Alcohol and Thiol Adsorption on (Oxy)hydroxide and Carbon Surfaces

Dec 13, 2012 - ... Sandia National Laboratories, Albuquerque, New Mexico 87185, United .... carbons for efficient oxygen reduction reaction in alkalin...
0 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/JPCC

Alcohol and Thiol Adsorption on (Oxy)hydroxide and Carbon Surfaces: Molecular Dynamics Simulation and Desorption Experiments Jeffery A. Greathouse,*,† David B. Hart,† and Margaret E. Ochs‡ †

Geochemistry Department and ‡Department of Materials, Devices, and Energy Technologies, Sandia National Laboratories, Albuquerque, New Mexico 87185, United States S Supporting Information *

ABSTRACT: Classical molecular dynamics simulations were used to investigate the interaction of methyl, ethyl, and n-propyl alcohols and thiols with the hydroxylated basal surfaces of aluminum hydroxide and iron oxyhydroxide, as well as a model graphite surface. Comparisons were made to concurrently run desorption experiments investigating the interaction of methyl, ethyl, and n-propyl alcohols with aluminum hydroxide and activated carbon. The metal (oxy)hydroxide surfaces represent the basal surfaces of the neutral end-member layered double hydroxides gibbsite and lepidocrocite, respectively, while the graphite surface is a simplified model of the pore walls in activated carbon used in the chemisorption experiment. Adsorption enthalpies obtained from simulations at infinite adsorbate dilution show that adsorption is greatly favored on the hydroxylated surfaces compared to the graphite surface, with the ethyl molecules adsorbing most favorably. Heats of desorption calculated from chemisorption experiments show the same increased interaction strength for the alcohols on the aluminum hydroxide surface compared with activated carbon, with the most favorable interaction being ethanol with the aluminum hydroxide surface. In general, simulations show that alcohols adsorb more strongly than thiols on the hydroxylated surfaces, while the reverse is true on the graphite surface. The structure of adsorbed monolayers was obtained from simulations of a liquidlike layer above each surface. As expected, monolayer surface densities decreased with increasing molecule size. The hydroxylated surfaces were found to be amphoteric with respect to both alcohol and thiol adsorption, and primary adsorption sites facilitate hydrogen bonding between the adsorbate and several surface hydroxyl groups. Alcohols and thiols adsorb at much larger distances to the graphite surface, resulting in the smaller adsorption enthalpies.

■

INTRODUCTION

LDH surfaces will aid in the design of new materials for the adsorption, storage, reaction, and degradation of organics. Applications include biointerfaces,1−5 catalysis,3,6−10 nanocomposites,11−13 and the adsorption and decomposition of contaminants.8,14−17 Additionally, the use of LDHs as coating materials in chemical sensing applications has grown in recent years.18 The use of computational chemistry to investigate interfacial structure and dynamics has increased rapidly in recent years with the availability of improved methods and computing power. These methods provide atomistic detail regarding

The fate of organic molecules adsorbed on inorganic surfaces plays a central role in a variety of applications. In particular, layered minerals such as clays are increasingly being used in chemical processes involving aqueous organic compounds.1 Clays are characterized by very large surface area, ion-exchange capacity, tunable layer charge, and structural stability over a range of environmental conditions. While many swelling clays bear a net negative charge, a subset of positively charged clays known as layered double hydroxides (LDHs) consist of metal (oxy)hydroxide layers separated by interlayer regions containing charge-balancing anions such as chloride or carbonate. The hydrophilic surfaces of LDHs are thus well suited for the adsorption of polar organic molecules. Fundamental studies to increase our understanding of such adsorption processes on © 2012 American Chemical Society

Received: May 30, 2012 Revised: November 9, 2012 Published: December 13, 2012 26756

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

As a validation of the simulation methods, desorption experiments are used to determine the heat of desorption of the C1−C3 alcohols on aluminum hydroxide and activated carbon. Aluminum hydroxide does not have an interlayer region or pores and therefore represents an experimental analogue to the idealized Al(OH)3 surface considered in the simulations. While the arrangement of surface hydroxyl groups on aluminum hydroxide will be more complex than the idealized Al(OH)3 surface used in the simulations, the primary mechanism for alcohol and thiol adsorptionhydrogen bondingis the same in each case. Results from the simulations therefore represent a lower bound to the experimental adsorption enthalpies, since interlayer regions/ pores and surface functionalization would likely result in larger adsorption enthalpies. To the best of our knowledge, no studies of the heat of desorption of alcohols on aluminum hydroxide and activated carbon have been reported.

