Adsorption of Water–Acetonitrile Mixtures to Model Silica Surfaces

Mar 13, 2013 - embraced work horse of chromatography, bare silica has been relegated to being the support ... layer,19 leaving an analyte to experienc...
1 downloads 0 Views 3MB Size
Article pubs.acs.org/JPCC

Adsorption of Water−Acetonitrile Mixtures to Model Silica Surfaces 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



S Supporting Information *

ABSTRACT: The retention mechanism of hydrophilic interaction chromatography relies on analyte partitioning from a mostly organic water−acetonitrile (W/ACN) mobile phase into an extended W-rich layer at the stationary phase. The formation of the W-rich layer is driven by the surface hydrophilicity of the stationary phase. We study the origin of the hydrophilicity of bare-silica stationary phases by molecular dynamics simulations of the adsorption of W/ACN mixtures from 99/1 to 2/98 (v/v) at three surfaces that each model one type of surface functional group: single silanol groups, geminal silanol groups, and siloxane bridges. Initiated by surface−W hydrogen bonding, the two silanol surfaces accumulate a dense W layer; their W surface excess adsorption isotherms remain positive over the whole W/ACN range. The siloxane bridges surface coordinates with an exceptionally tight layer of alternate W and ACN clusters; the mixed W/ACN surface layer is reflected in an S-shaped adsorption isotherm with W surface excess below 54/46 (v/v) W/ACN and ACN surface excess above 54/46 (v/v) W/ACN. Our results suggest that surface hydrophilicity is created largely by silanol groups, whereas siloxane bridges by offering adsorption sites to ACN molecules contribute in a different way to the retentive properties of bare-silica stationary phases.

I. INTRODUCTION The behavior of binary mixtures of water and an organic cosolvent at adsorbent surfaces is a pivotal point of many processes, such as the separation of chemical compounds (analytes) in liquid chromatography.1 A liquid aqueous− organic mobile phase transports the analytes through the chromatographic column, which contains the stationary phase on a solid support. Analytes are retained according to their interaction with the components of a layer of a few nanometers thickness at the adsorbent surface. As the transition between the solid support and the bulk mobile phase, the properties of this surface-adjacent layer lie between the solid and the liquid state, which is why it is referred to as the chromatographic interphase. Atomic-scale details of the retentive layer have been worked out for water−acetonitrile (W/ACN) mixtures at alkylmodified silica surfaces, that is, for reversed-phase liquid chromatography (RPLC),2−10 which still is the leading liquid chromatography mode. Because RPLC became the universally embraced work horse of chromatography, bare silica has been relegated to being the support structure for the actual stationary phase, the dense layer of hydrophobic alkyl chains,11,12 but the rise of hydrophilic interaction chromatography (HILIC)13−15 has sparked new interest in the stationary-phase properties of bare silica in combination with W/ACN mobile phases. HILIC targets hydrophilic analytes, including the many small, polar, uncharged compounds that are insufficiently separated by other liquid chromatography modes. The popularity of HILIC is based on its wide applicability and convenient operation, attributes that also apply to RPLC, but an intriguing aspect not © 2013 American Chemical Society

found in RPLC is that the selectivity can be manipulated over a rather wide range through changes in the mobile phase composition.16,17 This important and promising feature of HILIC is, however, scarcely understood, and even the retention mechanism of the simplest analytes on the simplest stationary phase, bare silica, remains a debated issue. Current consensus is that pure HILIC retention results from partitioning of the hydrophilic analytes from an ACN-rich mobile phase (usually ≥70 vol % ACN) into a W-rich layer at the stationary phase and from weak adsorption of the analytes to the stationary phase.18 The building of a HILIC knowledge base comparable to that for RPLC has just begun. Fundamental properties of the W-rich layer were recently elucidated by molecular dynamics (MD) simulations using a cylindrical pore carved into a β-cristobalite SiO2 block as a model for the porous silica particles that make up chromatographic beds.19,20 The pore surface was intended to approximate a bare-silica stationary phase, bearing single and geminal silanol groups as well as siloxane bridges at an overall surface hydroxylation of 8.0 to 8.5 μmol/m2; W/ACN mixtures at selected volumetric ratios represented typical HILIC mobile phases. It was found that over a distance of ca. 0.4 nm from the pore surface the solvent layer consisted of W molecules with nearly frozen translational mobility. The transition from this thin, surface-attached W layer to bulk solvent occurred not Received: December 19, 2012 Revised: March 9, 2013 Published: March 13, 2013 6620

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

silanol and siloxane groups are conjectures by the chromatography community rather than acquired knowledge because direct experimental evidence of the individual affinity of silanol and siloxane groups toward the components of a binary aqueous−organic mobile phase is lacking. In this work, we investigate by MD simulations how the components of a bare-silica surface contribute to its overall hydrophilicity and thus to the formation of the retentive W-rich layer required for HILIC separations. We employ three model silica surfaces derived from β-cristobalite SiO2 that each model one type of surface functional group: (i) the (111) face, which consists of regularly spaced, isolated, single silanol groups; (ii) the (001) face, which consists of geminal silanol groups, some of them connected to their vicinal neighbors by intrasurface hydrogen bonding; and (iii) a surface of siloxane bridges created upon the (111) face. The use of crystalline surfaces as models for the bare-silica surface has been rationalized by the similar density and refractivity as well as a degree of short-range order preserved in amorphous silica.11 The (111) face is the most widely accepted model for silica-based stationary phases: it has the average surface hydroxylation of bare silica and bears single silanol groups, presumably the major fraction among the functional groups at the amorphous silica surface. The (001) face models geminal silanol groups, estimated to account for 15−30% of all silanol groups on the amorphous silica surface. The third surface explores the effect of exposed, solventaccessible siloxane bridges; these are rarely considered in amorphous silica models,29 although chromatographers attribute to them the weak reversed-phase properties of bare silica exhibited with W-rich mobile phases.25−27 By comparison of the (111) and the (001) face we investigate the effect of density and configuration of surface OH groups, whereas by comparison of the silanol surfaces with the siloxane bridges surface we focus on the effects resulting from the presence or absence of surface−solvent hydrogen bonding. In contrast with our previous work,19,20 the MD simulations are not carried out for a cylindrical silica pore but for a silica slab placed between two solvent reservoirs. Apart from requiring shorter simulation times, this configuration is preferable for queries of a more general nature as it eliminates the effect of a nanoscale confinement on the liquid properties (particularly on the solvent mobility). Compared with the adsorption of W/ACN mixtures to silica surfaces,19,20,30,31 the bulk liquid properties of W/ACN mixtures32−35 and the adsorption of neat W36−42 and neat ACN43−47 to silica surfaces are well investigated. Through simulating the adsorption of W/ACN mixtures between 99/1 and 2/98 (v/v) to model surfaces with specified properties, we probe the solvent-selectivity of the different types of functional groups on the amorphous silica surface. In our study, we focus on the competitive behavior of W and ACN molecules; complementary neat solvent data show what happens when the competition between the two solvent species ceases. By establishing the W surface excess adsorption isotherms for each model silica surface, we provide a basis to estimate the contribution of each type of surface functional group to the retentive properties of bare-silica stationary phases in HILIC.

abruptly but through an interface region, populated predominantly (but not exclusively) by W molecules with restricted (but not frozen) mobility. The properties of the first, rigid W layer were entirely governed by the silica surface and remained insensitive toward a changing bulk W/ACN ratio; the properties of the second, diffuse W layer reflected a gradually declining surface influence and concurrent return to bulk values. Combined, rigid and diffuse parts of the W-rich layer extended for ca. 1.5 nm, sufficient for partitioning of small molecules. The reduced mobility in the W-rich layer was supposed to foster analyte retention by slowing down molecular motion and keeping the analytes close to the stationary phase. The retentive W-rich layer occupied >50 vol % of a typical-size (9 nm diameter) HILIC pore;20 a smallersized pore (3 nm diameter) was entirely filled by the W-rich layer,19 leaving an analyte to experience a very different environment inside than outside the porous silica particles of a chromatographic bed. The hydrophilicity of the pore surface was identified as the driving force for the formation of the Wrich layer and thus as the necessary condition for HILIC retention through the partitioning mechanism. In the strict sense of the word, the term “hydrophilic” describes the affinity of a molecule or a functional group for W. In chromatography especially, the terms “hydrophilic” and “polar” are often used synonymously, as Bicker et al.17 have pointed out. This practice is confusing not only because “polar” and “hydrophilic” are not interchangeable but also because “polarity” requires a specific context to have unequivocal meaning. A solvent molecule’s polarity is given by its dipole moment, measured in the gas phase, but a solvent’s polarity is judged by the dielectric constant (relative permittivity) of the bulk liquid, and the solvent’s elution strength in a given chromatographic mode is described by empirical parameters and depends on the solvent’s interaction with a particular adsorbent.21 ACN has ca. twice the dipole moment of W (μ = 3.9 D for ACN vs μ = 1.8 D for W) but only half of its permittivity (εr = 37.5 for ACN vs εr = 78.3 for W),22 because of the different organization of W and ACN molecules in the bulk liquid state. The hydrogen-bonding network of W molecules enhances the polarity of liquid W, whereas antiparallel dipole−dipole pairing reduces the polarity of liquid ACN.23,24 The dipole moment is a single-molecule property, and the dielectric constant is a molecular-ensemble property. When chemical surfaces are involved, the situation becomes complicated: Surface interaction is undergone by single solvent molecules, but solvent−solvent interactions remain present, so that none of the traditional connotations of solvent polarity apply. Chemical surfaces, especially those used as stationary phases, are also categorized as “polar” or “nonpolar”. A polar surface bears a majority of functional groups whose constituting atoms have a moderate to large electronegativity difference. All functional groups on a bare-silica surface are polar, but chromatographers consider silanol groups as hydrophilic and siloxane bridges as hydrophobic.25−27 The hydrophobic character of siloxane groups is deduced from the bonding in the Si−O−Si moiety, assumed to contain a contribution from dπ−pπ electron interaction,28 as a consequence of which the capability of the O atom to act as hydrogen-bond acceptor is lost. However, the notion of siloxane groups as hydrophobic is difficult to reconcile with their polar nature, particularly because true hydrophobic stationary phases (for RPLC) have alkylfunctionalized, nonpolar surfaces. The properties ascribed to

