You are on page 1of 19

Journal of Analytical and Applied Pyrolysis 60 (2001) 103 121 www.elsevier.

com/locate/jaap

Thermal degradation of polystyrene


T. Faravelli *, M. Pinciroli, F. Pisano, G. Bozzano, M. Dente, E. Ranzi
CIIC -Dipartimento di chimica Industriale e Ingegneria Chimica, Politecnico di Milano, Piazza L. da Vinci 32, 20133 Milano, Italy Received 14 June 2000; accepted 18 September 2000

Abstract Thermal degradation of plastic wastes offers the possibility of recovering energy and useful chemicals. Polyethylene and polypropylene pyrolysis have been discussed already in previous works (E. Ranzi, M. Dente, T. Faravelli, G. Bozzano, S. Fabini, R. Nava, V. Cozzani, L. Tognotti, J. Anal. Appl. Pyrol., 4041 (1997) 305 319 and T. Faravelli, G. Bozzano, C. Scassa, M. Perego, S. Fabini, E. Ranzi, M. Dente, J. Anal. Appl. Pyrol., 52 (1999) 87 103). This paper aims to develop a detailed kinetic model of polystyrene thermal degradation. The predictions of overall rates of degradation and volatile product distribution are compared with experimental results obtained by different authors at different pressure and temperature conditions. In order to reduce the computing times required by the numerical integration of the kinetic model, a exible lumping procedure has also been introduced. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Lumping procedure; Pyrolysis; Gasication

1. Introduction The fraction of plastics in municipal solid wastes (MSW) and in refuse derived fuels (RDF) is increasing continuously. In Western Europe, 610% of MSW is composed of plastics (9.3 million tons in 1992). The largest part (72%) is disposed of by landll, whereas the remaining part is incinerated or recycled in different ways [3].
* Corresponding author. Tel.: + 39-2-23993282; fax: + 39-2-70638173. E -mail address: tiziano.faravelli@polimi.it (T. Faravelli). 0165-2370/01/$ - see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 5 - 2 3 7 0 ( 0 0 ) 0 0 1 5 9 - 5

104

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Pyrolysis and gasication are now recognized as promising routes for the upgrading of solid wastes to more usable and energy dense materials, such as gas fuel and/or fuel oil, or to high value feed stocks for the chemical industry. The characterization of pyrolysis behavior of plastic wastes is then signicant in the optimization of pyrolysis processes for the recovery of valuable product fractions. Moreover, a pyrolysis step is always present in the initial stages of gasication and combustion processes. Literature reports several papers on pyrolysis and gasication of plastics. The goal of the major part of the works reported so far was to retrieve monomers or other valuable products through thermal processes in various types of reactors. They deal with the characterization of the rate of weight loss during the primary thermal degradation [4 7] as well as on the primary product characterization [8 11]. In the attempt of developing a model of plastic pyrolysis in full-scale systems, the rst step is to describe the thermal degradation of polymers in terms of an intrinsic kinetics, in which heat and mass transfer limitations are not included. Generally, kinetic models with apparent kinetic parameters are proposed in literature for plastics and biomasses. These models do not take into account the complete and more rigorous description of the chemistry of polymer thermal degradation and describe the pyrolysis process by means of a simplied reaction pathway. Each single reaction step is representative of a complex network of reactions. This approach proved adequate to describe the apparent kinetics, only in a narrow range of heating rates and operating conditions. In particular, a single step model is not able to cover, with the same kinetic parameters, a wide range of heating rates, temperatures and conversion levels. The possible presence of mass and heat transfer limitations, generally not taken into account in the identication of kinetic data, spreads the range of variation of these kinetic constants. The broad variations between the activation energies and pre-exponential factors found by various authors [4 6] are essentially due to two reasons differences in properties and characteristics (molecular weight, presence of weak links, additives) of polystyrene (PS), and differences in experimental conditions from which kinetic data are calculated; for example, anionic PS is thermally more stable than thermal PS, because of the greater number of weak links in the latter. As a result of the previous considerations, there comes out the interest in a mechanistic model able to account for the differences in starting material and also to describe the phenomenon in a wide range of reaction conditions (i.e. heating rates and temperatures). Furthermore, the mechanistic model would allow to predict the detail of gas product distribution and this is the most signicant step in the possibility of an upgrading of solid wastes toward chemical reactants. A mechanistic model for the polyethylene and polypropylene (PE and PP) degradation process was developed [1,2]. This work was prepared on very similar basis. In order to describe properly the phenomenon, particular attention has been paid to the reaction steps and to the physical aspects of the degradation process since both play an important role in the nal product distribution. As seen by many authors [4 15], the propagation step is the result of a competition between three

