How Ternary Mobile Phases Allow Tuning of Analyte

Aug 27, 2013 - When retention times increase sharply between 10/90 and 5/95 (v/v) W/ACN, intermediate retention values are stepwise accessible with a ...
2 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

How Ternary Mobile Phases Allow Tuning of Analyte Retention in Hydrophilic Interaction Liquid Chromatography Sergey M. Melnikov,†,‡ Alexandra Höltzel,† Andreas Seidel-Morgenstern,‡ and Ulrich Tallarek*,† †

Department of Chemistry, Philipps-Universität Marburg, Hans-Meerwein-Strasse, 35032 Marburg, Germany Max-Planck-Institut für Dynamik komplexer technischer Systeme, Sandtorstrasse 1, 39106 Magdeburg, Germany

‡

ABSTRACT: An attractive yet hardly explored feature of hydrophilic interaction liquid chromatography (HILIC) is the tuning of analyte retention through the addition of an alcohol to the water (W)−acetonitrile (ACN) mobile phase (MP). When retention times increase sharply between 10/90 and 5/95 (v/v) W/ACN, intermediate retention values are stepwise accessible with a ternary MP of 5/90/5 (v/v/v) W/ACN/alcohol by switching from methanol to ethanol to isopropyl alcohol. We investigate the physicochemical basis of this retention tuning by molecular dynamics simulations using a model of a 9 nm silica pore between two solvent reservoirs. Our simulations show that alcohol molecules insert themselves neatly into the retentive W-rich layer at the silica surface, without disrupting the layer’s structure or altering its essential properties. With the decreasing tendency of an alcohol (methanol > ethanol > isopropyl alcohol) to move toward the silica surface, the contrast between the W-rich layer and the bulk MP sharpens as the latter becomes more organic, while the W density near the silica surface remains high. Analyte retention increases with the ratio between the W mole fraction in the diffuse part of the W-rich layer and that in the bulk MP. We predict that tuning of HILIC retention is possible over a wide range through the choice of the third solvent in a W/ACN-based ternary MP, whereby the largest retention values can be expected from W-immiscible solvents that fully remain in the bulk MP.

H

This may or may not be desirable for the separation problem at hand, as analysis times should be as short as possible. For example, for a HILIC separation of four analytes on a hybrid silica SP, Grumbach et al.4 observed a retention leap between 10/90 and 5/95 (v/v) W/ACN that elongated the analysis time from ≈2.6 to 9.4 min. By substituting half of the W volume fraction in the MP with an alcohol (Alc), Grumbach et al.4 achieved retention factors and analysis times between the two extremes, tailoring the values by their choice of Alc: methanol (MeOH), ethanol (EtOH), or isopropyl alcohol (IPA). Figure 1 shows the retention factors obtained by Grumbach et al.4 with the cited binary and ternary MPs represented in a unified form as 5/90/5 (v/v/v) W/ACN/third solvent, where the third solvent is W, MeOH, EtOH, IPA, or ACN. The representation of the experimental data in Figure 1 shows that the third solvent in the MP determines overall retention, which increases in the order of W < MeOH < EtOH < IPA < ACN, with respect to the third solvent. The same eluotropic order was observed by Periat et al.,9 who used ternary W/ACN-based MPs in combination with various HILIC columns, including polar-bonded and zwitterionic SPs. The experimental data satisfy the rationale that retention is increased by diluting the strong solvent W with a solvent with a lesser elution strength. The retention factors in Figure 1 actually increase with decreasing elution strength of the third

