726 PREPARATIVE SEPARATIONS 15.1 INTRODUCTION The aim of preparative liquid chromatography (prep-LC) is the collection of one or more purified compounds from a mixture. The scale of prep-LC can range from micrograms (for compound identification) to tens of metric tons (for production of a pharmaceutical product, as in Section 13.9.2), but usually the goal is the recovery of milligrams to grams (the main emphasis in this chapter). Larger scale separations will receive only brief attention (Section 15.6). A closely allied separation technique, supercritical fluid chromatography [1], is beyond the scope of the present chapter. 15.1.1 Column Overload and Its Consequences Prep-LC involves mass-overload conditions, that is, sample weights that are large enough to affect peak widths and retention times. When developing an analytical separation, care is usually taken to ensure that the weight (or volume) of the sample does not exceed certain limits (Section 2.6), and that retention times and resolution will not vary with the amount of sample injected. As the weight of injected sample is increased, however, the detector will eventually become overloaded, retention times will decrease, and peaks will broaden and become distorted (compare Fig. 15.1a, b). A B C (a) (b) 5 10 15 ( min ) 0 A B C Small sample Column overload Figure 15.1 Hypothetical separations illustrating (a) an analytical separation and (b) a corre- sponding mass-overloaded touching-peak (T-P) separation. 15.1 INTRODUCTION 727 Table 15.1 Requirements for Purified Samples Objective Product Weight Required Tentative identification by instrumental methods ∼ 1mg Positive identification and confirmation of structure 1–100 mg Use as analytical standard (e.g., for calibrating an HPLC assay) 100 mg–2 g Toxicology testing 10–100 g Early phase 1 trials 200 g–2 kg Detector overload or nonlinearity is often the first consequence of column overload, for example, when the UV absorbance of a peak exceeds 1 or 2 absorbance units (AU; Section 4.2.5). Detector nonlinearity can often be circumvented by using a short-path-length flow cell (a preparative cell), or by changing the detection wavelength so that the compound of interest (referred to hereafter as the product) absorbs less strongly. A large enough sample can also lead to changes in the separation, which we will refer to as column overload (as in Fig. 15.1b). 15.1.2 Separation Scale The quantity of sample to be separated by prep-LC varies with the intended use of the purified product, as illustrated in Table 15.1 for small molecules (molecular weights <1000 Da) in pharmaceutical discovery and development. Note that column overload is determined by the weight of the largest peak of interest (usually the product peak), not the weight of the entire sample. For the purification of an impure product, however, where the initial purity is often > 80%, there is little difference between sample and product weights. When we speak of sample weight in this chapter, we mean the weight of the product to be purified. When the recovery of a few mg (or less) of a purified compound is required, the sample weight can be increased to the point where the space between the product peak and its nearest neighbor just disappears (i.e., corresponding to baseline reso- lution). The result is described as a touching-peak (or ‘‘touching-band’’) separation; touching-peak (T-P) separation corresponds to the largest quantity of sample that can be injected, while maintaining ≈100% recovery of the product with ≈100% purity. An illustrative T-P separation is shown in Figure 15.1b, where the sample weight has been increased (relative to the separation of Fig. 15.1a) until the major peak B expands to touch peak A. In this example the weight of peak A is not sufficient to affect either its retention time or peak width. Peak C is somewhat overloaded, but it has not broadened enough to touch peak B. It should be noted that the tail of both overloaded peaks elute near the analytical (small sample) retention times of Figure 15.1a (marked by the dashed lines in Fig. 15.1). A purified-product fraction (peak B in the example of Fig. 15.1b) can be collected as it leaves the column, using either manual or automated procedures (Section 15.2.4); solvent-free product can then be recovered by evaporation of the mobile phase (Section 15.2.5). 728 PREPARATIVE SEPARATIONS For the recovery of up to about 10 mg of purified product, more than a single sample injection may be required. When 10 mg of purified product are required, the large number of injection/collection cycles becomes inconvenient, especially if carried out manually. For larger sample weights, two different options exist: (1) use a larger diameter column (Section 15.1.2.1), or (2) optimize separation conditions for maximum resolution of the product peak (Section 15.1.2.2). 