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

105

different reaction mechanisms unzipping, intramolecular and intermolecular H transfer. As a consequence, these three pathways are introduced in the reaction scheme through the denition of two coefcients i and k, which indicate the fractions of the radicals involved, respectively, in intermolecular and intramolecular abstraction. The remaining radicals give unzipping reactions with a high production of the monomer. It is worth observing that this unzipping reaction constitutes a very relevant propagation mechanism in the usual conditions but it was not accounted, due to its lower importance, in PE and PP thermal degradation. The radical chain pyrolysis reactions here considered take place only in the liquid phase and are described on the basis of a very limited number of independent kinetic parameters. 2. Kinetic mechanism The thermal degradation of most of the polymers is a typical radical chain mechanism, where initiation, propagation and termination reactions are the relevant reaction classes. These radical reactions are described completely by a limited set of independent kinetic parameters, evaluated on the basis of structural contributions as well as similarity and analogy rules.

2.1. Initiation reactions


Initiation reactions determine a C C bond cleavage of polymer chains to form radicals; the following two types of different initiation reactions can be identied. a1. Random scission to form one primary radical (Rp) and one secondary benzyl radical (Rsb) with a strong benzylic resonance.

a2. Chain-end scission to form again one secondary benzyl (Rsb) and the resonantly stabilized allyl benzene radical (Ra).

This second type of initiation reactions has an increasing importance during degradation process, because of both the growth chain end positions with decreasing of the molecular weight and the formation of several unsaturated species, due to the propagating b-scission reactions (see b3).

106

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

We do not account for the presence of weak links, which can give surely an important contribution especially for radical polymerized polystyrene. Depending on the nature of the polymer, proper adaptive and corrective factors can force these initiation steps, but this aspect is beyond the main scope of this work. 2.2. Propagation reactions Propagation step consists of the sequence of H-abstraction and b-decomposition or unzipping reactions. There are following two types of H-abstraction reactions. b1. Intermolecular abstractions the radicals abstract the hydrogen from a different molecule:

Due to the higher stability of the long lived resonantly stabilized benzylic radicals formed, it is only considered the intermolecular abstraction on the tertiary carbons atoms with the formation of Rt. b2. Intramolecular abstractions the radicals Rsb and Rp can easily form ve-, six- or seven-membered ring intermediates, with the nal result of a 1 4, 1 5 or 1 6 isomerization reaction:

These reactions are also called back biting reactions. The six- or seven-membered ring reaction is favored by the energetic point of view (lower strain energy), while the 1 4 isomerization is favored from the entropic point of view, being lower the number of degrees of freedom (rotors) to be blocked. As a result of the sterical hindrance in the liquid phase, back biting 1 5 reaction is the favored one and the only one considered here. b3. The tertiary benzylic radical Rt undergoes a scission of the C C bond in b position to form a secondary benzylic radical and a polymer species with an unsaturated end:

As already mentioned, unzipping reactions are b-decomposition reactions of Rsb radicals with the formation of a monomer and another Rsb radical with a monomeric unit less. These reactions can be considered as the reverse of poly-addition reactions:

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

107

2.3. Termination reactions


Two different second-order termination reactions are considered. c1. Recombination reactions:

c2. Disproportionation reactions of radicals, like:

The main difference between these reactions is the formation of species with an unsaturated end in the disproportionation reaction.

