386 NORMAL-PHASE CHROMATOGRAPHY Non-localizing B-solvent Basic-localizing B solvent MTBE THF Non-basic localizing B-solvent Ethylacetate ACN CH 2 Cl 2 CHCl 3 1 263 45 7 Figure 8.15 Use of seven experiments for the optimization of solvent-type selectivity in NPC. A-solvent: hexane, heptane, or ethoxynonafluorobutane, or (for ε 0 > 0.30) CH 2 Cl 2 .Foravery strong localizing B-solvent use n-ori-propanol. • the purification of crude samples • the separation of isomers • orthogonal separation • samples that contain hydrophobic interferences • samples that contain very polar analytes (e.g., unretained in near-100% water by RPC) The purification (preparative separation) of organic-soluble samples is usually best carried out with NPC and a silica column. The removal of solvent from separated fractions is easier for the organic solvents used in NPC, compared with the higher boiling water used in RPC. Larger values of α for the solute to be purified are usually possible with NPC and silica columns, meaning that larger sample weights can be injected (other conditions the same). Finally, the use of TLC can be convenient for the preliminary assay of fractions from preparative NPC. As noted above (Section 8.3.5), isomeric compounds are usually much better separated by means of NPC with silica columns. However, many isomers are easily separated by RPC, so it may be preferable to try RPC first for such samples—unless prior experience with related samples suggests otherwise. Orthogonal separations (Section 6.3.6.2) can be used to check whether a proposed method has separated all of the compounds in a sample. In principle, the selectivity of an orthogonal method should be as different as possible from that of the original method. In the past some workers have assumed that separations based on a different principle (e.g., NPC vs. RPC) are more likely to provide a large enough difference in selectivity so that any pair of peaks overlapped in one method 8.4 METHOD-DEVELOPMENT SUMMARY 387 will be separated by an orthogonal method. While this suggests the use of NPC for orthogonal separations, it is also possible to design two orthogonal RPC methods (Section 6.3.6.2), especially for the case of ionic samples whose selectivity is more dependent on separation conditions (Section 7.3.2). Hydrophobic interferences in a sample are a less common reason for the use of NPC. Many samples contain hydrophilic additives or impurities that are much less retained in RPC than sample peaks of interest; these non-analytes then appear as ‘‘junk’’ or ‘‘garbage’’ peaks near t 0 (as in the example of Fig. 2.5b). If these hydrophilic non-analytes do not overlap peaks of interest, they may not need to be removed from the sample prior to RPC separation. When sample non-analytes are more hydrophobic than peaks of interest, they will elute much later in RPC, resulting in a very long run time (as in the example of Section 8.4.3 below). Either sample pretreatment will be required for their removal (Chapter 16) or gradient elution can be used to elute hydrophobic impurities within a reasonable time. Either of the latter two approaches may be inconvenient or ineffective for some samples. In such cases NPC separation may be the best solution, inasmuch as hydrophobic non-analytes will be retained less strongly than peaks of interest, and leave the column near t 0 . Very polar samples are often insufficiently retained by RPC (Section 6.6.1), whereas they are likely to be well retained by NPC. For this application, HILIC (Section 8.6) is often a first choice, especially for ionic samples. NPC method development proceeds in similar fashion as for RPC method development. The same seven steps described in Section 6.4 also apply for NPC: 1. define the goals of separation: resolution, run time, detection limits, and so on 2. carry out sample preparation (Chapter 16) 3. choose separation conditions 4. verify column reproducibility, choose alternative columns (Section 6.3.6.1) 5. develop a routine orthogonal separation (Section 6.3.6.2) 6. carry out method validation (Section 12.5) 7. develop a system suitability test (Section 12.3) For the case of preparative separations (Chapter 15), the choice of separation conditions is of primary importance, and method-development steps 1 and 4 through 7 can often be ignored. For all applications the primary difference in the case of NPC versus RPC is step 3, the choice of separation conditions. The following four steps apply for the development of any isocratic HPLC separation (see Section 9.3.10 for gradient methods): 1. select starting conditions 2. adjust %B for an acceptable retention range (e.g., 1 ≤ k ≤ 10) 3. optimize selectivity 4. optimize column length, flow rate, and (possibly) particle size for the best compromise between resolution and run time 388 NORMAL-PHASE CHROMATOGRAPHY Table 8.2 Method Development for Normal-Phase Chromatography Step Approach Comments 1. Starting conditions Choose column and A- and B-solvents from Table 8.1 A silica column is usually preferred; diol-silica is an alternative; hexane and methylene chloride are good first choices for the A- and B-solvents, respectively. 