Chromatographic test methods for characterizing alkylsiloxane-bonded silica columns for reversed-phase liquid chromatography

Journal of Chromatography B 1092 (2018) 207–219

Contents lists available at ScienceDirect

Journal of Chromatography B journal homepage: www.elsevier.com/locate/jchromb

Review

Chromatographic test methods for characterizing alkylsiloxane-bonded silica columns for reversed-phase liquid chromatography Colin F. Poole

T

⁎

Department of Chemistry, Wayne State University, Detroit, MI 48202, USA

A R T I C LE I N FO

A B S T R A C T

Keywords: Reversed-phase liquid chromatography Column classification Solvation parameter model Hydrophobic-subtraction model Prototypical test compounds System maps Alkylsiloxane-bonded silica stationary phases Selectivity

Major obstacles to formulating a simple retention mechanism for reversed-phase liquid chromatography have a direct impact on the development of experimental methods for column characterization as they limit our capability to understand observed differences in retention at a system level. These problems arise from the heterogeneous composition of the stationary phase, the difficulty of providing a working definition for the phase ratio, and uncertainty as to whether the distribution mechanism for varied compounds is a partition, adsorption or mixed (combination) of these models. Retention factor and separation factor measurements offer little guidance as they represent an average of various and variable contributing factors that can only be interpreted by assuming a specific model. Column characterization methods have tended to ignore these difficulties by inventing a series of terms to describe column properties, such as hydrophobicity, hydrophilicity, silanol activity, steric resistance, etc., without proper definition. This has allowed multiple scales to be proposed for the same property which generally are only weakly correlated. Against this background we review the major approaches for the characterization of alkylsiloxane-bonded silica stationary phases employing prototypical compounds, the hydrophobic-subtraction model and the solvation parameter model. Those methods using prototypical compounds are limited by the lack of compounds with a singular dominant interaction. The multivariate approaches that extract column characteristic properties from the retention of varied compounds are more hopeful but it is important to be more precise in defining the characteristic column properties and cognizant that general interpretation of these properties for varied columns cannot escape the problem of a poor understanding of the distribution mechanism.

1. Introduction Although the lexicon of available modes of liquid chromatography is now quite extensive, reversed-phase liquid chromatography (RPLC) remains the mode of choice for the separation of neutral and ionizable compounds in the biomedical and life sciences [1]. This dominance results from its capability to handle compounds of a wide range of size, polarity and ionicity. The combination of this versatility with rapid equilibration of the stationary phase with changes in mobile phase composition facilitating gradient elution and the availability of relatively straightforward, although largely empirical, approaches to method development add luster to the attraction [1, 2]. The inclusion of additives in the mobile phase allows exploitation of secondary chemical equilibria to control retention and selectivity by ion suppression, ionpair formation, and various complexation mechanisms. Reversed-phase liquid chromatography is now so widely employed in pharmaceutical, biomedical and life science research, product development, and quality control one can only speculate as to how these fields might look today if ⁎

Rm 185 Chemistry, Wayne State University, Detroit, MI 48202, USA. E-mail address: [email protected].

https://doi.org/10.1016/j.jchromb.2018.06.011 Received 31 March 2018; Received in revised form 3 June 2018; Accepted 4 June 2018 Available online 07 June 2018 1570-0232/ © 2018 Elsevier B.V. All rights reserved.

reversed-liquid chromatography had not arrived on the scene at the same time as exponential growth occurred in these areas. It would be an exaggeration to say that this growth was possible only because of the development of reversed-phase liquid chromatography, but it is certainly true that it was one of the major enabling techniques that facilitated the rapid progress made. From the beginning reversed-phase liquid chromatography has been synonymous with alkylsiloxane-bonded silica stationary phases, especially ocadecylsiloxane-bonded silica, or simply C18. Over time significant improvements in column performance, pH stability, and inertness with respect to the separation of basic compounds have been made [1, 3, 4]. More recent developments include sub-3 μm totally porous and core-shell particles and monolithic rods for fast separations or enhanced peak capacity [5–7]. Current developments in stationary phase chemistry are largely incremental and hidden from view to those, now the majority, of scientists who consider columns as a supply item and have passed the torch of innovation to the industrial sector for improvements through continuous product development. More than

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

2. Column chromatographic test methods

ever this has resulted in a crowded market place where competing companies launch new products at regular intervals supported by marketing information that often does not make chemical sense and serves to confuse as much as inform the customer. Columns are certainly more reliable, reproducible and robust compared to yesteryear, as can be gleaned from batch-to-batch studies of similar columns over multiple years [8, 9], but with so many columns of notionally similar type to choose from, the identification of columns with properties that closely resemble each other (selectivity equivalent) or columns with significantly different selectivity (orthogonal selectivity, although orthogonal is somewhat an exaggeration when applied to RPLC columns) suitable for the screening phase of method development have never been more difficult to identify. For typical alkylsiloxane-bonded silica columns the manufacturer will usually provide information for the type of bonded ligand, type of silica substrate (type-A, or -B), particle shape, the mean particle size, and mean pore diameter. In some cases they may provide information for the particle size distribution; the pore size distribution and pore volume; the specific surface area; percent carbon loading and/or degree of surface coverage (bonding density), concentration and identity of major metal impurities; and whether or not the stationary phase has been endcapped. Type-A silica refers to first generation silica substrates characterized by a relatively high concentration of metal impurities and type-B silica to modern high purity silica substrates with very low metal contamination. However, it is neither easy nor straightforward to relate the above properties to chromatographic retention and selectivity. Methods used to determine these properties remain non-standardized and they are not necessarily comparable when taken from different sources [2, 3, 9–12], The mean particle size and size distribution may be based on number, area or volume depending on the type of instrument used for the measurement and are not directly comparable, although this parameter is strongly correlated with the column kinetic performance (plate count or plate height) and the column pressure drop. Compared to the number particle size distribution, the area and volume distributions are more effected by large particles resulting in a distribution skewed towards the fraction of larger particles present in the packing. The surface area, pore size and pore volume are typically determined by either nitrogen BET or mercury porosity measurements with the aid of different models. The values obtained depend on the model employed and how micropores are treated in evaluating the experimental data. The percent carbon loading is influenced by whether the packing is endcapped or not and the extent of this reaction. For column packings with different surface areas the surface coverage (bonding density) is a more useful parameter but is calculated assuming the surface is covered with a single type of bonded organic moiety, in other words, does not handle endcapped packings correctly. In general, surface area and mean pore diameter values are given for the silica substrate prior to bonding and are not the values for the chemically bonded phase. Manufacturers typically withhold information for how a column packing is endcapped, or provide only general information, such as “by trimethylsilyl groups”, using a “polar endcapping reagent” or “endcapped to insure water compatibility”. Thus, in summary, typical physicochemical column parameters are not generally suitable or trustworthy for column selection and are only useful in a qualitative sense for a broad classification of column types. There is little impetus for individual laboratories to re-determine these parameters with a common standard scale since this means destroying (unpacking) the column. Also, the necessary equipment for these measurements may not be available in many analytical laboratories. From a customer perspective a more meaningful solution is the development of column chromatographic tests utilizing standard substances and prescribed separation conditions, typical instrumentation for liquid chromatography, and one or more chromatograms for data interpretation.

A number of column test methods have been proposed to assess different separation properties of alkylsiloxane-bonded silica stationary phases. The most prominent methods are the Tanaka [13], Engelhardt [14], the extended Engelhardt test proposed by Neue [9, 15] and the Hoogmartens' test (generally referred to as the Katholieke University Leuven method) [16, 17]. For these methods, specific compounds (prototypical compounds) with an assigned singular characteristic property are injected onto the column and their retention factors, separation factors, or in some cases peak asymmetry factors and column plate number, are determined with a specified mobile phase composition. These values are then used to rank columns according to the property inferred for the prototypical compounds. Typical column characteristic properties determined in this way are hydrophobicity (or hydrophobic selectivity), silanol activity (hydrogen-bonding interactions with neutral silanol groups), polar interactions, shape selectivity (or steric selectivity), metal contamination and ion-exchange capacity (electrostatic interactions with ionized silanol groups). Numerous studies have compared individual test methods leading to the selection of a smaller number of tests to define column properties. Euerby and coworkers [18–20], for example, used principle component analysis to define six test procedures to compare a large database of columns for the separation of pharmaceutical compounds. Likewise, Adams and coworkers [16, 17] selected four test parameters to rank columns in a large column database for the separation of pharmaceutical compounds. The object of these studies was to identify columns with similar separation properties (selectivity equivalent columns) that could be used as an alternative for a column specified in a method and to identify columns with significant selectivity differences for the column screening phase in method development. A different approach to column characterization based on quantitative retention relationships is represented by the hydrophobic-subtraction model and the solvation parameter model. These are modelbased, solute-independent approaches that isolate pre-defined characteristic column properties based on experimental retention factors for varied solutes with established capabilities (descriptor values) associated with the characteristic column properties. The hydrophobic subtraction model, Eq. (1), is an empirical model only applicable to retention in reversed-phase liquid chromatography [21–23]

log k = log kEB + η′H − σ ′S∗ + β′A + α′B + κ′C

(1)

in which k is the retention factor for a compound selected to facilitate calculation of the characteristic column properties; kEB is the retention factor for ethylbenzene (a reference compound to account for differences in the phase ratio for compared columns); η′, σ′, β′, α′ and κ′ are characteristic solute properties (descriptors); H, S⁎, A, B and C are the characteristic properties of the system (column, mobile phase and temperature). To characterize column properties, typically isocratic retention factors for 16 specified compounds and ethylbenzene are determined on each column with a mobile phase of 50% (v/v) acetonitrile-phosphate buffer at pH = 2.8 or 7.0 and a temperature of 35°C. It is further assumed that the column parameters (H, S⁎, A, B and C) are largely independent of mobile phase conditions except for pH and the solute parameters (η′, σ′, β′, α′ and κ′) are varied to allow fitting of the retention factors to the model for different column types [24]. The contributions to retention from hydrophobic interactions is identified as η′H, from steric interactions σ′S⁎, from hydrogen-bonding of basic solutes to silanol groups (stationary phase acting as a hydrogen-bond acid) β′A, from hydrogen bonding of acidic solutes to basic groups within the stationary phase (stationary phase acting as a hydrogenbond base) α′B, and ion-exchange interactions of ionized (protonated) bases at a mobile phase pH = 2.8 κ′C2.8 or pH = 7 κ′C7.0. The column properties (H, S⁎, A, B and C) are determined simultaneously by multiple linear regression analysis for the complete set of experimental retention factors for compounds selected to cover the descriptor space 208

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

Table 1 Column properties obtained from the hydrophobic-subtraction model. Parameter

Description

H S⁎

Hydrophobicity relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica stationary phase Steric resistance to insertion of bulky solutes into the stationary phase relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica stationary phase Column hydrogen-bond acidity (donor) capability relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica stationary phase Column hydrogen-bond basicity (acceptor) capability relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica stationary phase Column cation exchange activity relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica stationary phase with a mobile phase pH = 2.8 (C2.8) or pH = 7 (C7.0). For mobile phase preparation see [21, 24].

