Introduction to Modern Liquid Chromatography, Third Edition part 43 pps

To summarize, pronounced changes in relative retention with a change in %B solvent-strength selectivity are more likely for a sample that contains compounds substituted by different pola

The detection of sample bands in TLC as in Figure 8.8 normally requires the addition of a visualization agent to the plate For different means of band detection

in TLC, see [10]

Changes in retention order as a result of a change in separation conditions can be quite pronounced when NPC is used with a silica column—often much more so than in RPC This ability to exert a greater control over relative retention with NPC

is especially useful for preparative separations, where largeα values for a compound

to be purified allow much larger sample weights and a corresponding reduction in the effort and cost of sample purification (Section 15.3.2)

8.3.1 Solvent-Strength Selectivity

An example of solvent-strength selectivity can be seen in Figure 8.7 Although the relative retention of peaks 2, 3, 5, and 6 does not change much as %B is varied, peaks 1 and 4 (shaded) move toward the front of the chromatogram when %B is increased It is significant that molecules 2, 3, 5, and 6 are each substituted with the same polar group (–N=), whereas peak 1 is 2-aminonaphthalene (–NH2), and peak 4 is 4-nitrophenol (–NO2, –OH); that is, molecules of solutes 1 and 4 are

substituted by different polar groups The change in k with %B for peak 4 is greater

than for other compounds in this sample, and this can reasonably be attributed to the presence of two polar groups in the molecule (–NO2, –OH) compared with only a single polar group for the other compounds (–N= or –NH2); that is, n in Equation (8.4a) should equal 2 for peak 4 (vs n= 1 for the remaining peaks), and therefore its retention should change more for a given change in %B The somewhat different behavior of peak 1 compared with peaks 2, 3, 5, and 6 in this respect may

be due to its different functionality: a –NH2group for peak-1 as opposed to a –N= group for peaks 2, 3, 5, and 6

To summarize, pronounced changes in relative retention with a change in %B (solvent-strength selectivity) are more likely for a sample that contains compounds substituted by different polar groups, and especially for compounds with differing

numbers of polar groups in the solute molecule—and therefore different values of n

in Equation (8.4a)

8.3.2 Solvent-Type Selectivity

Before changing the B-solvent, %B should be varied so as to achieve 1≤ k ≤ 10, while at the same time maximizing resolution (as in Fig 8.7c) If further changes

in selectivity (α) and resolution are needed, the B-solvent can be changed while

maintaining the same solvent strength (same approximate value of ε in Fig 8.6)

for 1≤ k ≤ 10 A change of the (more-polar) B-solvent is usually the most effective

means for changing relative retention in NPC An example of solvent-type selectivity

is shown in Figure 8.9, for the separation of 12 naphthalene solutes that are substituted with different polar groups In each case the mobile-phase strength is approximately the same (ε = 0.25; see Fig 8.6 for the mobile phases of Figs 8.9a,b),

r 2 = 0.72 r 2 = 0.80 r 2 = 0.95

CH2Cl2vs ACN CH2Cl2vs MTBE

log k log k

log k

Time (min)

Time (min)

Time (min)

1 2

3

7 + 8

1 3

2 5

6 7 4

1 3

257

6104 +

(a)

(b)

(c)

53% CH2Cl2

Rs= 0.5

3.7% MTBE

Rs= 0.8

1.4% ACN

Rs= 0.3 10

ACN vs MTBE

Figure8.9 Example of solvent-type selectivity for normal-phase chromatography

Sam-ple: 1, 2-methoxynapthalene; 2, 1-nitronapthalene; 3, 1,2-dimethoxynapthalene; 4,

1,5-dinitronapthalene; 5, 1-naphthaldehyde; 6, methyl-1-naphthoate; 7, 2-naphthaldehyde;

8, 1-naphthylnitrile; 9, 1-hydroxynaphthalene; 10, 1-acetylnapthalene; 11, 2- acetylnaptha-lene; 12, 2-hydroxynaphthalene Conditions: 150 × 4.6-mm silica column (5-μm particles); mobile phases (%v) indicated in figure (50% water-saturated), except that (c) contains 6%

added CH2Cl2to achieve miscibility of ACN (hexane is the A-solvent in each case) 35◦C;

