Multi-response optimization of MIL-101 synthesis for selectively adsorbing N-compounds from fuels
- 288 Downloads
In this work, MIL-101, a metal organic framework, has been synthesized and examined in the adsorptive denitrogenation process. Due to the importance of adsorption capacity and selectivity, the effects of synthesis parameters including metal type, reagent ratio, time and temperature on the MIL-101 performance were investigated by measuring quinoline (QUI) separation from iso-octane. The optimum conditions were determined using a Taguchi experimental design and the multi-response optimization (multivariate statistical) method. Based on the arithmetic mean of normalized QUI adsorption capacity and QUI/dibenzothiophene (DBT) selectivity, as the objective function, the optimum value of synthesis parameters were found to be manganese as metal type in the structure, 180 °C for synthesis temperature, 15 h for synthesis time and 1.00 for reagent molar ratio. Under these conditions, QUI adsorption capacity and QUI/DBT selectivity were 19.3 mg-N/g-Ads. and 24.6, respectively. Accordingly, the arithmetic mean between normalized values of these measured parameters was equal to 1.10, which is in good agreement with the predicted value. The MIL-101 produced under optimum conditions was characterized by determining its specific surface area, X-ray powder diffraction patterns and Fourier transform infrared spectroscopy. Finally, isotherm and kinetic studies indicate that the Langmuir isotherm and pseudo-first-order model can successfully describe the experimental data.
KeywordsMIL-101 Adsorptive denitrogenation Taguchi experimental design Multi-response optimization
Metal–organic frameworks (MOFs) are one of the fastest growing materials that have attracted much attention. MOFs are tunable crystalline hybrid networks with high porosity, specific functionality and selective interaction (Li et al. 2010a, b). In the last decade, these materials have interested researchers as their properties such as pore size and shape, surface area and special physicochemical surface properties can easily be regulated and modified by altering the connectivity of organic–inorganic moieties (Hu et al. 2018). MOFs can be used with different roles in a large number of processes including gas adsorption and storage, adsorptive removal of hazardous materials, catalysis, drug delivery, carriers for nanomaterials, electrode materials, luminescence and magnetism (Li et al. 2010a, b).
Due to the problems of the presence of nitrogen-containing compounds (NCCs) in fuel such as NOx emission and spoiling hydrodesulfurization (HDS) process, the separation of NCCs before removing sulfur-containing compounds (SCCs) is crucial (Khan et al. 2017). Adsorptive denitrogenation (ADN) is one of the crucial technologies that are highly environmentally friendly and do not need any hydrogen usage for removing NCCs from fuel. Adsorption of NCCs from fuel is a promising process in which MOFs can be successfully utilized (Sentorun-Shalaby et al. 2013). Several MOFs have been used for ADN, among which MIL-101 had notable adsorption capacity and selectivity, for removing NCCs from fossil fuels (Ahmed et al. 2013a). Different techniques such as changing metal ions (Songolzadeh et al. 2019) and adding Lewis acid sites (Ahmed et al. 2013a; Ahmed and Jhung 2014) have been tested to improve ADN characteristics including adsorption capacity and selectivity. Varying and optimizing reaction conditions of MOF synthesis, so far not investigated for ADN, are other effective techniques that can be used to prepare superior adsorbents for this process. Several methods such as solvothermal, slow evaporation, microwave-assisted, sonochemical, mechanochemical and electrochemical synthesis are used for the synthesis of MOFs (Ahmed and Jhung 2016). Solvothermal, a simple and environmentally friendly method, has usually been applied for the MOF synthesis. Based on numerous studies (Taheri et al. 2018; Li et al. 2011; Stock and Biswas 2011), temperature and time of synthesis, reagent concentrations and ratios, solvent and metal are fundamental effective parameters in MOF synthesis. Consequently, an alteration in any of these synthetic factors can change the adsorbent properties and separation characteristics of MOFs.
The objective of the present work is to investigate the effects of synthetic factors on the capacity of MIL-101 in the ADN process and to optimize parameters for maximum adsorption efficiency by experimental design. In this study, two compositional factors, namely the type of metal ion and the reagent ratio, and two process parameters including time and temperature were considered as extremely crucial in order to produce a high-quality MIL-101 for the ADN process. Regarding the long time needed for and the difficulty of MOF synthesis, the Taguchi method was used to design experiments, decrease the number of experiments, minimize their error, evaluate the effects of variables and optimize the performance of this process. ADN performance was determined with adsorption capacity and selectivity of quinoline which is the most common basic N-compound in liquid fuel. Therefore, a combination of two responses consisting of adsorption capacity and selectivity was applied for statistical analysis.