15.1.2.1 Larger Diameter Columns For the use of a larger diameter (‘‘semi-preparative’’) column, the column length is usually unchanged, and the identical column packing should be used. The flow rate and sample volume are increased in proportion to column cross-sectional area or d 2 c (d c is the column’s inner diameter, i.d.); note that the equipment must be capable of this increase in flow rate. For example, when replacing a conventional analytical column (4.6-mm i.d.) with a 10-mm-i.d. column, both the flow rate and the sample volume should be increased by a factor of 10 2 /4.6 2 = 4.73. Under these conditions the same separation will be obtained for both small- and large-diameter columns (same retention times, peak widths, resolution, and column pressure P). The replacement of an analytical column by a larger diameter column in this way will be referred to as scale-up. Note that the column length can also be changed, in which case the sample size should be adjusted in proportion to column volume. While a longer column has a larger plate number N for analytical separations, this is less important in prep-LC, because N plays only a minor role in affecting separation (Section 15.3.1.2). 15.1.2.2 Optimized Conditions for Prep-LC A second option is a change in selectivity that provides a better separation of the product. Most analytical separations are designed for the baseline separation of all peaks of interest, as in the optimized, small-sample separation of Figure 15.2a.In prep-LC, where usually a single product peak is to be recovered, only the resolution of the product peak from adjacent impurity peaks is important; the resolution of the product peak should therefore be as large as possible. This is illustrated in Figure 15.2b, where selectivity has been optimized for just the recovery of product peak 8—using the same general approach (Section 2.5.2) as for the development of the analytical separation of Figure 15.2a (i.e., a change in separation conditions that improves selectivity). Although some impurity peaks now overlap in Figure 15.2b (peaks 2–3, 5–6), T-P separation for peak 8 allows a much larger sample to be injected—as illustrated in Figure 15.2c. If the same sample weight as in Figure 15.2c is injected for the separation conditions of Figure 15.2a (see Fig. 15.2d), product peak 8 will no longer be well separated from impurity peak 7. Thus a much larger sample can be separated (with ≈100% recovery of pure peak 8), when the conditions of Figure 15.2b are used rather than those of Figure 15.2a. 15.1.2.3 Other Considerations For the larger sample weights encountered in prep-LC, issues other than column dimensions or separation selectivity may also become important. One consequence of the chromatographic process is that components leaving the column are greatly 15.1 INTRODUCTION 729 Time (min) 1 2 0246810 Time (min) 0246810 Time (min) 0246810 Time (min) 0246810 3 4 5 6 7 8 (product) 9 Optimized analytical separation Optimized preparative separation Overloaded preparative separation Overloaded analytical separation (a) (b) (c) (d ) 1 3 2 4 5 + 6 8 9 7 1 3 2 4 5 + 6 8 9 10 10 10 7 8 1 2 3 4 5 9 10 7 6 Figure 15.2 Analytical and preparative conditions compared for the optimum separa- tion of a sample. Sample: a mixture of substituted anilines and benzoic acids. Conditions: 150 × 4.6-mm (5-μm) C 18 column; acetonitrile-buffer mobile phases; flow rate 2.0 mL/min; other conditions noted in figure. (a) Conditions optimized for the separation and analysis of all compounds in the sample (small sample); (b) conditions optimized for the prep-LC purifi- cation of the product peak 8 (small sample); (c) T-P separation of sample with prep-LC condi- tions of (b)(largesample);(d) injection of large sample as in (c), but with analytical conditions of (a). Reprinted from [2] with permission of Wiley-Interscience. diluted, and the ease of removing solvent from collected fractions is often a major issue. Solvent removal is generally easier for normal-phase chromatography (NPC) than for reversed-phase chromatography (RPC) because organic solvents are easier to evaporate than water. For the case of a few milligrams of product, dissolved in a few tens of milliliters of aqueous mobile phase from a RPC separation, solvent-free product can be recovered conveniently with a rotary evaporator. The removal of 730 PREPARATIVE SEPARATIONS larger amounts of aqueous solvent, however, requires much more effort and cost; for this reason many (but not all) prep-LC separations tend to be carried out by NPC, rather than by RPC. For sample weights > 10 mg, analytical