Research Article pubs.acs.org/acscatalysis
Demystifying Ziegler−Natta Catalysts: The Origin of Stereoselectivity Antonio Vittoria,† Anika Meppelder,‡ Nic Friederichs,‡ Vincenzo Busico,*,† and Roberta Cipullo*,† †
Laboratory of Stereoselective Polymerizations (LSP), Department of Chemical Sciences, Federico II University, Via Cintia, 80126 Naples, Italy ‡ SABIC, NL-6160 AH Geleen, The Netherlands
ACS Catal. 2017.7:4509-4518. Downloaded from pubs.acs.org by KAOHSIUNG MEDICAL UNIV on 11/12/18. For personal use only.
S Supporting Information *
ABSTRACT: Industrial Ziegler−Natta catalysts for polypropylene production are complex formulations with a reputation for being “black boxes”. In this paper, we report the results of an extensive investigation of the three latest commercial generations, performed with advanced high-throughput experimentation tools and methods. The thus obtained database of structure−property relationships, of extraordinary width and depth, provided a high-definition picture of the screened systems, allowing us to highlight important details of their inner workings. In particular, the delicate relationship between surface coverage and lateral steric pressure on the stereoselectivity of the catalytic species, as well as the role of chemisorbed donor and Al-alkyl species on said parameters for different formulations, was revealed. KEYWORDS: Ziegler−Natta catalysts, isotactic polypropylene, internal donor, external donor, stereoselectivity
■
INTRODUCTION Isotactic polypropylene (i-PP) and Ti-based Ziegler−Natta (ZN) catalysts have parallel success stories. With an annual consumption of more than 60 million tons, i-PP has become the second largest-volume polymer on the market after polyethylene.1 Progress in catalysis was also dramatic: activity increased from a few kilograms to several tons of polymer per gram of Ti and stereoselectivity from moderate to almost perfect.2−4 The first two catalyst generations, consisting of crystalline TiCl3 in a layered modification,3,4 were relatively simple. Stereoselectivity was a consequence of a peculiar crystal lattice with chirotopic Ti both in the bulk and on the side edges of the structural layers. The latter, in particular, exposed linear arrays of enantiomorphous Ti atoms amenable to Cl/R metathesis with an Al-alkyl compound.3−5 Elegant experimental6 and computational7,8 studies highlighted the surface constraints on the thus formed Ti-alkyls, conformationally locked at the first C−C bond to define chiral pockets in which the two propene enantiofaces could be distinguished at the insertion step.4,9 Supported homologues with greatly improved performance, but also much more complex formulations, were introduced in the 1970s.2−4 MgCl2 was serendipitously identified as the bestworking support for TiCl4 (the most frequently used Ti precursor).2 The addition of certain donor compounds as powerful stereoselectivity enhancers (Table 1), at the precatalyst preparation stage [“internal donors” (IDs)] or in combination with the AlR3 activator [“external donors” (EDs)], was also a trial-and-error achievement.2,3 © 2017 American Chemical Society
Unraveling the cross-interactions between the components of the different formulations that followed one another at intervals of approximately 10 years2,3,9 is a long-standing open problem, which generated the catalysts’ reputation for being “black boxes”. The similarity between the crystal lattices of MgCl2 and TiCl3 gave birth to the idea of an epitaxial relationship between at least part of the TiCl x adsorbates and the MgCl 2 substrate.2,3,9,10 Rather than representing a constructive input, this hypothesis triggered decades of flawed mechanistic speculation, including that of a competition between TiCl4 and the donors for selective chemisorption on the support.2,3,9,10 MgCl2(10l) crystal terminations, exposing pentacoordinated Mg, have long been thought to host Ti2Cl8 adducts that, once activated by an AlR3, would mimic the active sites of authentic TiCl3 catalysts. MgCl2(110) terminations, in turn, featuring tetracoordinated Mg and as such being more acidic, were postulated to be preferred targets for donor binding and home to nonstereoselective sites only. The successful introduction, in the 1990s, of 2,2-dialkyl-1,3-dimethoxypropanes as a class of IDs especially prone to chelating tetracoordinated Mg, thus supposedly hampering interaction of TiCl4 with MgCl2(110) facets,2,3,11 was presented as a compelling demonstration of the hypothesis, and even an achievement of molecular design.11,12 Received: April 16, 2017 Revised: May 24, 2017 Published: May 26, 2017 4509
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis
Table 1. Typical Formulations and Performance Ranges of MgCl2-Supported ZN Catalysts for i-PP Production2,3
a
generation
internal donor
external donor
productivitya
XSb (wt %)
Mw/Mn
third fourth fifth sixth
ethyl benzoate dialkyl phthalate 2,2′-dialkyl-1,3-dimethoxypropane dialkyl succinate
aromatic monoester alkoxysilane none or alkoxysilane alkoxysilane
0.5−0.8 1−2 >2 1−2
3−5 1−5 2−5 1−5
6−9 6−8 4−6 >8
In units of 103 kilograms of PP per gram of Ti. bXylene-soluble fraction.
