226 THE COLUMN C ≡ N OOHO OH O OH O OH O OH O CH 3 C 1 C 3 C 5 C 8 C 18 C 30 (TMS) (ODS) (a) Alkylsilica columns X OH O OH OH O OH OOHO (b) (c) (d) Phenyl columns Cyano column “Other”columns Column type X Sub-type Embedded-polar-group R-NH-(C=O)-O- Carbamate R-NH-(C=O)-NH- Urea R-(C=O)-NH- Amide Fluoro columns Perfluorophenyl (PFP) -CF 2 CF 2 CF 3 Fluoroalkyl FF FF F Figure 5.19 RPC columns classified according to the ligand (figures omit the connecting silane group [–Si(CH 3 ) 2 –]). simplified cartoons of Figure 5.19 (the—Si[CH 3 ] group is omitted in Fig. 5.19a–d). The ligand of a RPC column is often an alkyl group, for example, C 3 ,C 8 ,C 18 (Fig. 5.19a). Alternatively, the ligand may consist of phenylpropyl or phenylhexyl, called phenyl columns (Fig. 5.19b). If the ligand is –C 3 –C≡N (Fig. 5.19c), we have a cyano column. The alkyl group may also be substituted by other functional groups X (Fig. 5.19d), and this gives rise to the additional column types listed at the bottom of Figure 5.19. So-called embedded-polar-group (EPG) phases have been growing in popularity, because of their compatibility with low %B mobile phases, their reduced silanol interactions, and unique selectivity (Section 5.4.1); peak shape for basic solutes is usually quite good with these columns. The ligands in these phases contain amide, carbamate, urea (all of which are strong hydrogen-bond bases), or other polar functional groups embedded within the ligand structure. Some EPG 5.4 COLUMN SELECTIVITY 227 packings tend to be less stable than comparable alkyl or aryl columns. The nature of the ligand mainly determines column selectivity, which is the subject of following Section 5.4. 5.4 COLUMN SELECTIVITY Column selectivity can be important for different reasons. During method devel- opment a change of column may be necessary to improve selectivity and increase resolution (Sections 2.5.2, 5.4.3). For the latter application we must be able to iden- tify a second column with quite different selectivity. When a routine RPC procedure is used at different times and places, a replacement column from the same source may not be immediately available locally, or too costly, or impractical for other reasons (Sections 5.4.2, 6.3.6.1). In this case we must identify a column of equivalent (or at least similar) selectivity. For either situation, we require a quantitative procedure that allows us to compare column selectivity. Column selectivity is also related to certain problems that can arise during either method development or the routine use of an RPC procedure: peak tailing, the deterioration of a column during use, and ‘‘de-wetting’’ of the column when used with mobile phases that are predominantly aqueous (Section 5.3.2.3). Finally, knowledge concerning column selectivity helps us understand sample retention as a function of the column and solute molecular structure, in turn preparing us to better deal with various separation challenges. In the remainder of Section 5.4, we will first discuss the basis of column selectivity, which can be attributed to different interactions between solute molecules and the column. This will lead to quantitative values of those properties of a column that determine its selectivity. Finally, we will discuss the use of these column-selectivity properties for both method development and the routine use of an RPC procedure. 5.4.1 Basis of RPC Column Selectivity As discussed in Section 2.3.2.1, solute retention is determined by various interactions among the solute, mobile phase, and stationary phase (column). The relative impor- tance of different solute–column interactions—and column selectivity—depends on the composition of the stationary phase and the molecular structure of the solute. Figure 5.20 illustrates eight different interactions that can affect column selectivity: (a) hydrophobic interaction (b) steric exclusion of larger solute molecules from the stationary phase (here referred to as ‘‘steric interaction’’) (c) hydrogen bonding of an acceptor (basic) solute group by a donor (acidic) group within