Separation by Fixed-Bed Adsorption of Hexane Isomers in Zeolite

4316

Ind. Eng. Chem. Res. 2006, 45, 4316-4328

Separation by Fixed-Bed Adsorption of Hexane Isomers in Zeolite BETA Pellets Patrick S. Ba´ rcia,† Jose´ A. C. Silva,† and Alı´rio E. Rodrigues*,‡ Escola Superior de Tecnologia e Gesta˜ o, Instituto Polite´ cnico de Braganç a, Apartado 1134, 5301-857 Braganç a, Portugal, and Laboratory of Separation and Reaction Engineering, Departamento de Engenharia Quı´mica, Faculdade de Engenharia, UniVersidade do Porto, Rua do Dr. Roberto Frias, S/N, 4200-465 Porto, Portugal

An experimental study of single and binary fixed-bed adsorptions of hexane isomers n-hexane (nHEX), 3-methylpentane (3MP), 2,3-dimethylbutane (23DMB) and 2,2-dimethylbutane (22DMB) was performed in commercial pellets of zeolite BETA, covering the temperature range between 423 and 523 K and partial pressures up to 0.3 bar. The effect of partial pressure and temperature on the shape of the breakthrough curves was addressed. From these data, single and binary adsorption equilibrium isotherms were collected. On the basis of the analysis of sorption events at the molecular level, two different models were used to interpret, with good accuracy, the equilibrium data: dual-site Langmuir (DSL) for nHEX and 3MP and multisite Langmuir (MSL) for 23DMB and 22DMB. Thereafter, a dynamic adsorption model was developed and tested, predicting with good accuracy the behavior of the fixed-bed experiments. At the partial pressures studied, it was found that the affinity of the isomers to the zeolite is nHEX > 3MP > 23DMB > 22DMB. The selectivity between the isomers is higher at low partial pressures, decreasing as the amount adsorbed increases. The Ideal Adsorbed Solution Theory using the DSL model to describe the pure component equilibrium of nHEX and 3MP and the MSL model for the dibranched isomers 22DMB and 23DMB gives a good prediction of the mixture adsorption data. Introduction In a world aware of environmental concerns, it is time to research solutions that can provide maximum energy efficiency. The end of cheap oil will drive the refining industry to optimize the performances of their carburants even more, increasing the efficiency of vehicle motors.1 In the case of gasoline, the combustion quality is measured by the research octane number (RON). When the RON is high, combustion occurs as a smooth explosion instead of a detonation and the performance of the motor is improved. In the past 20 years, methyl tertiary butyl ether (MTBE) has been used in gasoline to replace lead as an octane enhancer. However, controversy has surrounded its use, most recently because of concern over contamination of drinking water supplies, leading to calls for restrictions on its use. The phaseout of MTBE in several parts of the U. S. stresses the need for alternative solutions.2 A proper alternative to additives is removing low-octane paraffins from the gasoline. For instance, n-hexane (nHEX) and 3-methylpentane (3MP) exhibit RONs of 24.8 and 74.5, respectively, whereas 2,2-dimethylbutane (22DMB) and 2,3dimethylbutane (23DMB) have RONs of 94.0 and 103.4, respectively. These four isomers are the major constituents of the output of Total Isomerization Processes (TIP), and their molecular kinetic diameters are similar, which makes the separation difficult. The TIP developed by Universal Oil Products consists of a hydroisomerization unit, which converts normal paraffin into branched ones by incomplete catalytic reactions. To improve this process, an adsorption unit packed with zeolite 5A capable of separating branched from linear components not converted * To whom correspondence should be addressed. Tel.: +351225081671. Fax: +351-225081674. E-mail address: [email protected]. † Instituto Politećnico de Bragança. ‡ Universidade do Porto.

in the hydroisomerization unit was added. Thus, the linear components are removed from the isomerate and recycled for a complete conversion in the catalytic reactor. However, there are still low-RON monobranched compounds in the isomerate. To overcome this problem, the present study focuses on the development of an adsorptive process to separate monobranched from dibranched hexane isomers, producing a high-octane, cleaner-burning gasoline. Zeolite BETA is a large-pore-size adsorbent of great importance in catalytic processes of the refining industry because of its thermal and acidic stability;3 however, its functionality as an adsorbent in separation processes is scarcely explored. Some studies of the equilibrium of adsorption on zeolite BETA suggest a relatively good selectivity for separating hexane isomers. Huddersman and Klimczyk4 indicate that zeolite BETA in cation form (H,Ba) is an effective separating media for branched hexane isomers, superior to silicalite. In a study of the sorption of hexane isomers in zeolite BETA, Ba´rcia et al.5 demonstrate that, in the low partial pressure range, the amount of nHEX and 3MP adsorbed is higher than that for the dibranched one. It is reported that in the zeolite BETA structure there are two types of mutually intersecting channels with different pore aperture dimensions similar to the kinetic diameter of the hexane isomers. This characteristic can be advantageous in carrying out the separation, exploiting the conformation of the molecules. Jolimaitre et al.6-8 studied experimentally in a fixed-bed adsorber the separation of C5 and C6 hydrocarbons with different degrees of branching in silicalite molecular sieves, concluding that the kinetics separation of di- from monobranched hydrocarbons is feasible on this type of zeolite. One of the goals of this work is to contribute to a better understanding of the adsorption of hexane isomers in this largeand medium-pore zeolite, using alkanes with considerable potential for commercial application in the petroleum industry, especially in schemes of total isomerization processes. Also, from an engineering point of view, it is important to obtain a