ydrophilic interaction liquid chromatography (HILIC) uses a hydrophilic stationary phase (SP) together with an aqueous−organic mobile phase (MP) to separate polar compounds. Formally, HILIC is complementary to reversedphase liquid chromatography (RPLC), where an aqueous− organic MP is combined with a hydrophobic SP to separate compounds of moderate to low polarity. Together, the two modes cover a wide, partially overlapping analyte spectrum, often with orthogonal retention order and selectivity between them.1,2 Whereas the organic component of the MP is the strong solvent in RPLC, water (W) is the strong solvent in HILIC so that analyte retention in HILIC increases with the fraction of the organic solvent, usually acetonitrile (ACN), in the MP. Although HILIC’s surge in popularity in both recent applications and research papers1 is due mostly to the orthogonality to RPLC and the advantages of an ACN-rich MP, the approach to HILIC solely from this perspective neglects features that bear the potential for developing the method further. One such feature is that analyte retention in HILIC (e.g., with a bare-silica SP) grows slowly over a wide range of W/ACN ratios before it takes a sudden leap and becomes highly sensitive to the ACN content of the MP.3−8 Because the extent of the leap also depends on analyte properties, retention and selectivity in HILIC can be comfortably manipulated via the ACN content of the MP, whereas manipulation of selectivity in RPLC requires a change of the MP solvents or the SP. Due to the high sensitivity of analyte retention to the ACN content of the MP, the time required for elution of all components of an analyte mixture may stretch considerably upon small changes in ACN content. © 2013 American Chemical Society

Received: July 12, 2013 Accepted: August 27, 2013 Published: August 27, 2013 8850

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

pierced silica block sandwiched between two solvent reservoirs feeding a W/ACN mixture of a defined ratio to the silica pore. Our results revealed a W-rich layer at the silica surface consisting of a rigid and a diffuse part, so named for the analogy to the electrostatic double layer formed by electrolyte solutions at charged surfaces. The thin, rigid W layer ( EtOH (ε° = 0.68) > IPA (ε° = 0.60) > ACN (ε° = 0.50),10 which is remarkable considering that the solvent elution strength concept was developed for adsorption (solid−liquid) chromatography, not HILIC.11 Another aspect revealed by Figure 1 is that later-eluting (longer-retained) analytes are more sensitive to the MP composition than earlier-eluting analytes. This observation leads directly to the model of the HILIC retention mechanism, which involves analyte partitioning and weak interactions with the SP surface.4 The polar, hydrophilic analytes partition from the ACN-rich MP into a “hydrophilic pillow”12 of W molecules accumulated at the polar SP; as a result, analytes may adsorb to the SP surface.13,14 (Other types of interaction can contribute to or even dominate retention under HILIC conditions, especially when the SP bears charges; the retention mechanism is then termed multi- or mixed-mode rather than HILICtype.15) The experimentally determined W surface excess adsorption isotherms of polar SPs under HILIC conditions5,16,17 confirmed the presence of surface W, but a molecular-detail picture could be obtained only through molecular dynamics (MD) simulations.18 For these MD simulations, we used a model that considered the experimental conditions in HILIC: a

■

SIMULATION DETAILS The design of the simulation box, the preparation of the silica pore’s surface, the choice of the force-field parameters, and the simulation scheme have been described in detail previously.18−20 We repeat only the salient points here, with an emphasis on the details that mimic the experimental conditions in HILIC. The quadrilateral simulation box (10.74 × 10.74 × 10 nm) consisted of a central, pierced block of β-cristobalite SiO2 (3.938 nm long) flanked at each side by a solvent reservoir (3.031 nm long). A cylindrical pore, carved along the β-

Figure 2. Front view of the pierced silica block (left) and a projection showing one-half of the pierced silica block over the whole pore length (right). The silica lattice is in gray to accentuate the surface groups, whose Si, O, and H atoms are shown in yellow, red, and white, respectively. 8851

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

used for data analysis. Table 1 lists how many molecules of each solvent species were used in the productive simulation runs.

