256 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES ‘‘normal-phase’’ chromatography (NPC). As RPC involves a less polar column and a more polar mobile phase, the two phases can be regarded as interchanged or ‘‘reversed.’’ The first RPC packings were made from silica particles that had been reacted with (CH 3 ) 2 SiCl 2 , so as to render the surface nonpolar. A nonpolar stationary phase was then used to coat the particles, allowing their use for reversed-phase liquid–liquid partition chromatography [2], a HPLC technique that is rarely used today. Bonded-phase columns represented the next major advance in ‘‘high- performance’’ RPC. Early C 18 columns of this type were prepared by the covalent attachment of polyoctadecylsiloxane to silica particles [3, 4]. This was followed a few years later by the introduction of columns where individual C 18 groups (rather than a C 18 polymer) were attached to the particle (Section 5.3.1). Beginning in the early 1970s, RPC columns of the latter type were used increasingly because of their many advantages. In the mid-1970s, the use of RPC underwent an explosive growth in popularity; several hundred separations are cited in [5] for the period 1976 to 1979. Whereas earlier applications of RPC emphasized the separation of more hydrophobic (‘‘lipophilic’’) samples, the classic paper by Horv ´ ath [6] demonstrated that RPC could also be used for the separation of relatively polar compounds, especially water-soluble samples of biochemical interest. A little later, RPC was further adapted for the separation of ionizable compounds (including enantiomers) by the addition of ion-pairing compounds to the mobile phase [7] (ion-pair chromatography, Section 7.4). In the late 1970s, several groups reported the first RPC separations of large peptides and proteins, using short-chain, wide-pore column packings. Today, RPC is the dominant HPLC mode and accounts for a substantial majority of all HPLC separations. 6.2 RETENTION Retention in RPC was discussed briefly in Section 2.2. Because very polar molecules interact more strongly with the polar mobile phase, these compounds are less retained and leave the column first. Similarly less polar compounds prefer the nonpolar stationary phase and leave the column last. Thus molecules of similar size are eluted in RPC approximately in order of decreasing polarity. An example is provided by Figure 6.1a, where it is seen that the more-polar benzonitrile (1) appears in the chromatogram first, followed by the increasingly less polar anisole (2), and finally toluene (3). A more detailed example of RPC retention as a function of solute polarity is provided by Figure 2.7c; see also the related discussion of RPC column selectivity in Section 5.4. Retention in RPC is largely the result of interactions between a solute molecule and either the mobile phase or the column (Section 2.3.2.1). An increase in %B (volume-% of organic solvent in the mobile phase) makes the mobile phase less polar (‘‘stronger’’) and increases the strength of interactions between solute and solvent molecules. The result is decreased retention for all solute molecules when %B is increased. This is illustrated by the separation of Figure 6.1b (with a mobile phase of 60% B) compared to that of Figure 6.1a (40% B). An increase in temperature weakens the interaction of the solute with both the mobile phase and column, and 6.2 RETENTION 257 C 18 column 40% ACN 30°C C 18 column 60% ACN 30°C C 18 column 40% ACN 70°C Cyano column 40% ACN 30°C (a) (b) (c) (d ) C≡N (1) OCH 3 (2) CH 3 (3) 1 2 3 3 2 1 1 2 3 0 2 4 6 8 10 12 14 (min) 0 2 4 6 8 10 12 14 (min) 0 2 4 6 8 10 12 14 (min) 0246810 12 14 (min) Figure 6.1 Retention in RPC as a function of temperature and the polarity of the solute, mobile phase and column. Sample: as indicated in figure. Conditions: (a-c) 150 × 4.0-mm 5-μm) Symmetry C 18 column, and (d) 150 × 4.6-mm (5-μm) Zorbax StableBond cyano column; 2.0 mL/min; mobile phase is acetonitrile/water, with mobile-phase composi- tion (%B) and temperature indicated in figure (bolded values represent changes from [a]). Chromatograms recreated from data of [8, 9]. decreases retention; compare Figure 6.1c (70 ◦ C) and Figure 6.1a (30 ◦ C). Finally, a decrease in column hydrophobicity weakens the interaction between the solute and column, and reduces retention; compare Figure 6.1d (more-polar cyano column) and Figure 6.1a (less-polar C 18 column). 