Characterization of a Novel Pyridinium Bromide Surface Confined Ionic Liquid Stationary Phase for High Performance Liquid Chromatography under Normal Phase Conditions via Linear Solvation Energy Relationships

Abstract

Utilizing Linear Solvation Free Energy Relationship methodology, a novel pyridinium bromide surface confined ionic liquid (SCIL) stationary phase was characterized under normal phase High Performance Liquid Chromatographic conditions. A limited set of neutral aromatic probe solutes were utilized to rapidly assess the utility of the LSER model, using mobile phases of hexane modified with 2-propanol. The excellent correlation of the global fit across the mobile phase composition range used in this study for the experimental and calculated retention values (R² = 0.994) indicates that the LSER model is an appropriate model of characterizing this polar bonded phase under normal phase conditions. For a limited subset of compounds, retention on the pyridinium bromide SCIL stationary phase is more highly correlated with that obtained on a cyano column than on a diol column under NP conditions.

1. Introduction

The distinct merits of reversed phase (RP) high performance liquid chromatography (HPLC) (e.g., broad applicability, general robustness) position it as the dominant mode of liquid chromatography. The introduction of new reversed phase media reinforces this dominance [1]. Recently, stationary phases containing covalently bound surface confined ionic liquid (SCIL) groups have been reported in the literature [2,3,4,5,6]. Their ability to separate individual classes of compounds under reversed phase conditions, and the physical characterization of these SCIL phases have been the primary focus of these reports [3].

Although a significant hurdle to widespread application of normal phase chromatography resides in reproducibility issues arising from energetically inhomogenous surfaces [7], bonded ligands such as aminopropyl [8,9] or cyano [10] provide potential electron donor/acceptor, dipole-dipole or hydrogen bonding interactions available under normal phase conditions that may offer advantages for the separation of polar analytes due to significant intermolecular interactions with the polar stationary phase [7]. Hence, the retention mechanism of the SCIL phase under normal phase conditions is of interest. In this work, the retention mechanism of a novel pyridinium bromide stationary phase is investigated using linear solvation energy relationships (LSER) under normal phase conditions.

Linear Solvation Energy Relationships (LSER) have been used extensively to elucidate retention mechanisms in reversed phase liquid chromatography [2,3,11,12]. In contrast, LSER-based characterization of normal phase chromatography is far less common. The LSER methodology [13,14,15], relating retention of an analyte on a stationary phase (log k) to a linear relationship of fundamental solute descriptors (Abraham’s solute descriptors) [11], usually takes the form of

$$\log \mathsf{k}=r{\mathsf{R}}_2+s{\pi }_2^{\ast }+b\sum {\beta }_2^{\mathrm{H}}+a\sum {\alpha }_2^{\mathrm{H}}+v{\mathsf{V}}_2+\log {\mathsf{k}}_{\mathsf{o}}$$ (1)

Recently, a different nomenclature has been adopted for the solute descriptors to simplify the model expressions [12,16,17], where the revised LSER equation takes the form of

$$\log \mathsf{k}=e\mathsf{E}+s\mathsf{S}+b\mathsf{B}+a\mathsf{A}+v\mathsf{V}+\log {\mathsf{k}}_{\mathsf{o}}$$ (2)

Specifically, R₂ or E represent the excess molar refraction, which reflects polarizability contributions from π- and n-electrons; π^*₂ or S are the solute dipolarity/polarizability; Σα^H₂ or A and Σβ^H₂ or B are the overall solute hydrogen bond acidity and basicity, respectively, which scale as the hydrogen bonding propensity of a solute to surrounding solvent molecules; V₂ or V are McGowan’s characteristic volume. The log k_o term is thought to include the phase ratio of the chromatographic system [2,3]. It should be noted that both of the stated models, Equations 1 and 2, fundamentally describe retention based upon global free energy principles directly related to the intermolecular interactions that occur between an analyte and the stationary/mobile phase system.

In the LSER equations, the coefficients (e, s, b, a, v and log k_o), extracted from multiple linear regression analysis of the retention data, represent the difference of a specific interaction of the solutes between the stationary and mobile phases [18,19]. Overall, both the sign and the magnitude of the coefficients are critical in assessing whether the stationary or the mobile phase displays the greater interaction with the solute via a specific interaction mode.

