Contents lists available at ScienceDirect

Journal of Analytical and Applied Pyrolysis

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

Group additive modeling of cyclopentane pyrolysis

Muralikrishna V. Khandavilli, Florence H. Vermeire, Ruben Van de Vijver, Marko Djokic,Hans-Heinrich Carstensen, Kevin M. Van Geem⁎, Guy B. MarinLaboratory for Chemical Technology, Ghent University, Technologiepark 914, 9052 Gent, Belgium

A R T I C L E I N F O

Keywords:CyclopentanePyrolysisSteam crackingKinetic modelUnimolecular decompositionPolycyclic aromatic hydrocarbonsCrackingAb-initoGroup additivity

A B S T R A C T

The pyrolysis of cyclopentane is not well established although it is an abundant compound in typical naphthafeedstocks and can be considered a model compound for cyclic fuels. The studies in literature so far have focusedprimarily on the initial decomposition of cyclopentane in shock tubes. This article therefore explores the pyr-olysis of cyclopentane in a continuous flow tubular reactor with pure cyclopentane feed at reactor conditions0.17 MPa, 973–1073 K, and a residence time of 0.5s. Conversions of 5% to 75% were realized while the productconcentrations were quantified using two dimensional gas chromatography. A mechanism composed of ele-mentary high pressure limit reactions has been generated using the automatic network generation tool“Genesys”. Kinetics of the reactions originate from high level ab-initio calculations and new group additivevalues derived from ab-initio kinetic data in literature. Overall the Genesys model outperforms the modelsavailable in literature and there is a good agreement between model calculated mass fraction profiles and ex-perimental data for 22 products ranging from hydrogen to naphthalene without any adjustments of the kineticparameters. Reaction path analysis reveals that cyclopentane consumption is initiated by the unimolecularisomerization to 1-pentene, but overall dominated by hydrogen abstraction reactions by allyl radicals and hy-drogen atoms to give cyclopentyl radicals, whose ring opening and further scissions lead to smaller molecules.Dominant routes for the major products are discussed.

1. Introduction

Research in pyrolysis of cyclic hydrocarbons is important for in-dustrial processes like fast pyrolysis and steam cracking [1]. There is ahuge pressure on the industry to produce fuels and lower olefins moreeconomically than ever before. This implies more optimal operatingconditions and using a cheaper feedstock. There is a limit to which theoperating conditions can be optimized within the boundary conditionsof metal corrosion and coke deposition [2]. The lucrative knob, which isalso the topic of one of the most relevant research areas is to explore acheaper feedstock. Typically, naphthenes and aromatics are undesirablefeed components for pyrolysis, hence petrochemical cuts or othersources containing higher quantities of those components are cheappyrolysis feeds. The tolerance toward naphthenes is relatively higherthan for aromatics because of their lower coking tendency. At the sametime, the pyrolysis of naphthenes is a relatively unexplored field com-pared to their linear and branched counterparts [3]. Among the sim-plest naphthenes, which are the single ring unsubstituted cycloalkanes,the ones most abundantly found in petrochemical feeds like lightnaphtha are cyclohexane and cyclopentane [4]. Out of these two, thepyrolysis of cyclohexane is a relatively mature research area [5–8].

Surprisingly, the one molecule whose pyrolysis behavior has not beenstudied in detail so far is cyclopentane. According to the authors’knowledge, no previous work has been published related to the pyr-olysis of cyclopentane, more so at steam cracking conditions.

The few literature studies involving pyrolysis of cyclopentane havefocused on experimental investigation of auto-ignition and initial de-composition products in shock tubes [9–11]. In order to explain theignition delay trends, kinetic models have been developed for cyclo-pentane oxidation. However, such models, though good for ignitiondelay time predictions, are generally not ideal for pyrolysis becausethey tend to focus on extremely short residence times, which impliesthat primarily the initial decomposition characteristics are well cap-tured. Also, the hydrocarbon feed is highly diluted and this is an un-realistic representation of an industrial pyrolysis reaction as the poly-cyclic aromatic hydrocarbons are not formed at high dilutions and lowresidence times to the extent that are usually formed in steam crackingfor example. Here we discuss a few relevant studies reported so far.

Around four decades ago, Tsang [11] did experiments to find outinitial decomposition rates of cyclopentane in a comparative-ratesingle-pulse shock-tube. In this, the initial cyclopentane decompositionrate was established after comparing the experiment with that of a

http://dx.doi.org/10.1016/j.jaap.2017.08.005Received 28 December 2016; Received in revised form 14 June 2017; Accepted 8 August 2017

⁎ Corresponding author.E-mail address: kevin.vangeem@ugent.be (K.M. Van Geem).

Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

Available online 12 August 20170165-2370/ © 2017 Elsevier B.V. All rights reserved.

MARK

standard reaction whose temperature decay characteristics were wellknown. This was the well-studied retro-Diels-Alder decomposition ofcyclohexene feed giving ethylene and 1,3-butadiene. The residencetime in the shock tube reached a maximum of 0.8 milli-second attemperatures 1000 K to 1200 K and pressures 2 bara to 6 bara with amaximum of 5% cyclopentane in feed. Cyclopentane conversions ob-tained were in the range of 0.01% to 1%. The main product detectedwas 1-pentene and based on product yields, the isomerization reactionof cyclopentane to give 1-pentene was hypothesized as the initialdominant decomposition step and a global rate coefficient for the samewas proposed. The shock tube experiments also detected cyclopropaneand a minor channel forming cyclopropane and ethylene from cyclo-pentane was proposed with a global rate coefficient, though at thetemperatures of interest for steam cracking, the mass yield of cyclo-propane was about 100–1000 times lower than that of 1-pentene. Tsangproposed a C5 biradical intermediate as the immediate product of cy-clopentane decyclization. This biradical was suggested to form majorly1-pentene through intra-molecular hydrogen abstraction, and minorlycyclopropane and ethylene by CeC beta scission.

About 30 years later, in 2006, Sirjean [12] did high level ab-initioCBS-QB3 calculations on ring opening of cycloalkanes. In that, cyclo-pentane was studied and a biradical C5 intermediate was envisaged,along the lines of Tsang [11]. Sirjean proposed CBS-QB3 based globalrate coefficients for cyclopentane conversion to 1-pentene and that tocyclopropane and ethylene. The rate coefficient of the reaction forming1-pentene matched well with that proposed by Tsang [11]. Cyclopro-pane was confirmed to be a minor product and the rate coefficient ofthe reaction forming cyclopropane and ethylene was about 1% of thatforming 1-pentene at 1000 K. In this study, ab-initio calculations re-vealed that the C5 biradical has a bond rotation energy barrier equal tothe activation energy for intra-molecular hydrogen abstraction to form1-pentene. Hence, as soon as the cyclopentane ring opens to form C5biradical, the C5-biradical undergoes bond rotation and a simultaneousand immediate conversion to 1-pentene. As a practical inference, for acomplete cyclopentane pyrolysis model, the initiation step can be re-presented as a single elementary step isomerization to 1-pentene. BothTsang [11] and Sirjean [12] studied the initial decomposition of cy-clopentane and not its complete decomposition to smaller hydrocarbonsand hydrogen nor the molecular growth to polycyclic aromatic hydro-carbons. Annesley and co-workers [13,14] measured cyclopentane ringopening kinetics in a Laser-Schlieren apparatus at 1500–2000 K at lowpressures (total pressures 40–400 mbar), with cyclopentane partialpressures 0.1–20 mbar. The rate expression for cyclopentane iso-merization to 1-pentene was derived from experiment and RRKMtheory was applied to make an extrapolation to the high pressure limit.The rate coefficient determined by Annesley et al. [13,14] matches theab-initio value of Sirjean et al. within 8% deviation at 1050 K. There-fore the value of Sirjean et al. was retained. The same authors also

concluded that no pressure dependence needs to be accounted for aspressure has a minimal effect on the rate coefficient for cyclopentaneisomerization to 1-pentene. The rate coefficient at 40 mbar is onlyabout 6% lower than the one at high pressure limit.