A B C

associated with the variation in η′, σ′, β′, α′ and κ′ values. The column properties are normalized by division of each column parameter H, S⁎, A, B and C by those for a hypothetical average type-B silica octadecylsiloxane-bonded column with the values for the reference column set to H = 1 and S⁎ = A = B = C = 0. See Table 1 for a description of the column parameters. It is notable that the hydrophobic-subtraction model contains no explicit term for interactions of a dipole-type, which are indicated as likely to be negligible in reversed-phase liquid chromatography [25]. The solvation parameter model is a fundamental model based on the parameterization of the cavity model of solvation with numerous applications in chemistry and environmental and biopartitioning studies to systems described by an equilibrium distribution constant [26–29]. In separation science, typical applications include the characterization of stationary phases in gas-liquid chromatography [30–32], micellar electrokinetic chromatography [33], and supercritical fluid chromatography [34]; liquid-liquid partition [35, 36]; as well as reversed-phase liquid chromatography [37–40]. For reversed-phase liquid chromatography the appropriate form of the solvation parameter model is [29, 40–42]

log k = c + eE + s S + a A + bB + v V

describes the difference for the defined property for the mobile and stationary phases. The system constants are determined by multiple linear regression analysis by inputting experimental isocratic retention factors for a number of varied compounds selected to cover a wide descriptor space and with minimal correlation between the descriptor values. The selection of compounds for column characterization is not prescriptive and compounds of sufficient number with favorable retention properties that match the above criteria can be freely selected from the large databases of descriptor values available [29, 35, 41, 42]. The equation constant, c term, is not a fundamental solvation parameter, but when the dependent variable is the retention factor, it contains information for the phase ratio of the system as well as unattributed contributions from the statistical fit of the model to the experimental retention factors. The solvation parameter model facilitates studies of the influence of solvent type, composition and temperature on column properties as system maps (single variable) or response surfaces (two variables) [40, 43–45]. The solvation parameter model contains no explicit terms to account for cation-exchange interactions or steric repulsion included in the hydrophobic subtraction model. It does, however, facilitate the detection of these interactions by comparison of the predicted retention factors for weak bases or bulky compounds for comparison with experimental retention factors. The general form of the solvation parameter model has been parameterized to include ion-exchange interactions but in this extended form has not been used for column characterization in reversed-phase liquid chromatography [46, 47]. The solvation parameter model also facilitates identification of solvent-effects on selectivity from correlation plots of the system constants on the compared columns [48]. The methodological details of the test methods described in this section focused on the choice of test compounds, experimental procedures and chemometric methods for column comparison have been reviewed several times [11, 15, 16, 22, 23, 29, 35, 40–42, 49–52]. There is little value in reassembling the information adequately collated in the cited works. Instead, we will focus on the unanswered questions that arise in the different approaches and the limitations of the interpretation. In other words, not how the column tests are performed and the extraction and comparison of information from the various column databases but how the tests have been interpreted and the limitations imposed by assumed retention mechanisms that may not be realistic. To help frame the discussion for the general reader a commentary on the processes that obfuscate a simple understanding of retention in reversed-phase liquid chromatography with an incipient impact on

(2)

where the upper case letters represent solute descriptors defined as the contribution of n- and π-electron lone pair interactions (or the additional contribution to dispersion interactions that arise from loosely bound electrons in polarizable molecules) E, the contribution of dipoletype interactions (orientation and induction) S, the contribution of solute hydrogen-bond acidity A, the contribution of solute hydrogen-bond basicity B, and the contribution of solute size (cavity formation) and dispersion interactions (both correlated with the solute size), V to the retention factor (log k). McGowan's characteristic volume V can be calculated from molecular structure and E can be calculated for liquids from its refractive index and characteristic volume [29, 41, 42]. The S, A and B descriptors are determined by experiment, typically using liquid-liquid partition, chromatography or solubility measurements. The E descriptor for solids can be calculated using estimated refractive index values or determined experimentally together with the other descriptors. The descriptors in the solvation parameter model are solute dependent and system independent in contrast to the hydrophobicsubtraction model. The complementary system constants (lower case letters in italics) characterize the column properties in reversed-phase liquid chromatography, as indicated in Table 2. Each system constant Table 2 Column properties obtained from the solvation parameter model. System constant

Description

e s a b v

Difference in the contribution of electron lone pair interactions in the solvated stationary phase and mobile phase at equilibrium Difference in the contribution of dipole-type interactions in the solvated stationary phase and mobile phase at equilibrium Differences in hydrogen-bond basicity of the solvated stationary phase and mobile phase at equilibrium Difference in hydrogen-bond acidity of the solvated stationary phase and mobile phase at equilibrium Difference in cohesion (cavity formation) in the solvated stationary phase and mobile phase at equilibrium and additional dispersion interactions that are not cancelled when the solute is transferred from the mobile phase to the solvated stationary phase. A surrogate measure of the system phase ratio when the stationary phase is in intimate contact and at equilibrium with the mobile phase. It is not a fundamental solvation parameter

c term

209

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

different property [60–62]. In reversed-phase liquid chromatography a common measure of “hydrophobicity” is the retention factor for any compound with water as a mobile phase (log kw) obtained by extrapolating the retention factors (log k) from a series of binary water-organic solvent mobile phase compositions to zero organic solvent [27, 60–63]. Alternative methods using gradient elution have been described for the high-throughput estimation of this parameter [64]. This is not a robust parameter from a measurement perspective, however, since values obtained by extrapolation rarely agree with those obtained by direct measurement and, in addition, the values depend on the choice of mobile phase compositions for the measurement [27, 61]. It is likely that log kw has no real physical meaning and its use as a measure of hydrophobicity is misplaced. A chromatographic hydrophobicity index has also been developed but is not relevant for column characterization [65] The above descriptions characterize hydrophobicity mainly as a compound property specifically related to an aqueous environment. Its use in column characterization implies that it is a column property. A hydrophobic surface is defined as one with a static water contact angle > 90° [66] but does not reflect the information captured in column characterization studies. The general use of the term in column characterization would imply contributions from a “phobia” due to a predominantly aqueous mobile phase and a “philia” due to favorable solvation in a predominantly non-aqueous environment represented by the column stationary phase. It is difficult to see how such a term would not depend on the mobile phase composition, and indeed, it is an open question whether hydrophobicity is the correct term to associate with aqueous-organic mobile phases with a large volume fraction of organic solvent. Even if the mobile phase at the microscopic level is considered to consist of solvent clusters of different composition, including exclusively water, only part of the driving force for transfer to a largely non-aqueous environment could be associated with the water clusters and competition with clusters in which water is not a dominant component would need to be considered [67, 68]. In terms of its general use in column test methods hydrophobicity clearly needs a more informative definition than those too casually presented in the past. A suitable compound that is purely hydrophobic according to the definition is then required to rank columns on a meaningful scale according to the retention properties of this compound. In theoretical studies inert gases or low-mass n-alkanes are used to represent the singular property of hydrophobicity [59]. In reversedphase liquid chromatography a range of low-polarity compounds with favorable uv-absorption properties for detection have been used for this purpose. Typical of these compounds are toluene and ethylbenzene (Engelhardt test [14]) with 55% (v/v) methanol-water at 40 °C); npentylbenzene (Tanaka test [13]) with 80% (v/v) methanol-water; naphthalene and acenaphthalene (Neue test [15]) with 65% (v/v) methanol-water at 23 °C; and biphenyl and naphthalene (Borges test [50]) with 70% (v/v) methanol-pH 7.0 buffer at 23 °C. The same compounds are used in several other test methods with the same or similar conditions [11, 16, 18, 49]. A convenient range of retention factors in reversed-phase liquid chromatography requires a suitable solvent strength, which for the test compounds used to measure hydrophobicity results in mobile phases with a large volume fraction of methanol (> 50% v/v). The contribution to hydrophobicity from the mobile phase is likely to be dampened by such large volume fractions of an organic solvent. The solute descriptors for the hydrophobicity test compounds are summarized in Table 3 and indicate their capability for general intermolecular interactions. None of these compounds are hydrogen-bond acids but all are dipolar/polarizable and hydrogen-bond basic. The homologs toluene, ethylbenzene and n-pentylbenzene have similar properties differing primarily in size (V descriptor) while the two-ring aromatic compounds are considerably more dipolar/polarizable and hydrogen-bond basic than the n-alkylbenzenes. Hydrophobicity determined by individual n-alkylbenzenes is expected to be highly correlated but significantly different to hydrophobicity

column test methods from the perspective of the interphase model for the stationary phase [48, 53], the complications arising from the selective adsorption of one (or more) components of the mobile phase at the sorbent surface due to the influence of surface forces (surface excess absorption) [54, 55] and fundamental difficulties in defining a retention mechanism (partition versus adsorption) and assigning a meaningful column phase ratio [56–58] is provided as electronic Supplementary material. Current thinking is that the retention mechanism is based on simultaneous multiple adsorption and partition interactions occurring in and close to the surface of the solvated stationary phase supplemented by specific interactions with the silica surface and embedded polar functional groups for polar compounds. The various sites are associated with different sorption energies and are subject to steric constraints by the limited flexibility of the chemically bonded ligands. It is further assumed that compounds distribute themselves between the different sorption sites of the solvated stationary phase and solvent clusters of variable size and composition known to exist in aqueous-organic solvent mixtures, each characterized by a separate distribution constant. In this case, the observed retention factor is an overall average of each distribution constant and their associated phase ratios, and does not pertain to a single retention mechanism. This limits the use of retention factors to assess the priority of different models of the distribution process. 3. Characteristic column properties Column chromatographic test methods based on prototypical compounds are indicated to rank column according to their hydrophobicity, steric or shape selectivity, silanol activity, cation-exchange capacity, and various miscellaneous properties, such as polar selectivity and metal contamination. The hydrophobic subtraction model ranks columns according to their hydrophobicity, steric interactions, hydrogenbond acid/base capability and ion-exchange capacity. The solvation parameter does not use these terms specifically but enables columns to be ranked according to the contribution of solute size to retention and contributions from electron lone pair, dipole-type, and hydrogenbonding interactions for any combination of chromatographic conditions and not just those specified by the test method. Since it is not explicitly a column test method it has the advantage that it allows comparison with other two-phase separation systems utilizing the same set of parameters. Although there is an expectation that columns ranked according to the same property determined by different test methods should be at least roughly equivalent, this is not generally the case, and we can most profitably attempt to establish the reasons for this discrepancy. 3.1. Hydrophobicity The use of terms thought to be commonly understood, but rarely, if ever, specifically defined is not unique to reversed-phase liquid chromatography. In such circumstances it is not unusual for multiple approaches to be justified to determine a specified property, but in reality, these approaches do not describe equivalent processes. Hydrophobicity is such a property. As found in most dictionaries it is defined in a medical context referring to a fear of water. Theoretical chemists discuss hydrophobicity in terms of the mechanism by which a hydrophobic compound modifies the structure of water by forming transient, semiordered, clathrate-like clusters around it, arising from enhanced water hydrogen bonding [59]. This enhancement is brought about by either strengthening or increasing the number of water hydrogen bonds. Dorsey and Khaledi describe the hydrophobic effect as the repulsion of hydrophobic solutes from an aqueous medium into an organic environment [60]. In pharmaceutical science hydrophobicity is often associated with the octanol-water partition coefficient (log Pow) and sometimes confused with lipophilicity, which should be considered a 210

