lable at ScienceDirect

Microporous and Mesoporous Materials 221 (2016) 187e196

Contents lists avai

Microporous and Mesoporous Materials

journal homepage: www.elsevier .com/locate/micromeso

High-performance SS-fiber@HZSM-5 coreeshell catalyst formethanol-to-propylene: A kinetic and modeling study

Ming Wen, Jia Ding, Chunzheng Wang, Yakun Li, Guofeng Zhao, Ye Liu, Yong Lu*

Shanghai Key Laboratory of Green Chemistry and Chemical Processes, School of Chemistry and Molecular Engineering, East China Normal University,Shanghai 200062, China

a r t i c l e i n f o

Article history:Received 18 August 2015Accepted 22 September 2015Available online 9 October 2015

Keywords:Methanol-to-propyleneKineticsModeling studyStructured catalystZSM-5 zeolite

* Corresponding author. Tel./fax: þ86 21 62233424E-mail address: ylu@chem.ecnu.edu.cn (Y. Lu).

http://dx.doi.org/10.1016/j.micromeso.2015.09.0391387-1811/© 2015 Elsevier Inc. All rights reserved.

a b s t r a c t

For the methanol-to-propylene process, a lumped kinetic model was developed on the basis of dual-cyclereaction mechanism, which attempted to reflect the main reaction paths with a combination to show theevolution of mole fraction of individual light olefins (C2

] , C3] and C4

]) with space time. The experi-ments were performed in a continuous flow fixed-bed reactor at 0.1 MPa as well as varied reactiontemperature from 400 to 480 �C and space time from 0.3 to 32.0 gcatalyst h mol�1, and the experimentaldata obtained on the structured SS-fiber@HZSM-5 and powdered HZSM-5 catalysts were fitted byMATLAB software based on the established model. The fitted results show that the lumped kinetic modelwell describes the product distribution and is identified to be suitable by model identification. Comparedto the powdered HZSM-5, the SS-fiber@HZSM-5 shows higher diffusion efficiency and narrower resi-dence time distribution, not only promoting the propylene formation but also improving the utilizationefficiency of the structured HZSM-5.

© 2015 Elsevier Inc. All rights reserved.

1. Introduction

Propylene is one of the most important raw materials inchemical industry and widely used for the production of poly-propylene, acrylonitrile, and propylene oxide, as well as the syn-thesis of plastics, rubber and many other daily necessities [1,2]. Inrecent years, with the ever-increasing dilemma between thecontinuous consumption of the finite petroleum reserves and thesharply increased demands for propylene and its derivatives, thetraditional petroleum-based production (such as steam thermalcracking of naphtha and refinery fluid catalytic cracking (FCC)) isdifficult to meet the requirement of propylene. Therefore, it be-comes urgent to develop economical and energy-efficient pro-cesses to replace the petroleum-based production of propylene[1,2].

Natural gas [3] and some gasifiable carbon-rich materials suchas coal [4] and biomass [5] can be firstly transformed into syngas(mixture of H2 and CO) and subsequently converted into methanolthrough the well-established technologies [6]. Methanol can betransformed into hydrocarbons (methanol-to-hydrocarbons,

.

namely MTH) in some acidic zeolite catalysts (e.g., HZSM-5 andHSAPO-34) under appropriate reaction conditions [7]. Dependingon the desired product, MTH process is classified into the followingtypes: methanol-to-gasoline (MTG, mainly producing gasoline)[8,9], methanol-to-aromatics (MTA, mainly aromatics) [10,11],methanol-to-olefins (MTO, mainly ethylene and propylene) [12,13]and methanol-to-propylene (MTP, mainly propylene) [14e16]. Theglobal demand for propylene is growing faster than ethylene [6],and it is thus particularly desirable to develop catalysts preferableto MTP process.

Since the first discovery onMTH process byMobil researchers in1976 [17], its mechanism has become one of the most controversialtopics in heterogeneous catalysis [2]. In the early 1990s, Dahl andKolboe proposed the “hydrocarbon pool” mechanism over HSAPO-34 [18e20]. Subsequently, Svelle et al. investigated the MTHpathway over HZSM-5 by 13C labeling strategy, which revealed that13C is divided into two groups in the products: one group of C3eC6olefins and the other group of ethylene and aromatics [21].Accordingly, dual-cycle mechanism consisting of aromatic-basedcycle (main products of ethylene and aromatics) and olefin-basedcycle (main products of C3eC6 olefins) were proposed and widelyaccepted as the dominant reaction mechanism of MTH process onHZSM-5 [22]. Notably, the general process of olefin-based cycle ismethylating light olefins to higher olefins and higher olefins

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196188

cracking into light olefins, and therefore the generation and con-sumption of propylene can be approximately regarded as aconsecutive reaction.

Compared with other MTO catalysts such as SAPO-34, ZSM-5favors competing formation of the propylene with relatively slowcoke-induced deactivation, making it possible to use fixed bedreactor with excess catalysts [23]. In this context, most attentionhas focused on tuning the acidity (strength and density) [24e27],size- and/or morphology-controllable synthesis [28e31] and hier-archical design of the pore structure [32,33] of the ZSM-5 zeolite. Ina case of ZSM-5 synthesized via the fluoride route [28], its lowdensity of strong acid sites in combination with the long diffusionpathway and few crystal defects delivers a good propylene toethylene ratio of ~5 with 66% selectivity toward C2eC4 olefins andcomplete methanol conversion at 350 �C. Moreover, very highpropylene to ethylene ratio of above 10 is also reported to beobtainable over the ZSM-5 zeolite catalysts with additives (ZrO2and H3PO4) by using dimethyl ether (DME) and N2 as feedstock(DME to N2 mole ratio of 1:5) due to the depression of ethyleneformation rather than promotion of the propylene selectivity [34].