adsorption processes that aid in the interpretation of experimental results and the prediction of materials properties involving molecular adsorption.19 While many experimental and theoretical studies exist on the adsorption of water and inorganic ions on oxide surfaces, much less attention has been paid to aqueous organics. Tunega et al. have used ab initio methods to study the structure and dynamics of acetic acid20,21 and agricultural chemicals22,23 on clay surfaces. Similar methods were used to show that the binding energies of small organics on a smectite clay are inversely proportional to the ionization potentials of the adsorbates.24 Classical simulation methods are more commonly used than electronic structure methods to study molecular adsorption onto mineral surfaces, particularly since the simulation of a liquidlike layer or interlayer region is still prohibitive by quantum methods. Skipper et al.25 used classical molecular dynamics (MD) simulation to show that the diffusion of methanol intercalated between hydrated smectite clay layers is similar to that in aqueous solution. MD simulation has also been used to study the interaction of biologically relevant organic molecules with the hydroxylated surfaces of LDHs.26−28 Chemisorption experiments for materials analysis are commonly used to determine a variety of surface-dependent characteristics from concentration of active sites to adsorption and desorption behavior.29−37 Temperature-controlled desorption techniques have been widely used to determine heats of desorption.38 Studies typically focus on catalysts; vapors of interest include water gas shift molecules such as methane, CO, and CO2, and titration of the acid sites is done with ammonia as the probe molecule.31,34,37 Fewer examples exist for chemisorption studies with desorption analysis outside this range; these investigations include N2O, acetylene, and other small organics.29,30,32,33 In this work, MD simulations are used to compare trends in the adsorption structure and thermodynamics of small alcohols and thiols onto idealized Al(OH)3 and FeOOH surfaces. Simulation results are compared with experimental results on the desorption of small alcohols on aluminum hydroxide. A comparison of adsorption properties on these inorganic surfaces with a hydrophobic surface representative of activated carbon (graphite) allows us to directly determine the effect of hydrophilic surface sites and hydrogen bonding on adsorption. Simulations of a single adsorbate molecule on each surface are used to compare adsorption enthalpies, while simulations of a liquid layer on each surface are used to determine monolayer coverages as well as structural properties of surface complexes. The Al(OH)3 and FeOOH surfaces considered here represent basal surfaces of commonly studied end-member LDHs (gibbsite and lepidocrocite, respectively), while the graphite surface represents activated carbon, a common sorbent materials for organic contaminants. Small alcohols and thiols (methyl, ethyl, and n-propyl; C1− C3) included in the present study contain functional groups found in many chemicals of industrial or technical interest, and their interaction with LDH surfaces represent the likely adsorption mechanism of larger organics with similar functional groups. A comparison of alcohol/thiol adsorption thermodynamics on idealized (oxy)hydroxide surfaces allows us to better understand the fundamental aspects of molecular adsorption as well as the influence of functional groups on adsorption. Thus, this work will aid in the prediction of adsorption phenomena of more complex organic compounds on similar surfaces.

■

SIMULATION METHODS The (oxy)hydroxide phases were modeled by use of ClayFF39 force field parameters without modification. ClayFF uses a nonbonded approach involving van der Waals and electrostatic interactions between atoms and a harmonic bond potential for the hydroxyl groups. ClayFF has been used to successfully reproduce bulk properties of layered metal oxy(hydroxides) such as Al(OH)3 and FeOOH.39 The van der Waals potential parameters for C atoms in graphite were taken from the literature.40 The alcohol and thiol adsorbates (C1−C3) were treated as fully flexible molecules according to the all-atom OPLS force field.41 The OPLS force field has been shown to accurately reproduce the bulk structural and thermodynamic properties of organic liquids such as those considered here.41 Geometric combination rules were used for adsorbate− adsorbate and adsorbate−surface interactions. All nonbonded parameters are given in Table 1. Structures for Al(OH)3, FeOOH, and graphite were taken from the literature42−44 and orthogonalized before creation of supercells as specified in Table 2. Models of naturally occurring layered mineral phases were used for Al(OH)3 (gibbsite) and FeOOH (lepidocrocite) so that the protonation state of each surface (001 and 101, respectively) was well-defined. The basal (001) surface of graphite was used. The lepidocrocite model was reoriented so that its hydroxylated surface was aligned with the z-direction. A vacuum gap was created above each surface and terminated by a Lennard-Jones wall at z = 50 Å to avoid unwanted interactions between adsorbate molecules and the adjacent surface. Two different adsorbate surface coverages were considered: one molecule representing infinite surface dilution, from which isosteric heats of adsorption were obtained, and a liquidlike layer (200 molecules for methyl molecules, 120 molecules for ethyl and propyl molecules) to accurately determine monolayer coverage, adsorption sites, and interfacial structure. After equilibration, the thickness of this liquid phase varied from 25 Å to 40 Å above the surface; the exact thickness of the liquidlike layer depends on the adsorbate. Snapshots of equilibrated ethanol layers above the Al(OH)3 and FeOOH surfaces are shown in Figure 1. Methanol yields a thinner layer than propanol; thiol layers are slightly thicker than the corresponding alcohol layers. While the organic molecules were fully flexible and mobile, only surface hydrogen atoms were allowed to move in simulations of the liquidlike layers near the Al(OH)3 and FeOOH surfaces to avoid any 26757

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

Table 1. Nonbonded Parameters for Pairwise Energy Terms between Atoms i and j Separated by Distance ra speciesb

q (e)

ε (kcal·mol−1)

σ (Å)

1.330 × 10−6 0.1554 0

4.271 3.166 0

9.028 × 10−7 0.1554 0.1554 0

4.906 3.166 3.166 0

0.0556

3.400

0.0660 0.0660 0.0660 0.1700 0.0300 0 0.0300

3.500 3.500 3.500 3.120 2.500 0 2.500

0.0660 0.0660 0.0660 0.0660 0.2500 0.0300 0

3.500 3.500 3.500 3.500 3.550 2.500 0

39

Al O(H) H(O) Fe O O(H) H(O) C (graphite) C, CH3OH, RCH2OH C, RCH3 C, R2CH2 O, ROH H, RH, alkanes H(O), ROH H(C), CH3OH C, CH3SH C, RCH2SH C. RCH3 C, R2CH2 S, RSH H, RH, alkanes H(S), RSH

Al(OH)3 1.575 −0.950 0.425 FeOOH39 1.575 −1.05 −0.95 0.425 Graphite40 0 Alcohols41 0.145 −0.180 −0.120 −0.683 0.060 0.418 0.040 Thiols41 0.000 0.060 −0.180 −0.120 −0.435 0.060 0.255

Figure 1. Snapshots (xz plane) showing 120 ethanol molecules above (a) Al(OH)3 and (b) FeOOH surfaces (H, white; O, red; Al, pink; Fe, blue; C, gray).

configuration of the liquid phase. After cooling for 50 ps to 300 K, the production stage was run for 1100 ps, with data from the final 1000 ps used for analysis. Equilibrium was confirmed by monitoring the fluctuations in the potential energy, which were less than 0.5% for any system. Additionally, one-dimensional atomic density profiles from the production simulation (Supporting Information) indicate a well-established adsorbed layer. Methanethiol has a boiling point of 279 K, which is slightly lower than our simulation temperature of 300 K, but this minor difference should not significantly impact the structural properties of adsorbed methanethiol layers.

