Special Reviews

Homogeneous Catalysis--New Approaches to Catalyst Separation, Recovery, and Recycling

See allHide authors and affiliations

Science  14 Mar 2003:
Vol. 299, Issue 5613, pp. 1702-1706
DOI: 10.1126/science.1081881


Homogeneous catalysts have many attractive properties, such as high selectivities. However, many homogeneous catalytic systems cannot be commercialized because of difficulties associated with separating the products from the catalyst. Recent approaches to tackling this problem are reviewed and compared.

Homogeneous catalysts offer a number of important advantages over their heterogeneous counterparts. For example, all catalytic sites are accessible because the catalyst is usually a dissolved metal complex. Furthermore, it is often possible to tune the chemoselectivity, regioselectivity, and/or enantioselectivity of the catalyst.

Despite these advantages, many homogeneous catalytic systems have not been commercialized because of one major disadvantage compared with heterogeneous catalysts: the difficulty encountered when trying to separate the reaction product from the catalyst and from any reaction solvent. This problem arises because the most commonly used separation method, distillation, requires elevated temperatures unless the product is very volatile. Most homogeneous catalysts are thermally sensitive, usually decomposing below 150°C. The thermal stress caused by product distillation even at reduced pressure will therefore decompose the often expensive catalyst. Other conventional processes such as chromatography or extraction also lead to catalyst loss. The homogeneous reactions that have been commercialized either involve volatile substrates and products or do not contain thermally sensitive organic ligands.

The hydroformylation of propene to butanal (1) is carried out in a continuous process (2) on a scale of 4 × 1012 to 5 × 1012 g/year with volatile Rh/PPh3 catalysts (Ph, phenyl); the volatile product is distilled directly from the reaction pot. A more thermally robust catalyst, [RhI2(CO)2], is used in the manufacture of acetic acid from the carbonylation of methanol. This catalyst can withstand relatively high temperatures, although in this case the distillation is carried out in a flash evaporator at low pressure, outside the reaction chamber, making it a batch continuous process (3). The use of low pressure during distillation often causes other problems. Because the catalysts have been optimized for stability, activity, and selectivity under the high-pressure reaction conditions, they may undergo undesirable side reactions under the reduced pressure conditions of the separator. In some specialized applications, alternative separation strategies have been commercially applied, but these—such as the removal of 1,4-butanediol, the product of the hydroformylation of allyl alcohol, from the toluene solvent by washing with water (4)—are often suitable only for a narrow range of reactions.

Many catalysts with highly desirable properties in terms of activity and selectivity have therefore not been commercialized for anything but the most valuable products. To overcome the separation problems, chemists and engineers are investigating a wide range of strategies other than distillation for recycling the catalyst (5, 6). In this article, I aim to provide an overview of the thinking, methodology, and progress in this area of research without attempting to be comprehensive. To allow comparison of the different approaches, I mainly use the hydroformylation of long-chain alkenes (typically 1-octene, Fig. 1), using rhodium-based catalysts, to yield aldehydes for the subsequent manufacture (several billion kilograms per year) of plasticizers, soaps, and detergents.

Figure 1

Hydroformylation of 1-octene.

Rhodium catalysts typically work under mild conditions (100°C, 25 bar), giving good activity and selectivity (80 to 90%) to the desired linear (l) aldehyde. Nonetheless, all commercial plants carrying out this reaction use cobalt catalysts, which require much harsher conditions (typically 200°C, 100 bar) and give poorer selectivities (1), because rhodium catalysts decompose when attempting to distill the product from them. Solving the product separation problem for the rhodium-catalyzed reaction in an effective and economically robust way would represent a major step forward in homogeneous catalysis. The lessons learned from this reaction will also be applicable to many others and could have far-reaching consequences.

The new processes under investigation can broadly be grouped into two types. In the first, the catalyst is anchored to some kind of soluble or insoluble support, and the separation is carried out by a filtration technique. This type of process is often referred to as heterogenizing homogeneous catalysts. The other type involves designing the catalyst so that it is solubilized in a solvent that, under some conditions, is immiscible with the reaction product. These reactions involve two phases and are often referred to as biphasic systems. A variety of systems have been designed for these new separation strategies (Fig. 2).

Figure 2

Structures of ligands, ionic liquids, and polyelectrolytes used for separation. See text for various meanings of n, R, and Y.

