396 NORMAL-PHASE CHROMATOGRAPHY Time (min) 020 1 2 3 4 5 6 7 40% H 2 O/ACN 30% H 2 O/ACN 20% H 2 O/ACN (a) (b) (c) 10 30 02010 30 02010 30 Figure 8.18 Separation of a mixture of derivatized oligosaccharides by HILIC with mobile phases of varying %-water. Conditions: 200 × 4.6-mm PolyHydroxyethyl A column (5-μm particles); mobile phases are water-acetonitrile as indicated in the figure; 2 mL/min. Values of n in the figure for each peak refer to the number of saccharide units in the corresponding oligosaccharide. The chromatograms are recreated from the data of [38]. 8.6.1 Retention Mechanism Although the ‘‘retention mechanism’’ for HILIC has received considerable attention in the literature, this topic is mainly of academic interest. Inasmuch as the subject has very limited practical application, the reader may wish to skip this section. Retention in HILIC is believed to involve a partitioning of the solute into a water layer that is formed on the surface of the column packing [38, 39, 42, 44]. For silica as column packing and HILIC conditions, it has been shown [42] that the column dead volume V m continues to decrease as the mobile-phase water increases from 0 to 30%; this has been attributed to the buildup of a layer of water on the silica surface. The observation that HILIC separations require at least 2–3% water in the mobile phase confirms the importance in HILIC separation of this water layer, which presumably comprises the stationary phase. It is also possible that the stationary phase includes some organic solvent (acetonitrile) from the mobile phase, with further contributions to solute retention from both the silica surface (silanols) and any column ligands that might interact with solute molecules. A distinction can be made between adsorption on and partition into the stationary phase, as in the case of RPC (Section 6.2.2.1). It is believed by some investigators that these two processes can be distinguished by a comparison of values of k as a function of the volume-fraction (φ)oftheB-solvent(waterinthiscase). In RPC, which is believed to involve a partition process (in at least some cases), approximately linear plots of log k versus φ are observed (Fig. 6.3). In NPC with silica as column packing, where adsorption is believed to prevail, more linear plots 8.6 HYDROPHILIC INTERACTION CHROMATOGRAPHY (HILIC) 397 –1.4 –1.2 –0.8 –0.4 1.0 0.5 0.0 log k log f H 2 O n = 2 n = 4 n = 5 n = 8, 7 n = 4 n = 9 n = 16 n = 54 n = 8 % H 2 O = 20510 3040 Figure 8.19 Changes in HILIC retention of various peptides with mobile-phase strength (varying %-water). Conditions: 250 × 4.6-mm TSK gel Amide-80 column; 1.0 mL/min; 40 ◦ C. The sample peptides can each be characterized by the number n of amino-acid sub-units in the molecule (values of n shown). Adapted from [41]. of log k versus log φ are found (Fig. 8.4). When the relationship of k to φ for HILIC separations is examined, no clear-cut relationship is found for all systems, but generally plots of log k versus log φ tend to be more linear; see the example of Figure 8.19 for several peptides (values of n in Fig. 8.19 indicate the number of amino-acid residues in the solute molecule). While linear log k–log φ plots suggest that HILIC retention involves adsorption rather than partition, the HILIC stationary phase—and the various possible interactions of the solute with the water layer, silica surface, and any ligands attached to the silica—is surely quite complex. Any conclusions as to whether ‘‘adsorption’’ or ‘‘partition’’ predominates in HILIC are therefore premature, as well as of little practical value. 