3. Kinetic parameters As discussed already in the case of polyethylene and polypropylene pyrolysis, rate constants determined in gas-phase pyrolysis of hydrocarbons constitute the starting point in the evaluation of kinetic parameters valid for liquid-phase degradation process [1]. Signicant corrections need to be applied to gas-phase kinetic parameters in order to account for the condensed state, because of the inhibition of molecular rotations of large C C segments [16]. Typically, reactions with low heat of reactions have a marginal correction when transported from the gas to the liquid phase, for this reason the propagation reactions are assumed with the same kinetic parameters in both the phases. On the contrary, chain initiation reactions require signicant corrections. This approach has been already tested and validated in the case of visbreaking process [17,18], as well as in PE and PP pyrolysis [1,2]. As already described, we consider two types of initiation reactions: a1. random scission {PS Rsb + Rp} ksr = 5 1013 exp

63700 RT

108

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

a2. allyl scission {PS Rsb + Ra} ksa = 5 1012 exp

58700 RT

These activation energies are taken directly for the equivalent C C bond cleavage in the gas phase [18] and contain already a correction of about 3800 cal mol 1 for the transposition to the liquid phase [1]. Further corrections are also required in the case of radical recombination reactions. In fact, the kinetic parameters of the termination reactions in the condensed phase become, kt = 1012.8 T Ev Vs exp b2 400 RT

 

Vs is the molar volume of the ux unit VS = PMS num0 = z z

where m0 is the molecular mass of monomer and z is the liquid density, which can be considered constant reasonably during the process and equal to 900 kg m 3, that is the estimated value (400C) starting from the polymer density of 1050 kg m 3 (25C) [19]. nu is the monomeric units of polystyrene characterizing the ux unit for the molecular momentum transfer and a value of nu = 7 is assumed, on the basis of Eyrings free volume theory [19]. Ev is the energy required for the mobility of the molecular ux unit, and b 2, the corrective factor that takes into account the symmetry, resonance, steric and surface effects [16]. In the case of polystyrene, the kinetic constant for chain termination reactions simply becomes, kt = 5 106T exp

14 000 RT

On the basis of the kinetic constant calculated for the initiation and termination reactions, it is then possible to evaluate the global concentration of radicals. We assume that all the different radicals (primary, secondary benzylic and tertiary, with different chain length) are equivalent to a unique lumped radical R. Initiation and termination reactions can be written as follows, initiation termination PS 2R
t 2R PS

ks

Assuming the steady state hypothesis, the concentration of this pseudo radical is evaluated from its mass balance, d[R] = 2ks[PS] 2kt[R]2 = 0 dt

[R] =

' '

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

109

ks[PS] kt

Taking into account both the random and the allyl initiation steps, the previous expression becomes, [R] = ksr[PSsr] + ksa[PSsa] kt

where [PSsr] and [PSsa] are, respectively, the concentration of the C C bonds, which could undergo random scission and allyl scission. The concentration of allyl and primary radicals is negligible, because they can be obtained only by initiation steps, whereas secondary benzyl radicals, which can also be formed by b-scission reactions, are the predominant ones. Radical chain mechanism is the result of propagation reactions of Rt and Rsb radicals. As far as the tertiary benzyl radicals are concerned, there are two possible paths, b-decomposition reactions{Rt PS + Rsb}; ki = 1013 exp

27 000 RT

H abstraction reactions {Rt + PS PS + R% t}; ker = 5 107 exp

16 500 RT

As far as the secondary benzyl radicals are concerned, there is the competition among three different reaction classes, H abstraction reactions of a tertiary benzyl hydrogen {Rsb + PS PS + Rt}; kef = 5 107 exp

13 500 RT

unzipping reactions {Rsb Styrene + Rsb}; ku = 1013 exp

26 000 RT

back biting reactions {Rsb Rt}. kbb(1,5) = 109 exp

16 000 RT

As mentioned already, the kinetic parameters above reported are taken directly from the analogous well dened gas phase reactions (inter and intra molecular H-abstractions, b-scissions).

