Introduction

The mechanism underlying the molecular recognition of solutes by sorbents plays a vital role in numerous chemical and biological systems. A crucial aspect of this is chiral recognition or enantiomer discrimination. Because biological systems are able to interact with enantiomers in different ways, pharmaceutical chiral separations continue to be an important aspect of drug development. The key to successful chiral separations often lies in the appropriate selection of a suitable stationary phase that has chiral recognition ability toward the target enantiomers. Therefore, knowledge of existing chiral stationary phases (CSPs) and their recognition ability is of major importance in this field. More than hundred CSPs have been discussed in the literature, of which only a few are used frequently [1,2,3]. Polysaccharide (PS) derivatives are one of the most widely used CSPs because of their higher-order structures, a high proportion of which have well-defined chirality. Substantial efforts have been made to understand the recognition mechanisms of these sorbents. However, the real understanding of the underlying mechanisms is behind their practical applications. There is a high demand for further research in this area.

Several recent investigations using various experimental techniques have extended the existing knowledge of PS-based CSP structures and their recognition ability [1, 2, 4,5,6,7,8,9,10,11,12,13,14,15,16]. The glucopyranose polymer backbones are helically wound and the radius of the helical twist is smaller for cellulose derivatives than for amylose derivatives. Kasat et al. used attenuated total reflection infrared spectroscopy (ATR-IR), X-ray diffraction (XRD), 13C cross-polarization/magic-angle spinning (CP/MAS), solid-state NMR, and density functional theory (DFT) modeling to study the effects of the backbone and side chain on the molecular environments of chiral cavities in three PS-based CSPs, viz. amylose tris(3,5-dimethylphenylcarbamate) (AD), cellulose tris(3,5-dimethylphenylcarbamate) (OD), and amylose tris[(S)-α-methylbenzylcarbamate] (AS) [4, 5]. Kasat et al. reported that following the adsorption of norephedrine, the helical pitch of both AS and AD was found to be 14.6 Å, whereas the structure of OD had a slightly larger helical pitch of 16.2 Å. They inferred that because of differences in the polymer backbones, the cavities of OD are slightly larger than those of AD, resulting in weaker intramolecular hydrogen bonding (H-bonding) in the former. The authors concluded that the molecular environments of the C=O, NH, and phenyl (Ph) groups of these polymers show considerable differences in terms of their molecular interactions.

Aboul-Enein and Ali studied the resolution of the enantiomers of econazole, miconazole, and sulconazole on three amylose columns: AD, AS, and amylose tris[(R)-α-methylbenzylcarbamate] (AR). The same elution orders were found for AS and AR [17]. They concluded that the molecular recognition of these CSPs was governed by hydrogen bonds and π–π interactions. Using a combination of high-performance liquid chromatography (HPLC) and ATR-IR, Wirz et al. studied the chiral recognitions of ethyl lactate and pantolactone (PL) with AS [18, 19]. The authors concluded that the C=O groups of the (R)-enantiomers form stronger hydrogen bonds with the polymer NH groups than those of the (S)-enantiomers. Similarly, studies on the chiral recognition of AS with four acyloin-containing chiral solutes have also been reported [6, 8]. The recognition mechanism was attributed to the formation of a strong H-bond as a result of nonenantioselective interaction between the solute and the sorbent C=O groups, whereas a weaker H-bond formed between the (R)-enantiomer and the sorbent NH groups. The notion that chiral recognition involves two interactions, a leading and secondary interaction, has been proposed several times in the literature [1, 20, 21]. When a chiral molecule approaches a sorbent binding site as the result of a strong attractive interaction, a weaker secondary interaction may result. The leading interaction is typically achieved through one or more attractive interactions, whereas the secondary interaction can be either attractive or repulsive, which leads to enantioselectivity.

Wu et al. performed DFT simulations to calculate the strength of H-bonding interactions between each achiral solute and a single 3,5-dimethylphenylcarbamate side chain [9]. Simulations of a C=O group of acetone H-bonded to the NH of the side chain yielded an energy difference of − 19.1 kJ/mol. The H-bonding energy of an isopropanol OH group H-bonded to the C=O group of the side chain was calculated to be − 18.3 kJ/mol. These values suggest similar strengths for both H-bonding interactions of a 3,5-dimethylphenylcarbamate side chain. Mitchell used a linear solvation energy relationship to study three PS-based CSPs: Chiralpak IA, IB, and IC in a mobile phase of 10/90 vol% ethanol/heptane [16]. The sorbent IA was amylose 3,5-dimethylphenylcarbamate immobilized onto silica and is analogous to the coated AD CSP. The sorbent IB was cellulose 3,5-dimethylphenylcarbamate immobilized onto silica and is analogous to the coated OD. The sorbent IC was cellulose 3,5-dichlorophenylcarbamate immobilized onto silica. The IC CSP has the strongest ability to accept a lone pair of electrons in a H-bond. This was attributed to an electron inductive effect caused by the presence of chlorine atoms on the side chain, whereas a minor difference was found for the IA and IB CSPs.

