Differential Adsorption of Phenol Derivatives onto a Polymeric Sorbent

Oct 7, 1999 - Center for Advanced Research in Biotechnology, National Institute of Standards and Technology, 9600 Gudelsky Drive, Rockville, Maryland ...
2 downloads 12 Views 120KB Size
J. Phys. Chem. B 1999, 103, 9879-9887

9879

Differential Adsorption of Phenol Derivatives onto a Polymeric Sorbent: A Combined Molecular Modeling and Experimental Study K. L. Mardis, A. J. Glemza, B. J. Brune, G. F. Payne, and M. K. Gilson* Center for AdVanced Research in Biotechnology, National Institute of Standards and Technology, 9600 Gudelsky DriVe, RockVille, Maryland 20850 and Department of Chemical and Biochemical Engineering and Center for Agricultural Biotechnology, UniVersity of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, Maryland 21250 ReceiVed: May 6, 1999; In Final Form: July 9, 1999

Oxygenated aromatic compounds (OACs) are used for the synthesis of a variety of commercial products. Lignin from wood and other plant products are potential sources of OACs, but it is difficult to separate the mixtures of OACs found in digests of these raw materials. One promising separation approach involves the use of selective adsorption resins, such as the acrylic ester sorbent XAD-7. It has been shown previously that this sorbent binds the para isomer of one OAC, methoxyphenol, more favorably that the ortho isomer when hexane is used as the solvent. The present study uses a combination of molecular modeling and experiment to elucidate the mechanism of this selectivity. The calculations yield good agreement with experimental binding affinities and indicate that hydrogen bonding is the dominant mode of adsorption of para-methoxyphenol onto XAD-7 from hexane. In contrast, ortho-methoxyphenol appears to form an intramolecular hydrogen bond that weakens the intermolecular hydrogen bond to the sorbent. As a consequence, ortho-methoxyphenol binds less strongly, and its association is dominated by van der Waals interactions and three-centered hydrogen bonds. This result is supported by quantum mechanical calculations and infrared spectroscopic experiments. It is also found that when water is the solvent, hydrogen bonding becomes an insignificant adsorption mechanism, and both molecules bind to the resin via nonpolar interactions. This explains the loss of selectivity that is observed in both experiment and calculation.

Introduction Oxygenated aromatic compounds (OACs) are used in the synthesis of a range of agricultural and pharmaceutical products. Petroleum is a good source for nonpolar, large-volume feedstock aromatics such as benzene, toluene, and xylene because of its ready availability and because well understood distillation-based separations exist to separate these compounds. However, partial oxidation of these aromatics to generate OACs is problematic, and typical reaction pathways use conditions that raise environmental or safety concerns. Environmentally safer and yet still economical methods are needed to produce OACs. A possible alternative to synthesis of OACs from petroleum is the recovery of preformed OACs from plant extracts or from lignin wastes produced by the pulp and paper industries. However, obtaining OACs from agricultural products poses its own challenges because specific compounds must be purified from complex plant digests. Distillation is not well suited to this task because OACs are polar, and the relatively strong intermolecular forces among them lead to low volatilities and thermodynamic nonidealities. As a consequence, alternative separation techniques that do not involve distillationssuch as adsorption and extractionsare being investigated.1-3 Since the sizes and shapes of plant-derived OACs are relatively uniform, the differential adsorption of these compounds must rely in large part upon differences among their charge distributions and hydrogen-bonding capabilities. In previous work,4-7 a porous polycarboxylic ester sorbent (XAD* To whom correspondence should be addressed: Fax: 301-738-6255. Email: [email protected].

Figure 1. Partial structure of the acrylic ester sorbent XAD-7. The circled “P” refers to additional polymer units connecting at these points. The area inside the dashed outline corresponds to ethyl propionate.

7, Rohm and Haas), shown in Figure 1, was identified which binds OACs from hexane with different affinities based on differences in their hydrogen-bonding abilities and polarizabilities. One of these studies focused on the preferential adsorption of positional isomers of methoxyphenol.7 It has been suggested that the methoxyphenols interact with XAD-7 by forming intermolecular hydrogen bonds between their hydroxyl hydrogens and the carboxylic ester groups believed to be present on the sorbent surface.8 The separation of methoxyphenol isomers is of interest for several reasons. The methoxyphenol moiety is common to many plant phenols, such as ferulic acid and isoeugenol, making it a simple test system for separating OACs. Additionally, methoxyphenols are important chemical intermediates in their own right. For example, ortho-methoxyphenol is an intermediate in the synthesis of several products, including vanillin and the

10.1021/jp991499q CCC: $18.00 © 1999 American Chemical Society Published on Web 10/07/1999

9880 J. Phys. Chem. B, Vol. 103, No. 45, 1999

Figure 2. Physical model of adsorption of para- and ortho-methoxyphenol onto XAD-7 from hexane, proposed in ref 7.