2 mL/min (a–c) Separations with indicated mobile phases; (d–f ) correlations of retention data from (a–c) Chromatograms recreated from data of [15].

as are the run times (7–10 min), so solvent-strength selectivity should be about the same Numerous changes in relative retention can be seen among these three separations, as a result of differences in solvent-type selectivity:

(53%CH2Cl2) 1< 2 < 3 < 4 < 5 < 6 < 7 = 8 < 9 < 10 < 11 < 12

(3.7%MTBE) 1 < 3 < 2 < 5 < 6 < 7 < 10 < 4 < 11 < 9 < 8 < 12

(1.4%ACN) 1 < 3 < 2 < 5 < 7 < 6 ≈ 10 < 4 = 11 < 8 < 9 < 12

Experimental studies [16] have shown that solvent-type selectivity in NPC depends mainly on the strength of the B-solvent (ε0

B) Asε0

Bincreases, the B-solvent

becomes more strongly attached to a specific silanol, resulting in localized adsorption

of the B-solvent Solvent fixation in this way is illustrated in Figure 8.3c by the arrow

that connects surface silanols with molecules of the B-solvent (THF) The latter (localized) interaction of silanol and solvent can be contrasted with the weaker, more diffuse interaction of silanols with a less-polar (nonlocalizing) solvent in

Figure 8.3a In the separations of Figure 8.9, the B-solvents are CH2Cl2 with

ε0

B= 0.30 (Fig 8.9a), methyl-t-butyl ether (MTBE) with ε0

B= 0.48 (Fig 8.9b),

and acetonitrile (ACN) withε0

B= 0.52 (Fig 8.9c); MTBE and ACN are localizing

solvents (as are mobile phases that contain these two solvents), while CH2Cl2is not

As the polarity of the solute increases, it too will become more strongly

attached to (or localized onto) one or more silanols, as in the example of phenol in Figure 8.3d Because the B-solvent and solute compete with each other for a place

on the silica surface, an increase in B-solvent localization will result in a relatively

greater reduction in k for solutes that are substituted by more-polar functional

groups (and are therefore more localized), compared to less-polar, less-localized groups That is, localized molecules of solute and B-solvent compete for the same positions (adsorption sites) on the silica surface Nonlocalizing solutes will be less affected by the localization of the mobile phase on individual silanols, resulting

in changes in relative retention (as in Fig 8.9) that are determined by the relative localization of different solute molecules See [6, 17] for a more detailed account of solvent and solute localization, and its effects on relative retention and solvent-type selectivity

Localizing B-solvents, such as ACN and MTBE, can exhibit smaller, but significant differences in selectivity For nine different localizing B-solvents [15–17] it was found that four of these solvents (nitromethane, ACN, acetone, and ethyl acetate) were very similar in terms of selectivity The remaining solvents (dimethylsulfoxide, triethylamine, THF, ethyl ether, and pyridine) were significantly different in this respect When these solvents are compared in terms of the solvent-selectivity triangle

of Figure 2.9, it is seen that the latter five solvents fall within the group labeled

‘‘basic solvents,’’ while the remaining four solvents are in the ‘‘dipolar solvents’’ (or nonbasic) group It therefore appears that solvents can be characterized in terms

of NPC selectivity as nonlocalizing (e.g., methylene chloride), basic localizing (e.g., MTBE), and nonbasic localizing (e.g., ethylacetate or ACN) Figure 2.9 allows other B-solvents to be classified as basic localizing or nonbasic localizing For example,

methyl-t-butylether is a commonly used (localizing) solvent in NPC; as seen in Figure

2.9, ethers such as MTBE are classified as basic

Relative retention is compared among the three B-solvents of Figure 8.9a–c in Figure 8.9d –f It is seen that the largest differences in relative retention (or change

in selectivity) occur for the nonlocalizing B-solvent methylene chloride rather than

for either the non-basic localizing ACN (Fig 8.9d) or the basic localizing MTBE (Fig 8.9e) That is, correlations for these plots are not very strong (0 72 ≤ r2≤ 0.80) When retention for the two localizing B-solvents is compared (Fig 8.9f ), relative retention is more similar for these two solvents (r2= 0.95), but still sufficiently different to result in useful differences in selectivity (as seen in Fig 8.9b vs Fig 8.9c).