The objective of this study was to elucidate the retention mechanism for four structurally related sorbents: IA, IB, IC, and amylose 3,5-dichlorophenylcarbamate (IE) (Fig. 1). IA and IB have the same side chains but different backbones. IA, being amylose-based, has α-(1 → 4)-glycosidic linkages, whereas IB, being cellulose-based, has β-(1 → 4)-glycosidic linkages. IA and IE have the same amylose backbone but different substituents on the aromatic ring. IC and IE have the same side chains but different backbones. These structural differences lead to distinct molecular environments and recognition mechanisms, and the mechanisms of these sorbents are not fully understood. To elucidate these mechanisms, retention behavior studies with five simple molecules, acetone (AC), tetrahydrofuran (THF), methanol (MET), tert-butanol (TBA), and benzene (BZN), as a function of isopropanol (IPA) concentration were performed. Frontal analysis of AC and IPA was also applied to understand the adsorption behaviors of these CSPs. Although PS-based CSPs were not employed to separate achiral solutes, these small molecules were used as probes to elucidate the interaction mechanisms of the sorbents. Identifying the retention mechanisms of these CSPs with distinct polymer structures represents a step toward understanding the chiral discrimination mechanisms of these polymers.

Fig. 1
figure 1

Molecular structures of the polymer repeat units of a IA, b IB, c IE, and d IC polymers, with R being the side chains

Full size image

Materials and Experimental Methods

Materials

Four 250-mm-long analytical Chiralpak columns, IA, IB, IC, and IE, each with a 4.6-mm column diameter and 5-μm particle diameter were purchased from Daicel Chemical Industries, Ltd. (Tokyo, Japan). tert-Butanol (purity, 99.77%) and HPLC-grade n-hexane (99%) were purchased from Fisher Scientific (Loughborough, UK). Acetone (99.9%) was purchased from Tedia Company, Inc. (Fairfield, OH, USA). Tetrahydrofuran (99.8%), methanol (99.9%), and HPLC-grade isopropanol (99.9%) were purchased from Echo Chemical co., Ltd. (Miaoli, Taiwan). Benzene (99%) was purchased from Showa Chemical co., Ltd. (Tokyo, Japan). 1,3,5-tri-tertbutylbenzene (TTBB) and (S)-pantolactone (98%) were purchased from AK Scientific (Union City, CA, USA). Naphthalene (NPL; 99%) was purchased from Acros Organics (Morris Plains, NJ, USA). (R)-Pantolactone (99%) was purchased from Sigma Aldrich (St. Louis, MO, USA).

HPLC Apparatus and Procedures

An HPLC apparatus was employed that comprised a variable UV wavelength detector (SPD-10A, Shimadzu Corporation, Kyoto, Japan), differential refractive index (RI) detector (RID-6A, Shimadzu Corporation, Kyoto, Japan), degassing unit (DG660B, GL Science Inc., Tokyo, Japan), solvent delivery system (LC-10AT, Shimadzu Corporation, Kyoto, Japan), and thermostat-equipped column oven (CO-966, JASCO Co. Ltd., Tokyo, Japan). The mobile phase used comprised a mixture of IPA and n-hexane. A flow rate of 1.0 mL/min and a pulse injection volume of 20 μL were employed in all pulse experiments. Using neat n-hexane as a mobile phase, the UV wavelengths used for solute detection were 278 nm for AC, 206 nm for IPA, 204 nm for THF, and 220 nm for PL, whereas the RI detector was used for the BZN and NPL detections. With IPA in n-hexane as the mobile phase, the RI detector was used for the MET, TBA, AC, THF, and BZN detections. Data acquisition rates were 2 and 10 Hz for the UV and RI detectors, respectively. Retention times were measured at least twice and then averaged.