II. SIMULATIONS Model Silica Surfaces. Silica slabs with three different surfaces were prepared from β-cristobalite SiO2 following an approach of Coasne et al.48 (Figure S1 in the Supporting Information shows views onto the three model silica surfaces.) 6621

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

Figure 1. Adsorption of a liquid W/ACN mixture (5/95, v/v) to model silica surfaces covered with (A) single silanol groups (SiS surface), (B) geminal silanol groups (GeS surface), or (C) siloxane bridges (SB surface). Snapshots were taken after 12 ns of simulation time. Atoms are colorcoded as follows: Si − yellow, O − red, H − white, N − blue, C and CH3 − cyan. The dimensions of the simulation box and of its compartments are indicated.

A surface of single and isolated silanol groups (SiS surface) was received after cutting the initial crystal structure parallel to the (111) face, removing all undercoordinated Si atoms and all unconnected O atoms, and saturating the dangling bonds of O atoms with H atoms. A surface of geminal silanol groups (GeS surface) was prepared similarly after cutting the crystal structure parallel to the (001) face. The silanol groups on the SiS surface are too far apart (0.55 nm) to form intrasurface hydrogen bonds (HBs), which are possible on the GeS surface between the closest neighbored (vicinal) silanol groups, whose O atoms are 0.253 nm apart.49 The hydroxylation of the SiS surface (4.5 OH groups/nm2 or 7.6 μmol OH groups/m2) corresponds to the average value reported for amorphous silica, whereas the hydroxylation of the GeS surface (7.8 OH groups/ nm2 or 13.3 μmol OH groups/m2) is at the high limit reported for amorphous silica.1 A surface of siloxane bridges (SB surface) was prepared by cutting the crystal structure parallel to the (111) face, followed by removing all undercoordinated Si atoms and all unconnected O atoms. Next, all O atoms with dangling bonds were pairwise substituted by placing an O atom at equal distance from the two surface Si atoms to which the original O atoms had belonged; this left each surface Si atom connected to three inner O atoms and one exposed O atom. In trial simulations, the ACN residence times at the SB surface were infinite within the time frame of the simulation (up to hundreds of nanoseconds). Although residence times in the millisecond range are not improbable from an experimental point of view, they are undesirable for simulations. To ensure that global equilibrium is reached within the simulation time frame, molecules should exchange two to three times even at the slowest-exchanging position. To achieve this, the first inner layer of O atoms (bearing negative partial charges) was moved 0.025 nm closer to the surface Si atoms, which helped to screen the positive partial charges of the solvent-accessible Si atoms. The slight modification decreased the ACN residence times at the SB surface to finite values, whereas the mobility of all other solvent molecules as well as the solvent density profiles,

hydrogen-bonding patterns, and orientational arrangement underwent only small, insignificant changes. Simulation Box and Force Field Parameters. The quadrilateral, fully periodic simulation boxes (with dimensions as shown in Figure 1) contained a central silica slab between two solvent reservoirs. A length of 2.5 nm was chosen for the solvent reservoirs to fully capture the transition of the liquid from surface-influenced to bulk behavior. For the model silica surfaces, we used the force-field parameters of Gulmen and Thompson50 because they provide van der Waals and electrostatic potential parameters for all atoms of the silica surface (Si, O, and H atoms of silanol and siloxane groups). Treating all surface atoms as possible solventinteraction sites with appropriate electrostatic and van der Waals potentials is a prerequisite to an unbiased investigation. The motion of the silica atoms was frozen during simulations, except for the free rotation of silanol H atoms. For the solvents, which were treated as rigid, three-site molecules, we used the simple point charge/extended (SPC/E) model51 for W and the united-atom version of the transferable potentials for phase equilibria (UA-TraPPE) force field52 for ACN. This combination has been validated to reproduce the experimental properties of W/ACN mixtures, particularly the liquid density and the extent of hydrogen bonding between W and ACN.32 Long-range electrostatic interactions were treated with the particle-mesh Ewald algorithm (with a real space cutoff of 1 nm). Nonbonded interactions were modeled with a 12−6 Lennard-Jones potential and truncated at 1 nm. Lennard-Jones parameters for unlike interactions were calculated using conventional combination rules. Molecular Dynamics Simulations. MD simulations were carried out with Gromacs 4.5.2.53 A Nosé-Hoover thermostat with a coupling constant of 0.1 ps was used for temperature control. The equations of motion were integrated with a time step of 1.0 fs. Each simulation run was conducted for a canonical NVT ensemble (constant number of particles, N, simulation box volume, V, and temperature, T), but manual adjustments of the solvent composition (i.e., changes in N) 6622

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

Hydrogen-Bonding Criteria. To determine whether two solvent molecules (W−W, W−ACN) or one solvent molecule and a surface OH group (surface−W, surface−ACN) or two surface OH groups (intrasurface) were positioned within hydrogen-bonding distance to each other, we applied three geometrical criteria:54 (i) distance between donor O atom and acceptor X atom (X = N, O) less than 0.35 nm, (ii) distance between donor H atom and acceptor X atom less than 0.25 nm, and (iii) angle between the direction of the O···X vector and the OH-bond vector of the donor less than 30°. Determination of Solvent Translational Mobility. The translational mobility of W and ACN molecules was determined for the bulk region (Z > 2.0 nm) and for each peak in the solvent density profiles (for a space interval of ±0.05 nm around the center coordinate). The translational molecular movement orthogonal to the surface (residence time τ) was calculated as the average time solvent molecules remained within a given space interval. To account for the perpetual molecular motion, a shift of ±0.1 nm around a molecule’s initial Z-coordinate was tolerated. Following an approach of Liu et al.,55 the translational molecular movement parallel to the surface (parallel diffusion coefficients, D∥) was calculated for each space interval from the mean-square displacements of the solvent molecules, ⟨r2(t)⟩, made parallel to the surface during the observation time, t. The linear slope of the curve during the time interval t = 4−20 ps was used to determine D∥ via the Einstein equation

were made between successive runs pertaining to the simulation for a given mixture composition and model silica surface. Simulations were carried out for eight mixture compositions, 99/1, 80/20, 50/50, 30/70, 20/80, 10/90, 5/95, and 2/98 (v/ v) W/ACN, as well as for neat W and neat ACN. At the beginning of a simulation, the respective number of solvent molecules was determined from the targeted W/ACN ratio. W and ACN molecules were then distributed randomly in the solvent reservoirs on the basis of nonoverlapping van der Waals radii. After energy minimization, initial velocities were randomly assigned according to a Maxwell−Boltzmann distribution before MD simulations were carried out at T = 300 K. In a chromatographic column, a continuously replenished mobile phase of defined composition percolates between the adsorbent silica particles. To mimic the environment of a silica particle in a chromatographic bed in our simulations, we pursued a constant solvent composition in the bulk region of the reservoirs through preliminary simulation runs. After equilibration of the system, which took between 4 and 6 ns, the actual W/ACN ratio in the bulk region of the reservoirs (at a distance of Z > 2.0 nm from the surface Si atoms) was checked and corrected manually by removing or adding the necessary solvent molecules. Four to five preliminary simulation runs (each 12 ns long) were necessary to steer the bulk solvent composition toward the targeted W/ ACN ratio (tolerating a deviation of ±1%). The respective number of W and ACN molecules used in the productive simulation runs for each nominal W/ACN ratio and model silica surface is listed in Table 1. Productive simulation runs