the stationary phase (usually a silanol –SiOH) (d) hydrogen bonding of a donor (acidic) solute group by an acceptor (basic) group within the stationary phase (represented here by a group ‘‘X’’) (e) cation-exchange or electrostatic interaction between a cationic solute and an ionized silanol (–SiO − ) within the stationary phase; also repulsion of an ionized acid (e.g., R–COO − ) 228 THE COLUMN (f) dipole–dipole interaction between a dipolar solute group (a nitro group in this example) and a dipolar group in the stationary phase (a nitrile group for a cyano column) (g, h) π –π interaction between an aromatic solute and either a phenyl group (phenyl column) (g), or a nitrile group (cyano column) (h) (a) Hydrophobic interaction OO COCH 3 Hydrogen bonding (acidic solute) (d) OO X O C OH (e) Cation exchange OO − O NH + (g) π-π interaction (phenyl column) O O 2 N NO 2 (h) π-π interaction (cyano column) O C = N O 2 N NO 2 (c) Hydrogen bonding (basic solute) OHOO N (b) Steric exclusion OOO (f) Dipole-dipole interaction O C = N + − O 2 N − + Figure 5.20 Solute-column interactions that determine column selectivity (figures omit the connecting silane group [–Si(CH 3 ) 2 –]). 5.4 COLUMN SELECTIVITY 229 O O M ++ N N Complexation of chelating solutes (i ) Figure 5.20 (Continued) (i) complexation between a chelating solute and metal contaminants on the particle surface Interactions (a–e) can be significant for every column; dipole interactions (f )are only important in the case of cyano columns, and π –π interactions (g, h) occur only for phenyl and cyano columns [57]. Both dipole and π –π interactions are inhibited by the use of acetonitrile as B-solvent, which further minimizes their importance for separations with acetonitrile. Complexation with surface metals (i) can result from the use of a less pure, type-A silica, leading to broad, tailing peaks (very undesirable); the chelating solute α,α-bipyridyl has been used to test columns for metal complexation. Because phenyl and cyano columns are used less often, and type-A columns are not recommended, we will emphasize interactions (a–e)inthis chapter (but see [57, 58]). 5.4.1.1 Solute–Column Interactions The various solute–column interactions of Figure 5.20, which determine column selectivity, have been understood in general terms since the 1980s; see [59] for a good discussion of recent attempts at characterizing column selectivity. However, only after 2000 did it become possible to reliably characterize RPC column selectivity in terms of these interactions. This was accomplished by the development and application of the hydrophobic-subtraction model [60–62], which recognizes that hydrophobic interactions are by far the most important contribution to RPC retention. If only hydrophobic interactions were significant, a plot of values of log k for one column against another would give a straight line with no scatter of data around the line. As seen in Figure 5.21, this is approximately the case for these two C 18 columns (Inertsil ODS-3 and Stablebond C18)—however, values of log k for aliphatic amides () and protonated strong bases ()fallbelow the best fit to these data. These latter deviations are due to interactions of these solute molecules with silanol groups (silanol interactions are more significant for the StableBond C18 column). These and other smaller deviations δ log k from this plot (see the expanded inset of Fig. 5.21) represent contributions to retention from nonhydrophobic interactions b–e of Figure 5.20. It is possible to analyze values of δ log k for the combination of different solutes and columns so as to separately evaluate the five interactions of Figure 5.20a–e. For columns other than phenyl or cyano (i.e., those for which only interactions a–e of Fig. 5.20 are significant), values 230 THE COLUMN δlog k 2.0 1.5 1.0 0.5 0.0 −0.5 −1.0 −1.5 −1.0 −0.5 0.0 1.00.5 1.5 log k (Inertsil ODS-3) lo g k (StableBond C18) y = 0.21 + 1.01 x r 2 = 0.995 amides protonated bases Figure 5.21 Comparison of retention on two different C 18 columns. Data for 90 different