Effect of MoS 2 Yolk-Shell Nanostructure on the Thiophene Separation Performance of PEG Membrane

Constructing facilitated transport based on π -complexation has been drawing more and more attention in mixed matrix membranes (MMMs) for pervaporative desulfurization. Herein, a unique molybdenum disul ﬁ de (MoS 2 ) yolk-shell nanostructure (MYNS) was prepared and incorporated into the polyethylene glycol (PEG) matrix to fabricate MMMs for model gasoline desulfurization by PV. Moreover, the e ﬀ ects of MYNS content, feed sulfur concentration, and feed temperature on the performance of PEG/MYNS MMMs were evaluated. It was found that there is good interfacial compatibility between the MYNS ﬁ ller and the PEG matrix, and the resultant MMMs show enhanced swelling resistance against thiophene. The PV results revealed that the as-fabricated MMMs are thiophene-selective, and their desulfurization performance in the pervaporative removal of thiophene from n -octane is remarkably evaluated due to the addition of MYNS. The MMMs display the highest sulfur enrichment factor of 4.02 with an associated permeation ﬂ ux of 2587g · m − 2 · h − 1 with the MYNS loading of 3 wt. % when carrying out in an n -octane and thiophene (500 μ g · g − 1 ) mixture at 343 K. Furthermore, a consistent increment in the permeation ﬂ ux accompanied with a continuous reduction in the enrichment factor was observed with increasing the feed sulfur concentration and feed temperature. This work may o ﬀ er great potential for practical gasoline desulfurization applications.


Introduction
With the increasing environmental pollution, people have put more stringent requirements on the sulfur content in petroleum products. The deep desulfurization of fuel has become a very important issue in the oil refining industry [1,2]. Until now, the most commonly used desulfurization technology is hydrodesulfurization (HDS). However, the HDS process is high cost due to high temperature, high pressure, and high hydrogen consumption and results in a significant loss in gasoline octane number [3]. Moreover, it provides a very efficacious method for separating thiols and sulfides in gasoline except thiophene. Thiophene is the main sulfurcontaining compound in gasoline, accounting for about 60-80% of the total sulfur content [4,5]. Consequently, it is desirable to explore other lower cost and high efficiency desulfurization methods.
Pervaporation (PV) desulfurization technology has received more and more attention worldwide [6]. Compared with traditional desulfurization technology, it has many potential advantages, including low energy consumption, low pollution, high separation efficiency, and simple equipment [3,[6][7][8][9]. In recent years, there are a large number of investigations on photovoltaic gasoline desulfurization. Most studies focus on introducing processes, amplifying technology, exploring membrane materials, optimizing operation parameters, etc. [10][11][12][13][14]. Recent attempts have focused on membrane modification in order to improve the separation performance [7]. Introducing adsorptive fillers into polymer matrices is an effective and flexible approach to enhance the desulfurization performance of membranes by improving the adsorption capacity of the target permeate components. To date, many adsorbents, such as metal oxides, zeolite, and activated carbon, have been adopted for the desulfurization of gasoline [13,15,16].
Transition metal sulfides, which are similar to graphene in structure, have attracted much attention due to their special optical, electrical, and chemical properties. Twodimensional (2D) layered molybdenum disulfide (MoS 2 ) exhibits specific photoelectric, catalytic, biological, and energy harvesting properties [17]. The direct band gap of monolayer MoS 2 reaches 1.8 eV, showing light absorption and photoluminescence characteristics [18,19]. Therefore, MoS 2 has many unique optical and physical and chemical properties and is widely used in photocatalysis, energy, electronic chips, biological detection, medical imaging, and other fields [20]. Recently, researchers have employed 2D materials (graphene and its derivatives and MoS 2 nanosheets) with high specific surface area as well as adjustable interlayer spacing as fillers to prepare MMMs. The functional groups available on the 2D nanosheet surface, which are used as facilitated transport carriers for thiophene, can effectively improve the PV desulfurization efficiency of membranes [21][22][23]. The reasons are the following two attributes: (i) the polymer chain packing is disrupted by the addition of 2D materials with a high aspect ratio which provides more diffusion pathways; (ii) the specific and reversible π-π interaction between the facilitated transport carriers and thiophene molecules results in a faster diffusion rate of thiophene and an increase of the overall selectivity towards thiophene [21,23]. However, the MMMs prepared by incorporating 2D materials have some drawbacks. Due to the poor compatibility between the 2D materials and the polymer materials, the performance of the MMMs is poor, including poor separation performance, low swelling resistance, and poor stability.
In this work, a MoS 2 yolk-shell nanostructure (MYNS) was prepared. MoS 2 was encapsulated in hollow mesoporous carbon nanosheets like a photo in a plastic bag. This threedimensional (3D) structure combines the advantages of 2D ultra-thin nanostructure and hollow porous nanostructure and can overcome the inherent defect of poor compatibility between traditional 2D materials and polymer materials. The MYNS was incorporated as fillers in polyethylene glycol (PEG) to prepare MMMs for improving the desulfurization performance via PV. The thus-prepared MMMs were utilized for the removal of thiophene from n-octane. Thanks to the high surface area, large pore volume, and excellent dispersion of the MYNS, the PEG/MYNS MMMs showed enhanced desulfurization performance. Also, the effects of the content of MYNS, feed sulfur concentration, and operating temperature on the separation performance were analyzed systematically.
2.2. Synthesis of MYNS. First, 30 mg MoS 2 nanosheets were dispersed in a 135 mL ethanol and water mixture (volume ratio = 8 : 1) under ultrasound for 10 min at 30°C. Subsequently, the dispersion was mixed with 0.3 g polyvinylpyrrolidone, 6 mL ammonia solution, and 4 mL tetraethyl orthosilicate. After stirring thereafter for 6 hours, centrifugation was used to collect the MoS 2 @silica (SiO 2 ) products, followed by washing with deionized water. The precipitation was dispersed into 30 mL deionized water. After that, the dispersion was combined with a 270 mL ethanol and water mixture (volume ratio = 1 : 2). After mixing with 1.8 mL tetraethyl orthosilicate, 0.9 mL formaldehyde, 0.6 g resorcinol, and 0.9 mL ethylenediamine, the mixture was left to stir at 40°C for one day. The dispersion was centrifuged, and the resultant intermediate MoS 2 @SiO 2 @carbon precursor polybenzoxazine (PB) was dried at 60°C over 12 hours. After carbonization at 800°C for 4 hours in a nitrogen (N 2 ) atmosphere, the obtained was washed with a 10% hydrofluoric acid solution for one day. Eventually, the resulting dispersion was centrifuged and washed with deionized water. The desired final product MYNS was obtained after further drying under vacuum overnight.