110

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

On this basis, a unique kinetic expression for the propagation step involving tertiary radicals can be derived, H-abstraction b-decomposition re-abstraction PSn + R Rtn + PS;
i Rtn Rs(n k) + PSk ; er PS + Rtn Rt + PSn.

kef

where R tn is the tertiary radical of length n, whilst Rt is a generic tertiary radical, with the same chain length as PS. The production of polymer species of length k (PSk ) can be expressed as, d[PSk ] = ki [Rtn ] dt where n]k+1

The steady-state assumption for the radicals Rtn becomes, d[Rtn ] = kef[PSn ][R] ki [Rtn ] ker[Rtn ][PS] = 0 dt [Rtn ] = Thus, d[PSk ] kefki = [R][PSn ] = kp[PSn ] dt ki + ker[PS] where kp = kefki [R ] ki + ker[PS] kef [R][PSn ] ki + ker[PS]

is the equivalent rate constant of the apparent propagation reactions involving the tertiary radicals. Two parameters i and k are useful to dene the fractions of secondary benzyl radicals which, respectively, follow H-abstraction and back biting reactions, i= k= kef[PS] ku + kef[PS] + kb

kb ku + kef[PS] + kb

The kb rate constant has been obtained in analogy with the previous kp but considering the back-biting as abstraction mechanism, kb = kbb(1 5) ki ki + ker[PS]

The remaining fraction (1-i -k ) of Rsb undergoes unzipping reactions [20]. In the usual conditions, with the proposed rate constant, i ranges between 0.1 and 0.2 and k is about 0.1.

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

111

4. Analytical and numerical solution of mass balance equations In the analysis of the kinetic mechanism of polystyrene thermal degradation, there is a progressive formation of unsaturations in the end positions of the molecules. Each molecule in the system is identied by the number of phenyl groups contained and on the basis of the different end unsaturations. Thus, it is possible to have in the system components with alkane backbone (P), without double bonds, alkene backbone (O) and a v dialkene backbone (D), respectively, with one and both ends unsaturated. However, these simple assumptions would not allow to distinguish molecules with similar structures. For example, O1 or the alkene backbone with only one phenyl group would include both styrene and h -metyl-styrene. It is then convenient to consider three types of chain end for each one of the previously considered species. This classication is shown schematically in Table 1. The total nine families of different species are reduced to ve, with the hypothesis that the initial polymer is constituted only by type I alkane backbone (PI). On the basis of the proposed mechanism the system is only composed by the following species PI and PIII alkane backbone, OI and OII alkene backbone, and DII dialkene backbone. As briey sketched in Fig. 1, both the alkene backbone families come from b-scission of tertiary radicals. DII is formed by the b-scission of both OI and OII. Finally, the H abstraction of secondary radicals produce either PI or PIII according to their structure. All the main degradation products observed experimentally are easily taken into account by these ve families.
Table 1 Families of species formed during polystyrene degradation

112

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Fig. 1. Sample of formation of the different backbone families, starting from the assumed original polymer structure.

Thermal degradation is a cracking process taking place in a liquid phase and reaction products go away as volatiles. Cracking reactions in gas phase are neglected and it is necessary to distinguish the molecules in the liquid phase from the gaseous ones. Clausius Clapeyron and Trouton Meissner equations [21] allow to dene, as a rst approximation, the lower limit of the number of monomeric units (L ) corresponding to species in the liquid phase as a function of system pressure (atm) and temperature (K), L= 1 T2 ln P 1 8 (136)2 10, 5

The good agreement found by this relationship in comparison with the experimental data is shown in Fig. 2, where the estimated boiling temperatures of hydrocarbons with different numbers of carbon atoms are compared with the experimental values. It has to be noted that at very high molecular weight, this