In spite of these promising results, their practical application ina fixed bed reactor is still significantly challenging, as macroscopicshape of microgranules or extruded pellets a fewmillimeters in sizeare required in the real-world forms rather than as-made powders.Therefore, some problems emerge in these cases of mass/heattransfer limitations, high pressure drop, non-regular flow patternand adverse effects of the used binders, which will reduce theintrinsic catalyst selectivity and activity. Recently, microfibrousstructured catalysts and reactors (MSCRs), as one primary kind ofstructured catalysts or catalytic reactors (SCRs), have been devel-oped and applied in order to achieve the aim of process intensifi-cation in chemical industry, because MSCRs and other SCRs can beprecisely designed in full detail up to the local surroundings of thecatalyst, and the exact shape and size of all column internals aredetermined by pre-design and calculation rather than trial anderror [35,36]. Therefore, these MSCRs exhibit great flexibility withrespect to different length scales (e.g., diffusion lengths and voi-dage) and allow exceptionally large rates and selectivities [36,37],and have many significant functional advantages over the con-ventional catalysts and reactors, providing broad prospects indesulfuration [38,39], air filtration [40,41], H2 fuel generation andcleanup [42e45], selective oxidation [46e52] and other fields[53e56], especially providing a new way of designing efficientcatalysts and reactors for mass and/or heat transfer limited re-actions [57]. In our previous studies [16,58], a new standing-freemicrofibrous-structured ZSM-5 zeolite catalyst was developed,being obtainable in a macroscopic scale by direct growth of theZSM-5 zeolite crystals onto a three-dimensional (3D) porousnetwork of sinter-locked metal microfibers. Thanks to the abovebeneficial properties, our structured SS-fiber@HZSM-5 coreeshellcatalyst demonstrated visible promotion on the selectivity to C2eC4olefins especially to propylenewith obviously prolonged lifetime instream, compared to the corresponding pure powdered HZSM-5catalyst. However, deep insight into such selectivity/stability pro-moting effect of the microfibrous-structured design is particularlydesirable to further improve the catalyst performance or designnext-generation catalyst with more advanced performance. Thiswork attempts to carry out kinetic and modeling study of MTPreaction over our structured SS-fiber@HZSM-5 catalyst as well aspowdered HZSM-5 catalyst for contrastive study, because it is auseful and helpful tool for us to accomplish this goal.

Due to the early rudimentary understanding of the complexityof reaction mechanism, rigorous kinetic treatments may neither bepracticable nor have much practical justification [59]. Subse-quently, some progress has been made in kinetic studies, but the

kinetic models suffer either from severe complexity or from over-simplification. For example, Park [60,61] formulated detailed ki-netic models at the elementary step level, including 726elementary steps, 142 olefins and 83 carbenium ions, which is socomplex that it is difficult to apply; while, in contrast, Aguayo [62]established a very simple kinetic model, only containing sevenlumps (oxygenates, n-butane, C2eC4 olefins, C2eC4 paraffins(without n-butane), C5eC10 fraction and methane), with regardingthe C2eC4 olefins as a lump, being unable to meet the requirementsof separately describing ethylene and propylene formation. It isthus necessary to further improve the model to study MTP reactionkinetics. Xiao [63,64] proposed a lumped kinetic model consistingof 17 reactions among 15 species, in which light olefins aredescribed separately, which could be applied to describe theirmonolith reactor behavior of MTP.

In the present work, we established a reaction network based onthe dual-cycle mechanism (olefin-based cycle and aromatic-basedcycle). This network not only simply described the rapid equilib-rium, olefin-based cycle, aromatic-based cycle, methanation, andthe generation of other alkanes and high-carbon hydrocarbons, butalso particularly described the individual formations of C2

], C3]

and C4]. Accordingly, a kinetic model (consisting of 19 reactions

involving 10 lumps: methanol, dimethyl ether, ethylene, propylene,butene, pentene, hexene, methane, C2eC6 alkane and C7þ) wasdeveloped. The experimental kinetic data obtained from ourstructured and powdered catalysts were fitted in our kinetic modelby MATLAB software, and the pre-exponential factors and activa-tion energies were obtained, indicating that the active center uti-lization efficiency and hexene cracking into propylene apparentreaction rate of the structured catalyst are higher than of thepowdered one. Therefore, the structured SS-fiber@HZSM-5 cor-eeshell catalyst exhibited longer lifetime and higher propyleneselectivity [16,58].

2. Experimental

2.1. Synthesis of the structured SS-fiber@HZSM-5 coreeshellcatalyst

The structured SS-fiber@HZSM-5 coreeshell catalyst was syn-thesized, as described elsewhere [58], by direct growth of ZSM-5crystals on macroscopic thin-sheet sinter-locked microfiber ofstainless steel 316-L fibers consisting of 15 vol% SS-fiber and 85 vol%voidage (SS-fiber, 20 mm in dia., purchased from Western MetalMaterial Co. Ltd., China). Firstly, circular chips (16.1 mm in dia.) ofSS-fiber substrate were soaked in HCl (1 wt%) aqueous solution for0.5 h, sonicated in acetone for 5 min, thoroughly washed usingdeionized water, and then dried at 80 �C in air. Secondly, the SS-fiber substrate chips were seeded with the ZSM-5 nanocrystalsusing dip-coating method. The dip-coating suspension was pre-pared by adding 1 wt% ZSM-5 seeds (SiO2/Al2O3 ratio ¼ 180; crystalsize: 100e200 nm) into a silica solegel (1 wt% SiO2) and its pH wasadjusted to 2.3 using HCl (1 wt%) aqueous solution prior to the dipcoating. Seeding of the ZSM-5 was performed by slowly dipping theSS-fiber substrates into the above suspension for 10 s, followed bydrying at 25 �C for 14 h and calcining at 450 �C for 2 h in air. Thirdly,the seeded SS-fiber circular chips (3 g) were placed in a Teflon-linedsteel autoclave (100 mL) filled with 70 mL synthesis gel consistingof tetraethylorthosilicate (TEOS, A.R.)/tetrapropylammonium(TPAOH)/NaOH (C.P.)/NaAlO2 (C.P.)/H2O with a mole ratio of 1/0.25/0.4/0.010/250 (corresponding to the gel SiO2/Al2O3 mole ratio of200), followed by gel-aging at room temperature for 4 h. Afterhydrothermal synthesis at 180 �C for 48 h, the resulting sampleswere washed thoroughly, dried at 100 �C overnight, and directlycalcined in air at 550 �C for 5 h to remove the organic template.

Fig. 1. The reactant conversion against the space time for the MTP reaction over thestructured SS-fiber@HZSM-5 catalyst (denoted as SS) and powdered HZSM-5 catalyst(denoted as Powder, hereinafter inclusive).

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196 189

Finally, the SS-fiber@ZSM-5 products were all converted to the H-form by the ion-exchanging with an aqueous NH4Cl (1 mol L�1)solution at 80 �C and calcining at 550 �C for 5 h in air, the HZSM-5zeolite content loaded on the catalyst was 19 wt%.

2.2. Synthesis of the powdered HZSM-5 catalyst

As previously noted, when using the synthesis gel with SiO2/Al2O3 mole ratio of 200, the SiO2/Al2O3 mole ratio of HZSM-5zeolite shell for the as-prepared structured SS-fiber@HZSM-5 is147 [58], while the SiO2/Al2O3mole ratio of the HZSM-5 collected atautoclave is consistent with that in the synthesis gel. Therefore, inorder to assure the comparative study of the structured andpowdered HZSM-5 catalysts at equivalent SiO2/Al2O3 mole ratio,the powdered HZSM-5 catalyst was hydrothermally synthesizedusing a synthesis gel of TEOS/TPAOH/NaOH/NaAlO2/H2O ¼ 1/0.25/0.4/0.014/250 (corresponding to the gel SiO2/Al2O3 moleratio ¼ 147) with other conditions unchanged. The real SiO2/Al2O3mole ratio for such obtained ZSM-5 zeolite was determined to be155 by means of inductively coupled plasma atomic emissionspectrometry (ICP-AES; Thermo IRIS Intrepid II XSP ICP spectrom-eter, USA). After ion-exchanging and calcining, the resulting HZSM-5 was pressed and then crushed and sieved to 100e300 mm as thepowdered catalyst.