Membrane Preparation.
The polymer PEG and the filler MYNS were mixed with the solvent 1-methyl-2-pyrrolidone. After adding the crosslinking agent maleic anhydride and the catalyst trimethylamine solution, the mixture was left to stir for 50 minutes. After degasification, the dope solution was cast on the PVDF support that had been immersed into deionized water for 3 more hours prior to casting to prevent dope solutions from penetrating. The as-prepared membranes were introduced into an oven at 90°C to allow for vaporization of the solvent and complete crosslinking, followed by drying in a vacuum oven for 24 hours at room temperature to ensure removal of residual solvent. The method for the preparation of the membrane samples that were used for sorption experiments was pouring the casting solution into Petri dishes instead of PVDF membranes. Advances in Polymer Technology ESCA, Perkin Elmer) using Mg Kα radiation source (hυ = 1253:6 eV). A homemade scanning tunneling microscopy (STM)/atomic force microscopy (AFM) system was applied to measure the surface roughness of membranes under an atmosphere of N 2 at room temperature. For assessing the thermal stability of membranes, the thermal gravimetric analysis (TGA) was conducted on a simultaneous thermal analyzer (STA 449 F3 Jupiter) in a temperature range from 30 to 800°C at a heating rate of 10°C·min −1 in a N 2 atmosphere. Sorption experiments were performed to evaluate the membrane swelling degree. The free-standing membrane samples were dried at 80°C in a vacuum oven for approximately 3 days, and their weights were recorded. Subsequently, they were immersed completely into thiophene solution and were left to soak for 48 hours at room temperature to achieve adsorption equilibrium. After that, the samples were removed and quickly wiped the excess solvent from their surface with filter papers and then weighed immediately. The swelling degree, SD, can be calculated as in which W d and W s represent the weights of initial dried and final swollen membrane samples, respectively. Three samples of each membrane were tested.

