Potassium promotion effects in carbon nanotube supported molybdenum sulfide catalysts for carbon monoxide hydrogenation

Catalysis Today 261 (2016) 137–145

Contents lists available at ScienceDirect

Catalysis Today journal homepage: www.elsevier.com/locate/cattod

Potassium promotion effects in carbon nanotube supported molybdenum sulfide catalysts for carbon monoxide hydrogenation Chang Liu, Mirella Virginie, Anne Griboval-Constant, Andrei Y. Khodakov ∗ Unité de Catalyse et de Chimie du Solide, UMR 8181 CNRS, USTL-ENSCL-EC Lille, Cité Scientifique, 59655 Villeneuve d’Ascq, France

a r t i c l e

i n f o

Article history: Received 15 April 2015 Received in revised form 10 July 2015 Accepted 13 July 2015 Keywords: Molybdenum sulfide Fischer–Tropsch synthesis Carbon nanotubes Olefins Oxygenates

a b s t r a c t The paper focuses on the effect of potassium promotion on the structure and catalytic performance of carbon nanotube supported molybdenum sulfide catalysts for carbon monoxide hydrogenation. A combination of characterization techniques showed the presence of MoO2 and mixed K-Mo oxides in the calcined catalysts. The sulfidation of oxide phases leads to MoS2 and K-Mo sulfides. MoO2 showed somewhat lower extent of sulfidation compared to other molybdenum oxide species. MoS2 was principally responsible for CH4 production, while lighter olefins, paraffins, alcohols and higher hydrocarbons were produced on the mixed K-Mo sulfides. The catalyst basicity seems to be one of the important factors controlling the reaction selectivity; moderate basicity is essential for higher rates of olefin and alcohol synthesis. © 2015 Elsevier B.V. All rights reserved.

1. Introduction The gasification of biomass and coal is considered as one of the most efficient ways to convert the energy embedded in fossil and renewable resources, which could be used as substitutes for oil-based fuels. The syngas produced in the gasification can be then converted to numerous value-added products (methane, olefins, long chain paraffins, methanol, dimethyl ether, higher alcohols, etc.). Light olefins represent important feedstocks for the petrochemical industry [1,2]. In the industry, olefins are typically obtained by nafta and ethane hydrocracking or via methanol-to-olefins (MTO) synthesis. The hydrocracking suffers from low olefin selectivity, while the catalyst stability is a major challenge in MTO synthesis. Direct conversion of syngas into light olefins via Fischer–Tropsch to olefins (FTO) process is an interesting alternative compared to petroleum cracking [3]. Sulfur is one of the most harmful impurities in syngas which can irreversibly contaminate the catalyst active sites and dramatically reduce the catalyst activity [4]. The works of Professor Bartholomew and his group have provided major insights into the mechanisms of deactivation of metallic Fischer–Tropsch (FT)

∗ Corresponding author. Tel.: +33 3 20 33 54 39; fax: +33 3 20 43 65 61. E-mail address: [email protected] (A.Y. Khodakov). http://dx.doi.org/10.1016/j.cattod.2015.07.003 0920-5861/© 2015 Elsevier B.V. All rights reserved.

catalysts and in particular in the presence of sulfur [5–11]. Design of sulfur resistant catalysts is therefore a major challenge for FT synthesis. In previous reports several sulfur tolerant catalysts have been used for this reaction, such as for example noble metals (Rh [12], Pd [13]) and transition metal sulfides (MoS2 [14], WS2 [15]). Molybdenum disulfide (MoS2 ) exhibits noticeable activity in carbon monoxide hydrogenation [16–18]. Because of the catalytic properties and excellent resistance to sulfur poisoning, molybdenum disulfide could be a potential candidate for valorization of syngas containing small amounts of sulfur. During the reaction, molybdenum sulfide losses sulfur. Thus the smooth operation of molybdenum sulfide requires the continuous presence of small amounts of sulfur in syngas [19]. The major challenge of molybdenum sulfide catalysts is however, the efficient selectivity control. In fact, the non-promoted MoS2 catalysts principally yields methane, [16,19] which is a cheap and often useless reaction product. Promotion is one of the methods to optimize the selectivity of carbon monoxide hydrogenation on molybdenum sulfide catalysts. Previous reports indicate higher selectivity of potassium-modified MoS2 -based catalysts to mixed alcohols at high pressures [20]. The potassium promoted molybdenum sulfide catalysts also exhibit high activity for water–gas shift (WGS) reaction and show significant tolerance to coke deposition [21]. The function of K seems to reduce on one hand, the catalyst hydrogenating ability and hydrocarbon production and on other hand, to favor alcohol formation [22]. In addition to the catalyst composition, the selectivity of

138

C. Liu et al. / Catalysis Today 261 (2016) 137–145

carbon monoxide hydrogenation of molybdenum sulfide catalysts is also affected by the reaction conditions (pressure, temperature, H2 /CO ratio). The linear alcohols are produced at 80–90 bar. Small amounts of olefins are also present during CO hydrogenation over molybdenum sulfide catalysts under higher pressure conditions [23]. While a significant number of papers have described alcohol synthesis on molybdenum sulfide catalysts which usually occurs at higher reaction pressures (80–90 bar), very few information is available about the catalytic performance at lower reaction pressures (20 bar) which could be more favorable for olefin synthesis. The catalytic performance of molybdenum sulfide catalysts can be also affected by the support. The supported catalysts have several advantages compared to the bulk catalysts such as mechanical, thermal and chemical stability. Recently [24] two types of active sites associated to unpromoted molybdenum sulfide and K-Mo-S species were uncovered in alumina supported molybdenum sulfide catalysts. The concentration of these sites varied as a function of potassium content in the catalysts. The use of carbon nanotubes (CNT) as supports for numerous catalysts has been recently drawing attention, due to their flexibility as support in tailoring the catalyst properties to specific needs [25]. Carbon nanotubes are resistant to acidic or basic media and stable at high temperatures. Moreover, they have several unique features, such as nanosized channels and sp2 -C-constructed surfaces [26]. They also display exceptionally high mechanical strength, high thermal conductivity, and medium or high specific surface areas. A few previous reports [27–29] suggest that the CNT supported K-MoS2 based catalysts exhibit a high C2+ alcohol productivity in carbon monoxide hydrogenation at higher reaction pressures. It is not clear however how the reaction rate and selectivity could be affected by different potassium contents in the catalysts supported by carbon nanotubes. The present work addresses the design of CNT supported molybdenum sulfide catalysts for synthesis of olefins from syngas. This paper more particularly focusses on the effect of promotion with K on the olefin selectivity over supported molybdenum sulfide catalysts. At different preparation stages, the catalysts were characterized by nitrogen adsorption, X-ray diffraction (XRD), Xray photoelectron spectroscopy (XPS), temperature programmed reduction of hydrogen (H2 -TPR) and temperature programmed desorption of CO2 (CO2 -TPD). The catalytic performance was evaluated in a fixed bed reactor as a function of K promotion. To enhance olefin production, the catalytic experiments were conducted at relatively low total pressure (20 bar) compared to the conditions typically used for alcohol synthesis.

