Methane-Assisted Catalytic Desulfurization in an Environmentally Benign Way

Petroleum is one of the most important natural resources for human beings, while the contained sulfur heteroatoms lead to a series of problems1. Therefore, a desulfurization process to reduce sulfur content is mandatory for clean petroleum utilization. Hydrodesulfurization is currently mature in industry, while this process is costly, energy intensive and environmentally unfriendly due to CO2 emission and H2S production2,3. Alternative cost-effective desulfurization process with environmentally benign sulfur-containing products remains unreported. Here we demonstrate that the desulfurization of a heavy oil model compound dibenzothiophene can be successfully achieved under methane environment over creatively designed dual catalyst system, generating a new sulfur-containing product CS2. Control experiments indicate that the presence of methane as well as catalyst components for direct desulfurization and methane activation are all required. The reaction process is better understood by extensive evidences from isotope labeling experiments, catalyst and product characterizations, density functional theory calculations and verication experiments, based on which a reasonable catalytic mechanism is proposed. It is found that methane-assisted desulfurization requires more stringent conditions, where sulfur vacancy abundance, methane activation capability and surface sulfur transfer are all indispensable. This study pioneers a transformational desulfurization route, which is more economically and environmentally attractive for petroleum processing industry.

for desulfurization process, the undervalued methane resources can be better utilized and the aforementioned reforming process can be avoided 16,17 . Besides, it would be highly preferred if other environmentally benign sulfur products rather than H 2 S can be generated. It is clear that the methaneassisted desulfurization process (MDS) can be a transformational route for clean oil production.
Dibenzothiophene (DBT) is one of the most commonly found sulfur-containing components in petroleum, especially in heavy oil 18,19 . It is widely accepted as a model compound for the study of heavy oil and is notably persistent to desulfurization due to the stable fused-ring structure 20 . Therefore, if DBT desulfurization is realized under methane environment, it is highly possible that other sulfur-containing species in crude oil can also be effectively converted. The chemistry involved could further provide valuable insights regarding possibilities and pathways to deal with unfavorable substances in a cost effective and environmentally benign way.
It is fully demonstrated in this proof-of-concept work that methane-assisted desulfurization is technically achievable over rationally designed dual catalyst system consisting of MDS1 catalyst (NiMoS/TiO 2 ) for direct desulfurization and MAC1 catalyst (AgGaCe/ZSM-5) for methane activation ( Fig. 1(a)). An environmentally benign sulfur-containing product carbon disul de (CS 2 ) is formed, which has never been publicly reported to the best of our knowledge. Control experiments prove that the presence of methane, MDS1 and MAC1 are all requisite for this ground-breaking desulfurization route. 13 C isotope labeling experiments suggest the participation of activated methane in the reaction, while 34 S isotope experiments indicate that the sulfur atom in DBT gets adsorbed and interchanged with sulfur vacancies in MDS1, triggering the subsequent reactions. Characterization of catalysts and products, extra con rmation experiments and DFT calculations are further performed and a validated reaction mechanism is proposed. It can be seen that different from hydrodesulfurization process, the methane-assisted desulfurization requires much more stringent conditions since the abundance of sulfur vacancies, the capability of methane activation and the surface sulfur transfer process are all indispensable. Therefore, unique catalyst design strategies are essential to this transformational MDS process, which has great potential to be much more economically feasible and environmentally friendly for petroleum industry.
DBT desulfurization under methane environment was carried out over a dual catalyst system (entry 1).
The design of the catalyst system is explained in Section 2.1 and 2.2 of the Supplementary Information. The overall analysis results with estimated error ranges are shown in Table S3. Acceptable mass, carbon and sulfur balances are veri ed in Table S4, and the detailed distribution of converted reactants and formed products are shown in Fig. 1(b). In entry 1, 1.4% methane conversion and 14.0% DBT conversion are realized, which cannot be fully attributed to experimental error, evidently showing the occurrence of reaction. In gas phase, ethane is observed as the main product and trace amount of H 2 is detected.
However, H 2 S is not detected in gas phase, which is particularly different from the traditional HDS process. In soluble phase, CS 2 is clearly observed as a product, which is distinctive in the MDS process.