For nonbonded energy Eij = qiqj/r + 4εij[(σij/r)12 − [(σij/r)6]. bR refers to an alkyl, alcohol, or thiol group.

a

Table 2. Dimensions of the Surface Models and Supercell Dimensions Used in the Simulations surface

supercell repeats

supercell dimensions (Å)

Al(OH)3 FeOOH C (graphite)

3×4×1 7×6×1 9×9×2

26.4 × 20.1 × 50.0 23.6 × 21.6 × 50.1 21.3 × 22.2 × 50.0

■

EXPERIMENTAL METHODS A Micromeritics Autochem II 2920 chemisorption unit was used with the vapor generator option to measure desorption temperatures for the C1−C3 alcohols at ramp rates ranging from 2 to 12 °C·min−1. Two sorbents, sample sizes of 0.5 g, were used: aluminum hydroxide hydrate, Al(OH)3 (Fisher, certified powder), and activated carbon (Fisher brand activated carbon charcoal, 50−200 mesh, 300−75 μm particle size). Xray analysis of the aluminum hydroxide revealed the structure was gibbsite. The activated carbon was, as expected, amorphous. Brunauer−Emmett−Teller (BET) surface area analysis of both samples showed 16.7 ± 0.4 m2/g for the aluminum hydroxide and 94.7 ± 2.9 m2/g for the activated carbon. Both BET isotherms were of type 2, indicating nonporous solid samples. The isotherms rose quickly at low relative pressures, then rose more moderately through intermediate relative pressures, before rising quickly again at high relative pressures, and displayed no hysteresis upon reducing the relative pressure. Each chemisorption experiment consisted of three parts: sample degas, dosing of vapor to sample, and temperature-programmed desorption. Helium was used as the preparation, carrier, and loop gases. During the degas step, the sample was heated to 350 °C at a ramp rate of 20 °C·min−1, then was held at 350 °C for 10 min before cooling to 25 °C via a house air line precooled in an ice bath that flowed over the exterior of the sample tube. The relatively low

nonphysical migration of surface groups (e.g., cations or hydroxyl groups) into the adsorbed layer. However, the entire top layer of Al(OH)3 and FeOOH was fully mobile in the simulations at infinite adsorbate dilution, so that the effect of momentum transfer between surface and adsorbate could be included in the calculation of adsorption enthalpies. All carbon atoms in the graphite surface model were fixed at their crystallographic coordinates. Although an approximate approach, this method has been used successfully to model adsorption phenomena on carbon surfaces.40,45−47 Constant-volume MD simulations were performed with the LAMMPS code48 at a thermostat temperature of 300 K. Shortrange interactions were evaluated every 0.5 fs with a real-space cutoff of 10.0 Å. Periodic boundary conditions were applied, and long-range electrostatic interactions were evaluated every 1.0 fs by use of the particle−particle particle−mesh (PPPM) summation algorithm49 with a precision of 1.0 × 10−4. Simulations at infinite adsorbate dilution were run for 550 ps, with data from the final 500 ps used for analysis. For simulations of the liquidlike layer, the thermostat temperature was set to 1000 K for 50 ps to remove any effect of the initial 26758

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

maximum degas temperature of 350 °C was chosen to avoid sintering or chemical changes in the samples, yet adsorbed vapor molecules were still removed from the surface. Fourier transform infrared spectroscopy (FT-IR) of both samples preand post-degas steps were completed. The IR spectrum of alumina hydroxide post-degas shows O−H stretching modes at 3619 and 3523 cm−1 that are slightly weaker than the predegassed sample. The spectroscopy of activated carbon, while somewhat difficult to resolve, showed small peaks in the 1800− 1600 cm−1 range, indicating alkenes, aldehydes, ketones, carboxylic acids, or esters that did not change upon degassing. Small broad peaks at 3500 cm−1 present on the as-received activated carbon sample were no longer present on the degassed sample. Such peaks are attributed to adsorbed water on the surface. During the vapor dosing step, the vapor flask was held at 55 °C and the refluxer at 45 °C; these temperatures were adequate for use with the three alcohols tested. The sample was maintained at 25 °C by use of the house air, which cools the exterior of the sample tube. During an injection, helium flowed through the vapor loop (volume 1 cm3) at 50 mL·min−1 carrying the vapor to the sample chamber. The vapor in the loop was injected 30 times, with 8 min between injections. A thermal conductivity detector monitored the outflowing vapor after the sample chamber. By confirming that peak volumes were identical for repetitions 27−30 of the dosing step of the experiment, the sample was said to be saturated with the vapor. The ramp rate during temperature-programmed desorption steps was set to 2, 5, 8, 10, and 12 °C·min−1 up to a maximum temperature of 350 °C, during which a thermal conductivity detector monitored any desorbing molecules. All desorption peaks in the study occurred between 90 °C and 125 °C. The maximum temperature was chosen to avoid chemical reactions of the samples; alumina is known to dehydrate alcohols at relevant reaction kinetics above 350 °C.50 The temperature at peak maximum for the desorbing peaks provides the data necessary to calculate the heat of desorption (Ed):38 ⎛ ⎞ ⎛E A⎞ E B ln⎜⎜ 2 ⎟⎟ = − d + ln⎜ d ⎟ ⎝ RC ⎠ RTp ⎝ Tp ⎠

Figure 2. Adsorption enthalpy at infinite dilution for C1−C3 alcohols and thiols. Error bars were obtained from block averages of the MD simulation energies.

