INFRARED SPECTROSCOPY -

A SECRET WINDOW TO THE HEART OF THE CATALYST?

E-mail: martin@ystenes.com, eilerts@chembio.ntnu.no, richard.blom@chem.sintef.no, err@statoil.com

Introduction

Spectroscopy as a tool for inferring molecular structures, has been put on the sideline by direct structural methods like X-ray crystallography. Nevertheless, indirect structural methods are still important for molecules in solution, for liquid and glassy phases and for supported species. In Ziegler-Natta catalysis, with its extension to the new organometallic single site catalysts, such situations are numerous. A survey of spectroscopic studies would be extremely extensive and well beyond the format of this work; several examples are given in the cited papers.

NMR has by far been the most extensively used spectroscopic technique to investigate these catalyst systems as well as the polymer product. The amount of interpretable information in NMR spectra is very high, with chemical shifts, coupling constants and relaxation times. All these are measurable with high accuracy, and their interpretation has an extensive experimental and theoretical basis.

For infrared spectroscopy (IR) the situation is partly very different. The spectra of condensed samples often show broad and overlapping peaks, and it may be impossible to establish complete assignments of spectra. Impurities in quantities up to percents level may escape detection under unlucky conditions, and for complex mixtures the spectra may be impossible to sort out. Quantum chemical calculations can support the interpretations, but anharmonicity and approximations in the calculations yield significant inaccuracies in calculated frequencies and intensities.

However, whereas the time-scale for NMR spectroscopy is long enough for several reactions to occur, the IR spectrum gives an almost instantaneous picture of the sample. NMR reveals ligand exchange, but not always the species involved in the exchange. For a reaction AB + B* º AB* + B, NMR can tell that it happens and that it is fast. IR spectroscopy may fail to do so, but in stead it can tell whether there is a true intermediate ABB*. If there is an intermediate, IR may indicate its concentration, if not IR may indicate an upper limit.

The aim of the present work is to show that there are still issues where other methods may be more powerful than those traditionally used. For some of those, IR may turn out to be a secret window to the heart of the catalyst, yielding new and important information.

Experimental methods.

Besides standard methods, the following techniques have been used in referred studies.

Pseudo Matrix Isolation (PMI): The sample was collected on a salt window placed in a vacuum chamber and cooled by liquid nitrogen. Two canulas allowed simultaneous deposition of vapours like ethyl benzoate (EB) and TiCl₄. A tubular furnace allowed evaporation of MgCl₂ for deposition on the sample window. The method is described by Ystenes and Rytter[1], and was used for the MgCl₂/EB/TiCl₄ catalyst system.

In situ liquid/solid cell: Two optical cells made by a pair of salt or Si/Ge windows were included in a circulation loop. Between one of the window pairs a solid could be fixed, whereas the other pair constituted a liquid only cell. The loop allowed manipulation of samples under inert conditions as spectra were recorded. The method is described by Bache and Ystenes [2] as used for components of the MgCl₂/TiCl₄/triethyl aluminium (TEA) catalyst systems, and by Eilertsen et. al [3] as used for zirconocene/methyl aluminoxane (MAO) systems.

Diffuse reflectance (DRIFTS): Incoming IR radiation was reflected from an open bed of catalyst powder, and the diffusely reflected radiation was recorded. The method is described by Bade and Blom [4], and was used for studies of gas phase polymerisation on chromium catalysts.

Results and discussion

Our IR studies started with the traditional MgCl₂/TiCl₄/EB catalysts, activated with triethyl aluminium (TEA) as co-catalyst, a system much studied in the literature. EB turned out to be a very useful indicator in IR studies of the catalyst components. Results from IR studies of EB and EB complexes are summarised in the Table 1 above.

The spectrum of EB is complicated and difficult to interpret, and a thorough investigation, including force field calculations, studies of isotopic congeners (deuterated and ¹⁸O-substitutions), different phases and a number of complexes - some of them partly deuterated or ¹⁸O-labelled - was needed as a basis for further work.[1]

PMI studies of the TiCl₄/EB compounds revealed spectra of dimeric 1:1 and monomeric 1:2 complex, both described by x-ray crystallography[1i]. With large excess of TiCl₄ and at liquid nitrogen temperatures, a new compound was formed, probably a 2:1 complex. The very low C=O stretching frequency indicated a structure with three Ti-Cl-Ti bridges.[1b,1d] Dissolving EB in liquid TiCl₄ at room temperature in the in situ cell gave a new, probably 1:1 monomeric complex with five-coordinated titanium.[2c]