8.6.2 Columns Almost all present HILIC columns are created from silica particles. Aminopropyl column packings (‘‘amino’’ columns) were used initially, mainly for carbohydrate separations; these columns are well suited for this application, as they prevent the formation of double peaks for each solute as a result of anomer resolution. Subsequently a variety of different bonded-silica packings have been employed for HILIC [39, 40, 45], which can be categorized as follows: bare silica, polar neutral (e.g., cyanopropyl), diol-bonded, amide-bonded, polypeptide-bonded, positively charged amine-bonded (anion-exchange), negatively charged (cation-exchange), and zwitterionic phases. While the differing selectivity of these column packings [46] can 398 NORMAL-PHASE CHROMATOGRAPHY YMC Sil C-2.8 = –0.15 Nucleosil silica C-2.8 = 0.32 Zorbax SIL C-2.8 = 1.4 8 0 5 15 (min)10 Figure 8.20 Tailing of pyrimidines on some HILIC columns. Conditions: 250 × 4.6-mm silica columns (YMC SIL, Nucleosil silica, and Zorbax SIL); mobile phase: 75% acetoni- trile/buffer (5-mM phosphoric acid); 1.0 mL/min; ambient. Values of C-2.8 were measured for C 18 RPC columns from same source. Adapted from [48]. prove useful during method development, unbonded silica appears to be the packing of choice for many HILIC separations when mass spectrometric detection (LC-MS) is used [39, 43, 47]—because of an absence of stationary-phase bleed. However, irreversible retention of some compounds on silica has been reported, in which case an amide-bonded column—a commonly used alternative [40]—can be used. For an extensive list and discussion of columns for HILIC, with manufacturer websites, see [39]. An early publication [48] noted significant peak tailing for the HILIC separation of basic pyrimidines at low pH on bare silica (Fig. 8.20). It is seen in Figure 8.20 that peak tailing increases for columns with larger values of the column parameter C-2.8 for the corresponding C 18 -bonded silica (Section 5.4.1). Values of C-2.8 > 0.25 for C 18 columns are indicative of a less-pure, more acidic, type-A silica (Section 5.2.2.2), suggesting that peak tailing as in Figure 8.20 is associated with the use of type-A silica. As most silica columns used today are type-B, tailing as in Figure 8.20b,c should be less likely for HILIC separations with type-B silica columns. Irreversible binding of the solute should also be less likely when using type-B silica columns. 8.6.3 HILIC Method Development Several publications have reviewed the effects of different experimental conditions on HILIC separation [38, 39, 45, 46, 44]. Method development for the HILIC separation of a small-molecule sample can be carried out in similar fashion as for RPC or NPC. A 100–150 × 4.6-mm HILIC column (3−μm particles) is first selected; as noted above, bare silica is often preferred as column packing, with amide 8.6 HYDROPHILIC INTERACTION CHROMATOGRAPHY (HILIC) 399 packings as a good alternative. An initial separation can be carried out with a strong mobile phase; for example, 60% acetonitrile buffer (acetonitrile is by far the most commonly used organic solvent for HILIC). The %-acetonitrile can then be increased in 10% increments, until the desired retention range is obtained (e.g., 1 ≤ k ≤ 10); see the similar procedure of Section 2.4.1 for RPC, as well as the HILIC example of Fig. 8.18. Alternatively, an initial gradient run may be more effective (Section 9.3.1). Formic acid, ammonium acetate or ammonium formate are commonly used as HILIC buffers for the separation of ionizable compounds, in concentrations of 2 to 20 mM; mobile-phase pH usually falls within a range of 3 to 7. The latter buffers are volatile (for MS detection) and soluble in all mixtures of organic and water. Once a value of %-water has been selected for acceptable retention, conditions can be varied to optimize selectivity. Changes in relative retention may occur as a result of a change in %B, as can be inferred from the plots of Figure 8.19. Thus the retention sequence for the last five peptides in the chromatogram is n = 54 > 16 > 8 > 9 > 4 for 15% H 2 O and n = 8 > 9 > 4 > 54 = 16 for 25% H 2 O Changes in selectivity as %-water is varied have also been observed by others [46, 44, 49, 50]. While such changes in relative