3932
Ind. Eng. Chem. Res. 2007, 46, 3932-3940
Effect of Internal Diffusion in Supported Ionic Liquid Catalysts: Interaction with Kinetics Jyri-Pekka Mikkola,* Johan Wa1 rnå, Pasi Virtanen, and Tapio Salmi Laboratory of Industrial Chemistry, Process Chemistry Centre, Åbo Akademi UniVersity, Biskopsgatan 8, FI-20500 Åbo-Turku, Finland
Kinetics and mass transfer for supported ionic liquid catalysts (SILCAs) were studied for complex catalytic hydrogenations of unsaturated aldehydes. The system can be characterized as a gas-liquid-liquid-solid system, where gaseous hydrogen and an unsaturated aldehyde, citral, diffuse from molecular solvent to a thin layer of ionic liquid containing the catalytically active species, metallic palladium nanoparticles. The intrinsic kinetics was determined at 353-423 K (80-150 °C) and pressures ranging from 10 to 50 bar. A kinetic model was proposed that was able to predict the experimental concentration data successfully. In order to obtain more understanding about the importance of mass-transfer and diffusion limitations for these kinds of processes, a number of models describing the concentration profiles of the reactants in the catalysts were developed. The developed reaction-diffusion models revealed important insight into the potential limiting phenomena for this type of reaction systems. Catalytic hydrogenation of unsaturated aldehydes is an important class of reactions in the production of fine and specialty chemicals. Typical examples of hydrogenation products are, e.g., components for perfumes and insect repellents, flavors, and pharmaceuticals.1-5 The factors affecting the hydrogenation selectivity of R,β-unsaturated aldehydes are multitude and have been reviewed by, e.g., Gazellot and Richard6 as well as Claus et al.7 These factors are typically such as the nature of active metal, support and ligand effects, metal-support interactions, size of active (metal) site and its morphology, influence of the poisons and reaction products (product poisoning), promoter effects (usually an add-on metal), and reaction conditions (temperature, pressure, external mass transfer) as well as steric effects. Moreover, the nature of catalyst, traditional powder (slurry, fixed bed, etc.) or more sophisticated structured ones nicely reviewed by Cybulski and Moulijn,8 bears importance especially from the selectivity point of view. Importantly, internal mass-transfer (inside the catalytic material) as well as coupled diffusion and reaction that can have major influence on the performance of, especially, an industrial process is often not studied in detail, and consequently, misleading conclusions can be drawn about the limiting phenomena involved. The interaction of reaction and diffusion in porous catalysts has been experimentally studied and modeled by, e.g., Salmi et al.9 and Toppinen et al.10,11 However, often the studies have been conducted on chemically relatively simple systems, except work by, e.g., Julcour et al.12 and Aumo et al.5 in which more complex hydrogenation systems were addressed and reaction-diffusion phenomena were quantitatively modeled. Ionic liquids (ILs) or room-temperature salt melts are a novel class of materials or neoteric solvents that can be applied in many fields, such as organic synthesis,13 material synthesis and nanotechnology,14 or entrapment and activation of catalytically active species (enzymatic and metallic) for catalytic applications.15,16 Generally, ionic liquids are associated with properties such as an extremely wide liquidus range (up to hundreds of degrees), very low vapor pressures, and nonflammability (however, certain ionic liquids will burn17). Nevertheless, an * To whom correspondence should be addressed. E-mail: jpmikkol@ abo.fi.
educated reader should, indeed, keep in mind that facts and fiction often blend in ionic liquid literature: generic statements are considered specific whereas specific are assumed as generic.18 The concept of immobilized catalytic species (e.g., enzymes, organometallic compounds, and metal nanoparticles) in ionic liquids has displayed its potential and was applied successfully in various catalytic transformations.19-23 Until now, application research in connection to ionic liquids has been somewhat retarded by the availability and cost of several ionic liquids. Although hundreds of ionic liquid preparation recipes have been published, not all laboratories focused on applied research have the expertise, work practices, and equipment necessary to carry out synthesis work. In addition, to prepare pure, dried ILs or to execute postsynthesis purification steps requires vacuum equipment and, sometimes, sophisticated procedures. On the other hand, commercial availability of several ionic liquids has clearly improved, although the cost of many liquids is still painfully high for applied engineering research. Therefore, the concept of immobilized ionic liquids entrapped, for instance, on the surface and pore-structure of various porous solid materials offers an attractive alternative to study the performance of ILs cost-effectively. Also, the higher relative viscosity of many ILs, compared to that of classical molecular solvents, promises enhanced performance for this concept: the established thin ionic liquid layer is an advantage, facilitating a more rapid diffusion and mass transfer, in comparison to cases when a bulk ionic liquid acts as a solvent. So far, in the case of ionic liquids no reaction-diffusion studies have emerged in the open literature, partially due to the lack of relevant physicochemical data. Although measures such as density, viscosity, and solubility of industrially most important gases have been determined for a few cation-anion combinations, the mere number of potential ionic liquids is so overwhelming, at least a million,24 that there is little hope for a complete mapping ever to be available. The issue is further complicated by the fact that especially the purity (halide contamination from IL synthesis), water content (virtually all ILs, even the “hydrophobic” ones, display a limited water miscibility), and eventual cosolvents or solutes can have a dramatic effect on the physicochemical properties.25,26 In this
