Clement G.
Collins Rice
,
Louis J.
Morris
,
Jean-Charles
Buffet
,
Zoë R.
Turner
and
Dermot
O'Hare
*
Chemistry Research Laboratory, Department of Chemistry, University of Oxford, 12 Mansfield Road, Oxford, OX1 3TA, UK. E-mail: dermot.ohare@chem.ox.ac.uk
First published on 6th December 2023
Using a highly active permethylindenyl-phenoxy (PHENI*) titanium catalyst, high to ultra-high molecular weight ethylene–linear-α-olefin (E/LAO) copolymers are prepared in high yields under mild conditions (2 bar, 30–90 °C). Controllable, efficient, and predictable comonomer enchainment provides access to a continuum of copolymer compositions and a vast range of material properties using a single monomer-agnostic catalyst. Multivariate statistical tools are employed that combine the tuneability of this system with the analytical and predictive power of data-derived models, this enables the targeting of polyolefins with designer properties directly through predictive alteration of reaction conditions.
Olefin polymerisation systems are typically high-complexity multicomponent systems. Unpredictable and nonlinear relationships exist between reaction conditions and the end composition and properties of the polymer. Furthermore, enchainment of ethylene is normally greatly preferred over LAOs.33,34 Thus, the targeted synthesis of an LLDPE material with pre-defined properties is non-trivial and reliant upon extensive screening and empirical optimisation of catalyst and reaction conditions. Klosin et al. recently noted that no single catalyst is able to fulfil the requirements of the increasingly broad range of polyolefin products, and highlighted the desirability of “programmable and predictable polymerisation performance”.35 A single catalyst that is efficiently able to copolymerise a range of olefins to high molecular weight polymers with readily and predictably tuneable composition is elusive, remaining a highly desirable industrial goal.
Traditional heterogeneous Ziegler–Natta catalysts have been used in the copolymerisation of ethylene with LAOs with high activities but relatively limited comonomer incorporation.36,37 Metallocenes have shown good homogeneous catalytic activity towards E/LAO copolymerisation when activated with methylaluminoxane (MAO), with a more uniform comonomer distribution than is possible with heterogeneous systems.38–40
With the development of post-metallocene catalysts such as the Constrained Geometry Complexes (CGCs),41 it was shown that ligands with electron-donating substituents had increased activities and afforded copolymers with increased molecular weights and high comonomer incorporations, up to 25 mol% C8.35,42 Quijada et al. showed that ansa-bridged metallocenes enabled higher LAO incorporation than unbridged analogues.40 Group four Phenoxy-Induced Complexes of Sumitomo (PHENICS) utilising Cp-,43,44 Ind-,45 and Flu-derived46 apical ligands have demonstrated LAO incorporations of up to 35 mol%, reported with the complex {(η5-Cp)Me2C(PhO)}TiCl2.44 Compared to CGCs, PHENICS complexes were found to result in greater E/C6 copolymerisation activities as well as high comonomer incorporation.44,45
Irwin et al. reported the “sterically expanded” complex Me2SB(Oct,tBuN)ZrCl2·OEt2 ({(η1-C29H36)Me2Si(tBuN)}ZrCl2·OEt), based on a monohapto octamethyloctahydrodibenzofluorenyl (Oct) ligand, which is the only example of a catalyst to demonstrate activities and C8 incorporations proportional to comonomer concentration.47 Initial activities as high as 81000 kgLLDPE molZr−1 h−1 bar−1 and incorporations of 75 mol% were reported for reactions conducted in neat 1-octene (6400 mM), for 50 seconds under 5.5 bar ethylene pressure at 75 °C. In this case an “unyielding comonomer effect” describes the phenomenon of increasing activities with comonomer concentration, and is attributed to the dominant electronic advantages of substituted α-olefins for an electron rich “sterically indiscriminate” complex.48 While this result is remarkable, the steric expansion also rendered termination processes facile, resulting in very low polymer molecular weights (Mw < 7 kDa),49 severely limiting the industrial utility of this catalyst. Varying the Oct ligand substituents was able to increase Mw but at the expense of activity and control, with Đ > 100 reported.49 Furthermore, the ease with which this complex reinserts olefinic macromonomers leads to long chain branching, reducing the predictability of the system.50,51
Single-site catalysts of sufficiently high activity towards LAO enchainment may allow access to a largely diffusion-limited monomer-agnostic regime, where copolymer composition is simply determined by the relative concentrations of comonomer. Recently, we reported solid-supported group four permethylindenyl-phenoxy (PHENI*) catalysts, which display outstanding performance for ethylene and propylene polymerisation, producing ultrahigh molecular weight homopolymers,52,53 and tuneable copolymers.54 In the latter case, ethylene–propylene copolymers were produced with polymer composition almost exactly corresponding to the feed ratio of gaseous monomers.