Reacting MgCl₂ and EB yielded a 1:2 complex, probably with chains of sidesharing MgCl₄ square planar units, and a 1:1 complex possibly with five-coordinated Mg.[1f] The stoichiometries were determined by thermogravimetric analyses (0.1K/min) combined with IR investigations of intermediates, and elemental analysis. Similar studies revealed two TiCl₃-EB complexes, both probably monomeric, with 1:1 and 1:3 stoichiometries.[1g]

TEA forms 1:1 complexes with EB, at all concentrations and concentration ratios. With diethyl aluminium chloride (DEAC) and ethylaluminium dichloride (EADC), species like TEA-DEAC-EB, (DEAC)_n-EB and (EADC)_n-EB are formed.[5] These results show that Al-Cl-Al bridges form readily, which is in agreement with earlier observations for a number of Al-Cl species.[6]

An early study showed the presence of a C=O stretching band at approx. 1550 cm^-1 in a catalyst prepared by comilling (TiCl₄EB)₂ with MgCl₂.[7] The intensity of this band was roughly proportional to the activity of the catalyst, and could be due to a precursor of the active center. The frequency of the band was close to that found for the (TiCl₄)₂EB complex [1b], and should therefore be anticipated to be due to species with a related structure.

This finding was investigated further by preparing TiCl₄-EB complexes in the PMI cell, and then depositing MgCl₂ vapour on this sample. Mostly this yielded MgCl₂-EB and TiCl₄-EB complexes, but in one case new bands appeared.[1h] The C=O stretching frequency of the new specie was close to the one found in the catalyst mentioned above, and was assigned to a six-coordinated Ti center, with three Ti-Cl-Mg bridges and one EB unit on Ti.

Catalyst preparation and activation was then investigated in situ. First solid MgCl₂ was reacted with TiCl₄ in heptane, and three weak bands 476, 461 and 450 cm^-1, obviously belonging to terminal Ti-Cl, were observed. When TEA was added, the bands disappeared quantitatively.[8] The catalyst thus prepared and activated polymerised propene in situ. [2a]

When using MgCl₂-EB instead of MgCl₂, a broad feature was observed at 476 cm^-1, probably a poorly resolved doublet. Also this band disappeared quantitatively upon activation with TEA. The spectrum of the MgCl₂/EB/TiCl₄ catalyst included a band at 1550 cm^-1,[8b] the same frequency as found for the band earlier assigned to a MgCl₂/TiCl₄/EB surface complex.[7]

Such a low C=O stretching frequency reflects a strong coordination of EB to the metal centre. Crystallographic studies of TiCl₄-EB complexes show a profound trans effect, and the structure of (TiCl₄-EB)₂ show that EB prefers to be positioned trans to a terminal Cl, not to a bridge.[1i]. The band at ca. 1550 cm^-1 therefore suggests that there are multiple Ti-Cl-Mg bridges on the active centre, which forces EB to be trans to a bridge. As a bridging Cl is a weaker ligand than terminal Cl, this causes a stronger Ti-EB bond and hence a lower C=O stretching frequency.

Upon activation, some TEA should be converted to DEAC, but by the actual concentrations, TEA and DEAC are difficult to observe directly by IR. The liquid phase of the catalyst system was therefore investigated by adding EB to see which complexes are formed. When EB was present from the beginning, DEAC-EB complexes were formed immediately. When EB was absent during the activation, but added at a later stage, then first only TEA-EB was seen; DEAC-EB appeared after a few minutes. Hence, DEAC is not released during the activation if EB is absent. [8b]

The surprising, but very robust, conclusion is that there are no terminal chlorides on the titanium centres on the activated catalyst. The only plausible explanation found for this observation is that terminal Ti-Cl is turned into Ti-Cl-Al bridges through reaction with TEA or DEAC. The observed lack of DEAC in the solvent after activation, supports this interpretation. These results, and the results from the Al-EB studies [5], indicate that this complexion may be quantitative.

A schematic model of the active centre that fits the observations is given above. One cannot exclude the possibility of a five-coordinated Ti (including the monomer site), or that there are only two Mg-Cl-Ti bridges instead of three. Furthermore one cannot exclude the possibility that the active centres are minority species not seen in the IR spectrum, but it is difficult to rationalise why these species should behave differently from those observed.