In 2015, Wang [15] proposed a comprehensive ab-initio based ki-netic model for propylene pyrolysis with extensive potential energysurface scans of species including cyclopentane. This model predictsanother initiation step for cyclopentane decomposition in addition toisomerization − that of CeH homolytic bond scission to give hydrogenatom and cyclopentyl radical. However, the rate coefficient for thisreaction was found to be negligible (about 200 times smaller at 1000 K)compared to that of isomerization to 1-pentene. The Wang [15] modelalso contained some reactions involving cyclopentane and cyclopentylradicals with ab-initio pressure dependent Chebyshev kinetics. Thoughoriginally intended for predicting propylene pyrolysis, it has the cap-ability to attempt prediction of cyclopentane pyrolysis too. However,the model uses single-step lumped reactions to form aromatics likebenzene, toluene, styrene, indene and naphthalene from 1,3-cyclo-pentadiene, so it is not a completely elementary reaction model. It has amixture of pressure dependent kinetics, high pressure limit rate coef-ficients, for some reactions altered kinetic parameters and globallumped kinetics for aromatics formation. Inspite of these features, it isthe most relevant model in literature that comes close to meet ourobjective of describing the complete pyrolysis of cyclopentane and notjust the initial decomposition trends. In addition, this model is also asource of ab-initio kinetic data especially of reactions involving cyclicswhose group additive values are not yet reported, and whose rateparameters can be used to derive new group additive values for cyclicreactions. The model is also a source of ab-initio thermodynamics ofcyclic species.

The other kinetic and thermodynamic data that may be required forthe present study is that involving molecular growth to form aromaticsand polycyclic aromatic hydrocarbons. These mechanisms involvemany complex species, sometimes bicyclic and tricyclic molecules andmay involve reaction families not usually relevant for pyrolysis ofsimple open chain molecules. It is generally believed that 1,3-cyclo-pentadiene is the precursor to the formation of aromatics and poly-aromatics [16]. Cyclopentane and 1,3-cyclopentadiene being of similarskeletal structure, it is expected that a significant amount of 1,3-cy-clopentadiene would be formed during cyclopentane pyrolysis whichcould in turn trigger aromatics formation and growth. Merchant'smodel [17] was reported to predict pyrolysis of 1,3-cyclopentadieneand ethylene feed by a model of more than 5000 elementary reactionsforming benzene, toluene, styrene, indene and naphthalene. Merchanthad done ab-initio investigation of kinetics of the most dominantpathways.

Among other studies on cyclopentane pyrolysis, Sirjean [10] didshock tube experiments with cyclopentane-oxygen-argon mixtures and

Nomenclature

List of symbols

k T( )TST Rate coefficient at temperature T, m3 mol−1 s−1 for bi-molecular, s−1 for unimolecular

χ T( ) Quantum mechanical tunneling correction factorkB Boltzmann’s constant, m2 kg s−2 K−1

h Planck’s constant, m2 kg s−1

RTp

Molar volume at 1 atm, m3

nΔ Molecularity of the reaction (2 for bimolecular, 1 for un-imolecular reactions)

GΔ ǂ Gibbs free energy difference between transition state andreactant(s) without the transitional mode, kJ

A Pre-exponential factor m3 mol−1 s−1 for bimolecular, s−1

for unimolecularà Single-event pre-exponential factor m3 mol−1 s−1 for bi-

molecular, s−1 for unimolecularEa Activation energy of a reaction, kJ/molΔGAV°(log Ã) Standard Δ group additive value for single-event pre-

exponential factorΔGAV°(Ea) Standard Δ group additive value for activation energy,

kJ/molne Number of single eventsNNI Non-nearest neighbor interaction (for logà and Ea)exo Exocyclic ring intra-molecular carbon centered radical

additionendo Endocyclic ring intra-molecular carbon centered radical

addition

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

438

cyclohexane-oxygen-argon mixtures, both with 0.5 or 1% hydrocarbonat temperatures 1230 K to 1840 K and pressures 7.3 atm to 9.5 atm. Thepurpose was to measure ignition delay times. The residence time in theshock tube was less than 1 milli-second by which time the ignition hadalready started. Here too, like Tsang [11], the cyclopentane was pyr-olyzed at extremely short residence times without allowing completepyrolysis and formation of polycyclic aromatic hydrocarbons. Alsothere was no analysis of products like in Tsang’s case, as the objectiveswere different here. In order to explain the ignition delay time mea-surement, Sirjean formulated a model using EXGAS [18–21] for cy-clopentane pyrolysis and oxidation. One important conclusion fromSirjean’s study was that cyclopentane is significantly more stablecompared to cyclohexane, as the ignition delay times of cyclopentanewere 10 times higher than those of cyclohexane.

Tian [9] also measured ignition delay times of cyclopentane/oxygenmixtures in a shock tube at 1150–1850 K and 1–10 atm. Tian qualita-tively confirmed what Sirjean reported about the stability of cyclo-pentane. However, in this study, cyclopentane was compared to methylcyclopentane. In order to explain the ignition delay times, Tian ex-tended the model of JetSurF 2.0 [22] to include cyclopentane pyrolysisinitiation kinetics to generate a more complete model than just theinitial decomposition reactions. Summarizing, in none of the studiesreported in literature was a complete pyrolysis experiment of cyclo-pentane done with analysis of full range of products. And none of themodels reported had consistently elementary step high pressure limitreactions with ab-initio or ab-initio based (group additive) kineticswithout parameter alterations.

In this article, for the first time, an experimental and kinetic mod-eling study of cyclopentane pyrolysis is presented. Experiments havebeen conducted in a continuous flow reactor at 0.17 MPa, 973–1073 Kand a residence time of 0.5 s with pure cyclopentane feed. The in-housemodel generator tool “Genesys” [23] was used to auto-generate anelementary step high pressure limit kinetic model for cyclopentanepyrolysis having ab-initio/group additive kinetics and thermodynamicswith no adjustment of kinetic parameters. Model trends were comparedto the experiment and three other models in literature, all involving oneor more global reactions to varying extent[9,10,15]. Reaction pathanalysis is presented detailing the dominant pathways to the majorproducts.

