Etienne
Grau
,
Jean-Pierre
Broyer
,
Christophe
Boisson
,
Roger
Spitz
and
Vincent
Monteil
*
Université de Lyon, Univ. Lyon 1, CPE Lyon, CNRS, UMR 5265 Laboratoire de Chimie Catalyse Polymères et Procédés (C2P2), LCPP team Bat 308F, 43 Bd du 11 novembre 1918, F-69616, Villeurbanne, France. E-mail: monteil@lcpp.cpe.fr; Fax: +33 4 72 43 17 68
First published on 28th July 2011
Ethylene polymerization is performed industrially either by radical polymerization under severe conditions (1000–4000 bar, 200–300 °C) or by a catalytic mechanism at lower temperatures (usually less than 100 °C) and pressures (below 50 bar). Standard radical polymerization conditions are too severe to permit a fine control of the macromolecular architecture. Under milder conditions (100 bar, 70 °C), radical ethylene polymerization is assumed ineffective which has been confirmed in bulk (supercritical ethylene). However, we have shown that the efficiency of this polymerization is strongly dependent on the solvent. This unusual activation by solvent has been rationalized using theoretical considerations. A second effect investigated is the influence of solvent on PE molecular weight. Indeed PE with either low molecular weight and high chain-end functionality or higher molecular weight can be synthesized according to the solvent used.
Free radical polymerization of ethylene is industrially conducted under high pressures (1000–4000 bar) and high temperatures (200–300 °C) in bulk. LDPE melts between 100 and 120 °C and exhibits a crystallinity in the range of 30–60%. We have investigated the efficiency of this reaction under milder slurry conditions: pressure up to 250 bar, T = 70 °C.5,6 These conditions are less efficient than the industrial process but easier to handle. The PE produced possesses higher crystallinities (∼70%) than the traditional industrial LDPE but molecular weights remain too low for applications (Mn < 5000 g mol−1). Moreover the influence of the solvent appears to be crucial for the free radical ethylene polymerization. For instance we managed to synthesize polyethylene down to 10 bar in a polar solvent (THF) by a free radical process with activity 6 times higher than in a nonpolar solvent (toluene).5 This unusual activation by solvent will be further studied in the present work. Indeed free radical polymerization of ethylene will be investigated in a wide range of solvents in order to improve the medium pressure process.
The influence of solvent on free radical polymerization of vinyl compounds was previously reported by Kamachi.7 For almost all monomers it is a tiny effect except for vinyl acetate8 and ethylene.9–12 For these two monomers the kinetics of radical polymerization could vary by a factor up to 10 depending on the solvent. The early studies for ethylene remain partial due to the experimental facilities available at this time (before 1980s). Machi13 suggested that the solubility of the growing polyethylene chain could induce the activation effect by solvent through a Trommsdorff–Norrish effect but his interpretation is still controversial.14 Myshkin8 assumed it was a fully different mechanism based on the dielectric constant ε of the solvent. For other monomers several explanations for this influence have been proposed but none of them is consistent with all sets of data. The diverse influences of solvent on the free radical polymerization are due to variations of kinetic constants according to solvent parameters. The termination rate15–17 was partially related to the viscosity of the solvent due to diffusion mechanisms. For the propagation rate,7,18,19 different origins have been proposed as polarity, interactions between polymer and solvent, interactions between monomer and solvent, and complexation between the propagating macroradical and the solvent. Other authors20–22 suggested that local monomer concentration could play a major role. In the present paper, we performed the radical polymerization of ethylene in a wide range of solvents and studied their influences on yield of polymerization and molecular weight of produced polyethylene. Thanks to theoretical studies, we suggest a new relation which links the yield of polymerization to a combination of key solvent parameters.
