Heterogenisation of a Carbonylation Catalyst on Dispersible Microporous Polymer Nanoparticles

The methanol carbonylation catalyst, cis-[Rh(CO) 2 I 2 ] – , has been heterogenised within a dispersible microporous polymer support bearing cationic functionality. The microporous polymer has a core-shell structure in which the porous and insoluble core (a co-polymer of divinylbenzene and 4-vinylpyridine) is suspended in solution by long hydrophilic poly(ethylene glycol) chains, allowing a stable suspension of the nanoparticles to form. Incorporation of 4-vinylpyridine as a co-monomer allows post-synthetic modification to generate N-methylpyridinium sites for electrostatic attachment of the anionic rhodium(I) complex. The dispersibility of the polymer-supported catalyst material facilitates the use of in situ transmission IR spectroscopy to obtain kinetic data for the oxidative addition of iodomethane to immobilised cis-[Rh(CO) 2 I 2 ] – (the rate-limiting step of the carbonylation cycle). Remarkably, the oxidative addition proceeds faster than for the homogeneous system (Bu 4 N + counter-ion, CH 2 Cl 2 , 25 °C). The polymer-supported catalyst was found to be active for methanol carbonylation, with a turnover frequency similar to that of the homogeneous analogue under the same conditions (10 bar CO, MeI/MeOH/CHCl 3 , 120 °C). The Abstract The methanol carbonylation catalyst, cis -[Rh(CO) 2 I 2 ] – , has been heterogenised within a dispersible microporous polymer support bearing cationic functionality. The microporous polymer has a core-shell structure in which the porous and insoluble core (a co-polymer of divinylbenzene and 4-vinylpyridine) is suspended in solution by long hydrophilic poly(ethylene glycol) chains, allowing a stable suspension of the nanoparticles to form. Incorporation of 4-vinylpyridine as a co-monomer allows post-synthetic modification to generate N -methylpyridinium sites for electrostatic attachment of the anionic rhodium(I) complex. The dispersibility of the polymer-supported catalyst material facilitates the use of in situ transmission IR spectroscopy to obtain kinetic data for the oxidative addition of iodomethane to immobilised cis -[Rh(CO) 2 I 2 ] – (the rate-limiting step of the carbonylation cycle). Remarkably, the oxidative addition proceeds faster than for the homogeneous system (Bu 4 N + counter-ion, CH 2 Cl 2 , 25 °C). The polymer-supported catalyst was found to be active for methanol carbonylation, with a turnover frequency similar to that of the homogeneous analogue under the same conditions (10 bar CO, MeI/MeOH/CHCl 3 , 120 °C). The supported catalyst is easily recovered and is shown to maintain comparable activity upon recycling. A rhodium carbonylation catalyst is immobilised on dispersible microporous polymer nanoparticles, facilitating recyclability and in situ kinetic measurements for a key step of the catalytic cycle.


