286 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES Choice of Separation Conditions (b) 1. Initial separation 1a. Vary %B for 1 ≤ k ≤ 10 (Section 6.2.1) or 1b. Start with gradient run (Section 9.3.1) 2. Separation problems? (Sections 2.4.2, 9.3.1.1) 3. Reconsider initial conditions: RPC (Chapter 6, or Section 7.3) IPC (Section 7.4) IEC (Section 7.5) NPC (Chapter 8) 4. Confirm peaks as acid, base, or neutral (Section 7.2) 5. Optimize selectivity (a) 6. Adjust column conditions (N, Section 2.5.3) Figure 6.21 (Continued) can be varied (step 6) for the purpose of either increasing resolution or reducing run time. Samples that are relatively easy to separate may require no more than the selection of a final value of %B, which involves only a few experiments in which %B is varied. If gradient elution is used during initial method development exper- iments, only a single experiment is needed in order to select a value of %B for 1 ≤ k ≤ 10 (Section 9.3.1). Many samples will require a further improvement of separation selectivity (preceding Section 6.3); for some samples this may involve the simultaneous change of two or more separation conditions. Various procedures for such multiple-variable optimization will be described next. 6.4.1 Multiple-Variable Optimization Multiple-variable optimization in each case relies on an experimental design:aplan for the required experiments, as illustrated in Figure 6.22 for certain combinations of conditions that affect selectivity for neutral samples. In each case it is assumed that %B has been varied initially, so as to define a range in %B that provides adequate retention of the sample, for example, 40–50%B, so that 0.5 ≤ k ≤ 20 for every peak (when varying %B for a change in selectivity, a wider k-range than the usual 1 ≤ k ≤ 10 is recommended). By way of illustration, first consider Figure 6.22a. Experiments 1 and 2 are carried out first (i.e., a change in %B only). These two 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY 287 low-% ACN high-% ACN low-% MeOH high-% MeOH (b) 12 3 4 56 (a) low %B, high T high %B, high T low %B, low T high %B, low T 2 4 1 3 ACN THF MeOH 153 2 46 7 (c) For each of the above runs in (a)-(c), %-water is varied to maintain 0.5 ≤ k ≤ 20. Figure 6.22 Experimental designs for the simultaneous optimization of various separation conditions for optimum selectivity. (a) Solvent strength (%B) and temperature (T); (b) solvent strength and solvent type (MeOH and ACN); (c) solvent type (MeOH, ACN, and THF). experiments may suggest successful separation for some final value of %B, where both 0.5 ≤ k ≤ 20 and resolution is adequate. If acceptable separation cannot be attained in this way, experiments 3 and 4 are carried out next (repeat experiments 1 and 2 at a higher temperature T). These four experiments can be interpolated (or extrapolated) to estimate values of T and %B that provide optimum selectivity and maximum resolution. A final experiment with these promising conditions is then carried out to confirm the predicted separation. Experimental-design experiments, because of the simultaneous variation of two (or occasionally more) different conditions, can be difficult to interpret (espe- cially for samples that contain a large number n of components; e.g., for n > 10). For this reason experimental design is often used in combination with computer simulation (Chapter 10). A reliable peak-tracking procedure will also be necessary (Section 2.7.4). 6.4.1.1 Mixtures of Different Organic Solvents Two organic solvents (B-solvents) have been noted previously as especially suitable for RPC: acetonitrile (ACN) and methanol (MeOH). Tetrahydrofuran (THF) is 288 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES H-B Basic (β 2 ) H-B acidic (α 2 H ) Dipolar (π*) MeOH ACN THF Figure 6.23 Preferred solvents for maximum change in solvent-type selectivity. used less often because of its higher UV cutoff, susceptibility to oxidation, slower column equilibration when changing the mobile phase (e.g., from THF/water to ACN/water), and incompatibility with PEEK tubing (Section 3.4.1.2). However, the oxidation of THF during use can be minimized by the addition of water [55], and many samples do not require detection below 230 nm, thus making the use of THF practical in some cases. Appendix I contains additional information on the properties of these three B-solvents. The remainder of Section 6.4.1.1 describes a procedure that is no longer commonly used; the reader may therefore wish to skip this discussion. The solvent-selectivity