While used extensively in gas chromatography and reversed phase HPLC [11,12], the application of the LSER method to normal phase HPLC [8,13,14,15,20,21] has been somewhat sparse, perhaps because of the more limited application range of normal phase HPLC relative to gas chromatography and reversed phase HPLC. Furthermore, because the retention mechanism of silica phases is based on adsorption, the LSER model, which is based on the cavity model of salvation [11], may not be entirely appropriate. Indeed, normal phase LSER studies on silica phases have shown reduced correlations [20,21,22] relative to reversed phase characterization. However, LSER has been somewhat successful at modeling normal phase retention on polar bonded phases [22].

As mentioned above, it should also be noted that interpretation of the terms extracted from the LSER equation are highly dependent on the mode of chromatography utilized. For instance, in reversed phase chromatography, the vV term is loosely related to the hydrophobicity of the stationary phase [2,3]. However, in normal phase studies, in the absence of water, the vV accounts for the ability of the mobile phase to participate in dispersive interactions and the relative energy required to form a “hole” in either the mobile phase or the stationary phase for the analyte [18].

In the present work, normal phase LSER studies were performed on a pyridinium bromide modified stationary phase, using a limited set of neutral aromatic probe solutes (Table 1). The limited set of solutes was used to rapidly assess the utility in characterizing the pyridinium bromide SCIL phase under normal phase conditions [23].

Table 1.

Solvation solute descriptors for the probe solutes

Probe solute		Descriptors
		E	S	A	B	V
1	Benzene	0.610	0.52	0.00	0.14	0.716
2	Anthracene	2.290	1.34	0.00	0.26	1.454
3	Toluene	0.601	0.52	0.00	0.14	0.857
4	Mesitylene	0.649	0.52	0.00	0.20	1.139
5	Ethylbenzene	0.613	0.51	0.00	0.15	0.998
6	Propylbenzene	0.599	0.50	0.00	0.15	1.139
7	Chlorobenzene	0.718	0.65	0.00	0.07	0.839
8	Phenol	0.805	0.89	0.60	0.30	0.775
9	Benzyl alcohol	0.803	0.87	0.33	0.56	0.916
10	2-phenyl ethanol	0.811	0.91	0.30	0.64	1.057
11	p-cresol	0.820	0.87	0.57	0.31	0.916
12	p-chlorophenol	0.915	1.08	0.67	0.20	0.898
13	Nitrobenzene	0.871	1.11	0.00	0.28	0.891
14	Benzonitrile	0.742	1.11	0.00	0.33	0.871
15	Benzaldehyde	0.820	1.00	0.00	0.39	0.873
16	Anisole	0.708	0.75	0.00	0.29	0.916
17	Fluorobenzene	0.477	0.57	0.00	0.10	0.734
18	Acetophenone	0.818	1.01	0.00	0.48	1.014
19	p-xylene	0.613	0.52	0.00	0.17	0.998
20	o-xylene	0.663	0.56	0.00	0.16	0.998
21	1-Napthol	1.520	1.05	0.61	0.37	1.144
22	Biphenyl	1.360	0.99	0.00	0.22	1.324
23	Bromobenzene	0.882	0.73	0.00	0.09	0.891

2. Experimental

2.1 Materials

For the NP studies, a set of 23 neutral aromatic probes solutes (Table 1), with low cross correlation in their molecular descriptors (Table 2), were selected [2,13]. Mobile phase components (HPLC-grade water, 99% pure HPLC-grade hexane, heptane, and methanol) were supplied by Tedia (Fairfield, OH); 2-propanol was supplied by Pharmco (Brookfield, CT). Most probe solutes were purchased from either Sigma-Aldrich Corp. (St. Louis, MO) or the Fisher Scientific Chemical Co. (Pittsburgh, PA). Anthracene was purchased from Eastman Organic Chemicals (Kingsport, TN), mesitylene from J.T. Baker Chemical Company, (Phillipsburgh, NJ), 2-phenylethanol from ICN Biomedicals (Irvine, CA), benzonitrile, biphenyl, and anisole from Acros Organics (Morris Plains, NJ).