retention are often minor and therefore less useful, they are easily recognized during the adjustment of %-H 2 Oforthe purpose of achieving acceptable retention (e.g., 1 ≤ k ≤ 10). A few studies have demonstrated changes in HILIC selectivity as a function of temperature [44, 50], suggesting that the simultaneous optimization of both %B and temperature may be a promising approach (similar to the case for RPC; Section 7.3.2.2). The effect of mobile-phase pH on solute ionization is predictable, as in the example of Figure 7.1a; as pH increases, ionization increases for acids and decreases for bases. Since HILIC retention increases for more ionized solutes, the retention of acids will increase with an increase in pH, while that of bases will decrease. When the mobile-phase pH is within ±1 unit of solute pK a values, and solute pK a values vary (frequently the case for ionizable solutes), large changes in selectivity should result from a change in pH. A change in column type is also expected to result in significant changes in selectivity, especially considering the wide range in stationary phase functionality that is commercially available; for some examples, see [40, 51]. A change in the organic A-solvent (from acetonitrile) is also possible, but infrequently used for a change in selectivity; other solvents are generally stronger (Fig. 8.21), which can result in insufficient retention of the sample; acetonitrile usually also results in higher values of N [44]. The usual changes in column length and flow rate are also available as a means for increasing resolution or decreasing run time (Section 2.5.3). The flow rate will be limited for acceptable pressure (Section 2.4.1), and smaller diameter columns are normally used for mass spectrometric detection (Section 4.14). The efficiencies of HILIC columns appear comparable to or even better than those of RPC columns [40]; a nice example is shown in the two separations of Figure 8.22. The average plate number for the separation of Figure 8.22a with a 150-mm column of 2.7-μm, superficially porous silica particles is N = 36,000, while that for the 450-mm column 400 NORMAL-PHASE CHROMATOGRAPHY Methanol i-Propanol THF Acetonitrile 0 25 (min)2015105 1 2 3 4 Figure 8.21 Effect of B-solvent on HILIC separation. Sample: epirubicin and its ana- logues. Conditions: 250 × 4.6 Kromasil KR100-5SIL (silica) column. Mobile phase: 90% organic/pH-2.9 buffer; 1.0 mL/min; ambient. Adapted from [52]. 0 2.5 5.0 (min) 0 5 15 (min)10 1 2 3 4 5 6 7 8 (a) (b) 150-mm column 1400 psi 450-mm column 4200 psi Figure 8.22 High-efficiency HILIC separation. Sample: 1, phenol; 2, 2-naphthalenesulfonic acid; 3, p-xylenesulfonic acid; 4,caffeine;5, nortriptyline; 6, diphenhydramine; 7, benzy- lamine; 8, procainamide. Conditions:150 × 4.6-mm Halo silica column (Advanced Mate- rials Technology, Wilmington, DE) (a); three columns as in (a) connected in series (b); 75% acetonitrile–pH-3.0 buffer; 1.0 mL/min; 30 ◦ C. Adapted from [51]. in Figure 8.22b is N = 103,000. This seems rather remarkable, considering the short run times. A review of the recent literature indicates that reported HILIC separations are often carried out with gradient elution and mass spectrometric detection. The comments above for isocratic separation generally apply to gradient elution as well; see Section 9.5.3 for details on HILIC separations based on gradient elution. REFERENCES 401 8.6.4 HILIC Problems Problems with peak shape (both fronting and tailing) seem to be somewhat more common when using HILIC, than for RPC [46, 50, 54]; this may reflect the fact that HILIC was introduced more recently and has not been used as extensively as RPC, as well as the possible use of type-A silica for some column packings—especially columns introduced before 1995 (e.g., see Fig. 8.20 and the related discussion). Further improvements in HILIC columns combined with a better understanding of how to manage these peak-shape