HPLC equipment is often too small, and its detectors too sensitive for prep-LC. For the separation of these larger samples, specialized equipment may be required that features high-flow pumps, automated sampling and fraction collection, and a detector fitted with a preparative flow cell. 15.2 EQUIPMENT FOR PREP-LC SEPARATION As noted above, many small-scale prep-LC separations can be carried out with analytical chromatography systems (Chapter 3), possibly with minor modifications for increased injection volumes or decreased detector sensitivity. As sample weight increases beyond a few mg, however, it is more convenient to increase column size (scale-up) than to use (time-consuming and tedious) multiple injections with an analytical-scale column. This may require a corresponding change in equipment to a dedicated prep-LC system that allows higher flow rates and the processing of larger sample weights. Table 15.2 summarizes approximate guidelines for column size, equipment type, and flow rates for different scales of operation. For separations at the gram scale, a dedicated prep-LC system is usually necessary because of the required flow rates. Semi-preparative equipment is essentially similar to an analytical HPLC unit, but with a higher flow-rate pump and some arrangement for fraction collection. Small- and laboratory-scale prep-LC systems are typically used for the isolation of tens of grams to kilograms. In these systems columns with internal diameters as large as 11 cm are often used; larger diameter columns, which are appropriate for multi-kilogram scale projects, are better used within a kilo-lab or pilot plant to handle the large volumes of solvent required in an explosion-proof environment. These larger-diameter-column systems are outside the scope of the present chapter. 15.2.1 Columns It is best to develop a prep-LC separation with an analytical column, using a column packing that is available in larger diameter columns. The use of small-diameter Table 15.2 Approximate Sizes of Columns and Equipment Used for Prep-LC on a Laboratory Scale Quantity Column Internal Equipment Scale-up Ratio for Flow Desired Diameter (mm) Rate and Sample Size <1 mg 4.6 Analytical (∼1 mL/min) (1) 1–100 mg 10 Analytical (∼5 mL/min) 4.7 100 mg–5 g 20–30 Semi-preparative (20–50 mL/min) 19–42 5–100 g 30–50 Small-scale preparative (50–150 mL/min) 42–120 200g–2 kg 50–110 Lab-scale preparative (100–600 mL/min) 120–570 15.2 EQUIPMENT FOR PREP-LC SEPARATION 731 columns during method development minimizes any unnecessary consumption of sample and mobile phase, as well as the need for a more expensive prep-LC system. The columns used for semi-prep and lab-scale prep-LC are closely similar to those employed for analytical separations (Chapter 5). For sample weights of <10 mg, the analytical column itself can often be used since—depending on the separation—such columns may be compatible with injections of several milligrams of sample. A prep-LC column should be packed with the identical column packing that was used for the analytical column prior to scale-up. This ensures that there will be no change in relative retention (selectivity) between the two columns, which can be especially important for prep-LC (Section 15.3.2). Note that it is also important that the particle size be the same for both the analytical and prep-scale columns, so that the same (also optimized) column efficiency and resolution found for the smaller column will be duplicated for the larger column. When moving to larger diameter columns (Table 5.2), with a corresponding increase in flow rate and sample volume, certain other considerations should be kept in mind. Be sure that the time lapse between the sample leaving the detector and entering the fraction collector is small, for both small- and large-diameter columns. Fractions are usually collected on the basis of the detector signal; if there is a significant volume of tubing between the detector cell and the fraction collector, there can be an appreciable time lag between detection of a peak and its collection; this can lead to mistakes in starting and ending fraction collection, with either a loss of product or its contamination by an adjacent impurity. Be aware of the tubing diameter that leads from the detector-cell outlet, which may be significantly larger than that used for the inlet. It is a simple matter to calculate the internal volume of the tubing, which with the flow rate determines the time lapse. With the higher flow rates used in prep-LC, the time lapse will be reduced proportionally—other factors equal. The higher flow rates used with larger columns will increase the pressure drop across connecting tubing, if the same (analytical) equipment is