In this work we describe the application of the PHENI* complex Me2SB(tBu2ArO,I*)TiCl2 ({(η5-C9Me6)Me2Si(2,4-tBu2-C6H2O)}TiCl2), supported on solid MAO (1; Fig. 1), for the copolymerisation of ethylene with LAOs. Virtually unprecedented levels of single-catalyst tuneability allowed statistical modelling to be applied as a predictive tool in the targeted synthesis of designer polyolefins.
n | [LAO]/mM | T p/°C | Activity/kgLLDPE mol−1 h−1 bar−1 | M w/kDa | Đ | x LAO (wt%) | T m/°C | α (%) | |η*|/kPa s |
---|---|---|---|---|---|---|---|---|---|
— | 0 | 60 | 3720 | 2088 | 5.2 | 0 | 133 | 78 | 280 |
8 | 20 | 60 | 2730 | 1619 | 4.2 | 2.1 | 123 | 63 | 300 |
8 | 40 | 60 | 2950 | 1231 | 4.7 | 3.3 | 116 | 45 | 220 |
8 | 80 | 60 | 4180 | 632 | 4.2 | 7.3 | 115 | 40 | 150 |
8 | 120 | 60 | 2820 | 206 | 3.0 | 17.1 | 101 | 27 | 59 |
8 | 160 | 60 | 2250 | 183 | 3.1 | 26.7 | 91 | 10 | 43 |
8 | 320 | 60 | 2960 | 166 | 2.5 | 51.3 | — | Amorph. | 27 |
8 | 640 | 60 | 5880 | 152 | 2.5 | 70.1 | — | Amorph. | 2 |
8 | 80 | 30 | 3060 | 1262 | 4.3 | 10.1 | 115 | 34 | n.d. |
8 | 80 | 40 | 4490 | 933 | 3.9 | 10.2 | 111 | 34 | n.d. |
8 | 80 | 50 | 3800 | 885 | 4.5 | 7.0 | 108 | 32 | n.d. |
8 | 80 | 70 | 2440 | 209 | 4.2 | 14.3 | 115 | 46 | n.d. |
8 | 80 | 80 | 2420 | 167 | 3.4 | 24.3 | 85 | 22 | n.d. |
8 | 80 | 90 | 2010 | 133 | 3.2 | 23.5 | 54 | 18 | n.d. |
6 | 50 | 60 | 4220 | 1071 | 4.9 | 5.0 | 113 | 20 | 260 |
6 | 100 | 60 | 2540 | 182 | 3.0 | 15.3 | 82 | 22 | 40 |
6 | 400 | 60 | 3040 | 162 | 2.7 | 46.5 | — | Amorph. | 23 |
12 | 56 | 60 | 6000 | 980 | 5.1 | 3.2 | 117 | 36 | 240 |
12 | 113 | 60 | 3180 | 206 | 3.1 | 22.5 | — | Amorph. | 39 |
12 | 451 | 60 | 7930 | 177 | 2.5 | 49.3 | — | Amorph. | 1 |
Williams et al. have recently reported activity enhancements in E/C6 copolymerisation activity when using the permethylindenyl (η5-C9Me6,I*) complex Me2SB(tBuN,I*)TiCl2 compared to the analogous Cp*-based CGC Me2SB(tBuN,Cp*)TiCl2, though lower comonomer incorporations were reported for the I* complex.56 Copolymerisation activity of 1 at Tp = 70 °C [C6] = 50 mM (4000 kgLLDPE molTi−1 h−1 bar−1) is comparable to the I* CGC (4400 kgLLDPE molTi−1 h−1 bar−1) and greater than that of Cp* CGC (2260 kgLLDPE molTi−1 h−1 bar−1) when supported on sMAO and tested under identical conditions.65 Albeit in less comparable conditions, the indenyl PHENICS complex Me2SB(tBu,MeArO,Ind)TiCl2/TIBA/[PhNMe2H][BArF4] has been reported with a moderate solution-phase activity of 8700 kgLLDPE molTi−1 h−1 bar−1 ([C6] = 96 mM; Tp = 40 °C; 6 bar).45 In preliminary high-throughput screening, we found a 19-fold increase in activity with 1/TIBA compared to the indenyl complex sMAO–Me2SB(tBu,MeArO,Ind)TiCl2/TIBA (3004 c.f. 157 kgLLDPE molTi−1 h−1 bar−1; [C6] = 400 mM, Tp = 40 °C, 8.3 bar; Table S1†).
