DOI: 10.1595/147106707X216855 Enhancement of Industrial Hydroformylation Processes by the Adoption of Rhodium-Based Catalyst: Part I DEVELOPMENT OF THE LP OXOSM PROCESS TO THE COMMERCIAL STAGE

By Richard Tudor* and Michael Ashley Davy Process Technology Ltd., 20 Eastbourne Terrace, London W2 6LE, U.K.; *E-mail: [email protected]

The adoption of a low-pressure rhodium-based catalyst system in place of high-pressure cobalt for the hydroformylation of propylene by reaction with carbon monoxide and hydrogen to produce butyraldehydes (an ‘oxo’reaction) has brought large cost benefits to oxo producers. The benefits derive from improved feedstock efficiency, lower energy usage and simpler and cheaper plant configurations. The technical and commercial merits of the ‘LP OxoSM Process’ for producing butyraldehydes have made it one of the best known applications of industrial- scale chemistry using a platinum group metal (pgm). Today, practically all butyraldehyde is made by rhodium catalysis, and this should provide convincing encouragement to researchers who are keen to exploit pgms as catalyst research materials, but are apprehensive as to the implications of their very high intrinsic value. It should also encourage developers and designers responsible for turning pgm chemistry into commercial processes, who may be daunted by problems such as containment and catalyst life. This article (Part I) reviews the background to the LP OxoSM Process, and its development to the point of first commercialisation. Part II, covering some of the key improvements made to the process and its use in non-propylene applications, will appear in a future issue of Platinum Metals Review.

2-Ethylhexanol (2EH) is the most widely used beginning in 1971. The principals were Johnson (‘workhorse’) plasticiser alcohol, and butanols – Matthey & Co. Ltd. (now Johnson Matthey PLC), the normal and iso isomers – are used as solvents The Power-Gas Corporation Ltd. (a former name or chemical intermediates. Both 2EH and of Davy Process Technology Ltd., now a sub- butanols are derivatives of butyraldehyde made sidiary of Johnson Matthey PLC) and Union from propylene by hydroformylation. From the Carbide Corporation (now a subsidiary of The early 1940s until the early 1980s, the world’s major Dow Chemical Company). Using rhodium-based producers of 2EH and butanols operated propy- catalysis, the LP OxoSM Process offered such great lene hydroformylation (often termed ‘oxo’) economic advantages over the established cobalt- processes for producing the required butyr- catalysed processes, as well as technical elegance, aldehyde using a cobalt catalyst system. This deliv- that many cobalt systems were replaced by brand ered poor conversion and low selectivity of the new plants. In the thirty years or so since the LP principal feedstock, propylene, to the desired OxoSM Process was first introduced, it has main- products, in complex and cumbersome plants tained its position as the world’s foremost oxo operating at high pressure. process, having undergone much improvement The ‘Low Pressure Oxo’ process (LP OxoSM and refinement. About two thirds of the world’s Process) was developed and then licensed to the butyraldehyde is now produced in LP OxoSM oxo industry through a tripartite collaboration plants. Most LP OxoSM systems are licensed