Ethylene polymerizations were done in a 160 mL stainless steel autoclave (equipped with safety valves, stirrer, oven) from Parr Instrument Co. The azobisisobutyronitrile (AIBN) was dissolved in 50 mL of desired solvent in a Schlenk tube under argon. The mixture was introduced through cannula into the reactor. Ethylene was introduced and the mixture was heated at the desired temperature under stirring (300 rpm). To manage safely polymerization over 50 bar of ethylene we use a 1.5 L intermediate tank. The tank was cooled down to −20 °C to liquefy ethylene at 35 bar. When thermodynamic equilibrium was reached, the intermediate tank was isolated and heated to reach up to 300 bar of ethylene pressure. This tank was used to charge the reactor and maintain the pressure of ethylene constant in the reactor by successive manual ethylene addition. After 4 hours of polymerization the reactor was slowly cooled down and degassed. The precipitated polymer was then dried under vacuum at 70 °C.
Runa | Solvent | Dielectric constant (ε) (at 20 °C)b | Dipole momentum (μ) (10−30 C m)b | Yield/g | Melting pointc/°C [crystallinity (%)] | M n d/g mol−1 | PDId |
---|---|---|---|---|---|---|---|
a The reactions were carried out using 50 mg of AIBN in 50 mL of purified solvent at 70 °C under 100 bar of ethylene pressure during 4 hours. b Obtained from ref. 24. c Determined by DSC. d Determined by HTSEC, nd = not determined. | |||||||
1 | Supercritical ethylene | — | 0 | 0.1 | 105.3 [46] | 3010 | 1.3 |
2 | Cyclohexane | 2.0 | 0 | 0.6 | 115.5 [58] | 4800 | 2.2 |
3 | Heptane | 1.9 | 0 | 0.65 | 116.7 [55] | 4700 | 2.1 |
4 | Toluene | 2.4 | 1 | 0.7 | 115.9 [63] | 2340 | 1.9 |
5 | Dimethylsulfoxide | 46.4 | 13.5 | 1 | 112.7 [43] | 1910 | 3.5 |
6 | Acetonitrile | 35.9 | 11.8 | 1.1 | 115.5 [59] | 1370 | 2.2 |
7 | Diethylcarbonate | 2.8 | 3.7 | 1.2 | 117.8 [62] | 7150 | 2.5 |
8 | N,N-Dimethylformamide | 36.7 | 10.8 | 1.3 | 108.5 [47] | 530 | 2.9 |
9 | Dibutylether | 3.1 | 3.9 | 1.3 | 109.0 [52] | 1370 | 1.4 |
10 | Ethanol | 24.5 | 5.8 | 1.4 | 117.6 [63] | 2130 | 2.4 |
11 | Acetone | 20.6 | 9 | 1.5 | 115.2 [62] | 1710 | 2.0 |
12 | Dimethylcarbonate | 3.2 | 3.7 | 1.6 | 117.9 [57] | 11720 | 2.5 |
13 | Butanone | 18.5 | 9.2 | 1.8 | 61 [nd] | 370 | 1.2 |
14 | Butyrolactone | 39.0 | 14.2 | 1.8 | nd [nd] | 570 | 1.4 |
15 | Butan-2-ol | 16.6 | 5.5 | 1.9 | 116.4 [68] | 2070 | 2.8 |
16 | Cyclohexanone | 16.0 | 10.2 | 2.1 | nd [nd] | 1760 | 1.5 |
17 | Butan-1-ol | 17.5 | 5.8 | 2.2 | 117.8 [58] | 4130 | 2.4 |
18 | Ethyl acetate | 6.0 | 6.1 | 2.3 | 115.2 [54] | 3760 | 3.3 |
19 | Dichloromethane | 8.9 | 5.2 | 2.7 | 105.1 [46] | 1050 | 1.6 |
20 | 1,4-Dioxane | 2.2 | 1.5 | 3.2 | 118.9 [65] | 1300 | 2.2 |
21 | Tetrahydrofuran | 7.6 | 5.8 | 3.9 | 115.2 [58] | 1190 | 1.9 |
In pure ethylene, without any solvent, the polymerization of ethylene is almost inefficient (run 1, 0.1 g corresponding to a conversion of 0.3%). The resulting polyethylene exhibits a low melting point, 105 °C, close to LDPE thermal properties, and a low molecular weight (3000 g mol−1). In all cases, a higher activity is measured in the presence of solvent. Therefore solvents seem to play a major role in the activation of the radical polymerization of ethylene. However, all solvents did not lead to the same activation (yield ranges from 0.6 g in cyclohexane to 3.9 g in THF) and in first approximation, it appears that nonpolar solvents are less efficient than polar ones.
