L-1904 ur Expanding Understanding and Use of Reversed-Phase Column Selectivity John W. Dolan Lloyd R. Snyder LC Resources, Inc. 1 for additional information, contact John Dolan: John.Dolan @LCResources.com Lloyd Snyder: snyder0036 @comcast.net
Abstract Reversed-phase columns differ in chemical composition, physical properties, and in other ways. As a result, relative retention (selectivity) varies from column to column. Efforts have been made to understand, measure, and utilize column selectivity for the past three decades. ne such study, based on the so-called hydrophobic-subtraction model, was recently summarized (1). The latter model quantitatively relates column selectivity to the five principal sample-column interactions that determine retention and selectivity, and allows the characterization of different columns in terms of five selectivity parameters (H, S*, A, B, and C). Consequently, it is possible to select columns of either similar or different selectivity. The present report extends this study in the following ways. First, we now have values of H, S*, etc. for almost 350 different reversedphase columns. An examination of this database has provided answers to several questions of practical interest; e.g., how can type-a and B alkylsilica columns be distinguished on the basis of measurement, as opposed to manufacturers claims? Is it possible to use this database to predict values of H, S*, etc. for columns not presently included in the database? Second, is the present column characterization in terms of values of H, S*, etc. complete? Are there additional interactions that are important for some columns that have been overlooked? Third, how well does the present column classification meet various practical needs, such as matching columns or choosing columns for orthogonal separation? Finally, we will consider some additional questions that presently stand in the way of a complete practical understanding and effective utilization of column selectivity. (1) L. R. Snyder, J. W. Dolan and P. W. Carr, J. Chromatogr. A, 1060 (2004) 77 2
An Experimental Study of Column Selectivity log (k/k EB ) log α = η H - σ S* + β A + α B + κ C hydrophobic steric hydrogen bonding cation exchange 150 different sample compounds of all kinds 150 different type-b alkyl-silica columns (C 1 C 18, but mainly C 8, C 18 ) values of k predicted with ±1% accuracy (1SD) 3 nly 5 parameters are required to describe reversed-phase column selectivity such that predictions of k within ±1% can be made.
Column Solute Interactions <==> <==> hydrophobic: η H steric restriction: σ S* B: H H X - BH + hydrogen bonding: β A, α B cation exchange: κ C 4 These sketches illustrate the 5 basic solute-column interactions that serve to define reversed-phase column selectivity.
Extension to ther Column Types type-a alkyl silica polar-embedded polar-end-capped cyano zirconia-base polymeric phenyl polymeric fluoro-substituted... a total of 300 + different columns 5 After testing more than 300 columns, the basic 5-term description of selectivity holds for reversed-phase columns. A small improvement in predictability can be achieved for cyano and phenyl columns with the addition of one or two additional terms to the equation.
Selecting an Equivalent Column F s = {(H 2 -H 1 ) 2 + (S* 2 -S* 1 ) 2 + (A 2 -A 1 ) 2 + (B 2 -B 1 ) 2 + (C 2 -C 1 ) 2 } 1/2 Luna C18(2) (original column) F s = 0 Need F s 3 for equivalent columns Prodigy DS (3) F s = 1 Inertsil DS-3 F s = 2 J Sphere H80 F s = 10 overlap! 6 For column equivalency, an F s value of 3 is necessary. Here, two replacement columns fit that requirement, and give equivalent separation.
Selecting an rthogonal Column 1 2 3 + 6 riginal method 0 10 20 30 40 50 * 4 5 1 2 rthogonal method F s =196 3 5 6 0 10 20 30 (min) 4 7 To identify a column that will give significantly different selectivity, i.e., orthogonal separation, a large difference in F s is necessary. We have found that a combination of F s > 65 and a change in mobile phase organic from MeH to ACN (or vice versa) gives a high probability of an orthogonal separation.