Thus solvent localization is the main source of solvent-type selectivity in NPC, but

localizing solvents can be further differentiated as either basic or nonbasic The commonly used solvents for NPC in Table 8.1 are characterized as nonlocalizing, basic localizing, or nonbasic localizing When exploring solvent-type selectivity, a B-solvent of each type should be tried—as in Figure 8.9 Alcohols are very strong, proton-donor solvents; while they can be classified as localizing and basic, they should provide a moderate, further change in selectivity—especially for samples that contain strong proton acceptors

Once separations have been carried out with nonlocalizing, basic-localizing, and nonbasic-localizing B-solvents as in Figure 8.9 (with %B adjusted to give

1≤ k ≤ 10 for each mobile phase), two or more of these mobile phases can

be blended to achieve an intermediate selectivity and further increase resolution; Figure 8.10 shows the best achievable resolution for the sample of Figure 8.9, as a

result of blending the mobile phases from the separations of Figure 8.9b,c (to give

0.03% ACN/0.1% CH2Cl2/3.7% MTBE/96% hexane) A systematic selection of the best mobile-phase mixture can be carried out in a similar way as was described for optimizing solvent-type selectivity in RPC (Fig 6.24); see Section 8.4.2 and Figure 8.15 for details

A best choice of A- and B-solvents for varying solvent-type selectivity depends

on several factors:

• B-solvent type: nonlocalizing, basic localizing, or nonbasic localizing

• solvent miscibility

• solvent UV cutoff (solutes that are weakly UV-absorbing may require detec-tion at lower wavelengths)

Potentially useful solvents for NPC are listed with their properties in Table 8.1 For a full exploration of solvent-type selectivity, three different B-solvents will be required: nonlocalizing, basic localizing, and nonbasic localizing For a nonlocalizing B-solvent, methylene chloride is preferable to chloroform by virtue of its lower UV absorbance and lower toxicity However, detection can only be carried out at wavelengths>230 nm Methylene chloride is miscible with all of the other solvents

in Table 8.1

There are five possible basic-localizing solvents listed in Table 8.1 Ethyl ether and tetrahydrofuran are susceptible to oxidation by air, and for this reason are not

recommended Both n- and i-propanol are quite strong ( ε0

B = 0.60), which means

that mobile phases withε0< 0.25 require propanol concentrations of <0.5% (as

seen in Fig 8.6); the use of very low concentrations of the B-solvent can create problems (Section 8.5) and should therefore be avoided if possible—especially when

Time (min)

1 3

2 5

6 7

10 4

2/98 blend of mobile phase (c) with (b) (optimum); Rs= 1.3

Figure8.10 Optimized selectivity and resolution for sample of Figure 8.9 Conditions as in

Figure 8.9, except that mobile phase is a 98/2 blend of mobile phases (b) and (c).

using bare-silica columns When choosing a basic-localizing B-solvent, it is suggested that MTBE be used for mobile phases withε0< 0.48, and either n- and i-propanol

forε0> 0.48 Neither MTBE nor propanol present miscibility or detection problems.

Ethyl acetate and ACN are each candidates for a nonbasic-localizing B-solvent Ethyl acetate suffers in terms of UV absorbance (detection with this solvent is only possible at>250 nm), while ACN and hexane are not fully miscible The addition

of a co-solvent (e.g., methylene chloride, as in Fig 8.9c) allows the use of hexane

with ACN But determining the required addition of the co-solvent is somewhat tedious, and the estimation ofε values for mixtures of these three solvents is not

straightforward [18] However, a computer program (LSChrom) is available for the calculation ofε-values [19, 20] An alternative to the need for a co-solvent is the use

of ethoxynonafluorobutane as the A-solvent [21] (available from 3M, and miscible with ACN or MeOH—but expensive) with ACN as the B-solvent—this allows UV detection at wavelengths≥ 220 nm Hexane and ethoxynonafluorobutane can be considered as interchangeable in terms of solvent strength (ε0≈ 0.00).