triangle of Figure 2.9 has been used to compare the solvent-type selectivity of different B-solvents. As suggested by Figure 6.23, sig- nificant differences in solvent selectivity exist among different mixtures of ACN, MeOH, and THF with water [56]. These three B-solvents can be used to optimize solvent-type selectivity by varying the proportions of each solvent in the mobile phase. Initial experiments with ACN as B-solvent will have identified a value of %B such that 0.5 ≤ k ≤ 20 for the sample. Corresponding values of %-MeOH and %-THF (for equal solvent strength or a similar range in k) can then be obtained from the nomograph of Figure 6.11. The resulting three binary-solvent mobile phases can be blended next in various proportions, as illustrated in Figure 6.22c.Anexample of the application of the experimental design of Figure 6.22c is shown in Figure 6.24 for the separation of a 9-component mixture of substituted naphthalenes. In this example, run 1 is 52% ACN/water; run 2 is 63% MeOH/water; and run 3 is 39% THF/water. While selectivity varies among runs 1 to 3 of Figure 6.24, two or more peaks are poorly resolved in each separation. The next step is to carry out further experiments in which the mobile phase is varied by blending equal portions of mobile phases 1 to 3. Thus a 1:1 (by volume) blend of mobile phases 1 and 2 results in mobile phase 4: a 26/32/42% mixture of ACN, MeOH, and water. Similarly mobile phases 5 to 7 are prepared as follows: 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY 289 Figure 6.24 Use of seven solvent-type-selectivity experiments for the separation of a mixture of nine substituted naphthalenes. Sample substituents are: 1,1-NHCOCH 3 ; 2,2-SO 2 CH 3 ; 3, 2-OH; 4,1-COCH 3 ; 5,1-NO 2 ;6,2-OCH 3 ; 7, -H (naphthalene); 8,1-SCH 3 ; 9, 1-Cl. Condi- tions: 150 × 4.6-mm C 8 column; 40 ◦ C; 2.0 mL/min. Mobile phases (circled): 1, exchange: 1, ACN; 2, MeOH; 2 exchange: 1, ACN; 2, MeOH; 3, 39% tetrahydrofuran/water; 4,1:1mix- ture of 1 and 2; 5, 1:1 mixture of 2 and 3; 6, 1:1 mixture of 1 and 3; 7, 1:1:1 mixture of 1, 2, and 3. Recreated from data of [57]. phase 5, a 1:1 blend of mobile phases 1 and 3 (26/20/54% ACN/THF/water); phase 6, a 1:1 blend of mobile phases 2 and 3 (32/20/48% MeOH/THF/water); and phase 7, a 1:1:1 blend of mobile phases 1, 2, and 3 (17/21/13/49% ACN/MeOH/THF/water). An examination of the latter four chromatograms in Figure 6.24 shows that baseline resolution is achieved with mobile phase 5 (ACN/THF). In most cases further improvements in selectivity and resolution are possible by blending mobile phases 1 to 7 to obtain intermediate mixtures. This approach can be simplified by eliminating unpromising mixtures. For example, runs 1 and 2 in Figure 6.24 show that peaks 6 and 7 overlap in each run; this suggests that no mixture of mobile phases 1 and 2 (e.g., mobile phase 4) is likely to improve the separation of peaks 6 and 7 (as seen in the separation of run 4). 290 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES The experimental design of Figures 6.22c and 6.24 can be further improved by using computer simulation to predict separation as a function of mobile phase composition [57]. Other work suggests expanding the number of experimental mobile phases from seven in Figure 6.22c to 15 [58], for a more reliable prediction of retention as a function of the mobile phase (when using computer simulation). However, the seven mobile phases of Figure 6.22c are usually sufficient for a visual (non–computer-assisted) selection of the final optimum mobile phase. By carrying out the experiments of Figure 6.24 sequentially (1 → 2, 2 → 3, etc.) and observing the result of each separation when completed, an acceptable separation may result without completing all seven experiments. For example, with this approach for the sample of Figure 6.24, only six experiments would be required (runs 1–6)—or five experiments if run 4 is bypassed for the reasons discussed above. When using this (or similar) multiple-solvent optimization approach, the reader is cautioned that more complex mobile phases are more prone to problems than simpler ones. For this reason an ‘‘adequate’’ ternary separation, as in run 5 of Figure 6.24, may be preferable (because of reliability) over a quaternary separation that uses all four solvents to achieve a slightly improved separation. 