Table 2.

Correlation coefficient matrix of solute descriptors

	E	S	A	B	V
E	1	0.722	0.173	0.186	0.186
S		1	0.346	0.538	0.340
A			1	0.363	-0.093
B				1	0.146
V					1

All reagents used in the synthesis of the stationary phase; hexachloroplatinic (IV) acid hydrate, 8-bromo-1-octene, trichlorosilane, chlorotrimethylsilane anhydrous toluene, 2,6-lutidine; were purchased from the Sigma-Aldrich Corp. (St. Louis, MO). The silica sorbent was a spherical 5μm, 100 Å pore Symmetry Silica provided by the Waters Corporation (Milford, MA).

2.2 Methods

2.2.1 Stationary Phase Synthesis

The pyridinium bromide modified silica phase was prepared by hydrosilyation of the alkenylbromide followed by immobilization of the trichlorosilane ligand onto the surface of the silica substrate. The phase was then endcapped with chlorotrimethylsilane and the pyridinium cation was subsequently attached. Elemental analysis (Galbraith Laboratories Inc., Knoxville, TN) revealed that loading of the linker was ∼3.4 μmol/m², the loading of the endcapping agent was ∼0.28 μmol/m² and the loading of the pyridinium cation was ∼1.6 μmol/m². The pyridinium-modified silica was packed into a stainless steel HPLC column (150 × 4.6 mm ID; Waters Corporation, Milford, MA).

2.2.2 HPLC Analysis

All HPLC studies were carried out at room temperature, using a Shimadzu LC-10AT solvent pump at a flow rate of 1 mL/min, and a Shimadzu SPD-10A UV detector set at 254 nm. Mobile phases were composed of mixtures of 2-propanol and either heptane or hexane (v/v). All chromatographic retention data was acquired with Chrom & Spec Chromatography Data System software (Ampersand International, Inc., Beachwood, OH).

Solute samples were prepared in the mobile phase. All samples were injected using a Rheodyne injection valve/loop system (20 μL) (Cotati, CA), and analyzed in triplicate. Because of the difficulty in identifying a solute which does not interact with the multimodal stationary phase, the void volume of the column (1.71 mL) was determined by measuring the weight difference of the column when filled with either dichloromethane or hexane [2]. Statistical and multiple linear regression analysis of the chromatographic data were carried out using Microsoft Excel.

3. Results and Discussion

When the LSER method is used for column characterization, care must be taken when interpreting the importance of a single coefficient (i.e. a, s, etc.), as the overall contribution to retention is in fact the product of the coefficient and the individual molecular descriptor (i.e. bB, discussed in Section 3.2). Hence, the general trends observed in the sign and magnitude of the individual coefficients will be discussed to interpret the chemical implications of each term in the LSER equation as it applies to the pyridinium bromide SCIL phase.

3.1 Evaluation of selected LSER molecular descriptors

To initially evaluate the suitability of the training set, a variance-covariance matrix was constructed for the solutes found in Table 1. The correlation matrix for the individual probe solutes (Table 2) indicates that there is a significant correlation (0.722) between the E and S molecular descriptor terms for the analytes used in this study [2]. This correlation has been noted previously and is expected because both terms reflect contributions to retention from the polarizability of the probe solutes [2,18]. In the present study, the chromatographic retention data obtained for the test solutes using hexane/2-propanol was subjected to a global multiple linear regression, including and individually omitting the eE and sS terms, to extract the LSER coefficients, using Equation 2. A student-t test showed that there is no significant difference in the data when the eE term is removed from the LSER model, yet the results from the global regression omitting the sS term was significantly different. Because the sS term incorporates ion-dipole and ion-induced dipole interactions, this term is expected to be important on the SCIL phase due to the presence of the anion/cation pair on the surface. Thus, as has been seen previously in literature for normal phase characterizations [13,14,15], the eE term is removed and the LSER equation becomes:

$$\log \mathsf{k}=s\mathsf{S}+b\mathsf{B}+a\mathsf{A}+v\mathsf{V}+\log {\mathsf{k}}_{\mathsf{o}}$$ (3)

From the retention data for the probe solutes, the system coefficients (s, a, b and v) and the intercept (log k_o) were calculated by simultaneous linear regression analysis with Equation 3.