compositions that are not used in industrial practice.17,27 In this sense, the aforementioned black box perception is justified. In this paper, we present the results of an extensive and thorough study of fourth-, fifth-, and sixth-generation ZN catalysts (Table 1), used in combination with AlEt3 and an array of alkoxysilane EDs with a wide range of structural diversity. The investigation, performed with advanced highthroughput experimentation (HTE) tools and methods,28 consisted of two parts. In the first part, catalyst performance (in terms of polymerization behavior and polymer properties) was determined using a fully automated secondary screening platform with 48 mini-reactors (Freeslate PPR48), integrated with a polymer characterization workflow, including gel permeation chromatography (GPC), analytical crystallization elution fractionation (A-CEF), and 13C nuclear magnetic resonance (NMR) spectroscopy. In a second part, another HTE platform (Freeslate Core Module) was used to follow the evolution in composition of the solid catalysts under conditions closely mimicking those of application. The resulting database of structure−property relationships, of extraordinary width and depth for a single investigation, provided a high-definition picture of the screened systems, allowing us to highlight for the first time important details of their inner workings. In particular, we explored the delicate relationship between surface coverage at saturation and lateral steric pressure on the stereoselectivity of the catalytic species and clarified the roles of chemisorbed donors and Al-alkyls on said parameters for different formulations (i.e., catalyst surface distributions); this is very important for further progress. In the latter respect, it should be recalled that a recent REACH ban on dialkyl phthalates for toxicity issues has generated a growing demand for their replacement;1 considering that fourth-generation ZN systems are the working horses of i-PP industry,2 the question is highly relevant and calls for urgent attention.
It was only several years later that more critical analyses of the experimental data and quantum mechanics modeling studies disproved the concept of MgCl 2 (10l) and MgCl2(110) as “good” and “bad” surfaces, respectively. In fact, it became impossible to ignore the unambiguous evidence that donor molecules have a direct and specific impact on polymer microstructure13 and therefore are, if not part of the catalytic species, at least in nonbonded contact with them.9 Several independent density functional theory (DFT) calculations, in turn, concluded that Ti 2Cl 8 adsorbates on MgCl2(10l)-like edges are not stable,14,15 and lately that TiCl4 chemisorption is feasible only in mononuclear form on MgCl2(110)-like edges.16,17 The current view is that the role of donors in MgCl2supported ZN catalyst systems is twofold. (i) The first is to stabilize the primary particles by strong chemisorption, lowering their surface energy.18−21 Mg/donor mole ratios in the range of 10−20 are not unusual,2,3 which points to lateral dimensions of the structural layers of only few unit cells,18,22 and surface areas of >150 m2 g−1 (unattainable for binary MgCl2/TiCl4 particles because TiCl4 adsorption is too weak16). (ii) The second is to impart the necessary steric hindrance to the inherently chiral but otherwise too open catalytic species, very much like ancillary ligands in molecular catalysts.9 A qualitative model for this function was proposed by some of us in ref 23 and still accounts for all known facts, including the complex stereoblock microstructure of (part of) the polymers.9 Although in principle one single donor might exert both functions, in most cases ID and ED roles are differentiated (Table 1).2,3,9 Surface stabilization is important already at the stage of primary particle formation; this usually entails chlorination of a Mg precursor compound with excess TiCl4 in the presence of the ID, which requires that the latter compounds be mutually unreactive. For the modification of the catalytic species, on the other hand, a proper steric demand, the preference for chemisorption on Mg rather than Ti, and the lack of reactivity with AlR3 are equally important conditions. Some well-functioning IDs (e.g., dialkyl phthalates) react irreversibly with the AlR3 activator and are extracted from the solid catalyst during polymerization;2,3,24−26 therefore, they need to be replaced by an ED. By far the most widely used EDs are sterically demanding alkoxysilanes;2,3 these react poorly with AlR3 compounds but do react with TiCl4, which prevents their use as IDs. While the general picture presented above is sound, what is still missing is an adequate understanding of the details that would allow true catalyst design. In particular, how the catalytic species look like and what determines their diverse behaviors in the different catalyst generations remain largely unanswered questions. Ironically, formulations that work well experimentally are not trivial to validate by means of computational modeling (as we shall see in the following sections), whereas some published models claiming a high stereoselectivity involve
■
EXPERIMENTAL SECTION The composition of the four (pre)catalysts, prepared according to ref 29, can be found in Table 2. The eight alkoxysilane EDs, in turn, are listed in Table 3. ED1 was purchased from Gelest Inc., and ED6 was custom-made; all others were obtained from Evonik AG. All semibatch propene polymerization experiments in a heptane slurry were performed in a Freeslate Parallel Pressure Table 2. Compositions of the Screened (Pre)catalysts internal donor (ID) C1 C2 C3 C4 4510
dibutyl phthalate 2,2-diisobutyl-1,3dimethoxypropane 2,2-dimethyl-1,3dimethoxypropane 2,3-diisopropyldiethylsuccinate
Ti (wt %)
Mg (wt %)
ID (wt %)
2.0 2.7
18.6 18.4
11.5 13.2
2.1
19.1
9.3
2.4
19.2
9.5
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis Table 3. Set of Screened Alkoxysilane EDs
Avance DRX 400 spectrometer operating at 400 MHz. Acquisition conditions were as follows: 5 mm probe; acquisition time, 3.0 s; relaxation delay, 5.0 s; pulse angle, 90°; spectral width, 10 ppm; eight transients. Resonance assignment was based on the literature and preliminary 1H NMR characterizations of the neat donor molecules. Quantitative determinations were based on peak integration against that of an aliquot of acetonitrile added as an internal standard (methyl peak at δ 2.05 downfield of TMS). ICP-OES analyses were performed using an Agilent 700-series spectrometer on water solutions of the solid phases treated in sequence with 2.0 mL of concentrated H2SO4, 2.0 mL of concentrated HNO3, and (when needed) 2.0 mL of 30 vol % H2O2 (total time of 16 h). The spectrometer was calibrated using commercial standard solutions (metal concentrations in the range of 1−100 ppm).