2 Materials and methods
Iso-octane (C8H18, 99.5%), terephthalic acid (TPA, 98.0%), N,N-dimethylformamide (DMF, 99.5%) and QUI (97.0%) were procured from Merck Company. Manganese (III) oxide (Mn2O3, 99.9%), vanadium pentoxide (V2O5, 99.0%) and chromium trioxide (CrO3) were provided by Sigma-Aldrich Company. In this study, all chemicals were used without further purification.
2.2 MOFs synthesis
Synthetic factors and their levels for synthesizing MIL-101 (Cr), MIL-101 (V) and MIL-101 (Mn)
No. of variables
Reagent molar ratio (molar ratio of metal oxide per TPA)
The X-ray diffraction (XRD) pattern was determined with a Philips PW 1730 diffractometer. NOVA series 1000-Quantachrome instruments were used to measure N2 sorption isotherms at −196 °C and calculate the specific surface area of the optimum synthesized MOFs. Before BET analysis, the samples were degassed down to 10−4 mbar at 150 °C for 5 h. FTIR patterns of optimum samples were recorded using a Nicolet Nexus 670 spectrometer FTIR at 4 cm−1 resolution within the range of 4000–400 cm−1.
2.4 Design of experiments
Selected experimental array (L9) and obtained results for adsorption capacity and selectivity
Reagent molar ratio
Adsorption capacity, mg-N/g-Ads.
Arithmetic mean of normalized capacity and selectivity
Adsorption capacity, mg-N/g-Ads.
Arithmetic mean of normalized capacity and selectivity
2.5 Adsorption experiments
3 Results and discussion
3.1 Analysis of variance results
Analysis of variance (ANOVA) for QUI adsorption capacity on MIL-101, QUI/DBT selectivity and arithmetic means
Sum of squares
Reagent molar ratio
Reagent molar ratio
Z1(arithmetic mean of normalized capacity and selectivity)
Reagent molar ratio
R2 and adjusted R2 of the QUI adsorption capacity model were obtained to be 82.8% and 75.2%, respectively. On the other hand, F value and p values, which are listed in Table 3, were used to determine the significance of each parameter. In general, the more significant parameter has a smaller p value and larger F value in ANOVA (Rahimi et al. 2015a). Based on Table 3, temperature with a p value of ≤ 0.001 and F value of 27.3 was the most significant parameter in the capacity model. Since p values ≥ 0.050 indicated the model terms that were not significant, it is concluded that changing the reagent molar ratio, with p values equal to 0.502, did not significantly affect the QUI adsorption capacity.
In case of QUI/DBT selectivity, the obtained R2 and Radj2 were 73.9% and 62.3%, respectively. According to Table 3, the rank order of the F value of each parameter follows: (1) temperature (14.25), (2) metal type (7.22), (3) time (2.48) and (4) reagent molar ratio (1.51). Therefore, it is clear that the temperature and metal type are the two most significant factors in the QUI/DBT selectivity. Furthermore, the importance of each parameter on the response can be evaluated by calculating and comparing their percentage contribution. Based on the ANOVA results, the two most influential factors in QUI/DBT adsorption selectivity were the temperature (56%) and metal type (28%). These parameters have the greatest effect on the QUI/DBT selectivity over MIL-101.
Comparing ANOVA results for four models shows that the described model on the arithmetic mean of responses with the biggest F value (54.64), R2 (85.86%) and adjusted R2 (79.57%) is the best and most impressive for this optimization. Similar to other models, temperature with F value of 33.09 and metal type with F value of 13.46 are the two most significant factors, which are significantly affected QUI adsorption capacity and QUI/DBT selectivity.
3.2 Effect of adsorption characteristic parameters
3.2.1 Effect of parameters on adsorption capacity
Figure 2b shows the effect of temperature on the QUI adsorption over MIL-101. Comparing Fig. 2b with a, c and d, it can be confirmed that temperature is a major and the most effective parameter on this response and QUI adsorption capacity increased dramatically (21%) when temperature was raised from 120 to 180 °C. This effect has been attributed to the fact that at higher temperatures a high solubility of reactants and large crystals of high quality would be obtained (Bag et al. 2015; Seetharaj et al. 2016; Zhang et al. 2013); therefore, synthesized MOFs in these conditions have better adsorption performance.
Figure 2c implies that the synthesis time was the parameter able to change the performance of the synthesized adsorbent. It was observed that increasing the synthesis time from 10 to 15 h leads to increasing QUI adsorption capacity from 14.46 to 16.76 mg-N/g-Ads. It is also evident that with an increase in the synthesis time, the probability of crystallization would be higher, which accelerates the QUI adsorption. However, with prolonging crystallization time, the behavior of QUI adsorption varied significantly, and it decreased sharply by increasing this time from 15 to 20 h (Aghaei and Haghighi 2015; Wee et al. 2012). This effect has been attributed to the relative large particle size that leads to long diffusion lengths and the hindrance of metal of these MOFs.