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

phase and its poor correlation with stationary phase type was considered a deficiency for column characterization without considering the alternative hypothesis. After correction for the column phase ratio, the retention of non-polar compounds on alkylsiloxane-bonded silica stationary phases appear to be quite similar, and if methylene group selectivity is promoted as a relevant measure of hydrophobicity, then the hydrophobicity of alkylsiloxane-bonded silica stationary phases are similar. The presumption that an effective scale of hydrophobicity would distinguish between stationary phases with different topology is an unnecessary limitation on this parameter. Water-methanol mobile phase compositions with a high volume fraction of methanol contribute to the limited range of methylene group selectivity by compressing the range of the methylene group selectivity scale [69]. The use of the average of the retention factors for toluene and ethylbenzene as a scale for column hydrophobicity is noted in passing but there is no obvious theoretical or practical advantage to justify such an approach [70]. The hydrophobic-subtraction model has the advantage that it does not associate the measurement of column hydrophobicity, H in Eq. (1), with the properties of a single prototypical compound. On the other hand, the meaning of hydrophobicity is only defined by its use in Eq. (1) [21, 23, 71, 72]. The column hydrophobicity H is determined by the contribution (η′H) to retention (log k) for a compound that is more or less hydrophobic than ethylbenzene (log kEB) when the other column or solute parameters are set to zero (steric resistance, hydrogen-bonding and cation-exchange interactions σ′S⁎, β′A, α′B and κ′C = 0) compared to an hypothetical average type-B silica octadecyldimethylsiloxanebonded silica stationary phase with a mobile phase of 50% (v/v) acetonitrile-aqueous buffer at 35°C. The model uses the conscript that the solute descriptor η′ varies with the type of stationary phase while H is constant suggesting that the solute hydrophobicity is dependent on its environment. It would seem more obvious that a solute has a certain capability to enter into hydrophobic interactions, which are only limited by changes to its environment. Given that the mobile phase is constant for the measurement of H then differences in the column properties should be the reason for the variation in H and not the reason to adjust the properties of the solutes. Ethylbenzene is dipolar/polarizable and hydrogen-bond basic, and to the extent that the column can participate in these interactions, must be subsumed by the value for H. The column parameters H, S*, etc. are relative to a hypothetical average octadecyldimethylsiloxane-bonded type-B silica column rather than an absolute measure of column properties, This makes it difficult to deduce what hydrophobicity indicated by H refers to in terms of fundamental intermolecular interactions. It is generally the most important parameter when combined with the retention factor for ethylbenzene in the prediction of retention for neutral compounds on alkylsiloxane-bonded silica columns [23, 71–73]. It is probably a composite terms that explains a large fraction of the variance associated with retention in reversed-phase liquid chromatography and further than it use in Eq. (1), it is difficult to assign to it an exact physicochemical meaning.

Table 3 Representative prototypical compounds used to determine the selectivity of column packings in reversed-phase liquid chromatography. Compound

Descriptors (solvation parameter model) V

E

S

A

B

(i) Hydrophobicity Toluene Ethylbenzene n-Pentylbenzene Naphthalene Acenaphthene Biphenyl

0.8573 0.9982 1.6359 1.0854 1.2156 1.3242

0.606 0.613 0.594 1.241 1.566 1.342

0.515 0.509 0.509 0.921 1.150 0.956

0 0 0 0 0 0

0.137 0.147 0.147 0.188 0.201 0.284

(ii) Shape selectivity Triphenylene o-Terphenyl Benzoic acid Sorbic acid

1.8234 1.9320 0.9317 0.9424

2.960 1.948 0.73 0.48

1.773 1.351 0.90 0.83

0 0 0.59 0.55

0.425 0.379 0.40 0.51

(iii) Silanol activity (neutral compounds) Phenol 0.7751 0.776 Caffeine 1.3632 1.643 Ethyl benzoate 1.2135 0.694 Benzyl alcohol 0.9160 0.804 Pyridine 0.6753 0.630 1,3-Dinitrobenzene 1.0648 0.985 Dimethyl phthalate 1.4288 0.795

0.772 1.728 0.897 0.872 0.838 1.715 1.502

0.703 0.031 0 0.409 0 0 0

0.317 1.243 0.451 0.557 0.459 0.424 0.798

determined by the two-ring aromatic hydrocarbons. Hydrophobicity determined with these probes is a composite term influenced by all intermolecular interactions, except for hydrogen-bond basicity, and as a single numerical value for different contributions from column interactions, can result in similar column rankings for columns with different properties. These compounds reflect a range of interactions characteristic of themselves and similar compounds but not a singular property identified with hydrophobicity unless hydrophobicity is defined in such a way as not to be a singular column property. Assuming the identification of a purely hydrophobic compound for column characterization was agreed upon then scales of hydrophobicity based on the observed retention factors for standard separation conditions would still be flawed. The retention factor is defined as the product of the distribution constant (if a single distribution mechanism could be assumed) and the column phase ratio. Since an independent determination of the column phase ratio is not straightforward, the column hydrophobicity scale is the product of two unrelated properties, a property of the stationary phase, its hydrophobicity, and a property of the system, the column phase ratio. Operationally this might be acceptable but would cause problems in arriving at a suitable definition for hydrophobicity from retention factor measurements. Some column characterization methods include a second measure of hydrophobicity referred to as hydrophobic selectivity [16, 18–20, 69]. This is typically determined as the separation factor for two consecutive members of a homologous series that differ by a methylene group (also referred to as methylene group selectivity). Typical test compounds are n-alkylbenzenes with n > 2 (where n is the carbon number for the alkyl chain), Hydrophobic selectivity has also been defined as the difference in the separation factor for propyl paraben (4-hydroxybenzoic acid propyl ester) and phenol corresponding to the retention increment associated with a propyl ester group [19]. Since the propyl ester group is dipolar/polarizable and hydrogen-bond basic while the methylene group is considered to exhibit no polar interactions the two hydrophobicity selectivity scales are not expected to assess the same column property. Methylene group selectivity would seem to offer some operational advantages as a measure of hydrophobicity (or hydrophobic selectivity) in the absence of steric resistance and if a single distribution mechanism is assumed. It was considered of limited use for distinguishing stationary phase topology [8, 15, 21]. Methylene group selectivity shows little dependence on the chain length of the bonded

3.2. Steric resistance and shape selectivity There are two common uses of column test methods to evaluate steric factors in reversed-phase liquid chromatography. One group of tests attempts to evaluate the capability of a stationary phase to enhance the separation of compounds of similar size with different conformations referred to as shape selectivity. For example, Sander and coworkers [74, 75] devised a test based on the separation factor for tetrabenzonaphthalene and benzo[a]pyrene with a mobile phase of 85% (v/v) acetonitrile-water, which facilitated the classification of octadecylsiloxane-boned silica columns according to their bonding chemistry (monomeric vs. polymeric) as well as providing insight into the selection of column types for the separation of isomeric polycyclic aromatic hydrocarbons. Length-to-breadth and thickness parameters were developed to explain the separation properties in conjunction with a slot model for the retention mechanism. This test, however, is not a general 211

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

test of shape selectivity as it applies explicitly to isomeric polycyclic aromatic hydrocarbons. The Tanaka test employed the separation factor for triphenylene and o-terphenyl in 80% (v/v) methanol-water to evaluate shape selectivity in a more general way [11, 13, 76]. Triphenylene has a rigid planar structure while o-terphenyl is twisted and not as easily accommodated in the stationary phase if the free volume between the chemically bonded ligands is restricted. Euerby et al. [18, 21, 50] proposed a similar test based on the separation factor for benzoic and sorbic acids and cinnamic and 3-phenylpropionic acids. These tests assume a retention mechanism based on partition and would have to be revisited if this is not the case. Further, it is assumed that each pair of compounds has similar hydrophobicity such that changes in the separation factor result almost entirely from differences in shape. As shown in Table 3, the chosen prototypical compounds have at least moderate differences in their capability for individual intermolecular interactions in addition to differences in their stereochemistry. These differences in solvation must contribute to the separation factors independent of steric factors as well as the possibility of partial exclusion from the interphase volume due to the difficulty that twisted molecules have in inserting themselves between the bound ligands of the solvated stationary phase. Differences in solvation of the prototypical compounds are stationary phase dependent and reduce the reliability of the separation factor scales to describe exclusively shape selectivity. This is likely the basis for why individual shape selectivity tests generally show poor correlation for the ranking of stationary phases [11, 16–20]. The ability to separate compounds with conformational differences is an important goal in separation science but whether this can be modeled in a comprehensive manner by the separation factors for a few prototypical compounds of an unrelated structure to the compounds of interest (the sample) is debatable. The hydrophobic-subtraction model includes a general term to account for steric resistance (σ′S⁎) to correct retention factors for the inability of compounds to penetrate completely into the solvated stationary phase volume due to its size and/or bulkiness [70, 77]. The poor correlation of values of S⁎ and the above methods using separation factors for prototypical compounds to estimate shape selectivity suggest that steric resistance and shape selectivity represent different processes [77]. The hydrophobic-subtraction model treats steric resistance as a small perturbation to the dominant contribution from hydrophobic interactions. All test compounds except for ethylbenzene are assigned a capability for this interaction, σ′ ≠ 0) but few stationary phases have a significant complementary capability compared with an average hypothetical octadecyldimethylsiloxane-bonded type-B silica stationary phase. Many of the low contributions to σ′S⁎ could be a result of overfitting of the model or from variations in the distribution mechanism while for a few stationary phases it is likely important. The interpretation of steric resistance is based on a partition model that may not be an appropriate model for the distribution process for all compound and stationary phase combinations allowing other explanations for steric interactions. Steric resistance in the stationary phase is a complex process and probably not easily captured by a simple single parameter term. The fact that it seems to be only a minor contribution which is not strongly dependent on stationary phase topology (compared to an average hypothetical octadecyldimethylsiloxane-bonded type-B silica stationary phase) facilitates the general application of Eq. (1) for column characterization. The solvation parameter model contains no explicit term to account for steric resistance but for the construction of system maps employs a general screening strategy for retention factors (log k) as a function of binary mobile phase compositions to identify compounds effected by steric resistance [40, 57, 78–81]. It is assumed that the retention factors as a function of mobile phase composition conform to Eq. (3) and that there is a general increase in retention with a decreasing volume fraction of organic solvent

log k = a0 + a1 ϕ + a2 ϕ2

Fig. 1. Representative plots of deviations from Eq. (3) for compounds affected by steric resistance indicated as category 2 and category 3 (see text for details). The stationary phase in this case was XTerra MS C18 and the compounds illustrating category 2 behavior biphenyl and category 3 behavior ethylbenzene.