6.2.1 Solvent Strength As noted in Section 2.4.1, retention in RPC varies with mobile phase %B as log k = log k w − Sφ (6.1) where k w refers to the (extrapolated) value of k for 0% B (water as mobile phase), S is a constant for a given solute when only %B is varied, and φ is the volume-fraction of organic solvent B in the mobile phase (φ ≡ 0.01% B). The value of %B selected 258 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES 80% ACN (k = 0.3) R s = 0.0 (a) (b) (c) 1 2 3 4 1 2 3 4 0.2 0.4 0.6 0.8 Time (min) t 0 1−4 50% ACN R s = 1.0 (1.9 ≤ k ≤ 4.2) 40% ACN R s = 1.4 (4 ≤ k ≤ 10) 0 4 Time (min) 2 6 0 2 Time (min) Figure 6.2 Separation of a mixture of four nitro-substituted benzenes as a function of solvent strength (%B). Sample: 1, nitrobenzene; 2, 4-nitrotoluene; 3, 3-nitrotoluene; 4, 2-nitro-1,3-xylene. Conditions: 100 × 4.6-mm (3-μm) Zorbax C8 column; mobile phase con- sists of acetonitrile-water mixtures (varying %B); 35 ◦ C; 2 mL/min. Chromatograms recreated from data of [10]. for the final separation should provide values of k for the sample that are within a desired range (e.g., 1 ≤ k ≤ 10), while at the same time maximizing solvent-strength selectivity (Section 6.3.1). A suitable value of %B can be obtained as described in Section 2.5.1: start with 80% B, and then reduce %B in steps until the desired retention range of 1 ≤ k ≤ 10 is obtained. Figure 6.2 provides an example of this approach. The first separation with 80% B (Fig. 6.2a) provides very little retention (k = 0.3), so a change to 50% B is tried for the next experiment (Fig. 6.2b). The retention range for the sample is now reasonable (1.9 ≤ k ≤ 4.2), but the resolution is inadequate (R s = 1.0). A further decrease in %B will usually (but not always) increase resolution; for samples with molecular weights < 500 Da, a 10%B decrease will increase values of k by a factor of about 2.5, which suggests a mobile phase of 40% B for the next experiment (Fig. 6.2c). Resolution is increased moderately (R s = 1.4) but is still inadequate. As a further decrease in %B will result in values of k > 10, some other means of further increasing resolution may be necessary: a change in selectivity (Section 6.3) or a change in column conditions (Section 2.5.3). It should be noted that Equation (6.1) is not an exact relationship but an approximation. For example, values of log k for a representative solute (4-nitrotoluene; compound-2 in Fig. 6.2) are plotted against %B in Figure 6.3 6.2 RETENTION 259 2.4 2.0 1.6 1.2 0.8 0.4 0.0 log k 020 6080 %B 40 MeOH ACN best linear fit to ACN data Figure 6.3 Variation of log k with %B. Sample is 4-nitrotoluene. Conditions: 250 × 4.6-mm (5-μm) Zorbax C8 column; mobile phase consists of organic/water mixtures; 35 ◦ C; 2 mL/min. Created from data taken from [10]. for both acetonitrile (ACN, ) and methanol (MeOH, • ) as the B-solvent. Whereas Equation (6.1) predicts a linear plot, a slightly curved plot results for ACN as B-solvent. The data for MeOH fall closer to the linear curve in Figure 6.3 that is fitted to these data. This behavior is typical of other samples and experimental conditions [11]; more linear plots are usually obtained for MeOH, compared to the use of ACN or other organics as B-solvent. However, over the usual range in k that is of interest (e.g., 1 ≤ k ≤ 10), Equation (6.1) is adequately accurate for either MeOH or ACN as the B-solvent. 