then sulfided in a mixture 10 mol% H2 S in H2 at a flow rate of 100 ml/min at 400 ◦ C for 2 h. After the sulfidation, the catalysts have been passivated in a flow of 1% O2 in argon at room temperature. The molybdenum content in the samples was 15 wt.% and potassium content was between 0 and 15 wt.%, respectively. The K/Mo molar ratio varied from 0.25 to 2.5. The catalysts are labelled as xK yM/CNT, where x designates potassium content (wt.%) in the catalysts and y stands for molybdenum loading (wt.%). 2.2. Characterization The BET surface area, pore volume, average pore diameter and pore size distribution of the catalysts were determined by N2 physisorption at −196 ◦ C using a Micromeritics ASAP 2010 apparatus. The pore size distribution curves were calculated from the desorption branches of the isotherms using the BJH method [32]. The ex-situ X-ray Diffraction (XRD) patterns were recorded by a Siemens D5000 diffractometer using a Cu K␣ source. The catalysts were scanned from 2Â of 5◦ to 70◦ with a scanning rate of 0.02◦ s−1 . The diffraction patterns were analyzed by Eva software (Bruker) and matched using the JCPDS database. The Temperature-Programmed Reduction (TPR) profiles of the sulfated samples were measured using AutoChem II 2920 (Micromeritics) with 10 vol% H2 diluted in argon stream. The total flow rate was 50 cm3 min−1 . The temperature was increased from room temperature to 1100 ◦ C with the ramping rate was 10 ◦ C min−1 . Then the temperature was kept at 1100 ◦ C for 1 h. The X-ray Photoelectron spectroscopy (XPS) spectra were recorded with a VG ESCALAB 220 XL spectrometer equipped with a monochromatic Al K␣ (E = 1486.6 eV) X-ray source. The binding energies (BE) of Mo 3d, Co 2p, C 1s and S 2p were determined by computer fitting of the measured spectra and referred to the Al 2p peak of the support at 74.6 eV, using Casa XPS software. The binding energies were estimated within ±0.2 eV. The Temperature Programmed Desorption of carbon dioxide (CO2 -TPD) was carried out in a quartz reactor connected with a mass spectrometer. The samples were first pre-treated in a flow of helium (40 cm3 /min) at 400 ◦ C for 1 h then the temperature was lowered to 30 ◦ C. CO2 was adsorbed on the sulfided samples using a pulse technique (0.49 cm3 of CO2 ) at 30 ◦ C. The CO2 desorption was measured during continuous temperature increase (10 ◦ C min−1 ) up to 800 ◦ C. TEM measurements were performed using a TECNAI microscope operating at a voltage of 200 kV. The sample powder was ultrasonically dispersed in ethanol and deposited on a copper grid prior to the measurements. 2.3. Catalytic measurements

2. Experimental 2.1. Catalyst preparation The CNT support (Multi-wall carbon nanotubes, diameter: 20–40 nm, length: 5–15 ␮m) was provided by Io.Li.Tech (Ion Liquid Technologies, Germany). CNTs were treated with 63% nitric acid before the catalyst synthesis to eliminate impurities. Molybdenum was deposed on the CNT support using incipient wetness impregnation with an aqueous solution of ammonium molybdate tetrahydrate (AMT, (NH4 )6 Mo7 O24 ·4H2 O, Sigma–Aldrich), which is a common precursor for MoS2 based catalysts [30]. The impregnated catalysts were dried at 60 ◦ C for 12 h, then milled mechanically with K2 CO3 precursor (Prolabo). Previously, it was found [31] that mechanical mixing of molybdenum sulfide with potassium resulted in better catalytic performance in carbon monoxide hydrogenation. The samples were calcined at 550 ◦ C under nitrogen flow for 2 h. The calcined samples were

The experimental catalytic tests were conducted using a highpressure milli-fixed bed stainless steel reactor (id = 1.4 mm). The reactor and tubing were coated with SulfInert® (Restek) to avoid sulfur loss and adsorption in the rig. The feed gas was composed of syngas with H2 /CO ratio of 2 and contained 14 ppm of H2 S. Carbon monoxide contained 5% of N2 (internal standard). The reaction was conducted at 360 ◦ C and 20 bar. The GHSV was 1050 cm3 /g/h. The gaseous reaction products were analyzed on-line using a GC-456 Bruker gas chromatograph equipped with a thermal conductivity and a flame ionization detectors. H2 , N2 , CO, CH4 and CO2 were separated with a Shincarbon (Resteck® ) column and analysed by a TCD detector. C1 to C5 hydrocarbons and light alcohols were separated by a Q-Plot (Bruker® ) column and measured by FID. The liquid products were collected in an off-line trap and analyzed by the same chromatograph. The selectivity of each product was calculated on carbon basis. The selectivity to hydrocarbons and alcohols takes into account CO2 production in this reaction.

C. Liu et al. / Catalysis Today 261 (2016) 137–145 Table 1 Textural proprieties of the CNT, calcined Mo/CNT and potassium promoted Mo/CNT catalysts. Samples

Area (m2 /g)

Pore volume (cm3 /g)

Pore diameter (nm)

CNT 15M/CNT 1.5K15M/CNT 3K15M/CNT 6K15M/CNT 9K15M/CNT 15K15M/CNT

163.3 68.6 49.4 45.9 43.5 42.7 40.7

0.56 0.32 0.20 0.21 0.18 0.26 0.24

17.2 16.2 17.6 15.3 15.7 17.6 18.4

139

Table 2 XPS surface atomic concentrations in sulfided K MoS2/CNT catalysts. K (%)

0 1.5 3 6 15

XPS Atomic concentrations (%) C

O

S

K

Mo

77.7 80.8 77.9 80.8 68.5

13.9 10.9 12.5 10.3 11.8

5.1 4.2 4.3 4.1 9.3

0 2.7 3.8 3.4 6.8

3.3 1.4 1.5 1.4 3.5

S/Mo ratio

K/Mo ratio

1.5 3.0 2.8 2.9 2.6

0 1.9 2.4 2.4 1.9

(JCPDF 87-0730, 2Â = 13.08◦ , 17.07◦ , 25.49◦ , 27.47◦ , 30.15◦ ). The K2 CO3 phase (JCPDF 87-0730, 2Â = 12.98◦ , 29.19◦ , 37.54◦ , 46.32◦ ) is also observed in the XRD patterns of the 6K 15M/CNT, 9K 15M/CNT and 15K 15M/CNT catalysts.