favorable on the LDH surfaces than on graphite. These molecules are able to form strong hydrogen bonds with hydroxyl groups on the LDH surfaces, while no such opportunity exists on the graphite surface. In general, alcohols adsorb more strongly on the LDH surfaces than the thiols, based on the electronegativity difference between oxygen (3.4) and sulfur (2.6). Clausen et al.24 noticed a slightly different trend from their quantum calculations of ethanol, ethyl acetate, and pyridine on the siloxane surface of a negatively charged clay. It should be noted, however, that their reported binding energies included the effect of an adsorbed sodium cation, while our model surfaces are charge-neutral. We also note that thiols adsorb more strongly than alcohols on the hydrophobic graphite surface, due to the enhanced van der Waals interactions with the sulfur atom. Goss and Schwarzenbach53 have established a general relationship between adsorption enthalpies on mineral surfaces and vaporization enthalpies of the bulk liquids. For the C1−C3 alcohols and thiols, experimental enthalpies of vaporization increase with carbon chain length, ranging from 8.9 to 11.3 kcal·mol−1 for alcohols and from 5.9 to 7.6 kcal·mol−1 for thiols.41 We would therefore expect the adsorption enthalpies of the C1−C3 alcohols and thiols to increase with increasing carbon chain length. We see from Figure 2 that ethanol and

(1)

where B is the temperature ramp rate, Tp is the temperature at peak maximum, Ed is the heat of desorption, R is the gas constant, A is the quantity of vapor adsorbed at saturation, and C is a constant related to the desorption rate. This equation is valid if it is assumed that there is no diffusion or readsorption during the temperature-programmed desorption.38,51 The effect of diffusion on the results was minimized by constraining the range of temperature ramp rates to be within 10 °C·min−1. The flow rate of helium was maintained at 50 mL·min−1 to avoid the readsorption of molecules.

■

RESULTS Adsorption enthalpies at infinite dilution were calculated according to52 ΔHads = ⟨Ugh⟩ − ⟨Ug⟩ − RT

(2)

where Ugh is the potential energy of the adsorbed guest molecule, Ug is the potential energy of the adsorbate in the ideal gas reference state, R is the gas constant, and T is the thermostat temperature. We see from Figure 2 that the adsorption of these small hydrophilic molecules is much more 26759

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

ethanethiol exhibit an unusually large adsorption enthalpy and that the n-propanethiol adsorption enthalpy on FeOOH is unusually small. The sizes of ethanol and ethanethiol molecules appear to produce especially stable surface complexes beyond that due merely to hydrogen bonding, but it is possible that the combination of the ClayFF and OPLS force fields results in unusually large adsorption enthalpies for these molecules. As we show below in a comparison of radial distribution functions, the hydrogen-bonding environment is nearly identical as additional carbon atoms are added to the adsorbate molecules (e.g., methanol has the same surface hydrogen bond lengths as propanol). Therefore, variations in adsorption enthalpies on Al(OH)3 and FeOOH surfaces must be due to van der Waals interactions with the surface that depend on molecular size relative to surface morphology. The magnitude of adsorption enthalpies on the graphite surface (2.0−3.3 kcal·mol−1) is similar to density functional theory (DFT) calculated adsorption energies reported for C1− C3 alcohols in carbon nanotubes (4.0−5.6 kcal·mol−1).54 A regular increase of approximately 0.7 kcal·mol−1 per alcohol CH2 group was seen in the DFT calculations,54 but such a regular increase in adsorption energy on the graphite surface is evident in Figure 2 only for the thiols. The use of a flat graphite surface rather than a nanotube may explain this difference in adsorption energy trends, but the rather weak interaction of alcohols and thiols on carbon surfaces is consistent. Chemisorption experiments matched the simulation results in both character and value. Figure 3 shows that the alcohols

Figure 3. Heats of desorption for C1−C3 alcohols calculated from chemisorption experiments. Error bars were derived from averaging over 5 runs.

Figure 4. Monolayer coverage for C1−C3 alcohols and thiols from MD simulations of a liquid layer on each surface. The monolayer surface density was determined from the area under the onedimensional density profile peak corresponding to adsorbed molecules (Supporting Information).

interact more strongly with the hydrophilic Al(OH)3 surface than the hydrophobic activated carbon surface. In good agreement with the simulation, ethanol binds more strongly than either methanol or n-propanol at 27 kcal·mol−1. The experimental desorption enthalpies on activated carbon, however, are larger than the simulated adsorption enthalpies on graphite by up to 12 kcal·mol−1.The adsorbate interacts with these carbon surfaces only through weak van der Waals terms, but these interactions are likely enhanced in activated carbon with irregular surface pockets similar to the kinetic diameter of a given adsorbate. Functionalization of activated carbon surfaces can also enhance adsorption.40 Simulations of a liquidlike phase of an alcohol or thiol near surfaces can be used to obtain structural properties at the interface. For example, the determination of monolayer coverage is helpful in estimating a material’s adsorption capacity for a given analyte. In this case, the first peak in the adsorbate one-dimensional atomic density profile (Supporting Information) was used to define the adsorbed monolayer, and

the area under this peak was used to quantify the monolayer density (Figure 4). Steric and molecular size effects are responsible for the decreasing monolayer densities with increasing chain length (C1 > C2 > C3), and for larger surface densities of the alcohols compared to the thiols. The larger surface densities on graphite reflect the packing density of adsorbate molecules, since only weak interactions exist between these adsorbates and the graphite surface. We use probability density plots such as those shown for ethanol in Figure 5 to identify key adsorption sites on the LDH surfaces. The theoretical upper limit of surface densities can be calculated from the availability of surface hydroxyl sites [6.8 and 8.3 sites·nm−2 for Al(OH)3 and FeOOH, respectively], which are much larger than the largest surface densities from the monolayer coverages (Figure 4). For the Al(OH)3 surface, the most favorable adsorption site is the cavity formed by four surface hydroxyl groups. The maximum surface density in this case is 4.5 sites·nm−2, in good agreement with the methanol monolayer density seen in Figure 4. Adsorption sites on the 26760

