6November/ December 2009 reasons. A major factor is one of time. In the transition from small to large particles there is a finite redevelopment time where the separation is modified to account for the lower column efficiency. Where the selectivity is high, this is not important, but for the more difficult separations there can be a significant loss in production rate. As most HPLC and SFC systems can cope with the pressures required to run semi- preparative columns at a reasonable flow rate the simplest and fastest option is to use the same particle size for the preparative separation as for the analytical scale column used for development. For larger scale separations the particle size becomes important as the column diameter is necessarily increased. For columns 10 cm id and above it is necessary to limit the operating pressure to prevent damage to the silica base particles since wall support for the chromatographic bed is lost in such wide diameter columns. Just as importantly, the production rate needs to be maximized for these larger scale separations to minimize the project duration and the costs. Larger particles of 10 to 20 microns diameter allow higher flow rates ( albeit at a loss in plate count, which for the higher selectivity separations is less important a parameter) which give higher production rates. Thus larger particle sizes are preferred as the scale of operation increases, with SMB processes optimally operating toward the 20 micron end of the range. The column technologies available for preparative chromatography have changed little over recent years. Axial compression technology, introduced in the 1980s4, revolutionized the preparative technique by allowing stable, high performance columns of diameters greater than 5 cm to be prepared from the small particles used in HPLC separations. Several variations on this theme have appeared more recently, but all such columns perform similarly with the compression technique compensating for the inevitable voiding and channeling that plagues large diameter columns. For columns 5 cm and less, there are several techniques used to pack high performance columns, some relying on axial compression schemes, others using more traditional high pressure slurry processes. For these, the performance of columns packed by different technologies is closely similar; a well- packed column has the same performance characteristics and lifetime regardless of how it is prepared. Supercritical Fluid Chromatography ( SFC). As noted above, SFC has supplanted HPLC as a preparative technique in many companies which are concerned with small scale separations at the discovery level. The reasons usually cited for this change in processing are the faster separations, due to the low mobile phase viscosity, and the reduction in organic solvent consumption which results in easier product recovery. The technique is promoted as being " green" in that it uses less solvent ( the carbon dioxide used in the systems is usually a by- product of other processes; its use in SFC separations merely delays its arrival in the atmosphere) and as such can make a small difference to the overall carbon emissions from the industry. Although SFC saves costs in terms of the low price of CO2, it must be remembered that it is more expensive to operate, as the pumps required for the CO2 are considerably larger than those required for similar flow rates of organic solvents and there are several phase changes through the cycle ( see below) which require energy input. Unlike the situation for HPLC, the mobile phase in SFC is a compressible fluid at high pressure which requires significant safety considerations to be taken into account in equipment design and operation. A schematic of a preparative supercritical fluid chromatographic system is shown in Figure 2. The key differences from HPLC systems lie in the use of carbon dioxide as the main component of the mobile phase. CO2is non polar and for almost all applications a mobile phase modifier has to be used to increase the overall solvent polarity to solubilise the sample and to allow elution from the column. The CO2 has to be in the supercritical fluid state ( or close to it) for the chromatographic step which means it has to be pressurized to greater than 73 bar at a temperature of greater than 31.1° C. In order to bring it to the required pressure it has to pumped, which means it needs to be in a liquid form at this point. This is usually accomplished either by using a cylinder with a dip tube or by condensing gaseous CO2by maintaining the pressure at around 50 bar and reducing the temperature to a few degrees above 0° C. Once the operating pressure is reached, the temperature is raised to bring the CO2to the supercritical state after which it is mixed with the mobile phase modifier. The sample, dissolved in the modifier, is introduced from a separate pump or from a loop injector. After the separation and the components are detected, the pressure is reduced in the back pressure regulator ( BPR) to bring the supercritical fluid to the gaseous state. This pressure reduction results in rapid cooling and the temperature has to be controlled to prevent the equipment from