CAC) with zinc acetate (ZnAc<sup>2</sup> ) with the factors and level for the Box–Behnken Design (BBD): molarity of 0.2–1.0 M ZnAc<sup>2</sup> solution, soaked temperature of 30–100 ◦C, and soaked time of 30–180 min. Two responses including the H2S adsorption capacity and the BET surface area were assessed using two-factor interaction (2FI) models. The interactions were examined by using the analysis of variance (ANOVA). Hence, the optimum point of molarity was 0.22 M ZnAc<sup>2</sup> solution, the soaked period was 48.82 min, and the soaked temperature was 95.08 ◦C obtained from the optimum point with the highest H2S adsorption capacity (2.37 mg H2S/g) and the optimum BET surface area (620.55 m2/g). Additionally, the comparison of the optimized and the non-optimized catalytic adsorbents showed an enhancement in the H2S adsorption capacity of up to 33%.

**Keywords:** adsorption; adsorbent; purification; H2S removal; response surface methodology (RSM)

## **1. Introduction**

Agricultural industries, livestock ranches, and fuel industries generally generate some natural wastewaters and wastewaters that have a tremendous effect on the debate and pollution of water [1]. The anaerobic digestion of natural wastewater and wastewater does not mitigate this degradation; instead, it creates biogas, fertilized solids, and filtered sewage for subsequent beneficial use [2–4]. For example, biogas can be efficiently used for heat and energy substitution for gasoline in transport applications [5]. The biogas composition typically consists of roughly 40–75% of methane (CH4), 25–40% of carbon dioxide (CO2), 0.5–2.5% of nitrogen (N2), 10–30 ppm(v) of ammonia (NH3), and 1000–3000 ppm(v) of hydrogen sulfide (H2S) [6,7]. These compositions, however, depend on the differential sources of the organic substrates.

In practice, the elimination of hydrogen sulfide (H2S) in the oil and gas or biogas processing industries remains one of the key obstacles to the sustainable growth of profitable technologies [8]. H2S is toxic at low concentrations (<1 ppm(v)), impacting the production of biogas and has life-threatening effects at higher concentrations (500 ppm(v)) [9,10]; hence, it is imperative to eliminate H2S in the early stages of the purification system [11]. Several methods have been implemented to eliminate H2S, such as the Clauss technique, which is primarily used in the oil and gas industries [12] that typically produce high concentrations

**Citation:** Zulkefli, N.N.; Masdar, M.S.; Wan Isahak, W.N.R.; Abu Bakar, S.N.H.; Abu Hasan, H.; Mohd Sofian, N. Application of Response Surface Methodology for Preparation of ZnAC2/CAC Adsorbents for Hydrogen Sulfide (H2S) Capture. *Catalysts* **2021**, *11*, 545. https:// doi.org/10.3390/catal11050545

Academic Editor: Daniela Barba

Received: 5 March 2021 Accepted: 19 April 2021 Published: 24 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

of H2S (>10,000 ppm(v)). Several technologies that are commonly used and commercialized for H2S removal include chemical absorption [13], physical adsorption [14], biological treatment [15–18], and membrane technology [19,20].

The H2S capture via a biological treatment is efficient and cost-effective; however, it needs a large upfront investment as compared to the dry-based processes. Even though this method is an environmentally friendly system, the separation and purification of H2S may be difficult to carry out. In contrast, the liquid-based and membrane techniques for H2S removal are not economically or energetically viable technologies [13]. However, the adsorption technology is the best and superior for H2S removal even at low concentrations and temperatures [21–23]. Adsorption is the most commonly used technique for both large-scale and small-scale applications. All of these technologies are summarized and compared with the most relevant and alternative technology for H2S removal in Table 1.



Adsorption techniques [25–27] to remove H2S typically involve mesoporous materials (activated carbon, zeolites, and/or silica) that are also widely known as catalysts because of their surface chemistry, high degree of microporosity, and developed surface area (which can exceed 1000 m2/g) [28]. On both the macro and nanoscales, these materials may have crystalline and/or amorphous structures [16], but they can be further changed to adjust their physicochemical properties, thus improving their adsorption ability against the target molecules. As commercial activated carbon (AC) is often impregnated to increase the capacity of the adsorbents to absorb the adsorbates, it is also subjected to surface modification. The improvement was primarily based on increasing the basic surface area and the porous structure of the mesoporous materials by using chemical activation methods.

Impregnated adsorbents such as catalytic adsorbents, widely applied several chemicals based on alkalis (NaOH, KOH, and KI) [29–33], carbonate compounds (Cu), transition metal oxide compounds (Zn, Fe, and Cu), or metal acetic acid compounds (Zn) [14,34] can be used as solid catalysts or be dispersed as small grains on the surface of a supporting material. Selecting the precursors of active components as well as any necessary promoters and stirring them in a solvent is the first step in producing a supported catalyst. In the end, the active metal or precursor from the solvents is dispersed on the adsorbents' surface. In contrast to raw activated carbon (AC), impregnated adsorbents with both of these chemicals have a higher specific surface area, smaller particle sizes, and increased H2S capability [32]. Despite this, a metal-supported catalyst (ZnAc2/CAC) demonstrates favorable associations between the adsorbent's capabilities in capturing the adsorbate and develops better surface area. The dispersion of ZnAc<sup>2</sup> on the CAC surface normally acts as active sites to capture the adsorbate particles efficiently. Moreover, ZnAc<sup>2</sup> leads to an increase in the specific surface area by decreasing the particle size, which results in an increase in the H2S adsorption capacity, as reported on the basis of the ZnO impregnated performance [35]. Impregnation from both chemicals (ZnAc<sup>2</sup> and ZnO) enhanced the adsorbent's capabilities through surface area and adsorption capacities.

For example, a study on the optimization of the CAC performance evaluated the optimal response using certain factors (molarity, time, humidity, temperature, and pH) and responses (adsorption capacity, surface area, selectivity, and percentage utilization). All the information obtained from these factors and responses can be used in an interaction study to determine the proposed optimization. Normally, the interaction parameters (condition variables) can be analyzed using two types of methods, namely univariate and multivariate optimization. Univariates have the slightest remedial effect relative to multivariates because of the capacity of the univariates to rely on one optimization variable; thus, the multivariate approach requires a design that adjusts all levels of variables simultaneously. For expository systems, this phase is crucial and the optimal operating conditions are determined using complex test designs, the Doehlert lattice (DM), central composite design (CCD), and threelevel designs such as the Box–Behnken design (BBD) [36–38]. The relationship between the explanatory variables and the response variables [39] can be evaluated graphically by using the empirical data sufficient for the optimal area, thereby allowing new models to be developed and identified and the current product designs to be updated [40].

Therefore, in this study, we applied the BBD by using response surface methodology (RSM) to assess the influence of factors with a minimal number of experiments by evaluating and controlling the Zn acetate CAC impregnation. The response to the selected factors determined the H2S adsorption capacity and characterizes the surface morphologies via the BET surface area of the impregnated CAC on the basis of the BBD recommendation.

#### **2. Materials and Methods**

#### *2.1. Adsorbent Preparation*

Effigen Carbon Sdn. Bhd, Malaysia, supplied granular commercial coconut activated carbon (CAC), which was sieved to obtain a particle size in the range of 3–5 mm. The selected CAC impregnation compound was zinc acetate (ZnC4H6O4), which was pur-

chased from Friendemann Schmidt Chemicals (Malaysia) and used as obtained without prior purification.

The impregnated CAC surface was prepared with 600 mL of distilled water for a 0.2–1.0 M zinc acetate solvent at 30–100 ◦C. In brief, 350 g of CAC was soaked into the solvent for 30–180 min before the distribution of the zinc acetate compound on the surface. The wet CAC was drained and dried at 120 ◦C overnight before being used for H2S adsorption testing and is indicated as ZnAc2/CAC. Moreover, the design of experiments (DOE) recommendation was submitted on the basis of the chosen molarity, soaked time, and soaked temperature for the preparation of the adsorbents.

#### *2.2. Characterization*

The surface area and the pore structure were analyzed by a Brunauer–Emmett–Teller (BET) surface area analysis using Micrometric ASAP 2010 Version 4.0.0. The surface area was obtained from the measurement of the BET isotherm, while the pore volumes and the standard pore volumes were calculated at P/Po of 0.98 by using the N<sup>2</sup> adsorption isotherm. Meanwhile, the micropore volume was calculated using the t-plot method. After degassing for 4 h at 150 ◦C, the textural properties of the sorbents were determined by N<sup>2</sup> adsorption–desorption at 196 ◦C with Quantachrome Autosorb 1 ◦C. The exact surface was extracted from the estimation of the BET.

The surface morphology and the chemical structure characterization for the optimized and non-optimized adsorbents were analyzed using the CARL ZEISS EVO MA10 and energy dispersive X-Ray analysis (EDX) with EDAX APOLLO X model. This characterization method was used to visualize the details of the adsorbent properties in terms of the structural morphology and to identify the elemental composition of the materials present on the surfaces of the adsorbent under an accelerating voltage of 10 kV.

