Cracking of plastic pyrolysis oil over FCC equilibrium catalysts to produce fuels: Kinetic modeling

Fuel 316 (2022) 123341 Available online 1 February 2022 0016-2361/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/bync/4.0/). Full Length Article Cracking of plastic pyrolysis oil over FCC equilibrium catalysts to produce fuels: Kinetic modeling Roberto Palos * , Elena Rodríguez, Alazne Guti´ errez, Javier Bilbao, Jos´ e M. Arandes Department of Chemical Engineering, University of the Basque Country UPV/EHU, PO Box 644, 48080 Bilbao, Spain ARTICLE INFO Keywords: Catalytic cracking Fuels Plastic pyrolysis oil Kinetic model Deactivation Waste refinery ABSTRACT The kinetics of the catalytic cracking of plastic pyrolysis oil (PPO) over three FCC (fluid catalytic cracking) equilibrium commercial catalysts has been modeled. The PPO comes from the fast pyrolysis of high-density polyethylene (HDPE). The cracking runs have been carried out in a laboratory-scale reactor under FCC conditions: 500–560 ◦C; catalyst/oil weight ratio of 5 g cat g PPO-1 ; and contact time of 1.5–6 s. Four different reaction schemes composed of six lumps have been compared and it has been obtained by statistical means that the simplest one is the most appropriate for describing the process. The differences in the kinetic parameters have been related to the properties of the catalysts. Among them, total acidity and mesoporous structure have a key role. The former for promoting the cracking reactions and the latter for limiting the diffusional restrictions of both the bulky compounds within the PPO and the formed coke precursors. This way, ECAT-3 that is the most acid and most mesoporous catalyst, maximizes the yields of naphtha (33.6 wt%) and liquefied petroleum gases (LPG) (18.9 wt%). In contrast, ECAT-1 and ECAT-2 should be chosen for producing light cycle oil (LCO). For ECAT-3, the apparent activation energies of the conversion of heavy cycle oil (HCO) into light cycle oil (LCO), LCO into naphtha, and LCO into LPG are 60.5 42.5 and 58.3 kJ mol -1 , respectively. In addition, those of the formation of coke from HCO, LPG and dry gas are 129.0, 4.4 and 40.7 kJ mol -1 , respectively. 1. Introduction The development and wellness of the humankind implies an increase of global pollution. One of the consequences is the increasing presence of waste plastics in the municipal solid wastes (MSW), which overflows the management capability of both public and private entities. Consequently, an unacceptable amount of these wastes ends up landfilled, causing the contamination of the soils and the aquifers [1]. In order to solve these problems, it is well established the interest on tertiary recycling by means of thermochemical processes, i.e. pyrolysis and gasification [2]. The fast pyrolysis of plastics is performed at low temperature, using high heating rates and short residence time for the volatiles. Moreover, it can be carried out in simple and versatile units equipped with different types of reactors (rotary kilns, screw reactors, fluidized or spouted beds, etc.) entailing a reduced environmental impact and with the possibility of tuning the operating conditions to adapt the production to the type of plastic fed [3,4]. The ideal goal of pyrolysis processes is the monomer recovery, which can be done with high yields in the pyrolysis of polystyrene [5] and polymethyl-methacrylate [6]. On the other hand, for polyolefinic plastics, which constitute two thirds of the plastic fraction found in the MSW [7], is of great interest the production of plastic pyrolysis oil (PPO) because of its possibilities to be used as an alternative fuel [8]. Based on its properties, the PPO has been considered as a potential fuel for diesel engines feeding it neatly or blended with commercial diesel [9]. Nevertheless, the PPO does not meet the tough requirements of commercial fuels and requires of physicochemical treatments to adapt its composition [10]. This situation has led to the proposal of integrating the fast pyrolysis of waste plastics with the upgrading of the PPO in refinery units (Waste-Refinery) [11]. The interest of the proposal lays on the capacity of refinery units for valorizing the PPO, either in ad hoc catalytic units or in already existing industrial units. The fluid catalytic cracking (FCC) units are the most appropriate ones in the short term, given their high capacity and versatility to manage unconventional feeds, such as diverse secondary refinery streams [12,13] or bio-oil [14]. Indeed, the chemical composition of the PPO (highly olefinic and free of aromatics) makes it appropriate to be fed to catalytic cracking units with the aim of producing fuels free of sulfur and nitrogen [15]. Furthermore, within the facilities available in the refineries, there are the required fractionation and conditioning units to obtain fuels similar to * Corresponding author. E-mail address: [email protected] (R. Palos). Contents lists available at ScienceDirect Fuel journal homepage: www.elsevier.com/locate/fuel https://doi.org/10.1016/j.fuel.2022.123341 Received 23 September 2021; Received in revised form 10 January 2022; Accepted 16 January 2022 Fuel 316 (2022) 123341 2 conventional ones. Among the advantages that the Waste–Refinery strategy offers, the following ones must be highlighted: (i) the recycling of petroleum-derived products with the subsequent savings of raw materials; (ii) the removal