Reactor setup with 48 reaction cells (PPR48), fully contained in a triple MBraun glovebox under nitrogen. The cells, with a working volume of 5.0 mL, featured an 800 rpm magnetically coupled stirring, and individual online reading/control of temperature, pressure, monomer uptake, and uptake rate. The setup and the operating protocol were described in full detail in ref 28. Polymerization conditions were as follows: T = 70 °C; p(C3H6) = 4.5 bar; p(H2) = 0.20 bar; [Al]/[Ti] = 160; [ED]/ [Al] = 0, 0.025, 0.050, 0.10, or 0.20; t = 30 min. The p(C3H6)/ p(H2) ratio was set at a value resulting in average polymer molecular weights in the typical commercial range. All experiments were performed in duplicate, for a total of 132 pairs. The polymerization products were characterized by hightemperature GPC with a Freeslate Rapid-GPC setup, analytical crystallization elution fractionation (A-CEF) with a Polymer Char setup, and quantitative 13C NMR with a Bruker DRX 400 spectrometer equipped with a high-temperature cryoprobe accommodating 5 mm OD tubes and a preheated robotic sample changer (see ref 28 and the Supporting Information for detailed operating protocols and conditions). A second high-throughput experimentation platform, namely a Freeslate Extended Core Module (XCM), also contained in a triple MBraun glovebox and allowing the robotic handling of solids, liquids, and slurries according to fully automated weighing, dispensing, and reaction protocols, was utilized to determine the composition of the four solid catalysts following the reaction with AlEt3/ED mixtures or AlEt3 alone in a heptane slurry under conditions mimicking those of polymerization (albeit in the absence of the monomer). For this study, aliquots of 15−20 mg of solid (pre)catalyst in a matrix of 24 × 8 mL pressure-tight vials were suspended in 3.8 mL of heptane, added with predefined amounts of AlEt3(/ED) solutions ([Al]/ [Ti] = 25, and [ED]/[Al] = 0.10), and allowed to react at 70 °C under magnetic stirring (800 rpm) for 30 min, and reactions were quenched when the mixtures were rapidly cooled to −15 °C. After centrifugation (1400 rpm) in a Savant SPD121P centrifugal drying station, the vials were opened and repositioned in the reaction deck, where the supernatants were removed robotically by aspiration through a needle. The solid phases were washed twice with 3.6 mL aliquots of heptane (applying the same sequence of decantation and supernatant aspiration) and once with 3.6 mL of pentane and finally dried under vacuum at 50 °C for 10 h in the aforementioned drying station. Each dry solid phase was then dissolved in 1.00 mL of methanol-d4; 0.60 mL was analyzed by 1H NMR to determine ID and ED contents, whereas 0.40 mL was dried again, mineralized, and analyzed by inductively coupled plasma optical emission spectrometry (ICP-OES) for Mg, Ti, and Al. Quantitative 1H NMR analyses were performed with a Bruker
■
RESULTS AND DISCUSSION
Catalyst Systems. The four ZN catalysts selected for this study (Table 2) belong in the three latest generations of commercial relevance (Table 1).2,3 Catalyst C1 (dibutyl phthalate as the ID) is a widely used fourth-generation representative.2,3 Catalysts C2 (2,2-diisobutyl-1,3-dimethoxypropane as the ID) and C3 (2,2-dimethyl-1,3-dimethoxypropane as the ID) are members of the fifth generation; the former found industrial application, whereas the latter is poorly stereoselective11,12 but very interesting from a mechanistic standpoint, as we shall see below. Catalyst C4 (2,3diisopropyldiethylsuccinate as the ID) was chosen as an example of the sixth generation, the most recent and also the least described in the scientific literature.2 Alkoxysilane EDs are employed with all three generations (Table 1), even though fifth-generation systems may also be used without.2,3,11 The ED set screened in the study presented here (Table 3) included dimethoxy-, trimethoxy-, and triethoxysilanes bearing linear, branched, and cyclic substituents with different steric demand; three of them (namely, ED1−ED3) were selected as “minimal structures” for comparative purposes, whereas the remaining five (ED4−ED8) are applied commercially. Polymerization Screening. All 132 duplicate pairs of slurry propene homopolymerization experiments were performed under the same conditions [T = 70 °C; p(C3H6) = 4.5 bar; p(H2) = 0.20 bar; [Al]/[Ti] = 160; t = 30 min], except for the [ED]/[Al] ratio that was varied stepwise ([ED]/[Al] = 0, 0.025, 0.050, 0.10, or 0.20). Catalyst deactivation was always negligible, and polymerization kinetics could be simply quantified in terms of average catalyst productivity (Rp, in kilograms of PP per gram of catalyst per hour). Polymer molecular weight, crystallinity, and stereosequence distributions were determined by rapid GPC, analytical crystallization elution 4511
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis Table 4. Experimental Results for Catalyst C1−C4-AlE3/EDx at an [ED]/[Al] Ratio of 0.10 Rp (kg g−1 h−1) none ED1 ED2 ED3 ED4 ED5 ED6 ED7 ED8
none ED1 ED2 ED3 ED4 ED5 ED6 ED7 ED8
none ED1 ED2 ED3 ED4 ED5 ED6 ED7 ED8
none ED1