## *2.3. H2S Adsorption Test*

In this study, the H2S adsorption test was implemented using a laboratory-scale set-up of a single stainless-steel column (height and diameter of 0.3 m and 0.06 m, respectively), as shown in Figure 1. In brief, 75 g of the impregnated adsorbent (ZnAc2/CAC) was loaded into the adsorber column and fed in with a commercial mixed gas H2S/N<sup>2</sup> (5000-ppm(v) H2S with balanced N2). The adsorption test operated at ambient temperature, the flow rate and pressure gauge were mounted at 5.5 L/min and 1 bar. Due to the tolerable range for the gas exposed to the atmosphere and fuel cell devices, the H2S breakthrough gas concentration at the outlet stream was set at 5–10 ppm(v) [41–43]. The outlet H2S gas was detected using a customized portable H2S analyzer (model GC310), which directly imported the data into the computer program. Then, the adsorption capacity of H2S for each DOE suggestion was calculated according to the equation reported by Zulkefli et al. [44].

**Figure 1.** H2S adsorption system [32]: (**a**) schematic diagram and (**b**) experimental H2S adsorption test set-up.

#### *2.4. Regeneration of Adsorbents*

The desorption process for the adsorbents was followed by set-up in a previous study by Zulkefli et al. [32]. The spent adsorbents underwent a three-step purging process. In the first step, the spent adsorbents were run through an air blower for 30 min at 150 ◦C and a flow rate of 100 L/min. Secondly, the same operating parameters were applied to the column at ambient temperature for 30 min. In the final step, the N<sup>2</sup> gas was introduced into the stream; it was fed at 5.5 L/min for 30 min to purge out and stabilized the active site on the adsorbent surface before use in the next adsorption operation up to several cycles of adsorption–desorption.

#### *2.5. Control Factors and Level Selection*

It is possible to test the effect of quadratic interactions by using a BBD combined with response surface modeling and quadratic programming. This experimental approach used the regression design to show the result as a predictive function of variables with an impartial and limited variance. In this strategy, the graphical profile illustrates the summary of the response surface being examined [45]. The effects of three preparation factors were investigated: (A) ZnAc<sup>2</sup> solution molarity (M), (B) soaked time (min), and (C) soaked temperature (◦C) on the CAC surface as well as the capture of the H2S gas. Two responses were used, namely the H2S adsorption capacity and the BET surface area, as a reference to the preparation factors.

The typical variables are coded separately as +1, −1, and 0 for the high, low, and center points; therefore, the units of the parameters are not relevant. Real variables (Xi) are coded by direct transformation as follows:

$$\chi\_{i} = \frac{\mathbf{x}\_{i} - \mathbf{x}\_{0}}{\Delta \mathbf{x}} \quad i = 1, 2, 3 \tag{1}$$

where *χ<sup>i</sup>* is the encoded value of an independent variable, *x<sup>i</sup>* is the actual value of an independent variable, *x*<sup>0</sup> is the actual value of a center point independent variable, and ∆*x* is the phase shift value of an independent variable [46]. The process factors and factor levels of the adsorbent preparation state are described in Table 2.


**Table 2.** Process factors and factor levels of the adsorbent preparation state.

The data from the BBD were analyzed by multiple regression to fit the following quadratic polynomial model:

$$\mathbf{Y} = \mathbf{a}\_0 + \sum\_{i=1}^3 \mathbf{a}\_i \mathbf{x}\_i + \sum\_{i=1}^3 \mathbf{a}\_{ii} \mathbf{x}\_i^2 + \sum\_{i=1}^2 \sum\_{j=2}^3 \mathbf{a}\_{ij} \mathbf{x}\_i \mathbf{x}\_j + \varepsilon \tag{2}$$

where Y is the response variables, α<sup>0</sup> is the model constant, α*<sup>i</sup>* represents the linear coefficient, α*ii* denotes the quadratic coefficient, α*ij* is the interaction coefficient, and *ε* is the statistical error. The least-squares method is used to solve this set of Equation (2). BBD is a common experimental design for the technique of response surfaces in statistics and is a type of second-order rotatable or nearly rotatable design based on three-level incomplete factorial designs. Each design is a combination of a two-level (full or fractional) factorial design with an incomplete block design [47].

Both combinations for factorial design are placed through a certain number of factors in each block, while the other factors are held at the central values. The BBD is a good design for this technique because (1) it enables the calculation of the parameters of the quadratic model, (2) there are no runs where all the variables are at either +1 or −1 levels, and (3) the number of experiments (N) needed for the BBD to evolve is defined as follows [48]:

$$N = 2k(k-1) + \mathbb{C}\_N \tag{3}$$

where the number of variables is k and the number of center points is *CN*. On the basis of Equation (3), the runs will be reduced to 17 with 3 major variables and 5 times the repetition in the center point to reduce the magnitude of error (*k* = 3 and *C<sup>N</sup>* = 5). Three conditions were investigated, namely the molarity state, soaked time, and the soaked temperature; hence, 17 runs were executed.

#### *2.6. Steps for Process Parameter Optimization*

The steps followed for process optimization are shown in the flow chart in Figure 2. In this optimization, the molarity, soaked period, and the soaked temperature of the adsorbents were entered as the explanatory variables and the optimal adsorption capacity with the BET surface area of the adsorbents as the response variable.

**Figure 2.** Process optimization.

Step 1. CAC impregnated with ZnAc<sup>2</sup> as suggested by the design tools.

Step 2. There were 17 run trials conducted, with the BBD matrix consisting of 12 different level combinations of the independent variables as well as three center point runs used to fit a second-order response surface and provide a measure of process stability and inherent variability [49]. The BBD design matrix along with the experimental values of the responses are shown in Table 2 (in terms of the coded factors).

Step 3. The adsorption capacity was calculated using an H2S adsorption test and was determined for each of the 17 runs of the adsorbents.

Step 4. The BET surface was determined for each of the 17 runs.

Step 5. RSM simulation, including second-order regression and analysis of variance (ANOVA), was conducted.

Step 6. The optimal conditions for the different molarities were traced on the basis of the contour and the surface plots of the RSM simulation.

Step 7. The simulation and experimental results were verified.

Step 8. The standard parameter conditions were duplicated for different molar ratios and impregnated materials.

#### **3. Results and Discussion**

#### *3.1. Box–Behnken Model Evaluation*

Based on Equation (3), with three main factors and five replications at the center point to reduce the magnitude of error (*k* = 3 and *C<sup>o</sup>* = 5), the runs were limited to 17, as detailed in Table 3. The obtained breakthrough curves for the three soaked periods, i.e., 30, 105, and 180 min, are presented in Figure 3a–c, respectively. The breakthrough curves are shown in three figures because of the large number of runs and for an easier understanding and interpretation of the obtained results. Thus, differences between the relative concentrations of the H2S curves of the runs could be understood more easily. To decide about the adequacy of the model for the H2S adsorption capacity, three different tests, namely the sequential model sum of squares, lack of fit test, and model summary statistics, were

carried out in the present study. The data of the H2S adsorption capacity in this research were subjected to a regression analysis to estimate the effect of the process variables.


**Table 3.** Box–Behnken and experimental data of responses' adsorption capacity and BET surface area.

The results shown in Table 3 can be compared with those of a previous study by Zulkefli et al. [32]. On the basis of [32], the adsorption capacity and the BET surface area were obtained at 1.83 mg H2S/g and 656.75 m2/g, respectively, at 0.2 M of ZnAc2, soaked temperature of 65 ◦C, and soaked period of 30 min. Under similar conditions, the obtained values in this study were slightly different at 1.75 mg H2S/g and 692.65 m2/g for the adsorption capacity and the BET surface area, respectively, which was probably because of the differences in the preparation process. The decrease in the BET surface area in the adsorbent was caused by the blocking of some micropores with the chemical compound of ZnAc2. This characterization of the pore structure influenced the adsorption profiles [50,51]. In contrast, the data in Table 3 show similar trends to those reported by Nakamura et al. [52]. The BET surface area decreased with an increase in the ZnAc<sup>2</sup> molarity, even though the BET surface area was slightly different for both cases because of the differences in the preparation conditions.

**Figure 3.** Breakthrough curve of H2S versus ZnAc<sup>2</sup> molarity for a constant soaked time: (**a**) 30 min, (**b**) 105 min, and (**c**) 180 min.

#### *3.2. ANOVA*

The results were analyzed using the analysis of variance (ANOVA), a regression model, coefficient of determination (R<sup>2</sup> ), adjusted R<sup>2</sup> , coefficient of variation (CV), and statistical–diagnostic and response plots. The analysis of variance (ANOVA) test is a robust and common statistical method in different applications. The ANOVA provides a statistical procedure that determines whether the means of several groups are equal or not. The Fisher's variance ratio, F-value, is used to test the significance of the model, individual variables, and their interactions [53,54]. Mean square (MSS) is the sum of squares divided by the degrees of freedom, for each source. The F-value is defined as MSSvariable/MSSresidual and shows the relative contribution of the sample variance to the residual variance [55]. If the ratio deviates increasingly from 1, the samples are not from the same population, with more confidence.

The results of the ANOVA based on experimental data are shown in Figure 4 and Table 4. The model summary statistics showed that the excluding cubic model was aliased and the 2FI model was found to have the maximum adjusted R<sup>2</sup> values. Therefore, the 2FI model was chosen for further analysis.