2. %B for 1 ≤ k ≤ 10 Vary % CH 2 Cl 2 -hexane Can use either TLC or HPLC; start with 100% CH 2 Cl 2 ; a 2-fold change in %B will change k by about 3-fold; if a stronger B-solvent is needed, try MTBE, then i-propanol. 3. Optimize α Change B-solvent If CH 2 Cl 2 /hexane can be used as mobile phase, use Figure 8.6 to choose a MTBE/hexane and/or ethyl acetate/hexane mobile phases of same ε value. a Blend B-solvents If further changes in selectivity are needed, blend above mobile phases as in Figure 8.15. a If CH 2 Cl 2 -hexane mixtures are too weak (k > 10), then try other mobile phases from Figure 8.6. Steps 1 to 3 for NPC are discussed below and summarized in Table 8.2. Step 4, the choice of final column conditions, is carried out in the same way as for RPC separation (Section 2.4.3). 8.4.1 Starting Conditions for NPC Method Development: Choice of Mobile-Phase Strength and Column Type (Steps 1, 2; Table 8.2) In RPC method development, it was recommended to begin with a 100 × 4.6-mm C 18 column (3-μm particles) and 80% ACN as mobile phase (Table 6.1), followed by the adjustment of % ACN so as to achieve 1 ≤ k ≤ 10. For NPC, either a silica or polar-bonded-phase column may represent a reasonable starting point. Silica columns are preferred for preparative separations and the resolution of isomers, while polar-bonded-phase columns present fewer problems for assay procedures, and are advantageous for wide-range samples that might require gradient elution if a silica column is used. The selection of an initial mobile phase involves a number of considera- tions, as reviewed in Section 8.3.2.1. A choice of mobile-phase strength (%B) is conveniently made on the basis of exploratory TLC separations with silica plates (Section 8.2.3). Alternatively, gradient elution with a polar-bonded-phase column can also be used (Section 9.3.2), especially if the latter column will be use for the final separation (instead of silica). In either case, the goal should be the estimation of a mobile-phase composition that will provide 1 ≤ k ≤ 10 in a subsequent column separation (equivalent to 0.1 ≤ R F ≤ 0.5 in a TLC separation). We recommend an initial TLC separation with a silica plate and 100% methylene chloride (B-solvent) as mobile phase. If values of R F are too high, then the mobile phase is too strong, and lower values of % methylene chloride (with hexane as A-solvent) can be 8.4 METHOD-DEVELOPMENT SUMMARY 389 investigated—using Section 8.2.2 as a guide. Values of R F that are too low with 100% methylene chloride as mobile phase will require a stronger B-solvent; in this case mixtures of MTBE and hexane are a good choice for subsequent TLC separa- tions. If a still stronger B-solvent is required, try mixtures of n-oriso-propanol with hexane as A-solvent. In each of the latter cases the solvent nomograph of Figure 8.6 can be useful in selecting values of %B for successive experiments. During the adjustment of %B for 1 ≤ k ≤ 10, a decrease in values of k by a factor of two requires an increase in ε by ≈0.05 units (a decrease in ε by ≈0.05 units will result in a 2-fold increase in k). For example, assume that a TLC separation with 100% methylene chloride as mobile phase gives 0.5 ≤ R F ≤ 0.8 for the sample bands, corresponding to 0.2 ≤ k ≤ 1 (Fig. 8.8 or Eq. 8.5). In this case we need an increase in k by 5- to 10-fold. A decrease in ε by 0.05, 0.10, or 0.15 units should increase values of k by a factor of about 2, 2 2 = 4, or 2 3 = 8, respectively. For this example, a decrease in ε 0 by 0.15 units is suggested. Referring to Figure 8.6, we see that ε for 100% methylene-chloride is 0.30, which means that the recommended mobile phase will have ε 0 = 0.15; a mobile phase of 18% methylene chloride-hexane would therefore be suggested by Figure 8.6. Because estimates of k as a function of %B can be less accurate in NPC, further adjustment of %B will usually be required to achieve the desired retention range of 1 ≤ k ≤ 10, as well as to