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

Figure 6. Probability distributions of (a) ethyl and (b) propyl adsorbate orientation relative to the Al(OH)3 and FeOOH basal surfaces. The angle is formed by the C−C vector (first and last carbon atoms) relative to the basal plane, as illustrated at top with ethanol for angles of −90°, 0°, and 90°. Data were analyzed from the MD trajectories of the liquidlike layer (120 molecules) near each surface. Only adsorbate molecules within 6 Å of surface oxygen atoms were included in the calculation. Figure 5. Two-dimensional probability density plots of a monolayer of adsorbed ethanol molecules (oxygen atoms) in the xy plane on (a) Al(OH)3 and (b) FeOOH surfaces (Al, pink; O, red; H, white; Fe, blue). The contour scales show ethanol surface densities in arbitrary units.

We gain additional insight into adsorbate orientation relative to the Al(OH)3 and FeOOH surfaces by plotting the angle formed by the adsorbate C−C vector (first and last carbon atoms) and the basal surface. Figure 6 reveals a similar orientation of ethyl and propyl adsorbates on the Al(OH)3 and FeOOH surfaces. Both the ethyl and propyl adsorbate molecules show two distinct orientations relative to the Al(OH)3 basal surface, with the C−C (or C−C−C) axis either parallel (∼0°) or nearly perpendicular to the basal surface (60− 70°). Examples of both of these predominant orientations can be seen in the equilibrium snapshot of ethanol molecules adsorbed on Al(OH)3 (Figure 1a). Figure 6a shows that ethanol has a distinct C−C orientation on the FeOOH surface (30−60°), while there is no angular preference for ethanethiol molecules over a very broad range (−20° to 75°). The orientation of propyl adsorbates on FeOOH is similar to the general trend seen on Al(OH)3, but the angles are shifted to slightly larger values. Looking at the orientation of propyl molecules (Figure 6b), all propyl molecules show a preferred orientation with the molecule tilted away from the surface (60°). Propanol and propanethiol also adsorb nearly parallel to the Al(OH)3 surface (0°), suggesting that these molecules form a natural fit above the cavities on that surface. Both of these orientations were seen in DFT studies of n-pentanol molecules on a hydroxylated cristobalite (111) surface, with the C−C axis nearly perpendicular to the surface (90°) at monolayer coverage.57 A completely perpendicular orientation is not seen in our

FeOOH surface are located in a grid pattern that facilitates hydrogen bonding with two surface hydroxyl groups. The identified sites correspond to the same theoretical surface density of 8.3 sites·nm−2, but steric and thermal effects prevent all of these sites from being occupied by even the smallest adsorbate (methanol) at 300 K. As expected, no primary adsorption sites exist on the graphite surface. A recent study of ethanol binding to calcite surfaces has shown that the ordering of ethanol molecules in the adsorbed monolayer coincides with the unit cell of the calcite surface.55 A similar ordering of adsorbed ethanol molecules is seen on the Al(OH)3 and FeOOH surfaces (Figure 5). Also, Sung et al.56 used sum frequency vibrational spectroscopy and found substantial ordering of ethanol molecules at a sapphire surface, but the second layer of adsorbed ethanol was much less organized. In our simulations, the first layer of adsorbed molecules is highly ordered compared to the second layer (Figure 1), and there is very little exchange of molecules between the first layer and subsequent layers near the Al(OH)3 and FeOOH surfaces. The static nature of the adsorbed monolayer contributes to the well-ordered arrangement of ethanol molecules seen in Figure 5. 26761

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

simulations because the adsorbed molecules also interact with adjacent liquidlike layers. Radial distribution functions (RDFs) can be used to quantify the hydrogen-bonding interactions between alcohols/thiols and the (oxy)hydroxide surfaces. Although well-defined criteria for quantifying hydrogen bonds exist, they are difficult to implement in simulations of bulk fluids with pairwise potentials.58 Because the present study involves six different adsorbates and two functional groups (OH and SH), we have not attempted a detailed analysis of surface−adsorbate hydrogen bonds. Rather, we compare average O−H and S− H distances obtained from RDFs between surface hydroxyl groups and adsorbate alcohol/thiol groups (Table 3). Table 3. Average Interatomic Distances for Alcohol and Thiol Surface Complexes on (Oxy)hydroxides from Radial Distribution Functionsa adsorbate

Osurf−Hads (Å)

MeOH MeSH EtOH EtSH PrOH PrSH

1.78 1.61 1.77 1.62 1.77 1.64

MeOH MeSH EtOH EtSH PrOH PrSH

1.79 1.62 1.77 1.61 1.77 1.60

Yads−Hsurf (Å)

Al(OH)3 Surface 1.85 2.38 1.86 2.36 1.89 2.36 FeOOH Surface 1.74 2.23 1.76 2.28 1.77 2.28

Osurf−Yads (Å) 2.79 3.09 2.79 3.09 2.78 3.10 2.75 3.12 2.74 3.05 2.75 3.10

Figure 7. Averaged RDFs of ethanol and ethanethiol surface complexes on (a) Al(OH)3 and (b) FeOOH surfaces. where Osurf = surface oxygen, Hads = adsorbate hydrogen, Hsurf = surface hydrogen, Oads = adsorbate oxygen, and Sads = adsorbate sulfur.

a

Me = methyl, Et = ethyl, Pr = n-propyl, Osurf = surface oxygen, Hads = adsorbate hydrogen, Hsurf = surface hydrogen, Yads = adsorbate oxygen or sulfur.