of economic barriers that entails the design and construction of new units, which correspond to the high cost of the equipment and of the marketing of non-conventional fuels that would compete against the conventional fuels; and, (iii) the rational organization of the plastics recycling, carrying out the pyrolysis process in a delocalized way in units located nearby of the waste plastics collection and segregation points. The PPO would be afterwards transported from different geographical areas to centralized refineries for its large-scale valorization. Feeding a liquid stream, such as the PPO, into a cracking unit, entails less technical difficulties that the feeding of pure polyolefins, the cracking of which has been also studied [16–18]. Nonetheless, these initiatives will require a rigorous control of the feeds, since their composition can be easily contaminated by the presence of different plastics and of additives and pollutants in the waste plastics. In a previous work it has been studied the effect of the properties of different FCC equilibrium catalysts on the production of fuel from PPO, operating at 500–560 ◦C and using a riser simulator reactor [19]. Interestingly, the yields of naphtha (highly olefinic and with a high octane rating) and light olefins were superior to 40 and 12 wt%, respectively. The proposed initiative is similar to that of cracking wax from Fischer-Tropsch process with the aim of producing high octane gasoline and light olefins [20,21]. Both for the simulation and optimization of an ad hoc designed reactor and for the feeding of the PPO to an industrial FCC unit, it is required a kinetic model capable of quantifying the products distribution. Traditionally, the efforts in the kinetic modeling have been focused on the cracking of vacuum gasoil (VGO). Moustafa and Froment [22] were pioneers in taking into account the heterogeneous composition of the VGO and they proposed a kinetic model with a complex reaction scheme that described the individual reactions involved and the formation of coke by means of elementary steps. This type of molecularlevel kinetic models has been also applied for the cracking of wax from Fischer-Tropsch process [21]. Nevertheless, most of the works have established lump-based kinetic models that simplify the computing and their posterior use in the design of the reactor [23–25]. Apart from the complexity of the reaction scheme, an additional difficulty for obtaining kinetic models is the extremely fast deactivation of the catalyst caused by coke deposition [26,27]. Kinetic models for the catalytic cracking of VGO consider between 3 and 17 lumps and have been collected by different authors [28,29]. These models assume that kinetic parameters are apparent values as a consequence of the diffusional restrictions caused by the components of the VGO (especially the heavier ones) [30]. In this work, it has been established a six-lump based kinetic model for the cracking of PPO obtained in the pyrolysis of high-density polyethylene from the experimental data obtained in a previous work [19]. The aim of the work is to provide a tool for quantifying the effects of the operating conditions on the yields of products of interest, such as fuels (gasoline and diesel) and commodities (light olefins). In the modeling, it has been taking into account the catalyst deactivation by coke deposition, which is extremely significant in cracking reactions. Moreover, the analysis of the kinetic parameters obtained for three different FCC equilibrium catalysts with different acidity and porous structure allows for assessing the effect of these properties on the different catalytic steps and on the formation of coke. 2. Material and methods 2.1. Materials The plastic pyrolysis oil (PPO) has been obtained at 500 ◦C under fast pyrolysis conditions by feeding virgin high-density polyethylene (HDPE) to a fountain confined conical spouted bed reactor [31]. Three different commercial equilibrium FCC catalysts (ECAT-1, ECAT-2 and ECAT-3) have been used in the work. The catalysts have been collected from the catalyst purge stream of industrial FCC units, specifically ECAT-1 from Petronor Refinery (Spain) and the other catalysts from Petrobras Refinery (Brazil). Consequently, they are equilibrium catalysts, since they have been submitted to numerous cycles composed of reaction, stripping and regeneration steps in their corresponding FCC units [32]. Nomenclature Abbreviations C/O catalyst to oil weight ratio FBP final boiling point FCC fluid catalytic cracking HCO heavy cycle oil IBP initial boiling point LCO light cycle oil LPG liquefied petroleum gases MSW municipal solid waste PPO plastic pyrolysis oil SSE mean squared error VGO vacuum gasoil Symbols A correlation coefficient C molar concentration (mol cm -3 ) E apparent activation energy (kJ mol -1 ) i certain lump j certain reaction k apparent kinetic parameter (m 6 kg cat-1 kmol -1 s -1 / m 3 kg cat1 s -1 ) k d catalyst deactivation kinetic parameter (s -1 ) m reaction order M HCO molecular weight of HCO lump (g mol -1 ) m cat mass of catalyst (g) m PPO mass of PPO (g) n e number of experiments n l number of lumps n p number of parameters p certain experiment r reaction rate R ideal gas constant (8.314 J mol -1 K -1 ) t contact time (s) T reaction temperature (◦C) T* reference temperature (◦C) V volume of the reactor (cm 3 ) y weight fraction Greek symbols α level of