where ϕ is the volume fraction of organic solvent and a0, a1, and a2 are regression coefficients not assigned any physical meaning. Application of Eq. (3) allows varied compounds to be sorted into three categories. Those in category 1 display retention factors conforming to Eq. (3) and is the only behavior observed for many stationary phases. In some cases discontinuities in the fit of the retention factors to Eq. (3) are observed as indicated by the labels category 2 and category 3 in Fig. 1 [81]. Each solute shows some region, typically at higher volume fractions of organic solvent that can be merged with the category 1 solutes for use in the solvation parameter model and other regions where the same compounds would be flagged as extreme values or are outliers. Steric resistance is more commonly observed for alkylsiloxane-bonded silica stationary phases with a high bonding density. The type and volume fraction of organic modifier is also important with steric resistance generally enhanced by water-methanol mobile phases and water-rich mobile phases in general. It is more common for angular compounds, rigid planar compounds, compounds with long alkyl chains, and for bulky compounds in general. Size alone, however, is a poor predictor of steric resistance and other factors such as conformational rigidity and the presence of hydrogen-bonding functional groups are also implicated. All contributions are negative (loss of retention) but the process is too complicated to be described by a simple solute descriptor and so is not modeled by the solvation parameter model. The screening mechanism does not require a prescribed distribution mechanism. In terms of partitioning, if a solute can only partially penetrate into the stationary phase the remaining portion must reside either in the mobile phase or aligned with the interface between phases where it experiences interactions different to those in the stationary phase. An alternative explanation is the result of a change in the distribution mechanism resulting from a solvent-mediated conformation change in the stationary phase structure. For compounds retained by a mixed retention mechanism the abrupt change in retention could be the result of a change in the contribution of partition and adsorption to the retention mechanism for different conformations of the interphase region. The observations are real but the explanations remain speculative. There are many stationary phases for which Eq. (3) is a good model of retention behavior for small molecules (those compatible with average retention for a specified mobile phase composition range) suggesting that steric resistance is not a universal property but rather a feature of some stationary phases, and in those cases where steric resistance is observed, insufficient knowledge of the microscopic structure of the interphase region prevents the development of a detailed mechanism. The use of the concept of steric resistance in the hydrophobic-subtraction model and with the solvation parameter model is probably related but not necessarily identical. The hydrophobic subtraction model considers steric resistance as a general stationary phase property observed as a minor perturbation effecting retention for many stationary phases while the screening approach employed with the solvation parameter model

(3) 212

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

views steric resistance as a property of a few stationary phases that generally results in a large loss of retention associated with a small range of mobile phase compositions. 3.3. Silanol activity Alkylsiloxane-bonded silica stationary phases contain residual silanol groups at the silica surface that are not completely removed by endcapping [2–5, 82–86]. These silanol groups are not all equivalent in their interactions with solute and solvent molecules. There are three types of neutral silanol groups characterized as geminal vicinal and isolated representing adsorption sites of different energy for neutral molecules and a small population of ionized silanol groups that can interact with bases through electrostatic (ion exchange) interactions. Changes in mobile phase pH alter the relative contribution of neutral and ionized silanol groups to the retention mechanism since ionized silanol groups are neutralized at low pH with a change in the capacity of the stationary phase for those interactions associated with neutral silanols. In addition, accessible silanol groups are modified by adsorption of solvent molecules so interactions between neutral solutes and these solvated neutral silanols are subject to change depending on the mobile phase composition. It is difficult to conceive of a single test for silanol activity as several uncorrelated parameters need to be considered to encompass the full range of possible silanol interactions as well as the contributions from neutral and acidic silanol groups cannot be considered equivalent. For this reason we have chosen to divide tests under the umbrella of silanol activity into two categories characterized as interactions associated with neutral silanols and neutral solutes and those occurring by electrostatic interactions between ionized silanols and protonated bases.

Fig. 3. A system map for the octadecylsiloxane-bonded silica column (Kinetex C-18) for small neutral compounds with acetonitrile-water mobile phase compositions. Reproduced from [87] with permission.

in general) are the dominant contribution to retention from cavity and dispersion interactions (v system constant) and by the hydrogen-bond acidity of the mobile phase (b system constant). In the latter case, the stationary phase only competes poorly with the mobile phase for interactions with hydrogen-bond bases. All interactions with a negative sign favor residence in the mobile phase and indicate that the solvated alkylsiloxane-bonded silica stationary phases are generally less competitive for interactions of a dipole-type (s system constant) and less hydrogen-bond basic (a system constant) than the mobile phase. Electron lone pair interactions are for the most part positive and weak, and favor retention by the stationary phase. However, all system constants represent a relative contribution in one phase over the other indicated by the sign of the system constant and its importance by its numerical value. It seems reasonable to associate the large v and b system constants with the presence of water in the mobile phase since water is the most cohesive and hydrogenbond acidic of the solvents typically used in reversed-phase liquid chromatography. Only in mobile phases containing an appreciable volume fraction of water is retention dominated by the relative ease of cavity formation and dispersion interactions and solute hydrogen-bond basicity. In terms of selectivity, however, the other interactions are not inconsequential for the separation of neutral compounds but is influenced less by mobile phase composition. The system constants for monomeric alkylsiloxane-bonded type-B silica stationary phases with different topology and morphology are summarized in Table 4 for a mobile phase composition of 50% (v/v) methanol-water [78–81, 87–94]. The surface-bonded groups include octadecyldimethylsiloxane, octyldimethylsiloxane, diisobutyloctadecylsiloxane, and alkylsiloxane-bonded phases containing a polar embedded group or mixed surface-bonded octadecyldimethylsiloxane and a short-chain ligand with a polar terminal functional group on totally porous, inorganic-organic hybrid, or superficially porous silica particles and one silica monolithic stationary phase. The twelve endcapped octadecyldimethylsiloxane-bonded silica stationary phases differ in their average pore size, surface area and bonding density but have quite similar system constants (average values v = 2.06 ± 0.09, e = 0.33 ± 0.05, s = −0.76 ± 0.05, a = −0.44 ± 0.05 and b = −1.70 ± 0.07). There is no obvious systematic trend in the system constants associated with differences in sorbent morphology. These stationary phases are expected to have a low concentration of silanol groups and solvent effects are likely to be of equal or greater importance to silanol activity in controlling retention. Thus, the contribution from dipole-type interactions and hydrogen-bond basicity cannot be assigned to silanol activity alone. In all cases, the sign of the system constants indicates that the stationary phase is a less polar environment than the mobile phase. The system constant ratios (e/v, s/v, a/v and b/v) are more relevant

3.3.1. Neutral silanols The solvation parameter model does not contain a specific term indicated as “silanol activity” but determines the range of interactions normally associated with neutral “silanol activity” as polar interactions together with polar interactions that result from the adsorption of mobile phase components by the interphase region. These interactions and their dependence on mobile phase composition for binary mobile phases are described by system maps [37, 40]. Typical examples for methanol-water and acetonitrile-water mobile phase compositions for an octadecyldimethylsiloxane-bonded silica stationary phase are shown in Figs. 2 and 3, respectively [87]. Major similarities in these plots (and also system maps

Fig. 2. A system map for the octadecylsiloxane-bonded silica column (Kinetex C-18) for small neutral compounds with methanol-water mobile phase compositions. Reproduced from [87] with permission. 213

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

Table 4 System constants for monomeric alkylsiloxane-bonded silica stationary phases with methanol-water (1:1) as the mobile phase at 45 °C. Stationary phase

System constants v

System constant ratios

Reference

e

s

a

b

c

e/v

s/v

a/v

b/v

(i) Endcapped octadecylsiloxane-bonded phases Ascentis C18 2.14 HYPURITY C18 2.08 Betasil C18 2.04 Synergy Hydro-RP 2.10 Kinetex C18 1.97 Kinetex XB-C18 2.07 SunFire C18 2.10 Discovery HS C18 1.96 Kinetex EVO C18 2.14 XTerra MS C18 1.90 XBridge C18 2.17 Chromolith Performance RP-18e 2.13

0.35 0.34 0.33 0.41 0.27 0.33 0.32 0.41 0.24 0.27 0.33 0.43

−0.82 −0.78 −0.78 −0.77 −0.69 −0.75 −0.77 −0.84 −0.70 −0.68 −0.76 −0.83

−0.44 −0.40 −0.44 −0.55 −0.38 −0.37 −0.46 −0.45 −0.39 −0.50 −0.41 −0.48

−1.70 −1.69 −1.79 −1.67 −1.65 −1.63 −1.70 −1.65 −1.76 −1.62 −1.82 −1.68

−0.35 −0.66 −0.28 −0.35 −0.47 −0.62 −0.32 −0.27 −0.58 −0.55 −0.61 −0.69

0.16 0.17 0.16 0.20 0.14 0.16 0.15 0.21 0.11 0.14 0.15 0.20

−0.38 −0.38 −0.38 −0.37 −0.35 −0.36 −0.36 −0.43 −0.33 −0.36 −0.35 −0.39

−0.21 −0.19 −0.21 −0.26 −0.19 −0.18 −0.22 −0.23 −0.18 −0.26 −0.19 −0.23

−0.79 −0.81 −0.88 −0.80 −0.84 −0.79 −0.81 −0.84 −0.82 −0.85 −0.84 −0.79

78 79 80 81 87 88 89 90 91 92 92 94

(ii) Endcapped octylsiloxane-bonded phases Kinetex C8 1.91 XBridge C8 1.79

0.26 0.13

−0.66 −0.56

−0.32 −0.28

−1.56 −1.44

−0.57 −0.57

0.14 0.07

−0.35 −0.31

−0.17 −0.16

−0.82 −0.80

88 93

(iii) Alkylsiloxane-bonded phases with a polar embedded group XBridge Shield RP18 2.04 0.25 −0.53 Synergi Fusion RP 2.01 0.20 −0.58

