Article pubs.acs.org/Langmuir
Adsorption of the Compounds Encountered in Monosaccharide Dehydration in Zeolite Beta Marta León, T. Dallas Swift, Vladimiros Nikolakis,* and Dionisios G. Vlachos Catalysis Center for Energy Innovation, Department of Chemical and Biomolecular Engineering, University of Delaware, 150 Academy Street, Newark, Delaware 19716, United States S Supporting Information *
ABSTRACT: A comprehensive study of the adsorption of the compounds involved in the reaction of dehydration of fructose to 5-hydroxymethyl furfural (HMF) on the zeolite H-BEA with SiO2/Al2O3 = 18 has been carried out. Furthermore, a method for the estimation of the real adsorption loading from the experimentally measured excess adsorption is developed and applied to calculate the adsorption isotherms both in the case of single-solute and multisolute mixtures. It was found that zeolite H-BEA adsorbs HMF and levulinic acid from water mixtures to greater extent than sugars and formic acid, which prefer to partition in the aqueous phase. HMF and levulinic acid adsorption isotherms could be fitted in a Redlich-Peterson isotherm model, while the adsorption of formic acid is better fitted using the Freundlich model and sugars via the Henry model. Adsorption loadings decreased with increasing temperature (0, 25, and 40 °C), which is characteristic of an exothermic process. From the temperature dependence of the isotherms, the limiting heat of adsorption at zero coverage was determined using van’t Hoff equation. Given the importance and the complexity of multicomponent systems, several experiments of adsorption of multisolute solutions have been carried out. In most of the cases, the ideal adsorbed solution theory (IAST) has been proven to satisfactorily predict adsorption from multisolute mixtures using as input the single-solute isotherms. fructose and HMF forming the so-called humins10 (Scheme 1). The possibility of tuning acidic and basic properties, hydrophilic and hydrophobic character, as well as adsorption and shape-selectivity properties make zeolites one of the most advantageous catalysts. Recently, our research group studied the dehydration of fructose to HMF using the protonated form of zeolite beta, with SiO2/Al2O3 ratio of 18.11 In order for a molecule to reach the catalyst actives sites, diffusion and adsorption have to occur first. The preferential adsorption of reactants, products, or byproducts could markedly affect the yield to the desired product. In this regard, studying the adsorption of the compounds present in the reaction media in the zeolite catalyst becomes crucial to understand the reactivity of the system. In addition, knowledge of the adsorption loadings is necessary to calculate the intrinsic reaction rate constants from kinetic data. It would be desirable to have the adsorption isotherms at the reaction temperatures, but unfortunately this is not possible since adsorption and reaction cannot be isolated. One way to overcome this difficulty is to measure the isotherms at low temperatures where the time required for the equilibration is much shorter than the time needed for reaction. By fitting such data with a suitable adsorption model, the temperature dependence can be estimated to make predictions of the adsorption at reaction
1. INTRODUCTION The synthesis of chemicals or fuels from renewable sources has recently gained a lot of interest. Biomass carbohydrates are the most abundant renewable resources available and have the potential to replace petroleum as a source of both energy and chemicals, leading to more environmentally friendly processes and reducing the dependence on crude oil. A promising reaction in this field is the dehydration of carbohydrates, such as fructose and glucose, toward 5-hydroxymethylfurfural (HMF).1 HMF and its furan derivatives are envisioned as substitutes of key petroleum-based building blocks for the synthesis of polymer precursors, pharmaceuticals, solvents, and numerous organic intermediates.2 The formation of HMF occurs by the loss of three water molecules in an acid-catalyzed conversion of fructose or glucose. This reaction can be catalyzed by either Brønsted or Lewis acids. It has been shown that nearly one hundred inorganic and organic acid compounds can successfully catalyze hexose dehydration reactions. H2SO4, H3PO4, and HCl are commonly used inexpensive catalysts.3 However, heterogeneous catalysts offer several advantages over homogeneous catalysts, like ease of catalyst recovery, reusability, minimization of equipment corrosion, etc.4 Some of the solid acid catalysts studied include zeolites,5,6 ion exchange resins,7 oxides,8 and phosphates.9 It has been demonstrated that the HMF formation chemistry is very complex; besides fructose or glucose dehydration, it includes a series of side reactions, such as HMF rehydration to form levulinic acid and formic acid, or polymerization of © XXXX American Chemical Society
Received: March 26, 2013 Revised: May 2, 2013
A
dx.doi.org/10.1021/la401138g | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Scheme 1. Reaction Network Corresponding to the Dehydration of Fructose to HMF
Table 1. Adsorption Loadings (mmol/g) of HMF and Sugars in Zeolites Reported in the Literaturea Ranjan et al.16
Schollner et al.12
Holtkamp et al.15
25 °C
HMF
30 °C
Fructose
30 °C
Glucose
BEA(12) BEA(100) ZSM-5(23) ZSM-5(280) Silicalite-1 FAU(30) FAU(60) FER(20) FER(55)
0.833 1.269 0.595 1.269 1.427 1.467 1.467 0.110 0.242
83CaNaY(2.6) 77CaNaY(2.6) 63CaNaY(2.6) 47CaNaY(2.6) 26CaNaY(2.6) NaY(2.6) 64CaNaY(2.55) 63CaNaY(2.6) 63CaNaY(2.42) 59CaNaX(1.4)
0.149 0.087 0.051 −0.025 −0.099 −0.238 1.440 0.489 0.253 −0.916
BEA(12.5) BEA(75) BEA(225) (4 °C) BEA(75) (50 °C) BEA(75)
0.355 0.360 0.089
Heper et al.13
0.467 0.467
Berensmeier and Buchholz14
50 °C
Fructose
Glucose
30 °C
Fructose
Glucose
NaY CaY MgY NH4Y
2.71 3.35 2.01 1.48
2.90 2.27 1.88 1.52
BEA(75)
0.486
0.328
a The equilibrium concentration of HMF is ∼0.12 mol/L and that of sugars ∼1.1 mmol/L except for Heper et al., who do not report it. The Si/Al ratio of each zeolite is shown in brackets after its name.
