236 THE COLUMN of a column-comparison function F s [61]: F s ={[12.5(H 2 − H 1 )] 2 + [100(S ∗ 2 − S ∗ 1 )] 2 + [30(A 2 − A 1 )] 2 + [143(B 2 − B 1 )] 2 + [83(C 2 − C 1 )] 2 } 0.5 (5.4) Values of H 1 and H 2 refertovaluesofH for columns 1 and 2, and similarly for the remaining column parameters S*, A,etc.AvalueofF s ≤ 3 indicates that the two columns are similar in selectivity and can be substituted for each other in any RPC separation (for less challenging separations, larger values of F s can be tolerated). The ability of Equation (5.4) to identify columns of equivalent selectivity has been demonstrated for a dozen different routine RPC separations that were developed and used in several different laboratories [68] (an example is given in Fig. 6.19). Software for the comparison of column selectivity by means of Equation (5.4), and values of H, S*, etc., for more than 400 RPC columns can be accessed at the US Pharmacopeia website (http://www.usp.org/USPNF/columnsDB.html). Alternatively, contact one of the authors for a current database of values of H, S*, etc., for different RPC columns. 5.4.3 Orthogonal Separation Just as it is sometimes necessary to identify two different columns of similar selectiv- ity, at other times we need a second column of different selectivity (usually combined with a change in mobile phase). Several different reasons for a change in selectivity can be identified. First, when developing a reversed-phase separation, a large change in separation selectivity may be needed in order to improve the resolution of certain peaks in the chromatogram. Second, during method development for samples of ini- tially unknown composition, there may be a concern that a minor component might be overlapped by a larger peak in the chromatogram—and therefore missed in the final analysis. Third, a similar situation may arise in the use of a routine procedure for future samples that might contain additional, unanticipated impurities. In either of the latter two cases it is desirable to have available an ‘‘orthogonal’’ separation, for which selectivity is different from that provided by the original assay procedure; if two peaks overlap in the routine separation, they are then more likely to be separated (and observable) in the orthogonal separation. The use of Equation (5.4) for identifying columns of different selectivity in the latter two applications has been demonstrated for a dozen different routine assay procedures [69] (see the example of Fig. 6.21). Fourth, orthogonal columns are required for the technique of thermally tuned tandem-column optimization [70], in which two columns connected in series are operated at different temperatures in order to better control selectivity. Finally, orthogonal columns may be required for two-dimensional separation, where two different columns are used sequentially for the separation of a given sample (Section 9.3.10). In all these applications two RPC columns of very different selectivity correspond to two columns with a large value of F s in Equation (5.4). It has been suggested [71] that the selection of an orthogonal column can be further improved by emphasizing the separation of neutral solutes (for which values of C are unimportant in determining column selectivity). For separations of neutral 5.4 COLUMN SELECTIVITY 237 solutes, the corresponding column selectivity function will be F s (−C) ={[12.5(H 2 − H 1 )] 2 + [100(S ∗ 2 − S ∗ 1 )] 2 + [30(A 2 − A 1 )] 2 + [143(B 2 − B 1 )] 2 (5.4a) For maximum column orthogonality, the simultaneous conditions F s (−C) ≥ 50 and F s ≥ 100 have been recommended [71]. 5.4.4 Other Applications of Column Selectivity Values of H, S*, etc., reflect the nature and properties of the stationary phase, which are also related to certain common problems associated with the column: • peak tailing • stationary-phase ‘‘de-wetting’’ • column degradation during routine use 5.4.4.1 Peak Tailing Tailing peaks were discussed in Section 2.3.2, primarily from the standpoint of their effect on resolution. Peak tailing in RPC occurs mainly for protonated basic solutes [72] and type-A alkylsilica columns [73]. Thus the use of type-B columns (values of C[2.8] ≤ 0.25) largely solves this problem. Additional means for reducing peak tailing for basic solutes are discussed in Section 7.3.4.2. Even for type-B columns, however, fully ionized compounds tend to tail whenever the sample size exceeds about 1 μg for a 4.6-mm-diameter column. The reason is the mutual repulsion of ionized molecules (of the same charge) when concentrated into the stationary phase (Section 15.3.2.1 and [72]). Non-ionized solutes do not tail until a 50-fold larger sample weight is injected. Tailing peaks may also occur (less frequently) when carboxylic acids are separated with low-pH mobile phases [73]. Such peak tailing is more likely for type-A columns but can also occur for type-B columns that are more basic (i.e., have larger values of B). Opposite to the case of protonated bases, non-ionized acids tend to tail for sample weights <1 μg but give symmetrical peaks for sample weights of 1to50μg. For further details, see [73]. 