For the frontal analysis and determination of the column void fractions, an alternative HPLC apparatus (Shimadzu Corporation, Kyoto, Japan) was used that consisted of a variable UV wavelength detector (SPD-20A), differential RI detector (410, Waters Corporation, MA, USA), degassing unit (DGU-20A), solvent delivery system (LC-20AT), and thermostat-equipped column oven (CTO-20A). The flow rate employed was 1.0 mL/min. For the AC frontal analysis, the UV detector with a wavelength of 278 nm and data acquisition rate of 2 Hz was used, whereas for the IPA frontal analysis, the RI detector with a data acquisition rate of 10 Hz was used. Feed concentrations of 1, 2, 5, 10, and 20 vol% of IPA/n-hexane and AC/n-hexane solutions were selected for the frontal analysis. Each feed was input for 25 min. Prior to each feed, the system was regarded as condition equilibrated in neat n-hexane when the chromatographic signal baseline became stable for 1 h after the frontal experiments involving 1, 2, and 5 vol% feed concentrations, and for two hours after the frontal experiments involving 10 and 20 vol% feed concentrations (see Fig. 2). The experiments were conducted in order of increasing concentration. The frontal experiments involving a 1 vol% feed concentration were repeated twice and the second chromatograph was used for the analysis. The extra-column dead volume in the frontal tests was measured and subtracted from the elution profile. The amount of adsorbed molecules at a given concentration was estimated from the frontal elution profile by applying the mass balance of the modifier between the hold-up volume and the elution volume where the concentration plateaued [22, 23]. The amount of adsorbed molecules at a given concentration was determined at least twice and averaged. The breakthrough times were determined from the mass center times of the frontal waves. The retention times tref of a nonretained solute, TTBB, were used as the reference time (see Table 1). The interparticle void fractions (\(\varepsilon_{\text{e}}\)) were estimated from pressure drop data using the Ergun equation [9, 24, 25]. The intraparticle void fractions (\(\varepsilon_{\text{p}}\)) were estimated from the values of tref and \(\varepsilon_{\text{p}}\). The ratios φ of solid volume to liquid volume in the column were estimated from the values of \(\varepsilon_{\text{e}}\) and \(\varepsilon_{\text{p}}\).

Fig. 2
figure 2

The chromatogram of the IPA frontal analysis using IE sorbent. Each feed was input for 25 min. The frontal experiments of 1 vol% were repeated twice and the second chromatograph was used for the analysis. The values of the IPA concentration were found to fit the expression \(C_{\text{IPA}} \left( {\text{M}} \right) = - 1.444 \times 10^{ - 14} \times I^{5} - 6.898 \times 10^{ - 12} \times I^{4} - 1.162 \times 10^{ - 9} \times I^{3} - 5.540 \times 10^{ - 8} \times I^{2} - 2.815 \times 10^{ - 6} \times I\), where I is the RI intensity. The amounts of adsorbed IPA molecules are equal to the hatched areas

Full size image
Table 1 Column parameters of the CSPs; tref is the retention time of a nonretained solute, TTBB, \(\varepsilon_{\text{e}}\) is the interparticle void fraction, \(\varepsilon_{\text{p}}\) is the intraparticle void fraction, and φ is the ratio of solid volume to liquid volume in the column
Full size table

Results and Discussion

Retention Behaviors in Neat n-Hexane

When neat n-hexane was used as the mobile phase, the retention factors of AC were found to increase in the order IA < IE < IB < IC (Fig. 3). For THF, the retention factors were in a slightly different order: IA < IB < IE < IC. The results indicated that retention factors were larger for the cellulose derivatives than the amylose derivatives and that the sorbents with Cl-substituted phenyl groups have larger retention factors than those with CH3-substituted phenyl groups. Because n-hexane does not form H-bonds and π–π interactions with the sorbent or solutes, any interactions between n-hexane and the sorbent and solutes were assumed to be negligible. For a pulse experiment of a solute in neat n-hexane, the retention factor k can be expressed as follows [1, 10, 11, 26]:

$$k = K_{\text{SL}} q_{\text{SL,max}} \varphi /C^{0}$$
(1)