−0.21 −0.34

−1.96 −1.71

−0.64 −0.56

0.12 0.10

−0.26 −0.29

−0.10 −0.17

−0.96 −0.85

92 81

for identifying selectivity differences than the absolute values of the system constants. Band spacing on two compared columns will vary if significant differences in at least two of the system constant ratios exist [95, 96]. The average system constant ratios for the octadecyldimethylsiloxane-bonded silica stationary phases and the diisobutyloctadecylsiloxane-bonded silica stationary phase are (e/v)AV = 0.16 ± 0.03, (s/v)AV = −0.34 ± 0.07, (a/v)AV = −0.21 ± 0.03 and (b/ v)AV = −0.82 ± 0.03. These can be compared to the mean values for the two octyldimethylsiloxane-bonded silica stationary phases e/ v = 0.11, s/v = −0.33, a/v = −0.17, and b/v = −0.81. The selectivity of the octyldimethylsiloxane stationary phases is similar to the twelve octadecyldimethylsiloxane-bonded silica stationary phases indicating a weak connection between selectivity and chain length. There are large differences in absolute retention for the compared columns but this is accounted for mainly by differences in the phase ratio (c term). The two alkylsiloxane-bonded phases with polar embedded groups are the most dissimilar to the twelve octadecyldimethylsiloxane-bonded silica stationary phases. The XBridge Shield RP18 stationary phase contains internal carbamate groups while Synergi Fusion RP contains co-bonded octadecyldimethylsiloxane-bonded and short-chain siloxane-bonded ligands with terminal polar groups (structure is not disclosed by the manufacturer). XBridge Shield RP18 is significantly more dipolar/polarizable, hydrogen-bond basic but less hydrogen-bond acidic than a typical octadecyldimethylsiloxane-bonded silica stationary phase. Correlation plots of the system constants as a function of mobile phase composition facilitate the identification of selectivity differences and the influence of the mobile phase composition on these differences. For comparison the system constants for XBridge Shield RP18 are compared with those for SunFire C18 (this stationary phase has system constants similar to the average of the twelve octadecyldimethylsiloxane-bonded silica stationary phases in Table 4) in Fig. 4. If the two columns were selectivity equivalent the system constants would fall on a straight line with a slope of one and an intercept of zero. This is only the case for v and e in Fig. 4. The s, a, and b system constants show a systematic change in slope compared with the line representing equivalent selectivity. Thus, for compounds capable of interactions of a dipole-type and/or with appreciable hydrogen-bond acidity or basicity the two columns will exhibit different band spacing depending on the relative magnitude of the complementary solute descriptors. The separation factor for pairs of prototypical compounds have been widely used to characterize interactions associated with neutral silanol groups. Representative methods classified as hydrogen-bonding

Fig. 4. Correlation plot of the system constants for XBridge Shield RP18 and SunFire C18 for methanol-water mobile phase compositions (10–70% v/v methanol). The range for each system constant is indicated by the lower case letters on the plot. The solid line indicates the theoretical result if the two columns were selectivity equivalent.

capacity, hydrophilicity, neutral polar interactions, aromatic selectivity and phenol selectivity are summarized in Table 5 [11, 13–20, 50]. In these examples, and others, little information is given as to why just those compounds were selected for column evaluation or how they are uniquely connected to the column property they are indicated to evaluate. The lack of a clear definition for the test property diminishes their value if the object is to understand how the singular indicated property affects the relative retention of varied compounds on different columns. The data for the solute descriptors in Table 3 can be used to characterize the contribution from different intermolecular interactions to the separation factors. The separation factor for caffeine and phenol is indicated as a measure of hydrogen bonding. The difference in solute descriptors for the two compounds (ΔV = 0.5881. ΔE = 0.867, ΔS = 0.956, ΔA = −0.672 and ΔB = 0.926) indicate that all solute properties contribute to the separation factor as well as the differences in values for the A and B solute descriptors. Given the typical range of system constants it suggests that solute size and hydrogen-bond basicity 214

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

in solute descriptors (ΔV = 0.0822. ΔE = 0.170, ΔS = 0.257, ΔA = 0.703 and ΔB = 0.180), (ΔV = −0.1409. ΔE = −0.028, ΔS = −0.040, ΔA = 0.294 and ΔB = −0.240) and (ΔV = −0.6537. ΔE = −0.019, ΔS = −0.730, ΔA = −0.703 and ΔB = −0.481), respectively. The significant variation in the differences for the solute descriptors indicates that the three tests are unlikely to determine a common stationary phase property. The separation factor for phenol and toluene is expected to be sensitive to differences in hydrogen-bond basicity for varied stationary phases although this might be compromised by the generally larger contribution for hydrogen-bond acidity to retention on typical reversed-phase stationary phases. The separation factors for phenol and benzyl alcohol should be sensitive to changes in hydrogen-bonding for varied stationary phases without distinguishing hydrogen-bond acidity from hydrogen-bond basicity as well as differences in contributions from cavity formation (modest value for ΔV). The separation factor for phenol and dimethyl phthalate is dependent on all interactions except for ΔE and would be difficult to relate to any specific property for varied stationary phases. None of the tests employing separation factors for prototypical compounds are likely to describe a singular characteristic column property unless that property is defined in some way as a weighted contribution from different intermolecular interactions. Such a scale would present considerable difficulty in giving meaning to column variations in the context of column characterization. The hydrophobic-subtraction model has terms representing stationary phase contributions from hydrogen-bond acidity and basicity but does not consider dipole-type interactions as relevant for reversedphase liquid chromatography [24, 97]. The zero point for the hydrogenbonding scales is ethylbenzene (α′ = β′ = 0). For the solute descriptors used in the solvation parameter model ethylbenzene is non‑hydrogenbond acidic but weakly hydrogen-bond basic. The solute descriptors for the hydrophobic-subtraction model have both positive and negative values and compounds without hydrogen-bond donor groups are assigned values for α′ [71]. This is unreasonable if α′ describes only solute hydrogen-bond acidity but may be due to the lack of a natural zero for the scale. In general, there is little correspondence between the A and B descriptors of the solvation parameter model and the α′ and β′ descriptors of the hydrophobic-subtraction model [70, 71, 98, 99]. In addition, the hydrogen-bonding descriptors for the hydrophobic-subtraction model are allowed to assume different values for stationary phase with different topologies to account for steric factors effecting solute-stationary phase interactions, which if accepted as logical, would make redundant any further comparison of the two solute descriptor scales. However, it seems unlikely that the two scales represent comparable measures of hydrogen-bonding properties [97–99]. In the hydrophobic-subtraction model it is assumed that the hydrogen-bond acidity of the stationary phase results from accessible and neutral silanol groups [70] and the hydrogen-bond basicity from the sorption of water molecules by the solvated stationary phase [99]. Vicinal silanol sites are identified as the likely source of enhanced basicity for compounds with adjacent donor and acceptor groups, such as benzoic acids, compared with phenols with only a single interaction site [98]. Since the stationary phase is suggested to possess sites with more than one binding energy the hydrogen-bond acidity of the stationary phase (B value) will be some type of average of the interactions represented by the test compounds, and is expected to vary depending on the selection and number of the test compounds. The hydrogen-bond acidity of the stationary phase is expected to be influenced by the pH of the mobile phase due to a change in the relative population of neutral and ionized accessible silanol groups [99]. Type-A silica alkylsiloxanebonded columns are expected to be more hydrogen-bond acidic and somewhat more hydrogen-bond basic compared with an average type-B alkylsiloxane-bonded silica column [99].

Table 5 Separation factors for prototypical compounds used to characterize column properties associated with neutral silanol groups (all measurements at 40 °C). Property

Separation factor

Typical conditions

Hydrogen bonding Hydrophilicity Neutral polar Aromatic selectivity

Caffeine/phenol Benzyl alcohol/toluene Phenol/ethyl benzoate 1,3-Diinitrobenzene/ toluene Phenol/Dimethyl phthalate

30% 65% 55% 50%

Phenol selectivity

Phenol/benzyl alcohol Phenol/toluene

(v/v) (v/v) (v/v) (v/v)

methanol-water methanol-water methanol-water methanol-water

30% (v/v) methanol-pH 2.7 buffer 30% (v/v) methanol-pH 2.7 buffer 65% (v/v) methanol-ph 2.7 buffer

will dominate the separation factor at the expense of hydrogen-bond acidity. The contribution of dipole-type interactions is also significant. The test result is a composite measure of multiple interactions and the single endpoint value (the separation factor) fails to provide insight into how these variations in interactions contributes to characteristic column properties. Hydrophilicity and neutral polarity would be expected to correlate with the capability of the stationary phase for interactions of a dipole-type (and possibly hydrogen-bonding as the definition of this term is unclear). The difference in solute descriptors for benzyl alcohol and toluene (ΔV = 0.0587. ΔE = 0.198, ΔS = 0.357, ΔA = −0.409 and ΔB = 0.420) and phenol and ethyl benzoate (ΔV = −0.438. ΔE = 0.082, ΔS = −0.125, ΔA = 0.703 and ΔB = −0.134) indicate that the two test determine multiple properties. Since ΔV is near zero for the benzyl alcohol and toluene separation factor (which should be considered a prerequisite for any test employing a separation factor to probe polar interactions) and the difference in solute descriptors for dipole-type and hydrogen-bonding interactions are numerically similar, this test is expected to be sensitive to interactions broadly defined as of a polar-type. However, for reversedphase liquid chromatography the b system constant is generally much larger than the s and a system constants and the separation factor for benzyl alcohol and toluene will likely be dominated by the difference in hydrogen-bond acidity for individual stationary phases. The separation factor for phenol and ethyl benzoate would not be expected to show high correlation with the value for phenol and toluene as the relative contribution from the differences in solute descriptors are dissimilar, and in the case of phenol and ethyl benzoate are significantly dependent on the ease of cavity formation in addition to relatively weak interactions of a polar-type, except for variation in hydrogen-bond basicity. It is proposed to estimate aromaticity (indicated as the differences in π-basicity) as the separation factor for 1,3-dinitrobenzene and toluene [20]. The corresponding differences in solute descriptors (ΔV = 0.2075. ΔE = 0.379, ΔS = 1.200, ΔA = 0 and ΔB = 0.287) indicates a strong dependence on interactions of a dipole-type (although this may not be the intended interaction of the test method) but given the relatively large values of the v and b system constants for reversed-phase liquid chromatography the separation factor is likely to be significantly influenced by the difference in size, ΔV, and hydrogen-bond basicity ΔB of the test compounds as well. Phenol selectivity is indicated as the enhanced retention of phenols compared to non-phenolic compounds [20]. Why the retention of phenols would be enhanced compared to other neutral compounds is not clear since phenols do not possess unique interactions that are different from other compounds. Phenols have interactions characteristic of phenols but no special interactions. It is difficult to describe this as an enhancement any more than it could be argued for other compounds belonging to a family group. Phenol selectivity is indicated by the separation factors for phenol and toluene, phenol and benzyl alcohol and phenol and dimethyl phthalate with the following differences