10.1021/ie061082o CCC: $37.00 © 2007 American Chemical Society Published on Web 11/22/2006
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 3933 Scheme 1 Reaction Network in the Hydrogenation of Citral on SILCA Catalysts
paper, we have made the first attempt, according to our best knowledge, to model the concentration profiles in a thin ionic liquid layer residing on a porous, solid support. With all of its shortcomings, this engineering approach is intended as a straightforward procedure to obtain insight about the internal life of the catalyst and, hopefully, in future function as a valid tool in the evaluation of the performance of SILCA-type catalysts. Experimental Section The model reaction, hydrogenation of citral, is an interesting one, since the multi-unsaturated model molecule contains both a carbon-carbon double bond and a carbonyl group. Upon hydrogenation, numerous parallel and consecutive reactions typically take place. The reaction scheme for the hydrogenation of this R,β-unsaturated aldehyde is presented in Scheme 1. The
supported ionic liquid catalysts (SILCAs) were prepared and characterized according to procedures described in our earlier publications.22,23,27 The procedure generally resulted in Pd0 nanoparticles that remained (immobilized) in the thin ionic liquid layer in which the reactions proceeded. A structural catalyst was prepared by means of an active carbon cloth (Kynol) attached as stirrer blades in a modified high-pressure laboratory autoclave operating in a semibatch mode (in terms of hydrogen gas). n-Hexane was chosen as an economical bulk solvent that is not comiscible with the ionic liquids ([bmim][PF6], [bmim][BF4], [A336][PF6], and [NB4MPy][BF4]). Figure 1 introduces the catalysts prepared by means of various magnifications, illustrating the different levels of organization of the material. The level of leaching of both Pd and ionic liquids into the liquid bulk phase was controlled and found to be negligible.22 The reaction mixture was analyzed by means of gas chromatography
Figure 1. Schematic illustration of the catalyst structure and different characteristic dimensions in the SILCA catalysts: (a) photograph of the carbon matt; (b) SEM image illustrating the knitted structure of the active carbon matt; (c) SEM image illustrating the single fibers; (d) EFTEM image of the Pd nanoparticles residing in the ionic liquid.
3934
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007
Table 1. Selected Physicochemical Data
ionic liquid [Bmim][BF4]
[Bmim][PF6]
molar mass M (g/mol) 226.02
284.19
molar vol (L/mol)
density F (g/mL)
0.188 (calcd with density of 1.20)
1.20d (293 K)
15.4k (154)
1.5 × 10-1 e (50 atm)
1.12b,c 1.21i
219l (298 K) 154n-201o 219q 233m (298 K) 48.9k (489) 450l (298 K)
0.86 × 10-3 f (1 atm, 293 K)
397q (298 K, H2O equil) 450q 312s 312m (298 K) 270p (298 K)
4.4 × 10-2 e (50 atm)
0.73 × 10-3 f (1 atm, 293 K)
1.184h
38.1k (381)
0.89 × 10-3 f 0.90 × 10-3 f (1 atm, 298 K)
0.9g
247m (298 K) 517h (283 K) 246h (293 K) 177h (298 K) 132h (303 K) 78h (313 K) 48h (323 K) 33h (333 K) 23h (343 K) 1500j (303 K)
0.209r 0.207 (calcd with density of 1.37)
1.37d 1.36b,c 1.37i
[NB4MPy][BF4]
[A336][PF6]
237.05
541.5
viscosity µ (cP ) mPa s)
0.200 (calcd with density of 1.184)
0.60 (calcd with density of 0.9)
H2 solubility (mol/L)
4.4 × 10-3 e (5 atm)
ACC + ILa specific surf area (m2/g)
ACC + ILa micropore vol (cm3/g)
1100
0.391
1365
0.485
8.9 × 10-1 e (100 atm)
a ACC + IL denotes the active carbon cloth impregnated with ionic liquid (fresh ACC: 1680 m2/g, 0.597 cm3/g); Dubinin surface area. b Reference 25. Reference 29. d Seddon; et al. Clean solVents: alternatiVe media for chemical reactions and processing; Abraham, M., Moens, L., Eds.; ACS Symposium Series 8/9; American Chemical Society: Washington, DC, 2002. e Reference 21. f For [buPy][Tf2N] and [BbumePy][Tf2N], ref 29. g Reference 23. h Crosthwaite, J. M.; Muldoon, M. J.; Dixon, J. K.; Anderson, J. L.; Brennecke, J. F. Phase transition and decomposition temperatures, heat capacities and viscosities of pyridinium ionic liquids. J. Chem. Thermodyn. 2005, 37, 559-568. For 1-butyl-3-methylpyridinium tetrafluoroborate equilibrated with water 327 ppm. i Poole, C. F. J. Chromatogr., A 2004, 1037, 49-82. j For Aliquat 336 (methyltrioctylammonium chloride). k Reference 37. Note: the values reported were presumably false by a factor of 10. l Reference 25. m www.sigma-aldrich.com/ionicliquids. n Reference 26. [Cl-] ) 0.01 mol/kg. o Reference 26. [Cl-] ) 0.5 mol/kg. p Dzyuba, S. V.; Bartsch, R. A. ChemPhysChem 2002, 3, 161; cited by ref 32. q Reference 25. r Suarez, P. A. Z.; Einloft, S.; Dullius, J. E. L.; de Souza, R. F.; Dupont, J. J. Chim. Phys. 1998, 95, 1626. s Reference 35. c