D = lim

t →∞

SiS surface

GeS surface

SB surface

(v/v)

NW

NACN

NW

NACN

NW

NACN

100/0 99/1 80/20 50/50 30/70 20/80 10/90 5/95 2/98 0/100

1935 1920 1640 1180 808 633 430 330 225 0

0 6 107 279 393 452 526 562 580 658

2020 2010 1725 1243 907 702 487 369 257 0

0 5 111 285 398 465 536 576 592 668

1985 1858 1490 1058 750 576 391 278 180 0

0 50 189 337 449 510 568 612 630 691

(1)

Only molecules that remained in the space interval for the whole observation time were counted. Determination of W Surface Excess Adsorption Isotherms. For the determination of the W surface excess adsorption isotherms, we followed the strict recommendations of Everett,56 thereby eliminating the need for a Gibbs dividing surface, which cannot be unequivocally defined for the surface adsorption of a binary liquid phase. The areal reduced surface excess of W, Γ(n) W , was calculated from

Table 1. Number of W (NW) and ACN Molecules (NACN) per Nominal W/ACN Ratio Used in the Productive Simulation Runs for Each Model Silica Surface W/ACN

⟨r 2(t )⟩ 4t

Γ(Wn) =

1 l l (nW xACN − nACNx W ) AS

(2)

where AS is the surface area (according to the lateral dimensions of each simulation box as shown in Figure 1), nW and nACN are the amounts of W and ACN, respectively, in the simulation box, and xlW and xlACN are the mole fractions of W and ACN, respectively, in the bulk liquid phase (i.e., the bulk region of the reservoirs, Z > 2 nm).

lasted 12 ns; configurations in the second half of the trajectory were saved every 2 ps for data analysis. For the calculation of solvent residence times, which required a finer time resolution, data were saved every 0.2 ps in trajectories of 400 ps. Longer simulations times (up to 300 ns) were necessary to estimate residence times at the SB surface, where solvent exchange was slow. In previous work,19,20 we have ascertained the independence of the equilibrium configuration from the starting conditions through trial simulations, in which the solvent molecules were initially placed opposite to their equilibrium positions. While requiring longer equilibration times, these biased starting conditions lead to the same equilibrium configurations as the random initial placement of solvent molecules.

III. RESULTS AND DISCUSSION Figure 1 shows simulation snapshots of the adsorption of a liquid W/ACN mixture (5/95, v/v) to each of the three model silica surfaces. Besides visualizing the topology of the surface models and the general design and dimensions of the simulation box, these snapshots capture the essence of the respective solvent selectivities of the three types of silica surfaces. Despite the low W supply in the system, the single and geminal silanol groups of the SiS and the GeS surface, respectively, are covered by a dense W curtain, which keeps ACN molecules away from the surface (Figure 1A,B). The solvent layer at the SB surface (Figure 1C) is composed of alternate ACN and W clusters, 6623

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

whose arrangement traces the Si−O−Si atomic pattern of the solvent-accessible siloxane bridges. In the following, atoms are designated as follows: SiS, surface silicon atom; OS, oxygen atom of surface OH group or exposed oxygen atom of siloxane bridge; HS, hydrogen atom of surface OH group; OW, oxygen atom of W; HW, hydrogen atom of W; CACN, central carbon atom of ACN; and NACN, nitrogen atom of ACN. The presented data refer to the average from both solvent reservoirs. Solvent Density Profiles. For a quantitative analysis of the adsorptive behavior of binary W/ACN mixtures to the three model silica surfaces, we first turn to the solvent density profiles of W and ACN, calculated from the atom number density ρn of OW for W and of CACN for ACN. Figures 2−4 show the local Figure 4. Solvent density profiles for W (ρn of OW) and ACN (ρn of CACN) at the siloxane bridges (SB) surface for nominal (bulk) solvent compositions of 99/1, 80/20, 50/50, 30/70, 20/80, 10/90, 5/95, and 2/98 (v/v) W/ACN as well as the neat solvents.

= 0.35 nm. The number density of this composed peak is governed by the surface, with the density of peak 1A being nearly unaffected by the actual W supply in the system and the density of peak 1B moving only little with the W/ACN ratio. Contrary to the W surface peak, the density of the next, broader peak (2) at Z = 0.51 nm reflects the influence of the W/ACN ratio already more than that of the surface. Beyond peak 2, W profiles start to level out toward the bulk values. The distance over which an influence of the surface on the solvent density profiles can be detected is ca. 1 nm for W and ca. 1.5 nm for ACN (Figure 2). ACN profiles start only beyond the W surface peak. (A few ACN molecules are present in the surface-W layer, but their number is too small to make a peak in the density profile.) The W-covered surface affects the ACN distribution as a physical barrier, causing up to three peaks in the ACN density profiles. The density of the first ACN peak (Z = 0.54 to 0.58 nm, the peak maximum shifts closer to the surface with decreasing W/ACN ratio) tops that of the second ACN peak (Z = 0.95 nm) only at 2/98 and 5/95 (v/v) W/ACN; that is, ACN approaches the W-covered surface only at low W supply in the system. In summary, the SiS surface proves to be very hydrophilic under HILIC conditions; as a consequence of the strong preference for W over ACN, the W density at the SiS surface is at each W/ACN ratio higher than in the bulk. The solvent density profiles of the GeS surface (Figure 3; SiS atoms at Z = 0 nm, OS atoms at Z = 0.09 nm, HS atoms at Z = 0.04 and 0.11 nm) closely resemble those of the SiS surface, allowing for the different configuration of single and geminal silanol groups. OS and HS atoms have a shorter distance to the SiS atom in geminal than in single silanol groups; the HS atoms of a geminal silanol group can turn back to the surface or stretch toward the bulk.41 Thus, solvent density profiles start at lower Z values compared with the SiS surface, and the W surface peak is split into two distinct components, 1A at Z = 0.17 nm and 1B at Z = 0.35 nm, in agreement with the two layers of HS atoms. For steric reasons, the second part of the surface peak (1B) carries the maximum density, except at very low W supply (2/98 (v/v) W/ACN). Peak 1B is followed by a shoulder (2 at Z = 0.45 nm) and a weakly defined third peak (3 at Z = 0.75 nm) before the W profiles flatten out. ACN profiles are essentially as observed for the SiS surface, with up to three peaks starting beyond the W-populated surface region. The first ACN peak (Z = 0.45 nm) gathers density and becomes skewed

Figure 2. Solvent density profiles for W (atom number density ρn of the oxygen atom, OW) and ACN (ρn of the central carbon atom, CACN) at the single silanols (SiS) surface for nominal (bulk) solvent compositions of 99/1, 80/20, 50/50, 30/70, 20/80, 10/90, 5/95, and 2/98 (v/v) W/ACN as well as the neat solvents.

Figure 3. Solvent density profiles for W (ρn of OW) and ACN (ρn of CACN) at the geminal silanols (GeS) surface for nominal (bulk) solvent compositions of 99/1, 80/20, 50/50, 30/70, 20/80, 10/90, 5/95, and 2/98 (v/v) W/ACN as well as the neat solvents.

solvent composition at the three model silica surfaces in dependence of the bulk (nominal) liquid composition, from 99/1 to 2/98 (v/v) W/ACN. For easy reference throughout the text, peaks in the W profiles are indexed. Figure 2 starts with the SiS surface (SiS, OS, and HS atoms at Z = 0, 0.16, and 0.2 nm, respectively). As seen in the simulation snapshot (Figure 1A), the solvent layer closest to the surface is made up entirely of W molecules, giving rise to a tall, sharply defined W peak (1A) at Z = 0.26 nm with a shoulder (1B) at Z 6624