expectorant guaifenesin.9 Since a current, petroleum-based, synthetic route from phenol to ortho-methoxyphenol is not completely regioselective,10 the desired ortho isomer must be separated from the byproduct para-methoxyphenol. Due to the relatively high boiling points of these isomers, it would be desirable to employ non-distillation techniques to separate the two. It has been shown that para-methoxyphenol adsorbs from hexane onto the acrylic ester sorbent XAD-7 19 times more favorably than ortho-methoxyphenol.7 Additional adsorption experiments suggested that steric limitations were not responsible for the observed suppression of adsorption. While the adsorption studies yielded only thermodynamic information and provided no mechanistic explanation for the lower adsorption of ortho-methoxyphenol, infrared (IR) spectroscopy could be used to explore the mechanism of the observed selectivity.7 IR spectra for hexane solutions containing para-methoxyphenol showed a peak at 3627 cm-1, which is attributable to the free O-H stretch. In contrast, spectra for hexane solutions of the ortho isomer did not have a peak in this free O-H stretching region but did have a slightly broadened peak at a lower wavenumber (3566 cm-1). This suggested that ortho-methoxyphenol in hexane exists exclusively as an intramolecularly hydrogen-bonded species.11 The interaction between the isomers and the ester sorbent was then studied with ethyl propionate (EP) as a hexane-soluble analogue of the sorbent. The IR spectra for a hexane solution containing para-methoxyphenol and EP showed a reduction in the intensity of the original free OH stretch peak at 3627 cm-1, and a new broad peak appeared at 3470 cm-1. The latter was attributed to the OH stretch in the hydrogen bonded:EP complex. When EP was added to a hexane solution containing ortho-methoxyphenol, a broad O-H peak appeared at essentially the same wavenumber (3469 cm-1) as that of the para-methoxyphenol:EP complex. However, a considerably higher EP concentration was required to observe this peak with ortho-methoxyphenol, indicating a less favorable interaction between the ortho isomer and EP. These IR data were interpreted according to the scheme shown in Figure 2 (Scheme 2 of Glemza et al.7). Specifically, para-methoxyphenol was proposed to bind to the sorbent (and to EP) by the formation of a hydrogen bond. For orthomethoxyphenol, it was hypothesized that hydrogen bonding with the sorbent required breakage of the intramolecular hydrogen bond. The strength of the proposed scheme was that it provided a plausible explanation for the experimental observation that the enthalpy of binding of para-methoxyphenol to XAD-7 is 13 kJ/mol more favorable than that of ortho-methoxyphenol to XAD-7. This difference was interpreted as the amount of enthalpy required to break the intramolecular hydrogen bond. A potential weakness of the proposed scheme for the binding of ortho-methoxyphenol to XAD-7 is that an expected minor peak was not observed in the spectrum. In particular, if the 19fold difference in binding of para- versus ortho-methoxyphenol results from the free energy cost of breaking the intramolecular hydrogen bond of ortho-methoxyphenol, then approximately

Mardis et al. 1/19 of the free species should lack an intramolecular hydrogen bond. However, the IR spectrum of free ortho-methoxyphenol does not show any evidence of such a species, as there is only one OH stretch band. Thus, the failure to observe an O-H stretching peak near 3625 cm-1 for a non-hydrogen-bonded species argues against the proposed scheme, as recognized by the brackets for the hypothetical intermediate in Figure 2. The present study uses a combination of computer modeling and experiment to examine this problem in greater detail. The molecular modeling technique used here is a novel one that efficiently identifies the stable conformations of a molecule or a complex and uses these conformations as the basis for calculations of binding free energies.12 It was found that the calculations yield realistic results for the relative binding affinities of the two methoxyphenol isomers in hexane. This supports the validity of the model and justifies its further application in an examination of the molecular mechanisms for selective binding. This examination suggests that the association of ortho-methoxyphenol with ethyl propionate is not accompanied by breakage of the intramolecular hydrogen bond, as was previously hypothesized.7 Additional ab initio quantum mechanics calculations and further IR studies support this computational result. As an additional investigation of the adsorption of para- and ortho-methoxyphenol onto the polymer sorbent XAD-7, their relative affinities were determined with water as the solvent instead of hexane. This experiment is of interest because water effectively weakens hydrogen bonds. The loss of selectivity that is observed when water is the solvent supports the concept that hydrogen bonding is indeed important for selectivity in hexane. Calculated binding affinities for adsorption from water also show this loss of specificity and indicate that nonpolar interactions, rather than hydrogen bonds, dominate adsorption from water. Methods Experimental. The affinity measurements for methoxyphenol isomer adsorption from hexane onto the acrylic ester sorbent (XAD-7), which are referred to in this paper, were previously reported by Glemza et al.7 In the present study, the affinities for adsorption of methoxyphenols from water onto the acrylic ester sorbent were measured. The porous acrylic ester sorbent XAD-7, manufactured by Rohm and Haas, was purchased from Sigma Chemical Co. The specific surface area is reported by the supplier to be approximately 450 m2/g. Before use, the sorbent was washed sequentially with water, methanol, acetone, hexane, acetone, methanol, and water to remove any chemical contaminants. To ensure that the pores remain wet, the sorbent was weighed wet, and the data were normalized using the observation that 1 g wet sorbent corresponds to 0.20 g dry sorbent. The para- and ortho-methoxyphenol isomers and the ethyl propionate were obtained from Aldrich, and 2,6-dimethoxyphenol was purchased from Sigma. All chemicals were 99% pure. Adsorption studies were conducted by equilibrating known amounts of the sorbent with water solutions containing a single solute for 2 days. The equilibrated concentration (C) of solute was determined by UV-visible spectrophotometry (Spectronics Genesys II). The amount of solute adsorbed per unit mass of sorbent (q) was calculated from the difference between the initial (Cinit) and equilibrated concentrations of the solute in water with the following expression:

q)