MgO surfaces,60,61 and the hydroxylated surface of kaolinite.62 While none of the comparison systems representsan ideal analogues to (oxy)hydroxide surfaces, coordinated or adsorbed alcohols and thiols form hydrogen bonds with surface oxygen atoms or hydroxyl groups in each case. For ethanol or ethanethiol complexes, Yads−Hsurf and Osurf−Hads distances reported in Table 3 are within 0.04 and 0.25 Å, respectively, of the p-cresol values.59 Similar calculations of methanol and ethanol adsorbed on MgO surface clusters60,61 showed Osurf− Hads distances of 1.67 and 1.65 Å, respectively, in excellent agreement with the values reported in Table 3 (1.78 and 1.77 Å). Finally, the calculated Oads−Hsurf and Oads−Osurf distances for methanol adsorbed on the hydroxylated surface of kaolinite were reported to be 1.73 and 2.69 Å, respectively,62 which are similar to our corresponding distances of 1.85 and 2.79 Å. We note that our MD simulations include thermal effects at 300 K, which result in slightly larger distances compared to the gasphase complexes studied by geometry optimization (without temperature effects). These comparisons further validate the force field method used in this work. Although the adsorption enthalpies were calculated from simulations at infinite dilution, an analysis of RDF data from the liquidlike layers on the surfaces can inform trends in these enthalpies. The orientation of a single adsorbate molecule on a surface may be somewhat different than a monolayer of adsorbed molecules, but the surface−adsorbate distances are very similar (e.g., O−H and S−H distances at infinite dilution are within 0.05 Å of the corresponding distances in the liquid− surface simulations). In general, trends in hydrogen-bond

Corresponding O−H and S−H RDFs for ethanol− and ethanethiol−surface distances are shown in Figure 7. Table 3 shows nearly identical adsorbate−surface distances for C1−C3 adsorbates, so only the ethyl RDFs are shown here. Corresponding RDFs for methyl and propyl adsorbates are shown in Supporting Information. Hydrogen bonds involving Yads−Hsurf pairs (Y = O or S) are slightly longer for molecules adsorbed on Al(OH)3 compared to FeOOH, but the Osurf−Hads distances for each alcohol or thiol are the same on each surface. Both Al(OH)3 and FeOOH surfaces are amphoteric, in agreement with quantum calculations of carboxylic acid adsorption on the aluminol surface of kaolinite.20 The shortest (and strongest) hydrogen bonds are formed with the adsorbate as the hydrogen donor. Of particular note is the very short Osurf−Hads distances for thiols (≈1.6 Å) compared to the much longer Sads−Hsurf distances (≈2.3 Å). It is interesting to note that,for alcohols adsorbed on Al(OH)3, the Osurf−Hads distances are about 0.1 Å shorter than the Oads−Hsurf distances, while on FeOOH the Oads−Hsurf distances are equal to or slightly shorter than the Osurf−Hads distances. These differences in O−H distances must be due to the unique arrangement of hydroxyl groups on each surface, since the interaction parameters for hydroxyl groups are identical for the two surfaces (Table 1). The interatomic distances from our simulations compare favorably with those obtained from electronic structure calculations of alcohol and thiol complexes with p-cresol,59 26762

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

Article

compared with the corresponding alcohols, which are oriented to promote intermolecular hydrogen bonding.

distances between different adsorbate−surface pairs correspond to trends in adsorption enthalpies. Specifically: (1) The unique arrangement of surface hydroxyl groups on the FeOOH surface (Figure 5) leads to short hydrogen bonds with ethanol and npropanol and unusually large adsorption enthalpies compared to ethanethiol and n-propanethiol. (2) In several cases (MeOH, MeSH, and PrOH), the Oads−Hsurf or Sads−Hsurf distances are shorter at the FeOOH surface than the Al(OH)3 surface, indicative of greater adsorption enthalpies at the FeOOH surface. (3) Despite the shorter Osurf−Hads distances for thiols than alcohols, ethanol and n-propanol have equal or greater adsorption enthalpies than their thiol counterparts. The most significant example of this trend can be seen by comparing npropanol and n-propanethiol on FeOOH. Although the Osurf− Hads distance is 0.17 Å shorter for n-propanethiol compared to n-propanol, the S−Hsurf distance is 0.51 Å longer than the O− Hsurf distance. As a result, the adsorption enthalpy of npropanol on FeOOH is 10 kcal·mol−1 greater than that of npropanethiol. Snapshots of representative ethanol and ethanethiol surface complexes are shown in Figure 8 and help to illustrate the

■

CONCLUSIONS The MD simulations presented here have been used to provide fundamental insight into the adsorption of small alcohols and thiols on idealized Al(OH)3, FeOOH, and graphite surfaces, with chemisorption experiments providing excellent corroboration for alcohol adsorption on aluminum hydroxide and activated carbon. The (oxy)hydroxide surfaces display amphoteric behavior with respect to adsorbed alcohols and thiols, and the increased electrostatic interaction between surface and adsorbate due to hydrogen bonding dominates the structure and energy of surface complexes. The large adsorption enthalpies calculated from simulations at infinite adsorbate dilution are also due to hydrogen bonding and are much larger than the adsorption enthalpies on the hydrophobic graphite or activated carbon surfaces. As expected, larger adsorbate molecules have a lower monolayer surface density, but the hydrogen-bonding structure of surface complexes is generally independent of adsorbate size. For example, hydrogen-bond lengths for methanol (methanethiol) surface complexes are nearly identical to those for n-propanol (n-propanethiol). Interestingly, the Osurf−Hads distances are significantly shorter for thiols compared to alcohols, even though the alcohols tend to have a larger adsorption enthalpy than the corresponding thiols. These results should inform future studies on the intercalation of polar organic molecules into LDH interlayers. Additional controlling factors that should be considered include the effect of adjacent surfaces and layer spacing, the role played by counterions, and the potential for surface modification to enhance adsorption. The simulated adsorption enthalpies presented here also provide insight into the adsorption of polar organic molecules on the pore walls of activated carbon. The increased adsorption enthalpies associated with hydrogen bonding to specific surface sites suggests that the performance of widely used activated carbon sorbents could be optimized through selective functionalization of pore walls.