3. Results and discussion 3.1. Catalyst characterization 3.1.1. Calcined catalysts The textural characteristics of the carbon nanotube support and calcined K Mo/CNT catalysts determined by BET/BJH method are shown in Table 1. The surface area and pore volume typically decrease after addition of molybdenum: from 163.3 to 68.6 m2 /g. and from 0.56 to 0.32 cm3 /g. The surface area still further decreases with addition of K content, while the pore volume remains in the average of 0.22 cm3 /g for the promoted catalysts. These results are in agreement with previous reports [33,34]. The decrease in surface area and pore volume could be due to pore plugging and CNT dilution particularly after addition of significant amounts of potassium carbonate. The CNT support and calcined K MoS2 /CNT catalysts were characterized by XRD. The patterns are shown in Fig. 1. The crystallized CNT phase can be detected (JCPDF 075-0621, 2Â = 26.23◦ , 42.21◦ , 44.36◦ ) with a good crystallinity on all catalysts. As expected, the intensity of the CNT patterns decreases with the increase in the K content. This suggests that the basic carbonate promoter could attack the CNT structure and result in formation of amorphous carbon. The MoO2 phase (JCPDF 032-0671, 2Â = 26.03◦ , 37.02◦ , 37.93◦ , 53.04◦ , 53.51◦ , 53.97◦ , 60.20◦ , 66.66◦ ) was clearly observed in the XRD patterns of all catalysts. The formation of this phase is due to the AMT precursor decomposition to MoO2 during calcination under N2 atmosphere [35]. With the increase in K content, the MoO2 XRD patterns become less intense and broader. On the 15K15M catalyst, the MoO2 phase is hardly present, indicating good MoO2 dispersion. The crystallized mixed Mo-K phases are detected at higher K loaded (more than 3%): K2 MoO4 (JCPDF 29-1011, 2Â = 18.87◦ , 26.27◦ , 30.65◦ , 39.42◦ , 45.83◦o ) and KMo4 O6

3.2. Sulfided catalysts The XRD patterns of the sulfided CNT supported MoS2 catalysts are presented in Fig. 2. The CNT crystalline phase can be observed on all catalysts, but the intensity of the phase decreases with the increase in potassium content, as seen previously for the calcined catalysts. The XRD patterns indicate the presence of MoS2 phase (JCPDF 89-3040, 2Â = 14.38◦ , 32.68◦ , 39.60◦ , 49.79◦ , 58.34◦ , 60.15◦ ); KMoS2 phase (JCPDF 18-1064, 2Â = 9.66◦ , 32.41◦ , 36.13◦ , 40.61◦ , 60.46◦ ) and K2 S phase (JCPDF 65-3001, 2Â = 34.21◦ , 49.16◦ , 61.26◦ ) in all the K-MoS2 /CNT catalysts. The intensity of the MoS2 patterns is low and the peaks are broad, indicating high MoS2 dispersion in the CNT. At higher potassium content (9K 15M/CNT and 15K 15 M/CNT), XRD shows the presence of K2 MoS4 (JCPDF 19-1001, 2Â = 17.55◦ , 24.23◦ , 29.36◦ , 41.19◦ , 47.05◦ , 58.40◦ ), K2 S3 (JCPDF 31-1095, 2Â = 28.28◦ , 30.02◦ , 32.70◦ ) and K2 S5 (JCPDF 300993, 2Â = 30.80◦ , 31.56◦ , 34.16◦ ). Note that the MoO2 phase is still present in all the catalysts except for 15K 15M/CNT. The intensity of this phase continues decreasing with the increase in K content. This indicates that MoO2 cannot be completely sulfided and seems to be well stabilized on the CNT support. Moreover, as the K content increases, the intensity of the XRD patterns attributed to this phase decreases, the peak becomes broader, indicating that the particle size of MoO2 became smaller. The XPS surface element composition of the sulfided K MoS2 /CNT catalysts is given in Table 2. Different elements (C, O, S, K, Al, Mo) were observed on the catalyst surface, and their ratio varied as a function of the K precursor percentage. In the 15M/CNT catalyst, the S/Mo ratio is 1.5, which is less than the stoichiometric ratio of the MoS2 phase equal to 2. This is indicative

15K15M_CNT 15K15M_CNT

6K15M_CNT 3K15M_CNT 1.5K15M_CNT

9K15M_CNT

Intensity (a.u.)

Intensity (a.u.)

9K15M_CNT

6K15M_CNT 3K15M_CNT 1.5K15M_CNT

15M_CNT

15M_CNT

CNT 0

10

20

30

40

50

60

70

o

2θ( ) Fig. 1. XRD patterns of the CNT support and calcined CNT supported catalysts. (䊉) MoO2 , (䊏) CNT, () KMo4 O6 , () K2 MoO4 .

0

10

20

30

40

50

60

70

o

2θ( ) Fig. 2. XRD patterns of the sulfided CNT supported catalysts. () MoS2 , (䊐) KMoS2 , (♦) K2 MoS4 , (䊉) MoO2 , (䊏) CNT.

140

C. Liu et al. / Catalysis Today 261 (2016) 137–145 Table 3 XPS atomic fractions of different Mo and S species in K Mo/CNT.

4+

Mo

A)

6+

Mo

5+

Mo

15K15M_CNT

Count (a.u.)

6K15M_CNT

3K15M_CNT

Catalyst

15M CNT 1.5K15M CNT 3K15M CNT 6K15M CNT 15K15M CNT

Mo (%)

S (%)

Mo4+

Mo5+

Mo6+

S2−

S2 2−

SO4 2−

90.0 84.6 73.9 82.6 85.2

6.4 4.5 6.9 6.5 11.3

3.6 10.9 19.2 10.9 3.5

60.7 55.5 61.5 64.6 92.9

24.9 17.3 12.2 10.3 0.7

14.3 27.2 26.3 25.1 6.4

1.5K15M_CNT

15M_CNT 220

225

230

235

240

Binding Energy (eV) B) 2-

S

2-

SO4

Count (a.u.)