8.3.3 Temperature Selectivity

The effect of temperature on NPC selectivity has received only limited attention; one

of a few reported examples is shown in Figure 8.11 In Figure 8.11a,b, separation

is shown for 22 and 55◦C, using methylene chloride as the (nonlocalizing) mobile phase It is seen that selectivity does not change appreciably with temperature, and this may be generally true for less-polar samples and nonlocalizing mobile

Time (min)

Time (min)

100% CH 2 Cl 2 mobile phase

1

2

5

1

2

5

2% ACN-hexane mobile phase

5

6 + 7

Figure8.11 Effect of temperature on relative retention in NPC Sample: 1, nitrobenzene; 2, methyl benzoate; 3, benzaldehyde; 4, acetophenone; 5, α-methyl benzyl alcohol; 6, benzyl alcohol; 7, 3-phenyl-1-propanol Conditions: 250 × 4.6-mm silica column (5-μm particles);

2 mL/min; mobile phase and temperature shown in figure Chromatograms recreated from data of [22]

phases For the separations of Figure 8.11c,d with a localizing mobile phase (2%

ACN-hexane), a change in temperature leads to a marked change in the relative retention of localized-solute peaks 6 and 7 Changes in selectivity with temperature

such as that of Figure 8.11c, d may be the result of the relatively reduced retention of

localizing solvents at higher temperatures, in which case similar changes in relative retention may result for a change from a localizing to a nonlocalizing B-solvent A change in temperature as a means of changing NPC selectivity is limited, in practice,

by the relatively low boiling points of the more useful solvents, and may have only marginal utility Alternatively, large changes in selectivity can be achieved by changing the B-solvent—minimizing the need for further changes in selectivity as by

a change of temperature

8.3.4 Column Selectivity

Because silica is most often used in NPC, and because a change of B-solvent is

an effective means for varying selectivity, a change of column is not often used for the sole purpose of changing NPC selectivity A more common reason for using a polar-bonded-phase column in preference to unbonded silica (primarily for assay methods) is to avoid the problems described in Section 8.5: poor separation reproducibility from run to run, slow column equilibration, or difficulty in using gradient elution for samples with a wide retention range Several studies have been reported [23–27] of retention for different test solutes, using each of the three types of polar-bonded-phase NPC columns (cyano, diol, amino) It appears that selectivity is a complex function of both column type and the choice of B-solvent [27], suggesting that trial-and-error changes in both column type and B-solvent can result in significant variations in relative retention

Figure 8.12 compares the separation of a mixture of aromatic compounds with hexane as mobile phase and (1) a cyano column, (2) a diol column, and (3) an

amino column The separation of the same sample on a silica column (Fig 8.12d)

can be estimated (very approximately) from other published data [1] A comparison

of these four separations suggests that run time (or ‘‘column strength’’) varies as

silica amino > diol > cyano Note the logarithmic time scale for the silica separation of Figure 8.12d, which

correctly implies that the range in retention (and selectivity) for a sample with a silica column can be far wider than for a polar-bonded-phase column

Separation selectivity or relative retention varies moderately from column to column in Figure 8.12, as expected; proton-donor solutes are retained more strongly

on amino columns relative to other solutes, and less strongly on cyano columns [26, 27] It should also be noted that the sample of Figure 8.12 can be separated

isocratically with any of the polar-bonded-phase columns (Fig 8.12a–c), whereas separation of this sample with a silica column (Fig 8.12d) would require gradient

elution (if gradient elution is even feasible; see Section 8.5.2) This observation

should be true for any sample: isocratic separation is more likely to be possible with

the use of a polar-bonded-phase column than with silica.