**Figure 4.** Design matrix evaluation for response surface of 2FI model.


**Table 4.** Adequacy of model for adsorption capacity response.

Next, the ANOVA of the adsorption capacity of H2S is summarized in Table 5. If the calculated value of F is greater than that in the F table at a specified probability level, a statistically significant factor or interaction is obtained [56,57]. The F is defined as F = MSF/MSE, where MSF and MSE are the mean square of factors (interactions) and the mean square of errors, respectively. The ANOVA test revealed that the factors A, B, and C, and the interactions A × C and B × C proved to have a statistically significant effect on the H2S adsorption capacity. The F value is an indication of the level of significance. A higher F denotes a more significant effect on the response.


**Table 5.** Analysis of variance (ANOVA) of adsorption capacity response.

We can compare the F-value from the calculations with the F-value obtained from the F-distribution table with the degree of freedom (DF) from the model and the error to discern the significance and the adequacy of the model [48]. An effect is statistically significant if the calculated F-value for the effect is greater than the F-value extracted from the table at the desired probability level. On the basis of the calculated *p*-value (prob > F), all the three factors, namely molarity, soaked period, and soaked temperature, and their interaction effects were found to be significant (Table 4). The regression equation obtained after the variance analysis yielded the level of the H2S adsorption capacity. It included a linear relationship between all the main effects and the response. The final quadratic polynomial equations in terms of the coded and the actual variables are presented as follows:

Adsorption capacity, Qcoded = 1.50 − 0.40*x<sup>A</sup>* + 0.16*x<sup>B</sup>* + 0.11*x<sup>C</sup>* + 0.13*xAx<sup>B</sup>* − 0.45*xAx<sup>C</sup>* − 0.34*xBx<sup>C</sup>* (4)

Adsorption capacity, Qactual = −0.21 + 0.65*x<sup>A</sup>* + 0.008*x<sup>B</sup>* + 0.04*x<sup>C</sup>* + 0.004*xAx<sup>B</sup>* − 0.03*xAx<sup>C</sup>* − 0.00013*xBx<sup>C</sup>* (5)

As seen in statistical studies, the values of prob > F below 0.05 signify that the model terms are significant. In this case, as shown in Table 5, models B and AC were significant with the value of prob > F of 0.05 and 0.03, respectively. The Fisher's F-value and the probability value of the regression model were found to be 1.34 and 0.32, respectively. This implied that the terms in the model had a significant effect on the response. The tabular F-value with the degree of freedom, DFmodel = 6 and DFerror = 10, respectively, at the significance level of 0.05 (F0.05,(6,10) = 3.22) was higher than the calculated F-value (F0.05,(6,10) = 1.34), implied that most of the variation in the response could not be explained by the regression equation.

Then, the coefficient of determination, R<sup>2</sup> , indicated the overall predictive capability of the model. From Table 6, the R<sup>2</sup> value of the model was determined to be 0.45. Therefore, we assumed that 45% of the total variations in the response can be explained by the model. However, this value of R<sup>2</sup> did not necessarily imply that the regression model was a suitable one. A negative prediction R<sup>2</sup> was defined as a better predictor of the H2S adsorption capacity response for the current model. In this case, an adequate R<sup>2</sup> value of 4.40 was more than 4 as the ratio desirability, which indicated that the model navigated the design space. As was observed, the adjusted R<sup>2</sup> was close to R<sup>2</sup> , emphasizing the high significance of the model. Another method to describe the variation of a model is to calculate the coefficient of variation (CV). While the values presented in Table 5 are not logically significant for the H2S adsorption capacity, the low value of the coefficient of variation (C.V.% = 40.14) might reflect the fact that this model could have high reliability and good fitness.


**Table 6.** Model reliability analysis of adsorption capacity response.

The response to the BET surface area also suggested the use of the 2FI model for further analysis through the ANOVA study based on the highest value obtained for the adjusted R<sup>2</sup> (0.35). Meanwhile, the F value obtained was an indication of the model significance level. As presented in Table 7, the 2FI model had the highest F value (2.65) of the considered models. Moreover, the highest values of the adjusted R<sup>2</sup> and the predicted R<sup>2</sup> could be a reason for the suggestion of the use of the 2FI model for further analysis.


**Table 7.** Adequacy of the model for BET surface area response.

Based on the calculated *p*-value (prob > F), all the three factors and their interaction effects were found to be significant, as presented in Table 8. The regression equation obtained after the variance analysis provided the level for the BET surface area response. It also included a linear relationship between all the main effects and the response. The factors A, B, and C, and the interactions A × C and B × C proved to have statistically significant effects on the BET surface area. The final quadratic polynomial equations of the coded and the actual variables are presented in the equation below:

BET surface area (coded) = 677.94 + 20.41*x<sup>A</sup>* − 73.97*x<sup>B</sup>* − 36.22*x<sup>C</sup>* − 11.63*xAx<sup>B</sup>* + 115.38*xAx<sup>C</sup>* − 60.50*xBx<sup>C</sup>* (6)

BET surface area (actual) = 957.86 − 443.97*x<sup>A</sup>* + 0.74*x<sup>B</sup>* − 3.56*x<sup>C</sup>* − 0.39*xAx<sup>B</sup>* + 8.24*xAx<sup>C</sup>* − 0.02*xBx<sup>C</sup>* (7)


The Fisher's F-value and the very low probability value of the regression model were found to be 2.44 and 0.10, respectively. This implied that the terms in the model had a significant effect on the response. The tabular F-value with a degree of freedom, DFmodel = 6 and DFerror = 10, respectively, at the significance level of 0.05 (F0.05,(6,10) = 3.22) was higher than the calculated F-value (F0.05,(6,10) = 2.44), indicating that the variation in the response was not significant.

As shown in Table 9, the R<sup>2</sup> value obtained was 0.59, which could be assumed to be 59% of the total variation in the BET surface area response. The coefficient of variation (CV) indicated a lower value than that of the H2S adsorption capacity response, which is 13.68% and had the highest chance for reliability and good fit of the model. The value of prediction R<sup>2</sup> = 0.21 was in reasonable agreement with the adjusted R<sup>2</sup> = 0.35. The adequate precision normally measures the signal-to-noise ratio. As shown in Table 9, the adequate precision marked at 5.10 and the ratios were more than 4, which indicated that the model was adequate for navigating the design space.

**Table 9.** Model reliability analysis of adsorption capacity response.


### *3.3. Contour Plots for H2S Adsorption Capacity and BET Surface Area Responses*

Response surface plots and contour plots are useful for the model equation image and perceiving the nature of the response surface. These plots are also useful in the study of the effect of process variables on the H2S adsorption capacity and the BET surface area in a wider range of preparation conditions of the adsorbents. Furthermore, they can be used for designing the optimum conditions for adsorbent synthesis. Equations (5) and (7) were used to construct the contour plots for the H2S adsorption capacity and the BET surface area against the molarity, soaked period, and soaked temperature, as shown in Figures 5 and 6. They depict the interaction of three main factors by keeping the other at its central level for two types of responses based on the refitted Equations (4) and (5) with the experimental data.

**Figure 5.** Contour plot describing the adsorption capacity response in soaked temperature function of ZnAc<sup>2</sup> molarity and soaked period (Soaked temperature: 65 ◦C).

**Figure 6.** Contour plot describing the BET surface area response in soaked temperature function of molarity and soaked period (soaked temperature: 65 ◦C).

As shown in Figure 6, the constant soaked temperature shows the increments of the H2S adsorption capacity with an increase in the ZnAc<sup>2</sup> molarity and the soaked period. The steepness of the increase in the H2S adsorption capacity ranged from 0.2 M to 1.0 M for the soaked period of 30 min to 180 min. Figure 6 shows the effect of the interaction of the factors with the constant soaked temperature on the BET surface area. The decreases in the molarity with the lowest soaked period resulted in the highest BET surface area of 721.04 m2/g, while at the lowest molarity (0.2 M) and a higher soaked period (>142.50 min), there was a reduction in the BET surface area.

Figure 7 presents the normal residual probability plot from the least squares fit, with both the predicted and the experimental data relatively similar to the straight line of 45◦ and the remaining points obeying the normal pattern of distribution. Hence, there was a high correlation and adequacy of the proposed model to predict the optimal conditions for preparing a highly efficient H2S adsorbent.

**Figure 7.** Normal probability plot: (**a**) adsorption capacity and (**b**) BET surface area response.

#### *3.4. Optimization and Validation*

The key goal of the optimization process was to identify the variable values at which the adsorption capacity of the H2S and the BET surface area were optimal. Consequently, the Behnken configuration box was used to evaluate the best operating mode. Figure 8 displays the proposed model, showing that the highest adsorption capacity was 2.52 mg H2S/g and the BET surface area was 620.55 m2/g at the optimum molarity of 0.22 M, soaked time of 48.82 min, and soaked temperature of 95.08 ◦C. The desirability factor was 1.0, as shown in Figure 9, which reflected the most favorable or perfect response value [58].

**Figure 8.** Optimum conditions according to the BBD statistical method.

**Figure 9.** Individual and combined desirability functions.

For a comparison that quantified the acceptability of the model, an experimental study on H2S adsorption was performed using the suggested optimum parameter conditions. The catalytic adsorbents were prepared using 350 g of CAC with a 0.22-M ZnAc<sup>2</sup> solution by soaking the CAC for up to 49 min at 95 ◦C. The experimental and theoretical verification was carried out using two responses, namely the H2S adsorption capacity and the BET surface area, as shown in Figure 10.

**Figure 10.** (**a**) Experimental plot for H2S adsorption and (**b**) 3D contour plot for optimum theoretical condition for H2S adsorption study.