Compared with the sterically expanded zirconocene reported by Irwin et al.,471/TIBA does not show linearly increasing activity as a function of [LAO] but notable enhancements are seen at dramatically lower concentrations, suggestive of a uniquely “monomer-agnostic” catalyst.
Using the Fineman–Ross method,67 and taking the concentration of ethylene in hexanes according to Kissin's equation, for partial pressure, p, in bar and absolute temperature T,68 it is possible to estimate the reactivity ratios (Fig. S28–S30†). At Tp = 60 °C, and including only results with soluble polymers (for a consistent kinetic regime) estimates were calculated: for E/C6 copolymerisation, rE = 10 and rC6 = 0.26 (rE × rC6 = 2.7); for E/C8, rE = 17 and rC8 = 0.73 (rE × rC8 = 13); and for E/12, rE = 13 and rC12 = 0.03 (rE × rC12 = 0.42). This is consistent with the formation of random (rE × rC6 ≈ 1) E/C6 copolymers, and blocky (rE × rC8 > 1) E/C8 copolymers as observed by 13C NMR spectroscopy.69 This may have interesting implications in the material properties, since olefinic block copolymers having hard and soft segments are highly desirable thermoplastic elastomers and compatibilisers.70–73 Moreover, the relatively small values of rE and large values for rLAO (c.f. (SBI)ZrCl2rE = 25, rC6 = 0.016;74 Dow-Exxon CGC rE = 7.90, rC8 = 0.10)75 are indicative of PHENI* having a very high comonomer affinity. Indeed, Irwin's “sterically expanded” fluorenyl-derived CGC displayed a rC10 value of 0.49, which was claimed by the authors to be the largest for any rα-olefin reported to date.49 Notwithstanding the assumptions of the Fineman–Ross method, and statistical uncertainties in the linear regression, there is strong evidence to suggest that the PHENI* catalyst 1 has a very high comonomer affinity, with rLAO comparable to or greater than the “sterically expanded” complex, and among the highest of previously reported molecular polyolefin catalysts.
Of the available data for slurry-phase sMAO-supported catalysts tested under the same conditions as 1, both the classic Cp* CGC and its I* analogues incorporate 1-hexene far less efficiently that PHENI*. At Tp = 70 °C and [C6] = 100 mM, incorporations of 3.5 mol% (sMAO–Me2SB(tBuN,Cp*)TiCl2), 2.4 mol% (sMAO–Me2SB(tBuN,3-EtI*)TiCl2), and 2.1 mol% (sMAO–Me2SB(tBuN,I*)TiCl2) are compared with 6.6 mol% (1) in the current study.65 In comparison to reported high comonomer efficiency titanium aryloxide and ketimide post-metallocenes, Nomura et al. reported up to 43 mol% C6 incorporation using Cp*TiCl2(ODipp) ([C6] = 1450 mM, 5 bar; rE = 2.29, rC6 = 0.13), up to 40 mol% with Cp*TiCl2(N=CtBu2) ([C6] = 2010 mM, 4 bar; rE = 6.1, rC6 = 0.085), and up to 60 mol% using the CGC complex (Me2SB(tBuN,Cp*)TiCl2) ([C6] = 1450 mM, 5 bar; rE = 3.42, rC6 = 0.29).76,77 Dankova et al. reported a series of unbridged 2-arylindenyl metallocenes with high comonomer selectivity, with bis(3′,5′-di-tert-butyl-2-phenylindenyl)HfCl2 incorporating 1-hexene as well as the CGC, up to 48 mol% in neat 1-hexene ([C6] = 8000 mM, 13 bar).78
The virtually linear dependence of incorporation on LAO concentration is suggestive that the rate of insertion of both monomers is greater than their rates of diffusion to the active site at 2 bar pressure. Such monomer-agnostic diffusional control allows for much greater comonomer incorporations at lower initial concentrations than are commonly seen in the literature, up to 67 wt% (40 mol%) C6 at 800 mM, 70 wt% (37 mol%) C8 at 637 mM, and 49 wt% (14 mol%) C12 at 451 mM. By contrast, forcing conditions (low ethylene pressure and neat 1-hexene) are typically required to produce comparably-incorporated LLDPE using existing high efficiency catalysts such as bis-indenyl hafnocenes and CGC.78 In addition to this ability to incorporate very large quantities of comonomer, it also enables a wide scope for facile tuneability of intermediate incorporations.