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

113

approach does not predict correctly the boiling temperature, but these temperature conditions are quite far from those of interest. Formed species with a number of monomeric units lower than L are considered directly pertaining to the gas phase. On the basis of the previous assumptions and hypotheses, it is then possible to obtain the following mass balance equations for the ve families in the system, dPIn = kp(n 1, 5) PIn + i RPIn dt dPIIIn = kp(n 2)PIIIn + i RPIIIn dt
1 1 dOIn = kp(n 2, 5)OIn + kp % PIj + kp % PIIIj + kp % OIj + i ROIn 2 n+1 2 n+2 dt n+1 dOIIn 1 1 = kp(n 2)OIIn + kp % PIj + kp % OIIj dt 2 n+1 2 n+2 dDIIn 1 1 = kp(n 3)DIIn + kp % OIj + kp % OIIj + kp % DIIj 2 n+1 2 n+1 dt n+2

where

Fig. 2. Comparison between predicted (line) and experimental (dots) boiling temperatures of aliphatic hydrocarbons. Boiling temperatures of styrene and dimer are also reported.

114

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121


1 1 RPIn = kp % PIj + kp % OIIj + (1-i -k )RPIn + 1 + k RPIn + 3 2 n+2 2 n+2 1 1 RPIIIn = kp % PIj + kp % PIIIj + kp % OIj + (1-i -k )RPIIIn + 1 + k RPIn + 3 2 n+1 2 n+2 n+2 1 1 ROIn = kp % OIj + kp % OIIj + kp % DIIj + (1-i -k )ROIn + 1 + k ROIn + 3 2 n+2 2 n+1 n+2

are the mass balance equations for secondary benzyl radicals of different types present in the system. n is the chain length and terms like (n 2) means that not all the positions in the molecule are equivalent and can be involved in the reaction, may be due to a different or lower reactivity (see for instance the end groups). The contributions of initial decomposition and termination reactions are negligible in the overall balance when compared with the chain propagation ones and, therefore, they have been neglected. The balance equations of the species in the liquid phase are characterized by a rst term of disappearance, whereas the ones of gaseous species contain only formation terms. Mass balance equations of styrene and of the trimer show the contributions of unzipping and back biting reactions,
dOI1 = (1-i -k ) %(RPIj + RPIIIj ) + % ROIj dt 2 3

dOI3 dOI3 = dt dt

  
0

+ k % (RPIj + RPIIIj + ROIj )


L+1

Initial conditions are needed to integrate the system of ordinary differential equations. It has been assumed that only PI species are present initially. Initial molecular weight distribution curve is assumed on the basis of Schultz most probable distribution [22], xi = 1 1 1 n n

i1

where xi is the mole fraction of molecules with degree of polymerization i, and n is the average degree of polymerization. The maximum length N is assumed on the basis of a total loss lower than 0.1% [20]. N becomes the upper limit of sums of mass balance equations. The resulting dimension of the overall differential system is 5 N. The solution is obtained after a numerical integration through an implicit multi-step Adams Moulton method [23]. Because of the heavy computing times (about 10-min on a PC) owing to the initial high molecular weights, a lumping procedure has been introduced [24 26]. It is based on the grouping of the longest species into lumps. It is possible to dene the critical length beyond which species are grouped and the number of species of each group. This approach strongly reduces the calculation time without a signicant effect on the predicted results.

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

115

Fig. 3. Isothermal degradation curves for PS. Comparison between experimental data found by Bockhorn et al. [4] and model results.