15K15M_CNT 2-

S2

6K15M_CNT 3K15M_CNT 1.5K15M_CNT

15M_CNT 155

160

165

170

175

Binding Energy (eV) Fig. 3. Mo3d (A) and S2p (B) XPS spectra and relevant peak decompositions.

of incomplete sulfidation of the Mo/CNT catalyst. This result is also consistent with the XRD observations. Note, however, that for the K promoted Mo/CNT catalysts, the S/Mo ratio is higher than the stoichiometric ratio of MoS2 phase. This could be due to the presence of sulfided potassium and mixed Mo-K sulfided phases on the catalyst surface. These phases were observed in the catalysts by XRD. The concentration of sulfur on the surface is therefore not only related to MoS2 , but a result of the contribution from potassium and Mo-K mixed sulfides. For all the promoted catalysts, the XPS K/Mo ratio is much lower than the atomic bulk ratio. This ratio remains relatively stable (2.15 ± 0.25) in spite of the increase in K content. Andersen et al. [36] studied the interaction of potassium and molybdenum sulfides using DFT. High mobility of K was observed on the MoS2 catalyst. At low potassium content, K is preferably present on the catalyst surface. But with the increase in K content, potassium atoms tend to form atom chains over the interstitial of MoS2 molecules. That could be a reason why in the 15K 15M/CNT catalyst, the K/Mo ratio on the catalyst surface does not increase. The Mo3d and S2p XPS spectra of the K MoS2 /CNT catalysts are shown in Fig. 3A and B. In the catalysts, a three-peak envelop of the Mo 3d signal is observed (Fig. 3A). It can be decomposed into three separate overlapping doublets. This suggests that molybdenum is present in three different Mo oxidation states: Mo4+ , Mo5+ and Mo6+ . The binding energy at 228.9 eV and 232.1 eV should correspond to Mo 3d5/2 and 3d3/2 of Mo4+ [37]. In the XRD patterns (Fig. 2), we have observed the presence of both MoO2 and MoS2

phases. In these two phases, molybdenum is presented in the oxidation state of +4 (Mo4+ ). The binding energies of the MoS2 phase are generally 228.9 eV and 232.1 eV and the binding energies of the MoO2 phase are at 229.3 eV and 232.4 eV [38]. Those two species are close in binding energy and are difficult to differentiate. In our case, we can observe that the XPS signal corresponds mainly to the MoS2 phase; however the presence of noticeable amounts of the MoO2 phase cannot be excluded. The binding energies of 230.6 eV and 233.7 eV could correspond to Mo5+ species [39], while the Mo6+ species were also detected with the binding energies of 232.8 eV and 235.9 eV [40]. Previous study [39] suggests that Mo5+ species are due to the formation of Mo oxy-sulfides such as the MoO2 S2 and MoOS2 phase. The Mo6+ species could be related to Mo oxides [34,35]. It was believed that Mo5+ and Mo6+ species could arise from [24]: (i) passivation step of catalysts after sulfidation, (ii) incomplete sulfidation of catalysts [41] and (iii) the presence of K-Mo mixed sulfides phases such as K2 MoS4 which can be also detected from XRD patterns (Fig. 2). The peak decomposition of S 2p XPS spectra is illustrated in Fig. 3B. Three different S species could be observed. The XPS peaks with the binding energies of 161.8 eV and 163.0 eV correspond to S2p3/2 and S2p1/2 of sulfide ions (S2− ) [42], which might be present in potassium sulfides, Mo disulfide and K-Mo mixed sulfides [43]. The binding energy around 163.5 eV and 164.7 eV could be possibly accounted by S 2p3/2 and S 2p1/2 of poly-sulfide ions (S2 2− ), which are present in K sulfides such as K2 S3 and K2 S5 . All those species have been detected by XRD. The S2 2− species can also present in the sulfur rich Mo sulfides (MoS2+x ) that is believed to have the same proprieties as MoS2 phase [44]. The S 2p XPS spectra also exhibit others broad peaks at 168.4 eV and 169.6 eV which are due to the presence of sulfates (SO4 2− ) [45]. Moreover, both the Mo 3d and S 2p XPS spectra of the K promoted catalysts exhibit a slight shift towards higher energies in the binding energies (0.3 eV) in comparison to the potassium-free 15M/CNT catalysts. The shift of binding energy for the catalysts with promotion should be due to the presence of K. In XPS, the binding energy of an element can be affected by neighboring atoms/groups. The presence of electrophile atoms/groups results in the decrease in the electron density, in this case, the binding energy of the analyzed element also becomes higher. In the K MoS2 catalysts, potassium presents as K+ ion. It is known the K+ is strong electrophile species, so the binding energies of Mo and S could get higher in the promoted catalysts. The results of the XPS peak decomposition showing the fractions of different Mo species or S species are presented in Table 3. The Mo4+ concentration is higher in 15M CNT than in any other promoted catalysts. Previous study [46] showed that addition of alkali metals could result in the decrease of the Mo4+ fraction, because of their inhibition effect on sulfidation. Indeed, with addition of K, the Mo4+ concentration decreases and the Mo6+ fraction increases significantly in 1.5K 15M/CNT, 3K 15M/CNT and 6K 15M/CNT. Thus, the presence of potassium could lead to lower fractions of Mo4+ species but to higher fraction of Mo6+ . Nevertheless, the 15K 15M/CNT catalyst exhibits lower concentration of Mo6+ species at the benefit of the fraction of Mo5+ .

C. Liu et al. / Catalysis Today 261 (2016) 137–145

141

Fig. 4. TEM images of (A and B) the 15M CNT catalyst and (C and D) 3K15M CNT catalyst.

For all the catalysts, the concentrations of S2− and S2 2− were much important than that of SO4 2− . It is noticeable that in the nonpromoted MoS2 /CNT catalyst, the concentration of poly-sulfides (S2 2− ) is higher than in the K promoted MoS2 /CNT catalysts. With addition of K, the percentage of S2 2− decreases for 1.5K 15M/CNT, 3K 15M/CNT and 6K 15M/CNT catalysts. The S2 2− phase could correspond to the Mo oxy-sulfides (MoOx Sy ) and poly-sulfides of alkali metal (such as K2 S3 ). In the catalyst with high K content, sulfidation leads to a higher fraction of sulfide species than polysulfides. As explained previously, the SO4 2− could arise from the passivation step. The S2− fraction becomes rather important, while S2 2− and SO4 2− fractions get rather low at high potassium content in the 15K 15M/CNT catalyst compared with other CNT supported catalysts. Transmission electron microscopy (TEM) imaging (Fig. 4) was conducted on 15M/CNT and 3K 15M/CNT catalysts to obtain information on molybdenum localization in the carbon nanotubes. The representative smooth parallel rolled up graphene layers are observed. The images of these two catalysts show also some black spots corresponding to the presence of MoS2 particles. Those particles, which have the form of multilayer sheets, can be well distinguished. Other small black spots could be also distinguished, which could correspond to the presence of small MoO2 particles. The MoS2 particles supported on the carbon nanotubes were highly