A Portion of ur Column-Selectivity Database Column H S* A B C(2.8) C(7.0) Company 18. SB-CN 0.50-0.11-0.22 0.04-0.15 1.05 Agilent 19. Eclipse CN 0.46-0.07-0.31 0.00 0.07 0.99 Agilent 20. Kromasil C18 1.05 0.04-0.07-0.02 0.04-0.06 Akzo 21. Kromasil C4 0.73 0.00-0.33 0.02 0.01 0.00 Akzo 22. Kromasil C8 0.86 0.01-0.21 0.02 0.05 0.00 Akzo 23. Kromasil CN 0.44-0.14-0.58-0.01 0.22 1.04 Akzo 24. Adsorbosphere (C18) 0.99-0.07 0.07-0.04 1.50 1.68 Alltech 8
Accuracy of Values of H, S*, A, B, C required accuracy for accurate F s values: F s <±3 test repeatability (identical columns, same batch): F s = ±1 comparisons of first- vs. secondgeneration columns... do they make sense? 9 If we are to measure values of F s that are as small as 3, we need to have errors in F s that are smaller than this. In a collaborative study, 40 columns were measured by 4 different laboratories. The data allowed us to calculate errors in F s of ±1 unit (1SD), which is adequate for our purposes.
Average Change in Column Parameters for 7 Second- vs. First-Generation Columns e.g., Supelcosil LC-DB vs. LC; Inertsil DS-3 vs. DS-2 Change in value 0.0-0.2-0.4 H S* B A -0.6 C(2.8) -0.8 C(7.0) The results from 7 column pairs of Type A vs Type B columns fits our expectations. Higher bonding should increase hydrophobicity (H) and make it more difficult for bulky molecules to penetrate the bonded phase (S*). Less acidic silica should mean lower values of A and C. Reduction in B may be due to reduced metal content of the silica, which would reduce the preferential binding of carboxylic acids. So there are logical explanations of experimental results. 10
Possible Uses of the Database selecting similar or different columns identifying column type Type-A A or Type-B proprietary designations ( aquapore( aquapore, Supersil acid, etc.) predicting values of H, S*, A, B, C flagging bad values of H, S*, etc. 11 ne possible use of the database is to determine if a column is Type-A or Type-B, as well as to classify proprietary columns into similar classes when adequate column descriptions are not publicized.
Summary of Column Types in Database Column Type Total C 30 C 18 C 8 <C 8 Type-B 163 6 98 42 19 Type-A 56 0 45 11 0 Embedded polar group 31 End-capped polar group 16 Cyano 18 Phenyl 16 ther 25 12 The following comparisons focus on the most commonly used columns the first 4 types listed above.
Cation Exchange: C(pH 2.8) 30 20 Type-B (95%) C(2.8) = 0.25 % of columns 10 0 15 10 5 0 Type-A (11%) -0.5 0.0 0.5 1.0 1.5 2.0 13 The value of cation-exchange capacity at ph 2.8, C(2.8), is a good way to distinguish between most Type A (C>2.5) and Type B (C<2.5) columns.
B and C(2.8) Values for Type-B B Columns 0.2 B 0.1 0.0 B = 0.05 95% A A combination of B (an indirect measure of the metal content of the column) and C(2.8) gives a high level of discrimination between Type A and Type B columns. Most Type B columns fall within the region defined by B<0.05 and C(2.8)<0.25. 95% B C(2.8) = 0.25 14-0.1-1 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 C(2.8)
Embedded-Polar Polar-Group (EPG) vs. Polar End-Capped Columns R X R CH 3 -Si-CCH 3 X Embedded End-capped (X is a proton-acceptor group) % 60 40 20 80 40 60 40 20 B = 0.05 type-b C 18 polar end-capped EPG (ether) The column basicity (B) also is useful to distinguish between the embedded-polargroup columns containing a basic function and other columns. Polar end-capped columns have similar basicity to other Type B columns, as do embedded-polar-group columns containing an ether moeity. 30 20 10 EPG (other) -0.1 0.0 0.1 0.2 0.3 0.4 B 15
Possible Uses of the Database selecting similar or different columns identifying column type Type-A A or Type-B proprietary designations ( aquapore( aquapore, Supersil acid, etc.) predicting values of H, S*, A, B, C flagging bad values of H, S*, etc. 16 Although there are >300 columns in the present database, there are probably hundreds of others that are not. It would be nice to be able to estimate the values of H, S*, etc. for untested columns based on the results for the existing database.
Correlations with Ligand Length n (type-b) H S* B C 18 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 n solutes appear partially excluded from C 30 17 These plots show the average values of H, S*, and B for related columns. Columns with similar ligand length have similar values of column parameters (as evidenced by tight error bars on each point in each plot). This suggests that it seems promising to estimate column parameters for other related columns.
Correlations with Ligand Length n (type-b) A C(2.8) C(7.0) 18 n 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 - silanols become more exposed for n < 10 - solutes held more strongly by longer n - highly ionized columns (C[7.0]) somewhat different The relationship of column acidity, A, and cation exchange capacity, C, is curious. The explanation of this behavior is open to speculation. However, in spite of an incomplete explanation of the underlying cause, the average values of A and C are consistent among similar columns.