Platinum Metals Rev., 2007, 51, (3), 116–126 116 plants, nearly all of which have been built under The engineering contracting company drew on its licences granted by Davy Process Technology (1) strong background in process engineering to working in cooperation with The Dow Chemical investigate the commercial potential for a low- Company (2); the remainder are plants owned and pressure route to butyraldehyde. With the operated by Dow’s Union Carbide subsidiary (3). publication of patents by Union Carbide (e.g. (4)) This article (Part I) reviews the background to and Johnson Matthey (e.g. (5)), the three parties the LP OxoSM Process, addressing some of the realised that they had a common interest, so in challenges that faced its developers and designers 1971 they launched a joint development pro- in planning the first commercial plant, and during gramme to convert the laboratory rhodium-oxo the period immediately following commercialis- chemistry into a commercial process with a view ation. Insights are given on the chemical function to exploiting its technical merit. of the homogeneous liquid-phase catalyst system. The focus of the collaboration was a process In Part II, some examples of advancements of the for the hydroformylation of propylene using a mix- technology in the years following the first round ture of carbon monoxide and hydrogen (in the of licensing will be outlined. form of synthesis gas) to produce normal butyraldehyde and iso-butyraldehyde according to The Beginnings of the Development Reaction (i): Commercialisation of the LP OxoSM Process was the culmination of an intensive joint effort in 2CH3CH=CH2 + 2CO + 2H2 chemistry and engineering by the three compa- → CH3CH2CH2CHO + (CH3)2C(H)CHO (i) nies, dating back to 1964. Early exploratory work normal butyraldehyde iso-butyraldehyde by the chemicals producer Union Carbide in the U.S.A. demonstrated promise for rhodium coordi- The normal butyraldehyde isomer is usually nation complexes in solution as hydroformylation more highly valued than iso-butyraldehyde. A catalysts at low pressure, yielding a high propor- much improved normal-to-iso yield ratio observed tion of the straight-chain aldehyde product and in the laboratory with rhodium catalysis, as com- with high enough catalyst productivity to justify pared with the then-current commercial cobalt examining the commercial potential of rhodium. systems (i.e. about ten as opposed to typically The company obtained a basic patent for this between three and four) was unquestionably a key work in 1970 (4). In operating a number of high- driver for collaborative development. The high pressure oxo plants, Union Carbide had become selectivity of conversion of propylene to normal very familiar with cobalt systems and their short- butyraldehyde has since become a hallmark of the comings, and viewed the potential for rhodium LP OxoSM Process. with guarded excitement. At that time, all industri- The collaborators’ success in exploiting their al oxo production used the classic high-pressure development efforts (8) would eventually result in cobalt process described below, or a modifi- the LP OxoSM Process becoming the technology cation of it. of choice for many of the world’s oxo producers, Meanwhile, independent research by the late with whom Davy Process Technology negotiated Professor Sir Geoffrey Wilkinson (5–7) (later to licences on behalf of the collaborators. The high win a Nobel Prize for Chemistry) at Imperial reputation which the process would acquire College, London, supported by the precious metal because of its operating excellence and low refiner and processor Johnson Matthey, produced production costs, and a sustained growth in the results using a suitable coordination complex of markets for the end products, drove investment in rhodium (e.g. (5)) which basically reproduced or continuing research and process development complemented the Union Carbide findings. programmes aimed at improving the technology Johnson Matthey in turn approached The Power- to ensure the long-term sustainability of Gas Corporation (now Davy Process Technology). the process.

Platinum Metals Rev., 2007, 51, (3) 117 The Uses and Market for which include alkyd resins and adhesives for lami- Butyraldehyde nated glass. Therefore, the range of product The major use of butyraldehydes was, and still applications linked to propylene hydroformylation is, for the production of 2EH and butanols; see is increasing, and the growth in global demand for Figure 1. Normal butyraldehyde has always been butyraldehyde is between about 2% and 3% per the more valuable of the two aldehyde isomers, year. because unlike iso-butyraldehyde, it can be used to produce 2EH, by a sequence involving an aldol The Classic Cobalt ‘Oxo’ Route condensation reaction followed by hydrogenation In 1938, the German chemist Otto Roelen, of the aldol product. Furthermore, normal working in the laboratories of Ruhrchemie AG, butanol, produced by the direct hydrogenation of discovered that it was possible to react a mixture normal butyraldehyde, usually offers solvent and of carbon monoxide and hydrogen with an olefin derivative value superior to that of iso-butanol. to form products containing oxygen. Roelen’s ini- The world production levels of 2EH and normal tial work identified aldehydes and ketones in the and iso-butanols combined are presently about product, and the reaction was named the ‘oxo’ 2.5 million and 4.5 million tonnes respectively (9). reaction. Later work established that using olefins In today’s marketplace, butanol and its deriva- other than ethylene, the product is principally an tives have gained prominence from the long-term aldehyde, with very little ketone formation, and growth potential of water-based coatings (such as the reaction was renamed ‘hydroformylation’. indoor paints), driven by environmental consider- Both names are in common use, but ‘oxo’ has ations, with demand for butyl acrylate and become the more convenient and more interna- methacrylate esters particularly strengthened by tionally recognisable name. this trend. Meanwhile, most of the world’s 2EH is The process was commercialised in Germany esterified with phthalic anhydride to produce during the early 1940s, and was then widely used di(2-ethylhexyl) phthalate (DEHP), often referred throughout the world from the late 1940s to as dioctylphthalate or DOP, a plasticiser in wide onwards. The classic oxo process uses a cobalt cat- use for the production of flexible PVC. DOP has alyst in solution, operating at very high pressure in been around for a long time, and its market is the range 200 to 450 bar and at temperatures in the somewhat mature. Increasing amounts of 2EH range 140 to 180ºC. The active compound is are, however, being esterified with acrylic acid to cobalt hydridocarbonyl HCo(CO)4. A very high produce 2-ethylhexyl acrylate, used for adhesives, CO pressure is needed to ensure catalyst stability resins for latex, paper coatings and textile finish- during hydroformylation. The catalyst has to be ing. 2EH is also used to produce 2-ethylhexyl decomposed before the reaction product can be nitrate, a diesel fuel additive, and also lubricant recovered; therefore the process involves a cum- additives. Propylene hydroformylation is increas- bersome and costly catalyst recovery cycle. Using ingly being used as the first step in the production propylene as feedstock, the ratio of normal to iso of 2-ethylhexanoic acid, the wide applications of products is typically between about three and four,