where \(K_{\text{SL}}\) is the equilibrium constant for the solute-sorbent interaction, \(q_{\text{SL,max}}\) (in mole/L of solid volume, or M′) is the saturation capacity of the adsorbed solute, and \(C^{0}\) (1 M) is the standard state concentration. The form of Eq. (1) can be deduced from the governing partial differential equations by making local equilibrium assumptions [27]. The adsorption mechanisms of solutes were modeled as a reversible equilibrium process for solute–sorbent interaction and were characterized by \(K_{\text{SL}}\) using the law of mass action. It was known that the structure of PS-based CSPs forms a regular arrangement of grooves along polymer backbone which serve as adsorption binding pockets. The binding sites of the side chains are buried inside the cavities near the backbone and are flanked by bulky aromatic functional groups which are located at the polymer surface and control the access to the binding site via steric factors [4, 6, 8]. The solutes were inferred to mainly bind with the sorbent side chains. Because IA and IE (or IE and IC) CSPs have the same side chain, the values of \(K_{\text{SL}}\) were assumed to be the same for IA and IB (or IE and IC). This suggests that the cellulose derivatives may have more NH binding sites available than the amylose derivatives. Similarly, for the alcohol IPA, the retention factors were larger for IB and IC, implying relatively larger saturation capacities of the C=O or C–O–C binding sites for IB and IC than for IA and IE. The polysaccharide derivatives have a number of H-bonding acceptors in the carbamate linkages between the polysaccharide backbone chain and side chains (see Fig. 1).

Fig. 3
figure 3

Comparisons of the van’t Hoff plots of relationships between ln k versus 1/T of the four sorbents for the five solutes a AC, b THF, c IPA, d BZN, and e NPL, in neat n-hexane. See also Table S1 in Supplementary Material

Full size image

It should be mentioned that the solutes might interact not only with the polymers but also with some silanol or 3-aminopropyl groups exposed on the silica surface. The immobilization process of the PS derivatives also requires a certain percentage of side chains containing reactive groups [3]. However, because the introduction of reactive groups on side chains for immobilization often reduced the recognition ability, it was known to minimize the amount of the reactive groups used. The implicit assumption here was that because all of the sorbents used in this study are the PS-based sorbents, the effects of these exposed residual and reactive groups on the solute retention behaviors were assumed to be the same. The difference in solute retention behaviors using distinct sorbents was attributed to the difference in polymer structures.

The retention factors of the BZN and NPL were found to increase in the order IE < IC ≈ IA < IB. The π-interactions were explained by the electrostatic and dispersion forces involved. Although aromatic groups do not have a net dipole moment, rich π electrons above and below the aromatic ring led to a strong quadrupole moment. Therefore, IC and IE were likely to have had low retention factors because of their dichlorophenyl moiety, which is π electron deficient. Notably, although the trends in retention factors were similar for BZN and NPL, the retention factors of NPL with IB were significantly greater than those using other sorbents. Strong π interactions of IB have been reported by Mitchell et al. using LSER to characterize IA, IB, and IC in a mobile phase of 10/90 ethanol/heptane [16]. Only IB had a measurable effect of intermolecular interactions associated with the polarizability of nonbonding n- and π-electrons, which includes π-electron stacking and induced dipole interactions.

The thermodynamic properties of solute adsorption can be obtained by measuring a retention factor k over a certain temperature range. The van’t Hoff plot of ln k versus 1/T has been widely used in the thermodynamic analysis of retention behaviors [1, 9, 28]. The relationship can be derived as

$$\ln k = - \frac{{\Delta H_{\text{SL}}^{0} }}{R}\left( {\frac{1}{T}} \right) + \frac{{\Delta S_{\text{SL}}^{0} }}{R} + \ln \left( {\varphi q_{\text{SL,max}} /C^{0} } \right) ,$$
(2)

where \(\Delta H_{\text{SL}}^{0}\) and \(\Delta S_{\text{SL}}^{0}\) are the standard state enthalpy and entropy changes of adsorption and R is the gas constant [9]. The thermodynamic parameters obtained from Eq. (2) are macroscopically lumped quantities that implicitly include the solvent effects and surface heterogeneity of CSPs. A study reported that the determined van’t Hoff enthalpy is strongly dependent on solvent modifier composition [11]. This was found to be a function of the fraction of solute molecules bound to modifier molecules and that of binding sites occupied by modifier molecules.

The temperature effect on the retention factors of the solutes in pure n-hexane was studied in the temperature range of 25–40 °C; where the temperature was incremented in steps of 5 °C (Fig. S1 in Supplementary Material). Because n-hexane is a nonpolar molecule, the solvent effects were assumed to be negligible. The plots of ln k versus 1/T produced straight lines, suggesting that the sorbent structure did not change substantially in this temperature range. H-bonding acceptors such as AC and THF bind primarily with sorbent H-bonding donors. Although the adsorption enthalpy changes of AC for the amylose-based sorbents were similar to those for the cellulose-based sorbents, the retention factors of AC for the cellulose-based sorbents were larger. This indicated a relatively larger saturation capacity for the cellulose CSPs, provided that the adsorption entropy changes of AC were the same (Table 2). The IE and IC CSPs had stronger adsorption energies than those of IA and IB, respectively. This is probably due to an electron inductive effect induced by the presence of chlorine atoms on the phenyl groups of IE and IC side chains [16].