The effect of reagent molar ratio on QUI adsorption capacity is shown in Fig. 2d. According to mechanisms of NCCs adsorption on MOFs, the QUI adsorption capacity did not significantly depend on the molar ratio of the reactants, because this parameter affected the topological pattern of MOFs and did not affect the metal cluster geometric structure.
3.2.2 Effect of parameters on QUI/DBT selectivity
It was observed that increasing temperature from 120 to 180 °C leads to raising QUI/DBT selectivity from 13.0 to 18.0 (Fig. 3b). It is also evident that with an increase in temperature, larger crystals with high quality are formed (Bag et al. 2015; Zhang et al. 2013; Seetharaj et al. 2016), which enhance the probability of selective QUI adsorption on the active sites of the adsorbent (metals in MOFs structure).
Figure 3c shows the effect of the synthesis time on the QUI/DBT selectivity over synthesized MOFs. It was considered that the maximum selectivity of QUI/DBT separation was related to the medium crystallization time (15 h). In case of increasing synthesis time from level 1 (10 h) to level 2 (15 h), the selectivity was increased by providing the crystals enough time to grow well-formed and consequently letting QUI and DBT to easily come in contact with metal sites of MOFs which can selectively adsorb QUI. However, when synthesis time increases from 15 to 20 h, the QUI/DBT selectivity decreases from 16.0 to 15.2. As mentioned in investigating the effect of synthesis time on adsorption capacity, by prolonging this time, after 15 h the obtained MOFs have a larger particle size, causing long diffusion lengths and hindrance of metal sites of these adsorbents.
The trend of changing selectivity with reagent molar ratio is similar to that with crystallization time. Based on ANOVA, the reagent molar ratio has a 6% contribution percentage in the selectivity model. Since Lewis acid–base and π–M interaction are two major mechanisms for NCC selective adsorption, varying the amount of metal in the MIL-101 structure, which occurred with changes in the reagent molar ratio, can affect QUI/DBT selectivity. Regarding Fig. 3d, the selectivity ratio increased sharply to 0.67 (level 2). It is evident that the possibility of Lewis acid–base and π–M interaction increases with increasing reagent molar ratio and thus with the amount of metal in the unit cell in the synthesis of MOFs. Moreover, at the high reagent molar ratio (> 0.67), QUI/DBT selectivity had an indirect correlation with this ratio. This is explained by the fact that the steric hindrance of metal clusters in MIL-101 increased by raising the metal ratio in MOFs synthesis (Feng and Xia 2018).
3.3 Optimization of NCC adsorption characteristics
As mentioned earlier, the QUI adsorption capacity, QUI/DBT selectivity and arithmetic mean were investigated as three objective functions to optimize effective parameters in MIL-101 synthesis for QUI adsorption from model fuel. Based on the obtained results, it seems that the arithmetic mean is the best response and the objective function for predicting experiment data and optimizing effective parameters in this case.
Based on the model of arithmetic mean of QUI capacity and QUI/DBT selectivity, the optimal conditions were obtained as metal type = Mn, synthesis temperature = 180 °C (which was the maximum studied temperature), synthesis time = 15 h and reagent molar ratio = 0.67. The QUI/DBT selectivity for synthesized adsorbent under optimum conditions was predicted as 23.5 by the model for this selectivity. Due to the optimum condition not presented in the 27 experiments of the orthogonal array, confirming experiments were carried out to judge the predicted results. For the synthesized adsorbent under optimum conditions, the experimental values of QUI adsorption capacity and QUI/DBT selectivity were obtained as 19.3 mg-N/g-Ads. and 24.6, respectively. According to these results, values of Z1, which were predicted to be 1.04, were calculated as 1.10. Therefore, it was shown that the experiment and model results were in good agreement with relative errors of just 4.2%. In this case, for simplification, a MIL-101 synthesized under the above optimal conditions was called optimized MIL-101 for ADN (OMA). In the previous studies (Wu et al. 2014; Ahmed and Jhung 2014), the maximum values of QUI adsorption capacity and QUI/DBT selectivity were, respectively, reported 8–11 mg-N/g-Ads and 3.1–12.2, which were much lower than the measured values for OMA.