PV Experiments.
PV experiments were made on the selfdesigned apparatus described in our previous work [24]. The feed mixture was a model gasoline system composed of noctane and thiophene, and the thiophene concentration was controlled in the range of 500-1700 μg·g −1 . Feed solutions were stirred constantly using a submerged pump to eliminate the concentration polarization. A recirculating water bath was used to realize the stable control of the feed temperature ranging from 343 to 383 K. Permeates were trapped by combining purging with nitrogen and vacuuming and keeping the pressure below 20 kPa. The pervaporative desulfurization performance can be expressed in terms of the permeation flux (J, g·m −2 ·h −1 ) and the enrichment factor (β) which are calculated as follows: where Q is the weight of the permeate, A refers to the effective membrane area, and t represents the experiment time, respectively. ω F and ω P refer to the weight fractions of thiophene in the feed and permeate, respectively. The composition of the permeate was measured by gas chromatographymass spectrometry (GC-MS, 5977B GC/MSD, Agilent).

Results and Discussion
3.1. Characterization of MYNS. The FTIR spectrum of the MYNS is represented in Figure 1. A strong peak at around 3425 cm −1 is attributable to O-H stretching, which is due to the adsorption of water on the MYNS surface [25]. The absorption peak at 1590 cm −1 can be attributed to the 3 Advances in Polymer Technology bending of the water molecule. The characteristic peak at 469 cm −1 arises from S-S stretching vibration in MoS 2 [26]. The weak peak at 535 cm −1 and the strong peak at 1107 cm −1 are assigned to Mo-S and Mo-O stretching, respectively. The characteristic peak at 809 cm −1 corresponds to C-S stretching [27]. In addition, a weak absorption peak can be observed at 1395 cm −1 , which belongs to the stretching of C=S bond [28], indicating the formation of chemical bonds between the MoS 2 and the porous carbon coated on it. It is suggested that the prepared MYNS has high structural stability. It is worth noting that typical characteristic peaks of SiO 2 are not shown in the spectrum. That is because the middle SiO 2 layer of the intermediate MoS 2 @-SiO 2 @PB was completely removed by etching with the hydrofluoric acid solution. It is concluded that the yolkshell nanoarchitecture of MYNS, MoS 2 , and carbon with rich nanopores was successfully prepared. Figure 2 shows the morphology of the MYNS particles. In Figure 2, it can be observed that MYNS particles are basically spherical and have an average size of approximately 100 nm with uniform size distribution. Nanoscale MYNS particles as filler are instructive in obtaining defect-free MMMs and evaluating membrane separation performance in PV. Besides, it is also can be seen that the SEM image of MYNS shows fewer agglomerates, and the MYNS particles are almost monodisperse. MYNS is expected to be well 10 20 Intensity (a.u.) 5 Advances in Polymer Technology dispersed in the PEG matrix, reducing the formation of undesirable nonselective voids at the interface between them, hence yielding an escalating thiophene permselectivity. Figure 3 represents the XRD pattern of MYNS. A broad diffraction peak 2θ of about 23°and a weak diffraction peak at 2θ of near 44°are the classic characteristic peaks of (002) and (100) crystal planes of carbon materials, thereby confirming the presence of carbon structure in the resultant MYNS. Moreover, the diffraction peaks at 2θ = 14:8°, 33.0°, 39.9°, 50.1°, and 58.6°can be designated in the (002), (100),  Figure 4 shows the XPS spectrum of MYNS over the binding energy range of 0-1100 eV that was calibrated by the C 1 s peak at 284.8 eV. From the full element survey, peaks of oxygen, nitrogen, carbon, molybdenum, and sulfur are observed, and the percentages of corresponding components are accordingly calculated to be 12.85, 4.07, 82.19, 0.20, and 0.69, respectively. This demonstrates that carbon  Advances in Polymer Technology was wrapped on the MoS 2 surface, forming the yolk-shell structure. Figure 5 presents the surface and cross-sectional morphologies of the pure PEG membrane and PEG/MYNS MMMs with different MYNS loading ranged in 1-5 wt. %. From the surface SEM images listed in the first column, it is noticeable that the pure PEG membrane surface is smooth and dense, and the PEG/ MYNS membrane surface becomes rough and uneven with increasing the MYNS loading. No pinholes or obvious cracks are found on the surface of all the membranes, which is important for membrane separation. From the crosssectional SEM images displayed in the second column, it can be found that the interface between MYNS fillers and PEG matrix is indistinct, indicating good interfacial compatibility between them. It is worth mentioning that MYNS still disperses well within the PEG matrix even when the MYNS loading is up to 3 wt. %. However, excessive filler aggregation is evident both on the membrane surface and in the membrane with the further increase of the MYNS loading (over 3 wt. %), inducing the formation of nonselective defects, hence resulting in the degradation of separation performance.