(GC), and the detailed description of analysis procedure can be found in earlier publications.22,23 The active carbon matt pieces were cut to pieces corresponding to 68 × 39 mm, with an average estimated thickness of 0.5 mm. The average fiber diameter in the knitted structure (Figure 1) and the Dubinin and BET surfaces as well as the micropore volumes of the catalysts were determined, and three different approaches were compared for the treatment of the catalyst in terms of catalyst efficiency: The first approach assumes that the ionic liquid layer is uniformly spread throughout the microporous structure of the catalyst (fresh carrier has a specific surface area around 1680 m2/g and a micropore volume of 0.597 cm3/g).22,23 Taking into account the measured densities of the ionic liquids and assuming cylindrical pore geometry, one can calculate that approximately 20-25% of the pore volume of the support was covered by the ionic liquid. Also, the reduction of the specific surface areas and specific pore volumes of the catalysts carriers (ACC matt)sas a result of the “incipientwetness impregnation” of the material with the ionic liquidss supports this assumption. Table 1 introduces the relevant data. Thus, one can estimate that the magnitude of order for the ionic liquid layer thickness is 0.5-1 nm. The other approaches are based on the hypothesis that only the outer layer of the matt structure is efficiently covered by the ionic liquids and, consequently, two different approximations can be selected: either the whole matt is considered as a flakelike geometry or all individual fibers are considered as very long cylinders with an average diameter of 9 µm. Also, one
should remember that due to the knitted structure, in average half of the fibers are perpendicular to each other. Consequently, the outer surface area of the catalyst carrier is around 5300 × 10-6 m2 (slab) or 0.53 m2 (long cylinders). Finally, on the basis of these simple calculations, one can estimate that the average ionic liquid layer thicknesses of the catalysts are assuming a layer-on-the-slab coverage 20-30 µm and a layer-on-the-fibers coverage 0.2-0.3 µm. This exercise bears importance, e.g., from the point of view of the catalyst efficiency and was intended via simulations to demonstrate the effect of the ionic liquid layer thickness, depending on that how the liquid is assumed to reside on the carrier. Physicochemical Data For meaningful calculation and simulation of the concentrations in the reaction zone, the ionic liquid layer, physicochemical data are required. Several parameters are needed in the model: hydrogen solubility, solvent viscosity, and molecular as well as effective diffusion coefficients. First, the solubility of hydrogen has been studied in very few ionic liquids only. Also, the results are not entirely in accordance with each other. For instance, Berger et al.28 published hydrogen solubility data in, e.g., [bmim][BF4] and [bmim][PF6] obtaining hydrogen solubilities that differ by 1 order of magnitude (1.5 × 10-1 vs 4.4 × 10-2 mol L-1 at RT (room temperature) and 50 atm). On the other hand, Dyson et al.29 reported Henry’s constants for the above-mentioned ILs to be 5.8 × 102 and 6.6 × 102 kH/MPa
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 3935 Table 2. Measured Diffusivities of Oxygen, Carbon Dioxide, and a Few Organics in the Ionic Liquid [bmim][PF6] at 303 K 25,a oxygen diffusivity
(m2/s)
0.22 ×
carbon dioxide
10-9
0.27 ×
10-9
ethylene 0.20 ×
10-9
propylene 0.11 ×
10-9
1,3-butadiene 0.12 ×
10-9
1-butene 0.08 ×
10-9
hydrogenb 0.33 ×
citralb
10-9
0.04 ×
dihydrocitronellalb
10-9
0.036 × 10-9
a Diffusion coefficients for hydrogen, citral, and dihydrocitronellal are calculated on the basis of the measured diffusion coefficients. b Calculated diffusion coefficients for hydrogen, citral, and dihydrocitronellal.
(at 293 K and 1 atm) indicating that there is no significant difference in the solubility of hydrogen in these ionic liquids. Interestingly, Dupont et al.30 propose that the 3-D molecular arrangement of imidazolium ILs forming chains of imidazolium rings generates channels in which anions are accommodated as chains. As a result, introduction of other molecules generates nanostructures with alternating polar/nonpolar regions. They also point out that polar and weakly coordinating solvents like ionic liquids can promote reactions involving charge-separated intermediates or transition statessa fact that further complicates the assessment of limiting phenomena. Chiappe et al.31 suggested that the formation of voids (cavities) in ILs and the ability of small molecules to move within them might enable H+ species to easily move between these voids and thus enable faster than the diffusion-limited movement of protons in ILs. Solvent density and viscosity play an important role from the diffusion velocity point of view. For the densities of ionic liquids, water equilibrated values were chosen, if available. Since the concentration of the solutes (e.g., citral and its reaction products) in the reactor was low (0.08 M solution in 250 mL n-hexane; the ionic liquids do not dissolve in n-hexane), it was assumed that the viscosities of the ionic liquids were not significantly affected by the dissolved species. However, a study by Wang et al.32 introduced measurements for several mixtures of ionic liquids and miscible organics. It was concluded that regardless of the organic liquid, the effect on IL viscosity was surprisingly similar and, thus, this approach could be utilized for more precise assessment of the viscosity of ionic liquid layers. Relevant physicochemical data in selected ionic liquids reported by several authors are displayed in Table 1. For the estimation of molecular diffusion coefficients, the classical approach of Wilke-Chang33 was adapted as an order-ofmagnitude engineering solution.