Mn (kDa)
Mw (kDa)
Mw/Mn
7.2 7.9 6.3 5.0 5.0 5.5 5.2 6.3 8.3 6.4 7.2 8.3 4.8 5.1 6.7 7.1 8.7 7.6
17 18 29 29 42 29 30 26 42 32 38 41 25 25 40 39 45 46
103 103 175 175 222 229 180 145 257 284 251 242 198 177 319 313 393 399
6.1 5.7 6.0 6.0 5.3 7.8 6.0 5.5 6.1 8.8 6.6 5.9 7.9 7.2 8.0 8.0 8.8 8.7
11.1 10.7 11.2 8.8 7.8 6.5 10.4 8.9 10.1 10.1 9.2 7.1 9.5 9.6 10.6 5.9 10.8 10.0
27 33 32 33 32 32 34 32 35 35 35 26 34 34 29 33 38 33
124 162 157 173 169 164 185 169 157 161 194 144 179 205 166 168 189 180
4.6 4.9 4.9 5.3 5.3 5.2 5.5 5.3 4.5 4.6 5.6 5.6 5.2 6.0 5.8 5.1 5.0 5.5
6.0 5.9 4.7 4.0 3.5 3.8 4.5 4.7 5.0 5.0 4.5 4.9 4.1 4.5 4.1 4.8 4.8 4.3
19 28 21 22 23 21 22 18 22 20 24 22 23 28 21 25 23 23
87 105 104 108 101 109 122 96 110 99 114 105 110 114 103 108 113 112
4.7 3.7 4.9 5.0 4.3 5.1 5.6 5.4 5.0 4.9 4.7 4.7 4.7 4.1 4.9 4.3 5.0 4.9
4.0 4.8 5.1
16 15 16
103 124 134
6.5 8.1 8.3
AF (%) (A) Catalyst C1 15.2 11.9 6.3 7.3 3.3 3.5 4.8 4.4 4.3 3.8 4.7 3.6 5.2 5.5 3.2 3.6 3.0 3.0 (B) Catalyst C2 3.9 4.4 3.5 4.1 2.8 2.4 3.6 3.6 3.7 2.9 3.1 3.4 3.7 4.4 3.6 4.9 3.6 3.8 (C) Catalyst C3 19.4 18.2 14.8 15.0 12.7 10.1 12.3 13.2 13.6 12.9 12.5 12.4 12.7 12.3 14.5 13.3 12.5 13.2 (D) Catalyst C4 9.9 9.6 8.5 4512
Tel,max (°C)
[mmmrrmmm] (%)
[mrrm] (%)
[rrrrrr] (%)
113.2 113.0 112.9 113.1 115.8 115.9 115.5 115.6 116.2 116.2 116.3 116.3 116.4 116.4 117.4 117.5 117.9 117.8
0.90 0.92 1.12 1.06 0.48 0.49 0.49 0.48 0.41 0.47 0.46 0.40 0.43 0.43 0.30 0.29 0.26 0.29
2.1 2.0 2.0 1.9 1.0 0.9 0.9 1.0 1.0 1.0 0.8 0.8 1.0 0.9 0.7 0.6 0.6 0.6
1.3 1.3 0.4 0.5 0.3 0.2 0.4 0.3 0.3 0.3 0.3 0.3 0.4 0.4 0.2 0.2 0.2 0.2
114.0 113.7 113.5 113.6 114.3 114.2 114.4 114.1 114.1 114.2 114.1 113.9 113.5 112.8 114.0 113.8 114.6 114.3
0.75 0.67 0.66 0.65 0.56 0.55 0.65 0.57 0.52 0.53 0.58 0.55 0.56 0.47 0.55 0.58 0.60 0.53
1.3 1.3 1.1 1.1 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.9 1.0 1.0 1.0 0.9
0.1 0.1 0.2 0.1 0.1 0.1 85% upon exposure
to AlEt3 as well as an AlEt3/ED mixture. In the latter case, only a modest amount of ED was adsorbed. (c) For a given catalyst, the incorporation of ED1 and ED8 (in moles per mole of Mg) 4514
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis
ended up with comparable degrees of coverage at saturation (in moles per mole of Mg) for the available surfaces of each given catalyst;20,31 the adsorption data of ED1 and ED8 (Figure 3) are compatible with such an assumption. If the hypothesis holds, then the lateral steric pressure experienced by the catalytic species, and hence their stereoselectivity according to the model described in ref 23 (see below), should grow with growing alkoxysilane steric demand and attain a characteristic plateau value for each ED once surface saturation is reached; considering Tables S1 and S4, this seems indeed to occur around an [ED]/[Al] ratio of ≈0.05. What does not seem to fit in the picture, on the other hand, is that the C1-AlEt3 and C4-AlEt3 systems turned out to be slightly more stereoselective than the C1-AlEt3/ED1 and C4AlEt3/ED1 systems (Tables 4A,D and Figures 1 and 2). Our explanation is that the chemisorbed Al species (Figure 3) acted as a surrogate for the ED as ID. AlEt3 and AlEt2Cl are strong Lewis acids, known for their self-dimerization equilibria;32,33 in the monomeric state, they can form heterodinuclear adducts with Al−Cl−Mg and Al−Cl−Ti bridges.2−4 Binding of Al-alkyl to TiCl3 with formation of doubly bridged Al−[(μ-Cl)(μ-Et)]− Ti moieties is strong,34,35 and likely one of the reasons for the low concentration of active Ti measured in quenched-flow studies.36,37 As far as binding to MgCl2 is concerned, a recent DFT study concluded that AlEt3−xClx chemisorption (x = 0 or 1) on MgCl2(104) facets is exergonic,17 which is in line with the results depicted in Figure 3. On the basis of our polymerization data (Figures 1 and 2), the ability of adsorbed Al-alkyls to enhance catalyst stereoselectivity is similar or even slightly better than that of small EDs (e.g., ED1) but much poorer than for best-in-class EDs (e.g., ED7 or ED8). It should be noted that the amount of chemisorbed Al on catalysts C1 and C4 was smaller in the presence of an alkoxysilane (Figure 3), which we interpret as evidence of a competition; in fact, according to computational modeling data, alkoxysilanes prevail over AlEt3−xClx species for adsorption on plain MgCl2 crystal terminations.17,20,27 A completely different picture emerged for systems based on catalysts C2 and C3, whose 1,3-dimethoxypropane IDs were the dominant donors in the adsorbate pool, leaving very limited room for ED action (Tables 4B,C, S2, S3, and S5 and Figures 1−3). On the basis of conventional wisdom, such ZN systems should be the easiest to interpret; in fact, the strong preference of their IDs for chemisorption on MgCl2(110) terminations11,12,31 is expected to determine the least differentiated surface environment [which indeed is consistent with the comparatively narrow molecular weight distribution (MWD) of the produced polymers]. However, our computational modeling studies indicated that 1,3-dimethoxypropanes on plain MgCl2(110) facets, irrespective