2.3. MTP reaction kinetic measurement

MTP kinetic experiment of the catalysts was conducted in acontinuous flow fixed bed reactor, using a quartz tube reactor (i.d.16 mm, length 765 mm) vertically placed in a tubular electricfurnace. Methanol (A.R., Sinopharm Chemical Reagent Co., Ltd.) wasfed continuously using a high-performance liquid chromatography(HPLC) pump. Electrical heating tape was wrapped around theentrance of the reactor to control the temperature of 130 �C, pre-heating the liquid methanol to vapor in the highly purified N2 flowto obtain a gaseous mixture feed with methanol concentration of30 vol%. Each catalyst was packed into the reactor with a constantHZSM-5 zeolite content of 0.2 g. Note that circular chips (16.1 mmdiameter) of the SS-fiber@HZSM-5 catalyst were packed layer-up-layer into the tube reactor, the diameter of 0.1 mm larger thanthe i.d. of the tubular reactor was retained deliberately to avoid theappearance of the gap between the reactor wall and the edges ofcatalyst chips thereby preventing the gas bypassing. The powderedHZSM-5 catalyst was diluted using 100e300 mm quartz sand toassure an equivalent bed volume as to the microfibrous structuredbed. The effluent gas was quantitatively analyzed by an on-line gaschromatography-flame ionization detector (GC-FID, Shimadzu2014C) with a 30-m HP-PLOT/Q capillary column.

At the beginning of the experiment, the reactor tube was heatedin the highly purified N2 flow (heating rate of 10 �C/min), after thetemperature of the catalyst bed reached to the set value and wasstable, turn on the pump to input methanol, regulate the flow ofmethanol and nitrogen according to the space time value ¼ 32.0gcatalyst h/mol, corresponding to the methanol weight hourly spacevelocity (WHSV, to ZSM-5 zeolite mass) of 1 h�1. Then the spacetime was gradually reduced (i.e., the WHSV was graduallyincreased) at intervals of 1 h by adjusting the flow rate of N2 andmethanol. The product distribution was measured at differentspace time, until the space time and the catalysts' conversion ofmethanol were reduced to very low.

2.4. Definition of reactant conversion and product selectivity

Given that the very fast transformation of methanol-to-DMEinitially takes place, and more particularly, DME subsequently

acts as the raw material for all the following reactions, the meth-anol conversion (C) and product selectivity (Sj) are calculated bytreating DME as reactant based on carbon number and expressed asthe following equations:

C ¼ 1� xCH3OH þ 2xDMEPNj¼1 njxj

(1)

Sj ¼njxjPN

j¼1 njxj � xCH3OH � 2xDME(2)

where N is the number of components, xj is the mole fraction ofcomponent j in the reaction effluent, xCH3OH and xDME are respec-tively the mole fractions of methanol and DME in the reactioneffluent, nj is the carbon number of component j.

3. Results and discussion

3.1. Kinetic experimental data for our structured and powderedHZSM-5 catalysts

Fig. 1 shows the space time dependent reactant conversion(reactant includes methanol and DME, see previous definition ofreactant conversion in the section of “Experimental”) in MTP pro-cess using the structured SS-fiber@HZSM-5 and powdered HZSM-5catalysts. Clearly, the structured SS-fiber@HZSM-5 catalyst is ableto deliver 100% conversion over a much wider range of space time,compared to the powdered HZSM-5 catalyst. Complete conversionis maintained over the SS-fiber@HZSM-5 along with the space timeuntil to 3.2 gcatalyst h/molCH3OH while until to 16.0 gcatalyst h/molCH3OH over the powdered HZSM-5 only. We believe that themicro-structured design is paramount to the enhanced conversionat very short contacting time rather than the zeolite acidity becauseof the equivalent SiO2/Al2O3 ratio of the powdered HZSM-5 toHZSM-5 shell over the structured catalyst (structured, 147;powdered, 155). The possible explanation is that the SS-fiber@HZSM-5 catalyst bed presents much higher acid site utiliza-tion efficiency than the powdered HZSM-5 catalyst bed. For the SS-fiber@HZSM-5 catalyst, zeolite crystals are directly grown on thecylindrical fiber core surface to form a uniform density membraneshell as thick as ~5 mm. For the powdered HZSM-5 catalyst, thepowders are nearly spherical with a diameter of 100e300 mm,which is much larger than zeolite shell thickness of ~5 mm for the

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196190

SS-fiber@HZSM-5 (Supporting Fig. S1). As a result, the structuredSS-fiber@HZSM-5 catalyst significantly improves the mass transferby nature. Actually, it can be estimated from the evolution of con-version with space time that the SS-fiber@HZSM-5 catalyst offersthe maximum utilization efficiency of acid site (it can be reflectedby the space time corresponding to the conversion beginning to fallfrom 100%) much higher than the powdered HZSM-5 catalyst(Fig. 1). The structured design of HZSM-5 crystals provides a verylow turning space time of 3.2, being 1/5 lower than that of 16.0 overthe powdered HZSM-5 catalyst, meaning that the acid site utiliza-tion efficiency of the structured SS-fiber@HZSM-5 catalyst is 4-foldhigher than of the powdered HZSM-5 catalyst.

Besides improving the zeolite utilization efficiency, our struc-tured design of HZSM-5 zeolite provides ability to promote thepropylene formation. Fig. 2 shows the selectivity evolution ofethylene and propylene in the MTP reaction along with the spacetime over the two catalysts, and both of them show similar ethyleneselectivity evolution against space time (Fig. 2A). Whereas theethylene selectivity is increased with the space time below ~5 gca-talyst h/molCH3OH and then remains almost unchanged with furtherincreasing the space time, the powdered HZSM-5 catalyst alwaysdelivers higher ethylene selectivity than the structured one in thewhole space time range studied. As previously noted, our struc-tured design shows ability, to some extent, to suppress thearomatic-based reaction cycle that is favorable for the ethyleneformation [16,58]. In addition, at the space time below ~5 gcatalyst h/molCH3OH, it is observed that the ethylene selectivity almost seems

Fig. 2. The ethylene (A) and propylene (B) selectivities against the space time for theMTP reaction over the structured SS-fiber@HZSM-5 and powdered HZSM-5 catalysts.Note: the ethylene and propylene selectivities over the powdered HZSM-5 catalystwere not obtained at the space time of below 1.1 gcatalyst h/molCH3OH, because of almostundetectable methanol conversion.

independent on the reaction temperature of both catalysts. Thisobservation is most likely due to that the reaction is wholly gov-erned by diffusion limitation in the whole temperature rangestudied.