Hydrogen n-Butyraldehyde Aldolisation Hydrogenation Product refining 2-Ethylhexanol Propylene LP OxoSM Syngas n-Butanol Hydrogenation Product refining n- + iso-Butyraldehyde iso-Butanol Hydrogen

Fig. 1 Schematic showing the production of oxo alcohols from propylene by the LP OxoSM Process

Platinum Metals Rev., 2007, 51, (3) 118 and the severe operating conditions mean that chemistry. For example, the absence of butanol in there is a high level of byproduct formation. the product meant that esters and acetals were not Derivatives of butanol present in the reaction formed – unlike with the cobalt process, for which product could adversely affect the environmental special measures were often needed to reduce their impact of the process. The two or three high- environmental impact. With LP OxoSM, the pressure cobalt plants remaining in operation for product could be worked up using a much simpler producing 2EH and butanols are very inefficient. system and, very significantly, the selectivity of They require considerable operator attention, are conversion of propylene to the preferred normal costly to maintain, and leave a poor impression on butyraldehyde was much better than with cobalt, the environment. the normal to iso ratio being improved about three- A modification of the classic cobalt process fold. These characteristics meant that propylene was commercialised in the 1960s, using as the could be converted to normal butyraldehyde more catalyst cobalt hydridocarbonyl trialkylphosphine, effectively and efficiently than had hitherto been

HCo(CO)3PR3. The process operates at a lower possible. The lower operating pressure compared pressure than the ‘classic’ process (around 50 bar), with cobalt eliminated or reduced the need for although a higher temperature is needed. With compression of the incoming synthesis gas, and propylene, the product shows much-improved lin- with a simpler distillation system needed to work earity, the normal-to-iso ratio being around seven. up the product butyraldehyde, overall energy The better selectivity to normal butyraldehyde is, demand was reduced. however, partly offset by an increase in reaction In the thirty years since rhodium was first used byproducts and an unavoidable production of commercially in hydroformylation, rhodium chem- alcohols during oxo synthesis. istry of one form or another has been adopted to meet at least 95% of world butyraldehyde demand. The Appeal of Rhodium-Catalysed First- and subsequent-generation LP OxoSM plants Hydroformylation account for more than 60% of this; see for exam- The first commercial plant to employ the ple Figure 2. (It is believed that the only remaining LP OxoSM Process to produce butyraldehydes suc- cobalt-based butyraldehyde production plants are cessfully started up in 1976. It was built by Union in Russia, all other cobalt plants having been shut Carbide at its petrochemical complex at Ponce, down, with many of them being replaced by Puerto Rico, with a capacity of 136,000 tonnes per LP OxoSM plants.) Rhodium catalysis has also made annum. As a result, the collaborators saw much inroads into non-propylene hydroformylation interest in the technology from both existing and applications, and the possibilities here may well new oxo producers. By the end of 1982, Davy increase with time. Some of these applications will Process Technology had licensed and designed ten be discussed in Part II. LP OxoSM plants that were built around the world. Several advantages of the LP OxoSM Process How the LP OxoSM Process was appealed at that time. The high activity and good Developed stability of the rhodium catalyst meant that it was The active catalyst used in the LP OxoSM not necessary to use the very high pressures need- Process is a hydridocarbonyl coordination com- ed with cobalt to retain catalyst integrity. The plex of rhodium, modified with triphenyl- LP OxoSM Process operated at less than 20 bar, phosphine (TPP) ligand. The catalyst is formed, and a lower reaction temperature of between 90 under process conditions, from rhodium and 100ºC resulted in less byproduct formation. acetylacetonato carbonyl triphenylphosphine