of the steric bulk of the alkyl substituents on C-2, cannot get close enough to adjacent TiCl2R catalytic species to make them highly stereoselective in propene insertion. The data in Figure 3 and Table S5, demonstrating unexpectedly that large amounts of Al species were chemisorbed on both catalysts C2 and C3, despite the presence of the ID and with little (catalyst C2) or practically no (catalyst C3) evidence of competition with the ED when used, can provide a solution to this puzzling problem, as will be illustrated in the next section. Improving the Computational Models of Catalytic Species. Let us now discuss how to interpret the phenomenological picture of the previous sections in terms of suitable models of ZN catalytic species. Different approaches
was practically the same. (d) A massive chemisorption of Al species was observed for all systems, including those in which the ID was retained. (e) All catalysts underwent a significant (20−50%) loss of Ti. Facts a and b are the aggregate of several concurring processes, on which the previous literature shed light only in part.2,3,10,11,24−26 ID extraction by the AlR3 may follow from an irreversible chemical reaction, or the formation of a strong Lewis acid−base adduct. On the basis of preliminary results of a wider (and still ongoing) kinetic study (Table S6), we concluded that the former is the case for dibutyl phthalate [for reduction by AlEt3 at 70 °C in a toluene solution, we measured first-order kinetics with a kobs value of (3.45 ± 0.05) × 10−3 s−1], but not for 2,3-diisopropyldiethylsuccinate [with a kobs of (1.95 ± 0.15) × 10−5 s−1]. 1,3-Dimethoxypropanes, in turn, do not react with AlR3 compounds but rapidly form adducts with them in solution.2,3 According to the first seminal papers on fifth-generation ZN catalyst systems,11,12 IDs with bulky substituents on C-2 (e.g., 2,2-diisobutyl-1,3-dimethoxypropane in C2) are strongly bound to MgCl 2 (110) terminations, and their extraction by AlR3 is marginal; on the other hand, less sterically demanding homologues (like 2,2dimethyl-1,3-dimethoxypropane in C3) would adsorb much more weakly, because in a large fraction of low-energy conformers the two O atoms are too far apart to chelate tetracoordinated Mg (O∩O distance of >3 Å). Our results (Figure 3 and Table S5) do not endorse such a discrimination; as a matter of fact, the ID of catalyst C3 turned out to be as extensively and firmly bound to the catalyst surface as that of catalyst C2 (which is also in line with recent DFT calculations21). We will come back to fact c at a later stage. With regard to fact d, our ICP-OES determinations were not informative with respect to the chemical nature of the Al adsorbates. TiCl4 reduction by AlR3 compounds is known to proceed with the formation of AlR3−xClx species (in particular, AlEt2Cl).2−4 In the case of catalyst C1, AlEt2OBu was also formed, as a reduction product of the dibutyl phthalate ID (see the Supporting Information). Last but not least, fact e confirms the rather weak chemisorption of TiCl4 onto MgCl2;16 we verified that the fraction of TiClx lost to the liquid phase had negligible activity in propene polymerization under the conditions used in this study, as was demonstrated by means of propene polymerization tests on the filtrates. Donor Structure−Property Analysis. Let us now examine the impact of individual donor structures on catalyst performance. As noted above, the screened EDs modulated very effectively the stereoselectivity of systems C1-AlEt3/EDx and, to a lower extent, C4-AlEt3/EDx ones (Tables 4A,D, S1, and S4 and Figures 1 and 2). It is plausible to trace this finding to the observed ID/ED exchange (Figure 3 and Table S5), which should result in a significant fraction of catalytic species in C1 and C4 with neighboring ED molecules. We note at this point that the x in the EDx identification codes of the screened alkoxysilanes was assigned ex-post, in such a way that a higher x corresponded to a higher stereoselectivity within the C1-AlEt3/ EDx series (Table 4A). Notably, a very similar ordering turned out to hold for the C4-AlEt3/EDx series, too (Table 4D and Figure 1). In a first approximation, the steric crowding at the Si atom, that is to say next to the surface once the ED molecules are adsorbed, also grows with an increase in x (Table 3). This correlation is less obvious than it may appear; in fact, it suggests that for all alkoxysilanes in the set similar chemisorption modes 4515
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis
Figure 4. Updated three-site model for ZN catalysts (see the text and ref 23). Mg and Cl atoms are colored violet and green, respectively. The large red spheres highlight the two active Ti sites according to the Cossee insertion mechanism; the light blue spheres highlight the surface Mg sites where the presence of an adsorbate would hinder one of the two octants (light gray) where the first chain C−C bond could be located. Chain propagation is predicted to be highly isotactic in case A, weakly isotactic in case B, and chain-end-controlled syndiotactic in case C.