Unlike the ethylene formation, the structured catalyst shows aspace time dependent propylene formation evolution quitedifferent from the powdered catalyst (Fig. 2B). Over the structuredSS-fiber@HZSM-5 catalyst, the propylene selectivity increasessharply and reaches the climax along with the space time from 0.3increased to ~1.1 gcatalyst h/molCH3OH, and then decreases graduallywith further increasing the space time. Over the powdered HZSM-5catalyst, such peaking feature of propylene selectivity against spacetime is not observed, instead of a monotonously decreasing ten-dency. At 400 �C, the propylene selectivity of these two catalysts isbasically identical. At 420 �C and 450 �C, propylene selectivity of thestructured catalyst is higher than that of the powdered catalystwhen the space time is low (<10.7 gcatalyst h/molCH3OH at 420 �C and<~19.0 gcatalyst h/molCH3OH at 450 �C), and much higher with thedecrease of the space time. Most notably, at 480 �C, our structuredHZSM-5 catalyst always delivers much higher propylene selectivitycompared to the powdered HZSM-5 catalyst. We believe that ourstructured design of the HZSM-5 is paramount to the improvedpropylene selectivity. A possible explanation is that the micro-structured catalytic bed provides a narrow resistance time distri-bution, being essential for the enhanced formation of propylenethat is the intermediate of the serial methylation/cracking (namelyolefin-based cycle) pathway in the MTP mechanism [22,65].

Clearly, the above results of the ethylene and propylene for-mation evolution against the space time confirms their separated-generation mechanism while the structured design of HZSM-5shows an ability for promoting the methylation/cracking pathwayand inhibit the aromatic-based cycle. Therefore, aromatic-basedcycle and olefin-based cycle, which are favorable to form ethyleneand propylene respectively, should be taken into account to theMTP reaction network establishment for distinguishing the pro-duction of ethylene and propylene.

3.2. Reaction network and kinetic model for MTP process

We herein proposed a MTP reaction network based on the dual-cycle mechanism (olefin-based cycle and aromatic-based cycle,including 10 lumps: methanol, dimethyl ether, ethylene, propylene,butene, pentene, hexene, methane, C2eC6 alkane and C7þ, Fig. 3),with a notice of that the stoichiometric coefficients of all reactionswere omitted to write. In the reaction network, the main carbo-naceous matters were taken into account, with ignoring hydrogen,water and other trace matters (such as CO2, coke, etc.).

The reactions are classified into five categories (including therapid equilibrium (between methanol and DME), olefin-based cy-cle, aromatic-based cycle, methanation, and generation of otheralkanes and high-carbon hydrocarbons, Table 1) with a total of 19reactions. Initially, methanol molecules are quickly transformedinto DME, which could achieve a rapid equilibrium, and then thefollowing four kinds of reactions subsequently take place usingDME as reactant. Olefin-based cycle is divided into methylatinglight olefins to higher olefins (C5

] and C6]) and higher olefins

cracking into light olefins (major C3] and C4

], and minor C2]).

Methylation reactions include a small amount of propylene pro-duced in the initial induction period reacting with DME to generatebutene, butenewith DME to pentene, pentene with DME to hexene,and hexene with DME to C7þ; cracking reactions include hexenecracking into propylene, hexene into ethylene/butene, pentene intoethylene/propylene, and two butene molecules polymerizing andthen cracking into propylene/pentene. There are many kinds ofintermediate species which cannot be separated and detected

Fig. 3. The proposed reaction network for MTP process based on the dual-cycle mechanism.

Table 1Classification and rate equation of each reaction of MTP process.

Category Reaction Rate equation

Rapid equilibrium 2CH3OH����!k1 CH3OCH3 þ H2O r1 ¼ k1xCH3OH

CH3OCH3 þ H2O����!k2 2CH3OH r2 ¼ k2xDME

Olefin-based cycle 2C3H6 þ CH3OCH3����!k3 2C4H8 þ H2O r3 ¼ k3xDMExC3H6

2C4H8 þ CH3OCH3����!k4 2C5H10 þ H2O r4 ¼ k4xDMExC4H8

2C5H10 þ CH3OCH3����!k5 2C6H12 þ H2O r5 ¼ k5xDMExC5H10

2C6H12 þ CH3OCH3����!k6 2C7þ r6 ¼ k6xDMExC6H12

C6H12����!k7 2C3H6 r7 ¼ k7xC6H12

2C3H6����!k8 C6H12 r8 ¼ k8xC3H6

C6H12����!k9 C2H4 þ C4H8 r9 ¼ k9xC6H12

C2H4 þ C4H8�����!k10 C6H12 r10 ¼ k10xC2H4xC4H8

C5H10�����!k11 C2H4 þ C3H6 r11 ¼ k11xC5H10

C2H4 þ C3H6�����!k12 C5H10 r12 ¼ k12xC2H4xC3H6

2C4H8�����!k13 C3H6 þ C5H10 r13 ¼ k13xC4H8

Aromatic-based cycle CH3OCH�����!k14 C2H4 þ H2O r14 ¼ k14xDME

Methanation CH3OCH3 þ 2H2�����!k15 2CH4þH2O r15 ¼ k15xDME

C5H10þH2�����!k16 C4H8þCH4 r16 ¼ k16xC5H10

Generation of other alkanes and high-carbon hydrocarbons C3H6�����!k17 C2 � C6 alkane r17 ¼ k17xC3H6

CH3OCH3�����!k18 C2 � C6 alkane r18 ¼ k18xDME

C6H12�����!k19 C7þ r19 ¼ k19xC6H12

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196 191

easily by the gas chromatography in the aromatic-based cycle,moreover some studies [63,66] show that aromatic-based cycle isnot dominant for HZSM-5 catalyst. Thus the process is simplifiedwith DME to ethylene, and the specific steps are omitted. Metha-nation reactions include DME-into-methane as well as pentene-into-butane/methane. Moreover, C2eC6 alkane lump is assumedto be derived from propylene and DME as well as C7þ lump directlyfrom hexene.

It should be noted that the reaction network is semi-empiricaland cannot fully express the real reaction process. As the elemen-tary reactions and intermediate species in MTP process arenumerous and difficult to measure completely and accurately byexperiment, so the reaction network simplifies the actual reactionprocess and makes some assumptions on the premise of fitting tothe experimental data very well. Although it does not represent thereaction mechanism fully and accurately, the reaction network andcorresponding kinetic model canwell predict the effects of reactionconditions on product distribution.