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

toward the surface for ratios ≤5/95 (v/v) W/ACN. This skewing, similar to the shifting of the first ACN peak toward the SiS surface (Figure 2), reflects how ACN molecules fill up the space at low W/ACN ratios, when the W curtain becomes thinner. Nevertheless, Figures 2 and 3 show that ACN molecules have no sizable density at the very hydrophilic silanol surfaces as long as W is present in the system. The solvent density profiles of the SB surface (Figure 4; SiS atoms at Z = 0 nm, exposed OS atoms at Z = 0.09 nm) present a contrasting picture to the situation at the two silanol surfaces, as both solvent species are present at the SB surface. ACN molecules form a major part of the surface layer, as evidenced by the tall and narrow ACN peak at Z = 0.27 nm. Except for a dip at the low ACN limit (99/1 (v/v) W/ACN), the density of the ACN surface peak changes only slowly with the W/ACN ratio. The density of the W surface peak (1 at Z = 0.30 nm) shows comparable sensitivity to the W/ACN ratio as the colocalized ACN peak. This perceived influence of the W/ACN ratio on the density of the surface peaks reflects that coordination of the SB surface involves both solvent species, as opposed to the silanol surfaces, where W is selected over ACN irrespective of the nominal W/ACN ratio. The mixed solvent layer at the SB surface is separated from the remaining solvent body by a distinctive “gap”, a region of low solvent density. This gap can also be recognized in the simulation snapshot (Figure 1C). W is the first solvent to recover its density after the gap, and, as opposed to ACN, W also maintains some density in the gap. The densities of the three remaining W peaks (2, 3, and 4 at Z = 0.57, 0.67, and 0.95 nm, respectively) reflect the gradually declining influence of the surface. With up to three further peaks, ACN profiles beyond the gap are highly similar to those of the SiS surface. Overall, the SB surface has approximately equal affinity for both solvent species. W and ACN molecules are sucked to the surface, so that even at low supply the density of each solvent species at the surface is always higher than in the bulk. We conclude our analysis of the solvent density profiles by briefly comparing the mixture data with those obtained for the neat solvents. The neat W profiles of the two silanol surfaces (Figures 2 and 3) simply top off the mixture profiles, whereas the neat ACN profiles differ substantially from the mixture profiles. In the absence of W, ACN molecules take over surface coordination, which is why neat ACN profiles start closer to the surface than the mixture profiles. The surface coordination by ACN molecules is also apparent in the split and doubly split peaks at the SiS and the GeS surfaces (Figures 2 and 3), respectively. The neat ACN profile of the SB surface (Figure 4) differs from its mixture counterparts mostly insofar as there is now necessarily some ACN density in the gap region, previously occupied by W rather than ACN molecules. In contrast with the silanol surfaces, the W profile of the SB surface changes visibly in the absence of ACN: the density of the surface W peak (1) is raised, as is the density of the next W peak (2) at the expense of its neighbor peak (3). The increased density of the surface peak upon changing from a W/ACN mixture to the neat solvent observed for W, but not for ACN, reveals that the SB surface can draw more W molecules from the neat solvent than from a binary mixture but not more ACN molecules. Apparently, the SB surface has a given capacity for ACN, which it tries to satisfy over most of the W/ACN range and cannot increase much in neat solvent due to an upper limit for the volume density of ACN molecules at the surface. The capacity of the SB surface for W is more flexible because W

molecules occupy only about one-third of the volume of an identical number of ACN molecules. With four connective sites for intermolecular hydrogen bonding, the small W molecules can be more easily accommodated at the surface than the larger, linear ACN molecules with their bulky methyl groups. Incidentally, the smaller molecular size and the higher versatility for intermolecular coordination is also the reason why in W/ACN mixtures W (not ACN) molecules span the gap between the solvent layer at the surface and the remaining solvent body (Figure 4). Hydrogen Bonding. As shown in Figure 5 for the SiS surface, the first W peak and its shoulder (1A and 1B, cf. Figure 2) consist predominantly of W molecules engaged in HBs with the surface.

Figure 5. Hydrogen bonding at the SiS surface for selected bulk compositions of 80/20 (red), 30/70 (green), and 5/95 (blue) W/ ACN (v/v). (A) Density profiles of surface-attached (dashed lines) and unattached (solid lines) W molecules. (B) Average number of hydrogen bonds per W molecule (HBW) for W−surface (dashed lines), W−W (thick solid lines), and W−ACN (thin solid lines) hydrogen bonds.

The number density of surface-attached W molecules is unaffected by the W/ACN ratio, which proves that the selectivity of the SiS surface for W over ACN rests on hydrogen bonding between W molecules and surface OH groups. The number of HBs per W molecule (HBW) with the surface drops from a value around HBW ≈ 1.7 in the first W peak (1A) to HBW ≈ 0.8 in the shoulder region (1B), whereas W−W HBs concurrently increase from HBW = 1.2 to 1.5 to HBW = 2 to 2.5 (depending on the W/ACN ratio). The number of W−W HBs reaches its maximum at Z = 0.51 nm (peak 2 in the W density profiles, cf. Figure 2), exactly at the point where HBs with the surface have ceased. With HBW = 2.5 to 3.2, W−W coordination around peak 2 remains high even as the density of this peak decreases at a sinking W fraction in the reservoirs. The explanation for this behavior is that surfacebound W molecules form further HBs with W rather than with ACN molecules (as witnessed by the slow increase of HBW for W−ACN HBs, shown as thin, solid lines in the bottom panel of Figure 5), so that ACN molecules are driven from the surface and reach sizable density only beyond peak 2. The formation of additional W−W HBs by the surface-attached W molecules supports the building of a closed W layer over the surface, but also expands the HB network beyond the immediate surface region so that the W density stays above the bulk value up to Z = 1.0 nm (cf. Figure 2). HB statistics complement the information given in Figure 5. The SiS surface maintains an average of two HBs per surface 6625

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

OH group (HBOH ≈ 2) to W molecules (Table 2). As expected from the solvent density profiles (cf. Figure 2), surface−ACN Table 2. Average Number of Hydrogen Bonds Per Surface OH Group (HBOH) at the SiS and the GeS Surfaces W/ACN

SiS surface

(v/v)

with W

100/0 80/20 30/70 5/95 0/100

2.12 2.11 2.06 2.00

GeS surface

with ACN 0.0016 0.002 0.005 0.94

with W 1.31 1.31 1.29 1.22

with ACN

intrasurface

0.0006 0.002 0.004 0.439

0.43 0.43 0.43 0.43 0.51

Figure 6. Hydrogen bonding at the GeS surface for selected bulk compositions of 80/20 (red), 30/70 (green), and 5/95 (blue) W/ ACN (v/v). (A) Density profiles of surface-attached (dashed lines) and unattached (solid lines) W molecules. (B) Average number of hydrogen bonds per W molecule (HBW) for W−surface (dashed lines), W−W (thick solid lines), and W−ACN (thin solid lines) hydrogen bonds.

interaction is negligible, except in neat ACN, when the surface donates HS atoms to NACN atoms (HBOH = 0.94). In W/ACN mixtures, the predominant configuration of a surface OH group at the SiS surface (Table 3) is a donor bond with one W Table 3. Hydrogen-Bond Configurations (Number of Donor (D) and Acceptor Bonds (A)) of Surface OH Groups at the SiS Surface W/ACN (v/v)

0D0A

1D0A

0D1A

1D1A

100/0 80/20 30/70 5/95 0/100

0.01 0.01 0.01 0.01 0.06

0.16 0.16 0.16 0.17 0.94

0.02 0.02 0.02 0.02

0.52 0.53 0.56 0.59

0D2A

the number of ACN molecules that can be accommodated at the surface is smaller than the number needed for complete coordination, ACN molecules are more often positioned within hydrogen-bonding distance to two neighboring silanol groups rather than forming one-to-one surface−ACN HBs. The lower accessibility of the OH groups at the GeS surface is also the reason why intrasurface and surface−W HBs combined (HBOH ≤ 1.74) do not reach the coordination of the SiS surface (HBOH ≤ 2.11) in W/ACN mixtures (Table 2). The SiS−OS bond vector is perpendicular to the SiS surface but inclined toward the GeS surface; the former configuration provides better approachability and thus enables a larger number of HBs per surface OH group than the latter configuration. At the SB surface, W molecules form HBW ≈ 1.5 with other W molecules and HBW ≈ 0.5 with ACN molecules (Figure 7). Because the SB surface blocks an expansion of the HB network to the solid side, the overall coordination of W molecules at the SB surface (total HBW ≈ 2, dominant configuration 1D1A) is smaller than that at the silanol surfaces (total HBW ≈ 3, preferred configuration 2D1A, cf. Figures 5 and 6). However, the number of W−W HBs increases immediately after peak 1 and reaches its maximum at peak 2 with HBW = 2.5 to 3.2, just as observed for the two silanol surfaces. (The dominant configuration for W molecules of peak 2 is 2D1A or 2D2A. See Tables S7 and S8 in the Supporting Information for statistics.) The high coordination of W molecules around peak 2 in the solvent density profiles of the silanol surfaces stems from the surface attachment of W molecules. Although the SB surface cannot form HBs, it is more than just a geometrical barrier to the liquid. Because of electrostatic attraction between specific solvent and surface atoms, that is, between OS and HW atoms and between SiS and NACN atoms, the SB surface sorts the two solvent species into (laterally) alternating clusters. The W−W HBs formed in the small W clusters propagate the preference for W−W over W−ACN HBs further into the liquid phase and so extend the dominance of W−W over W−ACN HBs beyond the immediate surface region. The analysis of the hydrogen-bonding patterns (Figures 5−7) has explained the salient features of the solvent density profiles at the three model silica surfaces (cf. Figures 2−4). The silica surfaces disturb the liquid structure either by selecting W over ACN (SiS and GeS surfaces) or by laterally separating W and ACN (SB surface). In each case, the disturbance initiated at