Di )
7.4 × 10-12xφMB/(g/mol)(T/K) 0.6
µ/cP VA
m2/s
(1)
With lack of better estimation, the dimensionless association factor, φ, could be assigned a value 1.9 (methanol) since ionic liquids are generally considered to have polarities in the range of methanol. However, a modified Wilke-Chang equation was chosen:33
7.48 × 10-12T(RMB)1/2
chosen were 3 and 0.5, respectively.34
Deff )
p D τp i
(3)
Morgan et al.35 recently introduced a few correlations for the estimation of gas diffusivities in ionic liquids. Although the correlations were developed on the basis of carbon dioxide and various hydrocarbons only, they are taken into this study as a first approximation. It was concluded that, in general, gas diffusion in ionic liquids (∼10-6 cm2/s) is slower than in hydrocarbon solvents or water. However, the dependence on viscosity is lower and the dependence of diffusivity on temperature and size of the solute gas is higher. The following two alternative formulas were introduced for the gas diffusion in ionic liquids, at 303 K:
Di ) 3.7 × 10-7
1 µ2
0.59(0.02
V1
1.00(0.07
F2
2.0(0.1
m2/s (4)
1 Di ) 2.66 × 10-7 0.66(0.03 1.04(0.08 m2/s µ2 V1
(5)
Here F2 is the density of IL (g/mL), µ2 the dynamic viscosity (cP), and V1 the molar volume in cm3/mol. Finally, they concluded that the dependence of diffusivity on viscosity (D R µ2-0.6) was lower than what Stokes-Einstein equation predicts. Moreover, the deviations between molecular solvents and ionic liquids might result from the relatively large differences in sizes: small solutes diffusing in an ocean of large IL solvent molecules. Huddleston et al.25 measured the diffusivities of oxygen, carbon dioxide, ethylene, propylene, 1,3-butadiene, and 1-butene at 303 K in [bmim][PF6] ionic liquid. The results are reproduced in Table 2. As we can see, the diffusivity of a small diatomic oxygen molecule is in the same order of magnitude as those of the hydrocarbons. Thus, the orders of magnitude for the diffusion coefficients of hydrogen, citral, and dihydrocitronellal (the main hydrogenation product), respectively, were calculated as follows:
VO20.6 V1,3butadiene0.6 DH2 ) DO2 0.6 Dcitral ) D1,3butadiene V H2 Vcitral0.6
(2)
V1,3butadiene0.6 Ddihydrocitronellal ) D1,3butadiene Vdihydrocitronellal0.6
Here the term (RMB)1/2 is used to account for solvent associations and size. MB is the molecular weight of the solvent, different ionic liquids (in the reaction zone), µ is the dynamic viscosity of the ionic liquids (cP) at temperature T (K), and VA is the molar volume of the liquid. Nevertheless, ionic liquids are actually electrolytes and, therefore, more advanced approaches in combination with actual diffusion coefficient measurements are anticipated in the future. For porous objects, the effective diffusion coefficient (Deff) is obtained from the molecular diffusion coefficient (Di) combined with the catalyst object porosity (p) and tortuosity (τp). The numerical values
Here the molar volumes of hydrogen, oxygen, citral, dihydrocitronellal, and 1,3 butadiene were 7.4, 14.8, 172, 179.6, and 24.5 cm3/mol, respectively. For the modeling, the molar volumes of the reactants and the reaction products were calculated on the basis of the Le Bas atomic increments36 method. For estimation of the temperature and pressure dependence of viscosity and hydrogen solubility, respectively, the following procedure was applied: Since data about the temperature dependence of the viscosity were available for [NB4MPy][BF4] only, the values for other ionic liquids were calculated on the basis of a scaling factor determined from a single value (Figure
Di )
µV h A0.6
m2/s R ) 0.15
3936
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007
Figure 2. Linearized viscosity vs temperature for the ionic liquids [Bmim][BF4], [Bmim][PF6], [NB4MPy][BF4], and [A336][PF6].
R3 ) kCOLcCALpH2rdeact/D
(8)
R4 ) kDCAL2cOCTpH2rdeact/D
(9)
R5 ) kDCAL1cCALpH2rdeact/D
(10)
R6 ) kTHG1cDCALpH2rdeact/D
(11)
R7 ) kTHG2cCOLpH2rdeact/D
(12)
(KHpH2)0.5)j. The exponent, j, was assigned values from 0 to 3. Although all models gave a better than 99% explanation of the experimental data, the value 2 was chosen due to the best fit. Moreover, the rate equations have a term, kdeact, that accounts for the slow deactivation of the catalyst in consecutive batch hydrogenations:
rdeact ) e-kdeacttcumu
(13)
Here kdeact is the deactivation rate constant and tcumu the cumulative time the catalyst has been exposed to the reaction conditions. The rest of the symbols are defined in the notation. The generation rates are, consequently, easily obtained by using the overall stoichiometry: rate 1 ) R1 - R2 rate 2 ) R1 - R3 - R5 rate 3 ) R2 - R4 rate 4 ) R3 - R7 rate 5 ) R4 + R5 - R6 rate 6 ) R6 + R7 Figure 3. Hydrogen solubilities for the ionic liquids [Bmim][BF4], [Bmim][PF6], and [NB4MPy][BF4].