According to the established reaction network, the rate equa-tions of all reactions are written in Table 1, where k is the rateconstant and x is themole fraction of the component in the reactioneffluent. The reaction orders are firstly obtained by assuming eachreaction is elementary reaction, and they are repeated modulatedto fit the experimental data and finally determined when thecalculated values and experimental values agree well. The rate

equation of the components can be expressed by the linear com-bination of the 19 reaction rate equations, where W is the mass ofHZSM-5, F0 is the feed mole flow rate of methanol (equations(3)e(12)). In order to fit more accurately, a few coefficients in frontof reaction rate constant in the kinetic equations are not in accor-dance with the stoichiometric coefficients of the chemical equa-tions in the reaction network, which may be due to the assumedreaction network simplifies the actual reaction mechanism, andcannot accurately represent the true reaction pathway. Therefore,some coefficients need to be modulated in order to make the dif-ference between the calculated and experimental values smaller.

rCH3OH ¼ dxCH3OH

dðW=F0Þ¼ �2r1 þ 2r2 (3)

rDME ¼ dxDME

dðW=F0Þ¼ r1 � r2 � r3 � r4 � r5 � r6 � r14 � r15 � r18

(4)

rC2H4¼ dxC2H4

dðW=F0Þ¼ r9 � r10 þ r11 � r12 þ r14 (5)

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196192

rC3H6¼ dxC3H6

dðW=F0Þ¼ �2r3 þ 2r7 � 2r8 þ r11 � r12 þ r13 � r17

(6)

rC4H8¼ dxC4H8

dðW=F0Þ¼ 2r3 � 2r4 þ r9 � r10 � 2r13 þ r16 (7)

rC5H10¼ dxC5H10

dðW=F0Þ¼ 3r4 � 2r5 � r11 þ r12 þ r13 � r16 (8)

rC6H12¼ dxC6H12

dðW=F0Þ¼ 3r5 � 2r6 � r7 þ r8 � r9 þ r10 � r19 (9)

rCH4¼ dxCH4

dðW=F0Þ¼ 2r15 þ r16 (10)

rC2�C6¼ dxC2�C6

dðW=F0Þ¼ r17 þ r18 (11)

rC7þ ¼ dxC7þdðW=F0Þ

¼ 2r6 þ 2r19 (12)

3.3. Numerical method for the kinetic study

The expression of the rate constant in Arrhenius equation is inthe form of reference temperature (723 K, i.e., 450 �C), that is theexponential term of reaction rate constant changes from the orig-inal ð1=TÞ to ðð1=TÞ � ð1=723ÞÞ, which can reduce the correlationbetween the pre-exponential factors and activation energies. It isexpressed as the equation (13) [67], where ki is the reaction rateconstant of reaction i, k0i is the pre-exponential factor of reaction i,k0i* is the pre-exponential factor of reaction i in the form of refer-ence temperature, Ei is the activation energy of reaction i, R is themolar gas constant, and T is the reaction temperature.

ki ¼ k0i exp�� EiRT

�

¼ k0i exp�� Ei

R

�1

723

��exp

�� Ei

R

�1T� 1723

��

¼ k*0i exp�� Ei

R

�1T� 1723

��(13)

Kinetic model parameter estimation is fitting the unknownparameters (the pre-exponential factors of each reaction rate con-stant for the reference temperature and the activation energies)with the experimental data based on rate equations of the com-ponents. In this work, we use MATLAB software to fit and estimatethe parameters. The process is to assume a set of initial parameters,then use the fourth order and fifth order variable step RungeeKuttamethod (i.e., MATLAB built-in ODE45 program) to solve the ordi-nary differential equations of components' rate and calculate eachcomponent's mole fraction in the reaction effluent correspondingto different temperatures and different space time. The objectivefunction is the residual sum of squares of calculated and experi-mental values:

OF ¼XMi¼1

XNj¼1

�xijc � xij

�2 (14)

where M is the number of experimental data sets, N is the numberof components, xij is the mole fraction of component j in the re-action effluent under the experimental condition i, while xijc rep-resents the calculated value corresponding to xij.

The purpose of parameter estimation is to minimize the objec-tive function value by appropriate optimization method. Thetermination condition of the optimization is that the relativeobjective function is changing by less than MATLAB default settingvalue of 10�6:OFn � OFn�1

OFn

� 10�6 (15)

where OFn is the objective function value of n-th optimization,OFn�1 is the objective function value of (n�1)-th optimization. If thetermination condition is satisfied, the assumed parameters areregarded as the final fitted values, otherwise the MATLAB built-inlsqnonlin program for solving the nonlinear least squares prob-lems is used to generate new parameters by optimization compu-tations, until the objective function meets the requirements.

3.4. Numerical simulation and promotion mechanism for catalyticperformance improvement

The experimental data (points) of the two catalysts at differenttemperatures were fitted by the above-described numericalmethod based on the proposed kinetic model, and the fitted curves(lines) were shown in Figs. 4 and 5. Subsequently, we adopted twoimportant statistics parameters for the model identification: r2 andF, and they are expressed as formulas (16) and (17), in which Ne isthe number of experimental points, Np is the dimension of themodel parameters, yei and yci are experimental value and calculatedvalue of the dependent variable for the i-th observation respec-tively. The larger the two values, the more reliable the model topredict the experimental results. For simplicity, r2 is generally set tobe >0.9 as well as F to be >10Fa (Np, NeeNp) (Fa value is obtainedfrom F-test table, significant level of a ¼ 5%, the Fa values of thesetwo catalysts are all less than 2) [68]. By calculating, for thestructured SS-fiber@HZSM-5 catalyst, r2 is 0.990 and F is 1250.9,while for the powdered HZSM-5 catalyst, r2 is 0.995 and F is 2161.9,which indicate all of the model identification parameters meet therequirements well. Moreover, the calculated points on the fittedcurves meet another requirement of that these points should beclose to and evenly distributed on both sides of the diagonal inparity plots (Figs. S2 and S3 in Supporting information) [68]. All theabove results indicate that the fitted curves for the evolution ofmole fraction of each component in the reaction effluent with spacetime wholly coincide well with the experimental data according toour kinetic model, and consequently our proposed kinetic modeland calculated parameter values are reliable for the two catalysts.

r2 ¼ 1�PNe

i¼1 ðyei � yciÞ2PNei¼1 y

2ei

(16)

F ¼

"PNei¼1 y

2ek �

PNei¼1 ðyei � yciÞ2

#=Np

PNei¼1 ðyei � yciÞ2=

�Ne � Np

� (17)

Table 2 lists the calculated kinetic parameters (pre-exponentialfactors at a reference temperature of 450 �C and activation en-ergies). It should be noted that, some calculated apparent activationenergies may be negative since the established reaction network isthe simplification of the actual reaction and each reaction is notelementary reaction, which has also been reported in other

Fig. 4. Comparison of calculated (lines) and experimental (points) values for the evolution of mole fraction of each component in the reaction effluent with space time over thestructured SS-fiber@HZSM-5 catalyst. A & B, 400 �C; C & D, 420 �C; E & F, 450 �C; G & H, 480 �C.