Supported Catalysts and Filtration

Insoluble supports. Ligands can be anchored onto solid materials such as inorganic oxides (often silica) or polymers. If the anchoring is covalent, it can be robust enough to withstand the rather harsh conditions of the catalytic reaction. Because the ligand for binding the metal resides only on accessible sites of the solid and can be designed to protrude into the solvent, all catalytically active sites are available for reaction, allowing rates and selectivities comparable to those obtained with analogous dissolved catalysts to be achieved. The main problem is that bonds between metal and ligand are often broken and reformed during catalytic reactions. If this happens, the catalyst may break away from the support and become dissolved. This “leaching” process leads to loss of activity of the catalyst when it is recovered by filtration and recycled, or is used with a continuous liquid flow. Reduced leaching has been observed when the catalyst was anchored inside the pores of zeolites (so-called ship-in-a-bottle catalysts) or of mesoporous solids.

Recently, a system for hydroformylation reactions was reported in which a member of the highly selective Xantphos family of ligands was derivatized with a hydrocarbon chain terminating in a triethoxysilyl group (7, 8). This ligand was incorporated into a sol-gel solution for the production of silica and the resulting supported ligand, 1 (Fig. 2), bound to rhodium. This catalyst shows high activity and selectivity, and a single sample has been used for a variety of different hydroformylation reactions under widely varying conditions over a period of more than a year. The catalyst still retains its activity and selectivity, providing a very rare example where leaching has been reduced to an acceptable level (8). It can be used in a continuous liquid flow process by incorporating monoliths of the sol-gel material suitably loaded with rhodium into the paddles of the autoclave stirrer (9).

This catalyst has also been applied to the hydroformylation of 1-octene in a gas-like continuous-flow system (10). The special properties of supercritical CO2 (scCO2)—gas-like diffusion combined with good solvating power (see below)—were used to provide the medium in which the substrates and products were flowed over the catalyst. 1-Octene, CO, and H2 are all fully soluble in the scCO2, which can reach all active catalytic sites. The product, nonanal, is also soluble in scCO2 and is thus removed from the catalyst in the scCO2 once it is formed. When the effluent from the reactor is depressurized, CO2forms a gas while the product is collected as a liquid. The only separation required is that of product from unreacted starting material and any by-products. The CO2 can be recompressed and recycled. Although the yields in this particular reaction are low, the principle is exciting because, even if the metal briefly dissociates from the ligand, it will not dissolve in scCO2 and leaching will be minimized.

To my knowledge, the only commercial example of a homogeneous catalyst heterogenized on a solid support is the carbonylation of methanol using [RhI2(CO)2]electrostatically bound to an ion exchange resin (11). However, in this case leaching cannot be avoided. The plant in Japan therefore uses a guard bed of fresh ion exchange resin downstream of the reaction bed to readsorb any [RhI2(CO)2] that dissolves. After a period of time, the guard-bed resin, now loaded with rhodium, is cycled to become the catalyst bed.

Soluble supports. The supports used to bind the catalyst may also be soluble in the reaction medium. This has the advantage that active catalytic sites are distributed throughout the reaction solution. The catalyst architecture can be similar to that of the efficient homogeneous catalyst that it is trying to mimic. The supports may be soluble polymers (12, 13). In a recent example, the anions of 2 were partially exchanged by the anions of3 (n = 0). The resulting polyelectrolyte was used to catalyze the hydroformylation of 1-octene, with the catalyst product separation being performed by ultrafiltration. Good rates were obtained and 93% of the catalyst could be recycled; the losses were attributed to ligand oxidation during the many manipulations required for the batch process.

Dendrimers are large (2 to 4 nm) tree-like molecules with a persistent globular shape, which makes them more suitable for ultrafiltration than soluble polymers, which may pass through filtration membranes more easily. The identical groups on the periphery of the dendrimer can form ligands to metal catalysts. The metal-binding groups are usually on the exterior of the dendrimer but can also be buried within shape-selective pockets. Because membranes for enzyme ultrafiltration have channels ∼1 nm in diameter, they allow the solvent and product of a catalytic reaction to pass through but reject the dendrimer-based catalyst. Advantageously, dendrimers may exhibit bidentate binding (through two donor atoms on the same dendrimer arm) to the metal. The chelate (ring-forming) effect will then ensure that leaching is minimized. Furthermore, if the metal separates from its dendrimer-bound ligand, it may be sequestered again rapidly by one of the many identical binding sites nearby.

Despite these advantages, reactions where ultrafiltration has been attempted have generally shown loss of activity upon recycling (14). This may not be because of genuine leaching, but because the membranes are not designed for use with organic solvents, high temperatures, and/or high pressures. Suitable membranes will probably need to be based on zeolites rather than the much less robust materials that are used for removing enzymes from aqueous solutions. One rather unexpected advantage of some dendrimers has emerged, however: They can show much higher selectivities to desired products than their small-molecule analogs. For example, a dendrimer bearing 16 PPh2 groups on its periphery, 4, gives linear:branched (l:b) ratios in the hydroformylation of octene of 13.9:1, compared with 3.8:1 for a small-molecule analog (15).