2). Also, hydrogen solubility data as a function of pressure were available only for [Bmim][PF6] and, consequently, the solubilities were again scaled on the basis of a single value (Figure 3). Since no measurements were available for the IL [A336][PF6], the best approximation is to use the values reported for [Bmim][PF6] (in fact, this IL has the same anion, [PF6]-, and for ionic liquids the anion is considered to have a decisive role in terms of various physicochemical properties). Quantitative Intrinsic Kinetics As can be depicted from Scheme 1, the citral hydrogenation reaction network follows the expected paths reported by a multitude of authors.37 The experiments were carried out in the kinetic regime; i.e., the efficient mixing (1800 rpm) eliminated any external mass transfer limitations and the classical (semi-) batch reactor model was applied. A Langmuir-Hinshelwoodtype kinetic model was developed. As expected, various mechanisms, including competitive and noncompetitive adsorption of hydrogen vs organics, did not differ significantly in terms of the degree of explanation obtained for the model fit. Surface reactions were presumed to control the overall hydrogenation rate, and consequently, the following set of rate equations was obtained:
R1 ) kCALcApH2rdeact/D
(6)
Here the denominator is defined as D ) (1 + KAcA +
R2 ) kOCTcApH2rdeact/D
(7)
citral consumption citronellal formation 3,7-dimethyl-2-octenal formation citronellol formation dihydrocitronellal formation tetrahydrogeraniol formation
Due to the assumption of the absence of mass transfer limitations, the kinetic model was simply coupled to the component mass balances of the batch reactor,
dci ) Fcatri dt
(14)
where t denotes the reaction time (in a batch), Fcat is the catalyst bulk density, and ri is the generation rate of component i. The numerical parameter estimation was carried out by means of combined Simplex-Levenberg-Marquardt method,38,39 minimizing the residual sum of squares between the experimentally recorded, ci, and estimated, ciest, concentrations through nonlinear regression:
Q ) Σ(ci(t) - ciest(t))2
(15)
The ordinary differential equations were solved numerically during the parameter estimation (backward-difference method40). The modeling was carried out by means of the ModEst (model estimation) software package for parameter estimation, simulation, and optimization.41 A sample model fit to the kinetic data is introduced in Figure 4. A more detailed treatment of the kinetics can be found in our other publication.42 Diffusion-Limited Kinetics The diffusion-limited kinetics was considered next. For a catalyst object with an arbitrary geometry, the mass balance for an infinitesimal volume element is written as
(NiA)m + riFp∆Vp ) (NiA)out +
dni dt
(16)
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 3937
the overall liquid mass that can be expressed in a form
dni ) NiAp dt
(18)
where Ap stands for the total outer surface area of the catalyst shape. Since the liquid volume, VLi, remains constant, we can write
dcLi dni ) VLi dt dt
(19)
Furthermore, after defining that Ap/VLi ) ap, we get
dci ) NiAp dt
Figure 4. Sample hydrogenation batch of citral over a Pd-[Bmim][BF4] SILCA catalyst (373 K, 10 bar of H2). The symbols represent the experimental concentrations, and the lines, the model fit.
where A denotes the outer surface of the object residing in the reference volume, Fp is the density of the catalyst object, ∆Vp is the volume element of the catalyst object, and ni stands for the amount of the compound in the reference volume. This equation (eq 14) can be rewritten9 into
(
)
d(Nirs) dci ) p-1 riFp - s dt r dr
(20)
where Ni is obtained from the concentration gradient at the surface,
Ni - Deff
( ) dci dr
r)R
(21)
For improved numerical accuracy, the concentration gradient at the surface was derived through integration of the concentration profile in the catalyst object, as discussed by Salmi et al.9 Simulation Results
(17)
where s denotes the shape factor of the object (s ) 0, 1, or 2 for slabs, infinite cylinders, and spheres, respectively). For nonideal particles the value of the shape factor is calculated from s + 1 ) (5/3 + (1 + R)R/L)/(1 + R); R ) 1/(3π)0.5 - 1/6. In our case, as expected, the value of shape factor became very close to 0 or 1, depending on the assumption whether the catalyst was considered to have a form of a slab or (multiple) cylinders, respectively. The model of the diffusion flux is inserted into the mass balance, in accordance with the law of Fick, Ni ) -Deff(dci/ dr), and a dimensionless coordinate (x ) r/R) is utilized. Moreover, it is assumed that the effective diffusion coefficient remains approximately constant throughout the catalyst object. The boundary consitions are ci ) cLi and dci/dt ) 0, at r ) 0. Consequently, the model of the catalyst object is coupled to
Calculations indicated that the molecular diffusion coefficient of hydrogen is around 0.33 × 10-9 and that of citral about 0.4 × 10-10 m2 s-1, at 303 K (Table 2). The diffusion coefficients of the hydrogenation products are close to that of citral; for instance, that of dihydrocitronellal is 0.36 × 10-10 m2 s-1, at 303 K (Table 2). The reaction-diffusion model was coupled to the intrinsic kinetics to simulate the situation in the thin ionic liquid layer. The simulations gave the internal concentration profiles of, e.g., hydrogen, citral, and the hydrogenation products in the stagnant ionic liquid layer. As can be seen, the internal mass transfer in the layer is not very pronounced, although some differences can be observed, depending on the ionic liquid in question. Moreover, regardless of the ionic liquid involved, the simulated values are in the same order of magnitude, at comparable conditions. Furthermore, neither the diffusion patterns nor the simulated values of the effective diffusion coefficients were differing from each other, depending on that whether the model chosen was the classical Wilke-Chang
Figure 5. Concentration profiles of hydrogen, citral, and dihydrocitronellal in the catalytically active ionic liquid layer containing Pd nanoparticles ((SILCA)Pd-[bmim][BF4]). Conditions: 423 and 373 K, 30 and 10 of bar H2; situation at around 50% conversion; ionic liquid layer thickness 20 µm. The x-axis is the dimensionless coordinate along the ionic liquid layer. Also, simulated values of the effective diffusion coefficient obtained according to the model presented in eq 4 are listed.