Comparing with H 2 S, CS 2 has much lower toxicity, which is widely used as an environmentally benign solvent and intermediate for organic synthesis [21][22][23] . Therefore, the production of CS 2 , although the amount is temporarily limited, deserves extra attentions. Besides, considerable quantities of other valuable products such as benzene, toluene and xylenes (BTX) are also witnessed. These products might originate from the decomposition of biphenyl, which is also observed as a product and widely reported in the direct desulfurization (DDS) pathway of the HDS process, indicating the presence of DDS reactions 24 . Other detectable products include benzene and biphenyl derivatives with various side chain substituents, which might be generated through alkylation reactions. However, typical products generated from the hydrogenation (HYD) pathway of the HDS process such as cyclohexylbenzene are not observed, con rming the absence of HYD reactions in the MDS process 24 . Besides, the presence of high boiling point soluble products con rmed by simulated distillation (Fig. S3) and coke con rmed by thermalgravimetric analysis (Fig. S4) suggest the occurrence of polymerization and coking reactions.
Control experiments (entry 2~4) were further performed and the overall analysis results are shown in Table S5. The distribution of converted reactants and formed products are further compared with entry 1 in Fig. 1(b). Comparing entry 1 and entry 2 (no CH 4 ), it is evident that much more types and amounts of soluble products are generated with the presence of methane, which cannot be realized under nitrogen environment, indicating the importance of methane to participate in the reaction and form valuable products. It is worth noting that the facilitation of DBT desulfurization after methane introduction leads to the co-utilization of methane and sulfur-containing compounds, converting intractable and low valueadded reactants to valuable products, which is highly meaningful in both economic and environmental aspects. Comparing entry 1 and entry 3 (no MAC1), it can be seen that methane conversion decreases dramatically without MAC1, suggesting the essential role of MAC1 for methane activation. Considering the inert nature of methane due to its stable structure and high C-H bond energy of 435 kJ mol -1 (the highest among hydrocarbons), this conclusion is undoubtably reasonable: different from the intrinsic high activity of H 2 in the HDS process, a rationally designed catalyst for methane activation is critical for triggering the MDS process 25,26 . The conclusion also explains the lack of reports on the MDS process so far, since the feasibility is extremely low if a conventional HDS catalyst is solely employed, which is not capable of methane activation. Comparing entry 1 and entry 4, although methane conversion can be even higher over MAC1 in the absence of MDS1, the activated methane molecules are almost fully converted to ethane since the mass of converted methane is similar to that of produced ethane. The predominant formation of ethane can be explained by the self-combination of methyl radicals after methane activation, which is commonly reported in our previous studies 27,28 . Looking back to entry 1, it can be deduced that the presence of MDS1 opens up a new desulfurization pathway, which competes with the formation of ethane. Therefore, MDS1 is also indispensable for effective conversion of DBT to environmentally benign CS 2 and other valuable sulfur-free products.
To further rationalize active components in MDS1 and MAC1, more control experiments were designed and the results are shown in Table S6 and Fig. 1(c). Entry 1-O is a control experiment using MDS1-O catalyst, which is the oxide form of MDS1. The product distribution in entry 1-O is similar to that in entry 4 while totally different from that in entry 1, suggesting that the sul de-form catalyst MDS1 cannot be replaced by an oxide-form counterpart. This might be due to the fact that appropriate active sites for desulfurization, which are widely reported as sulfur vacancies, only exist on the surface of sul de-form catalysts 29,30 . Besides, MAC2 (no Ag), MAC3 (no Ce) and MAC4 (no Ga) were evaluated to investigate the function of metals in MAC1. According to Fig. 1(c), MAC2 and MAC3 leads to similar conversion of reactants and product distribution. The yields of desulfurized products such as biphenyl and BTX are close to those in entry 1, indicating the desulfurization reaction still goes on smoothly. However, much lower CS 2 yield is observed in these two entries, suggesting the active roles of Ag and Ce for the formation of CS 2 . Moreover, methane conversion and the yields of desulfurized products are extremely limited in entry 1-MAC4, indicating methane is not fully activated without Ga. The conclusion that Ga is essential for methane activation is also repeatedly reported in previous work 31,32 .