(Cinit - C)V M

(1)

Adsorption of Phenols onto a Polymeric Sorbent

J. Phys. Chem. B, Vol. 103, No. 45, 1999 9881

Here, V is the volume of water, and M is the mass of dry sorbent. We report the adsorption affinity as the ratio of q to the concentration of solute in solution, q/C, computed with data in the linear part of the adsorption isotherm [cf. Figure 9]. During the course of the experiment, the appearance of the spectrum was monitored to ensure that the UV-visible absorption measured was due to the solute and not to materials leached from the sorbent. Fourier transform infrared (FTIR) spectroscopy was used to study intramolecular and intermolecular hydrogen bonding of the methoxyphenols. These measurements were carried out with a Nicolet Instrument Corporation 5DXC FTIR Spectrometer having a resolution of 4 cm-1, with KBr windows and 0.762mm spacers. Spectra were obtained for solutions of EP alone in hexane and for mixtures of EP with individual methoxyphenols. The spectrum for ethyl propionate alone was subtracted from the spectra of the mixtures. Signal-to-noise ratios were increased by averaging the spectra over 32 scans. Computational. Theory. We were interested in calculating the standard free energy of binding, ∆Gb°, of an isomer of methoxyphenol (MP) onto ethyl propionate (EP). The standard free energy of binding is related to the standard chemical potentials of the free and adsorbed species. Thus,

∆G°b ) µ°MP:EP - µ°EP - µ°MP

(2)

where µ°MP:EP, µ°EP, and µ°MP are the chemical potentials of the MP:EP complex, EP, and MP, respectively, each in a hypothetical ideal solution at standard concentration (1 mol/L). For molecular interactions that can be adequately described by classical statistical thermodynamics, the relevant contributions to the chemical potential may be written as integrals over molecular conformations. Thus13

µ° ) -RT ln

( ) 8π2 Z C°σ

(3)

where

Z)

∫ e-β(U(r)+W(r))dr

(4)

where β ) 1/kT; C° is the standard concentration; σ is the symmetry number of the molecule; U(r) and W(r) are the gasphase potential energy and the solvation energy, respectively, of the molecule as a function of its conformation; and r is a vector of internal coordinates that specify the conformation. Here, a mass factor that cancels upon calculation of ∆G° has not been included. Additionally, a factor of P°V h is neglected since it is very small at standard temperature and pressure.13 Evaluation of eq 4 was performed by the Mining Minima (MM) method, which has been described in a previous publication.12 It was implemented in a local version of the program UHBD.14 Briefly, this algorithm takes advantage of the fact that the largest contributions to the configuration integral are from regions of configuration space near energy minima. The algorithm proceeds by finding a minimum energy conformation, mapping out the extent of the potential energy well around this structure, and calculating the configuration integral for that well by Monte Carlo integration of the Boltzmann factor. The configuration integral is then approximated as the sum of the contributions of individual energy wells. New energy wells are included in the sum until, for 5 successive minima, either (a) contributions to the free energy drop to a fractional change of 0). Since bond lengths and angles were not varied in the sampling, it is possible that the calculations were biased by the bond lengths and angles used in the calculations. This is of particular concern for ortho-methoxyphenol, for which the formation of an intramolecular hydrogen bond is possible. The bond lengths and angles of the groups involved in this bond

9882 J. Phys. Chem. B, Vol. 103, No. 45, 1999

Mardis et al.

TABLE 1: Atomic Charge Parameters in Units of Electrons (1 Electron ) 1.602×10-19 C) chemical group

atom

charge

aromatic -C-OH

C O H -C O C H C H C O O C H C H C H

0.11 -0.54 0.43 0.21 -0.34 -0.14 0.09 -0.115 0.115 0.63 -0.52 -0.34 -0.05 0.09 -0.27 0.09 -0.08 0.09

aromatic -C-OCH3

aromatic -CH >CdO -O-CH2-CH3 -CH2-

might influence the apparent strength of this hydrogen bond. This possibility was examined by carrying out calculations with two sets of bond lengths and angles. One set was generated by energy-minimizing a conformation of ortho-methoxyphenol with an intramolecular hydrogen bond, and the other was generated by energy-minimizing a conformation with the hydroxyl group pointed away from the methoxy group, which breaks the intramolecular hydrogen bond. In both cases, the hydroxyl torsion was allowed to rotate during the binding calculation. The results of the two calculations agreed to within .84 kJ/mol, and similar distributions of bound conformations were obtained (see Results and Discussion). Energy Model. The energy in eq 4 is separated into the potential energy, U(r), and the solvation energy, W(r). Here, the potential energy was calculated using the CHARMm 2617 force field with the all-hydrogen representation. The EP force constants and charges were taken from the ethyl acetate parameters included in the September 1998 release.18 The methoxyphenol parameters were also obtained from the CHARMm 26 parameter set, except for those involving the methoxy oxygen. Since no ether oxygen parameters are available in CHARMm 26, parameters from an older versionsCHARMm 19swere used. These parameters are close to those for the chemically similar ester oxygen in CHARMm 26 and are expected to be compatible with the other CHARMm 26 parameters. The charges for all atoms are shown in Table 1. The solvation model approximates the solvation energy as the sum of the contributions from a two-step solvation process. The first step is the formation in the solvent of a Lennard-Jones cavity which has the shape of the solute. The nonpolar solvation energy associated with this step, ∆GNP, was approximated as linear in the solvent-accessible surface area of the solute. The constants of proportionality were set to 20.92 J/mol/Å2 for water19 and -69.87 J/mol/Å2 for hexane with an offset of -20.46 kJ/mol.15 The hexane value was obtained through a fit to the experimental data provided in the Supporting Information of a recent paper.20 These values are similar to those obtained for another nonpolar solvent, chloroform.15 While the physical basis for the offset is unknown, the exact value of the offset does not affect the relative binding affinities determined in this work. Because computing the surface area is time-consuming, the MM calculations were done without the nonpolar term, and the resulting energies were subsequently corrected for this contribution, as described below.