5.4.4.2 Stationary-Phase De-Wetting Column de-wetting (Section 5.3.2.3) is more likely for narrow-pore, more hydropho- bic columns; that is, columns with larger values of H. As seen in Table 5.8, the average value of H = 0.99 for a narrow-pore (≤ 12 nm), type-B C 18 column (the most popular column). Columns with lower values of H will be less likely to experience column de-wetting. Polar-end-capped columns (average H = 0.90) are specifically designed to minimize column de-wetting but are otherwise similar to type-B C 18 columns in terms of selectivity (because the effect of H on selectivity is less important, as can be seen from the weighting factors in Eq. 5.4). 238 THE COLUMN 5.4.4.3 Column Degradation When a column is used for routine analysis, the stationary phase is gradually lost due to attack by the mobile phase on either the silane–silica bond (at low pH) or the silica itself (at high pH). Either process results in an increase in the number of silanols (therefore increased values of the column-selectivity parameters A and C) and a decrease in column hydrophobicity (values of H) [73]. As a result column selectivity can be expected to gradually change with further use of the column. Usually a column is discarded and replaced by a new column when selectivity changes (or the plate number drops) to the point of peak overlap. 5.5 COLUMN HARDWARE 5.5.1 Column Fittings Stainless-steel columns with standard end fittings (Section 3.4.2) are available with a wide choice of column dimensions and different packings. The greatest efficiency, reproducibility, and ruggedness are found for columns of this type. Less-costly stainless-steel cartridge columns are also available with a wide range of packings, but these are mostly used for less-demanding routine assays. Cartridge columns are not supplied with end fittings, so they require reusable holders. Pressures as high as 15,000 psi (≈1000 bar or ≈100 MPa) are now achievable with some HPLC instruments. Columns should therefore be constructed to (1) withstand such pressures and (2) resist chemical attack by the mobile phase. To meet these goals, stainless-steel tubing is used for most columns. In rare cases where stainless steel might be attacked by the mobile phase or react with the sample, titanium or glass-lined steel columns are available. Fused-silica capillaries can be used with mass spectrometric detection, and heavy-wall glass tubing for pressures as high as 600 psi have been used for preparative separations. Polymer-based PEEK columns with very heavy walls are available for use with pressures up to 6000 psi. In all cases the inside walls of the column tubing blank should be smooth with a mirror finish; the condition of the column wall greatly influences the homogeneity of the packed bed and therefore the ultimate efficiency of the column. End fittings and connectors for the column must be designed to have a minimum dead-volume, in order not to contribute significantly to extra-column peak broadening (Sections 2.4.1.1, 3.9). Hardware such as compression fittings, tubing, connectors, detector cells, and so forth, should be assembled so as to minimize unswept corners or stagnant pools that can act as mixing vessels and broaden peaks. Metal-to-metal inlet seals are used for columns where input pressures exceed 6000 psi; some seals allow inlet pressures as high as 15,000 psi. For lower inlet pressures (e.g., <6000 psi), inlet compression fittings with seals composed of organic polymers such as PEEK can be used successfully. See Section 3.4 for further details. Introduction of the sample to the column as a sharp, minimum-volume ‘‘plug’’ is required for best results. Therefore porous frits or screens used at the column ends to retain the packing must minimize extra-column peak broadening. Porous stainless-steel, titanium, or Hastelloy frits about 0.2 mm thick are used most often. Thin, stainless-steel screens provide the least peak broadening, but they are more 5.5 COLUMN HARDWARE 239 difficult to use—especially for higher inlet pressures. The porosity of porous frits or screens must be substantially smaller than the size of the packing particles. This is especially true for a wider particle-size-distribution, where the smallest particles may be able to leak through or plug the frit. For example, 2-μm porosity frits are generally adequate for columns of 5-μm particles, or for 3-μm particles that have a narrow particle-size distribution. Most columns of 3-μm particles use 0.5-μmfrits, whereas sub-2-μm-particle columns use 0.2-μm outlet frits to retain the packing material in the column. Because 0.2- and 0.5-μm frits are susceptible to plugging by particulates, samples and mobile phases often require careful filtration when used with small-particle columns. Straight (rather than curved or coiled) columns are almost always used. Experience