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196 193

literatures [69,70]. Since our study is based on the apparent ki-netics, the influence of diffusion has been contained in the kineticparameters.

Fig. 6 displays the main steps of the olefin-based cycle and thecorresponding reaction rate constants of the two catalysts at 450 �C,the generation and consumption of propylene can be approximatelyregarded as a consecutive reaction. The generation rate of propylenefrom hexene cracking is much faster than its consumption rate, and

hence,with the increase of contacting time, propylenemole fractionrapidly increases at the beginning, and then gradually decreasesafter reaches a maximum value, forming a turning point (Figs. 4Eand 5E). Since the rate constant of hexene cracking into propylenefor the structured SS-fiber@HZSM-5 catalyst is much larger thanthat for the powdered HZSM-5 catalyst and the residence timedistribution of structured SS-fiber@HZSM-5 catalyst is narrower,the turning point for the structured SS-fiber@HZSM-5 catalyst is

Fig. 5. Comparison of calculated (lines) and experimental (points) values for the evolution of mole fraction of each component in the reaction effluent with space time over thepowdered HZSM-5 catalyst. A & B, 400 �C; C & D, 420 �C; E & F, 450 �C; G & H, 480 �C.

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196194

more obvious and the corresponding mole fraction of propylene(~45%, as shown in Fig. 4E) is also much higher than that (~30%, asshown in Fig. 5E) for the powdered HZSM-5 catalyst.

For the gasesolid heterogeneous catalysis, reactant moleculesfirstly pass through the laminar boundary layer outside the catalyst,reach the outer surface of the catalyst (i.e., external diffusion), andthen diffuse along the internal hole of the catalyst to its innersurface (i.e., internal diffusion) [71], its apparent performance is a

coupling process of flow, transfer and reaction. For both catalysts,the impact of external diffusion on reaction is approximately sameat the same space time, and moreover, reactant concentration ofthe gas phase and catalyst outer surface is basically consistent dueto the large linear velocity in the reaction tube at low space time,and the influence of external diffusion to reaction can be ignored.Therefore, different internal diffusions of the two catalysts directlyresult in the different apparent catalytic performances.

Table 2Kinetic parameters determined by fitting the kinetic model to the experimentalresults.

Reaction k0i*(molCH3OH/h/gcatalyst)

Ei (kJ/mol)

A B A B

2CH3OH����!k1 CH3OCH3þH2O 2.14 0.36 13.46 �1.77CH3OCH3þH2O����!k2 2CH3OH 3.70 0.64 13.46 �1.772C3H6þCH3OCH3����!k3 2C4H8þH2O 13.90 1.73 34.62 9.722C4H8þCH3OCH3����!k4 2C5H10þH2O 43.92 7.32 44.03 30.082C5H10þCH3OCH3����!k5 2C6H12þH2O 47.25 16.96 16.49 3.232C6H12þCH3OCH3����!k6 2C7þ 25.21 20.28 125.89 151.27C6H12����!k7 2C3H6 123.31 82.27 109.08 0.222C3H6����!k8 C6H12 4.39 1.25 0.63 �181.54C6H12����!k9 C2H4þC4H8 0.86 13.12 1.15 120.32C2H4þC4H8�����!k10 C6H12 0.27 1.57 �103.56 �48.92C5H10�����!k11 C2H4þC3H6 2.64 0.57 �61.72 �143.09C2H4þC3H6�����!k12 C5H10 3.05 0.50 �123.64 �200.002C4H8�����!k13 C3H6þC5H10 0.03 0.01 235.91 283.90CH3OCH3�����!k14 C2H4þH2O 0.33 0.02 94.96 133.90CH3OCH3þ2H2�����!k15 2CH4þH2O 0.10 0.03 141.02 99.48C5H10 þ H2�����!k16 C4H8 þ CH4 0.01 0.01 63.90 5.62C3H6�����!k17 C2 � C6 alkane 0.01 0.01 �46.72 �43.38CH3OCH3�����!k18 C2 � C6 alkane 1.13 0.15 26.24 16.58C6H12�����!k19 C7þ 0.11 0.13 73.92 127.89

A: the structured SS-fiber@HZSM-5 catalyst; B: the powdered HZSM-5 catalyst.

Fig. 6. Comparison of the calculated reaction rate constants (beside arrows, the unit ismolCH3OH/h/gcatalyst) of the MTP olefin-based cycle main steps at 450 �C for thestructured SS-fiber@HZSM-5 catalyst (upper numbers) and the powdered HZSM-5catalyst (lower numbers).

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196 195

Internal diffusion and reaction occur simultaneously, and due toovercoming the diffusion resistance and reaction consumption, thereactant concentration gradually decreases with the internaldiffusion, and the reaction rate is correspondingly reduced. Reac-tion mostly proceeds in outer shell of the catalyst with a certainthickness, while the inner part contributes little to the reaction dueto the low concentration of reactants, so the inner surface of cata-lyst has not been fully utilized [71]. The internal diffusion effectivefactor h is introduced to judge the influence degree of internaldiffusion on the reaction process in the chemical reaction engi-neering, which is equal to the ratio of apparent reaction rate andintrinsic reaction rate (equation (18)) [71]. The h of 1 indicates thatthe internal part of catalyst has been fully utilized, while h of 0.1indicates only 10 vol% of catalyst has been efficiently used [72]. Theinternal diffusion effective factor ratio (hstructured/hpowdered) of thestructured and powdered catalysts at 450 �C can be obtained bysubstituting the kinetic parameters of methanol into equation (18).Due to the intrinsic reaction rates of the two kinds of catalysts arethe same, the ratio of the effective factor is equal to the ratio of the

apparent reaction rate, and is also the ratio of the reaction rateconstant, as shown in equation (19).

h ¼ robservedrintrinsic

(18)

hmonolithichpowdered

� rmonolithicrpowdered

¼ kmonolithickpowdered

¼ 2:140:36

¼ 5:9 (19)

Thus, the h of the structured SS-fiber@HZSM-5 catalyst is 4.9-fold higher than that of the powdered HZSM-5 catalyst. This ex-plains the reason why the evolution of conversion with space timeof the two kinds of catalysts is very different in kinetic experiments.For traditional powdered catalyst, the utilization efficiency of thecatalyst is reduced since the transfer and diffusion limitationshinder the contacting of reactants and catalyst active sites; whilethe thin zeolite shell of the structured SS-fiber@HZSM-5 catalystpossibly shorten the diffusion path, and therefore keeps a highutilization efficiency of the active sites. The structured SS-fiber@HZSM-5 catalyst is still capable of transforming raw mate-rials completely when the space time is very low, while only whenthe space time is higher can the conversion of powdered HZSM-5catalyst achieve 100%. In the previous kinetic experiment sectionwe have revealed that the turning space time (corresponding to theconversion beginning to fall from 100%, Fig. 1) of the structured SS-fiber@HZSM-5 catalyst is 4-fold higher than that of the powderedHZSM-5 catalyst, it is close to the reciprocal of their internaldiffusion effective factor ratio. It shows our kineticmodel is reliable,and the calculated kinetic parameters can well explain the exper-imental phenomena.