significance φ activity term υ stoichiometric coefficient ν degrees of freedom R. Palos et al. Fuel 316 (2022) 123341 3 2.2. Experimental setup The catalytic cracking runs have been performed on a laboratory scale micro-riser reactor, specifically designed to mimic the conditions of the riser reactor of industrial FCC units [33]. A schematic representation of the experimental unit together with an explanation of the experimental procedure can be found elsewhere [34]. The operating conditions of riser simulator reactor have been: temperature, 500, 530 and 560 ◦C; catalyst to oil weight ratio (C/O), 3–7 g cat g PPO-1 ; and contact time, 1.5–6 s. 3. Kinetic model 3.1. Model description The catalytic cracking of the PPO has been described by means of a six-lump reaction network. The six lumps are heavy cycle oil (HCO, C 20+ ), light cycle oil (LCO, C 13 -C 20 ), naphtha (C 5 -C 12 ), liquefied petroleum gases (LPG, C 3 -C 4 ), dry gas (C 1 -C 2 ) and coke (carbonaceous material deposited on the catalyst). The reaction network in Fig. 1a corresponds to parent reaction network and it accounts for sixteen kinetic parameters, to which it must be added a parameter for catalyst deactivation. 3.2. Model equations and methodology The kinetic modeling methodology used is based on the one developed by Toch et al. [35] for catalytic processes with complex pathways and on that proposed by Cordero-Lanzac et al. [36,37], since they included the catalyst deactivation on it. Furthermore, the methodology has been adapted for handling the experimental data obtained in a batch reactor. Likewise, a molar balance to the micro riser reactor has been also required in order to properly describe the behavior of the different lumps [38]. According to the reaction network in Fig. 1a, the reaction rate equations that describe the evolution with contact time of the different lumps are listed below. dyHCO dt = − φ[( mPPO MHCO V)(k1+k2+k3+k4+k5)y2 HCO ]mcat V(1) dyLCO dt = φ[( mPPO MHCO V)(k1y2 HCO)− (k6+k7+k8+k9)yLCO ]mcat V(2) dyNaphtha dt = φ[( mPPO MHCO V)(k2y2 HCO)+(k6yLCO)− (k10 +k11 +k12)yNaphtha ]mcat V (3) dyLPG dt = φ[( mPPO MHCO V)(k3y2 HCO)+(k7yLCO)+(k10 yNaphtha)− (k13 +k14)yLPG ]mcat V (4) Fig. 1. Proposed reaction networks. R. Palos et al. Fuel 316 (2022) 123341 4 being y i the weight fraction of lump i, t the contact time between the reactants and the catalyst in the reactor, M HCO the molecular weight of the lump HCO, V the volume of the reactor, k j the apparent rate constant of reaction j, m PPO the mass of PPO fed and m cat the mass of catalyst used. One should observe that the catalytic cracking of the different lumps has been described using irreversible first-order reactions, with the exception of the cracking of HCO lump, which has been considered as an irreversible second-order reaction [28,39]. Additionally, it has been assumed that cracking reactions are non-selectively affected by catalyst deactivation and it has been quantified in Eqs. (1)-(6) by using the same activity term (φ), which has been defined as: φ=(−rj) (−rj)0 =exp ( − kdt)(7) where (−r j ) and (−r j ) 0 are the reaction rates of each step of the reaction network at t time and zero time, respectively, and k d is the deactivation parameter. The equation proposed for explaining the deactivation kinetics corresponds to a first-order exponential function, which is effective for describing the activity decay in the cracking reactions where a notably deactivation occurs for short contact times (<20 s) [40]. For computing the kinetic parameters, they have been expressed as a function of temperature by means of the reparameterized Arrhenius equation in order to avoid the regression issues derived from the strong correlation between the activation energy and the pre-exponential factor. kj=k∗ jexp [−Ej R(1 T−1 T∗)] (8) being k j * the kinetic reaction rate of the j reaction step at the reference temperature T* (500 ◦C), E j the corresponding apparent activation energy and R the universal gas constant. The system of differential equations that describes the catalytic cracking of PPO, Eqs. (1)-(6), has been solved using an in-house written MATLAB code. The code estimated the required kinetic rate constants and activation energies to fit the weight fraction of the different parameters to those experimentally obtained. To find the best fitting values, a loss function in which the mean squared error is minimized has been employed. Loss Function =1 ne ⋅∑ nl 1∑ p 1(ycal i,p−yexp i,p)2(9) where y ip is the weight fraction of lump i for experiment p, n l is the total number of lumps and n e the total number of experiments. Moreover, superscripts “cal” and “exp” denote the calculated and experimentally determined weight fractions, respectively. 