3.3.2. Ionized silanols The origin of ionized silanol groups is not absolutely clear. One 215

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

studies of column characterization [46]. A screening procedure for bases is applied in the construction of system maps to evaluate whether cation-exchange interactions prevent the co-modeling of weak bases with neutral compounds by Eq. (2) [40, 58, 78–81, 87–93]. A model is first constructed for varied neutral compounds and then used to predict the retention factors for the weak bases. If electrostatic interactions are not important the difference between the experimental retention factors for the bases and their neutral-model predicted values (Δlog k) will be small compared with the prediction error for the neutral-compound model (standard error of the estimate) and will have alternating signs. Larger differences on the same scale with a systematic positive sign indicate electrostatic interactions are important. It is noteworthy that for weak bases electrostatic interactions are often unimportant for methanol-water mobile phases while significant for acetonitrile-water mobile phases for the same column. For compounds which are partially ionized the solvation parameter model is not expected to accurately describe their retention under reversed-phase separation conditions unless their ionization is suppressed by a change in pH. Cation-exchange capacity has been estimated by the retention factor for specific bases or the separation factor of a base and a neutral compound at acid or near neutral pH in aqueous or buffered binary mobile phases [14–20, 50, 100]. The separation factor for benzylamine and phenol has been widely used [18, 19]. The test is usually performed at two pH values with the separation factor indicated as a measure of the total ion-exchange capacity at pH = 7.6 and the acidic ion-exchange capacity at pH = 2.7 with a 30% (v/v) methanol-buffer mobile phase at 40 °C [19]. Separation factors on alkylsiloxane-bonded silica stationary phases are generally larger at pH = 7.6. Benzylamine is assumed to be fully ionized at both pH values and the variation in separation factors to result from the change in concentration of acidic silanol groups. The separation factor, however, depends on non-ionic interactions as well as ionic interactions. The former are related to the difference in intermolecular interactions for the protonated benzylamine and phenol with the stationary phase. It is not clear that electrostatic interactions are the dominant contribution to the separation factor for benzylamine and phenol for typical measurement conditions. The Engelhardt test [14] employed retention factors for aniline and N.N-dimethylaniline and the separation of o-, m- and p-toluidine in a aqueous-methanol (49:51 or 55:45) to evaluate silanol activity. Stationary phases with weak silanol activity where characterized by poor selectivity of the toluidine isomers (with the assumption that o-, m- and p-toluidine have identical hydrophobicity and differ only in their pKb values, 4.4, 4.7. and 5.1, respectively), phenol eluted after aniline, and the asymmetry factors for aniline and phenol < 1.3 [14, 50]. Engelhardt indicated that the retention factor for the stronger base N,Ndimethylaniline was largely influenced by the hydrophobicity of the stationary phase and was not a good test compound for silanol group activity [14]. As an example of tests utilizing tests compounds thought to be more stringent for silanol activity Neue et al. [9, 15] proposed the use of the separation factors for propranolol and chlorpheniramine at acid pH and propranolol and amitriptyline at near neutral pH with toluamide (or in some cases acenaphthene) as the neutral reference compound. For varied stationary phases they observed a good correlation for the test compounds at acidic pH and a reasonable correlation at neutral pH. The authors characterized these results as unexpected but the test method would seem to be compromised as a measure of silanol activity. The large difference in physicochemical properties for the basic test compounds and toluamide are likely to overwhelm the contribution of electrostatic interactions to the retention factors.

theory has it that these arise predominantly from the activation of neutral silanols by metal impurities embedded in the silica substrate [3, 82, 86]. The general trend towards increasing use of high purity silica (type-B silica) substrates and improvements in bonding and endcapping chemistry have resulted in more inert column packings compared to the recent past. As columns have become more inert test methods have migrated from weak bases as test compounds to more stringent tests employing stronger bases [15, 20, 82, 86, 100]. Electrostatic interactions might contribute favorably to the separation of mixtures containing bases but usually of greater concern is that electrostatic interactions between the negatively charged stationary phase and positively charged bases might result in poor peak shapes, poor reproducibility, mass-dependent retention, and possibly partial adsorption of these compounds. Since many compounds of pharmaceutical interest are nitrogen bases this possibility has remained an important consideration for column selection in the life sciences. Typical column test methods for electrostatic interactions employ either aqueous or buffered mobile phases with a pH close to neutral (pH = 7) or an acidic mobile phase (pH = 2–3) as reference values. The selected measurement conditions and their stability are just as important as the selection of test compounds. As well as pH, cation-exchange interactions for bases depend on the ionic strength of the mobile phase and the identity of the buffer ions. At neutral pH the concentration of ionized silanol groups will be higher than at acidic pH due to suppression of silanol ionization. For these reasons different cation-exchange capacity tests are not interchangeable, especially for the ranking of columns with weak cationexchange properties [15, 16, 50]. Only the hydrophobic-subtraction model integrates the contribution from cation-exchange interactions into the general retention model and identifies the electrostatic interaction as an additional contribution to retention compared to neutral compounds [23, 70, 99, 101, 102]. There is some ambiguity concerning the descriptors for non-ionic interactions, whether they are the same, or can be the same, for the ionized and neutral form of a compound. The solute descriptor κ′ for the test compounds is determined by the charge on an ionized molecule and should be zero for neutral compounds and one for fully ionized compounds with a positive sign for cations and a negative sign for anions. Of the compounds suggested for column characterization only amitriptyline and nortriptyline have large positive values and must dominate the associated column cation-exchange capability using the hydrophobicsubtraction model. The neutral test compounds have small κ′ values with both positive and negative signs that do not make sense, but are unlikely to have much influence on the estimation of the C term. Negative values of κ′ are reasonable for carboxylic acids and allow the correction of retention factors for ion exclusion of anions by the negatively charged stationary phase [71]. Cation exchange capacity is observed to increase with the pH of the mobile phase and is approximately described by the relationship

C7.0 = C2.8 + log (k 7/ k2.8)

(4)

where C is the column cation-exchange capacity relative to an average hypothetical octadecyldimethylsiloxane-bonded type-B silica stationary phase at the reference pH of 7.0 or 2.8 and k7 and k2.8 are the retention factors for berberine (a quaternary ammonium compound) at the reference mobile phase pH values, respectively. For bases the contribution from cation-exchange interactions is often the second most important term in adjusting selectivity using Eq. (1), but is not important for the separation of neutral compounds. Cation-exchange properties are difficult to rationalize alongside those of reversed-phase liquid chromatography since the experimental variables affecting cation-exchange interactions are different to those used to control reversedphase separations while their interactions tend to confound simple models. The standard form of the solvation parameter model is not parameterized for ionic interactions. An extended form of the model has been developed for fully ionized compounds but has not been applied to

3.4. Metal chelation activity Metal ions residing in the silica substrate have been implicated in enhancing the acidity of silanol groups (see Section 3.3). In addition, metal impurities embedded or adsorbed on the silica surface have been associated with strong adsorption of chelating compounds resulting in 216

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

poor peak shapes and altered retention. Historic tests for metal impurities include dihydroxynaphthalene efficiency ratio test (DERT), the metal factor test, and asymmetry factor methods utilizing acetylacetone [16, 17, 25, 103, 104]. The DERT test compares the peak efficiency measured at the base of the two regioisomers 2,3- and 2,7-dihydroxynaphthalene with 25% (v/v) acetonirile- pH 7.2 buffer as mobile phase. Only the 2,3-isomer is capable of chelating with metal ions. Columns with low metal ion contamination have DERT values close to 1.0. The metal factor test is based on the ratio of the asymmetry factors for 2,2′-bipyridyl and 4,4′-bipyridyl with 49% (v/v) methanol-water as the mobile phase at 40 °C. 2,2′-Bipyridyl forms complexes with metal ions while 4,4′-bipyridyl is non-complexing but otherwise has similar stationary phase interactions to 2,2′-bipyridyl. For columns with a low metal content a metal factor close to 1 (or 100 if a scale factor of 100 is applied) is expected. Chelate formation with acetylacetone results in large asymmetry factors compared to values for low-metal contamination columns. Metal contamination tests have largely been discontinued in recent years as a component of column characterization studies. The general observation is that the results tend to reflect column history and storage conditions and do not accurately reflect the properties of the original stationary phase. The primary source of metal contamination in most cases is the result of mobile phase contamination and system components (particularly frits and screens) [103, 104].