2. Experimental setup and procedure

Cyclopentane pyrolysis experiments were carried out in a benchscale unit which consisted of three main sections − feed section, re-actor section and product analysis section. It has been described indetail elsewhere [16,24–26]. Only the specifics related to this work willbe given. Cyclopentane was procured from Sigma-Aldrich. The cyclo-pentane purity was determined to be 98.40wt% ± 0.05 based on threerepeat analysis. The main impurity is pentane (1.00 wt%), with also0.60 wt% 2,2-dimethyl butane being present. In the supporting in-formation, a detailed description of the experimental apparatus and theanalytical equipment, i.e. the details of the analytical procedures, thetemperature profiles and the columns used in the GC×GC analyses, isgiven. The length of the tubular reactor is 1.47 m with 6 mm internaldiameter, made of Incoloy 800HT (Ni, 30–35; Cr, 19–23; and Fe, >39.5 wt%). The reactor is placed vertically in a rectangular furnaceand heated electrically (supporting information). In all experiments, thereactor is operated nearly isothermally, i.e. with a steep temperatureincrease at the reactor inlet and a steep temperature drop at the outletof the reactor. Thermocouples monitor the process gas temperature ateight axial positions (supporting information).The measured tempera-ture profiles as a function of the axial reactor coordinate are given insupporting information. In the setup, Type K thermocouples are usedfor which the manufacturer calibrated accuracy is± 2.2 °C for tem-perature range 0–1250 °C or 0.75% of reading in °C whichever isgreater. The pressure in the reactor is controlled by an outlet pressure

restriction valve. Two manometers, situated at the inlet and outlet ofthe reactor, allow measuring the coil inlet pressure (CIP) and the coiloutlet pressure (COP), respectively. The pressure drop over the reactorwas found to be negligible, with the pressure remaining constant in thereactor at 1.7 bara. In the setup, Type K thermocouples are used forwhich the manufacturer calibrated accuracy is: for temperature range0–1250 °C (32–2300 °F):± 2.2 °C or 0.75% of reading in °C whicheveris greater. For the temperature range of interest, this comes to anaverage accuracy of± 4 °C. Pure cyclopentane was pumped through acoriolis mass flow controller at 240 gh−1 and fed to an electricallyheated tubular reactor made of Incoloy 800HT. No diluent was used,while nitrogen at 20 − 70 gh−1 was added to the reactor effluent andserved as an internal standard. The flow rate of cyclopentane waschosen so as to obtain a residence time in the order of 0.5 s. Thepressure was maintained at 0.17 MPa using an outlet pressure restric-tion valve situated at the exit of the effluent line. In all the experiments,the reactor was operated nearly isothermally in the temperature rangefrom 973 K to 1073 K, i.e. with a steep temperature increase at thereactor inlet and a steep temperature drop at the outlet of the reactor.Carbon balances closed within 1%.

The reactor effluent is sampled on-line, i.e. during operation, and athigh temperature (625 K). The heated sampling system that consists oftwo high temperature 6-port 2-way valves and that is kept at 575 K toprevent condensation of high molecular weight components [25]. Asshown by Van Geem et al. [27], the temperature at which samplingoccurs is well above the dew point of the effluent sample. The productanalysis section consisted of two different gas chromatographs for adetailed analysis of the reactor effluent: a so-called refinery gas ana-lyzer (RGA) and a GC × GC-FID/TOF-MS. The former enabled analysisof permanent gases and light hydrocarbons (C1–C4), while the latteranalyzed the whole product spectrum. Response factors for the formerwere determined by calibration with a synthetically created gas mixture(supplied by Air Liquide, Belgium). The response factors of the C5+

compounds were determined using the effective carbon number methodrelative to methane. GC × GC is an arrangement of 2 columns in serieswith a modulator in between them, all situated in a single oven withprogrammable temperature. The first GC typically separates com-pounds based on volatility and the second GC does so based on polarity.This enables a fast group-type analysis and provisional classifications ofunknown compounds [28]. Quantification in GC × GC has as mainadvantages over 1D-GC such as:

a Ordering which makes interference due to peak overlap less likely[29,30].

b Greater sensitivity or detectability due to the high speed of thesecond column. The resulting peaks are sharper and, therefore, ex-hibit a higher signal response [30].

c Reliable presence of a true baseline for peak integration [30].

Every experiment for a given temperature set point was repeatedthree to five times to establish repeatability. This yielded a relativeerror of less than 10% on the mass fractions of products.

3. Kinetic model generation

3.1. Reaction families and mechanism generation

The objective is to create a cyclopentane pyrolysis model consistingof only elementary reversible reactions whose kinetics and thermo-dynamics are based on ab-initio electronic structure calculations. Thefirst step is to select the relevant reaction families. For ethane cracking[31], Sabbe et al. used the following reaction families: hydrogen ab-straction, hydrogen atom addition, carbon radical addition and re-combination (and their reverse reaction families). Based on relatedliterature [12,15,17], the following reaction families and reactions alsohave been considered: Intra-molecular carbon centered radical

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

439

addition, Intra-cyclopentadiene hydrogen shift, Intra-molecular hy-drogen abstraction, cycloalkane isomerization to 1-alkene, Diels-Alderreaction, H2 elimination from cyclopentene. Table 1 shows the com-plete list of reaction families considered. With this, we were ready witha hopefully complete list of elementary reaction families to generate amechanism for cyclopentane pyrolysis.

The in-house automatic mechanism generation tool, “Genesys” [23]was used to generate the mechanism for cyclopentane pyrolysis.Genesys does not have pre-defined reaction families, but offers the usersflexibility to define the reaction families by creating reaction familyrecipes in a simple manner. Reaction recipes were created for the fa-milies listed in Table 1. Fig. 1 shows an example recipe for intra-cy-clopentadiene hydrogen shift having the 1,3-cyclopentadiene moiety. Itcan be seen that only 6 atoms are part of the recipe: carbon atomnumbers 1–5 and hydrogen atom number 6. All the 5 carbon atoms arecalled reactive centers as all of them are affected by the bond re-arrangement. The groups R1-R5 are called the nearest neighbors to thereactive centers. The feed molecule to Genesys is cyclopentane. In thefirst loop of the mechanism generation algorithm, this molecule istested for possible reactions included in Table 1. Accordingly, 1-pen-tene is generated by isomerization. In the next iteration, both 1-penteneand cyclopentane are checked for possible reactions among the reac-tions of Table 1, and CeC recombination (its reverse reaction) becomesactive on 1-pentene to generate new species. This procedure is con-tinued − the list of species present at any iteration being tested forpossible reactions to generate further species. In this way, the me-chanism grows until it reaches a constraint and there are no more newspecies generated. The constraint used for the present case was thecarbon number of 10 per species based on the idea to stop the moleculargrowth at naphthalene. Algorithmic details about mechanism genera-tion using Genesys, including schematics can be found in [23,32,33].However, a short description is provided here. Genesys is an in-housedeveloped automatic kinetic model generation tool, in which the size ofthe kinetic model is controlled by a rule based termination criterion. InGenesys, starting from the user-defined reaction families and the initialfeed molecules, an exhaustive mechanism is generated which is ter-minated by the user-defined constraints on product species, like con-straints on the maximum carbon number allowed, and constraints onthe reaction families, like limiting the size of the abstracting and addingspecies. The reaction families are defined by the user in Genesys bysupplying a reaction recipe, the possible reactive center by the user-friendly SMARTS language and constraints on the appearance of thereaction family or the products formed by this reaction family. Owingto the ever increasing power of computers and speed of ab-initio cal-culations, new types of elementary reactions are being discovered [34]and will be discovered in future. However, for the choice of elementaryreaction families in Genesys, current knowledge and experience is ap-plied. Fig. 2 shows the algorithm in Genesys for the generation of thekinetic model. The completeness of the mechanism generated byGenesys depends on the selected reaction families. The reaction familiesapplied in Genesys is a subjective choice of the user based on priorexperience, in the present case, those listed in Table 1. Genesys can nowalso be used using on the fly calculations [33].