4. Conclusions

A MTP reaction network and a corresponding lumped kineticmodel have been established based on the dual-cycle reactionmechanism, and the model not only simplifies the complex reac-tion mechanism but also shows each kind of light olefin's variationwith reaction conditions separately. Kinetic experimental datawere fitted and analyzed by MATLAB software, and the calculatedvalues agree well with the experimental ones, while the modelidentification parameters also meet the requirements, indicatingthat the established model is appropriate. The calculated macro-scopic kinetic parameters of different catalysts show that the pro-pylene generation rate is much faster than its consumption rate.Because the generation and consumption process of propylene areconsecutive reactions, and the narrower residence time distribu-tion is more favorable to the formation of propylene, therefore,propylene selectivity of the structured SS-fiber@HZSM-5 catalyst ishigher than of the powdered HZSM-5 catalyst when the space timeis low especially at high temperatures. Moreover, the calculatedresults also indicate a higher utilization efficiency of the structured-design HZSM-5 by a factor of ~5 than of the powdered HZSM-5. Inconclusion, compared with the conventional powdered catalyst,our structured SS-fiber@HZSM-5 catalyst has narrower residencetime distribution and higher diffusion efficiency, which couldinhibit ethylene formation to promote propylene selectivity andincrease the utilization efficiency to enhance the catalytic activity.

Acknowledgments

This work was funded by the “973 program” (2011CB201403)from theMOSTof China, and the NSF of China (21473057, 21273075,U146212, 21076083). We gratefully acknowledge the partialfinancial support from the Science and Technology Commission ofShanghai Municipality.

M. Wen et al. / Microporous and Mesoporous Materials 221 (2016) 187e196196

Appendix A. Supporting information

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.micromeso.2015.09.039.

References

[1] Y. Park, C.W. Lee, N.Y. Kang, W.C. Choi, S. Choi, S.H. Oh, D.S. Park, Catal. Surv.Asia 14 (2010) 75.

[2] K. Hemelsoet, J. Van der Mynsbrugge, K. De Wispelaere, M. Waroquier, V. VanSpeybroeck, ChemPhysChem 14 (2013) 1526.

[3] D.A. Hickman, L.D. Schmidt, Science 259 (1993) 343.[4] N.M. Laurendeau, Prog. Energy Combust. 4 (1978) 221.[5] J. Andersson, J. Lundgren, M. Marklund, Biomass Bioenergy 64 (2014) 256.[6] T. Mokrani, M. Scurrell, Catal. Rev. 51 (2009) 1.[7] U. Olsbye, S. Svelle, M. Bjørgen, P. Beato, T.V.W. Janssens, F. Joensen, S. Bordiga,

K.P. Lillerud, Angew. Chem. Int. Ed. 51 (2012) 5810.[8] J.H. Ahn, B. Temel, E. Iglesia, Angew. Chem. Int. Ed. 48 (2009) 3814.[9] S.R. Blaszkowski, R.A. van Santen, J. Am. Chem. Soc. 119 (1997) 5020.

[10] X.J. Niu, J. Gao, Q. Miao, M. Dong, G.F. Wang, W.B. Fan, Z.F. Qin, J.G. Wang,Micropor. Mesopor. Mater. 197 (2014) 252.

[11] T. Wang, X.P. Tang, X.F. Huang, W.Z. Qian, Y. Cui, X.Y. Hui, W. Yang, F. Wei,Catal. Today 233 (2014) 8.

[12] Z.M. Cui, Q. Liu, W.G. Song, L.J. Wan, Angew. Chem. Int. Ed. 45 (2006) 6512.[13] S.J. Kim, H. Jang, J.K. Lee, H. Min, S.B. Hong, G. Seo, Chem. Commun. 47 (2011)

9498.[14] H. Koempel, W. Liebner, Stud. Surf. Sci. Catal. 167 (2007) 261.[15] B.B. Jiang, X. Feng, L.X. Yan, Y.T. Jiang, Z.W. Liao, J.D. Wang, Y.R. Yang, Ind. Eng.

Chem. Res. 53 (2014) 4623.[16] X.Y. Wang, M. Wen, C.Z. Wang, J. Ding, Y. Sun, Y. Liu, Y. Lu, Chem. Commun. 50

(2014) 6343.[17] C.D. Chang, A.J. Silvestri, J. Catal. 47 (1977) 249.[18] I.M. Dahl, S. Kolboe, Catal. Lett. 20 (1993) 329.[19] I.M. Dahl, S. Kolboe, J. Catal. 149 (1994) 458.[20] I.M. Dahl, S. Kolboe, J. Catal. 161 (1996) 304.[21] S. Svelle, F. Joensen, J. Nerlov, U. Olsbye, K. Lillerud, S. Kolboe, M. Bjørgen,

J. Am. Chem. Soc. 128 (2006) 14770.[22] M. Bjørgen, S. Svelle, F. Joensen, J. Nerlov, S. Kolboe, F. Bonino, L. Palumbo,

S. Bordiga, U. Olsbye, J. Catal. 249 (2007) 195.[23] Y. Traa, Chem. Commun. 46 (2010) 2175.[24] S. Ivanova, C. Lebrun, E. Vanhaecke, C. Pham-Huu, B. Louis, J. Catal. 265 (2009)

1.[25] N.H. Xue, X.K. Chen, L. Nie, X.F. Guo, W.P. Ding, Y. Chen, M. Gu, Z.K. Xie, J. Catal.