6.2.2 Reversed-Phase Retention Process The following discussion provides further insight into the basis of RPC retention. However, it is mainly of academic interest, with little practical value for the application of RPC. A results-oriented reader may prefer to skip to following Section 6.3. The nature of RPC retention (the retention ‘‘mechanism’’) has been the subject of a large number of research studies, as summarized in several reviews and research publications [5, 12–16]. Some of the questions that this work has addressed include: • positioning of the solute molecule within the stationary phase (adsorption or partition?) 260 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES • dependence of solute retention on mobile-phase composition (k vs. φ) • conformation of the alkyl ligands that form the stationary phase (‘‘extended’’ vs. ‘‘collapsed’’) The positioning of the solute molecule within the stationary phase might occur in any of the ways pictured in Figures 6.4a–c for a C 8 column. Solvophobic interaction (Fig. 6.4a) assumes that the solute molecule aligns with and is attached to a ligand group (C 8 in this example). Adsorption (Fig. 6.4b)impliesthatthe solute molecule does not penetrate into the stationary phase, but is retained at the interface between the stationary and mobile phases. Partition (Fig. 6.4c) considers the stationary phase to be similar to a liquid phase, into which the solute molecule dissolves. Notice that the stationary phase consists of alkyl ligands plus an organic solvent that is preferentially extracted from the mobile phase by the C 8 groups of the stationary phase. In both solvophobic interaction and partition, the solute molecule lies within the stationary phase. When discussing the mechanism of RPC retention, Horv ´ ath’s solvophobic- interaction model [5] is commonly cited: (relatively) hydrophobic solute molecules prefer to adhere to the hydrophobic alkyl ligands—so-called hydrophobic retention. Soon after the introduction of RPC for HPLC, it was observed that RPC retention (values of k) correlates with partition coefficients P for the distribution of the solute between octanol and water (Fig. 6.4d; [17]); this suggests that a partition process best describes RPC retention. However, later studies showed that correlations of log P versus log k, as in Figure 6.4d (for amino acids) are less pronounced when the sample consists of molecules with more diverse structures, which makes the latter argument on behalf of partition less compelling. In the early 1980s [18] a surprising observation was made for the RPC retention of various homologous series (CH 3 –[CH 2 ] n − 1 –X), where X represents a functional group such as –OH or –CO 2 CH 3 . Plots of log k versus carbon-number n were found to exhibit a discontinuity for a value of n that is approximately equal to the carbon-number n  for the stationary-phase ligand (e.g., n  = 8for C 8 ≡ CH 3 –[CH 2 ] 7 –). This anomalous behavior is illustrated in Figure 6.5a by a hypothetical plot of log k versus n for a homologous series and a C 8 column; a discontinuity in the expected linear plot (dashed line) is observed (arrow) when n equals 8 for the solute (CH 3 –[CH 2 ] 7 –X). It was concluded from this observation that the contribution to retention for successive –CH 2 -groups in the solute becomes slightly smaller when the length of the solute molecule just exceeds the length of the alkyl ligand. The reason for this discontinuity in the plot of Figure 6.5a is visualized in Figure 6.5b,wheren = 12 for solute ii exceeds the value of n  = 8 for the column ligand (the situation for n = n  = 8 is illustrated by solute i). In the example of Figure 6.5b with n = 12, the end of the molecule likely folds back onto or into the stationary phase—rather than extending into the mobile phase as shown. Presumably there is a decreased interaction with the column for solute molecules that are too long to penetrate fully into the stationary phase (or attach to a single column ligand), with a corresponding decrease in the retention of –CH 2 -groups that ‘‘stick out of’’ the stationary phase. The contribution to retention of each –CH 2 -group can be defined as α CH2 = ratio of k values for