3.4 The characteristics of optimized adsorbent
By comparing Fig. 5a with b, it is clear that the FTIR pattern of OMA after QUI adsorption has two new peaks at 1737 cm−1 and 2922 cm−1. The FTIR spectroscopic results of QUI, which are obtained from results of Linstrom and Mallard (2018) and presented in Fig. 5c, were used to investigate the reasons of appearance of two new peaks in FTIR results of OMA after QUI adsorption. Except for the two highlighted peaks which appear at 1700–1800 cm−1 and 2900–3100 cm−1, respectively, other peaks of the QUI FTIR pattern approximately match major peaks of the FTIR pattern of OMA, and the QUI adsorption over the OMA can be determined with these two peaks. Therefore, it can be concluded that the difference between FTIR patterns of OMA before and after QUI adsorption can be attributed to the interaction of QUI with active sites of the adsorbent.
3.5 Adsorption isotherms
Langmuir and Freundlich isotherm and thermodynamic parameters for QUI adsorption over OMA
qmax, mg g−1
KL × 103, L mg−1
∆H°, kJ mol−1
∆S°, kJ mol−1 K−1
∆G°, kJ mol−1
3.6 Adsorption kinetics
The equations of these kinetic models (Eqs. 9 and 10) were fitted to the experimental data to define rate-limited mode and compare the accuracy of each kinetic model. The values of kinetic parameters were calculated to be kp1 = 7.29 × 10−4 s−1 and kp2 = 2.993 × 10−5 (g-Ads/mg-N) s−1. The R2 values for the pseudo-first-order and pseudo-second-order model were calculated as 0.986 and 0.772, respectively. Based on the obtained R2, it is concluded that the pseudo-first-order model is more accurate and suitable for studying QUI adsorption on OMA. These results demonstrate that QUI adsorption on OMA was controlled by physisorption which verified the results obtained in isotherm studies.
In order to improve NCCs removal from fuel, the MIL-101, as a classification of new metal organic frameworks (MOFs), has been synthesized, characterized and used for the ADN process. The Taguchi method and multi-response optimization were applied in synthesizing MIL-101 to find a superior adsorbent for the ADN process. Results of ANOVA showed that the model on the arithmetic mean (Z1 model) was the most valid one. Based on the Z1 model, the optimum conditions for MIL-101 synthesis were found to be Mn as metal type, 180 °C (which was the maximum studied temperature) as synthesis temperature, 15 h as synthesis time and 0.67 as reagent molar ratio, where the QUI adsorption capacity and QUI/DBT selectivity were obtained as 19.3 mg-N/g-Ads. and 24.6, respectively. In addition, temperature has the greatest effect of 56%–63% and the reagent molar ratio was not a striking factor in ADN characteristics. The Langmuir isotherm model successfully described the QUI adsorption in equilibrium conditions and indicates that this adsorption takes place at a monolayer coverage on a homogeneous adsorbent surface. Calculated thermodynamic parameters (especially ∆H° = 67.59 kJ mol−1) and kinetic studies results (experimental data following the pseudo-first-order model) demonstrate that QUI adsorption on MIL-101 might involve physisorption.
- Aghaei E, Haghighi M. Effect of crystallization time on properties and catalytic performance of nanostructured SAPO-34 molecular sieve synthesized at high temperatures for conversion of methanol to light olefins. Powder Technol. 2015;269:358–70. https://doi.org/10.1016/j.powtec.2014.09.036.CrossRefGoogle Scholar
- Antony J. Multi-response optimization in industrial experiments using Taguchi’s quality loss function and principal component analysis. Qual Reliab Eng Int. 2000;16(1):3–8. https://doi.org/10.1002/(SICI)1099-1638(200001/02)16:1%3c3:AID-QRE276%3e3.0.CO;2-W.CrossRefGoogle Scholar
- Gražulis S, Daškevič A, Merkys A, Chateigner D, Lutterotti L, Quiros M, et al. Crystallography Open Database (COD): an open-access collection of crystal structures and platform for world-wide collaboration. Nucleic Acids Res. 2011;40(D1):D420–7. https://doi.org/10.1093/nar/gkr900.CrossRefGoogle Scholar
- Heidari H, Razmi H. Multi-response optimization of magnetic solid phase extraction based on carbon coated Fe3O4 nanoparticles using desirability function approach for the determination of the organophosphorus pesticides in aquatic samples by HPLC–UV. Talanta. 2012;99:13–21. https://doi.org/10.1016/j.talanta.2012.04.023.CrossRefGoogle Scholar
- Kuptsov A, Zhizhin GN. Handbook of Fourier transform Raman and infrared spectra of polymers. Amsterdam: Elsevier; 1998.Google Scholar
- Linstrom PJ, Mallard WG. NIST Standard Reference Database Number 69. NIST Chemistry WebBook. 2018. https://doi.org/10.18434/t4d303.
- Mittal KL, Shah DO. Adsorption and aggregation of surfactants in solution. New York: Taylor & Francis; 2005.Google Scholar
- Montgomery DC. Design and analysis of experiments. Hoboken: Wiley; 2017.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.