The experimental data were collected through the synthesis of adsorbents based on the optimum parameter conditions as suggested at the end of BBD results. As a result, the adsorption capacity for H2S was 2.12 mg H2S/ g, whereas the theoretical data suggested a capacity of 2.52 mg H2S/g. Moreover, the experimental BET surface area was 649.56 m2/g, and the theoretical BET surface area was 620.55 m2/g. Then, the relative error between the experimental and the theoretical values was approximately 16.2% for the adsorption capacity and 4.7% for the BET surface area. Therefore, the results obtained were in the range of acceptance, as the adsorption capacity and BET surface area were closer at the optimum condition of the variable for both the experimental and the theoretical data.

#### *3.5. Adsorbent Characterization*

Figure 11 presents the SEM images for the exhausted adsorbents for two types of adsorbents, namely the optimized (ZnAc2/CAC\_O (E)) and the non-optimized (ZnAc2/CAC\_N (E)) adsorbents. The optimized adsorbents were prepared under the optimum conditions from the Box–Behnken model suggestion. While the non-optimized adsorbents were prepared using a 0.2 M ZnAc<sup>2</sup> solution with a soaked period of 30 min at 65 ◦C. Both sample syntheses were tested with a commercial mixed gas up to the exhausted point and analyzed. Next, the samples were assessed to visualize the details of the adsorbent properties in terms of the structural morphology images and the percentage of the elemental composition material presence on the adsorbent surface prepared.

**Figure 11.** SEM analysis of exhausted ZnAc2/CAC adsorbent: (**a**) ZnAc2/CAC\_O (E) and (**b**) ZnAc2/CAC\_N (E) at 2.5 k × (10 µm).

Table 10 indicates the weight percentage (wt. percentage) of the element composition in a particular region of the optimized and non-optimized adsorbents for fresh (F), exhausted (E) and after desorption (D) compared to that fresh CAC (without impregnation). The EDX analysis was conducted on elements C, Ca, Na, K, Zn, O, and S (Table 10). The C content was different because of the composition of the volume of chemicals coated on a particular adsorbent surface. The presence of the Ca, Na, and K elements in the ZnAc2/CAC adsorbents normally observed on the activated carbon as similar data was obtained in a previous study by Zulkefli et al. [32] and Moradi et al. [48]. Meanwhile, the difference of concentration between ZnAc<sup>2</sup> for optimized and non-optimized adsorbents were 0.22 M and 0.2 M, respectively, which is about 10% difference. For soaked time and soaked temperature, the difference was about 18.8 min and 30 ◦C, respectively. Based on these optimized conditions for optimized adsorbent, the Zn element increased about 80%. As a result, the ZnAc2/CAC\_O (E) had a slightly higher S element by about 50% compared to that ZnAc2/CAC\_N (E) at exhausted adsorbent as shown in Table 10. This could be due to the presence of higher Zn which can help to improve the interaction between adsorbent and H2S, and hence more H2S can be adsorbed than the non-optimized adsorbents.


**Table 10.** Semi-quantitative chemical analysis of selected points in weight percent through EDX analysis.

In the case of the exhausted adsorbents, the adsorbents were purged through the process of the desorption of air and N<sup>2</sup> gas, revealing the presence of sulfur (S) in the EDX analysis. The S element appeared on the surface for both optimized and non-optimized adsorbent as the H2S adsorb during adsorption process. Similar findings were reported by Isik-Gulsac et al. [59] because of the inclusion of the S element on the surface of the adsorbents. Based on Table 10, it indicated the increment of S element from the fresh (F) to the exhausted (E) adsorbents. After the desorption process, the S element decreased

to almost similar composition with the fresh adsorbent as indicated the physisorption occurred during the adsorption process.

Meanwhile, based on Table 10, the higher presence of the element composition of O on the adsorbent was obtained for optimized adsorbent which could be due to the increment of molarity of ZnAc2. Based on Rodriguez et al., the composition of O normally had electrostatic interactions between the dipole of H2S, and the ionic field generated by the charges in O might play a secondary role in accelerating and improving the adsorption process [60]. Hence, it would enhance the capability of the adsorbent to adsorb the H2S gas as shown in the higher S element for optimized adsorbent in Table 10. Moreover, the presence of moisture and oxygen might affect the adsorption capacity of the activated carbons, and numerous studies have investigated their impact on the H2S uptake. The presence of oxygen also increased the breakthrough time of H2S adsorption for the latter adsorbents [61–63].

As shown in Table 11, the analysis of the BET surface area was conducted for the ZnAc2/CAC\_O and ZnAc2/CAC\_N adsorbents. In order to determine the specific surface area and the pore size distribution, an analysis was carried out of the N<sup>2</sup> adsorption/desorption for the fresh and the exhausted samples denoted as ZnAc2/CAC\_O (F), ZnAc2/CAC\_N (F), ZnAc2/CAC\_O (E), and ZnAc2/CAC\_N (E). The surface area was calculated using a BET isotherm calculation, while the pore volume and the average pore volume were calculated at P/Po of 0.98 through the N<sup>2</sup> adsorption isotherm. The pores included all the micropore, mesopore, and macropore volumes.


**Table 11.** Porous properties for regeneration of optimized and non-optimized adsorbents.

Upon the adsorption–desorption of H2S, the BET surface area was influenced by the impregnation of ZnAc<sup>2</sup> as chemical compound in CAC and the presence of H2S and its elements on the adsorbent which cause pores blocking by the H2S components as previously observed [32]. The optimized adsorbents (ZnAc2/CAC\_O) showed a slightly lower BET surface area than the non-optimized adsorbents (ZnAc2/CAC\_N) because of the different parameter conditions for the prepared catalytic adsorbents. It is suggested the decrease in the BET surface area could be due to the increase of ZnAc<sup>2</sup> molarity used, hence blocking of some micropores on the adsorbent as mentioned previously. The exhausted adsorbents also showed a decrease in the BET surface area, total pore volume, volume ratio, and pore size, as a result of the interaction of H2S with adsorbents which cause a blocking of the pores.

#### *3.6. Performance of Adsorption–Desorption Cycle*

The performance of the adsorbents was investigated through the adsorption degradation in the adsorption–desorption regeneration cycle. As ZnAc<sup>2</sup> composited as a catalyst on the adsorbents' surfaces was synthesis and observed the adsorption capacity performance through adsorption–desorption regeneration cycle. The adsorption capacity was calculated using the adsorption breakthrough time with the concentration change known as the mass transfer zone through downwards within the bed till further away from the inlet stream.

The H2S adsorption capacity was compared between the optimized and the nonoptimized adsorbents in order to observe the adsorbents' performance, as illustrated in Table 12 and Figure 12. Thus, the optimized adsorbents (ZnAc2/CAC\_O) showed excellent performance based on the adsorption capacity, which was higher than that of the non-optimized adsorbents (ZnAc2/CAC\_N) with an adsorption capacity difference

of 49.3%. However, the performance of the optimized and the non-optimized adsorbents exhibited a degradation of up to 16% and 23% in the adsorption capacity throughout the regeneration cycle.


**Table 12.** Comparison of adsorption capacity in regeneration of adsorption–desorption H2S.

**Figure 12.** Regeneration adsorption–desorption curve for (**a**) optimized adsorbent and (**b**) nonoptimized adsorbent.

In actual operation, the impregnation of activated carbon could involve physisorption and chemisorption which are both important for accelerating the adsorption process through physical forces or to catalyze the oxidation. Since the chemisorption probably can happen during the adsorption process, the degradation of the adsorbent could occur [64]. However, in this study, the adsorbent can still be regenerated in several adsorption– desorption cycles as shown in the capability of the adsorbent to adsorb the H2S in the following cycles as shown in Figure 12. As discussed in a previous study, the presence of S elements throughout H2S adsorption–desorption cycle can effectively remove the S elements on the adsorbent's surface up to 98% [30]. In this study, based on previous EDX analysis as shown in Table 10, after the desorption process, the S element decreased to almost similar composition with the fresh adsorbent as indicated the physisorption occurred during adsorption process as mentioned previously. However, there are slightly remaining S elements (as compared to the fresh adsorbent) which could be due to insufficient desorption process, i.e., non-optimized conditions for the desorption process, and probably due to complex mechanisms that happen during H2S adsorption-desorption process. Hence, it probably could lead to a degradation of the H2S adsorption capabilities for the following cycle of adsorption–desorption [62] as shown in Table 12 and Figure 12.

However, the degradation was low in each cycle, and the H2S adsorption capacity could be enhanced by using a different desorption method in future works. Therefore, as proven by the previous characterization study, the performance of the optimized adsorbents improved the capabilities of the adsorbents as compared to the non-optimized adsorbents.