10.1021/ie0513954 CCC: $33.50 © 2006 American Chemical Society Published on Web 05/11/2006

Ind. Eng. Chem. Res., Vol. 45, No. 12, 2006 4317 Table 1. Physical Properties of Zeolite BETA Crystals and Mercury Porosimetry Data of Pellets Physical Properties of Crystals structure type H-BEA 150 Si/Al ratio (mol/mol) 150 particle dimensions (µm)a 0.25-0.40 oxygens in window 12 crystal type tetragonal channel size (nm) 〈100〉 0.66 × 0.67** b Tc [100] 0.56 × 0.56* Mercury Porosimetry Data of Pellets SBET (m2/g) 447.8 intrusion volume (cm3/g) 0.14 d (≈2-30 000 psia) apparent density, Fa (g/cm3) 1.18 solid density, Fs (g/cm3) 1.42 (30 000 psia) porosity, p 0.17 a Determined by SEM. b The number of asterisks indicates whether the channel system is one- or two-dimensional. c Interconnecting channel systems are separated by a double arrow (T), data from www.izastructure.org. d The mercury porosimetry was performed by LABGRAN with a Micrometrics Poresizer 9320, which operated at pressures between 0.5 and 30 000 psia.

good analytical description of pure component isotherms for linear and branched hexane isomers because they are the basis for the estimation of mixture sorption behavior from models such as Ideal Adsorbed Solution Theory (IAST) and Real Adsorbed Solution Theory.9 To reach the objectives, this work consists of (i) performing a set of single- and binary breakthrough curves in order to set up adsorption equilibrium isotherms, (ii) an analysis of the influence of partial pressure and temperature on the width of the mass-transfer zone of breakthrough curves and determination of the selectivities, and (iii) the development and validation of a dynamic model to be used in the implementation of an adsorption process. Experimental Section Characterization of the Adsorbent and the Hexane Isomers. The pellets investigated in this work are a cylindrical extrudate of zeolite BETA crystals, provided by SUD-CHEMIE AG, with diameter of 1/16 in. and a length of 5 mm approximately. The Si/Al ratio of the zeolite is 150. The SEM analysis (performed at CEMUP, University of Porto) of the adsorbent revealed that the crystals that compose the pellet are very small with a particle dimension ranging from 0.2 to 0.4

µm. In Table 1 are reported the mercury porosimetry results of the pellets performed by LABGRAN at the University of Coimbra. Figure 1 shows three-dimensional perspective views of the zeolite BETA framework. Its structure consists of straight 12membered-ring channels of free aperture 6.6 × 6.7 Å viewed along axes [100] and [010] and zigzag 10-membered-ring channels of 5.6 × 5.6 Å viewed along axis [001]. A perspective of the free aperture of the channels is also shown. Approximate three-dimensional structures of hexane isomers are also shown in Figure 1. It can be seen clearly in the structures that nHEX is linear, 3MP has one methyl branch, and 23DMB and 22DMB have two methyl branches, with 22DMB having the substitution of the hydrogen by the methyl groups in the same carbon atom. The kinetic diameter of the molecules is also indicated, because it can play an important role in the sorption events conditioning the access of the bulkier molecules to the zigzag channels of zeolite BETA. According to the values of the kinetic diameters of the hexane isomers and the channels aperture of zeolite BETA, 22DMB and 23DMB can only access the zigzag channels, while 3MP and nHEX can access both types of channels. All hydrocarbons used in this work were of analytical grade and supplied by Sigma-Aldrich. Experimental Setup. A schematic diagram of the apparatus developed to measure single breakthrough curves is shown in Figure 2a. It consists of three main sections: (i) a gas preparation section, (ii) an adsorption section, and (iii) an analytical section. In the gas preparation section, the carrier gas and the paraffins are introduced into the system. The carrier gas used in this work is helium and is introduced in the system in two different streams: the first line (1) passes through an electronic pressure controller (EPC) and the second line (2) through a mass flow controller (MFC). The hydrocarbon stream is continuously introduced with a syringe pump (SP) through an injection port into the carrier gas that flows through line 2. Both lines 1 and 2 run into a four-way crossover valve (V1). From the injection port to the flame ionization detector (FID), the transfer lines are thermally controlled to provide the best temperature stability and to avoid condensation. Valve position, oven temperature, and transfer-line temperature are completely automated. The adsorption column consists of a 4.6 mm i.d. stainless steel column 100 mm in length. During the studies of equilibrium of adsorption, the column is entirely filled with zeolite pellets. In the measurement of breakthrough curves, line 2 goes

Figure 1. Pictorial representation of a fixed-bed adsorption process showing the various levels of porosimetry: the fixed bed, the pellet, and the zeolite framework. The figure also includes a 3D view of hexane isomers and the respective kinetic diameters reported by Maloncy et al.24 and a perspective view of the zeolite BETA framework with the respective 12-ring pore apertures.