used. Many HPLC systems are constructed with narrow capillary tubing (0.005–0.010-in. i.d.), in order to reduce extra-column peak broadening (Section 3.4). Narrow tubing will lead to a higher pressure drop, with a possible shut-down of the system, when the original analytical system is used with a flow rate of 5 to 10 mL/min with a 10-mm column (instead of the normal 1–2 mL/min used for 4.6-mm columns). This is especially true when more-viscous solvents are used, as in RPC. Special attention should be given to the tubing from the detector outlet, since small-i.d. tubing combined with higher flow rates can lead to a higher back-pressure and damage to the detector flow cell. It may be necessary to replace the original tubing with wider diameter tubing, (as large as 0.020-in. for 20-mm-i.d. columns). At the same time, keep in mind the effect of such a change on the time lapse between peak detection and collection (as discussed above). Instrument manufacturers can also advise the user on how to set up their equipment for prep-LC applications. 15.2.2 Sample Introduction For small sample weights and the use of an analytical-scale HPLC system, sample introduction is usually carried out in one of two ways: (1) injection with the loop-injector that is part of the system, or (2) injection of the sample by means of a 732 PREPARATIVE SEPARATIONS separate pump. In dedicated prep-LC systems, the sample is usually introduced with a sample pump that is different from the mobile-phase pump(s). 15.2.2.1 Loop Injectors When using the standard injector that forms part of the analytical HPLC system, the maximum injection volume may be insufficient for prep-LC. The original sample loop can be replaced by a larger volume loop; these are available from a number of suppliers or can be made easily from bulk stainless-steel tubing. For systems fitted with autosamplers, large-sample-volume options are often available from the manufacturer. It is important to maintain the sample during injection as a cylindrical plug of approximately constant volume. Any dilution of the sample plug by mobile phase will increase the sample volume, which may compromise the separation (Section 15.3.2.2). Care should be taken when choosing a larger loop, as peak spreading in an open tube is proportional to the sixth power of tube i.d. Thus, when the loop diameter is increased, the injected sample plug may exhibit increased tailing and broadening. The use of a longer (vs. wider) sample loop will also increase the width and tailing of the sample plug, but usually to a lesser degree. The trailing edge of the sample plug will be much more spread out than the front of the plug because the tail of the plug (but not the front) must traverse the length of the sample loop for properly designed sample injection (Section 3.6.1.2). The extent of the trailing edge can be determined by injecting a nonretained, UV-absorbing compound (i.e., with k = 0), then observing the resulting peak that leaves the column. If the sample plug entering the column does not tail significantly, the latter nonretained peak will be symmetrical. One means of eliminating sample-tailing during injection is by partially emp- tying the sample loop. The filled sample loop is connected to the column for a time that is long enough to allow the required amount of sample to enter the column, but without introducing the end of the sample plug (that will be diluted with mobile phase). To achieve this result, the sample valve is switched from the inject position back to the load position before the sample loop is completely emptied. This ensures that the injected sample plug will not deteriorate the separation; however, the sample remaining in the loop may be lost. 15.2.2.2 Pump Injection This technique is more convenient and applicable for large injection-volumes. It requires a 2-pump (i.e., gradient) system for isocratic prep-LC separations; the sample is introduced to the column by means of one pump, with subsequent elution of sample by the second (mobile-phase) pump. Best results are obtained with a high-pressure-mixing gradient system (Section 3.5.2.1), because of its minimal dead-volume after the gradient mixer. One of the two pumps is used to supply the sample, while the other pump delivers the (pre-mixed) mobile phase. The sample pump is first primed with the sample solution, after which injection is accomplished by simply switching the flow from the mobile-phase pump to the sample pump for a length of time (depending on flow rate) that will supply the required sample volume. For pump injection with a low-pressure-mixing gradient system (Section 3.5.2.2), one of the solvent inlet lines to the mixer is used for delivering the 15.2 EQUIPMENT FOR PREP-LC