To better visualize the coefficients and their sensitivity to small changes in mobile phase composition, a histogram of the individual LSER coefficients is constructed, where positive values indicate a particular interaction mode is more favorable with the stationary phase than with the mobile phase (Figure 1). To evaluate the appropriateness of the selected molecular descriptors and LSER model, the log k predicted from the LSER analysis for each probe solute was plotted against the corresponding experimental log k for all mobile phase compositions (Figure 2). The overall linearity of the correlation plot with a slope near unity confirms that the LSER equation used in this study is indeed a suitable model for the selected solutes in this chromatographic system. The significance of the individual coefficients is discussed below.

Figure 1.

Histogram of the individual LSER coefficients for the pyridinium bromide column using hexane/2-propanol mobile phases.

Figure 2.

Plot displaying predicted log k versus experimental log k for the pyridinium column under normal phase conditions. (Slope = 0.994; R² = 0.994)

3.1.1 The s coefficient

The s term reflects the differences in dipole/dipole or ion/dipole interactions of the solute with the stationary and mobile phases. Consistent with reports of LSER investigations on other polar bonded phases [15], the s term is positive indicating that this particular interaction mode is stronger with the stationary phase (Figure 2). Overall, the s coefficient is relatively insensitive to mobile phase composition, becoming larger as the polar component of the mobile phase, 2-propanol, is increased. Previously, it was reported, on a related SCIL phase under high organic mobile phase compositions, a similar positive s coefficient was attributed to desolvation of the immobilized anion/cation pair [2]. Both of these findings are in contrast to previous work, where the positive s term is seen to decrease for various polar phases [13] or remains insensitive for bare silica [21] with an increase in the polar component of the mobile phase; this may be an important distinction between the synthesized SCIL phases and commercially available phases.

3.1.2 The a coefficient

The a coefficient reflects the extent to which an analyte can interact through hydrogen bond donation with the stationary and mobile phases [13]. The positive a coefficient is consistent with GC results [24] obtained by Anderson and co-workers. They reported that for ionic liquid stationary phases in gas chromatography, hydrogen bond basicity emanated from the ionic liquid anion [24]. The strong dependence upon mobile phase composition observed here for the a coefficient (Figure 2) is consistent with previous reports for normal phase LSER characterizations of bare silica phases [20,21]. In these previous reports of silica columns characterized with hexane/2-propanol mobile phases, the a coefficient decreases with an increase in the concentration of the polar component of the mobile phase and is attributed to increased adsorption of 2-propanol to the silica surface at high concentrations of 2-propanol in the mobile phase [21].

3.1.3 The b coefficient

The b coefficient reflects the extent to which analytes can interact through hydrogen bond acceptance with the stationary and mobile phases. Overall, a dependence upon mobile phase composition is observed for the positive b coefficient which decreases with increasing concentrations of 2-propanol in the mobile phase.

Because the immobilized pyridinium ligand has no hydrogen bond donation capability, the hydrogen bond donation of the stationary phase likely emanates from residual silanols, consistent with other reports of LSER investigations of bonded phases under normal phase conditions [13]. Indeed, studies on C₁₈ phases utilizing n-heptane/chloroform mobile phase highlighted the effect of residual surface silanol groups on the retention of analytes [25]. Increasing the nonpolar component of the mobile phase may slightly decrease the solvation of the stationary phase thereby leaving the surface silanol groups slightly more accessible for interactions with the solute through hydrogen bond donation.

3.1.4 The v coefficient

The v coefficient, is relatively insensitive to mobile phase composition and the negative sign indicates a higher affinity of the probe solutes for the mobile phase relative to the stationary phase is consistent with previous normal phase LSER reports [13,15,22]. The vV term reflects a combination of cavity effects and dispersive interactions.