organic compounds. Conditions: 15 × 4.6-mm columns; 50% acetonitrile-water, pH-2.8 phosphate buffer; 2.0 mL/min; 35 ◦ C. Adapted from [61]. of k can be related to the interactive properties of the solute and the column: log k k EB =η H − σ S ∗ + β A + α B + κ C (5.3) (i)(ii)(iii)(iv)(v) Here k and k EB are values of the retention factor for a given solute and the reference compound ethylbenzene (EB), respectively. Terms i–v of Equation (5.3) correspond, respectively, to the interactions of Figure 5.20a–e. Quantities η , σ , β , α ,andκ refer to properties of the solute molecule: hydrophobicity (η ), ‘‘bulkiness’’ (σ ), hydrogen-bond (H-B) basicity (β ), H-B acidity (α ), and effective ionic charge (κ ). Corresponding column parameters are of primary practical interest: H, hydrophobicity; S*, steric interaction, or resistance by the stationary phase to penetration by bulky solutes; A, H-B acidity; B, H-B basicity; and C, ion-exchange capacity or electrostatic (coulombic) interaction. Columns with similar values of H, S*, etc., will possess similar selectivity and provide a similar retention order for peaks within the chromatogram. Columns with different values of H, S*, etc., will differ in selectivity and provide changes in relative retention. The hydrophobic-subtraction model and Equation (5.3) best summarize our present understanding of RPC retention and column selectivity [59]. We will next relate terms i–v of Equation (5.3) to the interactions of Figure 5.20a–e. Hydrophobic interaction is illustrated in Figure 5.20a,bythe 5.4 COLUMN SELECTIVITY 231 interaction of the solute 2-n-octanone (CH 3 COC 6 H 13 ) with the ligand groups of aC 8 column. Values of column hydrophobicity H increase for longer ligands (e.g., C 18 vs. C 8 ), a higher concentration of ligand groups on the silica surface (μmoles/m 2 ), smaller pore diameters (e.g., 8- vs. 30-nm pores), and the presence of column end-capping. An increase in H increases the retention of more hydrophobic molecules, those with larger values of η . Steric exclusion or ‘‘steric interaction’’ is illustrated in Figure 5.20b by the retention of two polycyclic aromatic hydrocarbon (PAH) isomers: the narrow, long naphthacene and the more ‘‘bulky’’ triphenylene. Naphthacene is better able to squeeze between adjacent ligands, but if the spacing of column ligands is increased (lower ligand concentration), it becomes easier for the bulky triphenylene to enter the stationary phase. The column parameter S* measures the ‘‘tightness’’ of the stationary phase or the difficulty that bulky solute molecules experience in squeezing between the ligands; larger values of S* mean a ‘‘tighter’’ stationary phase and relatively less retention of bulky solute molecules. Values of S* increase for longer ligands, a higher concentration of the ligand (ligands closer together), and smaller pore diameters. Solute bulkiness is measured by its value of σ . Steric exclusion is a somewhat complex phenomenon; see Section 5.4.1.2 below for further insights. Hydrogen bonding of a non-ionized basic solute (e.g., pyridine) by a column silanol is illustrated in Figure 5.20c. The hydrogen-bond acidity A of the column is due to the presence of surface silanols, and therefore decreases when the column is end-capped (due to the removal of some silanols and blocking of others; see the example of Fig. 5.16d). The silanols of type-A columns are usually more acidic than those present in type-B columns; therefore values of A tend to be larger for type-A columns. The H-B basicity of the solute is measured by its value of α ; unprotonated amines and amides are more basic and have larger values of α , while nitriles and nitro compounds are much less basic and have smaller values of α . Most other polar compounds have intermediate H-B basicities and values of α . Hydrogen bonding of a solute that is a H-B acid (e.g., butyric acid) is illustrated in Figure 5.20d. The H-B basic group ‘‘X:’’ in the stationary phase is not specified because the nature of X differs for different kinds of RPC columns. For type-B