take advantage of solvent-strength selectivity. At the conclusion of exploratory TLC studies with silica plates, we need to choose a NPC column (unless TLC can be used for routine analysis). Silica columns provide larger values of α for preparative separations and the resolution of isomers, while polar-bonded-phase columns are less subject to the various problems described in Section 8.5, and are more suitable for samples whose retention range exceeds 0.5 ≤ k ≤ 20 (as estimated from preliminary TLC separations). Retention with polar-bonded-phase columns is much weaker than for silica (smaller values of k); if the mobile phase selected by means of TLC as above (with a silica plate) has ε<0.15, retention on a polar-bonded-column may be too weak to provide 1 ≤ k ≤ 10—even with pure hexane as mobile phase (ε 0 = 0.00). In this case the use of a polar-bonded-phase column is not an option. Once a mobile-phase composition has been selected from initial TLC studies (for 0.1 ≤ R F ≤ 0.5, or 1 ≤ k ≤ 10), this mobile phase can be used with a silica column. If a polar-bonded-phase column is used instead, then mobile-phase strength should be lowered by about 0.15 ε-units. For example, if the recommended mobile phase were to consist of 40% methylene chloride-hexane (for a silica column), the corresponding value of ε 0 = 0.22; for a polar-bonded-phase column, the value of ε 0 in Figure 8.6 should then be 0.22 − 0.15 ≈ 0.07. From Figure 8.6, a value of ε 0 = 0.07 corresponds to 5% methylene chloride-hexane, which can be used as mobile phase for an initial separation on a polar-bonded-phase column. Keep in mind that estimates of mobile-phase strength from Figure 8.6 for silica (based on a 2-fold change in k for a change in ε 0 by 0.05 units) are approximate, and the extension of this rule to polar-bonded-phase columns is even less reliable. 8.4.2 Strategies for Optimizing Selectivity (Step 3; Table 8.2) The determination of a suitable value of %B for 1 ≤ k ≤ 10 will usually involve experiments where %B is varied; solvent-strength selectivity (Section 8.3.1) can be 390 NORMAL-PHASE CHROMATOGRAPHY explored at the same time. If further changes in selectivity are required, solvent-type selectivity (Section 8.3.2) should be investigated next—because of its very large effect on relative retention and resolution. Three different B-solvents will be needed for a full exploration of solvent-type selectivity: nonlocalizing, basic localizing, and nonbasic localizing, as illustrated in Figure 8.9. These three binary-solvent mixtures can then be mixed with each other in various proportions to provide any intermediate selectivity (as in the example of Fig. 8.10). A general plan or ‘‘experimental design’’ for the latter approach is illustrated in Figure 8.15, with a list of possible B-solvents of each type. Initial experiments with CH 2 Cl 2 -hexane can be used to determine a value of %B (and ε) for acceptable retention (1 ≤ k ≤ 10). The latter separation corresponds to experiment 1 of Figure 8.15. If a further improvement in selectivity and resolution are desired, experiments 2 and 3 of Figure 8.15 are carried out (e.g., mixtures of MTBE-hexane and ethylacetate-hexane that have the same value of ε as in experiment 1; see Fig. 8.6). An examination of the latter three chromatograms will indicate whether further blending of these three mobile phases (experiments 4 to 7) can provide any improvement in selectivity and resolution. This procedure for optimizing solvent-type selectivity can be quite powerful, especially when silica columns are used. The plan of Figure 8.15 can be compared with the similar optimization of solvent-type selectivity for RPC in Figure 6.24 (see related text for details). When the mobile phases of experiments 1 to 3 of Figure 8.15 are blended, the resulting values of ε may not remain the same, requiring a change in the concentration of the A-solvent. A computer program for the more accurate prediction of values of ε as a function of mobile-phase composition has been reported [30], based on the procedure of [31]; a demo copy of the software is available [19]. 