cristobalite (001) vector, spans the length of the block and connected the two solvent reservoirs. The chosen pore diameter of 9 nm represents a typical average mesopore size of HILIC columns for small to medium-sized analytes. The inner and outer silica surfaces in contact with the solvent (the pore surface and the sides of the block facing the reservoirs) were prepared such21 that they bore single and geminal silanol groups as well as siloxane bridges at an overall hydroxylation of 4.8−5.2 OH groups/nm2 (8.0−8.5 μmol/m2), which is typical for chromatographic bare-silica particles. Figure 2 shows a front view of the pierced silica block next to a projection showing the curved pore surface, with external and internal surface groups in atomic detail. The curved surface of the cylindrical pore has an irregular roughness on the atomic level so that the radial positions of surface Si atoms are distributed rather than occurring at one value. The origin for measurements of the radial distance R from the pore surface (R = 0 nm) was placed 4.75 nm from the center of the initially carved pore (i.e., behind the pore radius of 4.5 nm) to obtain only positive R values for the solvent molecules. MD simulations were performed at 300 K with Gromacs 4.5.222 using 120 processors of Intel Xeon Central Processing Units. We used the force-field parameters of Gulmen and Thompson23 for all atoms of the silica surface; the silica frame was fixed during simulations, except for the free rotation of H atoms around the Si−O bond at fixed O−H distance and Si− O−H angle. Following the recommendation and example of Mountain,24,25 we used the SPC/E model26 for W and the TraPPE-UA force field 27 for ACN. This combination reproduces well the experimental properties of W−ACN mixtures. For MeOH, EtOH, and IPA, we also used the TraPPE-UA force field.28 Periodic boundary conditions were applied in all directions. A Nosé−Hoover thermostat kept the temperature at 300 K. Equations of motion were integrated with a 1 fs time step. Our simulation scheme accounts for the prerequisite column conditioning, during which the large surface of the porous silica particles that constitute the column bed is equilibrated with the MP. The hydrophilic silica surface attracts W from the W/ ACN-based MP, which is pumped from solvent reservoirs through the column bed. Once the W-rich layer has formed, the W/ACN ratio of the MP in the column remains constant, and the column is ready for sample injection. We mimicked the column conditioning by performing a series of preliminary simulations in which the W/ACN/third solvent ratio in the simulation box was adjusted. At the beginning, the respective number of molecules for each solvent species according to the targeted W/ACN/third solvent ratio was calculated, and the solvent molecules were then distributed in random positions in the simulation box on the basis of nonoverlapping van der Waals radii. After a simulation run, when W molecules had moved to the silica surface, the actual W/ACN/third solvent ratio was checked in the reservoirs (at a distance of >1.5 nm from the external surface) and inside the pore (at R > 2.5 nm) and manually corrected toward the targeted W/ACN/third solvent ratio by removing or adding the necessary number of molecules of a solvent species in the simulation box. This was repeated until the actual ratio deviated by less than 2% from the targeted ratio. Between three and five preliminary simulation runs (of 30 ns each) were required to condition the system for a productive simulation run (30 ns) of which the last 6 ns were

Table 1. Number of W (NW), ACN (NACN), and Alc (NAlc) Molecules Used in the Productive Simulation Runs for a Nominal MP of 5/90/5 (v/v/v) W/ACN/Third Solvent third solvent

NW

NACN

NAlc

W MeOH EtOH IPA ACN

5657 3440 3585 3715 3980

8850 8850 8850 8850 9280

901 522 354

Hydrogen bonds between solvent molecules or between surface OH groups and solvent molecules were assumed upon the fulfillment of three geometrical criteria.29 Solvent residence times were calculated as the average time that solvent molecules remained within a certain space interval, tolerating a shift of ±0.125 nm around the initial R coordinate to account for perpetual molecular motion. Solvent diffusion coefficients parallel to the pore surface were calculated via the Einstein equation from the mean square displacements of solvent molecules made parallel to the surface within a certain space and time interval, as described previously.20

■

RESULTS AND DISCUSSION Figure 3 shows the distribution of each solvent species inside the 9 nm silica pore for a MP of 5/90/5 (v/v/v) W/ACN/third

Figure 3. Radial number density profiles of ACN (central C atom), W (O atom), and Alc (O atom) for a MP of 5/90/5 (v/v/v) W/ACN/ third solvent, where the third solvent is W (black), MeOH (blue), EtOH (green), IPA (red), or ACN (gray). Dashed lines divide the rigid W layer (I), the diffuse W layer (II), and the pore bulk (III).