Solvent with high transfer ability can be used to obtain a high content of PE chain-end functionality. One of the most transferring solvents is THF, 26 times higher than cyclohexane. Transfer to solvent provided THF-ended polyethylenes which were fully identified by 13C NMR (see ESI†, Fig. S1).5 The 13C NMR spectrum exhibits standard chemical shift26 of a branched polyethylene (9 branches/1000C). Additionally two different structures, 1- and 2-polyethylenyl-THF, were identified. The chain-end functionalized PE obtained from transfer to solvent may be used further as macro-monomers. Chain-end labelling from transfer to solvent was also determined by 13C NMR for other solvents such as toluene, dioxane (run 20), dichloromethane (run 19, see ESI†, Fig. S2–S4).
An interesting solvent for further use of chain-end functionalized PE is butyrolactone (run 14). The resulting macromonomer could be used in copolymerization with lactonesviaring-opening polymerization in order to obtain polyester with PE branches. Otherwise, chloro-terminated PE from transfer to dichloromethane could also be used as starting reactant for nucleophilic substitution.
On the other hand, solvents with low transfer ability and high activity will provide non-functional polyethylenes. Usually, high transfer ability is related to high activities, except for ethyl acetate (run 18) which is a particularly poor transferring solvent (but still 4.9 times more transferring than cyclohexane), for butan-1-ol (run 17), and for carbonates (runs 7 and 12). These solvents provide PE with relatively higher molecular weights (over 10000 g mol−1).
Consequently, free radical polymerization of ethylene in solvent can provide either, functional/low-molecular weight polyethylenes or non-functional/higher-molecular weight polyethylenes.
To quantify the solvent effect we used the theory of the activated complex27 (eqn (1)) which links a kinetic constant in a solvent to a kinetic constant without a solvent. In this theory, the solvent effect is due to the preferential interactions between the solvent and the activated complex or the reactants. In the case of the free radical polymerization of ethylene, this stabilization is only due to van der Waals interactions, that is, Keesom (dipole–dipole), Debye (dipole–induced dipole) and London (instantaneous dipole–induced dipole) interactions28 (eqn (2)–(4)). Keesom interactions are assumed to be the principal interactions which stabilize the macroradical (EKeesom > EDebye and ELondon). Consequently, each kinetic constant of the polymerization (kd, kp, kt or ktot) exhibits a relation (eqn (5)) with different solvent properties (ε, μ). Thus, according to the free radical kinetic law (eqn (6)), ktot and therefore yield can be related to (μ/ε)2 (eqn (7)). Since only ktot is accessible in this experiment we cannot determine the relative dependency of kp, kt, kd with (μ/ε)2. However it is expected that kt is only slightly influenced by solvent properties as it is mostly controlled by diffusion processes.16,17
It should be noted that the transfer to solvent usually does not impact the kinetic of the polymerization. Indeed the transfer reaction does not modify the radical concentration (Steady States Approximation). However in specific cases the resulting radical cannot efficiently react with monomer and consequently slows down the polymerization. It appears not to be the case for ethylene polymerization as for example THF and dibutylether lead to similar radical (O–CH–CH2) but different activities (runs 9 and 21).