The lower temperature also brought other advan- (Rh(acac)(CO)PPh3 or ‘ROPAC’), or a suitable tages over cobalt catalysis. Overall, the product alternative catalyst precursor. From the outset, the mix from the reaction was much ‘cleaner’ and free process concept involved a homogeneous liquid- of many of the components formed using cobalt phase catalyst, in which the active catalyst species

Platinum Metals Rev., 2007, 51, (3) 119 Fig. 2 2-Ethylhexanol plant built by Sinopec Qilu Petrochemical Co. Ltd., China, employing the LP OxoSM Process are dissolved in the reaction mixture along with was able to optimise relationships among equip- reactants and reaction products, that is, normal and ment size and cost, reactant concentrations, iso-butyraldehyde and high-boiling aldol condens- feedstock consumption, and rhodium inventory, ation byproducts. The beauty of this route is that seeking the lowest possible overall produc- no extraneous solvent is necessary. A characteristic tion cost. of fully mixed homogeneous catalyst systems is that short molecular diffusion ranges encourage The Design of the First Commercial high reaction rates. These were achieved at labora- Process tory scale, suggesting that rhodium concentrations A key challenge to the developers and designers in the low hundreds of parts per million (ppm) of a first commercial LP OxoSM Process, resulting would be suitable. This in turn implied affordable from the intrinsic characteristics of the homoge- rhodium inventories for commercial-scale plants, neous catalyst system, was how best to separate the provided that the rhodium could be sufficiently butyraldehyde product from the reaction mixture. protected from poisoning and that excessive The solution adopted needed to address key fac- deactivation could be avoided. tors such as losses of unrecoverable reactants and In the early stages of development, Union product, energy usage and capital cost. There were Carbide needed to relate the rates of propylene however two very important additional considera- hydroformylation and of the main byproduct- tions that were directly linked to the use of a pgm forming reaction, i.e. the hydrogenation of of high intrinsic value: firstly, rhodium contain- propylene to propane, to the main process vari- ment, and secondly, the impact of process design ables. A statistical approach was used to design a on catalyst stability and catalyst life. For the for- set of laboratory experiments to develop kinetic mer, the physical loss of even relatively small models to determine these relationships. Models amounts of rhodium had to be avoided. As to the were also developed for the rate of formation of latter, much care had to be applied to the design of heavy byproducts resulting from aldehyde conden- the complete catalyst system, including the facili- sation reactions. Drawing upon these ties needed for preparing, handling, treating and mathematical models, Davy Process Technology processing the raw materials and the various