3.3. Statistical significance The discrimination between the different models proposed has been performed by means of a statistical significance test based on Fisher’s method. The procedure is well explained in the literature [41]. In brief, for two kinetic models with different degrees of freedom ( ν A ∕= ν B ) if model B shows a smaller mean squared error than model A (SSE B < SSE A ), the improvement offered by model B with respect to model A will be statistically significant when the following condition is fulfilled: FA−B= SSEA−SSEB SSEB ν A− ν B ν B >F1− α ( ν A− ν B, ν B)(10) being F 1α the critical value of the Fischer distribution function for a level of significance of 95% ( α =0.05). The degrees of freedom have been computed according to the following equation by taking into account the number of experiments (n e ), number of lumps (n l ) and number of parameters (n p ): ν = (ne⋅nl) − np(11) 4. Results 4.1. Characterization The main properties of the PPO are provided in Table 1. It consists of a mixture of hydrocarbons with a broad distillation range that can be divided into 82.0 wt% of HCO (heavy cycle oil), 12.5 wt% of LCO (light cycle oil) and 5.5 wt% of naphtha. These fractions have been defined according to the usual criteria followed by oil refiners: naphtha (C 5 –C 12 ), LCO (C 13 –C 20 ) and HCO (C 21+ ) [42]. Additionally, the chemical composition of the PPO obtained by chromatographic means has been already reported in our previous work [43]. Briefly, they are composed of 67.6 wt% of olefins and 32.4 wt% of paraffins. Even though a descriptive characterization of the catalysts has been already reported in a previous work [44], their main properties are shown in Table 2. In order to compare the properties of these industrial catalysts it must be taken into account its complex configuration, which is composed of an ultrastable Y zeolite (USY) embedded in a mesoand macroporous matrix (consisting of a mixture of clay, silica and alumina) [45]. The highest content of zeolite of ECAT-2 (21 wt%) is in concordance with its high micropore surface area (139 m 2 g -1 ). ECAT-3, in turn, has the highest matrix/USY zeolite ratio, which turns into the highest mesopore surface area and mesopore volume (111 m 2 g -1 and 172 cm 3 g - 1 , respectively). This way, its wide porous structure increases the accessibility of the NH 3 to the acid sites, making ECAT-3 the catalyst with the highest total acidity (124 μ mol NH3 g -1 ), acid strength (130 kJ mol NH3-1 ) and Brønsted/Lewis acid sites ratio (1.56). Furthermore, the high mesopore volume of the matrix will reduce diffusional constraints Table 1 Characterization results of the PPO [43]. Physical properties Specific gravity (◦API) 20.7 Viscosity at 100 ◦C (cSt) 2.2 Average molecular weight (g mol -1 ) 1430 Simulated distillation (◦C) IBP-FBP 157–618 T 50 -T 95 472–601 dyDry Gas dt = φ[( mPPO MHCO V)(k4y2 HCO)+ (k8yLCO) + (k11 yNaphtha)+ (k13 yLPG) − (k15 yDryGas)]mcat V(5) dyCoke dt = φ[( mPPO MHCO V)(k5y2 HCO)+ (k9yLCO) + (k12 yNaphtha)+ (k14 yLPG) + (k15 yDryGas)]mcat V(6) R. Palos et al. Fuel 316 (2022) 123341 5 of the long chains of PPO, easing its access to the external crystal surface of zeolites and, consequently, its posterior cracking on the channels of the zeolite. It is also remarkable the presence of rare earths in the catalysts, especially for ECAT-1 (2.50 wt%), since these elements increase the selectivity to naphtha lump [46]. The presence of P 2 O 5 (with a maximum value of 0.62 wt% for ECAT-1) aims the formation of light olefins. Metals, such as V, Fe and Ni, are irreversibly deposited on FCC catalysts in the successive reaction-regeneration cycles acting as poisons and causing a reduction in throughput by increasing coke formation [47]. 4.2. Model simplification In order to validate the model proposed, the results obtained with ECAT-1 for the parent reaction network (Fig. 1a) are shown below. The values of the apparent kinetic rate constants and activation energies that have minimized the mean squared error of the loss function (Eq. (9)) have been collected in Table 3. The values of the parameters provide a large amount of information about the relevancy of the different catalytic steps. This way, it can be seen that the steps that govern the catalytic cracking of PPO are those in which the cracking of HCO fraction is involved (steps #1 to #5 in Fig. 1a). The highest crackability of the compounds within the HCO fraction is a common result obtained in the catalytic cracking of hydrocarbon streams and it is coherent with the higher crackability of the high-molecular weight olefins [25,48]. However, the values of some of the kinetic rate constants, in particular those corresponding to steps #8 to #12, are so small that the contribution of these kinetic steps can be considered negligible. Therefore, three alternative reaction networks have been proposed (Fig. 1b, c and d) in which various simplifications have been made based on the results collected in Table 3. This way, in alternative network 2 (Fig. 1b) the naphtha fraction has been considered as a final product, i.e. steps #10, #11 and #12 have been removed from the parent network (Fig. 1a). In alternative network 3 (Fig. 1c), in turn, the steps removed have been those in which LCO fraction is converted into coke and into dry gas fractions (steps #8 and #9). Finally, alternative network 4, which is the simplest one from all the proposed, ignores all the routes removed in alternative networks 2 and 3. Thus, in alternative network 4 (Fig. 1d) steps #8 to #12 have been removed from the parent one (Fig. 1a). Consequently, the fitting of the experimental data obtained for the catalytic cracking of PPO with ECAT-1 has been also performed for the alternative reaction networks. Overall, good fitting results have been obtained for all of them. Hence, in order to perform an appropriate discrimination between the four networks, the statistical significance test based on Fisher’s method described in Section 3.3 has been applied. The results obtained have been tabulated in Table 4. Attending to the statistical