4318

Ind. Eng. Chem. Res., Vol. 45, No. 12, 2006

Figure 2. Schematic diagram of the (a) single- and (b) multicomponent breakthrough experimental apparatus. (Chr) chromatography column, (Col) packed column, (EPC) electronic pressure controller, (FID) detector of chromatograph, (IP) injection port, (MFC) mass flow controller, (PC) computer, (S) microliter syringe, (SL) sample loops, (SP) syringe pump, (V1) four-way crossover valve, (V2) six-way crossover valve, (1-7) streams.

directly into the packed column via line 3, whereas the pure carrier gas stream flowing through line 1 is directed to a vent. At the same time, the output of the column is directed to a FID and the signal continuously recorded on a personal computer. The oven where the column is placed is the one of the gas chromatograph with sufficient ventilation in order to keep the fixed-bed adsorption in isothermal conditions. The multicomponent breakthrough experiments were performed on the apparatus shown in Figure 2b. The gas preparation and the adsorption sections are similar to the ones for the singlecomponent experiments. During the saturation stage, the output of the packed column passes through a six-way crossover valve (V1) and then is directed to a heated sample collector with 10 loops (SL). The stream in line 3 comes back to valve V1 and flows through line 4, being directed to a four-way crossover valve (V2) and then to the FID. This setup allows us to monitor the packed column output signal, to collect samples during the most important stage of the breakthrough curve. When the saturation state is reached, the crossover valves (V1 and V2) are actuated and the composition of each loop is

evaluated by chromatography. Meanwhile, the carrier gas flows through line 2, cleaning the packed column. Experimental Procedure. The experimental procedure follows: A bed of adsorbent was operated in the adsorption by introducing, at the inlet, the feed containing the adsorbable species diluted in helium at a fixed partial pressure. The experiment consisted of measuring continuously the concentration profile at the outlet of the bed kept at a fixed temperature. The equilibrium loading was obtained by integrating the concentration profiles of the complete breakthrough. Before each run, the column was activated for at least 4 h at 573 K under a helium flow. Theoretical Analysis Determination of Experimental Adsorption Equilibrium Isotherms. Pure-component isotherms can be obtained from the breakthrough curves. The equilibrium loading in each experiment is obtained by integration of the adsorption curves as shown in Figure 3. In Figure 3i, A represents the number of moles of paraffins retained in the bed. At saturation, all of the

Ind. Eng. Chem. Res., Vol. 45, No. 12, 2006 4319

Figure 3. Schematic diagram showing the method used to calculate the amount adsorbed from breakthrough curves: (i) single- and (ii) binarycomponent experiment.

void volumes of the bed are filled by the feed. The amount adsorbed, q, in each experiment is then calculated by the following relation:

q ) area A

(1)

Binary isotherms can also be obtained from the breakthrough curves as shown in Figure 3ii. In graphical terms, the number of moles of compound 1 retained in the bed is equal to area A minus the roll-up area OS. For compound 2, the number of moles retained in the bed is equal to area A plus area B. The amount adsorbed of each compound is then calculated by the following relations:

q1 ) area A - area OS

(2)

of alkanes in silicalite. The MSL model assumes that the adsorbed molecule occupies a certain number of active sites on a single homogeneous surface, and it has been capable of correctly fitting data of n-paraffins in zeolite 5A (Silva and Rodrigues).13 Adsorption isotherm parameters can be obtained from sorption data by an appropriate numerical routine that minimizes DNav, defined as the average of the absolute values of the difference between experimental and calculated results.14 Thermodynamic consistency for model isotherms requires that, for every component, the saturation loading capacity, qm, must be kept constant at different temperatures.15 Predicting Binary Adsorption Isotherms by the Ideal Adsorbed Solution Theory. IAST, developed by Myers and Prausnitz,16 has been applied to predict the mixture adsorption equilibria from the single-component isotherms. An advantage of the IAST model is that it puts no restriction on the description of pure-component sorption. Consequently, different isotherm models can be combined for the different hexane isomers. This is a valuable tool of IAST relative to other models. In IAST, the equilibrium between the two-dimensional adsorbed phase, which is considered to be an ideal solution, and a perfect gas phase is described by an equation analogous to Raoult’s law for vapor-liquid equilibria:

Pyi ) Pi0xi

for i ) 1, 2, ..., n

(5)

and

xi )

qi q1 + q2 + ... + qn

(6)

where yi and xi denote the mole fractions of component i in the gas phase and in the adsorbed phase, respectively, and Pi0 is the hypothetical pressure of the pure component i, which yields the same spreading pressure, π, as that for the mixture. The surface potential is defined by the Gibbs adsorption isotherm: where A is the surface area per unit volume of the adsorbent

Aπ ) RT

∫0

Pi0

qi0 Pi0

dPi0

(7)

In the case of an equimolar mixture, the selectivity, S, is given

q2 ) area A + area B

(3)

by the quotient between the amounts adsorbed of the two species as follows:

q2 area A + area B S) ) q1 area A - area OS

and qi0 is the amount adsorbed from a pure component isotherm, which can be, for example, the MSL or the DSL models or a combination of the two. The total amount adsorbed, qmix, is obtained from n

(4)