Platinum Metals Rev., 2007, 51, (3) 120 rhodium-containing streams. The object was to rities that could either poison the rhodium or avoid design measures that might unduly harm the inhibit its performance. To put that problem into catalyst, thus shortening its useful life. perspective, it is useful to look at some data for the At the outset of commercialisation, there was scale of butyraldehyde production that was then considerable uncertainty as to the likely lifetime of being contemplated: based on predictive models a rhodium catalyst charge in a commercial plant. generated from laboratory results, a commercial Moving up from the laboratory to industrial scale plant designed to produce 100,000 tonnes per was not seen in itself as significantly influencing annum of normal butyraldehyde would need a catalyst life; the more salient issue was that during charge equivalent to about 50 kg of fresh rhodium. laboratory testing, it had not been possible to Given the rhodium metal price at the time, the replicate completely the operating regime to which replacement value of this rhodium was about the catalyst system would be subjected in a com- U.S.$1 million (allowing for the processing mercial plant, due to various limitations, and this charge). During one year of operation, each kilo- introduced its own uncertainties. Only a certain gram of rhodium, if it could last that long in amount could be learned in the laboratory about service, would be exposed to more than 2,500,000 the tendency for the catalyst to lose activity. times its own mass of commercial feedstocks. The It was recognised, for example, that the catalyst question was whether there could be present in stability observed in small-scale rigs using high- that huge quantity of raw materials enough harm- purity feedstocks could not reflect the effects on ful contaminants, albeit at low concentrations, to catalyst life of impurities present in commercial threaten serious damage to the catalyst, even feedstocks. Nor, with the limitations of rig design destroying its activity, within an unacceptably and scope, would it be possible to simulate the short period of plant operation. The answer was a long-term effects on the catalyst of operating con- resounding ‘yes’. ditions that could well occur in the plant but The poisoning studies carried out by Union cannot be reproduced in the laboratory. Such con- Carbide had shown that certain likely contami- ditions might negatively impact catalyst life. The nants such as hydrogen sulfide and carbonyl predictive models for deactivation rates based on sulfide (often found in commercial propylene and laboratory studies therefore had their limitations, synthesis gas streams), and organic chlorides often and considerable further effort would be needed seen in propylene, were definite catalyst poisons. here as the technology developed. Despite the Other impurities, in particular dienes present in uncertainties, the conceptual process design for propylene, had shown strong inhibiting effects on the first commercial application of the LP OxoSM the rhodium catalyst. Impurities that might catal- Process built in as much protection for the rhodi- yse the aldol condensation reaction had also been um as was thought desirable. The degree of considered. If this reaction were allowed to occur protection was based on the known science, or, to excess, it would produce too many high-boiling where there were large gaps in knowledge, on what byproducts in the reactor. Having identified target was considered intuitively correct, in either case impurities, and quantified the problem in terms of bearing in mind capital cost constraints. the permissible concentrations of those impurities The fact that the rhodium catalyst used in in raw material streams to be fed to commercial small-scale rigs was not seeing representative com- plants, new analytical techniques were required. mercial feedstocks, and the concerns this raised Their sensitivity and repeatability had to be suffi- with respect to catalyst life, had to be addressed cient to measure the target impurities present in before the flowsheet for a commercial plant could real feed streams down to sub-ppm levels. Armed be outlined. Early poisoning studies in the labora- with such analytical methods, Davy Process tory by Union Carbide had concluded that the Technology built laboratory rigs to develop and propylene and synthesis gas mixtures produced in characterise processing schemes, employing het- industrial-scale plants were likely to contain impu- erogeneous catalysts and adsorbents for removing

Platinum Metals Rev., 2007, 51, (3) 121 (to desired residual levels) the potentially trouble- ured in a loop, also containing a gas recycle com- some impurities likely to be found in commercial pressor, product condenser and liquid-vapour propylene and synthesis gas streams. separator. The catalyst solution, containing ligand- The impurity guard beds and other purification ed rhodium and excess triphenylphosphine (TPP) plant that Davy Process Technology developed for dissolved in the products and byproducts of commercial feedstocks ultimately featured in the hydroformylation, is retained in the stirred reactor. design of commercial LP OxoSM plants, and were The incoming fresh raw materials, after pretreat- to contribute to ensuring that catalyst deactivation ment to remove impurities, merge with recycled rates in the operating plants were within permissi- gas containing the chemical components of the ble limits. synthesis gas and vaporised organics from the reactor, to enter the base of the reactor through Using the Gas Recycle Principle distributor spargers. The gaseous reactants pass as To address the key challenge of how best to bubbles of small size (and hence large interfacial separate the products and byproducts of the oxo area) into the liquid phase, where reaction takes reaction from the catalyst, several distillation place at a closely controlled temperature, typically columns were proposed in an early LP OxoSM selected between 90 and 100ºC. While oxo synthe- flowsheet. However, it was felt that this scheme sis takes place in the reactor, the reaction products would only exacerbate concerns regarding catalyst are stripped from the catalyst solution by an deactivation. Thermodynamic modelling, in con- upward gas flow. Heat of reaction is taken out junction with the kinetic models, revealed that it partly via the latent heat of vaporisation of should be possible to remove from the catalyst aldehydes into the gas, and partly by circulating a solution the reaction products, including high-boil- coolant through coils inside the reactor. The prod- ing aldol condensation byproducts, by means of ucts are condensed from the gas/vapour effluent gas stripping. This emerged as the makings of the leaving the top of the reactor, and the resulting liq- ‘gas recycle flowsheet’ adopted for the first uid products are separated from the recycle gas. commercial LP OxoSM plant, and several sub- The gas/uncondensed vapour is then recom- sequent plants. pressed for recycling to the reactor. Operating The flowscheme of an early LP OxoSM plant conditions, in particular the gas recycle rate, are set employing the gas recycle principle is shown in so that all liquid products leave the system at the Figure 3. A stirred, back-mixed reactor is config- same rate at which they have been formed, so that