Figure 5. (A) DFT model of adjacent TiCl4 and 2,2-diisobutyl-1,3-dimethoxypropane co-adsorption on a MgCl2(110) edge. (B) Same as panel A after the adsorption of an AlEt2Cl molecule (see the text). Color key: Mg, violet; Ti, light gray; Al, pink; Cl, green; O, red; C, dark gray.
access to the Ti center with the favored enantioface and are under strong lateral pressure by the neighboring co-adsorbate pool, freezing diffusion phenomena41,42 or even hindered conformational motions. Should said steric pressure fade, the enantioselectivity will decrease, because the conformational constraints on the growing chain will weaken, and the chiral active pockets become too loose. Depending on the extent of said fading, and whether only one octant or both octants are involved, chain propagation will deteriorate to weakly isotactic (Figure 4B), or even chain-end-controlled syndiotactic (Figure 4C).9,23 In the case of a dynamic character of the interested surfaces, stereoblock chains might form.9,23 The distribution of the three basic cases of Figure 4 (i.e., close/close, close/open, and open/open octants) is a function of the adsorbate pool. In the previous sections, we provided evidence that said pool includes not only donors but also Alalkyls. Via examination of the recent literature,17,27 it appears that models of TiCl4 adsorbates at defective locations of MgCl2(104)-like edges exposing tetracoordinated Mg are not incompatible with the hypothesis of an effective steric modification by adjacent alkoxysilane EDs (even if explicit calculations are still pending). A case in which the cooperation of Al-alkyls is required, on the other hand, seems that of fifth-generation catalysts. Figure 5A shows a computational model (for full details, see the Supporting Information) of a portion of the plain MgCl2(110) edge accommodating a TiCl4 unit and an adjacent 2,2dimethyl-1,3-dimethoxypropane molecule in the minimum energy structure; it is evident upon inspection, and was confirmed by calculation, that the two co-adsorbates are too far apart to give rise to a catalytic species falling under the case of
have been reported to quantify the number of such species. Simulations of PP MWDs as summations of Schulz−Flory functions ended up with a minimum of four components;38 however, this method cannot discriminate between chemical and physical effects on the MWD, and the likelihood of overdetermined solutions is high because the poor resolution and limited precision of MWD data complicate the evaluation of model significance. A more robust approach, in our opinion, is based on the statistical analysis of high-resolution 13C NMR stereosequence distributions.9,23 This identified three basic types of stereosequences, namely, highly isotactic, weakly isotactic (“isotactoid”), and syndiotactic;9,23 the plausible assumption of a corresponding number of distinct families of catalytic species was translated into the three-site model of ref 23. Here we propose an updated version, assuming that all catalytic species are mononuclear Ti(III) surface adducts with the structure of Figure 4, in line with the indications of recent Raman39 and high-resolution ESR40 studies, and of the latest DFT calculations.16,17 The first coordination sphere of Ti is octahedral and C2-symmetric (like in crystalline TiCl3);9 steric hindrance in the second coordination sphere, on the other hand, can vary. Assuming a Cossee-type chain migratory insertion mechanism,5,9 highly isotactic chain propagation requires that the active sites be sterically constrained at two diagonal octants of the four where the first C−C bond of the growing polymer chain can be located in the 1,2-propene insertion TS, thus locking the chain conformation in the desired chiral orientation and ensuring site (pseudo)homotopicity (Figure 4A).9 Such a condition can be met when adsorbates with adequate bulk (vide inf ra) occupy the adjacent surface just at the limit still allowing fast monomer 4516
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis
design can be aimed to steer the formation of MgCl2(10l)-like or MgCl2(110)-like facets in case chemisorption is favored on one of the two,31 with interesting opportunities for, e.g., polymer MWD modulation. The ED search, in turn, can be performed with HTE methodologies28 or even be unnecessary because it is unlikely that a novel ID can lead to a (pre)catalyst with unprecedented MgCl2 surfaces, which implies that existing alkoxysilanes should always be an option. A superior versatility might be claimed as a further advantage of the ID/ED philosophy; diversifying the performance of one single (pre)catalyst by means of a broad ED portfolio looks appealing in fundamental terms, although plant engineers who know better about the drawbacks of transitions in huge continuous production lines may have a different opinion. As a final remark, we note that in this work we did not consider two very important and closely related properties of ZN systems for i-PP, namely, regioselectivity and H 2 response.2,3,9,37,43 We have recently implemented HTE protocols for determining such properties for the first time; the results will be presented and discussed in the future.