6.4.1.2 Simultaneous Variation of Solvent Strength and Type If the potential disadvantages of THF as B-solvent are considered unacceptable for a given sample, an alternative approach is the variation of the proportions of ACN, MeOH, and water in the mobile phase [26]. The experimental design for the latter procedure is shown in Figure 6.22b. Beginning with results for two mobile phases with varying %-ACN that cover a range in k of as much as 0.5 to 20 (runs 1 and 2), MeOH/water mobile phases of equivalent strength (runs 3 and 4) are selected with the help of Figure 6.11. Finally, mobile phases 5 and 6 are selected by blending mobile phases 1 and 3 (1:1) and 2 and 4 (1:1), respectively. The use of data collected according to the experimental design of Figure 6.22b then allows an optimum mobile phase to be selected by interpolation. A partial example of this approach for optimizing selectivity was described in Figures 6.8 to 6.10 (data were not reported there for runs 3 and 5 of Fig. 6.22b, which were poorly resolved). The use of all six runs of Figure 6.22b would not change the final optimum conditions of Figure 6.10d. A simpler, but less effective procedure for optimizing %B and solvent type is to successively vary %B for mobile phases that contain first one then another B-solvent (no blending of different B-solvents). An example is shown in Figure 6.25 for the separation of six steroids. In the initial two experiments (Fig. 6.26a,b)%-ACNis varied. As peaks 2 and 3 are unseparated in each run, ACN alone is unable to resolve this sample. The next two experiments (Fig. 6.25c,d) vary %-MeOH, using the nomograph of Figure 6.11 to choose values of %-MeOH that provide similar retention as for the separations of Figure 6.25a,b. Now all peaks can be separated with near-baseline resolution (R s = 1.4 for critical peak-pair 4/5 in Fig. 6.25d [35% B]). As the resolution of peak-pair 4/5 improves with increasing %-MeOH (while that of peak-pair 1/2 decreases), a slight increase in %-MeOH is suggested. For 37%-MeOH a resolution of R s = 1.5 (baseline resolution) was found, with a small reduction in separation time compared to 35%-MeOH (separation not shown). 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY 291 02468 Time (min) 1 2 + 3 4 5 6 020 Time (min) 1 2 + 3 4 5 6 0246810 Time (min) 1 2 4 5 6 3 020 Time (min) 1 2 4 5 6 3 020 Time ( min ) 1 2 5 6 4 3 0246 Time (min) 1 2 5 6 4 3 25% ACN 7 ≤ k ≤ 17 R s = 0.0 (b) 45% MeOH 2 ≤ k ≤ 6 R s = 1.1 35% MeOH 6 ≤ k ≤ 20 Rs = 1.4 25% THF 1.4 ≤ k ≤ 4 R s = 1.4 15% THF 6 ≤ k ≤ 17 R s = 2.2 (c) (d) (e) (f) 35% ACN 1.5 ≤ k ≤ 5 R s = 0.0 (a) Figure 6.25 Separation of six steroids by changes in solvent strength (%B) and type. Sam- ple: 1, prednisone; 2, hydrocortisone; 3, cortisone; 4, dexamethasone; 5, corticosterone; 6, cortexolone. Conditions: 250 × 4.6-mm C 8 (5-μm) column; different mobile phases are organic/water mixtures, as indicated in figure; 35 ◦ C; 2.0 mL/min. Chromatograms recreated from data of [59]. 292 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES 02468 Time (min) 024 Time (min) 13 2 4 5 6 024 Time (min) 1 2 3 4 + 5 6 1 2 3 4 5 6 02 Time (min) 1 2 + 3 4 + 5 6 (a) 45% B, 35°C 2 ≤ k ≤ 13 R s = 0.5 (b) 50% B, 35°C 1 ≤ k ≤ 7 R s = 0.6 (c) 45% B, 70°C 1.4 ≤ k ≤ 7 R s = 1.8 (d) 50% B, 70°C 1 ≤ k ≤ 4 R s = 0.2 Figure 6.26 Separation of a mixture of 6 organic compounds of diverse structure by changes in solvent strength (%B) and temperature. Sample: 1, methylbenzoate; 2, benzophenone; 3, toluene; 4, naphthalene; 5, phenothiazine; 6, 1,4-dichlorobenzene. Conditions: 125 × 3.0-mm C 18 column; mobile phase acetonitrile/water mixtures; 1.0 mL/min. Chromatograms recreated from data of [66]. Because a separation that is barely baseline resolved is considered marginal for a final method (assuming the use of a new, nondegraded column), THF as the B-solvent can be considered next (Fig. 6.25e,f). Even better resolution is obtained for 15%-THF (Fig. 6.25f , R s = 2.2), and this cannot be improved by further changes in %-THF without exceeding k = 20. The procedure of Figure 6.25 for optimizing solvent type and strength can be terminated at any step if an adequate resolution and run time are achieved. Thus all six experiments of Figure 6.25 will be unnecessary for some samples but may be insufficient for others. 