Key 1 Pretreatment 2 Reactor 8 3 Catalyst preparation 4 Condenser 5 Separator Purge gas 7 6 Stabiliser 7 Cycle compressor 8 Overhead compressor 4

Propylene Synthesis gas 1 2 5 6

3 Mixed aldehydes Fig. 3 Gas recycle flowsheet of an early LP OxoSM plant

Platinum Metals Rev., 2007, 51, (3) 122 the reactor inventory remains constant. Passive working regime for the catalyst, in terms of both components in the synthesis gas, such as nitro- loss prevention and deactivation, based on the gen, methane and carbon dioxide, along with ‘state of the art’ at the time. propane present in the propylene or formed by hydrogenation, are purged in a blow-off to a fuel Success from the First LP OxoSM Plant header, to prevent them from accumulating in the Having decided to build a commercial plant at system. Unreacted propylene, propane, CO and Ponce, Union Carbide erected a 200 tonnes per hydrogen dissolved in the condensed product annum gas recycle pilot plant at the same site to leaving the separator are removed from the prod- test the process on the feedstocks available there, uct in a stabiliser column, and recompressed and to provide scale-up data. While the pilot plant before being recycled to the reactor. was being built and commissioned, Davy Process The basic flowscheme of the LP OxoSM Technology started the process and basic engi- Process emerged as both simple and elegant. The neering design of the 136,000 tonnes per annum principle of using in situ gas stripping to separate full-scale unit. This was to be built almost along- product from the catalyst appeared sound, side the pilot plant. The process design was because the high molecular weight of the rhodi- refined and further developed once operating um catalyst complex should mean that the loss by data were available from the pilot plant, which vaporisation of rhodium in the product would be continued to operate for a short time after the practically zero. The rhodium catalyst was safely commercial unit first started up in January 1976. contained in the reactor, and provided sufficient The initial start-up of the full-scale Ponce energy could be imparted through the mixer plant was easier than anticipated. Excluding out- impeller, the catalyst would be exposed to operat- side interruptions, the plant was online for all but ing conditions more or less replicating those used one hour in its first month of operation. During in the laboratory. There was no reason for any its first year, its on-stream operational availability significant amount of rhodium to leave the reac- was greater than 99%. This contrasted with a typ- tor during day-to-day operation, provided leakage ical availability of about 90% for a conventional was avoided and the physical entrainment of cat- cobalt-based oxo plant, based on Union Carbide’s alyst solution in the reactor overhead gas stream own experience. The operation continued to be was minimised. Neither of these containment marked by what was until then unusual ease, requirements were expected to pose undue diffi- stability and smoothness. Design targets for pro- culties. Catalyst leakage could be virtually ductivity, selectivity, feedstock usage efficiency eliminated by good engineering practice, includ- and product quality were all met. The ratio of ing the careful selection of construction materials normal to iso-butyraldehyde was usually con- and mechanical seals for moving parts; entrain- trolled at around 10:1, but higher ratios up to 16:1 ment could be dealt with by using proprietary, but were achieved. Significantly, the costs attributable inexpensive, entrainment filters on the overhead to catalyst were lower than expected, and the life line from the reactor. The use of in situ stripping of the first catalyst charge exceeded one year. obviated the need to remove catalyst solution The reaction temperature was kept as low as from the reactor to separate product using exter- possible, and in the range of about 90 to 100ºC, nal distillation equipment, thereby eliminating any consistent with being able to achieve sufficient potential for increased catalyst deactivation due catalyst productivity from the volume of catalyst to concentrating the catalyst, and exposing it to solution available to meet the production higher temperatures than those used in the reac- demands, and being able to control the liquid lev- tor. els in the reactors. (Product stripping was easier at The adoption of the gas recycle principle not higher temperatures because of the higher vapour only led to a simple and affordable process flow- pressures of the products.) It was known that sheet, it also appeared to provide the best overall higher reaction temperatures would lead to an