dispersed. The particles seem to be mainly located outside the nanotubes. The coverage of the CNT support with MoS2 was not homogeneous and some areas remained uncovered. The H2 -TPR results for the CNT supported MoS2 catalysts with different potassium contents are shown in Fig. 5. Three groups of peaks can be detected at different temperatures. The low temperature H2 reduction peaks in the range of 200–430 ◦ C (area 1) could be related to the hydrogenation of extra sulfur atoms (chemisorbed H2 S or SH groups) or sulfur atoms that are weakly bonded to the catalyst surface [47]. Xiao et al. [48] suggested that those species are adsorbed on low coordinated edge/corner sites of Mo sulfides. The edge/corner sites of Mo sulfides may contain active sites for CO hydrogenation. In the presence of hydrogen, those sulfur species could be easily removed below 450 ◦ C. During the catalytic tests at 360 ◦ C, the chemisorbed sulfur species could be partly removed with the release of these surface sites. Ramachandran et al. [49] and Collins et al. [50] also observed those low temperature peaks for the Ni promoted MoS2 /Al2 O3 catalysts. They assigned these peaks to the sites responsible for the formation of hydrocarbons from syngas. They suggested however that these sites do not interfere into alcohol synthesis. The H2 -consumption relevant to Peak 1 is decreasing with the increase in K loading on the catalysts (Table 4). Potassium addition can cover the MoS2 surface and then block the adsorption of sulfur.

142

C. Liu et al. / Catalysis Today 261 (2016) 137–145

H2 consumption (ml/min)

2.5

15M_CNT 1.5K15M_CNT 3K15M_CNT 6K15M_CNT 15K15M_CNT

Peak 2

2.0

1.5

1.0

Peak 1 0.5

Peak 3

0.0 0

200

400

600

800

1000

isothermal

o

Temperature( C)

Fig. 6. CO2 -TPD on the sulfided K MoS2 /CNT catalysts.

Fig. 5. H2 -TPR on the sulfided K MoS2 /CNT catalysts.

The medium temperature reduction peak (Peak 2, between 575 and 750 ◦ C) could be related to the hydrogenation of molybdenum/potassium mixed sulfides (K-Mo-S phases). Note that this peak is absent in the TPR profile of the non-promoted CNT supported molybdenum (15Mo/CNT) catalyst which does not contain any potassium. The area of this peak significantly increases with the increase in K loading (Table 4). It can be suggested that Peak 2 is related with sulfided potassium phases. These species could be K-Mo-S and possibly potassium poly-sulfides ions, such as K2 S3 and K2 S5 which were detected in the XRD patterns. In addition, the sulfur atoms located at the edges or basal planes of molybdenum sulfide crystallites could also contribute to the hydrogen consumption in the temperature range of 575–750 ◦ C, in agreement with previous report by Toulhouat et al. [47]. Peak 3 (750–950 ◦ C) could be assigned to the reduction of MoO2 phases to metallic Mo. The MoO2 phase in the sulfided catalysts has been observed in the XRD patterns. Previously Brito et al. [51] reported for unsupported MoO3 catalysts, that MoO3 can be reduced to MoO2 at around 500 ◦ C and MoO2 can be reduced to Mo at around 800 ◦ C. Lai et al. [52] reported for active carbon supported catalysts that MoO2 can be reduced to the metallic phase at 720–830 ◦ C. The intensity of Peak 3 decreases with the increase in K content (Table 4). This peak is not detected in the TPR of 15K 15M/CNT. This indicates that no MoO2 could be present in this catalyst. This is consistent with the XRD results which show a decrease in the MoO2 concentration at higher potassium content. This peak cannot be attributed to the reduction of MoS2 to metallic Mo. Indeed, Mangnus et al. [53] reported that the reduction of crystallized Mo sulfide requires temperatures higher than 1050 ◦ C. The basicity of the catalysts was measured by CO2 -TPD. Fig. 6 presents the CO2 desorption curves of the sulfidated catalysts measured between 30 and 800 ◦ C. It is clearly observed that the 15M/CNT catalyst did not show any CO2 desorption in the temperature range from 30 to 800 ◦ C. This indicates that the unpromoted Table 4 H2 consumption in TPR experiments. Sample

15M/CNT 1.5K 15M/CNT 3K 15M/CNT 6K 15M/CNT 15K 15M/CNT

MoS2 catalyst does not contain any measurable concentration of basic sites. Promotion with potassium results in the appearance of basic sites which can be detected by CO2 desorption peaks. The 3K 15M/CNT and 6K 15M/CNT catalysts exhibit a small CO2 desorption peak between 200 and 450 ◦ C, and another CO2 peak between 420 and 560 ◦ C. These two peaks can be attributed to weaker and stronger catalyst basic sites, respectively. The first CO2 desorption peak was much broader in the 6K 15M/CNT catalyst. For the 15K 15M/CNT catalyst, no noticeable CO2 desorption peak at lower temperature could be observed. However, the CO2 desorption peak at higher temperature is more intense in 15K 15M/CNT than in any other catalysts. Table 5 displays the quantity of desorbed CO2 from the catalysts at different temperature range. The unpromoted MoS2 /CNT catalyst does not exhibit any basicity. The quantity of desorbed CO2 was slightly higher for the 6K 15M/CNT at low temperatures than for the 3K 15M/CNT and 15K 15M/CNT catalysts. At high temperature, the quantity of desorbed CO2 is much more important for 15K 15M/CNT catalyst than for the other catalysts. In summary, it is found that the basicity of CNT supported molybdenum sulfide catalysts depends on K content in the catalysts. Higher potassium loading results in higher concentration of basic sites. Higher potassium content also favors strong and very strong basic sites identified by CO2 desorption peaks at higher temperatures. 3.3. Catalytic tests The catalysts were tested in a fixed bed tubular reactor. The catalyst sulfidation was realized ex-situ before the catalytic tests. The reaction was conducted in the presence of 13.3 ppmv of H2 S in the syngas in order to maintain the catalysts in sulfided state [54]. The reaction pressure was fixed at 20 bar in order to favor hydrocarbon production [55]. Methane, olefins, paraffins, methanol, ethanol and carbon dioxide were the main products detected on alumina supported molybdenum catalysts. The CO conversion and product selectivities measured at the steady state

Table 5 CO2 desorption calculated from TPD on sulfided K-MoS2 /CNT catalysts.