1D2A 0.29 0.28 0.25 0.21

molecule and an acceptor bond with another (1D1A, where D stands for donor and A stands for acceptor bond). W molecules generally prefer the 2D1A configuration or at high W/ACN ratio the 2D2A configuration. Typically, a W molecule of peak 1A has one donor and one acceptor bond with the SiS surface and extends a further donor bond to another W molecule; a W molecule of the shoulder peak (1B) has an acceptor bond with the SiS surface and two further HBs with W molecules. (Tables S1−S3 in the Supporting Information contain comprehensive HB statistics for W molecules at the SiS surface.) The HB network at the GeS surface (Figure 6) displays the same characteristics as those at the SiS surface, adapted to two layers of HS atoms. The hydrogen-bond coordination of W molecules is also essentially as in the previous example (see Tables S4−S6 in the Supporting Information for statistics), but as opposed to the OH groups of the SiS surface, which clearly prefer the 1D1A configuration, the configuration of OH groups at the GeS surface is distributed over all possibilities (Table 4). The GeS surface has a lower number of surface−W HBs (HBOH = 1.22 to 1.31 between 5/95 and 100/0 (v/v) W/ACN) than the SiS surface (Table 2) because the GeS surface also maintains a significant percentage of intrasurface HBs (HBOH = 0.43). Interestingly, the percentage of intrasurface HBs is insensitive toward the W/ACN ratio as long as W remains present in the system. The fact that exposure to W does not weaken intrasurface hydrogen bonding has also been observed in previous simulations.42 In neat ACN, the number of intrasurface HBs is increased (HBOH = 0.51) in addition to HB formation with ACN molecules (HBOH = 0.44) to partially compensate the loss of HBs with W (Table 2). Nevertheless, 30% of OH groups remain inaccessible and uncoordinated in neat ACN (compared with only 6% at the SiS surface). Because 6626

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

Table 4. Hydrogen-Bond Configurations (Number of Donor (D) and Acceptor Bonds (A)) of Surface OH Groups at the GeS Surface W/ACN (v/v)

0D0A

1D0A

2D0Aa

0D1A

1D1A

2D1Aa

0D2A

1D2A

100/0 80/20 30/70 5/95 0/100

0.07 0.07 0.07 0.08 0.30

0.14 0.14 0.15 0.15 0.21

0.08 0.08 0.09 0.09 0.24

0.16 0.17 0.17 0.18 0.25b

0.20 0.20 0.19 0.19

0.13 0.13 0.13 0.13

0.20 0.19 0.18 0.17

0.02 0.02 0.02 0.01

a HS atoms of these surface OH groups are within hydrogen-bonding distance to two possible acceptor atoms. bIn neat ACN, acceptor bonds are intrasurface hydrogen bonds.

surface. W molecules of peak 1A at the GeS surface have highly similar orientation as their counterparts at the SiS surface, whereas W molecules of peaks 1B and 2 retain a bit more orientational preference at the GeS than at the SiS surface. The distance of peak 1B to the OS atoms of the GeS surface (0.35 nm − 0.09 nm = 0.26 nm) is larger than the corresponding distance at the SiS surface (0.35 nm − 0.16 nm = 0.19 nm). To facilitate an HB between a HW and an OS atom, most W molecules of peak 1B at the GeS surface therefore align one OH-bond vector nearly perpendicular to the surface. Peak 2, on the other hand, is located closer to the GeS surface than to the SiS surface; whereas surface−W HBs have tapered off before peak 2 at the SiS surface, peak 2 at the GeS surface contains a fraction of surface-tethered W molecules that contribute their orientational preference to the overall distribution (cf. Figures 5 and 6). The organization of W molecules at the SB surface is governed by the electrostatic attraction between HW and the exposed OS atoms and is supported by W−W and W−ACN hydrogen bonding. W molecules of peak 1 align their dipole vector nearly perpendicular to the SB surface so that both HW atoms are facing the OS atoms; W molecules of peaks 2 and 3 still point one HW atom to the SB surface. ACN orientation is also highly similar at the two silanol surfaces. ACN is organized in double layers of approximately antiparallel oriented molecules. The simulation snapshots of Figure 1A,B show the orientation of ACN molecules to be distributed over a wide range of angles, not as highly ordered as in a lipid bilayer. ACN molecules in the sublayer closer to the surface prefer to turn their NACN atoms toward it, whereas ACN molecules in the sublayer closer to the bulk liquid turn their methyl groups to the surface. Beyond the immediate surface region, the effect looses rapidly in strength but persists through the interface region (up to Z = 1.5 nm). Owing to the narrower spatial distribution of HS atoms at the SiS surface, ACN molecules have a higher degree of orientational order there than at the GeS surface. Surprisingly, ACN orientation in W/ACN mixtures remains essentially as observed, also in prior simulations and experiments,44−47 for neat ACN. In the absence of W, ACN molecules interact directly with the silanol surface through hydrogen bonding between HS and NACN atoms. This is reflected by the split surface peaks in the neat ACN density profiles of the silanol surfaces (cf. Figures 2 and 3, also observed by others45,46). The peak portion closer to the surface represents surface-tethered ACN molecules (at the GeS surface, this peak portion is split once again, in agreement with two layers of HS atoms); the peak portion farther from the surface represents oppositely oriented ACN molecules. In binary W/ACN mixtures, ACN molecules face the W curtain at the surface preferentially with their NACN atoms. This demonstrates that direct surface−ACN interaction is not necessary to instigate the observed orientational pattern of

Figure 7. Hydrogen bonding at the SB surface for selected bulk compositions of 80/20 (red), 30/70 (green), and 5/95 (blue) W/ ACN (v/v). (A) Density profiles of W molecules. (B) Average number of hydrogen bonds per W molecule (HBW) for W−W (thick solid lines) and W−ACN (thin solid lines) hydrogen bonds.

and by the surface extends beyond the immediate surface region and causes the preferred formation of W−W HBs over W−ACN HBs. The invariantly high density of the first W peak (1A) at the silanol surfaces demonstrates the immense driving force of surface−W HB formation. In the absence of W, surface coordination is attempted by ACN molecules, but the surface− solvent interaction is far from reaching the same strength as in W, because (i) surface−ACN HBs (HS···NACN) are weaker than surface−W HBs (OS···HW and HS···OW), (ii) the maximum number of HBs with the surface per solvent molecule is HBOH = 1 for ACN compared with HBOH = 3 for W, (iii) surfaceattached ACN molecules lack the cooperative effect of intermolecular hydrogen-bonding exercised by surface-attached W molecules, and (iv) the directionality of the surface−ACN interaction in conjunction with the comparatively large volume of ACN prohibit one-to-one coordination of all surface OH groups. Solvent Orientation. We now supplement the information gathered so far with details of the orientation of W and ACN molecules at the three model silica surfaces. (See Figures S2− S5 in the Supporting Information for quantitative data.) At the silanol surfaces, the orientation of W molecules is expectedly governed by surface−W hydrogen bonding. In general, W orientation at the silanol surfaces is overall well-defined in the first two density peaks (1A and 1B) but more broadly distributed in the next peak (2), as expected of surface-attached versus unattached W molecules. Most W molecules at the SiS surface (peak 1A) orient OW and one HW atom to the surface, thus facilitating one donor and one acceptor HB with the surface OH groups; the second HW atom faces the bulk liquid and forms the connection to other W molecules. W molecules of peak 1B point one HW atom to the surface, whereas the OW atom may or may not be available for hydrogen bonding to the 6627