(1) |
(2) |
(3) |
(4) |
(5) |
(6) |
(7) |
We plotted the conversion versus (μ/ε)2 (Fig. 1), in order to confirm this relation.
Fig. 1 Solvent influence due to Keesom interactions on radical polymerization of ethylene (labels correspond to run numbers in Table 1). ■: 50 mg AIBN, 50 mL of solvent 4 h at 70 °C under 100 bar of ethylene pressure. |
The curve obtained is unexpectedly Λ-shaped. A change of behavior is observed for Keesom interactions higher than the ones for THF ((μ/ε)2optimum ≈ 0.58 × 10−60 C2 m2). At lower value of (μ/ε)2, yield increases with this parameter, while it decreases over this value. Most of the solvents showed a good correlation between polymerization yield and (μ/ε)2.
For alcohols (runs 10, 15 and 17) such as ethanol an “over yield” is observed. This can be due to H-bond interactions which have been neglected in the theory. Indeed stabilization by solvent can take place not only by van der Waals interactions but also by H-bond interactions in this case.
Fig. 2 Solvent mixture composition influence related to Keesom interactions on radical polymerization of ethylene. 50 mg AIBN 4 h at 70 °C under 100 bar of ethylene pressure in 50 mL of ■: THF–toluene; : THF–DEC; : toluene–DEC; : THF–toluene–DEC mixture. |
Standard mixing rules29 respectively for relative permittivity εmixture = Σxiεi, with xi the volume fraction of solventi and εi the solventi relative permittivity and for dipole μmixture = ΣΣxixj(μiμj)1/2, with μi the dipole momentum of the solventi are used.
In all cases, whatever the mixture composition, the Λ-shaped curve is observed once again between conversion and values of (μ/ε)2. The maximum of activity (yield 4.1 g) was reached for (μ/ε)2optimum ≈ 0.65 × 10−60 C2 m2. Polymerizations in toluene–DEC mixture follow the same curve than toluene–THF and THF–DEC mixtures. So by tuning properly the proportion of toluene–DEC mixture we are able to provide the same activity than the ethylene polymerization in THF. This evidences that the solvent interaction with the alkyl radical is an exact average of the solvent composition and is not due to the nature of the solvent itself. In other words the solvation shell of the alkyl radical presents the same composition as that of the solvent composition, so there are not preferred interactions with one of the solvents.
Using solvent mixtures, conversion higher than in THF can be reached. Moreover the activation by the solvent appears to be uncoupled from the control of molecular weight by transfer to solvent as evidenced by the molecular weights of synthesized polyethylenes (see ESI†, Table S1). For example, a toluene/DEC 50/50 v/v mixture provides about the same polymerization activity as THF but does not lead to the same molecular weight: respectively 3200 g mol−1 and 1200 g mol−1. Therefore as the solvent activation effect is a global solvent effect only related to μ and ε, and molecular weight mostly controlled by the nature of the solvent (or solvent mixture) used, yield and average molecular weight can be tuned easily by choosing a suitable mixture of solvents.
Solvent | (μ/ε)2/10−60 C2 m2 | Global activation energy (Etot)b/kJ mol−1 | Global preexponential factor (ln (Atot))b |
---|---|---|---|
a Assuming the validity of Arrhenius law. b Determined by a linear regression from 15 experiments at 50, 70 and 90 °C and for each temperature at 50, 100, 150, 200, 250 bars. | |||
Toluene | 0.18 | 27.7 | 7.6 |
THF | 0.58 | 32.8 | 10.3 |
DEC | 1.72 | 40.0 | 12.2 |
Ideally, the determination of the Arrhenius parameters for each polymerization step should be performed, but this kind of study is currently incompatible with our experimental conditions of pressure (stopped flow or pulsed laser polymerization techniques cannot be easily used).