H2 consumption (mmolHydrogen /gcata ) Peak 1

Peak 2

Peak 3

1.24 2.65 2.02 2.02 1.69

<0.01 1.32 1.66 2.21 4.40

1.05 0.66 0.38 0.04 <0.01

Catalyst

15M/CNT 3K 15M/CNT 6K 15M/CNT 15K 15M/CNT

CO2 desorption (␮mol/gcatalyst ) Low temperature

High temperature

Total

0.0 0.03 0.14 0.03

0.0 0.05 0.08 0.42

0.0 0.08 0.22 0.45

C. Liu et al. / Catalysis Today 261 (2016) 137–145

143

Table 6 CO conversion and product selectivity on K MoS2 /CNT catalysts (T = 360 ◦ C, P = 20 bar, GHSV = 1050 ml g−1 h−1 , H2 /CO = 2). Catalyst

15M/CNT 1.5K15M/CNT 3K15M/CNT 6K15M/CNT 9K15M/CNT 15K15M/CNT

CO conversion (%)

37.4 19.1 17.7 16.3 16.8 9.3

Hydrocarbon selectivity (%)

Alcohol selectivity (%)

C1

C2 –C4 Olefin

C2 –C4 Paraffin

C5 +

C1

C2 +

24.7 22.9 20.7 17.4 20.4 17.1

0.1 2.5 9.6 5.4 3.7 2.9

27.3 26.7 11.9 8.4 12.2 1.8

0.0 1.6 6.3 10.8 18.2 34.0

0.0 7.0 10.1 11.9 7.6 14.3

0.0 1.6 3.7 5.6 6.7 8.6

at iso-GHSV for the non-promoted and potassium promoted CNT supported molybdenum sulfided catalysts are shown in Table 6. For the unpromoted 15M/CNT catalyst, the carbon monoxide is hydrogenated to produce mainly methane, C2 -C4 paraffins and carbon dioxide. In particular the catalyst showed a very high selectivity to methane and CO2 . Without K promotion, MoS2 was not active for production of alcohols, olefins or heavy products (C5+ hydrocarbons). Liu et al. [56] reported on non-promoted MoS2 catalysts, that the major products of carbon monoxide hydrogenation was methane and CO2 with very low concentration of ethane (selectivity <1%). Wang et al. [54] reported similar results. Note that the only active phase in the sulfided Mo catalysts was MoS2 . In our work, the selectivity to light paraffins (mainly ethane) can attain 27.3% even on the unpromoted catalyst. This can be possibly attributed to the presence of both MoS2 and MoO2 in 15 M/CNT (seen in XRD and H2 -TPR results). Tatumi et al. [57] suggested that during CO hydrogenation, MoO2 could be partly reduced to MoO2−x phase, on which CO was non-dissociatively absorbed. MoO2−x was considered to contain active sites for the C-C chemical bonds formation. This could be the reason why the selectivity to lighter paraffins was much higher on CNT supported catalysts compared to unsupported MoS2 . The catalytic performance of the CNT supported catalysts is strongly affected by potassium promotion. At GHSV = 1050 ml gcata −1 h−1 , the CO conversion decreases from 19.1% to 9.3% with the increase in K/Mo molar ratio from 0.2 to 2.5. Addition of potassium leads to higher selectivity to olefins and alcohols (Fig. 7). For light olefins, the best yield was obtained on 3K 15M/CNT (K/Mo = 0.5), while the highest productivity of C2+ alcohols was measured on 9K 15M/CNT (K/Mo = 1.5). The decrease in the CH4 and CO2 selectivities was also affected by the potassium addition, while the selectivity to long-chain hydrocarbons was significantly enhanced (Fig. 7). This result is consistent to the

Fig. 7. Production rates of hydrocarbons and oxygenates on CNT supported catalysts as a function of K/Mo atomic ratio.

CO2 Selectivity (%)

47.9 38.1 37.5 40.2 31.3 21.4

previous work of Surisetty et al. [27]. They reported that on K-MoS2 /CNT catalysts, for the same Mo content, the increase in potassium content leads to a decrease in selectivity of CH4 and CO2 . Extremely high potassium contents (15K 15M/CNT) do not contribute to higher CO conversion and are not in the favor of higher olefin and alcohol yields. Higher amounts of C5+ hydrocarbons are produced. The positive effect is the decrease of the selectivity to two undesirable products: methane and carbon dioxide (Table 6).

4. Discussion Carbon nanotube supported Mo disulfide catalysts were characterized using various techniques and tested in carbon monoxide hydrogenation. Our results show that addition of different amounts of K significantly influences the catalyst structure and catalytic performance in carbon monoxide hydrogenation. The XRD patterns showed that the peak intensity of CNT decreased significantly with potassium content, indicating the CNT crystallinity decreased. The surface area of the catalysts also decreased significantly with potassium content. In the sulfided K MoS2 /CNT catalysts, K-Mo mixed phases were detected in XRD patterns. With lower K loading (K% = 1.5%, 3%, 6%) only KMoS2 phase was observed while both KMoS2 and K2 MoS4 phases could be detected in the catalysts with higher K contents (K% = 9%, 15%). These phases could be reduced by hydrogen and exhibit peaks in the H2 -TPR profiles (Peak 2 in Fig. 5). Several previous reports [46,58,59] suggest that K-Mo-S phases possess the active sites for higher alcohol synthesis. Li et al. [60] found the CO was non-dissociatively adsorbed on K-Mo sulfides, and then C-C chemical bonds were formed. This is indicates that the presence of those mixed K-Mo phase coincides with higher selectivity to heavy products (C2+ alcohols and C5+ hydrocarbons). This hypothesis is consistent with our work. However, the rate of methane formation decreases and the rate of C5+ hydrocarbon production increases concurrently with an increase in the area of peak 2 in TPR profiles. Up to the optimum ratio K/Mo = 0.5, the active K-Mo-S phases were formed in the catalyst and the rate of olefin production is increased. After this optimum K/Mo ratio, more potassium poly-sulfided species were formed (observed by XRD). Those species are not active in CO hydrogenation but contribute to the increase in the basicity of the catalysts. The C5+ hydrocarbon selectivity was significantly higher on 9K 15M/CNT and 15K 15M/CNT catalysts, compared to other catalysts. We can speculate that the K2 MoS4 phase which was observed only in 9K 15M/CNT and 15K 15M/CNT can be more favorable for production of heavy products than the KMoS2 phase. Different Mo or S species detected and identified by XPS in the sulfided catalysts could be relevant to the catalytic performance. Woo et al. [41] investigated the relation between the XPS atomic concentration of different molybdenum species and alcohol production. They reported that higher alcohol selectivity could be attributed to the Mo4+ species and lower fraction of the

144

C. Liu et al. / Catalysis Today 261 (2016) 137–145

Fig. 8. Light olefin production rate as a function of Mo4+ fraction.