It should be noted that, in the case of adsorption from the gas phase, the amount of sorbate gas in the void volumes of the pellet and crystals can be neglected relative to the amount adsorbed in the solid phase. Modeling Adsorption Equilibrium. Two important adsorption equilibrium isotherms for the interpretation of sorption in zeolites are the dual-site Langmuir (DSL)10 and multisite Langmuir (MSL).11 Table 2 summarizes the mathematical relations of these models. Each model represents a different interpretation of adsorption equilibrium behavior. For instance, the DSL model represents adsorption on two different homogeneous types of sites, and it was successfully used by Krishna et al.12 to adequately model the adsorption isotherm inflection

qmix )

n

qi ) ∑ ∑ i)1 i)1

qi0(Pi0) xi

(8)

The loadings of the individual components in the mixtures can be obtained solving numerically the set of equations, eqs 5-8, in addition to the single-component sorption isotherm models selected for each component. Numerical Modeling of Breakthrough Experiments. Consider the adsorption system as an isothermal column packed with pellets of zeolite BETA through which an inert gas flows in a steady state. At time zero, a hexane isomer (or a mixture) of known composition and an inert are introduced at the bottom of the column. The following additional assumptions are made: (1) The ideal gas law applies. (2) The pressure drop through the bed is negligible. (3) The flow pattern is described by the axial dispersed plug flow model. (4) The main resistances

4320


Table 2. Isotherm Models and Respective Parameters model

equation

parameters

multisite Langmuir (MSL)

b)

dual-site Langmuir (DSL)

q ) qm1[(b1p)/(1 + b1p)] + qm2[(b2p)/(1 + b2p)]

1/

p[θ/(1

-

θ)n]

to mass transfer for adsorbable species are external fluid film resistance and macropore diffusion, as pointed out by Ba´rcia et al.5 in ZLC studies of the hexane isomers in pellets of zeolite BETA. External resistance and macropore diffusion can be combined in a global resistance according to a lumped model for the adsorbent particle, as suggested by Morbidelli et al.17 (5) The column is isothermal. According to these assumptions, the model equations are as follows. Mass Balance to Sorbate Species.

∂2ci ∂(uci) ∂ci ∂qji bDL 2 ) + b + (1 - b)Fa ∂z ∂t ∂t ∂z

(9)

where ci is the molar concentration of sorbate species i in the bulk gas phase, u is the superficial velocity, qji is the average adsorbed concentration in the pellet, Fa is the apparent adsorbent density, z and t are the axial coordinate and time, respectively, DL is the axial dispersion coefficient, and b is the bed porosity. OVerall Mass Balance.

∂qji ∂C ∂u + b + (1 - b)Fa ) 0 ∂z ∂t ∂t

C

(10)

where C is the total gas concentration. Mass Transfer Rate. Because the accumulation of mass in the gas phase can be neglected relative to the solid phase, the following equation is valid in a lumped particle system controlled by macropore diffusion7

∂qji ) apKgli(ci - cji) ∂t

Fa

(11)

where cji is the average concentration of sorbate species i in the macropores, ap is the specific area of the particle, and Kgl is the overall mass-transfer coefficient, which is defined by

1 1 1 ) + Kgl ke pki

(12)

where ke is the external film mass-transfer coefficient, ki is the internal mass-transfer coefficient,18 and p is the intraparticle porosity. This approach to the mass-transfer rate has been used with success in the study of the separation of n/iso paraffins in zeolite 5A by pressure-swing adsorption.19 Axial Dispersion and Internal and External Mass-Transfer Resistances. The axial mass dispersion coefficient (DL) was estimated from the correlation of Hsu and Haynes20

0.328 3.33 1 ) + Pe Re‚Sc 1 + 0.59(Re‚Sc)-1

(13)

where Pe is the particle Peclet number, Re is the Reynolds number, and Sc is the Schmidt number. The above correlation was obtained from experiments in a range of 0.08 < Re‚Sc
monobranched > linear). This behavior is similar to the one found for hexane isomers in the AFI structure.9 Consequently, the amounts adsorbed of all the isomers will be similar at a high partial pressure, and the selectivity between the hexane isomers decreases. According to the above considerations, DSL and MSL models were selected to describe experimental adsorption equilibrium

data. Thus, the DSL model was used for linear and monobranched molecules (nHEX and 3MP) and the MSL model for monobranched molecules (22DMB and 23DMB). Figure 4 shows the adsorption equilibrium isotherm models fitted to experimental data. It is clear that both the MSL and DSL models are reasonable in predicting the isotherm sorption behavior of dibranched and non-dibranched molecules, respectively, for the temperatures studied. The fitted parameters of the isotherm models are listed in Table 4, including the Henry’s constants. The values of the Henry’s constants clearly indicate that the affinity order of the hexane isomers at all temperatures and at low partial pressures is nHEX > 3MP > 23DMB > 22DMB. The maximum amount adsorbed is very similar for all of the isomers, ranging from 9.8 g/100 gads for 22DMB to 8.2 g/100 gads for 23DMB. Binary Adsorption Isotherms. The experiments were performed at 423, 473, and 523 K and partial pressures up to 0.06 bar. Parts a1-a3 of Figure 6 show the experimental binary adsorption isotherms for an equimolar mixture of 22DMB/3MP, whereas parts b1-b3 of Figure 6 do so for an equimolar mixture of 23DMB/3MP. These binary sorption isotherms were measured in the apparatus shown in Figure 2b from the breakthrough