The second solvation energy term is due to the electrostatic charging of the atomic partial charges inside the cavity. A relatively accurate way of estimating this electrostatic solvation term is to use detailed numerical solutions of the Poisson equation.21,22 However, this approach is too time-consuming to be used for the many conformations generated in the MM procedure. We therefore used a computationally rapid generalized Born (GB) approximation during the configurational search procedure23,24, in which the effective Born radius of each atom is computed via a charge-induced dipole interaction term.25 The resulting energies were subsequently adjusted for the differences between the GB model and detailed solutions of the Poisson equation, as described below. The corrections mentioned in the two previous paragraphs were implemented as follows.15,16 The MM method yields a list of the most stable conformationssenergy wellssfor a molecule or complex. Each conformation i is associated with a free energy G′i. After the MM procedure was complete, the surface area was computed for each conformation and was used to compute the nonpolar solvation energy of the conformation, ∆GNP i . This contribution was added to the free energy of the conformation. In addition, a detailed finite difference solution of the Poisson equation was used to compute a reference electrostatic solvation energy of each conformation, ∆Gelec,ref . i A GB calculation was done for the same conformation, yielding the approximate electrostatic solvation energy ∆Gelec,GB . The i free energy of conformation i was then corrected for the error of the GB result by adding ∆Gelec,ref - ∆Gelec,GB . Thus, the i i final calculated free energy of each well is Gi ) G′i + ∆GNP i + elec,GB ∆G . ∆Gelec,ref i i Computing solvation energies with this model requires that a cavity radius be assigned to each atom. Here, the atomic radius of each atom was set to the average of its CHARMm van der Waals radius (Rmin) and the solvent radius (2.0 Å for hexane and 1.4 Å for water). The solvent dielectric constant was set to 1.89 for hexane and 78.5 for water.26 Ethyl Propionate as a Model for XAD-7. The polymeric sorbent XAD-7 used in the adsorption studies is not suitable for infrared spectroscopy because of its poor solubility in nonpolar solvents. However, ethyl propionate possesses the same carboxylic ester group, as shown in Figure 1, and has been used previously as a model for XAD-7 in IR studies of the adsorption of methoxyphenols.7 Here, EP was used again as a model in both the calculations and the IR studies. In particular, we compared the computed relative binding affinities of the methoxyphenols for EP with the measured relative adsorption affinities of methoxyphenols for the acrylic ester sorbent. Such comparisons assume that EP is a good model for a representative patch of the surface of the sorbent. It is worth noting that only relatiVe binding affinities of two molecules for EP versus the sorbent can be compared because the number of such surface patches per unit mass of sorbent is not known. Results and Discussion Adsorption from Hexane. para-Methoxyphenol. A previous experimental study demonstrated that para-methoxyphenol has a higher adsorption affinity for the acrylic ester sorbent than does ortho-methoxyphenol when the solvent is hexane.7 The measured ratio of adsorption affinities was 19, based upon data at low concentrations of MP, for which saturation of binding sites is minimal. This result suggests that the affinity of paramethoxyphenol for a representative patch of surface on the acrylic ester sorbent is 19 times that of ortho-methoxyphenol. We used the computational method described above to compute

Adsorption of Phenols onto a Polymeric Sorbent

J. Phys. Chem. B, Vol. 103, No. 45, 1999 9883

TABLE 2: Comparison of Calculated Binding Constants (T ) 26.85 °C) and Experimental Adsorption Affinities (T ) 25 °C) for ortho- and para-Methoxyphenol in Hexane and Watera solvent

∆G°ortho

∆G°para

calculated binding ratiob

measured adsorption ratioc

hexane water

-2.6 -5.2

-9.2 -5.0

14 0.9

19d 1.2

a

Calculated ∆G°’s are given in units of kJ/mol for a 1 mol/L ideal standard state. b Kpara/Kortho. c (qpara/Cpara)/(qortho/Cortho). d From ref 7.