Platinum Metals Rev., 2007, 51, (3) 123 increased production of reaction byproducts and the presence of carbon monoxide and TPP. In an increased rate of catalyst deactivation; effec- this coordination complex the rhodium atom car- tive temperature control was therefore important. ries five labile-bonded ligands: two TPP, two The reaction temperature could be regulated very carbon monoxide and one hydrogen. In the first closely – to within ± 0.5ºC. The operating pres- reaction step, a propylene ligand is added to form sure of the reactors was also well controlled at complex B, which rearranges to the alkyl com- about 18 bar. plex C. This undergoes carbon monoxide The process characteristics and control insertion to form the acyl complex D. Oxidative systems used meant that the unit needed little addition of hydrogen gives the dihydroacyl com- day-to-day operator attention. Again, this plex E. Finally, hydrogen transfers to the acyl contrasted with experience on high-pressure group, and normal butyraldehyde is formed cobalt plants. The rhodium unit could quickly be together with complex F. Coordination of F with restarted from a full shutdown, and it was carbon monoxide regenerates complex A. possible to restore production following outages Some iso-butyraldehyde is produced along much more rapidly than had been the case with the normal butyraldehyde, but a high selec- with cobalt. tivity to the latter is ensured by exploiting a steric hindrance effect as follows. The reaction is car- How the Catalyst Works ried out in the presence of a large excess of TPP. The active rhodium species for the LP OxoSM Under the low-pressure conditions of the reac- Process is formed under hydroformylation reac- tion, the high TPP concentration suppresses the tion conditions, and there is no need for complex dissociation of complex A into one containing catalyst synthesis and handling steps. The proba- only a single phosphine ligand. If largely undisso- ble sequence of the reaction with propylene to ciated complex A is present, with its two bulky form normal butyraldehyde is shown in Figure 4. TPP ligands incontact with the propylene, then a Rhodium is introduced to the oxo reactor in high proportion of primary alkyl is favoured – if the form of a solution of ROPAC (a stable crys- fewer such ligands were present, then more talline compound) in butyraldehyde. Complex A propylene would form secondary alkyl groups, in Figure 4 is formed from the fresh rhodium in leading to more iso-butyraldehyde.

A B CH C 2=CHCH L L L H H CH2CH2CH3 + CH2=CHCH3 3 Rh CO Rh Rh CO CO L L CO CO L CO + CO CH2CH2CH3 H L L H H CO L

+ H2 Rh – CH3CH2CH2CHO Rh Rh

L L CO F CO E CO L D CO + L L CH2CH2CH3 H

Rh L

L CO Fig. 4 Probable reaction cycle for formation of normal butyraldehyde from propylene (L = triphenylphosphine (TPP); A: product of reaction of ROPAC with carbon monoxide and TPP; B: addition product of A and propylene; C: alkyl complex resulting from rearrangement of B; D: acyl complex resulting from carbon monoxide insertion to C; E: dihydroacyl complex resulting from oxidative addition of hydrogen to D; F: product of elimination of butyraldehyde from E)

Platinum Metals Rev., 2007, 51, (3) 124 Measures to Deal with Catalyst somewhat daunting. There were concerns about Deactivation handling and transporting such material in such The TPP-modified catalyst has a tendency to large quantities. With rhodium metal prices rising, deactivate over time due to the formation from the the logistics might put the security of, say, U.S.$2 monomeric rhodium species of rhodium clusters. million worth of rhodium at undue risk. There This type of deactivation is termed ‘intrinsic’, to were also uncertainties about what other sub- distinguish it from deactivation caused by an exter- stances might be present in the rhodium nal source such as catalyst poisons present in the concentrate that could cause Johnson Matthey feedstocks. Catalyst management models were processing problems. Although metals like iron developed to help operators of the LP OxoSM and nickel that are usually found in commercial Process to optimise the economic return from feedstocks could be anticipated, would metal con- their catalyst charges, in recognition that intrinsic tamination compromise rhodium recovery? The deactivation had to be tolerated to some extent. requirement for off-site rhodium recovery from For example, operating temperatures could not be bulk catalyst solution detracted from the elegance lowered to reduce catalyst deactivation if this also of the LP OxoSM Process. Fortunately, by the time reduced catalyst productivity to uneconomic or the first licensed plants actually started operation, unmanageable levels. Rhodium catalyst manage- Union Carbide had proven a catalyst reactivation ment guidelines from Union Carbide and Davy technique that would virtually obviate off-site Process Technology recommended operating recovery. adjustments to compensate for deactivation, in response to accumulated operating data which Catalyst Reactivation indicated the time evolution of catalyst activity. By the early 1980s, before any need had arisen The guidelines were couched so as to optimise the to resort to off-site rhodium recovery, Union balance between reaction rate, selectivity to nor- Carbide had developed a means to deal with the mal butyraldehyde and catalyst stability. While intrinsic deactivation – effectively by reversing it. operators felt some obligation to comply with the This involved concentrating the spent catalyst and licensor’s recommendations, at least until perfor- then treating the rhodium present in the resulting mance warranties had been met, it was interesting residue to convert it into a form capable of reacti- to observe how the long-term catalyst operating vation. The concentration process was carried out strategies adopted by plant owners varied so using specialised equipment (a proprietary evapo- widely between plants, depending on specific cir- rator) under very precise conditions, including cumstances and preferences. high vacuum, designed to prevent catalyst damage. Plant operators observed rates of catalyst deac- The overall process could conveniently be per- tivation that meant that a rhodium catalyst charge formed at the plant site, and required no chemical would typically last for about 18 to 24 months reagents. It resulted in a ‘declustering’ of rhodium before its activity had declined to the point when to enable the restoration of activity once the it would have to be discharged from the reactor treated residue had been returned to a hydro- and replaced by a fresh catalyst charge. formylation environment. Eventually, all operators The earliest LP OxoSM plants contained very either added reactivation equipment to their simple equipment to concentrate the discharged plants, or arranged to share facilities. Catalyst reac- spent catalyst solution. The idea was that concen- tivation was incorporated into the standard design trated catalyst, containing say 2000 ppm of of all new plants, and a measure of lost elegance rhodium, would be shipped to Johnson Matthey in was restored to the LP OxoSM Process! the U.K., who would then recover the rhodium in The catalyst reactivation technique was used to a form suitable for reprocessing to ROPAC. But carry out repeated reactivations of what was essen- the logistics of actually reprocessing around tially a single catalyst charge. This drastically 20 tonnes of concentrate for a typical plant were reduced the need for off-site recovery, which was