6.4.1.3 Simultaneous Variation of Solvent Strength and Temperature It was noted in Table 2.2 that changes in solvent strength (%B) and/or temperature (T) are less effective than a change in solvent type. However, the simultaneous 6.4 METHOD DEVELOPMENT AND STRATEGIES FOR OPTIMIZING SELECTIVITY 293 optimization of both %B and T is often adequate for achieving baseline reso- lution [60–65], while being more convenient and less tedious than alternative multiple-variable optimization schemes. The experimental design for the simulta- neous variation of %B and T is described by Figure 6.22a, while an example of its application is shown in Figure 6.26. In Figure 6.26a,b, %-B is varied for T = 35 ◦ C. There is considerable change in selectivity as %-B is varied, and a resolu- tion of R s = 1.4 is possible for 41% B, with a k-range of 3 ≤ k ≤ 20 (separation not shown). The temperature is next changed to 70 ◦ C, with the results of Figure 6.26c,d. A reasonable separation (R s = 1.8) is obtained for 45% B and 70 ◦ C. Further exper- iments varying %B and T did not increase resolution, as long as retention was maintained within the range 0.5 ≤ k ≤ 20. The use of a simultaneous variation of %B and T has become increasingly popular for the following reasons: (1) only four initial experiments are required, once a value of %B for a reasonable range in k has been established; (2) with on-line mixing of the A- and B-solvents, all four experiments can be carried out automatically, without operator intervention (assuming temperature control by the system controller); (3) there are none of the experimental problems associated with other means of optimizing selectivity [26, 67]; (4) this procedure is often adequate for achieving the desired selectivity and resolution of a sample; and (5) peak matching tends to be easier than for other experimental designs. For more details about this approach for optimizing selectivity and resolution, see [63]. For a variation on this technique, see also [64]. 6.4.1.4 Change of the Column with Variation of One or More Other Conditions It was seen in Figure 6.14 that the use of these four columns (with other conditions the same) resulted in a maximum resolution of R s = 0.8. Such a result for this particular sample is not unusual because a change in column does not allow the convenient use of ‘‘intermediate’’ conditions, as is the case for a change of other conditions in the examples of Figure 6.6 (%B), Figure 6.10 (blended B-solvents), or Figures 6.13 and 6.14 (temperature). A better approach, when changing the column, is to combine a change in column with a further change in one or more other conditions that affect selectivity. A procedure that we recommend (and that many laboratories now use) is a change in the column combined with simultaneous changes in %B and temperature (as in Fig. 6.22a). When this procedure was applied to the four columns of Figure 6.14, the optimized separations of Figure 6.27a–d resulted. Now the range in resolution for the three columns has been increased from 0.0 ≤ R s ≤ 0.8to1.2 ≤ R s ≤ 1.5. If N is increased for the Symmetry C18 column (Fig. 6.27a) by increasing column length (from 150 to 250 mm) and decreasing flow rate (from 2.0 to 1.0 mL/min), a resolution of R s = 2.0 is achieved, with no increase in pressure (but an increase in run time from 7.5 to 25 min). In many cases, simply optimizing %B and temperature for columns of different selectivity (large values of F s ) will lead to a satisfactory separation. For other, more demanding samples, a greater change in selectivity may be needed. For such cases solvent-type optimization as in Figure 6.22b might be combined with a change in column [68]. It was noted in Section 6.3.4 that it is inconvenient to vary column selectivity continuously, as by mixing two different column packings in different proportions to 294 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES 02468 Time (min) 5 2 3 1 4 6 7 8 9 10 02468 Time (min) 5 1 3 2 4 6 7 8 9 10 02468 Time (min) 5 1 3 2 4 6 7 8 9 10 0246 Time ( min ) 5 1 3 2 4 6 7 8 9 10 (a) Symmetry C18 (54% B, 33°C) 1.0 ≤ k ≤ 9, R s = 1.5 (b) Altima HP C18 amide (40.5% B, 40°C) 1 ≤ k ≤ 10, R s = 1.2 (c) Luna pheny - hexyl (46% B, 41°C) 1 ≤ k ≤ 9, R s = 1.4 (d) Spherisorb ODS-2 (50% B, 46°C) 1 ≤ k ≤ 7, R s = 1.4 Figure 6.27 Optimized separation of a mixture of 10 organic compounds of diverse structure on four different columns by varying solvent strength (%B) and temperature. Sample and con- ditions as in Figure 6.14, except as indicated in figure. Chromatograms recreated from data of [8, 9, 69]. form a single column. Two alternatives to the latter procedure have been suggested, neither of which has so far found much practical application. In one approach, small lengths of different columns are connected in series. By varying both column type and length, column selectivity can be varied in discontinuous fashion (called ‘‘phase-optimized liquid