problems seem likely. When peak-shape problems are encountered, an increase in mobile-phase buffer concentration should be tried first. Some samples may require a buffer concentration as high as 100 mM in order to achieve acceptable peak shapes. The sample should be dissolved in the mobile phase, but in some cases an increase in %-organic of the sample solvent can be beneficial. If peak-shape problems persist, a change in column or mobile-phase pH may be advisable. The mobile phase should always contain some water, preferably at least 3–5%. Column bleed has been observed for some bonded-phase columns [41] when using mass-spectrometric detection. A change to a silica column or a different bonded-phase column usually solves the problem. Irreversible sorption of some sample components has been reported for silica columns when using HILIC. If the problem arises, a change to a bonded-phase HILIC column may be advisable. Slow equilibration of the column with mobile phases that contain a different buffer or buffer concentration may also occur, but problems of this kind are much less likely for HILIC than for NPC with a silica column. REFERENCES 1. L. R. Snyder, Principles of Adsorption Chromatography. The Separation of Nonionic Organic Compounds, Ml Dekker, New York, 1968, chs. 8 and 10. 2. L AA. Truedsson and B. E. F. Smith, J. Chromatogr., 214 (1981) 291. 3. L. R. Snyder and D. L. Saunders, in Chromatography in Petroleum Analysis,K.H. Altgelt and T. H. Gouw, eds., Dekker, New York, 1979, ch. 10. 4. A. Kuksis, in Chromatography, E. Heftmann, ed., 5th ed., Elsevier, Amsterdam, 1993, p. B171. 5. L. R. Snyder and H. Poppe, J. Chromatogr., 184 (1980) 363. 6. L. R. Snyder, in High-Performance Liquid Chromatography: Advances and Perspectives, Vol. 3. C. Horv ´ ath, ed., Academic Press, New York, 1983, p. 157. tab. 1 7. E. Soczewinski, Anal. Chem., 41 (1969) 179. 8. E. Soczewinski, J. Chromatogr. A, 965 (2002) 109. 9. L. R. Snyder, J. Chromatogr., 63 (1971) 15. 10. B. Fried and J. Sherma. Thin-Layer Chromatography (Chromatographic Science,Vol. 81), Ml Dekker, New York, 1999. 11. R. P. W. Scott and P. Kucera, J. Chromatogr., 112 (1975) 425. 12. V. R. Meyer and M. D. Palamareva, J. Chromatogr., 641 (1993) 391. 13. Solvent Guide, Burdick and Jackson Labs., Muskegon, MI, 1980. 14. E. Soczewinski, T. Dzido, and W. Golkiewicz, J. Chromatogr., 131 (1977) 408. 15. J. L. Glajch, J. J. Kirkland, and L. R. Snyder, J. Chromatogr., 238 (1982) 269. 402 NORMAL-PHASE CHROMATOGRAPHY 16. L. R. Snyder, J. Chromatogr., 63 (1971) 15. 17. L. R. Snyder, J. Planar Chromatogr., 21 (2008) 315. 18. L. R. Snyder, J. L. Glajch, and J. J. Kirkland, J. Chromatogr., 218 (1981) 299. 19. M. D. Palamareva and H. E. Palamareva, J. Chromatogr., 477 (1989) 235. 20. Ch. E. Palamareva and M. D. Palamareva, LSChrom, Ver. 2.1, Demo version, 1999, http://www.lschrom.com. 21. M. Z. Kagan, J. Chromatogr. A, 918 (2001) 293. 22. L. R. Snyder and J. J. Kirkland, Introduction to Modern Liquid Chromatography, 2nd edn., Wiley-Interscience, New York, 1979, pp. 391–392. 23. L. R. Snyder and T. C. Schunk, Anal. Chem., 54 (1982) 1764. 24. E. L. Weiser, A. W. Salotto, S. M. Flach, and L. R. Snyder, J. Chromatogr., 303 (1984) 1. 25. W. T. Cooper and P. L. Smith, J. Chromatogr., 355 (1986) 57. 26. A.W.Salotto,E.L.Weiser,K.P.Caffey,R.L.Carty,S.C.Racine,andL.R.Snyder,J. Chromatogr., 498 (1990) 55. 27. P. L. Smith and W. T. Cooper, J. Chromatogr., 410 (1987) 249. 28. L. P. Hammett, Physical Organic Chemistry, McGraw-Hill, New York, 1940, ch. 7. 29. B. Stancher and F. Zonta, J. Chromatogr., 234 (1982) 244. 30. M. D. Palamareva and H. E. Palamareva, J. Chromatogr., 477 (1989) 225. 31. J. L. Glajch and L. R. Snyder, J. Chromatogr., 214 (1981) 21. 32. B. Vaisman, A. Shikanov, and A. J. Domb, J. Chromatogr. A, 1064 (2005) 85. 33. L. R. Snyder, J. Chromatogr. Sci., 21 (1983) 65. 34. V. R. Meyer, J. Chromatogr. A, 768 (1997) 315. 35. P. Jandera, J. Chromatogr. A, 965 (2002) 239. 