parameters, it can be seen that the number of experiments and lumps is the same for all the networks, but the number of parameters varies following the trend: n p,1 >n p,3 >n p,2 >n p,4 . Consequently, the degrees of freedom of the different reaction networks follows just the opposite order. On the other hand, the lowest value for the sum of squared errors has been obtained for scheme 3 (5.844 10 -3 ), whereas the values obtained for the other networks are slightly higher and follow the trend: SSE 2 <SSE 1 <SSE 4 . Therefore, since alternative reaction network 4 is the simplest one (less amount of parameters) and the worst fitting has been obtained with it, this one has been taken as reference for the statistical comparison. Thus, it has been assessed if the addition of more catalytic steps is statistically significant. It has been obtained that F 4–1 < F 1− α (0.841 <2.259), F 4–2 <F 1− α (1.569 <3.040) and F 4−3 <F 1− α (1.569 <2.649), meaning that neither parent network nor alternatives 2 and 3 improved in a statistically significant way the fitting of alternative scheme 4. 4.3. Kinetic parameters Based on all the previous, the alternative reaction network 4 has been also used for the fitting of the data obtained with both ECAT-2 and ECAT-3. The goodness of fit has been evaluated using parity plots Table 2 Properties of the commercial FCC equilibrium catalysts [44]. ECAT–1 ECAT–2 ECAT–3 USY zeolite content (wt%) 13.8 21.0 11.8 Zeolite UCS (Å) 24.26 24.23 24.27 Physical properties Apparent bulk density (g cm -3 ) 0.90 0.88 0.89 BET surface area (m 2 g -1 ) 124 190 174 Micropore surface area (m 2 g -1 ) 87 139 63 Micropore volume (cm 3 g -1 ) 0.04 0.06 0.03 Mesopore surface area (m 2 g -1 ) 37 50 111 Mesopore volume (cm 3 g -1 ) 112 147 172 Acidic properties Acidity ( μ mol NH3 g -1 ) 40 81 124 Average acid strength (kJ mol NH3-1 ) 100 126 130 Brønsted/Lewis acid sites ratio 0.66 0.91 1.56 Chemical composition Al 2 O 3 (wt%) a 53.8 52.2 62.8 SiO 2 (wt%) 38.5 44.0 35.1 SiO 2 /Al 2 O 3 ratio 0.72 0.84 0.56 Cu (ppm) 24 66 10 Ni (ppm) 741 106 53 V (ppm) 3335 244 67 Fe (wt%) 0.19 0.25 0.21 Na (wt%) 0.29 0.26 0.19 P 2 O 5 (wt%) 0.62 0.45 0.32 rare earths (wt%) 2.50 0.86 1.35 a Corresponding to the total content of alumina (in zeolite and matrix) Table 3 Apparent kinetic parameters at the reference temperature and activation energies for the reactions involved in the catalytic cracking of plastic pyrolysis waxes. Step Reaction involved k j, 773 K E j d – Deactivation, k da (3.2 ±0.2) 10 -2 136.0 ±7.1 1 HCO → LCO, k 1b 41.8 ±2.4 63.6 ±3.4 2 HCO → Naphtha, k 2b 14.4 ±0.8 86.6 ±4.1 3 HCO → LPG, k 3b 1.4 ±0.1 17.8 ±1.1 4 HCO → Dry gas, k 4b 0.8 ±0.2 22.2 ±1.5 5 HCO → Coke, k 5b 2.4 ±0.1 112.9 ±5.4 6 LCO → Naphtha, k 6c (3.5 ±0.2) 10 -4 96.2 ±4.8 7 LCO → LPG, k 7c (3.8 ±0.8) 10 -3 63.2 ±3.3 8 LCO → Dry gas, k 8c (3.3 ±0.1) 10 -8 56.9 ±3.4 9 LCO → Coke, k 9c (3.4 ±0.2) 10 -8 102.1 ±5.2 10 Naphtha → LPG, k 10c (3.2 ±0.4) 10 -6 73.2 ±3.8 11 Naphtha → Dry gas, k 11c (2.6 ±0.4) 10 -6 49.8 ±2.5 12 Naphtha → Coke, k 12c (2.8 ±0.6) 10 -9 140.6 ±7.7 13 LPG → Dry gas, k 13c (4.7 ±0.9) 10 -3 54.4 ±2.9 14 LPG → Coke, k 14c (5.9 ±0.5) 10 -4 89.5 ±4.8 15 Dry gas → Coke, k 15c (1.6 ±0.1) 10 -2 130.1 ±6.4 The parameters are measured in: a s -1 ; b m 6 kg cat-1 kmol -1 s -1 ; c m 3 kg cat-1 s -1 ; d kJ mol -1 Table 4 Statistical comparison of the four kinetic networks. Statistical parameter Reaction network 1 2 3 4 n e 36 36 36 36 n l 6 6 6 6 n p 16 13 14 11 ν 200 203 202 205 SSE 5.851 10 -3 5.848 10 -3 5.844 10 -3 5.974 10 -3 F 0.841 (F 4–P ) 1.569 (F 4–2 ) 1.569 (F 4–3 ) F 1α 2.259 3.040 2.649 R. Palos et al. Fuel 316 (2022) 123341 6 (Fig. S1 in the Supplementary Material) by evaluating the final fit of calculated data for the three catalysts against raw experimental data. As it can be seen, almost a perfect fit between the calculated and the experimental weight fraction has been obtained for all the catalysts, with the exception of the scattering of some points, especially for ECAT3. Nevertheless, those deviations do not exceed the 5% as they remain inside the region delimited by the dashed lines. In Table 5 have been collected the values computed for the apparent kinetic parameters and the activation energies of the kinetic steps involved in the catalytic cracking of PPO by the three catalysts. Overall, small differences have been obtained in the values of the kinetic parameters with all the catalysts. These differences lie in the properties of the catalysts, considering the effect of the acidity and porous structure on the activity, selectivity and deactivation of the catalysts [49]. This way, ECAT-2 has the highest value for the deactivation parameter (0.030 s -1 ) because of its moderate mesopore surface area (50 m 2 g -1 ) and mesopore volume (147 cm 3 g -1 ), which are not enough for easing the diffusion of coke precursors towards the external surface of catalyst particles. Likewise, the confinement of the precursors will block the micropores of the zeolite resulting in the ineffectiveness of its high content of zeolite [50]. It is well-established the role as coke precursors of light olefins in cracking processes, as they undergo oligomerization, aromatization and condensation reactions that are catalyzed by strong acid sites [51,52]. In the same line, do stand out the rate constants of the reactions that form coke from LPG and dry gas fractions (2.8 10 -3 