rate with alumina support is at a ratio K/Mo = 2.5, while the counterpart with carbon nanotube is at a ratio K/Mo = 0.5. The activity of both catalysts decreases with an increase in potassium content. This difference can be explained by the interactions of potassium and supports. Indeed, on alumina, K-Mo-Al mixed phases have been observed while on CNT, only K-Mo phases have been seen. Hence, a part of the added potassium diffused inside the alumina matrix and was unavailable for the molybdenum promotion, while in carbon nanotubes, potassium interacted only with molybdenum species. With CNT catalysts, potassium is on the surface and K-sulfided phases are more rapidly formed during the sulfidation step. In contrary, due to the diffusion of K through alumina, higher amount of potassium is needed before the formation of K-sulfided phases. In addition, the alumina supported catalysts showed higher carbon monoxide conversion rate that carbon nanotubes supported counterparts. The effect could be due to higher extent of molybdenum sulfidation in alumina supported catalysts. Furthermore, MoO2 has been detected in the CNT supported catalysts. MoO2 seems to be more difficult to be sulfided than MoO3 . Lower extent of sulfidation seems to the major reason of lower activity of MoS2 /CNT catalysts. Note that this MoO2 phase has not been detected in the catalyst synthesized on alumina support. The presence of noticeable concentrations of MoO2 in carbon nanotubes could hinder molybdenum sulfidation. Somewhat lower activity of CNT supported catalysts compared to the alumina-based counterparts can be attributed to the lower extent of sulfidation. 5. Conclusion

Fig. 9. Formation rate of light olefins on MoS2 /Al2 O3 and MoS2 /CNT catalysts with different K/Mo ratio ( Al2 O3 supported catalysts, ♦ CNT supported catalysts).

SO4 2− . Other papers [46,61] have reported similar conclusion. This is also consistent with our results. The fraction of sulfate phase decreases with increase in K content while the C2+ alcohol selectivity also increases. It is also interesting to mention the relation (Fig. 8) between light olefin production and amount of Mo4+ determined by XPS. The Mo4+ species correspond to MoS2 and MoO2 (not possible to differentiate those two phases by XPS). Light olefin production decreased with the increase in the fraction of Mo4+ species. With the lower amount of Mo4+ , the hydrogenating ability should be less important. Colley and al [62] concluded that catalysts with lower hydrogenating ability can improve the production of light olefins. This is consistent with the fact that C2 -C4 olefin selectivity increased with lower Mo4+ fraction. The catalyst basicity is another parameter which would affect the rate of secondary reactions such as re-adsorption and hydrogenation of olefins or oxygenated products [63]. In our catalysts, the highest yield of C2 -C4 olefins was obtained on the catalysts with a moderate basicity. However, when the catalyst basicity is too high, less light olefins and paraffins are produced for the benefit of C5+ hydrocarbons. Potassium sulfides (K2 S3 , K2 S5 ) were observed at higher potassium content. These phases with higher basicity seem to be favorable for higher selectivity to C5+ hydrocarbons. It is interesting to compare the activity of carbon nanotube supported samples with the activity of previously published alumina supported catalysts [24]. The catalysts were prepared using similar synthesis and activation procedure. The light olefin production rates are presented Fig. 9. Both alumina supported catalysts and CNT supported catalysts exhibit a maximum of olefin production rate as a function of K/Mo ratio. The maximum of olefin production

Potassium promotion in CNT supported MoS2 based catalysts leads to the production of light olefins, long-chain hydrocarbons and alcohols. The highest yield of light olefins was observed on the K-MoS2 /CNT catalyst with K/Mo atomic ratio of 0.5. Higher potassium contents lowered the light olefin productivity. Somewhat lower extent of sulfidation of carbon nanotube molybdenum supported catalysts can be due to the presence of MoO2 . Methane seems to be produced on the unpromoted MoS2 phase. K-Mo mixed sulfides were detected in the K-promoted catalysts. These phases probably contain active sites for the synthesis of light olefins, alcohols and heavy hydrocarbons. Addition of potassium increases the basicity of the catalyst. A moderate basicity seems favorable for producing light olefins, while higher catalyst basicity is favorable for the production of C5+ longer chain hydrocarbons. Acknowledgment The authors thank O. Gardoll, M. Frere-Trenteseaux, L Burylo and K. Cheng for assistance with measuring of TPR, XPS, XRD, and TPD data. This work was performed within the ANR-NSFC OLSYNCAT project. The authors gratefully acknowledge the support of the French National Research Agency (ANR-11-IS09-0003) and National Natural Science Foundation of China (No. 21161130522). References [1] H.T. Torres Galvis, K.P. de Jong, ACS Catal. 3 (2013) 2130–2149. [2] H.A. Wittcoff, B.G. Reuben, J.S. Plotkin, Industrial Organic Chemicals, second ed., John Wiley and Sons, Inc., Hoboken, NJ, 2004, pp. 662-626. [3] E. Schwab, A. Weck, J. Steiner, Bay.F K., Oil Gas Eur. Mag. 1 (2010) 44–47. [4] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044–4098. [5] J.L. Oliphant, R.W. Fowler, R.B. Pannell, C.H. Bartholomew, J. Catal. 51 (1978) 229–242. [6] R.B. Pannell, K.S. Chung, C.H. Bartholomew, J. Catal. 46 (1977) 340–347. [7] C.H. Bartholomew, G.D. Weatherbee, G.A. Jarvi, J. Catal. 60 (1979) 257–269. [8] C.H. Bartholomew, R.B. Pannell, Appl. Catal. 2 (1982) 39–49. [9] C.H. Bartholomew, A.H. Uken, Appl. Catal. 4 (1982) 19–29. [10] C.H. Bartholomew, R.M. Bowman, Appl. Catal. 15 (1985) 59–67. [11] C.H. Bartholomew, Stud. Surf. Sci. Catal. 34 (1987) 81–104. [12] M. Ichikawa, Bull. Chem. Soc. Jpn. 51 (1978) 2268–2272.