successive homologs, and this value is normally assumed constant (see discussion of Eq. 2.34). However, for 6.2 RETENTION 261 (a) (c) (b) (d ) solvophobic interaction adsorption partition 0 −1 −2 −3 −4 log P −2 −101 lo g k y = x − 2 sorbed mobile phase solute molecule Figure 6.4 Different possibilities for the retention of a solute molecule in reversed-phase chromatography. (a) Solvophobic interaction; (b) adsorption; (c) partition; (d) comparison of RPC retention (k) with octanol-water partition P [17]; sample; eight amino acids; column: C 8 ; mobile phase: aqueous buffer (pH-6.7); 70 ◦ C. excluded –CH 2 -groups, the value of α CH2 should be somewhat smaller. This ‘‘exclu- sion’’ effect is demonstrated experimentally in Figure 6.5c for a C 18 column, where α CH2 is plotted versus n. There is a distinct break in the plot in the vicinity of n = n  = 18 (dashed vertical line). Similar breaks are shown in the plots of Figures 6.5d,e for the C 8 and C 6 columns, respectively, but no significant penetration of the solute is possible for the C 1 column of Figure 6.5f . The data of Figures 6.5c–e suggest that small solute molecules can penetrate the stationary phase of a C 6 ,C 8 , or C 18 column, so this would rule out a purely adsorption process (as in Fig. 6.4b) for other small solutes. Figure 6.5 also supports the solvophobic-interaction model of Figure 6.4a. Figure 6.5 is suggestive of possible conclusions concerning retention in RPC, but more complicated arguments have been offered concerning the adsorption and partition processes [14, 15]. It should be noted that the nature of the stationary phase (and presumably the retention process) varies with experimental conditions. Increasing amounts of the B-solvent (e.g., acetonitrile) are taken up by the stationary 262 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES (b)(a) log k n n = n’ = 8 ii i X X C 18 5 10 15 20 25 1.40 1.38 1.36 α CH 2 n (c) (d ) (e) (f ) 1.30 1.28 1.26 1.24 5 10 15 20 25 n C 8 α CH 2 1.28 1.24 1.20 C 6 510152025 n α CH 2 C 1 1.21 1.17 1.13 510152025 n α CH 2 Figure 6.5 Retention as a function of alkyl chain length (for both the solute and column). (a) Illustrative plot of log k versus number of –CH 2 -groups n for a homologous series CH 3 –(CH 2 ) n−1 –X;C 8 column; (b) illustration of the ‘‘overlapping’’ of alkyl chains in the solute and column; (c–f) plots of experimental methylene selectivity α CH2 versus carbon num- ber n c for indicated columns of differing ligand length. Average data for several homologous series; 90% methanol-water as mobile phase; 25 ◦ C. Figures are adapted from [18]. 6.3 SELECTIVITY 263 phase as %B increases. Likewise some solutes may interact with underivatized silanols present on the particle surface (Section 5.4.1). For these and other reasons the precise nature of the retention process is likely to vary with the column, the solute, and experimental conditions. Horv ´ ath anticipated this situation early on [17], in noting that ‘‘a clear distinction between partition and adsorption in RPC of nonpolar [solutes] [and] with no apparent thermodynamic or practical significance [so that] this issue may not be worth further investigation.’’ The authors of this book find it difficult to argue with this conclusion. Speculation concerning the nature of RPC retention has also been based on retention as a function of mobile-phase composition (%B). While Equation (6.1) is a purely empirical relationship, several theory-based equations for k as a function of φ have been derived [11]. The resulting expressions for k versus φ are in some cases slightly more reliable than Equation (6.1)[19], largely because of additional fitting parameters. However, Equation (6.1) is generally adequate for practical application, and is much more convenient to use. The major assumptions involved in all previous theoretical derivations of k versus φ negate any value in their use for interpreting the nature of RPC retention. Stationary phase ligand conformation has been claimed to play