Table 2 Van’ t Hoff thermodynamic parameters of five solutes using neat n-hexane as a mobile phase; \(C^{0}\) (1 M) is the standard state concentration and R2 is the coefficient of determination
Full size table

Due to the electron-withdrawing effects of the aromatic chlorine substitution, these chlorine atoms may form halogen bonds with the O atoms of AC and THF. Although the electrostatic potential of chlorine overall is considered to be negative, studies showed that the halogens bound to carbon generally have a region of positive potential oriented outward [29, 30]. The electrostatic attraction between this positive potential and solute nucleophilic region is referred to as the halogen bond. Riley et al. use computational methods to investigate halogen bond of aromatic halogen substitution [29]. The bonding energy for chlorobenzene with AC was reported to be -3.14 kJ/mol. For the bi-Langmuir adsorption isotherm, Lin et al. showed that the apparent enthalpy change determined directly using van’t Hoff plot is an average value of enthalpy changes of sorbent sites I and sites II weighted by respective retention factors for the sites I and sites II [11]. Therefore, the fact that the IE and IC CSPs had stronger adsorption energies derived from van’t Hoff plots than those of IA and IB, respectively, implied a relatively small contribution of the halogen bonds to the retention factor and enthalpy change of adsorption.

The equilibrium binding constant \(K_{\text{SL}}\) (Eq. (1)) is related to the standard Gibbs free energy change of solute adsorption according to the following equation:

$$\Delta G_{\text{SL}}^{0} = - RT\ln K_{\text{SL}} .$$
(3)

The Gibbs free energy of an equilibrium process consists of enthalpic and entropic contributions. The difference in the retention factors and van’t Hoff enthalpy changes facilitates a comparison of the saturation capacity of solute adsorptions, the corresponding equation for which can be obtained from Eqs. (1) and (3) as

$$\Delta \left( {\ln k} \right) = \Delta \left( {\ln q_{\text{SL,max}} } \right) - \frac{{\Delta \Delta H_{\text{SL}}^{0} }}{RT} + \frac{{\Delta \Delta S_{\text{SL}}^{0} }}{R}.$$
(4)

Because AC is a small and rigid molecule, the entropy changes of AC adsorption in neat n-hexane were assumed to be similar for different sorbents (\(\Delta \Delta S_{\text{SL}}^{0} \approx 0)\). Using Eq. (4) with the \(\Delta H_{\text{SL}}^{0}\) and k data shows that the saturation capacities of the sorbents for AC adsorption in neat n-hexane were estimated to increase in the order of IE < IC < IA < IB.

For the IPA, similar values of \(\Delta H_{\text{SL}}^{0}\) were found for four sorbents, implying that the main difference in retention factors of the alcohols lies in the different saturation capacities of the sorbents. The results suggested that the cellulose CSPs may have more binding sites available on which to donate a lone pair of electrons in a hydrogen bond than the amylose CSPs. For BZN and NPL molecules, stronger adsorption energies were found for IB and IC than for IA and IE, respectively. This was attributed to the relatively larger binding cavities of cellulose derivatives compared with amylose derivatives [4, 5]. Notably, the binding energies for IA, IC, and IE have similar values for both BZN and NPL, whereas for IB, the binding energy of NPL (− 22.19 kJ/mol) is considerably larger than that of BZN (− 12.28 kJ/mol). The binding geometric configurations of π interactions are often classified into three types: face-to-face, parallel-displaced, and T-shaped configurations. Face-to-face and parallel-displaced configurations are stabilized by maximizing dispersion forces, which increase with a larger intermolecular contact area. Therefore, everything else being equal, this difference between the IB and other CSPs appears to imply that they have different binding geometric configurations.