For a given concentration, molar comonomer enchainment was found to be most efficient for 1-octene which is attributed to the slightly greater electron density compared with 1-hexene. At Tp = 60 °C incorporations of 22, 21 and 14 mol% are reported at concentrations of 400, 319, and 451 mM respectively for C6, C8, and C12. The relatively less efficient incorporation of 1-dodecene than either C6 or C8 is indicative of steric constraints becoming influential. At lower temperatures (Tp ≤ 50 °C), minimal C12 incorporations are observed, consistent with a sterically-induced increased energy barrier to comonomer insertion. However, the greater molecular weight of the C12 monomers results in a larger wt% incorporation for a given mol% which is likely to result in more significant perturbations to the physical polymer properties.
In addition to the overall comonomer incorporation, the branching density, calculated from GPC-IR as SCB/1000TC, provides an insight into the distribution and uniformity of polymer composition as a function of molecular weight. In this case, the traces are consistent with uniform copolymer composition and sample homogeneity, typical of well-controlled single-site molecular catalysts (Fig. S10–S12†).66,79
The physical and thermal properties of the synthesised LLDPEs depend on Mw and LAO incorporation, and therefore indirectly on the synthesis conditions. The increase in short-chain branching density afforded by LAO incorporation is expected to reduce the efficiency with which the polymer chains can pack together, reducing crystallinity and decreasing the melting point of the LLDPE relative to HDPE.29,81 As LAO incorporation increases, both the melting point (Tm) and crystallinity (α) of the LLDPE decreases, eventually forming amorphous materials that exhibit elastomeric or gel-like characteristics (Fig. 2d and S13–S16†). The UHMWPE homopolymer synthesised at 60 °C had a Tm of 133 °C and crystallinity of 78%, which is reduced to 91 °C and 10% respectively at a [C8] concentration of 159 mM, corresponding to an incorporation of 26.7 wt% (by 13C NMR spectroscopy). At still greater incorporations, above 40 wt%, LLDPE was found to be amorphous by DSC. By rheology it was shown that the melt-phase viscosity decreases as branching content increases (Fig. 2e and S17–S19†). When measured at 160 °C and an angular frequency, ω, of 1 rad s−1, UHMWPE had a complex viscosity, |η*|, of 2.8 × 105 Pa s, decreasing to 4.3 × 104 Pa s at a C8 incorporation of 26.7 wt% (α = 10%) and 2.3 × 103 Pa s at 70.1 wt% (Fig. 2e). This can also be related to an increase in tan(δ) as comonomer incorporation increases, with tan(δ) > 1 indicative of liquid-like behaviour dominated by viscous flow. This shows that beyond the amorphous limit of DSC detection, increasing SCB content continues to modify the physical properties of LLDPE by enhancing chain mobility.
Notably, for a given incorporation, there appears to be little correlation between either the thermal or rheological properties and the identity of the LAO comonomer (Fig. 2f and S20†).26,82 In this study, the SCB are of insufficient length for side chain crystallisation effects to become apparent.12 The physical properties of the copolymers are principally a function of the overall incorporation rather than any one experimental factor. This allows independent control of otherwise highly coupled parameters such as Mw and Tm which both depend on Tp and [LAO]. As a result, the PHENI*/E/LAO system offers virtually unprecedented tuneability across an extremely wide scope of LLDPEs of varying chemical, thermal, and physical properties. Though beyond the scope of this study, the reduction in crystallinity and melting point is expected to have a profound effect on many other physical and mechanical properties of the polymers.83,84 The ability to access a wide range of incorporations and physical characteristics from a single catalyst system is of significant industrial relevance.
Modelling of ethylene polymerisation systems generally focuses on quantitative structure–activity relationships (QSAR) informed by density functional theory (DFT) calculations on the precatalysts.85–87 This can give useful chemical and mechanistic insights, and inform future catalyst development.88 Beyond fundamental chemical research, the ultimate goal of such methodology is to construct a model that is able to predict experimental parameters to tune polymerisation properties. In real-world systems, many factors beyond precatalyst QSAR are significant, such as morphology, shear forces, diffusion, mixing, impurities, scavengers, and cocatalysts. Furthermore, the structures of components such as MAO and sMAO are not fully elucidated,89,90 and even the oxidation state of the active species is debated.91 Recently, models using multiple linear regression have been proposed in combination with large datasets obtained by high-throughput experimentation with reasonable explanatory and predictive abilities.92
Multivariate data analysis was carried out on the PHENI*/E/LAO copolymerisation dataset, with the aim of establishing a predictive model for the synthesis of designer polymers. To our knowledge this is the first application of such a regression model to olefin copolymerisations in the peer-reviewed literature.