[2] C.F. Poole, The Essence of Chromatography, Elsevier, Amsterdam, 2003. [3] K.K. Unger, R. Skudas, M.M. Schutto, Particle packed columns and monolithic columns in high-performance liquid chromatography – comparison and critical appraisal, J. Chromatogr. A 1184 (2008) 393–415. [4] J.J. Pesek, R.L. Boysen, M.T.W. Hearn, M.T. Matyska, Hydride-based HPLC stationary phases: a rapidly evolving technology for the development of new bioanalytical methods, Anal. Methods 6 (2014) 4496–4503. [5] R. Hayes, A. Ahmed, T. Edge, H. Zhang, Core-shell particles: preparation, fundamentals and applications in high performance liquid chromatography, J. Chromatogr. A 1357 (2014) 36–57. [6] E.M. Borges, M.A. Rostagno, N.A.A. Meireles, Sub-2 μm fully porous and partially porous (core-shell) stationary phases for reversed-phase liquid chromatography, RSC Adv. 4 (2014) 14–20. [7] T.H. Walter, R.W. Andrews, Recent innovations in UHPLC columns and instrumentation, Trends Anal. Chem. 63 (2014) 14–20. [8] M. Kele, G. Guiochon, Repeatability and reproducibility of retention data and band profiles on reversed-phase liquid chromatographic columns II. Results obtained with symmetry C18 columns, J. Chromatogr. A 830 (1999) 55–79. [9] U.D. Neue, E. Serowik, P. Iraneta, B.A. Alden, T.H. Walter, Universal procedure for assessment of the reproducibility and the characterization of silica-based reversedphase packings I. Assessment of the reproducibility of reversed-phase packings, J. Chromatogr. A 849 (1999) 87–100. [10] P. Myers, Column Technology, I.D. Wilson, E.R. Adlard, M. Cooke, C.F. Poole (Eds.), Encyclopedia of Separation Science, vol. 2, Academic Press, London, 2000, pp. 567–573. [11] E. Lesellier, C. West, Description and comparison of chromatographic tests and chemometric methods for packed columns, J. Chromatogr. A 1158 (2007) 329–360. [12] M. Kele, G. Guiochon, Repeatability and reproducibility of retention data and band profiles in chromatography on reversed-phase liquid chromatography columns I. Experimental protocol, J. Chromatogr. A 830 (1999) 41–54. [13] K. Kimata, K. Iwaguchi, S. Onishi, K. Jinno, R. Eksteen, K. Hosoya, M. Araki, N. Tanaka, Chromatographic characterization of silica-C-18 packing materials–correlation between a preparation method and retention behavior of stationary phase, J. Chromatogr. Sci. 27 (1989) 721–728. [14] H. Engelhardt, H. Low, W. Gotzinger, Chromatographic characterization of silicabased reversed phases, J. Chromatogr. 544 (1991) 311–379. [15] U.D. Neue, B.A. Alden, T.H. Walter, Universal procedure for the assessment of the reproducibility and the classification of silica-based reversed-phase packings II. Classification of reversed-phase packings, J. Chromatogr. A 849 (1999) 101–116. [16] T. Nemeth, E. Haghedooren, B. Noszal, J. Hoogmartens, E. Adams, Three methods to characterize reversed-phase liquid chromatographic columns applied to pharmaceutical separations, J. Chemom. 22 (2008) 168–175. [17] E. Haghedooren, T. Janssens, R. Nijs, S.K. Park, E. Farkas, S. Dragovic, B. Noszal, J. Hoogmartens, E. Adams, Selecting a suitable LC column for pharmaceutical separations using a column characterization system, J. Liq. Chromatogr. Relat. Technol. 32 (2009) 747–771. [18] M.R. Euerby, P. Petersson, Chromatographic classification and comparison of commercially available reversed-phase liquid chromatographic columns using principal component analysis, J. Chromatogr. A 994 (2003) 13–36. [19] M.R. Euerby, P. Petersson, Chromatographic classification and comparison of commercially available reversed-phase liquid chromatographic columns containing polar embedded groups/amino endcappings using principal component analysis, J. Chromatogr. A 1088 (2006) 1–15. [20] M.R. Euerby, M. James, P. Petersson, Practical applications of the “Tanaka” stationary phase characterization methodology using ultra high performance liquid chromatographic conditions, J. Chromatogr. A 1228 (2012) 165–174. [21] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, R.G. Wolcott, P.W. Carr, Column selectivity in reversed-phase liquid chromatography I. A general quantitative relationship, J. Chromatogr. A 961 (2003) 171–193. [22] L.R. Snyder, J.W. Dolan, D.H. Marchand, P.W. Carr, The hydrophobic-subtraction model in reversed-phase column selectivity, Adv. Chromatogr. 50 (2015) 297–376. [23] D.H. Marchand, L.R. Snyder, J.W. Dolan, Characterization and applications of reversed-phase column selectivity based on the hydrophobic-subtraction model, J. Chromatogr. A 1191 (2008) 2–20. [24] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, P.W. Carr, Column selectivity in reversed-phase liquid chromatography II. Effect of change of conditions, J. Chromatogr. A 961 (2002) 195–215. [25] D.H. Marchand, K. Croes, J.W. Dolan, L.R. Snyder, R.A. Henry, K.M.R. Kallury, S. Waite, P.W. Carr, Column selectivity in reversed-phase liquid chromatography VIII. Phenylalkyl- and fluoro-substituted columns, J. Chromatogr. A 1062 (2005) 65–78. [26] M.H. Abraham, J.M.R. Gola, R. Kumarsingh, J.E. Commetto-Muniz, W.S. Cain, Connection between chromatographic data and biological data, J. Chromatogr. B 745 (2000) 103–115. [27] C.F. Poole, A.D. Gunatilleka, S.K. Poole, In search of a chromatographic model for biopartitioning, Adv. Chromatogr. 40 (2000) 159–230. [28] S. Endo, K.-U. Goss, Applications of polyparameter linear free energy relationships in environmental chemistry, Environ. Sci. Technol. 48 (2014) 12477–12491. [29] C.F. Poole, T.C. Ariyasena, N. Lenca, Estimation of the environmental properties of compounds from chromatographic measurements and the solvation parameter model, J. Chromatogr. A 1317 (2013) 85–104. [30] M.H. Abraham, C.F. Poole, S.K. Poole, Classification of stationary phases and other materials by gas chromatography, J. Chromatogr. A 842 (1999) 79–114. [31] C.F. Poole, S.K. Poole, Separation characteristics of wall-coated open-tubular

4. Conclusions A variety of column tests based on retention or separation factors of prototypical compounds or multivariate methods employing varied compounds and applicable models to identify column interactions have been developed. Common column properties determined this way include hydrophobicity (or hydrophobic selectivity), silanol activity (hydrogen-bonding interactions with neutral silanol groups), polar interactions, steric resistance, ion-exchange capacity (electrostatic interactions with ionized silanol groups) and metal contamination. Competing scales used to rank the properties of alkysiloxane-bonded silica stationary phases notionally by the same property often show poor agreement. This arises because many of the terms used to describe column properties are poorly defined and the test compounds associated with these properties are equally poorly selected. Most scales based on prototypical compounds involve multiple stationary phase interactions and fail to rank columns by singular well-defined properties. The multivariate models are more hopeful as they extract characteristic column properties from retention factors for multiple compounds according to a model with a fixed number of pre-defined column interaction terms. The hydrophobic-subtraction model is limited specifically to column characterization and has been used to characterize a large number of columns. The solvation parameter model is a fundamental model with numerous applications in separation science including column characterization. The hydrophobic-subtraction model and solvation parameter model contain different terms and are only comparable in a broad sense with a limited possibility for a term by term comparison. All current methods are limited by the difficulty of parameterizing the heterogeneous structure of the interphase (active stationary phase) region and the uncertainty of the distribution mechanism for varied compounds transferred from the mobile to stationary phase. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jchromb.2018.06.011. References [1] S. Fanali, P.R. Haddad, C.F. Poole, M.-L. Riekkola (Eds.), Liquid Chromatography, 2nd edn, vol 1 & 2, Elsevier, Amsterdam, 2017.

217

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

performance liquid chromatography, Environ. Sci. Technol. 9 (2012) 564–570. [64] I. Kubik, W. Struck-Lewicka, R. Kaliszan, P. Wiezling, Simultaneous determination of hydrophobicity and dissociation constant for a large set of compounds by gradient reverse phase high performance chromatography-mass spectrometry, J. Chromatogr. A 1416 (2015) 31–37. [65] M. Roses, E. Bosch, C. Rafols, E. Fuguet, Chromatographic hydrophobicity index (CHI), Adv. Chromatogr. 50 (2012) 377–414. [66] K.-Y. Law, Water-surface interactions and definitions for hydrophobicity, hydrophilicity and superhydrophilicity, Pure Appl. Chem. 87 (2015) 759–765. [67] Y.C. Guillaume, C. Guinchard, Participation of cluster species in the solvation mechanism of a weak polar solute in a methanol-water mixture over a 0.2–0.7 water fraction range: high-performance liquid chromatography study, Anal. Chem. 70 (1998) 608–615. [68] Y. Marcus, The structure of and interactions in binary acetonitrile + water mixtures, J. Phys. Org. Chem. 25 (2012) 1072–1082. [69] S.C. Moldoveanu, E. Caiali, V. David, Results from solvophobic theory applied to methylene selectivity in reversed phase HPLC, J. Liq. Chromatogr. Relat. Technol. 41 (2018) 24–32. [70] J. Lehotay, B. Buszewski, Study of RP HPLC retention behaviours in analysis of carotenoids, Chromatographia 77 (2014) 1047–1057. [71] J.J. Gilroy, J.W. Dolan, L.R. Snyder, Column selectivity in reversed-phase liquid chromatography IV. Type-B alkyl-silica columns, J. Chromatogr. A 1000 (2003) 757–778. [72] N.G. Wilson, J. Gilroy, J.W. Dolan, L.R. Snyder, Column selectivity in reversedphase liquid chromatography VI. Columns with embedded or endcapped polar groups, J. Chromatogr. A 1026 (2004) 91–100. [73] Y. Wen, M. Telebi, R.I.J. Amos, R. Szucs, J.W. Dolan, C.A. Pohl, P.R. Haddad, Retention prediction in reversed phase high performance liquid chromatography using quantitative structure-retention relationships applied to the hydrophobic subtraction model, J. Chromatogr. A 1541 (2018) 1–11. [74] K.A. Lippa, C.A. Rimmer, L.C. Sander, Shape selectivity in reversed-phase liquid chromatography, Adv. Chromatogr. 46 (2008) 235–303. [75] J.O. Ona-Ruales, L.C. Sander, W.B. Wilson, S.A. Wise, Revisiting shape selectivity in liquid chromatography for polycyclic aromatic hydrocarbons (PAHs) – six-ring and seven-ring cata-condensed PAH isomers of molecular mass 328 Da and 378 Da, Anal. Bioanal. Chem. 410 (2018) 885–896. [76] E.M. Borges, How to select equivalent and complementary reversed phase liquid chromatography columns from column characterization databases, Anal. Chim. Acta 807 (2014) 143–152. [77] P.W. Carr, J.W. Dolan, U.D. Neue, L.R. Snyder, Contributions to reversed-phase column selectivity. I. Steric interaction, J. Chromatogr. A 1218 (2011) 1724–1742. [78] C.F. Poole, H. Ahmed, W. Kiridena, C. DeKay, W.W. Koziol, Contribution of steric repulsion to retention on an octadecylsiloxane-bonded silica stationary phase in reversed-phase liquid chromatography, Chromatographia 62 (2005) 553–561. [79] C.F. Poole, W. Kiridena, C. DeKay, W.W. Koziol, R.D. Rosencrans, Insights into the retention mechanism on an octadecylsiloxane-bonded silica stationary phase (HYPURITY C18) in reversed-phase liquid chromatography, J. Chromatogr. A 1115 (2006) 133–141. [80] C.F. Poole, H. Ahmed, W. Kiridena, C. DeKay, W.W. Koziol, Insights into the retention mechanisms on perfluorohexylpropylsiloxane-bonded (Fluophase-RP) and octadecylsiloxane-bonded (Betasil C18) stationary phases based on the same silica substrate in RP-LC, Chromatographia 65 (2007) 127–139. [81] S.N. Atapattu, C.F. Poole, Factors affecting the interpretation of selectivity on Synergi reversed-phase columns, Chromatographia 71 (2010) 185–193. [82] S.D. Rogers, J.G. Dorsey, Chromatographic silanol activity test procedures: the quest for a universal test, J. Chromatogr. A 892 (2000) 57–65. [83] U.D. Neue, J.E. O'Gara, A. Mandez, Selectivity in reversed-phase separations – influence of the stationary phase, J. Chromatogr. A 1127 (2006) 161–174. [84] P. Vajda, S. Bocian, B. Buzewski, A. Felinger, Effect of polar interactions on the nonlinear behavior of phenol and aniline in reversed-phase liquid chromatography, J. Chromatogr. A 1228 (2012) 155–164. [85] F. Gritti, G. Guiochon, Characteristics of the adsorption mechanism of acidic-basic compounds with two pK(a) in reversed-phase liquid chromatography, J. Chromatogr. A 1216 (2009) 6917–6930. [86] D.V. Mccalley, The challenges of the analysis of basic compounds by high-performance liquid chromatography: some possible approaches for improved separations, J. Chromatogr. A 1217 (2010) 858–880. [87] S.N. Atapattu, C.F. Poole, M.B. Praseuth, System maps for the retention of small neutral compounds on a superficially porous particle column in reversed-phase liquid chromatography, J. Chromatogr. A 1468 (2016) 250–256. [88] S.N. Atapattu, C.F. Poole, M.B. Praseuth, Insights into the retention mechanism of small neutral compounds on actylsiloxane-bonded and diisobutyloctadecylsiloxane-bonded silica stationary phases in reversed-phase liquid chromatography, Chromatographia 81 (2018) 373–385. [89] W. Kiridena, S.N. Atapattu, C.F. Poole, W.W. Koziol, System maps for RP-LC on an octadecylsiloxane-bonded silica stationary phases (SunFire C18), Chromatographia 68 (2008) 11–17. [90] W. Kiridena, C. DeKay, W.W. Koziol, Z. Ali, H. Ahmed, C.F. Poole, Insights into the retention mechanism on a pentafluorophenylpropylsiloxane-bonded silica stationary phase (Discovery HS F5) in RP-LC, Chromatographia 63 (2006) 407–417. [91] S.N. Atapattu, C.F. Poole, M.B. Praseuth, System maps for retention of small neutral compounds on a superficially porous ethyl-bridged octadecylsiloxanebonded silica stationary phase in reversed-phase liquid chromatography, Chromatographia 80 (2017) 1279–1286. [92] W. Kiridena, C.F. Poole, S.N. Atapattu, J. Qian, W.W. Koziol, Comparison of the separation characteristics of the organic-inorganic hybrid octadecyl stationary