8.4.3 Example of NPC Method Development A method was required for samples containing the polar drug paclitaxel in mixture with a more hydrophobic polymer (poly[sebacic-recinoleic ester-anhydride]) [32]. Structures of these two entities are shown at the bottom of Figure 8.16. Because the drug is more polar than the polymer, an assay by RPC would have required either prior separation of polymer from the drug (because of very late elution of the polymer in RPC) or gradient elution. For this reason NPC separation was explored, with the objective that the polymer would leave the column before the drug (in the vicinity of t 0 ) and thereby preclude a need for sample pretreatment. Initial studies were carried out by means of TLC with silica plates. The use of 100% methylene chloride yielded R F = 0.00 for the drug, so the stronger B-solvents tetrahydrofuran (THF, ε 0 = 0.53) and methanol (MeOH, ε 0 = 0.70) were investigated next, in mixture with methylene chloride. Mobile phases composed of 2–5% MeOH–CH 2 Cl 2 appeared promising from the TLC results presented in Table 8.3 (THF would be a less desirable choice; Section 8.3.2.1); 1.5% MeOH–CH 2 Cl 2 with a silica column provided the satisfactory separation of Figure 8.16a, with UV detection at 240 nm. Note that k ≈ 3 for the paclitaxel peak in the separation of Figure 8.16a, whereas TLC separation predicts k ≈ 6. Somewhat approximate predictions of NPC retention from TLC are expected, but such predictions can still be useful—as in the present example. 8.4 METHOD-DEVELOPMENT SUMMARY 391 Paclitaxel Polymer 0 2 4 6 8 10 Time (min) (a) (b) (c) Fresh sample Polymer Paclitaxel Degraded polymer Degraded sample NHO O OH HO O OH O O O O O O O H 3 CO O C H 2 O O O CH 3 O O CH 2 O O O CH 3 O 8 n m 8 O Figure 8.16 NPC assay of paclitaxel in the presence of a polymeric additive. Conditions: 250 × 4.0-mm silica column (5-μm particles); 1.5% methanol-methylene chloride; 25 ◦ C; 1mL/min.(a) Fresh sample; (b) degraded sample of polymer (stored at pH-7.4 and 37 ◦ C for 60 days); (c) degraded sample of paclitaxel plus polymer. Reprinted with permission from [32]. 392 NORMAL-PHASE CHROMATOGRAPHY Table 8.3 Exploratory TLC Separations of Paclitaxel-Polymer Samples MeOH–CH 2 Cl 2 THF–CH 2 Cl 2 %-MeOH R F Paclitaxel R F Polymer %-THF R F Paclitaxel R F Polymer 1 0.04 0.05 2 0.00 0.56 2 0.22 1.00 4 0.00 0.64 3 0.29 1.00 9 0.12 1.00 4 0.35 1.00 20 0.52 1.00 5 0.46 1.00 30 0.88 1.00 Source: Data from [32]. The method of Figure 8.16a was also intended for use with thermally stressed samples, as carried out in the experiments of Figure 8.16b,c. It appears that thermal degradation of the polymer (Fig. 8.16b) does not result in the formation of peaks that overlap the paclitaxel peak and thereby compromise its assay. For other details of this NPC method development, see [32]. 8.5 PROBLEMS IN THE USE OF NPC Several interrelated problems can occur during NPC separation with silica as column packing: • poor separation reproducibility (including extreme sensitivity to mobile-phase water content) • solvent demixing • slow column equilibration when changing the mobile phase • tailing peaks The first three problems arise from the very strong interaction of small, polar molecules with surface silanols; this in turn can have a dramatic effect on sample retention. 8.5.1 Poor Separation Reproducibility Sample retention times in NPC can vary from day to day, or even within the same day, as a result of significant variations in room humidity, and consequent changes in the water content of nominally ‘‘dry’’ mobile-phase solvents. This effect is illustrated in Figure 8.17a for the elution of benzanilide as the solute with methylene chloride as the mobile phase and a silica column. In this example, the mobile phase was prepared by blending different volumes of methylene chloride that were either water-free or water-saturated, so as to achieve different concentrations of water in the final mobile phase (see top of Fig. 8.17a). For example, blending equal volumes of the two solvents would result in 50% water saturation of the final mobile phase (or 0.08% water). As the water content of the mobile phase is increased, 8.5 PROBLEMS IN THE USE OF NPC 393 the retention of the solute decreases sharply (from k = 9tok = 2) because of the increasing coverage of the silica surface by adsorbed water (water interacts very strongly with silica). It can be appreciated from this example that small changes in room humidity can lead to significant changes in %-water saturation, which in turn can lead to variable sample retention. When a silica column is used, this variability of mobile-phase water-content is the most common cause of variable NPC retention—provided that the column has been properly equilibrated before samples are injected (Section 8.5.2). It is possible to minimize variations in mobile-phase water content and sample retention by controlling the water content of the mobile phase, as described above (blending the water-free with the water-saturated