solvent. The solvent density profiles were calculated as the number density (ρn) for the O atom of W, the central C atom of ACN, and the O atom of Alc as a function of the distance R from the silica pore’s surface. The three regions that can be distinguished according to their compositional, structural, and dynamic properties are also indicated in Figure 3: the immediate surface region (I, R < 0.425 nm) and adjacent 8852

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

interface region (II, R = 0.425−1.5 nm), which constitute rigid and diffuse parts of the W-rich layer, respectively, and the bulk region of the pore (R > 1.5 nm), whose properties match those of the outside bulk MP.18 (Note that a pore bulk region may not exist in pores with a smaller diameter.19) The comparison of the density profiles for binary and ternary MPs in Figure 3 reveals that the presence of 5 vol % Alc in the MP alters neither the W nor the ACN profiles substantially. ACN profiles are nearly identical for all MP compositions, which is unsurprising given that ACN molecules are pushed away from the hydrophilic surface and concentrated in the bulk region. W profiles are identical in region I, except for MeOH as the third solvent; in region II, the W density decreases in the order W > ACN > IPA > EtOH > MeOH, with respect to the third solvent. The change observed in the W profiles corresponds with the Alc profiles, which show that the maximum Alc density coincides with the W peak in region II. Of the three Alc species, MeOH has the largest effect on the W profiles; only with MeOH as the third solvent does the W density at the surface (region I) exhibit a small dip. The tendency of an Alc species to move from the predominantly organic MP into the W-rich layer at the silica surface decreases in the order MeOH > EtOH > IPA. Overall, Figure 3 shows that although Alc molecules are part of the solvent layer at and near the silica surface, this layer remains preferentially populated by W molecules, which justifies the term “W-rich layer” for the investigated ternary MPs as well. Key to HILIC is the hydrophilic SP that attracts and binds W. With a bare-silica SP and W/ACN-based MP, hydrogen bonding plays a central role in the formation of the W-rich layer: W molecules form hydrogen bonds (HBs) with surface OH groups, and these surface-attached W molecules then attract and bind further W molecules by extending their HB network.18 Thus, HBs are the “glue” between the molecules of the W-rich layer, as well as between the W-rich layer and the silica surface. As Alc molecules have an ability to form HBs with the surface OH groups similar though not equal to that of W, the addition of Alc to the W/ACN-based MP could possibly alter the structure of the W-rich layer. However, our simulations show that the Alc contribution to surface coordination is small to negligible. While 97−99% of the surface OH groups (which can form as much as 3 HBs per group) form HBs with W, only 7% form HBs with MeOH, and surface coordination by EtOH (2%) and IPA (0.7%) is as comparably low as that by ACN (≤1%). Thus, the number of surface−W HBs per surface OH group (HBOH) remains at an average of HBOH = 1.8 for all MP compositions, except for MeOH as the third solvent, when the average HBOH decreases slightly to 1.7. (Average HBOH = 0.08, 0.03, and 85%) and EtOH and IPA molecules exclusively donate their H atom to the surface. To facilitate HBs with the surface, Alc molecules in the rigid W layer uniformly orient their alkyl chains toward the bulk and their OH group to the silica surface; Alc molecules in the diffuse W layer are similarly but less strictly oriented. The alcohol orientation is recognizable in simulation snapshots and ⎯⎯⎯→ quantitatively traceable through the angle between the OH vector and the vector pointing from the pore center to the pore 8853

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

times and parallel diffusion coefficients. The data in Table 2 show that W mobility values are further decreased in ternary MPs, which means that the presence of Alc molecules reinforces the characteristically low mobility of the W-rich layer. In summary, the analysis of solvent density (Figure 3), hydrogen bonding (Figures 4 and 5), and translational mobility data (Table 2) has established that the characteristic properties of the W-rich layer remain unchanged for all investigated MP compositions. What changes is the difference of the W-rich layer compared to the bulk liquid, which can be quantified as the local-to-bulk ratio of the W mole fraction. Figure 6 shows

Figure 5. Hydrogen-bond network of Alc for a MP of 5/90/5 (v/v/v) W/ACN/Alc, where Alc is MeOH (blue), EtOH (green), or IPA (red). Number of hydrogen bonds per Alc molecule (HBAlc) for Alc− surface and Alc−W hydrogen bonds as a function of the distance R from the silica surface.