5. Experimental data and validation of the model In this work, three types of experimental data are compared with the model predictions isothermal data, TGA curves, and gas product distributions. Recent isothermal data at atmospheric pressure are reported by Bockhorn et al. [4]. Their comparison with model results are shown in Fig. 3. The agreement is very good at 360, 400 and 410C. The intermediate isothermal curves at 370, 380 and 390C are slightly underpredicted. Experimental data reported by Bouster et al. [5] are obtained in experimental conditions (temperatures and pressure) similar to those already discussed. They differ mainly in the polymer molecular weight (100 000 instead of 186 000 g mol 1). The comparisons with these data are shown in Fig. 4. Also in this case, the agreement is satisfactory and even better than in the previous example. Fig. 5 shows the comparison between predicted results and isothermal experimental data presented by Madorsky [6]. At 348C, the agreement is very good, but at lower temperatures, the model seems to forecast a faster decomposition. Due to the very low experimental pressure (about 10 5 mmHg), these discrepancies can be justied with only 2 and 4C, respectively, at 338 and 328C (i.e. within experimental uncertainty). Even if the comparison with the experimental data is generally good, the partial observed disagreement is conrmed by the differences between the overall kinetic parameters proposed in literature. The activation energies respectively proposed by Bockhorn, Bouster and Madorsky are 41, 49 and 55 kcal mol 1. None of these

116

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

experimental activation energies is able to cover all the temperature ranges investigated here and only a phenomenological approach can span over all the experiments with the same level of accuracy. The apparent activation energy calculated from the detailed kinetic model presented here is about 47 kcal mol 1. As a further comparison, Fig. 6 shows the model prediction and TG experimental data presented by Anderson and Freeman [7] under a vacuum of 1 mm Hg and a constant heating rate of 5C min 1. The agreement is especially good at the beginning of the degradation (until 390C). In the gure, there is also the curve obtained with the Bockhorns model. As mentioned already, the kinetic model was compared also with the experimental gas product distributions [8]. The experimental results obtained by Audisio and Bertini are reported in Table 2. Styrene is the most important product of the thermal degradation. The model is in quite good agreement with molecular weight effect, both for the monomer yield and also other secondary products. On the contrary, the agreement is not so good with the temperature variation. Moreover, the assumed kinetic mechanism does not explain the formation of benzene and light hydrocarbons, which are observed and measured in some experiments. Two are the possible explanations. From one side, it is possible to have secondary cross-linking reactions with additive substitutions of secondary or tertiary radicals on the different rings. A second explanation can refer to successive gas phase reactions. On the contrary, it is quite difcult to invoke an electrophilic attack on the aromatic ring, because the ionic mechanism is signicant only in the presence of acid catalysts, while the pure thermal degradation is governed by a radical depolymerization [27,28].

Fig. 4. Isothermal degradation curves for PS. Comparison between experimental data found by Bouster et al. [5] and model results.

Table 2 Product distribution (wt.%) from polystyrene pyrolysisa MW = 2100 MX = 7500 MX = 30 000 MX = 110 000 MW = 380 000

Pyrolysis product Experimental Calculated Experimental Calculated Experimental Calculated Experimental Calculated

Experimental Calculated

4.91 2.02 5.93 1.13 79.53 0.76 2.69 1.85 64.73 0.5 1.62 5.14 1.05 80.74 0.64 1.55 0.9 77.13 0.48 1.5 4.65 0.89 82.85 0.61 1.09 0.24 81.33 0.45 1.41 4.38 0.88 83.49 0.6

3.62

2.93

2.04

0.97 0.07 82.4 0.45

1.84 0.99 3.31 0.57 87.1 0.51

0.93 0.02 82.7 0.45

T = 600 C Light hydrocarbons Benzene Toluene Ethylbenzene Styrene a-Methylstyrene 5.93 5.27 7.11 1.6 70.17 1.9 2.15 1.5 65.88 0.38 4.99 5.73 1.22 73.68 1.63 1.1 0.92 82.34 0.23 4.1 5.4 1.13 77.79 1.5 0.59 0.26 88.19 0.2 4.25 3.15 2.75 3.67 4.74 1.01 82.5 1.19 0.45 0.07 89.74 0.19

2.12 2.58 3.81 0.71 85.59 1.11

0.41 0.02 90.17 0.19

T = 750 C Light hydrocarbons Benzene Toluene Ethylbenzene Styrene a-Methylstyrene

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Comparison between model predictions and experimental data of Audisio and Bertini [8].