mobile phase). An alternative, more convenient approach is the addition of small amounts of a very polar solvent to the mobile phase. As seen in Figure 8.17b, the addition of small amounts of methanol brings on a reduction in sample retention similar to that of added water—caused by a comparable deactivation of the silica surface. Presumably the addition of methanol renders the column less susceptible to variations in water content. The amount of methanol required for silica deactivation will vary with mobile-phase composition, and be less for weaker mobile phases (with smaller values of ε). The limited miscibility of methanol and hexane suggests that propanol should be substituted when necessary. Note that as the concentrations of either water or methanol in the mobile phase increase, sample retention decreases, and may call for adding hexane to the mobile phase to decrease ε. The ability of silica to separate isomers and other solutes will be compromised by excessive deactivation of the column. Variable retention due to changes in mobile-phase water content, as in Figure 8.17a, should be less pronounced for (1) more-polar mobile phases with larger values of ε or (2) less-polar, bonded-phase columns (because water is less tightly bound to such columns). Consequently polar-bonded-phase columns are less likely to be affected by problem than silica columns. Retention variability with silica columns may not even be a problem if the sample, mobile phase, the relative constancy of room humidity are controlled, and if solvents are transferred from their original bottles to the reservoir with minimum exposure to the atmosphere. 10 8 6 4 2 k 0 0.05 0.10 0.15 0.2 0 0.05 0.10 0.15 0 50 100 10 8 6 4 2 k (a)(b) %-water saturation % MeOH % H 2 O Figure 8.17 Effect of polar deactivators on NPC retention. Sample: benzanilide; conditions: type-B silica column; mobile phase, methylene chloride with varying concentrations of water (a) or methanol (b); 35 ◦ C. Adapted from [37]. 394 NORMAL-PHASE CHROMATOGRAPHY It is therefore prudent to wait until retention variability becomes an issue before adjusting the %-water-saturation of the mobile phase or adding a polar solvent such as methanol. A second possible cause of retention variability in NPC can arise from changes in mobile-phase composition as a result of helium sparging. The evaporation of the mobile phase that results can lead to a preferential loss of one solvent over the other [33], as well as changes in the water content of the mobile phase. Such changes in mobile-phase composition can be reduced by the use of on-line degassing of the mobile phase (Section 3.3.3). 8.5.2 Solvent Demixing and Slow Column Equilibration When dilute solutions of a polar solvent first contact a silica column or TLC plate, the polar solvent will be selectively taken up by the silica. This leaves a mobile phase that is depleted in the polar solvent, generally resulting in a lowering of observed R F values and larger values of k. Solvent demixing is mainly a problem for TLC or gradient elution, since isocratic separations are usually preceded by column equilibration (Section 2.7.1). Solvent demixing in TLC reduces the mobile-phase strength, so separations by TLC and column chromatography may no longer be equivalent. Consequently values of k estimated from TLC experiments may be too high, as in the example of Section 8.4.3 (which may or may not be due to solvent demixing). In gradient elution with a silica column, where the concentration of a polar B-solvent increases during the separation, solvent demixing can lead to an interruption of the gradient and a related deterioration of separation. For this reason gradient elution in NPC with a silica column is often avoided. Nevertheless, Meyer has shown [34] that gradients of hexane and methyl-t-butylether (ε 0 = 0.48) do not present problems of this kind. The same may be true of other gradients with silica columns, as long as the difference in ε 0 values of the A- and B-solvents is no greater than 0.5. See also the discussion of NPC gradient elution in [35]. For isocratic NPC separation, solvent demixing is not ordinarily a problem, but it may require a longer pre-equilibration of the column. Because of solvent demixing, when changing mobile phases the equilibration of the column will often be slower in NPC than in RPC; this will require a larger volume of mobile phase (for column equilibration) before injecting samples. Column equilibration in NPC can be especially slow for the case of less-polar mobile phases, if the %-water saturation of the new mobile phase is very different from that of the previous mobile phase. The reason is that the capacity of the column for water (weight water/weight stationary phase) is much greater than that of the mobile phase, so a large volume of mobile phase must pass through the column in order to transfer enough water between column and mobile phase to reach equilibrium. Polar-bonded-phase columns will be less subject to solvent demixing and slow column equilibration because polar solvents are less strongly retained by these columns (compared to silica columns); polar-bonded-phase columns are therefore more amenable for use with gradient elution. 