Figure 5. Schematic representation showing sorption sites and the location of hexane isomers in the zeolite BETA structure. Table 4. Isotherm Models’ Parameters for the Sorption of the Hexane Isomers on β Zeolite Pellets, Henry’s Constants, and Average Deviation between Model and Experimental Data 22DMB

23DMB

3MP

nHEX

isotherm models

MSL

MSL

DSL

DSL

qm -∆H qm1 -∆H1 qm2 -∆H2

(g/100 gads) (kJ/mol) (g/100 gads) (kJ/mol) (g/100 gads) (kJ/mol)

9.8 73.5

8.2 79.6 4.7 77.7 3.7 95.5

5.0 86.2 4.5 66.5

Ha b b1 b2 n DNavb

(g/gads bar) (bar -1) (bar -1) (bar -1) (-)

12.55

25.22

H b b1 b2 n DNav


H B b1 b2 n DNav ∑DNav


T ) 423 K 2.65 27.0

5.58 68.0

260 9.0 1.8 0.39

1.8 0.28

T ) 473 K 0.26 2.7

0.48 5.8

1.8 0.11

1.8 0.12

T ) 523 K 0.05 0.50

0.07 0.90

1.8 0.06 0.56

1.8 0.06 0.46

500 5.0

0.16

0.16

1.19

2.03

25.0 0.50

40.0 0.70

0.27

0.18

0.18

0.23

3.80 0.05

4.58 0.13

0.23 0.66

0.06 0.40

a H ) K q is Henry’s constant. b DN is defined as the average of the eq m av absolute values of the difference between experimental and calculated results (Valenzuela and Myers14).

experiments with the procedure previously described (eqs 2 and 3). The study of these binary mixtures is important because it is crucial for octane number enhancement in TIP, the separation of the monobranched from the dibranched paraffins.

For the mixture of 22DMB/3MP, parts a1-a3 of Figure show that there are significant differences in the amount adsorbed for each component. At all of the temperatures studied, the ratio between the amounts adsorbed of 3MP and 22DMB decreases as the partial pressure increases. This behavior is clearly shown in Figure 7, which represents the sorption selectivity calculated from eq 4 as a function of the total pressure. These differences can be explained by the consideration previously discussed, relative to the structures of the zeolite and the molecules. Relative to the equimolar mixture of 23DMB/3MP, parts b1b3 of Figure 6 demonstrate that the difference between the amounts adsorbed is smaller than that for the previous case. Figure 7 shows the sorption selectivity, where it can be seen that the selectivity between 3MP and 23DMB is very small, decreasing slightly with increasing partial pressure. IAST was applied to predict the mixture adsorption equilibria from the single-component isotherms. The IAST prediction (with the DSL model for 3MP and the MSL model for 22DMB and 23DMB) is represented by the lines in Figure 6, and in all cases, it gives a proper description of the binary adsorption data for the equimolar mixtures of 22DMB/3MP and 23DMB/3MP. This indicates that the single-component isotherm models used are adequate to describe the multicomponent equilibrium of the hexane isomers. These results show that a clear understanding of the singlecomponent sorption events with adequate modeling is crucial to the success of multicomponent sorption prediction with IAST. Influence of Temperature on Single-Component Breakthrough Curves of Hexane Isomers at Low Partial Pressures (p ) 0.06 bar). Parts a-d of Figure 8 show the influence of temperature on the breakthrough time of hexane isomers at partial pressures near 0.06 bar. We plot the breakthrough curves in terms of normalized concentrations of the sorbate c/c0, as a function of time. The stoichiometric time, tst, is also shown in the figures, and it is calculated by

tst )

(

)

Fbq0 L 1+ ui bc0

(16)

where ui is the interstitial velocity, L is the bed length, Fb is the bulk density, and q0 is the adsorbed concentration in equilibrium with the gas-phase concentration c0 at the inlet of the bed. Several important conclusions can be obtained from Figure 8: First, for the same flowrate, the stoichiometric time tst decreases significantly as the temperature increases; it being 61.7 min for nHEX at 423 K compared to 12.4 min at 523 K. For 22DMB, the component with the lower affinity to the adsorbent, it decreases from 47.1 min at 423 K to only 3.7 min at 523 K. The relative stoichiometric times (adsorption capacities) of the hexane isomers fit the following trend: nHEX > 3MP > 23DMB > 22DMB. This trend was observed at the three temperatures studied, indicating that the adsorption strengths of the hexane isomers to the adsorbent decrease as the degree of branching increases. This sorption hierarchy is similar to the one reported by Smit and Krishna9 for purecomponent and mixture sorption of hexane isomers in MFI zeolite, in studies performed by Monte Carlo simulations. An important feature is that the difference between the stoichiometric times, tst, of 22DMB and 23DMB at 523 K is practically null. More, from the parameters shown in Table 4 obtained from the fitting procedure of the isotherms, the MSL model predicts a selectivity based on Henry’s constants between 23DMB and 22DMB at 523 K of S ) HMSL(23DMB)/HMSL(22DMB) ) 0.07/0.05 ) 1.4, a value completely different at p ) 0.06

4324


Figure 6. Binary adsorption equilibrium isotherm for an equimolar mixture of (a) 22DMB/3MP and (b) 23DMB/3MP at (1) T ) 423 K, (2) T ) 473 K, and (3) T ) 523 K. The continuous lines represent the prediction of the IAST. The represented models are MSL for 22DMB and 23DMB and DSL for 3MP.