alkylsilica columns, there is so far no compelling explanation for what groups ‘‘X:’’ consist of. It has been suggested that water dissolved into the stationary phase corresponds to these groups ‘‘X:’’, on the basis of an inverse correlation of values of A and H—as well as other arguments. In the case of some type-A alkylsilica columns with large values of B, contaminating metals in the silica appear to comprise the ‘‘X:’’ groups. Alternatively, for embedded-polar-group columns, the polar group (which is usually a H-B base) very likely corresponds to the ‘‘X:’’ group. An increase in B leads to increased retention of carboxylic acids, which have large values of α ; the retention of other H-B acids, such as alcohols and phenols, is less dependent on B (smaller values of α )—although phenols are preferentially retained on embedded-polar-group columns (i.e., phenols appear as stronger H-B acids when these columns are used). Column ion-exchange capacity C is a measure of the ionization of the silica and the accessibility of ionized silanols. Silanol ionization and/or accessibility (and values of C) increase (1) as mobile-phase pH increases, (2) for non–end-capped columns, and (3) for type-A versus type-B columns. While the main effect of larger values of C is to increase the retention of protonated bases, it also results in a 232 THE COLUMN decrease in retention for ionized acids (because of electrostatic repulsion); the value of κ for a solute is approximately equal to its molecular charge (e.g., +1 for fully protonated bases, −1 for fully ionized acids). The main difference in selectivity for type-A versus type-B columns is determined by their low-pH values of C;type-B columns have values of C < 0.25 at pH 2.8, while type-A columns have C > 0.25. For columns with values of C < 0.00 at low pH, it is believed that these columns carry a net positive charge [63], presumably the result of protonated amine groups that are introduced during the manufacturing process for some columns. Values of H, S*, A,andB are assumed not to change with the pH of the mobile phase. Values of the column-selectivity parameters H, S*, etc., have been measured for over 400 different columns; see [64] for a partial listing, or for a current list of values contact one of the authors (or http://www.USP.org/USPNF/columnsDB.html). Average values of these column parameters are summarized at the top of Table 5.8 for several different kinds of RPC column. Within a given column type, there is also a significant variation in values of H, S*, etc., as illustrated at the bottom of Table 5.8 for several type-B C 18 columns. Consequently not all columns of a given kind can be regarded as equivalent in terms of selectivity. Apart from values of H and S*, for example, average retention as measured by values of k for ethylbenzene (last column of Table 5.8) increases with the surface area of the particle. 5.4.1.2 Shape Selectivity The following, minor digression examines two distinct forms of steric exclusion; for now, the reader may prefer to skip to Section 5.4.2. Two separate manifestations of steric exclusion have been described: steric interaction, as measured by term ii of Equation (5.3), and shape selectivity [65]. Differences between these two phenomena are illustrated in Figure 5.22 for the separation of two isomeric hydrocarbons on a polymeric column (Fig. 5.22a)and a monomeric column (Fig. 5.22b). The basis of shape selectivity is illustrated in Figure 5.22a for a ‘‘narrow’’ molecule i, a ‘‘wide’’ molecule j, and a polymeric alkylsilica column. The ‘‘wide’’ molecule j is excluded from part of the stationary phase because its minimum cross-section (double-headed arrow) exceeds the spacing between ligands (molecule j cannot ‘‘squeeze’’ between the ligands). In a monomeric column (Fig. 5.22b), the ligands are further apart, so as to allow access of both narrow and wide molecules (i and j) to the stationary phase. When a molecule has access to the stationary phase, steric exclusion affects retention in a different way; now the hydrodynamic diameter