36. R. W. Stout, J. J. DeStefano, and L. R. Snyder, J. Chromatogr., 282 (1983) 263. 37. J. J. Kirkland, C. H. Dilks, and J. J. DeStefano, J. Chromatogr., 635 (1993) 19. 38. A. Alpert, J. Chromatogr., 499 (1990) 177. 39. P. Hemstrom and K. Irgum, J. Sep. Sci., 29 (2006) 1784. 40. T. Ikegami, K. Tomomatsu, J. Takubo, K. Horie, and N. Tanaka, J. Chromatogr. A, 1184 (2008) 474. 41. T. Yoshida, J.Chromatogr. A, 811 (1998) 61. 42. D. V. McCalley and U. D. Neue, J.Chromatogr. A, 1192 (2008) 225. 43. D. V. McCalley, J. Chromatogr. A, 1171 (2007) 46. 44. Z. Hao, B. Xiao, and N. Weng, J. Sep. Sci. 31 (2008) 1449. 45. B. Dejaegher, D. Mangelings, and Y. Vander Heyden, J. Sep. Sci., 31 (2008) 1438. 46. Y. Guo and S. Gaiki, J. Chromatogr. A, 1074 (2005) 71. 47. W. Naidong, J. Chromatogr. B, 796 (2003) 209. 48. B. A. Olsen, J. Chromatogr. A, 913 (2001) 113. 49. X. Wang, W. Li, and H. T. Rasmussen, J. Chromatogr. A, 1083 (2005) 58. 50. C. Dell’Aversano, P. Hess, and M. A. Quilliam, J. Chromatogr. A, 1081 (2005) 190. 51. D. V. McCalley, J. Chromatogr. A, 1171 (2007) 46. 52. R. Li and J. Huang, J. Chromatogr. A, 1041 (2004) 163. 53. D. V. McCalley, J. Chromatogr. A, 1193 (2008) 85. 54. C. Dell’Aversan, G. K. Eaglesham, and M. A. Quilliam, J. Chromatogr. A, 1028 (2004) 155. CHAPTER NINE GRADIENT ELUTION 9.1 INTRODUCTION, 404 9.1.1 Other Reasons for the Use of Gradient Elution, 406 9.1.2 Gradient Shape, 407 9.1.3 Similarity of Isocratic and Gradient Elution, 409 9.2 EXPERIMENTAL CONDITIONS AND THEIR EFFECTS ON SEPARATION, 412 9.2.1 Effects of a Change in Column Conditions, 415 9.2.2 Effects of Changes in the Gradient, 418 9.2.3 ‘‘Irregular Samples’’, 428 9.2.4 Quantitative Relationships, 430 9.3 METHOD DEVELOPMENT, 434 9.3.1 Initial Gradient Separation, 437 9.3.2 Optimize k ∗ , 442 9.3.3 Optimize Gradient Selectivity α ∗ , 442 9.3.4 Optimizing Gradient Range, 444 9.3.5 Segmented (Nonlinear) Gradients, 445 9.3.6 Optimizing the Column Plate Number N ∗ , 445 9.3.7 Determine Necessary Column-Equilibration Time, 446 9.3.8 Method Reproducibility, 449 9.3.9 Peak Capacity and Fast Separation, 451 9.3.10 Comprehensive Two-Dimensional HPLC (with Peter Schoenmakers), 457 9.4 LARGE-MOLECULE SEPARATIONS, 464 9.5 OTHER SEPARATION MODES, 465 9.5.1 Theory, 465 9.5.2 Normal-Phase Chromatography (NPC), 466 9.5.3 Hydrophilic Interaction Chromatography (HILIC), 467 9.5.4 Ion-Exchange Chromatography (IEC), 470 9.6 PROBLEMS, 470 Introduction to Modern Liquid Chromatography, Third Edition, by Lloyd R. Snyder, Joseph J. Kirkland, and John W. Dolan Copyright © 2010 John Wiley & Sons, Inc. 403 404 GRADIENT ELUTION 9.6.1 Solvent Demixing, 470 9.6.2 Ghost Peaks, 470 9.6.3 Baseline Drift, 470 9.1 INTRODUCTION Gradient elution was introduced in Section 2.7.2 as a means for dealing with samples that are unsuitable for isocratic elution. The most common reason for the use of gradient elution is a sample whose retention range exceeds the preferred goal for isocratic separation (1 ≤ k ≤ 10). As discussed in Section 2.5.1, it is possible to expand this retention range somewhat, for example, to 0.5 ≤ k ≤ 20. However, many samples cover a much wider range in k-values, making gradient elution essential for their separation. An example of a sample that cannot be separated successfully by isocratic elution is shown in Figure 9.1a. Here a mixture of 14 toxicology standards is injected into a C 18 column, using a mobile phase of 50% acetonitrile buffer. The excessive retention range for this sample (k-values that range 0–50) results in the poor resolution of early peaks 1 to 6, and excessive retention times for later peaks 13 and 14. Later peaks are also very broad and therefore not very tall—consequently their measurement may be compromised (poor signal/noise ratio; Section 4.2.3). No single change in %-acetonitrile (%B) would result in the adequate separation of the entire sample; larger values of %B would mean