bar. These results show that selectivities based on Henry’s constants should be analyzed with caution in fixed-bed operations. These differences in selectivities have been already pointed out by Smit and Krishna9 whom demonstrate from molecular simulations that using Henry coefficients to estimate the sorption selectivity of mixtures can lead to completely wrong conclusions. Influence of Temperature on Single-Component Breakthrough Curves of Hexane Isomers at Medium Partial Pressures (p ) 0.12 bar). Parts a-d of Figure 9 show the influence of temperature on the breakthrough time of hexane isomers at partial pressures of 0.12 bar. An overall perspective of this figure shows that the width of the mass-transfer zone of the experiments is similar to the one found at a partial pressure of 0.06 bar, at all temperatures studied. However, in this case, we should note that at a fixed temperature the amounts adsorbed for all components tend to be closer. At this partial pressure, the difference tst(3MP) - tst(23DMB) at 423 K is 5.4 min;

thus, this difference has decreased relative to the previous case at low pressure (6.7 min). This is true for all other components. It is also important to note that Figure 9 shows that at 523 K there is now some selectivity between 23DMB and 22DMB. Influence of Temperature on Single-Component Breakthrough Curves of Hexane Isomers at High Partial Pressures (p ) 0.28 bar). Parts a-d of Figure 10 show the influence of temperature on the breakthrough time of hexane isomers at partial pressures of 0.28 bar. At this partial pressure, the difference tst(3MP) - tst(23DMB) at 423 K is only 1.7 min. We can conclude that at this partial pressure the breakthrough curves of the isomers are very close, and this can be explained by the previous discussion of the molecular structure of the zeolite BETA. At high loading, bulky molecules such as 22DMB and 23DMB are more easily filled in the large straight channel as a consequence of decreasing the linear dimension of the molecule, and this is reflected in the proximity between the


Figure 7. Sorption selectivity as a function of total system pressure. Mixtures are 50/50 of 3MP/22DMB and 3MP/23DMB at (a) 423 K, (b) 473 K, and (c) 523 K.

breakthrough curves at high partial pressures and low temperatures (higher adsorbent loading). Results of the Simulation of Breakthrough Experiments. The figures discussed previously also show the results of simulations (lines in the figures) with the numerical model developed. On the basis of the previous discussion, the adsorption equilibrium is predicted with the DSL model for nHEX and 3MP and with the MSL model for the branched isomers with the parameters shown in Table 4. We conclude from both figures that the shapes of the breakthrough curves

Figure 8. Single-component breakthrough curves at a partial pressure (p) of near 0.06 bar, for 22DMB, 23DMB, 3MP, and nHEX at T ) 423 K (O), T ) 473 K (4), and T ) 523 K ()). The experimental conditions and model parameters are specified in Table 3. The represented models are MSL for the branched isomers and DSL for 3MP and nHEX. The amount adsorbed (g/100 gads) in bold and the stoichiometric time (minutes) in italics are indicated next to the respective breakthrough curve. The points are experimental data, and the lines are model predictions.

are successfully predicted by the dynamic adsorption model, as well as the width of mass-transfer zone. These results are in agreement with the ones of Krishna and Baur23 who state that, in cases where macropore/micropore diffusion resistances are present, the nonlinearities of intracrystalline zeolite diffusion are diluted and the LDF-type approximations may be good ones. However, the predictions in the case of 3MP at 473 K show some deviations in the experimental results, which calls for a deeper study of the adsorption equilibrium.

4326



Breakthrough Curves of Binary Equimolar Mixtures of Hexane Isomers. Parts a and b of Figure 11 show binary breakthrough curves for an equimolar mixture of 22DMB/3MP and 23DMB/3MP, respectively, at 473 K. The experimental conditions as well as the parameters of the dynamic model are listed in Table 5. In the case of the 22DMB/3MP mixture, it can be seen that the degree of selectivity is significant. As expected, 3MP is the more strongly adsorbed component, mainly because it can access


different sites of the zeolite BETA. The selectivity obtained is acceptable for the development of a fixed-bed process between 423 and 523 K. Moreover, the dynamic mathematical model combined with IAST (using the DSL model for 3MP and the MSL model for 22DMB) is capable of describing the breakthrough curves very well. Note that, at high concentrations, rollup phenomena are also observed experimentally as a result of the nonlinearity of the adsorption isotherms. This does agree with the equilibrium data, which predict the same tendency.

Ind. Eng. Chem. Res., Vol. 45, No. 12, 2006 4327 Table 5. Experimental Conditions and Dynamic Model Parameters for Binary Breakthrough Curves in Zeolite β Pellets experimental conditions

run

temp (K)

helium flowrate (mL/min)

I II

473 473

5.7 5.7

model parameters

hydroc. flowrate (µmol/min) 22DMB

23DMB

3MP

ke (m/s)

5.5

5.5 5.5

0.1038 0.1038

5.4

apKgl (s-1)

ki (m/s)