3.2. Kinetics and thermodynamics assignment

After the mechanism generation is complete, Genesys assigns kineticparameters to the reactions and thermodynamic parameters to thespecies from user-defined databases. The kinetics databases containkinetics in the form of standard group additive values (ΔGAV°) of singleevent Arrhenius pre-exponential factors and activation energies. TheseΔGAV°s are defined as a function of groups (molecular fragments/branches, example R1-R5 in Fig. 1) attached to the reactive centers for aparticular reaction family. The ΔGAV°s for a few reaction families havealready been reported [35–37], and were used in the Genesys kineticsdatabases as such. The number of single events, based on symmetry

numbers and optical isomers of reactants and transition states is cal-culated by an auxiliary code of Genesys called SIGMA [38]. For ther-modynamics, ab-initio/group additive parameters exist [15,17,31] inthe form of NASA polynomial coefficients. These coefficients werecoded into the species thermodynamic database of Genesys so that theyare assigned when a particular species is found in the mechanism. Forspecies which are not included in the database, thermodynamics can beestimated by group additivity in Genesys. The ΔGAV° databases of thenew reaction families needed for cyclopentane pyrolysis model gen-eration (intra-cyclopentadiene hydrogen shift, intra-molecular carbonradical addition leading to mono, bi and tri-cyclic species and intra-molecular hydrogen abstraction within cyclic species) were completedfrom literature [15,17]. Tables 2 and 3 show the ΔGAV°s for the newreaction families: intra-cyclopentadiene hydrogen shift and intra-mo-lecular carbon radical addition. In these tables, it can be seen that atsome places, only a single reaction leads to a ΔGAV° value. Typicallymultiple reactions are used to derive a ΔGAV°. However, since the mainobjective of this article is not derivation of group additive values, thecurrent procedure is sufficient toward generating kinetics of importantreactions. The atom labeling for intra-cyclopentadiene hydrogen shiftand intra-molecular carbon centered radical addition are shown inFigs. 3 and 4 using illustrative examples. In Tables 2 and 3 the effect ofnon-nearest neighbors on the rate coefficient is negligible. This is in linewith the general philosophy of group additivity that the rate coefficientdepends only on the groups attached to the reactive centers and notthose far away from the place of action. An example of how Arrheniusparameters are calculated starting from group additive values is givenin the supporting information.

Based on these input databases, Genesys assigned kinetics to thereactions of the mechanism, yielding an automatically generated groupadditive model for cyclopentane pyrolysis. Genesys generated themodel as a CHEMKIN [39] readable input file and initial simulationswith cyclopentane feed pointed toward a few dominant reactions whosekinetics were worthwhile to be checked by a high level CBS-QB3computation [40] by the method shown next.

Table 1Reaction families considered in this work.

Reaction family Description/Example Reference

Hydrogen transferInter-molecular H-abstraction by C%

by H% [37], In-houseIntra-molecular H-abstraction

(acyclics)Ce(C)neC% ⇄ %Ce(C)neC [31]

Intra-molecular H-abstraction(cyclics)

[15,17]

Intra-CPD-H-shift [17]

AdditionInter-molecular C-radical addition C]C + %C ⇄ %CeCeC [35]Intra-molecular C-radical addition [15,17]

Hydrogen atom addition %CeCeH ⇄ C]C+ %H [36]

RecombinationCeC recombination %C + %C ⇄ CeC [15,17,31]CeH recombination %C + H ⇄ CeH [15,17,31]

H2 eliminationDirect release of H2 gas [17]

Diels-AlderMolecular mechanism –

rearrangement of doublebonds

[17]

IsomerizationRing opening of cycloalkanes [12]

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

440

3.3. CBS-QB3 computation of rate coefficients of dominant reactions

For dominant reactions (R2-R13 of Table 4), a CBS-QB3 method[40] was used to compute rate coefficients. The study was performedwith the Gaussian 09 revision D suite of programs [41] as implementedon the high-performance supercomputing facility at Ghent University.CBS-QB3 calculations give geometries, external moments of inertia,harmonic oscillator frequencies and the electronic energy at 0 K. Thefirst three properties are used to calculate entropies, heat capacities andthermal contributions to the enthalpy using statistical mechanics cal-culations. Except for internal rotations, which were treated separately,all internal modes were assumed to behave as harmonic oscillators anda scaling factor of 0.99 was applied. Internal modes that resemble ro-tations around a single bond were treated separately by replacing thecontributions of the corresponding oscillators to the partition functionwith numerically calculated partition functions for these hindered ro-tors. The required hindrance potentials were obtained from scans, inwhich the dihedral angle defining the rotation was varied from 0 to360 ° in steps of 10 ° while all other molecule parameters were allowedto optimize. The obtained hindrance potential was then expressed as aFourier series. Together with the reduced moment of inertia calculatedat the I(2,3) level as defined by East and Radom [42], the hindrancepotential was used to construct the Schrödinger equation for 1-D ro-tation. The eigenvalues of the solution to this Schrödinger equationrepresent the energy levels of this mode. They were used to determinethe partition function for this mode as a function of temperature. Aftercorrections for symmetry and optical isomers, the total partition func-tion was used to calculate the thermal contribution to the enthalpy,standard entropy and temperature-dependent heat capacity data.

Enthalpies of formation in CBS-QB3 were calculated with the ato-mization method [43]. Two additional corrections accounting for spin-orbit interactions [40,43,44] and systematic bond corrections (BAC)[40] significantly improve these values as has been shown in previouswork [45]. Bond Activity Corrections are applied to correct for sys-tematic errors in the CBS-QB3 calculations. The correction values areobtained by comparison with experimental data taken mainly from theNIST Chemistry webbook [46]. However, such corrections are onlyneeded to calculate the thermodynamic properties. All transition statecalculations used uncorrected enthalpy data because BACs are notknown for transition states. All data were stored as NASA polynomials.

Transition state theory expressed in terms of Gibbs free energies wasused to calculate the rate coefficients, as shown in Eq. (1):

⎜ ⎟= ⎛⎝

⎞⎠

−−k T χ T k T

hRTp

e( ) ( )· · ·TSTB

nG

RT

Δ 1Δ ǂ

(1)

GΔ ǂ is the Gibbs free energy difference between transition state withoutthe transitional mode and reactant(s), Δn is the molecularity of thereaction (2 for bimolecular and 1 for unimolecular reactions), andχ T( )accounts for quantum mechanical tunneling. Other symbols aredefined in the list of symbols. The asymmetric Eckart potential is usedto estimate tunneling contributions χ T( ). The Gibbs free energies wereobtained from the NASA polynomials. Rate coefficients were calculatedfor the temperature range 300 K to 2500 K in steps of 50 K and theresults were regressed to a simple Arrhenius expression. The majority ofcalculated reaction rate coefficients are believed to be within a factor of3 of experimental data. Rate coefficients calculated at the CBS-QB3level of theory have been usually found to be within a factor of 3 ofexperimental data. For the elementary reaction family of hydrogenabstraction, this has been established by multiple workers in the pastusing a comprehensive set of reactions involving various possibilities ofreactive sites [47,48]. For addition reactions too, this general agree-ment holds true [35]. This general agreement also holds for reactionsinvolving oxygenates [49].