and 1.0 10 -4 m 3 kg cat-1 s -1 , respectively) using ECAT-2. In contrast, the lower zeolite/matrix ratio of ECAT-3 that entails a higher mesopore surface area (111 m 2 g -1 ) and a higher mesopore volume (172 cm 3 g -1 ), will improve the diffusion of coke precursors, attenuating their confinement and, consequently, the blockage of the micropores. Moreover, the enhanced accessibility and diffusion of PPO chains to the active sites in ECAT-3 are in concordance with the high values of the kinetic parameters for the reactions that convert the components within the HCO lump into dry gases (1.16 m 6 kg cat-1 kmol -1 s -1 ) and the components within the LCO lump into naphtha (1.0 10 -3 m 3 kg cat-1 s -1 ). The similarities among the rest of the kinetic parameters for the different catalysts lie in the synergistic and parallel effects of the porosity and acidity that boost the extent of the cracking reactions. The values computed for the apparent activation energy of the different catalytic steps have been collected in Table 6. Unlike the kinetic parameters (Table 5), significant differences are observed between the activation energy required in some of the catalytic steps for the different catalysts. Likewise, the energy barrier that must be overcome for the deactivation step is very different depending on the catalyst used. The lower activation energy of the deactivation stage (79.2 kJ mol -1 ), added to the high value of the deactivation kinetic parameter obtained (3.0 10 -2 s -1 in Table 5) expose the high tendency of ECAT-2 to be deactivated. Equally, the high amount of acid sites on ECAT-3 explains the low activation energy required for the steps of formation of dry gas from HCO and LPG lumps (17.1 and 40.0 kJ mol -1 , respectively). In addition, the high matrix mesoporosity of ECAT-3, which is the other key feature of the catalysts, reduces the activation energy of the steps limited by the diffusivity of the components. This way, this catalyst reduces the energy involved in the steps that convert the HCO in LCO, LCO in naphtha and LCO in LPG (60.5, 42.5 and 58.3 kJ mol -1 , respectively), as well as in the formation of coke from LPG and dry gas (4.4 and 40.7 kJ mol -1 , respectively). 4.4. Yields of products The evolution of the yields of products with high commercial interest as fuels (LCO and naphtha) and with high content of olefins (LPG) has been obtained for the three catalysts by computing the kinetic model previously described and using the kinetic parameters collected in Tables 5 and 6. One should note that the PPO fed to the reactor has a content of 12.5 wt% of LCO and a 5.5 wt% of naphtha, as it has been previously detailed in Section 4.1. Those contents have not been taken into account for depicting the evolution of the yields, in order to assess the formation of these lumps in its real magnitude. On the other hand, in order to fully understand the obtained results, it should be taken into account that conversion has been defined as the ratio of mass of HCO converted to lighter products and to coke to the mass of HCO fed: Conversion =(mHCO)PPO − (mHCO)Products (mHCO)PPO 100 (12) The yield of each i lump of products has been defined as the mass of lump i referred to the total mass of lump HCO fed: Yieldi=mi mHCO 100 (13) Therefore, Figs. 2-4 compare the evolution with conversion of the yield of LCO, naphtha and LPG, respectively, obtained with the three catalysts at different temperatures. In a previous work [19] it has been detailed the composition of these lumps, stressing out the interest of the naphtha lump (research octane number up to 105) for being added to the stream of gasoline in refinery. It is also remarkable the propylene-rich LPG lump produced. The trend of the curves in Fig. 2 exposes the character of the LCO lump as an intermediate in the reaction network [53], as they go through a maximum at values of conversion of ca. 80 wt%. Furthermore, it can be seen that high temperatures promote the cracking reactions that convert the molecules within this lump into lighter molecules, Table 5 Apparent kinetic parameters at the reference temperature for the reactions involved in the catalytic cracking of PPO for the three catalysts. Step Reaction k i, 773 K ECAT-1 ECAT-2 ECAT-3 – Deactivation, k da (2.4 ±0.2) 10 - 2 (3.0 ±0.2) 10 - 2 (2.3 ±0.2) 10 - 2 1 HCO → LCO, k 1b 41.1 ±1.4 42.8 ±1.8 41.5 ±1.0 2 HCO → Naphtha, k 2b 14.3 ±1.1 14.7 ±1.5 14.7 ±1.7 3 HCO → LPG, k 3b 1.42 ±0.8 1.80 ±0.6 1.50 ±0.7 4 HCO → Dry gas, k 4b 0.79 ±0.1 0.72 ±0.3 1.16 ±0.5 5 HCO → Coke, k 5b 2.08 ±0.1 2.10 ±0.2 2.02 ±0.2 6 LCO → Naphtha, k 6c (2.4 ±0.4) 10 - 4 (4.5 ±0.3) 10 - 4 (1.0 ±0.1) 10 - 3 7 LCO → LPG, k 7c (3.9 ±0.5) 10 - 3 (3.8 ±0.4) 10 - 3 (4.1 ±0.6) 10 - 3 13 LPG → Dry gas, k 13c (4.5 ±0.5) 10 - 3 (4.9 ±0.9) 10 - 3 (4.5 ±0.4) 10 - 3 14 LPG → Coke, k 14c (1.7 ±0.2) 10 - 3 (2.8 ±0.7) 10 - 3 (1.1 ±0.3) 10 - 3 15 Dry gas → Coke, k 15c (1.9 ±0.1) 10 - 5 (1.2 ±0.3) 10 - 4 (4.7 ±0.6) 10 - 5 The parameters are measured in: a s -1 ; b m 6 kg cat-1 kmol -1 s -1 ; c m 3 kg cat-1 s -1 Table 6 Values of the apparent activation energy of the reactions involved in the catalytic cracking of PPO for the three catalysts. Step Reaction E i (kJ mol -1 ) ECAT-1 ECAT-2 ECAT-3 – Deactivation 163.2 ±8.0 79.2 ±3.8 120.3 ±5.9 1 HCO → LCO, E 1 65.3 ±3.4 72.8 ±3.4 60.5 ±2.7 2 HCO → Naphtha, E 2 97.3 ±4.8 96.2 ±4.0 89.4 ±4.1 3 HCO → LPG, E 3 17.7 ±1.0 18.4 ±2.0 16.6 ±1.3 