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

ACN molecules. Rather, the governing effect of the surface is conveyed through the W curtain, which presents a positive net charge to the ACN molecules. This explanation agrees with previous studies of liquid ACN at silica surfaces,44,47 where nonspecific electrostatic interactions (i.e., other than hydrogen bonding) were found to induce the observed orientational order. The comparative narrowness of the ACN surface peak in the solvent density profiles of the SB surface (Figure 4) already suggests a layer of strong orientational preference. ACN orientation is indeed strict, with the NACN atoms directed to the electropositive SiS atoms of the surface. As seen in the simulation snapshot (Figure 1C), ACN molecules cluster around SiS atoms, pointing their methyl ends toward the bulk liquid. Beyond the immediate surface region, ACN is organized as observed for the silanol surfaces, in double layers of roughly antiparallel oriented molecules, with ACN molecules closer to the surface preferring to turn their NACN atom toward it. In neat W, a closed solvent layer covers the SB surface. W molecules can occupy the default positions of ACN molecules due to the electrostatic attraction between SiS and Ow atoms and the support of the W−W hydrogen-bond network. ACN molecules are unable to form a closed solvent layer over the SB surface. In the absence of W, the ACN clusters at the SB surface fan out, but basically ACN molecules still collect at the SiS atoms. This somewhat holey surface layer is complemented by oppositely oriented ACN molecules (i.e., with their methyl groups toward the surface), which appear as peaks of low number density in the neat ACN profile (at Z = 0.41 and 0.49 nm in Figure 4). These peaks are absent from the mixture density profiles, where we instead observe a gap. The number density of the two gapfilling peaks is much lower than that of the surface peak because the surface coordination by bundles of ACN molecules does not leave enough space to insert an equally dense layer of oppositely oriented molecules. Instead, a high degree of interaction is achieved at low number density in a different way: molecules closer to the surface (around Z = 0.41 nm in Figure 4) are in contact with two or three surface-adsorbed ACN molecules and one OS atom; ACN molecules of the next layer (around Z = 0.49 nm in Figure 4) interact mainly with OS atoms, thus substituting for the absent W molecules. Thus, the organization of neat ACN at the SB surface differs in detail from that at the silanol surfaces, although in all three cases double layers of roughly antiparallel oriented molecules are involved. Translational Mobility of Solvent Molecules. Solvent structure and orientation at and near the surface determine the translational mobility of solvent molecules parallel and perpendicular to the surface (i.e., lateral and axial mobility, respectively). The data presented in Figures 8−10 show how the translational mobility of solvent molecules decreases when they approach each of the three silica surfaces. To connect the mobility data with the solvent density profiles, the profiles for a bulk composition of 30/70 (v/v) W/ACN are shown in the background as gray lines, and axial mobility (residence time τ) and lateral mobility (parallel diffusion coefficient D∥, cf. eq 1) data are indicated for the peaks in the density profiles (cf. Figures 2−4). In general, the axial and lateral mobility of solvent molecules that interact directly with the surface (W molecules with the silanol surfaces; W and ACN molecules with the SB surface) becomes severely limited. Surface-interacting molecules are retained at the surface longer the lower their fraction in the

Figure 8. Axial mobility (residence times τ) and lateral mobility (diffusion coefficients D∥, cf. eq 1) of W and ACN molecules at certain distances from the SiS surface for selected bulk compositions of 80/20 (red), 30/70 (green), and 5/95 (blue) W/ACN (v/v). For reference, the solvent density profile for 30/70 (v/v) W/ACN is shown in the background (gray lines).

Figure 9. Axial mobility (residence times τ) and lateral mobility (diffusion coefficients D∥, cf. eq 1) of W and ACN molecules at certain distances from the GeS surface for selected bulk compositions of 80/ 20 (red), 30/70 (green), and 5/95 (blue) W/ACN (v/v). For reference, the solvent density profile for 30/70 (v/v) W/ACN is shown in the background (gray lines).

Figure 10. Axial mobility (residence times τ) and lateral mobility (diffusion coefficients D∥, cf. eq 1) of W and ACN molecules at certain distances from the SB surface for selected bulk compositions of 80/20 (red), 30/70 (green), and 5/95 (blue) W/ACN (v/v). For reference, the solvent density profile for 30/70 (v/v) W/ACN is shown in the background (gray lines).

bulk liquid because surface coordination is of first importance and the surface solvent layer is preserved intact as long as possible. This can be seen in the axial mobility data, where the residence time of W molecules at all model silica surfaces increases with decreasing W/ACN ratio, whereas that of ACN 6628

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

of its larger density of OH groups, the GeS surface holds more excess W than the SiS surface. The effect is smaller than expected from the nominal surface hydroxylation values (7.8 vs 4.5 OH groups/nm2 for the GeS and the SiS surface) because intrasurface hydrogen bonding and a lower accessibility of OH groups at the GeS surface attenuate its capacity for W molecules. The adsorption isotherm of the SB surface has a 2 maximum (Γ(n) W = 10.5 μmol/m ) at 5/95 (v/v) W/ACN, just as observed for the silanol surfaces, but also a minimum (Γ(n) W = −4 μmol/m2) at 80/20 (v/v) W/ACN. The SB surface holds excess W when ACN constitutes the higher bulk fraction and excess ACN when W constitutes the higher bulk fraction. (The crossover point is at 54/46 (v/v) W/ACN.) In summary, the W surface excess adsorption isotherms of the three model silica surfaces (Figure 11) reflect their respective hydrophilicities, whose molecular origin was revealed in the solvent density profiles (Figures 2−4) and hydrogen-bonding data (Figures 5−7). Toward a binary W/ACN mixture, the silanol surfaces are very hydrophilic due to extensive surface−W hydrogen bonding, whereas the SB surface, due to its lack of hydrogenbonding capabilities, is only mildly hydrophilic. The calculated surface excess adsorption isotherms agree remarkably well with experimental data for bare-silica stationary phases. Their W surface excess adsorption isotherms are Sshaped, as observed for the SB surface, with a maximum around 20/80 (v/v) W/ACN and a minimum around 80/20 (v/v) W/ ACN.27,57−59 Even the observed experimental values for the maximum and minimum W surface excess (10.7 and −3.65 μmol/m2, respectively),27 are close to the corresponding simulated values. The close quantitative agreement between experiment and simulation should not be overemphasized as simulated and experimental data cannot be expected to be quantitatively comparable considering: (i) the difference between the idealized, regular structure of a surface model and the amorphous nature of the bare-silica surface, about whose detailed structural characteristics we still know little, and (ii) the difference between the experimental determination of surface excess adsorption isotherms and the rigorous calculation possible with the simulated data. With all due caution considering the inevitably idealized nature of the employed model silica surfaces, the simulated data in their entirety show convincingly why only siloxane moieties, not the hydrophilic silanol groups, produce S-shaped W surface excess adsorption isotherms.

molecules at the SB surface decreases concurrently. The ACN mobility data for the two silanol surfaces (Figures 8 and 9) are highly similar, as expected from their ACN density profiles. W lateral mobility is also essentially the same at both surfaces, allowing for the closer proximity of W molecules (peak 1A) to the GeS surface. W residence times are, however, higher at the SiS than at the GeS surface, where W molecules can easily switch their hydrogen-bond attachment between the two layers of surface OH groups. Compared with the silanol surfaces, solvent residence times at the SB surface are dramatically raised (Figure 10). The very low solvent mobility at the SB surface reflects the nature of the interaction between solvent and surface, namely, electrostatic attraction between specific atoms. ACN residence times are even tens of nanoseconds. (In neat ACN, the residence time at the SB surface decreases to τ = 400 ps. Mobility data for neat solvents are listed in Tables S9 and S10 of the Supporting Information.) The environment of ACN molecules at the SB surface is so different from the environment of ACN molecules beyond the gap that an exchange of position between a surfaceadsorbed ACN molecule with one from subsequent layers is unlikely to occur. This exchange is occasionally made by W molecules, enabled by the W hydrogen-bonding network spanning the gap. Solvent mobility at the SB surface is so low that the surface layer can be regarded not only as rigid but also as practically distinct from the remaining solvent body. The gap in the density profiles indicates an abrupt change of solvent properties, as opposed to the smoother transition observed in the solvent density profiles of the silanol surfaces. Considering that the tight, mixed solvent layer repeats the regular atomic pattern of the SB surface, the layer can be regarded as a surrogate surface itself. Contrary to the hydrophilic silanol surfaces, this surrogate surface is amphiphilic due to hydrophilic patches from the adsorbed W molecules and hydrophobic patches from the solvent-exposed methyl groups of the adsorbed ACN molecules. Surface Excess Adsorption Isotherms. Figure 11 displays the surface excess adsorption isotherms calculated for

IV. CONCLUSIONS By modeling the adsorption of binary W/ACN mixtures between 99/1 and 2/98 (v/v) to three surfaces that each model one type of functional group of bare-silica stationary phases, we have shown that the presence or absence of surface OH groups dictates the surface hydrophilicity and thus the form of the W surface excess adsorption isotherm. Single and geminal silanol groups produce surface W excess over the whole compositional W/ACN range, whereas siloxane bridges replicate the S-shaped W surface excess adsorption isotherms of bare-silica materials. Although we do not yet know enough about the structure of the amorphous silica surface and the solvent-selective properties of its elements to definitely exclude other possibilities, it is tempting to conclude that the minimum observed in the W surface excess adsorption isotherms of bare-silica stationary phase materials is due to siloxane groups. This would make siloxane bridges not only a structural but also a functional element of the adsorbent silica surface. The hydrophobic

Figure 11. Areal-reduced W surface excess adsorption isotherms (cf. eq 2) calculated from simulating adsorption of W/ACN mixtures with bulk compositions between 99/1 and 2/98 (v/v) W/ACN at the SiS surface (blue), the GeS surface (green), and the SB surface (red).