of the molecule becomes important, rather than its minimum cross-section. When the hydrodynamic diameter of the solute molecule is comparable in size to the spacing between ligands (as for molecule i), the retained molecule is restricted in its possible orientations within the stationary phase. This restriction of the solute molecule reduces its retention, in a similar way as for size-exclusion chromatography (Section 13.8.1). Note that shape selectivity and steric interaction lead to dissimilar effects on retention as a function of molecular shape; thus they clearly represent two different contributions to retention [61]. Because of the ‘‘either-or’’ nature of shape selectivity, it can result in relatively large changes in relative retention—whereas steric interaction has a smaller effect on RPC selectivity. Figures 5.22c, d illustrates the potential advantage of shape selectivity for the separation of polycyclic aromatic hydrocarbons (PAHs). The polymeric column 5.4 COLUMN SELECTIVITY 233 Table 5.8 Characterization of Column Selectivity by means of the Hydrophobic-Subtraction Model (Eq. 5.3) Column type HS*A BC(pH-2.8) C (pH-7.0) k EB Different column types C 1 (type-B) 0.41 −0.08 −0.08 0.02 0.04 0.66 1.2 C 3 (type-B) 0.60 −0.12 −0.08 0.04 −0.08 0.81 2.8 C 8 (type-B) 0.84 0.00 −0.12 0.02 −0.03 0.25 5.4 C 18 (type-B) 0.99 0.01 −0.01 0.00 0.00 0.24 8.8 C 18 (type-B, wide-pore) 0.95 0.01 −0.05 0.01 0.22 0.31 3.2 C 18 (type-B, monolith) 1.01 0.02 0.12 −0.02 0.11 0.31 3.2 C 18 (type-B, hybrid) 0.98 0.01 −0.14 −0.01 0.13 0.05 6.3 C 18 (polar end-capped) 0.90 −0.04 −0.02 0.02 −0.02 0.40 7.4 C 18 (type-A) 0.94 −0.05 0.14 0.01 0.79 1.18 6.4 C 30 (type-B) 1.05 −0.01 0.09 −0.02 −0.08 0.45 13.0 Embedded-polar-group 0.74 0.00 −0.22 0.12 −0.27 0.53 5.9 Phenyl (type-B) 0.63 −0.12 −0.20 0.02 0.13 0.68 2.7 Cyano (type-B) 0.43 −0.09 −0.49 0.00 0.02 0.72 1.0 Perfluorophenyl (PFP) 0.65 −0.11 −0.25 0.01 0.40 0.96 4.3 Fluoroalkyl 0.66 −0.07 −0.11 0.03 0.87 1.18 3.7 Zirconia-base 0.97 0.01 −0.62 0.00 2.01 2.01 0.8 Different narrow-pore, type-B C 18 columns a Halo-C18 b 1.11 0.05 0.01 −0.05 0.06 0.04 6.1 Zorbax StableBond C18 c 1.00 −0.03 0.26 0.00 0.14 1.04 7.6 Zorbax Eclipse XDB-C18 c 1.08 0.02 −0.06 −0.03 0.05 0.09 9.1 Kromasil 100–5C18 d 1.05 0.04 −0.07 −0.02 0.04 −0.06 12.5 ProntoSIL 120–5 C18 SH e 1.03 0.02 −0.11 −0.02 0.11 0.40 8.7 Inertsil ODS-3 f 0.99 0.02 −0.15 −0.02 −0.47 −0.33 10.9 Alltima C18 g 0.99 −0.01 0.04 −0.01 0.09 0.39 11.5 Nucleodur C18 Gravity h 1.06 0.04 −0.10 −0.02 −0.08 0.32 11.0 ACE 5 C18 i 1.00 0.03 −0.10 −0.01 0.14 0.10 7.9 Chromolith j 1.00 0.03 0.01 −0.01 0.10 0.19 3.1 Luna C18(2) k 1.00 0.02 −0.12 −0.01 −0.27 −0.17 9.6 Gemini C18 110A k 0.97 −0.01 0.03 0.01 −0.09 0.19 8.0 Discovery C18 l 0.98 0.03 −0.13 0.00 0.18 0.15 4.8 Hypurity C18 m 0.98 0.03 −0.09 0.00 0.19 0.17 5.6 Hypersil GOLD m 0.88 0.00 −0.02 0.04 0.16 0.48 3.9 Symmetry C18 n 1.05 0.06 0.02 −0.02 −0.30 0.12 9.8 Xterra MS C18 n 0.98 0.01 −0.14 −0.01 0.13 0.05 6.3 Sunfire C18 n 1.03 0.03 0.04 −0.01 −0.19 −0.10 9.9 TSKgel ODS-100Z o 1.03 0.02 −0.13 −0.03 −0.06 −0.16 11.6 a All columns 5-μm particles; data of [73]. b Advanced Materials Technology; c Agilent; d Akzo Nobel; e Bischoff; f GL Science; g Grace-Alltech; h Macherey Nagel; i ACT; j Merck; k Phenomenex; l Supelco; m Thermo/Hypersil; n Waters; o Tosoh Bioscience. 234 THE COLUMN (a)(b) (min) 0 10 20 30 4 ≥≥ k 16 (c) 10 15 ≥≥ 5 k 7 (d) Solute minimum cross-section (width) O OH OOOH Si Si-OH O OH OO Si Si-OH i j Solute hydrodynamic diameter O OH OH O OH OH O Si Si Si i j shape selectivity Polymeric column with closely spaced ligands steric interaction Monomeric column with widely spaced ligands steric interaction Monomeric C 18 column shape selectivity Polymeric C 18 column Figure 5.22 Different manifestations of steric exclusion. Shape selectivity (a) compared with steric interaction (b). (c) Separation of a mixture of 13 polycyclic aromatic hydrocarbons on a polymeric column. (d) Separation of same sample with same conditions on a monomeric column. (c)and(d) are adapted from [65]. for