■

ASSOCIATED CONTENT

* Supporting Information S

Figure 8. Equilibrium snapshots from MD simulations showing surface complexes of ethanol and ethanethiol on Al(OH)3, FeOOH, and graphite. Select hydrogen bond distances are indicated (Al, pink; O, red; H, white; Fe, blue; C, gray; S, yellow).

Four figures showing one-dimensional atomic density profiles and plots of radial distribution functions. This material is available free of charge via the Internet at http://pubs.acs.org.

■

interatomic distances given in Table 3. Hydrogen bonds shown in Figure 8 are defined according to the default criteria of the Materials Studio software (Accelrys, Inc.): maximum hydrogen−acceptor distance of 2.5 Å and a minimum donor− hydrogen−acceptor angle of 90°. Although examples of the ethyl molecules are shown, their adsorption structures are similar to the methyl and propyl molecules. The location of these molecules relative to the surface corresponds to primary adsorption sites shown in Figure 5. Weak van der Waals interactions between these adsorbates and the graphite surface result in less ordered surface complexes and greater adsorbate− surface distances. As seen in Figure 8, adsorbed ethanethiol is oriented to enhance van der Waals interactions between the sulfur atom and the graphite surface. This arrangement results in larger adsorption enthalpies for the C2 and C3 thiols

AUTHOR INFORMATION

Corresponding Author

*E-mail [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Defense Threat Reduction Agency (DTRA) and the U.S. Department of Energy, Office of Basic Energy Sciences, Geosciences Research Program. Sandia National Laboratories is a multiprogram laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under Contract DE-AC04-94AL85000. 26763

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764

The Journal of Physical Chemistry C

■

Article

(37) Song, H.; Ozkan, U. S. J. Phys. Chem. A 2010, 114, 3796−3801. (38) Webb, P. A.; Orr, C. Analytical Methods in Fine Particle Technology; Micromeritics Instrument Corp.: Norcross, GA, 1997. (39) Cygan, R. T.; Liang, J.-J.; Kalinichev, A. G. J. Phys. Chem. B 2004, 108, 1255−1266. (40) Shevade, A. V.; Jiang, S. Y.; Gubbins, K. E. J. Chem. Phys. 2000, 113, 6933−6942. (41) Jorgensen, W. L.; Maxwell, D. S.; TiradoRives, J. J. Am. Chem. Soc. 1996, 118, 11225−11236. (42) Saalfeld, H.; Wedde, M. Z. Kristallogr. 1974, 139, 129−135. (43) Zhukhlistov, A. P. Crystallogr. Rep. 2001, 46, 730−733. (44) Trucano, P.; Chen, R. Nature 1975, 258, 136−137. (45) Tolmachev, A. M.; Firsov, D. A.; Anuchin, K. M.; Fomkin, A. A. Colloid J. 2008, 70, 486−496. (46) Cosoli, P.; Fermeglia, M.; Ferrone, M. Mol. Simul. 2008, 34, 1321−1327. (47) Severson, B. L.; Snurr, R. Q. J. Chem. Phys. 2007, 126, 134708. (48) Plimpton, S. J. J. Comput. Phys. 1995, 117, 1−19. (49) Plimpton, S. J.; Pollock, R.; Stevens, M. Particle-mesh Ewald and rRESPA for parallel molecular dynamics simulations; Eighth SIAM Conference on Parallel Processing for Scientific Computing, 1997. (50) Adkins, H.; Perkins, P. P. J. Am. Chem. Soc. 1925, 47, 1163− 1167. (51) Characterization of Heterogeneous Catalysts; Delanny, F., Ed.; Marcel Dekker, Inc.: New York, 1984; Vol. 15, p 388. (52) Farrusseng, D.; Daniel, C.; Gaudillere, C.; Ravon, U.; Schuurman, Y.; Mirodatos, C.; Dubbeldam, D.; Frost, H.; Snurr, R. Q. Langmuir 2009, 25, 7383−7388. (53) Goss, K.-U.; Schwarzenbach, R. P. Environ. Sci. Technol. 1999, 33, 3390−3393. (54) Alldredge, E. S.; Badescu, S. C.; Bajwa, N.; Perkins, F. K.; Snow, E. S.; Reinecke, T. L.; Passmore, J. L.; Chang, Y. L. Phys. Rev. B 2008, 78, No. 161403. (55) Sand, K. K.; Yang, M.; Makovicky, E.; Cooke, D. J.; Hassenkam, T.; Bechgaard, K.; Stipp, S. L. S. Langmuir 2010, 26, 15239−15247. (56) Sung, J.; Waychunas, G. A.; Shen, Y. R. J. Phys. Chem. Lett. 2011, 2, 1831−1835. (57) Barnette, A. L.; Asay, D. B.; Janik, M. J.; Kim, S. H. J. Phys. Chem. C 2009, 113, 10632−10641. (58) Marti, J. J. Chem. Phys. 1999, 110, 6876−6886. (59) Biswal, H. S.; Shirhatti, P. R.; Wategaonkar, S. J. Phys. Chem. A 2010, 114, 6944−6955. (60) Branda, M. M.; Ferullo, R. M.; Belelli, P. G.; Castellani, N. J. Surf. Sci. 2003, 527, 89−99. (61) Branda, M. M.; Rodriguez, A. H.; Belelli, P. G.; Castellani, N. J. Surf. Sci. 2009, 603, 1093−1098. (62) Matusik, J.; Scholtzov, E.; Tunega, D. Clays Clay Miner. 2012, 60, 227−239.