columns for gas chromatography, J. Chromatogr. A 1184 (2008) 254–280. [32] C.F. Poole, N. Lenca, Gas chromatography on wall-coated open-tubular columns with ionic liquid stationary phases, J. Chromatogr. A 1357 (2014) 87–109. [33] S.K. Poole, C.F. Poole, Quantitative structure-retention (property) relationships in micellar electrokinetic chromatography, J. Chromatogr. A 1182 (2008) 1–24. [34] C. West, E. Lemesson, S. Berlin, P. Hennig, E. Lesellier, An improved classification of stationary phases for ultra-performance supercritical fluid chromatography, J. Chromatogr. A 1440 (2016) 212–228. [35] C.F. Poole, Partition constant database for totally organic biphasic systems, J. Chromatogr. A 1527 (2017) 18–32. [36] C.F. Poole, T. Karunasekara, Solvent classification for chromatography and extraction, J. Sep. Sci. 25 (2012) 190–199. [37] C.F. Poole, S.K. Poole, Column selectivity from the perspective of the solvation parameter model, J. Chromatogr. A 965 (2002) 263–299. [38] M.F. Vitha, P.W. Carr, The chemical interpretation and practice of linear solvation energy relationships in chromatography, J. Chromatogr. A 1126 (2006) 143–194. [39] S. Studzinska, B. Buszewski, Linear solvation energy relationships in the determination of specificity and selectivity of stationary phases, Chromatographia 75 (2012) 1235–1246. [40] C.F. Poole, N. Lenca, Applications of the solvation parameter model in reversedphase liquid chromatography, J. Chromatogr. A 1486 (2017) 2–19. [41] M.H. Abraham, A.M. Zissmos, Determination of sets of solute descriptors from chromatographic measurements, J. Chromatogr. A 1037 (2004) 29–47. [42] C.F. Poole, S.N. Atapattu, S.K. Poole, A.K. Bell, Determination of solute descriptors by chromatographic methods, Anal. Chim. Acta 652 (2009) 37–53. [43] D. Bolliet, C.F. Poole, Influence of temperature on retention and selectivity in reversed-phase liquid chromatography, Analyst 123 (1998) 295–299. [44] W. Kiridena, C.F. Poole, Structure-driven retention model for optimization of ternary solvent systems in reversed-phase liquid chromatography, Chromatographia 48 (1998) 607–614. [45] W. Kiridena, C.F. Poole, W.W. Koziol, Effect of solvent strength and temperature on retention for a polar-endcapped, octadecylsiloxane-bonded silica stationary phase with methanol-water mobile phases, J. Chromatogr. A 1060 (2004) 177–185. [46] M.H. Abraham, W.E. Acree, Descriptors for ions and ion-pairs in linear free energy relationships, J. Chromatogr. A 1430 (2016) 2–14. [47] M. Roses, D. Bolliet, C.F. Poole, Comparison of solute descriptors for predicting retention of ionic compounds (phenols) in reversed-phase liquid chromatography using the solvation parameter model, J. Chromatogr. A 829 (1998) 29–40. [48] C.F. Poole, Influence of solvent effects on retention in reversed-phase liquid chromatography. Chromatographia (in press). [49] D. Visky, Y. vander Heyden, T. Ivanyi, P. Baten, J. De Beer, Z. Kovacs, B. Noszal, E. Roets, D.L. Massart, J. Hoogmartens, Characterization of reversed-phase liquid chromatographic columns by chromatographic tests. Evaluation of 36 test parameters, repeatability, reproducibility and correlation, J. Chromatogr. A 977 (2002) 39–58. [50] E.M. Borges, M.R. Euerby, C.H. Collins, Comparison of classical chromatographic tests with chromatographic test applied to stationary phases prepared by thermal immobilization of poly(methyloctylsiloxane) onto silica, Anal. Bioanal. Chem. 404 (2012) 2985–3002. [51] D. Sykora, J. Vozka, E. Tesarova, Chromatographic methods enabling the characterization of stationary phases and prediction of retention in high-performance liquid chromatography and supercritical fluid chromatography, J. Sep. Sci. 39 (2016) 115–131. [52] F. Andric, K. Heberger, How to compare separation selectivity of high-performance liquid chromatographic columns properly, J. Chromatogr. A 1488 (2017) 45–56. [53] C.F. Poole, An interphase model for retention in chromatography, J. Planar Chromatogr. 28 (2015) 98–105. [54] Y.V. Kazakevich, R. Lobrutto, P. Chan, Interpretation of the excess adsorption isotherms on the surface of reversed-phase adsorbents – effect on the analyte retention, J. Chromatogr. A 913 (2001) 75–87. [55] F. Gritty, Y.V. Kazakevich, G. Guiochon, Effect of the surface coverage of endcapped C-18 silica on the excess adsorption isotherms of commonly used organic solvents from water in reversed-phase liquid chromatography, J. Chromatogr. A 1169 (2007) 111–124. [56] A. Vailaya, C. Horvath, Retention in reversed-phase liquid chromatography: partition or adsorption, J. Chromatogr. A 829 (1998) 1–37. [57] M. Wang, J. Mallette, J.F. Parcher, Fundamental mathematical relationships for retention volumes in liquid chromatography, J. Chromatogr. A 1213 (2008) 105–109. [58] C.F. Poole, S.K. Poole, Foundations of retention in partition chromatography, J. Chromatogr. A 1216 (2009) 1530–1550. [59] J. Grdadolnik, F. Meerzel, F. Avbelj, Origin of hydrophobicity and enhanced water hydrogen-bond strength near purely hydrophobic solutes, PNAS 114 (2017) 322–327. [60] J.G. Dorsey, M.G. Khaledi, Hydrophobicity estimations by reversed-liquid chromatography. Importance for biological partitioning processes, J. Chromatogr. A 656 (1993) 485–499. [61] R. Kaliszan, Quantitative structure-retention relationships applied to reversedphase high-performance liquid chromatography, J. Chromatogr. A 656 (1993) 417–435. [62] R. Kaliszan, Effect of separation conditions on chromatographic determination of hydrophobicity of acidic xenobiotics, J. Chromatogr. B 717 (1998) 125–134. [63] P. de Voogt, I. Zurano, P. Serne, J.J.H. Haffka, Experimental hydrophobicity parameters of perfluorinated alkylated substances from reversed-phase high-

218

Journal of Chromatography B 1092 (2018) 207–219

C.F. Poole

[93]

[94]

[95]

[96]

[97]

[98]

Chromatogr. A 1395 (2015) 57–64. [99] J.J. Gilroy, J.W. Dolan, P.W. Carr, L.R. Snyder, Column selectivity in reversedphase liquid chromatography V. Higher metal-content (type-A) alkyl-silica columns, J. Chromatogr. A 1026 (2004) 77–87. [100] K. Okusa, Y. Suita, M. Tahara, T. Ikegami, N. Tanaka, Test compounds for detecting the silanol effect on the elution of ionized amines in reversed-phase LC, J. Sep. Sci. 33 (2010) 348–358. [101] D.H. Marchand, P.W. Carr, D.V. McCalley, U.D. Neue, J.W. Dolan, L.R. Snyder, Contributions to reversed-phase column selectivity. II. Cation exchange, J. Chromatogr. A 1218 (2011) 7110–7129. [102] T. Undin, J. Samuelsson, A. Tomcrona, T. Fornstedt, Evaluation of a combined linear-nonlinear approach for column characterization using modern alkalinestable columns as model, J. Sep. Sci. 36 (2013) 1753–1761. [103] H. Engelhardt, T. Lobert, Chromatographic determination of metallic impurities in reversed-phase HPLC columns, Anal. Chem. 71 (1999) 1885–1892. [104] M.R. Euerby, C.M. Johnson, I.D. Rushin, D.A.S.S. Tennekoon, Investigation into the epimerization of tipredane ethyl sulfoxide diastereomers during chromatographic analysis on reversed phase silica. 2. The involvement of metals in commercially available C-18 silicas, J. Chromatogr. A 705 (1995) 229–245.

phases XTerra MS C18 and XBridge C18 and Shield RP18 in RPLC, Chromatographia 60 (2007) 453–460. W. Kiridena, S.N. Atapattu, C.F. Poole, W.W. Koziol, Comparison of the separation characteristics of the organic-inorganic hybrid stationary phases XBridge C8 and Phenyl and XTerra Phenyl in RP-LC, Chromatographia 68 (2008) 491–500. Y. Chu, C.F. Poole, System maps for retention of neutral organic compounds under isocratic conditions on a reversed-phase monolithic column, J. Chromatogr. A 1001 (2003) 113–121. M.H. Abraham, M. Roses, C.F. Poole, S.K. Poole, The characterization of reversedphase high performance liquid chromatographic C-18 stationary phases, J. Phys. Org. Chem. 10 (1997) 358–368. A. Wang, P.W. Carr, Comparative study of linear solvation energy relationship, linear solvent strength theory, and typical conditions model for retention prediction in reversed-phase liquid chromatography, J. Chromatogr. A 965 (2002) 3–23. K. Croes, A. Steffens, D.H. Marchand, L.R. Snyder, Relevance of π-π and dipoledipole interactions for retention on cyano and phenyl columns in reversed-phase liquid chromatography, J. Chromatogr. A 1098 (2005) 123–130. P.W. Crr, J.W. Dolan, J.G. Dorsey, L.R. Snyder, J.J. Kirkland, Contributions to reversed phase column selectivity III. Column hydrogen-bond basicity, J.

219