117

118

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Fig. 5. Isothermal degradation curves for PS. Comparison between experimental data found by Madorsky [6] and model results.

A better agreement was found in comparison with the experimental data proposed by Bouster et al. [9] and presented in Table 3. Toluene prediction agrees in this case with the experimental observations. In these data, the yields of dimer and trimer are also reported. The trends with the temperature and molecular weight are quite good. The relative yields of dimer and trimer are not reproduced correctly, even if their sum matches quite correctly the experimental results. It has to be noted that more experimental information is needed to characterize the model better. For instance, it is quite evident that the increase of the amount of 1,3 diphenylpropane with the temperature cannot be explained with the proposed model. Higher temperatures make easier the transformation of alkane chains in alkenes.

6. Conclusion In this paper, a detailed model of polystyrene thermal degradation has been presented. The model is able not only to describe the weight loss during the process, but overall to predict the gas phase composition. The kinetic parameters are derived from the well-known values proposed already for the gas phase pyrolysis, with the proper modications to be applied in the liquid phase. The results are encouraging even though not as accurate as in the detailed kinetic models of PE and PP. Nevertheless, the proposed model already allows reliable predictions. Some uncer-

Table 3 Product distribution (wt.%) from polystyrene pyrolysisa MW = 2200 Experimental Calculated Experimental Calculated Experimental Calculated MW = 8000 MW = 50 000 MW = 50 000 Experimental Calculated

Pyrolysis products

(a ) Effect of the molecular weight at T = 600 C Styrene 48.2 67.01 Dimer 1.7 1.3 Trimer 5.5 4.4 Toluene 1.4 3.0 1,3-Diphenylpropane 0.6 0.3 69.2 3.1 9.0 1.0 0.7 Trace 0.2 0.8 0.1 Trace 0.3 T = 700C 78.2 0.8 8.6 1.7 0.2 Trace Trace 76.4 3.4 5.0 0.7 1.6 0.1 0.5 85.4 0.3 6.2 1.0 0.1 0.01 Trace Trace Trace 76.3 1.0 6.1 2.1 0.2 72.7 3.2 10.0 0.7 0.6 78.2 0.8 8.6 1.7 0.2 0.1 0.2 2.4 0.3

76.5 3.5 10.5 0.8 0.5 Trace 0.2 T = 800C 78.4 1.9 1.3 0.9 2.0 0.3 0.9

78.2 0.8 8.2 1.7 0.2 0.1 0.001

Ethylbenzene a-Methylstyrene

(b ) Effect of pyrolysis temperature (MW = 100 000 g mol1) Pyrolysis products T = 500C T = 600C 65.6 3.1 12.7 3.0 0.6 0.1 0.01 Trace 0.3 72.7 3.2 10.0 0.7 0.6

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Styrene 66.8 Dimer 4.6 Trimer 15.0 Toluene 0.7 1,3-Diphenylpropane 0.3 Trace 0.2

89.6 0.1 4.5 0.6 0.1 0.01 Trace

Ethylbenzene a-Methylstyrene

Comparison between model predictions and experimental data [9].

119

120

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

Fig. 6. TG curves for PS. Comparison between experimental data of Anderson and Freeman [7] and model results.

tainties are still present like the difculty in predicting benzene formation as observed by some authors. At the same time, the difculties related with the experimental measures, the mass and heat transfer limitations, the possible presence of successive reactions in the gas products and the small amount of reliable data and experiments ask for further investigations. This work on the thermal degradation of poly-styrene adds a further step to the overall characterization of pyrolysis of plastics. Nowadays, polyethylene and polypropylene and polystyrene models are available. The major interest in this research activity is to found an alternative route for the upgrading of solid wastes to more usable and energy-dense materials.

Acknowledgements This work was supported by EU under the HALOCLEANCONVERSION project, contract n. G1RD-CT 1999-00082.