Adsorption Isotherms of AC and IPA

The adsorption isotherms of AC and IPA that were obtained using frontal analysis for concentration ranges of 0.14–2.7 M and 0.13–2.6 M, respectively, are shown in Fig. 4. The Langmuir (Eq. (5)) and Freundlich (Eq. (6)) isotherm models were assumed to estimate the adsorption parameters. Tables 3 and 4 list the fitted parameters of the models. The models are expressed as follows:

$$q_{\text{SL}} = \frac{{q_{\text{SL,max}} K_{\text{SL}} \left( {C_{\text{SL}} /C^{0} } \right) }}{{1 + K_{\text{SL}} \left( {C_{\text{SL}} /C^{0} } \right) }},$$
(5)
$$q_{\text{SL}} = A\left( {C_{\text{SL}} /C^{0} } \right)^{1/n} ,$$
(6)

where \(q_{\text{SL}}\) (in M′) is the concentration of the adsorbed molecule and A (in M′) and n are Freundlich constants for a given adsorbate and sorbent at a particular temperature. From the correlation coefficient (R2) of the model fit, adsorptions of AC on the four sorbents and IPA on the IA and IC sorbents were sufficiently modeled using the Langmuir isotherm. The R2 values of the fits are slightly worse for AC with IE (0.983) and IPA with IC (0.966). The Freundlich isotherm is more suitable for use in the adsorption of IPA on the IB and IE sorbents. Although the Freundlich model is known to be entirely empirical, it has been shown to successfully account for sorbents that have a heterogeneous adsorption surface [27].

Fig. 4
figure 4

Fits of adsorption isotherm models to the a\(q_{\text{AC}} \left( {C_{\text{AC}} } \right)\) and b\(q_{\text{IPA}} \left( {C_{\text{IPA}} } \right)\) data from the frontal analysis at 30 °C. For AC, the Langmuir isotherm model was used to fit the data. For IPA, the Langmuir isotherm model was used to fit the data using IA and IC sorbents, whereas the Freundlich isotherm was used for the adsorptions on the IB and IE sorbents. See also Table S2 in Supplementary Material

Full size image
Table 3 The determined parameters of the Langmuir and Freundlich isotherm models for the adsorption data of AC at 30 °C
Full size table
Table 4 The determined parameters of the Langmuir and Freundlich isotherm models for the adsorption data of IPA at 30 °C
Full size table

The saturation capacities \(q_{\text{SL,max}}\) of AC, 1.9 ± 0.1, 1.8 ± 0.1, 2.1 ± 0.2, and 1.9 ± 0.1 M′ for IA, IB, IE, and IC, respectively, were found to be similar. These results differ from those obtained from the pulse experiments in neat n-hexane. The discrepancy between the two methods is unclear. The use of high solute concentrations in the frontal analysis may have somewhat changed the sorbent structure, and probably also the adsorption mechanism. Kasat et al. used various techniques such as ATR-IR, XRD, CP/MAS, solid-state NMR, and DFT modeling to study the solvent effect on the structure of AD sorbent using n-hexane, MET, ETH, IPA, and acetonitrile [14]. The authors concluded that following adsorption of n-hexane, the structure of the polymer remained unchanged, whereas following adsorption of MET, ethanol, IPA, or acetonitrile, the polymer structure and the distribution of the H-bonding states of the polymer changed. Because the equilibrium constants \(K_{\text{SL}}\) obtained from AC-sorbent interactions, 1.7 ± 0.3, 2.0 ± 0.4, 1.9 ± 0.6, and 4.1 ± 0.4 for IA, IB, IE, and IC, respectively, have relatively larger standard errors than the differences between the \(K_{\text{SL}}\) values, all that can be concluded is that the binding groups of the IC sorbent might have the strongest binding strength with AC.

Except for IA, the fits of the Langmuir isotherm model are generally poor for the IPA adsorptions (Table 4), suggesting heterogeneous adsorption mechanisms. Although the estimated \(q_{\text{SL,max}}\) values of IPA, 1.8 ± 0.0, 1.4 ± 0.1, 1.9 ± 0.2, and 1.9 ± 0.1 M′ for IA, IB, IE, and IC, respectively, have the same order of magnitude as those of AC, the equilibrium constants \(K_{\text{SL}}\) that were obtained were considerably larger than those of AC, resulting in a steeper increase in solute adsorption at low concentrations in the liquid phase. The equilibrium constants \(K_{\text{SL}}\) of the cellulose derivatives IB (13.6 ± 6.3) and IC (15.0 ± 3.4) were larger than those of the amylose derivatives IA (7.80 ± 0.66) and IE (6.75 ± 2.97). Using the values of \(K_{\text{SL}}\) and \(\Delta H_{\text{SL}}^{0}\) determined by the frontal and pulse analyses, respectively, the values of \(\Delta S_{\text{SL}}^{0}\) were estimated and shown in Tables 3 and 4. The negative entropy changes suggested that AC and IPA adsorptions are driven by enthalpy changes.