248 (2007) 20.[26] H. Koller, A. W€olker, H. Eckert, C. Panz, P. Behrens, Angew. Chem. Int. Ed. 36

(1997) 2823.[27] C.A. Fyfe, D.H. Brouwer, A.R. Lewis, J. Ch�ezeau, J. Am. Chem. Soc. 123 (2001)

6882.[28] F.L. Bleken, S. Chavan, U. Olsbye, M. Boltz, F. Ocampo, B. Louis, Appl. Catal. A

Gen. 447e448 (2012) 178.[29] S. Mintova, N.H. Olson, V. Valtchev, T. Bein, Science 283 (1999) 958.[30] A. Zabala Ruiz, H. Li, G. Calzaferri, Angew. Chem. Int. Ed. 45 (2006) 5282.[31] M. Choi, K. Na, J. Kim, Y. Sakamoto, O. Terasaki, R. Ryoo, Nature 461 (2009)

246.[32] C. Mei, P. Wen, Z. Liu, H. Liu, Y. Wang, W. Yang, Z. Xie, W. Hua, Z. Gao, J. Catal.

258 (2008) 243.

[33] N.L. Michels, S. Mitchell, J. P�erez-Ramírez, ACS Catal. 4 (2014) 2409.[34] T.S. Zhao, T. Takemoto, N. Tsubaki, Catal. Commun. 7 (2006) 647.[35] M.T. Kreutzer, F. Kapteijn, J.A. Moulijn, Catal. Today 111 (2006) 111.[36] K. Pangarkar, T.J. Schildhauer, J.R. van Ommen, J. Nijenhuis, F. Kapteijn,

J.A. Moulijn, Ind. Eng. Chem. Res. 47 (2008) 3720.[37] F. Kapteijn, T.A. Nijhuis, J.J. Heiszwolf, J.A. Moulijn, Catal. Today 66 (2001) 133.[38] Y. Lu, N. Sathitsuksanoh, H.Y. Yang, B.K. Chang, A.P. Queen, B.J. Tatarchuk, in:

Y. Wang, J.D. Holladay (Eds.), Microreactor Technology and Process Intensi-fication, American Chemical Society, Washington, DC, 2005.

[39] B.K. Chang, Y. Lu, B.J. Tatarchuk, Chem. Eng. J. 115 (2006) 195.[40] D.M. Kennedy, D.R. Cahela, W.H. Zhu, K.C. Westrom, R.M. Nelms,

B.J. Tatarchuk, J. Power Sources 168 (2007) 391.[41] R.R. Kalluri, D.R. Cahela, B.J. Tatarchuk, Appl. Catal. B Environ. 90 (2009) 507.[42] Y. Lu, H. Wang, Y. Liu, Q.S. Xue, L. Chen, M.Y. He, Lab Chip 7 (2006) 133.[43] Y. Liu, H. Wang, J.F. Li, Y. Lu, H.H. Wu, Q.S. Xue, L. Chen, Appl. Catal. A Gen. 328

(2007) 77.[44] Y. Liu, H. Wang, J.F. Li, Y. Lu, Q.S. Xue, J.C. Chen, AIChE J. 53 (2007) 1845.[45] M. Ling, G.F. Zhao, W. Chen, M.M. Wang, Q.S. Xue, Y. Lu, Int. J. Hydrogen

Energy 36 (2011) 12833.[46] M.M. Deng, G.F. Zhao, Q.S. Xue, L. Chen, Y. Lu, Appl. Catal. B Environ. 99 (2010)

222.[47] J.P. Mao, M.M. Deng, L. Chen, Y. Liu, Y. Lu, AIChE J. 56 (2010) 1545.[48] G.F. Zhao, H.Y. Hu, M.M. Deng, Y. Lu, ChemCatChem 3 (2011) 1629.[49] G.F. Zhao, H.Y. Hu, M.M. Deng, M. Ling, Y. Lu, Green Chem. 13 (2011) 55.[50] G.F. Zhao, H.Y. Hu, M.M. Deng, Y. Lu, Chem. Commun. 47 (2011) 9642.[51] G.F. Zhao, M.M. Deng, Y.F. Jiang, H.Y. Hu, J. Huang, Y. Lu, J. Catal. 301 (2013) 46.[52] G.F. Zhao, Y.K. Li, Q.F. Zhang, M.M. Deng, F.H. Cao, Y. Lu, AIChE J. 60 (2014)

1045.[53] W. Chen, W.Q. Sheng, F.H. Cao, Y. Lu, Int. J. Hydrogen Energy 37 (2012) 18021.[54] W. Chen, W.Q. Sheng, G.F. Zhao, F.H. Cao, Q.S. Xue, L. Chen, Y. Lu, RSC Adv. 2

(2012) 3651.[55] Q.F. Zhang, Y.K. Li, L. Zhang, L. Chen, Y. Liu, Y. Lu, J. Catal. 317 (2014) 54.[56] Q.F. Zhang, Y.K. Li, L. Zhang, L. Chen, Y. Liu, Y. Lu, Green Chem. 16 (2014) 2992.[57] M. Ling, G.F. Zhao, F.H. Cao, Y. Lu, Chin. J. Catal. (2010) 717.[58] M. Wen, X.Y. Wang, L.P. Han, J. Ding, Y. Sun, Y. Liu, Y. Lu, Micropor. Mesopor.

Mater. 206 (2015) 8.[59] C.D. Chang, Catal. Rev. 25 (1983) 1.[60] T. Park, G.F. Froment, Ind. Eng. Chem. Res. 40 (2001) 4172.[61] T. Park, G.F. Froment, Ind. Eng. Chem. Res. 40 (2001) 4187.[62] A.T. Aguayo, D. Mier, A.G. Gayubo, M. Gamero, J. Bilbao, Ind. Eng. Chem. Res.

49 (2010) 12371.[63] W.Z. Wu, W.Y. Guo, W.D. Xiao, M. Luo, Chem. Eng. Sci. 66 (2011) 4722.[64] W.Y. Guo, W.Z. Wu, M. Luo, W.D. Xiao, Fuel Process. Technol. 108 (2013) 133.[65] D. Lesthaeghe, J. Van der Mynsbrugge, M. Vandichel, M. Waroquier, V. Van

Speybroeck, ChemCatChem 3 (2011) 208.[66] D. Lesthaeghe, A. Horr�e, M. Waroquier, G.B. Marin, V. Van Speybroeck,

Chem.dEur. J. 15 (2009) 10803.[67] J.R. Kittrell, Mathematical Modeling of Chemical Reactions, Academic Press,

1970.[68] H.J. Huang, Practical Computer Simulation of Chemical ProcessesdMATLAB's

Application in Chemical Engineering, Chemical Industry Press, Beijing, 2004.[69] U. Sedran, A. Mahay, H.I. De Lasa, Chem. Eng. Sci. 45 (1990) 1161.[70] J. Wei, Chem. Eng. Sci. 51 (1996) 2995.[71] S.F. Li, in: Reaction Engineering, second ed., Chemical Industry Press, Beijing,

2000.[72] J. Gascon, J.R. van Ommen, J.A. Moulijn, F. Kapteijn, Catal. Sci. Technol. 5

(2015) 807.