the three model silica surfaces (cf. eq 2). The curve reveals how much W is present at and near a silica surface compared with the amount of W in the bulk region (Z > 2 nm) at a given nominal W/ACN ratio. The calculation considers not only the immediate surface region occupied by a rather rigid solvent layer but also the adjacent interface region, where the surface influence is still present if gradually declining in force. The two silanol surfaces have excess W over the full range of W/ACN mixture compositions. The maximum of their adsorption isotherms is found at 5/95 (v/v) W/ACN, with 2 Γ(n) W = 14 (SiS surface) and 16 μmol/m (GeS surface). Because 6629

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

Notes

properties attributed to siloxane groups by chromatographers could stem from the hydrophobic patch created when ACN molecules upon adsorption to the siloxane sites expose their methyl groups to the bulk mobile phase. Considering the high residence times of surface-adsorbed ACN molecules, the solvent-exposed methyl groups would resemble a very short (C1-like) hydrophobic stationary phase, which would explain why bare silica exhibits weak reversed-phase properties at high W/ACN ratios. Although the residence times simulated for ACN molecules at the SB surface cannot be transferred quantitatively to ACN molecules adsorbed at the siloxane sites of an amorphous silica surface, it is nonetheless plausible that such molecules have far longer residence times than W molecules adsorbed at the silanol sites. Our results have shown that W molecules are extensively connected to the surface OH groups and to each other. Severing these bonds costs energy, which is why the residence times of W molecules at the silica surface are higher than in the bulk, but W molecules are never isolated; they are always in hydrogen-bonding contact with the solvent layers adjacent to the immediate surface region. In contrast, surface-adsorbed ACN molecules are rather isolated from ACN molecules in adjacent solvent layers. Because of the directed and specific interaction between SiS and NACN atoms and the comparative isolation of surfaceadsorbed ACN molecules, they are much less likely than W molecules to leave the silica surface. Our results suggest that all three types of surface functional groups contribute in their own way to the chromatographic properties of bare silica. Silanol groups create the surface hydrophilicity required for the formation of the W-rich layer that enables partitioning of hydrophilic analytes. Geminal silanol groups have essentially the same properties as single silanol groups but provide a higher surface hydroxylation; on a model silica surface bearing silanol and siloxane groups, geminal silanol groups help to target the experimental surface hydroxylation of bare silica. By offering adsorption sites to ACN molecules, siloxane bridges are not so much hydrophobic themselves but ultimately cause the presence of hydrophobic patches on the bare-silica surface. On the basis of conclusions reached by MD simulations employing model silica surfaces, we have put forward a plausible theory for the role of the individual functional groups on the bare-silica surface for its stationary-phase properties in HILIC. Support for this theory could come from spectroscopy when methods that can differentiate between the types of surface functional groups as well as measure their interaction with binary solvent mixtures become available.



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.



ASSOCIATED CONTENT

* Supporting Information S

Tables listing the hydrogen-bond configurations of W molecules at the three model silica surfaces. Tables containing translational mobility data for neat W and neat ACN at the three model silica surfaces. A Figure showing the three model silica surfaces. Figures presenting the orientational arrangement of W and ACN molecules at the three model silica surfaces. This material is available free of charge via the Internet at http://pubs.acs.org.



REFERENCES

(1) Neue, U. D. HPLC Columns: Theory, Technology, and Practice; Wiley-VCH: New York, 1997. (2) Sun, L.; Siepmann, J. I.; Schure, M. R. Monte Carlo Simulations of an Isolated n-Octadecane Chain Solvated in Water−Acetonitrile Mixtures. J. Chem. Theory Comput. 2007, 3, 350−357. (3) Fouqueau, A.; Meuwly, M.; Bemish, R. J. Adsorption of Acridine Orange at a C8,18/Water/Acetonitrile Interface. J. Phys. Chem. B 2007, 111, 10208−10216. (4) Braun, J.; Fouqueau, A.; Bemish, R. J.; Meuwly, M. Solvent Structures of Mixed Water/Acetonitrile Mixtures at Chromatographic Interfaces from Computer Simulations. Phys. Chem. Chem. Phys. 2008, 10, 4765−4777. (5) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. Influence of Residual Silanol Groups on Solvent and Ion Distribution at a Chemically Modified Silica Surface. J. Phys. Chem. C 2009, 113, 9230−9238. (6) Rafferty, J. L.; Siepmann, J. I.; Schure, M. R. Understanding the Retention Mechanism in Reversed-Phase Liquid Chromatography: Insights from Molecular Simulation. In Advances in Chromatography; Brown, P. R., Grushka, E., Eds.; Marcel Dekker: New York, 2010; Vol. 48, pp 1−55. (7) Orzechowski, M.; Meuwly, M. Dynamics of Water Filaments in Disordered Environments. J. Phys. Chem. B 2010, 114, 12203−12212. (8) Rafferty, J. L.; Sun, L.; Siepmann, J. I.; Schure, M. R. Investigation of the Driving Forces for Retention in Reversed-Phase Liquid Chromatography: Monte Carlo Simulations of Solute Partitioning Between n-Hexadecane and Various Aqueous−Organic Mixtures. Fluid Phase Equilib. 2010, 290, 25−35. (9) Rafferty, J. L.; Siepmann, J. I.; Schure, M. R. Mobile Phase Effects in Reversed-Phase Liquid Chromatography: A Comparison of Acetonitrile/Water and Methanol/Water Solvents as Studied by Molecular Simulation. J. Chromatogr. A 2011, 1218, 2203−2213. (10) Rafferty, J. L.; Siepmann, J. I.; Schure, M. R. Retention Mechanism for Polycyclic Aromatic Hydrocarbons in Reversed-Phase Liquid Chromatography with Monomeric Stationary Phases. J. Chromatogr. A 2011, 1218, 9183−9193. (11) Zhuravlev, N. D.; Siepmann, J. I.; Schure, M. R. Surface Coverages of Bonded-Phase Ligands on Silica: A Computational Study. Anal. Chem. 2001, 73, 4006−4011. (12) Gritti, F.; Guiochon, G. Effect of the Density of the C18 Surface Coverage on the Adsorption Mechanism of a Cationic Compound and on the Silanol Activity of the Stationary Phase in Reversed-Phase Liquid Chromatography. J. Chromatogr. A 2006, 1132, 51−66. (13) Alpert, A. J. Hydrophilic-Interaction Chromatography for the Separation of Peptides, Nucleic Acids, and Other Polar Compounds. J. Chromatogr. 1990, 499, 177−196. (14) Jandera, P. Stationary and Mobile Phases in Hydrophilic Interaction Chromatography: A Review. Anal. Chim. Acta 2011, 692, 1−25. (15) Buszewski, B.; Noga, S. Hydrophilic Interaction Liquid Chromatography (HILIC)A Powerful Separation Technique. Anal. Bioanal. Chem. 2012, 402, 231−247. (16) Grumbach, E. S.; Diehl, D. M.; Neue, U. D. The Application of Novel 1.7 μm Ethylene Bridged Hybrid Particles for Hydrophilic Interaction Chromatography. J. Sep. Sci. 2008, 31, 1511−1518. (17) Bicker, W.; Wu, J.; Lämmerhofer, M.; Lindner, W. Hydrophilic Interaction Chromatography in Nonaqueous Elution Mode for Separation of Hydrophilic Analytes on Silica-Based Packings with Noncharged Polar Bondings. J. Sep. Sci. 2008, 31, 2971−2987.