has shown that glass columns are rarely needed, even for separating sensitive biological samples such as some peptides and proteins. While columns of glass-lined stainless-steel and aluminum-clad rigid polymer (PEEK) columns are available, it is rare that such materials are required. Radial-compression columns (pp. 67–69 of [12]) consist of particles that are loosely packed into a soft, polymeric cylinder. Prior to use, hydraulic pressure is applied to the exterior of the column, so as to squeeze the particles together and form a compact bed. These columns are available for both analytical and preparative applications, with the advantages of lower cost and metal-free construction. Today these columns are not widely used because of awkward temperature control and other inconveniences, as well as added cost when connecting columns in series (for an increase in N). Axial-compression columns are packed loosely and then compressed by means of a close-fitting piston that enters one end of the column. These columns are used exclusively for large-scale preparative separations [74, 75]. 5.5.2 Column Configurations A wide range of column sizes are available, depending on the intended application. Table 5.9 summarizes some column dimensions that are commercially available for various types of column packings. Routine methods usually are developed with 3- to 4.6-mm i.d. columns, equipped with compression fitting and packed with 2.5- to 5-μm particles. Such columns represent a good compromise among convenience, efficient performance and adequate column lifetime. Columns of 2.1-mm i.d. often Table 5.9 Typical Column Configurations Type Inner Diameter (mm) Length (mm) Particle Size (μm) Analytical 1–4.6 30–250 1.5–10 Cartridge 3–4.6 50– 100 3–10 Microbore 1, 2.1 50–250 2–8 Semi-preparative 8–10 100– 250 5–200 Preparative 20–50 100– 250 50–200 Note: Stainless-steel columns; glass, glass-lined, plastic and PEEK are also available in some configura- tions. 240 THE COLUMN are used when interfacing with mass spectrometry, as the mobile-phase flow rates used with these columns are more compatible with the requirements of this detector (Section 4.14). Columns of ≤2.1-mm i.d. often exhibit 15–25% smaller plate numbers, because of (1) equipment extra-column peak broadening, and (2) difficulty in packing narrow-diameter columns. Automatic sample injectors can be a significant source of extra-column peak broadening when using narrow-bore columns, because of the required very small peak volumes. Columns of 1-mm i.d. and capillary columns as small as 50-μm i.d. sometimes are used with mass spectrometric detection, but they are generally not suited for routine application. Narrow-bore columns are more difficult to pack efficiently, and special equipment is needed when using these very small internal-diameter columns in order to minimize extra-column peak broadening. 5.6 COLUMN-PACKING METHODS For most readers, the following section will have little application to laboratory practice, and can therefore be bypassed. Today most chromatographers obtain columns from a commercial source, and have no need to pack their own columns. Commercial columns have been under development for several decades. Dupli- cating their efficiency, reliability, and reproducibility is beyond the capability of most laboratories that are responsible for HPLC method development and analysis. This chapter is not intended as a manual for the preparation of HPLC columns. Rather, we will review some principles for the packing of HPLC columns, so that the reader can at least appreciate how good columns are made. For those who might wish to pack their own columns for special applications, see [76, 77] for further details. 5.6.1 Dry-Packing Early columns for HPLC were packed with irregular particles in the 45- to 50-μm size range, using vibration procedures that had been used previously for gas chro- matography. With the introduction in the late 1960s of dense, ≈25-μm spherical particles with solid cores (both pellicular and superficially porous packings), a ‘‘tap-fill’’ dry-pack method could be used to produce efficient, stable columns [77]. The ‘‘tap-fill’’ method is still useful for the preparation of columns with larger parti- cles, especially for preparative chromatography (Chapter 15). With the subsequent introduction of much smaller (≤10-μm) particles for HPLC, dry-packing procedures were found to yield inefficient, unstable columns. Other column-packing approaches were therefore required. 