3938
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007
Figure 6. Concentration profiles of hydrogen, citral, and dihydrocitronellal in the catalytically active ionic liquid layer containing Pd nanoparticles ((SILCA)Pd-[bmim][PF6]). Conditions: 373 and 363 K, 5 and 2 bar of H2; situation at around 30% conversion; ionic liquid layer thickness 20 µm. The x-axis is the dimensionless coordinate along the ionic liquid layer. Also, simulated values of the effective diffusion coefficient obtained according to the model presented in eq 4 are listed.
Figure 7. Concentration profiles of hydrogen, citral, and dihydrocitronellal in the catalytically active ionic liquid layer containing Pd nanoparticles ((SILCA)Pd-[A336][PF6]). Conditions: 373 K, 10 and 2 of bar H2; situation at around 60% conversion; ionic liquid layer thickness 20 µm. The x-axis is the dimensionless coordinate along the ionic liquid layer. Also, simulated values of the effective diffusion coefficient obtained according to the model presented in eq 4 are listed.
Figure 8. Concentration profiles of hydrogen and citral in the catalytically active ionic liquid layer containing Pd nanoparticles ((SILCA)Pd-[NB4MPy][BF4]). Conditions: 393 and 353 K, 20 and 10 of bar H2; situation at around 60% conversion; ionic liquid layer thickness 20 µm. The x-axis is the dimensionless coordinate along the ionic liquid layer. Also, simulated values of the effective diffusion coefficient obtained according to the model presented in eq 4 are listed.
approach or one of the proposed new correlations (eqs 4 and 5). Sample concentration profiles of hydrogen, citral, and dihydrocitronellal (the primary product) in the catalytically active ionic liquid layer containing Pd nanoparticles are illustrated in Figures 5-8, at various conditions and in dimensionless coordinates along the ionic liquid layer. In addition, the simulated values of the diffusion coefficient obtained according
to the model presented in eq 4 are listed. The simulations performed with various diffusion models did not give results that would significantly differ from each other. Also the concentration profiles remained essentially unchanged, regardless of the assumed IL layer thickness (1 nm, 0.2 µm, or 20 µm). Therefore, only the profiles obtained for the thickest layer (20 µm) are shown here.
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007 3939
Discussion and Conclusions
∆Vp ) volume element
As more accurate and reliable data on ionic liquids becomes available, a more realistic description of the kinds of chemical systems introduced in this paper will be possible. The concept and model presented can be refined, and for instance, the recently introduced model by Qin and Prausnitz43 for the solubilities of ordinary gases and liquids in ionic liquids will be attempted in future. Also, in the paper of Camper et al.44 a correlation for hydrogen solubility in ionic liquids, on the basis of regular solution theory valid for the low-pressure regime, could be tried. Although our model coupling kinetics and diffusion phenomena could be characterized as a “first trial”, giving an order of magnitide estimation of the reality, it is an initial step and might prove useful for evaluation of potential performance of other similar systems. It is also evident that upon modeling diffusion in a very thin layer of ionic liquids, like in this case, it is hard to demonstrate any notable concentration gradients. The situation would be entirely different if ionic liquids were applied as bulk solvents, and future efforts will be aimed toward this direction. At the very least, the estimated values of the effective diffusion coefficients are logical: the higher the viscosity of the ionic liquid in question, the lower the hydrogen pressure and the lower the temperature during the isothermal kinetic experiment, and the smaller are the values obtained. All in all, the effect of diffusion phenomena on the activity and selectivity in a complex multiphase (gas-liquid-liquidsolid) hydrogenation system for citral was studied experimentally and numerically. Citral was the model molecule the hydrogenation of which can give rise to several reaction products. The model described well the behavior of the system and can potentially be useful in studies involving process development and intensification where the effect of various process and design parameters are studied.