REFERENCES

(1) Khan, A. I.; O’Hare, D. J. Mater. Chem. 2002, 12, 3191−3198. (2) del Hoyo, C. Appl. Clay Sci. 2007, 36, 103−121. (3) Di Cosimo, J. I.; Diez, V. K.; Xu, M.; Iglesia, E.; Apesteguia, C. R. J. Catal. 1998, 178, 499−510. (4) Ladewig, K.; Xu, Z. P.; Lu, G. Q. Expert Opin. Drug Delivery 2009, 6, 907−922. (5) Vallee, A.; Humblot, V.; Pradier, C.-M. Acc. Chem. Res. 2010, 43, 1297−1306. (6) Zhang, F. Z.; Xiang, X.; Li, F.; Duan, X. Catal. Surv. Asia 2008, 12, 253−265. (7) Zhu, K.; Wang, D. G.; Liu, J. Nano Res. 2009, 2, 1−29. (8) Oliver, S. R. J. Chem. Soc. Rev. 2009, 38, 1868−1881. (9) Debecker, D. P.; Gaigneaux, E. M.; Busca, G. Chem.Eur. J. 2009, 15, 3920−3935. (10) Sels, B. F.; De Vos, D. E.; Jacobs, P. A. Catal. Rev. 2001, 43, 443−488. (11) Ciardelli, F.; Coiai, S.; Passaglia, E.; Pucci, A.; Ruggeri, G. Polym. Int. 2008, 57, 805−836. (12) Costa, F. R.; Saphiannikova, M. Layered double hydroxide based polymer nanocomposites. In Wax Crystal Control: Nanocomposites: Stimuli-Responsive Polymers; Advances in Polymer Science, Vol. 210; Springer-Verlag: Berlin and Heidelberg, 2008; pp 101−168. (13) Leroux, F.; Taviot-Gueho, C. J. Mater. Chem. 2005, 15, 3628− 3642. (14) Srivastava, B.; Jhelum, V.; Basu, D. D.; Patanjali, P. K. J. Sci. Ind. Res. 2009, 68, 839−850. (15) Ahmad, T.; Rafatullah, M.; Ghazali, A.; Sulaiman, O.; Hashim, R.; Ahmad, A. J. Environ. Sci. Health C 2010, 28, 231−271. (16) Cornejo, J.; Celis, R.; Pavlovic, I.; Ulibarri, M. A. Clay Miner. 2008, 43, 155−175. (17) Goh, K. H.; Lim, T. T.; Dong, Z. Water Res. 2008, 42, 1343− 1368. (18) Mousty, C. Anal. Bioanal. Chem. 2010, 396, 315−325. (19) Cygan, R. T.; Greathouse, J. A.; Heinz, H.; Kalinichev, A. G. J. Mater. Chem. 2009, 19, 2470−2481. (20) Tunega, D.; Benco, L.; Haberhauer, G.; Gerzabek, M. H.; Lischka, H. J. Phys. Chem. B 2002, 106, 11515−11525. (21) Tunega, D.; Haberhauer, G.; Gerzabek, M. H.; Lischka, H. Langmuir 2002, 18, 139−147. (22) Tunega, D.; Haberkauer, G.; Gerzabek, M. H.; Lischka, H. Soil Sci. 2004, 169, 44−54. (23) Tunega, D.; Gerzabek, M. H.; Haberhauer, G.; Lischka, H. Eur. J. Soil Sci. 2007, 58, 680−691. (24) Clausen, P.; Andreoni, W.; Curioni, A.; Hughes, E.; Plummer, C. J. G. J. Phys. Chem. C 2009, 113, 12293−12300. (25) Skipper, N. T.; Lock, P. A.; Titiloye, J. O.; Swenson, J.; Mirza, Z. A.; Howells, W. S.; Fernandez-Alonso, F. Chem. Geol. 2006, 230, 182− 196. (26) Newman, S. P.; Di Cristina, T.; Coveney, P. V.; Jones, W. Langmuir 2002, 18, 2933−2939. (27) Mohanambe, L.; Vasudevan, S. J. Phys. Chem. B 2005, 109, 15651−15658. (28) Zhang, H.; Xu, Z. P.; Lu, G. Q.; Smith, S. C. J. Phys. Chem. C 2009, 113, 559−566. (29) Ates, A.; Reitzmann, A. J. Catal. 2005, 235, 164−174. (30) Ates, A.; Reitzmann, A. React. Kinet. Catal. L. 2005, 86, 11−20. (31) Azzam, K. G.; Babich, I. V.; Seshan, K.; Lefferts, L. J. Catal. 2007, 251, 163−171. (32) Berndt, H.; Briehn, V.; Evert, S. J. Mol. Catal. 1992, 73, 203− 214. (33) Diaz, E.; Ordonez, S.; Vega, A. J. Colloid Interface Sci. 2007, 305, 7−16. (34) Gutierrez-Ortiz, J. I.; de Rivas, B.; Lopez-Fonseca, R.; GonzalezVelasco, J. R. Appl. Catal., B 2006, 65, 191−200. (35) Li, P.; Zhao, T. J.; Zhou, J. H.; Sui, Z. J.; Dai, Y. C.; Yuan, W. K. Carbon 2005, 43, 2701−2710. (36) Reddy, K. H. P.; Anand, N.; Prasad, P. S. S.; Rao, K. S. R.; Raju, B. D. Catal. Commun. 2011, 12, 866−869. 26764

dx.doi.org/10.1021/jp305275q | J. Phys. Chem. C 2012, 116, 26756−26764