Characterization of Membranes.
The ATR FTIR spectrum of the PEG/MYNS MMM with 3 wt. % MYNS is presented in Figure 6. A weak absorption band at around 3500 cm −1 , which is characteristic of, is designated to hydroxyl (-OH) groups. The peaks at 2877 and 1467 cm −1 are characteristic for methylene (-CH 2 ) stretching and deformation vibration, respectively. The peaks at 1642 and 1242 cm −1 belong to C=C and C-O stretching vibrations, respectively. It is noted that the representative two absorbances at around 1800 and 1750 cm -1 originated from the stretching vibration of two neighboring carbonyl (C=O) groups of maleic anhydride are not observed in the spectrum. Conversely, an absorption peak at 1720 cm −1 related to C=O stretching vibration appears. It is indicated that the crosslinking reaction occurs between PEG and maleic anhydride. The schematic representation of the crosslinking reaction is shown in Scheme 1. The strong peak at 1107 cm −1 displayed in the spectrum proves the presence of MYNS.    [16] demonstrated that the membrane surface area has a positive correlation with the permeation flux. From this point of view, incorporating MYNS into the PEG will evidently improve the membrane permeation flux.
The swelling degree of as-fabricated pure PEG and PEG/ MYNS membranes with different MYNS contents was tested in thiophene, and the obtained results are given in Figure 8. A large reduction in the swelling degree for all the MYNS filled PEG MMMs is found as compared with the PEG pristine membrane, which indicates a higher antiswelling capac-ity of PEG membranes towards thiophene due to the incorporation of MYNS. Furthermore, the swelling resistance increases as the MYNS content increases. According to the study reported by Gonciaruk et al. [29], the introduction of MYNS into PEG has an effect on the polymer chain packing. The PEG chain fragments are arranged parallel to the MYNS sheets, leading to a decrease in the chain mobility and potentially preventing the membrane from swelling. The high swelling resistance in the desulfurization can ensure high performance stability during a long-term operation which is crucial in practical applications.
Thermal stability of the pure PEG membrane and PEG/ MYNS MMM with an MYNS loading of 3 wt. % was tested by TGA, and the resulting thermograms are presented in Figure 9. It can be found that the two weight loss curves exhibit a similar thermal degradation trend, confirming little influence on the thermal stability of PEG membrane by the incorporation of MYNS. Notably, both the pure PEG   Figure 11: Effect of thiophene concentration in feed on desulfurization performance of PEG/MYNS MMMs. 8 Advances in Polymer Technology membrane and PEG/MYNS MMM start to decompose at about 390°C. It is inferred that the membranes are thermally stable below 390°C. Therefore, the thermal stability of the obtained PEG/MYNS MMMs can completely satisfy the requirement of a PV operation for gasoline desulfurization.

PV Performance of Membranes
3.3.1. Effect of MYNS Content. The effect of MYNS content on the permeation flux and enrichment factor of PEG/ MYNS MMMs was investigated using a thiophene concentration in feed of 500 μg·g −1 and keeping operation at a temperature of 343 K. The obtained result is plotted in Figure 10.
Obviously, the permeation flux keeps an upward tendency, whereas the sulfur enrichment factor presents a parabolic increment with the temperature rising. When the content of MYNS is 3 wt. %, the permeation flux is 2587 g·m −2 ·h −1 , which is about two times that of the pristine PEG membrane (1355 g·m −2 ·h −1 ). Total flux gains are primarily attributed to the augment of the accessible free volume within membranes due to the incorporation of MYNS, which includes the inter-molecular spacing between the polymer chains, interface between PEG and MYNS, and the pores in MYNS, therefore providing more diffusional pathways for thiophene and noctane molecules transport. In addition, based on facilitated transport mechanism [30], the embedded MYNS served as facilitated transport efficacious carriers accelerates the mass transfer rate of thiophene through the reversible π-π interaction, which results in an increase in enrichment factor. However, when the embedded MYNS content exceeds 3 wt. %, enrichment factor declines due to the MYNS particle agglomeration that could create nonselective interfacial voids to foster n-octane transport through the membranes and cause the overlap of facilitated transport sites to weaken thiophene transport [21]. The maximum sulfur enrichment factor afforded by the PEG/MYNS MMM is 4.02 which corresponds to an MYNS content of 3 wt. %. 9 Advances in Polymer Technology Figure 11. As shown in Figure 11, permeation flux increases continuously with the increasing of the thiophene concentration within the range of 500-1700 μg·g −1 , while the enrichment factor follows the reverse order. Because of the larger affinity of the MMMs towards thiophene, more thiophene molecules are adsorbed within the membrane with increasing of the thiophene concentration, thus exacerbating the swelling of the membrane which imparts larger polymeric interchain spacing and subsequently improves the free volume. Accordingly, the permeability of thiophene and noctane is elevated simultaneously. Yet, given that n-octane molecules have a relatively larger kinetic diameter of about 0.63 than thiophene molecules (circa 0.53 nm) [31], the noctane flux increases more dramatically and therefore, the selectivity towards thiophene declines. Additionally, it is known that the partial vapor pressure difference increases along with the increment of feed concentration, therefore improving the transport driving force for the target component through the membrane. Nevertheless, it is limited by the fact that the thiophene concentration in feed is of an exceptionally low level compared with n-octane. Therefore, the contribution from the gain of thiophene flux to the increment of the total flux is infinitesimal. Generally, for all these reasons, an increasing permeation flux but a decreasing enrichment factor is observed.