Introduction
Homogeneous transition metal catalysts are widely used for organic transformations on laboratory and bulk scales. 1 They typically benefit from high activity and selectivity under relatively mild conditions, and can be tuned by choice of metal, ligand, solvent and reaction conditions. Compared with heterogeneous catalysts, however, they suffer from the need to separate product from a solution phase containing the dissolved catalyst. Typically, a distillation is required under conditions that can lead to catalyst degradation and incur significant costs. 2 In order to combine the benefits of both homogeneous and heterogeneous catalysis, a common approach is to immobilise a soluble metal complex on an insoluble support such as an inorganic oxide, zeolite, metal-organic framework or organic polymer. 3 One of the most significant commercial bulk processes that uses homogeneous transition metal catalysis is the carbonylation of methanol to produce acetic acid (global demand ~20 million tonnes p.a. 4 ), employing rhodium-or iridium-based catalysts. [5][6][7][8][9] The Monsanto process, developed in the 1960s, employs a rhodium catalyst and an iodide promoter, achieving very high selectivity (> 99% based on MeOH). 10 The active rhodium complex is a square-planar Rh(I) species, cis-[Rh(CO) 2 I 2 ] -, which undergoes oxidative addition of methyl iodide, derived from the methanol feedstock (Scheme 1).
To maintain high activity and catalyst stability, as well as to improve solubility, a relatively high concentration of water (~10% wt.) is used in the reaction medium. At lower water concentrations, an inactive Rh(III) species, [Rh(CO) 2 I 4 ] -, can accumulate as an intermediate in a competing water gas shift (WGS) reaction. 11 This species can dissociate CO and lead to precipitation of RhI 3 in parts of the plant that have lower CO partial pressure. Although high water concentration alleviates this, it increases the cost of product purification, by distillation, to produce dry acetic acid. Variations of the Monsanto process have been introduced that allow operation at lower water concentration.
For example, the Celanese Acid Optimisation (AO Plus) technology employs lithium iodide as an additive that stabilises the Rh catalyst and maintains high carbonylation activity at significantly lower water concentration. 12 16 BP Chemicals' Cativa™ process uses a promoted iridium/iodide homogeneous catalyst system that achieves high activity and selectivity for methanol carbonylation at low water concentration. 17,18 Another approach to minimise problems with solubility at low water concentration is to heterogenise the catalyst by immobilisation on a solid support. 5,16,19 A range of support materials that bind the Rh catalyst directly through a covalent interaction have been investigated, including inorganic oxides, 20, 21 zeolites, [22][23][24][25] polymers, 26-31 carbon 32 and covalent triazine frameworks. 33 Many of these systems exhibit reaction rates slower than that of the homogeneous process, likely due (in part) to the covalent tethering of the catalyst, which modifies the first coordination sphere of the Rh complex. Leaching of catalyst due to lability of the Rh-support interaction can also be problematic.
An alternative strategy for heterogenisation, which preserves the first coordination sphere of the catalyst, is immobilisation of the anionic Rh complex on a cationic support by electrostatic interactions. [34][35][36][37] This approach has been observed to result in less catalyst leaching compared with covalent tethering. In 1981, Drago et al. described the effective immobilisation of cis-[Rh(CO) 2 I 2 ]on polymer supports based on N-methylated polyvinylpyridines. 38 The carbonylation activity was equal to the homogeneous system at 120 °C with minimal leaching of the supported catalyst. This ionic attachment approach was adopted in the Acetica™ process developed by Chiyoda and UOP which employs a polyvinylpyridine resin tolerant of catalytic reaction conditions. 39 The pyridyl groups are quaternised in the presence of iodomethane, and the anionic catalyst, cis-[Rh(CO) 2 I 2 ]is bound electrostatically to the support. The supported catalyst showed no deactivation after continuous operation for 7,000 hours and its activity is competitive with the homogeneous process, with decreased by-product formation via the WGS reaction due to the lower water concentration employed.
The catalytic mechanism for this type of supported system was investigated by Haynes et al.
using cross-linked poly(4-vinylpyridine-co-styrene-co-divinylbenzene) in the form of thin films suitable for in situ transmission IR spectroscopy. 40  Microporous polymers have also been widely investigated for heterogeneous catalysis, [41][42][43] including their use to encapsulate metal catalysts electrostatically. [44][45][46] Recently, a new class of dispersible porous polymer has been reported, synthesised using Reversible Addition-Fragmentation chain Transfer-mediated Polymerisation-Induced Self-Assembly (RAFT-mediated PISA). 47 Co-polymerisation of divinylbenzene (DVB) and fumaronitrile (FN) was mediated using a hydrophilic macromolecular chain transfer agent (macro-CTA) in an anti-solvent for the growing polymer chain. This resulted in the formation of polymer nanoparticles with a core-shell morphology. The high degree of crosslinking within the core of the sample means that the material

Synthesis and characterisation of supported catalyst
Synthesis of the dispersible porous polymeric material 1 employed a RAFT-mediated PISA approach analogous to that reported by James et al. 47 whereby DVB and 4VP monomers were copolymerised in a 1:1 (w/w) ethanol:water solvent mixture using a PEG-based macro-CTA (Scheme  (Table S1). The IR spectra of all three materials (Figure 1  differential pore size distribution (offset successively by 0.1 cm 3 g -1 , micropore region highlighted) of 1 (black), 2 (red) and 3 (blue). Table 1. Surface areas and pore volumes of polymer nanoparticle samples. To assess the porosity and surface area of 1-3, volumetric gas sorption studies using nitrogen gas at 77 K were performed on each material (Figure 2 and S1). Resulting parameters for 1-3 are given in Table 1. The specific BET surface area (SA BET ) of 1 was determined to be 366 m 2 g -1 , with both micropores and mesopores present in the material. After quaternisation of 1 with MeI, the surface area of 2 is lower (166 m 2 g -1 ) and the micropores are lost. A further reduction in surface area (to 114 m 2 g -1 ) is observed upon incorporation of the Rh catalyst in 3. Pore size analysis of 2 suggests that PSM of 1 results in a loss of the micropores of the material. Initially, the micropore volume (V 0.1 ) of 1 is 0.15 cm 3 g -1 but this falls to 0.06 cm 3 g -1 upon quaternisation of the material to yield 2. A further and more minimal decrease of micropore volume to 0.04 cm 3 g -1 is observed upon incorporation of the Rh complex to yield 3. However, the mesopores are completely lost in 3 which suggests occupation by the Rh complex. The ratio of micropore: total pore volume is unchanged upon transformation of 2 to 3, further suggesting that the Rh complex is situated in the mesopores rather than the micropores. This correlates with the loss of hysteresis at ~ 0.8 P/P 0 initially present on the desorption branch of the isotherm of sample 1 indicting a further loss in mesoporosity. Finally, the large uptake of gas at high relative pressure (> 0.9 p/p 0 ) is attributed to the aggregated morphology of the material and shows that even after two PSM steps the morphology of the material remains intact, as further evidenced by transmission electron microscopy (TEM, see below).
The materials are able to form stable dispersions in a wide variety of solvents as previously reported for the PEG-DVB/FN system. 47 This is attributed to the presence of the long solubilising PEG chains which allow the insoluble porous DVB/4VP core network to be dispersed in solution.
The stability of an undisturbed dispersion of 1 in CHCl 3 was monitored by UV-Vis spectroscopy for 72 h ( Figure S2). Any change in signal intensity is attributed to the material settling out of solution so correlates directly to loss of stability of the dispersion. After 72 h it was found that only ~15 % of the sample had settled out of solution and the majority of the polymer was still present as a stable dispersion, indicating suitability for use in a catalytic system as proposed.
Small-angle X-ray scattering (SAXS) studies were performed on 5% w/w dispersions of samples 1-3 in MeOH (Figure 3a) to gain insight into the size and morphology of nanoparticles in each sample. The data were successfully fitted using the same two-population model as used previously for PEG-FN/DVB dispersible porous polymers, 47 which suggests the presence of aggregates of smaller primary assemblies, presumably due to both inter-and intra-particle covalent crosslinking ( Figure S3). Fitting these data indicated that mean diameters for the primary particles (D 1 ) ranged from 25 nm to 34 nm, with very broad size distributions in each case (standard deviation ≥ 60%).
Remarkably, the mean diameter of the overall aggregates (D 2 ) remained relatively constant for each sample, with values of 101-104 nm being determined (Table 2). Additionally, the standard deviation in the D 2 values in all cases was significantly lower (≤ 20%) than that observed for D 1 .
Importantly, this proposed hierarchical structure is supported by TEM images (Figure 3b-d).

Reaction of supported catalyst 3 with iodomethane
After soaking a sample of 3 in neat MeI overnight, the IR spectrum of the resulting material displays only one n(CO) absorption in the terminal metal carbonyl region, at 2058 cm -1 , as well as a broad band at ~1700 cm -1 (Figure 4a). This is consistent with formation of a Rh(III) acetyl   The stability of dispersions of the polymer-supported complex mean that the organometallic reaction steps can be monitored in situ using transmission IR spectroscopy. Figure 4b Figure 4c shows a typical absorbance vs. time plot generated from one of these experiments with the data fitted to a first-order decay curve. It is apparent from this plot that the data are not perfectly fitted by a single exponential decay function. There appear to be two phases during the reaction with an initial faster decay followed by a slower downward drift.
Empirically it was found that the kinetic data are better described using a double exponential decay, with contributions from two different pseudo first-order rate constants (k obs 1 and k obs 2 ) as illustrated in Figure 4d. Analogous experiments were repeated across a range of MeI concentrations ( Figures   S4-S8), giving the pseudo first-order rate constants listed in Table S2. This behaviour differs from that of an insoluble polymer-supported system studied previously, which gave a rate constant very similar to that measured in the solution phase (2.6 × 10 -5 mol -1 dm 3 s -1 ). 40 Hence the dispersibility of the polymer support in 3 appears to have a significant effect on the oxidative addition rate.

Catalytic methanol carbonylation
To demonstrate catalytic turnover for methanol carbonylation, in situ high-pressure IR (HPIR) spectroscopy was used to monitor a dispersion of supported catalyst 3 in chloroform containing methanol and iodomethane at 120 °C under 10 bar CO pressure. IR spectra collected over the course of four hours showed the growth of an absorption at 1741 cm -1 due to the n(C=O) of methyl acetate, formed by esterification of the carbonylation product, acetic acid, in the presence of excess methanol (overall 2MeOH + CO → MeCO2Me + H2O, Scheme 5). A plot of absorbance vs. time for this band (Figure 6a) is approximately linear and a value of the catalytic turnover frequency (TOF) of 11.4 h -1 can be determined from the slope. Data from an experiment using Bu 4 N[Rh(CO) 2 I 2 ] under the same conditions, also plotted in Figure 6a, give a TOF of 21.8 h -1 . The catalytic rate is lower for the polymer-supported system, despite the faster oxidative addition of MeI reported above, but the different temperature and solvent system in the model and catalytic reactions may be significant. The polymer-supported catalyst was recovered after the reaction by centrifugation and a subsequent experiment using the recovered catalyst showed that it retained activity comparable with the initial run (Figure 6b), thus demonstrating its recyclability. Polymer-supported catalysts of this type present the opportunity to simplify product/catalyst separation and remove catalyst solubility constraints, thereby enhancing process productivity and efficiency. More generally, we anticipate that these support materials will be applicable in the design of a range of hybrid homogeneous/heterogeneous processes that combine the benefits of each approach. The efficient mixing in solution and ability to incorporate molecular catalysts should ensure high catalytic activity, while the ease of recovery removes the need for more costly recycling procedures that can lead to catalyst degradation.

Synthesis of PEG-based macromolecular chain transfer agent (macro-CTA)
The PEG-based macro-CTA was synthesised in an identical procedure to that previously reported. 47 Dodecane thiol (0.60 mL, 2.5 mmol, 1 eq.) was added to a stirred suspension of K 3 PO 4 (0.53 g, 2.5 mmol, 1 eq.) in acetone (50 mL) and stirred for 10 min. Carbon disulfide (0.36 mL, 6 mmol, 2.5 eq.) was added to the suspension and left to stir for a further 10 min. PEG 113 -Br (10 g, 2 mmol, 0.8 eq.) in acetone (30 mL) was added to the suspension, which was left to stir overnight at room temperature. The solution was concentrated in vacuo and the crude product was precipitated by addition of n-hexane (100 mL). This was isolated by filtration, dissolved in acetone and the precipitation procedure repeated once more using n-hexane and then once using diethyl

Post-synthetic modification of 1 to form 2
A suspension of 1 (596 mg) in a mixture of CHCl 3 (60 mL) and MeI (6 mL

Kinetic measurements on reaction of 3 with MeI
A dispersion of 3 (~10 mg) in CH 2 Cl 2 (3 mL) was generated by sonication of the mixture for 30 min. The required amount of MeI was added to a 2 mL volumetric flask and this was made up to the mark with the suspension of 3 in CH 2 Cl 2 and shaken. A sample of the reaction mixture was injected into an IR liquid cell (CaF 2 windows, 0.5 mm path length) fitted with a thermostatted jacket (25 °C). IR spectra in the region 2200-1500 cm -1 were recorded at regular intervals, using a spectrum of the appropriate solvent mixture as the background reference. Absorbance vs. time data for the frequencies of interest were analysed using Origin software.

Catalytic carbonylation experiments
Reactions were monitored in situ by high-pressure IR spectroscopy using a cylindrical internal reflectance (CIR) 54 cell comprising an autoclave (Parr) modified to accommodate a crystalline silicon CIR rod, as described previously. 18,55,56 Spectra were recorded using a Perkin-Elmer Spectrum GX FTIR spectrometer fitted with an MCT detector. The cell was placed directly in the spectrometer sample compartment and aligned to maximise IR energy throughput using a tilt table.
A background spectrum was recorded using the appropriate solvent mixture at 120 °C. In a typical procedure, a dispersion of the supported catalyst 3 was generated by addition of dry CHCl 3 (5 mL)

Conflicts of interest
The authors declare no conflicts of interest.

S1. Materials and synthetic methods
Dry CH 2 Cl 2 , CHCl 3 , toluene and n-hexane were obtained from a Grubbs solvent purification system in which the solvents were degassed prior to being passed through activated alumina and a supported copper catalyst to remove protic contaminants and trace oxygen respectively.

S2. Instrumentation
Infra-red spectra of polymeric materials and supported catalyst were collected using a Perkin-Elmer 100 FTIR spectrometer. Samples were prepared by grinding with a 20× excess of pure KBr pre-dried overnight in a vacuum oven at 80 °C. The sample was pelletised at high pressure (10 tons) and analysed as a thin transparent disc.
Nitrogen gas sorption isotherms were collected at 77 K using approximately 100 mg of sample on an ASAP 2020 Micromeritics volumetric adsorption analyser. Prior to analysis all samples except 3 were degassed for at least 16 h at 120 °C under a vacuum of at least 10 -5 bar.
3 was degassed under the same vacuum level but without heating to avoid any decomposition of the supported Rh complex. BET surface areas were calculated over a relative pressure range of 0.01 -0.15 p/p 0 . Pore size distributions and pore volumes were calculated from the adsorption isotherms and modelled using the nonlocal density functional theory model (NL-DFT) for N 2 on carbon slit pores found within the micromeritics ASAP software.
Solid-state NMR samples were packed into 4 mm zirconia rotors and transferred to a Bruker Avance III HD spectrometer. 1D 1 H-13 C cross-polarisation magic angle spinning (CP/MAS) NMR experiments were measured at 125.76 MHz (500.13 MHz 1H) at a MAS rate of 10.0 kHz. The 1 H π/2 pulse was 3.4 μs, and two-pulse phase modulation (TPPM) decoupling was used during the acquisition. The Hartmann-Hahn condition was set using hexamethylbenzene.
The spectra were measured using a contact time of 2.0 ms. The relaxation delay, D 1 , for each sample was individually determined from the proton T 1 measurement (D 1 = 5 × T 1 ). Scans were collected until a sufficient signal to noise ratio was obtained, typically greater than 1094 scans.
The values of the chemical shifts are referred to that of TMS.
Carbon, hydrogen, nitrogen and sulphur elemental analysis was performed by combustion of an amount of sample in a stream of pure oxygen. The sample was placed in a tin capsule and introduced into the combustion tube of the Elementar Vario MICRO Cube CHN/S analyser via a stream of helium. Combustion products were analysed through after being passed through a copper tube to remove excess oxygen and reduce any NO x to N 2 . Gases were separated using a Thermal Programmed Desorption column and detected using a Thermal Conductivity Detector.
Iodide analysis was performed using the Schöninger flask combustion method in which an amount of sample is combusted in an oxygen-enriched environment, the resultant gases are absorbed and a titration is conducted to determine the iodide concentration.
Transmission electron microscopy (TEM) studies were conducted using a Philips CM 100 instrument operating at 100 kV and equipped with a Gatan 1 k CCD camera. A diluted solution of the polymer material (0.10% w/w) was placed on carbon-coated copper grids, allowed to dry and then exposed to ruthenium(VIII) oxide vapor for 7 min at 20 °C prior to analysis. The ruthenium(VIII) oxide was prepared as follows: Ruthenium(IV) oxide (0.30 g) was added to water (50 g) to form a black slurry; addition of sodium periodate (2.0 g) with stirring produced a yellow solution of ruthenium(VIII) oxide within 1 min. 4 SAXS patterns were recorded at a synchrotron source (Diamond Light Source, station I22, Didcot, UK; Experiment ID SM23501) using monochromatic X-ray radiation (X-ray wavelength λ = 0.999 Å, with scattering vector q ranging from 0.0027 to 0.25 Å -1 , where q = 4π sin θ/λ and θ is one-half of the scattering angle) and a 2D Pilatus 2M pixel detector (Dectris, Switzerland). Scattering data were reduced and normalised, with glassy carbon being used for the absolute intensity calibration utilising standard routines available at the beamline 5 and further analysed (background subtraction and data modelling) using Irena SAS macros for Igor Pro. 6 Figure S1. Cumulative surface area vs. pore volume for 1 (black), 2 (red) and 3 (blue).    Fits to a double exponential decay used equation S1 to give values of k obs 1 and k obs 2 .

S8. Catalytic carbonylation reactions -determination of TOF values
For each experiment, the rate of formation of methyl acetate was determined from the slope of