Both global activation energy and pre-exponential factor increase with (μ/ε)2. On one hand, lower global activation energy is usually linked to a more favorable reaction. In all solvents, the polymerization mechanism is considered to be the same, so the change in the global activation energy is only due to the relative stabilization of intermediate and activated states, which differ from one solvent to the other.30 Solvation by toluene provides a lower energy barrier than in THF and DEC. On the other hand, the global pre-exponential factor is proportional to the frequency of efficient collisions. With a higher pre-exponential factor the probability of the mechanism involved is supposed to increase. Differences in geometry of activated states26 in toluene, in THF and in DEC could explain the difference of pre-exponential factors. Toluene is less electron donor than THF, more toluene molecules are necessary to stabilize the radical. Thus the radical should have a denser solvation shell in toluene than in THF. This could explain why preexponential factor is higher in THF than in toluene. The same reason could be applied for DEC. For these three solvents, a linear relationship seems to exist between Etot and (μ/ε)2, in the same way ln (Atot) vs. (ε/μ)2 is linear. These two relations allow to calculate the Arrhenius parameters for every (μ/ε)2 and to predict the optimum of solvent activation. The optimum depends on the temperature (in Kelvin the relation is (μ/ε)2optimum ≈ 0.03 T1/2 × 10−60 C2 m2).31 Therefore at 70 °C the predict optimum is (μ/ε)2optimum ≈ 0.56 × 10−60 C2 m2.
Otherwise the optimum corresponds to the radical stemming from monomer, not to the radical fragment issued from AIBN. Therefore the initiation (first addition of the monomer) is not the determining step in the solvent activation effect. Indeed if it was the case the optimum (μ/ε)2 would correspond to the (μ/ε)2 ≈ 0.02 × 10−60 C2 m2 of the radical resulting from AIBN dissociation (μ ≈ 1.6 × 10−30 C m and ε ≈ 25—see ESI†, Fig. S6 and Table S4). Consequently it must be the propagation and/or termination steps which are influenced by the solvent.
It should be noted that for standard monomers (MMA, Sty, BuA), the solvent effect remains tiny.7,10 These monomers possess higher propagation rate and lower termination rate than the monomers (ethylene, vinyl acetate) which exhibit a solvent activation effect. So only monomers, which possess low propagation rate or high termination rate, seem to express a high solvent effect.
Moreover, the crucial importance of solvent transfer capacities on the nature of synthesized PE was evidenced. Indeed since the alkyl radical possesses a high reactivity, the transfer constants to solvents are high and the molecular weight of the PE synthesized is controlled by transfer to solvent. This transfer to solvent can be used to functionalize PE chain-end. Carbonates are the less transferring solvents and Mn values up to 15000 g mol−1 are reached. This molecular weight far over the entanglement molecular weight should lead to some attractive mechanical properties. Activation by solvent and molecular weight control by transfer appear to be unlinked.
We observe that the major factor to explain the solvent activation effect is the Keesom interaction of the solvent on the macroradical and therefore demonstrated a good correlation between yield of polymerization and solvent properties ((μ/ε)2). This Λ-shaped relation demonstrates that an optimal solvent exists for ethylene radical polymerization and possesses a (μ/ε)2 close to 0.6 × 10−60 C2 m2. The same optimum has been identified using two other techniques: solvent mixtures and Arrhenius parameters. Moreover, this optimum is correlated to the corresponding (μ/ε)2 of the propagating radical. Finally, we have demonstrated that this solvent activation effect is not a punctual effect of the solvent but a global average effect of the solvation shell of the growing radical and is only dependent on the average μ and ε of the solvent mixture.
Footnote |
† Electronic supplementary information (ESI) available: Text detailing calculations; NMR spectra of PE synthesized in THF, toluene, dioxane and dichloromethane; tables showing the influence of the mixture composition and the temperature on the free radical polymerization of ethylene, and GAUSSIAN 03 output for 1-hexyl radical and AIBN radical fragment. See DOI: 10.1039/c1py00200g |
This journal is © The Royal Society of Chemistry 2011 |