the affinities of both isomers of MP for the compound EP, which models the binding site of the acrylic ester. As shown in Table 2, the computed standard free energies of binding are -9.2 and -2.6 kJ/mol for para- and ortho-methoxyphenol, respectively. These values correspond to a ratio of single-site binding constants of 14, which is similar to the adsorption affinity ratio of 19. Thus, the calculations reproduce the experimental data well. It is therefore reasonable to use the calculations to examine the mechanisms by which the MPs adsorb onto the acrylic ester sorbent as well as the physical basis for the reduced affinity of ortho-methoxyphenol relative to para-methoxyphenol. The calculations indicate that the mechanism by which paramethoxyphenol binds onto EP is straightforward. As shown in the scheme in Figure 2, the para isomer forms an intermolecular hydrogen bond between the hydroxyl hydrogen of methoxyphenol and the carbonyl oxygen of EP. This is illustrated in Figure 4a, which shows the most stable conformation found in the calculations for the complex of para-methoxyphenol with EP. Here, the hydroxyl hydrogen is 1.76 Å from the carbonyl oxygen of the EP molecule, with an O-H-O angle of 174.5°. The dominance of hydrogen-bonded conformations in the complex is illustrated further in Figures 5a and 6a, which plot conformational free energy versus the OH bond distance and the O-H-O angle for the most stable (bottom of the well) conformations found in the calculations. Except for one stacked conformation, the hydrogen bond lengths are all within 0.02 Å of 1.78 Å, and the angles are within 16° of an ideal linear hydrogen bond. ortho-Methoxyphenol. In contrast, the lowest-energy conformation found for the ortho isomer complexed with EP does not form an intermolecular hydrogen bond. Indeed, as shown in Figure 4b, the intermolecular O-H distance of this conformation is 5.9 Å. The distributions of hydrogen-acceptor distances and angles are shown in Figures 5b and 6b. The most stable conformation is a “stacked” conformation, where the primary intermolecular interactions are van der Waals forces rather than hydrogen bonds. However, there is a small but distinct population of hydrogen-bonded conformations having distances of about 1.95 Å and angles of about 145°. The most stable representative of this group is only 1.7 kJ/mol less stable than the optimal stacked conformation. This energy difference is less than kT for room temperature, so the hydrogen-bonded conformation is expected to be at least weakly populated. A striking feature of the hydrogen-bonded conformations of ortho-methoxyphenol with EP is that the hydrogen bonds are three-centered. That is, the hydroxyl hydrogen participates in both an intermolecular and an intramolecular hydrogen bond, as illustrated in Figure 4c. In these three-centered bonds, the intramolecular hydrogen bond is lengthened relative to the noncomplexed ortho-methoxyphenol by the movement of the hydrogen atom out of the plane of the benzene ring by about 16°. The sharing of the hydrogen causes the computed intermolecular hydrogen bonds of ortho-methoxyphenol to deviate

Figure 4. Lowest-free energy conformations for the complexes of EP and (a) para-methoxyphenol, (b) ortho-methoxyphenol stacked, and (c) ortho-methoxyphenol in the three-centered bond arrangement in hexane. The dashed lines show the inter- and intramolecular hydrogen bonds. The distances are in Angstroms and are measured between the hydroxyl hydrogen of the methoxyphenol and the carbonyl oxygen of EP for the intermolecular hydrogen bond and between the ether oxygen and hydroxyl hydrogen for the intramolecular hydrogen bond. The bold arrows indicate the direction of the motion of the hydrogen atom in the O-H stretching mode.

further from ideal geometry than those formed by paramethoxyphenol (Figures 5b and 6b). The calculations yielded no conformations in which the intramolecular hydrogen bond of ortho-methoxyphenol is broken, indicating that it is quite stable. Indeed, calculations for ortho-methoxyphenol in hexane without EP indicate that the energy cost of breaking the intramolecularly hydrogenbonded conformation is 12-21 kJ/mol, depending upon the particular bond lengths and angles used (see Methods). This is consistent with experimental work showing that intramolecularly hydrogen bonded ortho-methoxyphenol is present at nearly 100% in nonpolar solvents.11 Thus, the calculations suggest that ortho-methoxyphenol associates less favorably with the acrylic

9884 J. Phys. Chem. B, Vol. 103, No. 45, 1999

Mardis et al.

Figure 7. 2,6-Dimethoxyphenol.

Figure 5. Intermolecular hydrogen bond distance between EP and para-methoxyphenol (a) and ortho-methoxyphenol (b) in hexane.

Figure 6. Intermolecular O‚‚‚H-O angle between EP and paramethoxyphenol (a) and ortho-methoxyphenol (b) in hexane.

ester sorbent than does para-methoxyphenol because the persistent intramolecular hydrogen bond effectively weakens the intermolecular hydrogen bond with the carbonyl group of the sorbent. It is worth remarking that three-centered hydrogen bonds have been observed in atomic-resolution structures of small molecules. Thus, a 1984 survey of the Cambridge Structural

Figure 8. Infrared spectra of ortho-methoxyphenol (- -) and 2,6dimethoxyphenol (s) alone and of ortho-methoxyphenol (-‚-) and 2,6-dimethoxyphenol (‚‚‚) with 2 mol/L EP in hexane. The peaks near 3560 cm-1 are due to the phenol O-H stretch. In the presence of EP, a new peak at 3460 cm-1 appears. The concentrations of orthomethoxyphenol and 2,6-dimethoxyphenol alone in hexane are 12 mM and 11.8 mM, respectively. Their concentrations in the presence of EP are 9.24 mM and 9.13 mM, respectively. Spectra were obtained at room temperature.