C. Liu et al. / Catalysis Today 261 (2016) 137–145 [13] N. Koizumi, K. Murai, S. Tamayama, T. Ozaki, M. Yamada, Energy Fuels 17 (2003) 829–835. [14] G. Quarderer, K. Cochran, European Paten 0119609B1, 1991. [15] J. Lee, S. Kim, K. Lee, I. Nam, J. Chung, Y. Kim, H. Woo, Appl. Catal. A: Gen. 110 (1994) 11–25. [16] E. Bernard, L. Greg, H. Calvin, J. Catal. 89 (1984) 536–541. [17] J. Happei, M.A. Hnatow, U.S. Patent 4320030, 1982. [18] G.J. Quarderer, K.A. Cochran, European Patent Application 0119609, 1984. [19] B. Wang, G. Ding, Y. Shang, J. Lv, H. Wang, E. Wang, Z. Li, M. Xinbin, S. Qin, Sun.F Q., Appl. Catal. A: Gen. 431–432 (2012) 144–150. [20] G.J. Quarderer, K.A. Cochran, European Paten: 0119609B1, 1991. [21] L. Feng, X. Li, D.B. Dadyburjor, E.L. Kugler, J. Catal. 190 (2000) 1–13. [22] H.C. Woo, K.Y. Park, Appl. Catal. 75 (1991) 267–280. [23] J. Christensen, P. Jensen, A. Jensen, Ind. Eng. Chem. Res. 50 (2011) 7949–7963. [24] C. Liu, M. Virginie, A. Griboval-Constant, A. Khodakov, Appl. Catal. A: Gen. (2015), http://dx.doi.org/10.1016/j.apcata.01.031 [25] K.J. Smith, R.G. Herman, K. Klier, Chem. Eng. Sci. 45 (1990) 2639–2646. [26] K.P. De Jong, J.W. Geus, Catal. Rev. Sci. Eng. 42 (2000) 481–510. [27] V. Surisetty, A. Tavasoli, A. Dalai, Appl. Catal. A: Gen. 365 (2009) 243–251. [28] J. Christensenm, P. Mortensen, R. Trane, P. Jensen, A. Jensen, Appl. Catal. A: Gen. 366 (2009) 29–43. [29] Z. Li, Y. Fua, J. Bao, M. Jiang, T. Hu, T. Liu, Y. Xie, Appl. Catal. A: Gen. 220 (2001) 21–30. [30] R. Cattaneo, T. Shido, R. Prins, J. Catal. 185 (1999) 199–212. [31] D. Ferrari, G. Budroni, L. Bisson, N.J. Rane, B.D. Dickie, J.H. Kang, S.J. Rozeveld, Appl. Catal. A 462–463 (2013) 302–309. [32] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373–380. [33] A. Tavasoli, S. Karimi, M. Davari, N. Nasrollahi, T. Nematian, J. Ind. Eng. Chem. 20 (2014) 674–681. [34] J. Wang, J. Xie, Y. Huang, B. Chen, G. Lin, H. Zhang, Appl. Catal. A: Gen. 468 (2013) 44–51. [35] E. Liakakoua, E. Heracleousa, K. Triantafyllidisa, A. Lemonidou, Appl. Catal. B: Environ. 165 (2015) 296–305. [36] A. Andersen, S. Kathmann, M. Lilga, K. Albrecht, R. Hallen, D. Mei, J. Phys. Chem. C 115 (2011) 9025–9040. [37] T. Patterson, J. Carver, D. Leyden, D. Hercules, J. Phys. Chem. 80 (1976) 1700–1708. [38] A. Cimino, B. De, Angelis, J. Catal. 36 (1975) 11–22.

145

[39] L. Benois, D. Gonbeau, G. Pfister-Guillouzo, E. Schmidt, G. Meunier, A. Levasseur, Thin Solid Films 258 (1995) 110–114. [40] S. Andersson, R. Howe, J. Phys. Chem. 93 (1989) 4913–4920. [41] H. Woo, I. Nam, J. Lee, J. Chung, K. Lee, Y. Kim, J. Catal. 138 (1992) 525–535. [42] D. Zingg, L. Makovsky, R. Tischer, F. Brown, D. Hercules, J. Phys. Chem. 84 (1980) 2898–2906. [43] J. Hensley, S. Pylypenko, D. Ruddy, J. Catal. 309 (2014) 199–211. [44] N. Mcintyre, T. Chan, P. Spevack, Appl. Catal. 63 (1990) 391–401. [45] X. Yu, F. Liu, Z. Wang, Y. Chen, J. Electron Spectrosc. Relat. Phenom. 50 (1990) 159–166. [46] M. Jiang, G. Bian, Y. Fu, J. Catal. 146 (1994) 144–154. [47] H. Toulhouat, S. Kasztelan, in: M.J. Phillips, M. Ternan (Eds.), Proceedings, 9th International Congress on Catalysis Calgary (Chem.), vol. 1, Ottawa, Institute of Canada, 1988, pp. 152–153. [48] H. Xiao, D. Lia, W. Li, Y. Sun, Fuel Process. Technol. 91 (2010) 383–387. [49] R. Ramachandran, F. Massoth, Canad. J. Chem. Eng. 60 (1982) 17–22. [50] R. Burch, A. Collins, Appl. Catal. 18 (1985) 373–387. [51] J.L. Brito, J. Laine, J. Catal. 139 (1993) 540–550. [52] W. Lai, L. Pang, J. Zheng, J. Li, Z. Wu, X. Yi, W. Fang, L. Jia, Fuel Process. Technol. 110 (2013) 8–16. [53] P. Mangnus, A. Riezebos, A. Vanlangeveld, J. Moulijn, J. Catal. 151 (1995) 178–191. [54] B. Wang, G. Ding, Y. Shang, J. Lv, H. Wang, E. Wang, Z. Li, M. Xinbin, S. Qin, Q. Sun, Appl. Catal. A: Gen. 431–432 (2012) 144–150. [55] H. Galvis, J. Bitter, C. Khare, M. Ruitenbeek, A. Dugulan, K. de Jong, Science 335 (2012) 835–838. [56] J. Liu, E. Wang, J. Lv, Z. Li, B. Wang, X. Ma, S. Qin, Q. Sun, Fuel Process. Technol. 110 (2013) 249–257. [57] A. Muramatsu, T. Tatstmi, H. Tominaga, J. Phys. Chem. 96 (1992) 1334–1340. [58] N. Koizumi, G. Bian, K. Murai, T. Ozaki, M. Yamada, J. Mol. Catal. A: Chem. 207 (2004) 173–182. [59] S. DeCanio, M. Cataloo, E. DeCanio, D. Storm, J. Catal. 119 (1989) 256–260. [60] D. Li, C. Yang, W.H. Li, Y.H. Sun, B. Zhong, Top. Catal. 32 (2005) 233–242. [61] Z. Li, Y. Fu, M. Jiang, Appl. Catal. A: Gen. 187 (1999) 187–199. [62] S. Colley, R. Copperthwaite, G. Hutchings, G. Foulds, N. Coville, Appl. Catal. A: Gen. 84 (1992) 1. [63] E.E. Lowenthal, S. Schwarz, H.C. Foley, Catal. Today 51 (1999) 187–193.