8.5.3 Tailing Peaks Peak tailing has traditionally been a more important problem for NPC than for RPC. Similarly the column plate number N is often smaller in NPC than in RPC, possibly 8.6 HYDROPHILIC INTERACTION CHROMATOGRAPHY (HILIC) 395 because of slower diffusion on the silica surface (stationary-phase diffusion [36]). Just as in the case of RPC (Sections 5.2.2.2, 7.3.4.2), column performance for NPC is much improved by the use of type-B silica in place of the older type-A silica [37]; peak shapes are generally better, and plate numbers higher. For this reason the use of type-B silica is strongly recommended for analytical separations by means of NPC. For preparative separations, the higher cost of type-B silica may not be justified, especially as these separations often tolerate lower values of N (Section 15.4.1.1). 8.6 HYDROPHILIC INTERACTION CHROMATOGRAPHY (HILIC) Hydrophilic interaction chromatography (HILIC) can be regarded as normal-phase chromatography with an aqueous-organic mobile phase [38–40]; for this reason it is sometimes referred to as ‘‘aqueous normal-phase chromatography.’’ HILIC columns are more polar than RPC columns, and the more-polar water serves as the stronger B-solventinHILIC—sothatanincreasein%-waterresultsinadecrease in sample retention (the opposite of RPC behavior). An example is provided in Figure 8.18 for the separation of a mixture of neutral oligosaccharides by HILIC, using mobile phases of water and acetonitrile; as the %-water increases, retention decreases (see also the similar example of Fig. 8.19 for the HILIC separation of several peptides). More-polar and/or ionized solutes tend to be more strongly retained in HILIC, other factors equal (again, the opposite of RPC, but typical of NPC). In most HILIC separations the mobile phase is varied from 3 to 40% water. There is usually little retention (k ≈ 0) for water concentrations > 40%, although occasionally—for some solutes and columns—retention can begin to increase as the water concentration increases beyond 40% (i.e., onset of RPC behavior) [42]. The preferential retention of polar solutes in HILIC means that many samples that exhibit poor retention in RPC (k ≈ 0) can be better separated by HILIC. HILIC is also characterized by several other potentially advantageous features [43]: • good peak shape for basic solutes • enhanced mass-spectrometer sensitivity • possibility of direct injection of samples that are dissolved in a primarily organic solvent (which would be unsuitable for RPC; Section 2.6.1) • higher flow rates (or lower column pressures) possible, because of the lower viscosity of the mobile phase (Table I.4 in Appendix I) While relative retention for NPC with a silica column tends to be the reverse of RPC retention (Fig. 8.1b), HILIC retention is often intermediate between these two extremes. The latter observation may reflect the fact that solute ‘‘polarity’’ is a complex function of (1) molecular structure and (2) the kinds of sample-column interactions that are important for retention and especially selectivity. Many (perhaps most) HILIC separations are carried out by means of gradient elution. However, the following discussion for isocratic HILIC separations is equally applicable for gradient elution with HILIC; see the further discussion of Section 9.5.3. . required to achieve the desired retention range of 1 ≤ k ≤ 10, as well as to take advantage of solvent-strength selectivity. At the conclusion of exploratory TLC studies with silica plates, we need to. hexane to the mobile phase to decrease ε. The ability of silica to separate isomers and other solutes will be compromised by excessive deactivation of the column. Variable retention due to changes. often tolerate lower values of N (Section 15.4.1.1). 8.6 HYDROPHILIC INTERACTION CHROMATOGRAPHY (HILIC) Hydrophilic interaction chromatography (HILIC) can be regarded as normal-phase chromatography