Greek Symbols
Acknowledgment The financial support from the Academy of Finland is gratefully acknowledged (Decision Nos. 209391 and 211463). This work is part of the activities at the Åbo Akademi Process Chemistry Centre (PCC) within the Finnish Centre of Excellence Programmes (2000-2005 and 2006-2011) by the Academy of Finland. Notation Ionic Liquid Nomenclature [Bmim][BF4] ) 1-butyl-3-methylimidazolium tetrafluoroborate [Bmim][PF6] ) 1-butyl-3-methylimidazolium hexafluorophosphate [A336][PF6] ) Aliquat 336 hexafluorophosphate [NB4MPy][BF4] ) N-butyl-4-methylpyridinium tetrafluoroborate Symbols A ) outer surface area of the catalyst object Ap ) total surface area of the catalyst object c ) concentration Di ) molecular diffusion coefficient Deff ) effective diffusion coefficient KA ) adsorption coefficient, citral MB ) molar mass of the solvent (ionic liquid) Ni ) molar flux of component i R ) reaction rate expression VA ) molar volume of the solved molecule
R ) association factor constant determined for ionic liquids32 p ) catalyst object porosity µB ) dynamic viscosity of the solvent Fcat ) catalyst bulk density τp ) tortuosity Φ ) association factor Sub- and Superscripts A ) citral H2 ) hydrogen CAL ) citronellal COL ) citronellol DCALi ) dihydrocitronellal (i denotes route 1 or 2) OCT ) 3,7-dimethyloctanol THGi ) tetrahydrogeraniol (i denotes route 1 or 2) Literature Cited (1) De Simone, R. S.; Gradeff, P. S. Process for the hydrogenation of citral to citronellal and of citronellal to citronellol suing chromium-promoted Raeny-nickel catalyst. U.S. Patent 4,029,709, 1977. (2) Marin, A. B.; Butler, J. F. Method for repelling fire ants and horn flies and compositions for repelling fire ants and horn flies and acting as anti-feedants for fire ants and horn flies. U.S. Patent 5,753,686, 1998. (3) Mohammadi, F.; Vargas, A. Cosmetic composition for stressed skin under extreme conditions. U.S. Patent 6,649,178, 2003. (4) Warren, C. B.; Butler, J. F.; Wilson, R. A.; Mookherjee, B. D.; Smith, L. C.; Marin, A. B.; Narula, A. P. S.; Boden, R. M. Insect repellent compositions and methods for using same. U.S. Patent 5,633,236, 1997. (5) Aumo, J. Structured Catalysts in Complex Three-Phase Hydrogenations. Doctoral Thesis, Åbo Akademi, Åbo-Turku, Finland, 2005; ISBN 952-12-1503-8. (6) Gazellot, P.; Richard, D. Selective hydrogenation of R,β-unsaturated aldehydes. Catal. ReV.sSci. Eng. 1998, 40, 81-126. (7) Claus, P.; Schimpf, S.; Gaube, J. Selective hydrogenation of multiple unsaturated compounds. In Basic principles in applied catalysis; Baerns, M., Ed.; Springer-Verlag: Berlin, Germany, 2004; p 85. (8) Cybulski, A., Moulijn, J. A., Eds.; Structured catalysts and reactors; Marcel Dekker: New York, 1998; ISBN 0-8247-9921-6. (9) Salmi, T.; Rantakyla¨, T.-K.; Wa¨rnå, J.; Ma¨ki-Arvela, P.; Kuusisto, J.; Martinez, I. Modelling of kinetic and transport effects in aldol hydrogenation over metal catalysts. Chem. Eng. Sci. 2002, 57, 1793-1803. (10) Toppinen, S.; Salmi, T.; Rantakyla¨, T.-K.; Aittamaa, J. Kinetics of the liquid-phase hydrogenation of benzenen and some monosubstituted alkylbenzenes over a nickel catalyst. Ind. Eng. Chem. Res. 1996, 35, 18241833. (11) Toppinen, S.; Rantakyla¨, T.-K.; Salmi, T.; Aittamaa, J. The liquid phase hydrogenation of benzene and substituted alkylbenzenes over a nickel catalyst in a semi-batch reactor. Catal. Today 1997, 38, 23-30. (12) Julcour, C.; Le Lann, J. M.; Wilhelm, A. M.; Delmas, H. Dynamics of internal diffusion during the hydrogenation of 1,5,9-cyclododecatriene on Pd/Al2O3. Catal. Today 1999, 48, 147-159. (13) Wasserscheid, P., Welton, T., Eds. Ionic Liquids in Synthesis; Wiley-VCH: Weinheim, Germany, 2003. (14) Zhou, Y. Recent Advances in Ionic Liquids for Synthesis of Inorganic Nanomaterials. Curr. Nanosci. 2005, 1, 35-42. (15) van Rantwijk, F.; Lau, R. M.; Sheldon, R. A. Biocatalytic transformations in ionic liquids. Trends Biotechnol. 2003, 21, 131-138. (16) Dupont, J.; Fonseca, G. S.; Umpierre, A. P.; Fichtner, P. F. P.; Teixeira, S. R. Transition-Metal Nanoparticles in Imidazolium Ionic Liquids: Recycable Catalysts for Biphasic Hydrogenation Reactions. J. Am. Chem. Soc. 2002, 124, 4228-4229. (17) Smiglak, M.; Reichert, W. M.; Holbrey, J. D.; Wilkes, J. S.; Sun, L.; Trasher, J. S.; Kirichenko, K.; Singh, S.; Katrizky, A. R.; Rogers, R. D. Combustible ionic liquids by design: is laboratory safety another ionic liquid myth? Chem. Commun. 2006, 2554-2556. (18) Deetlefs, M.; Seddon, K. R. Chim. Oggi 2006, 24, 2, 16-23. (19) Sheldon, R. A.; Lau, R. M.; Sorgedrager, M. J.; van Rantwijk, F.; Seddon, K. R. Biocatalysis in ionic liquids. Green Chem. 2002, 4, 147151.