The cavity term relates to the amount of energy required to create a cavity in the solution to accommodate the analyte relative to the amount of energy required to accommodate the analyte in the stationary phase. In the case of nonpolar analytes dissolved in hydroorganic mobile phases, the hydrogen bond network of the solvent imposes a significant barrier to cavity formation. However, in the case of hexane/2-propanol mobile phases, the predominately dispersive contributions to solvent cohesion imposes a greatly reduced barrier to cavity formation. Thus, reorganization of the tethered ligands and associated counter-ions imposes a more significant barrier to “cavity formation”; hence, the v coefficient is negative.

3.1.5 The log k_o coefficient

The log k_o coefficient, related to the system phase ratio [11,13], also appears to be insensitive to mobile phase composition, within experimental error. The fact that this term has a negative sign and is relatively constant suggests that mobile phase sorption into the stationary phase is somewhat uniform across the compositions used in this study.

3.2 LSER derived separation mechanisms

While the histogram shows the relative magnitudes and signs of the various LSER coefficients, it should be noted that the contribution of individual types of interactions to retention for a specific solute is the product of the LSER coefficient and the corresponding molecular descriptor. Examination of these products for individual analytes can provide insight into the interactions responsible for the separation of selected analytes (Table 3). For instance, comparison of the individual products for phenol and p-cresol at 80% hexane MP composition reveal that the most significant difference resides in the vV term, related to cavity formation and dispersion interactions. At 98% hexane, while the difference in the v term still dominates the separation, it is augmented by the aA term. In contrast, the separation of the o- and p- isomers of xylene is driven by the dipole/dipole and ion/dipole interactions (sS term) at both 80% and 98% hexane conditions. Thus, the ability of the pyridinium bromide phase to distinguish between the isomers arises from differences in their dipolarity/polarizability.

Table 3.

Products of LSER terms for selected compounds

Compound	Calculated LSER terms
	S	A	B	V	log ko	log k
Phenol a	0.86	1.47	0.37	-0.52	1.18	1.102
p-cresol a	0.84	1.40	0.38	-0.61	1.00	1.036
p-Xylene a	0.50	0.00	0.21	-0.66	-0.96	-0.997
o-Xylene a	0.54	0.00	0.19	-0.66	-0.93	-0.931
Phenol b	1.06	0.68	0.20	-0.61	0.18	0.128
p-cresol b	1.03	0.65	0.21	-0.72	0.01	0.032
p-Xylene b	0.62	0.00	0.11	-0.79	-1.27	-1.211
o-Xylene b	0.67	0.00	0.11	-0.79	-1.18	-1.170

^a98% hexane:2% 2-propanol (v/v)

^b80% hexane:20% 2-propanol (v/v)

3.3 Comparison with conventional columns

In further discussion of the pyridinium bromide SCIL stationary phase, it is useful to compare this novel phase to commercially available conventional phases (cyano and diol), utilized under identical normal phase conditions. From a plot of retention data on the pyridinium bromide phase vs. data obtained from the literature [15] for a limited subset of compounds (Figure 3), it appears that the retention on the pyridinium bromide SCIL stationary phase is more highly correlated with that obtained on a cyano column than on a diol column [15]. Given the availability of π—electrons on the cyano column and noting that π-π interactions are a primary mode of retention for ionic liquids in HPLC [26] as well as the greater dipole-dipole and charge transfer interactions [15,22] that are possible on the cyano column versus the diol column, it is perhaps not unreasonable that the retention of the neutral aromatic probe solutes on the synthesized SCIL phase correlate more closely with a cyano phase than on a diol phase.

Figure 3.

Plot of retention data on the pyridinium bromide column versus retention data on a cyano and a diol column.

4. Conclusion

A novel SCIL stationary phase containing a pyridinium cation was synthesized and characterized utilizing LSER methodology. The excellent correlation of the global fit between experimental and calculated retention across the mobile phase composition range used in this study (R² = 0.994), despite the complexity of the novel stationary phase, supports the utility of the LSER model to describe retention for this limited set of solutes under normal phase conditions. It appeared that retention on the pyridinium SCIL phase was more closely correlated to that of a cyano phase than that of a diol phase.