being encased in a block of ice. Once the CO2is a gas, the solubility of both the samples and the mobile phase modifier becomes extremely small and these components drop out of solution as a fine mist. Collection of the organic components is usually done in a cyclone collector which efficiently separates out the mist, condensing the product as a solution in the mobile phase modifier. The carbon dioxide is then either vented to the atmosphere or is recycled back to the pump through a stripper to remove remnants of modifier or solutes. In the latter case, the pressure downstream from the backpressure regulator is maintained at around 50 bar and the gaseous CO2is condensed by cooling the stream. One aspect of SFC that is currently problematic lies in sample introduction. The sample is usually introduced into the mobile phase stream with a loop injector or a sample pump as a solution in the organic modifier. This results in band distortion when the sample volume is large because the pulse of strong solvent causes premature elution of the solute molecules within it as it mixes with the mobile phase. This distortion can limit the injection volume that can be used. An alternative, to introduce the sample into the modifier stream before mixing with the CO2, results in broader injection bands, especially when the modifier concentration is low. Another problem that can arise is that of sample solubility. A not infrequent situation is where the sample, or a sample component, is less soluble in the supercritical mobile phase than it is in the modifier. As the mobile phase and injected sample mix, the sample - or the insoluble component - may precipitate prior to reaching the column inlet. This often results in pressure increases on injection and can result in blocked and distorted frits, which destroys the column ( Figure 3). The ideal solution, to dissolve the sample in the supercritical mobile phase, is not easily implemented and is not offered in commercial systems. High Performance Liquid Chromatography. HPLC has been around for many years and although at the small scale end of preparative chromatography it is being supplanted by SFC, nevertheless it remains the more important technique at larger scale. This is partly due to the size, availability and cost of large scale SFC equipment, as well as the services and costs required to run it. In labs at Figure 2. Schematic of a Preparative SFC unit. Figure 3. Result of Inlet Frit Blockage and Consequent Over- pressure in an SFC Column. CHIRALPAK AD- H, 250 x 50 mm.

7 Chiral Technologies we screen both HPLC and SFC screens are conducted, choosing the technique which gives the most economical solution, although HPLC becomes the preferred methodology for projects in excess of around 1 kg. As the simpler process, HPLC retains several advantages over SFC, in that sample cannot be lost during fraction collection, safety precautions are less stringent as the mobile phase is an incompressible liquid and as there are no phase changes, heating and cooling services are not necessary. Separations in HPLC are usually scaled to " touching band" level, where the sample load is increased to the point where the front of the second band starts where the tail of the first eluted component reaches the baseline. Displacement effects are not as strong in chiral separations as in many achiral situations for a number of reasons and recovery of valuable material is often a priority so heavier loading is rarely used. Simulated Moving Bed Chromatography ( SMB). SMB as a process for the pharmaceutical industry was implemented in the mid 1990s as an adaptation of the large scale processes for p- xyleneand high fructose corn syrup. SMB is a multi- column, countercurrent continuous binary separation process and is preferred on the basis of process economics as the scale of the separation increases toward production . There are currently several enantiomerically pure pharmaceutical products that are produced at a manufacturing scale ( ie multi- MTA) using this technique. Although at first sight it appears to be complex, it is based on simple chromatographic concepts. As bands separate in a column they move at differing speeds. If we could move the stationary phase as well as the mobile phase, then moving it in the opposite direction at a speed intermediate between the two band speeds would result in the slower moving band being transported with the stationary phase while the faster one would move with the mobile phase. If nothing else happens, the two bands would move further apart with time, leaving an unused space in the centre of the column. This means that one can introduce the feed continuously into the centre of the column and the two components would continue to separate. The products are removed by bleeding off material from the pure zones at the outer ends of the band. As the stationary phase cannot be moved while maintaining a well-packed bed, the entire column must move. This is accomplished by using multiple columns in series, with movement affected not by moving the columns but by moving the inlet and outlet positions instead. Unlike the situation in HPLC, where it is straightforward to design the preparative