5.6.2 Slurry-Packing of Rigid Particles Early HPLC particles had a relatively wide particle-size distribution, and in some cases, an irregular shape. Such particles apparently tend to segregate across the column cross-section according to size, leading to an inefficient and unstable column bed. Methods were therefore required to minimize this problem. Packing columns with small, rigid particles is best done by filling the column blank under pressure with a slurry of particles in some liquid [26, 76, 77]. Three different variations of 5.6 COLUMN-PACKING METHODS 241 this approach can be used, in each case making use of conditions that minimize the size-separation of particles during column-packing. • balanced-density liquids • high-viscosity liquids • low-viscosity unbalanced slurry To suspend the packing, the balanced-density procedure uses a slurry liquid (e.g., mixtures of tetrabromoethane with a less-dense solvent) whose density is similar to that of the liquid-filled particle. The high-viscosity procedure makes use of a high-viscosity slurrying liquid (e.g., glycerin). Both of these approaches impede the settling and sizing of particles during the packing process. However, both methods require more time and generally do not produce the higher column efficiency of the following method. The subsequent availability of porous silica microspheres (≤5 μm) with a relatively narrow particle-size distribution led to the low-viscosity unbalanced-slurry method. Because particle sizing during column packing is not so critical for very small particles, a balanced-density or high-viscosity slurry is no longer needed. Low-viscosity liquids such as tetrahydrofuran, chloroform, or mixtures of these with other low-viscosity liquids can be used, allowing the rapid production of packed columns that are both efficient and stable. 5.6.2.1 Selection of Slurry Liquid Packing columns by means of any of the slurry procedures is relatively simple: slurry the particles with a liquid, place the slurry in a reservoir, attach this reservoir to a high pressure pump, then force the slurry into an empty column blank (with the outlet frit and end-fitting in place) by starting the pump (Fig. 5.24). During this Solvent reservoir Pump Slurry reservoir Column Drain End-fitting & frit Figure 5.24 Schematic of equipment for packing columns by the slurry procedure. Adapted from [76]. 242 THE COLUMN process particles are held at the column outlet by means of a porous frit or screen, while the slurry solvent passes through. When the column blank is full, the column is removed from the packing apparatus and a second porous frit or screen is placed at the inlet of the column. Despite the simplicity of this description, many important issues must be addressed for the production of efficient, stable, and reproducible columns [26, 76]. A critical factor is the need for suspension of the particles in the slurry liquid without aggregation. Thus the selection of the slurry liquid is critical. The suitability of a liquid for avoiding particle aggregation usually can be determined by means of the following, simple test. Particles are added to the slurry liquid, followed by inserting the mixture into an ultrasonic bath. The mixture is then examined with an ordinary microscope. If the particles are freely dispersed and nonaggregated, as illustrated by the cartoon of Figure 5.25a, the liquid may be acceptable. However, if the particles tend to aggregate as in Figure 5.25b, the liquid is likely a poor choice. Even when the particles are properly dispersed, a liquid may not be optimum for a particular column packing; the only real test is the performance of the resulting column. The selection of a low-viscosity slurry liquid depends on the nature of the pack- ing particles. To prevent particle aggregation, the interaction of the particle with the slurry liquid should be stronger than interactions between particles. For example, C 8 -orC 18 -modified particles (which contain polar, unreacted silanol groups) should be packed with a slightly polar liquid such as tetrahydrofuran, methyl-t-butyl ether—or mixtures such as acetonitrile/chloroform, chloroform/methanol, or chlo- roform/acetone (the use of hydrocarbons such as hexane usually is not successful). More polar column packings (unmodified silica or silica that is bonded with polar ligands such as amino or diol) require methanol or some other polar liquid for the slurry. When packing capillary columns with small particles, it is not clear whether particle nonaggregation in the slurry is required for good columns; further study is needed. However, for conventional columns with internal diameters ≥1 mm, a nonaggregating slurry liquid is strongly recommended. Once a nonaggregating slurry liquid has been chosen, some further aspects of slurry packing should be considered. Packing particle-size. Particles ≥3 μm in diameter are relatively easy to pack into efficient, stable beds. Resulting columns should exhibit minimum reduced plate heights h of 2 to 2.5, and values of the tailing factor <1.2 (for small, nonpolar (a) Dispersed (b) Aggregated Figure 5.25 Comparison of dispersed and aggregated particles (different slurry-packing solvents). 