SEPARATION 733 sample. The sample is loaded by programming a step-gradient that switches from the mobile-phase line to the sample line and back again. Because of the larger dead-volume of low-pressure gradient systems (Section 3.5.2.2), it is advisable to measure the extent of peak broadening during injection, as described above (Section 15.2.2.1). Significant sample losses may occur with low-pressure gradient systems as a result of priming the pump and tubing; these sample losses can be both substantial and difficult to avoid. Dedicated prep-LC units usually have a sample-injection pump that is separate from the mobile-phase pump(s). Where the sample volume is limited and the system volume is large, it is better to use manual injection—or aspirate the sample through a small tube directly into the sample pump. Sample injection is often operated in a stopped-flow mode; the mobile-phase pump is stopped, and the feed pump is actuated to pump the required sample volume directly to the column. This direct, on-column injection with a feed pump eliminates the tailing that may be seen in alternative systems where an injection valve with a large sample loop is used with injection of the entire contents of the loop. 15.2.3 Detectors A general description of HPLC detectors is provided in Chapter 4. The present section will emphasize detectors and their characteristics that are most relevant for prep-LC. 15.2.3.1 UV Detectors In most cases, the same UV detector can be used for both analytical and prep-LC applications. However, it is advisable to fit the detector with a short-path-length (≈1 mm) flow cell that allows operation at the optimum wavelength without detector overload. Some detector cells are available with variable path-lengths that can be selected for different separations. Despite the use of prep-LC flow cells, the sample absorbance can still exceed 1 to 2 AU, with peaks that are off scale and chromatograms that do not return to baseline—so that monitoring the separation becomes difficult or impossible. Excess detection sensitivity can be reduced by selecting a suitable (non-optimum, usually longer) wavelength, but note the possibility that impurities may absorb much more strongly than the product at the new wavelength, with resulting problems in recognizing the product peak for fraction collection. An example is shown in Figure 15.3, where the small-sample chromatogram (Fig. 15.3a, with detection at 280 nm) does not indicate any impurities with significant UV absorption at this wavelength (only two major peaks). When the sample load is increased (Fig. 15.3b), however, the detector signal is quickly overloaded at 280 nm. For detection at a longer detector wavelength (375 nm), in an attempt to bring the product peak on scale, the impurity peaks are relatively enhanced—to the point that the major components are no longer clearly identifiable. For such a sample another detector may be necessary. 15.2.3.2 Other Detectors The refractive-index (RI) detector is not often sufficiently sensitive for analytical use, but it can be quite useful for prep-LC applications—precisely because of this 734 PREPARATIVE SEPARATIONS min 0 2 4 6 8 10 mAU 0 500 1000 1500 2000 (a) 280-nm detection min 0 2 4 6 8 10 mAU 0 500 1000 1500 2000 (b) 375-nm 280-nm Figure 15.3 Difficulty in monitoring a prep-LC separation when the UV wavelength is changed to decrease detection sensitivity. (a) Analytical chromatogram, 280 nm; (b) prepar- ative chromatogram, 280 and 375 nm. insensitivity. Because the refractive index of the mobile phase can exceed that of some sample components, negative peaks are possible—a problem that need not prove serious if the fractions are collected manually, or as long as any automation software used for fraction collection can function correctly when negative peaks are present. For related components (which comprise the majority of samples) the RI detector provides similar detection sensitivity and a better representation of relative concentration than the UV detector—thus largely avoiding the problem of Figure 15.3b. In principle, any detector can be used for prep-LC (Chapter 4). The size of the detector flow cell may preclude its use with the higher flow rates that are common in prep-LC, in which case a stream splitter can be used to bypass the flow cell. A low-dead-volume tee is inserted into the outlet line from the column and connected by short, small-diameter tubing to the detector. The flow through the cell is then controlled by the length and diameter of the tubing from the other branch of the tee to the fraction collector. Care should be taken in balancing the flow rates to ensure that the detector output is synchronized with the peaks entering the fraction collector. Ideally the volumes of the two tubes downstream of the splitter should be in the same ratio as the volumetric flow rates through them, in order to ensure that the peak arrives at the fraction collector and detector at the same time. Stream splitters should also be used for detectors that are destructive of the sample (evaporative light scattering or mass spectrometry [MS]). The use of MS detection for small-scale prep-LC is increasing, mainly for complex mixtures where compounds are collected based on their molecular weight. 