Figure 4A. Figure 5B shows the same fragment with an additional AlEt2Cl molecule chemisorbed between the two aforementioned adsorbates; the calculated free energy of adsorption (ΔGads) was −7.9 kcal mol−1. Notably, even small alkoxysilane molecules like ED1 were estimated to be too bulky to effectively compete with the AlEt 2 Cl moiety for chemisorption at that specific surface vacancy. For the insertion of propene at the catalytic species formed by alkylation and reduction of the TiCl4 precursors in panels A and B of Figure 5, we calculated ΔGre/si values of ≈0 and 1.5 kcal mol−1, respectively; the latter is in good agreement with experiment (Table 4C). Conformational interlocking of ID and AlEt2Cl enhanced stereorigidity; with bulkier 1,3-dimethoxypropanes, this can only be more severe (calculations are running). The presence of ED molecules at distal surface locations can also contribute to enforce the necessary lateral steric pressure (Figures 1−3).
■
CONCLUSIONS In this paper, we revisited the polymerization behavior of the three latest generations of industrial MgCl2-supported ZN catalysts for i-PP production with a comprehensive integrated HTE approach. The extensive structure−property database implemented under rigorously controlled experimental conditions allowed us to analyze and compare catalyst behaviors at an unprecedented level of detail; as a result, we highlighted the roles of all individual catalyst components and how surface coverage and lateral steric pressure on the active sites can be modulated by means of co-adsorbed donor and Al-alkyl molecules for different surface distributions. A full molecular description of these systems is still difficult to imagine, if only because we are far from sorting out quantitatively the competing interactions between all components of successful formulations. As long as this is the case, deterministic catalyst design remains unfeasible. On the other hand, the working principles of such formulations are now understood, and searching for novel and/or better ones is possible with a combination of HTE and statistical modeling tools (e.g., of the QSAR type). In particular with regard to the implementation of new donors, the pros and cons of the two viable options (i.e., one being single donors that can play the two being key functions of surface stabilization and active site modification, vs functionspecific ID/ED pairs2,3) can be better evaluated in light of our new results. The single-donor solution is undoubtedly more elegant and in principle more simple from the perspective of catalyst application. However, donor design is extremely demanding, because the aforementioned functions call for different (albeit not incompatible) requirements; combining all of them without compromising on ultimate performance is virtually impossible, as is demonstrated by the 1,3-dimethoxypropane case history. Moreover, the approach is difficult to implement in HTE mode, because it includes the stage of precatalyst synthesis, involving a harsh chemistry with several lengthy kinetically controlled steps.2,3 The ID/ED pair solution, in turn, looks more complicated on paper, because it entails two independent searches instead of one, and the optimization of additional process variables (in particular, [Ti]/[ED] and [Al]/ [ED] ratios as a function of polymerization temperature). However, compared with the single-donor approach, each of the two searches is simpler, because the number of performance criteria to be met by the donor is lower. Moreover, extra functions can be introduced more easily; as an example, ID
■
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.7b01232. Chemical specifications, full results of polymerization screening, full results of catalyst composition screening, kinetic data for the reaction of dibutyl phthalate and 2,3diisopropyldiethylsuccinate with AlEt3 at 70 °C, and computational modeling details (PDF)
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail:
[email protected]. *E-mail:
[email protected]. ORCID
Roberta Cipullo: 0000-0003-3846-1999 Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors are grateful to Emanuele Breuza and Andrea Correa for the computational modeling results of Figure 5, and Peter H. M. Budzelaar for useful discussions.
■
REFERENCES