separation from a series of mass- overloaded injections, SMB requires a more complex procedure; usually computer simulations are used to develop operating conditions suitable for the separation followed by experiment to " fine- tune" the conditions thus developed. The data from the HPLC loading study is used to determine the parameters for the adsorption isotherms of the components which are then used in the computer simulations. Empirical determination of the operating conditions, although it is somewhat slower, is fortunately not too exacting a task and is normally used for the situations where the adsorption isotherms are not well described by a theoretical model. An excellent account of the development of a large scale manufacturing process by SMB has been written and although it is not the purpose here to go deeply into a description of such a procedure, there are some basic principles that can be noted. At the laboratory scale, the most precious resources are time and manpower. Thus, separations are generally designed to take the shortest possible time in the equipment available and the emphasis is on the rate of production of the desired enantiomer. In a manufacturing process, the emphasis is on cost of the product ( in $ per kg, etc) and this may change the way in which the process is run. Where the final product is valuable, the rate of production remains critical but cost considerations can result in a non- optimum process ( from the chromatographic viewpoint) being preferred. For a production process it is worth spending the time to optimize the separation using all possible stationary phase and mobile phase combinations - and also to calculate the economic consequences of several options to determine the best. It is essential to test intermediates at all points in the synthetic process downstream of the introduction of the chiral centre where there is no possibility of racemisation in processes still further downstream to find the best point at which to run the chromatographic resolution. This may be self- selecting in some cases where the chiral centre is introduced late in the synthesis, while in others there can be a genuine best point at which to introduce the resolution. Although at the production scale the recovery of solvent can reach over 99.9%, the cost of some solvents ( such as acetonitrile under the present economic climate) may influence the choice of one separation option over another. At present, manufacturing scale SMB processes are generally outsourced to a CMO with this capability. There are several companies in the world with such equipment ( eg Ampac, Daicel, Johnson Matthey, Novasep and SAFC) where large scale separations may be carried out. This is because one of the greater costs of SMB processing is the investment in equipment and infrastructure. If a new crystallization process is envisioned for a pharmaceutical product, there are usually sufficient tanks in a manufacturing plant to accommodate it. Most companies do not have SMB equipment in place and this extra investment can militate against implementation of a process even where it has longer term economic advantage. Another advantage of outsourcing such processes is that the CMOs have good experience in design, running and maintaining them which is not generally available. Conclusion Preparative enantioselective chromatography is a fast and efficient way to produce highly pure enantiomers from racemic ( or enriched) mixtures. Where there is a critical need to prepare pure enantiomers in the shortest possible time ( for example in the pharmaceutical industry from early discovery to the point where the product is moving through Phase 1 and perhaps Phase IIa clinical trials) the most effective route is generally through chromatographic resolution of the racemate. It is easy to develop a small scale separation of a few hundred milligrams of racemate and to progress to having operating conditions for isolation of kilogram quantities and even a production scale process within a few weeks. Once the first few kg of enantiomer have been prepared, the pressure to have material quickly is reduced so there is time, perhaps, to compare the chromatographic route with alternatives. This does not imply that enantioselective chromatographic processing is not used at the manufacturing scale; the imperative is to find the most cost-effective process. It should, of course, be remembered that the cost of chromatography for the first few grams of material is very different from that for the first 10 metric tons; the costs decrease quickly with scale and with further optimization of the separation process. Chromatographic processing should always be considered as one of the options for the manufacture of pure enantiomers. Reference L R Snyder, J W Dolan and G B Cox, J Chromatogr., 1989, 483, 63. J W Priegnitz and B McCulloch, US Patent 5518625. P Franco and T Zhang, J Chromatogr., B, 2008, 875, 48. J Dingenen, Analusis, 1998, 26, M18. R M Nicoud, Pharma. Technol. Eur., 1999, March/ April. M. Hamende, " Case Study in Production- Scale Multicolumn Continuous Chromatography" in Preparative Enantioselective Chromatography, Ed G B Cox, Blackwell, Oxford, 2005.