Effect of Feed
Temperature. The effect of feed temperature on the desulfurization performance of PEG/MYNS MMMs containing 3 wt. % with a 500 μg·g −1 thiophene and n-octane solution is shown in Figure 12. It can be seen that the permeation flux increases with a trend of continuous enlargement over a temperature range from 343 to 383 K. This result mainly derived from the following: (i) the enhanced chain mobility of PEG, increasing the fractional free volume in the membrane [22], (ii) the improved saturated vapor pressure difference between the feed and permeate sides because of the elevation of vapor pressure on the feed side, increasing the transport driving force [23], and (iii) the accelerated molecular diffusivity [32]. On the contrary, the sulfur enrichment factor decreases monotonically with rising the feed temperature. The amplitude of thiophene increment, as depicted in Figure 12(b), is appreciably lower than that of n-octane, which weakened its contribution to total flux, resulting in a significant decrement in sulfur enrichment factor.
The correlation between thiophene and octane fluxes and feed temperature can be expressed by Arrhenius equation as follows: where J P , A i , E P , T, and R represent the thiophene or noctane flux, preexponential factor, apparent activation energy, feed temperature, and gas constant, respectively. Figure 13 plots ln J P versus 1000/T. It can be found that there is a favorable linear relationship between them. On the basis of the slope of the fitted straight lines, the apparent activation energy of n-octane calculated is 13.4 kJ·mol −1 , which is nine to ten times that of thiophene of 1.4 kJ·mol −1 . It is confirmed that the permeation of n-octane is more sensitive to the feed temperature than that of thiophene. This is consistent with the results of the considerable increase in the permeation flux of n-octane but inconspicuous variation in that of thiophene as the feed temperature goes from 343 to 383 K. As a result, a growing permeation flux but a declining enrichment factor was achieved.  10 Advances in Polymer Technology possesses good and acceptable comprehensive performance although it does not have the highest permeation flux or the highest sulfur enrichment factor. In addition, it can be found that its overall desulfurization performance obviously outperforms those of other pure polymeric membranes reported. Accordingly, it is inferred that the PEG/MYNS MMMs may be prospective membrane materials for practical pervaporative desulfurization application.

Conclusion
In this work, MYNS was prepared and doped into the PEG matrix to prepare MMMs which were utilized in the desulfurization of model gasoline by PV. The results manifest that the MYNS particles have high interfacial compatibility with the PEG matrix, and the incorporation of MYNS endows the PEG membrane with improved swelling resistance. Moreover, the as-prepared MMMs filled with MYNS showed thiophene-selective, and the overall PV performance in terms of both permeability and selectivity is enhanced for the thiophene removal from n-octane as compared to the pure PEG membrane. The MMMs containing 3 wt. % MYNS deliver the highest sulfur enrichment factor of 4.02 coupled with a total permeation flux of 2587 g·m −2 ·h −1 when separating the 500 μg·g −1 thiophene-containing n-octane solution at 343 K. In addition, the total permeation flux is improved as increasing the thiophene concentration in feed and feed temperature, whereas the enrichment factor decreases all along. This work may provide useful insights of MMMs for practical gasoline desulfurization applications.

Data Availability
The data used to support the findings of this study are included within the article.

Conflicts of Interest
The authors declare that they have no conflicts of interest.