In one recent study, dendrimer wedges (molecules that have dendrimer-like properties but only form part of a sphere) were anchored to beads of silica (16) or a polymer, 5, using approaches developed for solid-phase organic synthesis (17,18). They have been used for hydroformylation and can easily be removed by conventional filtration. This procedure combines the advantages of the controlled environment of the dendrimer-bound catalyst with the ease of separation of the supported catalyst. In this application, the metal-binding phosphine groups may be placed at the end of or along the arms of the dendrimer. Both types of binding allow good rates for the hydroformylation of styrene and related substrates, but dendrimers in which the rhodium is more deeply buried show better recyclability (18).

Membranes with even smaller pore sizes are also being developed. These nanofiltration membranes rely on the difference in size between a simple organic molecule and a metal complex. They still allow solvent and product molecules to pass through, but retain the organometallic catalyst (19). The advantage of these systems is that optimized catalysts can be used without modification. This approach has been successfully applied to a system in which scCO2 is the solvent. The scCO2 transports the products through the membrane while the catalyst remains in the reactor (20).

Biphasic Systems

Aqueous biphasic systems. One biphasic system that has been commercialized for the hydroformylation of propene uses as ligand a sodium salt of sulfonated triphenylphosphine (3, n = 0) (21) to make the rhodium-based catalyst soluble in water. Because most organic compounds do not mix with water, the reaction can be carried out in two phases with rapid mixing to ensure maximum contact between the catalyst and the substrate. After the reaction, the mixture is allowed to settle and the product is decanted, leaving the catalyst in the aqueous phase. However, because of the very low solubility of long-chain alkenes in water, the rates of reaction are too low for commercialization of this process for the production of detergent-range aldehydes.

A recently reported approach solves this problem while retaining the advantages of the aqueous biphasic systems (22). Instead of being sulfonated, triphenylphosphine was derivatized with a polyethylene glycol chain (6). At room temperature, the rhodium complex of this phosphine is soluble in water but not in organic solvents. Upon heating, however, the polyether side chains undergo a phase transition and the complex becomes more soluble in the organic phase than in water, so that at the reaction temperature all the required components are dissolved in the organic phase. Upon cooling, the phase transition is reversed and the catalyst returns to the aqueous phase; the organic phase, now devoid of rhodium complex, can be decanted.

A biphasic system is exploited in the commercial Shell Higher Olefins Process (SHOP). The oligomerization of ethene in the presence of a specially designed ionic nickel catalyst is carried out in 1,4-butanediol. The very high polarity of the solvent ensures that the medium- to long-chain alkene products phase-separate and can be removed by decanting (23).

Fluorous biphasic systems. The problem of differential solubility experienced in the aqueous biphasic systems led Horváth to propose the use of fluorous-organic biphasic systems. Fluorous and organic solvents mix at the typical reaction temperatures used for hydroformylation reactions, but they phase-separate at room temperature (24). Using rhodium complexes of 7 in a mixture of perfluorocyclohexane and toluene, Horváth showed that good rates for hydroformylation of 1-octene could be obtained and that leaching of rhodium into the organic phase was very limited (4.2% after nine runs). The l:b ratio could be high (8:1), but ∼10% of the starting alkene was lost through isomerization. Even higher rates can be obtained if the toluene is omitted and the ligand is replaced by 8 (n = 0). In this case, rhodium leaching is reduced to 0.05% per run (25). The lack of toluene removes the energetic requirement for fractional distillation to separate the catalyst from a solvent.

Supercritical fluids. Supercritical fluids (compressed gases above their critical temperature) dissolve many low- to medium-polarity organic molecules and are fully miscible with permanent gases. If a catalyst can be dissolved, truly homogeneous catalysis can occur, because all participants are fully dissolved in one phase and no phase transfer problems arise. Most metal-containing complexes, particularly those with aryl-containing ligands, are poorly soluble in scCO2, the most commonly used supercritical fluid, but can be rendered soluble by incorporation of fluorocarbon chains (8, n = 1) (26) or by using trialkylphosphines (6).

Although the supercritical solvent can very simply be removed by decompression to a gas, this does not overcome the main separation problem: that of the catalyst from the product. One elegant example where this has been achieved is a hydrogenation reaction in which the iridium-based catalyst 9 is soluble in scCO2 in the presence of the substrate, but precipitates once all the substrate has been used up (27). In an alternative process, temperature and pressure swings are used to precipitate the catalyst selectively; the product is then recovered by decompression. This process has been successfully used for hydroformylation with ligands such as 8 (n = 1) (26).