(1) Ali, S. Catal. Rev. 2014, 27, 7−14. (2) Cecchin, G.; Morini, G.; Piemontesi, F.; Seidel, A. Kirk-Othmer Encyclopedia of Chemical Technology; Wiley-Interscience: New York, 2007; Vol. 26. (3) Pasquini, N., Ed. Polypropylene Handbook, 2nd ed.; Hanser Publishers: Munich, 2005. (4) Busico, V. In Giulio Natta and the Development of Stereoselective Propene Polymerization in Polyolefins: 50 years after Ziegler and Natta; Kaminsky, W., Ed.; Springer: Heidelberg, Germany, 2013; Vol. 257, pp 37−58. (5) Arlman, E. J.; Cossee, P. J. Catal. 1964, 3, 99−104. (6) Zambelli, A.; Sacchi, M. C.; Locatelli, P.; Zannoni, G. Macromolecules 1982, 15, 211−212. (7) Corradini, P.; Barone, V.; Fusco, R.; Guerra, G. Eur. Polym. J. 1979, 15, 1133−1141. (8) Corradini, P.; Guerra, G.; Fusco, R.; Barone, V. Eur. Polym. J. 1980, 16, 835−842. 4517
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518
Research Article
ACS Catalysis (9) Busico, V.; Cipullo, R. Prog. Polym. Sci. 2001, 26, 443−533. (10) Corradini, P.; Barone, V.; Fusco, R.; Guerra, G. Gazz. Chim. Ital. 1983, 113, 601−607. (11) Albizzati, E.; Giannini, U.; Morini, G.; Galimberti, M.; Barino, L.; Scordamaglia, R. Macromol. Symp. 1995, 89, 73−89. (12) Scordamaglia, R.; Barino, L. Macromol. Theory Simul. 1998, 7 (4), 399−405. (13) See, e.g.: Morini, G.; Albizzati, E.; Balbontin, G.; Mingozzi, I.; Sacchi, M. C.; Forlini, F.; Tritto, I. Macromolecules 1996, 29 (18), 5770−5776. (14) Seth, M.; Margl, P. M.; Ziegler, T. Macromolecules 2002, 35, 7815−7829. (15) Boero, M.; Parrinello, M.; Weiss, H.; Hüffer, S. J. Phys. Chem. A 2001, 105, 5096−5105. (16) D’Amore, M.; Credendino, R.; Budzelaar, P. H. M.; Causá, M.; Busico, V. J. Catal. 2012, 286, 103−110. (17) Credendino, R.; Liguori, D.; Fan, Z.; Morini, G.; Cavallo, L. ACS Catal. 2015, 5, 5431−5435. (18) Busico, V.; Causà, M.; Cipullo, R.; Credendino, R.; Cutillo, F.; Friederichs, N.; Lamanna, R.; Segre, A.; Van Axel Castelli, V. J. Phys. Chem. C 2008, 112, 1081−1089. (19) Credendino, R.; Pater, J. T. M.; Correa, A.; Morini, G.; Cavallo, L. J. Phys. Chem. C 2011, 115, 13322−13328. (20) Capone, F.; Rongo, L.; D’Amore, M.; Budzelaar, P. H. M.; Busico, V. J. Phys. Chem. C 2013, 117, 24345−24353. (21) Kuklin, M. S.; Bazhenov, A. S.; Denifl, P.; Leinonen, T.; Linnolahti, M.; Pakkanen, T. A. Surf. Sci. 2015, 635, 5−10. (22) D’Amore, M.; Thushara, K. S.; Piovano, A.; Causà, M.; Bordiga, S.; Groppo, E. ACS Catal. 2016, 6, 5786−5796. (23) Busico, V.; Cipullo, R.; Monaco, G.; Talarico, G.; Vacatello, M.; Chadwick, J. C.; Segre, A. L.; Sudmeijer, O. Macromolecules 1999, 32, 4173−4182. (24) Busico, V.; Corradini, P.; De Martino, L.; Proto, A.; Savino, V.; Albizzati, E. Makromol. Chem. 1985, 186, 1279−1288. (25) Busico, V.; Corradini, P.; Demartino, L.; Proto, A.; Albizzati, E. Makromol. Chem. 1986, 187, 1115−1124. (26) Noristi, L.; Barbè, P. C.; Baruzzi, G. Makromol. Chem. 1991, 192, 1115−1127. (27) Correa, A.; Credendino, R.; Pater, J. T. M.; Morini, G.; Cavallo, L. Macromolecules 2012, 45, 3695−3701. (28) Busico, V.; Cipullo, R.; Mingione, A.; Rongo, L. Ind. Eng. Chem. Res. 2016, 55, 2686−2695. (29) WO 2007 134851 to SABIC. (30) For third-generation catalysts, some examples are: (a) Zhou, Q.; Zheng, T.; Li, H.; Li, Q.; Zhang, Y.; Zhang, L.; Hu, Y. Ind. Eng. Chem. Res. 2014, 53, 17929−17936. (b) Kissin, Y. V.; Zhou, Q.; Li, H.; Zhang, L. J. Catal. 2015, 332, 156−163. (31) Correa, A.; Piemontesi, F.; Morini, G.; Cavallo, L. Macromolecules 2007, 40, 9181−9189. (32) Shreve, A. P.; Muelhaupt, R.; Fultz, W.; Calabrese, J.; Robbins, W.; Ittel, S. D. Organometallics 1988, 7, 409−416. (33) Ehm, C.; Antinucci, G.; Budzelaar, P. H. M.; Busico, V. J. Organomet. Chem. 2014, 772−773, 161−171. (34) Bahri-Laleh, N.; Correa, A.; Mehdipour-Ataei, S.; Arabi, H.; Haghighi, M. N.; Zohuri, G.; Cavallo, L. Macromolecules 2011, 44, 778−783. (35) Kumawat, J.; Gupta, V. K.; Vanka, K. ChemCatChem 2016, 8, 1809−1818. (36) Chammingkwan, P.; Thang, V.; Terano, M.; Taniike, T. Top. Catal. 2014, 57, 911−917. (37) Yu, Y.; Busico, V.; Budzelaar, P. H. M.; Vittoria, A.; Cipullo, R. Angew. Chem., Int. Ed. 2016, 55, 8590−8594. (38) Kissin, Y. V.; Chadwick, J. C.; Mingozzi, I.; Morini, G. Macromol. Chem. Phys. 2006, 207, 1344−1350. (39) Brambilla, L.; Zerbi, G.; Piemontesi, F.; Nascetti, S.; Morini, G. J. Phys. Chem. C 2010, 114, 11475−11484. (40) Morra, E.; Giamello, E.; Van Doorslaer, S.; Antinucci, G.; D’Amore, M.; Busico, V.; Chiesa, M. Angew. Chem., Int. Ed. 2015, 54, 4857−4860.
(41) Credendino, R.; Pater, J. T. M.; Liguori, D.; Morini, G.; Cavallo, L. J. Phys. Chem. C 2012, 116, 22980−22986. (42) Credendino, R.; Liguori, D.; Morini, G.; Cavallo, L. J. Phys. Chem. C 2014, 118, 8050−8058. (43) Chadwick, J. C.; van der Burgt, F.; Rastogi, S.; Busico, V.; Cipullo, R.; Talarico, G.; Heere, J. J. R. Macromolecules 2004, 37, 9722−9727.
4518
DOI: 10.1021/acscatal.7b01232 ACS Catal. 2017, 7, 4509−4518