References
[1] E. Ranzi, M. Dente, T. Faravelli, G. Bozzano, S. Fabini, R. Nava, V. Cozzani, L. Tognotti, J. Anal. Appl. Pyrol. 40 41 (1997) 305 319. [2] T. Faravelli, G. Bozzano, C. Scassa, M. Perego, S. Fabini, E. Ranzi, M. Dente, J. Anal. Appl. Pyrol. 52 (1999) 87 103. [3] R. Zevenhoven, M. Karlsson, M. Frankenhaeuser, M. Hupa, Laboratory scale characterization of plastic-derived fuels, Borealis Polymer Oy, Report 95/3, Borga, Finland, 1995.

T. Fara6elli et al. / J. Anal. Appl. Pyrolysis 60 (2001) 103 121

121

[4] H. Bockhorn, A. Hornung, U. Hornung, in: The Twenty-seventh International Symposium on Combustion, University of Colorado, Boulder, 1998. [5] C. Bouster, P. Vermande, J. Veron, J. Anal. Appl. Pyrol. 1 (1980) 297 313. [6] S.L. Madorsky, Thermal Degradation of Organic Polymers, Interscience, New York, 1964. [7] D.A. Anderson, E.S. Freeman, J. Polym. Sci. 54 (1961) 253 260. [8] G. Audisio, F. Bertini, J. Anal. Appl. Pyrol. 24 (1992) 61 74. [9] C. Bouster, P. Vermande, J. Veron, J. Anal. Appl. Pyrol. 15 (1989) 249 259. [10] M.M. Shapi, A. Hesso, J. Anal. Appl. Pyrol. 18 (1990) 143 161. [11] M.T. Sousa Pessoa De Amorim, C. Bouster, P. Vermande, J. Veron, J. Anal. Appl. Pyrol. 3 (1981) 19 34. [12] A. Guyot, Poly. Deg. Stab. 15 (1986) 219 235. [13] L. Costa, G. Camino, A. Guyot, M. Bert, A. Chiotis, Poly. Deg. Stab. 4 (1982) 245 260. [14] L. Costa, G. Camino, A. Guyot, M. Bert, G. Clouet, J. Brossas, Poly. Deg. Stab. 14 (1986) 85 93. [15] M. Swistek, G. Nguyen, D. Nicole, J. Anal. Appl. Pyrol. 37 (1996) 15 26. [16] S. Benson, The Foundations of Chemical Kinetics, McGraw-Hill, New York, 1960. [17] M. Dente, G. Bozzano, M. Rossi, in: Proceedings of the First Conference on Chemical and Process Engineering, Florence, Italy, 1993, pp. 163 172. [18] M. Dente, G. Bozzano, G. Bussani, Comp. Chem. Eng. 21 (1997) 1125 1234. [19] D.L. Van Krevelen, Properties of Polymers, Elsevier, Amsterdam, 1990. [20] M. Pinciroli, F. Pisano, Degradazione Termica del Polistirene, Thesis, Politecnico di Milano, 1999. [21] O.A. Hougen, K.M. Watson, R.A. Ragatz, Chemical Process Principles, Wiley, New York, 1954. [22] U.W. Gedde, Polymer Physics, Chapman & Hall, London, 1995. [23] A. Hindmarsh, in: R.S. Stepleman, et al. (Eds.), ODEPACK, A Systematized Collection of Ode Solvers in Scientic Computing, North-Holland, Amsterdam, 1983, pp. 55 64. [24] P.G. Coxson, K.B. Bischoff, Ind. Eng. Chem. Res. 26 (1987) 1239 1248. [25] P.G. Coxson, K.B. Bischoff, Ind. Eng. Chem. Res. 26 (1987) 1251 1257. [26] J. Wei, J.C.W. Kuo, Ind. Eng. Chem. Fundam. 8 (1969) 114 123. [27] P. Carniti, A. Gervasini, P.L. Beltrame, G. Audisio, F. Bertini, Appl. Catal. A 127 (1995) 139 155. [28] R. Lin, R.L. White, J. Appl. Polym. Sci. 63 (1997) 1287 1298.

You might also like