Finally, it is important to note that the adsorption isotherms may depend on the regeneration time in neat n-hexane. Feeds of the IPA solutions after equilibration with different column volumes in neat n-hexane were therefore evaluated (data not shown). A long regeneration protocol was necessary to ensure the reproducibility of the results (see Sect. 2.2 and Fig. 2 for the protocol used in this study). This implies that the sorbent structures gradually changed to a thermodynamically stable state in neat n-hexane. Moreover, it was suspected that during each frontal analysis, the sorbent structure might be transitioning from a thermodynamically stable state in neat n-hexane to a stable state in adsorbate/n-hexane solution. Acknowledging the possible kinetic transition of sorbent structures adds another level of difficulty in characterizing sorbent adsorption mechanisms. Nevertheless, the adsorption isotherms of sorbents obtained from the frontal analysis may not correspond to the thermodynamically stable state in neat n-hexane or in an adsorbate/n-hexane solution. Further work needs to be conducted to elucidate this issue.

Retention behaviors as a function of IPA content

The retention factors of the solutes associated with H-bonding groups decreased as the IPA concentration increased from 0.00 to 7.85 M (see Fig. 5). For AC, MET, and TBA, the plots of \(\ln k\) versus \(\ln C_{\text{IPA}}\) show slightly concave downward retention curves (Fig. S2 in Supplementary Material). It was inferred that this nonlinearity of retention curves resulted from the competitive adsorption equilibrium of IPA and the solute-IPA complexation equilibrium [31, 32]. The retention factors of BZN were found to decrease, reach a minimum, and then increase as the IPA concentration increased. Previous research has found that the π interactions involved both electrostatic and dispersion interactions, with complementary effects on the retention behavior when the solvent composition was changed [9, 12]. For alcohols, the retention curves showed the strongest dependence on the IPA content. The slope values of the \(\ln k\) versus \(\ln C_{\text{IPA}}\) plots were related to the number of IPAs displaced from the sorbent surface and the solute–IPA complex following solute adsorption [32, 33]. Therefore, the high absolute values of slopes for alcohol solutes may be due to the potential of IPA hydroxyl groups to form multiple H-bonds on a single hydroxyl functional group of alcohol solutes [31, 34, 35].

Fig. 5
figure 5

Plots of ln k versus \(\ln C_{\text{IPA}}^{'}\) at 30 °C for solutes a AC, b THF, c MET, d TBA, and e BZN using four sorbents.\(C_{\text{IPA}}^{'}\) is the dimensionless IPA concentration and is equal to \(C_{\text{IPA}} /C^{0}\); \(C^{0}\) (1 M) is the standard state concentration. See also Table S3 in Supplementary Material

Full size image

In the presence of IPA, the retention factors of AC and THF for the four columns were found to increase in the orders of IA < IB < IE < IC and IA ≈ IB < IE < IC, respectively. The results were consistent with findings using neat n-hexane as the mobile phase in that the cellulose derivatives have larger retention factors than amylose derivatives, and larger retention factors were found for sorbents with Cl-substituted phenyl groups. It was inferred that the cellulose derivatives may have more available binding sites than the amylose derivatives and that the IE and IC CSPs have stronger adsorption energies because of the presence of chlorine atoms on the phenyl groups of side chains. The retention curves of the \(\ln k\) versus \(\ln C_{\text{IPA}}\) plots for IA, IE, and IC were found to be in parallel, as described in the retention models proposed by Tsui et al. [31, 32]. Because AC and THF molecules presumably bind primarily with the sorbent H-bonding donors and do not compete with IPA for H-bonding acceptors, the H-bond between the solute O atom and IPA OH groups was expected to be the dominant factor affecting solute retention behaviors. According to the proposed model, the plots of \(\ln k\) versus \(\ln C_{\text{IPA}}\) of AC for different sorbents should only differ in the intercepts, which is controlled by the equilibrium constant for the solute-sorbent interaction and the saturation capacity of the adsorbed solute. For IB, the discrepancy in the retention behavior is unclear, probably due to the progressive sorbent structure change that occurs with increasing IPA content, or strong solvophobic interactions that were not considered in the model and would have opposite effects on retention behaviors with increasing solvent polarity [12, 36].