4 HCO → Dry gas, E 4 21.5 ±1.1 17.1 ±1.4 26.8 ±1.0 5 HCO → Coke, E 5 125.9 ±6.3 133.0 ±6.6 129.0 ±6.0 6 LCO → Naphtha, E 6 122.5 ±6.0 83.9 ±4.8 42.5 ±2.6 7 LCO → LPG, E 7 63.9 ±3.3 69.4 ±5.5 58.3 ±3.1 13 LPG → Dry gas, E 13 57.4 ±2.9 40.0 ±3.5 72.9 ±4.0 14 LPG → Coke, E 14 31.8 ±1.4 27.3 ±1.4 4.4 ±0.5 15 Dry gas → Coke, E 15 105.2 ±5.2 91.7 ±4.8 40.7 ±2.1 R. Palos et al. Fuel 316 (2022) 123341 7 Fig. 2. Model prediction (lines) and experimental data (symbols) of the evolution of the LCO yield with the level of conversion for the three catalysts. Fig. 3. Model prediction (lines) and experimental data (symbols) of the evolution of the naphtha yield with the level of conversion for the three catalysts. Fig. 4. Model prediction (lines) and experimental data (symbols) of the evolution of the LPG yield with the level of conversion for the three catalysts. R. Palos et al. Fuel 316 (2022) 123341 8 resulting in lower yields of LCO. Comparing the results obtained with the three catalysts, similarities are observed between the results obtained. This way, with ECAT-1 and ECAT-2 higher values than with ECAT-3 are obtained, yielding up to 46.5 wt% with the former catalysts at 500 ◦C. Attending to the evolution of the yields of naphtha and LPG lumps (Figs. 3 and 4, respectively), both are end-products in the reaction network since their yield increases continuously with the extent of conversion. Nonetheless, in spite of the evolution obtained for LPG lump (Fig. 4), the molecules within this lump are cracked to dry gas and condensed to coke as it has been previously obtained in the reaction network (Fig. 1d). With regard to the evolution of the yield of naphtha (Fig. 3), high temperatures promote the production of this lump, since the cracking of molecules within HCO and LCO lumps is boosted. Furthermore, higher yields of naphtha have been obtained with ECAT-3, yielding up to 33.6 wt% at 560 ◦C. However, the maximum values obtained with ECAT-2 and ECAT-1 have been slightly inferiors (31.1 and 29.6 wt%, respectively). The evolution of the different yields (Figs. 2-4) strongly depends on the properties of the catalyst used (Table 2) and can be correlated with the values of the apparent kinetic parameters reported on Table 5. This way, ECAT-3 is by far the catalyst with the highest and strongest acidity (124 μ mol NH3 g −1 and 130 kJ mol NH3-1 , respectively), which turns into the catalyst with the highest cracking activity. Moreover, it is the catalyst with the highest mesoporosity that eases the access of the bulky molecules within the HCO and LCO lumps to the acid sites located in the inside of the porous structure of the catalyst. Therefore, the highest yields of naphtha and LPG, together with the lowest yield of LCO should be expected using this catalyst. Even though ECAT-2 possesses a lower amount of acid sites available (81 μ mol NH3 g −1 ), their strength is quite remarkable (126 kJ mol NH3-1 ) making a priori this catalyst a serious candidate for maximizing the yield of naphtha and LPG lumps. However, its microporous nature and its, subsequent, shortness in mesopores are unsuitable for boosting the access of the heavy molecules to inner acid sites. Consequently, the behavior of ECAT-2 is only comparable with ECAT-3 at 560 ◦C as an increase in temperature increases the diffusivity [54]. Nonetheless, ECAT-2 promotes the formation of LPG instead of naphtha, which can be attributed to the overcracking reaction that takes place within the micropores of the zeolite as a consequence of the higher residence time of the reactants. Its low content of rare earths will also presumably contribute to obtain the aforementioned results [46]. Furthermore, the narrower porous structure of ECAT-2 will lead to a faster activity decay of the catalyst. Finally, the configuration and composition of ECAT-1 are the less favorable ones to promote the cracking reactions. Indeed, ECAT-1 has the lowest superficial area (124 m 2 g −1 ), the lowest acidity (40 μ mol NH3 g −1 ) and the weakest acid strength (100 kJ mol NH3-1 ). In addition, the high concentration of impurity metals detected on ECAT-1, especially of vanadium (3335 ppm), will also contribute to deteriorate the properties of the catalyst. Consequently, slightly lower yields of both naphtha and LPG lumps (Figs. 3 and 4) have been obtained with ECAT-1. 4.5. Selectivity to products The selectivity to each lump i has been defined as the mass of lump i formed respect to that of all the products: Selectivityi=mi mLCO +mNaphtha +mLPG +mDryGas +mCoke 100 (14) Taking into account that naphtha and LPG lumps are the ones with the highest commercial interest, the evolution with conversion of the selectivity to them has been depicted in Fig. 5. Overall, it can be seen how different the selectivity to each lump is. This way, the selectivity to naphtha lump is barely affected by the extent of conversion. The trend followed by the selectivity to naphtha curves depends on the catalyst. This way, it can be seen that with ECAT-1, which is the less active catalysts, the selectivity to naphtha remains almost steady for values of conversion below 80%, to increase exponentially at higher values. For ECAT-3, in turn, the growth can be noticed for values of conversion above 55%. Furthermore, the differences between the performances of the catalysts are more evident at high temperatures and high values of conversion. This way, the selectivity to