#### **4. Conclusions**

The performance of catalytic adsorbents for H2S captured using the adsorption technique was examined through the impregnation of ZnAc<sup>2</sup> on the activated carbon surfaces. The optimization was carried out using RSM and the Box–Behnken experimental design to determine the optimum conditions for the adsorbent synthesis. Several factors and levels were evaluated, including the ZnAc<sup>2</sup> molarity, soaked period, and soaked temperature,

along with the response of the H2S adsorption capacity and the BET surface area. From the statistical analysis, the optimum points for ZnAc<sup>2</sup> molarity, soaked period, and soaked temperature were obtained as 0.22 M, 48.82 min, and 95.08 ◦C, respectively. Furthermore, the optimized adsorbents (ZnAc2/CAC\_O) improved the adsorbent efficiency by up to 49% of the adsorption capacity as compared to the non-optimized adsorbents (ZnAc2/CAC\_N). The optimized ZnAc<sup>2</sup> as the active catalyst dispersed onto the microporous materials of the activated carbon and improve the interaction of H2S on adsorbent during the adsorption process. It was observed that the S element increase with the exhausted adsorbent from fresh adsorbent and the S element of optimized adsorbent was higher compared to that non-optimized. Based on the adsorption–desorption cycle, it was revealed the adsorbent slightly degraded by referring the calculated H2S adsorption capacity up to 16% and 23% for optimized and non-optimized adsorbents, respectively throughout the cycles. It is suggested the degradation could be due to insufficient desorption process, i.e., non-optimized conditions, and probably due to complex mechanisms that happen during the adsorption–desorption process. Hence, comprehensive studies are required in the future to analyze the adsorbent degradation by optimizing the conditions of the desorption process and analyze the mechanism of adsorption-desorption of H2S on the adsorbent.

**Author Contributions:** For research articles with several authors, the following statements should be used Conceptualization, N.N.Z. and M.S.M.; methodology, N.N.Z. and M.S.M.; software, N.N.Z. and S.N.H.A.B.; validation, M.S.M., H.A.H., and W.N.R.W.I.; formal analysis, N.N.Z., M.S.M. and N.M.S.; investigation N.N.Z.; resources, N.N.Z. and M.S.M.; data curation, N.N.Z. and M.S.M.; writing original draft preparation, N.N.Z.; writing—review and editing, M.S.M., H.A.H., and W.N.R.W.I.; visualization, N.N.Z. and M.S.M.; supervision, M.S.M., H.A.H., and W.N.R.W.I.; project administration, M.S.M.; funding acquisition, M.S.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Ministry of Higher Education, Malaysia, grant number FRGS/1/2020/TK0/UKM/02/4 and Universiti Kebangsaan Malaysia, grant numbers GUP-2018-042 & PP-FKAB-2021.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All relevant data are contained in the present manuscript. Other inherent data are available on request from the corresponding author.

**Acknowledgments:** This research was supported by the Ministry of Higher Education, Malaysia under research code FRGS/1/2020/TK0/UKM/02/4 and Universiti Kebangsaan Malaysia under research code GUP-2018-042 & PP-FKAB-2021.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Nomenclature List**



#### **References**


## *Article* **Polyoxometalate Dicationic Ionic Liquids as Catalyst for Extractive Coupled Catalytic Oxidative Desulfurization**

**Jingwen Li, Yanwen Guo, Junjun Tan and Bing Hu \***

School of Materials and Chemical Engineering, Hubei University of Technology, Wuhan 430068, China; lijingwen@hbut.edu.cn (J.L.); guoyanwen@hbut.edu.cn (Y.G.); tanjunjun2011@hbut.edu.cn (J.T.) **\*** Correspondence: hubing@hbut.edu.cn; Tel.: +86-13667257353

**Abstract:** Wettability is an important factor affecting the performance of catalytic oxidative desulfurization. In order to develop an efficient catalyst for the extractive coupled catalytic oxidative desulfurization (ECODS) of fuel oil by H2O<sup>2</sup> and acetonitrile, a novel family of imidazole-based polyoxometalate dicationic ionic liquids (POM-DILs) [Cn(MIM)<sup>2</sup> ]PW12O<sup>40</sup> (n = 2, 4, 6) was synthesized by modifying phosphotungstic acid (H3PW12O40) with double imidazole ionic liquid. These kinds of catalysts have good dispersity in oil phase and H2O<sup>2</sup> , which is conducive to the deep desulfurization of fuel oil. The catalytic performance of the catalysts was studied under different conditions by removing aromatic sulfur compound dibenzothiophene (DBT) from model oil. Results showed that [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O<sup>40</sup> had excellent desulfurization efficiency, and more than 98% of DBT was removed under optimum conditions. In addition, it also exhibited good recyclability, and activity with no significant decline after seven reaction cycles. Meanwhile, dibenzothiophene sulfone (DBTO<sup>2</sup> ), the only oxidation product of DBT, was confirmed by Gas Chromatography-Mass Spectrometry (GC-MS), and a possible mechanism of the ECODS process was proposed.

**Keywords:** polyoxometalate; dicationic ionic liquids; extraction; oxidative desulfurization; dibenzothiophene

## **1. Introduction**

With the development of the economy, traditional fossil fuels still occupy a large proportion of supply and demand in the market [1]. Some sulfur compounds contained in fuel oil, such as mercaptan, thioether, thiophene and their derivatives, will produce sulfur oxides during combustion, which can lead to a series of environmental problems such as acid rain and haze [2–4]. Therefore, many countries are constantly strengthening the control standards of sulfur content in fuel oil. Improving the technology to produce high quality fuel oil in accordance with the standards has become a top priority for refineries [5,6]. Hydrodesulfurization (HDS) is the most mature technique and has been applied in industry [7–10]. It can efficiently eliminate aliphatic sulfur compounds, such as mercaptan and thioether. However, in addition to the harsh operation conditions, HDS is not effective for removing aromatic sulfur compounds and their derivatives with steric hindrance [11,12]. In this context, as a nonhydrodesulfurization technology that can achieve deep desulfurization of fuel oil under mild conditions, the ECODS process has become a main focus due to its simplicity and effectiveness [13–16]. Although various types of solvents and oxidants have been used in the desulfurization process, and also play an important role, the biggest challenge for a successful ECODS process is to use catalysts with high activity.

Polyoxometalates (POMs), represented by H3PW12O40, have been widely used as the catalysts for oxidative desulfurization under mild conditions of the model oil system due to their strong Bronsteic acidities and redox properties [17–20]. On the other hand, ionic liquids (ILs), as a prominent catalyst/extractant with low vapor pressure, good thermal stability, recyclability and environmental friendliness, are also usually

**Citation:** Li, J.; Guo, Y.; Tan, J.; Hu, B. Polyoxometalate Dicationic Ionic Liquids as Catalyst for Extractive Coupled Catalytic Oxidative Desulfurization. *Catalysts* **2021**, *11*, 356. https://doi.org/10.3390/ catal11030356

Academic Editor: Oxana Kholdeeva

Received: 1 February 2021 Accepted: 6 March 2021 Published: 9 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

used in desulfurization reactions [21]. However, the shortcomings of POMs and ILs are the main obstacles to their industrial application. For example, the small specific surface area of POMs (<10 m2/g) makes their catalytic activity low [22], and the liquid properties of ILs make them difficult to separate and recover. In order to solve those problem, according to the characteristics that specific ILs with different properties can be designed by the combination of different cations and anions [23], and the catalytic performance of POMs can be regulated by the electrostatic interaction and hydrogen bonding between the cations of specific ionic liquids and the anions of POMs [24], a new type of POM-IL with different physical and chemical properties, which is formed by the combination of heteropolyanions and organic cations, has attracted widespread attention. Huang et al., synthesized a heteropolyanionic-based ionic liquid catalyst [(3 sulfonic acid) propylpyridine]3PW12O40·2H2O ([PSPy]3PW12O40·2H2O) by the reaction of N-Propanesulfone pyridinium with an aqueous solution of H3PW12O40, which showed high catalytic activity and excellent recyclability in the oxidative desulfurization of fuel oil [25]. Our groups successively synthesized a kind of POM-IL, [Hmim]5PMo10V2O<sup>40</sup> [26], [C3H3N2(CH3)(CnH2n)]5PMo10V2O<sup>40</sup> ([Cnmim]PMoV n = 2, 4 and 6) [27], by the reaction of molybdovanadophosphoric acid (H5PMo10O40) with N-methylimidazole and imidazole bromides, respectively, and applied it as a catalyst to the desulfurization process with H2O<sup>2</sup> as the oxidant. The results showed that 99.1% and 100% of dibenzothiophene (DBT) in the model oil are removed, respectively, and the catalytic activity of POM-ILs decreased slightly after six cycles. However, these systems still need more catalysts and a relatively long reaction time to achieve ideal desulfurization efficiency. Therefore, it is necessary to find other methods to improve the economic applicability and effectiveness of the catalysts. In recent years, many other types of POM-ILs have been used as catalysts for oxidative desulfurization, such as [C11H9N(CH2)4SO3H]3PW12O<sup>40</sup> (PhPyBs-PW) [28], [3- (pyridine-1-ium-1-yl)propane-1-sulfonate]3(NH4)3Mo7O24·4H2O ([PyPS]3(NH4)3Mo7O24) [C6H5NO2CH2(CH2)2CH3]7PMo12O<sup>40</sup> ([29] and (NKBu)7PMo12O42) [30], which can effectively improve the desulfurization efficiency in a short period of time with a small amount of catalyst. However, it is rarely reported that POM-based dicationic ionic liquids with higher thermal stability, good wettability and high activity are used as catalysts for oxidative desulfurization. DILs are a new type of ionic liquid compound with higher stability and lower toxicity, which consists of two monomers linked by alkyl or aryl groups [31–34]. Compared with the traditional monocationic ILs, DILs have a larger cationic volume, which makes the π-π interaction between cations and aromatic sulfides stronger, and can effectively remove aromatic sulfides in fuel oil. In addition, through electrostatic interaction and hydrogen bonding, the double cation can be well connected with the anion, so as to improve the overall catalytic activity of the catalyst. Due to their unique properties, DILs have been successfully used and achieved ideal effects in esterification, supercapacitor, biodiesel catalysis and extractive desulfurization as eutectic solvents/catalysts, electrolyte additives, catalysts and extractant, respectively [35–38]. In our group's recent research results, a series of novel binuclear magnetic ionic liquids (MILs) [Cn(MIM)2]Cl2/mFeCl<sup>3</sup> (n = 2, 4 or 6 and m = 1, 2 or 3) were synthesized and used as catalysts for the desulfurization with oxidant H2O<sup>2</sup> and extractant acetonitrile [39]. The results showed that 97.07% desulfurization efficiency can be achieved in 10 min, showing ultrahigh catalytic activity of MILs.