To further reveal the evolution of methane in the reaction process, an isotope labeling reaction entry1-13C was carried out using 13 C concentrated methane. The soluble product was analyzed by gas chromatography-mass spectrometry (GC-MS) and 13 C nuclear magnetic resonance (NMR), and the results are illustrated in Fig. 2, Fig. S5 and Table S7. According to the mass spectra in Fig. 2(a~c), the distributions of m/z for monocyclic aromatic products clearly shift to the higher end in entry 1-13C compared to those in entry 1, indicating methane incorporation in these products. The signal increase rate follows the order xylene (310 %) > toluene (234%) > benzene (181%), suggesting the 13 C atoms from 13 CH 4 are incorporated preferably on the side chain as methyl groups rather than in the aromatic ring.
This conclusion is further supported by liquid 13 C NMR analysis in Fig. S7. However, the signal increase rate of CS 2 is only 11%, implying that the carbon atoms in CS 2 should not be fully from methane ( Fig.   2(d)). Moreover, no signal increase is observed for biphenyl, indicating that the carbon atoms in biphenyl might fully come from DBT (Fig.2 (e)). The unique 13 C distribution provides extra evidence for the reaction mechanisms.
To further substantiate the important role of sul de catalyst in desulfurization process, entry 1-S and entry 1-34S were carried out using the catalyst derived from natural elemental S and 34 S, respectively. According to Fig. 2(f), the ratios of m/z intensities 78/76 (0.049) and 80/78 (0.042) for CS 2 in entry 1-S conform well with natural sulfur isotope abundance ( 34 S: 32 S=0.043). However, it is clearly seen that the intensities of m/z at 78 and 80 (correspond to C 32 S 34 S and C 34 S 2 ) increase in entry 1-34S dramatically, indicating the incorporation of S atoms from catalyst in the nal CS 2 product, which can be attributed to the sulfur exchange between 32 S captured from DBT and 34 S initially present in the isotopelabeled catalyst. This sulfur exchange is also reported to be critical for the HDS process in previous studies 33,34 .
The fresh and used MDS1 and MAC1 catalysts were systematically characterized and the results are summarized in Fig. 3 and Fig. S6~S14. The sulfur capturing capability of MDS1 is further con rmed by X-ray adsorption ne structure (XAFS) study as shown in Fig. 3(a) and Fig. S6. According to Fig. 3(a), the patterns of Mo K-edge in sample MDS1-O are similar to those in MoO 3 standard, con rming its oxide nature. However, the features in MDS1 and MDS1-Used are different from those in MoO 3 and MoS 2 , which seem to be a combination of both. The tting results in Table S8 substantiate that the coordination number of Mo-S in MDS1 is much lower than the theoretical value of 6, suggesting the presence of sulfur vacancies. Meanwhile, the coordination number of Mo-S increases signi cantly in MDS1-Used, proving that sulfur atoms from DBT have been captured by the sulfur vacancies of MDS1. Similar conclusions can be drawn from Ni K-edge and S K-edge spectra (Fig. S6 and Table S9). Microstructural analysis of molybdenum sul de in fresh and used MDS1 catalysts is further discussed in Fig. S9~S12.
X-ray photoelectron spectroscopy (XPS) results in Fig. 3(b) clearly suggest that Ag in MAC1-Used is sulfurized. Besides, local energy-dispersive X-ray spectroscopy (EDX) mapping results validate the coexistences of Ag-S and Ce-S in MAC1-Used (Fig. 3(c~e) and Fig. S13(b)). Therefore, although desulfurization occurs primarily over MDS1, a sulfur transfer process from MDS1 to MAC1 also takes place with the assistance of Ag and Ce during the reaction.