5.6 COLUMN-PACKING METHODS 243 solutes). Values of h ≤ 1.5 have been reported for superficially porous particles of 2.7-μm diameter [5, 13]. Particles with sizes of ≤2 μm are more difficult to pack, and less efficient columns may result. Particle-size distribution. Many commercial packings possess relatively narrow particle-size distributions, with standard deviations of 15–20% from the average. Superficially porous, 2.7-μm particles have been described with a standard deviation of only 5%, and this may be in part responsible for the higher efficiency of these columns (Section 5.2.2.1). Slurry concentration. A particle concentration of 7–15% usually works best. The exact concentration of particles in the slurry liquid (for best results) depends on particle type, the slurry liquid, and the packing apparatus—and must be determined empirically. Slurry-apparatus design. The design and configuration of the reservoir to deliver the packing into the column blank can be important. Best results are obtained when the packing is delivered though tubes that maintain the same internal diameter from the bottom outlet of the reservoir into the column blank inlet (as in Fig. 5.24). Constant-pressure pumps (e.g., Haskell pumps) are favored over constant-flow pumps for this operation. Packing pressure. The pressure required for a dense column bed depends on the size of the particle, column length, and internal diameter. For stable columns, the packing pressure should be ≥50% higher than the maximum pressure at which the column will be used, and the delivery of slurry into the column blank must be a fast as possible. Surface finish of the column-blank wall. The walls of the column expand or ‘‘balloon’’ slightly during the pressure loading process, and the particles of packing are also compressed [76]. After the column has been filled and the pressure released, the particles decompress and some are forced to move along the column wall. Consequently the inside wall of the column should have a smooth, mirror finish, in order to avoid abrading particles and creating fines that can result in a lower column efficiency. Particle strength. Particles are compressed during the pressurization process and deform elastically; some of this compression will compensate for internal stresses encountered during use of the column at a higher pressure or temper- ature. If particles do not have sufficient strength, they will eventually fracture or crush, resulting in an inefficient column with a higher than normal pressure. Some columns are rated at a maximum pressure of 9000 to 15,000 psi; the particles in such columns must be strong enough to maintain their structure during slurry packing at even higher pressures. 5.6.2.2 Rigid Polymeric Particles The balanced-density slurry-packing procedure is also used for preparing columns of hard, polymeric particles (‘‘gels’’) such as cross-linked polystyrene. The same hardware and general procedure is used as for rigid particles. In contrast to rigid particles, hard gels must first be allowed to swell in the liquid in which they are to be packed. As organic gels have a lower density than rigid solids, lower density slurring liquids are used (e.g., acetone/perchlororethylene mixtures). Because 244 THE COLUMN hard-gel particles are not as strong as rigid particles, the packing pressures must be lower. Polymeric particles (‘‘resins’’) for ion-exchange also can be slurry-packed by the balanced-density method, using aqueous liquids. The recommended procedure is as follows: A thick slurry of the swollen, ion-exchange resin is prepared in a salt solution such as calcium chloride, whose density can be matched to that of the resin. The slurry is then forced by high pressure into the column blank, using the same apparatus and technique as for rigid particles. The packing pressure depends on the strength of the resin particles, which is generally a function of the degree of cross-linking of the resin. A pressure of ∼5000 psi can be used for the strongest ion-exchange resins. Packed columns must be carefully flushed with the mobile phase to ensure removal of the salt before use. 5.6.3 Soft Gels Soft-gel particles such as those used for gel filtration (size exclusion, Section 13.8) cannot be dry-packed, nor can the high-pressure slurry-packing method be used (soft column packings compress and deform at relatively low pressures). Therefore columns of soft gels usually are packed using a gravity-based, slurry-sedimentation method. It is best to follow the procedure recommended by the manufacturer. 