surface (data not shown). Overall, the data on surface−solvent and solvent−solvent hydrogen bonding show that Alc molecules insert themselves into the W-rich layer without disrupting its structure. Next, we look at solvent mobility data to judge how the presence of Alc molecules affects the solvent mobility in the Wrich layer. Table 2 lists average residence times τ (mobility perpendicular to the pore surface) and diffusion coefficients parallel to the pore surface (D∥) of W and Alc molecules for three different regions: the rigid W layer (region I, R < 0.425 nm), the location of maximum W−W coordination and Alc density in the diffuse W layer (region II, R = 0.55−0.75 nm), and the pore bulk (R > 2 nm). Alc residence times in the rigid W layer are much lower than those of W, in agreement with the different degrees of surface attachment: W forms multiple surface attachments per molecule (maximum HBW = 2.3, top panel of Figure 4B), whereas Alc molecules donate only their H atom to the surface; the percentage of surface-attached molecules is also higher for W (92%) than for Alc (53−77%, depending on the Alc species). In the diffuse W layer, however, Alc residence times top those of W, indicating the favorite location of Alc molecules inside the silica pore. In the diffuse W layer, W−W and Alc−W hydrogen bonding are at their maximum, and the translation of an Alc molecule requires breaking and re-forming multiple HBs between the surrounding W molecules. Alc parallel diffusion coefficients in the rigid and diffuse W layer are less than those of W, reflecting the hindered movement of Alc molecules within the highly organized HB network. Apart from the mobility differences between W and Alc molecules, Table 2 also monitors the effect of the MP composition on W residence

W Figure 6. Local-to-bulk ratio of the W mole fraction χW local/χbulk for the rigid W layer (region I), the diffuse W layer (region II), and the complete W layer (regions I + II) depending on the third solvent in a MP of 5/90/5 (v/v/v) W/ACN/third solvent.

this ratio for the complete W-rich layer, as well as for its rigid and diffuse part. The local-to-bulk ratio of the W mole fraction for the rigid W layer nearly doubles when the W volume fraction in the MP is decreased by half upon switching from W to MeOH as the third solvent; the subsequent increases are much smaller. This behavior reflects the fact that the rigid W layer consists almost exclusively of W molecules (cf. Figure 3) so that its local W mole fraction remains highly similar for all MPs containing 5 vol % W. The local-to-bulk ratio of the W mole fraction for the rigid W layer depends therefore much more on the W volume fraction in the MP than on the third solvent species. The diffuse W layer shows a different sensitivity to the MP composition than its rigid counterpart. The local-to-bulk ratio of the W mole fraction for the diffuse W layer (region II) increases more upon exchanging W for MeOH and IPA for ACN as the third solvent (i.e., upon switching between binary and ternary MPs) than upon exchanging one Alc species for

Table 2. Translational Mobility Perpendicular (Residence Time τ) and Parallel (D∥) to the Silica Pore’s Surface of W and Alc Molecules for a MP of 5/90/5 (v/v/v) W/ACN/Third Solvent R < 0.425 nm third solvent W MeOH EtOH IPA ACN a

W W MeOH W EtOH W IPA W

R = 0.55−0.75 nm

R > 2 nm

τ (ps)

D∥ (10−5 cm2/s)

τ (ps)

D∥ (10−5 cm2/s)

τ (ps)

D∥ (10−5 cm2/s)

730 1250 170 1170 130 1010 nda 830

0.021 0.017 0.007 0.018 0.003 0.021 nda 0.020

19.5 21.9 125 21.4 170 18.5 177 19.6

0.52 0.39 0.17 0.45 0.15 0.53 0.14 0.52

3.1 2.9 3.3 2.9 3.7 2.9 4.6 2.8

2.16 2.62 2.90 2.58 2.60 2.64 2.50 2.70

Mobility data were not determined for IPA molecules because their center of mass remains outside the limits of region I. 8854