A standard full-factorial model (M1) was used, regressing the response variables (activity A, polymer melting point Tm, crystallinity α, molecular weight Mw, dispersity PDI, Đ and comonomer incorporation x) against full-factorial polynomial combinations of the explanatory variables up to quadratics: temperature of polymerisation Tp, LAO length n, and LAO concentration c (Tp, c, n, Tp2, c2, n2, Tpn, Tpc, cn; Fig. 3). On the basis of likelihood ratio tests, Tp and c explain much of the variation, with −log10(p-values) of 33.0 and 24.9 respectively, and all of the polynomial terms apart from n2 have significant relationships at the 0.01 level of hypothesis testing. The model is predictive with R2 values of: A (0.65), Tm (0.85), α (0.81), Mw (0.96), PDI (0.50), and x (0.85).
Modelling of this kind enables the delineation of interrelated variables. Incorporation is determined almost linearly by c, alongside a contribution from Tp. While Tm is determined principally by the temperature-concentration couple, crystallinity depends more strongly on the side chain length, with the predictors nc, n2 and Tpn all having statistically significant contributions. This is consistent with physical expectations: increased comonomer concentration (and therefore, incorporation) increases the degree of branching, which reduces the intermolecular forces between polymer chains and lowers the melting point.93 The branches are generally excluded from the crystalline lamellae, disrupt chain folding and lead to defective crystallisation,94 with the length of the side chain influencing crystallinity.95
The anticipated dependency of Mw on both Tp and c is reflected in the model, and of the cross terms, Tpn has the greatest effect, showing that temperature-chain length coupling is a more important factor than concentration-chain length. This is suggestive of a mechanistic interpretation, with the larger energy barriers associated with larger monomers interacting with the thermal energy in the system.
To explore this, the model was optimised with respect to three sets of “designer” copolymer properties, with the model converging on reaction conditions expected to produce the desired materials. A detailed discussion of methodology and results may be found in the ESI.† There was variable agreement between the desired, predicted, and experimental values, though the polymer properties that were interpolated within the model data were well predicted. In particular, sample P1 (optimised towards the desired properties of Tm = 110 °C and Mw = 1 MDa) resulted in very well predicted values for activity (M1 3424 kg mol−1 h−1 bar−1; expt. 3180 ± 270 kg mol−1 h−1 bar−1), melting point (M1 112 °C; expt. 112 ± 3 °C), crystallinity (M1 39%; expt. 40 ± 7%), molecular weight (M1 846 kDa, expt. 606 ± 281 kDa), and dispersity (M1 4.1; expt. 4.1 ± 0.4) (Fig. 4). Additionally, the polymer melting temperature closely fitted the value that was desired through programmed synthesis. The relatively large uncertainty in measured Mw results for the reaction being at the boundary of the soluble and insoluble regimes, with poor reproducibility between runs. Surprisingly, and despite the well-predicted macroscopic thermal properties, the incorporation itself was relatively poorly predicted in this case, despite being one of the simplest, best correlated aspects of the model. The generally good agreement between the desired, predicted, and experimental parameters highlights the power and utility of statistical modelling for the synthesis of designer polymers.
Sample P2 was optimised with respect to four parameters simultaneously which led to poorer convergence of the model, though the experimentally measured parameters, including incorporation, were generally well predicted by M1 (Fig. S33†). Finally, sample P3 was extrapolated beyond the scope of the dataset, with a poor optimisation and generally inaccurate predictions highlighting the limitations of this method much beyond the property-space of the data used to construct M1.
This pilot study demonstrates the underlying utility of multivariate modelling, combined with relatively large datasets, in delineating and predicting the outcomes of complex chemical reactions. It is envisaged that including mechanical and material characterisation in the modelling would dramatically expand the scope of this methodology towards the application-directed precision synthesis of “designer” polyolefins. The utility of such a methodology may have far-reaching applications in many areas of chemistry.
The potential of synthesising industrially relevant designer polyolefins using a single catalyst based on statistical models has been demonstrated in principle. The advent of parallelised high-throughput experiments now enables efficient dataset acquisition, and the inclusion of additional parameters such as pressure, and mechanical and material properties would further enable a dramatically expanded scope of tunability and control, with the ultimate goal of entirely application-directed “programmable and predictable” synthesis.35
Footnote |
† Electronic supplementary information (ESI) available: Experimental procedures, polymerisation data, polymer characterisation data, regression modelling and detailed discussions. See DOI: https://doi.org/10.1039/d3sc04861f |
This journal is © The Royal Society of Chemistry 2024 |