15.2 EQUIPMENT FOR PREP-LC SEPARATION 735 15.2.4 Fraction Collection Fraction collection can be carried out manually, by using the detector signal to determine when to begin and end fraction collection. For repetitive separations, however, operator fatigue rapidly ensues—with collection of the wrong fractions, or diversion of product to waste. When more than an occasional, small-scale prep-LC separation is contemplated, the use of a fraction collector is recommended. For some prep-LC systems the fraction collector will form an integral part of the system. It is also possible to purchase an add-on fraction collector, the most useful of which can be programmed to collect according to the detector signal. By means of a combination of fraction time window, peak threshold, and baseline slope, the correct assignment of fractions to the product container can be accomplished. These fraction collectors can function automatically and run continuously, or until the desired amount of sample has been processed. Dedicated semi-preparative and preparative units have built-in fraction collec- tors. Small-scale prep-LC fraction collectors may use 96-well plates, while larger units generally have a fixed set of fraction-collection valves mounted in a manifold, to allow fractions to be selected and collected. Systems which use fixed-volume fraction-collection devices such as a 96-well plate or a multi-tube collector will gen- erally collect on a time basis, to be sure that the fraction volumes are not exceeded. It is then the responsibility of the operator to combine those fractions that contain the purified product. Where collection occurs via a manifold of collection valves, the fractions can be collected in any suitably sized container, the required volume of which can be calculated from the peak width, the flow rate, and the number of injections to be made. It is important to choose a fraction collector that meets the likely requirements of the prep-LC facility. When only a single product is recovered—as is often the case—only a few fraction-collection ports are needed. When more complex samples are separated into multiple fractions, more fraction-collection ports will be required. In the latter case a multi-tube fraction collector (similar to an autosampler) will usually be preferable—although care must be taken not to exceed the volume of the collection vessels. Fortunately, very complex samples are more likely to be encountered for small-scale separations in a research laboratory, while larger scale separations generally involve collection of only one or two components so that no more than 5 or 10 fraction ports are necessary. Because it is often found that small impurities may elute at the front or on the tail of the major peaks, it is common practice, at least during the early stages of a prep-LC separation, to collect narrow fractions at the front and tail of the peak to ensure that the desired purity is reached. Thus, as many as three fraction ports may be required for each compound to be purified. It is always better to collect too many fractions across a peak, and then combine the pure fractions, than to collect a single fraction that is too broad, and thus less pure. 15.2.5 Product Recovery (Removal of the Mobile Phase) Product recovery can influence the entire plan for prep-LC method development, for example, the initial choice between RPC or NPC. When the separation results in several grams of product dissolved in several liters of an aqueous RPC mobile phase (possibly containing a nonvolatile buffer), separation of the product from the mobile . be given to the tubing from the detector outlet, since small-i.d. tubing combined with higher flow rates can lead to a higher back-pressure and damage to the detector flow cell. It may be necessary to. form an integral part of the system. It is also possible to purchase an add-on fraction collector, the most useful of which can be programmed to collect according to the detector signal. By means of. of the tee to the fraction collector. Care should be taken in balancing the flow rates to ensure that the detector output is synchronized with the peaks entering the fraction collector. Ideally