Considering the mixture of 23DMB/3MP, Figure 11b shows that the degree of separation is not as good as that in the case of 22DMB/3MP. However, it is important to note that the breakthrough curves do not overlap. Breakthrough Curves of Quaternary Equimolar Mixtures of Hexane Isomers. Quaternary breakthrough experiments were performed with equimolar mixtures of hexane isomers. Figure 12 shows an experimental quaternary breakthrough curve for an equimolar mixture of nHEX/3MP/23DMB/22DMB in zeolite BETA at 423 K and a mixture partial pressure, pmix, close to 0.01 bar. The retention times are in the following order: nHEX > 3MP > 23DMB > 22DMB. This experiment shows a complete separation between the isomers, which indicates that the separation of the hexane isomers in zeolite BETA is feasible. The results of this study are now being used in the development of a cyclic process for the separation of mono- and dibranched hexane isomers by appropriate technology. Conclusions This work shows clearly that zeolite BETA can be an effective adsorbent for the separation of hexane isomers. Several breakthrough curves of hexane isomers in commercial pellets of

Figure 11. Binary breakthrough curves for mixtures of (a) 22DMB-3MP and (b) 23DMB-3MP at 473 K in zeolite BETA. The experimental conditions and model parameters are specified in Table 5. The lines represent IAST predictions, and the points are experimental data.

22DMB

23DMB

3MP

22DMB

0.0935 0.0935

20.8

0.0296

0.0340

23DMB

3MP

DL (m2/s)

18.3

52.3 52.3

1.2×10-4 1.3×10-4

zeolite BETA were performed, with the objective of measuring adsorption equilibrium isotherms and the development of a predictive dynamic fixed-bed adsorption model. The equilibrium data obtained show that a significant amount of hexane isomers can be adsorbed in the zeolites, with values at 423 K and at a partial pressure of 0.3 bar reaching values of 7 g/100 gads; however, with increasing temperature, the values decrease, reaching approximately 2 g/100 gads at 523 K for the same pressure. Two different models were used to fit the experimental data named: dual-site-Langmuir and multisite-Langmuir. On the basis of the fitting procedure and an analysis of sorption events at the molecular level, it is shown that the DSL model is the best for nHEX and 3MP and the MSL model is suitable for predicting the equilibrium data of branched isomers. From the analysis of breakthrough curves, the following points and conclusions emerge: the stoichiometric time of the hexane isomers decreases in the order nHEX > 3MP > 23DMB > 22DMB at the three temperatures studied, indicating that the adsorption strengths of the hexane isomers to the adsorbent decrease as the degree of branching increases; at temperature T ) 423 K and partial pressure p ) 0.28 bar, the difference between the stoichiometric times of each isomer is very small (high adsorbent loadings); however, this difference increases at 523 K for all of the partial pressures studied. The results of the simulations clearly show that the agreement between model and experimental data is reasonably good when we consider the wide range of temperatures, partial pressures, and hexane isomers studied. The binary mixture adsorption equilibrium was measured at 423, 473, and 523 K. Zeolite BETA demonstrates a significant degree of selectivity for the equimolar mixture of 22DMB/3MP, giving a good perspective regarding future work. Moreover, the IAST predicts well the binary isotherm, and the dynamical model using IAST with the DSL model for 3MP and the MSL model for 22DMB and 23DMB gives a good description of the binary breakthrough curves. The results of this study are now being used in the development of a cyclic process for the separation of mono- and dibranched hexane isomers.

Figure 12. Experimental quaternary breakthrough curve for an equimolar mixture of nHEX/3MP/23DMB/22DMB in zeolite BETA at 423 K and a mixture partial pressure, pmix, of close to 0.01 bar.

4328


Nomenclature ap ) specific area of particle (m-1) A ) specific surface area per unit volume of the adsorbent (m2 m-3) b ) parameter in adsorption equilibrium isotherms (Pa-1) c0 ) molar gas concentration at the inlet of the fixed bed (mol m-3) ci ) molar concentration of sorbate species i in the bulk gas phase (mol m-3) cji ) average concentration of sorbate species i in the macropores (mol m-3) C ) total molar gas concentration in bulk gas phase (mol m-3) DL ) axial dispersion coefficient in fixed bed (m2 s) Dp ) macropore diffusivity (m s-2) H ) Henry’s law coefficient (mol kg-1 Pa-1) ke ) external film mass-transfer coefficient (m s-1) ki ) internal mass-transfer coefficient (m s-1) Kgl ) overall mass-transfer coefficient (m s-1) L ) fixed bed length (m) n ) parameter representing the number of active sites occupied by adsorbed molecules in the MSL isotherm model (-) p ) partial pressure (Pa) Pe ) particle Peclet number (-) q ) loading concentration of sorbate in the adsorbent (mol kg-1) q0 ) loading concentration of sorbate in equilibrium with the gas-phase concentration c0 at the inlet of the fixed bed (mol kg-1) qm ) saturation loading concentration of sorbate in the adsorbent (mol kg-1) qji ) average adsorbed concentration in adsorbent particle (mol m-3) Re ) Reynolds number (-) Rp ) pellet radius (m) S ) sorption selectivity (-) Sc ) Schmidt number (-) Sh ) Sherwood number (-) t ) time (s) tst ) stoichiometric time (s) u ) superficial velocity in packed bed (m s-1) ui ) interstitial velocity in packed bed (m s-1) z ) distance coordinate along fixed bed (m) Greek Letters b ) fixed-bed porosity (-) p ) pellet porosity (-) π ) spreading pressure (Joule m-2) θ ) fractional surface occupancy of the adsorbent particle (-) Fa ) apparent adsorbent density (mol m-3) Fb ) bulk density of the fixed bed (mol m-3) Literature Cited (1) Campbell, C. Association for the Study of Peak Oil & Gas (ASPO). http://www.peakoil.net/ (accessed Dec 2005).