Database for NH‚‚‚OdC bonds found 20% of the hydrogen bonds to be three-centered.27 Such shared three-centered bonds have also been observed in carbohydrates,28 amino acids,29-30 and in a variety of small biological molecules.31 Thus, the mechanism of association of ortho-methoxyphenol with the acrylic ester sorbent suggested by the present calculations is physically reasonable. The earlier concept that binding of ortho-methoxyphenol to EP involves breakage of the intramolecular hydrogen bond (Figure 1) was based on IR spectra showing similar broad O-H stretching bands for both para- and ortho-methoxyphenol in the presence of EP. It was presumed that these spectra would be markedly different if they did not both result from similar two-centered intermolecular hydrogen bonds. However, since the calculations described above suggest that ortho-methoxyphenol forms a three-centered hydrogen bond, we wished to determine whether IR spectra can, in fact, distinguish between two-centered and three-centered hydrogen bonds in this system. This issue was examined both experimentally and theoretically, as is now described. The experimental test involved the use of 2,6-dimethoxyphenol, in which the hydroxyl group is flanked by two methoxy groups (structure shown in Figure 7). Whereas ortho-methoxyphenol can break its intramolecular hydrogen bond while retaining a planar OH conformation, 2,6-dimethoxyphenol can only break this bond by rotating the OH bond out of plane. Since phenol hydroxyls strongly prefer to remain in the plane of the aromatic ring,32 2,6-dimethoxyphenol is expected to retain its intramolecular hydrogen bond whether or not EP is present. We find that, in the absence of EP, the O-H stretch peak is nearly identical to that of the solitary O-H peak of orthomethoxyphenol (Figure 8) at 3560 cm-1. This result confirms that ortho-methoxyphenol exists primarily in a hydrogen-bonded state in hexane. In the presence of EP, the original peak at 3560 cm-1 is reduced in magnitude, and both 2,6-dimethoxyphenol and ortho-methoxyphenol show a new O-H stretching peak at

Adsorption of Phenols onto a Polymeric Sorbent a lower frequency, 3460 cm-1. The appearance and position of this new peak is consistent with the formation of new hydrogen bonds involving the O-H hydrogen.33 The persistence of the original peak at reduced magnitude suggests that a population of ortho-methoxyphenol molecules hydrogen-bonded to EP is in equilibrium with another population that lacks intermolecular hydrogen bonds. For 2,6-dimethoxyphenol, the new peak is presumably associated with a three-centered hydrogen bond, given that it is expected to retain its intramolecular hydrogen bond even in the presence of EP. The similarity of the hydrogen-bonded peak of 2,6-dimethoxyphenol to that of ortho-methoxyphenol suggests that this compound also forms a three-centered hydrogen bond when it binds to EP. Thus, this IR analysis is consistent with our calculations suggesting the formation of a three-centered hydrogen bond between ortho-methoxyphenol and EP. We were surprised that the stretch frequency of a hydrogen in a two-centered hydrogen bond (ortho-methoxyphenol:EP) should be virtually indistinguishable from that of a hydrogen in a three-centered hydrogen bond (ortho-methoxyphenol:EP and 2,6-dimethoxyphenol:EP). We therefore examined the reasonableness of this concept further by means of ab initio quantum mechanics calculations. The program GAMESS34 was used to carry out gas-phase optimization of the most stable hydrogen-bonded EP complexes found in the mining minima calculations described above. The para-methoxyphenol:EP complex used has a two-centered hydrogen bond and the orthomethoxyphenol:EP complex has a three-centered hydrogen bond. Energies were calculated at the restricted Hartree-Fock level using the 6-31G(d,p) split valence basis set. The optimizations did not markedly alter the conformations of the complexes. As compared to the minimum energy structures used in the free energy calculations, the minimized ab initio structures had slightly longer intermolecular O‚‚‚H distances, 0.11 Å for ortho-methoxyphenol and 0.23 Å for paramethoxyphenol. For ortho-methoxyphenol, the intramolecular O‚‚‚H distance lengthened by 0.08 Å. The intermolecular O‚‚ HO angle also increased by 7°, while the H-O-C-C dihedral angle (measuring how planar the O-H bond is with respect to the benzene ring) became more planar, increasing from 165° to 171°. After optimization, a harmonic wavenumber calculation was performed for each complex. This yielded scaled frequencies of 3673 and 3622 cm-1 for EP complexes with para- and ortho-methoxyphenol, respectively (scale factor ) 0.8935). Although the absolute frequencies calculated at the restricted Hartree-Fock level are not expected to correspond well with experimental frequencies, their differences are interpretable. The calculated difference of 51 cm-1 is small relative to the widths of these broad peaks (Figure 8) and also relative to the 200 cm-1 shift from the uncomplexed species. Thus, these calculations further support the concept that the IR frequency shifts cannot be used to distinguish between two-centered and threecentered hydrogen bonds in this system. Further insight into the dynamics of the O-H stretch motions in these two cases can be gained by examining the direction of the hydrogen motion. The harmonic normal modes for the paramethoxyphenol:EP and ortho-methoxyphenol:EP complexes are determined from the same ab initio calculations as the frequencies. As shown by the bold arrows in Figure 4, the hydrogen in both cases vibrates along a vector joining the hydroxy oxygen with the carbonyl oxygen of EP, despite the neighboring methoxy oxygen in the case of ortho-methoxyphenol. No other atoms move significantly during this vibration. This supports

J. Phys. Chem. B, Vol. 103, No. 45, 1999 9885

Figure 9. Comparison of experimental adsorption affinities of paramethoxyphenol in water (4), ortho-methoxyphenol in water (O), paramethoxyphenol in hexane (+), and ortho-methoxyphenol in hexane ()) onto the acrylic ester sorbent. The limit of the solid curves as concentration (C) goes to zero gives the adsorption affinity q/C in units of L/g. For ortho in water, q/C ) 0.85, for para in water q/C ) 1.04. Measurements were taken at 25 °C.