3940
Ind. Eng. Chem. Res., Vol. 46, No. 12, 2007
(20) Cassol, C. C.; Umpierre, A. P.; Machado, G.; Wolke, S. I.; Dupont, J. Role of Pd nanoparticles in ionic liquid in the Heck reaction. J. Am. Chem. Soc. 2005, 127, 3298-3299. (21) Berger, A.; de Souza, R. F.; Delgado, M. R.; Dupont, J. Ionic liquidphase asymmetric catalytic hydrogenation: hydrogen concentration effects on enantioselectivity. Tetrahedron: Asymmetric 2001, 12, 1825-1828. (22) Mikkola, J.-P.; Virtanen, P.; Korda´s, K.; Karhu, H.; Salmi, T. SILCA- Supported Ionic Liquid Catalysts for Fine Chemicals. Special issue of the XX SICAT-Simpo´sio Ibero-Americano de Cata´lise. Catal. Today 2006. (23) Mikkola, J.-P.; Virtanen, P.; Karhu, H.; Murzin, D.; Yu Salmi, T. Supported Ionic Liquids Catalysts for Fine Chemicals: citral hydrogenation. Green Chem. 2006, 8, 197-205. (24) Rogers, R. D.; Seddon, K. R. Ionic Liquids-Solvents of the Future? Science 2003, 203, 729-293. (25) Huddleston, J. G.; Visser, A. E.; Matthew, R.; Willauer, H. D.; Broker, G. A.; Rogers, R. D. Characterization and comparison of hydrophilic and hydrophobic room temperature ionic liquids incorporating the imidazolium cation. Green Chem. 2001, 3, 156-164. (26) Seddon, K. R.; Stark, A.; Torres, M.-J. Influence of chloride, water and organic solvents on the physical properties of ionic liquids. Pure Appl. Chem. 2000, 12, 2275-2287. (27) Mikkola, J.-P.; Virtanen, P.; Sjo¨holm, R. Aliquat 336R - A versatile and affordable cation source for an entirely new family of hydrophobic ionic liquids. Green Chem. 2006, 8, 250-255 (28) Berger, A.; de Souza, R. F.; Delgado, M. R.; Dupont, J. Ionic liquidphase asymmetric catalytic hydrogenation: hydrogen concentration effect on enantioselectivity. Tetrahedron: Asymmetry 2001, 12, 825-1828. (29) Dyson, P. J.; Laurenczy, G.; Ohlin, C. A.; Vallance, J.; Welton, T. Determination of hydrogen concentration in ionic liquids and the effect (or lack of) on rates of hydrogenation. Chem. Commun. 2003, 2418-2419. (30) Dupont, J.; Suarrez, P. A. Z. Physico-chemical processes in imidazolium ionic liquids. Phys. Chem. Chem. Phys. 2006, 8, 2441-2452. (31) Chiappe, C.; Pieraccini, D. Ionic liquids: solvent properties and organic reactivity. J. Phys. Org. Chem. 2005, 18, 275-297. (32) Wang, J.; Zhu, A.; Zhao, Y.; Zhuo, K. Excess molar volumes and excess logarithm viscosities for binary mixtures of the ionic liquid 1-butyl-
3-methylimidazolium hexafluoro phosphate with some organic compounds. J. Solution Chem. 2005, 34, 5, 585-596. (33) Wilke, C. Am. Inst. Chem. Eng. J. 1955, 1, 264 (34) Satterfield, C. N. Mass transfer in heterogeneous catalysis; MIT Press: Cambridge, MA, 1970. (35) Morgan, D.; Ferguson, L.; Scovazzo, P. Diffusivities of gases in room-temeperature ionic liquids: Data and Correlations obtained using a lag-time technique. Ind. Eng. Chem. Res. 2005, 44, 4815-4823. (36) Reid, R. C.; Prausnitz, J. M.; Poling, B. E. The properties of gases and liquids, 4th ed.; McGraw Hill: New York, 1998. (37) Anderson, K.; Goodrich, P.; Hardacre, C.; Rooney, D. W. Heterogeneously catalysed selective hydrogenation reactions in ionic liquids. Green Chem. 2003, 5, 448-453. (38) Marquardt, D. W. An algorithm for least squares estimation on non-linear parameters. SIAM J. 1963, 11, 431-441. (39) Nelder, J. A.; Mead, R. A simple method for function minimization. Comput. J. 1962, 7, 308-313. (40) Henrici, A. C. Discrete Variable methods in ordinary differential equations; Prentice Hall: Englewood Cliffs, NJ, 1983. (41) Haario, H. Modest Users Manual; Profmath Oy, Helsinki, Finland, 1998. (42) Virtanen, P.; Mikkola, J.-P.; Salmi, T. Kinetics of citral hydrogenation on supported ionic liquid catalysts (SILCA). Green SolVents for Processes Symposium, Oct 8-11, 2006, Lake Constance, Friedrikshafen, Germany; Dechema: Frankfurt/Main, Germany, in press. (43) Qin, Y.; Prausnitz, J. M. Solubilities in ionic liquids and molten salts from a simple perturbed-hard-sphere theory. Ind. Eng. Chem. Res. 2006, 45, 5518-5523. (44) Camper, D.; Scovazzo, P.; Koval, C.; Noble, R. Gas solubilities in room-temperature ionic liquids. Ind. Eng. Chem. Res. 2004, 43, 30493054.
ReceiVed for reView August 15, 2006 ReVised manuscript receiVed October 5, 2006 Accepted October 9, 2006 IE061082O