(2) McCarthy, J. E.; Tiemann, M. MTBE in Gasoline: Clean Air and Drinking Water Issues. CRS: Report for Congress, 2000. (3) Davis, M. E.; Zones, S. I. Synthesis of Porous Materials: Zeolites, Clays and Nanostuctures; Marcel Dekker: New York, 1996. (4) Huddersman, K.; Klimczyk, M. Separation of Branched Hexane Isomers Using Zeolite Molecular Sieves. AIChE J. 1996, 42, 405. (5) Ba´rcia, P. S.; Silva, J. A. C.; Rodrigues, A. E. Adsorption Equilibrium and Kinetics of Branched Hexane Isomers in Pellets of BETA Zeolite. Microporous Mesoporous Mater. 2005, 79, 145. (6) Jolimaitre, E. EÄ tude et Mode´lisation de l’Adsorption et du Transfert de Matie`re Dans les Zeólithes de Structure MFI. Application a` la Se´paration des Hydrocarbures Sature´s Mono et Di-branche´s. Ph.D. Thesis, LAGEP, Villeurbanne Cedex, France, 1999. (7) Jolimaitre, E.; Fayolle, M. T.; Jallut, C.; Ragil, K. Determination of Mass Transfer and Thermodynamic Properties of Branched Paraffins in Silicalite by Inverse Chromatography Technique. Ind. Eng. Chem. Res. 2001, 40, 914. (8) Jolimaitre, E.; Ragil, K.; Fayolle, M. T.; Jallut, C. Separation of Mono- and Dibranched Hydrocarbons on Silicalite. AIChE J. 2002, 48, 1927. (9) Smit, B.; Krishna, R. Molecular Simulations in Zeolitic Process Design. Chem. Eng. Sci. 2003, 58, 557. (10) Doetsch, I. H.; Ruthven, D. M.; Loughlin, K. F. Sorption and Diffusion of n-Heptane in 5A Zeolite. Can. J. Chem. 1974, 52, 2117. (11) Nitta, T.; Shigetomi, T.; Kuro-Oka, M.; Katayama, M. J. An Adsorption Isotherm of Multi-site Occupancy Model for Homogeneous Surface. J. Chem. Eng. Jpn. 1984, 17, 39. (12) Krishna, R.; Vlugt, T. J. H.; Smit, B. Influence of Isotherm Inflection on Diffusion in Silicalite. Chem. Eng. Sci. 1999, 54, 1751. (13) Silva, J. A. C.; Rodrigues, A. E. Multisite Langmuir Model Applied to the Interpretation of Sorption of n-Paraffins in 5A Zeolite. Ind. Eng. Chem. Res. 1999, 38, 2432. (14) Valenzuela, D. P.; Myers, A. L. Adsorption Equilibrium Data Handbook; Prentice Hall Inc.: Englewood Cliffs, New Jersey, 1989. (15) Sircar, S. Influence of Adsorbate Size and Adsorbent Heterogeneity of IAST. AIChE J. 1995, 41, 1135. (16) Myers, A. L.; Prausnitz, J. M. Thermodynamics of Mixed-Gas Adsorption. AIChE J. 1965, 11, 121. (17) Morbidelli, M.; Servida, A.; Storti, G.; Carra´, S. Simulation of Multicomponent Adsorption Beds. Model Analysis and Numerical Solution. Ind. Eng. Chem. Fundam. 1982, 21, 123. (18) Glueckauf, E. Formulae for Diffusion into Spheres and Their Application to Chromatography. J. Chem. Soc. 1955, 51, 1540. (19) Silva, J. A. C.; Rodrigues, A. E. Separation of n/iso Paraffins by PSA. Sep. Purif. Technol. 2000, 20, 97. (20) Hsu, L. K. P.; Haynes, H. M. Effective Diffusivity by the Gas Chromatography Technique: Analysis and Application to Measurements of Diffusion of Various Hydrocarbons in Zeolite NaY. AIChE J. 1981, 27, 91. (21) Wakao, N.; Funazkri, T. Effect of Fluid Dispersion Coefficients on Fluid-to-Particle Mass Transfer Coefficients in Packed Beds. Chem. Eng. Sci. 1978, 33, 1375. (22) Villadsen, J. V.; Michelsen, M. L. Solution of Differential Equation Models by Polynomial Approximation; Prentice Hall Inc.: Englewood Cliffs, New Jersey, 1978. (23) Krishna, R.; Baur, R. Modelling Issues in Zeolite Based Separation Processes. Sep. Purif. Technol. 2003, 33, 213. (24) Maloncy, M. L.; Gora, L.; Jansen, J. C.; Maschmeyer, T. Conceptual Processes for Zeolite Membrane Based Hydroisomerization of Light Alkanes. Ars Separatoria Acta 2003, 2, 18.

ReceiVed for reView December 14, 2005 ReVised manuscript receiVed March 24, 2006 Accepted April 10, 2006 IE0513954

Separation by Fixed-Bed Adsorption of Hexane Isomers in Zeolite

Recommend Documents