the idea that the stretch motion is dominated by the intermolecular hydrogen bond rather than the intramolecular hydrogen bond. Adsorption from Water. If hydrogen bonding is important in the selectivity of the acrylic ester sorbent for para-methoxyphenol over ortho-methoxyphenol in hexane, then this selectivity is likely to disappear in water, a solvent that competes with and effectively weakens intersolute hydrogen bonds. New measurements, presented in Figure 9, show that the adsorption of para- and ortho-methoxyphenol in water onto the acrylic ester sorbent are indeed nearly equal. As Table 2 indicates, the ratio of adsorption affinities (para:ortho), based on the low concentration data, is 1.2, implying very weak selectivity. This result is consistent with the concept that hydrogen bonds are important for the binding of methoxyphenols onto the acrylic ester from hexane but not from water. Calculations of the standard binding free energies for the methoxyphenols onto EP from water yield -5.0 and -5.2 kJ/ mol for para and ortho, respectively. These values correspond to a ratio of binding constants of 0.90, in good agreement with the measured ratio of adsorption affinities. In contrast with the results for hexane, neither of the low-energy conformations of either isomer with EP show intermolecular hydrogen bonding (Figure 10). Instead, the MP molecules stack against EP to maximize nonpolar interactions. This mode of binding is consistent with the effective weakening of hydrogen bonds by the aqueous solvent. Previous adsorption studies have also provided phenomenological evidence that the adsorption mechanism changes with the solvent.36 It is of interest to compare the affinities of the methoxyphenols for the acrylic ester sorbent in water versus those in hexane. The binding affinity measurements in hexane taken from ref 7 are displayed with the water measurements in Figure 9. It is evident that the adsorption affinities of both isomers in water are similar to that of para-methoxyphenol in hexane, and that ortho-methoxyphenol in hexane has a significantly lower affinity. In contrast, the calculations suggest that the affinities for EP of both isomers in water are similar to that of orthomethoxyphenol in hexane (∆G°bind ≈ -2.6 to -5.2 kJ/mol), and that para-methoxyphenol in hexane has a significantly higher affinity (∆G°bind ≈ -9.2 kJ/mol). This discrepancy could result from differences between EP, which was used in the calculations, and the actual polymer sorbent used in the experiments.

9886 J. Phys. Chem. B, Vol. 103, No. 45, 1999

Mardis et al. has practical significance because it implies that a nonpolar medium must be used in order to separate positional isomers via differential hydrogen bonding. Indeed, maximal selectivity should be achieved by decreasing the polarity of the solvent and increasing the hydrogen bond-accepting ability of the sorbent. More generally, the present study indicates that significant insights into the mechanisms of adsorption can be gained by an investigative approach that combines experiment with molecular modeling. The strength of modeling is that it provides much greater molecular detail than can be obtained by adsorption studies and IR analysis alone. However, it remains important to test the results of modeling studies experimentally. A tremendous array of computational methods have been developed for simulating biomolecules such as proteins, nucleic acids, and the smaller molecules that interact with them. We believe that such techniques can also be useful in elucidating the molecular mechanisms underlying chemical separation processes and, ultimately, in developing separation processes that are of practical value.

Figure 10. Lowest free energy conformations for (a) para- and (b) ortho-methoxyphenol with EP in water.

While EP is a reasonable model for the hydrogen bonding sites of the acrylic ester, EP may under-represent the number of hydrophobic sites available on the sorbent for binding methoxyphenols from water. In this case, the binding calculations with EP would underestimate the true ratios of q/C in water to q/C in hexane, as observed here. Conclusions The computational and experimental studies described here provide further insight into the mechanisms by which small molecules bind to the surface of the acrylic ester sorbent XAD7. In particular, we provide support for the concept, previously presented, that hydrogen bonding plays a critical role in the adsorption of hydrogen bond donors from hexane. For, orthomethoxyphenol competition between intramolecular and intermolecular hydrogen bonding lowers the affinity, resulting in selective adsorption of para-methoxyphenol. The calculations also suggest that van der Waals “stacking” interactions appear to be significant in the binding of ortho-methoxyphenol onto the sorbent from hexane. Interestingly, the calculations suggest that breakage of the intramolecular hydrogen bond of ortho-methoxyphenol is not required for formation of an intermolecular hydrogen bond with the sorbent. Rather, the calculations indicate that orthomethoxyphenol forms a three-centered hydrogen bond that produces a weaker link with the acceptor carbonyl than that produced by the two-centered hydrogen bond of para-methoxyphenol. This computational result was tested by an IR study of 2,6-dimethoxyphenol and by quantum mechanical calculations of the vibrational spectra of the two- and three-centered hydrogen bonds in question. These studies are consistent with the existence of a three-centered hydrogen bond between orthomethoxyphenol and the acrylic ester sorbent. The binding of methoxyphenols to the sorbent from water appears to be dominated by nonpolar interactions rather than by hydrogen bonding, as was the case in hexane. This presumably results from the weakening of solute-solute hydrogen bonds by water. Thus, the medium appears to have a significant effect upon the mechanism of adsorption. This result