3.4. Kinetics evaluation

As stated previously, rate coefficients for the reactions were as-signed from existing group additive values and new group additivevalues derived from available literature data. Ab-initio thermodynamicparameters were assigned to the species. In total, there were 757 re-actions between 220 species in the model. The geometrical dimensionsof the reactor, its operating conditions, i.e. pressure, temperature pro-file and cyclopentane mass flow rate, are the real cyclopentane feedcomposition are fed to the plug flow reactor model of CHEMKIN-PROfor simulations. Based on initial simulations, a few reactions were foundto be dominant, and their kinetics were re-calculated in-house at CBS-QB3 level of theory. Leading to the final model. Table 4 lists the mostimportant reactions of the model and the source of their kinetics in theGenesys model. It also lists the rate coefficients of the reactions at

Fig. 1. A sample reaction recipe in Genesys.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

441

1050 K for the Genesys model as well as those for the models fromliterature − Wang [15], Sirjean [10] and Tian [9]. The Genesys me-chanism consists of 220 species taking part in 757 reactions. Out of the220 species, there are 141 radical and 79 molecular species. Among thecompeting models, Sirjean model has 215 species and 1103 reactions.Tian model has 394 species and 2282 reactions, while Wang model has552 species and 8565 reactions. In this Table, reactions R1 to R13 areimportant reactions for initial decomposition of cyclopentane andpropagation leading to the formation of major products with carbonnumbers equal to or less than 5. Reactions R14 to R22 are important forthe formation of molecular growth leading to aromatics and polycyclicaromatic hydrocarbons. It can be noted that the cyclopentane pyrolysismodels in the literature do not have reactions R15 to R22 because theseare elementary reactions while the literature models, if at all they formaromatics, do so by global lumped reactions. A few initial decomposi-tion reactions were also missing in the literature models Sirjean [10]and Tian [9], as shown shaded. From Table 4, it can also be seen thatthe rate coefficients of important reactions in Genesys are differentfrom those of literature models. At 1050 K, the discrepancy is at

maximum a factor of 20. Overall, it can be concluded that Genesysmodel has a more complete reaction list compared to the models inliterature. Also, its CBS-QB3 based rate coefficients are different inmagnitude from those found in literature models. Moreover, it is theonly model that is completely elementary reaction-based and has con-sistently ab-initio based high pressure limit rate coefficients with noalteration of kinetic or thermodynamic parameters. In the next section,experimental results and model performance would be discussed.

4. Results and discussion

4.1. Experimental data

Table S1 in supporting information lists the experimental resultswhile Fig. 5 shows the GC × GC image of the effluent of cyclopentanecracking. 34 molecules were detected and quantified in the reactor ef-fluent during experiments spanning six different temperature set pointsof the reactor between 973 K and 1073 K, at 20 K intervals. As can beseen in the supporting information, the process gas temperature profile

Fig. 2. Algorithm of the generation of a kinetic model with the use of the rule based termination criterion in Genesys.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

442

is not perfectly linear. This is due to the changes in heat flux over thelength of the reactor, the endothermic character of the pyrolysis reac-tions and the number of heating zones that are used (in this case 8).However, the measured temperature profile is fed as input to the plugflow reactor model in CHEMKIN simulations. The typical reactorpressure drop is< 0.01 bar. This has negligible effect on simulationresults. Fig. 6 shows the experimental data (including model trendsdiscussed later) of 9 main products on plots, while Fig. S1 in supportinginformation shows the trends for 22 major and minor products. FromFig. 6 and Fig. S1, it can be seen that experimental ethylene yield in-creases monotonically to 25 wt% at the highest temperature, whilepropylene yield slightly flattens at the conditions. It is to be noted thatunlike cyclopentane, for linear alkanes, the propylene yield is a lotlower [31]. For example for ethane cracking, it is about 10 wt% forsimilar operating conditions. 1,3-Cyclopentadiene product exhibits amaximum of approximately 8 wt% yield at 1053 K. Cyclopentene yieldhas a maximum of 3 wt% at 1013 K. Benzene yield increases ex-ponentially with temperature and reaches a maximum of 3 wt%, withtoluene and styrene behaving similarly and both reaching 0.8 wt%. Amaximum of 1 wt% each of indene and naphthalene are formed at thehighest temperature. A maximum of 2 wt% ethane is formed at thehighest temperature while propane yield stays below 1 wt%. Hydrogendata shows some scatter but the trend shows a monotonic increase to1.5 wt%. A maximum of 4 wt% methane is formed at the highesttemperature. These are the 14 products whose maximum experimentalyield is more than 0.5 wt%. There are other minor products with lesserthan 0.5 wt% maximum yield − like 1- and 2-butenes, acetylene, me-thyl acetylene, allene (propadiene/PD), 2-pentene, 1,4-pentadiene and1,3-butadiene. Carbon balance was closed within 1% across all ex-periments, with H/C mole ratio at exit calculated for each temperature.Based on that, the mass balance is excellent for the lower temperatures,while there is some deviation at the high temperatures. At the highest

temperature of 1073 K, the inlet H/C mole ratio is 2, while the outletratio is 2.18.

4.2. Kinetic model evaluation

Reactor simulations using the Genesys model were conducted withthe CHEMKIN PRO package using the plug flow reactor module [39].Residence time is considered to be the time spent by the cyclopentanemolecule in the isothermal region. This isothermal region has a lengthof 100 cm from 20 cm to 120 cm, accounting for a reactor volume of28.3 cm3. At 1073 K (the highest temperature experiment), this implies0.00053853 mol cyclopentane in the isothermal region at any timeinstant. As the feed flow is 0.066667 g s−1, the residence time is cal-culated to be 0.57 s. At the lowest temperature experiment (973 K), thisis 0.62 s. The formula to calculate residence time is given below in Eq.(2):

∫=⎡⎣ ⎤⎦θ

f

PVRT

(2)

P is the total pressure (1.7 bara), V is the volume in the isothermalregion (28.3 cm3), R is the ideal gas constant, T is the isothermaltemperature and f is the cyclopentane mole flow rate at the inlet of thereactor. In Fig. 6, cyclopentane conversion is plotted against reactortemperature. For all other plots, product yield is plotted against cy-clopentane conversion. The reason for plotting product yield againstcyclopentane conversion is to be able to more easily compare the dif-ferent kinetic models at the same feed conversions. From Fig. 6, it canbe seen that there is quantitative and qualitative agreement betweenthe model and experiments for all plots. The major disagreements areonly with some minor products (in Fig. S1) where the maximum ex-perimental yield is less than 0.5 wt%. A reaction path analysis wasperformed on major products at the experimental reactor temperatureset point of 1053 K and a schematic showing the dominant reactions ofthe Genesys mechanism is given in Fig. 7. Rate of production andsensitivity analyses reveal the following dominant pathways:

Cyclopentane: Cyclopentane is consumed at the initial temperatureramp majorly by the isomerization reaction to 1-pentene. However,across a substantial part of the reactor, the predominant consumptionroute is through the hydrogen abstraction by allyl radicals followed bythe abstraction by hydrogen atoms. The allyl radical concentration isvery high along the reactor length for cyclopentane pyrolysis. Thereason for this is − once the isomerization to 1-pentene takes place, 1-pentene undergoes a homolytic CeC scission to give ethyl radicals andallyl radicals. These radicals abstract hydrogen from cyclopentane togive cyclopentyl radical. Cyclopentyl in turn undergoes ring opening togive pent-1-en-5-yl radical. The latter primarily decomposes via acarbon centered beta-scission to give ethylene and a new allyl radical.Since the resonance stabilized allyl radical is produced immediately inthe primary decomposition routes of cyclopentane, its concentration ishigh and hence, the allyl-assisted hydrogen abstraction is the pre-dominant conversion route of cyclopentane. Hydrogen atoms are gen-erated from the CeH beta scission of the cyclopentyl ring giving cy-clopentene. This is also released from the ethyl radical produced from1-pentene decomposition, giving ethylene, hence hydrogen-atom-as-sisted abstraction is one of the most important consumers of cyclo-pentane. The rate of consumption is shown in Fig. 8. A sensitivityanalysis is added to the supporting information.