Inspired by the above research, we are deeply interested in the preparation of novel POM-DILs and their application in the field of catalytic oxidative desulfurization. In this work, in order to give full play to the advantages of POMs and DILs, we prepared a new kind of POM-DIL catalyst [Cn(MIM)2]PW12O<sup>40</sup> (n = 2, 4, 6) with H3PW12O<sup>40</sup> modified by imidazole-based DILs, and the catalyst was further applied in ECODS system, which was constructed with H2O<sup>2</sup> as the oxidant and acetonitrile as the extractant. The optimum reaction conditions and parameters were determined. In addition to their high thermal stability, the catalysts also showed high catalytic activity for the removal of DBT from model oil due to their excellent wettability. At the same time, the recycling performance of the catalyst was also explored. Finally, based on the corresponding characterization results, the possible mechanism of the desulfurization process was proposed.

#### **2. Results and Discussion**

#### *2.1. Characterization of Catalyst*

Fourier transform infrared spectroscopy (FT-IR) is a suitable technique to prove the success of catalyst synthesis. The FT-IR spectra (wavelength range from 4000 to 500 cm−<sup>1</sup> ) of catalysts [C2(MIM)2]PW12O40, [C4(MIM)2]PW12O40, [C6(MIM)2]PW12O<sup>40</sup> are shown in Figure 1a. The absorption peak near 2950 cm−<sup>1</sup> was attributed to the C-H stretching vibration of imidazole ring. The stretching vibration absorption peaks around 1630 and 1564 cm−<sup>1</sup> were attributed to C=C and C=N skeleton vibrations on the aromatic ring, respectively. The peaks at 1463, 1408 cm−<sup>1</sup> , 1341 and 1253 cm−<sup>1</sup> were related to C–N heterocycles. In the range of 870-1100 cm−<sup>1</sup> , four characteristic absorption peaks of Keggin structure were observed in all three catalysts, which were 1084 (P-O), 983 (W=O), 898 (W-Oc-W corner-sharing) and 800 cm−<sup>1</sup> (W-Oe-W edge-sharing), respectively. This was consistent with the characteristic peaks of H3PW12O<sup>40</sup> in Figure 1b, which showed that the catalysts still had the Keggin structure of PW12O<sup>40</sup> <sup>3</sup>−.

**Figure 1.** *Cont.*

**Figure 1.** Characterizations of the samples. (**a**) FT-IR spectra of catalysts; (**b**) FT-IR spectra of H3PW12O40; (**c**) XRD patterns of catalysts; (**d**) XRD patterns of H3PW12O40; (**e**) Ultraviolet visible (UV-vis) spectra of catalysts; (**f**) Thermogravimetric (TG) curve of catalysts.

X-ray diffraction (XRD) also provided strong evidence to support the successful synthesis of catalysts. Compared with the XRD pattern (2θ from 5◦ to 50◦ ) of catalysts (Figure 1c) and H3PW12O<sup>40</sup> (Figure 1d), some diffractive peaks of the three catalysts were obtained at near 10◦ , 21◦ and 32.8◦ . This was due to the disappearance of coordination water H5O<sup>2</sup> <sup>+</sup> and H3O<sup>+</sup> which interact with the anions of the Keggin structure by replacing the secondary structure proton of H3PW12O<sup>40</sup> with [Cn(MIM)2] 2+ (n = 2, 4, 6).

UV-vis spectroscopy is a rapid and accurate method to determine the molecular structure of organic compounds and the charge transfer behavior of catalysts. The UV-vis spectra of the catalysts are shown in Figure 1e. There were two characteristic peaks in the range of 190–400 nm near the ultraviolet region, which were related to the electronic properties of the center metal atoms in the anions of the catalyst structure. This structure was similar to [PW12O40] <sup>3</sup><sup>−</sup> [40]. The absorption peaks at 203, 196 and 191 nm of the three catalysts were caused by O→P transition, and the strong absorption peaks at 266, 266 and 267 nm of the three catalysts were considered to charge the transfer of metal atoms (O2−→W6+), where W atoms were located in W-Oe-W intrabridges between edge-sharing WO<sup>6</sup> octahedra in the Keggin units.

By recording the TG curve of the synthesized catalysts, the thermal stability of the catalysts can be clearly displayed in Figure 1f. The first mass loss occurred at 100 ◦C, which was caused by the disappearance of physical water and crystallization water in the catalyst. With the increase in temperature, the weight loss within the range of 300 to 800 ◦C was related to the decomposition of catalysts. The [Cn(MIM)2] 2+ cation was decomposed first; the initial decomposition temperature of the three POM-DIL catalysts was 495 ◦C for [C2(MIM)2]PW12O40, 420 ◦C for [C4(MIM)2]PW12O<sup>40</sup> and 427 ◦C for [C6(MIM)2]PW12O40. Then, the PW12O<sup>40</sup> <sup>3</sup><sup>−</sup> anion was decomposed at about 600 ◦C [25]. From the results, it can be concluded that the reasons for the high thermal stability of POM-DIL catalysts were not only the cation symmetric structure of the double imidazole ring but also the Keggin structure of the anion. At the same time, the carbon chain length of cations had no obvious effect on the thermal stability of the catalysts.

In order to further accurately determine the moisture in the samples, the content of free water in the samples was determined by the Karl Fischer Titrator (KFT) Ipol method with the Karl Fischer Moisture Titrator (870 KF Titrino plus), and the results are listed in Table 1. From the test results, the moisture content in the sample was extremely small, and the purity of each sample could reach 99%.


**Table 1.** The moisture in the samples.

X-ray photoelectron spectroscopy (XPS) characterization is a powerful technique to determine the composition, content and molecular structure of catalysts [41]. The composition of the catalyst [C2(MIM)2]PW12O<sup>40</sup> was analyzed by XPS, as can be seen from the survey spectrum of the sample (Figure 2a), the catalyst was mainly composed of C, N, O, P and W elements. The contents of each element are listed in Table 2. The results showed that [C2(MIM)2]PW12O<sup>40</sup> was mainly composed of C and O elements. At the same time, the higher O content indicated that there were abundant oxygen-containing groups in the catalyst, which was consistent with the characterization results of FT-IR. The XPS spectrum of C1s (Figure 2b) can be fitted into three peaks with binding energies of 285.3, 284.5 and 283.3 eV, respectively, which were attributed to C=N–C, C–C/C=C and C=N. In the XPS spectra of N 1s, O 1s and P 2s (Figure 2c–e), the corresponding binding energies were 400, 529.5 and 132.9 eV, which correspond to C-N, O2<sup>−</sup> and P-O, respectively. The two peaks at 36.7 and 34.6 eV (Figure 3f) were attributed to W 4f5/2 and W 4f7/2. Compared with the existing literature [42,43], due to the electrostatic interaction between the cation and the anion, significant electron shift was observed in the W4f spectra. The results showed that the surface electron density of the catalysts was effectively enhanced, which was conducive to the formation of hydroxyl radicals, thus improving the overall ECODS performance.

**Table 2.** Element composition and content of [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40.


**Figure 2.** *Cont.*

**Figure 2.** XPS spectra of [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40; (**a**) Survey of the catalyst; (**b**) C 1s; (**c**) N 1s; (**d**) O 1s; (**e**) P 2p; (**f**) W 4f.

Figure 3 shows the results of hydrophilicity and hydrophobicity tests of the catalysts. The instantaneous contact angles of a water droplet on the three POM-DIL catalysts were all less than 90◦ . The contact angle of [C2(MIM)2]PW12O<sup>40</sup> was also measured with n-octane as the testing droplet. The results showed that the instantaneous contact angle between the oil droplet and the catalyst surface was almost 0. These results indicated that the catalysts have good wettability for both H2O<sup>2</sup> and n-octane, which can effectively improve the utilization rate of the oxidant and the overall desulfurization efficiency.

#### *2.2. Catalytic Activity of Catalyst*

Table 3 summarizes and compares the effects of different kinds of POM-IL catalysts on the DBT removal effect in fuel oil under major reaction parameters, such as the H2O2/DBT molar ratio (n(H2O2)/n(S)), reaction temperature and reaction time. The results showed that increasing the length of the carbon chain in catalyst cation has no obvious effect on the desulfurization effect under the same reaction conditions. However, when the volume of cation increased, the desulfurization effect was obviously improved. On one hand, the catalyst had good wettability in the oil phase and H2O2, which can rapidly interaction with H2O<sup>2</sup> and oil. When the catalyst was fully contacted with the oxidant, it could decompose more active substances [44], which was more conducive to the removal of sulfide. On the other hand, the large cations had higher aromatic π electron density, which could effectively enhance the π-π interaction between the double imidazole ring and the thiophene ring, thus making POM-DILs have excellent desulfurization performance. Considering the economic factors and desulfurization effect, [C2(MIM)2]PW12O<sup>40</sup> was selected for the subsequent desulfurization research.