For the solute MET, the retention factors were slightly larger for IB and IC than for IA and IE, respectively. By contrast, for TBA, the retention factors were larger for IA and IE than for IB and IC, respectively. This difference between MET and TBA is likely to be because MET has a smaller steric bulk, which reduces steric restriction and increases the accessibility of the binding sites. For cellulose derivatives such as IB and IC, it was inferred that the cavities of these sorbents are slightly larger than those of amylose derivatives. This implies that a small molecule MET may be able to interact with sorbent C–O–C groups buried deep inside the cavities near the polymer backbone.

In the presence of IPA, the retention factors of the BZN were found to increase in the order of IE < IC < IA < IB and thus were the same as those found in neat n-hexane. The results suggest that the side chains with Cl-substituted phenyl groups have somewhat weakened the π interactions. Larger retention factors were found for the sorbents with cellulose backbones. This was probably due to their larger cavities, which allowed more access to the aromatic binding sites.

Retention Behaviors of Pantolactone Enantiomers

To further elucidate the effects of the sorbent structure on the retention and recognition mechanisms, the retention behaviors of (R)-pantolactone (R-PL) and (S)-pantolactone (S-PL) for the four CSPs using 10 vol% IPA in n-hexane as a mobile phase were studied (Fig. 6). The retention factors of R-PL using distinct columns were found to increase in an order of IB < IA < IE < IC, whereas for S-PL, a slightly different order of retention factors was obtained: IB ≈ IA < IE < IC. Generally, the retention factors were higher for the sorbents with Cl-substituted phenyl groups, and the enantiomeric elution order somewhat depended on the polymer backbone.

Fig. 6
figure 6

HPLC results for retention times of (R)- and (S)-pantolactone enantiomers with Chiralpak IA, IB, IC, and IE sorbents using 10 vol% IPA in n-hexane as a mobile phase at 30 °C

Full size image

The retention factors, 1.82 and 1.85, of S-PL enantiomer were approximately the same for IA and IB, respectively. The enantioselectivity (\(\alpha \equiv k_{\text{R}} /k_{\text{S}}\)) for IB was found to be 0.97 and no clear chiral resolution was observed, whereas the highest \(\alpha\) value (1.35) was obtained for IA. These results suggested that for the sorbents with CH3-substituted phenyl groups, the chiral recognition of PL enantiomers by IA involved an attractive secondary interaction, which might be attributed to the smaller cavity size of the amylose CSPs than that of the cellulose CSPs.

For the sorbents with amylose backbones, the \(\alpha\) value was smaller for IE (1.16) than that for IA (1.35), whereas for the cellulose sorbents, a successful enantiomeric separation of PL by IC sorbent (\(\alpha = 0.82\)) was achieved. The presence of chlorine atoms on the phenyl groups of side chains seemed to enhance the chiral discrimination ability of the cellulose CSP but reduce that of the amylose CSP. Therefore, all else being equal, it was speculated that the interactions between PL and the sorbent NH groups are enantioselective for the cellulose CSPs but nonenantioselective for the amylose CSPs.

Conclusions

For the retention behaviors of AC and THF, larger retention factors were found for the cellulose derivatives than the amylose derivatives, and the retention factors for sorbents with Cl-substituted phenyl groups were larger than those with CH3-substituted phenyl groups. The van’t Hoff analysis showed that the sorbents with Cl-substituted phenyl groups have stronger adsorption energies. The saturation capacities of the sorbents for AC adsorption in neat n-hexane were estimated to be higher for the cellulose CSPs.

Similarly, for alcohol molecules in neat n-hexane, the retention factors and the values of \(\Delta H_{\text{SL}}^{0}\) suggest that the cellulose CSPs may have more binding sites available than the amylose CSPs. In the presence of IPA, the retention factors of MET were slightly larger for the cellulose CSPs than for the amylose CSPs, whereas the opposite results were found for TBA. The difference was attributed to the small molecular size of MET and the larger cavity size of the cellulose CSPs in comparison with amylose CSPs.

The data on the retention behaviors of the aromatic solutes suggest that the side chains with Cl-substituted phenyl groups weakened the π interactions, and larger retention factors and stronger adsorption energies were found for the sorbents with cellulose backbones.

The adsorption isotherms of AC and IPA were determined using frontal analysis. Adsorptions of AC on the four sorbents were sufficiently modeled using the Langmuir isotherms, whereas heterogeneous adsorption mechanisms were suggested for the adsorptions of IPA.