the separation of Figure 5.22c exhibits greater shape selectivity and therefore provides a much greater differentiation (and better separation) of these different isomeric C 22 PAHs, versus the corresponding separation in Figure 5.22d with a monomeric column (where shape selectivity is minimal). Long, narrow molecules (compared to those that are short and wide) are preferentially retained when shape selectivity is more important, while short, wide solute molecules (of similar molecular weight) are more retained when steric interaction is dominant. As a rule, we can say that shape selectivity is more important when C 30 or polymeric columns are used, and sample molecules are both large and have very different ratios of length to width. Most RPC separations are carried out with monomeric columns other than C 30 , in which case steric interaction and values of S* largely define the effect of steric 5.4 COLUMN SELECTIVITY 235 exclusion on column selectivity. For further details on the practical utility of shape selectivity, see Section 6.3.5.2. 5.4.2 Column Reproducibility and ‘‘Equivalent’’ Columns Column manufacturers try to ensure that each column (e.g., Waters Symmetry C18) has similar properties and will perform satisfactorily and reproducibly in a routine RPC assay. Consequently the plate number N and column pressure drop for each column usually is measured prior to its sale (Section 5.7); columns whose values of N fall below some minimum value are discarded. Similarly other tests are carried out by the manufacturer (Section 5.7) to ensure that column selectivity stays the same from one batch to the next of the column packing (similar to the measurements of values of H, S*, etc.). An example is shown in Figure 5.23 for several successive batches of Zorbax R Rx-C 18 , where the retention times for dimethylaniline and toluene are plotted against the batch number. Values of k for the two solutes vary by ±4% (1 SD), mainly as a result of small, unimportant differences in the surface areas of the silica particles. The ratio of these two k-values (α) is a more direct measure of column selectivity (primarily the important column-selectivity parameter C of Eq. 5.3); values of α vary by only ±0.5%. Consequently it appears that the selectivity of different batches of this column packing should be similar, especially for separations that involve protonated basic compounds. During the 1970s and 1980s HPLC column manufacturing had not yet devel- oped to its present advanced state, and column selectivity often varied significantly among different batches of a particular column. More recently column repro- ducibility has improved (e.g., see [67]), but it is still possible that one column batch will differ enough from another in terms of selectivity to result in a failed separation—especially for demanding separations. Various means exist for dealing with the latter problem (Section 6.3.6.1), one of which is to select an ‘‘equivalent’’ column from a different source. Using values of H, S*, etc., as in Table 5.8b,itis possible to identify one or more columns with similar values of H, S*,etc.,bymeans 5 4 3 2 1 α 0 k 1 3 5 7 9 11 13 Lot number CH 3 N(CH 3 ) 2 Figure 5.23 Monitoring different batches of column packing for possible changes in selectiv- ity. Sample: dimethylaniline and toluene. Conditions: 150 × 4.6-mm Zorbax Rx-C 18 columns; 50% acetonitrile-water plus pH-7 phosphate buffer; 1.6 mL/min; 22 ◦ C. Adapted from [66]. . 5.22b), the ligands are further apart, so as to allow access of both narrow and wide molecules (i and j) to the stationary phase. When a molecule has access to the stationary phase, steric exclusion. contributions to retention from nonhydrophobic interactions b–e of Figure 5.20. It is possible to analyze values of δ log k for the combination of different solutes and columns so as to separately evaluate. surface area of the particle. 5.4.1.2 Shape Selectivity The following, minor digression examines two distinct forms of steric exclusion; for now, the reader may prefer to skip to Section 5.4.2. Two