Ethylene: Ethylene is majorly produced by 2 routes: carbon centeredbeta scission of pent-1-en-5-yl to give ethylene and allyl radical, and of1-propyl radical to give ethylene and methyl radical. 1-propyl radical isformed by hydrogen atom addition to propylene. Hence, propylene canbe converted to ethylene in this way. As expected, ethylene yield in-creases monotonically with cyclopentane conversion (Fig. 6).

Propylene: Across the entire reactor length, propylene is mainlyformed by the allyl-assisted hydrogen abstraction of cyclopentane.

Table 2Group Additive Values for Intra-CPD H-shift reaction family ( ).

REFERENCE REACTION (A, n, Ea) = 4.8 × 1013, 0,107.8 kJ, mol, s

Reaction Group name ΔGAV°[log(Ã)] ΔGAV°[Ea]

C1-C 0.317 0.92

C1-Cd 0.053 −12.59

C5-C 1.070 14.84

C5-Cd 0.415 11.97

C3-C (or C3-Cd)

0.618 8.45

C1-C. −0.728 −38.79

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

443

However, at high concentrations of propylene, it is decomposed byhydrogen atom addition and subsequent conversion to ethylene andmethyl radical. This is the reason why the propylene trend flattens outtoward high conversions (Fig. 6).

Cyclopentene: Cyclopentene is formed mainly by the CeH betascission of the cyclopentyl radical. The hydrogen atom hence releasedassists in the further conversion of cyclopentane through hydrogenabstraction. So, we might expect a high conversion of cyclopentane tocorrespond to a high monotonically increasing yield of cyclopentene.However, it can be seen from Fig. 6 that the cyclopentene yield goesthrough a maximum. This is because at high cyclopentene concentra-tions, there is a molecular mechanism (as opposed to radical me-chanism) leading to release of hydrogen gas and 1,3-cyclopentadieneformation, as will be discussed next.

1,3-Cyclopentadiene: Cyclopentene which is formed majorly by theCeH beta-scission of the cyclopentyl radical preferably undergoes amolecular mechanism to form 1,3-cyclopentadiene and hydrogen gas.This is the dominant route of 1,3-cyclopentadiene formation. A minorroute from cyclopentene is through the hydrogen abstraction at the

allylic position and eventual CeH beta scission to give 1,3-cyclo-pentadiene. Hence, 1,3-cyclopentadiene is formed at the expense ofcyclopentene, giving rise to the maximum feature seen in cyclopentene.However, even 1,3-cyclopentadiene tends to flatten out at the highconversions of cyclopentane. This is because it is the main precursor forthe formation of polycyclic aromatic hydrocarbons.

Methane: Methane is formed mainly by the hydrogen abstraction bymethyl radicals on cyclopentane. Methyl radicals are in turn generatedpredominantly by the CeC beta scission of 1-propyl radical which isformed by hydrogen addition to propylene. Hence, there is a trade-offbetween propylene and methane yields. Hence, it comes as no surprisethat while propylene is a bit over-predicted by the model, methane is abit under-predicted.

Hydrogen: At the initial part of the reactor, hydrogen is mainlygenerated by 2 routes: hydrogen-atom-assisted hydrogen abstraction ofcyclopentane and H2 gas release from cyclopentene to give 1,3-cyclo-pentadiene. As the 1,3-cyclopentadiene yield increases at higher con-versions, hydrogen is also abstracted from it by hydrogen atoms, re-leasing H2 gas. This third route gets activated at high concentrations of

Table 3Group Additive Values for Intra-molecular carbon centered radical addition ( ).

Reaction Group Name ΔGAV°[log(Ã)] ΔGAV°[Ea] Nearest neighbors

C4,exo REFERENCE (1.35 × 1012, 0, 45.5), kJ, mol, s

−0.04 −25.7 (C2-Cd)

0.88 112.4 (C2-Cd) (C1=allyl)

0.88 112.4 (C2-Cd) (C1=allyl)

0.23 −4.4 (C2-Cd)(C1=CPDyl)

C5,endo REFERENCE (3.09 × 1010, 0, 69.5), kJ, mol, s

0.00 0.0 C5,endo

0.00 0.0 C5,endo

0.00 0.0 C5,endo

0.00 0.0 (C1=allyl)

0.00 0.0 (C1=allyl)

0.25 −10.2 (C1-C)

0.25 −10.2 (C2-C)

0.25 −10.2 (C2-C) (C1=allyl)

2.51 90.6 (C2-Cd) (C1=allyl)

2.51 90.6 (C2-Cd) (C1=allyl)

C5,exo REFERENCE (1.79 × 1011, 0, 67.7), kJ, mol, s

0.44 −8.3 (C2-Cd)

C6,endo REFERENCE (5.85 × 109, 0, 35), kJ, mol, s

0.00 0.0 (C1=allyl)

0.00 0.0 (C1-C) (C1=allyl)

2.13 55.7 (C2-C) (C1=CPDyl)

C6,exo REFERENCE (2.06 × 1010, 0, 26.2), kJ, mol, s

2.00 6.4 (C2=benzene) (C1d)

C7,exo REFERENCE (5.55 × 1010, 0, 59.3), kJ, mol, s

1.17 26.5 (C1=allyl)

1.17 26.5 (C1=allyl)

1.17 26.5 (C1=allyl)

1.17 26.5 (C2-C) (C1=allyl)

2.72 −10.9 (C2-Cd) (C1d)

2.58 105.5 (C2-Cd) (C3-C)

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

444

1,3-cyclopentadiene typically towards the end of the reactor.Ethane: Ethane trend in Fig. 6 shows a monotonic increase with

respect to cyclopentane conversion. This is because the main routeforming ethane is the hydrogen abstraction of cyclopentane by ethylradical. The ethyl radical is in turn generated by the homolytic CeCscission of 1-pentene to give ethyl and allyl radical. At the start of thereactor where there is excess ethyl produced, on one hand it attackscyclopentane assisting in its conversion, while on the other hand, itundergoes CeH beta scission to give ethylene.

Propane: Propane is produced predominantly across the reactor bythe hydrogen abstraction of cyclopentane by 2-propyl radical. 2-propylradical is in turn formed by the hydrogen atom addition to propylene.Hydrogen atom addition to propylene can yield 1-propyl as well as 2-propyl. As discussed earlier, 1-propyl leads ultimately to ethylene andmethane. However, 2-propyl preferably leads to propane because itcannot undergo a CeC beta scission like 1-propyl due to the positioningof the radical center and a 1–2 radical shift is not energetically favored.Hence, there is a trade-off between propylene and propane.

Benzene: Hydrogen abstraction on 1,3-cyclopentadiene leads to thehighly stabilized resonance radical 1,3-cyclopentadien-5-yl. This ra-dical recombines with ethyl radical to give ethyl cyclopentadiene. Thismolecule undergoes intra-cyclopentadiene hydrogen shifts to give adifferent location of the ethyl substituent. The ethyl substituent’s CeC

bond undergoes homolytic CeC scission to give CH2-cyclopentadienylradical, radical center being at the primary carbon atom, This againundergoes intra-cyclopentadiene hydrogen shifts and then intra-mole-cular CeC addition (ring formation) to give a six-membered bicyclicradical. This radical undergoes a CeC beta scission to expand the ringfrom a 5-membered one to a six membered one: cyclohexadienyl ra-dical. This radical undergoes CeH beta scission to give benzene andhydrogen atom.

Indene and Naphthalene: Indene and naphthalene formation goesby a series of elementary steps starting from 1,3-cyclopentadiene andthe resonance stabilized cyclopentadienyl radical.