al.,15 who obtained similar results. Finally, Ranjan et al.16 reported isotherms of HMF adsorption from water on zeolites beta, ZSM-5, Y, and ferrierite of different Si/Al ratios at 25 °C. Adsorption loadings from solution are usually estimated from changes of the solute concentration assuming that the solution volume is not affected by adsorption. This quantity is called excess adsorption loading and sometimes it is quite different from the real loading. The concept of excess adsorption is discussed in detail in section 2, along with the models found in the literature for determining the real loading from the corresponding excess values. In the present work we study the adsorption of fructose, glucose, HMF, and levulinic and formic acid from aqueous solutions in the protonated form of zeolite beta with SiO2/ Al2O3 ratio of 18. For first time, the effect of temperature (0, 25, and 40 °C) and the presence of more than one solute in the mixtures are examined. A method for the estimation of the real loading from the excess adsorption based on the zeolite pore volume is developed. It can be applied in the case of multisolute systems and it is herein used to analyze our experimental data. The experimental isotherms were fitted to adsorption models with one (Henry), two (Freundlich and Langmuir), and three adjustable parameters (Sips, Redlich-Peterson, and Toth). The best isotherm model for each solute was used as input to the ideal adsorbed solution theory (IAST) that was used to predict
temperatures. Single-solute adsorption constitutes an essential first step in the study of adsorption from solutions. However, during sugar dehydration new compounds appear and their relative concentrations change with time. Therefore, understanding how the presence of additional solutes affects adsorption is also extremely important. However, it is impractical to experimentally study all possible combinations of components. Thus, models that can predict the adsorption from multisolute mixtures, based on single-solute isotherms, constitute considerably helpful tools. Only a small number of publications report the adsorption of components involved in monosaccharide dehydration to HMF, with most of the existing data being limited to sugars and HMF. Furthermore, in most of the previous studies the motivation was either to separate glucose from fructose or to remove fermentation inhibitors. Results from these previous studies are summarized in Table 1. Schollner et al.12 studied the effect of the cation on the adsorption of monosaccharides on ionexchanged (Ca and Na) X and Y zeolites. The adsorption kinetics of fructose, glucose, and mixtures thereof in ionexchanged (Na, Ca, Mg, and NH4) Y zeolites were studied by Heper et al.13 However, the adsorption data reported in these two publications are qualitatively different (Schollner et al. found that water adsorbs preferentially on NaY, whereas Heper et al., the opposite). The adsorption of sugars in zeolite beta was studied by Berensmeier and Buchholz14 and Holtkamp et B
dx.doi.org/10.1021/la401138g | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Finally, the excess adsorption can be calculated from changes in concentration expressed per volume or as mole fractions. Both definitions are not equivalent (see Supporting Information, Section S3), giving rise to different values for the same system, which has to be taken into account when comparing literature data. The use of the excess adsorption is supported by the IUPAC since it can be determined directly from experimental measurements and provides important information on the adsorption at the solid−liquid interface.18 However, if the magnitude of interest is the actual loading on the solid in equilibrium with the liquid, the excess adsorption provides a reasonable approximation only in certain cases. The maximum concentration at which the real adsorption can be approximated by the excess adsorption depends on the relative adsorption of solute and solvent,19 but this issue is not always taken into account in the analysis of liquid phase adsorption. Several reviews on adsorption from solutions onto solids dealt with the concept of excess adsorption.19−22 Since it is not possible to relate the composition of the adsorbed phase with experimentally measurable quantities, a model that describes the adsorbed phase is needed. Some of the different approaches proposed in the literature (e.g., works of Schay and Nagy,23−26 Everett,27,28 Schollner et al.,12 Berensmeier and Buchholz,14 and Holtkamp et al.,15) are reviewed in the Supporting Information (Section S1). Two models commonly used consider the adsorbed phase either confined to a single molecular layer next to the surface of the adsorbent or as filling the available pore space. This second approach is known as the pore-filling model and is generally applied for adsorption in solids with relatively narrow pores, like the zeolite micropores, when it can be assumed that the pores are equally accessible by all components.29 However, the assumptions of the existing methods are not valid for the multisolute systems of our study. Below, we propose a model that can overcome the aforementioned challenges.
multisolute adsorption. Finally, the limiting heat of adsorption at zero coverage was determined by the van’t Hoff equation.