Finally, large values ofα (and the possibility of baseline resolution) are more

likely when a silica column is used Similarly changes in selectivity with change

Cyano column

Diol column

Amino column

Silica column (estimate)

0

Time (min)

1

2 3 4

5 + 6

7

3

Time (min)

Time (min)

1

2

4 5 6

7

3 + 4

Time (min)

Time (min)

1

2

7

(a)

(b)

(c)

(d )

Time (min)

1

2 3 4

5 +

Figure8.12 Comparison of retention and selectivity among different NPC columns

Sample: 1, chrysene; 2, perylene; 3, 1-nitronaphthalene; 4, 1-cyanonaphthalene; 5,

2-acetonaphthalene; 6, naphthalene-2,7-dimethylcarboxylate; 7, benzyl alcohol

Condi-tions: 150× 4.6-mm columns (column type indicated in figure); hexane mobile phase; 35◦C;

2.0 mL/min Chromatograms (a − c) reconstructed from data of [26]; (d) estimated from data

of [1] (note extreme change in retention range for silica column d vs polar-bonded columns a–c).

in the B-solvent (solvent-type selectivity) are much more pronounced for silica columns—compared to polar-bonded-phase columns Consequently a change from silica to a polar-bonded-phase column, with further variation of the mobile phase (change in %B and/or B-solvent), is unlikely to lead to a better resolution of a peak-pair that has not been separated after varying solvent-type selectivity with a silica column

8.3.5 Isomer Separations

We have noted that isomers are generally better separated by NPC than by RPC Similarly limited data [15, 26] suggest that isomer separation is generally more

pronounced on silica columns, compared to polar-bonded-phase columns

How-ever, the separations of Figure 8.2b–d with a cyano column make clear that polar-bonded-phase columns are able to separate some isomers Isomer-selectivity,

using NPC with a silica column, can be attributed to (at least) three possible characteristics of isomeric molecules:

• steric hindrance of a polar substituent by an adjacent nonpolar substituent

• electron donation or withdrawal from a polar group by a second substituent

in the solute molecule

• the relative positions of different polar groups within the molecule, and the planarity of the solute molecule

Figure 8.13 illustrates each of these three effects In Figure 8.13a,b it is seen

that for the separation of two methylaniline isomers the interaction of the polar –NH2 group with a surface silanol will be interfered with by an o-methyl group (Fig 8.13b), because of steric hindrance, but not by a p-methyl group (Fig 8.13a) The steric hindrance created by the o-methyl group will be further enhanced by

the adjacent silica surface Consequently a molecule with a sterically hindered polar substituent should be less retained than an isomer in which steric hindrance is

absent or less pronounced Figure 8.2b provides an experimental example of the contribution of steric resistance to isomer selectivity, where p-methylaniline (peak 3) is more strongly retained than o-methylaniline (peak 1) in this NPC separation.

In Figure 8.13c,d, steric hindrance of the –NH2group by the methyl substituent

does not occur for either the m- or p-isomer However, the methyl group in the

para position is more effective at transferring electrons to the –NH2group, in turn

increasing its hydrogen-bond basicity and retention; that is, a p-CH3 group has a more negative value of the Hammettσ -parameter [28] than does a m–CH3group

Therefore the p-methyl isomer should be more retained than the m-methyl isomer

in this example This is confirmed in Figure 8.2b where p-methylaniline (peak 3)

is more retained than m-methylaniline (peak 2) Similar examples of the effects of methyl substitution on retention are provided by the separations of Figure 8.2c,d.

In Figure 8.13e,f the relative retention of these two dimethoxyethylene

iso-mers will be affected by which isomer can better position itself adjacent to the

surface, so as to allow each polar methoxy group in the molecule to interact

with an adjacent silanol group on the surface In this example it appears that

cis-1, 2-dimethoxyethylene will be more strongly retained, but any such prediction

must be regarded as tentative; silanol groups are distributed randomly about the silica surface, and the solute molecule is free to adapt various positions on the silica surface Consequently there is usually little basis for predicting which of two posi-tional isomers will have its polar substituents more closely matched to the positions

of neighboring silanols In some cases adjacent polar groups within a solute molecule

(as in cis-1,2-dihydroxyethylene) can interact intra-molecularly (Fig 8.13g), possibly

competing with and reducing inter-molecular interactions that increase retention

These and other contributions to isomer separation on silica (as in Fig 8.13a–f ) are

examined in detail in Chapter 11 of [1]