The above discussion shows the dominant paths at low as well ashigh temperatures and address the yield patterns seen such as maximaand plateau wherever applicable. The cyclopentadiene yield shows aplateau at higher temperatures and does not show a monotonic risebecause it is the precursor to the formation of aromatics. It will leadindeed to the build-up of naphthalene and indene, the formation ofwhich involves cyclopentadiene and cyclopentadienyl radical. The cy-clopentene yield indeed undergoes a maximum. At lower temperatures,the CeH βscission of a cyclopentyl radical leads to an increase in yieldof cyclopentene, however at high enough cyclopentene concentrations,temperatures higher than 1023 K, the dehydrogenation reaction is fa-vored directly giving cyclopentadiene. Hence, the maximum feature

Fig. 3. Group nomenclature for intra-cyclopentadiene hydrogen shift family.

Fig. 4. Atom labels for endocyclic and exocyclicintra-molecular carbon centered radical addition.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

445

seen in cyclopentene yield. According to Fig. 8, the most importantreactions for cyclopentane consumption are: ring-opening isomeriza-tion to 1-pentene, hydrogen abstraction by allyl radical and hydrogenabstraction by hydrogen atom. At 993 K (low temperature experiment),near the start of the reactor (10 cm), isomerization reactions dominatethe cyclopentane consumption. However, at the same temperature, asthe reaction proceeds, the abstraction reactions take over and near theend of the reactor (100 cm), hydrogen abstraction by allyl radical is themost dominant reaction followed by hydrogen abstraction by hydrogenatoms, the relative contribution of these two for cyclopentane

consumption is a factor of 3 different, with the contribution of otherreactions being negligible. This relative ratio of allyl/H abstractioncontribution remains the same at higher temperatures (1053 K), al-though the contribution of methyl and ethyl assisted abstractions be-come non-negligible.

A sensitivity analysis is performed with respect to the cyclopentanemole fraction at 993 K and 1053 K near the end of the reactor (100 cm).The results and a detailed discussion are added to the supporting in-formation. Sensitive reactions are hydrogen abstraction reactions fromthe fuel molecule and the ring-opening isomerization reaction. Other

Table 4Most important reactions in the Genesys cyclopentane model – comparison to literature models [9,10,15].

Fig. 5. GC × GC image of effluent of cyclopentane pyrolysis at 993 K (at 240 g.h−1 hydrocarbon feed, 20 g.h−1 N2 internal standard, 0.17 MPa) and 1053 K (at 240 g.h−1 hydrocarbonfeed, 70 g.h−1 N2 internal standard, 0.17 MPa).

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

446

sensitive reactions that do not contain the fuel molecule are the com-petitive β-scission/ring opening of the cyclopentyl radical and the CeHβ-scission with the formation of cyclopentene. At higher temperatures,latter reaction has a positive sensitivity coefficient because of the highcyclopentene mass fraction in the reactor. Also the bond scission re-action of 1-pentene, with the formation of ethyl and allyl radicals is asensitive reaction.

The 1.6 wt% impurity in cyclopentane does not affect the yield patternsin any major way, as shown in the supporting information. In summary itcan be stated that the Genesys model predictions are better than that of theother models for most products. The Wang model [15] is reasonably goodfor non-polycyclic aromatic hydrocarbon products. As the intention of theother models was not to match pyrolysis data of cyclopentane, it is strictlyspeaking not a drawback for them, and they may be applicable to their caseof interest. It is to be noted that the available models are developed forcombustion and hence, using them at pyrolysis conditions, is obviously outof their comfort zone and out of the range where they are developed for.Nevertheless, overall they do a reasonable job. However, as an elementaryreaction pyrolysis model based consistently on ab-initio derived kinetics, theGenesys developed model is a first.

5. Conclusions

The pyrolysis of cyclopentane has been studied for the first time inthe temperature range of 973 K to 1073 K, in a continuous flow reactor.34 products were identified and quantified using 2 dimensional gaschromatography, ranging from hydrogen to polycyclic aromatic hy-drocarbons. A kinetic model has been generated consisting of 757 re-actions between 220 species based on mainly CBS-QB3 kinetics andthermodynamics. This compact model predicts the trends of all majorproducts and minor products well, significantly better than the cur-rently established models. This is attributed to the use of accurate ab-initio derived kinetics and a more complete kinetic model consistingonly of elementary steps. Reaction path analysis reveals the dominantpathways to the most important products such as ethylene, propylene,methane, hydrogen, ethane, cyclopentene, 1,3-cyclopentadiene, ben-zene, indene and naphthalene. Cyclopentane pyrolysis initiation pri-marily starts via an isomerization pathway to 1-pentene. However, asthe reaction proceeds, radicals are generated by subsequent decom-position which then abstract hydrogen from cyclopentane. This hy-drogen abstraction eventually becomes the most dominant conversion

Fig. 6. Experimental pyrolysis yields and model predictions for 9 main products in continuous flow experiments at 0.5 s residence time, 0.17 MPa pressure, pure cyclopentane feed. (Experiment, Genesys, Wang, Sirjean, Tian).

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

447

route. Between 973 K and 1073 K (the temperature range of experi-ments), the global decomposition of cyclopentane has a pre-exponentialfactor of 1.28 × 108 (/s) and an activation energy of 163438 J/mol.

Acknowledgements

The authors acknowledge the financial support from the Long TermStructural Methusalem Funding by the Flemish Government, theEuropean Research Council under the European Union’s SeventhFramework Programme (FP7/2007-2013)/ERC grant agreement n°

290793, the Research Board of Ghent University (BOF), the Fund forScientific Research Flanders (FWO), the SBO proposal “ARBOREF”supported by the Institute for promotion of Innovation through Scienceand Technology in Flanders (IWT) and the SBO proposal “Bioleum”supported by the Institute for promotion of Innovation through Scienceand Technology in Flanders (IWT). Computational support was pro-vided by the STEVIN Supercomputer Infrastructure at Ghent University,funded by Ghent University, the Flemish Supercomputer Center (VSC),the Hercules Foundation and the Flemish Government – departmentEWI.

Fig. 7. Schematic of the dominant reaction pathwaysin cyclopentane pyrolysis based on Rate ofProduction Analysis at 1053 K, 100 cm reactorlength.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

448

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in theonline version, at http://dx.doi.org/10.1016/j.jaap.2017.08.005.

References

[1] J. Speight, Handbook of Industrial Hydrocarbon Processes, Gulf ProfessionalPublishing, 2010.

[2] K.M. Van Geem, M.F. Reyniers, G.B. Marin, J. Song, W.H. Green, D.M. Matheu,AIChE J. 52 (2006) 718.

[3] Z. Wang, Z. Cheng, W. Yuan, J. Cai, L. Zhang, F. Zhang, F. Qi, J. Wang, Combust.Flame 159 (2012) 2243.

[4] M.K.K., B.R. Nautiyal, R.K. Joshi, B.S. Rawat, S.M. Nanoti, G. Prasad, D. Paul, M.O.Garg, Separation of cyclopentane from light naphtha fraction, 2010, at: Petrotech,New Delhi, January 9–12.

[5] Z. Serinyel, O. Herbinet, O. Frottier, P. Dirrenberger, V. Warth, P.A. Glaude,F. Battin-Leclerc, Combust. Flame 160 (2013) 2319.

[6] I.G. Zsely, T. Varga, T. Nagy, M. Cserhati, T. Turanyi, S. Peukert, M. Braun-Unkhoff,C. Naumann, U. Riedel, Energy 43 (2012) 85.

[7] W. Hui, H.F. Yang, X.Q. Ran, Q.Z. Shi, Z.Y. Wen, J. Mol. Struct.-Theochem. 678(2004) 39.