smaller values of k and even poorer resolution of early peaks in the chromatogram, while smaller values of %B would further increase values of k and run time—and make the measurement of later peaks still more difficult. Figure 9.1a is a good example of the general elution problem for samples with a wide range of k-values, which is the main reason for gradient elution. If adjacent groups of peaks from Figure 9.1a could be processed separately (with different values of %B), the improved isocratic separations of Figure 9.1b–f would result. Thus the use of 10% B as mobile phase for peaks 1 to 3 (Fig. 9.1b)resultsin an average value of k ≈ 3 for these peaks, and their baseline resolution. Similarly the separation of peaks 4 to 6 with a mobile phase of 25% B (Fig. 9.1c)alsoprovides an average value of k ≈ 3 with good resolution. Likewise the separations of peaks 7 to 8, 9 to 11, and 12 to 14 with 45, 62, and 75% B, respectively (Figs. 9.1d–f), result in k ≈ 3 for each group of peaks. Thus each of these sample-fractions can be separated isocratically with reasonable resolution (R s ≥ 2) and separation time (6–8 min), as well as providing narrower, taller peaks for improved detection. The only requirement is a mobile phase that provides k ≈ 3 for each group of peaks; however, as different values of %B are required for each set of peaks, isocratic separation of the total sample with 1 ≤ k ≤ 10 is not possible. Gradient elution (Fig. 9.1g) is a means of realizing the benefits shown in the isocratic separations of Figures 9.1b–f by means of a single run. Thus at the 9.1 INTRODUCTION 405 0 204060 Time (min) 1-3 7-8 9-10 12 13 14 isocratic, 50% B (0 ≤ k ≤ 50) (a) 4 5 + 6 11 2468 2 4 6 4 6 8 46846 Time (min) Time (min) Time (min) 10% B 25% B 45% B 52% B 75% B (b) (c) (d) (e) (f ) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Time (min) Time (r min) (g) gradient 5-90% B in 13 min (k* = 3) 0462 8 10 12 Time (min) 4-6 1-3 7-8 9-11 t 0 100% B 80% 60% 40% 20% 0% 12-14 Figure 9.1 Example of the general elution problem. Sample: 14 toxicology standards. Con- ditions: 250 × 4.6-mm (5-μm) C 18 column; mobile phase is ACN (B) and pH-2.5 phosphate buffer (A); 65 ◦ C; 2.0 mL/min. (a) isocratic separation with 50% B; (b − f), isocratic separa- tion of indicated compounds (peaks) with 10%, 25%, 45%, 52%, and 75% B, respectively (k ≈ 3); (c) gradient elution as indicated. Chromatograms are computer simulations based on the experimental data of [1]. beginning of the gradient, peaks 1 to 3 move through the column with an average value of k ≈ 3, while peaks 4 to 14 lag behind near the column inlet. As %B continues to increase during the gradient (indicated by dashed line, marking %B at column outlet vs. time), later peaks become less strongly retained and then also move through the column, again, with average values of k ≈ 3. As we will see in Section 9.1.3.2, values of k in gradient elution tend to be similar for all peaks in the chromatogram, and can be easily controlled by the choice of gradient time and flow rate. Unless stated otherwise, the present chapter refers to RPC separation; however, the same general principles and conclusions apply for other HPLC separations (ion-exchange, normal-phase, etc.). For a more comprehensive and detailed account of gradient elution than is presented in this book, see [2]. . Chromatography (HILIC), 467 9.5.4 Ion-Exchange Chromatography (IEC), 470 9.6 PROBLEMS, 470 Introduction to Modern Liquid Chromatography, Third Edition, by Lloyd R. Snyder, Joseph J. Kirkland, and. 1999, http://www.lschrom.com. 21. M. Z. Kagan, J. Chromatogr. A, 918 (2001) 293. 22. L. R. Snyder and J. J. Kirkland, Introduction to Modern Liquid Chromatography, 2nd edn., Wiley-Interscience, New York,. Chromatogr. A, 1064 (2005) 85. 33. L. R. Snyder, J. Chromatogr. Sci., 21 (1983) 65. 34. V. R. Meyer, J. Chromatogr. A, 768 (1997) 315. 35. P. Jandera, J. Chromatogr. A, 965 (2002) 239. 36. R. W. Stout,