Estimating Values of H, S*, A, B, C Average values for a particular column type: for C18 columns, F s accurate to ± 15 units Calculations from column properties : for C18 columns, F s accurate to ± 12 units Comparison of values for columns from a single manufacturer comparison of values for C8 vs. C18 comparison of values for 30-nm vs, 10-nm ther column comparisons The use of average column parameters for columns of a particular type give F s values within about 15 units. By using column properties (pore diameter, ligand concentration, etc.), this is slightly improved to about 12 units. These are not sufficient for accurate comparisons, but certainly are better than nothing. 19 Comparison based on other factors are discussed in the following slides.
Comparison of Values for Specific Columns from a Single Manufacturer Comparison of values for C 18 vs. C 8 columns Change in value F s accurate to ± 10 units B C(7.0) H S* A C(2.8) 20 Here, comparisons of C 18 vs. C 8 columns from the same manufacturer (e.g., StableBond, Kromasil) demonstrate an additional slight improvement in predictability of F s to about 10 units; still not adequate for accurate description of column equivalency.
Comparison of Values for Specific Columns from a Single Manufacturer Comparison of values for wide-pore vs. narrow-pore columns Change in value B H S* A C(2.8) C(7.0) F s accurate to ± 10 units (low ph) The same small improvement in predictability is made when comparing columns of different pore size. 21
Type-B B Phenyl Column Comparisons log (k/k EB ) log α = η H - σ S* + β A + α B + κ C + π P N 2 Avg. values 2 N H 0.64 S* -0.14 A -0.22 B 0.02 C(2.8) 0.13 C(7.0) 0.69 P 1.0 n = 16 F s accurate to ± 11 units (low ph) 22 Phenyl columns have an additional kind of selectivity due to π-π interactions between the phenyl groups and aromatic compounds (π P). With only 16 phenyl columns in the database, there is a greater need for estimates of phenyl column parameters. The predictability of selectivity for phenyl columns (±11) is better than C 18 columns (±15), because there seems to be less manufacturer-tomanufacturer variation in phenyl columns.
Type-B Cyano Column Comparisons log (k/k EB ) log α = η H - σ S* + β A + α B + κ C + π P + µ D Avg. values - + 2 N-R C=N + - H 0.43 S* -0.09 A -0.48 B 0.00 C(2.8) 0.02 C(7.0) 0.75 P 1.0 D 1.0 n = 18 F s accurate to ± 10 units (low ph) 23 Cyano columns are capable of π-π interactions as well as interactions with molecules with large dipole moments (µ D). Cyano columns (18 in database) present a similar need for predictability as phenyl columns. There is less variability between cyano columns in the database and thus a better predictive value of the properties from the average values.
Estimating Values of H, S*, etc. A Summary Rough estimates can still be useful R s = 2.0 R s = 6 (need F s <±3) (need F s <±25) 24 Avg. values for columns of a certain kind -- C 18 : F s ± 15 -- cyano, phenyl: F s ± 10-11 Difference in values for related columns -- C 18 vs. C 8: F s ± 10-11 -- wide-pore vs. narrow-pore: F s ± 10-11 Although estimates of column parameters from average values is not accurate within 3 units of F s necessary to identify equivalent columns for challenging separations (e.g., R s <2), many separations are not as challenging. For the case of R s = 6, F s values within 25 are adequate, so estimates with errors of 10-15 units should be adequate.
Possible Uses of the Database selecting similar or different columns identifying column type Type-A A or Type-B proprietary designations ( aquapore( aquapore, Supersil acid, etc.) predicting values of H, S*, A, B, C flagging bad values of H, S*, etc. A final possible use of the database is to identify columns for which bad values of the column parameters had been tabulated, such as through experimental or tabulation errors. Although we have found some apparently discrepant values using this technique, only a few have changed upon retesting the columns. And even in these cases, we cannot be sure that batch-to-batch differences or shelf-life effects were not responsible. So the database may be of limited utility for this purpose. 25
Summary the H-S model is now well developed and appears useful for selecting alternative columns 300+ column database; new columns being added column manufacturers are invited to submit additional columns for testing and addition to the database data for additional columns can be estimated, but their accuracy limits their value somewhat the database and supporting software will be available on the USP website reprints: www.lcresources.com 26