naphtha has been maximized at 560 ◦C with ECAT-3 reaching a value of 35.8 wt%, whereas under the same conditions a selectivity of 31.5 wt% has been obtained with ECAT1. ECAT-2, in turn, offers an intermediate result and a selectivity to naphtha of 32.8 wt%. In contrast, selectivity to LPG lump grows exponentially with conversion since the very beginning of the reaction. The effect of the temperature is less marked but the opposite in the case of the LPG. Likewise, an increase from 500 to 560 ◦C for a fixed value of conversion entails just a reduction of the selectivity to LPG of ca. 2.5 wt%. Focusing on the performance of the catalysts, ECAT-2 offers the highest selectivity at 560 ◦C but also the lowest at 530 and 500 ◦C. This result is characteristic of a partially deactivated catalyst, in which thermal cracking plays a more important role that in the case of ECAT-1 and ECAT-3. Fig. 5. Model prediction of the evolution of the selectivity to naphtha (dashed lines) and to LPG (solid lines) lumps with the level of conversion for the three catalysts. R. Palos et al. Fuel 316 (2022) 123341 9 Attending to the results collected on Figs. 2-5, operating under the conditions that allow for reaching conversions levels within the range 60–80% would be the optimal considering the possibility of varying the reaction temperature between 500 and 560 ◦C. This way, the conversion of HCO would be promoted keeping under control the overcracking reactions that would lead to obtain too much dry gas. Furthermore, ECAT-3 should be the selected one for turning the production to LPG and naphtha lumps, whereas ECAT-1 and ECAT-2 would increase the yield of LCO lump in detriment to the yield of naphtha. 4.6. Catalyst deactivation Since catalyst deactivation has a notable impact on the results collected in Sections 4.4 and 4.5 about the yields and selectivity, the evolution of the activity term (φ) of the three catalysts with contact time at 500 ◦C has been plotted on Fig. 6. One should note that these curves have been obtained by applying the previously proposed deactivation equation (Eq. (7)) and using the corresponding kinetic parameters (Tables 5 and 6). It can be seen that for a contact time of 6 s, all the catalysts maintain good activity levels as they are above of the 83% of the initial activity. This result is very different to that obtained in the cracking of VGO (benchmark feed in FCC) [55], where the catalyst was totally deactivated. This result exposes the crucial role that the composition of the stream fed to cracking reactor plays in catalyst deactivation and, therefore, in products yield and distribution. This way, the heterogeneity of the VGO, with high contents of aromatics and the presence of polyaromatics, is more prone to the formation of coke than the olefins that predominate in the composition of the PPO [43]. This low deactivation is an interesting result for adopting different cracking strategies for the PPO, such as being co-fed with other refinery streams that deactivate the catalysts in a large extent. Comparing the evolution followed by the catalysts, ECAT-1 and ECAT-3 show almost identical curves of activity vs. time as a difference of<1% for a contact time of 6 s (ca. 87%) has been obtained. However, ECAT-2 suffers from a higher and more severe activity decay since the very beginning of the reaction. Indeed, the final value for activity obtained for this latter catalyst is of 83.5%. Undoubtedly, the deactivation suffered by ECAT-2 lies in the porous structure of the catalyst, which is by far more microporous (Table 2) than the structure of the other catalysts. Consequently, the coke formed during the reaction will more easily block the channels of the zeolite reducing the accessibility of hydrocarbon species to the catalyst inner micropore network [56]. To offer another perspective of the deactivation results, Fig. 7 depicts the evolution of the activity of the three catalysts with the content of coke deposited. Clearly, the amount of coke deposited on ECAT-2 is higher than that deposited on ECAT-1 or ECAT-3. Consequently, ECAT-2 suffers from a higher activity decay than the other catalysts. In spite of that, attending to the accelerated decrease of the activity obtained for all the catalysts, it can be concluded that the deactivation mechanism is highly affected by the micropore blocking caused by coke deposition. This phenomenon will also restrict the access of the reactants to the acid sites located in the inner crystals of the zeolite. This result is in concordance with the hypothesis of the key role of the matrix mesopores for attenuating the catalyst deactivation by delaying the aforementioned phenomenon. 5. Conclusions A lumped kinetic modeling method has been applied to the experimental data of the catalytic cracking of plastic pyrolysis oil (PPO) over three commercial FCC equilibrium catalysts. By means of a statistical data analysis, it has been obtained that from the four different reaction networks proposed, the simplest one was the most appropriate for describing the process. From the kinetic parameters obtained in the fitting of the results, it has been obtained that both total acidity and acid Fig. 6. Model prediction of the evolution of the catalysts activity with contact time at 500 ◦C. Fig. 7. Model prediction of the evolution of the catalyst activity with the content of coke deposited on the three catalysts. R. Palos et al.