Ionic liquids. Ionic liquids are salts (for example,10) that are liquid at room temperature, or at least at the reaction temperature. They have an extremely low vapor pressure and, depending on the design of the ionic liquid, can dissolve or reject organic compounds. They also dissolve ionic catalysts. The first hydroformylation reactions carried out in ionic liquids used rhodium complexes of 3 (n = 2). However, conversions were low, probably because the high lattice energies of the sodium salts make them rather insoluble in the ionic liquid (28). More recent work has shown excellent activity and selectivity with ligands such as11 (29) or 12 (30). The products can be decanted from 10 (R = C4H9, Y = PF6) or extracted with nonpolar organic solvents. For ligand 12, high rates and selectivities with minimal leaching (<5 parts per billion) are obtained in ionic liquid 10 (R = C4H9, Y = PF6), and it is possible to envisage processes involving decanting or extraction of the products into an organic solvent in a batch continuous mode.

Another way of using ionic liquids that has the potential for continuous-flow liquid operation is to support the ionic liquid as a film on a solid (such as silica, sometimes derivatized with molecules similar to the ionic liquid). The catalyst is dissolved in the ionic liquid film, with the advantage that the surface area of the ionic liquid is greatly enhanced relative to its volume and the substrate can readily diffuse to the catalyst. These systems have recently been used for rhodium-catalyzed hydroformylation with ligands such as3 (n = 0) or 13 (n = 0, R = C4H9). The ionic liquids were 10(R = C4H9, Y = PF6 or BF4). Leaching was reduced by working at low conversions so that the polarity of the substrate/product phase was kept low to minimize catalyst extraction from the ionic liquid film (31). One potential disadvantage of using ionic liquids containing PF6 or BF4 is that they react with traces of water to give species such as O2PF2 and the very highly reactive and corrosive HF, which can destroy the catalyst (32).

Supercritical fluid–ionic liquid biphasic systems.An alternative to extracting products from ionic liquids with organic solvents is to combine the favorable properties of ionic liquids with those of supercritical fluids. scCO2 has been shown to be miscible with certain ionic liquids (33) and will extract many organic compounds from ionic liquids (34), allowing a genuinely continuous process to be developed. An ionic catalyst is dissolved in the ionic liquid in a stirred tank reactor. The substrate (alkene for hydroformylation), permanent gases (CO and hydrogen), and scCO2 are then passed into the reactor either separately or mixed. The reaction takes place and the products flow out of the reactor dissolved in scCO2, which is decompressed to release the products (Fig. 3). The CO2 containing any excess CO and H2 can be recompressed for an emissionless and continuous process that requires no separation of the product from the solvent. The first demonstration of such a system was in hydroformylation (32) and showed rather low rates, but these have been much improved by using ligand 13 (n = 2, R = C3H7) and an alternative ionic liquid,10 [R = C8H17, Y = (CF3SO2)2N] (35). The approach has also been used for hydrovinylation of styrene (36) (but using CO2 below its critical temperature), for Wacker oxidation (37) reactions, and for kinetic resolution using enzymes (38,39).

Figure 3

Schematic diagram of a continuous-flow hydroformylation reaction in a supercritical fluid–ionic liquid biphasic system.


A number of innovative solutions to the problem of separation of catalysts from products in homogeneous catalytic reactions have been proposed and demonstrated (Table 1) (1). Some are genuinely continuous with built-in product separation; others can be carried out in batch continuous mode, with the separation being carried out in an external chamber by a nondistillation process.

Table 1

Rhodium-catalyzed hydroformylation of 1-octene, using a variety of separation methods and catalyst systems, compared with results from homogeneous commercial systems (first four rows). Abbreviations: nr, not recorded; n/a, not applicable; R, low rate; P, high pressure; S, low selectivity to linear aldehyde; L, high catalyst leaching; U, ultrafiltration not attempted; E, expensive ligand and/or solvent.

View this table:

The question of which of these processes should be commercialized cannot be answered at present. All of the processes have some disadvantages, and most of them do not yet show the activity required for commercial application. The catalyst, solvent, or process may be very expensive, catalyst leaching may be too high, the solvent or ligand may be too expensive, or the pressure may be too high. Many approaches show considerable promise, but few detailed cost analyses have been carried out. They will be essential before any commercialization can be contemplated. An important property of some of the processes may be reduced environmental impact, as a result of recycling most of the components and getting away from volatile organic solvents with their polluting properties. The environmental impact of some solvents (fluorous, ionic liquid) is, however, still unknown.


Stay Connected to Science

Navigate This Article