chromatography’’ or POPLC ® from Bischoff). Experiments with individual columns define sample retention for each column type, and it is then possible to predict retention for different combinations of columns and column length [71]. A second approach uses two different columns in series, with separate control of the temperature for each column (called ‘‘thermally tuned tandem-column approach’’ [72]). Because sample retention decreases at higher temperatures, the relative contribution of either of the two columns to overall selectivity can be 6.5 NONAQUEOUS REVERSED-PHASE CHROMATOGRAPHY (NARP) 295 increased by lowering its temperature relative to that of the other column. Each of these two procedures appears somewhat complicated for routine application. 6.4.2 Optimizing Column Conditions Column conditions (column length L and diameter d c ,particlesized p , flow rate F) are preferably optimized after other conditions are varied for optimized selectivity and maximum resolution (step 6 of Fig. 6.21b). Particle size, column length and diameter, and flow rate are usually selected prior to the start of method development (e.g., as recommended in Table 6.1) to provide a sufficient plate number for the separation of the sample under study. Following the selection of other conditions (column, mobile phase, temperature), column length and flow rate can be varied to either increase resolution (at the cost of increased run time) or decrease run time (when the initial resolution  2). Figure 6.28 shows examples of each of these cases. The separation of Figure 6.28a has marginal resolution (R s = 1.1), which must be increased. This is most effectively done by an increase in column length, as illustrated by Figure 6.28b for an increase in column length from 150 to 250 mm. Baseline resolution is now (barely) achieved (R s = 1.5), at the cost of a 2/3 increase in run time and pressure. An even greater increase in resolution is desirable; however, the latter example illustrates that an increase in column length comes at a cost of increased run time and pressure, which effectively limits the possible increase in resolution. A decrease in flow rate is an alternative option for increasing resolution, but this is generally not worthwhile. For the example of Figure 6.28a a reduction in flow rate from 2.0 to 1.0 mL/min (not shown) results in an insignificant increase in resolution (to R s = 1.2), while run time is doubled to 10 min. Alternatively, the particle size can be reduced, for example, with a 150 × 4.6-mm column of 3-μm particles. In this caseruntimeremainsthesameasinFigure6.28a, pressure increases to 2900 psi, and R s = 1.6. In each of these examples, resolution, run time, and pressure can be varied to meet some final goal. A reasonable overall compromise might be achieved with a 250 × 4.6-mm, 3-μm column and a flow rate of 1.0 mL/min: R s = 2.1, 2400 psi, with a run time of 15 minutes. See also Section 2.4.1.2. Figure 6.28c,d provides an example of a decrease in run time when the initial resolution is more than adequate. In Figure 6.28c the run time is 30 minutes and R s = 4.8. By decreasing column length and increasing flow rate, run time can be shortened drastically, while maintaining R s > 2.0 and a pressure < 2000 psi. In Figure 6.28d these combined changes in column length and flow rate result in a 10-fold decrease in run time, with R s = 2.1. By decreasing column length in the identical proportion as flow rate is increased, the pressure can be maintained constant (not done in the example of Fig. 6.28c,d). 6.5 NONAQUEOUS REVERSED-PHASE CHROMATOGRAPHY (NARP) Separation by nonaqueous reversed-phase chromatography (NARP) is reserved for very hydrophobic samples that are retained strongly and not eluted by 100% ACN as mobile phase (lipids, synthetic polymers, etc. [74–77]). The mobile phase for NARP separations will therefore consist of a mixture of more polar (A-solvent) and . in flow rate from 2.0 to 1.0 mL/min (not shown) results in an insignificant increase in resolution (to R s = 1.2), while run time is doubled to 10 min. Alternatively, the particle size can be reduced,. respectively. The use of data collected according to the experimental design of Figure 6.22b then allows an optimum mobile phase to be selected by interpolation. A partial example of this approach for optimizing. benzophenone; 3, toluene; 4, naphthalene; 5, phenothiazine; 6, 1,4-dichlorobenzene. Conditions: 125 × 3.0-mm C 18 column; mobile phase acetonitrile/water mixtures; 1.0 mL/min. Chromatograms recreated from