Ethylene Oxychlorination on CuCl2 based Catalysts: Mechanism and Kinetics
Abstract
Ethylene oxychlorination is one of the major industrial sources of vinyl chloride, the monomer of poly-vinyl chloride (PVC). The oxychlorination reaction is catalysed by alumina supported, CuCl2 based catalysts which undergoes a redox reaction cycling between cupric and cuprous copper. Due to deactivation phenomena such as the volatilization of cuprous copper and the highly exothermic reaction, industrial catalysts are always doped to a varying extent, typically by alkali, alkali earth and/or rare earth metals. Thus, dopant effects on the individual redox reactions are of great importance as it controls the conversion, reactor temperature and catalyst deactivation as well as the quantity and properties of the active sites.
While the catalyst oxidation state can be inferred from the cumulative reactant consumption in step transient experiments, this is not the case when ethylene, oxygen and hydrogen chloride are simultaneously present (co-feed). In the past, X-ray absorption spectroscopies (XAS) have been used to track the active sites in the presence of one or all reactants. Although effective, synchrotron availability and laborious data treatment have restricted the application. To combat this, we have taken advantage of the fact that contrary to metallic or cuprous copper, d-d transitions are allowed in cupric copper ([Ar]3d9) which absorbs light in the near infrared spectrum. Following the step transient reduction (C2H4 or H2) and oxidation with operando UV-Vis-NIR spectroscopy through a fibre optic probe, the intensity followed the cumulative reactant consumption, and hence Cu oxidation state, with a high degree of linearity.
After normalizing the intensity with respect to the fully the oxidized and reduced catalyst, the intensity was related to the concentration of cupric copper, forming the calibration curve. Thus, at co-feed conditions, the catalyst oxidation state is quantitatively obtained from the normalized intensity with respect to time on stream or axial position with sub-second resolution. While unable to differentiate the various Cu2+ species through the d-d transition, operando UV-Vis-NIR spectroscopy is a cheap, fast and portable technique compared to XAS. The technique is expected to be transferrable to other systems where the catalyst partakes in redox reactions and the oxidation state can be inferred by the absorption of visible light.
A methodology of step transient reduction and oxidation experiments was adopted in which dopant effects on the number of active sites, reaction rate and turnover frequency could be explored. By co-plotting the step transient reduction and oxidation reaction rates as a function of the catalyst oxidation state, the two non-linear lines intersect at a given oxidation state where the reduction and oxidation rates are equal. The intersection thus indicates the steady-state operating point (Cu oxidation state and reaction rate) at co-feed conditions given a fast and kinetically irrelevant hydrochlorination step. The rate diagram showed that K addition decreases the reducibility and increases the oxidation rate while Ce addition increased the rate of both redox reactions. Compared to the neat catalyst, both additives increased the fraction of reducible Cu (number of active sites) and favoured an increased oxidation state at co-feed conditions. Finally, the notion of a rate determining step should be used with caution as increasing either of the redox rates will shift the steady-state operating point towards higher co-feed conversion.
Complementary to the quantitative UV-Vis-NIR technique, CuCl2 supported on a hydrotalcite derived mixed MgAl oxide (HT700) was used as a model system where a small shoulder in the UV region could qualitatively differentiate CuCl2 from Cu2OCl2. It showed that heating in an inert atmosphere alone was not enough to completely convert all Cu to the active CuCl2 phase which was only achieved after exposure to HCl. The shoulder was further used to estimate a hydrochlorination energy barrier considerably smaller than the redox steps while the co-feed spectrum indicated CuCl2 as the dominating cupric phase. The K doped, HT700 supported catalyst further increased stability at similar conversion to the neat catalysts on both supports, while the K doped, alumina supported catalyst was considerably more active. It is likely that both the Cu and K loading on HT700 could be further optimized.
The non-linearity of the rate diagram, supported by deconvolution, TPR and DFT calculations, strongly suggests at least two different active sites present in both redox reactions: i) four- and three coordinated Cu during the reduction and ii) two- and three coordinated Cu during the oxidation. Especially with respect to the reduction, DFT suggest that gas phase ethylene simultaneously extracts two Cl from the four coordinated site while absorbing on a Cu atom and successively attracting two Cl atoms on the three coordinated. Thus, the kinetic model needs to account for both gas- and solid phase reactions in order to capture the highly dynamic system. In both redox steps on the modelled neat catalyst, the tree-coordinated site was not initially present, bur formed as the reaction proceeded. It had the highest activation barrier compared to the respective four- or two coordinated sites, but the reduction activation energy was larger than the oxidation barrier. Despite this, the oxidation rate is still the slower step for most oxidation states due to the small pre-exponential constant, indicating that the oxidation is an entropy driven process. Likely explained by the loss of entropy during the absorption, dissociation and incorporation of oxygen from the gas phase into the solid catalyst.
Finally, the co-feed operating point obtained by operando UV-Vis-NIR spectroscopy was found to correlate well with those predicted by the rate diagram and kinetic model for catalysts not suffering from volatilization of cuprous copper. As the kinetic parameters was obtained from step transient experiments and not further fitted predict the co-feed performance, both the model and the experimental approaches mutually support each other’s validity and conclusion.