[8] H. Wang, H.F. Yang, X.Q. Ran, Q.Z. Shi, Z.Y. Wen, J. Mol. Struct.-Theochem. 581(2002) 187.

[9] Z.M. Tian, C.L. Tang, Y.J. Zhang, J.X. Zhang, Z.H. Huang, Energy Fuels 29 (2015)428.

[10] B. Sirjean, F. Buda, H. Hakka, P.A. Glaude, R. Fournet, V. Warth, F. Battin-Leclerc,M. Ruiz-Lopez, Proc. Combust. Inst. 31 (2007) 277.

[11] W. Tsang, Int. J. Chem. Kinet. 10 (1978) 599.[12] B. Sirjean, P.A. Glaude, M.F. Ruiz-Lopez, R. Fournet, J. Phys. Chemi. A 110 (2006)

12693.[13] C.J. Annesley, C.F. Goldsmith, R.S. Tranter, Phys. Chem. Chem. Phys. 16 (2014)

7241.[14] J.B. Randazzo, C.J. Annesley, K. Bell, R.S. Tranter, Proc. Combust. Inst. 36 (2017)

273.[15] K. Wang, S.M. V, A.M. Dean, Combust. Flame 162 (2015) 4456.[16] M.R. Djokic, K.M. Van Geem, C. Cavallotti, A. Frassoldati, E. Ranzi, G.B. Marin,

Combust. Flame 161 (2014) 2739.[17] S. Merchant, Molecules to Engines: Combustion Chemistry of Alcohols and Their

Application to Advanced Engines, MIT, 2015.[18] V. Warth, N. Stef, P.A. Glaude, F. Battin-Leclerc, G. Scacchi, G.M. Come, Combust.

Flame 114 (1998) 81.[19] F. Buda, R. Bounaceur, V. Warth, P. Glaude, R. Fournet, F. Battin-Leclerc, Combust.

Flame 142 (2005) 170.[20] S. Touchard, F. Buda, G. Dayma, P.A. Glaude, R. Fournet, F. Battin-Leclerc, Int. J.

Chem. Kinet. 37 (2005) 451.[21] M. Yahyaoui, N. Djebaili-Chaumeix, C.E. Paillard, S. Touchard, R. Fournet,

P.A. Glaude, F. Battin-Leclerc, Proc. Combust. Inst. 30 (2005) 1137.[22] E.D. H. Wang, B. Sirjean, D. A. Sheen, R. Tango, A. Violi, J. Y. W. Lai, F. N.

Egolfopoulos, D. F. Davidson, R. K. Hanson, C. T. Bowman, C. K. Law, W. Tsang, N.P. Cernansky, D. L. Miller, R. P. Lindstedt, in, http://web.stanford.edu/group/haiwanglab/JetSurF/JetSurF2.0/index.html, 2010, p. A high.

[23] N.M. Vandewiele, K.M. Van Geem, M.F. Reyniers, G.B. Marin, Chem. Eng. J. 207(2012) 526.

[24] S.P. Pyl, K.M. Van Geem, P. Puimege, M.K. Sabbe, M.F. Reyniers, G.B. Marin,Energy 43 (2012) 146.

[25] C.M.S. Steven, P. Pyl, M. Kevin Van Geem, Marie-Françoise Reyniers,Joeri Vercammen, Jan Beens Guy, B. Marin, J. Chrom A. 1218 (2011) 3217.

[26] S.P. Pyl, C.M. Schietekat, M.F. Reyniers, R. Abhari, G.B. Marin, K.M. Van Geem,Chem. Eng. J. 176 (2011) 178.

[27] I.D.K.M. Van Geem, S. Prokopiev, M.F. Reyniers, D. Viennet, G.B. Marin, Ind. Eng.Chem. Res. 48 (2009) 10343.

[28] J. Dalluge, J. Beens, U.A.T. Brinkman, J. Chromatogr. A 1000 (2003) 69.[29] Z.Y. Liu, J.B. Phillips, J. Chromatogr. Sci. 29 (1991) 227.[30] J.B. Phillips, J. Beens, J. Chromatogr. A 856 (1999) 331.[31] M.K. Sabbe, K.M. Van Geem, M.F. Reyniers, G.B. Marin, AIChE J. 57 (2011) 482.[32] N. Vandewiele, Kinetic Model Construction Using Chemoinformatics, Universiteit

Gent, 2014 LCT, UGent.[33] R.V. d. Vijver, Automatic Ab-initio Calculations for Kinetic Model Generation of Gas

Phase Processes, Universiteit Gent, 2017 LCT, Universiteit Gent.[34] Suleimanov, Green, J. Chem. Theory Comput. 11 (2015) 4248.[35] M.K. Sabbe, M.F. Reyniers, V. Van Speybroeck, M. Waroquier, G.B. Marin,

Chemphyschem 9 (2008) 124.[36] M.K. Sabbe, M.F. Reyniers, M. Waroquier, G.B. Marin, Chemphyschem 11 (2010)

195.[37] M.K. Sabbe, A.G. Vandeputte, M.F. Reyniers, M. Waroquier, G.B. Marin, Phys.

Chem. Chem. Phys. 12 (2010) 1278.[38] R. Van de Vijver, N.M. Vandewiele, A.G. Vandeputte, K.M. Van Geem,

M.F. Reyniers, W.H. Green, G.B. Marin, Chem. Eng. J. 278 (2015) 385.[39] ChemKin, Reaction Design, San Diego, 2013.[40] G.A. Petersson, D.K. Malick, W.G. Wilson, J.W. Ochterski, J.A. Montgomery,

M.J. Frisch, J. Chem. Phys. 109 (1998) 10570.[41] G.W.T. M. J. Frisch, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G.

Scalmani, V. Barone, B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li,H. P. Hratchian, A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M.Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O.Kitao, H. Nakai, T. Vreven, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro, M.Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Kobayashi, J.

Fig. 8. Rate of consumption of cyclopentane by various elementary reactions, % at 993 K at 10 and 100 cm reactor length and 1053 K at 100 cm reactor length.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

449

Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M.Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B. Cross, V. Bakken, C. Adamo,J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C.Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma, V. G. Zakrzewski, G. A. Voth,P. Salvador, J. J. Dannenberg, S. Dapprich, A. D. Daniels, Ö. Farkas, J. B. Foresman,J. V. Ortiz, J. Cioslowski, and D. J. Fox, Gaussian, Wallingford CT, 2009.

[42] A.L.L. East, L. Radom, J. Chem. Phys. 106 (1997) 6655.[43] L.A. Curtiss, K. Raghavachari, P.C. Redfern, J.A. Pople, J. Chem. Phys. 106 (1997)

1063.

[44] L.A. Curtiss, K. Raghavachari, P.C. Redfern, V. Rassolov, J.A. Pople, J. Chem. Phys.109 (1998) 7764.

[45] M.K. Sabbe, M. Saeys, M.F. Reyniers, G.B. Marin, V. Van Speybroeck, M. Waroquier,J. Phys. Chem. A 109 (2005) 7466.

[46] NIST chemistry webbook, webbook.nist.gov/chemistry.[47] H.H. Carstensen, A.M. Dean, J. Phys. Chem. A 113 (2009) 367.[48] H.H. Carstensen, C.V. Naik, A.M. Dean, J. Phys. Chem. A 109 (2005) 2264.[49] H.H. Carstensen, A.M. Dean, Int. J. Chem. Kinet. 44 (2012) 75.

M.V. Khandavilli et al. Journal of Analytical and Applied Pyrolysis 128 (2017) 437–450

450