AUTHOR INFORMATION

Corresponding Author

*E-mail: tallarek@staff.uni-marburg.de. 6630

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631

The Journal of Physical Chemistry C

Article

(18) Bicker, W.; Wu, J.; Yeman, H.; Albert, K.; Lindner, W. Retention and Selectivity Effects Caused by Bonding of a Polar Urea-Type Ligand to Silica: A Study on Mixed-Mode Retention Mechanisms and the Pivotal Role of Solute−Silanol Interactions in the Hydrophilic Interaction Chromatography Elution Mode. J. Chromatogr., A 2011, 1218, 882−895. (19) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. Composition, Structure, and Mobility of Water−Acetonitrile Mixtures in a Silica Nanopore Studied by Molecular Dynamics Simulations. Anal. Chem. 2011, 83, 2569−2575. (20) Melnikov, S. M.; Höltzel, A.; Seidel-Morgenstern, A.; Tallarek, U. A Molecular Dynamics Study on the Partitioning Mechanism in Hydrophilic Interaction Chromatography. Angew. Chem., Int. Ed. 2012, 51, 6251−6254. (21) Ruta, J.; Rudaz, S.; McCalley, D. V.; Veuthey, J.-L.; Guillarme, D. A Systematic Investigation of the Effect of Sample Diluent on Peak Shape in Hydrophilic Interaction Liquid Chromatography. J. Chromatogr., A 2010, 1217, 8230−8240. (22) Reimers, J. R.; Hall, L. E. The Solvation of Acetonitrile. J. Am. Chem. Soc. 1999, 121, 3730−3744. (23) Srinivasan, K. R.; Kay, R. L. The Pressure Dependence of the Dielectric Constant and Density of Acetonitrile at Three Temperatures. J. Solution Chem. 1977, 6, 357−367. ́ (24) Jadżyn, J.; Swiergiel, J. On Intermolecular Dipolar Coupling in Two Strongly Polar Liquids: Dimethyl Sulfoxide and Acetonitrile. J. Phys. Chem. B 2011, 115, 6623−6628. (25) Bidlingmeyer, B. A.; Del Rios, J. K.; Kropi, J. Separation of Organic Amine Compounds on Silica Gel with Reversed-Phase Eluents. Anal. Chem. 1982, 54, 442−447. (26) Nawrocki, J. The Silanol Group and Its Role in Liquid Chromatography. J. Chromatogr., A 1997, 779, 29−71. (27) Gritti, F.; dos Santos Pereira, A.; Sandra, P.; Guiochon, G. Comparison of the Adsorption Mechanisms of Pyridine in Hydrophilic Interaction Chromatography and in Reversed-Phase Aqueous Liquid Chromatography. J. Chromatogr., A 2009, 1216, 8496−8504. (28) West, R.; Whatley, L. S.; Lake, K. J. Hydrogen Bonding Studies. V. The Relative Basicities of Ethers, Alkoxysilanes and Siloxanes and the Nature of the Silicon−Oxygen Bond. J. Am. Chem. Soc. 1961, 83, 761−764. (29) Zhuravlev, L. T. The Surface Chemistry of Amorphous Silica. Zhuravlev Model. Colloids Surf., A 2000, 173, 1−38. (30) Rodriguez, J.; Elola, M. D.; Laria, D. Polar Mixtures under Nanoconfinement. J. Phys. Chem. B 2009, 113, 12744−12749. (31) Rodriguez, J.; Elola, M. D.; Laria, D. Confined Polar Mixtures within Cylindrical Nanocavities. J. Phys. Chem. B 2010, 114, 7900− 7908. (32) Mountain, R. D. Microstructure and Hydrogen Bonding in Water−Acetonitrile Mixtures. J. Phys. Chem. B 2010, 114, 16460− 16464. (33) Lange, K. M.; Hodeck, K. F.; Schade, U.; Aziz, E. F. Nature of the Hydrogen Bond of Water in Solvents of Different Polarities. J. Phys. Chem. B 2010, 114, 16997−17001. (34) Lange, K. M.; Könnecke, R.; Soldatov, M.; Golnak, R.; Rubensson, J. E.; Soldatov, A.; Aziz, E. F. On the Origin of the Hydrogen-Bond-Network Nature of Water: X-Ray Absorption and Emission Spectra of Water−Acetonitrile Mixtures. Angew. Chem., Int. Ed. 2011, 50, 10621−10625. (35) Huang, N.; Nordlund, D.; Huang, C.; Bergmann, U.; Weiss, T. M.; Pettersson, L. G. M.; Nilsson, A. X-Ray Raman Scattering Provides Evidence for Interfacial Acetonitrile−Water Dipole Interactions in Aqueous Solutions. J. Chem. Phys. 2011, 135, 164509−1−164509−6. (36) Argyris, D.; Tummala, N. R.; Striolo, A.; Cole, D. R. Molecular Structure and Dynamics in Thin Water Films at the Silica and Graphite Surfaces. J. Phys. Chem. C 2008, 112, 13587−13599. (37) Argyris, D.; Cole, D. R.; Striolo, A. Dynamic Behavior of Interfacial Water at the Silica Surface. J. Phys. Chem. C 2009, 113, 19591−19600. (38) Argyris, D.; Cole, D. R.; Striolo, A. Hydration Structure on Crystalline Silica Substrates. Langmuir 2009, 25, 8025−8035.

(39) Romero-Vargas Castrillón, S.; Giovambattista, N.; Aksay, I. A.; Debenedetti, P. G. Evolution from Surface-Influenced to Bulk-Like Dynamics in Nanoscopically Confined Water. J. Phys. Chem. B 2009, 113, 7973−7976. (40) Bonnaud, P. A.; Coasne, B.; Pellenq, R. J.-M. Molecular Simulation of Water Confined in Nanoporous Silica. J. Phys.: Condens. Matter 2010, 22, 284110−1−284110−15. (41) Ho, T. A.; Argyris, D.; Papavassiliou, D. V; Striolo, A.; Lee, L. L.; Cole, D. R. Interfacial Water on Crystalline Silica: A Comparative Molecular Dynamics Simulation Study. Mol. Simul. 2011, 37, 172− 195. (42) Musso, F.; Mignon, P.; Ugliengo, P.; Sodupe, M. Cooperative Effects at Water−Crystalline Silica Interfaces Strengthen Surface Silanol Hydrogen Bonding. An ab initio Molecular Dynamics Study. Phys. Chem. Chem. Phys. 2012, 14, 10507−10514. (43) Gulmen, T. S.; Thompson, W. H. Grand Canonical Monte Carlo Simulations of Acetonitrile Filling of Silica Pores of Varying Hydrophilicity/Hydrophobicity. Langmuir 2009, 25, 1103−1111. (44) Morales, C. M.; Thompson, W. H. Simulations of Infrared Spectra of Nanoconfined Liquids: Acetonitrile Confined in Nanoscale, Hydrophilic Silica Pores. J. Phys. Chem. A 2009, 113, 1922−1933. (45) Ding, F.; Hu, Z.; Zhong, Q.; Manfred, K.; Gattass, R. R.; Brindza, M. R.; Fourkas, J. T.; Walker, R. A.; Weeks, J. D. Interfacial Organization of Acetonitrile: Simulation and Experiment. J. Phys. Chem. C 2010, 114, 17651−17659. (46) Hu, Z.; Weeks, J. D. Acetonitrile on Silica Surfaces and at Its Liquid−Vapor Interface: Structural Correlations and Collective Dynamics. J. Phys. Chem. C 2010, 114, 10202−10211. (47) Cheng, L.; Morrone, J. A.; Berne, B. J. Structure and Dynamics of Acetonitrile Confined in a Silica Nanopore. J. Phys. Chem. C 2012, 116, 9582−9593. (48) Coasne, B.; Di Renzo, F.; Galarneau, A.; Pellenq, R. J.-M. Adsorption of Simple Fluid on Silica Surface and Nanopore: Effect of Surface Chemistry and Pore Shape. Langmuir 2008, 24, 7285−7293. (49) Nangia, S.; Washton, N. M.; Mueller, K. T.; Kubicki, J. D.; Garrison, B. J. Study of a Family of 40 Hydroxylated β-Cristobalite Surfaces Using Empirical Potential Energy Functions. J. Phys. Chem. C 2007, 111, 5169−5177. (50) Gulmen, T. S.; Thompson, W. H. Testing a Two-State Model of Nanoconfined Liquids: Conformational Equilibrium of Ethylene Glycol in Amorphous Silica Pores. Langmuir 2006, 22, 10919−10923. (51) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. The Missing Term in Effective Pair Potentials. J. Phys. Chem. 1987, 91, 6269−6271. (52) Wick, C. D.; Stubbs, J. M.; Rai, N.; Siepmann, J. I. Transferable Potentials for Phase Equilibria. 7. Primary, Secondary, and Tertiary Amines, Nitroalkanes and Nitrobenzene, Nitriles, Amides, Pyridine, and Pyrimidine. J. Phys. Chem. B 2005, 109, 18974−18982. (53) van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C. GROMACS: Fast, Flexible, and Free. J. Comput. Chem. 2005, 26, 1701−1718. (54) Marti, J. Analysis of the Hydrogen Bonding and Vibrational Spectra of Supercritical Model Water by Molecular Dynamics Simulations. J. Chem. Phys. 1999, 110, 6876−6886. (55) Liu, P.; Harder, E.; Berne, B. J. On the Calculation of Diffusion Coefficients in Confined Fluids and Interfaces with an Application to the Liquid−Vapor Interface of Water. J. Phys. Chem. B 2004, 108, 6595−6602. (56) Everett, D. H. Reporting Data on Adsorption From Solution at the Solid−Solution Interface (Recommendations 1986). Pure Appl. Chem. 1986, 58, 967−984. (57) Tani, K.; Suzuki, Y. Effect of Pore Size on the Surface Excess Isotherm of Silica Packings. J. Chromatogr. 1990, 515, 159−168. (58) Tani, K. Concept of Surface Excess Amount and Applications for Native and Modified Surfaces. Chromatography 2008, 29, 7−11. (59) Vajda, P.; Felinger, A.; Cavazzini, A. Adsorption Equilibria of Proline in Hydrophilic Interaction Chromatography. J. Chromatogr., A 2010, 1217, 5965−5970.

6631

dx.doi.org/10.1021/jp312501b | J. Phys. Chem. C 2013, 117, 6620−6631