**Figure 3.** Contact angles of water droplets on the surface of the catalyst: (**a**) [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40; (**b**) [C<sup>4</sup> (MIM)<sup>2</sup> ]PW12O<sup>40</sup> and (**c**) [C<sup>6</sup> (MIM)<sup>2</sup> ]PW12O40. Contact angles of n-octane droplets on the surface of the catalyst: (**d**) [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40.

**Table 3.** Comparison of different catalysts for removal of dibenzothiophene (DBT) in model fuel.


The solubility of catalysts in different solvents is an important factor to be considered in their application. Therefore, the solubility of the catalyst in model oil and acetonitrile was tested. According to the results in Table 4, the catalysts were slightly dissolved in n-octane and acetonitrile, and tended to be dissolved in acetonitrile with relatively strong polarity. Combined with the desulfurization effect in Table 3, the partial dissolution of the catalyst in solvent had little effect on desulfurization efficiency and oil quality, which can be almost ignored.


**Table 4.** Solubility of catalyst in n-octane and acetonitrile.

Condition: m(catalyst) = 0.005 g; V(solvent) = 5 mL; T=50 ◦C; t = 60 min.

After the ECODS system was established, the initial reaction conditions were optimized to ensure the best desulfurization efficiency. The effect of reaction temperature on the removal of sulfide DBT from model oil was investigated, and the results are displayed in Figure 4. When the temperature increased from 20 to 80 ◦C, the desulfurization efficiency showed a trend of increasing first and then decreasing. This was because the oxidation reaction was limited by kinetics and POM-DILs could not effectively catalyze desulfurization at low temperature [48]. With the increase in reaction temperature, the activity of the catalyst and oxidant was gradually enhanced, and the oxidation rate of DBT into DBTO<sup>2</sup> was accelerated. However, with the further increase in reaction temperature, the desulfurization efficiency would decrease slightly due to the gradual decomposition of H2O<sup>2</sup> and the deactivation of active components [16,49]. Therefore, the best reaction temperature was determined to be 50 ◦C.

**Figure 4.** Effect of reaction temperature on desulfurization. Reaction conditions: V(oil) = 5 mL; n(catalyst)/n(S) = 0.025; n(H2O<sup>2</sup> )/n(S) = 6; V(acetonitrile) = 0.5 mL; t = 60 min.

It can be seen from Figure 5, the desulfurization efficiency increased greatly in the initial stage of the reaction. After the reaction for 60 min, the desulfurization efficiency reached the maximum. This may be because the sulfide content in the model oil was highest in the initial reaction stage, so both the extraction rate and oxidation rate were much higher than the other reaction stage. Then, with the increase in reaction time, the overall desulfurization reaction tended to equilibrium, and the desulfurization efficiency did not increase significantly, or even slightly decreased, which was due to the partial volatilization of n-octane in a longer reaction time. Therefore, taking into account the effect of reaction time, the t = 60 min was selected to be used in the rest of the experiments.

**Figure 5.** Effect of reaction time on desulfurization. Reaction conditions: V(oil) = 5 mL; n(catalyst)/n(S) = 0.025; n(H2O<sup>2</sup> )/n(S) = 6; V(acetonitrile) = 0.5 mL; T = 50 ◦C.

The influence of the catalyst amount on the desulfurization effect was also considered. As shown in Figure 6, without catalyst, the desulfurization effect was 67.11%. When the molar ratio of catalyst to sulfide was increased to 0.025, the desulfurization effect was increased to 98.35%. According to the reported literature [50], the high catalytic efficiency of H3PW12O<sup>40</sup> composites in oxidative desulfurization (ODS) was mainly due to the existence of catalytic active center W=O. Therefore, we speculated that with the increase in the amount of catalyst, the active sites provided for oxidative desulfurization increased, which promoted the effective removal of DBT. When the molar ratio of catalyst to sulfide was further increased, the desulfurization effect did not change obviously. Therefore, 0.025 was a suitable molar ratio of catalyst to sulfide.

**Figure 6.** Effect of the mole ratio of catalyst to sulfide on desulfurization. Reaction conditions: V(oil) = 5 mL; n(H2O<sup>2</sup> )/n(S) = 6; V(acetonitrile) = 0.5 mL; t = 60 min; T=50 ◦C.

The effect of oxidant dosage on the removal of sulfide DBT from model oil was presented in Figure 7. When the desulfurization system was carried out in the absence of H2O2, the desulfurization efficiency was only about 67.48%. However, when the n(H2O2)/n(S) was increased from 0 to 4, the conversion of DBT was improved significantly. After further increasing the n(H2O2)/n(S) to 6, the removal of DBT could reach 98.35%. According to the stoichiometric reaction, the oxidation of 1 mol DBT to the corresponding sulfones requires 2 mol of H2O<sup>2</sup> [51]. In theory, the excessive amount of H2O<sup>2</sup> was beneficial to fully oxidize DBT to sulfones. In practice, considering the desulfurization efficiency and excessive oxidant may cause oil pollution, n(H2O2)/n(S) = 6 was appropriate.

**Figure 7.** Effect of the mole ratio of oxidant to sulfide on desulfurization. Reaction conditions: V(oil) = 5 mL; n(catalyst)/n(S) = 0.025; V(acetonitrile) = 0.5 mL; t = 60 min; T = 50 ◦C.

The effect of the extractant dosage on DBT removal from model oil is shown in Figure 8. When the desulfurization test was conducted without acetonitrile as the extractant, the removal rate of DBT in the model oil reached 69.49%, which was the result of the combination of oxidant and catalyst. With the increase in the dosage of extractant, the desulfurization effect was significantly improved. This was because when the catalyst, oxidant and extractant were added to the reactor, an environment similar to the emulsion was formed. This environment could effectively increase the contact among the catalyst, oxidant and sulfide, so as to improve the desulfurization efficiency. When the amount of extractant became greater than 0.5 mL, the desulfurization efficiency was slightly improved. Therefore, 0.5 mL extractant was suitable for sulfide extraction.

**Figure 8.** Effect of extractant dosage on desulfurization efficiency. Reaction conditions: V(oil) = 5 mL; n(catalyst)/n(S) = 0.025; n(H2O<sup>2</sup> )/n(S) = 6; t = 60 min; T = 50 ◦C.

Many ILs have been developed which are used as both the catalyst and extractant. Usually, the cation side of ILs influences the extraction ability of this material for DBT removal [38]. Therefore, the extraction ability of DILs was investigated. The results in Table 5 show that the DILs also exhibited a certain extraction ability. Due to the higher aromatic π-electron density of DILs, a stronger π-π interaction could be formed between the cations of DILs and aromatic sulfides, so DILs have a higher extraction efficiency than monocationic ionic liquids and ordinary organic solvents.


**Table 5.** Effect of different extractants on the removal of DBT.

Reaction conditions: V(oil) = 5 mL; V(extractant) = 0.5 mL; t = 60 min; T = 50 ◦C.

#### *2.3. Recycling of Catalysts*

The recyclability of the POM-DIL catalyst in the reaction system was further investigated from the perspective of economic cost. After the reaction, the upper oil phase was taken for analysis. The solvent in the system was separated by the decanting method, and the catalyst was dried at 100 ◦C to remove the residual H2O2, acetonitrile and model oil. Then, the catalyst was reused for the next cycle by the addition of fresh model oil, oxidant and extraction agent. Each group of data was repeated at least three times, and the standard deviation was less than 1.2. As results show in Figure 9, the desulfurization rate was still up to 89.42% after seven cycles of the catalyst. Compared with the catalytic performance of the fresh catalyst, the desulfurization effect was only slightly reduced (<10%), indicating that the catalyst [C2(MIM)2]PW12O<sup>40</sup> had good recycling performance. In addition, the compounds in the oil phase before and after desulfurization were tested by gas chromatography-flame ionization detection (GC-FID). It can be seen from the results in Figure 10 that the oxidation product DBTO<sup>2</sup> of DBT was detected in the oil phase after the reaction, which indicated that some oxidation products were still in the oil phase. It could be inferred from the results that with the increase in the number of cycles, the decrease in desulfurization efficiency might be partly due to the presence of oxidation product in the oil phase, which resulted in the blocking of mass transfer in the reaction process. Figure 11 shows the infrared spectrum analysis of the recycled catalyst and fresh catalyst. After recycled tests, the structure of the catalyst was not destroyed, indicating that the catalyst has excellent stability.

**Figure 9.** Desulfurization efficiency of recycling system. Reaction conditions: V(oil) = 5 mL; n(catalyst)/n(S) = 0.025; n(H2O<sup>2</sup> )/n(S) = 6; t = 60 min; T = 50 ◦C; V(acetonitrile) = 0.5 mL.