Finally, approximately planar molecules are more easily matched to the (roughly) planar silica surface so that more planar isomers are preferentially retained

An example is provided in Figure 8.14 for the separation of five isomers of the

com-pound retinol In the RPC separation of Figure 8.14a, there is little separation of

these isomers, whereas in the NPC separation of Figure 8.14b every peak is at least partly resolved Peak-5 (the all-trans) isomer is more nearly planar and is preferentially retained Solutes with an increasing number of cis-linkages tend to be

increasingly less planar and less retained

Let us now compare all of the contributions above to isomer separation for

a silica column versus separation on (1) polar-bonded-phase NPC columns or (2) RPC columns In the case of polar-bonded-phase NPC columns, steric hindrance

effects (Figs 8.13a,b) will be less important because the silica surface is further

removed from the polar cyano, diol, or amino group of the stationary phase—hence contributing less to steric hindrance between the solute and the stationary phase Similarly the matching of polar groups in the solute molecule with polar groups in

the stationary phase (Figs 8.13e,f ) will be easier for a polar-bonded-phase column

(with less effect on isomer selectivity) because the cyano, diol, or amino groups are not rigidly positioned on the surface but are connected to the silica surface by a flexible –CH2–CH2–CH2–linkage Finally, the attraction of polar groups in the solute molecule to the polar stationary phase is weaker for polar-bonded-phase columns than for silica, which in turn reduces the effect of each of the

con-tributions to isomer separation in Figure 8.13 Consequently isomer separations

2 N

CH3 H

O

H O

H O

H O

H

O

H O

H O

H O

H O

H O

H2N

CH3

e –

C=C H OCH3

CH3O

H

H O

H O

C=C H OCH3

CH3O H

H

O

H O

H O

H O

C=C H OH O

H H

H2N

H O

H O

Figure8.13 Factors that contribute to isomer selectivity for NPC separation on silica

columns (a, b) Steric hindrance; (c, d) electron donation; (e, f ) relative positions of polar groups within the solute molecule; (g) intramolecular hydrogen bonding of two polar groups.

9 11 13

all-trans retinol

1 11,13-di-cis retinol

2 13-cis retinol

3 9,13-di-cis retinol

4 9-cis retinol

5 all trans retinol

15 20 15 20 (min) 1

2 +

+

2 3

4 5

Figure8.14 Comparison of isomer selectivity for separation by RPC (a) and NPC (b) Sample shown in figure (retinal isomers) Conditions: (a) 200 × 4.4-mm C18column; 80% methanol-water; 1.0 mL/min; 40◦C (b) 250 × 4.0-mm silica column; 8% dioxane-hexane;

1.0 mL/min Figure is adapted from [29]

on polar-bonded-phase columns will usually be less pronounced, compared to separation on silica.

For the case of RPC separation, the interaction of polar solute groups with the stationary phase is much weaker than in NPC, which minimizes each of the effects of Figure 8.13 (as in the case of polar-bonded-phase columns), and reduces isomer selectivity While corresponding interactions with the polar mobile phase are possible, the latter interactions are generally weaker than corresponding interactions

of solute and mobile phase with silica silanols, and less subject to steric effects There

is one exception to this conclusion for RPC, however, in the case of separations

on cyclodextrin columns (Section 6.3.5) Isomer resolution is more pronounced for the latter columns, and this may be interpreted as follows: First, the cyclodextrin molecule possesses a cavity into which a solute molecule can enter (see Figs 14.17, 14.18), and some solute molecules may fit this cavity better than others Second, the cyclodextrin molecule possesses multiple –OH substituents with fixed positions within the molecule So far as their effect on isomer separation, these cyclodextrin—OH groups may play a similar role in RPC as for silanols in NPC (as

in Fig 8.13e,f ).

The first step in NPC method development should consist of a review of the goals of separation, including reasons why NPC is being considered Unless some problem can be anticipated for the use of RPC—or has been experienced in prior RPC separations of the sample—RPC is normally a best first choice at the beginning

of method development Some applications for which NPC might be considered initially include: