Methanol carbonylation revisited: thirty years on

Monsanto initiated development of its rhodium- and iodide-catalysed process for the carbonylation of methanol to acetic acid in 1966. Ownership of the technology was acquired in 1986 by BP Chemicals who have further extended it. The work of the Sheffield group, in developing a deeper understanding of the mechanism of the process, is reviewed. The rate-determining step in the rhodium–iodide catalysed reaction is the oxidative addition of methyl iodide to Rh(CO)2I2–1a: the product from this reaction, the reactive intermediate MeRh(CO)2I3–2a has been detected and fully characterised spectroscopically. The rates of the reversible reactions linking 1a, 2a and the acetyl complex (MeCO)Rh(CO)I3–3a, as well as activation parameters for several of the processes involved, have been measured. The efficiency of methanol carbonylation arises primarily from rapid conversion of 2a into 3a, leading to a low standing concentration of 2a, and minimising side reactions such as methane formation. By contrast, in the iridium-catalysed carbonylation, for which similar cycles can be written, the reaction of MeIr(CO)2I3–2b with CO to give (MeCO)Ir(CO)2I3–4b is rate determining. Model studies show that while kRh/kIr is ca. 1 : 150 for the oxidative addition, it is ca. 105–106 : 1 for migratory CO insertion. The migratory insertion for iridium can be substantially accelerated by adding either methanol or a Lewis acid (SnI2); both appear to facilitate substitution of an iodide ligand by CO, resulting in easier methyl migration. The carbonylation of higher alcohols (ROH) has also been successfully modelled; the corresponding alkyl iodides (RI) react much more slowly than MeI with M(CO)2I2–, but again give the acyls (RCO)M(CO)I3– for M = Rh and the alkyls RM(CO)2I3– for M = Ir. The greater stability of MeIr(CO)2I3– compared with MeRh(CO)2I3– accounts for the very different characters of the reactions catalysed by the two metals. It is suggested that the broad features of the Rh/Ir reactivities can be rationalised since the M–C bond to a 5d metal is generally stronger than that to the corresponding 4d metal; thus if metal–ligand bond making plays a key role in a step, then the 5d metal is more likely to react faster (e.g. in the oxidative addition), but if a metal–ligand bond-weakening or -breaking step plays a key role in a process (e.g. in the migration), it is likely that the 4d metal will be faster.