Platinum Metals Rev., 2007, 51, (3) 125 normally deployed only on rhodium that could no 2 The Dow Chemical Company: http://www.dow.com/ longer be reactivated economically. In that case, 3 Union Carbide Corporation: the recovery could be performed on residues typi- http://www.unioncarbide.com/ cally containing about 8000 ppm of rhodium, four 4 R. L. Pruett and J. A. Smith, Union Carbide Corporation, ‘Hydroformylation Process’, U.S. Patent 3,527,809; 1970 times the concentration initially envisaged, thus 5 G. Wilkinson, Johnson Matthey, ‘Improvements in improving the logistics and reducing the cost of Catalytic Hydrogenation or Hydroformylation’, British off-site processing. Patent 1,219,763; 1971 6 M. L. H. Green and W. P. Griffith, Platinum Metals Rev., Conclusion 1998, 42, (4), 168 This article (Part I) has sought to demonstrate 7 W. P. Griffith, Platinum Metals Rev., 2007, 51, (3), 150 the initial promise of the LP OxoSM Process, 8 F. J. Smith, Platinum Metals Rev., 1975, 19, (3), 93 employing rhodium-based catalysis, in terms of 9 Production estimates provided by RXN Petrochemical Consulting Inc.: http://rxnpetrochem.com/page4.html high availability, selectivity and productivity, low environmental impact and low maintenance. Part Further Reading II, to be published in a future issue of Platinum Metals Review, will address subsequent key improve- ‘Low-pressure oxo process yields a better product mix’, ments to the process, and its use in non-propylene Chem. Eng. (New York), 5th December, 1977, 84, (26), 110; marking the award of the 1977 Kirkpatrick Chemical applications. Engineering Achievement Award to the winners: Union Carbide Corporation, Davy Powergas Ltd., and Johnson LP OxoSM is a service mark of The Dow Chemical Company. Matthey & Co. Ltd. References J. L. Stewart, ‘LP OxoSM process – a success story’, 1 Davy Process Technology Ltd.: Indications, Winter 1982/83; the international journal of http://www.davyprotech.com/ Davy McKee

The Authors Richard Tudor is a chartered chemical Mike Ashley spent many years with John engineer. He has played a leading part in Brown, involved with process technology and Davy Process Technology’s oxo licensing business development, before joining Davy activities for over thirty years, firstly as Process Technology. He is now concerned Process Manager, and then as Business with business analysis, technology Manager after a period as Licensing acquisition, marketing, website development Manager. As a Vice President of sales and and all aspects of public relations. marketing, he now has overall responsibility for the oxo business.

Platinum Metals Rev., 2007, 51, (3) 126