Acknowledgment. This work was supported by the National Institute of Standards and Technology, the United States Department of Agriculture through Grant 98-35504-6357, the National Science Foundation through Grant CTS-9531812, and REU supplements to this grant. K.L.M. was supported by a National Research Council Research Associateship. Certain commercial equipment or materials are identified in this paper in order to specify the methods adequately. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose. References and Notes (1) Farrier, D. S.; Hines, A. L.; Wang, S. W. J. Colloid Interface Sci. 1979, 69, 233. (2) Furuya, E.; Takeuchi, Y.; Noll, K. E. J. Chem. Eng. Jpn. 1989, 22, 670. (3) Winkler, K.; Radeke, K.-H.; Stach, H. Chem. Technol. (Leipzig) 1996, 48, 249. (4) Maity, N.; Payne, G. F.; Chipchosky, J. L. Ind. Eng. Chem. Res. 1991, 30, 2456. (5) Brune, B. J.; Payne, G. F.; Chaubal, M. V. Langmuir 1997, 13, 5766. (6) Chaubal, M. V.; Payne, G. F. Biotechnol. Prog. 1995, 11, 468. (7) Glemza, A. J.; Koehler, J. A.; Brune, B. J.; Payne, G. F. Ind. Eng. Chem. Res. 1998, 37, 3685. (8) Maity, N.; Payne, G. F.; Ernest, M. V.; Albright, R. L. Reac. Polym. 1992, 17, 273. (9) Robbers, J. E.; Speedie, M. K.; Tyler, V. E. Pharmacognosy and Pharmacobiotechnology; Williams and Wilkens: Baltimore, MD, 1996. (10) Tenulkar, S. B.; Tambe, S. S.; Chandra, I.; Rao, P. V.; Naik, R. V.; Kulkarni, B. D. Ind. Eng. Chem. Res. 1998, 37, 2081. (11) Berthelot, M.; Laurence, C.; Lucon, M.; Rossignol, C.; Taft, R. W. J. Phys. Org. Chem. 1996, 9, 626. (12) Head, M. S.; Given, J. A.; Gilson, M. K. J. Phys. Chem. A 1997, 101, 1609. (13) Gilson, M. K.; Given, J. A.; Bush, B. L.; McCammon, J. A. Biophys. J. 1997, 79, 5333. (14) Davis, M. E.; Madura, J. D.; Luty, B. A.; McCammon, J. A. Comput. Phys. Commun. 1991, 62, 187. (15) Luo, R.; Head, M. S.; Given, J. A.; Gilson, M. K. Biophys. Chem. 1999, 78, 183. (16) David, L.; Luo, R.; Head, M. S.; Gilson, M. K. J. Phys. Chem. B 1999, 103, 1031. (17) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4, 187-217. (18) Schlenkrich, M.; Brickmann, J.; A. D. MacKerrell, J.; Karplus, M. In Biological Membranes: A Molecular PerspectiVe from Computation and Experiment; Merz, K. M., Roux, B., Ed. Birkhauser: Boston, 1996; pp 3181.

Adsorption of Phenols onto a Polymeric Sorbent (19) Sitkoff, D.; Sharp, K. A.; Honig, B. J. Phys. Chem. 1994, 98, 1978. (20) Hawkins, G. D.; Cramer, C. J.; Truhlar, D. G. J. Phys. Chem. B 1998, 102, 3257. (21) Gilson, M. K.; Sharp, K. A.; Honig, B. H. J. Comput. Chem. 1988, 9, 327-335. (22) Gilson, M. K.; Honig, B. Proteins: Struct. Funct. Genet. 1988, 4, 7-18. (23) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990, 112, 6127. (24) Qiu, D.; Shenkin, P. S.; Hollinger, F. P.; Still, W. C. J. Phys. Chem. 1997, 101, 3005. (25) Gilson, M. K.; Honig, B. J. Comput.-Aided Drug Des. 1990, 5, 5. (26) Handbook of Chemistry and Physics; Weast, R. C., Ed.; CRC Press: Boca Raton, FL, 1989. (27) Taylor, R.; Kennard, O.; Versichel, W. J. Am. Chem. Soc. 1984, 106, 244.

J. Phys. Chem. B, Vol. 103, No. 45, 1999 9887 (28) Ceccarelli, C.; Jeffrey, G. A.; Taylor, R. J. Mol. Struct. 1981, 70, 255. (29) Jeffrey, G. A.; Maluszynska, H. Int. J. Biol. Macromol. 1982, 7, 336. (30) Jeffrey, G. A.; Mitra, J. J. Am. Chem. Soc. 1984, 106, 5546. (31) Jeffrey, G. A.; Maluszynska, H. Acta Crystallogr. 1990, B46, 546. (32) Wright, J. S.; Carpenter, D. J.; McKay, D. J.; Ingold, K. U. J. Am. Chem. Soc. 1997, 119, 4245. (33) Takasuka, M.; Matsui, Y. J. Chem. Soc., Perkin Trans. 2 1979, 1743. (34) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S. J.; Windus, T. L.; Dupuis, M.; Montgomery, J. A. J. Comput. Chem 1993, 14, 1347. (35) Scott, A. P.; Radom, L. J. Phys. Chem. 1996, 100, 16502. (36) Maity, N.; Payne, G. F. Langmuir 1991, 7, 1247.