2. CONCEPT OF EXCESS ADSORPTION FROM SOLUTION AND ITS LIMITATIONS The adsorption on solids from liquid solutions is often treated in a similar fashion to adsorption from the gas-phase. However, this approach is not necessarily correct. In the case of adsorption from a single gas-phase, it is possible to measure the adsorption loading. On the other hand, this is not possible in the case of adsorption from liquids. The most common method for determining the adsorption uptake in solid−liquid systems is the so-called batch method. The solid is brought in contact with a liquid and the change in liquid composition is measured after the equilibrium is reached. The adsorption of the solute is often assumed to be equal to the change in concentration (expressed as mass or moles per volume) multiplied by the initial volume of the solution. This value, called “adsorption excess”,17 does not always correspond to the real loading of the solute in the solid, since the volume of the solution after equilibration may be different from that of the initial solution. This change in volume is due to coadsorption of solute and the solvent in the solid phase, but can also be due to changes of the liquid phase density. The difference between the excess and the real adsorption depends on the relative adsorption of the solute and solvent. Three characteristic examples are shown is shown in Figure 1, (plotted using eq S9;
3. EXPERIMENTAL SECTION
Figure 1. Three representative cases of excess isotherms (solid lines) as a function of equilibrium mole fraction compared to the corresponding real adsorption of the solute (dashed lines) and the solvent (dotted lines). Left graph: only the solute adsorbs; middle graph: both the solute and the solvent adsorb; right graph: the solvent is preferentially adsorbed over the solute.
The protonated form of zeolite beta, herein named H-BEA, was used in this study. It was obtained by thermal treatment in air (1 h at 95 °C and 8 h at 450 °C with heating rates of 2 °C/min) of the commercial ammonium form (CP 814N, powder, Zeolyst International). The SiO2/Al2O3 of the zeolite is 18 according to the provider’s specifications. The adsorbates, HMF (≥99%), levulinic acid (98%), formic acid (98%), fructose (≥99%), and glucose (≥99.5%), were purchased from Sigma-Aldrich and used as supplied. The morphological characterization of the adsorbent was carried out by N2 physisorption at 77 K in a Micromeritics ASAP 2020 instrument. The micropore volume was calculated using the t-plot method on the N2 adsorption isotherm. The adsorption experiments were carried out in a batch system. Degassed zeolite samples (0.2 g) were suspended in solutions (1 mL) of known concentration of the adsorbates in DI water and shaken in a thermostatic bath (at 0, 25, 40 °C) for at least 5 h. Initial experiments of adsorption with time were carried out and it was found that 5 h was enough to reach equilibrium. Then the zeolite was removed by filtration (0.2 μm nylon membrane filter with modified acrylic housing, Nalgene) and the clear liquid was analyzed via HPLC (Waters e2695 with Aminex HPX-87H column and 2414 Refractive Index Detector) to determine the equilibrium concentration. A 5 mM aqueous solution of sulfuric acid was used as mobile phase with a flow rate 0.65 mL/ min. The temperature of the column and the detector was 65 and 35 °C, respectively. Each adsorption experiment was performed in triplicate and the values reported correspond to the average of the measurements. Adsorption from single solute mixtures was measured from solutions of different initial concentrations. In the case of multisolute
see Supporting Information). The graph on the left represents a system where only the solute, but not the solvent, is adsorbed. For dilute solutions, the excess and the real isotherms are equivalent, but as the concentration of the solute increases, the excess adsorption passes through a maximum. In the middle graph, both the solute and the solvent adsorb. The excess isotherm of the solute is close to the real one in the dilute region, although this region spans a smaller concentration range than in the previous case. At some point the adsorption loading of the solvent becomes higher than that of the solute and the excess adsorption acquires negative values. Finally, the graph on the right corresponds to a system in which the solvent is preferentially adsorbed over the solute. In this case, the excess isotherm is negative over the entire composition range. Considering the former remarks, the real adsorption can be approximated by the corresponding excess values for sufficiently dilute solutions and when the adsorption of the solvent is not considerably higher than the adsorption of the solute. The excess adsorption is still of practical interest since it is a measure of the relative adsorption of the components. C
dx.doi.org/10.1021/la401138g | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Langmuir33 (qsat and bL); and three for Sips34 (qsat, bS, and nS), Redlich-Peterson35 (KRP, bRP, and nRP), and Toth36 (KT, aT, and tT). In our analysis we assume that the activity coefficients of the solutes are equal to 1 (using as reference the chemical potentials at infinite dilution) since we study dilute systems (molar fraction