**Figure 10.** GC-FID of the model oil containing DBT before and after desulfurization in the proposed system.

**Figure 11.** FT-IR spectra of fresh and recycled [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40.

#### *2.4. Oxidation Product and Reaction Mechanism*

The oxidation product of DBT was further verified by Gas Chromatography-Mass Spectrometry (GC-MS), and the results are shown in Figure 12. From the analysis results, it can be concluded that the oxidation product of DBT in model oil was DBTO<sup>2</sup> (*m*/*z* = 216.0). Based on our research and the related literature reports [52], we hypothesized the reaction mechanism of ECODS, as shown in Scheme 1. It was assumed that DBT was first extracted into the extraction phase by acetonitrile and POM-DILs with the extraction function. In the process of catalytic oxidation, [PW12O40] <sup>3</sup><sup>−</sup> in the catalyst was oxidized by H2O<sup>2</sup> to the intermediate product [PO4{W(O)(O2)2}4] <sup>3</sup><sup>−</sup> with strong oxidation, and H2O<sup>2</sup> was activated to form ·OH (the catalytically active O species). Then, by a redox reaction between DBT and the intermediate product and hydroxyl radical, DBT was selectively oxidized to DBTO2, and [PO4{W(O)(O2)2}4] <sup>3</sup><sup>−</sup> was reduced to the original [PW12O40] <sup>3</sup>−. As the reaction proceeded, DBT was continuously oxidized and extracted, and the content of DBT in the oil phase was continuously reduced, while the oxidation product DBTO<sup>2</sup> was continuously accumulated in the extraction phase until the end of the reaction.

**Figure 12.** Gas Chromatography-Mass Spectrometry (GC-MS) analysis of catalyst after reaction.

**Scheme 1.** Proposed mechanism for the oxidation of DBT in the ECODS.

#### **3. Materials and Methods**

#### *3.1. Materials*

1-Methylimidazole, 1,2-dichloroethane, 1,4-dichlorobutane, 1,6-dichlorohexane, DBT and H3PW12O<sup>40</sup> were purchased from MACKLIN (Shanghai, China). Acetone, H2O<sup>2</sup> (30%), acetonitrile, ethanol and n-octane were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). All the reagents and chemicals were directly used in experiments without any purification.

#### *3.2. Synthesis of Catalyst*

POM-DIL catalysts were prepared by a two-step method. Take [C2(MIM)2]PW12O<sup>40</sup> as an example, and the synthesis steps are shown in Scheme 2. 1-methylimidazole (3.284 g, 0.04 mol) and 1,2-dichloroethane (1.9792 g, 0.02 mol) were charged into a 100 mL roundbottomed flask with a condensation reflux device. Under solvent-free conditions, the reaction mixture was stirred at 90 ◦C for 2–4 h until white solid appeared. The white solid intermediate product [C2(MIM)2]Cl<sup>2</sup> was washed repeatedly with acetone to remove nonionic residues and dried at 80 ◦C in a vacuum for 6 h. Then, [C2(MIM)2]Cl<sup>2</sup> (0.3946 g, 1.5 mmol) and H3PW12O<sup>40</sup> (2.8800 g, 1.0 mmol) were dissolved in 30 mL distilled water and stirred for 2 h at room temperature. After centrifugation and drying, the final catalyst [C2(MIM)2]PW12O<sup>40</sup> was obtained. Preparation of [C4(MIM)2]PW12O<sup>40</sup> and [C6(MIM)2]PW12O<sup>40</sup> was similar to that of [C2(MIM)2]PW12O40. The results of <sup>1</sup>H nuclear magnetic resonance (1H NMR) spectroscopy (400 MHz, DMSO) characterization were as follows: [C2(MIM)2]PW12O40: δ 8.98 (s, 2H), 7.72 (s, 2H), 7.57 (s, 2H), 4.65 (s, 4H), 3.86 (s, 6H), 3.37 (s, 6H), 2.52 (s, 2H). [C4(MIM)2]PW12O40: δ 9.07 (s, 2H), 7.72 (s, 2H), 4.20 (s, 4H), 3.86 (s, 6H), 3.37 (s, 6H), 2.50 (s, 2H). 1.79 (s, 2H). [C6(MIM)2]PW12O40: δ 9.07 (s, 2H), 7.72 (s, 2H), 4.20 (s, 4H), 3.86 (s, 6H), 3.48 (s, 6H), 2.52 (s, 4H), 1.81 (s, 2H), 1.30 (s, 2H).

**Scheme 2.** Synthetic route of [C<sup>2</sup> (MIM)<sup>2</sup> ]PW12O40.

#### *3.3. Characterization*

<sup>1</sup>H NMR spectra were obtained on BRUKER AVANCE 400 (Karlsruhe, Germany) using dimethyl sulfoxide (DMSO) as the solvent. Fourier transform infrared spectroscopy (FT-IR) analyses were performed on a Nicolet 6700 FT-IR spectrometer (Thermo Fisher, Waltham, MA, USA) using KBr pellets at room temperature. XRD was performed on an Empyrean X-ray diffractometer (PANalytical B.V., Almelo, The Netherlands) equipped with Cu-Kα source. The scan speed and step size were 5◦/min and 0.02◦ , respectively. UVvis spectra were obtained with a UV-vis spectrometer (UVmini-1280, Shimadzu, Suzhou, China) in acetonitrile. TG analyses were carried out on Microcomputer differential thermal balance HCT-3 instrument (Beijing Hengjiu Scientific Instrument Factory, Beijing, China) from 35 to 800 ◦C, with a heating rate of 10 ◦C /min in N<sup>2</sup> atmosphere. The moisture content was determined on a Karl Fischer Moisture Titrator (870 KF Titrino plus, Heirishau, Switzerland). The main components of KF reagent were I2, SO2, pyridine (buffer) and methanol (solvent). XPS was carried out on ESCALAB250xi (Thermo Scientifle, Waltham, MA, USA) with a monochromatic Mg-Kα source with 1487 eV of energy to explore the surface composition. The contact angle tests were conducted on a contact angle instrument (JC2000D, Shanghai Zhongchen Digital Technic Apparatus Co. Ltd., Shanghai, China). The oxidation product of DBT in the model oil was measured by GC-MS (Agilent 5975C, Santa Clara, CA, USA) and the signals were collected from 4 to 16 min.

#### *3.4. Oxidative Desulfurization Process*

Model oil containing specific sulfide was selected for the test. In this experiment, DBT (500 mg/L) was used as model oil substrate and n-octane as solvent to form model oil. Firstly, a certain amount of catalyst, H2O<sup>2</sup> (30%) and acetonitrile were added to a 100 mL round bottomed flask containing 5 mL of model oil. Then, the mixture was magnetically stirred for a period of time in a thermostatic water bath. After the reaction, the mixture precipitated for a certain time, the upper oil phase was taken and the sulfide content was determined by GC-FID (VF-1column type; 30 m × 0.25 mm × 0.25 µm; column temperature: 230 ◦C; the temperature was raised from 100 to 230 ◦C at the rate of 20 ◦C /min and kept for 2 min; injection temperature: 300 ◦C; detector temperature: 320 ◦C). Each group of data was repeated at least three times. According to the initial and final sulfur content in the model oil, the desulfurization efficiency can be calculated as follows: S-removal efficiency (%) = (1 − S1/S0) × 100%, where S<sup>0</sup> is the initial sulfur content in model oil (mg/L); S<sup>1</sup> is the final sulfur content in the model oil (mg/L).

#### **4. Conclusions**

In this study, three kinds of imidazole-based POM-DIL catalysts were synthesized by a two-step method and applied to the removal of DBT from model oil with acetonitrile as the extractant and H2O<sup>2</sup> as the oxidant. Compared with the previous POM-IL desulfurization

system, the experimental conditions of the present work were greatly optimized. Under the optimum reaction conditions—n([C2(MIM)2]PW12O40)/n(S) = 0.025; n(H2O2)/n(S) = 6; V(acetonitrile) = 0.5 mL; T = 50 ◦C; t = 60 min—the removal rate of DBT reached above 98%. Meanwhile, [C2(MIM)2]PW12O<sup>40</sup> could be reused seven times without obvious catalytic activity loss, and the original Keggin structure of the POM-DIL catalyst was undamaged. The results showed that the catalyst displayed good catalytic activity, stability and recycling performance. This is because the double cation can effectively regulate the interaction between the catalyst and sulfide, oxidant, which has a great impact on enhancing its desulfurization performance. In addition, increasing the carbon chain lengths of the catalysts with short carbon chain displayed no significant effect on the desulfurization efficiency due to the good wettability for both H2O<sup>2</sup> and model oil. Finally, the oxidation product of DBT was proved to be DBTO<sup>2</sup> by GC-MS. Detailed analysis of the ECODS mechanism found that the W=O in the catalyst was the active center to activate H2O<sup>2</sup> to generate ·OH, which was beneficial to desulfurization process. This study provided a new reference for the development of an efficient catalyst for catalytic oxidative desulfurization.

**Author Contributions:** Conceptualization, B.H. and J.T.; methodology, B.H.; software, J.L.; validation, B.H.; formal analysis, J.L.; investigation, J.L. and Y.G.; resources, J.L.; data curation, J.L.; writing original draft preparation, J.L.; writing—review and editing, B.H. and J.T.; visualization, B.H.; supervision, B.H.; project administration, B.H.; funding acquisition, B.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Hubei Natural Science Foundation Project (2013CKB032) and The Doctoral Foundation Project of Hubei University of Technology (0701).

**Data Availability Statement:** The data presented in this study are available in article.

**Acknowledgments:** We would like to thank the teachers of the School of Materials and Chemical Engineering for their support in the inspection equipment used for experiments.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