a role in the ‘‘mechanism’’ of RPC retention. The use of mobile phases that are predominantly aqueous (φ ≈ 0) can lead to greatly reduced sample retention—the opposite of that predicted by Equation (6.1). When first observed, this reduced retention was attributed to ‘‘phase collapse,’’ whereby alkyl ligands clump together and tend toward a horizontal rather than vertical alignment with the particle surface. This retention behavior was subsequently shown to arise not from phase collapse but rather from exclusion of mobile phase and sample molecules from particle pores as a result of surface-tension effects (‘‘de-wetting,’’ Section 5.3.2.3 and [2, 21]). More recent studies [22, 22a] suggest that ligand conformation does not change as a function of either mobile-phase composition or the relative coverage of the particle surface by ligands. 6.3 SELECTIVITY The most effective way to improve the resolution (or speed) of a chromatographic separation is to initiate a change in relative retention (selectivity). For the separation of non-ionic samples by RPC, changes in selectivity can be achieved by a change in solvent strength (%B), temperature, solvent type (e.g., ACN vs. MeOH as the organic solvent), or column type (e.g., C 18 vs. cyano). The relative effectiveness of a change in conditions for a change in selectivity varies roughly as temperature (least effective) < %B < solvent type ≈ column type (most effective) However, each of the four conditions above for changing selectivity can be useful for different samples or separation goals, as discussed next. 6.3.1 Solvent-Strength Selectivity In the examples of Figures 6.1 and 6.2, relative retention does not change when solvent strength (%B) is varied. As %B decreases (and k increases), the resolution 264 REVERSED-PHASE CHROMATOGRAPHY FOR NEUTRAL SAMPLES 0 2 Time (min) 1 2 + 3 4 5 6 7 8 50% ACN 2 ≤ k ≤ 6 R s = 0.0 (a) 45% ACN 3 ≤ k ≤ 6 R s = 1.5 (c) 024 Time ( min ) 1 2 3 4 5 6 7 8 4 40% ACN 4 ≤ k ≤ 14 R s = 1.0 (b) 02468 Time (min) 1 2 5 6 7 8 3 Figure 6.6 Separation of a moderately irregular sample (mixture of eight nitro-aromatic compounds) as a function of solvent strength (%B). Sample: 1, nitrobenzene; 2, 2,6-dinitrobenzene; 3, benzene (shaded peak); 4, 2-nitrotoluene; 5, 3-nitrotoluene; 6, toluene; 7, 2-nitro-1,3-xylene; 8, 1,3-xylene. Conditions: 100 × 4.6-mm (3-μm) Zorbax C8 column; mobile phase consists of acetonitrile/water mixtures; 35 ◦ C; 2 mL/min. Chromatograms recre- ated from data of [10]. of all peaks improves, but their relative spacing stays essentially the same. In Section 2.5.2.1 we defined samples as in Figures 6.1 and 6.2 as regular. Figure 6.6 shows the separation of a sample where %B is varied, but relative retention does not remain the same. An initial separation with 50% B (Fig. 6.6a) shows a complete overlap of peaks 2 and 3 (shaded). When the mobile phase is changed to 40% B (Fig. 6.6b), peak 3 moves toward peak 4 and partially overlaps it. From these two experiments it can be seen that an intermediate value of %B should result in an improved separation, which is observed for the separation of Figure 6.6c (45% B). Samples such as this, where relative retention changes with solvent strength, are referred to as irregular. Regular samples are often composed of structurally similar molecules; for example, in the separation of Figure 6.2 the sample is a mixture of mono-nitro alkylbenzenes. The sample of Figure 6.6, on the other hand, exhibits a somewhat greater molecular diversity: it is a mixture of alkylbenzenes that contain 0, 1, or 2 nitro-substituents. Changes in %B often lead to significant changes in relative retention, with maximum resolution occurring for an intermediate value of %B. Despite this obvious fact, practical workers often overlook solvent-strength selectivity as a useful tool 6.3 SELECTIVITY 265 for optimizing relative retention and resolution. To take maximum advantage of solvent-strength selectivity, the allowable retention range can be expanded from 1 ≤ k ≤ 