Until autumn 1998, we assumed that MAO in solution consisted of a multitude of molecular species in equilibrium with each other, and with the average molecular weight decreasing with increasing TMA content. The first test of this hypothesis by the in situ technique proved this assumption to be wrong. The MAO spectrum (commercial MAO) was uninfluenced by the TMA, the spectrum of the added TMA was just superimposed on the MAO spectrum.

MAO solutions thus turned out to be a mixture of two independent components: A "true MAO" with a CH₃/Al ratio close to 1.5, and free TMA. The results were consistent for several different MAO batches, including commercial and TMA-depleted MAO, both at room temperature (20-25°C) and at temperatures about 80°C.[3] We were never able to make toluene-soluble MAO with CH₃/Al ratios significantly below 1.5. "True MAO" itself is probably a mixture of several related, stable species.

The IR spectra also reveal that MAO includes both terminal and bridging methyls. Only bridging methyls could be substituted with chlorides by adding dimethyl aluminium chloride (DMAC). The product, a chlorinated MAO (MAO-Cl), was inactive as cocatalyst for Cp*₂ZrCl₂. Adding TMA gave no effect, but MAO-Cl did not block the effect of later additions of MAO. Hence the methyl bridges must have been the active part of the MAO.[9]

The infrared spectra of zirconocenes have the drawback of having few bands that are influenced by activation. Bands that could be assigned directly to the Zr-CH₃ group would have been very useful, but all such bands in the frequency range studied were weak and overlapping with other bands, hence none were suitable. However, one band turned out to be diagnostic, a band at slightly above 800 cm^-1, assigned to out-of-plane hydrogen deformation on the cyclopentadienyl ring. In another paper in this proceeding [10] it is described how this band have been utilised to investigate the activation of the catalyst with MAO and other co-catalysts.

The results of this study appear to verify that MAO causes the formation of Zr-Me-Al bridges, and that a monomer is needed to break this bridge and make the zirconocene catalyst active. This result has profound implications for the understanding of the kinetics of the polymerization.

So far we have not been able to observe directly monomer complexion on Ziegler-Natta catalysts. However, ethene complexed on chromium catalysts has been observed through DRIFTS. An example is given below. The diagnostic feature is a band at 2997-2999 cm^-1 in the C-H stretching range, and it is clearly seen how the monomer band appears before any polymer can be observed. The C=C stretching band is difficult to observe, but was found at a weak feature at 1584 cm^-1.[4]

Concluding remarks

The field of olefin polymerisation catalysis would not have reached anything as far as it has, if IR spectroscopy were the only molecular spectroscopy tool available. Fortunately this is not so, but unfortunately one may have relied too much on NMR alone. Obviously there are blind areas in the sight of NMR, which can better be studied with other methods. Here are some results that were easily obtained through IR, but where other methods have failed:

1. The complex stoichiometries in DEAC and EADC complexes with EB.
2. The absence of any terminal Ti-Cl bands in TEA-activated MgCl₂/TiCl₄ catalysts.
3. The presence of molecular species in MAO with structures and CH₃/Al ratios that are unaffected by addition or removal of TMA.

The significance of the last result for the understanding of the effect of MAO as co-catalyst is obvious. If TMA does not modify the MAO structure, then its influence on the polymerisation must be direct and independent from the effect of MAO.

These results were published at the Hamburg conference in 1998[3b], and we immediately started to look for a structure that was compatible with the new experimental results. To explain the results, we needed a structure with a particular stability relative to addition or removal of TMA, but which also allowed rapid methyl exchange with TMA. With this new piece of information at hand, we quickly arrived at a model that fitted all experimental results.[9] Two other groups have independently arrived at similar conclusions about the most likely structure of MAO,[11a,b] and a fourth has earlier suggested a similar structure for a fraction of tert-butyl aluminoxane.[11c]

The effect of the second result on the modelling of the active centre of traditional TiCl₄-based catalysts is profound. Turning the terminal chlorides into bridges should influence both the strength of the monomer complexion and the steric discrimination of the insertion. Models based on centres with terminal Ti-Cl ligands should simply not be expected to yield correct results.

The first finding is less dramatic, but may be important for the modelling of the chemical equilibria of the system.

It may therefore not be impertinent to suggest that there may still be other important information undisclosed by the work done so far but ready to be unveiled by IR spectroscopy.

Acknowledgement

Financial support from the Norwegian Research Council (NFR) under the Polymer Science Programme is gratefully acknowledged. Several scientist have contributed to the work by quantum chemical calculations that allowed proper interpretations of the data. They are found as coauthors of the referred papers.

References

(theses are available from corresponding author)