5.7 COLUMN SPECIFICATIONS 5.7.1 Manufacturing Standards As a means of ensuring reproducible column performance, manufacturers set stan- dards or specifications for each column (defined by model number) that they produce—however, there is little uniformity in these standards from one manu- facturer to another. Many manufacturers supply a written specification for each column, including target values for various column characteristics and actual per- formance data. Other manufacturers assume that data for one or more columns from a manufacturing lot or batch will be representative for all columns in that lot. Data reported for the column can vary from one manufacturer to the next, or even between product lines for a single manufacturer. It is important to discriminate between data that are specific to an individual column, and data that depend on properties of the bulk packing (a batch test). Column-specific data are the plate number N, peak asymmetry A s (or tailing factor), and column pressure P. A test chromatogram for each individual column should report values of N, A s ,andP, together with the detailed test conditions, so that users can confirm these results on their own equipment. However, values of N, A s , and P determined by the user can differ somewhat because of differences in HPLC equipment. Information from the batch test includes physicochemical data (e.g., surface area, pore size, pore volume, particle size, carbon content, μmoles/m 2 of ligand) and/or a batch-test chromatogram. Physicochemical data are not of immediate use to most chromatographers, but these data may be helpful for future troubleshooting, and should therefore be kept on file. The test solutes used for the batch-test chromatogram should include compounds of varied functionality (e.g., acids, bases, 5.7 COLUMN SPECIFICATIONS 245 neutrals) whose retention can indicate undesirable changes in batch-to-batch column selectivity (Section 5.4.2). Retention data for the test solutes (values of k and α) should fall within a narrow range of values (usually specified by the manufacturer). It is important to keep these data and chromatograms on file, for use if problems arise with a routine procedure. Not all manufacturers provide all of the data for each column. Sometimes, only average values for each column model number are reported, and this information may be found in different places: the column insert, the manufacturer’s website, company literature, or scientific publications. Some suppliers warrant their columns against defects for a period of use so that the user can be assured of a certain column lifetime. Additional useful data can also be found in care-and-use manuals, brochures, or websites; e.g., recommended operating conditions such as pH and temperature ranges, as well as solvents that are not recommended for a particular column. 5.7.2 Column Plate Number The value of N reported is usually for separation conditions that are close to ’’ideal’’ (low-viscosity mobile phase, a small, neutral solute molecule, near-optimum flow rate). This value of N will often differ from that found for other solutes and/or operating conditions, for reasons described in Section 2.4.1. For columns of totally porous particles, the following equation can be used to estimate the plate number for a well-packed column and conditions that have been optimized for maximum N: N ≈ 500L d p (5.5) Here the column length L is in mm, and the particle diameter d p is in μm. Table 5.10 shows typical plate-number values (neutral solute molecules with molecular weights of ≈200 Da) for well-packed HPLC columns of various lengths, particle sizes and types. The values in Table 5.10 assume a column diameter of 4.6 mm; a column diameter ≤2 mm can result in values of N that are lower, possibly because of the less-efficient packing of small-diameter columns, but mainly because of extra-column peak broadening. The conditions used to measure N for a given column are usually specified by the manufacturer so that the user can confirm this value of N when poor column performance is suspected. If the measured plate number is less than 80% of the reported value, if the performance of the HPLC system has been verified, and if the column has not been mistreated or used extensively, the column should be returned to the supplier. However, it is unusual for a column not to meet the manufacturer’s specifications. Lower plate numbers and higher asymmetry values measured by users for a new column are almost always the result of a system that has excessive extra-column peak broadening—which will be larger for small-diameter and/or short columns, especially those with particles smaller than 5 μm. When carrying out a routine HPLC analysis over a period of time, it is desirable to dedicate one or more columns to this assay. It can also be useful to keep track of values of N andpeakasymmetry(A s or TF) for one or more of the sample compounds, during routine analysis. These values of N and peak asymmetry can then be compared with values originally determined for the method, when using a . consist of particles that are loosely packed into a soft, polymeric cylinder. Prior to use, hydraulic pressure is applied to the exterior of the column, so as to squeeze the particles together. pack- ing particles. To prevent particle aggregation, the interaction of the particle with the slurry liquid should be stronger than interactions between particles. For example, C 8 -orC 18 -modified particles. as illustrated by the cartoon of Figure 5.25a, the liquid may be acceptable. However, if the particles tend to aggregate as in Figure 5.25b, the liquid is likely a poor choice. Even when the particles are