10 to 0.5 ≤ k ≤ 20. When conditions are varied so as to change selectivity, it is important to keep track of which peak is which. The numbering of each peak in the chromatogram (as in Fig. 6.6) may not be obvious; in such cases peak tracking will be required (Section 2.6.4). The remainder of this section, which expands on the discussion above of regular and irregular samples, is somewhat detailed; the reader may prefer to skip (or skim) this discussion and go on to Section 6.3.2. The dependence of retention on %B for regular as opposed to irregular samples is further illustrated in Figure 6.7. Figure 6.7a shows plots of log k against %B for a regular sample; a mixture of nine herbicides of similar molecular structure (see Fig. 2.6 for the separation of several of these compounds as a function of %B). The slope of each plot increases for more retained solutes (in the order 1 < 2 < 3 ). For plots of log k versus %B for regular samples, this results in a characteristic, ‘‘fan-like’’ appearance—with no intersection of one plot by another (no change in relative retention). Values of α and resolution increase continuously for regular samples as %B is decreased. Another way to describe the behavior of regular samples is in terms of Equation (6.1). For regular samples, the slopes S of plots as in Figure 6.7a are highly correlated with extrapolated values of log k for water as mobile phase (log k w ). Figure 6.7b shows such a plot for the data of Figure 6.7a; an excellent correlation is noted (r 2 = 1.00). Corresponding correlations of log k w versus S for the regular samples of Figures 6.1a,b and 6.2 give r 2 = 0.99 for each. A similar treatment as in Figures 6.7a,b for a regular sample is shown in Figures 6.7c,d for the irregular sample of Figure 6.6. In Figure 6.7c, plots of log k versus %B frequently intersect (marked by • ), unlike the behavior of the regular sample in Figure 6.7a. As a result several peak reversals occur over this range in %B (e.g., peaks 3–4). Similarly a plot of S versus log k w for the irregular sample in Figure 6.7d shows a somewhat poorer correlation (r 2 = 0.87). Samples that contain acidic and/or basic compounds (unlike the examples of Fig. 6.7) tend to be more irregular; However, most samples—whether ‘‘ionic’’ or ‘‘neutral’’—exhibit some degree of irregularity and solvent-strength selectivity can be a useful tool for improving the resolution of such samples. 6.3.2 Solvent-Type Selectivity Changes in %B may fail to achieve adequate resolution of a given sample. An illustration is shown in Figure 6.8 for the separation of a mixture of substituted benzenes. In this example, a change in mobile phase from 46 to 34% ACN results in some change in relative retention (e.g., peaks 2–3, 8–9), but that has no significant effect on the separation of overlapped peaks 3 and 4. Consequently some other change in conditions that affect selectivity will be necessary—for example, a change from ACN to MeOH as the B-solvent. When 61% MeOH is used in place of 46% ACN (Fig. 6.9a), solvent strength is about the same (similar run times), but several further changes in relative retention are seen due to solvent-type selectivity. Although previously unresolved peaks 3 and 4 are now well separated, peaks 1 and 2 (which were well resolved with ACN as B-solvent) are overlapped. With results such as this for two mobile phases with different B-solvents, a mixture of the two . coat the particles, allowing their use for reversed-phase liquid liquid partition chromatography [2], a HPLC technique that is rarely used today. Bonded-phase columns represented the next major. polyoctadecylsiloxane to silica particles [3, 4]. This was followed a few years later by the introduction of columns where individual C 18 groups (rather than a C 18 polymer) were attached to the particle. the interface between the stationary and mobile phases. Partition (Fig. 6.4c) considers the stationary phase to be similar to a liquid phase, into which the solute molecule dissolves. Notice that