The formation of CS 2 is further supported by advanced scanning/transmission electron microscopy (STEM). In order to visualize the beam-sensitive porous structure of ZSM-5 and identify chemical compositions of absorbents in MAC1, integrated differential phase contrast (iDPC) STEM technique was employed to image fresh and used MAC1 catalysts. The framework of ZSM-5 is clearly seen from the [010] projection with atomic resolution in Fig. 3(g~h), which is in good agreement with theoretical model and previous TEM work 35 . In contrast to empty channels of the fresh catalyst in Fig. 3(g), channels of the used one are almost fully occupied by light absorbent molecules. The carbon and sulfur signals can be clearly identi ed in both integrated STEM-EELS (electron energy loss spectroscopy, Fig. 3(f)) and EDX spectra (Fig. S14(c)), which were collected from the ZSM-5 areas without sulfurized Ag or Ce compounds (Fig. S14(a~b)). These observations rationalize that the absorbent molecules should be carbon-sulfur compounds, suggesting that CS 2 might be formed and the process might occur preferably over MAC1.
The reaction mechanism was further explored by carefully designed veri cation experiments (Table  S10~S16). Vial-separated experiments (Table S10) where MDS1 and MAC1 are physically separated further differentiate the function of MDS1 and MAC1: the yields of BTX and biphenyl are higher over MDS1 than over MAC1, while the yields of other aromatics and coke are higher over MAC1 than over MDS1. These results suggest that DBT desulfurization and biphenyl decomposition mainly occur over MDS1, while alkylation take place preferably over MAC1. This conclusion is further corroborated by 13 C{ 1 H} solid-state cross-polarization magic angle spinning nuclear magnetic resonance (SS CP/MAS NMR) spectra (Fig. 3(i)) and other veri cation experiments (Table S11~S13). The preferred adsorptions of DBT over MDS1 and CH 4 over MAC1 are also supported by density functional theory (DFT) calculations ( Fig. 4(a~b)). Besides, the absence of CS 2 in vial-separated experiments also suggests that sulfur transfer from MDS1 to MAC1 is a surface process, which does not occur in the gas phase.
Based on all experiments, characterization and calculation results, a catalytic reaction mechanism is proposed as shown in Fig. 1(a). First, DBT is adsorbed on the sulfur vacancies over MDS1 (supported by Fig. 3(a), 4(a), S6, Table S8 and S9) and methane is activated by Ga over MAC1 (supported by Fig. 3(i), 4(b) and Table S11). Then, the radicals generated upon methane activation interact with DBT over MDS1 through the DDS pathway (supported by Fig. 4(c) and Table S15) to form an important sulfur-free product biphenyl and S atoms (supported by Fig. S6 and Table S10). Later, biphenyl participates in a series of reactions including decomposition and alkylation, resulting in the formation of other aromatic products such as BTX (supported by Fig. 3(i) and Table S12). Meanwhile, a surface sulfur transfer process from MDS1 to MAC1 occurs with the assistance of Ag and Ce (supported by Fig. 3(b~e) and Table S10). Subsequently, methyl group generated over MAC1 further triggers hydrogen transfer reactions and produces free C atoms, and the latter react with transferred S atoms to form the nal sulfur-containing product CS 2 (supported by Fig. 3(f~h), Table S14 and S16). Besides, it should be noted that a series of side reactions, such as ethane formation (supported by Table S11), coking (supported by Table S10~S14), methylation (supported by Table S10 and S13), methyl exchange and ring expansioncontraction of aromatics (supported by Fig. 2(a~c)) are also going on during the process, leading to more diverse product distribution as shown in Fig. 1(b). These side reactions are not shown in Fig. 1(a) for clarity. From this surface reaction network, it is clear that MDS1 and MAC1 are also indispensable for successful MDS process. It is worth noting that it is the creative catalyst design that enables the possibility of methane-assisted DBT desulfurization, which pioneers a new pathway for dealing with sulfur heteroatoms and generating new sulfur-containing product CS 2 with unique economic and environmental advantages.  Mass spectra of main soluble products. (a~e) benzene, toluene, xylene, carbon disul de and biphenyl in 13C isotope labeling experiment. (f) carbon disul de in 34S isotope labeling experiment. The numbers above the column indicate the increase rate of signal intensities in isotope labeled entry compared to those in entry with natural isotope abundance.