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

fraction in the MP with Alc, the bulk liquid becomes more organic, while the functional SP loses little of its W content. This conclusion allows a prediction about the effect of third solvents other than the studied Alc species: HILIC retention due to partitioning increases in the amount that the third solvent decreases the elution strength of the MP, while the W density remains concentrated in the W-rich layer. Via exchange of Alc for solvents with less hydrogen-bond capability, such as acetone or tetrahydrofuran (which can accept but not donate H atoms), retention may be increased above the values obtained with IPA. The largest retention values should come from the apolar solvents used in normal-phase chromatography, such as chloroform, diisopropylether, or cyclopentane. These Wimmiscible solvents would remain in the ACN-rich MP and thus maximize its contrast to the W-rich, functional SP. The idea of using W-immiscible solvents finally leads back to the possible earliest report of a retention mechanism related to HILIC (long before Alpert defined and named this chromatographic mode3), as has been pointed out by McCalley,30 when Martin and Synge31 used W-saturated chloroform as the MP to separate amino acids on a silica SP.

another. The same dependence on the MP composition was observed for the experimental retention factors (Figure 1). Comparison of Figure 6 with Figure 1 reveals that the local-tobulk ratio of the W mole fraction for the diffuse W layer (region II) reproduces well the trend of the experimental retention factors. That the diffuse part is a better indicator of the effect of the MP composition on analyte retention than the complete Wrich layer can be attributed to the inflexible properties of its rigid part. When partitioning is the dominant retention mechanism of a HILIC separation, the main effect of the rigid W layer is to provide the structure for the formation of the diffuse W layer. In this study, we have concentrated on the conditions that analytes meet inside a silica pore when Alc is the third component of a ternary W/ACN-based MP. From an alternative viewpoint, the Alc species could also be considered as simple examples of uncharged, hydrophilic analytes. In this case, the results of our simulations also provide a glimpse into the HILIC retention behavior of this analyte type: molecules will accumulate in the diffuse rather than the rigid W layer, where only small analytes will be able to penetrate; once they arrive in the highly organized W-rich layer, the low mobility there will keep analyte molecules from returning quickly into the bulk MP, as evidenced by the high Alc residence times in the diffuse W layer (cf. Table 2).

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: tallarek@staff.uni-marburg.de. Fax: +49-(0)6421-2827065. Tel.: +49-(0)6421-28-25727.

■

CONCLUSIONS Our simulations have shown that the essential properties of the retentive W-rich layer at the bare-silica SP are conserved upon switching from a binary MP of 10/90 (v/v) W/ACN to a ternary MP of 5/90/5 (v/v/v) W/ACN/Alc: high W density, large extent of hydrogen bonding, and low translational mobility. Alc molecules insert themselves neatly into the Wrich layer, where their presence enforces its rigidity. The low extent of surface coordination by Alc proves that Alc molecules, due to their larger molecular size and lower versatility in intermolecular hydrogen bonding, cannot compete with W molecules for attachment to surface OH groups. Alc molecules tend to occupy the diffuse over the rigid part of the W-rich layer, more so in proportion to their size. At the same time, the Alc density in the W-rich layer decreases with the size of an Alc molecule; that is, the smaller the Alc molecule, the higher its hydrophilicity and presence in the W-rich layer. Alc molecules diversify the W-rich layer, providing alternative sites for hydrophobic interaction with analyte molecules. This could contribute to retention, but the main retentive effect originates (just as for binary W−ACN MPs) from the different W content in the W-rich layer compared with the predominantly organic MP. The experimental retention factors of analytes in the investigated HILIC separation show a dependence on the third solvent in a MP of 5/90/5 (v/v/v) W/ACN/third solvent that is similar to the ratio between the W mole fraction in the diffuse part of the W-rich layer and the bulk W mole fraction in our simulations. The identification of the diffuse rather than the rigid part of the W-rich layer as the main location of analyte retention tends to confirm partitioning as the dominant retention mechanism in the investigated HILIC separation. Beyond explaining the experimental data of a particular HILIC separation, our results hold implications for HILIC separations in general. Because partitioning is central to the HILIC retention mechanism, the W-rich layer at the surface of the nominal (solid) SP must be understood as an integral part of the functional SP. Via substitution of half of the W volume

Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS This work was supported by the Deutsche Forschungsgemeinschaft DFG (Bonn, Germany) under grant TA 268/7-1. REFERENCES

(1) Buszewski, B.; Noga, S. Anal. Bioanal. Chem. 2012, 402, 231−247. (2) Jandera, P. Anal. Chim. Acta 2011, 692, 1−25. (3) Alpert, A. J. J. Chromatogr., A 1990, 499, 177−196. (4) Grumbach, E. S.; Diehl, D. M.; Neue, U. D. J. Sep. Sci. 2008, 31, 1511−1518. (5) Gritti, F.; dos Santos Pereira, A.; Sandra, P.; Guiochon, G. J. Chromatogr., A 2010, 1217, 683−688. (6) Chauve, B.; Guillarme, D.; Cléon, P.; Veuthey, J.-L. J. Sep. Sci. 2010, 33, 752−764. (7) Karatapanis, A. E.; Fiamegos, Y. E.; Stalikas, C. D. J. Chromatogr., A 2011, 1218, 2871−2879. (8) Kumar, A.; Heaton, J. C.; McCalley, D. V. J. Chromatogr., A 2013, 1276, 33−46. (9) Periat, A.; Debrus, B.; Rudaz, S.; Guillarme, D. J. Chromatogr., A 2013, 1282, 72−83. (10) Mayer, V. R. Practical High-Performance Liquid Chromatography, 5th ed.; John Wiley & Sons, Ltd.: New York, 2010; p 82. (11) Neue, U. D. HPLC Columns: Theory, Technology and Practice; Wiley-VCH: New York, 1997; pp 164−174. (12) Gazała-Kopciuch, R.; Buszewski, B. J. Sep. Sci. 2003, 26, 1273− 1283. (13) Gritti, F.; Guiochon, G. J. Chromatogr., A 2013, 1297, 85−95. (14) Gritti, F.; Guiochon, G. J. Chromatogr., A 2013, 1302, 55−64. (15) Bicker, W.; Wu, J.; Yeman, H.; Albert, K.; Lindner, W. J. Chromatogr., A 2011, 1218, 882−895. (16) Vajda, P.; Felinger, A.; Cavazzini, A. J. Chromatogr., A 2010, 1217, 5965−5970. (17) Noga, S.; Bocian, S.; Buszewski, B. J. Chromatogr., A 2013, 1278, 89−97. (18) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. Angew. Chem., Int. Ed. 2012, 51, 6251−6254.

8855

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856

Analytical Chemistry

Article

(19) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. Anal. Chem. 2011, 83, 2569−2575. (20) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. J. Phys. Chem. C 2013, 117, 6620−6631. (21) Coasne, B.; Di Renzo, F.; Galarneau, A.; Pellenq, R. J.-M. Langmuir 2008, 24, 7285−7293. (22) van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C. J. Comput. Chem. 2005, 26, 1701−1718. (23) Gulmen, T. S.; Thompson, W. H. Langmuir 2006, 22, 10919− 10923. (24) Mountain, R. D. J. Phys. Chem. B 2010, 114, 16460−16464. (25) Mountain, R. D. J. Phys. Chem. C 2013, 117, 3923−3929. (26) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. J. Phys. Chem. 1987, 91, 6269−6271. (27) Wick, C. D.; Stubbs, J. M.; Rai, N.; Siepmann, J. I. J. Phys. Chem. B 2005, 109, 18974−18982. (28) Chen, B.; Potoff, J. J.; Siepmann, J. I. J. Phys. Chem. B 2001, 105, 3093−3104. (29) Marti, J. J. Chem. Phys. 1999, 110, 6876−6886. (30) McCalley, D. V. J. Chromatogr., A 2010, 1217, 3408−3417. (31) Martin, A. J. P.; Synge, R. L. M. Biochem. J. 1941, 35, 1358− 1368.

8856

dx.doi.org/10.1021/ac402123a | Anal. Chem. 2013, 85, 8850−8856